1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Multicast support for IPv6
4 * Linux INET6 implementation
9 * Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
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.
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>
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>
44 #include <linux/workqueue.h>
46 #include <linux/netfilter.h>
47 #include <linux/netfilter_ipv6.h>
49 #include <net/net_namespace.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,
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,
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); \
120 pmc = sock_dereference(pmc->next, sk))
122 #define for_each_pmc_rcu(np, pmc) \
123 for (pmc = rcu_dereference((np)->ipv6_mc_list); \
125 pmc = rcu_dereference(pmc->next))
127 #define for_each_psf_mclock(mc, psf) \
128 for (psf = mc_dereference((mc)->mca_sources, mc->idev); \
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); \
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); \
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); \
145 mc = mc_dereference(mc->next, idev))
147 #define for_each_mc_rcu(idev, mc) \
148 for (mc = rcu_dereference((idev)->mc_list); \
150 mc = rcu_dereference(mc->next))
152 #define for_each_mc_tomb(idev, mc) \
153 for (mc = mc_dereference((idev)->mc_tomb, idev); \
155 mc = mc_dereference(mc->next, idev))
157 static int unsolicited_report_interval(struct inet6_dev *idev)
161 if (mld_in_v1_mode(idev))
162 iv = idev->cnf.mldv1_unsolicited_report_interval;
164 iv = 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);
180 if (!ipv6_addr_is_multicast(addr))
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))
189 mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
195 mc_lst->addr = *addr;
199 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
205 dev = __dev_get_by_index(net, ifindex);
208 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
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);
223 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
227 mc_lst->next = np->ipv6_mc_list;
228 rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
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);
257 if (!ipv6_addr_is_multicast(addr))
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;
269 dev = __dev_get_by_index(net, mc_lst->ifindex);
271 struct inet6_dev *idev = __in6_dev_get(dev);
273 ip6_mc_leave_src(sk, mc_lst, idev);
275 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
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);
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,
294 struct net_device *dev = NULL;
295 struct inet6_dev *idev = NULL;
298 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
305 dev = __dev_get_by_index(net, ifindex);
310 idev = __in6_dev_get(dev);
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);
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);
333 struct inet6_dev *idev = __in6_dev_get(dev);
335 ip6_mc_leave_src(sk, mc_lst, idev);
337 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
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))
356 __ipv6_sock_mc_close(sk);
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);
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))
380 idev = ip6_mc_find_dev_rtnl(net, group, pgsr->gsr_interface);
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)
390 if (ipv6_addr_equal(&pmc->addr, group))
393 if (!pmc) { /* must have a prior join */
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) {
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);
410 psl = sock_dereference(pmc->sflist, sk);
413 goto done; /* err = -EADDRNOTAVAIL */
415 for (i = 0; i < psl->sl_count; i++) {
416 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
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) {
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];
438 /* else, add a new source to the filter */
440 if (psl && psl->sl_count >= sysctl_mld_max_msf) {
444 if (!psl || psl->sl_count == psl->sl_max) {
445 struct ip6_sf_socklist *newpsl;
446 int count = IP6_SFBLOCK;
449 count += psl->sl_max;
450 newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, count),
456 newpsl->sl_max = count;
457 newpsl->sl_count = count - IP6_SFBLOCK;
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),
466 rcu_assign_pointer(pmc->sflist, psl);
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. */
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;
479 /* update the interface list */
480 ip6_mc_add_src(idev, group, omode, 1, source, 1);
482 mutex_unlock(&idev->mc_lock);
484 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
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);
500 group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
502 if (!ipv6_addr_is_multicast(group))
504 if (gsf->gf_fmode != MCAST_INCLUDE &&
505 gsf->gf_fmode != MCAST_EXCLUDE)
508 idev = ip6_mc_find_dev_rtnl(net, group, gsf->gf_interface);
514 if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
519 for_each_pmc_socklock(inet6, sk, pmc) {
520 if (pmc->ifindex != gsf->gf_interface)
522 if (ipv6_addr_equal(&pmc->addr, group))
525 if (!pmc) { /* must have a prior join */
529 if (gsf->gf_numsrc) {
530 newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr,
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);
548 mutex_unlock(&idev->mc_lock);
549 sock_kfree_s(sk, newpsl, struct_size(newpsl, sl_addr,
553 mutex_unlock(&idev->mc_lock);
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);
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),
570 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
572 mutex_unlock(&idev->mc_lock);
573 rcu_assign_pointer(pmc->sflist, newpsl);
574 pmc->sfmode = gsf->gf_fmode;
578 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
582 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
583 struct sockaddr_storage __user *p)
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 int i, count, copycount;
591 group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
593 if (!ipv6_addr_is_multicast(group))
596 /* changes to the ipv6_mc_list require the socket lock and
597 * rtnl lock. We have the socket lock, so reading the list is safe.
600 for_each_pmc_socklock(inet6, sk, pmc) {
601 if (pmc->ifindex != gsf->gf_interface)
603 if (ipv6_addr_equal(group, &pmc->addr))
606 if (!pmc) /* must have a prior join */
607 return -EADDRNOTAVAIL;
609 gsf->gf_fmode = pmc->sfmode;
610 psl = sock_dereference(pmc->sflist, sk);
611 count = psl ? psl->sl_count : 0;
613 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
614 gsf->gf_numsrc = count;
616 for (i = 0; i < copycount; i++, p++) {
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_user(p, &ss, sizeof(ss)))
630 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
631 const struct in6_addr *src_addr)
633 struct ipv6_pinfo *np = inet6_sk(sk);
634 struct ipv6_mc_socklist *mc;
635 struct ip6_sf_socklist *psl;
639 for_each_pmc_rcu(np, mc) {
640 if (ipv6_addr_equal(&mc->addr, mc_addr))
647 psl = rcu_dereference(mc->sflist);
649 rv = mc->sfmode == MCAST_EXCLUDE;
653 for (i = 0; i < psl->sl_count; i++) {
654 if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
657 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
659 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
667 /* called with mc_lock */
668 static void igmp6_group_added(struct ifmcaddr6 *mc)
670 struct net_device *dev = mc->idev->dev;
671 char buf[MAX_ADDR_LEN];
673 if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
674 IPV6_ADDR_SCOPE_LINKLOCAL)
677 if (!(mc->mca_flags&MAF_LOADED)) {
678 mc->mca_flags |= MAF_LOADED;
679 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
680 dev_mc_add(dev, buf);
683 if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
686 if (mld_in_v1_mode(mc->idev)) {
687 igmp6_join_group(mc);
692 /* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
693 * should not send filter-mode change record as the mode
694 * should be from IN() to IN(A).
696 if (mc->mca_sfmode == MCAST_EXCLUDE)
697 mc->mca_crcount = mc->idev->mc_qrv;
699 mld_ifc_event(mc->idev);
702 /* called with mc_lock */
703 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
705 struct net_device *dev = mc->idev->dev;
706 char buf[MAX_ADDR_LEN];
708 if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
709 IPV6_ADDR_SCOPE_LINKLOCAL)
712 if (mc->mca_flags&MAF_LOADED) {
713 mc->mca_flags &= ~MAF_LOADED;
714 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
715 dev_mc_del(dev, buf);
718 if (mc->mca_flags & MAF_NOREPORT)
722 igmp6_leave_group(mc);
724 if (cancel_delayed_work(&mc->mca_work))
725 refcount_dec(&mc->mca_refcnt);
729 * deleted ifmcaddr6 manipulation
730 * called with mc_lock
732 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
734 struct ifmcaddr6 *pmc;
736 /* this is an "ifmcaddr6" for convenience; only the fields below
737 * are actually used. In particular, the refcnt and users are not
738 * used for management of the delete list. Using the same structure
739 * for deleted items allows change reports to use common code with
740 * non-deleted or query-response MCA's.
