printf: Remove unused 'bprintf'
[drm/drm-misc.git] / net / ipv6 / mcast.c
blobb244dbf61d5f3969c8886a999166ad5ef68b746d
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Multicast support for IPv6
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
12 /* Changes:
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
25 #include <linux/module.h>
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/string.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/jiffies.h>
32 #include <linux/net.h>
33 #include <linux/in.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/route.h>
38 #include <linux/init.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/slab.h>
42 #include <linux/pkt_sched.h>
43 #include <net/mld.h>
44 #include <linux/workqueue.h>
46 #include <linux/netfilter.h>
47 #include <linux/netfilter_ipv6.h>
49 #include <net/net_namespace.h>
50 #include <net/sock.h>
51 #include <net/snmp.h>
53 #include <net/ipv6.h>
54 #include <net/protocol.h>
55 #include <net/if_inet6.h>
56 #include <net/ndisc.h>
57 #include <net/addrconf.h>
58 #include <net/ip6_route.h>
59 #include <net/inet_common.h>
61 #include <net/ip6_checksum.h>
63 /* Ensure that we have struct in6_addr aligned on 32bit word. */
64 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
65 BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
66 BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
67 BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
70 static struct workqueue_struct *mld_wq;
71 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
73 static void igmp6_join_group(struct ifmcaddr6 *ma);
74 static void igmp6_leave_group(struct ifmcaddr6 *ma);
75 static void mld_mca_work(struct work_struct *work);
77 static void mld_ifc_event(struct inet6_dev *idev);
78 static bool mld_in_v1_mode(const struct inet6_dev *idev);
79 static int sf_setstate(struct ifmcaddr6 *pmc);
80 static void sf_markstate(struct ifmcaddr6 *pmc);
81 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
82 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
83 int sfmode, int sfcount, const struct in6_addr *psfsrc,
84 int delta);
85 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
86 int sfmode, int sfcount, const struct in6_addr *psfsrc,
87 int delta);
88 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
89 struct inet6_dev *idev);
90 static int __ipv6_dev_mc_inc(struct net_device *dev,
91 const struct in6_addr *addr, unsigned int mode);
93 #define MLD_QRV_DEFAULT 2
94 /* RFC3810, 9.2. Query Interval */
95 #define MLD_QI_DEFAULT (125 * HZ)
96 /* RFC3810, 9.3. Query Response Interval */
97 #define MLD_QRI_DEFAULT (10 * HZ)
99 /* RFC3810, 8.1 Query Version Distinctions */
100 #define MLD_V1_QUERY_LEN 24
101 #define MLD_V2_QUERY_LEN_MIN 28
103 #define IPV6_MLD_MAX_MSF 64
105 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
106 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
109 * socket join on multicast group
111 #define mc_dereference(e, idev) \
112 rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock))
114 #define sock_dereference(e, sk) \
115 rcu_dereference_protected(e, lockdep_sock_is_held(sk))
117 #define for_each_pmc_socklock(np, sk, pmc) \
118 for (pmc = sock_dereference((np)->ipv6_mc_list, sk); \
119 pmc; \
120 pmc = sock_dereference(pmc->next, sk))
122 #define for_each_pmc_rcu(np, pmc) \
123 for (pmc = rcu_dereference((np)->ipv6_mc_list); \
124 pmc; \
125 pmc = rcu_dereference(pmc->next))
127 #define for_each_psf_mclock(mc, psf) \
128 for (psf = mc_dereference((mc)->mca_sources, mc->idev); \
129 psf; \
130 psf = mc_dereference(psf->sf_next, mc->idev))
132 #define for_each_psf_rcu(mc, psf) \
133 for (psf = rcu_dereference((mc)->mca_sources); \
134 psf; \
135 psf = rcu_dereference(psf->sf_next))
137 #define for_each_psf_tomb(mc, psf) \
138 for (psf = mc_dereference((mc)->mca_tomb, mc->idev); \
139 psf; \
140 psf = mc_dereference(psf->sf_next, mc->idev))
142 #define for_each_mc_mclock(idev, mc) \
143 for (mc = mc_dereference((idev)->mc_list, idev); \
144 mc; \
145 mc = mc_dereference(mc->next, idev))
147 #define for_each_mc_rcu(idev, mc) \
148 for (mc = rcu_dereference((idev)->mc_list); \
149 mc; \
150 mc = rcu_dereference(mc->next))
152 #define for_each_mc_tomb(idev, mc) \
153 for (mc = mc_dereference((idev)->mc_tomb, idev); \
154 mc; \
155 mc = mc_dereference(mc->next, idev))
157 static int unsolicited_report_interval(struct inet6_dev *idev)
159 int iv;
161 if (mld_in_v1_mode(idev))
162 iv = READ_ONCE(idev->cnf.mldv1_unsolicited_report_interval);
163 else
164 iv = READ_ONCE(idev->cnf.mldv2_unsolicited_report_interval);
166 return iv > 0 ? iv : 1;
169 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
170 const struct in6_addr *addr, unsigned int mode)
172 struct net_device *dev = NULL;
173 struct ipv6_mc_socklist *mc_lst;
174 struct ipv6_pinfo *np = inet6_sk(sk);
175 struct net *net = sock_net(sk);
176 int err;
178 ASSERT_RTNL();
180 if (!ipv6_addr_is_multicast(addr))
181 return -EINVAL;
183 for_each_pmc_socklock(np, sk, mc_lst) {
184 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
185 ipv6_addr_equal(&mc_lst->addr, addr))
186 return -EADDRINUSE;
189 mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
191 if (!mc_lst)
192 return -ENOMEM;
194 mc_lst->next = NULL;
195 mc_lst->addr = *addr;
197 if (ifindex == 0) {
198 struct rt6_info *rt;
199 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
200 if (rt) {
201 dev = rt->dst.dev;
202 ip6_rt_put(rt);
204 } else
205 dev = __dev_get_by_index(net, ifindex);
207 if (!dev) {
208 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
209 return -ENODEV;
212 mc_lst->ifindex = dev->ifindex;
213 mc_lst->sfmode = mode;
214 RCU_INIT_POINTER(mc_lst->sflist, NULL);
217 * now add/increase the group membership on the device
220 err = __ipv6_dev_mc_inc(dev, addr, mode);
222 if (err) {
223 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
224 return err;
227 mc_lst->next = np->ipv6_mc_list;
228 rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
230 return 0;
233 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
235 return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
237 EXPORT_SYMBOL(ipv6_sock_mc_join);
239 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
240 const struct in6_addr *addr, unsigned int mode)
242 return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
246 * socket leave on multicast group
248 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
250 struct ipv6_pinfo *np = inet6_sk(sk);
251 struct ipv6_mc_socklist *mc_lst;
252 struct ipv6_mc_socklist __rcu **lnk;
253 struct net *net = sock_net(sk);
255 ASSERT_RTNL();
257 if (!ipv6_addr_is_multicast(addr))
258 return -EINVAL;
260 for (lnk = &np->ipv6_mc_list;
261 (mc_lst = sock_dereference(*lnk, sk)) != NULL;
262 lnk = &mc_lst->next) {
263 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
264 ipv6_addr_equal(&mc_lst->addr, addr)) {
265 struct net_device *dev;
267 *lnk = mc_lst->next;
269 dev = __dev_get_by_index(net, mc_lst->ifindex);
270 if (dev) {
271 struct inet6_dev *idev = __in6_dev_get(dev);
273 ip6_mc_leave_src(sk, mc_lst, idev);
274 if (idev)
275 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
276 } else {
277 ip6_mc_leave_src(sk, mc_lst, NULL);
280 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
281 kfree_rcu(mc_lst, rcu);
282 return 0;
286 return -EADDRNOTAVAIL;
288 EXPORT_SYMBOL(ipv6_sock_mc_drop);
290 static struct inet6_dev *ip6_mc_find_dev_rtnl(struct net *net,
291 const struct in6_addr *group,
292 int ifindex)
294 struct net_device *dev = NULL;
295 struct inet6_dev *idev = NULL;
297 if (ifindex == 0) {
298 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
300 if (rt) {
301 dev = rt->dst.dev;
302 ip6_rt_put(rt);
304 } else {
305 dev = __dev_get_by_index(net, ifindex);
308 if (!dev)
309 return NULL;
310 idev = __in6_dev_get(dev);
311 if (!idev)
312 return NULL;
313 if (idev->dead)
314 return NULL;
315 return idev;
318 void __ipv6_sock_mc_close(struct sock *sk)
320 struct ipv6_pinfo *np = inet6_sk(sk);
321 struct ipv6_mc_socklist *mc_lst;
322 struct net *net = sock_net(sk);
324 ASSERT_RTNL();
326 while ((mc_lst = sock_dereference(np->ipv6_mc_list, sk)) != NULL) {
327 struct net_device *dev;
329 np->ipv6_mc_list = mc_lst->next;
331 dev = __dev_get_by_index(net, mc_lst->ifindex);
332 if (dev) {
333 struct inet6_dev *idev = __in6_dev_get(dev);
335 ip6_mc_leave_src(sk, mc_lst, idev);
336 if (idev)
337 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
338 } else {
339 ip6_mc_leave_src(sk, mc_lst, NULL);
342 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
343 kfree_rcu(mc_lst, rcu);
347 void ipv6_sock_mc_close(struct sock *sk)
349 struct ipv6_pinfo *np = inet6_sk(sk);
351 if (!rcu_access_pointer(np->ipv6_mc_list))
352 return;
354 rtnl_lock();
355 lock_sock(sk);
356 __ipv6_sock_mc_close(sk);
357 release_sock(sk);
358 rtnl_unlock();
361 int ip6_mc_source(int add, int omode, struct sock *sk,
362 struct group_source_req *pgsr)
364 struct in6_addr *source, *group;
365 struct ipv6_mc_socklist *pmc;
366 struct inet6_dev *idev;
367 struct ipv6_pinfo *inet6 = inet6_sk(sk);
368 struct ip6_sf_socklist *psl;
369 struct net *net = sock_net(sk);
370 int i, j, rv;
371 int leavegroup = 0;
372 int err;
374 source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
375 group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
377 if (!ipv6_addr_is_multicast(group))
378 return -EINVAL;
380 idev = ip6_mc_find_dev_rtnl(net, group, pgsr->gsr_interface);
