2 * Linux NET3: Internet Group Management Protocol [IGMP]
4 * This code implements the IGMP protocol as defined in RFC1112. There has
5 * been a further revision of this protocol since which is now supported.
7 * If you have trouble with this module be careful what gcc you have used,
8 * the older version didn't come out right using gcc 2.5.8, the newer one
9 * seems to fall out with gcc 2.6.2.
12 * Alan Cox <alan@lxorguk.ukuu.org.uk>
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 * Alan Cox : Added lots of __inline__ to optimise
22 * the memory usage of all the tiny little
24 * Alan Cox : Dumped the header building experiment.
25 * Alan Cox : Minor tweaks ready for multicast routing
26 * and extended IGMP protocol.
27 * Alan Cox : Removed a load of inline directives. Gcc 2.5.8
28 * writes utterly bogus code otherwise (sigh)
29 * fixed IGMP loopback to behave in the manner
30 * desired by mrouted, fixed the fact it has been
31 * broken since 1.3.6 and cleaned up a few minor
34 * Chih-Jen Chang : Tried to revise IGMP to Version 2
35 * Tsu-Sheng Tsao E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
36 * The enhancements are mainly based on Steve Deering's
37 * ipmulti-3.5 source code.
38 * Chih-Jen Chang : Added the igmp_get_mrouter_info and
39 * Tsu-Sheng Tsao igmp_set_mrouter_info to keep track of
40 * the mrouted version on that device.
41 * Chih-Jen Chang : Added the max_resp_time parameter to
42 * Tsu-Sheng Tsao igmp_heard_query(). Using this parameter
43 * to identify the multicast router version
44 * and do what the IGMP version 2 specified.
45 * Chih-Jen Chang : Added a timer to revert to IGMP V2 router
46 * Tsu-Sheng Tsao if the specified time expired.
47 * Alan Cox : Stop IGMP from 0.0.0.0 being accepted.
48 * Alan Cox : Use GFP_ATOMIC in the right places.
49 * Christian Daudt : igmp timer wasn't set for local group
50 * memberships but was being deleted,
51 * which caused a "del_timer() called
52 * from %p with timer not initialized\n"
54 * Christian Daudt : removed del_timer from
55 * igmp_timer_expire function (960205).
56 * Christian Daudt : igmp_heard_report now only calls
57 * igmp_timer_expire if tm->running is
59 * Malcolm Beattie : ttl comparison wrong in igmp_rcv made
60 * igmp_heard_query never trigger. Expiry
61 * miscalculation fixed in igmp_heard_query
62 * and random() made to return unsigned to
63 * prevent negative expiry times.
64 * Alexey Kuznetsov: Wrong group leaving behaviour, backport
65 * fix from pending 2.1.x patches.
66 * Alan Cox: Forget to enable FDDI support earlier.
67 * Alexey Kuznetsov: Fixed leaving groups on device down.
68 * Alexey Kuznetsov: Accordance to igmp-v2-06 draft.
69 * David L Stevens: IGMPv3 support, with help from
73 #include <linux/module.h>
74 #include <linux/slab.h>
75 #include <linux/uaccess.h>
76 #include <linux/types.h>
77 #include <linux/kernel.h>
78 #include <linux/jiffies.h>
79 #include <linux/string.h>
80 #include <linux/socket.h>
81 #include <linux/sockios.h>
83 #include <linux/inet.h>
84 #include <linux/netdevice.h>
85 #include <linux/skbuff.h>
86 #include <linux/inetdevice.h>
87 #include <linux/igmp.h>
88 #include <linux/if_arp.h>
89 #include <linux/rtnetlink.h>
90 #include <linux/times.h>
91 #include <linux/pkt_sched.h>
92 #include <linux/byteorder/generic.h>
94 #include <net/net_namespace.h>
97 #include <net/protocol.h>
98 #include <net/route.h>
100 #include <net/checksum.h>
101 #include <net/inet_common.h>
102 #include <linux/netfilter_ipv4.h>
103 #ifdef CONFIG_IP_MROUTE
104 #include <linux/mroute.h>
106 #ifdef CONFIG_PROC_FS
107 #include <linux/proc_fs.h>
108 #include <linux/seq_file.h>
111 #ifdef CONFIG_IP_MULTICAST
112 /* Parameter names and values are taken from igmp-v2-06 draft */
114 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT (400*HZ)
115 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT (400*HZ)
116 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL (10*HZ)
117 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL (1*HZ)
118 #define IGMP_QUERY_RESPONSE_INTERVAL (10*HZ)
119 #define IGMP_QUERY_ROBUSTNESS_VARIABLE 2
122 #define IGMP_INITIAL_REPORT_DELAY (1)
124 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
125 * IGMP specs require to report membership immediately after
126 * joining a group, but we delay the first report by a
127 * small interval. It seems more natural and still does not
128 * contradict to specs provided this delay is small enough.
131 #define IGMP_V1_SEEN(in_dev) \
132 (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
133 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
134 ((in_dev)->mr_v1_seen && \
135 time_before(jiffies, (in_dev)->mr_v1_seen)))
136 #define IGMP_V2_SEEN(in_dev) \
137 (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
138 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
139 ((in_dev)->mr_v2_seen && \
140 time_before(jiffies, (in_dev)->mr_v2_seen)))
142 static int unsolicited_report_interval(struct in_device *in_dev)
144 int interval_ms, interval_jiffies;
146 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
147 interval_ms = IN_DEV_CONF_GET(
149 IGMPV2_UNSOLICITED_REPORT_INTERVAL);
151 interval_ms = IN_DEV_CONF_GET(
153 IGMPV3_UNSOLICITED_REPORT_INTERVAL);
155 interval_jiffies = msecs_to_jiffies(interval_ms);
157 /* _timer functions can't handle a delay of 0 jiffies so ensure
158 * we always return a positive value.
160 if (interval_jiffies <= 0)
161 interval_jiffies = 1;
162 return interval_jiffies;
165 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
166 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im);
167 static void igmpv3_clear_delrec(struct in_device *in_dev);
168 static int sf_setstate(struct ip_mc_list *pmc);
169 static void sf_markstate(struct ip_mc_list *pmc);
171 static void ip_mc_clear_src(struct ip_mc_list *pmc);
172 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
173 int sfcount, __be32 *psfsrc, int delta);
175 static void ip_ma_put(struct ip_mc_list *im)
177 if (refcount_dec_and_test(&im->refcnt)) {
178 in_dev_put(im->interface);
183 #define for_each_pmc_rcu(in_dev, pmc) \
184 for (pmc = rcu_dereference(in_dev->mc_list); \
186 pmc = rcu_dereference(pmc->next_rcu))
188 #define for_each_pmc_rtnl(in_dev, pmc) \
189 for (pmc = rtnl_dereference(in_dev->mc_list); \
191 pmc = rtnl_dereference(pmc->next_rcu))
193 static void ip_sf_list_clear_all(struct ip_sf_list *psf)
195 struct ip_sf_list *next;
204 #ifdef CONFIG_IP_MULTICAST
210 static void igmp_stop_timer(struct ip_mc_list *im)
212 spin_lock_bh(&im->lock);
213 if (del_timer(&im->timer))
214 refcount_dec(&im->refcnt);
217 im->unsolicit_count = 0;
218 spin_unlock_bh(&im->lock);
221 /* It must be called with locked im->lock */
222 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
224 int tv = prandom_u32() % max_delay;
227 if (!mod_timer(&im->timer, jiffies+tv+2))
228 refcount_inc(&im->refcnt);
231 static void igmp_gq_start_timer(struct in_device *in_dev)
233 int tv = prandom_u32() % in_dev->mr_maxdelay;
234 unsigned long exp = jiffies + tv + 2;
236 if (in_dev->mr_gq_running &&
237 time_after_eq(exp, (in_dev->mr_gq_timer).expires))
240 in_dev->mr_gq_running = 1;
241 if (!mod_timer(&in_dev->mr_gq_timer, exp))
245 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
247 int tv = prandom_u32() % delay;
249 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
253 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
255 spin_lock_bh(&im->lock);
256 im->unsolicit_count = 0;
257 if (del_timer(&im->timer)) {
258 if ((long)(im->timer.expires-jiffies) < max_delay) {
259 add_timer(&im->timer);
261 spin_unlock_bh(&im->lock);
264 refcount_dec(&im->refcnt);
266 igmp_start_timer(im, max_delay);
267 spin_unlock_bh(&im->lock);
272 * Send an IGMP report.