742 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
746 pmc->idev = im->idev;
748 pmc->mca_addr = im->mca_addr;
749 pmc->mca_crcount = idev->mc_qrv;
750 pmc->mca_sfmode = im->mca_sfmode;
751 if (pmc->mca_sfmode == MCAST_INCLUDE) {
752 struct ip6_sf_list *psf;
754 rcu_assign_pointer(pmc->mca_tomb,
755 mc_dereference(im->mca_tomb, idev));
756 rcu_assign_pointer(pmc->mca_sources,
757 mc_dereference(im->mca_sources, idev));
758 RCU_INIT_POINTER(im->mca_tomb, NULL);
759 RCU_INIT_POINTER(im->mca_sources, NULL);
761 for_each_psf_mclock(pmc, psf)
762 psf->sf_crcount = pmc->mca_crcount;
765 rcu_assign_pointer(pmc->next, idev->mc_tomb);
766 rcu_assign_pointer(idev->mc_tomb, pmc);
769 /* called with mc_lock */
770 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
772 struct ip6_sf_list *psf, *sources, *tomb;
773 struct in6_addr *pmca = &im->mca_addr;
774 struct ifmcaddr6 *pmc, *pmc_prev;
777 for_each_mc_tomb(idev, pmc) {
778 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
784 rcu_assign_pointer(pmc_prev->next, pmc->next);
786 rcu_assign_pointer(idev->mc_tomb, pmc->next);
790 im->idev = pmc->idev;
791 if (im->mca_sfmode == MCAST_INCLUDE) {
792 tomb = rcu_replace_pointer(im->mca_tomb,
793 mc_dereference(pmc->mca_tomb, pmc->idev),
794 lockdep_is_held(&im->idev->mc_lock));
795 rcu_assign_pointer(pmc->mca_tomb, tomb);
797 sources = rcu_replace_pointer(im->mca_sources,
798 mc_dereference(pmc->mca_sources, pmc->idev),
799 lockdep_is_held(&im->idev->mc_lock));
800 rcu_assign_pointer(pmc->mca_sources, sources);
801 for_each_psf_mclock(im, psf)
802 psf->sf_crcount = idev->mc_qrv;
804 im->mca_crcount = idev->mc_qrv;
806 in6_dev_put(pmc->idev);
807 ip6_mc_clear_src(pmc);
812 /* called with mc_lock */
813 static void mld_clear_delrec(struct inet6_dev *idev)
815 struct ifmcaddr6 *pmc, *nextpmc;
817 pmc = mc_dereference(idev->mc_tomb, idev);
818 RCU_INIT_POINTER(idev->mc_tomb, NULL);
820 for (; pmc; pmc = nextpmc) {
821 nextpmc = mc_dereference(pmc->next, idev);
822 ip6_mc_clear_src(pmc);
823 in6_dev_put(pmc->idev);
827 /* clear dead sources, too */
828 for_each_mc_mclock(idev, pmc) {
829 struct ip6_sf_list *psf, *psf_next;
831 psf = mc_dereference(pmc->mca_tomb, idev);
832 RCU_INIT_POINTER(pmc->mca_tomb, NULL);
833 for (; psf; psf = psf_next) {
834 psf_next = mc_dereference(psf->sf_next, idev);
840 static void mld_clear_query(struct inet6_dev *idev)
844 spin_lock_bh(&idev->mc_query_lock);
845 while ((skb = __skb_dequeue(&idev->mc_query_queue)))
847 spin_unlock_bh(&idev->mc_query_lock);
850 static void mld_clear_report(struct inet6_dev *idev)
854 spin_lock_bh(&idev->mc_report_lock);
855 while ((skb = __skb_dequeue(&idev->mc_report_queue)))
857 spin_unlock_bh(&idev->mc_report_lock);
860 static void mca_get(struct ifmcaddr6 *mc)
862 refcount_inc(&mc->mca_refcnt);
865 static void ma_put(struct ifmcaddr6 *mc)
867 if (refcount_dec_and_test(&mc->mca_refcnt)) {
868 in6_dev_put(mc->idev);
873 /* called with mc_lock */
874 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
875 const struct in6_addr *addr,
878 struct ifmcaddr6 *mc;
880 mc = kzalloc(sizeof(*mc), GFP_KERNEL);
884 INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);
886 mc->mca_addr = *addr;
887 mc->idev = idev; /* reference taken by caller */
889 /* mca_stamp should be updated upon changes */
890 mc->mca_cstamp = mc->mca_tstamp = jiffies;
891 refcount_set(&mc->mca_refcnt, 1);
893 mc->mca_sfmode = mode;
894 mc->mca_sfcount[mode] = 1;
896 if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
897 IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
898 mc->mca_flags |= MAF_NOREPORT;
904 * device multicast group inc (add if not found)
906 static int __ipv6_dev_mc_inc(struct net_device *dev,
907 const struct in6_addr *addr, unsigned int mode)
909 struct ifmcaddr6 *mc;
910 struct inet6_dev *idev;
914 /* we need to take a reference on idev */
915 idev = in6_dev_get(dev);
925 mutex_lock(&idev->mc_lock);
926 for_each_mc_mclock(idev, mc) {
927 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
929 ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
930 mutex_unlock(&idev->mc_lock);
936 mc = mca_alloc(idev, addr, mode);
938 mutex_unlock(&idev->mc_lock);
943 rcu_assign_pointer(mc->next, idev->mc_list);
944 rcu_assign_pointer(idev->mc_list, mc);
948 mld_del_delrec(idev, mc);
949 igmp6_group_added(mc);
950 mutex_unlock(&idev->mc_lock);
955 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
957 return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
959 EXPORT_SYMBOL(ipv6_dev_mc_inc);
962 * device multicast group del
964 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
966 struct ifmcaddr6 *ma, __rcu **map;
970 mutex_lock(&idev->mc_lock);
971 for (map = &idev->mc_list;
972 (ma = mc_dereference(*map, idev));
974 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
975 if (--ma->mca_users == 0) {
978 igmp6_group_dropped(ma);
979 ip6_mc_clear_src(ma);
980 mutex_unlock(&idev->mc_lock);
985 mutex_unlock(&idev->mc_lock);
990 mutex_unlock(&idev->mc_lock);
994 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
996 struct inet6_dev *idev;
1001 idev = __in6_dev_get(dev);
1005 err = __ipv6_dev_mc_dec(idev, addr);
1009 EXPORT_SYMBOL(ipv6_dev_mc_dec);
1012 * check if the interface/address pair is valid
1014 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
1015 const struct in6_addr *src_addr)
1017 struct inet6_dev *idev;
1018 struct ifmcaddr6 *mc;
1022 idev = __in6_dev_get(dev);
1024 for_each_mc_rcu(idev, mc) {
1025 if (ipv6_addr_equal(&mc->mca_addr, group))
1029 if (src_addr && !ipv6_addr_any(src_addr)) {
1030 struct ip6_sf_list *psf;
1032 for_each_psf_rcu(mc, psf) {
1033 if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1037 rv = psf->sf_count[MCAST_INCLUDE] ||
1038 psf->sf_count[MCAST_EXCLUDE] !=
1039 mc->mca_sfcount[MCAST_EXCLUDE];
1041 rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1043 rv = true; /* don't filter unspecified source */
1050 /* called with mc_lock */
1051 static void mld_gq_start_work(struct inet6_dev *idev)
1053 unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1055 idev->mc_gq_running = 1;
1056 if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2))
1060 /* called with mc_lock */
1061 static void mld_gq_stop_work(struct inet6_dev *idev)
1063 idev->mc_gq_running = 0;
1064 if (cancel_delayed_work(&idev->mc_gq_work))
1065 __in6_dev_put(idev);
1068 /* called with mc_lock */
1069 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
1071 unsigned long tv = prandom_u32() % delay;
1073 if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2))
1077 /* called with mc_lock */
1078 static void mld_ifc_stop_work(struct inet6_dev *idev)
1080 idev->mc_ifc_count = 0;
1081 if (cancel_delayed_work(&idev->mc_ifc_work))
1082 __in6_dev_put(idev);
1085 /* called with mc_lock */
1086 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
1088 unsigned long tv = prandom_u32() % delay;
1090 if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2))
1094 static void mld_dad_stop_work(struct inet6_dev *idev)
1096 if (cancel_delayed_work(&idev->mc_dad_work))
1097 __in6_dev_put(idev);
1100 static void mld_query_stop_work(struct inet6_dev *idev)
1102 spin_lock_bh(&idev->mc_query_lock);
1103 if (cancel_delayed_work(&idev->mc_query_work))
1104 __in6_dev_put(idev);
1105 spin_unlock_bh(&idev->mc_query_lock);
1108 static void mld_report_stop_work(struct inet6_dev *idev)
1110 if (cancel_delayed_work_sync(&idev->mc_report_work))
1111 __in6_dev_put(idev);