381 if (!idev)
382 return -ENODEV;
384 err = -EADDRNOTAVAIL;
386 mutex_lock(&idev->mc_lock);
387 for_each_pmc_socklock(inet6, sk, pmc) {
388 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
389 continue;
390 if (ipv6_addr_equal(&pmc->addr, group))
391 break;
393 if (!pmc) { /* must have a prior join */
394 err = -EINVAL;
395 goto done;
397 /* if a source filter was set, must be the same mode as before */
398 if (rcu_access_pointer(pmc->sflist)) {
399 if (pmc->sfmode != omode) {
400 err = -EINVAL;
401 goto done;
403 } else if (pmc->sfmode != omode) {
404 /* allow mode switches for empty-set filters */
405 ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
406 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
407 pmc->sfmode = omode;
410 psl = sock_dereference(pmc->sflist, sk);
411 if (!add) {
412 if (!psl)
413 goto done; /* err = -EADDRNOTAVAIL */
414 rv = !0;
415 for (i = 0; i < psl->sl_count; i++) {
416 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
417 if (rv == 0)
418 break;
420 if (rv) /* source not found */
421 goto done; /* err = -EADDRNOTAVAIL */
423 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
424 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
425 leavegroup = 1;
426 goto done;
429 /* update the interface filter */
430 ip6_mc_del_src(idev, group, omode, 1, source, 1);
432 for (j = i+1; j < psl->sl_count; j++)
433 psl->sl_addr[j-1] = psl->sl_addr[j];
434 psl->sl_count--;
435 err = 0;
436 goto done;
438 /* else, add a new source to the filter */
440 if (psl && psl->sl_count >= sysctl_mld_max_msf) {
441 err = -ENOBUFS;
442 goto done;
444 if (!psl || psl->sl_count == psl->sl_max) {
445 struct ip6_sf_socklist *newpsl;
446 int count = IP6_SFBLOCK;
448 if (psl)
449 count += psl->sl_max;
450 newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, count),
451 GFP_KERNEL);
452 if (!newpsl) {
453 err = -ENOBUFS;
454 goto done;
456 newpsl->sl_max = count;
457 newpsl->sl_count = count - IP6_SFBLOCK;
458 if (psl) {
459 for (i = 0; i < psl->sl_count; i++)
460 newpsl->sl_addr[i] = psl->sl_addr[i];
461 atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
462 &sk->sk_omem_alloc);
464 rcu_assign_pointer(pmc->sflist, newpsl);
465 kfree_rcu(psl, rcu);
466 psl = newpsl;
468 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
469 for (i = 0; i < psl->sl_count; i++) {
470 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
471 if (rv == 0) /* There is an error in the address. */
472 goto done;
474 for (j = psl->sl_count-1; j >= i; j--)
475 psl->sl_addr[j+1] = psl->sl_addr[j];
476 psl->sl_addr[i] = *source;
477 psl->sl_count++;
478 err = 0;
479 /* update the interface list */
480 ip6_mc_add_src(idev, group, omode, 1, source, 1);
481 done:
482 mutex_unlock(&idev->mc_lock);
483 if (leavegroup)
484 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
485 return err;
488 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
489 struct sockaddr_storage *list)
491 const struct in6_addr *group;
492 struct ipv6_mc_socklist *pmc;
493 struct inet6_dev *idev;
494 struct ipv6_pinfo *inet6 = inet6_sk(sk);
495 struct ip6_sf_socklist *newpsl, *psl;
496 struct net *net = sock_net(sk);
497 int leavegroup = 0;
498 int i, err;
500 group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
502 if (!ipv6_addr_is_multicast(group))
503 return -EINVAL;
504 if (gsf->gf_fmode != MCAST_INCLUDE &&
505 gsf->gf_fmode != MCAST_EXCLUDE)
506 return -EINVAL;
508 idev = ip6_mc_find_dev_rtnl(net, group, gsf->gf_interface);
509 if (!idev)
510 return -ENODEV;
512 err = 0;
514 if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
515 leavegroup = 1;
516 goto done;
519 for_each_pmc_socklock(inet6, sk, pmc) {
520 if (pmc->ifindex != gsf->gf_interface)
521 continue;
522 if (ipv6_addr_equal(&pmc->addr, group))
523 break;
525 if (!pmc) { /* must have a prior join */
526 err = -EINVAL;
527 goto done;
529 if (gsf->gf_numsrc) {
530 newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr,
531 gsf->gf_numsrc),
532 GFP_KERNEL);
533 if (!newpsl) {
534 err = -ENOBUFS;
535 goto done;
537 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
538 for (i = 0; i < newpsl->sl_count; ++i, ++list) {
539 struct sockaddr_in6 *psin6;
541 psin6 = (struct sockaddr_in6 *)list;
542 newpsl->sl_addr[i] = psin6->sin6_addr;
544 mutex_lock(&idev->mc_lock);
545 err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
546 newpsl->sl_count, newpsl->sl_addr, 0);
547 if (err) {
548 mutex_unlock(&idev->mc_lock);
549 sock_kfree_s(sk, newpsl, struct_size(newpsl, sl_addr,
550 newpsl->sl_max));
551 goto done;
553 mutex_unlock(&idev->mc_lock);
554 } else {
555 newpsl = NULL;
556 mutex_lock(&idev->mc_lock);
557 ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
558 mutex_unlock(&idev->mc_lock);
561 mutex_lock(&idev->mc_lock);
562 psl = sock_dereference(pmc->sflist, sk);
563 if (psl) {
564 ip6_mc_del_src(idev, group, pmc->sfmode,
565 psl->sl_count, psl->sl_addr, 0);
566 atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
567 &sk->sk_omem_alloc);
568 } else {
569 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
571 rcu_assign_pointer(pmc->sflist, newpsl);
572 mutex_unlock(&idev->mc_lock);
573 kfree_rcu(psl, rcu);
574 pmc->sfmode = gsf->gf_fmode;
575 err = 0;
576 done:
577 if (leavegroup)
578 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
579 return err;
582 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
583 sockptr_t optval, size_t ss_offset)
585 struct ipv6_pinfo *inet6 = inet6_sk(sk);
586 const struct in6_addr *group;
587 struct ipv6_mc_socklist *pmc;
588 struct ip6_sf_socklist *psl;
589 unsigned int count;
590 int i, copycount;
592 group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
594 if (!ipv6_addr_is_multicast(group))
595 return -EINVAL;
597 /* changes to the ipv6_mc_list require the socket lock and
598 * rtnl lock. We have the socket lock, so reading the list is safe.
601 for_each_pmc_socklock(inet6, sk, pmc) {
602 if (pmc->ifindex != gsf->gf_interface)
603 continue;
604 if (ipv6_addr_equal(group, &pmc->addr))
605 break;
607 if (!pmc) /* must have a prior join */
608 return -EADDRNOTAVAIL;
610 gsf->gf_fmode = pmc->sfmode;
611 psl = sock_dereference(pmc->sflist, sk);
612 count = psl ? psl->sl_count : 0;
614 copycount = min(count, gsf->gf_numsrc);
615 gsf->gf_numsrc = count;
616 for (i = 0; i < copycount; i++) {
617 struct sockaddr_in6 *psin6;
618 struct sockaddr_storage ss;
620 psin6 = (struct sockaddr_in6 *)&ss;
621 memset(&ss, 0, sizeof(ss));
622 psin6->sin6_family = AF_INET6;
623 psin6->sin6_addr = psl->sl_addr[i];
624 if (copy_to_sockptr_offset(optval, ss_offset, &ss, sizeof(ss)))
625 return -EFAULT;
626 ss_offset += sizeof(ss);
628 return 0;
631 bool inet6_mc_check(const struct sock *sk, const struct in6_addr *mc_addr,
632 const struct in6_addr *src_addr)
634 const struct ipv6_pinfo *np = inet6_sk(sk);
635 const struct ipv6_mc_socklist *mc;
636 const struct ip6_sf_socklist *psl;
637 bool rv = true;
639 rcu_read_lock();
640 for_each_pmc_rcu(np, mc) {
641 if (ipv6_addr_equal(&mc->addr, mc_addr))
642 break;
644 if (!mc) {
645 rcu_read_unlock();
646 return inet6_test_bit(MC6_ALL, sk);
648 psl = rcu_dereference(mc->sflist);
649 if (!psl) {
650 rv = mc->sfmode == MCAST_EXCLUDE;
651 } else {
652 int i;
654 for (i = 0; i < psl->sl_count; i++) {
655 if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
656 break;
658 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
659 rv = false;
660 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
661 rv = false;
663 rcu_read_unlock();
665 return rv;
668 /* called with mc_lock */
669 static void igmp6_group_added(struct ifmcaddr6 *mc)
671 struct net_device *dev = mc->idev->dev;
672 char buf[MAX_ADDR_LEN];
674 if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
675 IPV6_ADDR_SCOPE_LINKLOCAL)
676 return;
678 if (!(mc->mca_flags&MAF_LOADED)) {
679 mc->mca_flags |= MAF_LOADED;
680 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
681 dev_mc_add(dev, buf);
684 if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
685 return;
687 if (mld_in_v1_mode(mc->idev)) {
688 igmp6_join_group(mc);
689 return;
691 /* else v2 */
693 /* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
694 * should not send filter-mode change record as the mode
695 * should be from IN() to IN(A).
697 if (mc->mca_sfmode == MCAST_EXCLUDE)
698 mc->mca_crcount = mc->idev->mc_qrv;
700 mld_ifc_event(mc->idev);
703 /* called with mc_lock */
704 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
706 struct net_device *dev = mc->idev->dev;
707 char buf[MAX_ADDR_LEN];
709 if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
710 IPV6_ADDR_SCOPE_LINKLOCAL)
711 return;
713 if (mc->mca_flags&MAF_LOADED) {
714 mc->mca_flags &= ~MAF_LOADED;
715 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
716 dev_mc_del(dev, buf);
719 if (mc->mca_flags & MAF_NOREPORT)
720 return;
722 if (!mc->idev->dead)
723 igmp6_leave_group(mc);
725 if (cancel_delayed_work(&mc->mca_work))
726 refcount_dec(&mc->mca_refcnt);
730 * deleted ifmcaddr6 manipulation
731 * called with mc_lock
733 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
735 struct ifmcaddr6 *pmc;
737 /* this is an "ifmcaddr6" for convenience; only the fields below
738 * are actually used. In particular, the refcnt and users are not
739 * used for management of the delete list. Using the same structure
740 * for deleted items allows change reports to use common code with
741 * non-deleted or query-response MCA's.
