xref: /linux/net/ipv6/mcast.c (revision 348f968b89bfeec0bb53dd82dba58b94d97fbd34)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Multicast support for IPv6
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *
9  *	Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
10  */
11 
12 /* Changes:
13  *
14  *	yoshfuji	: fix format of router-alert option
15  *	YOSHIFUJI Hideaki @USAGI:
16  *		Fixed source address for MLD message based on
17  *		<draft-ietf-magma-mld-source-05.txt>.
18  *	YOSHIFUJI Hideaki @USAGI:
19  *		- Ignore Queries for invalid addresses.
20  *		- MLD for link-local addresses.
21  *	David L Stevens <dlstevens@us.ibm.com>:
22  *		- MLDv2 support
23  */
24 
25 #include <linux/module.h>
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/string.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/jiffies.h>
32 #include <linux/net.h>
33 #include <linux/in.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_addr.h>
37 #include <linux/if_arp.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/init.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/slab.h>
44 #include <linux/pkt_sched.h>
45 #include <net/mld.h>
46 #include <linux/workqueue.h>
47 
48 #include <linux/netfilter.h>
49 #include <linux/netfilter_ipv6.h>
50 
51 #include <net/net_namespace.h>
52 #include <net/netlink.h>
53 #include <net/sock.h>
54 #include <net/snmp.h>
55 
56 #include <net/ipv6.h>
57 #include <net/protocol.h>
58 #include <net/if_inet6.h>
59 #include <net/ndisc.h>
60 #include <net/addrconf.h>
61 #include <net/ip6_route.h>
62 #include <net/inet_common.h>
63 
64 #include <net/ip6_checksum.h>
65 
66 /* Ensure that we have struct in6_addr aligned on 32bit word. */
67 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
68 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
69 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
70 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
71 };
72 
73 static struct workqueue_struct *mld_wq;
74 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
75 
76 static void igmp6_join_group(struct ifmcaddr6 *ma);
77 static void igmp6_leave_group(struct ifmcaddr6 *ma);
78 static void mld_mca_work(struct work_struct *work);
79 
80 static void mld_ifc_event(struct inet6_dev *idev);
81 static bool mld_in_v1_mode(const struct inet6_dev *idev);
82 static int sf_setstate(struct ifmcaddr6 *pmc);
83 static void sf_markstate(struct ifmcaddr6 *pmc);
84 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
85 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
86 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
87 			  int delta);
88 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
89 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
90 			  int delta);
91 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
92 			    struct inet6_dev *idev);
93 static int __ipv6_dev_mc_inc(struct net_device *dev,
94 			     const struct in6_addr *addr, unsigned int mode);
95 
96 #define MLD_QRV_DEFAULT		2
97 /* RFC3810, 9.2. Query Interval */
98 #define MLD_QI_DEFAULT		(125 * HZ)
99 /* RFC3810, 9.3. Query Response Interval */
100 #define MLD_QRI_DEFAULT		(10 * HZ)
101 
102 /* RFC3810, 8.1 Query Version Distinctions */
103 #define MLD_V1_QUERY_LEN	24
104 #define MLD_V2_QUERY_LEN_MIN	28
105 
106 #define IPV6_MLD_MAX_MSF	64
107 
108 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
109 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
110 
111 /*
112  *	socket join on multicast group
113  */
114 #define mc_dereference(e, idev) \
115 	rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock))
116 
117 #define sock_dereference(e, sk) \
118 	rcu_dereference_protected(e, lockdep_sock_is_held(sk))
119 
120 #define for_each_pmc_socklock(np, sk, pmc)			\
121 	for (pmc = sock_dereference((np)->ipv6_mc_list, sk);	\
122 	     pmc;						\
123 	     pmc = sock_dereference(pmc->next, sk))
124 
125 #define for_each_pmc_rcu(np, pmc)				\
126 	for (pmc = rcu_dereference((np)->ipv6_mc_list);		\
127 	     pmc;						\
128 	     pmc = rcu_dereference(pmc->next))
129 
130 #define for_each_psf_mclock(mc, psf)				\
131 	for (psf = mc_dereference((mc)->mca_sources, mc->idev);	\
132 	     psf;						\
133 	     psf = mc_dereference(psf->sf_next, mc->idev))
134 
135 #define for_each_psf_rcu(mc, psf)				\
136 	for (psf = rcu_dereference((mc)->mca_sources);		\
137 	     psf;						\
138 	     psf = rcu_dereference(psf->sf_next))
139 
140 #define for_each_psf_tomb(mc, psf)				\
141 	for (psf = mc_dereference((mc)->mca_tomb, mc->idev);	\
142 	     psf;						\
143 	     psf = mc_dereference(psf->sf_next, mc->idev))
144 
145 #define for_each_mc_mclock(idev, mc)				\
146 	for (mc = mc_dereference((idev)->mc_list, idev);	\
147 	     mc;						\
148 	     mc = mc_dereference(mc->next, idev))
149 
150 #define for_each_mc_rcu(idev, mc)				\
151 	for (mc = rcu_dereference((idev)->mc_list);             \
152 	     mc;                                                \
153 	     mc = rcu_dereference(mc->next))
154 
155 #define for_each_mc_tomb(idev, mc)				\
156 	for (mc = mc_dereference((idev)->mc_tomb, idev);	\
157 	     mc;						\
158 	     mc = mc_dereference(mc->next, idev))
159 
unsolicited_report_interval(struct inet6_dev * idev)160 static int unsolicited_report_interval(struct inet6_dev *idev)
161 {
162 	int iv;
163 
164 	if (mld_in_v1_mode(idev))
165 		iv = READ_ONCE(idev->cnf.mldv1_unsolicited_report_interval);
166 	else
167 		iv = READ_ONCE(idev->cnf.mldv2_unsolicited_report_interval);
168 
169 	return iv > 0 ? iv : 1;
170 }
171 
__ipv6_sock_mc_join(struct sock * sk,int ifindex,const struct in6_addr * addr,unsigned int mode)172 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
173 			       const struct in6_addr *addr, unsigned int mode)
174 {
175 	struct net_device *dev = NULL;
176 	struct ipv6_mc_socklist *mc_lst;
177 	struct ipv6_pinfo *np = inet6_sk(sk);
178 	struct net *net = sock_net(sk);
179 	int err;
180 
181 	ASSERT_RTNL();
182 
183 	if (!ipv6_addr_is_multicast(addr))
184 		return -EINVAL;
185 
186 	for_each_pmc_socklock(np, sk, mc_lst) {
187 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
188 		    ipv6_addr_equal(&mc_lst->addr, addr))
189 			return -EADDRINUSE;
190 	}
191 
192 	mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
193 
194 	if (!mc_lst)
195 		return -ENOMEM;
196 
197 	mc_lst->next = NULL;
198 	mc_lst->addr = *addr;
199 
200 	if (ifindex == 0) {
201 		struct rt6_info *rt;
202 		rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
203 		if (rt) {
204 			dev = rt->dst.dev;
205 			ip6_rt_put(rt);
206 		}
207 	} else
208 		dev = __dev_get_by_index(net, ifindex);
209 
210 	if (!dev) {
211 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
212 		return -ENODEV;
213 	}
214 
215 	mc_lst->ifindex = dev->ifindex;
216 	mc_lst->sfmode = mode;
217 	RCU_INIT_POINTER(mc_lst->sflist, NULL);
218 
219 	/*
220 	 *	now add/increase the group membership on the device
221 	 */
222 
223 	err = __ipv6_dev_mc_inc(dev, addr, mode);
224 
225 	if (err) {
226 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
227 		return err;
228 	}
229 
230 	mc_lst->next = np->ipv6_mc_list;
231 	rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
232 
233 	return 0;
234 }
235 
ipv6_sock_mc_join(struct sock * sk,int ifindex,const struct in6_addr * addr)236 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
237 {
238 	return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
239 }
240 EXPORT_SYMBOL(ipv6_sock_mc_join);
241 
ipv6_sock_mc_join_ssm(struct sock * sk,int ifindex,const struct in6_addr * addr,unsigned int mode)242 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
243 			  const struct in6_addr *addr, unsigned int mode)
244 {
245 	return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
246 }
247 
248 /*
249  *	socket leave on multicast group
250  */
ipv6_sock_mc_drop(struct sock * sk,int ifindex,const struct in6_addr * addr)251 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
252 {
253 	struct ipv6_pinfo *np = inet6_sk(sk);
254 	struct ipv6_mc_socklist *mc_lst;
255 	struct ipv6_mc_socklist __rcu **lnk;
256 	struct net *net = sock_net(sk);
257 
258 	ASSERT_RTNL();
259 
260 	if (!ipv6_addr_is_multicast(addr))
261 		return -EINVAL;
262 
263 	for (lnk = &np->ipv6_mc_list;
264 	     (mc_lst = sock_dereference(*lnk, sk)) != NULL;
265 	      lnk = &mc_lst->next) {
266 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
267 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
268 			struct net_device *dev;
269 
270 			*lnk = mc_lst->next;
271 
272 			dev = __dev_get_by_index(net, mc_lst->ifindex);
273 			if (dev) {
274 				struct inet6_dev *idev = __in6_dev_get(dev);
275 
276 				ip6_mc_leave_src(sk, mc_lst, idev);
277 				if (idev)
278 					__ipv6_dev_mc_dec(idev, &mc_lst->addr);
279 			} else {
280 				ip6_mc_leave_src(sk, mc_lst, NULL);
281 			}
282 
283 			atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
284 			kfree_rcu(mc_lst, rcu);
285 			return 0;
286 		}
287 	}
288 
289 	return -EADDRNOTAVAIL;
290 }
291 EXPORT_SYMBOL(ipv6_sock_mc_drop);
292 
ip6_mc_find_dev_rtnl(struct net * net,const struct in6_addr * group,int ifindex)293 static struct inet6_dev *ip6_mc_find_dev_rtnl(struct net *net,
294 					      const struct in6_addr *group,
295 					      int ifindex)
296 {
297 	struct net_device *dev = NULL;
298 	struct inet6_dev *idev = NULL;
299 
300 	if (ifindex == 0) {
301 		struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
302 
303 		if (rt) {
304 			dev = rt->dst.dev;
305 			ip6_rt_put(rt);
306 		}
307 	} else {
308 		dev = __dev_get_by_index(net, ifindex);
309 	}
310 
311 	if (!dev)
312 		return NULL;
313 	idev = __in6_dev_get(dev);
314 	if (!idev)
315 		return NULL;
316 	if (idev->dead)
317 		return NULL;
318 	return idev;
319 }
320 
__ipv6_sock_mc_close(struct sock * sk)321 void __ipv6_sock_mc_close(struct sock *sk)
322 {
323 	struct ipv6_pinfo *np = inet6_sk(sk);
324 	struct ipv6_mc_socklist *mc_lst;
325 	struct net *net = sock_net(sk);
326 
327 	ASSERT_RTNL();
328 
329 	while ((mc_lst = sock_dereference(np->ipv6_mc_list, sk)) != NULL) {
330 		struct net_device *dev;
331 
332 		np->ipv6_mc_list = mc_lst->next;
333 
334 		dev = __dev_get_by_index(net, mc_lst->ifindex);
335 		if (dev) {
336 			struct inet6_dev *idev = __in6_dev_get(dev);
337 
338 			ip6_mc_leave_src(sk, mc_lst, idev);
339 			if (idev)
340 				__ipv6_dev_mc_dec(idev, &mc_lst->addr);
341 		} else {
342 			ip6_mc_leave_src(sk, mc_lst, NULL);
343 		}
344 
345 		atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
346 		kfree_rcu(mc_lst, rcu);
347 	}
348 }
349 
ipv6_sock_mc_close(struct sock * sk)350 void ipv6_sock_mc_close(struct sock *sk)
351 {
352 	struct ipv6_pinfo *np = inet6_sk(sk);
353 
354 	if (!rcu_access_pointer(np->ipv6_mc_list))
355 		return;
356 
357 	rtnl_lock();
358 	lock_sock(sk);
359 	__ipv6_sock_mc_close(sk);
360 	release_sock(sk);
361 	rtnl_unlock();
362 }
363 
ip6_mc_source(int add,int omode,struct sock * sk,struct group_source_req * pgsr)364 int ip6_mc_source(int add, int omode, struct sock *sk,
365 	struct group_source_req *pgsr)
366 {
367 	struct in6_addr *source, *group;
368 	struct ipv6_mc_socklist *pmc;
369 	struct inet6_dev *idev;
370 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
371 	struct ip6_sf_socklist *psl;
372 	struct net *net = sock_net(sk);