1115 * IGMP handling (alias multicast ICMPv6 messages)
1116 * called with mc_lock
1118 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1120 unsigned long delay = resptime;
1122 /* Do not start work for these addresses */
1123 if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1124 IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1127 if (cancel_delayed_work(&ma->mca_work)) {
1128 refcount_dec(&ma->mca_refcnt);
1129 delay = ma->mca_work.timer.expires - jiffies;
1132 if (delay >= resptime)
1133 delay = prandom_u32() % resptime;
1135 if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
1136 refcount_inc(&ma->mca_refcnt);
1137 ma->mca_flags |= MAF_TIMER_RUNNING;
1140 /* mark EXCLUDE-mode sources
1141 * called with mc_lock
1143 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1144 const struct in6_addr *srcs)
1146 struct ip6_sf_list *psf;
1150 for_each_psf_mclock(pmc, psf) {
1151 if (scount == nsrcs)
1153 for (i = 0; i < nsrcs; i++) {
1154 /* skip inactive filters */
1155 if (psf->sf_count[MCAST_INCLUDE] ||
1156 pmc->mca_sfcount[MCAST_EXCLUDE] !=
1157 psf->sf_count[MCAST_EXCLUDE])
1159 if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1165 pmc->mca_flags &= ~MAF_GSQUERY;
1166 if (scount == nsrcs) /* all sources excluded */
1171 /* called with mc_lock */
1172 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1173 const struct in6_addr *srcs)
1175 struct ip6_sf_list *psf;
1178 if (pmc->mca_sfmode == MCAST_EXCLUDE)
1179 return mld_xmarksources(pmc, nsrcs, srcs);
1181 /* mark INCLUDE-mode sources */
1184 for_each_psf_mclock(pmc, psf) {
1185 if (scount == nsrcs)
1187 for (i = 0; i < nsrcs; i++) {
1188 if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1196 pmc->mca_flags &= ~MAF_GSQUERY;
1199 pmc->mca_flags |= MAF_GSQUERY;
1203 static int mld_force_mld_version(const struct inet6_dev *idev)
1205 /* Normally, both are 0 here. If enforcement to a particular is
1206 * being used, individual device enforcement will have a lower
1207 * precedence over 'all' device (.../conf/all/force_mld_version).
1210 if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1211 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1213 return idev->cnf.force_mld_version;
1216 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1218 return mld_force_mld_version(idev) == 2;
1221 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1223 return mld_force_mld_version(idev) == 1;
1226 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1228 if (mld_in_v2_mode_only(idev))
1230 if (mld_in_v1_mode_only(idev))
1232 if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1238 static void mld_set_v1_mode(struct inet6_dev *idev)
1240 /* RFC3810, relevant sections:
1241 * - 9.1. Robustness Variable
1242 * - 9.2. Query Interval
1243 * - 9.3. Query Response Interval
1244 * - 9.12. Older Version Querier Present Timeout
1246 unsigned long switchback;
1248 switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1250 idev->mc_v1_seen = jiffies + switchback;
1253 static void mld_update_qrv(struct inet6_dev *idev,
1254 const struct mld2_query *mlh2)
1256 /* RFC3810, relevant sections:
1257 * - 5.1.8. QRV (Querier's Robustness Variable)
1258 * - 9.1. Robustness Variable
1261 /* The value of the Robustness Variable MUST NOT be zero,
1262 * and SHOULD NOT be one. Catch this here if we ever run
1263 * into such a case in future.
1265 const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1266 WARN_ON(idev->mc_qrv == 0);
1268 if (mlh2->mld2q_qrv > 0)
1269 idev->mc_qrv = mlh2->mld2q_qrv;
1271 if (unlikely(idev->mc_qrv < min_qrv)) {
1272 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1273 idev->mc_qrv, min_qrv);
1274 idev->mc_qrv = min_qrv;
1278 static void mld_update_qi(struct inet6_dev *idev,
1279 const struct mld2_query *mlh2)
1281 /* RFC3810, relevant sections:
1282 * - 5.1.9. QQIC (Querier's Query Interval Code)
1283 * - 9.2. Query Interval
1284 * - 9.12. Older Version Querier Present Timeout
1285 * (the [Query Interval] in the last Query received)
1287 unsigned long mc_qqi;
1289 if (mlh2->mld2q_qqic < 128) {
1290 mc_qqi = mlh2->mld2q_qqic;
1292 unsigned long mc_man, mc_exp;
1294 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1295 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1297 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1300 idev->mc_qi = mc_qqi * HZ;
1303 static void mld_update_qri(struct inet6_dev *idev,
1304 const struct mld2_query *mlh2)
1306 /* RFC3810, relevant sections:
1307 * - 5.1.3. Maximum Response Code
1308 * - 9.3. Query Response Interval
1310 idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1313 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1314 unsigned long *max_delay, bool v1_query)
1316 unsigned long mldv1_md;
1318 /* Ignore v1 queries */
1319 if (mld_in_v2_mode_only(idev))
1322 mldv1_md = ntohs(mld->mld_maxdelay);
1324 /* When in MLDv1 fallback and a MLDv2 router start-up being
1325 * unaware of current MLDv1 operation, the MRC == MRD mapping
1326 * only works when the exponential algorithm is not being
1327 * used (as MLDv1 is unaware of such things).
1329 * According to the RFC author, the MLDv2 implementations
1330 * he's aware of all use a MRC < 32768 on start up queries.
1332 * Thus, should we *ever* encounter something else larger
1333 * than that, just assume the maximum possible within our
1337 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1339 *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1341 /* MLDv1 router present: we need to go into v1 mode *only*
1342 * when an MLDv1 query is received as per section 9.12. of
1343 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1344 * queries MUST be of exactly 24 octets.
1347 mld_set_v1_mode(idev);
1349 /* cancel MLDv2 report work */
1350 mld_gq_stop_work(idev);
1351 /* cancel the interface change work */
1352 mld_ifc_stop_work(idev);
1353 /* clear deleted report items */
1354 mld_clear_delrec(idev);
1359 static void mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1360 unsigned long *max_delay)
1362 *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1364 mld_update_qrv(idev, mld);
1365 mld_update_qi(idev, mld);
1366 mld_update_qri(idev, mld);
1368 idev->mc_maxdelay = *max_delay;
1373 /* called with rcu_read_lock() */
1374 int igmp6_event_query(struct sk_buff *skb)
1376 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1386 spin_lock_bh(&idev->mc_query_lock);
1387 if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) {
1388 __skb_queue_tail(&idev->mc_query_queue, skb);
1389 if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0))
1392 spin_unlock_bh(&idev->mc_query_lock);
1397 static void __mld_query_work(struct sk_buff *skb)
1399 struct mld2_query *mlh2 = NULL;
1400 const struct in6_addr *group;
1401 unsigned long max_delay;
1402 struct inet6_dev *idev;
1403 struct ifmcaddr6 *ma;
1404 struct mld_msg *mld;
1409 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1412 /* compute payload length excluding extension headers */
1413 len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1414 len -= skb_network_header_len(skb);
1417 * Upon reception of an MLD message that contains a Query, the node
1418 * checks if the source address of the message is a valid link-local
1419 * address, if the Hop Limit is set to 1, and if the Router Alert
1420 * option is present in the Hop-By-Hop Options header of the IPv6
1421 * packet. If any of these checks fails, the packet is dropped.