743 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
744 if (!pmc)
745 return;
747 pmc->idev = im->idev;
748 in6_dev_hold(idev);
749 pmc->mca_addr = im->mca_addr;
750 pmc->mca_crcount = idev->mc_qrv;
751 pmc->mca_sfmode = im->mca_sfmode;
752 if (pmc->mca_sfmode == MCAST_INCLUDE) {
753 struct ip6_sf_list *psf;
755 rcu_assign_pointer(pmc->mca_tomb,
756 mc_dereference(im->mca_tomb, idev));
757 rcu_assign_pointer(pmc->mca_sources,
758 mc_dereference(im->mca_sources, idev));
759 RCU_INIT_POINTER(im->mca_tomb, NULL);
760 RCU_INIT_POINTER(im->mca_sources, NULL);
762 for_each_psf_mclock(pmc, psf)
763 psf->sf_crcount = pmc->mca_crcount;
766 rcu_assign_pointer(pmc->next, idev->mc_tomb);
767 rcu_assign_pointer(idev->mc_tomb, pmc);
770 /* called with mc_lock */
771 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
773 struct ip6_sf_list *psf, *sources, *tomb;
774 struct in6_addr *pmca = &im->mca_addr;
775 struct ifmcaddr6 *pmc, *pmc_prev;
777 pmc_prev = NULL;
778 for_each_mc_tomb(idev, pmc) {
779 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
780 break;
781 pmc_prev = pmc;
783 if (pmc) {
784 if (pmc_prev)
785 rcu_assign_pointer(pmc_prev->next, pmc->next);
786 else
787 rcu_assign_pointer(idev->mc_tomb, pmc->next);
790 if (pmc) {
791 im->idev = pmc->idev;
792 if (im->mca_sfmode == MCAST_INCLUDE) {
793 tomb = rcu_replace_pointer(im->mca_tomb,
794 mc_dereference(pmc->mca_tomb, pmc->idev),
795 lockdep_is_held(&im->idev->mc_lock));
796 rcu_assign_pointer(pmc->mca_tomb, tomb);
798 sources = rcu_replace_pointer(im->mca_sources,
799 mc_dereference(pmc->mca_sources, pmc->idev),
800 lockdep_is_held(&im->idev->mc_lock));
801 rcu_assign_pointer(pmc->mca_sources, sources);
802 for_each_psf_mclock(im, psf)
803 psf->sf_crcount = idev->mc_qrv;
804 } else {
805 im->mca_crcount = idev->mc_qrv;
807 in6_dev_put(pmc->idev);
808 ip6_mc_clear_src(pmc);
809 kfree_rcu(pmc, rcu);
813 /* called with mc_lock */
814 static void mld_clear_delrec(struct inet6_dev *idev)
816 struct ifmcaddr6 *pmc, *nextpmc;
818 pmc = mc_dereference(idev->mc_tomb, idev);
819 RCU_INIT_POINTER(idev->mc_tomb, NULL);
821 for (; pmc; pmc = nextpmc) {
822 nextpmc = mc_dereference(pmc->next, idev);
823 ip6_mc_clear_src(pmc);
824 in6_dev_put(pmc->idev);
825 kfree_rcu(pmc, rcu);
828 /* clear dead sources, too */
829 for_each_mc_mclock(idev, pmc) {
830 struct ip6_sf_list *psf, *psf_next;
832 psf = mc_dereference(pmc->mca_tomb, idev);
833 RCU_INIT_POINTER(pmc->mca_tomb, NULL);
834 for (; psf; psf = psf_next) {
835 psf_next = mc_dereference(psf->sf_next, idev);
836 kfree_rcu(psf, rcu);
841 static void mld_clear_query(struct inet6_dev *idev)
843 struct sk_buff *skb;
845 spin_lock_bh(&idev->mc_query_lock);
846 while ((skb = __skb_dequeue(&idev->mc_query_queue)))
847 kfree_skb(skb);
848 spin_unlock_bh(&idev->mc_query_lock);
851 static void mld_clear_report(struct inet6_dev *idev)
853 struct sk_buff *skb;
855 spin_lock_bh(&idev->mc_report_lock);
856 while ((skb = __skb_dequeue(&idev->mc_report_queue)))
857 kfree_skb(skb);
858 spin_unlock_bh(&idev->mc_report_lock);
861 static void mca_get(struct ifmcaddr6 *mc)
863 refcount_inc(&mc->mca_refcnt);
866 static void ma_put(struct ifmcaddr6 *mc)
868 if (refcount_dec_and_test(&mc->mca_refcnt)) {
869 in6_dev_put(mc->idev);
870 kfree_rcu(mc, rcu);
874 /* called with mc_lock */
875 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
876 const struct in6_addr *addr,
877 unsigned int mode)
879 struct ifmcaddr6 *mc;
881 mc = kzalloc(sizeof(*mc), GFP_KERNEL);
882 if (!mc)
883 return NULL;
885 INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);
887 mc->mca_addr = *addr;
888 mc->idev = idev; /* reference taken by caller */
889 mc->mca_users = 1;
890 /* mca_stamp should be updated upon changes */
891 mc->mca_cstamp = mc->mca_tstamp = jiffies;
892 refcount_set(&mc->mca_refcnt, 1);
894 mc->mca_sfmode = mode;
895 mc->mca_sfcount[mode] = 1;
897 if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
898 IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
899 mc->mca_flags |= MAF_NOREPORT;
901 return mc;
905 * device multicast group inc (add if not found)
907 static int __ipv6_dev_mc_inc(struct net_device *dev,
908 const struct in6_addr *addr, unsigned int mode)
910 struct ifmcaddr6 *mc;
911 struct inet6_dev *idev;
913 ASSERT_RTNL();
915 /* we need to take a reference on idev */
916 idev = in6_dev_get(dev);
918 if (!idev)
919 return -EINVAL;
921 if (idev->dead) {
922 in6_dev_put(idev);
923 return -ENODEV;
926 mutex_lock(&idev->mc_lock);
927 for_each_mc_mclock(idev, mc) {
928 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
929 mc->mca_users++;
930 ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
931 mutex_unlock(&idev->mc_lock);
932 in6_dev_put(idev);
933 return 0;
937 mc = mca_alloc(idev, addr, mode);
938 if (!mc) {
939 mutex_unlock(&idev->mc_lock);
940 in6_dev_put(idev);
941 return -ENOMEM;
944 rcu_assign_pointer(mc->next, idev->mc_list);
945 rcu_assign_pointer(idev->mc_list, mc);
947 mca_get(mc);
949 mld_del_delrec(idev, mc);
950 igmp6_group_added(mc);
951 mutex_unlock(&idev->mc_lock);
952 ma_put(mc);
953 return 0;
956 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
958 return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
960 EXPORT_SYMBOL(ipv6_dev_mc_inc);
963 * device multicast group del
965 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
967 struct ifmcaddr6 *ma, __rcu **map;
969 ASSERT_RTNL();
971 mutex_lock(&idev->mc_lock);
972 for (map = &idev->mc_list;
973 (ma = mc_dereference(*map, idev));
974 map = &ma->next) {
975 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
976 if (--ma->mca_users == 0) {
977 *map = ma->next;
979 igmp6_group_dropped(ma);
980 ip6_mc_clear_src(ma);
981 mutex_unlock(&idev->mc_lock);
983 ma_put(ma);
984 return 0;
986 mutex_unlock(&idev->mc_lock);
987 return 0;
991 mutex_unlock(&idev->mc_lock);
992 return -ENOENT;
995 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
997 struct inet6_dev *idev;
998 int err;
1000 ASSERT_RTNL();
1002 idev = __in6_dev_get(dev);
1003 if (!idev)
1004 err = -ENODEV;
1005 else
1006 err = __ipv6_dev_mc_dec(idev, addr);
1008 return err;
1010 EXPORT_SYMBOL(ipv6_dev_mc_dec);
1013 * check if the interface/address pair is valid
1015 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
1016 const struct in6_addr *src_addr)
1018 struct inet6_dev *idev;
1019 struct ifmcaddr6 *mc;
1020 bool rv = false;
1022 rcu_read_lock();
1023 idev = __in6_dev_get(dev);
1024 if (idev) {
1025 for_each_mc_rcu(idev, mc) {
1026 if (ipv6_addr_equal(&mc->mca_addr, group))
1027 break;
1029 if (mc) {
1030 if (src_addr && !ipv6_addr_any(src_addr)) {
1031 struct ip6_sf_list *psf;
1033 for_each_psf_rcu(mc, psf) {
1034 if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1035 break;
1037 if (psf)
1038 rv = psf->sf_count[MCAST_INCLUDE] ||
1039 psf->sf_count[MCAST_EXCLUDE] !=
1040 mc->mca_sfcount[MCAST_EXCLUDE];
1041 else
1042 rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1043 } else
1044 rv = true; /* don't filter unspecified source */
1047 rcu_read_unlock();
1048 return rv;
1051 /* called with mc_lock */
1052 static void mld_gq_start_work(struct inet6_dev *idev)
1054 unsigned long tv = get_random_u32_below(idev->mc_maxdelay);
1056 idev->mc_gq_running = 1;
1057 if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2))
1058 in6_dev_hold(idev);
1061 /* called with mc_lock */
1062 static void mld_gq_stop_work(struct inet6_dev *idev)
1064 idev->mc_gq_running = 0;
1065 if (cancel_delayed_work(&idev->mc_gq_work))
1066 __in6_dev_put(idev);
1069 /* called with mc_lock */
1070 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
1072 unsigned long tv = get_random_u32_below(delay);
1074 if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2))
1075 in6_dev_hold(idev);
1078 /* called with mc_lock */
1079 static void mld_ifc_stop_work(struct inet6_dev *idev)
1081 idev->mc_ifc_count = 0;
1082 if (cancel_delayed_work(&idev->mc_ifc_work))
1083 __in6_dev_put(idev);
1086 /* called with mc_lock */
1087 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
1089 unsigned long tv = get_random_u32_below(delay);
1091 if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2))
1092 in6_dev_hold(idev);
1095 static void mld_dad_stop_work(struct inet6_dev *idev)
1097 if (cancel_delayed_work(&idev->mc_dad_work))
1098 __in6_dev_put(idev);
1101 static void mld_query_stop_work(struct inet6_dev *idev)
1103 spin_lock_bh(&idev->mc_query_lock);
1104 if (cancel_delayed_work(&idev->mc_query_work))
1105 __in6_dev_put(idev);
1106 spin_unlock_bh(&idev->mc_query_lock);
1109 static void mld_report_stop_work(struct inet6_dev *idev)
1111 if (cancel_delayed_work_sync(&idev->mc_report_work))
1112 __in6_dev_put(idev);
1116 * IGMP handling (alias multicast ICMPv6 messages)
1117 * called with mc_lock
1119 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1121 unsigned long delay = resptime;
1123 /* Do not start work for these addresses */
1124 if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1125 IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1126 return;
1128 if (cancel_delayed_work(&ma->mca_work)) {
1129 refcount_dec(&ma->mca_refcnt);
1130 delay = ma->mca_work.timer.expires - jiffies;
1133 if (delay >= resptime)
1134 delay = get_random_u32_below(resptime);
1136 if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
1137 refcount_inc(&ma->mca_refcnt);
1138 ma->mca_flags |= MAF_TIMER_RUNNING;
1141 /* mark EXCLUDE-mode sources
1142 * called with mc_lock
1144 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1145 const struct in6_addr *srcs)
1147 struct ip6_sf_list *psf;
1148 int i, scount;
1150 scount = 0;
1151 for_each_psf_mclock(pmc, psf) {
1152 if (scount == nsrcs)
1153 break;
1154 for (i = 0; i < nsrcs; i++) {
1155 /* skip inactive filters */
1156 if (psf->sf_count[MCAST_INCLUDE] ||
1157 pmc->mca_sfcount[MCAST_EXCLUDE] !=
1158 psf->sf_count[MCAST_EXCLUDE])
1159 break;
1160 if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1161 scount++;
1162 break;
1166 pmc->mca_flags &= ~MAF_GSQUERY;
1167 if (scount == nsrcs) /* all sources excluded */
1168 return false;
1169 return true;
1172 /* called with mc_lock */
1173 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1174 const struct in6_addr *srcs)
1176 struct ip6_sf_list *psf;
1177 int i, scount;
1179 if (pmc->mca_sfmode == MCAST_EXCLUDE)
1180 return mld_xmarksources(pmc, nsrcs, srcs);
1182 /* mark INCLUDE-mode sources */
1184 scount = 0;
1185 for_each_psf_mclock(pmc, psf) {
1186 if (scount == nsrcs)
1187 break;
1188 for (i = 0; i < nsrcs; i++) {
1189 if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1190 psf->sf_gsresp = 1;
1191 scount++;
1192 break;
1196 if (!scount) {
1197 pmc->mca_flags &= ~MAF_GSQUERY;
1198 return false;
1200 pmc->mca_flags |= MAF_GSQUERY;
1201 return true;
1204 static int mld_force_mld_version(const struct inet6_dev *idev)
1206 const struct net *net = dev_net(idev->dev);
1207 int all_force;
1209 all_force = READ_ONCE(net->ipv6.devconf_all->force_mld_version);
1210 /* Normally, both are 0 here. If enforcement to a particular is
1211 * being used, individual device enforcement will have a lower
1212 * precedence over 'all' device (.../conf/all/force_mld_version).
