xref: /linux/net/ipv6/mcast.c (revision 56d95b0adfa224bb1c67733dbcad30dd8debd39e)
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_arp.h>
37 #include <linux/route.h>
38 #include <linux/init.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/slab.h>
42 #include <linux/pkt_sched.h>
43 #include <net/mld.h>
44 #include <linux/workqueue.h>
45 
46 #include <linux/netfilter.h>
47 #include <linux/netfilter_ipv6.h>
48 
49 #include <net/net_namespace.h>
50 #include <net/sock.h>
51 #include <net/snmp.h>
52 
53 #include <net/ipv6.h>
54 #include <net/protocol.h>
55 #include <net/if_inet6.h>
56 #include <net/ndisc.h>
57 #include <net/addrconf.h>
58 #include <net/ip6_route.h>
59 #include <net/inet_common.h>
60 
61 #include <net/ip6_checksum.h>
62 
63 /* Ensure that we have struct in6_addr aligned on 32bit word. */
64 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
65 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
66 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
67 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
68 };
69 
70 static struct workqueue_struct *mld_wq;
71 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
72 
73 static void igmp6_join_group(struct ifmcaddr6 *ma);
74 static void igmp6_leave_group(struct ifmcaddr6 *ma);
75 static void mld_mca_work(struct work_struct *work);
76 
77 static void mld_ifc_event(struct inet6_dev *idev);
78 static bool mld_in_v1_mode(const struct inet6_dev *idev);
79 static int sf_setstate(struct ifmcaddr6 *pmc);
80 static void sf_markstate(struct ifmcaddr6 *pmc);
81 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
82 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
83 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
84 			  int delta);
85 static int ip6_mc_add_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_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
89 			    struct inet6_dev *idev);
90 static int __ipv6_dev_mc_inc(struct net_device *dev,
91 			     const struct in6_addr *addr, unsigned int mode);
92 
93 #define MLD_QRV_DEFAULT		2
94 /* RFC3810, 9.2. Query Interval */
95 #define MLD_QI_DEFAULT		(125 * HZ)
96 /* RFC3810, 9.3. Query Response Interval */
97 #define MLD_QRI_DEFAULT		(10 * HZ)
98 
99 /* RFC3810, 8.1 Query Version Distinctions */
100 #define MLD_V1_QUERY_LEN	24
101 #define MLD_V2_QUERY_LEN_MIN	28
102 
103 #define IPV6_MLD_MAX_MSF	64
104 
105 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
106 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
107 
108 /*
109  *	socket join on multicast group
110  */
111 #define mc_dereference(e, idev) \
112 	rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock))
113 
114 #define sock_dereference(e, sk) \
115 	rcu_dereference_protected(e, lockdep_sock_is_held(sk))
116 
117 #define for_each_pmc_socklock(np, sk, pmc)			\
118 	for (pmc = sock_dereference((np)->ipv6_mc_list, sk);	\
119 	     pmc;						\
120 	     pmc = sock_dereference(pmc->next, sk))
121 
122 #define for_each_pmc_rcu(np, pmc)				\
123 	for (pmc = rcu_dereference((np)->ipv6_mc_list);		\
124 	     pmc;						\
125 	     pmc = rcu_dereference(pmc->next))
126 
127 #define for_each_psf_mclock(mc, psf)				\
128 	for (psf = mc_dereference((mc)->mca_sources, mc->idev);	\
129 	     psf;						\
130 	     psf = mc_dereference(psf->sf_next, mc->idev))
131 
132 #define for_each_psf_rcu(mc, psf)				\
133 	for (psf = rcu_dereference((mc)->mca_sources);		\
134 	     psf;						\
135 	     psf = rcu_dereference(psf->sf_next))
136 
137 #define for_each_psf_tomb(mc, psf)				\
138 	for (psf = mc_dereference((mc)->mca_tomb, mc->idev);	\
139 	     psf;						\
140 	     psf = mc_dereference(psf->sf_next, mc->idev))
141 
142 #define for_each_mc_mclock(idev, mc)				\
143 	for (mc = mc_dereference((idev)->mc_list, idev);	\
144 	     mc;						\
145 	     mc = mc_dereference(mc->next, idev))
146 
147 #define for_each_mc_rcu(idev, mc)				\
148 	for (mc = rcu_dereference((idev)->mc_list);             \
149 	     mc;                                                \
150 	     mc = rcu_dereference(mc->next))
151 
152 #define for_each_mc_tomb(idev, mc)				\
153 	for (mc = mc_dereference((idev)->mc_tomb, idev);	\
154 	     mc;						\
155 	     mc = mc_dereference(mc->next, idev))
156 
157 static int unsolicited_report_interval(struct inet6_dev *idev)
158 {
159 	int iv;
160 
161 	if (mld_in_v1_mode(idev))
162 		iv = READ_ONCE(idev->cnf.mldv1_unsolicited_report_interval);
163 	else
164 		iv = READ_ONCE(idev->cnf.mldv2_unsolicited_report_interval);
165 
166 	return iv > 0 ? iv : 1;
167 }
168 
169 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
170 			       const struct in6_addr *addr, unsigned int mode)
171 {
172 	struct net_device *dev = NULL;
173 	struct ipv6_mc_socklist *mc_lst;
174 	struct ipv6_pinfo *np = inet6_sk(sk);
175 	struct net *net = sock_net(sk);
176 	int err;
177 
178 	ASSERT_RTNL();
179 
180 	if (!ipv6_addr_is_multicast(addr))
181 		return -EINVAL;
182 
183 	for_each_pmc_socklock(np, sk, mc_lst) {
184 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
185 		    ipv6_addr_equal(&mc_lst->addr, addr))
186 			return -EADDRINUSE;
187 	}
188 
189 	mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
190 
191 	if (!mc_lst)
192 		return -ENOMEM;
193 
194 	mc_lst->next = NULL;
195 	mc_lst->addr = *addr;
196 
197 	if (ifindex == 0) {
198 		struct rt6_info *rt;
199 		rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
200 		if (rt) {
201 			dev = rt->dst.dev;
202 			ip6_rt_put(rt);
203 		}
204 	} else
205 		dev = __dev_get_by_index(net, ifindex);
206 
207 	if (!dev) {
208 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
209 		return -ENODEV;
210 	}
211 
212 	mc_lst->ifindex = dev->ifindex;
213 	mc_lst->sfmode = mode;
214 	RCU_INIT_POINTER(mc_lst->sflist, NULL);
215 
216 	/*
217 	 *	now add/increase the group membership on the device
218 	 */
219 
220 	err = __ipv6_dev_mc_inc(dev, addr, mode);
221 
222 	if (err) {
223 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
224 		return err;
225 	}
226 
227 	mc_lst->next = np->ipv6_mc_list;
228 	rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
229 
230 	return 0;
231 }
232 
233 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
234 {
235 	return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
236 }
237 EXPORT_SYMBOL(ipv6_sock_mc_join);
238 
239 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
240 			  const struct in6_addr *addr, unsigned int mode)
241 {
242 	return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
243 }
244 
245 /*
246  *	socket leave on multicast group
247  */
248 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
249 {
250 	struct ipv6_pinfo *np = inet6_sk(sk);
251 	struct ipv6_mc_socklist *mc_lst;
252 	struct ipv6_mc_socklist __rcu **lnk;
253 	struct net *net = sock_net(sk);
254 
255 	ASSERT_RTNL();
256 
257 	if (!ipv6_addr_is_multicast(addr))
258 		return -EINVAL;
259 
260 	for (lnk = &np->ipv6_mc_list;
261 	     (mc_lst = sock_dereference(*lnk, sk)) != NULL;
262 	      lnk = &mc_lst->next) {
263 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
264 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
265 			struct net_device *dev;
266 
267 			*lnk = mc_lst->next;
268 
269 			dev = __dev_get_by_index(net, mc_lst->ifindex);
270 			if (dev) {
271 				struct inet6_dev *idev = __in6_dev_get(dev);
272 
273 				ip6_mc_leave_src(sk, mc_lst, idev);
274 				if (idev)
275 					__ipv6_dev_mc_dec(idev, &mc_lst->addr);
276 			} else {
277 				ip6_mc_leave_src(sk, mc_lst, NULL);
278 			}
279 
280 			atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
281 			kfree_rcu(mc_lst, rcu);
282 			return 0;
283 		}
284 	}
285 
286 	return -EADDRNOTAVAIL;
287 }
288 EXPORT_SYMBOL(ipv6_sock_mc_drop);
289 
290 static struct inet6_dev *ip6_mc_find_dev_rtnl(struct net *net,
291 					      const struct in6_addr *group,
292 					      int ifindex)
293 {
294 	struct net_device *dev = NULL;
295 	struct inet6_dev *idev = NULL;
296 
297 	if (ifindex == 0) {
298 		struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
299 
300 		if (rt) {
301 			dev = rt->dst.dev;
302 			ip6_rt_put(rt);
303 		}
304 	} else {
305 		dev = __dev_get_by_index(net, ifindex);
306 	}
307 
308 	if (!dev)
309 		return NULL;
310 	idev = __in6_dev_get(dev);
311 	if (!idev)
312 		return NULL;
313 	if (idev->dead)
314 		return NULL;
315 	return idev;
316 }
317 
318 void __ipv6_sock_mc_close(struct sock *sk)
319 {
320 	struct ipv6_pinfo *np = inet6_sk(sk);
321 	struct ipv6_mc_socklist *mc_lst;
322 	struct net *net = sock_net(sk);
323 
324 	ASSERT_RTNL();
325 
326 	while ((mc_lst = sock_dereference(np->ipv6_mc_list, sk)) != NULL) {
327 		struct net_device *dev;
328 
329 		np->ipv6_mc_list = mc_lst->next;
330 
331 		dev = __dev_get_by_index(net, mc_lst->ifindex);
332 		if (dev) {
333 			struct inet6_dev *idev = __in6_dev_get(dev);
334 
335 			ip6_mc_leave_src(sk, mc_lst, idev);
336 			if (idev)
337 				__ipv6_dev_mc_dec(idev, &mc_lst->addr);
338 		} else {
339 			ip6_mc_leave_src(sk, mc_lst, NULL);
340 		}
341 
342 		atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
343 		kfree_rcu(mc_lst, rcu);
344 	}
345 }
346 
347 void ipv6_sock_mc_close(struct sock *sk)
348 {
349 	struct ipv6_pinfo *np = inet6_sk(sk);
350 
351 	if (!rcu_access_pointer(np->ipv6_mc_list))
352 		return;
353 
354 	rtnl_lock();
355 	lock_sock(sk);
356 	__ipv6_sock_mc_close(sk);
357 	release_sock(sk);
358 	rtnl_unlock();
359 }
360 
361 int ip6_mc_source(int add, int omode, struct sock *sk,
362 	struct group_source_req *pgsr)
363 {
364 	struct in6_addr *source, *group;
365 	struct ipv6_mc_socklist *pmc;
366 	struct inet6_dev *idev;
367 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
368 	struct ip6_sf_socklist *psl;
369 	struct net *net = sock_net(sk);
370 	int i, j, rv;
371 	int leavegroup = 0;
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_rtnl(net, group, pgsr->gsr_interface);
381 	if (!idev)
382 		return -ENODEV;
383 
384 	err = -EADDRNOTAVAIL;
385 
386 	mutex_lock(&idev->mc_lock);
387 	for_each_pmc_socklock(inet6, sk, pmc) {
388 		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
389 			continue;
390 		if (ipv6_addr_equal(&pmc->addr, group))
391 			break;
392 	}
