xref: /kernel/linux/linux-5.10/net/ipv6/mcast.c (revision 8c2ecf20)
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/times.h>
33#include <linux/net.h>
34#include <linux/in.h>
35#include <linux/in6.h>
36#include <linux/netdevice.h>
37#include <linux/if_arp.h>
38#include <linux/route.h>
39#include <linux/init.h>
40#include <linux/proc_fs.h>
41#include <linux/seq_file.h>
42#include <linux/slab.h>
43#include <linux/pkt_sched.h>
44#include <net/mld.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. */
64static 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
70static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
71
72static void igmp6_join_group(struct ifmcaddr6 *ma);
73static void igmp6_leave_group(struct ifmcaddr6 *ma);
74static void igmp6_timer_handler(struct timer_list *t);
75
76static void mld_gq_timer_expire(struct timer_list *t);
77static void mld_ifc_timer_expire(struct timer_list *t);
78static void mld_ifc_event(struct inet6_dev *idev);
79static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
80static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
81static void mld_clear_delrec(struct inet6_dev *idev);
82static bool mld_in_v1_mode(const struct inet6_dev *idev);
83static int sf_setstate(struct ifmcaddr6 *pmc);
84static void sf_markstate(struct ifmcaddr6 *pmc);
85static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
86static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
87			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
88			  int delta);
89static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
90			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
91			  int delta);
92static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
93			    struct inet6_dev *idev);
94static int __ipv6_dev_mc_inc(struct net_device *dev,
95			     const struct in6_addr *addr, unsigned int mode);
96
97#define MLD_QRV_DEFAULT		2
98/* RFC3810, 9.2. Query Interval */
99#define MLD_QI_DEFAULT		(125 * HZ)
100/* RFC3810, 9.3. Query Response Interval */
101#define MLD_QRI_DEFAULT		(10 * HZ)
102
103/* RFC3810, 8.1 Query Version Distinctions */
104#define MLD_V1_QUERY_LEN	24
105#define MLD_V2_QUERY_LEN_MIN	28
106
107#define IPV6_MLD_MAX_MSF	64
108
109int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
110int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
111
112/*
113 *	socket join on multicast group
114 */
115
116#define for_each_pmc_rcu(np, pmc)				\
117	for (pmc = rcu_dereference(np->ipv6_mc_list);		\
118	     pmc != NULL;					\
119	     pmc = rcu_dereference(pmc->next))
120
121static int unsolicited_report_interval(struct inet6_dev *idev)
122{
123	int iv;
124
125	if (mld_in_v1_mode(idev))
126		iv = idev->cnf.mldv1_unsolicited_report_interval;
127	else
128		iv = idev->cnf.mldv2_unsolicited_report_interval;
129
130	return iv > 0 ? iv : 1;
131}
132
133static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
134			       const struct in6_addr *addr, unsigned int mode)
135{
136	struct net_device *dev = NULL;
137	struct ipv6_mc_socklist *mc_lst;
138	struct ipv6_pinfo *np = inet6_sk(sk);
139	struct net *net = sock_net(sk);
140	int err;
141
142	ASSERT_RTNL();
143
144	if (!ipv6_addr_is_multicast(addr))
145		return -EINVAL;
146
147	rcu_read_lock();
148	for_each_pmc_rcu(np, mc_lst) {
149		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
150		    ipv6_addr_equal(&mc_lst->addr, addr)) {
151			rcu_read_unlock();
152			return -EADDRINUSE;
153		}
154	}
155	rcu_read_unlock();
156
157	mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
158
159	if (!mc_lst)
160		return -ENOMEM;
161
162	mc_lst->next = NULL;
163	mc_lst->addr = *addr;
164
165	if (ifindex == 0) {
166		struct rt6_info *rt;
167		rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
168		if (rt) {
169			dev = rt->dst.dev;
170			ip6_rt_put(rt);
171		}
172	} else
173		dev = __dev_get_by_index(net, ifindex);
174
175	if (!dev) {
176		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
177		return -ENODEV;
178	}
179
180	mc_lst->ifindex = dev->ifindex;
181	mc_lst->sfmode = mode;
182	rwlock_init(&mc_lst->sflock);
183	mc_lst->sflist = NULL;
184
185	/*
186	 *	now add/increase the group membership on the device
187	 */
188
189	err = __ipv6_dev_mc_inc(dev, addr, mode);
190
191	if (err) {
192		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
193		return err;
194	}
195
196	mc_lst->next = np->ipv6_mc_list;
197	rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
198
199	return 0;
200}
201
202int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
203{
204	return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
205}
206EXPORT_SYMBOL(ipv6_sock_mc_join);
207
208int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
209			  const struct in6_addr *addr, unsigned int mode)
210{
211	return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
212}
213
214/*
215 *	socket leave on multicast group
216 */
217int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
218{
219	struct ipv6_pinfo *np = inet6_sk(sk);
220	struct ipv6_mc_socklist *mc_lst;
221	struct ipv6_mc_socklist __rcu **lnk;
222	struct net *net = sock_net(sk);
223
224	ASSERT_RTNL();
225
226	if (!ipv6_addr_is_multicast(addr))
227		return -EINVAL;
228
229	for (lnk = &np->ipv6_mc_list;
230	     (mc_lst = rtnl_dereference(*lnk)) != NULL;
231	      lnk = &mc_lst->next) {
232		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
233		    ipv6_addr_equal(&mc_lst->addr, addr)) {
234			struct net_device *dev;
235
236			*lnk = mc_lst->next;
237
238			dev = __dev_get_by_index(net, mc_lst->ifindex);
239			if (dev) {
240				struct inet6_dev *idev = __in6_dev_get(dev);
241
242				(void) ip6_mc_leave_src(sk, mc_lst, idev);
243				if (idev)
244					__ipv6_dev_mc_dec(idev, &mc_lst->addr);
245			} else
246				(void) ip6_mc_leave_src(sk, mc_lst, NULL);
247
248			atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
249			kfree_rcu(mc_lst, rcu);
250			return 0;
251		}
252	}
253
254	return -EADDRNOTAVAIL;
255}
256EXPORT_SYMBOL(ipv6_sock_mc_drop);
257
258/* called with rcu_read_lock() */
259static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
260					     const struct in6_addr *group,
261					     int ifindex)
262{
263	struct net_device *dev = NULL;
264	struct inet6_dev *idev = NULL;
265
266	if (ifindex == 0) {
267		struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
268
269		if (rt) {
270			dev = rt->dst.dev;
271			ip6_rt_put(rt);
272		}
273	} else
274		dev = dev_get_by_index_rcu(net, ifindex);
275
276	if (!dev)
277		return NULL;
278	idev = __in6_dev_get(dev);
279	if (!idev)
280		return NULL;
281	read_lock_bh(&idev->lock);
282	if (idev->dead) {
283		read_unlock_bh(&idev->lock);
284		return NULL;
285	}
286	return idev;
287}
288
289void __ipv6_sock_mc_close(struct sock *sk)
290{
291	struct ipv6_pinfo *np = inet6_sk(sk);
292	struct ipv6_mc_socklist *mc_lst;
293	struct net *net = sock_net(sk);
294
295	ASSERT_RTNL();
296
297	while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
298		struct net_device *dev;
299
300		np->ipv6_mc_list = mc_lst->next;
301
302		dev = __dev_get_by_index(net, mc_lst->ifindex);
303		if (dev) {
304			struct inet6_dev *idev = __in6_dev_get(dev);
305
306			(void) ip6_mc_leave_src(sk, mc_lst, idev);
307			if (idev)
308				__ipv6_dev_mc_dec(idev, &mc_lst->addr);
309		} else
310			(void) ip6_mc_leave_src(sk, mc_lst, NULL);
311
312		atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
313		kfree_rcu(mc_lst, rcu);
314	}
315}
316
317void ipv6_sock_mc_close(struct sock *sk)
318{
319	struct ipv6_pinfo *np = inet6_sk(sk);
320
321	if (!rcu_access_pointer(np->ipv6_mc_list))
322		return;
323	rtnl_lock();
324	__ipv6_sock_mc_close(sk);
325	rtnl_unlock();
326}
327
328int ip6_mc_source(int add, int omode, struct sock *sk,
329	struct group_source_req *pgsr)
330{
331	struct in6_addr *source, *group;
332	struct ipv6_mc_socklist *pmc;
333	struct inet6_dev *idev;
334	struct ipv6_pinfo *inet6 = inet6_sk(sk);
335	struct ip6_sf_socklist *psl;
336	struct net *net = sock_net(sk);
337	int i, j, rv;
338	int leavegroup = 0;
339	int pmclocked = 0;
340	int err;
341
342	source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
343	group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
344
345	if (!ipv6_addr_is_multicast(group))
346		return -EINVAL;
347
348	rcu_read_lock();
349	idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
350	if (!idev) {
351		rcu_read_unlock();
352		return -ENODEV;
353	}
354
355	err = -EADDRNOTAVAIL;
356
357	for_each_pmc_rcu(inet6, pmc) {
358		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
359			continue;
360		if (ipv6_addr_equal(&pmc->addr, group))
361			break;
362	}
363	if (!pmc) {		/* must have a prior join */
364		err = -EINVAL;
365		goto done;
366	}
367	/* if a source filter was set, must be the same mode as before */
368	if (pmc->sflist) {
369		if (pmc->sfmode != omode) {
370			err = -EINVAL;
371			goto done;
372		}
373	} else if (pmc->sfmode != omode) {
374		/* allow mode switches for empty-set filters */
375		ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
376		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
377		pmc->sfmode = omode;
378	}
379
380	write_lock(&pmc->sflock);
381	pmclocked = 1;
382
383	psl = pmc->sflist;
384	if (!add) {
385		if (!psl)
386			goto done;	/* err = -EADDRNOTAVAIL */
387		rv = !0;
388		for (i = 0; i < psl->sl_count; i++) {
389			rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