1423 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1424 ipv6_hdr(skb)->hop_limit != 1 ||
1425 !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1426 IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1429 idev = in6_dev_get(skb->dev);
1433 mld = (struct mld_msg *)icmp6_hdr(skb);
1434 group = &mld->mld_mca;
1435 group_type = ipv6_addr_type(group);
1437 if (group_type != IPV6_ADDR_ANY &&
1438 !(group_type&IPV6_ADDR_MULTICAST))
1441 if (len < MLD_V1_QUERY_LEN) {
1443 } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1444 err = mld_process_v1(idev, mld, &max_delay,
1445 len == MLD_V1_QUERY_LEN);
1448 } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1449 int srcs_offset = sizeof(struct mld2_query) -
1450 sizeof(struct icmp6hdr);
1452 if (!pskb_may_pull(skb, srcs_offset))
1455 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1457 mld_process_v2(idev, mlh2, &max_delay);
1459 if (group_type == IPV6_ADDR_ANY) { /* general query */
1460 if (mlh2->mld2q_nsrcs)
1461 goto out; /* no sources allowed */
1463 mld_gq_start_work(idev);
1466 /* mark sources to include, if group & source-specific */
1467 if (mlh2->mld2q_nsrcs != 0) {
1468 if (!pskb_may_pull(skb, srcs_offset +
1469 ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1472 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1479 if (group_type == IPV6_ADDR_ANY) {
1480 for_each_mc_mclock(idev, ma) {
1481 igmp6_group_queried(ma, max_delay);
1484 for_each_mc_mclock(idev, ma) {
1485 if (!ipv6_addr_equal(group, &ma->mca_addr))
1487 if (ma->mca_flags & MAF_TIMER_RUNNING) {
1488 /* gsquery <- gsquery && mark */
1490 ma->mca_flags &= ~MAF_GSQUERY;
1492 /* gsquery <- mark */
1494 ma->mca_flags |= MAF_GSQUERY;
1496 ma->mca_flags &= ~MAF_GSQUERY;
1498 if (!(ma->mca_flags & MAF_GSQUERY) ||
1499 mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1500 igmp6_group_queried(ma, max_delay);
1511 static void mld_query_work(struct work_struct *work)
1513 struct inet6_dev *idev = container_of(to_delayed_work(work),
1516 struct sk_buff_head q;
1517 struct sk_buff *skb;
1518 bool rework = false;
1521 skb_queue_head_init(&q);
1523 spin_lock_bh(&idev->mc_query_lock);
1524 while ((skb = __skb_dequeue(&idev->mc_query_queue))) {
1525 __skb_queue_tail(&q, skb);
1527 if (++cnt >= MLD_MAX_QUEUE) {
1529 schedule_delayed_work(&idev->mc_query_work, 0);
1533 spin_unlock_bh(&idev->mc_query_lock);
1535 mutex_lock(&idev->mc_lock);
1536 while ((skb = __skb_dequeue(&q)))
1537 __mld_query_work(skb);
1538 mutex_unlock(&idev->mc_lock);
1544 /* called with rcu_read_lock() */
1545 int igmp6_event_report(struct sk_buff *skb)
1547 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1557 spin_lock_bh(&idev->mc_report_lock);
1558 if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) {
1559 __skb_queue_tail(&idev->mc_report_queue, skb);
1560 if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0))
1563 spin_unlock_bh(&idev->mc_report_lock);
1568 static void __mld_report_work(struct sk_buff *skb)
1570 struct inet6_dev *idev;
1571 struct ifmcaddr6 *ma;
1572 struct mld_msg *mld;
1575 /* Our own report looped back. Ignore it. */
1576 if (skb->pkt_type == PACKET_LOOPBACK)
1579 /* send our report if the MC router may not have heard this report */
1580 if (skb->pkt_type != PACKET_MULTICAST &&
1581 skb->pkt_type != PACKET_BROADCAST)
1584 if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1587 mld = (struct mld_msg *)icmp6_hdr(skb);
1589 /* Drop reports with not link local source */
1590 addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1591 if (addr_type != IPV6_ADDR_ANY &&
1592 !(addr_type&IPV6_ADDR_LINKLOCAL))
1595 idev = in6_dev_get(skb->dev);
1600 * Cancel the work for this group
1603 for_each_mc_mclock(idev, ma) {
1604 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1605 if (cancel_delayed_work(&ma->mca_work))
1606 refcount_dec(&ma->mca_refcnt);
1607 ma->mca_flags &= ~(MAF_LAST_REPORTER |
1618 static void mld_report_work(struct work_struct *work)
1620 struct inet6_dev *idev = container_of(to_delayed_work(work),
1623 struct sk_buff_head q;
1624 struct sk_buff *skb;
1625 bool rework = false;
1628 skb_queue_head_init(&q);
1629 spin_lock_bh(&idev->mc_report_lock);
1630 while ((skb = __skb_dequeue(&idev->mc_report_queue))) {
1631 __skb_queue_tail(&q, skb);
1633 if (++cnt >= MLD_MAX_QUEUE) {
1635 schedule_delayed_work(&idev->mc_report_work, 0);
1639 spin_unlock_bh(&idev->mc_report_lock);
1641 mutex_lock(&idev->mc_lock);
1642 while ((skb = __skb_dequeue(&q)))
1643 __mld_report_work(skb);
1644 mutex_unlock(&idev->mc_lock);
1650 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1651 int gdeleted, int sdeleted)
1654 case MLD2_MODE_IS_INCLUDE:
1655 case MLD2_MODE_IS_EXCLUDE:
1656 if (gdeleted || sdeleted)
1658 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1659 if (pmc->mca_sfmode == MCAST_INCLUDE)
1661 /* don't include if this source is excluded
1664 if (psf->sf_count[MCAST_INCLUDE])
1665 return type == MLD2_MODE_IS_INCLUDE;
1666 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1667 psf->sf_count[MCAST_EXCLUDE];
1670 case MLD2_CHANGE_TO_INCLUDE:
1671 if (gdeleted || sdeleted)
1673 return psf->sf_count[MCAST_INCLUDE] != 0;
1674 case MLD2_CHANGE_TO_EXCLUDE:
1675 if (gdeleted || sdeleted)
1677 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1678 psf->sf_count[MCAST_INCLUDE])
1680 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1681 psf->sf_count[MCAST_EXCLUDE];
1682 case MLD2_ALLOW_NEW_SOURCES:
1683 if (gdeleted || !psf->sf_crcount)
1685 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1686 case MLD2_BLOCK_OLD_SOURCES:
1687 if (pmc->mca_sfmode == MCAST_INCLUDE)
1688 return gdeleted || (psf->sf_crcount && sdeleted);
1689 return psf->sf_crcount && !gdeleted && !sdeleted;
1695 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1697 struct ip6_sf_list *psf;
1700 for_each_psf_mclock(pmc, psf) {
1701 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1708 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1709 struct net_device *dev,
1710 const struct in6_addr *saddr,
1711 const struct in6_addr *daddr,
1714 struct ipv6hdr *hdr;
1716 skb->protocol = htons(ETH_P_IPV6);
1719 skb_reset_network_header(skb);
1720 skb_put(skb, sizeof(struct ipv6hdr));
1721 hdr = ipv6_hdr(skb);
1723 ip6_flow_hdr(hdr, 0, 0);
1725 hdr->payload_len = htons(len);
1726 hdr->nexthdr = proto;
1727 hdr->hop_limit = inet6_sk(sk)->hop_limit;
1729 hdr->saddr = *saddr;
1730 hdr->daddr = *daddr;
1733 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1735 u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
1736 2, 0, 0, IPV6_TLV_PADN, 0 };
1737 struct net_device *dev = idev->dev;
1738 int hlen = LL_RESERVED_SPACE(dev);
1739 int tlen = dev->needed_tailroom;
1740 struct net *net = dev_net(dev);
1741 const struct in6_addr *saddr;
1742 struct in6_addr addr_buf;
1743 struct mld2_report *pmr;
1744 struct sk_buff *skb;
1749 sk = net->ipv6.igmp_sk;
1750 /* we assume size > sizeof(ra) here
1751 * Also try to not allocate high-order pages for big MTU
1753 size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen;
1754 skb = sock_alloc_send_skb(sk, size, 1, &err);
1758 skb->priority = TC_PRIO_CONTROL;
1759 skb_reserve(skb, hlen);
1760 skb_tailroom_reserve(skb, mtu, tlen);
1762 if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1763 /* <draft-ietf-magma-mld-source-05.txt>:
1764 * use unspecified address as the source address
1765 * when a valid link-local address is not available.