393 	if (!pmc) {		/* must have a prior join */
394 		err = -EINVAL;
395 		goto done;
396 	}
397 	/* if a source filter was set, must be the same mode as before */
398 	if (rcu_access_pointer(pmc->sflist)) {
399 		if (pmc->sfmode != omode) {
400 			err = -EINVAL;
401 			goto done;
402 		}
403 	} else if (pmc->sfmode != omode) {
404 		/* allow mode switches for empty-set filters */
405 		ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
406 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
407 		pmc->sfmode = omode;
408 	}
409 
410 	psl = sock_dereference(pmc->sflist, sk);
411 	if (!add) {
412 		if (!psl)
413 			goto done;	/* err = -EADDRNOTAVAIL */
414 		rv = !0;
415 		for (i = 0; i < psl->sl_count; i++) {
416 			rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
417 			if (rv == 0)
418 				break;
419 		}
420 		if (rv)		/* source not found */
421 			goto done;	/* err = -EADDRNOTAVAIL */
422 
423 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
424 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
425 			leavegroup = 1;
426 			goto done;
427 		}
428 
429 		/* update the interface filter */
430 		ip6_mc_del_src(idev, group, omode, 1, source, 1);
431 
432 		for (j = i+1; j < psl->sl_count; j++)
433 			psl->sl_addr[j-1] = psl->sl_addr[j];
434 		psl->sl_count--;
435 		err = 0;
436 		goto done;
437 	}
438 	/* else, add a new source to the filter */
439 
440 	if (psl && psl->sl_count >= sysctl_mld_max_msf) {
441 		err = -ENOBUFS;
442 		goto done;
443 	}
444 	if (!psl || psl->sl_count == psl->sl_max) {
445 		struct ip6_sf_socklist *newpsl;
446 		int count = IP6_SFBLOCK;
447 
448 		if (psl)
449 			count += psl->sl_max;
450 		newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, count),
451 				      GFP_KERNEL);
452 		if (!newpsl) {
453 			err = -ENOBUFS;
454 			goto done;
455 		}
456 		newpsl->sl_max = count;
457 		newpsl->sl_count = count - IP6_SFBLOCK;
458 		if (psl) {
459 			for (i = 0; i < psl->sl_count; i++)
460 				newpsl->sl_addr[i] = psl->sl_addr[i];
461 			atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
462 				   &sk->sk_omem_alloc);
463 		}
464 		rcu_assign_pointer(pmc->sflist, newpsl);
465 		kfree_rcu(psl, rcu);
466 		psl = newpsl;
467 	}
468 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
469 	for (i = 0; i < psl->sl_count; i++) {
470 		rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
471 		if (rv == 0) /* There is an error in the address. */
472 			goto done;
473 	}
474 	for (j = psl->sl_count-1; j >= i; j--)
475 		psl->sl_addr[j+1] = psl->sl_addr[j];
476 	psl->sl_addr[i] = *source;
477 	psl->sl_count++;
478 	err = 0;
479 	/* update the interface list */
480 	ip6_mc_add_src(idev, group, omode, 1, source, 1);
481 done:
482 	mutex_unlock(&idev->mc_lock);
483 	if (leavegroup)
484 		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
485 	return err;
486 }
487 
488 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
489 		    struct sockaddr_storage *list)
490 {
491 	const struct in6_addr *group;
492 	struct ipv6_mc_socklist *pmc;
493 	struct inet6_dev *idev;
494 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
495 	struct ip6_sf_socklist *newpsl, *psl;
496 	struct net *net = sock_net(sk);
497 	int leavegroup = 0;
498 	int i, err;
499 
500 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
501 
502 	if (!ipv6_addr_is_multicast(group))
503 		return -EINVAL;
504 	if (gsf->gf_fmode != MCAST_INCLUDE &&
505 	    gsf->gf_fmode != MCAST_EXCLUDE)
506 		return -EINVAL;
507 
508 	idev = ip6_mc_find_dev_rtnl(net, group, gsf->gf_interface);
509 	if (!idev)
510 		return -ENODEV;
511 
512 	err = 0;
513 
514 	if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
515 		leavegroup = 1;
516 		goto done;
517 	}
518 
519 	for_each_pmc_socklock(inet6, sk, pmc) {
520 		if (pmc->ifindex != gsf->gf_interface)
521 			continue;
522 		if (ipv6_addr_equal(&pmc->addr, group))
523 			break;
524 	}
525 	if (!pmc) {		/* must have a prior join */
526 		err = -EINVAL;
527 		goto done;
528 	}
529 	if (gsf->gf_numsrc) {
530 		newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr,
531 						      gsf->gf_numsrc),
532 				      GFP_KERNEL);
533 		if (!newpsl) {
534 			err = -ENOBUFS;
535 			goto done;
536 		}
537 		newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
538 		for (i = 0; i < newpsl->sl_count; ++i, ++list) {
539 			struct sockaddr_in6 *psin6;
540 
541 			psin6 = (struct sockaddr_in6 *)list;
542 			newpsl->sl_addr[i] = psin6->sin6_addr;
543 		}
544 		mutex_lock(&idev->mc_lock);
545 		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
546 				     newpsl->sl_count, newpsl->sl_addr, 0);
547 		if (err) {
548 			mutex_unlock(&idev->mc_lock);
549 			sock_kfree_s(sk, newpsl, struct_size(newpsl, sl_addr,
550 							     newpsl->sl_max));
551 			goto done;
552 		}
553 		mutex_unlock(&idev->mc_lock);
554 	} else {
555 		newpsl = NULL;
556 		mutex_lock(&idev->mc_lock);
557 		ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
558 		mutex_unlock(&idev->mc_lock);
559 	}
560 
561 	mutex_lock(&idev->mc_lock);
562 	psl = sock_dereference(pmc->sflist, sk);
563 	if (psl) {
564 		ip6_mc_del_src(idev, group, pmc->sfmode,
565 			       psl->sl_count, psl->sl_addr, 0);
566 		atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
567 			   &sk->sk_omem_alloc);
568 	} else {
569 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
570 	}
571 	rcu_assign_pointer(pmc->sflist, newpsl);
572 	mutex_unlock(&idev->mc_lock);
573 	kfree_rcu(psl, rcu);
574 	pmc->sfmode = gsf->gf_fmode;
575 	err = 0;
576 done:
577 	if (leavegroup)
578 		err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
579 	return err;
580 }
581 
582 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
583 		  sockptr_t optval, size_t ss_offset)
584 {
585 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
586 	const struct in6_addr *group;
587 	struct ipv6_mc_socklist *pmc;
588 	struct ip6_sf_socklist *psl;
589 	unsigned int count;
590 	int i, copycount;
591 
592 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
593 
594 	if (!ipv6_addr_is_multicast(group))
595 		return -EINVAL;
596 
597 	/* changes to the ipv6_mc_list require the socket lock and
598 	 * rtnl lock. We have the socket lock, so reading the list is safe.
599 	 */
600 
601 	for_each_pmc_socklock(inet6, sk, pmc) {
602 		if (pmc->ifindex != gsf->gf_interface)
603 			continue;
604 		if (ipv6_addr_equal(group, &pmc->addr))
605 			break;
606 	}
607 	if (!pmc)		/* must have a prior join */
608 		return -EADDRNOTAVAIL;
609 
610 	gsf->gf_fmode = pmc->sfmode;
611 	psl = sock_dereference(pmc->sflist, sk);
612 	count = psl ? psl->sl_count : 0;
613 
614 	copycount = min(count, gsf->gf_numsrc);
615 	gsf->gf_numsrc = count;
616 	for (i = 0; i < copycount; i++) {
617 		struct sockaddr_in6 *psin6;
618 		struct sockaddr_storage ss;
619 
620 		psin6 = (struct sockaddr_in6 *)&ss;
621 		memset(&ss, 0, sizeof(ss));
622 		psin6->sin6_family = AF_INET6;
623 		psin6->sin6_addr = psl->sl_addr[i];
624 		if (copy_to_sockptr_offset(optval, ss_offset, &ss, sizeof(ss)))
625 			return -EFAULT;
626 		ss_offset += sizeof(ss);
627 	}
628 	return 0;
629 }
630 
631 bool inet6_mc_check(const struct sock *sk, const struct in6_addr *mc_addr,
632 		    const struct in6_addr *src_addr)
633 {
634 	const struct ipv6_pinfo *np = inet6_sk(sk);
635 	const struct ipv6_mc_socklist *mc;
636 	const struct ip6_sf_socklist *psl;
637 	bool rv = true;
638 
639 	rcu_read_lock();
640 	for_each_pmc_rcu(np, mc) {
641 		if (ipv6_addr_equal(&mc->addr, mc_addr))
642 			break;
643 	}
644 	if (!mc) {
645 		rcu_read_unlock();
646 		return inet6_test_bit(MC6_ALL, sk);
647 	}
648 	psl = rcu_dereference(mc->sflist);
649 	if (!psl) {
650 		rv = mc->sfmode == MCAST_EXCLUDE;
651 	} else {
652 		int i;
653 
654 		for (i = 0; i < psl->sl_count; i++) {
655 			if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
656 				break;
657 		}
658 		if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
659 			rv = false;
660 		if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
661 			rv = false;
662 	}
663 	rcu_read_unlock();
664 
665 	return rv;
666 }
667 
668 /* called with mc_lock */
669 static void igmp6_group_added(struct ifmcaddr6 *mc)
670 {
671 	struct net_device *dev = mc->idev->dev;
672 	char buf[MAX_ADDR_LEN];
673 
674 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
675 	    IPV6_ADDR_SCOPE_LINKLOCAL)
676 		return;
677 
678 	if (!(mc->mca_flags&MAF_LOADED)) {
679 		mc->mca_flags |= MAF_LOADED;
680 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
681 			dev_mc_add(dev, buf);
682 	}
683 
684 	if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
685 		return;
686 
687 	if (mld_in_v1_mode(mc->idev)) {
688 		igmp6_join_group(mc);
689 		return;
690 	}
691 	/* else v2 */
692 
693 	/* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
694 	 * should not send filter-mode change record as the mode
695 	 * should be from IN() to IN(A).
696 	 */
697 	if (mc->mca_sfmode == MCAST_EXCLUDE)
698 		mc->mca_crcount = mc->idev->mc_qrv;
699 
700 	mld_ifc_event(mc->idev);
701 }
702 
703 /* called with mc_lock */
704 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
705 {
706 	struct net_device *dev = mc->idev->dev;
707 	char buf[MAX_ADDR_LEN];
708 
709 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
710 	    IPV6_ADDR_SCOPE_LINKLOCAL)
711 		return;
712 
713 	if (mc->mca_flags&MAF_LOADED) {
714 		mc->mca_flags &= ~MAF_LOADED;
715 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
716 			dev_mc_del(dev, buf);
717 	}
718 
719 	if (mc->mca_flags & MAF_NOREPORT)
720 		return;
721 
722 	if (!mc->idev->dead)
723 		igmp6_leave_group(mc);
724 
725 	if (cancel_delayed_work(&mc->mca_work))
726 		refcount_dec(&mc->mca_refcnt);
727 }
728 
729 /*
730  * deleted ifmcaddr6 manipulation
731  * called with mc_lock
732  */
733 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
734 {
735 	struct ifmcaddr6 *pmc;
736 
737 	/* this is an "ifmcaddr6" for convenience; only the fields below
738 	 * are actually used. In particular, the refcnt and users are not
739 	 * used for management of the delete list. Using the same structure
740 	 * for deleted items allows change reports to use common code with
741 	 * non-deleted or query-response MCA's.