1214 return all_force ?: READ_ONCE(idev->cnf.force_mld_version);
1217 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1219 return mld_force_mld_version(idev) == 2;
1222 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1224 return mld_force_mld_version(idev) == 1;
1227 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1229 if (mld_in_v2_mode_only(idev))
1230 return false;
1231 if (mld_in_v1_mode_only(idev))
1232 return true;
1233 if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1234 return true;
1236 return false;
1239 static void mld_set_v1_mode(struct inet6_dev *idev)
1241 /* RFC3810, relevant sections:
1242 * - 9.1. Robustness Variable
1243 * - 9.2. Query Interval
1244 * - 9.3. Query Response Interval
1245 * - 9.12. Older Version Querier Present Timeout
1247 unsigned long switchback;
1249 switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1251 idev->mc_v1_seen = jiffies + switchback;
1254 static void mld_update_qrv(struct inet6_dev *idev,
1255 const struct mld2_query *mlh2)
1257 /* RFC3810, relevant sections:
1258 * - 5.1.8. QRV (Querier's Robustness Variable)
1259 * - 9.1. Robustness Variable
1262 /* The value of the Robustness Variable MUST NOT be zero,
1263 * and SHOULD NOT be one. Catch this here if we ever run
1264 * into such a case in future.
1266 const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1267 WARN_ON(idev->mc_qrv == 0);
1269 if (mlh2->mld2q_qrv > 0)
1270 idev->mc_qrv = mlh2->mld2q_qrv;
1272 if (unlikely(idev->mc_qrv < min_qrv)) {
1273 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1274 idev->mc_qrv, min_qrv);
1275 idev->mc_qrv = min_qrv;
1279 static void mld_update_qi(struct inet6_dev *idev,
1280 const struct mld2_query *mlh2)
1282 /* RFC3810, relevant sections:
1283 * - 5.1.9. QQIC (Querier's Query Interval Code)
1284 * - 9.2. Query Interval
1285 * - 9.12. Older Version Querier Present Timeout
1286 * (the [Query Interval] in the last Query received)
1288 unsigned long mc_qqi;
1290 if (mlh2->mld2q_qqic < 128) {
1291 mc_qqi = mlh2->mld2q_qqic;
1292 } else {
1293 unsigned long mc_man, mc_exp;
1295 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1296 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1298 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1301 idev->mc_qi = mc_qqi * HZ;
1304 static void mld_update_qri(struct inet6_dev *idev,
1305 const struct mld2_query *mlh2)
1307 /* RFC3810, relevant sections:
1308 * - 5.1.3. Maximum Response Code
1309 * - 9.3. Query Response Interval
1311 idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1314 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1315 unsigned long *max_delay, bool v1_query)
1317 unsigned long mldv1_md;
1319 /* Ignore v1 queries */
1320 if (mld_in_v2_mode_only(idev))
1321 return -EINVAL;
1323 mldv1_md = ntohs(mld->mld_maxdelay);
1325 /* When in MLDv1 fallback and a MLDv2 router start-up being
1326 * unaware of current MLDv1 operation, the MRC == MRD mapping
1327 * only works when the exponential algorithm is not being
1328 * used (as MLDv1 is unaware of such things).
1330 * According to the RFC author, the MLDv2 implementations
1331 * he's aware of all use a MRC < 32768 on start up queries.
1333 * Thus, should we *ever* encounter something else larger
1334 * than that, just assume the maximum possible within our
1335 * reach.
1337 if (!v1_query)
1338 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1340 *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1342 /* MLDv1 router present: we need to go into v1 mode *only*
1343 * when an MLDv1 query is received as per section 9.12. of
1344 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1345 * queries MUST be of exactly 24 octets.
1347 if (v1_query)
1348 mld_set_v1_mode(idev);
1350 /* cancel MLDv2 report work */
1351 mld_gq_stop_work(idev);
1352 /* cancel the interface change work */
1353 mld_ifc_stop_work(idev);
1354 /* clear deleted report items */
1355 mld_clear_delrec(idev);
1357 return 0;
1360 static void mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1361 unsigned long *max_delay)
1363 *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1365 mld_update_qrv(idev, mld);
1366 mld_update_qi(idev, mld);
1367 mld_update_qri(idev, mld);
1369 idev->mc_maxdelay = *max_delay;
1371 return;
1374 /* called with rcu_read_lock() */
1375 void igmp6_event_query(struct sk_buff *skb)
1377 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1379 if (!idev || idev->dead)
1380 goto out;
1382 spin_lock_bh(&idev->mc_query_lock);
1383 if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) {
1384 __skb_queue_tail(&idev->mc_query_queue, skb);
1385 if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0))
1386 in6_dev_hold(idev);
1387 skb = NULL;
1389 spin_unlock_bh(&idev->mc_query_lock);
1390 out:
1391 kfree_skb(skb);
1394 static void __mld_query_work(struct sk_buff *skb)
1396 struct mld2_query *mlh2 = NULL;
1397 const struct in6_addr *group;
1398 unsigned long max_delay;
1399 struct inet6_dev *idev;
1400 struct ifmcaddr6 *ma;
1401 struct mld_msg *mld;
1402 int group_type;
1403 int mark = 0;
1404 int len, err;
1406 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1407 goto kfree_skb;
1409 /* compute payload length excluding extension headers */
1410 len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1411 len -= skb_network_header_len(skb);
1413 /* RFC3810 6.2
1414 * Upon reception of an MLD message that contains a Query, the node
1415 * checks if the source address of the message is a valid link-local
1416 * address, if the Hop Limit is set to 1, and if the Router Alert
1417 * option is present in the Hop-By-Hop Options header of the IPv6
1418 * packet. If any of these checks fails, the packet is dropped.
1420 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1421 ipv6_hdr(skb)->hop_limit != 1 ||
1422 !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1423 IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1424 goto kfree_skb;
1426 idev = in6_dev_get(skb->dev);
1427 if (!idev)
1428 goto kfree_skb;
1430 mld = (struct mld_msg *)icmp6_hdr(skb);
1431 group = &mld->mld_mca;
1432 group_type = ipv6_addr_type(group);
1434 if (group_type != IPV6_ADDR_ANY &&
1435 !(group_type&IPV6_ADDR_MULTICAST))
1436 goto out;
1438 if (len < MLD_V1_QUERY_LEN) {
1439 goto out;
1440 } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1441 err = mld_process_v1(idev, mld, &max_delay,
1442 len == MLD_V1_QUERY_LEN);
1443 if (err < 0)
1444 goto out;
1445 } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1446 int srcs_offset = sizeof(struct mld2_query) -
1447 sizeof(struct icmp6hdr);
1449 if (!pskb_may_pull(skb, srcs_offset))
1450 goto out;
1452 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1454 mld_process_v2(idev, mlh2, &max_delay);
1456 if (group_type == IPV6_ADDR_ANY) { /* general query */
1457 if (mlh2->mld2q_nsrcs)
1458 goto out; /* no sources allowed */
1460 mld_gq_start_work(idev);
1461 goto out;
1463 /* mark sources to include, if group & source-specific */
1464 if (mlh2->mld2q_nsrcs != 0) {
1465 if (!pskb_may_pull(skb, srcs_offset +
1466 ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1467 goto out;
1469 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1470 mark = 1;
1472 } else {
1473 goto out;
1476 if (group_type == IPV6_ADDR_ANY) {
1477 for_each_mc_mclock(idev, ma) {
1478 igmp6_group_queried(ma, max_delay);
1480 } else {
1481 for_each_mc_mclock(idev, ma) {
1482 if (!ipv6_addr_equal(group, &ma->mca_addr))
1483 continue;
1484 if (ma->mca_flags & MAF_TIMER_RUNNING) {
1485 /* gsquery <- gsquery && mark */
1486 if (!mark)
1487 ma->mca_flags &= ~MAF_GSQUERY;
1488 } else {
1489 /* gsquery <- mark */
1490 if (mark)
1491 ma->mca_flags |= MAF_GSQUERY;
1492 else
1493 ma->mca_flags &= ~MAF_GSQUERY;
1495 if (!(ma->mca_flags & MAF_GSQUERY) ||
1496 mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1497 igmp6_group_queried(ma, max_delay);
1498 break;
1502 out:
1503 in6_dev_put(idev);
1504 kfree_skb:
1505 consume_skb(skb);
1508 static void mld_query_work(struct work_struct *work)
1510 struct inet6_dev *idev = container_of(to_delayed_work(work),
1511 struct inet6_dev,
1512 mc_query_work);
1513 struct sk_buff_head q;
1514 struct sk_buff *skb;
1515 bool rework = false;
1516 int cnt = 0;
1518 skb_queue_head_init(&q);
1520 spin_lock_bh(&idev->mc_query_lock);
1521 while ((skb = __skb_dequeue(&idev->mc_query_queue))) {
1522 __skb_queue_tail(&q, skb);
1524 if (++cnt >= MLD_MAX_QUEUE) {
1525 rework = true;
1526 break;
1529 spin_unlock_bh(&idev->mc_query_lock);
1531 mutex_lock(&idev->mc_lock);
1532 while ((skb = __skb_dequeue(&q)))
1533 __mld_query_work(skb);
1534 mutex_unlock(&idev->mc_lock);
1536 if (rework && queue_delayed_work(mld_wq, &idev->mc_query_work, 0))
1537 return;
1539 in6_dev_put(idev);
1542 /* called with rcu_read_lock() */
1543 void igmp6_event_report(struct sk_buff *skb)
1545 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1547 if (!idev || idev->dead)
1548 goto out;
1550 spin_lock_bh(&idev->mc_report_lock);
1551 if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) {
1552 __skb_queue_tail(&idev->mc_report_queue, skb);
1553 if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0))
1554 in6_dev_hold(idev);
1555 skb = NULL;
1557 spin_unlock_bh(&idev->mc_report_lock);
1558 out:
1559 kfree_skb(skb);
1562 static void __mld_report_work(struct sk_buff *skb)
1564 struct inet6_dev *idev;
1565 struct ifmcaddr6 *ma;
1566 struct mld_msg *mld;
1567 int addr_type;
1569 /* Our own report looped back. Ignore it. */
1570 if (skb->pkt_type == PACKET_LOOPBACK)
1571 goto kfree_skb;