275 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
278 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
279 int gdeleted, int sdeleted)
282 case IGMPV3_MODE_IS_INCLUDE:
283 case IGMPV3_MODE_IS_EXCLUDE:
284 if (gdeleted || sdeleted)
286 if (!(pmc->gsquery && !psf->sf_gsresp)) {
287 if (pmc->sfmode == MCAST_INCLUDE)
289 /* don't include if this source is excluded
292 if (psf->sf_count[MCAST_INCLUDE])
293 return type == IGMPV3_MODE_IS_INCLUDE;
294 return pmc->sfcount[MCAST_EXCLUDE] ==
295 psf->sf_count[MCAST_EXCLUDE];
298 case IGMPV3_CHANGE_TO_INCLUDE:
299 if (gdeleted || sdeleted)
301 return psf->sf_count[MCAST_INCLUDE] != 0;
302 case IGMPV3_CHANGE_TO_EXCLUDE:
303 if (gdeleted || sdeleted)
305 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
306 psf->sf_count[MCAST_INCLUDE])
308 return pmc->sfcount[MCAST_EXCLUDE] ==
309 psf->sf_count[MCAST_EXCLUDE];
310 case IGMPV3_ALLOW_NEW_SOURCES:
311 if (gdeleted || !psf->sf_crcount)
313 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
314 case IGMPV3_BLOCK_OLD_SOURCES:
315 if (pmc->sfmode == MCAST_INCLUDE)
316 return gdeleted || (psf->sf_crcount && sdeleted);
317 return psf->sf_crcount && !gdeleted && !sdeleted;
323 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
325 struct ip_sf_list *psf;
328 for (psf = pmc->sources; psf; psf = psf->sf_next) {
329 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
336 /* source address selection per RFC 3376 section 4.2.13 */
337 static __be32 igmpv3_get_srcaddr(struct net_device *dev,
338 const struct flowi4 *fl4)
340 struct in_device *in_dev = __in_dev_get_rcu(dev);
343 return htonl(INADDR_ANY);
346 if (fl4->saddr == ifa->ifa_local)
348 } endfor_ifa(in_dev);
350 return htonl(INADDR_ANY);
353 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
358 struct igmpv3_report *pig;
359 struct net *net = dev_net(dev);
361 int hlen = LL_RESERVED_SPACE(dev);
362 int tlen = dev->needed_tailroom;
363 unsigned int size = mtu;
366 skb = alloc_skb(size + hlen + tlen,
367 GFP_ATOMIC | __GFP_NOWARN);
374 skb->priority = TC_PRIO_CONTROL;
376 rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
378 IPPROTO_IGMP, 0, dev->ifindex);
384 skb_dst_set(skb, &rt->dst);
387 skb_reserve(skb, hlen);
388 skb_tailroom_reserve(skb, mtu, tlen);
390 skb_reset_network_header(skb);
392 skb_put(skb, sizeof(struct iphdr) + 4);
395 pip->ihl = (sizeof(struct iphdr)+4)>>2;
397 pip->frag_off = htons(IP_DF);
399 pip->daddr = fl4.daddr;
402 pip->saddr = igmpv3_get_srcaddr(dev, &fl4);
405 pip->protocol = IPPROTO_IGMP;
406 pip->tot_len = 0; /* filled in later */
407 ip_select_ident(net, skb, NULL);
408 ((u8 *)&pip[1])[0] = IPOPT_RA;
409 ((u8 *)&pip[1])[1] = 4;
410 ((u8 *)&pip[1])[2] = 0;
411 ((u8 *)&pip[1])[3] = 0;
413 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
414 skb_put(skb, sizeof(*pig));
415 pig = igmpv3_report_hdr(skb);
416 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
424 static int igmpv3_sendpack(struct sk_buff *skb)
426 struct igmphdr *pig = igmp_hdr(skb);
427 const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
429 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
431 return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
434 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
436 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
439 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
440 int type, struct igmpv3_grec **ppgr, unsigned int mtu)
442 struct net_device *dev = pmc->interface->dev;
443 struct igmpv3_report *pih;
444 struct igmpv3_grec *pgr;
447 skb = igmpv3_newpack(dev, mtu);
451 pgr = skb_put(skb, sizeof(struct igmpv3_grec));
452 pgr->grec_type = type;
453 pgr->grec_auxwords = 0;
455 pgr->grec_mca = pmc->multiaddr;
456 pih = igmpv3_report_hdr(skb);
457 pih->ngrec = htons(ntohs(pih->ngrec)+1);
462 #define AVAILABLE(skb) ((skb) ? skb_availroom(skb) : 0)
464 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
465 int type, int gdeleted, int sdeleted)
467 struct net_device *dev = pmc->interface->dev;
468 struct net *net = dev_net(dev);
469 struct igmpv3_report *pih;
470 struct igmpv3_grec *pgr = NULL;
471 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
472 int scount, stotal, first, isquery, truncate;
475 if (pmc->multiaddr == IGMP_ALL_HOSTS)
477 if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
480 mtu = READ_ONCE(dev->mtu);
481 if (mtu < IPV4_MIN_MTU)
484 isquery = type == IGMPV3_MODE_IS_INCLUDE ||
485 type == IGMPV3_MODE_IS_EXCLUDE;
486 truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
487 type == IGMPV3_CHANGE_TO_EXCLUDE;
491 psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
496 pih = skb ? igmpv3_report_hdr(skb) : NULL;
498 /* EX and TO_EX get a fresh packet, if needed */
500 if (pih && pih->ngrec &&
501 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
503 igmpv3_sendpack(skb);
504 skb = igmpv3_newpack(dev, mtu);
509 for (psf = *psf_list; psf; psf = psf_next) {
512 psf_next = psf->sf_next;
514 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
519 /* Based on RFC3376 5.1. Should not send source-list change
520 * records when there is a filter mode change.
522 if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) ||
523 (!gdeleted && pmc->crcount)) &&
524 (type == IGMPV3_ALLOW_NEW_SOURCES ||
525 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount)
526 goto decrease_sf_crcount;
528 /* clear marks on query responses */
532 if (AVAILABLE(skb) < sizeof(__be32) +
533 first*sizeof(struct igmpv3_grec)) {
534 if (truncate && !first)
535 break; /* truncate these */
537 pgr->grec_nsrcs = htons(scount);
539 igmpv3_sendpack(skb);
540 skb = igmpv3_newpack(dev, mtu);
545 skb = add_grhead(skb, pmc, type, &pgr, mtu);
550 psrc = skb_put(skb, sizeof(__be32));
551 *psrc = psf->sf_inaddr;
553 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
554 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
557 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
559 psf_prev->sf_next = psf->sf_next;
561 *psf_list = psf->sf_next;
571 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
572 type == IGMPV3_BLOCK_OLD_SOURCES)
574 if (pmc->crcount || isquery) {
575 /* make sure we have room for group header */
576 if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
577 igmpv3_sendpack(skb);
578 skb = NULL; /* add_grhead will get a new one */
580 skb = add_grhead(skb, pmc, type, &pgr, mtu);
584 pgr->grec_nsrcs = htons(scount);
587 pmc->gsquery = 0; /* clear query state on report */
591 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
593 struct sk_buff *skb = NULL;
594 struct net *net = dev_net(in_dev->dev);
599 for_each_pmc_rcu(in_dev, pmc) {
600 if (pmc->multiaddr == IGMP_ALL_HOSTS)
602 if (ipv4_is_local_multicast(pmc->multiaddr) &&
603 !net->ipv4.sysctl_igmp_llm_reports)
605 spin_lock_bh(&pmc->lock);
606 if (pmc->sfcount[MCAST_EXCLUDE])
607 type = IGMPV3_MODE_IS_EXCLUDE;
609 type = IGMPV3_MODE_IS_INCLUDE;
610 skb = add_grec(skb, pmc, type, 0, 0);
611 spin_unlock_bh(&pmc->lock);
615 spin_lock_bh(&pmc->lock);
616 if (pmc->sfcount[MCAST_EXCLUDE])
617 type = IGMPV3_MODE_IS_EXCLUDE;
619 type = IGMPV3_MODE_IS_INCLUDE;
620 skb = add_grec(skb, pmc, type, 0, 0);