742 	 */
743 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
744 	if (!pmc)
745 		return;
746 
747 	pmc->idev = im->idev;
748 	in6_dev_hold(idev);
749 	pmc->mca_addr = im->mca_addr;
750 	pmc->mca_crcount = idev->mc_qrv;
751 	pmc->mca_sfmode = im->mca_sfmode;
752 	if (pmc->mca_sfmode == MCAST_INCLUDE) {
753 		struct ip6_sf_list *psf;
754 
755 		rcu_assign_pointer(pmc->mca_tomb,
756 				   mc_dereference(im->mca_tomb, idev));
757 		rcu_assign_pointer(pmc->mca_sources,
758 				   mc_dereference(im->mca_sources, idev));
759 		RCU_INIT_POINTER(im->mca_tomb, NULL);
760 		RCU_INIT_POINTER(im->mca_sources, NULL);
761 
762 		for_each_psf_mclock(pmc, psf)
763 			psf->sf_crcount = pmc->mca_crcount;
764 	}
765 
766 	rcu_assign_pointer(pmc->next, idev->mc_tomb);
767 	rcu_assign_pointer(idev->mc_tomb, pmc);
768 }
769 
770 /* called with mc_lock */
771 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
772 {
773 	struct ip6_sf_list *psf, *sources, *tomb;
774 	struct in6_addr *pmca = &im->mca_addr;
775 	struct ifmcaddr6 *pmc, *pmc_prev;
776 
777 	pmc_prev = NULL;
778 	for_each_mc_tomb(idev, pmc) {
779 		if (ipv6_addr_equal(&pmc->mca_addr, pmca))
780 			break;
781 		pmc_prev = pmc;
782 	}
783 	if (pmc) {
784 		if (pmc_prev)
785 			rcu_assign_pointer(pmc_prev->next, pmc->next);
786 		else
787 			rcu_assign_pointer(idev->mc_tomb, pmc->next);
788 	}
789 
790 	if (pmc) {
791 		im->idev = pmc->idev;
792 		if (im->mca_sfmode == MCAST_INCLUDE) {
793 			tomb = rcu_replace_pointer(im->mca_tomb,
794 						   mc_dereference(pmc->mca_tomb, pmc->idev),
795 						   lockdep_is_held(&im->idev->mc_lock));
796 			rcu_assign_pointer(pmc->mca_tomb, tomb);
797 
798 			sources = rcu_replace_pointer(im->mca_sources,
799 						      mc_dereference(pmc->mca_sources, pmc->idev),
800 						      lockdep_is_held(&im->idev->mc_lock));
801 			rcu_assign_pointer(pmc->mca_sources, sources);
802 			for_each_psf_mclock(im, psf)
803 				psf->sf_crcount = idev->mc_qrv;
804 		} else {
805 			im->mca_crcount = idev->mc_qrv;
806 		}
807 		in6_dev_put(pmc->idev);
808 		ip6_mc_clear_src(pmc);
809 		kfree_rcu(pmc, rcu);
810 	}
811 }
812 
813 /* called with mc_lock */
814 static void mld_clear_delrec(struct inet6_dev *idev)
815 {
816 	struct ifmcaddr6 *pmc, *nextpmc;
817 
818 	pmc = mc_dereference(idev->mc_tomb, idev);
819 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
820 
821 	for (; pmc; pmc = nextpmc) {
822 		nextpmc = mc_dereference(pmc->next, idev);
823 		ip6_mc_clear_src(pmc);
824 		in6_dev_put(pmc->idev);
825 		kfree_rcu(pmc, rcu);
826 	}
827 
828 	/* clear dead sources, too */
829 	for_each_mc_mclock(idev, pmc) {
830 		struct ip6_sf_list *psf, *psf_next;
831 
832 		psf = mc_dereference(pmc->mca_tomb, idev);
833 		RCU_INIT_POINTER(pmc->mca_tomb, NULL);
834 		for (; psf; psf = psf_next) {
835 			psf_next = mc_dereference(psf->sf_next, idev);
836 			kfree_rcu(psf, rcu);
837 		}
838 	}
839 }
840 
841 static void mld_clear_query(struct inet6_dev *idev)
842 {
843 	struct sk_buff *skb;
844 
845 	spin_lock_bh(&idev->mc_query_lock);
846 	while ((skb = __skb_dequeue(&idev->mc_query_queue)))
847 		kfree_skb(skb);
848 	spin_unlock_bh(&idev->mc_query_lock);
849 }
850 
851 static void mld_clear_report(struct inet6_dev *idev)
852 {
853 	struct sk_buff *skb;
854 
855 	spin_lock_bh(&idev->mc_report_lock);
856 	while ((skb = __skb_dequeue(&idev->mc_report_queue)))
857 		kfree_skb(skb);
858 	spin_unlock_bh(&idev->mc_report_lock);
859 }
860 
861 static void mca_get(struct ifmcaddr6 *mc)
862 {
863 	refcount_inc(&mc->mca_refcnt);
864 }
865 
866 static void ma_put(struct ifmcaddr6 *mc)
867 {
868 	if (refcount_dec_and_test(&mc->mca_refcnt)) {
869 		in6_dev_put(mc->idev);
870 		kfree_rcu(mc, rcu);
871 	}
872 }
873 
874 /* called with mc_lock */
875 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
876 				   const struct in6_addr *addr,
877 				   unsigned int mode)
878 {
879 	struct ifmcaddr6 *mc;
880 
881 	mc = kzalloc(sizeof(*mc), GFP_KERNEL);
882 	if (!mc)
883 		return NULL;
884 
885 	INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);
886 
887 	mc->mca_addr = *addr;
888 	mc->idev = idev; /* reference taken by caller */
889 	mc->mca_users = 1;
890 	/* mca_stamp should be updated upon changes */
891 	mc->mca_cstamp = mc->mca_tstamp = jiffies;
892 	refcount_set(&mc->mca_refcnt, 1);
893 
894 	mc->mca_sfmode = mode;
895 	mc->mca_sfcount[mode] = 1;
896 
897 	if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
898 	    IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
899 		mc->mca_flags |= MAF_NOREPORT;
900 
901 	return mc;
902 }
903 
904 /*
905  *	device multicast group inc (add if not found)
906  */
907 static int __ipv6_dev_mc_inc(struct net_device *dev,
908 			     const struct in6_addr *addr, unsigned int mode)
909 {
910 	struct ifmcaddr6 *mc;
911 	struct inet6_dev *idev;
912 
913 	ASSERT_RTNL();
914 
915 	/* we need to take a reference on idev */
916 	idev = in6_dev_get(dev);
917 
918 	if (!idev)
919 		return -EINVAL;
920 
921 	if (idev->dead) {
922 		in6_dev_put(idev);
923 		return -ENODEV;
924 	}
925 
926 	mutex_lock(&idev->mc_lock);
927 	for_each_mc_mclock(idev, mc) {
928 		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
929 			mc->mca_users++;
930 			ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
931 			mutex_unlock(&idev->mc_lock);
932 			in6_dev_put(idev);
933 			return 0;
934 		}
935 	}
936 
937 	mc = mca_alloc(idev, addr, mode);
938 	if (!mc) {
939 		mutex_unlock(&idev->mc_lock);
940 		in6_dev_put(idev);
941 		return -ENOMEM;
942 	}
943 
944 	rcu_assign_pointer(mc->next, idev->mc_list);
945 	rcu_assign_pointer(idev->mc_list, mc);
946 
947 	mca_get(mc);
948 
949 	mld_del_delrec(idev, mc);
950 	igmp6_group_added(mc);
951 	mutex_unlock(&idev->mc_lock);
952 	ma_put(mc);
953 	return 0;
954 }
955 
956 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
957 {
958 	return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
959 }
960 EXPORT_SYMBOL(ipv6_dev_mc_inc);
961 
962 /*
963  * device multicast group del
964  */
965 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
966 {
967 	struct ifmcaddr6 *ma, __rcu **map;
968 
969 	ASSERT_RTNL();
970 
971 	mutex_lock(&idev->mc_lock);
972 	for (map = &idev->mc_list;
973 	     (ma = mc_dereference(*map, idev));
974 	     map = &ma->next) {
975 		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
976 			if (--ma->mca_users == 0) {
977 				*map = ma->next;
978 
979 				igmp6_group_dropped(ma);
980 				ip6_mc_clear_src(ma);
981 				mutex_unlock(&idev->mc_lock);
982 
983 				ma_put(ma);
984 				return 0;
985 			}
986 			mutex_unlock(&idev->mc_lock);
987 			return 0;
988 		}
989 	}
990 
991 	mutex_unlock(&idev->mc_lock);
992 	return -ENOENT;
993 }
994 
995 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
996 {
997 	struct inet6_dev *idev;
998 	int err;
999 
1000 	ASSERT_RTNL();
1001 
1002 	idev = __in6_dev_get(dev);
1003 	if (!idev)
1004 		err = -ENODEV;
1005 	else
1006 		err = __ipv6_dev_mc_dec(idev, addr);
1007 
1008 	return err;
1009 }
1010 EXPORT_SYMBOL(ipv6_dev_mc_dec);
1011 
1012 /*
1013  *	check if the interface/address pair is valid
1014  */
1015 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
1016 			 const struct in6_addr *src_addr)
1017 {
1018 	struct inet6_dev *idev;
1019 	struct ifmcaddr6 *mc;
1020 	bool rv = false;
1021 
1022 	rcu_read_lock();
1023 	idev = __in6_dev_get(dev);
1024 	if (!idev)
1025 		goto unlock;
1026 	for_each_mc_rcu(idev, mc) {
1027 		if (ipv6_addr_equal(&mc->mca_addr, group))
1028 			break;
1029 	}
1030 	if (!mc)
1031 		goto unlock;
1032 	if (src_addr && !ipv6_addr_any(src_addr)) {
1033 		struct ip6_sf_list *psf;
1034 
1035 		for_each_psf_rcu(mc, psf) {
1036 			if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1037 				break;
1038 		}
1039 		if (psf)
1040 			rv = READ_ONCE(psf->sf_count[MCAST_INCLUDE]) ||
1041 				READ_ONCE(psf->sf_count[MCAST_EXCLUDE]) !=
1042 				READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]);
1043 		else
1044 			rv = READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]) != 0;
1045 	} else {
1046 		rv = true; /* don't filter unspecified source */
1047 	}
1048 unlock:
1049 	rcu_read_unlock();
1050 	return rv;
1051 }
1052 
1053 /* called with mc_lock */
1054 static void mld_gq_start_work(struct inet6_dev *idev)
1055 {
1056 	unsigned long tv = get_random_u32_below(idev->mc_maxdelay);
1057 
1058 	idev->mc_gq_running = 1;
1059 	if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2))
1060 		in6_dev_hold(idev);
1061 }
1062 
1063 /* called with mc_lock */
1064 static void mld_gq_stop_work(struct inet6_dev *idev)
1065 {
1066 	idev->mc_gq_running = 0;
1067 	if (cancel_delayed_work(&idev->mc_gq_work))
1068 		__in6_dev_put(idev);
1069 }
1070 
1071 /* called with mc_lock */
1072 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
1073 {
1074 	unsigned long tv = get_random_u32_below(delay);
1075 
1076 	if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2))
1077 		in6_dev_hold(idev);
1078 }
1079 
1080 /* called with mc_lock */
1081 static void mld_ifc_stop_work(struct inet6_dev *idev)
1082 {
1083 	idev->mc_ifc_count = 0;
1084 	if (cancel_delayed_work(&idev->mc_ifc_work))
1085 		__in6_dev_put(idev);
1086 }
1087 
1088 /* called with mc_lock */
1089 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
1090 {
1091 	unsigned long tv = get_random_u32_below(delay);
1092 
1093 	if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2))
1094 		in6_dev_hold(idev);
1095 }
1096 
1097 static void mld_dad_stop_work(struct inet6_dev *idev)
1098 {
1099 	if (cancel_delayed_work(&idev->mc_dad_work))
1100 		__in6_dev_put(idev);
1101 }
1102 
1103 static void mld_query_stop_work(struct inet6_dev *idev)
1104 {
1105 	spin_lock_bh(&idev->mc_query_lock);
1106 	if (cancel_delayed_work(&idev->mc_query_work))
1107 		__in6_dev_put(idev);
1108 	spin_unlock_bh(&idev->mc_query_lock);
1109 }
1110 
1111 static void mld_report_stop_work(struct inet6_dev *idev)
1112 {
1113 	if (cancel_delayed_work_sync(&idev->mc_report_work))
1114 		__in6_dev_put(idev);
1115 }
1116 
1117 /*
1118  * IGMP handling (alias multicast ICMPv6 messages)
1119  * called with mc_lock
1120  */
1121 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1122 {
1123 	unsigned long delay = resptime;
1124 
1125 	/* Do not start work for these addresses */
1126 	if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1127 	    IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1128 		return;
1129 
1130 	if (cancel_delayed_work(&ma->mca_work)) {
1131 		refcount_dec(&ma->mca_refcnt);
1132 		delay = ma->mca_work.timer.expires - jiffies;
1133 	}
1134 
1135 	if (delay >= resptime)
1136 		delay = get_random_u32_below(resptime);
1137 
1138 	if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
1139 		refcount_inc(&ma->mca_refcnt);
1140 	ma->mca_flags |= MAF_TIMER_RUNNING;
1141 }
1142 
1143 /* mark EXCLUDE-mode sources
1144  * called with mc_lock
1145  */
1146 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1147 			     const struct in6_addr *srcs)
1148 {
1149 	struct ip6_sf_list *psf;
1150 	int i, scount;
1151 
1152 	scount = 0;
1153 	for_each_psf_mclock(pmc, psf) {
1154 		if (scount == nsrcs)
1155 			break;
1156 		for (i = 0; i < nsrcs; i++) {
1157 			/* skip inactive filters */
1158 			if (psf->sf_count[MCAST_INCLUDE] ||
1159 			    pmc->mca_sfcount[MCAST_EXCLUDE] !=
1160 			    psf->sf_count[MCAST_EXCLUDE])
1161 				break;
1162 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1163 				scount++;
1164 				break;
1165 			}
1166 		}
1167 	}
1168 	pmc->mca_flags &= ~MAF_GSQUERY;
1169 	if (scount == nsrcs)	/* all sources excluded */
1170 		return false;
1171 	return true;
1172 }
1173 
1174 /* called with mc_lock */
1175 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1176 			    const struct in6_addr *srcs)
1177 {
1178 	struct ip6_sf_list *psf;
1179 	int i, scount;
1180 
1181 	if (pmc->mca_sfmode == MCAST_EXCLUDE)
1182 		return mld_xmarksources(pmc, nsrcs, srcs);
1183 
1184 	/* mark INCLUDE-mode sources */
1185 
1186 	scount = 0;
1187 	for_each_psf_mclock(pmc, psf) {
1188 		if (scount == nsrcs)
1189 			break;
1190 		for (i = 0; i < nsrcs; i++) {
1191 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1192 				psf->sf_gsresp = 1;
1193 				scount++;
1194 				break;
1195 			}
1196 		}
1197 	}
1198 	if (!scount) {
1199 		pmc->mca_flags &= ~MAF_GSQUERY;
1200 		return false;
1201 	}
1202 	pmc->mca_flags |= MAF_GSQUERY;
1203 	return true;
1204 }
1205 
1206 static int mld_force_mld_version(const struct inet6_dev *idev)
1207 {
1208 	const struct net *net = dev_net(idev->dev);
1209 	int all_force;
1210 
1211 	all_force = READ_ONCE(net->ipv6.devconf_all->force_mld_version);
1212 	/* Normally, both are 0 here. If enforcement to a particular is
1213 	 * being used, individual device enforcement will have a lower
1214 	 * precedence over 'all' device (.../conf/all/force_mld_version).
1215 	 */
1216 	return all_force ?: READ_ONCE(idev->cnf.force_mld_version);
1217 }
1218 
1219 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1220 {
1221 	return mld_force_mld_version(idev) == 2;
1222 }
1223 
1224 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1225 {
1226 	return mld_force_mld_version(idev) == 1;
1227 }
1228 
1229 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1230 {
1231 	if (mld_in_v2_mode_only(idev))
1232 		return false;
1233 	if (mld_in_v1_mode_only(idev))
1234 		return true;
1235 	if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1236 		return true;
1237 
1238 	return false;
1239 }
1240 
1241 static void mld_set_v1_mode(struct inet6_dev *idev)
1242 {
1243 	/* RFC3810, relevant sections:
1244 	 *  - 9.1. Robustness Variable
1245 	 *  - 9.2. Query Interval
1246 	 *  - 9.3. Query Response Interval
1247 	 *  - 9.12. Older Version Querier Present Timeout
1248 	 */
1249 	unsigned long switchback;
1250 
1251 	switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1252 
1253 	idev->mc_v1_seen = jiffies + switchback;
1254 }
1255 
1256 static void mld_update_qrv(struct inet6_dev *idev,
1257 			   const struct mld2_query *mlh2)
1258 {
1259 	/* RFC3810, relevant sections:
1260 	 *  - 5.1.8. QRV (Querier's Robustness Variable)
1261 	 *  - 9.1. Robustness Variable
1262 	 */
1263 
1264 	/* The value of the Robustness Variable MUST NOT be zero,
1265 	 * and SHOULD NOT be one. Catch this here if we ever run
1266 	 * into such a case in future.