373 	int i, j, rv;
374 	int leavegroup = 0;
375 	int err;
376 
377 	source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
378 	group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
379 
380 	if (!ipv6_addr_is_multicast(group))
381 		return -EINVAL;
382 
383 	idev = ip6_mc_find_dev_rtnl(net, group, pgsr->gsr_interface);
384 	if (!idev)
385 		return -ENODEV;
386 
387 	err = -EADDRNOTAVAIL;
388 
389 	mutex_lock(&idev->mc_lock);
390 	for_each_pmc_socklock(inet6, sk, pmc) {
391 		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
392 			continue;
393 		if (ipv6_addr_equal(&pmc->addr, group))
394 			break;
395 	}
396 	if (!pmc) {		/* must have a prior join */
397 		err = -EINVAL;
398 		goto done;
399 	}
400 	/* if a source filter was set, must be the same mode as before */
401 	if (rcu_access_pointer(pmc->sflist)) {
402 		if (pmc->sfmode != omode) {
403 			err = -EINVAL;
404 			goto done;
405 		}
406 	} else if (pmc->sfmode != omode) {
407 		/* allow mode switches for empty-set filters */
408 		ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
409 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
410 		pmc->sfmode = omode;
411 	}
412 
413 	psl = sock_dereference(pmc->sflist, sk);
414 	if (!add) {
415 		if (!psl)
416 			goto done;	/* err = -EADDRNOTAVAIL */
417 		rv = !0;
418 		for (i = 0; i < psl->sl_count; i++) {
419 			rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
420 			if (rv == 0)
421 				break;
422 		}
423 		if (rv)		/* source not found */
424 			goto done;	/* err = -EADDRNOTAVAIL */
425 
426 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
427 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
428 			leavegroup = 1;
429 			goto done;
430 		}
431 
432 		/* update the interface filter */
433 		ip6_mc_del_src(idev, group, omode, 1, source, 1);
434 
435 		for (j = i+1; j < psl->sl_count; j++)
436 			psl->sl_addr[j-1] = psl->sl_addr[j];
437 		psl->sl_count--;
438 		err = 0;
439 		goto done;
440 	}
441 	/* else, add a new source to the filter */
442 
443 	if (psl && psl->sl_count >= sysctl_mld_max_msf) {
444 		err = -ENOBUFS;
445 		goto done;
446 	}
447 	if (!psl || psl->sl_count == psl->sl_max) {
448 		struct ip6_sf_socklist *newpsl;
449 		int count = IP6_SFBLOCK;
450 
451 		if (psl)
452 			count += psl->sl_max;
453 		newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, count),
454 				      GFP_KERNEL);
455 		if (!newpsl) {
456 			err = -ENOBUFS;
457 			goto done;
458 		}
459 		newpsl->sl_max = count;
460 		newpsl->sl_count = count - IP6_SFBLOCK;
461 		if (psl) {
462 			for (i = 0; i < psl->sl_count; i++)
463 				newpsl->sl_addr[i] = psl->sl_addr[i];
464 			atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
465 				   &sk->sk_omem_alloc);
466 		}
467 		rcu_assign_pointer(pmc->sflist, newpsl);
468 		kfree_rcu(psl, rcu);
469 		psl = newpsl;
470 	}
471 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
472 	for (i = 0; i < psl->sl_count; i++) {
473 		rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
474 		if (rv == 0) /* There is an error in the address. */
475 			goto done;
476 	}
477 	for (j = psl->sl_count-1; j >= i; j--)
478 		psl->sl_addr[j+1] = psl->sl_addr[j];
479 	psl->sl_addr[i] = *source;
480 	psl->sl_count++;
481 	err = 0;
482 	/* update the interface list */
483 	ip6_mc_add_src(idev, group, omode, 1, source, 1);
484 done:
485 	mutex_unlock(&idev->mc_lock);
486 	if (leavegroup)
487 		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
488 	return err;
489 }
490 
ip6_mc_msfilter(struct sock * sk,struct group_filter * gsf,struct sockaddr_storage * list)491 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
492 		    struct sockaddr_storage *list)
493 {
494 	const struct in6_addr *group;
495 	struct ipv6_mc_socklist *pmc;
496 	struct inet6_dev *idev;
497 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
498 	struct ip6_sf_socklist *newpsl, *psl;
499 	struct net *net = sock_net(sk);
500 	int leavegroup = 0;
501 	int i, err;
502 
503 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
504 
505 	if (!ipv6_addr_is_multicast(group))
506 		return -EINVAL;
507 	if (gsf->gf_fmode != MCAST_INCLUDE &&
508 	    gsf->gf_fmode != MCAST_EXCLUDE)
509 		return -EINVAL;
510 
511 	idev = ip6_mc_find_dev_rtnl(net, group, gsf->gf_interface);
512 	if (!idev)
513 		return -ENODEV;
514 
515 	err = 0;
516 
517 	if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
518 		leavegroup = 1;
519 		goto done;
520 	}
521 
522 	for_each_pmc_socklock(inet6, sk, pmc) {
523 		if (pmc->ifindex != gsf->gf_interface)
524 			continue;
525 		if (ipv6_addr_equal(&pmc->addr, group))
526 			break;
527 	}
528 	if (!pmc) {		/* must have a prior join */
529 		err = -EINVAL;
530 		goto done;
531 	}
532 	if (gsf->gf_numsrc) {
533 		newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr,
534 						      gsf->gf_numsrc),
535 				      GFP_KERNEL);
536 		if (!newpsl) {
537 			err = -ENOBUFS;
538 			goto done;
539 		}
540 		newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
541 		for (i = 0; i < newpsl->sl_count; ++i, ++list) {
542 			struct sockaddr_in6 *psin6;
543 
544 			psin6 = (struct sockaddr_in6 *)list;
545 			newpsl->sl_addr[i] = psin6->sin6_addr;
546 		}
547 		mutex_lock(&idev->mc_lock);
548 		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
549 				     newpsl->sl_count, newpsl->sl_addr, 0);
550 		if (err) {
551 			mutex_unlock(&idev->mc_lock);
552 			sock_kfree_s(sk, newpsl, struct_size(newpsl, sl_addr,
553 							     newpsl->sl_max));
554 			goto done;
555 		}
556 		mutex_unlock(&idev->mc_lock);
557 	} else {
558 		newpsl = NULL;
559 		mutex_lock(&idev->mc_lock);
560 		ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
561 		mutex_unlock(&idev->mc_lock);
562 	}
563 
564 	mutex_lock(&idev->mc_lock);
565 	psl = sock_dereference(pmc->sflist, sk);
566 	if (psl) {
567 		ip6_mc_del_src(idev, group, pmc->sfmode,
568 			       psl->sl_count, psl->sl_addr, 0);
569 		atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
570 			   &sk->sk_omem_alloc);
571 	} else {
572 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
573 	}
574 	rcu_assign_pointer(pmc->sflist, newpsl);
575 	mutex_unlock(&idev->mc_lock);
576 	kfree_rcu(psl, rcu);
577 	pmc->sfmode = gsf->gf_fmode;
578 	err = 0;
579 done:
580 	if (leavegroup)
581 		err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
582 	return err;
583 }
584 
ip6_mc_msfget(struct sock * sk,struct group_filter * gsf,sockptr_t optval,size_t ss_offset)585 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
586 		  sockptr_t optval, size_t ss_offset)
587 {
588 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
589 	const struct in6_addr *group;
590 	struct ipv6_mc_socklist *pmc;
591 	struct ip6_sf_socklist *psl;
592 	unsigned int count;
593 	int i, copycount;
594 
595 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
596 
597 	if (!ipv6_addr_is_multicast(group))
598 		return -EINVAL;
599 
600 	/* changes to the ipv6_mc_list require the socket lock and
601 	 * rtnl lock. We have the socket lock, so reading the list is safe.
602 	 */
603 
604 	for_each_pmc_socklock(inet6, sk, pmc) {
605 		if (pmc->ifindex != gsf->gf_interface)
606 			continue;
607 		if (ipv6_addr_equal(group, &pmc->addr))
608 			break;
609 	}
610 	if (!pmc)		/* must have a prior join */
611 		return -EADDRNOTAVAIL;
612 
613 	gsf->gf_fmode = pmc->sfmode;
614 	psl = sock_dereference(pmc->sflist, sk);
615 	count = psl ? psl->sl_count : 0;
616 
617 	copycount = min(count, gsf->gf_numsrc);
618 	gsf->gf_numsrc = count;
619 	for (i = 0; i < copycount; i++) {
620 		struct sockaddr_in6 *psin6;
621 		struct sockaddr_storage ss;
622 
623 		psin6 = (struct sockaddr_in6 *)&ss;
624 		memset(&ss, 0, sizeof(ss));
625 		psin6->sin6_family = AF_INET6;
626 		psin6->sin6_addr = psl->sl_addr[i];
627 		if (copy_to_sockptr_offset(optval, ss_offset, &ss, sizeof(ss)))
628 			return -EFAULT;
629 		ss_offset += sizeof(ss);
630 	}
631 	return 0;
632 }
633 
inet6_mc_check(const struct sock * sk,const struct in6_addr * mc_addr,const struct in6_addr * src_addr)634 bool inet6_mc_check(const struct sock *sk, const struct in6_addr *mc_addr,
635 		    const struct in6_addr *src_addr)
636 {
637 	const struct ipv6_pinfo *np = inet6_sk(sk);
638 	const struct ipv6_mc_socklist *mc;
639 	const struct ip6_sf_socklist *psl;
640 	bool rv = true;
641 
642 	rcu_read_lock();
643 	for_each_pmc_rcu(np, mc) {
644 		if (ipv6_addr_equal(&mc->addr, mc_addr))
645 			break;
646 	}
647 	if (!mc) {
648 		rcu_read_unlock();
649 		return inet6_test_bit(MC6_ALL, sk);
650 	}
651 	psl = rcu_dereference(mc->sflist);
652 	if (!psl) {
653 		rv = mc->sfmode == MCAST_EXCLUDE;
654 	} else {
655 		int i;
656 
657 		for (i = 0; i < psl->sl_count; i++) {
658 			if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
659 				break;
660 		}
661 		if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
662 			rv = false;
663 		if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
664 			rv = false;
665 	}
666 	rcu_read_unlock();
667 
668 	return rv;
669 }
670 
671 /* called with mc_lock */
igmp6_group_added(struct ifmcaddr6 * mc)672 static void igmp6_group_added(struct ifmcaddr6 *mc)
673 {
674 	struct net_device *dev = mc->idev->dev;
675 	char buf[MAX_ADDR_LEN];
676 
677 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
678 	    IPV6_ADDR_SCOPE_LINKLOCAL)
679 		return;
680 
681 	if (!(mc->mca_flags&MAF_LOADED)) {
682 		mc->mca_flags |= MAF_LOADED;
683 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
684 			dev_mc_add(dev, buf);
685 	}
686 
687 	if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
688 		return;
689 
690 	if (mld_in_v1_mode(mc->idev)) {
691 		igmp6_join_group(mc);
692 		return;
693 	}
694 	/* else v2 */
695 
696 	/* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
697 	 * should not send filter-mode change record as the mode
698 	 * should be from IN() to IN(A).
699 	 */
700 	if (mc->mca_sfmode == MCAST_EXCLUDE)
701 		mc->mca_crcount = mc->idev->mc_qrv;
702 
703 	mld_ifc_event(mc->idev);
704 }
705 
706 /* called with mc_lock */
igmp6_group_dropped(struct ifmcaddr6 * mc)707 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
708 {
709 	struct net_device *dev = mc->idev->dev;
710 	char buf[MAX_ADDR_LEN];
711 
712 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
713 	    IPV6_ADDR_SCOPE_LINKLOCAL)
714 		return;
715 
716 	if (mc->mca_flags&MAF_LOADED) {
717 		mc->mca_flags &= ~MAF_LOADED;
718 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
719 			dev_mc_del(dev, buf);
720 	}
721 
722 	if (mc->mca_flags & MAF_NOREPORT)
723 		return;
724 
725 	if (!mc->idev->dead)
726 		igmp6_leave_group(mc);
727 
728 	if (cancel_delayed_work(&mc->mca_work))
729 		refcount_dec(&mc->mca_refcnt);
730 }
731 
732 /*
733  * deleted ifmcaddr6 manipulation
734  * called with mc_lock
735  */
mld_add_delrec(struct inet6_dev * idev,struct ifmcaddr6 * im)736 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
737 {
738 	struct ifmcaddr6 *pmc;
739 
740 	/* this is an "ifmcaddr6" for convenience; only the fields below
741 	 * are actually used. In particular, the refcnt and users are not
742 	 * used for management of the delete list. Using the same structure
743 	 * for deleted items allows change reports to use common code with
744 	 * non-deleted or query-response MCA's.