390			if (rv == 0)
391				break;
392		}
393		if (rv)		/* source not found */
394			goto done;	/* err = -EADDRNOTAVAIL */
395
396		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
397		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
398			leavegroup = 1;
399			goto done;
400		}
401
402		/* update the interface filter */
403		ip6_mc_del_src(idev, group, omode, 1, source, 1);
404
405		for (j = i+1; j < psl->sl_count; j++)
406			psl->sl_addr[j-1] = psl->sl_addr[j];
407		psl->sl_count--;
408		err = 0;
409		goto done;
410	}
411	/* else, add a new source to the filter */
412
413	if (psl && psl->sl_count >= sysctl_mld_max_msf) {
414		err = -ENOBUFS;
415		goto done;
416	}
417	if (!psl || psl->sl_count == psl->sl_max) {
418		struct ip6_sf_socklist *newpsl;
419		int count = IP6_SFBLOCK;
420
421		if (psl)
422			count += psl->sl_max;
423		newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
424		if (!newpsl) {
425			err = -ENOBUFS;
426			goto done;
427		}
428		newpsl->sl_max = count;
429		newpsl->sl_count = count - IP6_SFBLOCK;
430		if (psl) {
431			for (i = 0; i < psl->sl_count; i++)
432				newpsl->sl_addr[i] = psl->sl_addr[i];
433			sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
434		}
435		pmc->sflist = psl = newpsl;
436	}
437	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
438	for (i = 0; i < psl->sl_count; i++) {
439		rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
440		if (rv == 0) /* There is an error in the address. */
441			goto done;
442	}
443	for (j = psl->sl_count-1; j >= i; j--)
444		psl->sl_addr[j+1] = psl->sl_addr[j];
445	psl->sl_addr[i] = *source;
446	psl->sl_count++;
447	err = 0;
448	/* update the interface list */
449	ip6_mc_add_src(idev, group, omode, 1, source, 1);
450done:
451	if (pmclocked)
452		write_unlock(&pmc->sflock);
453	read_unlock_bh(&idev->lock);
454	rcu_read_unlock();
455	if (leavegroup)
456		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
457	return err;
458}
459
460int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
461		    struct sockaddr_storage *list)
462{
463	const struct in6_addr *group;
464	struct ipv6_mc_socklist *pmc;
465	struct inet6_dev *idev;
466	struct ipv6_pinfo *inet6 = inet6_sk(sk);
467	struct ip6_sf_socklist *newpsl, *psl;
468	struct net *net = sock_net(sk);
469	int leavegroup = 0;
470	int i, err;
471
472	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
473
474	if (!ipv6_addr_is_multicast(group))
475		return -EINVAL;
476	if (gsf->gf_fmode != MCAST_INCLUDE &&
477	    gsf->gf_fmode != MCAST_EXCLUDE)
478		return -EINVAL;
479
480	rcu_read_lock();
481	idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
482
483	if (!idev) {
484		rcu_read_unlock();
485		return -ENODEV;
486	}
487
488	err = 0;
489
490	if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
491		leavegroup = 1;
492		goto done;
493	}
494
495	for_each_pmc_rcu(inet6, pmc) {
496		if (pmc->ifindex != gsf->gf_interface)
497			continue;
498		if (ipv6_addr_equal(&pmc->addr, group))
499			break;
500	}
501	if (!pmc) {		/* must have a prior join */
502		err = -EINVAL;
503		goto done;
504	}
505	if (gsf->gf_numsrc) {
506		newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
507							  GFP_ATOMIC);
508		if (!newpsl) {
509			err = -ENOBUFS;
510			goto done;
511		}
512		newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
513		for (i = 0; i < newpsl->sl_count; ++i, ++list) {
514			struct sockaddr_in6 *psin6;
515
516			psin6 = (struct sockaddr_in6 *)list;
517			newpsl->sl_addr[i] = psin6->sin6_addr;
518		}
519		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
520			newpsl->sl_count, newpsl->sl_addr, 0);
521		if (err) {
522			sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
523			goto done;
524		}
525	} else {
526		newpsl = NULL;
527		(void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
528	}
529
530	write_lock(&pmc->sflock);
531	psl = pmc->sflist;
532	if (psl) {
533		(void) ip6_mc_del_src(idev, group, pmc->sfmode,
534			psl->sl_count, psl->sl_addr, 0);
535		sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
536	} else
537		(void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
538	pmc->sflist = newpsl;
539	pmc->sfmode = gsf->gf_fmode;
540	write_unlock(&pmc->sflock);
541	err = 0;
542done:
543	read_unlock_bh(&idev->lock);
544	rcu_read_unlock();
545	if (leavegroup)
546		err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
547	return err;
548}
549
550int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
551	struct sockaddr_storage *p)
552{
553	int err, i, count, copycount;
554	const struct in6_addr *group;
555	struct ipv6_mc_socklist *pmc;
556	struct inet6_dev *idev;
557	struct ipv6_pinfo *inet6 = inet6_sk(sk);
558	struct ip6_sf_socklist *psl;
559	struct net *net = sock_net(sk);
560
561	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
562
563	if (!ipv6_addr_is_multicast(group))
564		return -EINVAL;
565
566	rcu_read_lock();
567	idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
568
569	if (!idev) {
570		rcu_read_unlock();
571		return -ENODEV;
572	}
573
574	err = -EADDRNOTAVAIL;
575	/* changes to the ipv6_mc_list require the socket lock and
576	 * rtnl lock. We have the socket lock and rcu read lock,
577	 * so reading the list is safe.
578	 */
579
580	for_each_pmc_rcu(inet6, pmc) {
581		if (pmc->ifindex != gsf->gf_interface)
582			continue;
583		if (ipv6_addr_equal(group, &pmc->addr))
584			break;
585	}
586	if (!pmc)		/* must have a prior join */
587		goto done;
588	gsf->gf_fmode = pmc->sfmode;
589	psl = pmc->sflist;
590	count = psl ? psl->sl_count : 0;
591	read_unlock_bh(&idev->lock);
592	rcu_read_unlock();
593
594	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
595	gsf->gf_numsrc = count;
596	/* changes to psl require the socket lock, and a write lock
597	 * on pmc->sflock. We have the socket lock so reading here is safe.
598	 */
599	for (i = 0; i < copycount; i++, p++) {
600		struct sockaddr_in6 *psin6;
601		struct sockaddr_storage ss;
602
603		psin6 = (struct sockaddr_in6 *)&ss;
604		memset(&ss, 0, sizeof(ss));
605		psin6->sin6_family = AF_INET6;
606		psin6->sin6_addr = psl->sl_addr[i];
607		if (copy_to_user(p, &ss, sizeof(ss)))
608			return -EFAULT;
609	}
610	return 0;
611done:
612	read_unlock_bh(&idev->lock);
613	rcu_read_unlock();
614	return err;
615}
616
617bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
618		    const struct in6_addr *src_addr)
619{
620	struct ipv6_pinfo *np = inet6_sk(sk);
621	struct ipv6_mc_socklist *mc;
622	struct ip6_sf_socklist *psl;
623	bool rv = true;
624
625	rcu_read_lock();
626	for_each_pmc_rcu(np, mc) {
627		if (ipv6_addr_equal(&mc->addr, mc_addr))
628			break;
629	}
630	if (!mc) {
631		rcu_read_unlock();
632		return np->mc_all;
633	}
634	read_lock(&mc->sflock);
635	psl = mc->sflist;
636	if (!psl) {
637		rv = mc->sfmode == MCAST_EXCLUDE;
638	} else {
639		int i;
640
641		for (i = 0; i < psl->sl_count; i++) {
642			if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
643				break;
644		}
645		if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
646			rv = false;
647		if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
648			rv = false;
649	}
650	read_unlock(&mc->sflock);
651	rcu_read_unlock();
652
653	return rv;
654}
655
656static void igmp6_group_added(struct ifmcaddr6 *mc)
657{
658	struct net_device *dev = mc->idev->dev;
659	char buf[MAX_ADDR_LEN];
660
661	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
662	    IPV6_ADDR_SCOPE_LINKLOCAL)
663		return;
664
665	spin_lock_bh(&mc->mca_lock);
666	if (!(mc->mca_flags&MAF_LOADED)) {
667		mc->mca_flags |= MAF_LOADED;
668		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
669			dev_mc_add(dev, buf);
670	}
671	spin_unlock_bh(&mc->mca_lock);
672
673	if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
674		return;
675
676	if (mld_in_v1_mode(mc->idev)) {
677		igmp6_join_group(mc);
678		return;
679	}
680	/* else v2 */
681
682	/* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
683	 * should not send filter-mode change record as the mode
684	 * should be from IN() to IN(A).
685	 */
686	if (mc->mca_sfmode == MCAST_EXCLUDE)
687		mc->mca_crcount = mc->idev->mc_qrv;
688
689	mld_ifc_event(mc->idev);
690}
691
692static void igmp6_group_dropped(struct ifmcaddr6 *mc)
693{
694	struct net_device *dev = mc->idev->dev;
695	char buf[MAX_ADDR_LEN];
696
697	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
698	    IPV6_ADDR_SCOPE_LINKLOCAL)
699		return;
700
701	spin_lock_bh(&mc->mca_lock);
702	if (mc->mca_flags&MAF_LOADED) {
703		mc->mca_flags &= ~MAF_LOADED;
704		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
705			dev_mc_del(dev, buf);
706	}
707
708	spin_unlock_bh(&mc->mca_lock);
709	if (mc->mca_flags & MAF_NOREPORT)
710		return;
711
712	if (!mc->idev->dead)
713		igmp6_leave_group(mc);
714
715	spin_lock_bh(&mc->mca_lock);
716	if (del_timer(&mc->mca_timer))
717		refcount_dec(&mc->mca_refcnt);
718	spin_unlock_bh(&mc->mca_lock);
719}
720
721/*
722 * deleted ifmcaddr6 manipulation
723 */
724static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
725{
726	struct ifmcaddr6 *pmc;
727
728	/* this is an "ifmcaddr6" for convenience; only the fields below
729	 * are actually used. In particular, the refcnt and users are not
730	 * used for management of the delete list. Using the same structure
731	 * for deleted items allows change reports to use common code with
732	 * non-deleted or query-response MCA's.