1267 	 */
1268 	const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1269 	WARN_ON(idev->mc_qrv == 0);
1270 
1271 	if (mlh2->mld2q_qrv > 0)
1272 		idev->mc_qrv = mlh2->mld2q_qrv;
1273 
1274 	if (unlikely(idev->mc_qrv < min_qrv)) {
1275 		net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1276 				     idev->mc_qrv, min_qrv);
1277 		idev->mc_qrv = min_qrv;
1278 	}
1279 }
1280 
1281 static void mld_update_qi(struct inet6_dev *idev,
1282 			  const struct mld2_query *mlh2)
1283 {
1284 	/* RFC3810, relevant sections:
1285 	 *  - 5.1.9. QQIC (Querier's Query Interval Code)
1286 	 *  - 9.2. Query Interval
1287 	 *  - 9.12. Older Version Querier Present Timeout
1288 	 *    (the [Query Interval] in the last Query received)
1289 	 */
1290 	unsigned long mc_qqi;
1291 
1292 	if (mlh2->mld2q_qqic < 128) {
1293 		mc_qqi = mlh2->mld2q_qqic;
1294 	} else {
1295 		unsigned long mc_man, mc_exp;
1296 
1297 		mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1298 		mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1299 
1300 		mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1301 	}
1302 
1303 	idev->mc_qi = mc_qqi * HZ;
1304 }
1305 
1306 static void mld_update_qri(struct inet6_dev *idev,
1307 			   const struct mld2_query *mlh2)
1308 {
1309 	/* RFC3810, relevant sections:
1310 	 *  - 5.1.3. Maximum Response Code
1311 	 *  - 9.3. Query Response Interval
1312 	 */
1313 	idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1314 }
1315 
1316 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1317 			  unsigned long *max_delay, bool v1_query)
1318 {
1319 	unsigned long mldv1_md;
1320 
1321 	/* Ignore v1 queries */
1322 	if (mld_in_v2_mode_only(idev))
1323 		return -EINVAL;
1324 
1325 	mldv1_md = ntohs(mld->mld_maxdelay);
1326 
1327 	/* When in MLDv1 fallback and a MLDv2 router start-up being
1328 	 * unaware of current MLDv1 operation, the MRC == MRD mapping
1329 	 * only works when the exponential algorithm is not being
1330 	 * used (as MLDv1 is unaware of such things).
1331 	 *
1332 	 * According to the RFC author, the MLDv2 implementations
1333 	 * he's aware of all use a MRC < 32768 on start up queries.
1334 	 *
1335 	 * Thus, should we *ever* encounter something else larger
1336 	 * than that, just assume the maximum possible within our
1337 	 * reach.
1338 	 */
1339 	if (!v1_query)
1340 		mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1341 
1342 	*max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1343 
1344 	/* MLDv1 router present: we need to go into v1 mode *only*
1345 	 * when an MLDv1 query is received as per section 9.12. of
1346 	 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1347 	 * queries MUST be of exactly 24 octets.
1348 	 */
1349 	if (v1_query)
1350 		mld_set_v1_mode(idev);
1351 
1352 	/* cancel MLDv2 report work */
1353 	mld_gq_stop_work(idev);
1354 	/* cancel the interface change work */
1355 	mld_ifc_stop_work(idev);
1356 	/* clear deleted report items */
1357 	mld_clear_delrec(idev);
1358 
1359 	return 0;
1360 }
1361 
1362 static void mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1363 			   unsigned long *max_delay)
1364 {
1365 	*max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1366 
1367 	mld_update_qrv(idev, mld);
1368 	mld_update_qi(idev, mld);
1369 	mld_update_qri(idev, mld);
1370 
1371 	idev->mc_maxdelay = *max_delay;
1372 
1373 	return;
1374 }
1375 
1376 /* called with rcu_read_lock() */
1377 void igmp6_event_query(struct sk_buff *skb)
1378 {
1379 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1380 
1381 	if (!idev || idev->dead)
1382 		goto out;
1383 
1384 	spin_lock_bh(&idev->mc_query_lock);
1385 	if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) {
1386 		__skb_queue_tail(&idev->mc_query_queue, skb);
1387 		if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0))
1388 			in6_dev_hold(idev);
1389 		skb = NULL;
1390 	}
1391 	spin_unlock_bh(&idev->mc_query_lock);
1392 out:
1393 	kfree_skb(skb);
1394 }
1395 
1396 static void __mld_query_work(struct sk_buff *skb)
1397 {
1398 	struct mld2_query *mlh2 = NULL;
1399 	const struct in6_addr *group;
1400 	unsigned long max_delay;
1401 	struct inet6_dev *idev;
1402 	struct ifmcaddr6 *ma;
1403 	struct mld_msg *mld;
1404 	int group_type;
1405 	int mark = 0;
1406 	int len, err;
1407 
1408 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1409 		goto kfree_skb;
1410 
1411 	/* compute payload length excluding extension headers */
1412 	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1413 	len -= skb_network_header_len(skb);
1414 
1415 	/* RFC3810 6.2
1416 	 * Upon reception of an MLD message that contains a Query, the node
1417 	 * checks if the source address of the message is a valid link-local
1418 	 * address, if the Hop Limit is set to 1, and if the Router Alert
1419 	 * option is present in the Hop-By-Hop Options header of the IPv6
1420 	 * packet.  If any of these checks fails, the packet is dropped.
1421 	 */
1422 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1423 	    ipv6_hdr(skb)->hop_limit != 1 ||
1424 	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1425 	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1426 		goto kfree_skb;
1427 
1428 	idev = in6_dev_get(skb->dev);
1429 	if (!idev)
1430 		goto kfree_skb;
1431 
1432 	mld = (struct mld_msg *)icmp6_hdr(skb);
1433 	group = &mld->mld_mca;
1434 	group_type = ipv6_addr_type(group);
1435 
1436 	if (group_type != IPV6_ADDR_ANY &&
1437 	    !(group_type&IPV6_ADDR_MULTICAST))
1438 		goto out;
1439 
1440 	if (len < MLD_V1_QUERY_LEN) {
1441 		goto out;
1442 	} else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1443 		err = mld_process_v1(idev, mld, &max_delay,
1444 				     len == MLD_V1_QUERY_LEN);
1445 		if (err < 0)
1446 			goto out;
1447 	} else if (len >= MLD_V2_QUERY_LEN_MIN) {
1448 		int srcs_offset = sizeof(struct mld2_query) -
1449 				  sizeof(struct icmp6hdr);
1450 
1451 		if (!pskb_may_pull(skb, srcs_offset))
1452 			goto out;
1453 
1454 		mlh2 = (struct mld2_query *)skb_transport_header(skb);
1455 
1456 		mld_process_v2(idev, mlh2, &max_delay);
1457 
1458 		if (group_type == IPV6_ADDR_ANY) { /* general query */
1459 			if (mlh2->mld2q_nsrcs)
1460 				goto out; /* no sources allowed */
1461 
1462 			mld_gq_start_work(idev);
1463 			goto out;
1464 		}
1465 		/* mark sources to include, if group & source-specific */
1466 		if (mlh2->mld2q_nsrcs != 0) {
1467 			if (!pskb_may_pull(skb, srcs_offset +
1468 			    ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1469 				goto out;
1470 
1471 			mlh2 = (struct mld2_query *)skb_transport_header(skb);
1472 			mark = 1;
1473 		}
1474 	} else {
1475 		goto out;
1476 	}
1477 
1478 	if (group_type == IPV6_ADDR_ANY) {
1479 		for_each_mc_mclock(idev, ma) {
1480 			igmp6_group_queried(ma, max_delay);
1481 		}
1482 	} else {
1483 		for_each_mc_mclock(idev, ma) {
1484 			if (!ipv6_addr_equal(group, &ma->mca_addr))
1485 				continue;
1486 			if (ma->mca_flags & MAF_TIMER_RUNNING) {
1487 				/* gsquery <- gsquery && mark */
1488 				if (!mark)
1489 					ma->mca_flags &= ~MAF_GSQUERY;
1490 			} else {
1491 				/* gsquery <- mark */
1492 				if (mark)
1493 					ma->mca_flags |= MAF_GSQUERY;
1494 				else
1495 					ma->mca_flags &= ~MAF_GSQUERY;
1496 			}
1497 			if (!(ma->mca_flags & MAF_GSQUERY) ||
1498 			    mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1499 				igmp6_group_queried(ma, max_delay);
1500 			break;
1501 		}
1502 	}
1503 
1504 out:
1505 	in6_dev_put(idev);
1506 kfree_skb:
1507 	consume_skb(skb);
1508 }
1509 
1510 static void mld_query_work(struct work_struct *work)
1511 {
1512 	struct inet6_dev *idev = container_of(to_delayed_work(work),
1513 					      struct inet6_dev,
1514 					      mc_query_work);
1515 	struct sk_buff_head q;
1516 	struct sk_buff *skb;
1517 	bool rework = false;
1518 	int cnt = 0;
1519 
1520 	skb_queue_head_init(&q);
1521 
1522 	spin_lock_bh(&idev->mc_query_lock);
1523 	while ((skb = __skb_dequeue(&idev->mc_query_queue))) {
1524 		__skb_queue_tail(&q, skb);
1525 
1526 		if (++cnt >= MLD_MAX_QUEUE) {
1527 			rework = true;
1528 			break;
1529 		}
1530 	}
1531 	spin_unlock_bh(&idev->mc_query_lock);
1532 
1533 	mutex_lock(&idev->mc_lock);
1534 	while ((skb = __skb_dequeue(&q)))
1535 		__mld_query_work(skb);
1536 	mutex_unlock(&idev->mc_lock);
1537 
1538 	if (rework && queue_delayed_work(mld_wq, &idev->mc_query_work, 0))
1539 		return;
1540 
1541 	in6_dev_put(idev);
1542 }
1543 
1544 /* called with rcu_read_lock() */
1545 void igmp6_event_report(struct sk_buff *skb)
1546 {
1547 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1548 
1549 	if (!idev || idev->dead)
1550 		goto out;
1551 
1552 	spin_lock_bh(&idev->mc_report_lock);
1553 	if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) {
1554 		__skb_queue_tail(&idev->mc_report_queue, skb);
1555 		if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0))
1556 			in6_dev_hold(idev);
1557 		skb = NULL;
1558 	}
1559 	spin_unlock_bh(&idev->mc_report_lock);
1560 out:
1561 	kfree_skb(skb);
1562 }
1563 
1564 static void __mld_report_work(struct sk_buff *skb)
1565 {
1566 	struct inet6_dev *idev;
1567 	struct ifmcaddr6 *ma;
1568 	struct mld_msg *mld;
1569 	int addr_type;
1570 
1571 	/* Our own report looped back. Ignore it. */
1572 	if (skb->pkt_type == PACKET_LOOPBACK)
1573 		goto kfree_skb;
1574 
1575 	/* send our report if the MC router may not have heard this report */
1576 	if (skb->pkt_type != PACKET_MULTICAST &&
1577 	    skb->pkt_type != PACKET_BROADCAST)
1578 		goto kfree_skb;
1579 
1580 	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1581 		goto kfree_skb;
1582 
1583 	mld = (struct mld_msg *)icmp6_hdr(skb);
1584 
1585 	/* Drop reports with not link local source */
1586 	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1587 	if (addr_type != IPV6_ADDR_ANY &&
1588 	    !(addr_type&IPV6_ADDR_LINKLOCAL))
1589 		goto kfree_skb;
1590 
1591 	idev = in6_dev_get(skb->dev);
1592 	if (!idev)
1593 		goto kfree_skb;
1594 
1595 	/*
1596 	 *	Cancel the work for this group