745 	 */
746 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
747 	if (!pmc)
748 		return;
749 
750 	pmc->idev = im->idev;
751 	in6_dev_hold(idev);
752 	pmc->mca_addr = im->mca_addr;
753 	pmc->mca_crcount = idev->mc_qrv;
754 	pmc->mca_sfmode = im->mca_sfmode;
755 	if (pmc->mca_sfmode == MCAST_INCLUDE) {
756 		struct ip6_sf_list *psf;
757 
758 		rcu_assign_pointer(pmc->mca_tomb,
759 				   mc_dereference(im->mca_tomb, idev));
760 		rcu_assign_pointer(pmc->mca_sources,
761 				   mc_dereference(im->mca_sources, idev));
762 		RCU_INIT_POINTER(im->mca_tomb, NULL);
763 		RCU_INIT_POINTER(im->mca_sources, NULL);
764 
765 		for_each_psf_mclock(pmc, psf)
766 			psf->sf_crcount = pmc->mca_crcount;
767 	}
768 
769 	rcu_assign_pointer(pmc->next, idev->mc_tomb);
770 	rcu_assign_pointer(idev->mc_tomb, pmc);
771 }
772 
773 /* called with mc_lock */
mld_del_delrec(struct inet6_dev * idev,struct ifmcaddr6 * im)774 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
775 {
776 	struct ip6_sf_list *psf, *sources, *tomb;
777 	struct in6_addr *pmca = &im->mca_addr;
778 	struct ifmcaddr6 *pmc, *pmc_prev;
779 
780 	pmc_prev = NULL;
781 	for_each_mc_tomb(idev, pmc) {
782 		if (ipv6_addr_equal(&pmc->mca_addr, pmca))
783 			break;
784 		pmc_prev = pmc;
785 	}
786 	if (pmc) {
787 		if (pmc_prev)
788 			rcu_assign_pointer(pmc_prev->next, pmc->next);
789 		else
790 			rcu_assign_pointer(idev->mc_tomb, pmc->next);
791 	}
792 
793 	if (pmc) {
794 		im->idev = pmc->idev;
795 		if (im->mca_sfmode == MCAST_INCLUDE) {
796 			tomb = rcu_replace_pointer(im->mca_tomb,
797 						   mc_dereference(pmc->mca_tomb, pmc->idev),
798 						   lockdep_is_held(&im->idev->mc_lock));
799 			rcu_assign_pointer(pmc->mca_tomb, tomb);
800 
801 			sources = rcu_replace_pointer(im->mca_sources,
802 						      mc_dereference(pmc->mca_sources, pmc->idev),
803 						      lockdep_is_held(&im->idev->mc_lock));
804 			rcu_assign_pointer(pmc->mca_sources, sources);
805 			for_each_psf_mclock(im, psf)
806 				psf->sf_crcount = idev->mc_qrv;
807 		} else {
808 			im->mca_crcount = idev->mc_qrv;
809 		}
810 		in6_dev_put(pmc->idev);
811 		ip6_mc_clear_src(pmc);
812 		kfree_rcu(pmc, rcu);
813 	}
814 }
815 
816 /* called with mc_lock */
mld_clear_delrec(struct inet6_dev * idev)817 static void mld_clear_delrec(struct inet6_dev *idev)
818 {
819 	struct ifmcaddr6 *pmc, *nextpmc;
820 
821 	pmc = mc_dereference(idev->mc_tomb, idev);
822 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
823 
824 	for (; pmc; pmc = nextpmc) {
825 		nextpmc = mc_dereference(pmc->next, idev);
826 		ip6_mc_clear_src(pmc);
827 		in6_dev_put(pmc->idev);
828 		kfree_rcu(pmc, rcu);
829 	}
830 
831 	/* clear dead sources, too */
832 	for_each_mc_mclock(idev, pmc) {
833 		struct ip6_sf_list *psf, *psf_next;
834 
835 		psf = mc_dereference(pmc->mca_tomb, idev);
836 		RCU_INIT_POINTER(pmc->mca_tomb, NULL);
837 		for (; psf; psf = psf_next) {
838 			psf_next = mc_dereference(psf->sf_next, idev);
839 			kfree_rcu(psf, rcu);
840 		}
841 	}
842 }
843 
mld_clear_query(struct inet6_dev * idev)844 static void mld_clear_query(struct inet6_dev *idev)
845 {
846 	struct sk_buff *skb;
847 
848 	spin_lock_bh(&idev->mc_query_lock);
849 	while ((skb = __skb_dequeue(&idev->mc_query_queue)))
850 		kfree_skb(skb);
851 	spin_unlock_bh(&idev->mc_query_lock);
852 }
853 
mld_clear_report(struct inet6_dev * idev)854 static void mld_clear_report(struct inet6_dev *idev)
855 {
856 	struct sk_buff *skb;
857 
858 	spin_lock_bh(&idev->mc_report_lock);
859 	while ((skb = __skb_dequeue(&idev->mc_report_queue)))
860 		kfree_skb(skb);
861 	spin_unlock_bh(&idev->mc_report_lock);
862 }
863 
mca_get(struct ifmcaddr6 * mc)864 static void mca_get(struct ifmcaddr6 *mc)
865 {
866 	refcount_inc(&mc->mca_refcnt);
867 }
868 
ma_put(struct ifmcaddr6 * mc)869 static void ma_put(struct ifmcaddr6 *mc)
870 {
871 	if (refcount_dec_and_test(&mc->mca_refcnt)) {
872 		in6_dev_put(mc->idev);
873 		kfree_rcu(mc, rcu);
874 	}
875 }
876 
877 /* called with mc_lock */
mca_alloc(struct inet6_dev * idev,const struct in6_addr * addr,unsigned int mode)878 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
879 				   const struct in6_addr *addr,
880 				   unsigned int mode)
881 {
882 	struct ifmcaddr6 *mc;
883 
884 	mc = kzalloc(sizeof(*mc), GFP_KERNEL);
885 	if (!mc)
886 		return NULL;
887 
888 	INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);
889 
890 	mc->mca_addr = *addr;
891 	mc->idev = idev; /* reference taken by caller */
892 	mc->mca_users = 1;
893 	/* mca_stamp should be updated upon changes */
894 	mc->mca_cstamp = mc->mca_tstamp = jiffies;
895 	refcount_set(&mc->mca_refcnt, 1);
896 
897 	mc->mca_sfmode = mode;
898 	mc->mca_sfcount[mode] = 1;
899 
900 	if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
901 	    IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
902 		mc->mca_flags |= MAF_NOREPORT;
903 
904 	return mc;
905 }
906 
inet6_ifmcaddr_notify(struct net_device * dev,const struct ifmcaddr6 * ifmca,int event)907 static void inet6_ifmcaddr_notify(struct net_device *dev,
908 				  const struct ifmcaddr6 *ifmca, int event)
909 {
910 	struct inet6_fill_args fillargs = {
911 		.portid = 0,
912 		.seq = 0,
913 		.event = event,
914 		.flags = 0,
915 		.netnsid = -1,
916 		.force_rt_scope_universe = true,
917 	};
918 	struct net *net = dev_net(dev);
919 	struct sk_buff *skb;
920 	int err = -ENOMEM;
921 
922 	skb = nlmsg_new(NLMSG_ALIGN(sizeof(struct ifaddrmsg)) +
923 			nla_total_size(sizeof(struct in6_addr)) +
924 			nla_total_size(sizeof(struct ifa_cacheinfo)),
925 			GFP_KERNEL);
926 	if (!skb)
927 		goto error;
928 
929 	err = inet6_fill_ifmcaddr(skb, ifmca, &fillargs);
930 	if (err < 0) {
931 		WARN_ON_ONCE(err == -EMSGSIZE);
932 		nlmsg_free(skb);
933 		goto error;
934 	}
935 
936 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MCADDR, NULL, GFP_KERNEL);
937 	return;
938 error:
939 	rtnl_set_sk_err(net, RTNLGRP_IPV6_MCADDR, err);
940 }
941 
942 /*
943  *	device multicast group inc (add if not found)
944  */
__ipv6_dev_mc_inc(struct net_device * dev,const struct in6_addr * addr,unsigned int mode)945 static int __ipv6_dev_mc_inc(struct net_device *dev,
946 			     const struct in6_addr *addr, unsigned int mode)
947 {
948 	struct ifmcaddr6 *mc;
949 	struct inet6_dev *idev;
950 
951 	ASSERT_RTNL();
952 
953 	/* we need to take a reference on idev */
954 	idev = in6_dev_get(dev);
955 
956 	if (!idev)
957 		return -EINVAL;
958 
959 	if (idev->dead) {
960 		in6_dev_put(idev);
961 		return -ENODEV;
962 	}
963 
964 	mutex_lock(&idev->mc_lock);
965 	for_each_mc_mclock(idev, mc) {
966 		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
967 			mc->mca_users++;
968 			ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
969 			mutex_unlock(&idev->mc_lock);
970 			in6_dev_put(idev);
971 			return 0;
972 		}
973 	}
974 
975 	mc = mca_alloc(idev, addr, mode);
976 	if (!mc) {
977 		mutex_unlock(&idev->mc_lock);
978 		in6_dev_put(idev);
979 		return -ENOMEM;
980 	}
981 
982 	rcu_assign_pointer(mc->next, idev->mc_list);
983 	rcu_assign_pointer(idev->mc_list, mc);
984 
985 	mca_get(mc);
986 
987 	mld_del_delrec(idev, mc);
988 	igmp6_group_added(mc);
989 	inet6_ifmcaddr_notify(dev, mc, RTM_NEWMULTICAST);
990 	mutex_unlock(&idev->mc_lock);
991 	ma_put(mc);
992 	return 0;
993 }
994 
ipv6_dev_mc_inc(struct net_device * dev,const struct in6_addr * addr)995 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
996 {
997 	return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
998 }
999 EXPORT_SYMBOL(ipv6_dev_mc_inc);
1000 
1001 /*
1002  * device multicast group del
1003  */
__ipv6_dev_mc_dec(struct inet6_dev * idev,const struct in6_addr * addr)1004 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
1005 {
1006 	struct ifmcaddr6 *ma, __rcu **map;
1007 
1008 	ASSERT_RTNL();
1009 
1010 	mutex_lock(&idev->mc_lock);
1011 	for (map = &idev->mc_list;
1012 	     (ma = mc_dereference(*map, idev));
1013 	     map = &ma->next) {
1014 		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
1015 			if (--ma->mca_users == 0) {
1016 				*map = ma->next;
1017 
1018 				igmp6_group_dropped(ma);
1019 				inet6_ifmcaddr_notify(idev->dev, ma,
1020 						      RTM_DELMULTICAST);
1021 				ip6_mc_clear_src(ma);
1022 				mutex_unlock(&idev->mc_lock);
1023 
1024 				ma_put(ma);
1025 				return 0;
1026 			}
1027 			mutex_unlock(&idev->mc_lock);
1028 			return 0;
1029 		}
1030 	}
1031 
1032 	mutex_unlock(&idev->mc_lock);
1033 	return -ENOENT;
1034 }
1035 
ipv6_dev_mc_dec(struct net_device * dev,const struct in6_addr * addr)1036 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
1037 {
1038 	struct inet6_dev *idev;
1039 	int err;
1040 
1041 	ASSERT_RTNL();
1042 
1043 	idev = __in6_dev_get(dev);
1044 	if (!idev)
1045 		err = -ENODEV;
1046 	else
1047 		err = __ipv6_dev_mc_dec(idev, addr);
1048 
1049 	return err;
1050 }
1051 EXPORT_SYMBOL(ipv6_dev_mc_dec);
1052 
1053 /*
1054  *	check if the interface/address pair is valid
1055  */
ipv6_chk_mcast_addr(struct net_device * dev,const struct in6_addr * group,const struct in6_addr * src_addr)1056 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
1057 			 const struct in6_addr *src_addr)
1058 {
1059 	struct inet6_dev *idev;
1060 	struct ifmcaddr6 *mc;
1061 	bool rv = false;
1062 
1063 	rcu_read_lock();
1064 	idev = __in6_dev_get(dev);
1065 	if (!idev)
1066 		goto unlock;
1067 	for_each_mc_rcu(idev, mc) {
1068 		if (ipv6_addr_equal(&mc->mca_addr, group))
1069 			break;
1070 	}
1071 	if (!mc)
1072 		goto unlock;
1073 	if (src_addr && !ipv6_addr_any(src_addr)) {
1074 		struct ip6_sf_list *psf;
1075 
1076 		for_each_psf_rcu(mc, psf) {
1077 			if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1078 				break;
1079 		}
1080 		if (psf)
1081 			rv = READ_ONCE(psf->sf_count[MCAST_INCLUDE]) ||
1082 				READ_ONCE(psf->sf_count[MCAST_EXCLUDE]) !=
1083 				READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]);
1084 		else
1085 			rv = READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]) != 0;
1086 	} else {
1087 		rv = true; /* don't filter unspecified source */
1088 	}
1089 unlock:
1090 	rcu_read_unlock();
1091 	return rv;
1092 }
1093 
1094 /* called with mc_lock */
mld_gq_start_work(struct inet6_dev * idev)1095 static void mld_gq_start_work(struct inet6_dev *idev)
1096 {
1097 	unsigned long tv = get_random_u32_below(idev->mc_maxdelay);
1098 
1099 	idev->mc_gq_running = 1;
1100 	if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2))
1101 		in6_dev_hold(idev);
1102 }
1103 
1104 /* called with mc_lock */
mld_gq_stop_work(struct inet6_dev * idev)1105 static void mld_gq_stop_work(struct inet6_dev *idev)
1106 {
1107 	idev->mc_gq_running = 0;
1108 	if (cancel_delayed_work(&idev->mc_gq_work))
1109 		__in6_dev_put(idev);
1110 }
1111 
1112 /* called with mc_lock */
mld_ifc_start_work(struct inet6_dev * idev,unsigned long delay)1113 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
1114 {
1115 	unsigned long tv = get_random_u32_below(delay);
1116 
1117 	if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2))
1118 		in6_dev_hold(idev);
1119 }
1120 
1121 /* called with mc_lock */
mld_ifc_stop_work(struct inet6_dev * idev)1122 static void mld_ifc_stop_work(struct inet6_dev *idev)
1123 {
1124 	idev->mc_ifc_count = 0;
1125 	if (cancel_delayed_work(&idev->mc_ifc_work))
1126 		__in6_dev_put(idev);
1127 }
1128 
1129 /* called with mc_lock */
mld_dad_start_work(struct inet6_dev * idev,unsigned long delay)1130 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
1131 {
1132 	unsigned long tv = get_random_u32_below(delay);
1133 
1134 	if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2))
1135 		in6_dev_hold(idev);
1136 }
1137 
mld_dad_stop_work(struct inet6_dev * idev)1138 static void mld_dad_stop_work(struct inet6_dev *idev)
1139 {
1140 	if (cancel_delayed_work(&idev->mc_dad_work))
1141 		__in6_dev_put(idev);
1142 }
1143 
mld_query_stop_work(struct inet6_dev * idev)1144 static void mld_query_stop_work(struct inet6_dev *idev)
1145 {
1146 	spin_lock_bh(&idev->mc_query_lock);
1147 	if (cancel_delayed_work(&idev->mc_query_work))
1148 		__in6_dev_put(idev);
1149 	spin_unlock_bh(&idev->mc_query_lock);
1150 }
1151 
mld_report_stop_work(struct inet6_dev * idev)1152 static void mld_report_stop_work(struct inet6_dev *idev)
1153 {
1154 	if (cancel_delayed_work_sync(&idev->mc_report_work))
1155 		__in6_dev_put(idev);
1156 }
1157 
1158 /*
1159  * IGMP handling (alias multicast ICMPv6 messages)
1160  * called with mc_lock
1161  */
igmp6_group_queried(struct ifmcaddr6 * ma,unsigned long resptime)1162 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1163 {
1164 	unsigned long delay = resptime;
1165 
1166 	/* Do not start work for these addresses */
1167 	if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1168 	    IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1169 		return;
1170 
1171 	if (cancel_delayed_work(&ma->mca_work)) {
1172 		refcount_dec(&ma->mca_refcnt);
1173 		delay = ma->mca_work.timer.expires - jiffies;
1174 	}
1175 
1176 	if (delay >= resptime)
1177 		delay = get_random_u32_below(resptime);
1178 
1179 	if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
1180 		refcount_inc(&ma->mca_refcnt);
1181 	ma->mca_flags |= MAF_TIMER_RUNNING;
1182 }
1183 
1184 /* mark EXCLUDE-mode sources
1185  * called with mc_lock
1186  */
mld_xmarksources(struct ifmcaddr6 * pmc,int nsrcs,const struct in6_addr * srcs)1187 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1188 			     const struct in6_addr *srcs)
1189 {
1190 	struct ip6_sf_list *psf;
1191 	int i, scount;
1192 
1193 	scount = 0;
1194 	for_each_psf_mclock(pmc, psf) {
1195 		if (scount == nsrcs)
1196 			break;
1197 		for (i = 0; i < nsrcs; i++) {
1198 			/* skip inactive filters */
1199 			if (psf->sf_count[MCAST_INCLUDE] ||
1200 			    pmc->mca_sfcount[MCAST_EXCLUDE] !=
1201 			    psf->sf_count[MCAST_EXCLUDE])
1202 				break;
1203 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1204 				scount++;
1205 				break;
1206 			}
1207 		}
1208 	}
1209 	pmc->mca_flags &= ~MAF_GSQUERY;
1210 	if (scount == nsrcs)	/* all sources excluded */
1211 		return false;
1212 	return true;
1213 }
1214 
1215 /* called with mc_lock */
mld_marksources(struct ifmcaddr6 * pmc,int nsrcs,const struct in6_addr * srcs)1216 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1217 			    const struct in6_addr *srcs)
1218 {
1219 	struct ip6_sf_list *psf;
1220 	int i, scount;
1221 
1222 	if (pmc->mca_sfmode == MCAST_EXCLUDE)
1223 		return mld_xmarksources(pmc, nsrcs, srcs);
1224 
1225 	/* mark INCLUDE-mode sources */
1226 
1227 	scount = 0;
1228 	for_each_psf_mclock(pmc, psf) {
1229 		if (scount == nsrcs)
1230 			break;
1231 		for (i = 0; i < nsrcs; i++) {
1232 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1233 				psf->sf_gsresp = 1;
1234 				scount++;
1235 				break;
1236 			}
1237 		}
1238 	}
1239 	if (!scount) {
1240 		pmc->mca_flags &= ~MAF_GSQUERY;
1241 		return false;
1242 	}
1243 	pmc->mca_flags |= MAF_GSQUERY;
1244 	return true;
1245 }
1246 
mld_force_mld_version(const struct inet6_dev * idev)1247 static int mld_force_mld_version(const struct inet6_dev *idev)
1248 {
1249 	const struct net *net = dev_net(idev->dev);
1250 	int all_force;
1251 
1252 	all_force = READ_ONCE(net->ipv6.devconf_all->force_mld_version);
1253 	/* Normally, both are 0 here. If enforcement to a particular is
1254 	 * being used, individual device enforcement will have a lower
1255 	 * precedence over 'all' device (.../conf/all/force_mld_version).