733	 */
734	pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
735	if (!pmc)
736		return;
737
738	spin_lock_bh(&im->mca_lock);
739	spin_lock_init(&pmc->mca_lock);
740	pmc->idev = im->idev;
741	in6_dev_hold(idev);
742	pmc->mca_addr = im->mca_addr;
743	pmc->mca_crcount = idev->mc_qrv;
744	pmc->mca_sfmode = im->mca_sfmode;
745	if (pmc->mca_sfmode == MCAST_INCLUDE) {
746		struct ip6_sf_list *psf;
747
748		pmc->mca_tomb = im->mca_tomb;
749		pmc->mca_sources = im->mca_sources;
750		im->mca_tomb = im->mca_sources = NULL;
751		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
752			psf->sf_crcount = pmc->mca_crcount;
753	}
754	spin_unlock_bh(&im->mca_lock);
755
756	spin_lock_bh(&idev->mc_lock);
757	pmc->next = idev->mc_tomb;
758	idev->mc_tomb = pmc;
759	spin_unlock_bh(&idev->mc_lock);
760}
761
762static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
763{
764	struct ifmcaddr6 *pmc, *pmc_prev;
765	struct ip6_sf_list *psf;
766	struct in6_addr *pmca = &im->mca_addr;
767
768	spin_lock_bh(&idev->mc_lock);
769	pmc_prev = NULL;
770	for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
771		if (ipv6_addr_equal(&pmc->mca_addr, pmca))
772			break;
773		pmc_prev = pmc;
774	}
775	if (pmc) {
776		if (pmc_prev)
777			pmc_prev->next = pmc->next;
778		else
779			idev->mc_tomb = pmc->next;
780	}
781	spin_unlock_bh(&idev->mc_lock);
782
783	spin_lock_bh(&im->mca_lock);
784	if (pmc) {
785		im->idev = pmc->idev;
786		if (im->mca_sfmode == MCAST_INCLUDE) {
787			swap(im->mca_tomb, pmc->mca_tomb);
788			swap(im->mca_sources, pmc->mca_sources);
789			for (psf = im->mca_sources; psf; psf = psf->sf_next)
790				psf->sf_crcount = idev->mc_qrv;
791		} else {
792			im->mca_crcount = idev->mc_qrv;
793		}
794		in6_dev_put(pmc->idev);
795		ip6_mc_clear_src(pmc);
796		kfree(pmc);
797	}
798	spin_unlock_bh(&im->mca_lock);
799}
800
801static void mld_clear_delrec(struct inet6_dev *idev)
802{
803	struct ifmcaddr6 *pmc, *nextpmc;
804
805	spin_lock_bh(&idev->mc_lock);
806	pmc = idev->mc_tomb;
807	idev->mc_tomb = NULL;
808	spin_unlock_bh(&idev->mc_lock);
809
810	for (; pmc; pmc = nextpmc) {
811		nextpmc = pmc->next;
812		ip6_mc_clear_src(pmc);
813		in6_dev_put(pmc->idev);
814		kfree(pmc);
815	}
816
817	/* clear dead sources, too */
818	read_lock_bh(&idev->lock);
819	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
820		struct ip6_sf_list *psf, *psf_next;
821
822		spin_lock_bh(&pmc->mca_lock);
823		psf = pmc->mca_tomb;
824		pmc->mca_tomb = NULL;
825		spin_unlock_bh(&pmc->mca_lock);
826		for (; psf; psf = psf_next) {
827			psf_next = psf->sf_next;
828			kfree(psf);
829		}
830	}
831	read_unlock_bh(&idev->lock);
832}
833
834static void mca_get(struct ifmcaddr6 *mc)
835{
836	refcount_inc(&mc->mca_refcnt);
837}
838
839static void ma_put(struct ifmcaddr6 *mc)
840{
841	if (refcount_dec_and_test(&mc->mca_refcnt)) {
842		in6_dev_put(mc->idev);
843		kfree(mc);
844	}
845}
846
847static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
848				   const struct in6_addr *addr,
849				   unsigned int mode)
850{
851	struct ifmcaddr6 *mc;
852
853	mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
854	if (!mc)
855		return NULL;
856
857	timer_setup(&mc->mca_timer, igmp6_timer_handler, 0);
858
859	mc->mca_addr = *addr;
860	mc->idev = idev; /* reference taken by caller */
861	mc->mca_users = 1;
862	/* mca_stamp should be updated upon changes */
863	mc->mca_cstamp = mc->mca_tstamp = jiffies;
864	refcount_set(&mc->mca_refcnt, 1);
865	spin_lock_init(&mc->mca_lock);
866
867	mc->mca_sfmode = mode;
868	mc->mca_sfcount[mode] = 1;
869
870	if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
871	    IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
872		mc->mca_flags |= MAF_NOREPORT;
873
874	return mc;
875}
876
877/*
878 *	device multicast group inc (add if not found)
879 */
880static int __ipv6_dev_mc_inc(struct net_device *dev,
881			     const struct in6_addr *addr, unsigned int mode)
882{
883	struct ifmcaddr6 *mc;
884	struct inet6_dev *idev;
885
886	ASSERT_RTNL();
887
888	/* we need to take a reference on idev */
889	idev = in6_dev_get(dev);
890
891	if (!idev)
892		return -EINVAL;
893
894	write_lock_bh(&idev->lock);
895	if (idev->dead) {
896		write_unlock_bh(&idev->lock);
897		in6_dev_put(idev);
898		return -ENODEV;
899	}
900
901	for (mc = idev->mc_list; mc; mc = mc->next) {
902		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
903			mc->mca_users++;
904			write_unlock_bh(&idev->lock);
905			ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
906			in6_dev_put(idev);
907			return 0;
908		}
909	}
910
911	mc = mca_alloc(idev, addr, mode);
912	if (!mc) {
913		write_unlock_bh(&idev->lock);
914		in6_dev_put(idev);
915		return -ENOMEM;
916	}
917
918	mc->next = idev->mc_list;
919	idev->mc_list = mc;
920
921	/* Hold this for the code below before we unlock,
922	 * it is already exposed via idev->mc_list.
923	 */
924	mca_get(mc);
925	write_unlock_bh(&idev->lock);
926
927	mld_del_delrec(idev, mc);
928	igmp6_group_added(mc);
929	ma_put(mc);
930	return 0;
931}
932
933int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
934{
935	return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
936}
937EXPORT_SYMBOL(ipv6_dev_mc_inc);
938
939/*
940 *	device multicast group del
941 */
942int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
943{
944	struct ifmcaddr6 *ma, **map;
945
946	ASSERT_RTNL();
947
948	write_lock_bh(&idev->lock);
949	for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
950		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
951			if (--ma->mca_users == 0) {
952				*map = ma->next;
953				write_unlock_bh(&idev->lock);
954
955				igmp6_group_dropped(ma);
956				ip6_mc_clear_src(ma);
957
958				ma_put(ma);
959				return 0;
960			}
961			write_unlock_bh(&idev->lock);
962			return 0;
963		}
964	}
965	write_unlock_bh(&idev->lock);
966
967	return -ENOENT;
968}
969
970int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
971{
972	struct inet6_dev *idev;
973	int err;
974
975	ASSERT_RTNL();
976
977	idev = __in6_dev_get(dev);
978	if (!idev)
979		err = -ENODEV;
980	else
981		err = __ipv6_dev_mc_dec(idev, addr);
982
983	return err;
984}
985EXPORT_SYMBOL(ipv6_dev_mc_dec);
986
987/*
988 *	check if the interface/address pair is valid
989 */
990bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
991			 const struct in6_addr *src_addr)
992{
993	struct inet6_dev *idev;
994	struct ifmcaddr6 *mc;
995	bool rv = false;
996
997	rcu_read_lock();
998	idev = __in6_dev_get(dev);
999	if (idev) {
1000		read_lock_bh(&idev->lock);
1001		for (mc = idev->mc_list; mc; mc = mc->next) {
1002			if (ipv6_addr_equal(&mc->mca_addr, group))
1003				break;
1004		}
1005		if (mc) {
1006			if (src_addr && !ipv6_addr_any(src_addr)) {
1007				struct ip6_sf_list *psf;
1008
1009				spin_lock_bh(&mc->mca_lock);
1010				for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
1011					if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1012						break;
1013				}
1014				if (psf)
1015					rv = psf->sf_count[MCAST_INCLUDE] ||
1016						psf->sf_count[MCAST_EXCLUDE] !=
1017						mc->mca_sfcount[MCAST_EXCLUDE];
1018				else
1019					rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1020				spin_unlock_bh(&mc->mca_lock);
1021			} else
1022				rv = true; /* don't filter unspecified source */
1023		}
1024		read_unlock_bh(&idev->lock);
1025	}
1026	rcu_read_unlock();
1027	return rv;
1028}
1029
1030static void mld_gq_start_timer(struct inet6_dev *idev)
1031{
1032	unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1033
1034	idev->mc_gq_running = 1;
1035	if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1036		in6_dev_hold(idev);
1037}
1038
1039static void mld_gq_stop_timer(struct inet6_dev *idev)
1040{
1041	idev->mc_gq_running = 0;
1042	if (del_timer(&idev->mc_gq_timer))
1043		__in6_dev_put(idev);
1044}
1045
1046static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1047{
1048	unsigned long tv = prandom_u32() % delay;
1049
1050	if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1051		in6_dev_hold(idev);
1052}
1053
1054static void mld_ifc_stop_timer(struct inet6_dev *idev)
1055{
1056	idev->mc_ifc_count = 0;
1057	if (del_timer(&idev->mc_ifc_timer))
1058		__in6_dev_put(idev);
1059}
1060
1061static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1062{
1063	unsigned long tv = prandom_u32() % delay;
1064
1065	if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1066		in6_dev_hold(idev);
1067}
1068
1069static void mld_dad_stop_timer(struct inet6_dev *idev)
1070{
1071	if (del_timer(&idev->mc_dad_timer))
1072		__in6_dev_put(idev);
1073}
1074
1075/*
1076 *	IGMP handling (alias multicast ICMPv6 messages)
1077 */
1078
1079static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1080{
1081	unsigned long delay = resptime;
1082
1083	/* Do not start timer for these addresses */
1084	if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1085	    IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1086		return;
1087
1088	if (del_timer(&ma->mca_timer)) {
1089		refcount_dec(&ma->mca_refcnt);
1090		delay = ma->mca_timer.expires - jiffies;
1091	}
1092
1093	if (delay >= resptime)
1094		delay = prandom_u32() % resptime;
1095
1096	ma->mca_timer.expires = jiffies + delay;
1097	if (!mod_timer(&ma->mca_timer, jiffies + delay))
1098		refcount_inc(&ma->mca_refcnt);
1099	ma->mca_flags |= MAF_TIMER_RUNNING;
1100}
1101
1102/* mark EXCLUDE-mode sources */
1103static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1104			     const struct in6_addr *srcs)
1105{
1106	struct ip6_sf_list *psf;
1107	int i, scount;
1108
1109	scount = 0;
1110	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1111		if (scount == nsrcs)
1112			break;
1113		for (i = 0; i < nsrcs; i++) {
1114			/* skip inactive filters */
1115			if (psf->sf_count[MCAST_INCLUDE] ||
1116			    pmc->mca_sfcount[MCAST_EXCLUDE] !=
1117			    psf->sf_count[MCAST_EXCLUDE])
1118				break;
1119			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1120				scount++;
1121				break;
1122			}
1123		}
1124	}
1125	pmc->mca_flags &= ~MAF_GSQUERY;
1126	if (scount == nsrcs)	/* all sources excluded */
1127		return false;
1128	return true;
1129}
1130
1131static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1132			    const struct in6_addr *srcs)
1133{
1134	struct ip6_sf_list *psf;
1135	int i, scount;
1136
1137	if (pmc->mca_sfmode == MCAST_EXCLUDE)
1138		return mld_xmarksources(pmc, nsrcs, srcs);
1139
1140	/* mark INCLUDE-mode sources */
1141
1142	scount = 0;
1143	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1144		if (scount == nsrcs)
1145			break;
1146		for (i = 0; i < nsrcs; i++) {
1147			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1148				psf->sf_gsresp = 1;
1149				scount++;
1150				break;
1151			}
1152		}
1153	}
1154	if (!scount) {
1155		pmc->mca_flags &= ~MAF_GSQUERY;
1156		return false;
1157	}
1158	pmc->mca_flags |= MAF_GSQUERY;
1159	return true;
1160}
1161
1162static int mld_force_mld_version(const struct inet6_dev *idev)
1163{
1164	/* Normally, both are 0 here. If enforcement to a particular is
1165	 * being used, individual device enforcement will have a lower
1166	 * precedence over 'all' device (.../conf/all/force_mld_version).
1167	 */
1168
1169	if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1170		return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1171	else
1172		return idev->cnf.force_mld_version;
1173}
1174
1175static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1176{
1177	return mld_force_mld_version(idev) == 2;
1178}
1179
1180static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1181{
1182	return mld_force_mld_version(idev) == 1;
1183}
1184
1185static bool mld_in_v1_mode(const struct inet6_dev *idev)
1186{
1187	if (mld_in_v2_mode_only(idev))
1188		return false;
1189	if (mld_in_v1_mode_only(idev))
1190		return true;
1191	if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1192		return true;
1193
1194	return false;
1195}
1196
1197static void mld_set_v1_mode(struct inet6_dev *idev)
1198{
1199	/* RFC3810, relevant sections:
1200	 *  - 9.1. Robustness Variable
1201	 *  - 9.2. Query Interval
1202	 *  - 9.3. Query Response Interval
1203	 *  - 9.12. Older Version Querier Present Timeout
1204	 */
1205	unsigned long switchback;
1206
1207	switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1208
1209	idev->mc_v1_seen = jiffies + switchback;
1210}
1211
1212static void mld_update_qrv(struct inet6_dev *idev,
1213			   const struct mld2_query *mlh2)
1214{
1215	/* RFC3810, relevant sections:
1216	 *  - 5.1.8. QRV (Querier's Robustness Variable)
1217	 *  - 9.1. Robustness Variable
1218	 */
1219
1220	/* The value of the Robustness Variable MUST NOT be zero,
1221	 * and SHOULD NOT be one. Catch this here if we ever run
1222	 * into such a case in future.