621 spin_unlock_bh(&pmc->lock);
625 return igmpv3_sendpack(skb);
629 * remove zero-count source records from a source filter list
631 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
633 struct ip_sf_list *psf_prev, *psf_next, *psf;
636 for (psf = *ppsf; psf; psf = psf_next) {
637 psf_next = psf->sf_next;
638 if (psf->sf_crcount == 0) {
640 psf_prev->sf_next = psf->sf_next;
642 *ppsf = psf->sf_next;
649 static void kfree_pmc(struct ip_mc_list *pmc)
651 ip_sf_list_clear_all(pmc->sources);
652 ip_sf_list_clear_all(pmc->tomb);
656 static void igmpv3_send_cr(struct in_device *in_dev)
658 struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
659 struct sk_buff *skb = NULL;
663 spin_lock_bh(&in_dev->mc_tomb_lock);
667 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
668 pmc_next = pmc->next;
669 if (pmc->sfmode == MCAST_INCLUDE) {
670 type = IGMPV3_BLOCK_OLD_SOURCES;
671 dtype = IGMPV3_BLOCK_OLD_SOURCES;
672 skb = add_grec(skb, pmc, type, 1, 0);
673 skb = add_grec(skb, pmc, dtype, 1, 1);
676 if (pmc->sfmode == MCAST_EXCLUDE) {
677 type = IGMPV3_CHANGE_TO_INCLUDE;
678 skb = add_grec(skb, pmc, type, 1, 0);
681 if (pmc->crcount == 0) {
682 igmpv3_clear_zeros(&pmc->tomb);
683 igmpv3_clear_zeros(&pmc->sources);
686 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
688 pmc_prev->next = pmc_next;
690 in_dev->mc_tomb = pmc_next;
691 in_dev_put(pmc->interface);
696 spin_unlock_bh(&in_dev->mc_tomb_lock);
699 for_each_pmc_rcu(in_dev, pmc) {
700 spin_lock_bh(&pmc->lock);
701 if (pmc->sfcount[MCAST_EXCLUDE]) {
702 type = IGMPV3_BLOCK_OLD_SOURCES;
703 dtype = IGMPV3_ALLOW_NEW_SOURCES;
705 type = IGMPV3_ALLOW_NEW_SOURCES;
706 dtype = IGMPV3_BLOCK_OLD_SOURCES;
708 skb = add_grec(skb, pmc, type, 0, 0);
709 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */
711 /* filter mode changes */
713 if (pmc->sfmode == MCAST_EXCLUDE)
714 type = IGMPV3_CHANGE_TO_EXCLUDE;
716 type = IGMPV3_CHANGE_TO_INCLUDE;
717 skb = add_grec(skb, pmc, type, 0, 0);
720 spin_unlock_bh(&pmc->lock);
726 (void) igmpv3_sendpack(skb);
729 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
736 struct net_device *dev = in_dev->dev;
737 struct net *net = dev_net(dev);
738 __be32 group = pmc ? pmc->multiaddr : 0;
743 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
744 return igmpv3_send_report(in_dev, pmc);
746 if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
749 if (type == IGMP_HOST_LEAVE_MESSAGE)
750 dst = IGMP_ALL_ROUTER;
754 rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
756 IPPROTO_IGMP, 0, dev->ifindex);
760 hlen = LL_RESERVED_SPACE(dev);
761 tlen = dev->needed_tailroom;
762 skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
767 skb->priority = TC_PRIO_CONTROL;
769 skb_dst_set(skb, &rt->dst);
771 skb_reserve(skb, hlen);
773 skb_reset_network_header(skb);
775 skb_put(skb, sizeof(struct iphdr) + 4);
778 iph->ihl = (sizeof(struct iphdr)+4)>>2;
780 iph->frag_off = htons(IP_DF);
783 iph->saddr = fl4.saddr;
784 iph->protocol = IPPROTO_IGMP;
785 ip_select_ident(net, skb, NULL);
786 ((u8 *)&iph[1])[0] = IPOPT_RA;
787 ((u8 *)&iph[1])[1] = 4;
788 ((u8 *)&iph[1])[2] = 0;
789 ((u8 *)&iph[1])[3] = 0;
791 ih = skb_put(skb, sizeof(struct igmphdr));
796 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
798 return ip_local_out(net, skb->sk, skb);
801 static void igmp_gq_timer_expire(unsigned long data)
803 struct in_device *in_dev = (struct in_device *)data;
805 in_dev->mr_gq_running = 0;
806 igmpv3_send_report(in_dev, NULL);
810 static void igmp_ifc_timer_expire(unsigned long data)
812 struct in_device *in_dev = (struct in_device *)data;
814 igmpv3_send_cr(in_dev);
815 if (in_dev->mr_ifc_count) {
816 in_dev->mr_ifc_count--;
817 igmp_ifc_start_timer(in_dev,
818 unsolicited_report_interval(in_dev));
823 static void igmp_ifc_event(struct in_device *in_dev)
825 struct net *net = dev_net(in_dev->dev);
826 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
828 in_dev->mr_ifc_count = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
829 igmp_ifc_start_timer(in_dev, 1);
833 static void igmp_timer_expire(unsigned long data)
835 struct ip_mc_list *im = (struct ip_mc_list *)data;
836 struct in_device *in_dev = im->interface;
838 spin_lock(&im->lock);
841 if (im->unsolicit_count) {
842 im->unsolicit_count--;
843 igmp_start_timer(im, unsolicited_report_interval(in_dev));
846 spin_unlock(&im->lock);
848 if (IGMP_V1_SEEN(in_dev))
849 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
850 else if (IGMP_V2_SEEN(in_dev))
851 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
853 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
858 /* mark EXCLUDE-mode sources */
859 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
861 struct ip_sf_list *psf;
865 for (psf = pmc->sources; psf; psf = psf->sf_next) {
868 for (i = 0; i < nsrcs; i++) {
869 /* skip inactive filters */
870 if (psf->sf_count[MCAST_INCLUDE] ||
871 pmc->sfcount[MCAST_EXCLUDE] !=
872 psf->sf_count[MCAST_EXCLUDE])
874 if (srcs[i] == psf->sf_inaddr) {
881 if (scount == nsrcs) /* all sources excluded */
886 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
888 struct ip_sf_list *psf;
891 if (pmc->sfmode == MCAST_EXCLUDE)
892 return igmp_xmarksources(pmc, nsrcs, srcs);
894 /* mark INCLUDE-mode sources */
896 for (psf = pmc->sources; psf; psf = psf->sf_next) {
899 for (i = 0; i < nsrcs; i++)
900 if (srcs[i] == psf->sf_inaddr) {
914 /* return true if packet was dropped */
915 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
917 struct ip_mc_list *im;
918 struct net *net = dev_net(in_dev->dev);
920 /* Timers are only set for non-local groups */
922 if (group == IGMP_ALL_HOSTS)
924 if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
928 for_each_pmc_rcu(in_dev, im) {
929 if (im->multiaddr == group) {
938 /* return true if packet was dropped */
939 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
942 struct igmphdr *ih = igmp_hdr(skb);
943 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
944 struct ip_mc_list *im;
945 __be32 group = ih->group;
948 struct net *net = dev_net(in_dev->dev);
953 /* Alas, old v1 router presents here. */
955 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
956 in_dev->mr_v1_seen = jiffies +
957 IGMP_V1_ROUTER_PRESENT_TIMEOUT;
960 /* v2 router present */
961 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
962 in_dev->mr_v2_seen = jiffies +
963 IGMP_V2_ROUTER_PRESENT_TIMEOUT;
965 /* cancel the interface change timer */
966 in_dev->mr_ifc_count = 0;
967 if (del_timer(&in_dev->mr_ifc_timer))
968 __in_dev_put(in_dev);
969 /* clear deleted report items */
970 igmpv3_clear_delrec(in_dev);
971 } else if (len < 12) {
972 return true; /* ignore bogus packet; freed by caller */
973 } else if (IGMP_V1_SEEN(in_dev)) {
974 /* This is a v3 query with v1 queriers present */
975 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
977 } else if (IGMP_V2_SEEN(in_dev)) {
978 /* this is a v3 query with v2 queriers present;
979 * Interpretation of the max_delay code is problematic here.
980 * A real v2 host would use ih_code directly, while v3 has a
981 * different encoding. We use the v3 encoding as more likely
982 * to be intended in a v3 query.