1597 	 */
1598 
1599 	for_each_mc_mclock(idev, ma) {
1600 		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1601 			if (cancel_delayed_work(&ma->mca_work))
1602 				refcount_dec(&ma->mca_refcnt);
1603 			ma->mca_flags &= ~(MAF_LAST_REPORTER |
1604 					   MAF_TIMER_RUNNING);
1605 			break;
1606 		}
1607 	}
1608 
1609 	in6_dev_put(idev);
1610 kfree_skb:
1611 	consume_skb(skb);
1612 }
1613 
1614 static void mld_report_work(struct work_struct *work)
1615 {
1616 	struct inet6_dev *idev = container_of(to_delayed_work(work),
1617 					      struct inet6_dev,
1618 					      mc_report_work);
1619 	struct sk_buff_head q;
1620 	struct sk_buff *skb;
1621 	bool rework = false;
1622 	int cnt = 0;
1623 
1624 	skb_queue_head_init(&q);
1625 	spin_lock_bh(&idev->mc_report_lock);
1626 	while ((skb = __skb_dequeue(&idev->mc_report_queue))) {
1627 		__skb_queue_tail(&q, skb);
1628 
1629 		if (++cnt >= MLD_MAX_QUEUE) {
1630 			rework = true;
1631 			break;
1632 		}
1633 	}
1634 	spin_unlock_bh(&idev->mc_report_lock);
1635 
1636 	mutex_lock(&idev->mc_lock);
1637 	while ((skb = __skb_dequeue(&q)))
1638 		__mld_report_work(skb);
1639 	mutex_unlock(&idev->mc_lock);
1640 
1641 	if (rework && queue_delayed_work(mld_wq, &idev->mc_report_work, 0))
1642 		return;
1643 
1644 	in6_dev_put(idev);
1645 }
1646 
1647 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1648 		  int gdeleted, int sdeleted)
1649 {
1650 	switch (type) {
1651 	case MLD2_MODE_IS_INCLUDE:
1652 	case MLD2_MODE_IS_EXCLUDE:
1653 		if (gdeleted || sdeleted)
1654 			return false;
1655 		if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1656 			if (pmc->mca_sfmode == MCAST_INCLUDE)
1657 				return true;
1658 			/* don't include if this source is excluded
1659 			 * in all filters
1660 			 */
1661 			if (psf->sf_count[MCAST_INCLUDE])
1662 				return type == MLD2_MODE_IS_INCLUDE;
1663 			return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1664 				psf->sf_count[MCAST_EXCLUDE];
1665 		}
1666 		return false;
1667 	case MLD2_CHANGE_TO_INCLUDE:
1668 		if (gdeleted || sdeleted)
1669 			return false;
1670 		return psf->sf_count[MCAST_INCLUDE] != 0;
1671 	case MLD2_CHANGE_TO_EXCLUDE:
1672 		if (gdeleted || sdeleted)
1673 			return false;
1674 		if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1675 		    psf->sf_count[MCAST_INCLUDE])
1676 			return false;
1677 		return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1678 			psf->sf_count[MCAST_EXCLUDE];
1679 	case MLD2_ALLOW_NEW_SOURCES:
1680 		if (gdeleted || !psf->sf_crcount)
1681 			return false;
1682 		return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1683 	case MLD2_BLOCK_OLD_SOURCES:
1684 		if (pmc->mca_sfmode == MCAST_INCLUDE)
1685 			return gdeleted || (psf->sf_crcount && sdeleted);
1686 		return psf->sf_crcount && !gdeleted && !sdeleted;
1687 	}
1688 	return false;
1689 }
1690 
1691 static int
1692 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1693 {
1694 	struct ip6_sf_list *psf;
1695 	int scount = 0;
1696 
1697 	for_each_psf_mclock(pmc, psf) {
1698 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1699 			continue;
1700 		scount++;
1701 	}
1702 	return scount;
1703 }
1704 
1705 static void ip6_mc_hdr(const struct sock *sk, struct sk_buff *skb,
1706 		       struct net_device *dev, const struct in6_addr *saddr,
1707 		       const struct in6_addr *daddr, int proto, int len)
1708 {
1709 	struct ipv6hdr *hdr;
1710 
1711 	skb->protocol = htons(ETH_P_IPV6);
1712 	skb->dev = dev;
1713 
1714 	skb_reset_network_header(skb);
1715 	skb_put(skb, sizeof(struct ipv6hdr));
1716 	hdr = ipv6_hdr(skb);
1717 
1718 	ip6_flow_hdr(hdr, 0, 0);
1719 
1720 	hdr->payload_len = htons(len);
1721 	hdr->nexthdr = proto;
1722 	hdr->hop_limit = READ_ONCE(inet6_sk(sk)->hop_limit);
1723 
1724 	hdr->saddr = *saddr;
1725 	hdr->daddr = *daddr;
1726 }
1727 
1728 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1729 {
1730 	u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
1731 		     2, 0, 0, IPV6_TLV_PADN, 0 };
1732 	struct net_device *dev = idev->dev;
1733 	int hlen = LL_RESERVED_SPACE(dev);
1734 	int tlen = dev->needed_tailroom;
1735 	struct net *net = dev_net(dev);
1736 	const struct in6_addr *saddr;
1737 	struct in6_addr addr_buf;
1738 	struct mld2_report *pmr;
1739 	struct sk_buff *skb;
1740 	unsigned int size;
1741 	struct sock *sk;
1742 	int err;
1743 
1744 	sk = net->ipv6.igmp_sk;
1745 	/* we assume size > sizeof(ra) here
1746 	 * Also try to not allocate high-order pages for big MTU
1747 	 */
1748 	size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen;
1749 	skb = sock_alloc_send_skb(sk, size, 1, &err);
1750 	if (!skb)
1751 		return NULL;
1752 
1753 	skb->priority = TC_PRIO_CONTROL;
1754 	skb_reserve(skb, hlen);
1755 	skb_tailroom_reserve(skb, mtu, tlen);
1756 
1757 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1758 		/* <draft-ietf-magma-mld-source-05.txt>:
1759 		 * use unspecified address as the source address
1760 		 * when a valid link-local address is not available.
1761 		 */
1762 		saddr = &in6addr_any;
1763 	} else
1764 		saddr = &addr_buf;
1765 
1766 	ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1767 
1768 	skb_put_data(skb, ra, sizeof(ra));
1769 
1770 	skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1771 	skb_put(skb, sizeof(*pmr));
1772 	pmr = (struct mld2_report *)skb_transport_header(skb);
1773 	pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1774 	pmr->mld2r_resv1 = 0;
1775 	pmr->mld2r_cksum = 0;
1776 	pmr->mld2r_resv2 = 0;
1777 	pmr->mld2r_ngrec = 0;
1778 	return skb;
1779 }
1780 
1781 static void mld_sendpack(struct sk_buff *skb)
1782 {
1783 	struct ipv6hdr *pip6 = ipv6_hdr(skb);
1784 	struct mld2_report *pmr =
1785 			      (struct mld2_report *)skb_transport_header(skb);
1786 	int payload_len, mldlen;
1787 	struct inet6_dev *idev;
1788 	struct net *net = dev_net(skb->dev);
1789 	int err;
1790 	struct flowi6 fl6;
1791 	struct dst_entry *dst;
1792 
1793 	rcu_read_lock();
1794 	idev = __in6_dev_get(skb->dev);
1795 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
1796 
1797 	payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1798 		sizeof(*pip6);
1799 	mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1800 	pip6->payload_len = htons(payload_len);
1801 
1802 	pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1803 					   IPPROTO_ICMPV6,
1804 					   csum_partial(skb_transport_header(skb),
1805 							mldlen, 0));
1806 
1807 	icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1808 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1809 			 skb->dev->ifindex);
1810 	dst = icmp6_dst_alloc(skb->dev, &fl6);
1811 
1812 	err = 0;
1813 	if (IS_ERR(dst)) {
1814 		err = PTR_ERR(dst);
1815 		dst = NULL;
1816 	}
1817 	skb_dst_set(skb, dst);
1818 	if (err)
1819 		goto err_out;
1820 
1821 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1822 		      net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1823 		      dst_output);
1824 out:
1825 	if (!err) {
1826 		ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1827 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1828 	} else {
1829 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1830 	}
1831 
1832 	rcu_read_unlock();
1833 	return;
1834 
1835 err_out:
1836 	kfree_skb(skb);
1837 	goto out;
1838 }
1839 
1840 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1841 {
1842 	return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1843 }
1844 
1845 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1846 	int type, struct mld2_grec **ppgr, unsigned int mtu)
1847 {
1848 	struct mld2_report *pmr;
1849 	struct mld2_grec *pgr;
1850 
1851 	if (!skb) {
1852 		skb = mld_newpack(pmc->idev, mtu);
1853 		if (!skb)
1854 			return NULL;
1855 	}
1856 	pgr = skb_put(skb, sizeof(struct mld2_grec));
1857 	pgr->grec_type = type;
1858 	pgr->grec_auxwords = 0;
1859 	pgr->grec_nsrcs = 0;
1860 	pgr->grec_mca = pmc->mca_addr;	/* structure copy */
1861 	pmr = (struct mld2_report *)skb_transport_header(skb);
1862 	pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1863 	*ppgr = pgr;
1864 	return skb;
1865 }
1866 
1867 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
1868 
1869 /* called with mc_lock */
1870 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1871 				int type, int gdeleted, int sdeleted,
1872 				int crsend)
1873 {
1874 	struct ip6_sf_list *psf, *psf_prev, *psf_next;
1875 	int scount, stotal, first, isquery, truncate;
1876 	struct ip6_sf_list __rcu **psf_list;
1877 	struct inet6_dev *idev = pmc->idev;
1878 	struct net_device *dev = idev->dev;
1879 	struct mld2_grec *pgr = NULL;
1880 	struct mld2_report *pmr;
1881 	unsigned int mtu;
1882 
1883 	if (pmc->mca_flags & MAF_NOREPORT)
1884 		return skb;
1885 
1886 	mtu = READ_ONCE(dev->mtu);
1887 	if (mtu < IPV6_MIN_MTU)
1888 		return skb;
1889 
1890 	isquery = type == MLD2_MODE_IS_INCLUDE ||
1891 		  type == MLD2_MODE_IS_EXCLUDE;
1892 	truncate = type == MLD2_MODE_IS_EXCLUDE ||
1893 		    type == MLD2_CHANGE_TO_EXCLUDE;
1894 
1895 	stotal = scount = 0;
1896 
1897 	psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1898 
1899 	if (!rcu_access_pointer(*psf_list))
1900 		goto empty_source;
1901 
1902 	pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1903 
1904 	/* EX and TO_EX get a fresh packet, if needed */
1905 	if (truncate) {
1906 		if (pmr && pmr->mld2r_ngrec &&
1907 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1908 			if (skb)
1909 				mld_sendpack(skb);
1910 			skb = mld_newpack(idev, mtu);
1911 		}
1912 	}
1913 	first = 1;
1914 	psf_prev = NULL;
1915 	for (psf = mc_dereference(*psf_list, idev);
1916 	     psf;
1917 	     psf = psf_next) {
1918 		struct in6_addr *psrc;
1919 
1920 		psf_next = mc_dereference(psf->sf_next, idev);
1921 
1922 		if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1923 			psf_prev = psf;
1924 			continue;
1925 		}
1926 
1927 		/* Based on RFC3810 6.1. Should not send source-list change
1928 		 * records when there is a filter mode change.