1573 /* send our report if the MC router may not have heard this report */
1574 if (skb->pkt_type != PACKET_MULTICAST &&
1575 skb->pkt_type != PACKET_BROADCAST)
1576 goto kfree_skb;
1578 if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1579 goto kfree_skb;
1581 mld = (struct mld_msg *)icmp6_hdr(skb);
1583 /* Drop reports with not link local source */
1584 addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1585 if (addr_type != IPV6_ADDR_ANY &&
1586 !(addr_type&IPV6_ADDR_LINKLOCAL))
1587 goto kfree_skb;
1589 idev = in6_dev_get(skb->dev);
1590 if (!idev)
1591 goto kfree_skb;
1594 * Cancel the work for this group
1597 for_each_mc_mclock(idev, ma) {
1598 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1599 if (cancel_delayed_work(&ma->mca_work))
1600 refcount_dec(&ma->mca_refcnt);
1601 ma->mca_flags &= ~(MAF_LAST_REPORTER |
1602 MAF_TIMER_RUNNING);
1603 break;
1607 in6_dev_put(idev);
1608 kfree_skb:
1609 consume_skb(skb);
1612 static void mld_report_work(struct work_struct *work)
1614 struct inet6_dev *idev = container_of(to_delayed_work(work),
1615 struct inet6_dev,
1616 mc_report_work);
1617 struct sk_buff_head q;
1618 struct sk_buff *skb;
1619 bool rework = false;
1620 int cnt = 0;
1622 skb_queue_head_init(&q);
1623 spin_lock_bh(&idev->mc_report_lock);
1624 while ((skb = __skb_dequeue(&idev->mc_report_queue))) {
1625 __skb_queue_tail(&q, skb);
1627 if (++cnt >= MLD_MAX_QUEUE) {
1628 rework = true;
1629 break;
1632 spin_unlock_bh(&idev->mc_report_lock);
1634 mutex_lock(&idev->mc_lock);
1635 while ((skb = __skb_dequeue(&q)))
1636 __mld_report_work(skb);
1637 mutex_unlock(&idev->mc_lock);
1639 if (rework && queue_delayed_work(mld_wq, &idev->mc_report_work, 0))
1640 return;
1642 in6_dev_put(idev);
1645 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1646 int gdeleted, int sdeleted)
1648 switch (type) {
1649 case MLD2_MODE_IS_INCLUDE:
1650 case MLD2_MODE_IS_EXCLUDE:
1651 if (gdeleted || sdeleted)
1652 return false;
1653 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1654 if (pmc->mca_sfmode == MCAST_INCLUDE)
1655 return true;
1656 /* don't include if this source is excluded
1657 * in all filters
1659 if (psf->sf_count[MCAST_INCLUDE])
1660 return type == MLD2_MODE_IS_INCLUDE;
1661 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1662 psf->sf_count[MCAST_EXCLUDE];
1664 return false;
1665 case MLD2_CHANGE_TO_INCLUDE:
1666 if (gdeleted || sdeleted)
1667 return false;
1668 return psf->sf_count[MCAST_INCLUDE] != 0;
1669 case MLD2_CHANGE_TO_EXCLUDE:
1670 if (gdeleted || sdeleted)
1671 return false;
1672 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1673 psf->sf_count[MCAST_INCLUDE])
1674 return false;
1675 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1676 psf->sf_count[MCAST_EXCLUDE];
1677 case MLD2_ALLOW_NEW_SOURCES:
1678 if (gdeleted || !psf->sf_crcount)
1679 return false;
1680 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1681 case MLD2_BLOCK_OLD_SOURCES:
1682 if (pmc->mca_sfmode == MCAST_INCLUDE)
1683 return gdeleted || (psf->sf_crcount && sdeleted);
1684 return psf->sf_crcount && !gdeleted && !sdeleted;
1686 return false;
1689 static int
1690 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1692 struct ip6_sf_list *psf;
1693 int scount = 0;
1695 for_each_psf_mclock(pmc, psf) {
1696 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1697 continue;
1698 scount++;
1700 return scount;
1703 static void ip6_mc_hdr(const struct sock *sk, struct sk_buff *skb,
1704 struct net_device *dev, const struct in6_addr *saddr,
1705 const struct in6_addr *daddr, int proto, int len)
1707 struct ipv6hdr *hdr;
1709 skb->protocol = htons(ETH_P_IPV6);
1710 skb->dev = dev;
1712 skb_reset_network_header(skb);
1713 skb_put(skb, sizeof(struct ipv6hdr));
1714 hdr = ipv6_hdr(skb);
1716 ip6_flow_hdr(hdr, 0, 0);
1718 hdr->payload_len = htons(len);
1719 hdr->nexthdr = proto;
1720 hdr->hop_limit = READ_ONCE(inet6_sk(sk)->hop_limit);
1722 hdr->saddr = *saddr;
1723 hdr->daddr = *daddr;
1726 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1728 u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
1729 2, 0, 0, IPV6_TLV_PADN, 0 };
1730 struct net_device *dev = idev->dev;
1731 int hlen = LL_RESERVED_SPACE(dev);
1732 int tlen = dev->needed_tailroom;
1733 struct net *net = dev_net(dev);
1734 const struct in6_addr *saddr;
1735 struct in6_addr addr_buf;
1736 struct mld2_report *pmr;
1737 struct sk_buff *skb;
1738 unsigned int size;
1739 struct sock *sk;
1740 int err;
1742 sk = net->ipv6.igmp_sk;
1743 /* we assume size > sizeof(ra) here
1744 * Also try to not allocate high-order pages for big MTU
1746 size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen;
1747 skb = sock_alloc_send_skb(sk, size, 1, &err);
1748 if (!skb)
1749 return NULL;
1751 skb->priority = TC_PRIO_CONTROL;
1752 skb_reserve(skb, hlen);
1753 skb_tailroom_reserve(skb, mtu, tlen);
1755 if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1756 /* <draft-ietf-magma-mld-source-05.txt>:
1757 * use unspecified address as the source address
1758 * when a valid link-local address is not available.
1760 saddr = &in6addr_any;
1761 } else
1762 saddr = &addr_buf;
1764 ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1766 skb_put_data(skb, ra, sizeof(ra));
1768 skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1769 skb_put(skb, sizeof(*pmr));
1770 pmr = (struct mld2_report *)skb_transport_header(skb);
1771 pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1772 pmr->mld2r_resv1 = 0;
1773 pmr->mld2r_cksum = 0;
1774 pmr->mld2r_resv2 = 0;
1775 pmr->mld2r_ngrec = 0;
1776 return skb;
1779 static void mld_sendpack(struct sk_buff *skb)
1781 struct ipv6hdr *pip6 = ipv6_hdr(skb);
1782 struct mld2_report *pmr =
1783 (struct mld2_report *)skb_transport_header(skb);
1784 int payload_len, mldlen;
1785 struct inet6_dev *idev;
1786 struct net *net = dev_net(skb->dev);
1787 int err;
1788 struct flowi6 fl6;
1789 struct dst_entry *dst;
1791 rcu_read_lock();
1792 idev = __in6_dev_get(skb->dev);
1793 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
1795 payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1796 sizeof(*pip6);
1797 mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1798 pip6->payload_len = htons(payload_len);
1800 pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1801 IPPROTO_ICMPV6,
1802 csum_partial(skb_transport_header(skb),
1803 mldlen, 0));
1805 icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1806 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1807 skb->dev->ifindex);
1808 dst = icmp6_dst_alloc(skb->dev, &fl6);
1810 err = 0;
1811 if (IS_ERR(dst)) {
1812 err = PTR_ERR(dst);
1813 dst = NULL;
1815 skb_dst_set(skb, dst);
1816 if (err)
1817 goto err_out;
1819 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1820 net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1821 dst_output);
1822 out:
1823 if (!err) {
1824 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1825 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1826 } else {
1827 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1830 rcu_read_unlock();
1831 return;
1833 err_out:
1834 kfree_skb(skb);
1835 goto out;
1838 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1840 return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1843 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1844 int type, struct mld2_grec **ppgr, unsigned int mtu)
1846 struct mld2_report *pmr;
1847 struct mld2_grec *pgr;
1849 if (!skb) {
1850 skb = mld_newpack(pmc->idev, mtu);
1851 if (!skb)
1852 return NULL;
1854 pgr = skb_put(skb, sizeof(struct mld2_grec));
1855 pgr->grec_type = type;
1856 pgr->grec_auxwords = 0;
1857 pgr->grec_nsrcs = 0;
1858 pgr->grec_mca = pmc->mca_addr; /* structure copy */
1859 pmr = (struct mld2_report *)skb_transport_header(skb);
1860 pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1861 *ppgr = pgr;
1862 return skb;
1865 #define AVAILABLE(skb) ((skb) ? skb_availroom(skb) : 0)
1867 /* called with mc_lock */
1868 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1869 int type, int gdeleted, int sdeleted,
1870 int crsend)
1872 struct ip6_sf_list *psf, *psf_prev, *psf_next;
1873 int scount, stotal, first, isquery, truncate;
1874 struct ip6_sf_list __rcu **psf_list;
1875 struct inet6_dev *idev = pmc->idev;
1876 struct net_device *dev = idev->dev;
1877 struct mld2_grec *pgr = NULL;
1878 struct mld2_report *pmr;
1879 unsigned int mtu;
1881 if (pmc->mca_flags & MAF_NOREPORT)
1882 return skb;
1884 mtu = READ_ONCE(dev->mtu);
1885 if (mtu < IPV6_MIN_MTU)
1886 return skb;
1888 isquery = type == MLD2_MODE_IS_INCLUDE ||
1889 type == MLD2_MODE_IS_EXCLUDE;
1890 truncate = type == MLD2_MODE_IS_EXCLUDE ||
1891 type == MLD2_CHANGE_TO_EXCLUDE;
1893 stotal = scount = 0;
1895 psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1897 if (!rcu_access_pointer(*psf_list))
1898 goto empty_source;
1900 pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1902 /* EX and TO_EX get a fresh packet, if needed */
1903 if (truncate) {
1904 if (pmr && pmr->mld2r_ngrec &&
1905 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1906 if (skb)
1907 mld_sendpack(skb);
1908 skb = mld_newpack(idev, mtu);
1911 first = 1;
1912 psf_prev = NULL;
1913 for (psf = mc_dereference(*psf_list, idev);
1914 psf;
1915 psf = psf_next) {
1916 struct in6_addr *psrc;
1918 psf_next = mc_dereference(psf->sf_next, idev);
1920 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1921 psf_prev = psf;
1922 continue;
1925 /* Based on RFC3810 6.1. Should not send source-list change
1926 * records when there is a filter mode change.