984 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
986 max_delay = 1; /* can't mod w/ 0 */
988 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
991 ih3 = igmpv3_query_hdr(skb);
993 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
994 + ntohs(ih3->nsrcs)*sizeof(__be32)))
996 ih3 = igmpv3_query_hdr(skb);
999 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
1001 max_delay = 1; /* can't mod w/ 0 */
1002 in_dev->mr_maxdelay = max_delay;
1004 in_dev->mr_qrv = ih3->qrv;
1005 if (!group) { /* general query */
1007 return true; /* no sources allowed */
1008 igmp_gq_start_timer(in_dev);
1011 /* mark sources to include, if group & source-specific */
1012 mark = ih3->nsrcs != 0;
1016 * - Start the timers in all of our membership records
1017 * that the query applies to for the interface on
1018 * which the query arrived excl. those that belong
1019 * to a "local" group (224.0.0.X)
1020 * - For timers already running check if they need to
1022 * - Use the igmp->igmp_code field as the maximum
1026 for_each_pmc_rcu(in_dev, im) {
1029 if (group && group != im->multiaddr)
1031 if (im->multiaddr == IGMP_ALL_HOSTS)
1033 if (ipv4_is_local_multicast(im->multiaddr) &&
1034 !net->ipv4.sysctl_igmp_llm_reports)
1036 spin_lock_bh(&im->lock);
1038 im->gsquery = im->gsquery && mark;
1041 changed = !im->gsquery ||
1042 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
1043 spin_unlock_bh(&im->lock);
1045 igmp_mod_timer(im, max_delay);
1051 /* called in rcu_read_lock() section */
1052 int igmp_rcv(struct sk_buff *skb)
1054 /* This basically follows the spec line by line -- see RFC1112 */
1056 struct net_device *dev = skb->dev;
1057 struct in_device *in_dev;
1059 bool dropped = true;
1061 if (netif_is_l3_master(dev)) {
1062 dev = dev_get_by_index_rcu(dev_net(dev), IPCB(skb)->iif);
1067 in_dev = __in_dev_get_rcu(dev);
1071 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1074 if (skb_checksum_simple_validate(skb))
1079 case IGMP_HOST_MEMBERSHIP_QUERY:
1080 dropped = igmp_heard_query(in_dev, skb, len);
1082 case IGMP_HOST_MEMBERSHIP_REPORT:
1083 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1084 /* Is it our report looped back? */
1085 if (rt_is_output_route(skb_rtable(skb)))
1087 /* don't rely on MC router hearing unicast reports */
1088 if (skb->pkt_type == PACKET_MULTICAST ||
1089 skb->pkt_type == PACKET_BROADCAST)
1090 dropped = igmp_heard_report(in_dev, ih->group);
1093 #ifdef CONFIG_IP_PIMSM_V1
1094 return pim_rcv_v1(skb);
1096 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1099 case IGMP_HOST_LEAVE_MESSAGE:
1101 case IGMP_MTRACE_RESP:
1119 * Add a filter to a device
1122 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1124 char buf[MAX_ADDR_LEN];
1125 struct net_device *dev = in_dev->dev;
1127 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1128 We will get multicast token leakage, when IFF_MULTICAST
1129 is changed. This check should be done in ndo_set_rx_mode
1130 routine. Something sort of:
1131 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1134 if (arp_mc_map(addr, buf, dev, 0) == 0)
1135 dev_mc_add(dev, buf);
1139 * Remove a filter from a device
1142 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1144 char buf[MAX_ADDR_LEN];
1145 struct net_device *dev = in_dev->dev;
1147 if (arp_mc_map(addr, buf, dev, 0) == 0)
1148 dev_mc_del(dev, buf);
1151 #ifdef CONFIG_IP_MULTICAST
1153 * deleted ip_mc_list manipulation
1155 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1157 struct ip_mc_list *pmc;
1158 struct net *net = dev_net(in_dev->dev);
1160 /* this is an "ip_mc_list" for convenience; only the fields below
1161 * are actually used. In particular, the refcnt and users are not
1162 * used for management of the delete list. Using the same structure
1163 * for deleted items allows change reports to use common code with
1164 * non-deleted or query-response MCA's.
1166 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1169 spin_lock_init(&pmc->lock);
1170 spin_lock_bh(&im->lock);
1171 pmc->interface = im->interface;
1172 in_dev_hold(in_dev);
1173 pmc->multiaddr = im->multiaddr;
1174 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1175 pmc->sfmode = im->sfmode;
1176 if (pmc->sfmode == MCAST_INCLUDE) {
1177 struct ip_sf_list *psf;
1179 pmc->tomb = im->tomb;
1180 pmc->sources = im->sources;
1181 im->tomb = im->sources = NULL;
1182 for (psf = pmc->sources; psf; psf = psf->sf_next)
1183 psf->sf_crcount = pmc->crcount;
1185 spin_unlock_bh(&im->lock);
1187 spin_lock_bh(&in_dev->mc_tomb_lock);
1188 pmc->next = in_dev->mc_tomb;
1189 in_dev->mc_tomb = pmc;
1190 spin_unlock_bh(&in_dev->mc_tomb_lock);
1194 * restore ip_mc_list deleted records
1196 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1198 struct ip_mc_list *pmc, *pmc_prev;
1199 struct ip_sf_list *psf;
1200 struct net *net = dev_net(in_dev->dev);
1201 __be32 multiaddr = im->multiaddr;
1203 spin_lock_bh(&in_dev->mc_tomb_lock);
1205 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1206 if (pmc->multiaddr == multiaddr)
1212 pmc_prev->next = pmc->next;
1214 in_dev->mc_tomb = pmc->next;
1216 spin_unlock_bh(&in_dev->mc_tomb_lock);
1218 spin_lock_bh(&im->lock);
1220 im->interface = pmc->interface;
1221 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1222 if (im->sfmode == MCAST_INCLUDE) {
1223 swap(im->tomb, pmc->tomb);
1224 swap(im->sources, pmc->sources);
1225 for (psf = im->sources; psf; psf = psf->sf_next)
1226 psf->sf_crcount = im->crcount;
1228 in_dev_put(pmc->interface);
1231 spin_unlock_bh(&im->lock);
1235 * flush ip_mc_list deleted records
1237 static void igmpv3_clear_delrec(struct in_device *in_dev)
1239 struct ip_mc_list *pmc, *nextpmc;
1241 spin_lock_bh(&in_dev->mc_tomb_lock);
1242 pmc = in_dev->mc_tomb;
1243 in_dev->mc_tomb = NULL;
1244 spin_unlock_bh(&in_dev->mc_tomb_lock);
1246 for (; pmc; pmc = nextpmc) {
1247 nextpmc = pmc->next;
1248 ip_mc_clear_src(pmc);
1249 in_dev_put(pmc->interface);
1252 /* clear dead sources, too */
1254 for_each_pmc_rcu(in_dev, pmc) {
1255 struct ip_sf_list *psf;
1257 spin_lock_bh(&pmc->lock);
1260 spin_unlock_bh(&pmc->lock);
1261 ip_sf_list_clear_all(psf);
1267 static void igmp_group_dropped(struct ip_mc_list *im)
1269 struct in_device *in_dev = im->interface;
1270 #ifdef CONFIG_IP_MULTICAST
1271 struct net *net = dev_net(in_dev->dev);
1277 ip_mc_filter_del(in_dev, im->multiaddr);
1280 #ifdef CONFIG_IP_MULTICAST
1281 if (im->multiaddr == IGMP_ALL_HOSTS)
1283 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1286 reporter = im->reporter;
1287 igmp_stop_timer(im);
1289 if (!in_dev->dead) {
1290 if (IGMP_V1_SEEN(in_dev))
1292 if (IGMP_V2_SEEN(in_dev)) {
1294 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1298 igmpv3_add_delrec(in_dev, im);
1300 igmp_ifc_event(in_dev);
1305 static void igmp_group_added(struct ip_mc_list *im)
1307 struct in_device *in_dev = im->interface;
1308 #ifdef CONFIG_IP_MULTICAST
1309 struct net *net = dev_net(in_dev->dev);
1312 if (im->loaded == 0) {
1314 ip_mc_filter_add(in_dev, im->multiaddr);
1317 #ifdef CONFIG_IP_MULTICAST
1318 if (im->multiaddr == IGMP_ALL_HOSTS)
1320 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1325 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1326 spin_lock_bh(&im->lock);
1327 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1328 spin_unlock_bh(&im->lock);
1333 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1334 igmp_ifc_event(in_dev);
1340 * Multicast list managers
1343 static u32 ip_mc_hash(const struct ip_mc_list *im)
1345 return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1348 static void ip_mc_hash_add(struct in_device *in_dev,
1349 struct ip_mc_list *im)
1351 struct ip_mc_list __rcu **mc_hash;
1354 mc_hash = rtnl_dereference(in_dev->mc_hash);
1356 hash = ip_mc_hash(im);
1357 im->next_hash = mc_hash[hash];
1358 rcu_assign_pointer(mc_hash[hash], im);
1362 /* do not use a hash table for small number of items */
1363 if (in_dev->mc_count < 4)
1366 mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1371 for_each_pmc_rtnl(in_dev, im) {
1372 hash = ip_mc_hash(im);
1373 im->next_hash = mc_hash[hash];
1374 RCU_INIT_POINTER(mc_hash[hash], im);
1377 rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1380 static void ip_mc_hash_remove(struct in_device *in_dev,
1381 struct ip_mc_list *im)
1383 struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1384 struct ip_mc_list *aux;
1388 mc_hash += ip_mc_hash(im);
1389 while ((aux = rtnl_dereference(*mc_hash)) != im)
1390 mc_hash = &aux->next_hash;
1391 *mc_hash = im->next_hash;
1396 * A socket has joined a multicast group on device dev.