1223	 */
1224	const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1225	WARN_ON(idev->mc_qrv == 0);
1226
1227	if (mlh2->mld2q_qrv > 0)
1228		idev->mc_qrv = mlh2->mld2q_qrv;
1229
1230	if (unlikely(idev->mc_qrv < min_qrv)) {
1231		net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1232				     idev->mc_qrv, min_qrv);
1233		idev->mc_qrv = min_qrv;
1234	}
1235}
1236
1237static void mld_update_qi(struct inet6_dev *idev,
1238			  const struct mld2_query *mlh2)
1239{
1240	/* RFC3810, relevant sections:
1241	 *  - 5.1.9. QQIC (Querier's Query Interval Code)
1242	 *  - 9.2. Query Interval
1243	 *  - 9.12. Older Version Querier Present Timeout
1244	 *    (the [Query Interval] in the last Query received)
1245	 */
1246	unsigned long mc_qqi;
1247
1248	if (mlh2->mld2q_qqic < 128) {
1249		mc_qqi = mlh2->mld2q_qqic;
1250	} else {
1251		unsigned long mc_man, mc_exp;
1252
1253		mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1254		mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1255
1256		mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1257	}
1258
1259	idev->mc_qi = mc_qqi * HZ;
1260}
1261
1262static void mld_update_qri(struct inet6_dev *idev,
1263			   const struct mld2_query *mlh2)
1264{
1265	/* RFC3810, relevant sections:
1266	 *  - 5.1.3. Maximum Response Code
1267	 *  - 9.3. Query Response Interval
1268	 */
1269	idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1270}
1271
1272static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1273			  unsigned long *max_delay, bool v1_query)
1274{
1275	unsigned long mldv1_md;
1276
1277	/* Ignore v1 queries */
1278	if (mld_in_v2_mode_only(idev))
1279		return -EINVAL;
1280
1281	mldv1_md = ntohs(mld->mld_maxdelay);
1282
1283	/* When in MLDv1 fallback and a MLDv2 router start-up being
1284	 * unaware of current MLDv1 operation, the MRC == MRD mapping
1285	 * only works when the exponential algorithm is not being
1286	 * used (as MLDv1 is unaware of such things).
1287	 *
1288	 * According to the RFC author, the MLDv2 implementations
1289	 * he's aware of all use a MRC < 32768 on start up queries.
1290	 *
1291	 * Thus, should we *ever* encounter something else larger
1292	 * than that, just assume the maximum possible within our
1293	 * reach.
1294	 */
1295	if (!v1_query)
1296		mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1297
1298	*max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1299
1300	/* MLDv1 router present: we need to go into v1 mode *only*
1301	 * when an MLDv1 query is received as per section 9.12. of
1302	 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1303	 * queries MUST be of exactly 24 octets.
1304	 */
1305	if (v1_query)
1306		mld_set_v1_mode(idev);
1307
1308	/* cancel MLDv2 report timer */
1309	mld_gq_stop_timer(idev);
1310	/* cancel the interface change timer */
1311	mld_ifc_stop_timer(idev);
1312	/* clear deleted report items */
1313	mld_clear_delrec(idev);
1314
1315	return 0;
1316}
1317
1318static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1319			  unsigned long *max_delay)
1320{
1321	*max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1322
1323	mld_update_qrv(idev, mld);
1324	mld_update_qi(idev, mld);
1325	mld_update_qri(idev, mld);
1326
1327	idev->mc_maxdelay = *max_delay;
1328
1329	return 0;
1330}
1331
1332/* called with rcu_read_lock() */
1333int igmp6_event_query(struct sk_buff *skb)
1334{
1335	struct mld2_query *mlh2 = NULL;
1336	struct ifmcaddr6 *ma;
1337	const struct in6_addr *group;
1338	unsigned long max_delay;
1339	struct inet6_dev *idev;
1340	struct mld_msg *mld;
1341	int group_type;
1342	int mark = 0;
1343	int len, err;
1344
1345	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1346		return -EINVAL;
1347
1348	/* compute payload length excluding extension headers */
1349	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1350	len -= skb_network_header_len(skb);
1351
1352	/* RFC3810 6.2
1353	 * Upon reception of an MLD message that contains a Query, the node
1354	 * checks if the source address of the message is a valid link-local
1355	 * address, if the Hop Limit is set to 1, and if the Router Alert
1356	 * option is present in the Hop-By-Hop Options header of the IPv6
1357	 * packet.  If any of these checks fails, the packet is dropped.
1358	 */
1359	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1360	    ipv6_hdr(skb)->hop_limit != 1 ||
1361	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1362	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1363		return -EINVAL;
1364
1365	idev = __in6_dev_get(skb->dev);
1366	if (!idev)
1367		return 0;
1368
1369	mld = (struct mld_msg *)icmp6_hdr(skb);
1370	group = &mld->mld_mca;
1371	group_type = ipv6_addr_type(group);
1372
1373	if (group_type != IPV6_ADDR_ANY &&
1374	    !(group_type&IPV6_ADDR_MULTICAST))
1375		return -EINVAL;
1376
1377	if (len < MLD_V1_QUERY_LEN) {
1378		return -EINVAL;
1379	} else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1380		err = mld_process_v1(idev, mld, &max_delay,
1381				     len == MLD_V1_QUERY_LEN);
1382		if (err < 0)
1383			return err;
1384	} else if (len >= MLD_V2_QUERY_LEN_MIN) {
1385		int srcs_offset = sizeof(struct mld2_query) -
1386				  sizeof(struct icmp6hdr);
1387
1388		if (!pskb_may_pull(skb, srcs_offset))
1389			return -EINVAL;
1390
1391		mlh2 = (struct mld2_query *)skb_transport_header(skb);
1392
1393		err = mld_process_v2(idev, mlh2, &max_delay);
1394		if (err < 0)
1395			return err;
1396
1397		if (group_type == IPV6_ADDR_ANY) { /* general query */
1398			if (mlh2->mld2q_nsrcs)
1399				return -EINVAL; /* no sources allowed */
1400
1401			mld_gq_start_timer(idev);
1402			return 0;
1403		}
1404		/* mark sources to include, if group & source-specific */
1405		if (mlh2->mld2q_nsrcs != 0) {
1406			if (!pskb_may_pull(skb, srcs_offset +
1407			    ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1408				return -EINVAL;
1409
1410			mlh2 = (struct mld2_query *)skb_transport_header(skb);
1411			mark = 1;
1412		}
1413	} else {
1414		return -EINVAL;
1415	}
1416
1417	read_lock_bh(&idev->lock);
1418	if (group_type == IPV6_ADDR_ANY) {
1419		for (ma = idev->mc_list; ma; ma = ma->next) {
1420			spin_lock_bh(&ma->mca_lock);
1421			igmp6_group_queried(ma, max_delay);
1422			spin_unlock_bh(&ma->mca_lock);
1423		}
1424	} else {
1425		for (ma = idev->mc_list; ma; ma = ma->next) {
1426			if (!ipv6_addr_equal(group, &ma->mca_addr))
1427				continue;
1428			spin_lock_bh(&ma->mca_lock);
1429			if (ma->mca_flags & MAF_TIMER_RUNNING) {
1430				/* gsquery <- gsquery && mark */
1431				if (!mark)
1432					ma->mca_flags &= ~MAF_GSQUERY;
1433			} else {
1434				/* gsquery <- mark */
1435				if (mark)
1436					ma->mca_flags |= MAF_GSQUERY;
1437				else
1438					ma->mca_flags &= ~MAF_GSQUERY;
1439			}
1440			if (!(ma->mca_flags & MAF_GSQUERY) ||
1441			    mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1442				igmp6_group_queried(ma, max_delay);
1443			spin_unlock_bh(&ma->mca_lock);
1444			break;
1445		}
1446	}
1447	read_unlock_bh(&idev->lock);
1448
1449	return 0;
1450}
1451
1452/* called with rcu_read_lock() */
1453int igmp6_event_report(struct sk_buff *skb)
1454{
1455	struct ifmcaddr6 *ma;
1456	struct inet6_dev *idev;
1457	struct mld_msg *mld;
1458	int addr_type;
1459
1460	/* Our own report looped back. Ignore it. */
1461	if (skb->pkt_type == PACKET_LOOPBACK)
1462		return 0;
1463
1464	/* send our report if the MC router may not have heard this report */
1465	if (skb->pkt_type != PACKET_MULTICAST &&
1466	    skb->pkt_type != PACKET_BROADCAST)
1467		return 0;
1468
1469	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1470		return -EINVAL;
1471
1472	mld = (struct mld_msg *)icmp6_hdr(skb);
1473
1474	/* Drop reports with not link local source */
1475	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1476	if (addr_type != IPV6_ADDR_ANY &&
1477	    !(addr_type&IPV6_ADDR_LINKLOCAL))
1478		return -EINVAL;
1479
1480	idev = __in6_dev_get(skb->dev);
1481	if (!idev)
1482		return -ENODEV;
1483
1484	/*
1485	 *	Cancel the timer for this group
1486	 */
1487
1488	read_lock_bh(&idev->lock);
1489	for (ma = idev->mc_list; ma; ma = ma->next) {
1490		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1491			spin_lock(&ma->mca_lock);
1492			if (del_timer(&ma->mca_timer))
1493				refcount_dec(&ma->mca_refcnt);
1494			ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1495			spin_unlock(&ma->mca_lock);
1496			break;
1497		}
1498	}
1499	read_unlock_bh(&idev->lock);
1500	return 0;
1501}
1502
1503static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1504		  int gdeleted, int sdeleted)
1505{
1506	switch (type) {
1507	case MLD2_MODE_IS_INCLUDE:
1508	case MLD2_MODE_IS_EXCLUDE:
1509		if (gdeleted || sdeleted)
1510			return false;
1511		if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1512			if (pmc->mca_sfmode == MCAST_INCLUDE)
1513				return true;
1514			/* don't include if this source is excluded
1515			 * in all filters
1516			 */
1517			if (psf->sf_count[MCAST_INCLUDE])
1518				return type == MLD2_MODE_IS_INCLUDE;
1519			return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1520				psf->sf_count[MCAST_EXCLUDE];
1521		}
1522		return false;
1523	case MLD2_CHANGE_TO_INCLUDE:
1524		if (gdeleted || sdeleted)
1525			return false;
1526		return psf->sf_count[MCAST_INCLUDE] != 0;
1527	case MLD2_CHANGE_TO_EXCLUDE:
1528		if (gdeleted || sdeleted)
1529			return false;
1530		if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1531		    psf->sf_count[MCAST_INCLUDE])
1532			return false;
1533		return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1534			psf->sf_count[MCAST_EXCLUDE];
1535	case MLD2_ALLOW_NEW_SOURCES:
1536		if (gdeleted || !psf->sf_crcount)
1537			return false;
1538		return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1539	case MLD2_BLOCK_OLD_SOURCES:
1540		if (pmc->mca_sfmode == MCAST_INCLUDE)
1541			return gdeleted || (psf->sf_crcount && sdeleted);
1542		return psf->sf_crcount && !gdeleted && !sdeleted;
1543	}
1544	return false;
1545}
1546
1547static int
1548mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1549{
1550	struct ip6_sf_list *psf;
1551	int scount = 0;
1552
1553	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1554		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1555			continue;
1556		scount++;
1557	}
1558	return scount;
1559}
1560
1561static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1562		       struct net_device *dev,
1563		       const struct in6_addr *saddr,
1564		       const struct in6_addr *daddr,
1565		       int proto, int len)
1566{
1567	struct ipv6hdr *hdr;
1568
1569	skb->protocol = htons(ETH_P_IPV6);
1570	skb->dev = dev;
1571
1572	skb_reset_network_header(skb);
1573	skb_put(skb, sizeof(struct ipv6hdr));
1574	hdr = ipv6_hdr(skb);
1575
1576	ip6_flow_hdr(hdr, 0, 0);
1577
1578	hdr->payload_len = htons(len);
1579	hdr->nexthdr = proto;
1580	hdr->hop_limit = inet6_sk(sk)->hop_limit;
1581
1582	hdr->saddr = *saddr;
1583	hdr->daddr = *daddr;
1584}
1585
1586static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1587{
1588	struct net_device *dev = idev->dev;
1589	struct net *net = dev_net(dev);
1590	struct sock *sk = net->ipv6.igmp_sk;
1591	struct sk_buff *skb;
1592	struct mld2_report *pmr;
1593	struct in6_addr addr_buf;
1594	const struct in6_addr *saddr;
1595	int hlen = LL_RESERVED_SPACE(dev);
1596	int tlen = dev->needed_tailroom;
1597	unsigned int size = mtu + hlen + tlen;
1598	int err;
1599	u8 ra[8] = { IPPROTO_ICMPV6, 0,
1600		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
1601		     IPV6_TLV_PADN, 0 };
1602
1603	/* we assume size > sizeof(ra) here */
1604	skb = sock_alloc_send_skb(sk, size, 1, &err);
1605	if (!skb)
1606		return NULL;
1607
1608	skb->priority = TC_PRIO_CONTROL;
1609	skb_reserve(skb, hlen);
1610	skb_tailroom_reserve(skb, mtu, tlen);
1611
1612	if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1613		/* <draft-ietf-magma-mld-source-05.txt>:
1614		 * use unspecified address as the source address
1615		 * when a valid link-local address is not available.