1256 	 */
1257 	return all_force ?: READ_ONCE(idev->cnf.force_mld_version);
1258 }
1259 
mld_in_v2_mode_only(const struct inet6_dev * idev)1260 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1261 {
1262 	return mld_force_mld_version(idev) == 2;
1263 }
1264 
mld_in_v1_mode_only(const struct inet6_dev * idev)1265 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1266 {
1267 	return mld_force_mld_version(idev) == 1;
1268 }
1269 
mld_in_v1_mode(const struct inet6_dev * idev)1270 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1271 {
1272 	if (mld_in_v2_mode_only(idev))
1273 		return false;
1274 	if (mld_in_v1_mode_only(idev))
1275 		return true;
1276 	if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1277 		return true;
1278 
1279 	return false;
1280 }
1281 
mld_set_v1_mode(struct inet6_dev * idev)1282 static void mld_set_v1_mode(struct inet6_dev *idev)
1283 {
1284 	/* RFC3810, relevant sections:
1285 	 *  - 9.1. Robustness Variable
1286 	 *  - 9.2. Query Interval
1287 	 *  - 9.3. Query Response Interval
1288 	 *  - 9.12. Older Version Querier Present Timeout
1289 	 */
1290 	unsigned long switchback;
1291 
1292 	switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1293 
1294 	idev->mc_v1_seen = jiffies + switchback;
1295 }
1296 
mld_update_qrv(struct inet6_dev * idev,const struct mld2_query * mlh2)1297 static void mld_update_qrv(struct inet6_dev *idev,
1298 			   const struct mld2_query *mlh2)
1299 {
1300 	/* RFC3810, relevant sections:
1301 	 *  - 5.1.8. QRV (Querier's Robustness Variable)
1302 	 *  - 9.1. Robustness Variable
1303 	 */
1304 
1305 	/* The value of the Robustness Variable MUST NOT be zero,
1306 	 * and SHOULD NOT be one. Catch this here if we ever run
1307 	 * into such a case in future.
1308 	 */
1309 	const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1310 	WARN_ON(idev->mc_qrv == 0);
1311 
1312 	if (mlh2->mld2q_qrv > 0)
1313 		idev->mc_qrv = mlh2->mld2q_qrv;
1314 
1315 	if (unlikely(idev->mc_qrv < min_qrv)) {
1316 		net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1317 				     idev->mc_qrv, min_qrv);
1318 		idev->mc_qrv = min_qrv;
1319 	}
1320 }
1321 
mld_update_qi(struct inet6_dev * idev,const struct mld2_query * mlh2)1322 static void mld_update_qi(struct inet6_dev *idev,
1323 			  const struct mld2_query *mlh2)
1324 {
1325 	/* RFC3810, relevant sections:
1326 	 *  - 5.1.9. QQIC (Querier's Query Interval Code)
1327 	 *  - 9.2. Query Interval
1328 	 *  - 9.12. Older Version Querier Present Timeout
1329 	 *    (the [Query Interval] in the last Query received)
1330 	 */
1331 	unsigned long mc_qqi;
1332 
1333 	if (mlh2->mld2q_qqic < 128) {
1334 		mc_qqi = mlh2->mld2q_qqic;
1335 	} else {
1336 		unsigned long mc_man, mc_exp;
1337 
1338 		mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1339 		mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1340 
1341 		mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1342 	}
1343 
1344 	idev->mc_qi = mc_qqi * HZ;
1345 }
1346 
mld_update_qri(struct inet6_dev * idev,const struct mld2_query * mlh2)1347 static void mld_update_qri(struct inet6_dev *idev,
1348 			   const struct mld2_query *mlh2)
1349 {
1350 	/* RFC3810, relevant sections:
1351 	 *  - 5.1.3. Maximum Response Code
1352 	 *  - 9.3. Query Response Interval
1353 	 */
1354 	idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1355 }
1356 
mld_process_v1(struct inet6_dev * idev,struct mld_msg * mld,unsigned long * max_delay,bool v1_query)1357 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1358 			  unsigned long *max_delay, bool v1_query)
1359 {
1360 	unsigned long mldv1_md;
1361 
1362 	/* Ignore v1 queries */
1363 	if (mld_in_v2_mode_only(idev))
1364 		return -EINVAL;
1365 
1366 	mldv1_md = ntohs(mld->mld_maxdelay);
1367 
1368 	/* When in MLDv1 fallback and a MLDv2 router start-up being
1369 	 * unaware of current MLDv1 operation, the MRC == MRD mapping
1370 	 * only works when the exponential algorithm is not being
1371 	 * used (as MLDv1 is unaware of such things).
1372 	 *
1373 	 * According to the RFC author, the MLDv2 implementations
1374 	 * he's aware of all use a MRC < 32768 on start up queries.
1375 	 *
1376 	 * Thus, should we *ever* encounter something else larger
1377 	 * than that, just assume the maximum possible within our
1378 	 * reach.
1379 	 */
1380 	if (!v1_query)
1381 		mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1382 
1383 	*max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1384 
1385 	/* MLDv1 router present: we need to go into v1 mode *only*
1386 	 * when an MLDv1 query is received as per section 9.12. of
1387 	 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1388 	 * queries MUST be of exactly 24 octets.
1389 	 */
1390 	if (v1_query)
1391 		mld_set_v1_mode(idev);
1392 
1393 	/* cancel MLDv2 report work */
1394 	mld_gq_stop_work(idev);
1395 	/* cancel the interface change work */
1396 	mld_ifc_stop_work(idev);
1397 	/* clear deleted report items */
1398 	mld_clear_delrec(idev);
1399 
1400 	return 0;
1401 }
1402 
mld_process_v2(struct inet6_dev * idev,struct mld2_query * mld,unsigned long * max_delay)1403 static void mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1404 			   unsigned long *max_delay)
1405 {
1406 	*max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1407 
1408 	mld_update_qrv(idev, mld);
1409 	mld_update_qi(idev, mld);
1410 	mld_update_qri(idev, mld);
1411 
1412 	idev->mc_maxdelay = *max_delay;
1413 
1414 	return;
1415 }
1416 
1417 /* called with rcu_read_lock() */
igmp6_event_query(struct sk_buff * skb)1418 void igmp6_event_query(struct sk_buff *skb)
1419 {
1420 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1421 
1422 	if (!idev || idev->dead)
1423 		goto out;
1424 
1425 	spin_lock_bh(&idev->mc_query_lock);
1426 	if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) {
1427 		__skb_queue_tail(&idev->mc_query_queue, skb);
1428 		if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0))
1429 			in6_dev_hold(idev);
1430 		skb = NULL;
1431 	}
1432 	spin_unlock_bh(&idev->mc_query_lock);
1433 out:
1434 	kfree_skb(skb);
1435 }
1436 
__mld_query_work(struct sk_buff * skb)1437 static void __mld_query_work(struct sk_buff *skb)
1438 {
1439 	struct mld2_query *mlh2 = NULL;
1440 	const struct in6_addr *group;
1441 	unsigned long max_delay;
1442 	struct inet6_dev *idev;
1443 	struct ifmcaddr6 *ma;
1444 	struct mld_msg *mld;
1445 	int group_type;
1446 	int mark = 0;
1447 	int len, err;
1448 
1449 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1450 		goto kfree_skb;
1451 
1452 	/* compute payload length excluding extension headers */
1453 	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1454 	len -= skb_network_header_len(skb);
1455 
1456 	/* RFC3810 6.2
1457 	 * Upon reception of an MLD message that contains a Query, the node
1458 	 * checks if the source address of the message is a valid link-local
1459 	 * address, if the Hop Limit is set to 1, and if the Router Alert
1460 	 * option is present in the Hop-By-Hop Options header of the IPv6
1461 	 * packet.  If any of these checks fails, the packet is dropped.
1462 	 */
1463 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1464 	    ipv6_hdr(skb)->hop_limit != 1 ||
1465 	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1466 	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1467 		goto kfree_skb;
1468 
1469 	idev = in6_dev_get(skb->dev);
1470 	if (!idev)
1471 		goto kfree_skb;
1472 
1473 	mld = (struct mld_msg *)icmp6_hdr(skb);
1474 	group = &mld->mld_mca;
1475 	group_type = ipv6_addr_type(group);
1476 
1477 	if (group_type != IPV6_ADDR_ANY &&
1478 	    !(group_type&IPV6_ADDR_MULTICAST))
1479 		goto out;
1480 
1481 	if (len < MLD_V1_QUERY_LEN) {
1482 		goto out;
1483 	} else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1484 		err = mld_process_v1(idev, mld, &max_delay,
1485 				     len == MLD_V1_QUERY_LEN);
1486 		if (err < 0)
1487 			goto out;
1488 	} else if (len >= MLD_V2_QUERY_LEN_MIN) {
1489 		int srcs_offset = sizeof(struct mld2_query) -
1490 				  sizeof(struct icmp6hdr);
1491 
1492 		if (!pskb_may_pull(skb, srcs_offset))
1493 			goto out;
1494 
1495 		mlh2 = (struct mld2_query *)skb_transport_header(skb);
1496 
1497 		mld_process_v2(idev, mlh2, &max_delay);
1498 
1499 		if (group_type == IPV6_ADDR_ANY) { /* general query */
1500 			if (mlh2->mld2q_nsrcs)
1501 				goto out; /* no sources allowed */
1502 
1503 			mld_gq_start_work(idev);
1504 			goto out;
1505 		}
1506 		/* mark sources to include, if group & source-specific */
1507 		if (mlh2->mld2q_nsrcs != 0) {
1508 			if (!pskb_may_pull(skb, srcs_offset +
1509 			    ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1510 				goto out;
1511 
1512 			mlh2 = (struct mld2_query *)skb_transport_header(skb);
1513 			mark = 1;
1514 		}
1515 	} else {
1516 		goto out;
1517 	}
1518 
1519 	if (group_type == IPV6_ADDR_ANY) {
1520 		for_each_mc_mclock(idev, ma) {
1521 			igmp6_group_queried(ma, max_delay);
1522 		}
1523 	} else {
1524 		for_each_mc_mclock(idev, ma) {
1525 			if (!ipv6_addr_equal(group, &ma->mca_addr))
1526 				continue;
1527 			if (ma->mca_flags & MAF_TIMER_RUNNING) {
1528 				/* gsquery <- gsquery && mark */
1529 				if (!mark)
1530 					ma->mca_flags &= ~MAF_GSQUERY;
1531 			} else {
1532 				/* gsquery <- mark */
1533 				if (mark)
1534 					ma->mca_flags |= MAF_GSQUERY;
1535 				else
1536 					ma->mca_flags &= ~MAF_GSQUERY;
1537 			}
1538 			if (!(ma->mca_flags & MAF_GSQUERY) ||
1539 			    mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1540 				igmp6_group_queried(ma, max_delay);
1541 			break;
1542 		}
1543 	}
1544 
1545 out:
1546 	in6_dev_put(idev);
1547 kfree_skb:
1548 	consume_skb(skb);
1549 }
1550 
mld_query_work(struct work_struct * work)1551 static void mld_query_work(struct work_struct *work)
1552 {
1553 	struct inet6_dev *idev = container_of(to_delayed_work(work),
1554 					      struct inet6_dev,
1555 					      mc_query_work);
1556 	struct sk_buff_head q;
1557 	struct sk_buff *skb;
1558 	bool rework = false;
1559 	int cnt = 0;
1560 
1561 	skb_queue_head_init(&q);
1562 
1563 	spin_lock_bh(&idev->mc_query_lock);
1564 	while ((skb = __skb_dequeue(&idev->mc_query_queue))) {
1565 		__skb_queue_tail(&q, skb);
1566 
1567 		if (++cnt >= MLD_MAX_QUEUE) {
1568 			rework = true;
1569 			break;
1570 		}
1571 	}
1572 	spin_unlock_bh(&idev->mc_query_lock);
1573 
1574 	mutex_lock(&idev->mc_lock);
1575 	while ((skb = __skb_dequeue(&q)))
1576 		__mld_query_work(skb);
1577 	mutex_unlock(&idev->mc_lock);
1578 
1579 	if (rework && queue_delayed_work(mld_wq, &idev->mc_query_work, 0))
1580 		return;
1581 
1582 	in6_dev_put(idev);
1583 }
1584 
1585 /* called with rcu_read_lock() */
igmp6_event_report(struct sk_buff * skb)1586 void igmp6_event_report(struct sk_buff *skb)
1587 {
1588 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1589 
1590 	if (!idev || idev->dead)
1591 		goto out;
1592 
1593 	spin_lock_bh(&idev->mc_report_lock);
1594 	if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) {
1595 		__skb_queue_tail(&idev->mc_report_queue, skb);
1596 		if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0))
1597 			in6_dev_hold(idev);
1598 		skb = NULL;
1599 	}
1600 	spin_unlock_bh(&idev->mc_report_lock);
1601 out:
1602 	kfree_skb(skb);
1603 }
1604 
__mld_report_work(struct sk_buff * skb)1605 static void __mld_report_work(struct sk_buff *skb)
1606 {
1607 	struct inet6_dev *idev;
1608 	struct ifmcaddr6 *ma;
1609 	struct mld_msg *mld;
1610 	int addr_type;
1611 
1612 	/* Our own report looped back. Ignore it. */