1767 saddr = &in6addr_any;
1771 ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1773 skb_put_data(skb, ra, sizeof(ra));
1775 skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1776 skb_put(skb, sizeof(*pmr));
1777 pmr = (struct mld2_report *)skb_transport_header(skb);
1778 pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1779 pmr->mld2r_resv1 = 0;
1780 pmr->mld2r_cksum = 0;
1781 pmr->mld2r_resv2 = 0;
1782 pmr->mld2r_ngrec = 0;
1786 static void mld_sendpack(struct sk_buff *skb)
1788 struct ipv6hdr *pip6 = ipv6_hdr(skb);
1789 struct mld2_report *pmr =
1790 (struct mld2_report *)skb_transport_header(skb);
1791 int payload_len, mldlen;
1792 struct inet6_dev *idev;
1793 struct net *net = dev_net(skb->dev);
1796 struct dst_entry *dst;
1799 idev = __in6_dev_get(skb->dev);
1800 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1802 payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1804 mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1805 pip6->payload_len = htons(payload_len);
1807 pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1809 csum_partial(skb_transport_header(skb),
1812 icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1813 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1815 dst = icmp6_dst_alloc(skb->dev, &fl6);
1822 skb_dst_set(skb, dst);
1826 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1827 net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1831 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1832 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1834 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1845 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1847 return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1850 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1851 int type, struct mld2_grec **ppgr, unsigned int mtu)
1853 struct mld2_report *pmr;
1854 struct mld2_grec *pgr;
1857 skb = mld_newpack(pmc->idev, mtu);
1861 pgr = skb_put(skb, sizeof(struct mld2_grec));
1862 pgr->grec_type = type;
1863 pgr->grec_auxwords = 0;
1864 pgr->grec_nsrcs = 0;
1865 pgr->grec_mca = pmc->mca_addr; /* structure copy */
1866 pmr = (struct mld2_report *)skb_transport_header(skb);
1867 pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1872 #define AVAILABLE(skb) ((skb) ? skb_availroom(skb) : 0)
1874 /* called with mc_lock */
1875 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1876 int type, int gdeleted, int sdeleted,
1879 struct ip6_sf_list *psf, *psf_prev, *psf_next;
1880 int scount, stotal, first, isquery, truncate;
1881 struct ip6_sf_list __rcu **psf_list;
1882 struct inet6_dev *idev = pmc->idev;
1883 struct net_device *dev = idev->dev;
1884 struct mld2_grec *pgr = NULL;
1885 struct mld2_report *pmr;
1888 if (pmc->mca_flags & MAF_NOREPORT)
1891 mtu = READ_ONCE(dev->mtu);
1892 if (mtu < IPV6_MIN_MTU)
1895 isquery = type == MLD2_MODE_IS_INCLUDE ||
1896 type == MLD2_MODE_IS_EXCLUDE;
1897 truncate = type == MLD2_MODE_IS_EXCLUDE ||
1898 type == MLD2_CHANGE_TO_EXCLUDE;
1900 stotal = scount = 0;
1902 psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1904 if (!rcu_access_pointer(*psf_list))
1907 pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1909 /* EX and TO_EX get a fresh packet, if needed */
1911 if (pmr && pmr->mld2r_ngrec &&
1912 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1915 skb = mld_newpack(idev, mtu);
1920 for (psf = mc_dereference(*psf_list, idev);
1923 struct in6_addr *psrc;
1925 psf_next = mc_dereference(psf->sf_next, idev);
1927 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1932 /* Based on RFC3810 6.1. Should not send source-list change
1933 * records when there is a filter mode change.
1935 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1936 (!gdeleted && pmc->mca_crcount)) &&
1937 (type == MLD2_ALLOW_NEW_SOURCES ||
1938 type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1939 goto decrease_sf_crcount;
1941 /* clear marks on query responses */
1945 if (AVAILABLE(skb) < sizeof(*psrc) +
1946 first*sizeof(struct mld2_grec)) {
1947 if (truncate && !first)
1948 break; /* truncate these */
1950 pgr->grec_nsrcs = htons(scount);
1953 skb = mld_newpack(idev, mtu);
1958 skb = add_grhead(skb, pmc, type, &pgr, mtu);
1963 psrc = skb_put(skb, sizeof(*psrc));
1964 *psrc = psf->sf_addr;
1966 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1967 type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1968 decrease_sf_crcount:
1970 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1972 rcu_assign_pointer(psf_prev->sf_next,
1973 mc_dereference(psf->sf_next, idev));
1975 rcu_assign_pointer(*psf_list,
1976 mc_dereference(psf->sf_next, idev));
1977 kfree_rcu(psf, rcu);
1986 if (type == MLD2_ALLOW_NEW_SOURCES ||
1987 type == MLD2_BLOCK_OLD_SOURCES)
1989 if (pmc->mca_crcount || isquery || crsend) {
1990 /* make sure we have room for group header */
1991 if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1993 skb = NULL; /* add_grhead will get a new one */
1995 skb = add_grhead(skb, pmc, type, &pgr, mtu);
1999 pgr->grec_nsrcs = htons(scount);
2002 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
2006 /* called with mc_lock */
2007 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
2009 struct sk_buff *skb = NULL;
2013 for_each_mc_mclock(idev, pmc) {
2014 if (pmc->mca_flags & MAF_NOREPORT)
2016 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2017 type = MLD2_MODE_IS_EXCLUDE;
2019 type = MLD2_MODE_IS_INCLUDE;
2020 skb = add_grec(skb, pmc, type, 0, 0, 0);
2023 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2024 type = MLD2_MODE_IS_EXCLUDE;
2026 type = MLD2_MODE_IS_INCLUDE;
2027 skb = add_grec(skb, pmc, type, 0, 0, 0);
2034 * remove zero-count source records from a source filter list
2035 * called with mc_lock
2037 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
2039 struct ip6_sf_list *psf_prev, *psf_next, *psf;
2042 for (psf = mc_dereference(*ppsf, idev);
2045 psf_next = mc_dereference(psf->sf_next, idev);
2046 if (psf->sf_crcount == 0) {
2048 rcu_assign_pointer(psf_prev->sf_next,
2049 mc_dereference(psf->sf_next, idev));
2051 rcu_assign_pointer(*ppsf,
2052 mc_dereference(psf->sf_next, idev));
2053 kfree_rcu(psf, rcu);
2060 /* called with mc_lock */
2061 static void mld_send_cr(struct inet6_dev *idev)
2063 struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
2064 struct sk_buff *skb = NULL;
2069 for (pmc = mc_dereference(idev->mc_tomb, idev);
2072 pmc_next = mc_dereference(pmc->next, idev);
2073 if (pmc->mca_sfmode == MCAST_INCLUDE) {
2074 type = MLD2_BLOCK_OLD_SOURCES;
2075 dtype = MLD2_BLOCK_OLD_SOURCES;
2076 skb = add_grec(skb, pmc, type, 1, 0, 0);
2077 skb = add_grec(skb, pmc, dtype, 1, 1, 0);
2079 if (pmc->mca_crcount) {
2080 if (pmc->mca_sfmode == MCAST_EXCLUDE) {
2081 type = MLD2_CHANGE_TO_INCLUDE;
2082 skb = add_grec(skb, pmc, type, 1, 0, 0);
2085 if (pmc->mca_crcount == 0) {
2086 mld_clear_zeros(&pmc->mca_tomb, idev);
2087 mld_clear_zeros(&pmc->mca_sources, idev);
2090 if (pmc->mca_crcount == 0 &&
2091 !rcu_access_pointer(pmc->mca_tomb) &&
2092 !rcu_access_pointer(pmc->mca_sources)) {
2094 rcu_assign_pointer(pmc_prev->next, pmc_next);
2096 rcu_assign_pointer(idev->mc_tomb, pmc_next);
2097 in6_dev_put(pmc->idev);
2098 kfree_rcu(pmc, rcu);
2104 for_each_mc_mclock(idev, pmc) {
2105 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2106 type = MLD2_BLOCK_OLD_SOURCES;
2107 dtype = MLD2_ALLOW_NEW_SOURCES;
2109 type = MLD2_ALLOW_NEW_SOURCES;
2110 dtype = MLD2_BLOCK_OLD_SOURCES;
2112 skb = add_grec(skb, pmc, type, 0, 0, 0);
2113 skb = add_grec(skb, pmc, dtype, 0, 1, 0); /* deleted sources */
2115 /* filter mode changes */
2116 if (pmc->mca_crcount) {
2117 if (pmc->mca_sfmode == MCAST_EXCLUDE)
2118 type = MLD2_CHANGE_TO_EXCLUDE;
2120 type = MLD2_CHANGE_TO_INCLUDE;
2121 skb = add_grec(skb, pmc, type, 0, 0, 0);
2127 (void) mld_sendpack(skb);
2130 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
2132 struct net *net = dev_net(dev);
2133 struct sock *sk = net->ipv6.igmp_sk;
2134 struct inet6_dev *idev;
2135 struct sk_buff *skb;
2136 struct mld_msg *hdr;
2137 const struct in6_addr *snd_addr, *saddr;
2138 struct in6_addr addr_buf;
2139 int hlen = LL_RESERVED_SPACE(dev);
2140 int tlen = dev->needed_tailroom;
2141 int err, len, payload_len, full_len;
2142 u8 ra[8] = { IPPROTO_ICMPV6, 0,
2143 IPV6_TLV_ROUTERALERT, 2, 0, 0,
2146 struct dst_entry *dst;
2148 if (type == ICMPV6_MGM_REDUCTION)
2149 snd_addr = &in6addr_linklocal_allrouters;
2153 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2154 payload_len = len + sizeof(ra);
2155 full_len = sizeof(struct ipv6hdr) + payload_len;
2158 IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2159 IPSTATS_MIB_OUT, full_len);
2162 skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2166 IP6_INC_STATS(net, __in6_dev_get(dev),
2167 IPSTATS_MIB_OUTDISCARDS);
2171 skb->priority = TC_PRIO_CONTROL;
2172 skb_reserve(skb, hlen);
2174 if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2175 /* <draft-ietf-magma-mld-source-05.txt>:
2176 * use unspecified address as the source address
2177 * when a valid link-local address is not available.