1616		 */
1617		saddr = &in6addr_any;
1618	} else
1619		saddr = &addr_buf;
1620
1621	ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1622
1623	skb_put_data(skb, ra, sizeof(ra));
1624
1625	skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1626	skb_put(skb, sizeof(*pmr));
1627	pmr = (struct mld2_report *)skb_transport_header(skb);
1628	pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1629	pmr->mld2r_resv1 = 0;
1630	pmr->mld2r_cksum = 0;
1631	pmr->mld2r_resv2 = 0;
1632	pmr->mld2r_ngrec = 0;
1633	return skb;
1634}
1635
1636static void mld_sendpack(struct sk_buff *skb)
1637{
1638	struct ipv6hdr *pip6 = ipv6_hdr(skb);
1639	struct mld2_report *pmr =
1640			      (struct mld2_report *)skb_transport_header(skb);
1641	int payload_len, mldlen;
1642	struct inet6_dev *idev;
1643	struct net *net = dev_net(skb->dev);
1644	int err;
1645	struct flowi6 fl6;
1646	struct dst_entry *dst;
1647
1648	rcu_read_lock();
1649	idev = __in6_dev_get(skb->dev);
1650	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1651
1652	payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1653		sizeof(*pip6);
1654	mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1655	pip6->payload_len = htons(payload_len);
1656
1657	pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1658					   IPPROTO_ICMPV6,
1659					   csum_partial(skb_transport_header(skb),
1660							mldlen, 0));
1661
1662	icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1663			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1664			 skb->dev->ifindex);
1665	dst = icmp6_dst_alloc(skb->dev, &fl6);
1666
1667	err = 0;
1668	if (IS_ERR(dst)) {
1669		err = PTR_ERR(dst);
1670		dst = NULL;
1671	}
1672	skb_dst_set(skb, dst);
1673	if (err)
1674		goto err_out;
1675
1676	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1677		      net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1678		      dst_output);
1679out:
1680	if (!err) {
1681		ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1682		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1683	} else {
1684		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1685	}
1686
1687	rcu_read_unlock();
1688	return;
1689
1690err_out:
1691	kfree_skb(skb);
1692	goto out;
1693}
1694
1695static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1696{
1697	return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1698}
1699
1700static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1701	int type, struct mld2_grec **ppgr, unsigned int mtu)
1702{
1703	struct mld2_report *pmr;
1704	struct mld2_grec *pgr;
1705
1706	if (!skb) {
1707		skb = mld_newpack(pmc->idev, mtu);
1708		if (!skb)
1709			return NULL;
1710	}
1711	pgr = skb_put(skb, sizeof(struct mld2_grec));
1712	pgr->grec_type = type;
1713	pgr->grec_auxwords = 0;
1714	pgr->grec_nsrcs = 0;
1715	pgr->grec_mca = pmc->mca_addr;	/* structure copy */
1716	pmr = (struct mld2_report *)skb_transport_header(skb);
1717	pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1718	*ppgr = pgr;
1719	return skb;
1720}
1721
1722#define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
1723
1724static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1725	int type, int gdeleted, int sdeleted, int crsend)
1726{
1727	struct inet6_dev *idev = pmc->idev;
1728	struct net_device *dev = idev->dev;
1729	struct mld2_report *pmr;
1730	struct mld2_grec *pgr = NULL;
1731	struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1732	int scount, stotal, first, isquery, truncate;
1733	unsigned int mtu;
1734
1735	if (pmc->mca_flags & MAF_NOREPORT)
1736		return skb;
1737
1738	mtu = READ_ONCE(dev->mtu);
1739	if (mtu < IPV6_MIN_MTU)
1740		return skb;
1741
1742	isquery = type == MLD2_MODE_IS_INCLUDE ||
1743		  type == MLD2_MODE_IS_EXCLUDE;
1744	truncate = type == MLD2_MODE_IS_EXCLUDE ||
1745		    type == MLD2_CHANGE_TO_EXCLUDE;
1746
1747	stotal = scount = 0;
1748
1749	psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1750
1751	if (!*psf_list)
1752		goto empty_source;
1753
1754	pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1755
1756	/* EX and TO_EX get a fresh packet, if needed */
1757	if (truncate) {
1758		if (pmr && pmr->mld2r_ngrec &&
1759		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1760			if (skb)
1761				mld_sendpack(skb);
1762			skb = mld_newpack(idev, mtu);
1763		}
1764	}
1765	first = 1;
1766	psf_prev = NULL;
1767	for (psf = *psf_list; psf; psf = psf_next) {
1768		struct in6_addr *psrc;
1769
1770		psf_next = psf->sf_next;
1771
1772		if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1773			psf_prev = psf;
1774			continue;
1775		}
1776
1777		/* Based on RFC3810 6.1. Should not send source-list change
1778		 * records when there is a filter mode change.
1779		 */
1780		if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1781		     (!gdeleted && pmc->mca_crcount)) &&
1782		    (type == MLD2_ALLOW_NEW_SOURCES ||
1783		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1784			goto decrease_sf_crcount;
1785
1786		/* clear marks on query responses */
1787		if (isquery)
1788			psf->sf_gsresp = 0;
1789
1790		if (AVAILABLE(skb) < sizeof(*psrc) +
1791		    first*sizeof(struct mld2_grec)) {
1792			if (truncate && !first)
1793				break;	 /* truncate these */
1794			if (pgr)
1795				pgr->grec_nsrcs = htons(scount);
1796			if (skb)
1797				mld_sendpack(skb);
1798			skb = mld_newpack(idev, mtu);
1799			first = 1;
1800			scount = 0;
1801		}
1802		if (first) {
1803			skb = add_grhead(skb, pmc, type, &pgr, mtu);
1804			first = 0;
1805		}
1806		if (!skb)
1807			return NULL;
1808		psrc = skb_put(skb, sizeof(*psrc));
1809		*psrc = psf->sf_addr;
1810		scount++; stotal++;
1811		if ((type == MLD2_ALLOW_NEW_SOURCES ||
1812		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1813decrease_sf_crcount:
1814			psf->sf_crcount--;
1815			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1816				if (psf_prev)
1817					psf_prev->sf_next = psf->sf_next;
1818				else
1819					*psf_list = psf->sf_next;
1820				kfree(psf);
1821				continue;
1822			}
1823		}
1824		psf_prev = psf;
1825	}
1826
1827empty_source:
1828	if (!stotal) {
1829		if (type == MLD2_ALLOW_NEW_SOURCES ||
1830		    type == MLD2_BLOCK_OLD_SOURCES)
1831			return skb;
1832		if (pmc->mca_crcount || isquery || crsend) {
1833			/* make sure we have room for group header */
1834			if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1835				mld_sendpack(skb);
1836				skb = NULL; /* add_grhead will get a new one */
1837			}
1838			skb = add_grhead(skb, pmc, type, &pgr, mtu);
1839		}
1840	}
1841	if (pgr)
1842		pgr->grec_nsrcs = htons(scount);
1843
1844	if (isquery)
1845		pmc->mca_flags &= ~MAF_GSQUERY;	/* clear query state */
1846	return skb;
1847}
1848
1849static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1850{
1851	struct sk_buff *skb = NULL;
1852	int type;
1853
1854	read_lock_bh(&idev->lock);
1855	if (!pmc) {
1856		for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1857			if (pmc->mca_flags & MAF_NOREPORT)
1858				continue;
1859			spin_lock_bh(&pmc->mca_lock);
1860			if (pmc->mca_sfcount[MCAST_EXCLUDE])
1861				type = MLD2_MODE_IS_EXCLUDE;
1862			else
1863				type = MLD2_MODE_IS_INCLUDE;
1864			skb = add_grec(skb, pmc, type, 0, 0, 0);
1865			spin_unlock_bh(&pmc->mca_lock);
1866		}
1867	} else {
1868		spin_lock_bh(&pmc->mca_lock);
1869		if (pmc->mca_sfcount[MCAST_EXCLUDE])
1870			type = MLD2_MODE_IS_EXCLUDE;
1871		else
1872			type = MLD2_MODE_IS_INCLUDE;
1873		skb = add_grec(skb, pmc, type, 0, 0, 0);
1874		spin_unlock_bh(&pmc->mca_lock);
1875	}
1876	read_unlock_bh(&idev->lock);
1877	if (skb)
1878		mld_sendpack(skb);
1879}
1880
1881/*
1882 * remove zero-count source records from a source filter list
1883 */
1884static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1885{
1886	struct ip6_sf_list *psf_prev, *psf_next, *psf;
1887
1888	psf_prev = NULL;
1889	for (psf = *ppsf; psf; psf = psf_next) {
1890		psf_next = psf->sf_next;
1891		if (psf->sf_crcount == 0) {
1892			if (psf_prev)
1893				psf_prev->sf_next = psf->sf_next;
1894			else
1895				*ppsf = psf->sf_next;
1896			kfree(psf);
1897		} else
1898			psf_prev = psf;
1899	}
1900}
1901
1902static void mld_send_cr(struct inet6_dev *idev)
1903{
1904	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1905	struct sk_buff *skb = NULL;
1906	int type, dtype;
1907
1908	read_lock_bh(&idev->lock);
1909	spin_lock(&idev->mc_lock);
1910
1911	/* deleted MCA's */
1912	pmc_prev = NULL;
1913	for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1914		pmc_next = pmc->next;
1915		if (pmc->mca_sfmode == MCAST_INCLUDE) {
1916			type = MLD2_BLOCK_OLD_SOURCES;
1917			dtype = MLD2_BLOCK_OLD_SOURCES;
1918			skb = add_grec(skb, pmc, type, 1, 0, 0);
1919			skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1920		}
1921		if (pmc->mca_crcount) {
1922			if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1923				type = MLD2_CHANGE_TO_INCLUDE;
1924				skb = add_grec(skb, pmc, type, 1, 0, 0);
1925			}
1926			pmc->mca_crcount--;
1927			if (pmc->mca_crcount == 0) {
1928				mld_clear_zeros(&pmc->mca_tomb);
1929				mld_clear_zeros(&pmc->mca_sources);
1930			}
1931		}
1932		if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1933		    !pmc->mca_sources) {
1934			if (pmc_prev)
1935				pmc_prev->next = pmc_next;
1936			else
1937				idev->mc_tomb = pmc_next;
1938			in6_dev_put(pmc->idev);