1613 	if (skb->pkt_type == PACKET_LOOPBACK)
1614 		goto kfree_skb;
1615 
1616 	/* send our report if the MC router may not have heard this report */
1617 	if (skb->pkt_type != PACKET_MULTICAST &&
1618 	    skb->pkt_type != PACKET_BROADCAST)
1619 		goto kfree_skb;
1620 
1621 	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1622 		goto kfree_skb;
1623 
1624 	mld = (struct mld_msg *)icmp6_hdr(skb);
1625 
1626 	/* Drop reports with not link local source */
1627 	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1628 	if (addr_type != IPV6_ADDR_ANY &&
1629 	    !(addr_type&IPV6_ADDR_LINKLOCAL))
1630 		goto kfree_skb;
1631 
1632 	idev = in6_dev_get(skb->dev);
1633 	if (!idev)
1634 		goto kfree_skb;
1635 
1636 	/*
1637 	 *	Cancel the work for this group
1638 	 */
1639 
1640 	for_each_mc_mclock(idev, ma) {
1641 		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1642 			if (cancel_delayed_work(&ma->mca_work))
1643 				refcount_dec(&ma->mca_refcnt);
1644 			ma->mca_flags &= ~(MAF_LAST_REPORTER |
1645 					   MAF_TIMER_RUNNING);
1646 			break;
1647 		}
1648 	}
1649 
1650 	in6_dev_put(idev);
1651 kfree_skb:
1652 	consume_skb(skb);
1653 }
1654 
mld_report_work(struct work_struct * work)1655 static void mld_report_work(struct work_struct *work)
1656 {
1657 	struct inet6_dev *idev = container_of(to_delayed_work(work),
1658 					      struct inet6_dev,
1659 					      mc_report_work);
1660 	struct sk_buff_head q;
1661 	struct sk_buff *skb;
1662 	bool rework = false;
1663 	int cnt = 0;
1664 
1665 	skb_queue_head_init(&q);
1666 	spin_lock_bh(&idev->mc_report_lock);
1667 	while ((skb = __skb_dequeue(&idev->mc_report_queue))) {
1668 		__skb_queue_tail(&q, skb);
1669 
1670 		if (++cnt >= MLD_MAX_QUEUE) {
1671 			rework = true;
1672 			break;
1673 		}
1674 	}
1675 	spin_unlock_bh(&idev->mc_report_lock);
1676 
1677 	mutex_lock(&idev->mc_lock);
1678 	while ((skb = __skb_dequeue(&q)))
1679 		__mld_report_work(skb);
1680 	mutex_unlock(&idev->mc_lock);
1681 
1682 	if (rework && queue_delayed_work(mld_wq, &idev->mc_report_work, 0))
1683 		return;
1684 
1685 	in6_dev_put(idev);
1686 }
1687 
is_in(struct ifmcaddr6 * pmc,struct ip6_sf_list * psf,int type,int gdeleted,int sdeleted)1688 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1689 		  int gdeleted, int sdeleted)
1690 {
1691 	switch (type) {
1692 	case MLD2_MODE_IS_INCLUDE:
1693 	case MLD2_MODE_IS_EXCLUDE:
1694 		if (gdeleted || sdeleted)
1695 			return false;
1696 		if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1697 			if (pmc->mca_sfmode == MCAST_INCLUDE)
1698 				return true;
1699 			/* don't include if this source is excluded
1700 			 * in all filters
1701 			 */
1702 			if (psf->sf_count[MCAST_INCLUDE])
1703 				return type == MLD2_MODE_IS_INCLUDE;
1704 			return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1705 				psf->sf_count[MCAST_EXCLUDE];
1706 		}
1707 		return false;
1708 	case MLD2_CHANGE_TO_INCLUDE:
1709 		if (gdeleted || sdeleted)
1710 			return false;
1711 		return psf->sf_count[MCAST_INCLUDE] != 0;
1712 	case MLD2_CHANGE_TO_EXCLUDE:
1713 		if (gdeleted || sdeleted)
1714 			return false;
1715 		if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1716 		    psf->sf_count[MCAST_INCLUDE])
1717 			return false;
1718 		return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1719 			psf->sf_count[MCAST_EXCLUDE];
1720 	case MLD2_ALLOW_NEW_SOURCES:
1721 		if (gdeleted || !psf->sf_crcount)
1722 			return false;
1723 		return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1724 	case MLD2_BLOCK_OLD_SOURCES:
1725 		if (pmc->mca_sfmode == MCAST_INCLUDE)
1726 			return gdeleted || (psf->sf_crcount && sdeleted);
1727 		return psf->sf_crcount && !gdeleted && !sdeleted;
1728 	}
1729 	return false;
1730 }
1731 
1732 static int
mld_scount(struct ifmcaddr6 * pmc,int type,int gdeleted,int sdeleted)1733 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1734 {
1735 	struct ip6_sf_list *psf;
1736 	int scount = 0;
1737 
1738 	for_each_psf_mclock(pmc, psf) {
1739 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1740 			continue;
1741 		scount++;
1742 	}
1743 	return scount;
1744 }
1745 
ip6_mc_hdr(const struct sock * sk,struct sk_buff * skb,struct net_device * dev,const struct in6_addr * saddr,const struct in6_addr * daddr,int proto,int len)1746 static void ip6_mc_hdr(const struct sock *sk, struct sk_buff *skb,
1747 		       struct net_device *dev, const struct in6_addr *saddr,
1748 		       const struct in6_addr *daddr, int proto, int len)
1749 {
1750 	struct ipv6hdr *hdr;
1751 
1752 	skb->protocol = htons(ETH_P_IPV6);
1753 	skb->dev = dev;
1754 
1755 	skb_reset_network_header(skb);
1756 	skb_put(skb, sizeof(struct ipv6hdr));
1757 	hdr = ipv6_hdr(skb);
1758 
1759 	ip6_flow_hdr(hdr, 0, 0);
1760 
1761 	hdr->payload_len = htons(len);
1762 	hdr->nexthdr = proto;
1763 	hdr->hop_limit = READ_ONCE(inet6_sk(sk)->hop_limit);
1764 
1765 	hdr->saddr = *saddr;
1766 	hdr->daddr = *daddr;
1767 }
1768 
mld_newpack(struct inet6_dev * idev,unsigned int mtu)1769 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1770 {
1771 	u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
1772 		     2, 0, 0, IPV6_TLV_PADN, 0 };
1773 	struct net_device *dev = idev->dev;
1774 	int hlen = LL_RESERVED_SPACE(dev);
1775 	int tlen = dev->needed_tailroom;
1776 	const struct in6_addr *saddr;
1777 	struct in6_addr addr_buf;
1778 	struct mld2_report *pmr;
1779 	struct sk_buff *skb;
1780 	unsigned int size;
1781 	struct sock *sk;
1782 	struct net *net;
1783 
1784 	/* we assume size > sizeof(ra) here
1785 	 * Also try to not allocate high-order pages for big MTU
1786 	 */
1787 	size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen;
1788 	skb = alloc_skb(size, GFP_KERNEL);
1789 	if (!skb)
1790 		return NULL;
1791 
1792 	skb->priority = TC_PRIO_CONTROL;
1793 	skb_reserve(skb, hlen);
1794 	skb_tailroom_reserve(skb, mtu, tlen);
1795 
1796 	rcu_read_lock();
1797 
1798 	net = dev_net_rcu(dev);
1799 	sk = net->ipv6.igmp_sk;
1800 	skb_set_owner_w(skb, sk);
1801 
1802 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1803 		/* <draft-ietf-magma-mld-source-05.txt>:
1804 		 * use unspecified address as the source address
1805 		 * when a valid link-local address is not available.
1806 		 */
1807 		saddr = &in6addr_any;
1808 	} else
1809 		saddr = &addr_buf;
1810 
1811 	ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1812 
1813 	rcu_read_unlock();
1814 
1815 	skb_put_data(skb, ra, sizeof(ra));
1816 
1817 	skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1818 	skb_put(skb, sizeof(*pmr));
1819 	pmr = (struct mld2_report *)skb_transport_header(skb);
1820 	pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1821 	pmr->mld2r_resv1 = 0;
1822 	pmr->mld2r_cksum = 0;
1823 	pmr->mld2r_resv2 = 0;
1824 	pmr->mld2r_ngrec = 0;
1825 	return skb;
1826 }
1827 
mld_sendpack(struct sk_buff * skb)1828 static void mld_sendpack(struct sk_buff *skb)
1829 {
1830 	struct ipv6hdr *pip6 = ipv6_hdr(skb);
1831 	struct mld2_report *pmr =
1832 			      (struct mld2_report *)skb_transport_header(skb);
1833 	int payload_len, mldlen;
1834 	struct inet6_dev *idev;
1835 	struct net *net = dev_net(skb->dev);
1836 	int err;
1837 	struct flowi6 fl6;
1838 	struct dst_entry *dst;
1839 
1840 	rcu_read_lock();
1841 	idev = __in6_dev_get(skb->dev);
1842 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
1843 
1844 	payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1845 		sizeof(*pip6);
1846 	mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1847 	pip6->payload_len = htons(payload_len);
1848 
1849 	pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1850 					   IPPROTO_ICMPV6,
1851 					   csum_partial(skb_transport_header(skb),
1852 							mldlen, 0));
1853 
1854 	icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1855 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1856 			 skb->dev->ifindex);
1857 	dst = icmp6_dst_alloc(skb->dev, &fl6);
1858 
1859 	err = 0;
1860 	if (IS_ERR(dst)) {
1861 		err = PTR_ERR(dst);
1862 		dst = NULL;
1863 	}
1864 	skb_dst_set(skb, dst);
1865 	if (err)
1866 		goto err_out;
1867 
1868 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1869 		      net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1870 		      dst_output);
1871 out:
1872 	if (!err) {
1873 		ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1874 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1875 	} else {
1876 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1877 	}
1878 
1879 	rcu_read_unlock();
1880 	return;
1881 
1882 err_out:
1883 	kfree_skb(skb);
1884 	goto out;
1885 }
1886 
grec_size(struct ifmcaddr6 * pmc,int type,int gdel,int sdel)1887 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1888 {
1889 	return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1890 }
1891 
add_grhead(struct sk_buff * skb,struct ifmcaddr6 * pmc,int type,struct mld2_grec ** ppgr,unsigned int mtu)1892 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1893 	int type, struct mld2_grec **ppgr, unsigned int mtu)
1894 {
1895 	struct mld2_report *pmr;
1896 	struct mld2_grec *pgr;
1897 
1898 	if (!skb) {
1899 		skb = mld_newpack(pmc->idev, mtu);
1900 		if (!skb)
1901 			return NULL;
1902 	}
1903 	pgr = skb_put(skb, sizeof(struct mld2_grec));
1904 	pgr->grec_type = type;
1905 	pgr->grec_auxwords = 0;
1906 	pgr->grec_nsrcs = 0;
1907 	pgr->grec_mca = pmc->mca_addr;	/* structure copy */
1908 	pmr = (struct mld2_report *)skb_transport_header(skb);
1909 	pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1910 	*ppgr = pgr;
1911 	return skb;
1912 }
1913 
1914 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
1915 
1916 /* called with mc_lock */
add_grec(struct sk_buff * skb,struct ifmcaddr6 * pmc,int type,int gdeleted,int sdeleted,int crsend)1917 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1918 				int type, int gdeleted, int sdeleted,
1919 				int crsend)
1920 {
1921 	struct ip6_sf_list *psf, *psf_prev, *psf_next;
1922 	int scount, stotal, first, isquery, truncate;
1923 	struct ip6_sf_list __rcu **psf_list;
1924 	struct inet6_dev *idev = pmc->idev;
1925 	struct net_device *dev = idev->dev;
1926 	struct mld2_grec *pgr = NULL;
1927 	struct mld2_report *pmr;
1928 	unsigned int mtu;
1929 
1930 	if (pmc->mca_flags & MAF_NOREPORT)
1931 		return skb;
1932 
1933 	mtu = READ_ONCE(dev->mtu);
1934 	if (mtu < IPV6_MIN_MTU)
1935 		return skb;
1936 
1937 	isquery = type == MLD2_MODE_IS_INCLUDE ||
1938 		  type == MLD2_MODE_IS_EXCLUDE;
1939 	truncate = type == MLD2_MODE_IS_EXCLUDE ||
1940 		    type == MLD2_CHANGE_TO_EXCLUDE;
1941 
1942 	stotal = scount = 0;
1943 
1944 	psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1945 
1946 	if (!rcu_access_pointer(*psf_list))
1947 		goto empty_source;
1948 
1949 	pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1950 
1951 	/* EX and TO_EX get a fresh packet, if needed */
1952 	if (truncate) {
1953 		if (pmr && pmr->mld2r_ngrec &&
1954 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1955 			if (skb)
1956 				mld_sendpack(skb);
1957 			skb = mld_newpack(idev, mtu);
1958 		}
1959 	}
1960 	first = 1;
1961 	psf_prev = NULL;
1962 	for (psf = mc_dereference(*psf_list, idev);
1963 	     psf;
1964 	     psf = psf_next) {
1965 		struct in6_addr *psrc;
1966 
1967 		psf_next = mc_dereference(psf->sf_next, idev);
1968 
1969 		if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1970 			psf_prev = psf;
1971 			continue;
1972 		}
1973 
1974 		/* Based on RFC3810 6.1. Should not send source-list change
1975 		 * records when there is a filter mode change.