1929 		 */
1930 		if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1931 		     (!gdeleted && pmc->mca_crcount)) &&
1932 		    (type == MLD2_ALLOW_NEW_SOURCES ||
1933 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1934 			goto decrease_sf_crcount;
1935 
1936 		/* clear marks on query responses */
1937 		if (isquery)
1938 			psf->sf_gsresp = 0;
1939 
1940 		if (AVAILABLE(skb) < sizeof(*psrc) +
1941 		    first*sizeof(struct mld2_grec)) {
1942 			if (truncate && !first)
1943 				break;	 /* truncate these */
1944 			if (pgr)
1945 				pgr->grec_nsrcs = htons(scount);
1946 			if (skb)
1947 				mld_sendpack(skb);
1948 			skb = mld_newpack(idev, mtu);
1949 			first = 1;
1950 			scount = 0;
1951 		}
1952 		if (first) {
1953 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
1954 			first = 0;
1955 		}
1956 		if (!skb)
1957 			return NULL;
1958 		psrc = skb_put(skb, sizeof(*psrc));
1959 		*psrc = psf->sf_addr;
1960 		scount++; stotal++;
1961 		if ((type == MLD2_ALLOW_NEW_SOURCES ||
1962 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1963 decrease_sf_crcount:
1964 			psf->sf_crcount--;
1965 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1966 				if (psf_prev)
1967 					rcu_assign_pointer(psf_prev->sf_next,
1968 							   mc_dereference(psf->sf_next, idev));
1969 				else
1970 					rcu_assign_pointer(*psf_list,
1971 							   mc_dereference(psf->sf_next, idev));
1972 				kfree_rcu(psf, rcu);
1973 				continue;
1974 			}
1975 		}
1976 		psf_prev = psf;
1977 	}
1978 
1979 empty_source:
1980 	if (!stotal) {
1981 		if (type == MLD2_ALLOW_NEW_SOURCES ||
1982 		    type == MLD2_BLOCK_OLD_SOURCES)
1983 			return skb;
1984 		if (pmc->mca_crcount || isquery || crsend) {
1985 			/* make sure we have room for group header */
1986 			if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1987 				mld_sendpack(skb);
1988 				skb = NULL; /* add_grhead will get a new one */
1989 			}
1990 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
1991 		}
1992 	}
1993 	if (pgr)
1994 		pgr->grec_nsrcs = htons(scount);
1995 
1996 	if (isquery)
1997 		pmc->mca_flags &= ~MAF_GSQUERY;	/* clear query state */
1998 	return skb;
1999 }
2000 
2001 /* called with mc_lock */
2002 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
2003 {
2004 	struct sk_buff *skb = NULL;
2005 	int type;
2006 
2007 	if (!pmc) {
2008 		for_each_mc_mclock(idev, pmc) {
2009 			if (pmc->mca_flags & MAF_NOREPORT)
2010 				continue;
2011 			if (pmc->mca_sfcount[MCAST_EXCLUDE])
2012 				type = MLD2_MODE_IS_EXCLUDE;
2013 			else
2014 				type = MLD2_MODE_IS_INCLUDE;
2015 			skb = add_grec(skb, pmc, type, 0, 0, 0);
2016 		}
2017 	} else {
2018 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2019 			type = MLD2_MODE_IS_EXCLUDE;
2020 		else
2021 			type = MLD2_MODE_IS_INCLUDE;
2022 		skb = add_grec(skb, pmc, type, 0, 0, 0);
2023 	}
2024 	if (skb)
2025 		mld_sendpack(skb);
2026 }
2027 
2028 /*
2029  * remove zero-count source records from a source filter list
2030  * called with mc_lock
2031  */
2032 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
2033 {
2034 	struct ip6_sf_list *psf_prev, *psf_next, *psf;
2035 
2036 	psf_prev = NULL;
2037 	for (psf = mc_dereference(*ppsf, idev);
2038 	     psf;
2039 	     psf = psf_next) {
2040 		psf_next = mc_dereference(psf->sf_next, idev);
2041 		if (psf->sf_crcount == 0) {
2042 			if (psf_prev)
2043 				rcu_assign_pointer(psf_prev->sf_next,
2044 						   mc_dereference(psf->sf_next, idev));
2045 			else
2046 				rcu_assign_pointer(*ppsf,
2047 						   mc_dereference(psf->sf_next, idev));
2048 			kfree_rcu(psf, rcu);
2049 		} else {
2050 			psf_prev = psf;
2051 		}
2052 	}
2053 }
2054 
2055 /* called with mc_lock */
2056 static void mld_send_cr(struct inet6_dev *idev)
2057 {
2058 	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
2059 	struct sk_buff *skb = NULL;
2060 	int type, dtype;
2061 
2062 	/* deleted MCA's */
2063 	pmc_prev = NULL;
2064 	for (pmc = mc_dereference(idev->mc_tomb, idev);
2065 	     pmc;
2066 	     pmc = pmc_next) {
2067 		pmc_next = mc_dereference(pmc->next, idev);
2068 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
2069 			type = MLD2_BLOCK_OLD_SOURCES;
2070 			dtype = MLD2_BLOCK_OLD_SOURCES;
2071 			skb = add_grec(skb, pmc, type, 1, 0, 0);
2072 			skb = add_grec(skb, pmc, dtype, 1, 1, 0);
2073 		}
2074 		if (pmc->mca_crcount) {
2075 			if (pmc->mca_sfmode == MCAST_EXCLUDE) {
2076 				type = MLD2_CHANGE_TO_INCLUDE;
2077 				skb = add_grec(skb, pmc, type, 1, 0, 0);
2078 			}
2079 			pmc->mca_crcount--;
2080 			if (pmc->mca_crcount == 0) {
2081 				mld_clear_zeros(&pmc->mca_tomb, idev);
2082 				mld_clear_zeros(&pmc->mca_sources, idev);
2083 			}
2084 		}
2085 		if (pmc->mca_crcount == 0 &&
2086 		    !rcu_access_pointer(pmc->mca_tomb) &&
2087 		    !rcu_access_pointer(pmc->mca_sources)) {
2088 			if (pmc_prev)
2089 				rcu_assign_pointer(pmc_prev->next, pmc_next);
2090 			else
2091 				rcu_assign_pointer(idev->mc_tomb, pmc_next);
2092 			in6_dev_put(pmc->idev);
2093 			kfree_rcu(pmc, rcu);
2094 		} else
2095 			pmc_prev = pmc;
2096 	}
2097 
2098 	/* change recs */
2099 	for_each_mc_mclock(idev, pmc) {
2100 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2101 			type = MLD2_BLOCK_OLD_SOURCES;
2102 			dtype = MLD2_ALLOW_NEW_SOURCES;
2103 		} else {
2104 			type = MLD2_ALLOW_NEW_SOURCES;
2105 			dtype = MLD2_BLOCK_OLD_SOURCES;
2106 		}
2107 		skb = add_grec(skb, pmc, type, 0, 0, 0);
2108 		skb = add_grec(skb, pmc, dtype, 0, 1, 0);	/* deleted sources */
2109 
2110 		/* filter mode changes */
2111 		if (pmc->mca_crcount) {
2112 			if (pmc->mca_sfmode == MCAST_EXCLUDE)
2113 				type = MLD2_CHANGE_TO_EXCLUDE;
2114 			else
2115 				type = MLD2_CHANGE_TO_INCLUDE;
2116 			skb = add_grec(skb, pmc, type, 0, 0, 0);
2117 			pmc->mca_crcount--;
2118 		}
2119 	}
2120 	if (!skb)
2121 		return;
2122 	(void) mld_sendpack(skb);
2123 }
2124 
2125 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
2126 {
2127 	struct net *net = dev_net(dev);
2128 	struct sock *sk = net->ipv6.igmp_sk;
2129 	struct inet6_dev *idev;
2130 	struct sk_buff *skb;
2131 	struct mld_msg *hdr;
2132 	const struct in6_addr *snd_addr, *saddr;
2133 	struct in6_addr addr_buf;
2134 	int hlen = LL_RESERVED_SPACE(dev);
2135 	int tlen = dev->needed_tailroom;
2136 	int err, len, payload_len, full_len;
2137 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
2138 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
2139 		     IPV6_TLV_PADN, 0 };
2140 	struct flowi6 fl6;
2141 	struct dst_entry *dst;
2142 
2143 	if (type == ICMPV6_MGM_REDUCTION)
2144 		snd_addr = &in6addr_linklocal_allrouters;
2145 	else
2146 		snd_addr = addr;
2147 
2148 	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2149 	payload_len = len + sizeof(ra);
2150 	full_len = sizeof(struct ipv6hdr) + payload_len;
2151 
2152 	rcu_read_lock();
2153 	IP6_INC_STATS(net, __in6_dev_get(dev), IPSTATS_MIB_OUTREQUESTS);
2154 	rcu_read_unlock();
2155 
2156 	skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2157 
2158 	if (!skb) {
2159 		rcu_read_lock();
2160 		IP6_INC_STATS(net, __in6_dev_get(dev),
2161 			      IPSTATS_MIB_OUTDISCARDS);
2162 		rcu_read_unlock();
2163 		return;
2164 	}
2165 	skb->priority = TC_PRIO_CONTROL;
2166 	skb_reserve(skb, hlen);
2167 
2168 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2169 		/* <draft-ietf-magma-mld-source-05.txt>:
2170 		 * use unspecified address as the source address
2171 		 * when a valid link-local address is not available.
2172 		 */
2173 		saddr = &in6addr_any;
2174 	} else
2175 		saddr = &addr_buf;
2176 
2177 	ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2178 
2179 	skb_put_data(skb, ra, sizeof(ra));
2180 
2181 	hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2182 	hdr->mld_type = type;
2183 	hdr->mld_mca = *addr;
2184 
2185 	hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2186 					 IPPROTO_ICMPV6,
2187 					 csum_partial(hdr, len, 0));
2188 
2189 	rcu_read_lock();
2190 	idev = __in6_dev_get(skb->dev);
2191 
2192 	icmpv6_flow_init(sk, &fl6, type,
2193 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2194 			 skb->dev->ifindex);
2195 	dst = icmp6_dst_alloc(skb->dev, &fl6);
2196 	if (IS_ERR(dst)) {
2197 		err = PTR_ERR(dst);
2198 		goto err_out;
2199 	}
2200 
2201 	skb_dst_set(skb, dst);
2202 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2203 		      net, sk, skb, NULL, skb->dev,
2204 		      dst_output);
2205 out:
2206 	if (!err) {
2207 		ICMP6MSGOUT_INC_STATS(net, idev, type);
2208 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2209 	} else
2210 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2211 
2212 	rcu_read_unlock();
2213 	return;
2214 
2215 err_out:
2216 	kfree_skb(skb);
2217 	goto out;
2218 }
2219 
2220 /* called with mc_lock */
2221 static void mld_send_initial_cr(struct inet6_dev *idev)
2222 {
2223 	struct sk_buff *skb;
2224 	struct ifmcaddr6 *pmc;
2225 	int type;
2226 
2227 	if (mld_in_v1_mode(idev))
2228 		return;
2229 
2230 	skb = NULL;
2231 	for_each_mc_mclock(idev, pmc) {
2232 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2233 			type = MLD2_CHANGE_TO_EXCLUDE;
2234 		else
2235 			type = MLD2_ALLOW_NEW_SOURCES;
2236 		skb = add_grec(skb, pmc, type, 0, 0, 1);
2237 	}
2238 	if (skb)
2239 		mld_sendpack(skb);
2240 }
2241 
2242 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2243 {
2244 	mutex_lock(&idev->mc_lock);
2245 	idev->mc_dad_count = idev->mc_qrv;
2246 	if (idev->mc_dad_count) {
2247 		mld_send_initial_cr(idev);
2248 		idev->mc_dad_count--;
2249 		if (idev->mc_dad_count)
2250 			mld_dad_start_work(idev,
2251 					   unsolicited_report_interval(idev));
2252 	}
2253 	mutex_unlock(&idev->mc_lock);
2254 }
2255 
2256 static void mld_dad_work(struct work_struct *work)
2257 {
2258 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2259 					      struct inet6_dev,
2260 					      mc_dad_work);
2261 	mutex_lock(&idev->mc_lock);
2262 	mld_send_initial_cr(idev);
2263 	if (idev->mc_dad_count) {
2264 		idev->mc_dad_count--;
2265 		if (idev->mc_dad_count)
2266 			mld_dad_start_work(idev,
2267 					   unsolicited_report_interval(idev));
2268 	}
2269 	mutex_unlock(&idev->mc_lock);
2270 	in6_dev_put(idev);
2271 }
2272 
2273 /* called with mc_lock */
2274 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2275 	const struct in6_addr *psfsrc)
2276 {
2277 	struct ip6_sf_list *psf, *psf_prev;
2278 	int rv = 0;
2279 
2280 	psf_prev = NULL;
2281 	for_each_psf_mclock(pmc, psf) {
2282 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2283 			break;
2284 		psf_prev = psf;
2285 	}
2286 	if (!psf || psf->sf_count[sfmode] == 0) {
2287 		/* source filter not found, or count wrong =>  bug */
2288 		return -ESRCH;
2289 	}
2290 	WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] - 1);
2291 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2292 		struct inet6_dev *idev = pmc->idev;
2293 
2294 		/* no more filters for this source */
2295 		if (psf_prev)
2296 			rcu_assign_pointer(psf_prev->sf_next,
2297 					   mc_dereference(psf->sf_next, idev));
2298 		else
2299 			rcu_assign_pointer(pmc->mca_sources,
2300 					   mc_dereference(psf->sf_next, idev));
2301 
2302 		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2303 		    !mld_in_v1_mode(idev)) {
2304 			psf->sf_crcount = idev->mc_qrv;
2305 			rcu_assign_pointer(psf->sf_next,
2306 					   mc_dereference(pmc->mca_tomb, idev));
2307 			rcu_assign_pointer(pmc->mca_tomb, psf);
2308 			rv = 1;
2309 		} else {
2310 			kfree_rcu(psf, rcu);
2311 		}
2312 	}
2313 	return rv;
2314 }
2315 
2316 /* called with mc_lock */
2317 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2318 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2319 			  int delta)
2320 {
2321 	struct ifmcaddr6 *pmc;
2322 	int	changerec = 0;
2323 	int	i, err;
2324 
2325 	if (!idev)
2326 		return -ENODEV;
2327 
2328 	for_each_mc_mclock(idev, pmc) {
2329 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2330 			break;
2331 	}
2332 	if (!pmc)
2333 		return -ESRCH;
2334 
2335 	sf_markstate(pmc);
2336 	if (!delta) {
2337 		if (!pmc->mca_sfcount[sfmode])
2338 			return -EINVAL;
2339 
2340 		pmc->mca_sfcount[sfmode]--;
2341 	}
2342 	err = 0;
2343 	for (i = 0; i < sfcount; i++) {
2344 		int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2345 
2346 		changerec |= rv > 0;
2347 		if (!err && rv < 0)
2348 			err = rv;
2349 	}
2350 	if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2351 	    pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2352 	    pmc->mca_sfcount[MCAST_INCLUDE]) {