1939			kfree(pmc);
1940		} else
1941			pmc_prev = pmc;
1942	}
1943	spin_unlock(&idev->mc_lock);
1944
1945	/* change recs */
1946	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1947		spin_lock_bh(&pmc->mca_lock);
1948		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1949			type = MLD2_BLOCK_OLD_SOURCES;
1950			dtype = MLD2_ALLOW_NEW_SOURCES;
1951		} else {
1952			type = MLD2_ALLOW_NEW_SOURCES;
1953			dtype = MLD2_BLOCK_OLD_SOURCES;
1954		}
1955		skb = add_grec(skb, pmc, type, 0, 0, 0);
1956		skb = add_grec(skb, pmc, dtype, 0, 1, 0);	/* deleted sources */
1957
1958		/* filter mode changes */
1959		if (pmc->mca_crcount) {
1960			if (pmc->mca_sfmode == MCAST_EXCLUDE)
1961				type = MLD2_CHANGE_TO_EXCLUDE;
1962			else
1963				type = MLD2_CHANGE_TO_INCLUDE;
1964			skb = add_grec(skb, pmc, type, 0, 0, 0);
1965			pmc->mca_crcount--;
1966		}
1967		spin_unlock_bh(&pmc->mca_lock);
1968	}
1969	read_unlock_bh(&idev->lock);
1970	if (!skb)
1971		return;
1972	(void) mld_sendpack(skb);
1973}
1974
1975static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1976{
1977	struct net *net = dev_net(dev);
1978	struct sock *sk = net->ipv6.igmp_sk;
1979	struct inet6_dev *idev;
1980	struct sk_buff *skb;
1981	struct mld_msg *hdr;
1982	const struct in6_addr *snd_addr, *saddr;
1983	struct in6_addr addr_buf;
1984	int hlen = LL_RESERVED_SPACE(dev);
1985	int tlen = dev->needed_tailroom;
1986	int err, len, payload_len, full_len;
1987	u8 ra[8] = { IPPROTO_ICMPV6, 0,
1988		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
1989		     IPV6_TLV_PADN, 0 };
1990	struct flowi6 fl6;
1991	struct dst_entry *dst;
1992
1993	if (type == ICMPV6_MGM_REDUCTION)
1994		snd_addr = &in6addr_linklocal_allrouters;
1995	else
1996		snd_addr = addr;
1997
1998	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
1999	payload_len = len + sizeof(ra);
2000	full_len = sizeof(struct ipv6hdr) + payload_len;
2001
2002	rcu_read_lock();
2003	IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2004		      IPSTATS_MIB_OUT, full_len);
2005	rcu_read_unlock();
2006
2007	skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2008
2009	if (!skb) {
2010		rcu_read_lock();
2011		IP6_INC_STATS(net, __in6_dev_get(dev),
2012			      IPSTATS_MIB_OUTDISCARDS);
2013		rcu_read_unlock();
2014		return;
2015	}
2016	skb->priority = TC_PRIO_CONTROL;
2017	skb_reserve(skb, hlen);
2018
2019	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2020		/* <draft-ietf-magma-mld-source-05.txt>:
2021		 * use unspecified address as the source address
2022		 * when a valid link-local address is not available.
2023		 */
2024		saddr = &in6addr_any;
2025	} else
2026		saddr = &addr_buf;
2027
2028	ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2029
2030	skb_put_data(skb, ra, sizeof(ra));
2031
2032	hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2033	hdr->mld_type = type;
2034	hdr->mld_mca = *addr;
2035
2036	hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2037					 IPPROTO_ICMPV6,
2038					 csum_partial(hdr, len, 0));
2039
2040	rcu_read_lock();
2041	idev = __in6_dev_get(skb->dev);
2042
2043	icmpv6_flow_init(sk, &fl6, type,
2044			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2045			 skb->dev->ifindex);
2046	dst = icmp6_dst_alloc(skb->dev, &fl6);
2047	if (IS_ERR(dst)) {
2048		err = PTR_ERR(dst);
2049		goto err_out;
2050	}
2051
2052	skb_dst_set(skb, dst);
2053	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2054		      net, sk, skb, NULL, skb->dev,
2055		      dst_output);
2056out:
2057	if (!err) {
2058		ICMP6MSGOUT_INC_STATS(net, idev, type);
2059		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2060	} else
2061		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2062
2063	rcu_read_unlock();
2064	return;
2065
2066err_out:
2067	kfree_skb(skb);
2068	goto out;
2069}
2070
2071static void mld_send_initial_cr(struct inet6_dev *idev)
2072{
2073	struct sk_buff *skb;
2074	struct ifmcaddr6 *pmc;
2075	int type;
2076
2077	if (mld_in_v1_mode(idev))
2078		return;
2079
2080	skb = NULL;
2081	read_lock_bh(&idev->lock);
2082	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2083		spin_lock_bh(&pmc->mca_lock);
2084		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2085			type = MLD2_CHANGE_TO_EXCLUDE;
2086		else
2087			type = MLD2_ALLOW_NEW_SOURCES;
2088		skb = add_grec(skb, pmc, type, 0, 0, 1);
2089		spin_unlock_bh(&pmc->mca_lock);
2090	}
2091	read_unlock_bh(&idev->lock);
2092	if (skb)
2093		mld_sendpack(skb);
2094}
2095
2096void ipv6_mc_dad_complete(struct inet6_dev *idev)
2097{
2098	idev->mc_dad_count = idev->mc_qrv;
2099	if (idev->mc_dad_count) {
2100		mld_send_initial_cr(idev);
2101		idev->mc_dad_count--;
2102		if (idev->mc_dad_count)
2103			mld_dad_start_timer(idev,
2104					    unsolicited_report_interval(idev));
2105	}
2106}
2107
2108static void mld_dad_timer_expire(struct timer_list *t)
2109{
2110	struct inet6_dev *idev = from_timer(idev, t, mc_dad_timer);
2111
2112	mld_send_initial_cr(idev);
2113	if (idev->mc_dad_count) {
2114		idev->mc_dad_count--;
2115		if (idev->mc_dad_count)
2116			mld_dad_start_timer(idev,
2117					    unsolicited_report_interval(idev));
2118	}
2119	in6_dev_put(idev);
2120}
2121
2122static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2123	const struct in6_addr *psfsrc)
2124{
2125	struct ip6_sf_list *psf, *psf_prev;
2126	int rv = 0;
2127
2128	psf_prev = NULL;
2129	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2130		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2131			break;
2132		psf_prev = psf;
2133	}
2134	if (!psf || psf->sf_count[sfmode] == 0) {
2135		/* source filter not found, or count wrong =>  bug */
2136		return -ESRCH;
2137	}
2138	psf->sf_count[sfmode]--;
2139	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2140		struct inet6_dev *idev = pmc->idev;
2141
2142		/* no more filters for this source */
2143		if (psf_prev)
2144			psf_prev->sf_next = psf->sf_next;
2145		else
2146			pmc->mca_sources = psf->sf_next;
2147		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2148		    !mld_in_v1_mode(idev)) {
2149			psf->sf_crcount = idev->mc_qrv;
2150			psf->sf_next = pmc->mca_tomb;
2151			pmc->mca_tomb = psf;
2152			rv = 1;
2153		} else
2154			kfree(psf);
2155	}
2156	return rv;
2157}
2158
2159static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2160			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2161			  int delta)
2162{
2163	struct ifmcaddr6 *pmc;
2164	int	changerec = 0;
2165	int	i, err;
2166
2167	if (!idev)
2168		return -ENODEV;
2169	read_lock_bh(&idev->lock);
2170	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2171		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2172			break;
2173	}
2174	if (!pmc) {
2175		/* MCA not found?? bug */
2176		read_unlock_bh(&idev->lock);
2177		return -ESRCH;
2178	}
2179	spin_lock_bh(&pmc->mca_lock);
2180	sf_markstate(pmc);
2181	if (!delta) {
2182		if (!pmc->mca_sfcount[sfmode]) {
2183			spin_unlock_bh(&pmc->mca_lock);
2184			read_unlock_bh(&idev->lock);
2185			return -EINVAL;
2186		}
2187		pmc->mca_sfcount[sfmode]--;
2188	}
2189	err = 0;
2190	for (i = 0; i < sfcount; i++) {
2191		int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2192
2193		changerec |= rv > 0;
2194		if (!err && rv < 0)
2195			err = rv;
2196	}
2197	if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2198	    pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2199	    pmc->mca_sfcount[MCAST_INCLUDE]) {
2200		struct ip6_sf_list *psf;
2201
2202		/* filter mode change */
2203		pmc->mca_sfmode = MCAST_INCLUDE;
2204		pmc->mca_crcount = idev->mc_qrv;
2205		idev->mc_ifc_count = pmc->mca_crcount;
2206		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2207			psf->sf_crcount = 0;
2208		mld_ifc_event(pmc->idev);
2209	} else if (sf_setstate(pmc) || changerec)
2210		mld_ifc_event(pmc->idev);
2211	spin_unlock_bh(&pmc->mca_lock);
2212	read_unlock_bh(&idev->lock);
2213	return err;
2214}
2215
2216/*
2217 * Add multicast single-source filter to the interface list
2218 */
2219static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2220	const struct in6_addr *psfsrc)
2221{
2222	struct ip6_sf_list *psf, *psf_prev;
2223
2224	psf_prev = NULL;
2225	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2226		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2227			break;
2228		psf_prev = psf;
2229	}
2230	if (!psf) {
2231		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2232		if (!psf)
2233			return -ENOBUFS;
2234
2235		psf->sf_addr = *psfsrc;
2236		if (psf_prev) {
2237			psf_prev->sf_next = psf;
2238		} else
2239			pmc->mca_sources = psf;
2240	}
2241	psf->sf_count[sfmode]++;
2242	return 0;
2243}
2244
2245static void sf_markstate(struct ifmcaddr6 *pmc)
2246{
2247	struct ip6_sf_list *psf;
2248	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2249
2250	for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2251		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2252			psf->sf_oldin = mca_xcount ==
2253				psf->sf_count[MCAST_EXCLUDE] &&
2254				!psf->sf_count[MCAST_INCLUDE];
2255		} else
2256			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2257}
2258
2259static int sf_setstate(struct ifmcaddr6 *pmc)
2260{
2261	struct ip6_sf_list *psf, *dpsf;
2262	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2263	int qrv = pmc->idev->mc_qrv;
2264	int new_in, rv;
2265
2266	rv = 0;