1976 		 */
1977 		if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1978 		     (!gdeleted && pmc->mca_crcount)) &&
1979 		    (type == MLD2_ALLOW_NEW_SOURCES ||
1980 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1981 			goto decrease_sf_crcount;
1982 
1983 		/* clear marks on query responses */
1984 		if (isquery)
1985 			psf->sf_gsresp = 0;
1986 
1987 		if (AVAILABLE(skb) < sizeof(*psrc) +
1988 		    first*sizeof(struct mld2_grec)) {
1989 			if (truncate && !first)
1990 				break;	 /* truncate these */
1991 			if (pgr)
1992 				pgr->grec_nsrcs = htons(scount);
1993 			if (skb)
1994 				mld_sendpack(skb);
1995 			skb = mld_newpack(idev, mtu);
1996 			first = 1;
1997 			scount = 0;
1998 		}
1999 		if (first) {
2000 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
2001 			first = 0;
2002 		}
2003 		if (!skb)
2004 			return NULL;
2005 		psrc = skb_put(skb, sizeof(*psrc));
2006 		*psrc = psf->sf_addr;
2007 		scount++; stotal++;
2008 		if ((type == MLD2_ALLOW_NEW_SOURCES ||
2009 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
2010 decrease_sf_crcount:
2011 			psf->sf_crcount--;
2012 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
2013 				if (psf_prev)
2014 					rcu_assign_pointer(psf_prev->sf_next,
2015 							   mc_dereference(psf->sf_next, idev));
2016 				else
2017 					rcu_assign_pointer(*psf_list,
2018 							   mc_dereference(psf->sf_next, idev));
2019 				kfree_rcu(psf, rcu);
2020 				continue;
2021 			}
2022 		}
2023 		psf_prev = psf;
2024 	}
2025 
2026 empty_source:
2027 	if (!stotal) {
2028 		if (type == MLD2_ALLOW_NEW_SOURCES ||
2029 		    type == MLD2_BLOCK_OLD_SOURCES)
2030 			return skb;
2031 		if (pmc->mca_crcount || isquery || crsend) {
2032 			/* make sure we have room for group header */
2033 			if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
2034 				mld_sendpack(skb);
2035 				skb = NULL; /* add_grhead will get a new one */
2036 			}
2037 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
2038 		}
2039 	}
2040 	if (pgr)
2041 		pgr->grec_nsrcs = htons(scount);
2042 
2043 	if (isquery)
2044 		pmc->mca_flags &= ~MAF_GSQUERY;	/* clear query state */
2045 	return skb;
2046 }
2047 
2048 /* called with mc_lock */
mld_send_report(struct inet6_dev * idev,struct ifmcaddr6 * pmc)2049 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
2050 {
2051 	struct sk_buff *skb = NULL;
2052 	int type;
2053 
2054 	if (!pmc) {
2055 		for_each_mc_mclock(idev, pmc) {
2056 			if (pmc->mca_flags & MAF_NOREPORT)
2057 				continue;
2058 			if (pmc->mca_sfcount[MCAST_EXCLUDE])
2059 				type = MLD2_MODE_IS_EXCLUDE;
2060 			else
2061 				type = MLD2_MODE_IS_INCLUDE;
2062 			skb = add_grec(skb, pmc, type, 0, 0, 0);
2063 		}
2064 	} else {
2065 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2066 			type = MLD2_MODE_IS_EXCLUDE;
2067 		else
2068 			type = MLD2_MODE_IS_INCLUDE;
2069 		skb = add_grec(skb, pmc, type, 0, 0, 0);
2070 	}
2071 	if (skb)
2072 		mld_sendpack(skb);
2073 }
2074 
2075 /*
2076  * remove zero-count source records from a source filter list
2077  * called with mc_lock
2078  */
mld_clear_zeros(struct ip6_sf_list __rcu ** ppsf,struct inet6_dev * idev)2079 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
2080 {
2081 	struct ip6_sf_list *psf_prev, *psf_next, *psf;
2082 
2083 	psf_prev = NULL;
2084 	for (psf = mc_dereference(*ppsf, idev);
2085 	     psf;
2086 	     psf = psf_next) {
2087 		psf_next = mc_dereference(psf->sf_next, idev);
2088 		if (psf->sf_crcount == 0) {
2089 			if (psf_prev)
2090 				rcu_assign_pointer(psf_prev->sf_next,
2091 						   mc_dereference(psf->sf_next, idev));
2092 			else
2093 				rcu_assign_pointer(*ppsf,
2094 						   mc_dereference(psf->sf_next, idev));
2095 			kfree_rcu(psf, rcu);
2096 		} else {
2097 			psf_prev = psf;
2098 		}
2099 	}
2100 }
2101 
2102 /* called with mc_lock */
mld_send_cr(struct inet6_dev * idev)2103 static void mld_send_cr(struct inet6_dev *idev)
2104 {
2105 	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
2106 	struct sk_buff *skb = NULL;
2107 	int type, dtype;
2108 
2109 	/* deleted MCA's */
2110 	pmc_prev = NULL;
2111 	for (pmc = mc_dereference(idev->mc_tomb, idev);
2112 	     pmc;
2113 	     pmc = pmc_next) {
2114 		pmc_next = mc_dereference(pmc->next, idev);
2115 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
2116 			type = MLD2_BLOCK_OLD_SOURCES;
2117 			dtype = MLD2_BLOCK_OLD_SOURCES;
2118 			skb = add_grec(skb, pmc, type, 1, 0, 0);
2119 			skb = add_grec(skb, pmc, dtype, 1, 1, 0);
2120 		}
2121 		if (pmc->mca_crcount) {
2122 			if (pmc->mca_sfmode == MCAST_EXCLUDE) {
2123 				type = MLD2_CHANGE_TO_INCLUDE;
2124 				skb = add_grec(skb, pmc, type, 1, 0, 0);
2125 			}
2126 			pmc->mca_crcount--;
2127 			if (pmc->mca_crcount == 0) {
2128 				mld_clear_zeros(&pmc->mca_tomb, idev);
2129 				mld_clear_zeros(&pmc->mca_sources, idev);
2130 			}
2131 		}
2132 		if (pmc->mca_crcount == 0 &&
2133 		    !rcu_access_pointer(pmc->mca_tomb) &&
2134 		    !rcu_access_pointer(pmc->mca_sources)) {
2135 			if (pmc_prev)
2136 				rcu_assign_pointer(pmc_prev->next, pmc_next);
2137 			else
2138 				rcu_assign_pointer(idev->mc_tomb, pmc_next);
2139 			in6_dev_put(pmc->idev);
2140 			kfree_rcu(pmc, rcu);
2141 		} else
2142 			pmc_prev = pmc;
2143 	}
2144 
2145 	/* change recs */
2146 	for_each_mc_mclock(idev, pmc) {
2147 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2148 			type = MLD2_BLOCK_OLD_SOURCES;
2149 			dtype = MLD2_ALLOW_NEW_SOURCES;
2150 		} else {
2151 			type = MLD2_ALLOW_NEW_SOURCES;
2152 			dtype = MLD2_BLOCK_OLD_SOURCES;
2153 		}
2154 		skb = add_grec(skb, pmc, type, 0, 0, 0);
2155 		skb = add_grec(skb, pmc, dtype, 0, 1, 0);	/* deleted sources */
2156 
2157 		/* filter mode changes */
2158 		if (pmc->mca_crcount) {
2159 			if (pmc->mca_sfmode == MCAST_EXCLUDE)
2160 				type = MLD2_CHANGE_TO_EXCLUDE;
2161 			else
2162 				type = MLD2_CHANGE_TO_INCLUDE;
2163 			skb = add_grec(skb, pmc, type, 0, 0, 0);
2164 			pmc->mca_crcount--;
2165 		}
2166 	}
2167 	if (!skb)
2168 		return;
2169 	(void) mld_sendpack(skb);
2170 }
2171 
igmp6_send(struct in6_addr * addr,struct net_device * dev,int type)2172 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
2173 {
2174 	const struct in6_addr *snd_addr, *saddr;
2175 	int err, len, payload_len, full_len;
2176 	struct in6_addr addr_buf;
2177 	struct inet6_dev *idev;
2178 	struct sk_buff *skb;
2179 	struct mld_msg *hdr;
2180 	int hlen = LL_RESERVED_SPACE(dev);
2181 	int tlen = dev->needed_tailroom;
2182 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
2183 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
2184 		     IPV6_TLV_PADN, 0 };
2185 	struct dst_entry *dst;
2186 	struct flowi6 fl6;
2187 	struct net *net;
2188 	struct sock *sk;
2189 
2190 	if (type == ICMPV6_MGM_REDUCTION)
2191 		snd_addr = &in6addr_linklocal_allrouters;
2192 	else
2193 		snd_addr = addr;
2194 
2195 	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2196 	payload_len = len + sizeof(ra);
2197 	full_len = sizeof(struct ipv6hdr) + payload_len;
2198 
2199 	skb = alloc_skb(hlen + tlen + full_len, GFP_KERNEL);
2200 
2201 	rcu_read_lock();
2202 
2203 	net = dev_net_rcu(dev);
2204 	idev = __in6_dev_get(dev);
2205 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
2206 	if (!skb) {
2207 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2208 		rcu_read_unlock();
2209 		return;
2210 	}
2211 	sk = net->ipv6.igmp_sk;
2212 	skb_set_owner_w(skb, sk);
2213 
2214 	skb->priority = TC_PRIO_CONTROL;
2215 	skb_reserve(skb, hlen);
2216 
2217 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2218 		/* <draft-ietf-magma-mld-source-05.txt>:
2219 		 * use unspecified address as the source address
2220 		 * when a valid link-local address is not available.
2221 		 */
2222 		saddr = &in6addr_any;
2223 	} else
2224 		saddr = &addr_buf;
2225 
2226 	ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2227 
2228 	skb_put_data(skb, ra, sizeof(ra));
2229 
2230 	hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2231 	hdr->mld_type = type;
2232 	hdr->mld_mca = *addr;
2233 
2234 	hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2235 					 IPPROTO_ICMPV6,
2236 					 csum_partial(hdr, len, 0));
2237 
2238 	icmpv6_flow_init(sk, &fl6, type,
2239 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2240 			 skb->dev->ifindex);
2241 	dst = icmp6_dst_alloc(skb->dev, &fl6);
2242 	if (IS_ERR(dst)) {
2243 		err = PTR_ERR(dst);
2244 		goto err_out;
2245 	}
2246 
2247 	skb_dst_set(skb, dst);
2248 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2249 		      net, sk, skb, NULL, skb->dev,
2250 		      dst_output);
2251 out:
2252 	if (!err) {
2253 		ICMP6MSGOUT_INC_STATS(net, idev, type);
2254 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2255 	} else
2256 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2257 
2258 	rcu_read_unlock();
2259 	return;
2260 
2261 err_out:
2262 	kfree_skb(skb);
2263 	goto out;
2264 }
2265 
2266 /* called with mc_lock */
mld_send_initial_cr(struct inet6_dev * idev)2267 static void mld_send_initial_cr(struct inet6_dev *idev)
2268 {
2269 	struct sk_buff *skb;
2270 	struct ifmcaddr6 *pmc;
2271 	int type;
2272 
2273 	if (mld_in_v1_mode(idev))
2274 		return;
2275 
2276 	skb = NULL;
2277 	for_each_mc_mclock(idev, pmc) {
2278 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2279 			type = MLD2_CHANGE_TO_EXCLUDE;
2280 		else
2281 			type = MLD2_ALLOW_NEW_SOURCES;
2282 		skb = add_grec(skb, pmc, type, 0, 0, 1);
2283 	}
2284 	if (skb)
2285 		mld_sendpack(skb);
2286 }
2287 
ipv6_mc_dad_complete(struct inet6_dev * idev)2288 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2289 {
2290 	mutex_lock(&idev->mc_lock);
2291 	idev->mc_dad_count = idev->mc_qrv;
2292 	if (idev->mc_dad_count) {
2293 		mld_send_initial_cr(idev);
2294 		idev->mc_dad_count--;
2295 		if (idev->mc_dad_count)
2296 			mld_dad_start_work(idev,
2297 					   unsolicited_report_interval(idev));
2298 	}
2299 	mutex_unlock(&idev->mc_lock);
2300 }
2301 
mld_dad_work(struct work_struct * work)2302 static void mld_dad_work(struct work_struct *work)
2303 {
2304 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2305 					      struct inet6_dev,
2306 					      mc_dad_work);
2307 	mutex_lock(&idev->mc_lock);
2308 	mld_send_initial_cr(idev);
2309 	if (idev->mc_dad_count) {
2310 		idev->mc_dad_count--;
2311 		if (idev->mc_dad_count)
2312 			mld_dad_start_work(idev,
2313 					   unsolicited_report_interval(idev));
2314 	}
2315 	mutex_unlock(&idev->mc_lock);
2316 	in6_dev_put(idev);
2317 }
2318 
2319 /* called with mc_lock */
ip6_mc_del1_src(struct ifmcaddr6 * pmc,int sfmode,const struct in6_addr * psfsrc)2320 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2321 	const struct in6_addr *psfsrc)
2322 {
2323 	struct ip6_sf_list *psf, *psf_prev;
2324 	int rv = 0;
2325 
2326 	psf_prev = NULL;
2327 	for_each_psf_mclock(pmc, psf) {
2328 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2329 			break;
2330 		psf_prev = psf;
2331 	}
2332 	if (!psf || psf->sf_count[sfmode] == 0) {
2333 		/* source filter not found, or count wrong =>  bug */
2334 		return -ESRCH;
2335 	}
2336 	WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] - 1);
2337 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2338 		struct inet6_dev *idev = pmc->idev;
2339 
2340 		/* no more filters for this source */
2341 		if (psf_prev)
2342 			rcu_assign_pointer(psf_prev->sf_next,
2343 					   mc_dereference(psf->sf_next, idev));
2344 		else
2345 			rcu_assign_pointer(pmc->mca_sources,
2346 					   mc_dereference(psf->sf_next, idev));
2347 
2348 		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2349 		    !mld_in_v1_mode(idev)) {
2350 			psf->sf_crcount = idev->mc_qrv;
2351 			rcu_assign_pointer(psf->sf_next,
2352 					   mc_dereference(pmc->mca_tomb, idev));
2353 			rcu_assign_pointer(pmc->mca_tomb, psf);
2354 			rv = 1;
2355 		} else {
2356 			kfree_rcu(psf, rcu);
2357 		}
2358 	}
2359 	return rv;
2360 }
2361 
2362 /* called with mc_lock */
ip6_mc_del_src(struct inet6_dev * idev,const struct in6_addr * pmca,int sfmode,int sfcount,const struct in6_addr * psfsrc,int delta)2363 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2364 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2365 			  int delta)
2366 {
2367 	struct ifmcaddr6 *pmc;
2368 	int	changerec = 0;
2369 	int	i, err;
2370 
2371 	if (!idev)
2372 		return -ENODEV;
2373 
2374 	for_each_mc_mclock(idev, pmc) {
2375 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2376 			break;
2377 	}
2378 	if (!pmc)
2379 		return -ESRCH;
2380 
2381 	sf_markstate(pmc);
2382 	if (!delta) {
2383 		if (!pmc->mca_sfcount[sfmode])
2384 			return -EINVAL;
2385 
2386 		pmc->mca_sfcount[sfmode]--;
2387 	}
2388 	err = 0;
2389 	for (i = 0; i < sfcount; i++) {
2390 		int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2391 
2392 		changerec |= rv > 0;
2393 		if (!err && rv < 0)
2394 			err = rv;
2395 	}
2396 	if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2397 	    pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2398 	    pmc->mca_sfcount[MCAST_INCLUDE]) {
2399 		struct ip6_sf_list *psf;
2400 
2401 		/* filter mode change */
2402 		pmc->mca_sfmode = MCAST_INCLUDE;
2403 		pmc->mca_crcount = idev->mc_qrv;