1399 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1401 struct ip_mc_list *im;
1402 #ifdef CONFIG_IP_MULTICAST
1403 struct net *net = dev_net(in_dev->dev);
1408 for_each_pmc_rtnl(in_dev, im) {
1409 if (im->multiaddr == addr) {
1411 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1416 im = kzalloc(sizeof(*im), GFP_KERNEL);
1421 im->interface = in_dev;
1422 in_dev_hold(in_dev);
1423 im->multiaddr = addr;
1424 /* initial mode is (EX, empty) */
1425 im->sfmode = MCAST_EXCLUDE;
1426 im->sfcount[MCAST_EXCLUDE] = 1;
1427 refcount_set(&im->refcnt, 1);
1428 spin_lock_init(&im->lock);
1429 #ifdef CONFIG_IP_MULTICAST
1430 setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1431 im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1434 im->next_rcu = in_dev->mc_list;
1436 rcu_assign_pointer(in_dev->mc_list, im);
1438 ip_mc_hash_add(in_dev, im);
1440 #ifdef CONFIG_IP_MULTICAST
1441 igmpv3_del_delrec(in_dev, im);
1443 igmp_group_added(im);
1445 ip_rt_multicast_event(in_dev);
1449 EXPORT_SYMBOL(ip_mc_inc_group);
1451 static int ip_mc_check_iphdr(struct sk_buff *skb)
1453 const struct iphdr *iph;
1455 unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1457 if (!pskb_may_pull(skb, offset))
1462 if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1465 offset += ip_hdrlen(skb) - sizeof(*iph);
1467 if (!pskb_may_pull(skb, offset))
1472 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1475 len = skb_network_offset(skb) + ntohs(iph->tot_len);
1476 if (skb->len < len || len < offset)
1479 skb_set_transport_header(skb, offset);
1484 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1486 unsigned int len = skb_transport_offset(skb);
1488 len += sizeof(struct igmpv3_report);
1490 return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1493 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1495 unsigned int len = skb_transport_offset(skb);
1497 len += sizeof(struct igmphdr);
1502 if (skb->len != len) {
1504 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1505 if (skb->len < len || !pskb_may_pull(skb, len))
1509 /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1510 * all-systems destination addresses (224.0.0.1) for general queries
1512 if (!igmp_hdr(skb)->group &&
1513 ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1519 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1521 switch (igmp_hdr(skb)->type) {
1522 case IGMP_HOST_LEAVE_MESSAGE:
1523 case IGMP_HOST_MEMBERSHIP_REPORT:
1524 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1527 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1528 return ip_mc_check_igmp_reportv3(skb);
1529 case IGMP_HOST_MEMBERSHIP_QUERY:
1530 return ip_mc_check_igmp_query(skb);
1536 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1538 return skb_checksum_simple_validate(skb);
1541 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1544 struct sk_buff *skb_chk;
1545 unsigned int transport_len;
1546 unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1549 transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1551 skb_chk = skb_checksum_trimmed(skb, transport_len,
1552 ip_mc_validate_checksum);
1556 if (!pskb_may_pull(skb_chk, len))
1559 ret = ip_mc_check_igmp_msg(skb_chk);
1564 *skb_trimmed = skb_chk;
1565 /* free now unneeded clone */
1566 else if (skb_chk != skb)
1572 if (ret && skb_chk && skb_chk != skb)
1579 * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1580 * @skb: the skb to validate
1581 * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1583 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1584 * skb transport header accordingly and returns zero.
1586 * -EINVAL: A broken packet was detected, i.e. it violates some internet
1588 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1589 * -ENOMEM: A memory allocation failure happened.
1591 * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1592 * to NULL): After parsing an IGMP packet successfully it will point to
1593 * an skb which has its tail aligned to the IP packet end. This might
1594 * either be the originally provided skb or a trimmed, cloned version if
1595 * the skb frame had data beyond the IP packet. A cloned skb allows us
1596 * to leave the original skb and its full frame unchanged (which might be
1597 * desirable for layer 2 frame jugglers).
1599 * Caller needs to set the skb network header and free any returned skb if it
1600 * differs from the provided skb.
1602 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1604 int ret = ip_mc_check_iphdr(skb);
1609 if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1612 return __ip_mc_check_igmp(skb, skb_trimmed);
1614 EXPORT_SYMBOL(ip_mc_check_igmp);
1617 * Resend IGMP JOIN report; used by netdev notifier.
1619 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1621 #ifdef CONFIG_IP_MULTICAST
1622 struct ip_mc_list *im;
1624 struct net *net = dev_net(in_dev->dev);
1628 for_each_pmc_rtnl(in_dev, im) {
1629 if (im->multiaddr == IGMP_ALL_HOSTS)
1631 if (ipv4_is_local_multicast(im->multiaddr) &&
1632 !net->ipv4.sysctl_igmp_llm_reports)
1635 /* a failover is happening and switches
1636 * must be notified immediately
1638 if (IGMP_V1_SEEN(in_dev))
1639 type = IGMP_HOST_MEMBERSHIP_REPORT;
1640 else if (IGMP_V2_SEEN(in_dev))
1641 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1643 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1644 igmp_send_report(in_dev, im, type);
1650 * A socket has left a multicast group on device dev
1653 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1655 struct ip_mc_list *i;
1656 struct ip_mc_list __rcu **ip;
1660 for (ip = &in_dev->mc_list;
1661 (i = rtnl_dereference(*ip)) != NULL;
1662 ip = &i->next_rcu) {
1663 if (i->multiaddr == addr) {
1664 if (--i->users == 0) {
1665 ip_mc_hash_remove(in_dev, i);
1668 igmp_group_dropped(i);
1672 ip_rt_multicast_event(in_dev);
1681 EXPORT_SYMBOL(ip_mc_dec_group);
1683 /* Device changing type */
1685 void ip_mc_unmap(struct in_device *in_dev)
1687 struct ip_mc_list *pmc;
1691 for_each_pmc_rtnl(in_dev, pmc)
1692 igmp_group_dropped(pmc);
1695 void ip_mc_remap(struct in_device *in_dev)
1697 struct ip_mc_list *pmc;
1701 for_each_pmc_rtnl(in_dev, pmc) {
1702 #ifdef CONFIG_IP_MULTICAST
1703 igmpv3_del_delrec(in_dev, pmc);
1705 igmp_group_added(pmc);
1709 /* Device going down */
1711 void ip_mc_down(struct in_device *in_dev)
1713 struct ip_mc_list *pmc;
1717 for_each_pmc_rtnl(in_dev, pmc)
1718 igmp_group_dropped(pmc);
1720 #ifdef CONFIG_IP_MULTICAST
1721 in_dev->mr_ifc_count = 0;
1722 if (del_timer(&in_dev->mr_ifc_timer))
1723 __in_dev_put(in_dev);
1724 in_dev->mr_gq_running = 0;
1725 if (del_timer(&in_dev->mr_gq_timer))
1726 __in_dev_put(in_dev);
1729 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1732 void ip_mc_init_dev(struct in_device *in_dev)
1734 #ifdef CONFIG_IP_MULTICAST
1735 struct net *net = dev_net(in_dev->dev);
1739 #ifdef CONFIG_IP_MULTICAST
1740 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1741 (unsigned long)in_dev);
1742 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1743 (unsigned long)in_dev);
1744 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1747 spin_lock_init(&in_dev->mc_tomb_lock);
1750 /* Device going up */
1752 void ip_mc_up(struct in_device *in_dev)
1754 struct ip_mc_list *pmc;
1755 #ifdef CONFIG_IP_MULTICAST
1756 struct net *net = dev_net(in_dev->dev);
1761 #ifdef CONFIG_IP_MULTICAST
1762 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1764 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1766 for_each_pmc_rtnl(in_dev, pmc) {
1767 #ifdef CONFIG_IP_MULTICAST
1768 igmpv3_del_delrec(in_dev, pmc);
1770 igmp_group_added(pmc);
1775 * Device is about to be destroyed: clean up.