2179 saddr = &in6addr_any;
2183 ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2185 skb_put_data(skb, ra, sizeof(ra));
2187 hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2188 hdr->mld_type = type;
2189 hdr->mld_mca = *addr;
2191 hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2193 csum_partial(hdr, len, 0));
2196 idev = __in6_dev_get(skb->dev);
2198 icmpv6_flow_init(sk, &fl6, type,
2199 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2201 dst = icmp6_dst_alloc(skb->dev, &fl6);
2207 skb_dst_set(skb, dst);
2208 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2209 net, sk, skb, NULL, skb->dev,
2213 ICMP6MSGOUT_INC_STATS(net, idev, type);
2214 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2216 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2226 /* called with mc_lock */
2227 static void mld_send_initial_cr(struct inet6_dev *idev)
2229 struct sk_buff *skb;
2230 struct ifmcaddr6 *pmc;
2233 if (mld_in_v1_mode(idev))
2237 for_each_mc_mclock(idev, pmc) {
2238 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2239 type = MLD2_CHANGE_TO_EXCLUDE;
2241 type = MLD2_ALLOW_NEW_SOURCES;
2242 skb = add_grec(skb, pmc, type, 0, 0, 1);
2248 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2250 mutex_lock(&idev->mc_lock);
2251 idev->mc_dad_count = idev->mc_qrv;
2252 if (idev->mc_dad_count) {
2253 mld_send_initial_cr(idev);
2254 idev->mc_dad_count--;
2255 if (idev->mc_dad_count)
2256 mld_dad_start_work(idev,
2257 unsolicited_report_interval(idev));
2259 mutex_unlock(&idev->mc_lock);
2262 static void mld_dad_work(struct work_struct *work)
2264 struct inet6_dev *idev = container_of(to_delayed_work(work),
2267 mutex_lock(&idev->mc_lock);
2268 mld_send_initial_cr(idev);
2269 if (idev->mc_dad_count) {
2270 idev->mc_dad_count--;
2271 if (idev->mc_dad_count)
2272 mld_dad_start_work(idev,
2273 unsolicited_report_interval(idev));
2275 mutex_unlock(&idev->mc_lock);
2279 /* called with mc_lock */
2280 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2281 const struct in6_addr *psfsrc)
2283 struct ip6_sf_list *psf, *psf_prev;
2287 for_each_psf_mclock(pmc, psf) {
2288 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2292 if (!psf || psf->sf_count[sfmode] == 0) {
2293 /* source filter not found, or count wrong => bug */
2296 psf->sf_count[sfmode]--;
2297 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2298 struct inet6_dev *idev = pmc->idev;
2300 /* no more filters for this source */
2302 rcu_assign_pointer(psf_prev->sf_next,
2303 mc_dereference(psf->sf_next, idev));
2305 rcu_assign_pointer(pmc->mca_sources,
2306 mc_dereference(psf->sf_next, idev));
2308 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2309 !mld_in_v1_mode(idev)) {
2310 psf->sf_crcount = idev->mc_qrv;
2311 rcu_assign_pointer(psf->sf_next,
2312 mc_dereference(pmc->mca_tomb, idev));
2313 rcu_assign_pointer(pmc->mca_tomb, psf);
2316 kfree_rcu(psf, rcu);
2322 /* called with mc_lock */
2323 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2324 int sfmode, int sfcount, const struct in6_addr *psfsrc,
2327 struct ifmcaddr6 *pmc;
2334 for_each_mc_mclock(idev, pmc) {
2335 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2343 if (!pmc->mca_sfcount[sfmode])
2346 pmc->mca_sfcount[sfmode]--;
2349 for (i = 0; i < sfcount; i++) {
2350 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2352 changerec |= rv > 0;
2356 if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2357 pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2358 pmc->mca_sfcount[MCAST_INCLUDE]) {
2359 struct ip6_sf_list *psf;
2361 /* filter mode change */
2362 pmc->mca_sfmode = MCAST_INCLUDE;
2363 pmc->mca_crcount = idev->mc_qrv;
2364 idev->mc_ifc_count = pmc->mca_crcount;
2365 for_each_psf_mclock(pmc, psf)
2366 psf->sf_crcount = 0;
2367 mld_ifc_event(pmc->idev);
2368 } else if (sf_setstate(pmc) || changerec) {
2369 mld_ifc_event(pmc->idev);
2376 * Add multicast single-source filter to the interface list
2377 * called with mc_lock
2379 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2380 const struct in6_addr *psfsrc)
2382 struct ip6_sf_list *psf, *psf_prev;
2385 for_each_psf_mclock(pmc, psf) {
2386 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2391 psf = kzalloc(sizeof(*psf), GFP_KERNEL);
2395 psf->sf_addr = *psfsrc;
2397 rcu_assign_pointer(psf_prev->sf_next, psf);
2399 rcu_assign_pointer(pmc->mca_sources, psf);
2402 psf->sf_count[sfmode]++;
2406 /* called with mc_lock */
2407 static void sf_markstate(struct ifmcaddr6 *pmc)
2409 struct ip6_sf_list *psf;
2410 int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2412 for_each_psf_mclock(pmc, psf) {
2413 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2414 psf->sf_oldin = mca_xcount ==
2415 psf->sf_count[MCAST_EXCLUDE] &&
2416 !psf->sf_count[MCAST_INCLUDE];
2418 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2423 /* called with mc_lock */
2424 static int sf_setstate(struct ifmcaddr6 *pmc)
2426 struct ip6_sf_list *psf, *dpsf;
2427 int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2428 int qrv = pmc->idev->mc_qrv;
2432 for_each_psf_mclock(pmc, psf) {
2433 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2434 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2435 !psf->sf_count[MCAST_INCLUDE];
2437 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2439 if (!psf->sf_oldin) {
2440 struct ip6_sf_list *prev = NULL;
2442 for_each_psf_tomb(pmc, dpsf) {
2443 if (ipv6_addr_equal(&dpsf->sf_addr,
2450 rcu_assign_pointer(prev->sf_next,
2451 mc_dereference(dpsf->sf_next,
2454 rcu_assign_pointer(pmc->mca_tomb,
2455 mc_dereference(dpsf->sf_next,
2457 kfree_rcu(dpsf, rcu);
2459 psf->sf_crcount = qrv;
2462 } else if (psf->sf_oldin) {
2463 psf->sf_crcount = 0;
2465 * add or update "delete" records if an active filter
2469 for_each_psf_tomb(pmc, dpsf)
2470 if (ipv6_addr_equal(&dpsf->sf_addr,
2474 dpsf = kmalloc(sizeof(*dpsf), GFP_KERNEL);
2478 rcu_assign_pointer(dpsf->sf_next,
2479 mc_dereference(pmc->mca_tomb, pmc->idev));
2480 rcu_assign_pointer(pmc->mca_tomb, dpsf);
2482 dpsf->sf_crcount = qrv;
2490 * Add multicast source filter list to the interface list
2491 * called with mc_lock
2493 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2494 int sfmode, int sfcount, const struct in6_addr *psfsrc,
2497 struct ifmcaddr6 *pmc;
2504 for_each_mc_mclock(idev, pmc) {
2505 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2512 isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2514 pmc->mca_sfcount[sfmode]++;