1928 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1929 (!gdeleted && pmc->mca_crcount)) &&
1930 (type == MLD2_ALLOW_NEW_SOURCES ||
1931 type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1932 goto decrease_sf_crcount;
1934 /* clear marks on query responses */
1935 if (isquery)
1936 psf->sf_gsresp = 0;
1938 if (AVAILABLE(skb) < sizeof(*psrc) +
1939 first*sizeof(struct mld2_grec)) {
1940 if (truncate && !first)
1941 break; /* truncate these */
1942 if (pgr)
1943 pgr->grec_nsrcs = htons(scount);
1944 if (skb)
1945 mld_sendpack(skb);
1946 skb = mld_newpack(idev, mtu);
1947 first = 1;
1948 scount = 0;
1950 if (first) {
1951 skb = add_grhead(skb, pmc, type, &pgr, mtu);
1952 first = 0;
1954 if (!skb)
1955 return NULL;
1956 psrc = skb_put(skb, sizeof(*psrc));
1957 *psrc = psf->sf_addr;
1958 scount++; stotal++;
1959 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1960 type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1961 decrease_sf_crcount:
1962 psf->sf_crcount--;
1963 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1964 if (psf_prev)
1965 rcu_assign_pointer(psf_prev->sf_next,
1966 mc_dereference(psf->sf_next, idev));
1967 else
1968 rcu_assign_pointer(*psf_list,
1969 mc_dereference(psf->sf_next, idev));
1970 kfree_rcu(psf, rcu);
1971 continue;
1974 psf_prev = psf;
1977 empty_source:
1978 if (!stotal) {
1979 if (type == MLD2_ALLOW_NEW_SOURCES ||
1980 type == MLD2_BLOCK_OLD_SOURCES)
1981 return skb;
1982 if (pmc->mca_crcount || isquery || crsend) {
1983 /* make sure we have room for group header */
1984 if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1985 mld_sendpack(skb);
1986 skb = NULL; /* add_grhead will get a new one */
1988 skb = add_grhead(skb, pmc, type, &pgr, mtu);
1991 if (pgr)
1992 pgr->grec_nsrcs = htons(scount);
1994 if (isquery)
1995 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1996 return skb;
1999 /* called with mc_lock */
2000 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
2002 struct sk_buff *skb = NULL;
2003 int type;
2005 if (!pmc) {
2006 for_each_mc_mclock(idev, pmc) {
2007 if (pmc->mca_flags & MAF_NOREPORT)
2008 continue;
2009 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2010 type = MLD2_MODE_IS_EXCLUDE;
2011 else
2012 type = MLD2_MODE_IS_INCLUDE;
2013 skb = add_grec(skb, pmc, type, 0, 0, 0);
2015 } else {
2016 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2017 type = MLD2_MODE_IS_EXCLUDE;
2018 else
2019 type = MLD2_MODE_IS_INCLUDE;
2020 skb = add_grec(skb, pmc, type, 0, 0, 0);
2022 if (skb)
2023 mld_sendpack(skb);
2027 * remove zero-count source records from a source filter list
2028 * called with mc_lock
2030 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
2032 struct ip6_sf_list *psf_prev, *psf_next, *psf;
2034 psf_prev = NULL;
2035 for (psf = mc_dereference(*ppsf, idev);
2036 psf;
2037 psf = psf_next) {
2038 psf_next = mc_dereference(psf->sf_next, idev);
2039 if (psf->sf_crcount == 0) {
2040 if (psf_prev)
2041 rcu_assign_pointer(psf_prev->sf_next,
2042 mc_dereference(psf->sf_next, idev));
2043 else
2044 rcu_assign_pointer(*ppsf,
2045 mc_dereference(psf->sf_next, idev));
2046 kfree_rcu(psf, rcu);
2047 } else {
2048 psf_prev = psf;
2053 /* called with mc_lock */
2054 static void mld_send_cr(struct inet6_dev *idev)
2056 struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
2057 struct sk_buff *skb = NULL;
2058 int type, dtype;
2060 /* deleted MCA's */
2061 pmc_prev = NULL;
2062 for (pmc = mc_dereference(idev->mc_tomb, idev);
2063 pmc;
2064 pmc = pmc_next) {
2065 pmc_next = mc_dereference(pmc->next, idev);
2066 if (pmc->mca_sfmode == MCAST_INCLUDE) {
2067 type = MLD2_BLOCK_OLD_SOURCES;
2068 dtype = MLD2_BLOCK_OLD_SOURCES;
2069 skb = add_grec(skb, pmc, type, 1, 0, 0);
2070 skb = add_grec(skb, pmc, dtype, 1, 1, 0);
2072 if (pmc->mca_crcount) {
2073 if (pmc->mca_sfmode == MCAST_EXCLUDE) {
2074 type = MLD2_CHANGE_TO_INCLUDE;
2075 skb = add_grec(skb, pmc, type, 1, 0, 0);
2077 pmc->mca_crcount--;
2078 if (pmc->mca_crcount == 0) {
2079 mld_clear_zeros(&pmc->mca_tomb, idev);
2080 mld_clear_zeros(&pmc->mca_sources, idev);
2083 if (pmc->mca_crcount == 0 &&
2084 !rcu_access_pointer(pmc->mca_tomb) &&
2085 !rcu_access_pointer(pmc->mca_sources)) {
2086 if (pmc_prev)
2087 rcu_assign_pointer(pmc_prev->next, pmc_next);
2088 else
2089 rcu_assign_pointer(idev->mc_tomb, pmc_next);
2090 in6_dev_put(pmc->idev);
2091 kfree_rcu(pmc, rcu);
2092 } else
2093 pmc_prev = pmc;
2096 /* change recs */
2097 for_each_mc_mclock(idev, pmc) {
2098 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2099 type = MLD2_BLOCK_OLD_SOURCES;
2100 dtype = MLD2_ALLOW_NEW_SOURCES;
2101 } else {
2102 type = MLD2_ALLOW_NEW_SOURCES;
2103 dtype = MLD2_BLOCK_OLD_SOURCES;
2105 skb = add_grec(skb, pmc, type, 0, 0, 0);
2106 skb = add_grec(skb, pmc, dtype, 0, 1, 0); /* deleted sources */
2108 /* filter mode changes */
2109 if (pmc->mca_crcount) {
2110 if (pmc->mca_sfmode == MCAST_EXCLUDE)
2111 type = MLD2_CHANGE_TO_EXCLUDE;
2112 else
2113 type = MLD2_CHANGE_TO_INCLUDE;
2114 skb = add_grec(skb, pmc, type, 0, 0, 0);
2115 pmc->mca_crcount--;
2118 if (!skb)
2119 return;
2120 (void) mld_sendpack(skb);
2123 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
2125 struct net *net = dev_net(dev);
2126 struct sock *sk = net->ipv6.igmp_sk;
2127 struct inet6_dev *idev;
2128 struct sk_buff *skb;
2129 struct mld_msg *hdr;
2130 const struct in6_addr *snd_addr, *saddr;
2131 struct in6_addr addr_buf;
2132 int hlen = LL_RESERVED_SPACE(dev);
2133 int tlen = dev->needed_tailroom;
2134 int err, len, payload_len, full_len;
2135 u8 ra[8] = { IPPROTO_ICMPV6, 0,
2136 IPV6_TLV_ROUTERALERT, 2, 0, 0,
2137 IPV6_TLV_PADN, 0 };
2138 struct flowi6 fl6;
2139 struct dst_entry *dst;
2141 if (type == ICMPV6_MGM_REDUCTION)
2142 snd_addr = &in6addr_linklocal_allrouters;
2143 else
2144 snd_addr = addr;
2146 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2147 payload_len = len + sizeof(ra);
2148 full_len = sizeof(struct ipv6hdr) + payload_len;
2150 rcu_read_lock();
2151 IP6_INC_STATS(net, __in6_dev_get(dev), IPSTATS_MIB_OUTREQUESTS);
2152 rcu_read_unlock();
2154 skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2156 if (!skb) {
2157 rcu_read_lock();
2158 IP6_INC_STATS(net, __in6_dev_get(dev),
2159 IPSTATS_MIB_OUTDISCARDS);
2160 rcu_read_unlock();
2161 return;
2163 skb->priority = TC_PRIO_CONTROL;
2164 skb_reserve(skb, hlen);
2166 if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2167 /* <draft-ietf-magma-mld-source-05.txt>:
2168 * use unspecified address as the source address
2169 * when a valid link-local address is not available.
2171 saddr = &in6addr_any;
2172 } else
2173 saddr = &addr_buf;
2175 ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2177 skb_put_data(skb, ra, sizeof(ra));
2179 hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2180 hdr->mld_type = type;
2181 hdr->mld_mca = *addr;
2183 hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2184 IPPROTO_ICMPV6,
2185 csum_partial(hdr, len, 0));
2187 rcu_read_lock();
2188 idev = __in6_dev_get(skb->dev);
2190 icmpv6_flow_init(sk, &fl6, type,
2191 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2192 skb->dev->ifindex);
2193 dst = icmp6_dst_alloc(skb->dev, &fl6);
2194 if (IS_ERR(dst)) {
2195 err = PTR_ERR(dst);
2196 goto err_out;
2199 skb_dst_set(skb, dst);
2200 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2201 net, sk, skb, NULL, skb->dev,
2202 dst_output);
2203 out:
2204 if (!err) {
2205 ICMP6MSGOUT_INC_STATS(net, idev, type);
2206 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2207 } else
2208 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2210 rcu_read_unlock();
2211 return;
2213 err_out:
2214 kfree_skb(skb);
2215 goto out;
2218 /* called with mc_lock */
2219 static void mld_send_initial_cr(struct inet6_dev *idev)
2221 struct sk_buff *skb;
2222 struct ifmcaddr6 *pmc;
2223 int type;
2225 if (mld_in_v1_mode(idev))
2226 return;
2228 skb = NULL;
2229 for_each_mc_mclock(idev, pmc) {
2230 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2231 type = MLD2_CHANGE_TO_EXCLUDE;
2232 else
2233 type = MLD2_ALLOW_NEW_SOURCES;
2234 skb = add_grec(skb, pmc, type, 0, 0, 1);
2236 if (skb)
2237 mld_sendpack(skb);
2240 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2242 mutex_lock(&idev->mc_lock);
2243 idev->mc_dad_count = idev->mc_qrv;
2244 if (idev->mc_dad_count) {
2245 mld_send_initial_cr(idev);
2246 idev->mc_dad_count--;
2247 if (idev->mc_dad_count)
2248 mld_dad_start_work(idev,
2249 unsolicited_report_interval(idev));
2251 mutex_unlock(&idev->mc_lock);
2254 static void mld_dad_work(struct work_struct *work)
2256 struct inet6_dev *idev = container_of(to_delayed_work(work),
2257 struct inet6_dev,
2258 mc_dad_work);
2259 mutex_lock(&idev->mc_lock);
2260 mld_send_initial_cr(idev);
2261 if (idev->mc_dad_count) {
2262 idev->mc_dad_count--;
2263 if (idev->mc_dad_count)
2264 mld_dad_start_work(idev,
2265 unsolicited_report_interval(idev));
2267 mutex_unlock(&idev->mc_lock);
2268 in6_dev_put(idev);
2271 /* called with mc_lock */
2272 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2273 const struct in6_addr *psfsrc)
2275 struct ip6_sf_list *psf, *psf_prev;
2276 int rv = 0;
2278 psf_prev = NULL;
2279 for_each_psf_mclock(pmc, psf) {
2280 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2281 break;
2282 psf_prev = psf;
2284 if (!psf || psf->sf_count[sfmode] == 0) {
2285 /* source filter not found, or count wrong => bug */
2286 return -ESRCH;
2288 psf->sf_count[sfmode]--;
2289 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2290 struct inet6_dev *idev = pmc->idev;
2292 /* no more filters for this source */
2293 if (psf_prev)
2294 rcu_assign_pointer(psf_prev->sf_next,
2295 mc_dereference(psf->sf_next, idev));
2296 else
2297 rcu_assign_pointer(pmc->mca_sources,
2298 mc_dereference(psf->sf_next, idev));
2300 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2301 !mld_in_v1_mode(idev)) {
2302 psf->sf_crcount = idev->mc_qrv;
2303 rcu_assign_pointer(psf->sf_next,
2304 mc_dereference(pmc->mca_tomb, idev));
2305 rcu_assign_pointer(pmc->mca_tomb, psf);
2306 rv = 1;
2307 } else {
2308 kfree_rcu(psf, rcu);
2311 return rv;
2314 /* called with mc_lock */
2315 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2316 int sfmode, int sfcount, const struct in6_addr *psfsrc,
2317 int delta)
2319 struct ifmcaddr6 *pmc;
2320 int changerec = 0;
2321 int i, err;
2323 if (!idev)
2324 return -ENODEV;
2326 for_each_mc_mclock(idev, pmc) {
2327 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2328 break;
2330 if (!pmc)
2331 return -ESRCH;
2333 sf_markstate(pmc);
2334 if (!delta) {
2335 if (!pmc->mca_sfcount[sfmode])
2336 return -EINVAL;
2338 pmc->mca_sfcount[sfmode]--;
2340 err = 0;
2341 for (i = 0; i < sfcount; i++) {
2342 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2344 changerec |= rv > 0;
2345 if (!err && rv < 0)
2346 err = rv;