1778 void ip_mc_destroy_dev(struct in_device *in_dev)
1780 struct ip_mc_list *i;
1784 /* Deactivate timers */
1786 #ifdef CONFIG_IP_MULTICAST
1787 igmpv3_clear_delrec(in_dev);
1790 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1791 in_dev->mc_list = i->next_rcu;
1798 /* RTNL is locked */
1799 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1801 struct net_device *dev = NULL;
1802 struct in_device *idev = NULL;
1804 if (imr->imr_ifindex) {
1805 idev = inetdev_by_index(net, imr->imr_ifindex);
1808 if (imr->imr_address.s_addr) {
1809 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1815 struct rtable *rt = ip_route_output(net,
1816 imr->imr_multiaddr.s_addr,
1824 imr->imr_ifindex = dev->ifindex;
1825 idev = __in_dev_get_rtnl(dev);
1831 * Join a socket to a group
1834 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1837 struct ip_sf_list *psf, *psf_prev;
1841 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1842 if (psf->sf_inaddr == *psfsrc)
1846 if (!psf || psf->sf_count[sfmode] == 0) {
1847 /* source filter not found, or count wrong => bug */
1850 psf->sf_count[sfmode]--;
1851 if (psf->sf_count[sfmode] == 0) {
1852 ip_rt_multicast_event(pmc->interface);
1854 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1855 #ifdef CONFIG_IP_MULTICAST
1856 struct in_device *in_dev = pmc->interface;
1857 struct net *net = dev_net(in_dev->dev);
1860 /* no more filters for this source */
1862 psf_prev->sf_next = psf->sf_next;
1864 pmc->sources = psf->sf_next;
1865 #ifdef CONFIG_IP_MULTICAST
1866 if (psf->sf_oldin &&
1867 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1868 psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1869 psf->sf_next = pmc->tomb;
1879 #ifndef CONFIG_IP_MULTICAST
1880 #define igmp_ifc_event(x) do { } while (0)
1883 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1884 int sfcount, __be32 *psfsrc, int delta)
1886 struct ip_mc_list *pmc;
1893 for_each_pmc_rcu(in_dev, pmc) {
1894 if (*pmca == pmc->multiaddr)
1898 /* MCA not found?? bug */
1902 spin_lock_bh(&pmc->lock);
1904 #ifdef CONFIG_IP_MULTICAST
1909 if (!pmc->sfcount[sfmode])
1911 pmc->sfcount[sfmode]--;
1914 for (i = 0; i < sfcount; i++) {
1915 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1917 changerec |= rv > 0;
1921 if (pmc->sfmode == MCAST_EXCLUDE &&
1922 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1923 pmc->sfcount[MCAST_INCLUDE]) {
1924 #ifdef CONFIG_IP_MULTICAST
1925 struct ip_sf_list *psf;
1926 struct net *net = dev_net(in_dev->dev);
1929 /* filter mode change */
1930 pmc->sfmode = MCAST_INCLUDE;
1931 #ifdef CONFIG_IP_MULTICAST
1932 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1933 in_dev->mr_ifc_count = pmc->crcount;
1934 for (psf = pmc->sources; psf; psf = psf->sf_next)
1935 psf->sf_crcount = 0;
1936 igmp_ifc_event(pmc->interface);
1937 } else if (sf_setstate(pmc) || changerec) {
1938 igmp_ifc_event(pmc->interface);
1942 spin_unlock_bh(&pmc->lock);
1947 * Add multicast single-source filter to the interface list
1949 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1952 struct ip_sf_list *psf, *psf_prev;
1955 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1956 if (psf->sf_inaddr == *psfsrc)
1961 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1964 psf->sf_inaddr = *psfsrc;
1966 psf_prev->sf_next = psf;
1970 psf->sf_count[sfmode]++;
1971 if (psf->sf_count[sfmode] == 1) {
1972 ip_rt_multicast_event(pmc->interface);
1977 #ifdef CONFIG_IP_MULTICAST
1978 static void sf_markstate(struct ip_mc_list *pmc)
1980 struct ip_sf_list *psf;
1981 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1983 for (psf = pmc->sources; psf; psf = psf->sf_next)
1984 if (pmc->sfcount[MCAST_EXCLUDE]) {
1985 psf->sf_oldin = mca_xcount ==
1986 psf->sf_count[MCAST_EXCLUDE] &&
1987 !psf->sf_count[MCAST_INCLUDE];
1989 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1992 static int sf_setstate(struct ip_mc_list *pmc)
1994 struct ip_sf_list *psf, *dpsf;
1995 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1996 int qrv = pmc->interface->mr_qrv;
2000 for (psf = pmc->sources; psf; psf = psf->sf_next) {
2001 if (pmc->sfcount[MCAST_EXCLUDE]) {
2002 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2003 !psf->sf_count[MCAST_INCLUDE];
2005 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2007 if (!psf->sf_oldin) {
2008 struct ip_sf_list *prev = NULL;
2010 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
2011 if (dpsf->sf_inaddr == psf->sf_inaddr)
2017 prev->sf_next = dpsf->sf_next;
2019 pmc->tomb = dpsf->sf_next;
2022 psf->sf_crcount = qrv;
2025 } else if (psf->sf_oldin) {
2027 psf->sf_crcount = 0;
2029 * add or update "delete" records if an active filter
2032 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
2033 if (dpsf->sf_inaddr == psf->sf_inaddr)
2036 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2040 /* pmc->lock held by callers */
2041 dpsf->sf_next = pmc->tomb;
2044 dpsf->sf_crcount = qrv;
2053 * Add multicast source filter list to the interface list
2055 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2056 int sfcount, __be32 *psfsrc, int delta)
2058 struct ip_mc_list *pmc;
2065 for_each_pmc_rcu(in_dev, pmc) {
2066 if (*pmca == pmc->multiaddr)
2070 /* MCA not found?? bug */
2074 spin_lock_bh(&pmc->lock);
2077 #ifdef CONFIG_IP_MULTICAST
2080 isexclude = pmc->sfmode == MCAST_EXCLUDE;
2082 pmc->sfcount[sfmode]++;
2084 for (i = 0; i < sfcount; i++) {
2085 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2093 pmc->sfcount[sfmode]--;
2094 for (j = 0; j < i; j++)
2095 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2096 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2097 #ifdef CONFIG_IP_MULTICAST
2098 struct ip_sf_list *psf;
2099 struct net *net = dev_net(pmc->interface->dev);
2100 in_dev = pmc->interface;
2103 /* filter mode change */
2104 if (pmc->sfcount[MCAST_EXCLUDE])
2105 pmc->sfmode = MCAST_EXCLUDE;
2106 else if (pmc->sfcount[MCAST_INCLUDE])
2107 pmc->sfmode = MCAST_INCLUDE;
2108 #ifdef CONFIG_IP_MULTICAST
2109 /* else no filters; keep old mode for reports */
2111 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2112 in_dev->mr_ifc_count = pmc->crcount;
2113 for (psf = pmc->sources; psf; psf = psf->sf_next)
2114 psf->sf_crcount = 0;
2115 igmp_ifc_event(in_dev);
2116 } else if (sf_setstate(pmc)) {
2117 igmp_ifc_event(in_dev);
2120 spin_unlock_bh(&pmc->lock);
2124 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2126 struct ip_sf_list *tomb, *sources;
2128 spin_lock_bh(&pmc->lock);
2131 sources = pmc->sources;
2132 pmc->sources = NULL;
2133 pmc->sfmode = MCAST_EXCLUDE;
2134 pmc->sfcount[MCAST_INCLUDE] = 0;
2135 pmc->sfcount[MCAST_EXCLUDE] = 1;
2136 spin_unlock_bh(&pmc->lock);
2138 ip_sf_list_clear_all(tomb);
2139 ip_sf_list_clear_all(sources);
2142 /* Join a multicast group
2145 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2147 __be32 addr = imr->imr_multiaddr.s_addr;
2148 struct ip_mc_socklist *iml, *i;
2149 struct in_device *in_dev;
2150 struct inet_sock *inet = inet_sk(sk);
2151 struct net *net = sock_net(sk);
2158 if (!ipv4_is_multicast(addr))
2161 in_dev = ip_mc_find_dev(net, imr);
2169 ifindex = imr->imr_ifindex;
2170 for_each_pmc_rtnl(inet, i) {
2171 if (i->multi.imr_multiaddr.s_addr == addr &&
2172 i->multi.imr_ifindex == ifindex)
2177 if (count >= net->ipv4.sysctl_igmp_max_memberships)
2179 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2183 memcpy(&iml->multi, imr, sizeof(*imr));
2184 iml->next_rcu = inet->mc_list;
2186 iml->sfmode = MCAST_EXCLUDE;
2187 rcu_assign_pointer(inet->mc_list, iml);
2188 ip_mc_inc_group(in_dev, addr);
2193 EXPORT_SYMBOL(ip_mc_join_group);
2195 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2196 struct in_device *in_dev)
2198 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2202 /* any-source empty exclude case */
2203 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2204 iml->sfmode, 0, NULL, 0);
2206 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2207 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2208 RCU_INIT_POINTER(iml->sflist, NULL);
2209 /* decrease mem now to avoid the memleak warning */
2210 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2211 kfree_rcu(psf, rcu);
2215 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2217 struct inet_sock *inet = inet_sk(sk);
2218 struct ip_mc_socklist *iml;
2219 struct ip_mc_socklist __rcu **imlp;
2220 struct in_device *in_dev;
2221 struct net *net = sock_net(sk);
2222 __be32 group = imr->imr_multiaddr.s_addr;
2224 int ret = -EADDRNOTAVAIL;
2228 in_dev = ip_mc_find_dev(net, imr);
2229 if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2233 ifindex = imr->imr_ifindex;