2404 		idev->mc_ifc_count = pmc->mca_crcount;
2405 		for_each_psf_mclock(pmc, psf)
2406 			psf->sf_crcount = 0;
2407 		mld_ifc_event(pmc->idev);
2408 	} else if (sf_setstate(pmc) || changerec) {
2409 		mld_ifc_event(pmc->idev);
2410 	}
2411 
2412 	return err;
2413 }
2414 
2415 /*
2416  * Add multicast single-source filter to the interface list
2417  * called with mc_lock
2418  */
ip6_mc_add1_src(struct ifmcaddr6 * pmc,int sfmode,const struct in6_addr * psfsrc)2419 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2420 	const struct in6_addr *psfsrc)
2421 {
2422 	struct ip6_sf_list *psf, *psf_prev;
2423 
2424 	psf_prev = NULL;
2425 	for_each_psf_mclock(pmc, psf) {
2426 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2427 			break;
2428 		psf_prev = psf;
2429 	}
2430 	if (!psf) {
2431 		psf = kzalloc(sizeof(*psf), GFP_KERNEL);
2432 		if (!psf)
2433 			return -ENOBUFS;
2434 
2435 		psf->sf_addr = *psfsrc;
2436 		if (psf_prev) {
2437 			rcu_assign_pointer(psf_prev->sf_next, psf);
2438 		} else {
2439 			rcu_assign_pointer(pmc->mca_sources, psf);
2440 		}
2441 	}
2442 	WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] + 1);
2443 	return 0;
2444 }
2445 
2446 /* called with mc_lock */
sf_markstate(struct ifmcaddr6 * pmc)2447 static void sf_markstate(struct ifmcaddr6 *pmc)
2448 {
2449 	struct ip6_sf_list *psf;
2450 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2451 
2452 	for_each_psf_mclock(pmc, psf) {
2453 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2454 			psf->sf_oldin = mca_xcount ==
2455 				psf->sf_count[MCAST_EXCLUDE] &&
2456 				!psf->sf_count[MCAST_INCLUDE];
2457 		} else {
2458 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2459 		}
2460 	}
2461 }
2462 
2463 /* called with mc_lock */
sf_setstate(struct ifmcaddr6 * pmc)2464 static int sf_setstate(struct ifmcaddr6 *pmc)
2465 {
2466 	struct ip6_sf_list *psf, *dpsf;
2467 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2468 	int qrv = pmc->idev->mc_qrv;
2469 	int new_in, rv;
2470 
2471 	rv = 0;
2472 	for_each_psf_mclock(pmc, psf) {
2473 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2474 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2475 				!psf->sf_count[MCAST_INCLUDE];
2476 		} else
2477 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2478 		if (new_in) {
2479 			if (!psf->sf_oldin) {
2480 				struct ip6_sf_list *prev = NULL;
2481 
2482 				for_each_psf_tomb(pmc, dpsf) {
2483 					if (ipv6_addr_equal(&dpsf->sf_addr,
2484 					    &psf->sf_addr))
2485 						break;
2486 					prev = dpsf;
2487 				}
2488 				if (dpsf) {
2489 					if (prev)
2490 						rcu_assign_pointer(prev->sf_next,
2491 								   mc_dereference(dpsf->sf_next,
2492 										  pmc->idev));
2493 					else
2494 						rcu_assign_pointer(pmc->mca_tomb,
2495 								   mc_dereference(dpsf->sf_next,
2496 										  pmc->idev));
2497 					kfree_rcu(dpsf, rcu);
2498 				}
2499 				psf->sf_crcount = qrv;
2500 				rv++;
2501 			}
2502 		} else if (psf->sf_oldin) {
2503 			psf->sf_crcount = 0;
2504 			/*
2505 			 * add or update "delete" records if an active filter
2506 			 * is now inactive
2507 			 */
2508 
2509 			for_each_psf_tomb(pmc, dpsf)
2510 				if (ipv6_addr_equal(&dpsf->sf_addr,
2511 				    &psf->sf_addr))
2512 					break;
2513 			if (!dpsf) {
2514 				dpsf = kmalloc(sizeof(*dpsf), GFP_KERNEL);
2515 				if (!dpsf)
2516 					continue;
2517 				*dpsf = *psf;
2518 				rcu_assign_pointer(dpsf->sf_next,
2519 						   mc_dereference(pmc->mca_tomb, pmc->idev));
2520 				rcu_assign_pointer(pmc->mca_tomb, dpsf);
2521 			}
2522 			dpsf->sf_crcount = qrv;
2523 			rv++;
2524 		}
2525 	}
2526 	return rv;
2527 }
2528 
2529 /*
2530  * Add multicast source filter list to the interface list
2531  * called with mc_lock
2532  */
ip6_mc_add_src(struct inet6_dev * idev,const struct in6_addr * pmca,int sfmode,int sfcount,const struct in6_addr * psfsrc,int delta)2533 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2534 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2535 			  int delta)
2536 {
2537 	struct ifmcaddr6 *pmc;
2538 	int	isexclude;
2539 	int	i, err;
2540 
2541 	if (!idev)
2542 		return -ENODEV;
2543 
2544 	for_each_mc_mclock(idev, pmc) {
2545 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2546 			break;
2547 	}
2548 	if (!pmc)
2549 		return -ESRCH;
2550 
2551 	sf_markstate(pmc);
2552 	isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2553 	if (!delta)
2554 		WRITE_ONCE(pmc->mca_sfcount[sfmode],
2555 			   pmc->mca_sfcount[sfmode] + 1);
2556 	err = 0;
2557 	for (i = 0; i < sfcount; i++) {
2558 		err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2559 		if (err)
2560 			break;
2561 	}
2562 	if (err) {
2563 		int j;
2564 
2565 		if (!delta)
2566 			WRITE_ONCE(pmc->mca_sfcount[sfmode],
2567 				   pmc->mca_sfcount[sfmode] - 1);
2568 		for (j = 0; j < i; j++)
2569 			ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2570 	} else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2571 		struct ip6_sf_list *psf;
2572 
2573 		/* filter mode change */
2574 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2575 			pmc->mca_sfmode = MCAST_EXCLUDE;
2576 		else if (pmc->mca_sfcount[MCAST_INCLUDE])
2577 			pmc->mca_sfmode = MCAST_INCLUDE;
2578 		/* else no filters; keep old mode for reports */
2579 
2580 		pmc->mca_crcount = idev->mc_qrv;
2581 		idev->mc_ifc_count = pmc->mca_crcount;
2582 		for_each_psf_mclock(pmc, psf)
2583 			psf->sf_crcount = 0;
2584 		mld_ifc_event(idev);
2585 	} else if (sf_setstate(pmc)) {
2586 		mld_ifc_event(idev);
2587 	}
2588 	return err;
2589 }
2590 
2591 /* called with mc_lock */
ip6_mc_clear_src(struct ifmcaddr6 * pmc)2592 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2593 {
2594 	struct ip6_sf_list *psf, *nextpsf;
2595 
2596 	for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
2597 	     psf;
2598 	     psf = nextpsf) {
2599 		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2600 		kfree_rcu(psf, rcu);
2601 	}
2602 	RCU_INIT_POINTER(pmc->mca_tomb, NULL);
2603 	for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
2604 	     psf;
2605 	     psf = nextpsf) {
2606 		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2607 		kfree_rcu(psf, rcu);
2608 	}
2609 	RCU_INIT_POINTER(pmc->mca_sources, NULL);
2610 	pmc->mca_sfmode = MCAST_EXCLUDE;
2611 	pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2612 	/* Paired with the READ_ONCE() from ipv6_chk_mcast_addr() */
2613 	WRITE_ONCE(pmc->mca_sfcount[MCAST_EXCLUDE], 1);
2614 }
2615 
2616 /* called with mc_lock */
igmp6_join_group(struct ifmcaddr6 * ma)2617 static void igmp6_join_group(struct ifmcaddr6 *ma)
2618 {
2619 	unsigned long delay;
2620 
2621 	if (ma->mca_flags & MAF_NOREPORT)
2622 		return;
2623 
2624 	igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2625 
2626 	delay = get_random_u32_below(unsolicited_report_interval(ma->idev));
2627 
2628 	if (cancel_delayed_work(&ma->mca_work)) {
2629 		refcount_dec(&ma->mca_refcnt);
2630 		delay = ma->mca_work.timer.expires - jiffies;
2631 	}
2632 
2633 	if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
2634 		refcount_inc(&ma->mca_refcnt);
2635 	ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2636 }
2637 
ip6_mc_leave_src(struct sock * sk,struct ipv6_mc_socklist * iml,struct inet6_dev * idev)2638 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2639 			    struct inet6_dev *idev)
2640 {
2641 	struct ip6_sf_socklist *psl;
2642 	int err;
2643 
2644 	psl = sock_dereference(iml->sflist, sk);
2645 
2646 	if (idev)
2647 		mutex_lock(&idev->mc_lock);
2648 
2649 	if (!psl) {
2650 		/* any-source empty exclude case */
2651 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2652 	} else {
2653 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2654 				     psl->sl_count, psl->sl_addr, 0);
2655 		RCU_INIT_POINTER(iml->sflist, NULL);
2656 		atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
2657 			   &sk->sk_omem_alloc);
2658 		kfree_rcu(psl, rcu);
2659 	}
2660 
2661 	if (idev)
2662 		mutex_unlock(&idev->mc_lock);
2663 
2664 	return err;
2665 }
2666 
2667 /* called with mc_lock */
igmp6_leave_group(struct ifmcaddr6 * ma)2668 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2669 {
2670 	if (mld_in_v1_mode(ma->idev)) {
2671 		if (ma->mca_flags & MAF_LAST_REPORTER) {
2672 			igmp6_send(&ma->mca_addr, ma->idev->dev,
2673 				ICMPV6_MGM_REDUCTION);
2674 		}
2675 	} else {
2676 		mld_add_delrec(ma->idev, ma);
2677 		mld_ifc_event(ma->idev);
2678 	}
2679 }
2680 
mld_gq_work(struct work_struct * work)2681 static void mld_gq_work(struct work_struct *work)
2682 {
2683 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2684 					      struct inet6_dev,
2685 					      mc_gq_work);
2686 
2687 	mutex_lock(&idev->mc_lock);
2688 	mld_send_report(idev, NULL);
2689 	idev->mc_gq_running = 0;
2690 	mutex_unlock(&idev->mc_lock);
2691 
2692 	in6_dev_put(idev);
2693 }
2694 
mld_ifc_work(struct work_struct * work)2695 static void mld_ifc_work(struct work_struct *work)
2696 {
2697 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2698 					      struct inet6_dev,
2699 					      mc_ifc_work);
2700 
2701 	mutex_lock(&idev->mc_lock);
2702 	mld_send_cr(idev);
2703 
2704 	if (idev->mc_ifc_count) {
2705 		idev->mc_ifc_count--;
2706 		if (idev->mc_ifc_count)
2707 			mld_ifc_start_work(idev,
2708 					   unsolicited_report_interval(idev));
2709 	}
2710 	mutex_unlock(&idev->mc_lock);
2711 	in6_dev_put(idev);
2712 }
2713 
2714 /* called with mc_lock */
mld_ifc_event(struct inet6_dev * idev)2715 static void mld_ifc_event(struct inet6_dev *idev)
2716 {
2717 	if (mld_in_v1_mode(idev))
2718 		return;
2719 
2720 	idev->mc_ifc_count = idev->mc_qrv;
2721 	mld_ifc_start_work(idev, 1);
2722 }
2723 
mld_mca_work(struct work_struct * work)2724 static void mld_mca_work(struct work_struct *work)
2725 {
2726 	struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
2727 					    struct ifmcaddr6, mca_work);
2728 
2729 	mutex_lock(&ma->idev->mc_lock);
2730 	if (mld_in_v1_mode(ma->idev))
2731 		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2732 	else
2733 		mld_send_report(ma->idev, ma);
2734 	ma->mca_flags |=  MAF_LAST_REPORTER;
2735 	ma->mca_flags &= ~MAF_TIMER_RUNNING;
2736 	mutex_unlock(&ma->idev->mc_lock);
2737 
2738 	ma_put(ma);
2739 }
2740 
2741 /* Device changing type */
2742 
ipv6_mc_unmap(struct inet6_dev * idev)2743 void ipv6_mc_unmap(struct inet6_dev *idev)
2744 {
2745 	struct ifmcaddr6 *i;
2746 
2747 	/* Install multicast list, except for all-nodes (already installed) */
2748 
2749 	mutex_lock(&idev->mc_lock);
2750 	for_each_mc_mclock(idev, i)
2751 		igmp6_group_dropped(i);
2752 	mutex_unlock(&idev->mc_lock);
2753 }
2754 
ipv6_mc_remap(struct inet6_dev * idev)2755 void ipv6_mc_remap(struct inet6_dev *idev)
2756 {
2757 	ipv6_mc_up(idev);
2758 }
2759 
2760 /* Device going down */
ipv6_mc_down(struct inet6_dev * idev)2761 void ipv6_mc_down(struct inet6_dev *idev)
2762 {
2763 	struct ifmcaddr6 *i;
2764 
2765 	mutex_lock(&idev->mc_lock);
2766 	/* Withdraw multicast list */
2767 	for_each_mc_mclock(idev, i)
2768 		igmp6_group_dropped(i);
2769 	mutex_unlock(&idev->mc_lock);
2770 
2771 	/* Should stop work after group drop. or we will
2772 	 * start work again in mld_ifc_event()
2773 	 */
2774 	mld_query_stop_work(idev);
2775 	mld_report_stop_work(idev);
2776 
2777 	mutex_lock(&idev->mc_lock);
2778 	mld_ifc_stop_work(idev);
2779 	mld_gq_stop_work(idev);
2780 	mutex_unlock(&idev->mc_lock);
2781 
2782 	mld_dad_stop_work(idev);
2783 }
2784 
ipv6_mc_reset(struct inet6_dev * idev)2785 static void ipv6_mc_reset(struct inet6_dev *idev)
2786 {
2787 	idev->mc_qrv = sysctl_mld_qrv;
2788 	idev->mc_qi = MLD_QI_DEFAULT;
2789 	idev->mc_qri = MLD_QRI_DEFAULT;
2790 	idev->mc_v1_seen = 0;
2791 	idev->mc_maxdelay = unsolicited_report_interval(idev);
2792 }
2793 
2794 /* Device going up */
2795 
ipv6_mc_up(struct inet6_dev * idev)2796 void ipv6_mc_up(struct inet6_dev *idev)
2797 {
2798 	struct ifmcaddr6 *i;
2799 
2800 	/* Install multicast list, except for all-nodes (already installed) */
2801 
2802 	ipv6_mc_reset(idev);
2803 	mutex_lock(&idev->mc_lock);
2804 	for_each_mc_mclock(idev, i) {
2805 		mld_del_delrec(idev, i);
2806 		igmp6_group_added(i);
2807 	}
2808 	mutex_unlock(&idev->mc_lock);
2809 }
2810 
2811 /* IPv6 device initialization. */
2812 
ipv6_mc_init_dev(struct inet6_dev * idev)2813 void ipv6_mc_init_dev(struct inet6_dev *idev)
2814 {
2815 	idev->mc_gq_running = 0;
2816 	INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work);
2817 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
2818 	idev->mc_ifc_count = 0;
2819 	INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work);
2820 	INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work);
2821 	INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work);
2822 	INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work);
2823 	skb_queue_head_init(&idev->mc_query_queue);
2824 	skb_queue_head_init(&idev->mc_report_queue);
2825 	spin_lock_init(&idev->mc_query_lock);
2826 	spin_lock_init(&idev->mc_report_lock);
2827 	mutex_init(&idev->mc_lock);
2828 	ipv6_mc_reset(idev);
2829 }
2830 
2831 /*
2832  *	Device is about to be destroyed: clean up.