2348 if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2349 pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2350 pmc->mca_sfcount[MCAST_INCLUDE]) {
2351 struct ip6_sf_list *psf;
2353 /* filter mode change */
2354 pmc->mca_sfmode = MCAST_INCLUDE;
2355 pmc->mca_crcount = idev->mc_qrv;
2356 idev->mc_ifc_count = pmc->mca_crcount;
2357 for_each_psf_mclock(pmc, psf)
2358 psf->sf_crcount = 0;
2359 mld_ifc_event(pmc->idev);
2360 } else if (sf_setstate(pmc) || changerec) {
2361 mld_ifc_event(pmc->idev);
2364 return err;
2368 * Add multicast single-source filter to the interface list
2369 * called with mc_lock
2371 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2372 const struct in6_addr *psfsrc)
2374 struct ip6_sf_list *psf, *psf_prev;
2376 psf_prev = NULL;
2377 for_each_psf_mclock(pmc, psf) {
2378 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2379 break;
2380 psf_prev = psf;
2382 if (!psf) {
2383 psf = kzalloc(sizeof(*psf), GFP_KERNEL);
2384 if (!psf)
2385 return -ENOBUFS;
2387 psf->sf_addr = *psfsrc;
2388 if (psf_prev) {
2389 rcu_assign_pointer(psf_prev->sf_next, psf);
2390 } else {
2391 rcu_assign_pointer(pmc->mca_sources, psf);
2394 psf->sf_count[sfmode]++;
2395 return 0;
2398 /* called with mc_lock */
2399 static void sf_markstate(struct ifmcaddr6 *pmc)
2401 struct ip6_sf_list *psf;
2402 int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2404 for_each_psf_mclock(pmc, psf) {
2405 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2406 psf->sf_oldin = mca_xcount ==
2407 psf->sf_count[MCAST_EXCLUDE] &&
2408 !psf->sf_count[MCAST_INCLUDE];
2409 } else {
2410 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2415 /* called with mc_lock */
2416 static int sf_setstate(struct ifmcaddr6 *pmc)
2418 struct ip6_sf_list *psf, *dpsf;
2419 int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2420 int qrv = pmc->idev->mc_qrv;
2421 int new_in, rv;
2423 rv = 0;
2424 for_each_psf_mclock(pmc, psf) {
2425 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2426 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2427 !psf->sf_count[MCAST_INCLUDE];
2428 } else
2429 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2430 if (new_in) {
2431 if (!psf->sf_oldin) {
2432 struct ip6_sf_list *prev = NULL;
2434 for_each_psf_tomb(pmc, dpsf) {
2435 if (ipv6_addr_equal(&dpsf->sf_addr,
2436 &psf->sf_addr))
2437 break;
2438 prev = dpsf;
2440 if (dpsf) {
2441 if (prev)
2442 rcu_assign_pointer(prev->sf_next,
2443 mc_dereference(dpsf->sf_next,
2444 pmc->idev));
2445 else
2446 rcu_assign_pointer(pmc->mca_tomb,
2447 mc_dereference(dpsf->sf_next,
2448 pmc->idev));
2449 kfree_rcu(dpsf, rcu);
2451 psf->sf_crcount = qrv;
2452 rv++;
2454 } else if (psf->sf_oldin) {
2455 psf->sf_crcount = 0;
2457 * add or update "delete" records if an active filter
2458 * is now inactive
2461 for_each_psf_tomb(pmc, dpsf)
2462 if (ipv6_addr_equal(&dpsf->sf_addr,
2463 &psf->sf_addr))
2464 break;
2465 if (!dpsf) {
2466 dpsf = kmalloc(sizeof(*dpsf), GFP_KERNEL);
2467 if (!dpsf)
2468 continue;
2469 *dpsf = *psf;
2470 rcu_assign_pointer(dpsf->sf_next,
2471 mc_dereference(pmc->mca_tomb, pmc->idev));
2472 rcu_assign_pointer(pmc->mca_tomb, dpsf);
2474 dpsf->sf_crcount = qrv;
2475 rv++;
2478 return rv;
2482 * Add multicast source filter list to the interface list
2483 * called with mc_lock
2485 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2486 int sfmode, int sfcount, const struct in6_addr *psfsrc,
2487 int delta)
2489 struct ifmcaddr6 *pmc;
2490 int isexclude;
2491 int i, err;
2493 if (!idev)
2494 return -ENODEV;
2496 for_each_mc_mclock(idev, pmc) {
2497 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2498 break;
2500 if (!pmc)
2501 return -ESRCH;
2503 sf_markstate(pmc);
2504 isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2505 if (!delta)
2506 pmc->mca_sfcount[sfmode]++;
2507 err = 0;
2508 for (i = 0; i < sfcount; i++) {
2509 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2510 if (err)
2511 break;
2513 if (err) {
2514 int j;
2516 if (!delta)
2517 pmc->mca_sfcount[sfmode]--;
2518 for (j = 0; j < i; j++)
2519 ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2520 } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2521 struct ip6_sf_list *psf;
2523 /* filter mode change */
2524 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2525 pmc->mca_sfmode = MCAST_EXCLUDE;
2526 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2527 pmc->mca_sfmode = MCAST_INCLUDE;
2528 /* else no filters; keep old mode for reports */
2530 pmc->mca_crcount = idev->mc_qrv;
2531 idev->mc_ifc_count = pmc->mca_crcount;
2532 for_each_psf_mclock(pmc, psf)
2533 psf->sf_crcount = 0;
2534 mld_ifc_event(idev);
2535 } else if (sf_setstate(pmc)) {
2536 mld_ifc_event(idev);
2538 return err;
2541 /* called with mc_lock */
2542 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2544 struct ip6_sf_list *psf, *nextpsf;
2546 for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
2547 psf;
2548 psf = nextpsf) {
2549 nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2550 kfree_rcu(psf, rcu);
2552 RCU_INIT_POINTER(pmc->mca_tomb, NULL);
2553 for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
2554 psf;
2555 psf = nextpsf) {
2556 nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2557 kfree_rcu(psf, rcu);
2559 RCU_INIT_POINTER(pmc->mca_sources, NULL);
2560 pmc->mca_sfmode = MCAST_EXCLUDE;
2561 pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2562 pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2565 /* called with mc_lock */
2566 static void igmp6_join_group(struct ifmcaddr6 *ma)
2568 unsigned long delay;
2570 if (ma->mca_flags & MAF_NOREPORT)
2571 return;
2573 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2575 delay = get_random_u32_below(unsolicited_report_interval(ma->idev));
2577 if (cancel_delayed_work(&ma->mca_work)) {
2578 refcount_dec(&ma->mca_refcnt);
2579 delay = ma->mca_work.timer.expires - jiffies;
2582 if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
2583 refcount_inc(&ma->mca_refcnt);
2584 ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2587 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2588 struct inet6_dev *idev)
2590 struct ip6_sf_socklist *psl;
2591 int err;
2593 psl = sock_dereference(iml->sflist, sk);
2595 if (idev)
2596 mutex_lock(&idev->mc_lock);
2598 if (!psl) {
2599 /* any-source empty exclude case */
2600 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2601 } else {
2602 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2603 psl->sl_count, psl->sl_addr, 0);
2604 RCU_INIT_POINTER(iml->sflist, NULL);
2605 atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
2606 &sk->sk_omem_alloc);
2607 kfree_rcu(psl, rcu);
2610 if (idev)
2611 mutex_unlock(&idev->mc_lock);
2613 return err;
2616 /* called with mc_lock */
2617 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2619 if (mld_in_v1_mode(ma->idev)) {
2620 if (ma->mca_flags & MAF_LAST_REPORTER) {
2621 igmp6_send(&ma->mca_addr, ma->idev->dev,
2622 ICMPV6_MGM_REDUCTION);
2624 } else {
2625 mld_add_delrec(ma->idev, ma);
2626 mld_ifc_event(ma->idev);
2630 static void mld_gq_work(struct work_struct *work)
2632 struct inet6_dev *idev = container_of(to_delayed_work(work),
2633 struct inet6_dev,
2634 mc_gq_work);
2636 mutex_lock(&idev->mc_lock);
2637 mld_send_report(idev, NULL);
2638 idev->mc_gq_running = 0;
2639 mutex_unlock(&idev->mc_lock);
2641 in6_dev_put(idev);
2644 static void mld_ifc_work(struct work_struct *work)
2646 struct inet6_dev *idev = container_of(to_delayed_work(work),
2647 struct inet6_dev,
2648 mc_ifc_work);
2650 mutex_lock(&idev->mc_lock);
2651 mld_send_cr(idev);
2653 if (idev->mc_ifc_count) {
2654 idev->mc_ifc_count--;
2655 if (idev->mc_ifc_count)
2656 mld_ifc_start_work(idev,
2657 unsolicited_report_interval(idev));
2659 mutex_unlock(&idev->mc_lock);
2660 in6_dev_put(idev);
2663 /* called with mc_lock */
2664 static void mld_ifc_event(struct inet6_dev *idev)
2666 if (mld_in_v1_mode(idev))
2667 return;
2669 idev->mc_ifc_count = idev->mc_qrv;
2670 mld_ifc_start_work(idev, 1);
2673 static void mld_mca_work(struct work_struct *work)
2675 struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
2676 struct ifmcaddr6, mca_work);
2678 mutex_lock(&ma->idev->mc_lock);
2679 if (mld_in_v1_mode(ma->idev))
2680 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2681 else
2682 mld_send_report(ma->idev, ma);
2683 ma->mca_flags |= MAF_LAST_REPORTER;
2684 ma->mca_flags &= ~MAF_TIMER_RUNNING;
2685 mutex_unlock(&ma->idev->mc_lock);
2687 ma_put(ma);
2690 /* Device changing type */
2692 void ipv6_mc_unmap(struct inet6_dev *idev)
2694 struct ifmcaddr6 *i;
2696 /* Install multicast list, except for all-nodes (already installed) */
2698 mutex_lock(&idev->mc_lock);
2699 for_each_mc_mclock(idev, i)
2700 igmp6_group_dropped(i);
2701 mutex_unlock(&idev->mc_lock);
2704 void ipv6_mc_remap(struct inet6_dev *idev)
2706 ipv6_mc_up(idev);
2709 /* Device going down */
2710 void ipv6_mc_down(struct inet6_dev *idev)
2712 struct ifmcaddr6 *i;
2714 mutex_lock(&idev->mc_lock);
2715 /* Withdraw multicast list */
2716 for_each_mc_mclock(idev, i)
2717 igmp6_group_dropped(i);
2718 mutex_unlock(&idev->mc_lock);
2720 /* Should stop work after group drop. or we will
2721 * start work again in mld_ifc_event()
2723 mld_query_stop_work(idev);
2724 mld_report_stop_work(idev);
2726 mutex_lock(&idev->mc_lock);
2727 mld_ifc_stop_work(idev);
2728 mld_gq_stop_work(idev);
2729 mutex_unlock(&idev->mc_lock);
2731 mld_dad_stop_work(idev);
2734 static void ipv6_mc_reset(struct inet6_dev *idev)
2736 idev->mc_qrv = sysctl_mld_qrv;
2737 idev->mc_qi = MLD_QI_DEFAULT;
2738 idev->mc_qri = MLD_QRI_DEFAULT;
2739 idev->mc_v1_seen = 0;
2740 idev->mc_maxdelay = unsolicited_report_interval(idev);
2743 /* Device going up */
2745 void ipv6_mc_up(struct inet6_dev *idev)
2747 struct ifmcaddr6 *i;
2749 /* Install multicast list, except for all-nodes (already installed) */
2751 ipv6_mc_reset(idev);
2752 mutex_lock(&idev->mc_lock);
2753 for_each_mc_mclock(idev, i) {
2754 mld_del_delrec(idev, i);
2755 igmp6_group_added(i);
2757 mutex_unlock(&idev->mc_lock);
2760 /* IPv6 device initialization. */
2762 void ipv6_mc_init_dev(struct inet6_dev *idev)
2764 idev->mc_gq_running = 0;
2765 INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work);
2766 RCU_INIT_POINTER(idev->mc_tomb, NULL);
2767 idev->mc_ifc_count = 0;
2768 INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work);
2769 INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work);
2770 INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work);
2771 INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work);
2772 skb_queue_head_init(&idev->mc_query_queue);
2773 skb_queue_head_init(&idev->mc_report_queue);
2774 spin_lock_init(&idev->mc_query_lock);
2775 spin_lock_init(&idev->mc_report_lock);
2776 mutex_init(&idev->mc_lock);
2777 ipv6_mc_reset(idev);
2781 * Device is about to be destroyed: clean up.