2234 for (imlp = &inet->mc_list;
2235 (iml = rtnl_dereference(*imlp)) != NULL;
2236 imlp = &iml->next_rcu) {
2237 if (iml->multi.imr_multiaddr.s_addr != group)
2240 if (iml->multi.imr_ifindex != ifindex)
2242 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2243 iml->multi.imr_address.s_addr)
2246 (void) ip_mc_leave_src(sk, iml, in_dev);
2248 *imlp = iml->next_rcu;
2251 ip_mc_dec_group(in_dev, group);
2253 /* decrease mem now to avoid the memleak warning */
2254 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2255 kfree_rcu(iml, rcu);
2261 EXPORT_SYMBOL(ip_mc_leave_group);
2263 int ip_mc_source(int add, int omode, struct sock *sk, struct
2264 ip_mreq_source *mreqs, int ifindex)
2267 struct ip_mreqn imr;
2268 __be32 addr = mreqs->imr_multiaddr;
2269 struct ip_mc_socklist *pmc;
2270 struct in_device *in_dev = NULL;
2271 struct inet_sock *inet = inet_sk(sk);
2272 struct ip_sf_socklist *psl;
2273 struct net *net = sock_net(sk);
2277 if (!ipv4_is_multicast(addr))
2282 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2283 imr.imr_address.s_addr = mreqs->imr_interface;
2284 imr.imr_ifindex = ifindex;
2285 in_dev = ip_mc_find_dev(net, &imr);
2291 err = -EADDRNOTAVAIL;
2293 for_each_pmc_rtnl(inet, pmc) {
2294 if ((pmc->multi.imr_multiaddr.s_addr ==
2295 imr.imr_multiaddr.s_addr) &&
2296 (pmc->multi.imr_ifindex == imr.imr_ifindex))
2299 if (!pmc) { /* must have a prior join */
2303 /* if a source filter was set, must be the same mode as before */
2305 if (pmc->sfmode != omode) {
2309 } else if (pmc->sfmode != omode) {
2310 /* allow mode switches for empty-set filters */
2311 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2312 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2314 pmc->sfmode = omode;
2317 psl = rtnl_dereference(pmc->sflist);
2320 goto done; /* err = -EADDRNOTAVAIL */
2322 for (i = 0; i < psl->sl_count; i++) {
2323 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2328 if (rv) /* source not found */
2329 goto done; /* err = -EADDRNOTAVAIL */
2331 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2332 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2337 /* update the interface filter */
2338 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2339 &mreqs->imr_sourceaddr, 1);
2341 for (j = i+1; j < psl->sl_count; j++)
2342 psl->sl_addr[j-1] = psl->sl_addr[j];
2347 /* else, add a new source to the filter */
2349 if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2353 if (!psl || psl->sl_count == psl->sl_max) {
2354 struct ip_sf_socklist *newpsl;
2355 int count = IP_SFBLOCK;
2358 count += psl->sl_max;
2359 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2364 newpsl->sl_max = count;
2365 newpsl->sl_count = count - IP_SFBLOCK;
2367 for (i = 0; i < psl->sl_count; i++)
2368 newpsl->sl_addr[i] = psl->sl_addr[i];
2369 /* decrease mem now to avoid the memleak warning */
2370 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2371 kfree_rcu(psl, rcu);
2373 rcu_assign_pointer(pmc->sflist, newpsl);
2376 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2377 for (i = 0; i < psl->sl_count; i++) {
2378 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2383 if (rv == 0) /* address already there is an error */
2385 for (j = psl->sl_count-1; j >= i; j--)
2386 psl->sl_addr[j+1] = psl->sl_addr[j];
2387 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2390 /* update the interface list */
2391 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2392 &mreqs->imr_sourceaddr, 1);
2395 err = ip_mc_leave_group(sk, &imr);
2399 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2402 struct ip_mreqn imr;
2403 __be32 addr = msf->imsf_multiaddr;
2404 struct ip_mc_socklist *pmc;
2405 struct in_device *in_dev;
2406 struct inet_sock *inet = inet_sk(sk);
2407 struct ip_sf_socklist *newpsl, *psl;
2408 struct net *net = sock_net(sk);
2411 if (!ipv4_is_multicast(addr))
2413 if (msf->imsf_fmode != MCAST_INCLUDE &&
2414 msf->imsf_fmode != MCAST_EXCLUDE)
2419 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2420 imr.imr_address.s_addr = msf->imsf_interface;
2421 imr.imr_ifindex = ifindex;
2422 in_dev = ip_mc_find_dev(net, &imr);
2429 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2430 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2435 for_each_pmc_rtnl(inet, pmc) {
2436 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2437 pmc->multi.imr_ifindex == imr.imr_ifindex)
2440 if (!pmc) { /* must have a prior join */
2444 if (msf->imsf_numsrc) {
2445 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2451 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2452 memcpy(newpsl->sl_addr, msf->imsf_slist,
2453 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2454 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2455 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2457 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2462 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2463 msf->imsf_fmode, 0, NULL, 0);
2465 psl = rtnl_dereference(pmc->sflist);
2467 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2468 psl->sl_count, psl->sl_addr, 0);
2469 /* decrease mem now to avoid the memleak warning */
2470 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2471 kfree_rcu(psl, rcu);
2473 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2475 rcu_assign_pointer(pmc->sflist, newpsl);
2476 pmc->sfmode = msf->imsf_fmode;
2480 err = ip_mc_leave_group(sk, &imr);
2484 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2485 struct ip_msfilter __user *optval, int __user *optlen)
2487 int err, len, count, copycount;
2488 struct ip_mreqn imr;
2489 __be32 addr = msf->imsf_multiaddr;
2490 struct ip_mc_socklist *pmc;
2491 struct in_device *in_dev;
2492 struct inet_sock *inet = inet_sk(sk);
2493 struct ip_sf_socklist *psl;
2494 struct net *net = sock_net(sk);
2498 if (!ipv4_is_multicast(addr))
2501 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2502 imr.imr_address.s_addr = msf->imsf_interface;
2503 imr.imr_ifindex = 0;
2504 in_dev = ip_mc_find_dev(net, &imr);
2510 err = -EADDRNOTAVAIL;
2512 for_each_pmc_rtnl(inet, pmc) {
2513 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2514 pmc->multi.imr_ifindex == imr.imr_ifindex)
2517 if (!pmc) /* must have a prior join */
2519 msf->imsf_fmode = pmc->sfmode;
2520 psl = rtnl_dereference(pmc->sflist);
2525 count = psl->sl_count;
2527 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2528 len = copycount * sizeof(psl->sl_addr[0]);
2529 msf->imsf_numsrc = count;
2530 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2531 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2535 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2542 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2543 struct group_filter __user *optval, int __user *optlen)
2545 int err, i, count, copycount;
2546 struct sockaddr_in *psin;
2548 struct ip_mc_socklist *pmc;
2549 struct inet_sock *inet = inet_sk(sk);
2550 struct ip_sf_socklist *psl;
2554 psin = (struct sockaddr_in *)&gsf->gf_group;
2555 if (psin->sin_family != AF_INET)
2557 addr = psin->sin_addr.s_addr;
2558 if (!ipv4_is_multicast(addr))
2561 err = -EADDRNOTAVAIL;
2563 for_each_pmc_rtnl(inet, pmc) {
2564 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2565 pmc->multi.imr_ifindex == gsf->gf_interface)
2568 if (!pmc) /* must have a prior join */
2570 gsf->gf_fmode = pmc->sfmode;
2571 psl = rtnl_dereference(pmc->sflist);
2572 count = psl ? psl->sl_count : 0;
2573 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2574 gsf->gf_numsrc = count;
2575 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2576 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2579 for (i = 0; i < copycount; i++) {
2580 struct sockaddr_storage ss;
2582 psin = (struct sockaddr_in *)&ss;
2583 memset(&ss, 0, sizeof(ss));
2584 psin->sin_family = AF_INET;
2585 psin->sin_addr.s_addr = psl->sl_addr[i];
2586 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2595 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2597 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr,
2600 struct inet_sock *inet = inet_sk(sk);
2601 struct ip_mc_socklist *pmc;
2602 struct ip_sf_socklist *psl;
2607 if (!ipv4_is_multicast(loc_addr))
2611 for_each_pmc_rcu(inet, pmc) {
2612 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2613 (pmc->multi.imr_ifindex == dif ||
2614 (sdif && pmc->multi.imr_ifindex == sdif)))
2620 psl = rcu_dereference(pmc->sflist);
2621 ret = (pmc->sfmode == MCAST_EXCLUDE);
2625 for (i = 0; i < psl->sl_count; i++) {
2626 if (psl->sl_addr[i] == rmt_addr)
2630 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2632 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2642 * A socket is closing.