2267	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2268		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2269			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2270				!psf->sf_count[MCAST_INCLUDE];
2271		} else
2272			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2273		if (new_in) {
2274			if (!psf->sf_oldin) {
2275				struct ip6_sf_list *prev = NULL;
2276
2277				for (dpsf = pmc->mca_tomb; dpsf;
2278				     dpsf = dpsf->sf_next) {
2279					if (ipv6_addr_equal(&dpsf->sf_addr,
2280					    &psf->sf_addr))
2281						break;
2282					prev = dpsf;
2283				}
2284				if (dpsf) {
2285					if (prev)
2286						prev->sf_next = dpsf->sf_next;
2287					else
2288						pmc->mca_tomb = dpsf->sf_next;
2289					kfree(dpsf);
2290				}
2291				psf->sf_crcount = qrv;
2292				rv++;
2293			}
2294		} else if (psf->sf_oldin) {
2295			psf->sf_crcount = 0;
2296			/*
2297			 * add or update "delete" records if an active filter
2298			 * is now inactive
2299			 */
2300			for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2301				if (ipv6_addr_equal(&dpsf->sf_addr,
2302				    &psf->sf_addr))
2303					break;
2304			if (!dpsf) {
2305				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2306				if (!dpsf)
2307					continue;
2308				*dpsf = *psf;
2309				/* pmc->mca_lock held by callers */
2310				dpsf->sf_next = pmc->mca_tomb;
2311				pmc->mca_tomb = dpsf;
2312			}
2313			dpsf->sf_crcount = qrv;
2314			rv++;
2315		}
2316	}
2317	return rv;
2318}
2319
2320/*
2321 * Add multicast source filter list to the interface list
2322 */
2323static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2324			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2325			  int delta)
2326{
2327	struct ifmcaddr6 *pmc;
2328	int	isexclude;
2329	int	i, err;
2330
2331	if (!idev)
2332		return -ENODEV;
2333	read_lock_bh(&idev->lock);
2334	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2335		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2336			break;
2337	}
2338	if (!pmc) {
2339		/* MCA not found?? bug */
2340		read_unlock_bh(&idev->lock);
2341		return -ESRCH;
2342	}
2343	spin_lock_bh(&pmc->mca_lock);
2344
2345	sf_markstate(pmc);
2346	isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2347	if (!delta)
2348		pmc->mca_sfcount[sfmode]++;
2349	err = 0;
2350	for (i = 0; i < sfcount; i++) {
2351		err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2352		if (err)
2353			break;
2354	}
2355	if (err) {
2356		int j;
2357
2358		if (!delta)
2359			pmc->mca_sfcount[sfmode]--;
2360		for (j = 0; j < i; j++)
2361			ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2362	} else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2363		struct ip6_sf_list *psf;
2364
2365		/* filter mode change */
2366		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2367			pmc->mca_sfmode = MCAST_EXCLUDE;
2368		else if (pmc->mca_sfcount[MCAST_INCLUDE])
2369			pmc->mca_sfmode = MCAST_INCLUDE;
2370		/* else no filters; keep old mode for reports */
2371
2372		pmc->mca_crcount = idev->mc_qrv;
2373		idev->mc_ifc_count = pmc->mca_crcount;
2374		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2375			psf->sf_crcount = 0;
2376		mld_ifc_event(idev);
2377	} else if (sf_setstate(pmc))
2378		mld_ifc_event(idev);
2379	spin_unlock_bh(&pmc->mca_lock);
2380	read_unlock_bh(&idev->lock);
2381	return err;
2382}
2383
2384static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2385{
2386	struct ip6_sf_list *psf, *nextpsf;
2387
2388	for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2389		nextpsf = psf->sf_next;
2390		kfree(psf);
2391	}
2392	pmc->mca_tomb = NULL;
2393	for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2394		nextpsf = psf->sf_next;
2395		kfree(psf);
2396	}
2397	pmc->mca_sources = NULL;
2398	pmc->mca_sfmode = MCAST_EXCLUDE;
2399	pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2400	pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2401}
2402
2403
2404static void igmp6_join_group(struct ifmcaddr6 *ma)
2405{
2406	unsigned long delay;
2407
2408	if (ma->mca_flags & MAF_NOREPORT)
2409		return;
2410
2411	igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2412
2413	delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2414
2415	spin_lock_bh(&ma->mca_lock);
2416	if (del_timer(&ma->mca_timer)) {
2417		refcount_dec(&ma->mca_refcnt);
2418		delay = ma->mca_timer.expires - jiffies;
2419	}
2420
2421	if (!mod_timer(&ma->mca_timer, jiffies + delay))
2422		refcount_inc(&ma->mca_refcnt);
2423	ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2424	spin_unlock_bh(&ma->mca_lock);
2425}
2426
2427static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2428			    struct inet6_dev *idev)
2429{
2430	int err;
2431
2432	write_lock_bh(&iml->sflock);
2433	if (!iml->sflist) {
2434		/* any-source empty exclude case */
2435		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2436	} else {
2437		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2438				iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2439		sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2440		iml->sflist = NULL;
2441	}
2442	write_unlock_bh(&iml->sflock);
2443	return err;
2444}
2445
2446static void igmp6_leave_group(struct ifmcaddr6 *ma)
2447{
2448	if (mld_in_v1_mode(ma->idev)) {
2449		if (ma->mca_flags & MAF_LAST_REPORTER)
2450			igmp6_send(&ma->mca_addr, ma->idev->dev,
2451				ICMPV6_MGM_REDUCTION);
2452	} else {
2453		mld_add_delrec(ma->idev, ma);
2454		mld_ifc_event(ma->idev);
2455	}
2456}
2457
2458static void mld_gq_timer_expire(struct timer_list *t)
2459{
2460	struct inet6_dev *idev = from_timer(idev, t, mc_gq_timer);
2461
2462	idev->mc_gq_running = 0;
2463	mld_send_report(idev, NULL);
2464	in6_dev_put(idev);
2465}
2466
2467static void mld_ifc_timer_expire(struct timer_list *t)
2468{
2469	struct inet6_dev *idev = from_timer(idev, t, mc_ifc_timer);
2470
2471	mld_send_cr(idev);
2472	if (idev->mc_ifc_count) {
2473		idev->mc_ifc_count--;
2474		if (idev->mc_ifc_count)
2475			mld_ifc_start_timer(idev,
2476					    unsolicited_report_interval(idev));
2477	}
2478	in6_dev_put(idev);
2479}
2480
2481static void mld_ifc_event(struct inet6_dev *idev)
2482{
2483	if (mld_in_v1_mode(idev))
2484		return;
2485	idev->mc_ifc_count = idev->mc_qrv;
2486	mld_ifc_start_timer(idev, 1);
2487}
2488
2489static void igmp6_timer_handler(struct timer_list *t)
2490{
2491	struct ifmcaddr6 *ma = from_timer(ma, t, mca_timer);
2492
2493	if (mld_in_v1_mode(ma->idev))
2494		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2495	else
2496		mld_send_report(ma->idev, ma);
2497
2498	spin_lock(&ma->mca_lock);
2499	ma->mca_flags |=  MAF_LAST_REPORTER;
2500	ma->mca_flags &= ~MAF_TIMER_RUNNING;
2501	spin_unlock(&ma->mca_lock);
2502	ma_put(ma);
2503}
2504
2505/* Device changing type */
2506
2507void ipv6_mc_unmap(struct inet6_dev *idev)
2508{
2509	struct ifmcaddr6 *i;
2510
2511	/* Install multicast list, except for all-nodes (already installed) */
2512
2513	read_lock_bh(&idev->lock);
2514	for (i = idev->mc_list; i; i = i->next)
2515		igmp6_group_dropped(i);
2516	read_unlock_bh(&idev->lock);
2517}
2518
2519void ipv6_mc_remap(struct inet6_dev *idev)
2520{
2521	ipv6_mc_up(idev);
2522}
2523
2524/* Device going down */
2525
2526void ipv6_mc_down(struct inet6_dev *idev)
2527{
2528	struct ifmcaddr6 *i;
2529
2530	/* Withdraw multicast list */
2531
2532	read_lock_bh(&idev->lock);
2533
2534	for (i = idev->mc_list; i; i = i->next)
2535		igmp6_group_dropped(i);
2536
2537	/* Should stop timer after group drop. or we will
2538	 * start timer again in mld_ifc_event()
2539	 */
2540	mld_ifc_stop_timer(idev);
2541	mld_gq_stop_timer(idev);
2542	mld_dad_stop_timer(idev);
2543	read_unlock_bh(&idev->lock);
2544}
2545
2546static void ipv6_mc_reset(struct inet6_dev *idev)
2547{
2548	idev->mc_qrv = sysctl_mld_qrv;
2549	idev->mc_qi = MLD_QI_DEFAULT;
2550	idev->mc_qri = MLD_QRI_DEFAULT;
2551	idev->mc_v1_seen = 0;
2552	idev->mc_maxdelay = unsolicited_report_interval(idev);
2553}
2554
2555/* Device going up */
2556
2557void ipv6_mc_up(struct inet6_dev *idev)
2558{
2559	struct ifmcaddr6 *i;
2560
2561	/* Install multicast list, except for all-nodes (already installed) */
2562
2563	read_lock_bh(&idev->lock);
2564	ipv6_mc_reset(idev);
2565	for (i = idev->mc_list; i; i = i->next) {
2566		mld_del_delrec(idev, i);
2567		igmp6_group_added(i);
2568	}
2569	read_unlock_bh(&idev->lock);
2570}
2571
2572/* IPv6 device initialization. */
2573
2574void ipv6_mc_init_dev(struct inet6_dev *idev)
2575{
2576	write_lock_bh(&idev->lock);
2577	spin_lock_init(&idev->mc_lock);
2578	idev->mc_gq_running = 0;
2579	timer_setup(&idev->mc_gq_timer, mld_gq_timer_expire, 0);
2580	idev->mc_tomb = NULL;
2581	idev->mc_ifc_count = 0;
2582	timer_setup(&idev->mc_ifc_timer, mld_ifc_timer_expire, 0);
2583	timer_setup(&idev->mc_dad_timer, mld_dad_timer_expire, 0);
2584	ipv6_mc_reset(idev);
2585	write_unlock_bh(&idev->lock);
2586}
2587
2588/*
2589 *	Device is about to be destroyed: clean up.
2590 */
2591
2592void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2593{
2594	struct ifmcaddr6 *i;
2595
2596	/* Deactivate timers */
2597	ipv6_mc_down(idev);
2598	mld_clear_delrec(idev);
2599
2600	/* Delete all-nodes address. */
2601	/* We cannot call ipv6_dev_mc_dec() directly, our caller in
2602	 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2603	 * fail.