2784 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2786 struct ifmcaddr6 *i;
2788 /* Deactivate works */
2789 ipv6_mc_down(idev);
2790 mutex_lock(&idev->mc_lock);
2791 mld_clear_delrec(idev);
2792 mutex_unlock(&idev->mc_lock);
2793 mld_clear_query(idev);
2794 mld_clear_report(idev);
2796 /* Delete all-nodes address. */
2797 /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2798 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2799 * fail.
2801 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2803 if (idev->cnf.forwarding)
2804 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2806 mutex_lock(&idev->mc_lock);
2807 while ((i = mc_dereference(idev->mc_list, idev))) {
2808 rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
2810 ip6_mc_clear_src(i);
2811 ma_put(i);
2813 mutex_unlock(&idev->mc_lock);
2816 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2818 struct ifmcaddr6 *pmc;
2820 ASSERT_RTNL();
2822 mutex_lock(&idev->mc_lock);
2823 if (mld_in_v1_mode(idev)) {
2824 for_each_mc_mclock(idev, pmc)
2825 igmp6_join_group(pmc);
2826 } else {
2827 mld_send_report(idev, NULL);
2829 mutex_unlock(&idev->mc_lock);
2832 static int ipv6_mc_netdev_event(struct notifier_block *this,
2833 unsigned long event,
2834 void *ptr)
2836 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2837 struct inet6_dev *idev = __in6_dev_get(dev);
2839 switch (event) {
2840 case NETDEV_RESEND_IGMP:
2841 if (idev)
2842 ipv6_mc_rejoin_groups(idev);
2843 break;
2844 default:
2845 break;
2848 return NOTIFY_DONE;
2851 static struct notifier_block igmp6_netdev_notifier = {
2852 .notifier_call = ipv6_mc_netdev_event,
2855 #ifdef CONFIG_PROC_FS
2856 struct igmp6_mc_iter_state {
2857 struct seq_net_private p;
2858 struct net_device *dev;
2859 struct inet6_dev *idev;
2862 #define igmp6_mc_seq_private(seq) ((struct igmp6_mc_iter_state *)(seq)->private)
2864 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2866 struct ifmcaddr6 *im = NULL;
2867 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2868 struct net *net = seq_file_net(seq);
2870 state->idev = NULL;
2871 for_each_netdev_rcu(net, state->dev) {
2872 struct inet6_dev *idev;
2873 idev = __in6_dev_get(state->dev);
2874 if (!idev)
2875 continue;
2877 im = rcu_dereference(idev->mc_list);
2878 if (im) {
2879 state->idev = idev;
2880 break;
2883 return im;
2886 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2888 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2890 im = rcu_dereference(im->next);
2891 while (!im) {
2892 state->dev = next_net_device_rcu(state->dev);
2893 if (!state->dev) {
2894 state->idev = NULL;
2895 break;
2897 state->idev = __in6_dev_get(state->dev);
2898 if (!state->idev)
2899 continue;
2900 im = rcu_dereference(state->idev->mc_list);
2902 return im;
2905 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2907 struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2908 if (im)
2909 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2910 --pos;
2911 return pos ? NULL : im;
2914 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2915 __acquires(RCU)
2917 rcu_read_lock();
2918 return igmp6_mc_get_idx(seq, *pos);
2921 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2923 struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2925 ++*pos;
2926 return im;
2929 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2930 __releases(RCU)
2932 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2934 if (likely(state->idev))
2935 state->idev = NULL;
2936 state->dev = NULL;
2937 rcu_read_unlock();
2940 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2942 struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2943 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2945 seq_printf(seq,
2946 "%-4d %-15s %pi6 %5d %08X %ld\n",
2947 state->dev->ifindex, state->dev->name,
2948 &im->mca_addr,
2949 im->mca_users, im->mca_flags,
2950 (im->mca_flags & MAF_TIMER_RUNNING) ?
2951 jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0);
2952 return 0;
2955 static const struct seq_operations igmp6_mc_seq_ops = {
2956 .start = igmp6_mc_seq_start,
2957 .next = igmp6_mc_seq_next,
2958 .stop = igmp6_mc_seq_stop,
2959 .show = igmp6_mc_seq_show,
2962 struct igmp6_mcf_iter_state {
2963 struct seq_net_private p;
2964 struct net_device *dev;
2965 struct inet6_dev *idev;
2966 struct ifmcaddr6 *im;
2969 #define igmp6_mcf_seq_private(seq) ((struct igmp6_mcf_iter_state *)(seq)->private)
2971 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2973 struct ip6_sf_list *psf = NULL;
2974 struct ifmcaddr6 *im = NULL;
2975 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2976 struct net *net = seq_file_net(seq);
2978 state->idev = NULL;
2979 state->im = NULL;
2980 for_each_netdev_rcu(net, state->dev) {
2981 struct inet6_dev *idev;
2982 idev = __in6_dev_get(state->dev);
2983 if (unlikely(idev == NULL))
2984 continue;
2986 im = rcu_dereference(idev->mc_list);
2987 if (likely(im)) {
2988 psf = rcu_dereference(im->mca_sources);
2989 if (likely(psf)) {
2990 state->im = im;
2991 state->idev = idev;
2992 break;
2996 return psf;
2999 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
3001 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3003 psf = rcu_dereference(psf->sf_next);
3004 while (!psf) {
3005 state->im = rcu_dereference(state->im->next);
3006 while (!state->im) {
3007 state->dev = next_net_device_rcu(state->dev);
3008 if (!state->dev) {
3009 state->idev = NULL;
3010 goto out;
3012 state->idev = __in6_dev_get(state->dev);
3013 if (!state->idev)
3014 continue;
3015 state->im = rcu_dereference(state->idev->mc_list);
3017 psf = rcu_dereference(state->im->mca_sources);
3019 out:
3020 return psf;
3023 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
3025 struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
3026 if (psf)
3027 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
3028 --pos;
3029 return pos ? NULL : psf;
3032 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
3033 __acquires(RCU)
3035 rcu_read_lock();
3036 return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
3039 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3041 struct ip6_sf_list *psf;
3042 if (v == SEQ_START_TOKEN)
3043 psf = igmp6_mcf_get_first(seq);
3044 else
3045 psf = igmp6_mcf_get_next(seq, v);
3046 ++*pos;
3047 return psf;
3050 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
3051 __releases(RCU)
3053 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3055 if (likely(state->im))
3056 state->im = NULL;
3057 if (likely(state->idev))
3058 state->idev = NULL;
3060 state->dev = NULL;
3061 rcu_read_unlock();
3064 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
3066 struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
3067 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3069 if (v == SEQ_START_TOKEN) {
3070 seq_puts(seq, "Idx Device Multicast Address Source Address INC EXC\n");
3071 } else {
3072 seq_printf(seq,
3073 "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
3074 state->dev->ifindex, state->dev->name,
3075 &state->im->mca_addr,
3076 &psf->sf_addr,
3077 psf->sf_count[MCAST_INCLUDE],
3078 psf->sf_count[MCAST_EXCLUDE]);
3080 return 0;
3083 static const struct seq_operations igmp6_mcf_seq_ops = {
3084 .start = igmp6_mcf_seq_start,
3085 .next = igmp6_mcf_seq_next,
3086 .stop = igmp6_mcf_seq_stop,
3087 .show = igmp6_mcf_seq_show,
3090 static int __net_init igmp6_proc_init(struct net *net)
3092 int err;
3094 err = -ENOMEM;
3095 if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
3096 sizeof(struct igmp6_mc_iter_state)))
3097 goto out;
3098 if (!proc_create_net("mcfilter6", 0444, net->proc_net,
3099 &igmp6_mcf_seq_ops,
3100 sizeof(struct igmp6_mcf_iter_state)))
3101 goto out_proc_net_igmp6;
3103 err = 0;
3104 out:
3105 return err;
3107 out_proc_net_igmp6:
3108 remove_proc_entry("igmp6", net->proc_net);
3109 goto out;
3112 static void __net_exit igmp6_proc_exit(struct net *net)
3114 remove_proc_entry("mcfilter6", net->proc_net);
3115 remove_proc_entry("igmp6", net->proc_net);
3117 #else
3118 static inline int igmp6_proc_init(struct net *net)
3120 return 0;
3122 static inline void igmp6_proc_exit(struct net *net)
3125 #endif
3127 static int __net_init igmp6_net_init(struct net *net)
3129 int err;
3131 err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
3132 SOCK_RAW, IPPROTO_ICMPV6, net);
3133 if (err < 0) {
3134 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
3135 err);
3136 goto out;
3139 inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
3140 net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL;
3142 err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
3143 SOCK_RAW, IPPROTO_ICMPV6, net);
3144 if (err < 0) {
3145 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
3146 err);
3147 goto out_sock_create;
3150 err = igmp6_proc_init(net);
3151 if (err)
3152 goto out_sock_create_autojoin;
3154 return 0;
3156 out_sock_create_autojoin:
3157 inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3158 out_sock_create:
3159 inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3160 out:
3161 return err;
3164 static void __net_exit igmp6_net_exit(struct net *net)
3166 inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3167 inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3168 igmp6_proc_exit(net);
3171 static struct pernet_operations igmp6_net_ops = {
3172 .init = igmp6_net_init,
3173 .exit = igmp6_net_exit,
3176 int __init igmp6_init(void)
3178 int err;
3180 err = register_pernet_subsys(&igmp6_net_ops);
3181 if (err)
3182 return err;
3184 mld_wq = create_workqueue("mld");
3185 if (!mld_wq) {
3186 unregister_pernet_subsys(&igmp6_net_ops);
3187 return -ENOMEM;
3190 return err;
3193 int __init igmp6_late_init(void)
3195 return register_netdevice_notifier(&igmp6_netdev_notifier);
3198 void igmp6_cleanup(void)
3200 unregister_pernet_subsys(&igmp6_net_ops);
3201 destroy_workqueue(mld_wq);
3204 void igmp6_late_cleanup(void)
3206 unregister_netdevice_notifier(&igmp6_netdev_notifier);