2516 for (i = 0; i < sfcount; i++) {
2517 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2525 pmc->mca_sfcount[sfmode]--;
2526 for (j = 0; j < i; j++)
2527 ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2528 } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2529 struct ip6_sf_list *psf;
2531 /* filter mode change */
2532 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2533 pmc->mca_sfmode = MCAST_EXCLUDE;
2534 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2535 pmc->mca_sfmode = MCAST_INCLUDE;
2536 /* else no filters; keep old mode for reports */
2538 pmc->mca_crcount = idev->mc_qrv;
2539 idev->mc_ifc_count = pmc->mca_crcount;
2540 for_each_psf_mclock(pmc, psf)
2541 psf->sf_crcount = 0;
2542 mld_ifc_event(idev);
2543 } else if (sf_setstate(pmc)) {
2544 mld_ifc_event(idev);
2549 /* called with mc_lock */
2550 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2552 struct ip6_sf_list *psf, *nextpsf;
2554 for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
2557 nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2558 kfree_rcu(psf, rcu);
2560 RCU_INIT_POINTER(pmc->mca_tomb, NULL);
2561 for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
2564 nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2565 kfree_rcu(psf, rcu);
2567 RCU_INIT_POINTER(pmc->mca_sources, NULL);
2568 pmc->mca_sfmode = MCAST_EXCLUDE;
2569 pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2570 pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2573 /* called with mc_lock */
2574 static void igmp6_join_group(struct ifmcaddr6 *ma)
2576 unsigned long delay;
2578 if (ma->mca_flags & MAF_NOREPORT)
2581 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2583 delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2585 if (cancel_delayed_work(&ma->mca_work)) {
2586 refcount_dec(&ma->mca_refcnt);
2587 delay = ma->mca_work.timer.expires - jiffies;
2590 if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
2591 refcount_inc(&ma->mca_refcnt);
2592 ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2595 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2596 struct inet6_dev *idev)
2598 struct ip6_sf_socklist *psl;
2601 psl = sock_dereference(iml->sflist, sk);
2604 mutex_lock(&idev->mc_lock);
2607 /* any-source empty exclude case */
2608 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2610 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2611 psl->sl_count, psl->sl_addr, 0);
2612 RCU_INIT_POINTER(iml->sflist, NULL);
2613 atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
2614 &sk->sk_omem_alloc);
2615 kfree_rcu(psl, rcu);
2619 mutex_unlock(&idev->mc_lock);
2624 /* called with mc_lock */
2625 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2627 if (mld_in_v1_mode(ma->idev)) {
2628 if (ma->mca_flags & MAF_LAST_REPORTER) {
2629 igmp6_send(&ma->mca_addr, ma->idev->dev,
2630 ICMPV6_MGM_REDUCTION);
2633 mld_add_delrec(ma->idev, ma);
2634 mld_ifc_event(ma->idev);
2638 static void mld_gq_work(struct work_struct *work)
2640 struct inet6_dev *idev = container_of(to_delayed_work(work),
2644 mutex_lock(&idev->mc_lock);
2645 mld_send_report(idev, NULL);
2646 idev->mc_gq_running = 0;
2647 mutex_unlock(&idev->mc_lock);
2652 static void mld_ifc_work(struct work_struct *work)
2654 struct inet6_dev *idev = container_of(to_delayed_work(work),
2658 mutex_lock(&idev->mc_lock);
2661 if (idev->mc_ifc_count) {
2662 idev->mc_ifc_count--;
2663 if (idev->mc_ifc_count)
2664 mld_ifc_start_work(idev,
2665 unsolicited_report_interval(idev));
2667 mutex_unlock(&idev->mc_lock);
2671 /* called with mc_lock */
2672 static void mld_ifc_event(struct inet6_dev *idev)
2674 if (mld_in_v1_mode(idev))
2677 idev->mc_ifc_count = idev->mc_qrv;
2678 mld_ifc_start_work(idev, 1);
2681 static void mld_mca_work(struct work_struct *work)
2683 struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
2684 struct ifmcaddr6, mca_work);
2686 mutex_lock(&ma->idev->mc_lock);
2687 if (mld_in_v1_mode(ma->idev))
2688 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2690 mld_send_report(ma->idev, ma);
2691 ma->mca_flags |= MAF_LAST_REPORTER;
2692 ma->mca_flags &= ~MAF_TIMER_RUNNING;
2693 mutex_unlock(&ma->idev->mc_lock);
2698 /* Device changing type */
2700 void ipv6_mc_unmap(struct inet6_dev *idev)
2702 struct ifmcaddr6 *i;
2704 /* Install multicast list, except for all-nodes (already installed) */
2706 mutex_lock(&idev->mc_lock);
2707 for_each_mc_mclock(idev, i)
2708 igmp6_group_dropped(i);
2709 mutex_unlock(&idev->mc_lock);
2712 void ipv6_mc_remap(struct inet6_dev *idev)
2717 /* Device going down */
2718 void ipv6_mc_down(struct inet6_dev *idev)
2720 struct ifmcaddr6 *i;
2722 mutex_lock(&idev->mc_lock);
2723 /* Withdraw multicast list */
2724 for_each_mc_mclock(idev, i)
2725 igmp6_group_dropped(i);
2726 mutex_unlock(&idev->mc_lock);
2728 /* Should stop work after group drop. or we will
2729 * start work again in mld_ifc_event()
2732 mld_query_stop_work(idev);
2733 mld_report_stop_work(idev);
2734 mld_ifc_stop_work(idev);
2735 mld_gq_stop_work(idev);
2736 mld_dad_stop_work(idev);
2739 static void ipv6_mc_reset(struct inet6_dev *idev)
2741 idev->mc_qrv = sysctl_mld_qrv;
2742 idev->mc_qi = MLD_QI_DEFAULT;
2743 idev->mc_qri = MLD_QRI_DEFAULT;
2744 idev->mc_v1_seen = 0;
2745 idev->mc_maxdelay = unsolicited_report_interval(idev);
2748 /* Device going up */
2750 void ipv6_mc_up(struct inet6_dev *idev)
2752 struct ifmcaddr6 *i;
2754 /* Install multicast list, except for all-nodes (already installed) */
2756 ipv6_mc_reset(idev);
2757 mutex_lock(&idev->mc_lock);
2758 for_each_mc_mclock(idev, i) {
2759 mld_del_delrec(idev, i);
2760 igmp6_group_added(i);
2762 mutex_unlock(&idev->mc_lock);
2765 /* IPv6 device initialization. */
2767 void ipv6_mc_init_dev(struct inet6_dev *idev)
2769 idev->mc_gq_running = 0;
2770 INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work);
2771 RCU_INIT_POINTER(idev->mc_tomb, NULL);
2772 idev->mc_ifc_count = 0;
2773 INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work);
2774 INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work);
2775 INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work);
2776 INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work);
2777 skb_queue_head_init(&idev->mc_query_queue);
2778 skb_queue_head_init(&idev->mc_report_queue);
2779 spin_lock_init(&idev->mc_query_lock);
2780 spin_lock_init(&idev->mc_report_lock);
2781 mutex_init(&idev->mc_lock);
2782 ipv6_mc_reset(idev);
2786 * Device is about to be destroyed: clean up.