2833  */
2834 
ipv6_mc_destroy_dev(struct inet6_dev * idev)2835 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2836 {
2837 	struct ifmcaddr6 *i;
2838 
2839 	/* Deactivate works */
2840 	ipv6_mc_down(idev);
2841 	mutex_lock(&idev->mc_lock);
2842 	mld_clear_delrec(idev);
2843 	mutex_unlock(&idev->mc_lock);
2844 	mld_clear_query(idev);
2845 	mld_clear_report(idev);
2846 
2847 	/* Delete all-nodes address. */
2848 	/* We cannot call ipv6_dev_mc_dec() directly, our caller in
2849 	 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2850 	 * fail.
2851 	 */
2852 	__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2853 
2854 	if (idev->cnf.forwarding)
2855 		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2856 
2857 	mutex_lock(&idev->mc_lock);
2858 	while ((i = mc_dereference(idev->mc_list, idev))) {
2859 		rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
2860 
2861 		ip6_mc_clear_src(i);
2862 		ma_put(i);
2863 	}
2864 	mutex_unlock(&idev->mc_lock);
2865 }
2866 
ipv6_mc_rejoin_groups(struct inet6_dev * idev)2867 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2868 {
2869 	struct ifmcaddr6 *pmc;
2870 
2871 	ASSERT_RTNL();
2872 
2873 	mutex_lock(&idev->mc_lock);
2874 	if (mld_in_v1_mode(idev)) {
2875 		for_each_mc_mclock(idev, pmc)
2876 			igmp6_join_group(pmc);
2877 	} else {
2878 		mld_send_report(idev, NULL);
2879 	}
2880 	mutex_unlock(&idev->mc_lock);
2881 }
2882 
ipv6_mc_netdev_event(struct notifier_block * this,unsigned long event,void * ptr)2883 static int ipv6_mc_netdev_event(struct notifier_block *this,
2884 				unsigned long event,
2885 				void *ptr)
2886 {
2887 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2888 	struct inet6_dev *idev = __in6_dev_get(dev);
2889 
2890 	switch (event) {
2891 	case NETDEV_RESEND_IGMP:
2892 		if (idev)
2893 			ipv6_mc_rejoin_groups(idev);
2894 		break;
2895 	default:
2896 		break;
2897 	}
2898 
2899 	return NOTIFY_DONE;
2900 }
2901 
2902 static struct notifier_block igmp6_netdev_notifier = {
2903 	.notifier_call = ipv6_mc_netdev_event,
2904 };
2905 
2906 #ifdef CONFIG_PROC_FS
2907 struct igmp6_mc_iter_state {
2908 	struct seq_net_private p;
2909 	struct net_device *dev;
2910 	struct inet6_dev *idev;
2911 };
2912 
2913 #define igmp6_mc_seq_private(seq)	((struct igmp6_mc_iter_state *)(seq)->private)
2914 
igmp6_mc_get_first(struct seq_file * seq)2915 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2916 {
2917 	struct ifmcaddr6 *im = NULL;
2918 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2919 	struct net *net = seq_file_net(seq);
2920 
2921 	state->idev = NULL;
2922 	for_each_netdev_rcu(net, state->dev) {
2923 		struct inet6_dev *idev;
2924 		idev = __in6_dev_get(state->dev);
2925 		if (!idev)
2926 			continue;
2927 
2928 		im = rcu_dereference(idev->mc_list);
2929 		if (im) {
2930 			state->idev = idev;
2931 			break;
2932 		}
2933 	}
2934 	return im;
2935 }
2936 
igmp6_mc_get_next(struct seq_file * seq,struct ifmcaddr6 * im)2937 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2938 {
2939 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2940 
2941 	im = rcu_dereference(im->next);
2942 	while (!im) {
2943 		state->dev = next_net_device_rcu(state->dev);
2944 		if (!state->dev) {
2945 			state->idev = NULL;
2946 			break;
2947 		}
2948 		state->idev = __in6_dev_get(state->dev);
2949 		if (!state->idev)
2950 			continue;
2951 		im = rcu_dereference(state->idev->mc_list);
2952 	}
2953 	return im;
2954 }
2955 
igmp6_mc_get_idx(struct seq_file * seq,loff_t pos)2956 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2957 {
2958 	struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2959 	if (im)
2960 		while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2961 			--pos;
2962 	return pos ? NULL : im;
2963 }
2964 
igmp6_mc_seq_start(struct seq_file * seq,loff_t * pos)2965 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2966 	__acquires(RCU)
2967 {
2968 	rcu_read_lock();
2969 	return igmp6_mc_get_idx(seq, *pos);
2970 }
2971 
igmp6_mc_seq_next(struct seq_file * seq,void * v,loff_t * pos)2972 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2973 {
2974 	struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2975 
2976 	++*pos;
2977 	return im;
2978 }
2979 
igmp6_mc_seq_stop(struct seq_file * seq,void * v)2980 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2981 	__releases(RCU)
2982 {
2983 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2984 
2985 	if (likely(state->idev))
2986 		state->idev = NULL;
2987 	state->dev = NULL;
2988 	rcu_read_unlock();
2989 }
2990 
igmp6_mc_seq_show(struct seq_file * seq,void * v)2991 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2992 {
2993 	struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2994 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2995 
2996 	seq_printf(seq,
2997 		   "%-4d %-15s %pi6 %5d %08X %ld\n",
2998 		   state->dev->ifindex, state->dev->name,
2999 		   &im->mca_addr,
3000 		   im->mca_users, im->mca_flags,
3001 		   (im->mca_flags & MAF_TIMER_RUNNING) ?
3002 		   jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0);
3003 	return 0;
3004 }
3005 
3006 static const struct seq_operations igmp6_mc_seq_ops = {
3007 	.start	=	igmp6_mc_seq_start,
3008 	.next	=	igmp6_mc_seq_next,
3009 	.stop	=	igmp6_mc_seq_stop,
3010 	.show	=	igmp6_mc_seq_show,
3011 };
3012 
3013 struct igmp6_mcf_iter_state {
3014 	struct seq_net_private p;
3015 	struct net_device *dev;
3016 	struct inet6_dev *idev;
3017 	struct ifmcaddr6 *im;
3018 };
3019 
3020 #define igmp6_mcf_seq_private(seq)	((struct igmp6_mcf_iter_state *)(seq)->private)
3021 
igmp6_mcf_get_first(struct seq_file * seq)3022 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
3023 {
3024 	struct ip6_sf_list *psf = NULL;
3025 	struct ifmcaddr6 *im = NULL;
3026 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3027 	struct net *net = seq_file_net(seq);
3028 
3029 	state->idev = NULL;
3030 	state->im = NULL;
3031 	for_each_netdev_rcu(net, state->dev) {
3032 		struct inet6_dev *idev;
3033 		idev = __in6_dev_get(state->dev);
3034 		if (unlikely(idev == NULL))
3035 			continue;
3036 
3037 		im = rcu_dereference(idev->mc_list);
3038 		if (likely(im)) {
3039 			psf = rcu_dereference(im->mca_sources);
3040 			if (likely(psf)) {
3041 				state->im = im;
3042 				state->idev = idev;
3043 				break;
3044 			}
3045 		}
3046 	}
3047 	return psf;
3048 }
3049 
igmp6_mcf_get_next(struct seq_file * seq,struct ip6_sf_list * psf)3050 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
3051 {
3052 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3053 
3054 	psf = rcu_dereference(psf->sf_next);
3055 	while (!psf) {
3056 		state->im = rcu_dereference(state->im->next);
3057 		while (!state->im) {
3058 			state->dev = next_net_device_rcu(state->dev);
3059 			if (!state->dev) {
3060 				state->idev = NULL;
3061 				goto out;
3062 			}
3063 			state->idev = __in6_dev_get(state->dev);
3064 			if (!state->idev)
3065 				continue;
3066 			state->im = rcu_dereference(state->idev->mc_list);
3067 		}
3068 		psf = rcu_dereference(state->im->mca_sources);
3069 	}
3070 out:
3071 	return psf;
3072 }
3073 
igmp6_mcf_get_idx(struct seq_file * seq,loff_t pos)3074 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
3075 {
3076 	struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
3077 	if (psf)
3078 		while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
3079 			--pos;
3080 	return pos ? NULL : psf;
3081 }
3082 
igmp6_mcf_seq_start(struct seq_file * seq,loff_t * pos)3083 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
3084 	__acquires(RCU)
3085 {
3086 	rcu_read_lock();
3087 	return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
3088 }
3089 
igmp6_mcf_seq_next(struct seq_file * seq,void * v,loff_t * pos)3090 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3091 {
3092 	struct ip6_sf_list *psf;
3093 	if (v == SEQ_START_TOKEN)
3094 		psf = igmp6_mcf_get_first(seq);
3095 	else
3096 		psf = igmp6_mcf_get_next(seq, v);
3097 	++*pos;
3098 	return psf;
3099 }
3100 
igmp6_mcf_seq_stop(struct seq_file * seq,void * v)3101 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
3102 	__releases(RCU)
3103 {
3104 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3105 
3106 	if (likely(state->im))
3107 		state->im = NULL;
3108 	if (likely(state->idev))
3109 		state->idev = NULL;
3110 
3111 	state->dev = NULL;
3112 	rcu_read_unlock();
3113 }
3114 
igmp6_mcf_seq_show(struct seq_file * seq,void * v)3115 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
3116 {
3117 	struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
3118 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3119 
3120 	if (v == SEQ_START_TOKEN) {
3121 		seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
3122 	} else {
3123 		seq_printf(seq,
3124 			   "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
3125 			   state->dev->ifindex, state->dev->name,
3126 			   &state->im->mca_addr,
3127 			   &psf->sf_addr,
3128 			   READ_ONCE(psf->sf_count[MCAST_INCLUDE]),
3129 			   READ_ONCE(psf->sf_count[MCAST_EXCLUDE]));
3130 	}
3131 	return 0;
3132 }
3133 
3134 static const struct seq_operations igmp6_mcf_seq_ops = {
3135 	.start	=	igmp6_mcf_seq_start,
3136 	.next	=	igmp6_mcf_seq_next,
3137 	.stop	=	igmp6_mcf_seq_stop,
3138 	.show	=	igmp6_mcf_seq_show,
3139 };
3140 
igmp6_proc_init(struct net * net)3141 static int __net_init igmp6_proc_init(struct net *net)
3142 {
3143 	int err;
3144 
3145 	err = -ENOMEM;
3146 	if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
3147 			sizeof(struct igmp6_mc_iter_state)))
3148 		goto out;
3149 	if (!proc_create_net("mcfilter6", 0444, net->proc_net,
3150 			&igmp6_mcf_seq_ops,
3151 			sizeof(struct igmp6_mcf_iter_state)))
3152 		goto out_proc_net_igmp6;
3153 
3154 	err = 0;
3155 out:
3156 	return err;
3157 
3158 out_proc_net_igmp6:
3159 	remove_proc_entry("igmp6", net->proc_net);
3160 	goto out;
3161 }
3162 
igmp6_proc_exit(struct net * net)3163 static void __net_exit igmp6_proc_exit(struct net *net)
3164 {
3165 	remove_proc_entry("mcfilter6", net->proc_net);
3166 	remove_proc_entry("igmp6", net->proc_net);
3167 }
3168 #else
igmp6_proc_init(struct net * net)3169 static inline int igmp6_proc_init(struct net *net)
3170 {
3171 	return 0;
3172 }
igmp6_proc_exit(struct net * net)3173 static inline void igmp6_proc_exit(struct net *net)
3174 {
3175 }
3176 #endif
3177 
igmp6_net_init(struct net * net)3178 static int __net_init igmp6_net_init(struct net *net)
3179 {
3180 	int err;
3181 
3182 	err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
3183 				   SOCK_RAW, IPPROTO_ICMPV6, net);
3184 	if (err < 0) {
3185 		pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
3186 		       err);
3187 		goto out;
3188 	}
3189 
3190 	inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
3191 	net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL;
3192 
3193 	err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
3194 				   SOCK_RAW, IPPROTO_ICMPV6, net);
3195 	if (err < 0) {
3196 		pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
3197 		       err);
3198 		goto out_sock_create;
3199 	}
3200 
3201 	err = igmp6_proc_init(net);
3202 	if (err)
3203 		goto out_sock_create_autojoin;
3204 
3205 	return 0;
3206 
3207 out_sock_create_autojoin:
3208 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3209 out_sock_create:
3210 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3211 out:
3212 	return err;
3213 }
3214 
igmp6_net_exit(struct net * net)3215 static void __net_exit igmp6_net_exit(struct net *net)
3216 {
3217 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3218 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3219 	igmp6_proc_exit(net);
3220 }
3221 
3222 static struct pernet_operations igmp6_net_ops = {
3223 	.init = igmp6_net_init,
3224 	.exit = igmp6_net_exit,
3225 };
3226 
igmp6_init(void)3227 int __init igmp6_init(void)
3228 {
3229 	int err;
3230 
3231 	err = register_pernet_subsys(&igmp6_net_ops);
3232 	if (err)
3233 		return err;
3234 
3235 	mld_wq = create_workqueue("mld");
3236 	if (!mld_wq) {
3237 		unregister_pernet_subsys(&igmp6_net_ops);
3238 		return -ENOMEM;
3239 	}
3240 
3241 	return err;
3242 }
3243 
igmp6_late_init(void)3244 int __init igmp6_late_init(void)
3245 {
3246 	return register_netdevice_notifier(&igmp6_netdev_notifier);
3247 }
3248 
igmp6_cleanup(void)3249 void igmp6_cleanup(void)
3250 {
3251 	unregister_pernet_subsys(&igmp6_net_ops);
3252 	destroy_workqueue(mld_wq);
3253 }
3254 
igmp6_late_cleanup(void)3255 void igmp6_late_cleanup(void)
3256 {
3257 	unregister_netdevice_notifier(&igmp6_netdev_notifier);
3258 }
3259