2353 		struct ip6_sf_list *psf;
2354 
2355 		/* filter mode change */
2356 		pmc->mca_sfmode = MCAST_INCLUDE;
2357 		pmc->mca_crcount = idev->mc_qrv;
2358 		idev->mc_ifc_count = pmc->mca_crcount;
2359 		for_each_psf_mclock(pmc, psf)
2360 			psf->sf_crcount = 0;
2361 		mld_ifc_event(pmc->idev);
2362 	} else if (sf_setstate(pmc) || changerec) {
2363 		mld_ifc_event(pmc->idev);
2364 	}
2365 
2366 	return err;
2367 }
2368 
2369 /*
2370  * Add multicast single-source filter to the interface list
2371  * called with mc_lock
2372  */
2373 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2374 	const struct in6_addr *psfsrc)
2375 {
2376 	struct ip6_sf_list *psf, *psf_prev;
2377 
2378 	psf_prev = NULL;
2379 	for_each_psf_mclock(pmc, psf) {
2380 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2381 			break;
2382 		psf_prev = psf;
2383 	}
2384 	if (!psf) {
2385 		psf = kzalloc(sizeof(*psf), GFP_KERNEL);
2386 		if (!psf)
2387 			return -ENOBUFS;
2388 
2389 		psf->sf_addr = *psfsrc;
2390 		if (psf_prev) {
2391 			rcu_assign_pointer(psf_prev->sf_next, psf);
2392 		} else {
2393 			rcu_assign_pointer(pmc->mca_sources, psf);
2394 		}
2395 	}
2396 	WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] + 1);
2397 	return 0;
2398 }
2399 
2400 /* called with mc_lock */
2401 static void sf_markstate(struct ifmcaddr6 *pmc)
2402 {
2403 	struct ip6_sf_list *psf;
2404 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2405 
2406 	for_each_psf_mclock(pmc, psf) {
2407 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2408 			psf->sf_oldin = mca_xcount ==
2409 				psf->sf_count[MCAST_EXCLUDE] &&
2410 				!psf->sf_count[MCAST_INCLUDE];
2411 		} else {
2412 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2413 		}
2414 	}
2415 }
2416 
2417 /* called with mc_lock */
2418 static int sf_setstate(struct ifmcaddr6 *pmc)
2419 {
2420 	struct ip6_sf_list *psf, *dpsf;
2421 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2422 	int qrv = pmc->idev->mc_qrv;
2423 	int new_in, rv;
2424 
2425 	rv = 0;
2426 	for_each_psf_mclock(pmc, psf) {
2427 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2428 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2429 				!psf->sf_count[MCAST_INCLUDE];
2430 		} else
2431 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2432 		if (new_in) {
2433 			if (!psf->sf_oldin) {
2434 				struct ip6_sf_list *prev = NULL;
2435 
2436 				for_each_psf_tomb(pmc, dpsf) {
2437 					if (ipv6_addr_equal(&dpsf->sf_addr,
2438 					    &psf->sf_addr))
2439 						break;
2440 					prev = dpsf;
2441 				}
2442 				if (dpsf) {
2443 					if (prev)
2444 						rcu_assign_pointer(prev->sf_next,
2445 								   mc_dereference(dpsf->sf_next,
2446 										  pmc->idev));
2447 					else
2448 						rcu_assign_pointer(pmc->mca_tomb,
2449 								   mc_dereference(dpsf->sf_next,
2450 										  pmc->idev));
2451 					kfree_rcu(dpsf, rcu);
2452 				}
2453 				psf->sf_crcount = qrv;
2454 				rv++;
2455 			}
2456 		} else if (psf->sf_oldin) {
2457 			psf->sf_crcount = 0;
2458 			/*
2459 			 * add or update "delete" records if an active filter
2460 			 * is now inactive
2461 			 */
2462 
2463 			for_each_psf_tomb(pmc, dpsf)
2464 				if (ipv6_addr_equal(&dpsf->sf_addr,
2465 				    &psf->sf_addr))
2466 					break;
2467 			if (!dpsf) {
2468 				dpsf = kmalloc(sizeof(*dpsf), GFP_KERNEL);
2469 				if (!dpsf)
2470 					continue;
2471 				*dpsf = *psf;
2472 				rcu_assign_pointer(dpsf->sf_next,
2473 						   mc_dereference(pmc->mca_tomb, pmc->idev));
2474 				rcu_assign_pointer(pmc->mca_tomb, dpsf);
2475 			}
2476 			dpsf->sf_crcount = qrv;
2477 			rv++;
2478 		}
2479 	}
2480 	return rv;
2481 }
2482 
2483 /*
2484  * Add multicast source filter list to the interface list
2485  * called with mc_lock
2486  */
2487 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2488 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2489 			  int delta)
2490 {
2491 	struct ifmcaddr6 *pmc;
2492 	int	isexclude;
2493 	int	i, err;
2494 
2495 	if (!idev)
2496 		return -ENODEV;
2497 
2498 	for_each_mc_mclock(idev, pmc) {
2499 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2500 			break;
2501 	}
2502 	if (!pmc)
2503 		return -ESRCH;
2504 
2505 	sf_markstate(pmc);
2506 	isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2507 	if (!delta)
2508 		WRITE_ONCE(pmc->mca_sfcount[sfmode],
2509 			   pmc->mca_sfcount[sfmode] + 1);
2510 	err = 0;
2511 	for (i = 0; i < sfcount; i++) {
2512 		err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2513 		if (err)
2514 			break;
2515 	}
2516 	if (err) {
2517 		int j;
2518 
2519 		if (!delta)
2520 			WRITE_ONCE(pmc->mca_sfcount[sfmode],
2521 				   pmc->mca_sfcount[sfmode] - 1);
2522 		for (j = 0; j < i; j++)
2523 			ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2524 	} else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2525 		struct ip6_sf_list *psf;
2526 
2527 		/* filter mode change */
2528 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2529 			pmc->mca_sfmode = MCAST_EXCLUDE;
2530 		else if (pmc->mca_sfcount[MCAST_INCLUDE])
2531 			pmc->mca_sfmode = MCAST_INCLUDE;
2532 		/* else no filters; keep old mode for reports */
2533 
2534 		pmc->mca_crcount = idev->mc_qrv;
2535 		idev->mc_ifc_count = pmc->mca_crcount;
2536 		for_each_psf_mclock(pmc, psf)
2537 			psf->sf_crcount = 0;
2538 		mld_ifc_event(idev);
2539 	} else if (sf_setstate(pmc)) {
2540 		mld_ifc_event(idev);
2541 	}
2542 	return err;
2543 }
2544 
2545 /* called with mc_lock */
2546 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2547 {
2548 	struct ip6_sf_list *psf, *nextpsf;
2549 
2550 	for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
2551 	     psf;
2552 	     psf = nextpsf) {
2553 		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2554 		kfree_rcu(psf, rcu);
2555 	}
2556 	RCU_INIT_POINTER(pmc->mca_tomb, NULL);
2557 	for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
2558 	     psf;
2559 	     psf = nextpsf) {
2560 		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2561 		kfree_rcu(psf, rcu);
2562 	}
2563 	RCU_INIT_POINTER(pmc->mca_sources, NULL);
2564 	pmc->mca_sfmode = MCAST_EXCLUDE;
2565 	pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2566 	/* Paired with the READ_ONCE() from ipv6_chk_mcast_addr() */
2567 	WRITE_ONCE(pmc->mca_sfcount[MCAST_EXCLUDE], 1);
2568 }
2569 
2570 /* called with mc_lock */
2571 static void igmp6_join_group(struct ifmcaddr6 *ma)
2572 {
2573 	unsigned long delay;
2574 
2575 	if (ma->mca_flags & MAF_NOREPORT)
2576 		return;
2577 
2578 	igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2579 
2580 	delay = get_random_u32_below(unsolicited_report_interval(ma->idev));
2581 
2582 	if (cancel_delayed_work(&ma->mca_work)) {
2583 		refcount_dec(&ma->mca_refcnt);
2584 		delay = ma->mca_work.timer.expires - jiffies;
2585 	}
2586 
2587 	if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
2588 		refcount_inc(&ma->mca_refcnt);
2589 	ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2590 }
2591 
2592 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2593 			    struct inet6_dev *idev)
2594 {
2595 	struct ip6_sf_socklist *psl;
2596 	int err;
2597 
2598 	psl = sock_dereference(iml->sflist, sk);
2599 
2600 	if (idev)
2601 		mutex_lock(&idev->mc_lock);
2602 
2603 	if (!psl) {
2604 		/* any-source empty exclude case */
2605 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2606 	} else {
2607 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2608 				     psl->sl_count, psl->sl_addr, 0);
2609 		RCU_INIT_POINTER(iml->sflist, NULL);
2610 		atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
2611 			   &sk->sk_omem_alloc);
2612 		kfree_rcu(psl, rcu);
2613 	}
2614 
2615 	if (idev)
2616 		mutex_unlock(&idev->mc_lock);
2617 
2618 	return err;
2619 }
2620 
2621 /* called with mc_lock */
2622 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2623 {
2624 	if (mld_in_v1_mode(ma->idev)) {
2625 		if (ma->mca_flags & MAF_LAST_REPORTER) {
2626 			igmp6_send(&ma->mca_addr, ma->idev->dev,
2627 				ICMPV6_MGM_REDUCTION);
2628 		}
2629 	} else {
2630 		mld_add_delrec(ma->idev, ma);
2631 		mld_ifc_event(ma->idev);
2632 	}
2633 }
2634 
2635 static void mld_gq_work(struct work_struct *work)
2636 {
2637 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2638 					      struct inet6_dev,
2639 					      mc_gq_work);
2640 
2641 	mutex_lock(&idev->mc_lock);
2642 	mld_send_report(idev, NULL);
2643 	idev->mc_gq_running = 0;
2644 	mutex_unlock(&idev->mc_lock);
2645 
2646 	in6_dev_put(idev);
2647 }
2648 
2649 static void mld_ifc_work(struct work_struct *work)
2650 {
2651 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2652 					      struct inet6_dev,
2653 					      mc_ifc_work);
2654 
2655 	mutex_lock(&idev->mc_lock);
2656 	mld_send_cr(idev);
2657 
2658 	if (idev->mc_ifc_count) {
2659 		idev->mc_ifc_count--;
2660 		if (idev->mc_ifc_count)
2661 			mld_ifc_start_work(idev,
2662 					   unsolicited_report_interval(idev));
2663 	}
2664 	mutex_unlock(&idev->mc_lock);
2665 	in6_dev_put(idev);
2666 }
2667 
2668 /* called with mc_lock */
2669 static void mld_ifc_event(struct inet6_dev *idev)
2670 {
2671 	if (mld_in_v1_mode(idev))
2672 		return;
2673 
2674 	idev->mc_ifc_count = idev->mc_qrv;
2675 	mld_ifc_start_work(idev, 1);
2676 }
2677 
2678 static void mld_mca_work(struct work_struct *work)
2679 {
2680 	struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
2681 					    struct ifmcaddr6, mca_work);
2682 
2683 	mutex_lock(&ma->idev->mc_lock);
2684 	if (mld_in_v1_mode(ma->idev))
2685 		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2686 	else
2687 		mld_send_report(ma->idev, ma);
2688 	ma->mca_flags |=  MAF_LAST_REPORTER;
2689 	ma->mca_flags &= ~MAF_TIMER_RUNNING;
2690 	mutex_unlock(&ma->idev->mc_lock);
2691 
2692 	ma_put(ma);
2693 }
2694 
2695 /* Device changing type */
2696 
2697 void ipv6_mc_unmap(struct inet6_dev *idev)
2698 {
2699 	struct ifmcaddr6 *i;
2700 
2701 	/* Install multicast list, except for all-nodes (already installed) */
2702 
2703 	mutex_lock(&idev->mc_lock);
2704 	for_each_mc_mclock(idev, i)
2705 		igmp6_group_dropped(i);
2706 	mutex_unlock(&idev->mc_lock);
2707 }
2708 
2709 void ipv6_mc_remap(struct inet6_dev *idev)
2710 {
2711 	ipv6_mc_up(idev);
2712 }
2713 
2714 /* Device going down */
2715 void ipv6_mc_down(struct inet6_dev *idev)
2716 {
2717 	struct ifmcaddr6 *i;
2718 
2719 	mutex_lock(&idev->mc_lock);
2720 	/* Withdraw multicast list */
2721 	for_each_mc_mclock(idev, i)
2722 		igmp6_group_dropped(i);
2723 	mutex_unlock(&idev->mc_lock);
2724 
2725 	/* Should stop work after group drop. or we will
2726 	 * start work again in mld_ifc_event()
2727 	 */
2728 	mld_query_stop_work(idev);
2729 	mld_report_stop_work(idev);
2730 
2731 	mutex_lock(&idev->mc_lock);
2732 	mld_ifc_stop_work(idev);
2733 	mld_gq_stop_work(idev);
2734 	mutex_unlock(&idev->mc_lock);
2735 
2736 	mld_dad_stop_work(idev);
2737 }
2738 
2739 static void ipv6_mc_reset(struct inet6_dev *idev)
2740 {
2741 	idev->mc_qrv = sysctl_mld_qrv;
2742 	idev->mc_qi = MLD_QI_DEFAULT;
2743 	idev->mc_qri = MLD_QRI_DEFAULT;
2744 	idev->mc_v1_seen = 0;
2745 	idev->mc_maxdelay = unsolicited_report_interval(idev);
2746 }
2747 
2748 /* Device going up */
2749 
2750 void ipv6_mc_up(struct inet6_dev *idev)
2751 {
2752 	struct ifmcaddr6 *i;
2753 
2754 	/* Install multicast list, except for all-nodes (already installed) */
2755 
2756 	ipv6_mc_reset(idev);
2757 	mutex_lock(&idev->mc_lock);
2758 	for_each_mc_mclock(idev, i) {
2759 		mld_del_delrec(idev, i);
2760 		igmp6_group_added(i);
2761 	}
2762 	mutex_unlock(&idev->mc_lock);
2763 }
2764 
2765 /* IPv6 device initialization. */
2766 
2767 void ipv6_mc_init_dev(struct inet6_dev *idev)
2768 {
2769 	idev->mc_gq_running = 0;
2770 	INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work);
2771 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
2772 	idev->mc_ifc_count = 0;
2773 	INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work);
2774 	INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work);
2775 	INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work);
2776 	INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work);
2777 	skb_queue_head_init(&idev->mc_query_queue);
2778 	skb_queue_head_init(&idev->mc_report_queue);
2779 	spin_lock_init(&idev->mc_query_lock);
2780 	spin_lock_init(&idev->mc_report_lock);
2781 	mutex_init(&idev->mc_lock);
2782 	ipv6_mc_reset(idev);
2783 }
2784 
2785 /*
2786  *	Device is about to be destroyed: clean up.