2789 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2791 struct ifmcaddr6 *i;
2793 /* Deactivate works */
2795 mutex_lock(&idev->mc_lock);
2796 mld_clear_delrec(idev);
2797 mutex_unlock(&idev->mc_lock);
2798 mld_clear_query(idev);
2799 mld_clear_report(idev);
2801 /* Delete all-nodes address. */
2802 /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2803 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2806 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2808 if (idev->cnf.forwarding)
2809 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2811 mutex_lock(&idev->mc_lock);
2812 while ((i = mc_dereference(idev->mc_list, idev))) {
2813 rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
2815 ip6_mc_clear_src(i);
2818 mutex_unlock(&idev->mc_lock);
2821 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2823 struct ifmcaddr6 *pmc;
2827 mutex_lock(&idev->mc_lock);
2828 if (mld_in_v1_mode(idev)) {
2829 for_each_mc_mclock(idev, pmc)
2830 igmp6_join_group(pmc);
2832 mld_send_report(idev, NULL);
2834 mutex_unlock(&idev->mc_lock);
2837 static int ipv6_mc_netdev_event(struct notifier_block *this,
2838 unsigned long event,
2841 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2842 struct inet6_dev *idev = __in6_dev_get(dev);
2845 case NETDEV_RESEND_IGMP:
2847 ipv6_mc_rejoin_groups(idev);
2856 static struct notifier_block igmp6_netdev_notifier = {
2857 .notifier_call = ipv6_mc_netdev_event,
2860 #ifdef CONFIG_PROC_FS
2861 struct igmp6_mc_iter_state {
2862 struct seq_net_private p;
2863 struct net_device *dev;
2864 struct inet6_dev *idev;
2867 #define igmp6_mc_seq_private(seq) ((struct igmp6_mc_iter_state *)(seq)->private)
2869 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2871 struct ifmcaddr6 *im = NULL;
2872 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2873 struct net *net = seq_file_net(seq);
2876 for_each_netdev_rcu(net, state->dev) {
2877 struct inet6_dev *idev;
2878 idev = __in6_dev_get(state->dev);
2882 im = rcu_dereference(idev->mc_list);
2891 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2893 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2895 im = rcu_dereference(im->next);
2897 state->dev = next_net_device_rcu(state->dev);
2902 state->idev = __in6_dev_get(state->dev);
2905 im = rcu_dereference(state->idev->mc_list);
2910 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2912 struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2914 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2916 return pos ? NULL : im;
2919 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2923 return igmp6_mc_get_idx(seq, *pos);
2926 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2928 struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2934 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2937 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2939 if (likely(state->idev))
2945 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2947 struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2948 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2951 "%-4d %-15s %pi6 %5d %08X %ld\n",
2952 state->dev->ifindex, state->dev->name,
2954 im->mca_users, im->mca_flags,
2955 (im->mca_flags & MAF_TIMER_RUNNING) ?
2956 jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0);
2960 static const struct seq_operations igmp6_mc_seq_ops = {
2961 .start = igmp6_mc_seq_start,
2962 .next = igmp6_mc_seq_next,
2963 .stop = igmp6_mc_seq_stop,
2964 .show = igmp6_mc_seq_show,
2967 struct igmp6_mcf_iter_state {
2968 struct seq_net_private p;
2969 struct net_device *dev;
2970 struct inet6_dev *idev;
2971 struct ifmcaddr6 *im;
2974 #define igmp6_mcf_seq_private(seq) ((struct igmp6_mcf_iter_state *)(seq)->private)
2976 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2978 struct ip6_sf_list *psf = NULL;
2979 struct ifmcaddr6 *im = NULL;
2980 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2981 struct net *net = seq_file_net(seq);
2985 for_each_netdev_rcu(net, state->dev) {
2986 struct inet6_dev *idev;
2987 idev = __in6_dev_get(state->dev);
2988 if (unlikely(idev == NULL))
2991 im = rcu_dereference(idev->mc_list);
2993 psf = rcu_dereference(im->mca_sources);
3004 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
3006 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3008 psf = rcu_dereference(psf->sf_next);
3010 state->im = rcu_dereference(state->im->next);
3011 while (!state->im) {
3012 state->dev = next_net_device_rcu(state->dev);
3017 state->idev = __in6_dev_get(state->dev);
3020 state->im = rcu_dereference(state->idev->mc_list);
3024 psf = rcu_dereference(state->im->mca_sources);
3030 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
3032 struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
3034 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
3036 return pos ? NULL : psf;
3039 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
3043 return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
3046 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3048 struct ip6_sf_list *psf;
3049 if (v == SEQ_START_TOKEN)
3050 psf = igmp6_mcf_get_first(seq);
3052 psf = igmp6_mcf_get_next(seq, v);
3057 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
3060 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3062 if (likely(state->im))
3064 if (likely(state->idev))
3071 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
3073 struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
3074 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3076 if (v == SEQ_START_TOKEN) {
3077 seq_puts(seq, "Idx Device Multicast Address Source Address INC EXC\n");
3080 "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
3081 state->dev->ifindex, state->dev->name,
3082 &state->im->mca_addr,
3084 psf->sf_count[MCAST_INCLUDE],
3085 psf->sf_count[MCAST_EXCLUDE]);
3090 static const struct seq_operations igmp6_mcf_seq_ops = {
3091 .start = igmp6_mcf_seq_start,
3092 .next = igmp6_mcf_seq_next,
3093 .stop = igmp6_mcf_seq_stop,
3094 .show = igmp6_mcf_seq_show,
3097 static int __net_init igmp6_proc_init(struct net *net)
3102 if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
3103 sizeof(struct igmp6_mc_iter_state)))
3105 if (!proc_create_net("mcfilter6", 0444, net->proc_net,
3107 sizeof(struct igmp6_mcf_iter_state)))
3108 goto out_proc_net_igmp6;
3115 remove_proc_entry("igmp6", net->proc_net);
3119 static void __net_exit igmp6_proc_exit(struct net *net)
3121 remove_proc_entry("mcfilter6", net->proc_net);
3122 remove_proc_entry("igmp6", net->proc_net);
3125 static inline int igmp6_proc_init(struct net *net)
3129 static inline void igmp6_proc_exit(struct net *net)
3134 static int __net_init igmp6_net_init(struct net *net)
3138 err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
3139 SOCK_RAW, IPPROTO_ICMPV6, net);
3141 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
3146 inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
3147 net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL;
3149 err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
3150 SOCK_RAW, IPPROTO_ICMPV6, net);
3152 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
3154 goto out_sock_create;
3157 err = igmp6_proc_init(net);
3159 goto out_sock_create_autojoin;
3163 out_sock_create_autojoin:
3164 inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3166 inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3171 static void __net_exit igmp6_net_exit(struct net *net)
3173 inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3174 inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3175 igmp6_proc_exit(net);
3178 static struct pernet_operations igmp6_net_ops = {
3179 .init = igmp6_net_init,
3180 .exit = igmp6_net_exit,
3183 int __init igmp6_init(void)
3187 err = register_pernet_subsys(&igmp6_net_ops);
3191 mld_wq = create_workqueue("mld");
3193 unregister_pernet_subsys(&igmp6_net_ops);
3200 int __init igmp6_late_init(void)
3202 return register_netdevice_notifier(&igmp6_netdev_notifier);
3205 void igmp6_cleanup(void)
3207 unregister_pernet_subsys(&igmp6_net_ops);
3208 destroy_workqueue(mld_wq);
3211 void igmp6_late_cleanup(void)
3213 unregister_netdevice_notifier(&igmp6_netdev_notifier);