2604	 */
2605	__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2606
2607	if (idev->cnf.forwarding)
2608		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2609
2610	write_lock_bh(&idev->lock);
2611	while ((i = idev->mc_list) != NULL) {
2612		idev->mc_list = i->next;
2613
2614		write_unlock_bh(&idev->lock);
2615		ip6_mc_clear_src(i);
2616		ma_put(i);
2617		write_lock_bh(&idev->lock);
2618	}
2619	write_unlock_bh(&idev->lock);
2620}
2621
2622static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2623{
2624	struct ifmcaddr6 *pmc;
2625
2626	ASSERT_RTNL();
2627
2628	if (mld_in_v1_mode(idev)) {
2629		read_lock_bh(&idev->lock);
2630		for (pmc = idev->mc_list; pmc; pmc = pmc->next)
2631			igmp6_join_group(pmc);
2632		read_unlock_bh(&idev->lock);
2633	} else
2634		mld_send_report(idev, NULL);
2635}
2636
2637static int ipv6_mc_netdev_event(struct notifier_block *this,
2638				unsigned long event,
2639				void *ptr)
2640{
2641	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2642	struct inet6_dev *idev = __in6_dev_get(dev);
2643
2644	switch (event) {
2645	case NETDEV_RESEND_IGMP:
2646		if (idev)
2647			ipv6_mc_rejoin_groups(idev);
2648		break;
2649	default:
2650		break;
2651	}
2652
2653	return NOTIFY_DONE;
2654}
2655
2656static struct notifier_block igmp6_netdev_notifier = {
2657	.notifier_call = ipv6_mc_netdev_event,
2658};
2659
2660#ifdef CONFIG_PROC_FS
2661struct igmp6_mc_iter_state {
2662	struct seq_net_private p;
2663	struct net_device *dev;
2664	struct inet6_dev *idev;
2665};
2666
2667#define igmp6_mc_seq_private(seq)	((struct igmp6_mc_iter_state *)(seq)->private)
2668
2669static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2670{
2671	struct ifmcaddr6 *im = NULL;
2672	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2673	struct net *net = seq_file_net(seq);
2674
2675	state->idev = NULL;
2676	for_each_netdev_rcu(net, state->dev) {
2677		struct inet6_dev *idev;
2678		idev = __in6_dev_get(state->dev);
2679		if (!idev)
2680			continue;
2681		read_lock_bh(&idev->lock);
2682		im = idev->mc_list;
2683		if (im) {
2684			state->idev = idev;
2685			break;
2686		}
2687		read_unlock_bh(&idev->lock);
2688	}
2689	return im;
2690}
2691
2692static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2693{
2694	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2695
2696	im = im->next;
2697	while (!im) {
2698		if (likely(state->idev))
2699			read_unlock_bh(&state->idev->lock);
2700
2701		state->dev = next_net_device_rcu(state->dev);
2702		if (!state->dev) {
2703			state->idev = NULL;
2704			break;
2705		}
2706		state->idev = __in6_dev_get(state->dev);
2707		if (!state->idev)
2708			continue;
2709		read_lock_bh(&state->idev->lock);
2710		im = state->idev->mc_list;
2711	}
2712	return im;
2713}
2714
2715static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2716{
2717	struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2718	if (im)
2719		while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2720			--pos;
2721	return pos ? NULL : im;
2722}
2723
2724static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2725	__acquires(RCU)
2726{
2727	rcu_read_lock();
2728	return igmp6_mc_get_idx(seq, *pos);
2729}
2730
2731static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2732{
2733	struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2734
2735	++*pos;
2736	return im;
2737}
2738
2739static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2740	__releases(RCU)
2741{
2742	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2743
2744	if (likely(state->idev)) {
2745		read_unlock_bh(&state->idev->lock);
2746		state->idev = NULL;
2747	}
2748	state->dev = NULL;
2749	rcu_read_unlock();
2750}
2751
2752static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2753{
2754	struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2755	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2756
2757	seq_printf(seq,
2758		   "%-4d %-15s %pi6 %5d %08X %ld\n",
2759		   state->dev->ifindex, state->dev->name,
2760		   &im->mca_addr,
2761		   im->mca_users, im->mca_flags,
2762		   (im->mca_flags&MAF_TIMER_RUNNING) ?
2763		   jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2764	return 0;
2765}
2766
2767static const struct seq_operations igmp6_mc_seq_ops = {
2768	.start	=	igmp6_mc_seq_start,
2769	.next	=	igmp6_mc_seq_next,
2770	.stop	=	igmp6_mc_seq_stop,
2771	.show	=	igmp6_mc_seq_show,
2772};
2773
2774struct igmp6_mcf_iter_state {
2775	struct seq_net_private p;
2776	struct net_device *dev;
2777	struct inet6_dev *idev;
2778	struct ifmcaddr6 *im;
2779};
2780
2781#define igmp6_mcf_seq_private(seq)	((struct igmp6_mcf_iter_state *)(seq)->private)
2782
2783static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2784{
2785	struct ip6_sf_list *psf = NULL;
2786	struct ifmcaddr6 *im = NULL;
2787	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2788	struct net *net = seq_file_net(seq);
2789
2790	state->idev = NULL;
2791	state->im = NULL;
2792	for_each_netdev_rcu(net, state->dev) {
2793		struct inet6_dev *idev;
2794		idev = __in6_dev_get(state->dev);
2795		if (unlikely(idev == NULL))
2796			continue;
2797		read_lock_bh(&idev->lock);
2798		im = idev->mc_list;
2799		if (likely(im)) {
2800			spin_lock_bh(&im->mca_lock);
2801			psf = im->mca_sources;
2802			if (likely(psf)) {
2803				state->im = im;
2804				state->idev = idev;
2805				break;
2806			}
2807			spin_unlock_bh(&im->mca_lock);
2808		}
2809		read_unlock_bh(&idev->lock);
2810	}
2811	return psf;
2812}
2813
2814static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2815{
2816	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2817
2818	psf = psf->sf_next;
2819	while (!psf) {
2820		spin_unlock_bh(&state->im->mca_lock);
2821		state->im = state->im->next;
2822		while (!state->im) {
2823			if (likely(state->idev))
2824				read_unlock_bh(&state->idev->lock);
2825
2826			state->dev = next_net_device_rcu(state->dev);
2827			if (!state->dev) {
2828				state->idev = NULL;
2829				goto out;
2830			}
2831			state->idev = __in6_dev_get(state->dev);
2832			if (!state->idev)
2833				continue;
2834			read_lock_bh(&state->idev->lock);
2835			state->im = state->idev->mc_list;
2836		}
2837		if (!state->im)
2838			break;
2839		spin_lock_bh(&state->im->mca_lock);
2840		psf = state->im->mca_sources;
2841	}
2842out:
2843	return psf;
2844}
2845
2846static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2847{
2848	struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2849	if (psf)
2850		while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2851			--pos;
2852	return pos ? NULL : psf;
2853}
2854
2855static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2856	__acquires(RCU)
2857{
2858	rcu_read_lock();
2859	return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2860}
2861
2862static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2863{
2864	struct ip6_sf_list *psf;
2865	if (v == SEQ_START_TOKEN)
2866		psf = igmp6_mcf_get_first(seq);
2867	else
2868		psf = igmp6_mcf_get_next(seq, v);
2869	++*pos;
2870	return psf;
2871}
2872
2873static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2874	__releases(RCU)
2875{
2876	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2877	if (likely(state->im)) {
2878		spin_unlock_bh(&state->im->mca_lock);
2879		state->im = NULL;
2880	}
2881	if (likely(state->idev)) {
2882		read_unlock_bh(&state->idev->lock);
2883		state->idev = NULL;
2884	}
2885	state->dev = NULL;
2886	rcu_read_unlock();
2887}
2888
2889static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2890{
2891	struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2892	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2893
2894	if (v == SEQ_START_TOKEN) {
2895		seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2896	} else {
2897		seq_printf(seq,
2898			   "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2899			   state->dev->ifindex, state->dev->name,
2900			   &state->im->mca_addr,
2901			   &psf->sf_addr,
2902			   psf->sf_count[MCAST_INCLUDE],
2903			   psf->sf_count[MCAST_EXCLUDE]);
2904	}
2905	return 0;
2906}
2907
2908static const struct seq_operations igmp6_mcf_seq_ops = {
2909	.start	=	igmp6_mcf_seq_start,
2910	.next	=	igmp6_mcf_seq_next,
2911	.stop	=	igmp6_mcf_seq_stop,
2912	.show	=	igmp6_mcf_seq_show,
2913};
2914
2915static int __net_init igmp6_proc_init(struct net *net)
2916{
2917	int err;
2918
2919	err = -ENOMEM;
2920	if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
2921			sizeof(struct igmp6_mc_iter_state)))
2922		goto out;
2923	if (!proc_create_net("mcfilter6", 0444, net->proc_net,
2924			&igmp6_mcf_seq_ops,
2925			sizeof(struct igmp6_mcf_iter_state)))
2926		goto out_proc_net_igmp6;
2927
2928	err = 0;
2929out:
2930	return err;
2931
2932out_proc_net_igmp6:
2933	remove_proc_entry("igmp6", net->proc_net);
2934	goto out;
2935}
2936
2937static void __net_exit igmp6_proc_exit(struct net *net)
2938{
2939	remove_proc_entry("mcfilter6", net->proc_net);
2940	remove_proc_entry("igmp6", net->proc_net);
2941}
2942#else
2943static inline int igmp6_proc_init(struct net *net)
2944{
2945	return 0;
2946}
2947static inline void igmp6_proc_exit(struct net *net)
2948{
2949}
2950#endif
2951
2952static int __net_init igmp6_net_init(struct net *net)
2953{
2954	int err;
2955
2956	err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2957				   SOCK_RAW, IPPROTO_ICMPV6, net);
2958	if (err < 0) {
2959		pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2960		       err);
2961		goto out;
2962	}
2963
2964	inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2965
2966	err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2967				   SOCK_RAW, IPPROTO_ICMPV6, net);
2968	if (err < 0) {
2969		pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2970		       err);
2971		goto out_sock_create;
2972	}
2973
2974	err = igmp6_proc_init(net);
2975	if (err)
2976		goto out_sock_create_autojoin;
2977
2978	return 0;
2979
2980out_sock_create_autojoin:
2981	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2982out_sock_create:
2983	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2984out:
2985	return err;
2986}
2987
2988static void __net_exit igmp6_net_exit(struct net *net)
2989{
2990	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2991	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2992	igmp6_proc_exit(net);
2993}
2994
2995static struct pernet_operations igmp6_net_ops = {
2996	.init = igmp6_net_init,
2997	.exit = igmp6_net_exit,
2998};
2999
3000int __init igmp6_init(void)
3001{
3002	return register_pernet_subsys(&igmp6_net_ops);
3003}
3004
3005int __init igmp6_late_init(void)
3006{
3007	return register_netdevice_notifier(&igmp6_netdev_notifier);
3008}
3009
3010void igmp6_cleanup(void)
3011{
3012	unregister_pernet_subsys(&igmp6_net_ops);
3013}
3014
3015void igmp6_late_cleanup(void)
3016{
3017	unregister_netdevice_notifier(&igmp6_netdev_notifier);
3018}
3019