2787  */
2788 
2789 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2790 {
2791 	struct ifmcaddr6 *i;
2792 
2793 	/* Deactivate works */
2794 	ipv6_mc_down(idev);
2795 	mutex_lock(&idev->mc_lock);
2796 	mld_clear_delrec(idev);
2797 	mutex_unlock(&idev->mc_lock);
2798 	mld_clear_query(idev);
2799 	mld_clear_report(idev);
2800 
2801 	/* Delete all-nodes address. */
2802 	/* We cannot call ipv6_dev_mc_dec() directly, our caller in
2803 	 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2804 	 * fail.
2805 	 */
2806 	__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2807 
2808 	if (idev->cnf.forwarding)
2809 		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2810 
2811 	mutex_lock(&idev->mc_lock);
2812 	while ((i = mc_dereference(idev->mc_list, idev))) {
2813 		rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
2814 
2815 		ip6_mc_clear_src(i);
2816 		ma_put(i);
2817 	}
2818 	mutex_unlock(&idev->mc_lock);
2819 }
2820 
2821 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2822 {
2823 	struct ifmcaddr6 *pmc;
2824 
2825 	ASSERT_RTNL();
2826 
2827 	mutex_lock(&idev->mc_lock);
2828 	if (mld_in_v1_mode(idev)) {
2829 		for_each_mc_mclock(idev, pmc)
2830 			igmp6_join_group(pmc);
2831 	} else {
2832 		mld_send_report(idev, NULL);
2833 	}
2834 	mutex_unlock(&idev->mc_lock);
2835 }
2836 
2837 static int ipv6_mc_netdev_event(struct notifier_block *this,
2838 				unsigned long event,
2839 				void *ptr)
2840 {
2841 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2842 	struct inet6_dev *idev = __in6_dev_get(dev);
2843 
2844 	switch (event) {
2845 	case NETDEV_RESEND_IGMP:
2846 		if (idev)
2847 			ipv6_mc_rejoin_groups(idev);
2848 		break;
2849 	default:
2850 		break;
2851 	}
2852 
2853 	return NOTIFY_DONE;
2854 }
2855 
2856 static struct notifier_block igmp6_netdev_notifier = {
2857 	.notifier_call = ipv6_mc_netdev_event,
2858 };
2859 
2860 #ifdef CONFIG_PROC_FS
2861 struct igmp6_mc_iter_state {
2862 	struct seq_net_private p;
2863 	struct net_device *dev;
2864 	struct inet6_dev *idev;
2865 };
2866 
2867 #define igmp6_mc_seq_private(seq)	((struct igmp6_mc_iter_state *)(seq)->private)
2868 
2869 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2870 {
2871 	struct ifmcaddr6 *im = NULL;
2872 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2873 	struct net *net = seq_file_net(seq);
2874 
2875 	state->idev = NULL;
2876 	for_each_netdev_rcu(net, state->dev) {
2877 		struct inet6_dev *idev;
2878 		idev = __in6_dev_get(state->dev);
2879 		if (!idev)
2880 			continue;
2881 
2882 		im = rcu_dereference(idev->mc_list);
2883 		if (im) {
2884 			state->idev = idev;
2885 			break;
2886 		}
2887 	}
2888 	return im;
2889 }
2890 
2891 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2892 {
2893 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2894 
2895 	im = rcu_dereference(im->next);
2896 	while (!im) {
2897 		state->dev = next_net_device_rcu(state->dev);
2898 		if (!state->dev) {
2899 			state->idev = NULL;
2900 			break;
2901 		}
2902 		state->idev = __in6_dev_get(state->dev);
2903 		if (!state->idev)
2904 			continue;
2905 		im = rcu_dereference(state->idev->mc_list);
2906 	}
2907 	return im;
2908 }
2909 
2910 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2911 {
2912 	struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2913 	if (im)
2914 		while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2915 			--pos;
2916 	return pos ? NULL : im;
2917 }
2918 
2919 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2920 	__acquires(RCU)
2921 {
2922 	rcu_read_lock();
2923 	return igmp6_mc_get_idx(seq, *pos);
2924 }
2925 
2926 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2927 {
2928 	struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2929 
2930 	++*pos;
2931 	return im;
2932 }
2933 
2934 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2935 	__releases(RCU)
2936 {
2937 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2938 
2939 	if (likely(state->idev))
2940 		state->idev = NULL;
2941 	state->dev = NULL;
2942 	rcu_read_unlock();
2943 }
2944 
2945 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2946 {
2947 	struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2948 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2949 
2950 	seq_printf(seq,
2951 		   "%-4d %-15s %pi6 %5d %08X %ld\n",
2952 		   state->dev->ifindex, state->dev->name,
2953 		   &im->mca_addr,
2954 		   im->mca_users, im->mca_flags,
2955 		   (im->mca_flags & MAF_TIMER_RUNNING) ?
2956 		   jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0);
2957 	return 0;
2958 }
2959 
2960 static const struct seq_operations igmp6_mc_seq_ops = {
2961 	.start	=	igmp6_mc_seq_start,
2962 	.next	=	igmp6_mc_seq_next,
2963 	.stop	=	igmp6_mc_seq_stop,
2964 	.show	=	igmp6_mc_seq_show,
2965 };
2966 
2967 struct igmp6_mcf_iter_state {
2968 	struct seq_net_private p;
2969 	struct net_device *dev;
2970 	struct inet6_dev *idev;
2971 	struct ifmcaddr6 *im;
2972 };
2973 
2974 #define igmp6_mcf_seq_private(seq)	((struct igmp6_mcf_iter_state *)(seq)->private)
2975 
2976 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2977 {
2978 	struct ip6_sf_list *psf = NULL;
2979 	struct ifmcaddr6 *im = NULL;
2980 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2981 	struct net *net = seq_file_net(seq);
2982 
2983 	state->idev = NULL;
2984 	state->im = NULL;
2985 	for_each_netdev_rcu(net, state->dev) {
2986 		struct inet6_dev *idev;
2987 		idev = __in6_dev_get(state->dev);
2988 		if (unlikely(idev == NULL))
2989 			continue;
2990 
2991 		im = rcu_dereference(idev->mc_list);
2992 		if (likely(im)) {
2993 			psf = rcu_dereference(im->mca_sources);
2994 			if (likely(psf)) {
2995 				state->im = im;
2996 				state->idev = idev;
2997 				break;
2998 			}
2999 		}
3000 	}
3001 	return psf;
3002 }
3003 
3004 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
3005 {
3006 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3007 
3008 	psf = rcu_dereference(psf->sf_next);
3009 	while (!psf) {
3010 		state->im = rcu_dereference(state->im->next);
3011 		while (!state->im) {
3012 			state->dev = next_net_device_rcu(state->dev);
3013 			if (!state->dev) {
3014 				state->idev = NULL;
3015 				goto out;
3016 			}
3017 			state->idev = __in6_dev_get(state->dev);
3018 			if (!state->idev)
3019 				continue;
3020 			state->im = rcu_dereference(state->idev->mc_list);
3021 		}
3022 		psf = rcu_dereference(state->im->mca_sources);
3023 	}
3024 out:
3025 	return psf;
3026 }
3027 
3028 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
3029 {
3030 	struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
3031 	if (psf)
3032 		while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
3033 			--pos;
3034 	return pos ? NULL : psf;
3035 }
3036 
3037 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
3038 	__acquires(RCU)
3039 {
3040 	rcu_read_lock();
3041 	return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
3042 }
3043 
3044 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3045 {
3046 	struct ip6_sf_list *psf;
3047 	if (v == SEQ_START_TOKEN)
3048 		psf = igmp6_mcf_get_first(seq);
3049 	else
3050 		psf = igmp6_mcf_get_next(seq, v);
3051 	++*pos;
3052 	return psf;
3053 }
3054 
3055 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
3056 	__releases(RCU)
3057 {
3058 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3059 
3060 	if (likely(state->im))
3061 		state->im = NULL;
3062 	if (likely(state->idev))
3063 		state->idev = NULL;
3064 
3065 	state->dev = NULL;
3066 	rcu_read_unlock();
3067 }
3068 
3069 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
3070 {
3071 	struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
3072 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3073 
3074 	if (v == SEQ_START_TOKEN) {
3075 		seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
3076 	} else {
3077 		seq_printf(seq,
3078 			   "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
3079 			   state->dev->ifindex, state->dev->name,
3080 			   &state->im->mca_addr,
3081 			   &psf->sf_addr,
3082 			   READ_ONCE(psf->sf_count[MCAST_INCLUDE]),
3083 			   READ_ONCE(psf->sf_count[MCAST_EXCLUDE]));
3084 	}
3085 	return 0;
3086 }
3087 
3088 static const struct seq_operations igmp6_mcf_seq_ops = {
3089 	.start	=	igmp6_mcf_seq_start,
3090 	.next	=	igmp6_mcf_seq_next,
3091 	.stop	=	igmp6_mcf_seq_stop,
3092 	.show	=	igmp6_mcf_seq_show,
3093 };
3094 
3095 static int __net_init igmp6_proc_init(struct net *net)
3096 {
3097 	int err;
3098 
3099 	err = -ENOMEM;
3100 	if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
3101 			sizeof(struct igmp6_mc_iter_state)))
3102 		goto out;
3103 	if (!proc_create_net("mcfilter6", 0444, net->proc_net,
3104 			&igmp6_mcf_seq_ops,
3105 			sizeof(struct igmp6_mcf_iter_state)))
3106 		goto out_proc_net_igmp6;
3107 
3108 	err = 0;
3109 out:
3110 	return err;
3111 
3112 out_proc_net_igmp6:
3113 	remove_proc_entry("igmp6", net->proc_net);
3114 	goto out;
3115 }
3116 
3117 static void __net_exit igmp6_proc_exit(struct net *net)
3118 {
3119 	remove_proc_entry("mcfilter6", net->proc_net);
3120 	remove_proc_entry("igmp6", net->proc_net);
3121 }
3122 #else
3123 static inline int igmp6_proc_init(struct net *net)
3124 {
3125 	return 0;
3126 }
3127 static inline void igmp6_proc_exit(struct net *net)
3128 {
3129 }
3130 #endif
3131 
3132 static int __net_init igmp6_net_init(struct net *net)
3133 {
3134 	int err;
3135 
3136 	err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
3137 				   SOCK_RAW, IPPROTO_ICMPV6, net);
3138 	if (err < 0) {
3139 		pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
3140 		       err);
3141 		goto out;
3142 	}
3143 
3144 	inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
3145 	net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL;
3146 
3147 	err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
3148 				   SOCK_RAW, IPPROTO_ICMPV6, net);
3149 	if (err < 0) {
3150 		pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
3151 		       err);
3152 		goto out_sock_create;
3153 	}
3154 
3155 	err = igmp6_proc_init(net);
3156 	if (err)
3157 		goto out_sock_create_autojoin;
3158 
3159 	return 0;
3160 
3161 out_sock_create_autojoin:
3162 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3163 out_sock_create:
3164 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3165 out:
3166 	return err;
3167 }
3168 
3169 static void __net_exit igmp6_net_exit(struct net *net)
3170 {
3171 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3172 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3173 	igmp6_proc_exit(net);
3174 }
3175 
3176 static struct pernet_operations igmp6_net_ops = {
3177 	.init = igmp6_net_init,
3178 	.exit = igmp6_net_exit,
3179 };
3180 
3181 int __init igmp6_init(void)
3182 {
3183 	int err;
3184 
3185 	err = register_pernet_subsys(&igmp6_net_ops);
3186 	if (err)
3187 		return err;
3188 
3189 	mld_wq = create_workqueue("mld");
3190 	if (!mld_wq) {
3191 		unregister_pernet_subsys(&igmp6_net_ops);
3192 		return -ENOMEM;
3193 	}
3194 
3195 	return err;
3196 }
3197 
3198 int __init igmp6_late_init(void)
3199 {
3200 	return register_netdevice_notifier(&igmp6_netdev_notifier);
3201 }
3202 
3203 void igmp6_cleanup(void)
3204 {
3205 	unregister_pernet_subsys(&igmp6_net_ops);
3206 	destroy_workqueue(mld_wq);
3207 }
3208 
3209 void igmp6_late_cleanup(void)
3210 {
3211 	unregister_netdevice_notifier(&igmp6_netdev_notifier);
3212 }
3213