2645 void ip_mc_drop_socket(struct sock *sk)
2647 struct inet_sock *inet = inet_sk(sk);
2648 struct ip_mc_socklist *iml;
2649 struct net *net = sock_net(sk);
2655 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2656 struct in_device *in_dev;
2658 inet->mc_list = iml->next_rcu;
2659 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2660 (void) ip_mc_leave_src(sk, iml, in_dev);
2662 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2663 /* decrease mem now to avoid the memleak warning */
2664 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2665 kfree_rcu(iml, rcu);
2670 /* called with rcu_read_lock() */
2671 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2673 struct ip_mc_list *im;
2674 struct ip_mc_list __rcu **mc_hash;
2675 struct ip_sf_list *psf;
2678 mc_hash = rcu_dereference(in_dev->mc_hash);
2680 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2682 for (im = rcu_dereference(mc_hash[hash]);
2684 im = rcu_dereference(im->next_hash)) {
2685 if (im->multiaddr == mc_addr)
2689 for_each_pmc_rcu(in_dev, im) {
2690 if (im->multiaddr == mc_addr)
2694 if (im && proto == IPPROTO_IGMP) {
2698 spin_lock_bh(&im->lock);
2699 for (psf = im->sources; psf; psf = psf->sf_next) {
2700 if (psf->sf_inaddr == src_addr)
2704 rv = psf->sf_count[MCAST_INCLUDE] ||
2705 psf->sf_count[MCAST_EXCLUDE] !=
2706 im->sfcount[MCAST_EXCLUDE];
2708 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2709 spin_unlock_bh(&im->lock);
2711 rv = 1; /* unspecified source; tentatively allow */
2716 #if defined(CONFIG_PROC_FS)
2717 struct igmp_mc_iter_state {
2718 struct seq_net_private p;
2719 struct net_device *dev;
2720 struct in_device *in_dev;
2723 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2725 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2727 struct net *net = seq_file_net(seq);
2728 struct ip_mc_list *im = NULL;
2729 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2731 state->in_dev = NULL;
2732 for_each_netdev_rcu(net, state->dev) {
2733 struct in_device *in_dev;
2735 in_dev = __in_dev_get_rcu(state->dev);
2738 im = rcu_dereference(in_dev->mc_list);
2740 state->in_dev = in_dev;
2747 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2749 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2751 im = rcu_dereference(im->next_rcu);
2753 state->dev = next_net_device_rcu(state->dev);
2755 state->in_dev = NULL;
2758 state->in_dev = __in_dev_get_rcu(state->dev);
2761 im = rcu_dereference(state->in_dev->mc_list);
2766 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2768 struct ip_mc_list *im = igmp_mc_get_first(seq);
2770 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2772 return pos ? NULL : im;
2775 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2779 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2782 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2784 struct ip_mc_list *im;
2785 if (v == SEQ_START_TOKEN)
2786 im = igmp_mc_get_first(seq);
2788 im = igmp_mc_get_next(seq, v);
2793 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2796 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2798 state->in_dev = NULL;
2803 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2805 if (v == SEQ_START_TOKEN)
2807 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2809 struct ip_mc_list *im = (struct ip_mc_list *)v;
2810 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2814 #ifdef CONFIG_IP_MULTICAST
2815 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2816 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2822 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2823 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2824 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2827 delta = im->timer.expires - jiffies;
2829 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2830 im->multiaddr, im->users,
2832 im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2838 static const struct seq_operations igmp_mc_seq_ops = {
2839 .start = igmp_mc_seq_start,
2840 .next = igmp_mc_seq_next,
2841 .stop = igmp_mc_seq_stop,
2842 .show = igmp_mc_seq_show,
2845 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2847 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2848 sizeof(struct igmp_mc_iter_state));
2851 static const struct file_operations igmp_mc_seq_fops = {
2852 .owner = THIS_MODULE,
2853 .open = igmp_mc_seq_open,
2855 .llseek = seq_lseek,
2856 .release = seq_release_net,
2859 struct igmp_mcf_iter_state {
2860 struct seq_net_private p;
2861 struct net_device *dev;
2862 struct in_device *idev;
2863 struct ip_mc_list *im;
2866 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2868 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2870 struct net *net = seq_file_net(seq);
2871 struct ip_sf_list *psf = NULL;
2872 struct ip_mc_list *im = NULL;
2873 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2877 for_each_netdev_rcu(net, state->dev) {
2878 struct in_device *idev;
2879 idev = __in_dev_get_rcu(state->dev);
2880 if (unlikely(!idev))
2882 im = rcu_dereference(idev->mc_list);
2884 spin_lock_bh(&im->lock);
2891 spin_unlock_bh(&im->lock);
2897 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2899 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2903 spin_unlock_bh(&state->im->lock);
2904 state->im = state->im->next;
2905 while (!state->im) {
2906 state->dev = next_net_device_rcu(state->dev);
2911 state->idev = __in_dev_get_rcu(state->dev);
2914 state->im = rcu_dereference(state->idev->mc_list);
2918 spin_lock_bh(&state->im->lock);
2919 psf = state->im->sources;
2925 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2927 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2929 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2931 return pos ? NULL : psf;
2934 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2938 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2941 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2943 struct ip_sf_list *psf;
2944 if (v == SEQ_START_TOKEN)
2945 psf = igmp_mcf_get_first(seq);
2947 psf = igmp_mcf_get_next(seq, v);
2952 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2955 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2956 if (likely(state->im)) {
2957 spin_unlock_bh(&state->im->lock);
2965 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2967 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2968 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2970 if (v == SEQ_START_TOKEN) {
2971 seq_puts(seq, "Idx Device MCA SRC INC EXC\n");
2975 "0x%08x %6lu %6lu\n",
2976 state->dev->ifindex, state->dev->name,
2977 ntohl(state->im->multiaddr),
2978 ntohl(psf->sf_inaddr),
2979 psf->sf_count[MCAST_INCLUDE],
2980 psf->sf_count[MCAST_EXCLUDE]);
2985 static const struct seq_operations igmp_mcf_seq_ops = {
2986 .start = igmp_mcf_seq_start,
2987 .next = igmp_mcf_seq_next,
2988 .stop = igmp_mcf_seq_stop,
2989 .show = igmp_mcf_seq_show,
2992 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2994 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2995 sizeof(struct igmp_mcf_iter_state));
2998 static const struct file_operations igmp_mcf_seq_fops = {
2999 .owner = THIS_MODULE,
3000 .open = igmp_mcf_seq_open,
3002 .llseek = seq_lseek,
3003 .release = seq_release_net,
3006 static int __net_init igmp_net_init(struct net *net)
3008 struct proc_dir_entry *pde;
3011 pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
3014 pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
3015 &igmp_mcf_seq_fops);
3018 err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
3019 SOCK_DGRAM, 0, net);
3021 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
3029 remove_proc_entry("mcfilter", net->proc_net);
3031 remove_proc_entry("igmp", net->proc_net);
3036 static void __net_exit igmp_net_exit(struct net *net)
3038 remove_proc_entry("mcfilter", net->proc_net);
3039 remove_proc_entry("igmp", net->proc_net);
3040 inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
3043 static struct pernet_operations igmp_net_ops = {
3044 .init = igmp_net_init,
3045 .exit = igmp_net_exit,
3049 static int igmp_netdev_event(struct notifier_block *this,
3050 unsigned long event, void *ptr)
3052 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3053 struct in_device *in_dev;
3056 case NETDEV_RESEND_IGMP:
3057 in_dev = __in_dev_get_rtnl(dev);
3059 ip_mc_rejoin_groups(in_dev);
3067 static struct notifier_block igmp_notifier = {
3068 .notifier_call = igmp_netdev_event,
3071 int __init igmp_mc_init(void)
3073 #if defined(CONFIG_PROC_FS)
3076 err = register_pernet_subsys(&igmp_net_ops);
3079 err = register_netdevice_notifier(&igmp_notifier);
3081 goto reg_notif_fail;
3085 unregister_pernet_subsys(&igmp_net_ops);
3088 return register_netdevice_notifier(&igmp_notifier);