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 <asm/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 (atomic_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 atomic_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 atomic_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 atomic_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 = (struct igmpv3_grec *)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 = (__be32 *)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 = (struct igmphdr *)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 in_device *in_dev = __in_dev_get_rcu(skb->dev);
1058 bool dropped = true;
1063 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1066 if (skb_checksum_simple_validate(skb))
1071 case IGMP_HOST_MEMBERSHIP_QUERY:
1072 dropped = igmp_heard_query(in_dev, skb, len);
1074 case IGMP_HOST_MEMBERSHIP_REPORT:
1075 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1076 /* Is it our report looped back? */
1077 if (rt_is_output_route(skb_rtable(skb)))
1079 /* don't rely on MC router hearing unicast reports */
1080 if (skb->pkt_type == PACKET_MULTICAST ||
1081 skb->pkt_type == PACKET_BROADCAST)
1082 dropped = igmp_heard_report(in_dev, ih->group);
1085 #ifdef CONFIG_IP_PIMSM_V1
1086 return pim_rcv_v1(skb);
1088 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1091 case IGMP_HOST_LEAVE_MESSAGE:
1093 case IGMP_MTRACE_RESP:
1111 * Add a filter to a device
1114 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1116 char buf[MAX_ADDR_LEN];
1117 struct net_device *dev = in_dev->dev;
1119 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1120 We will get multicast token leakage, when IFF_MULTICAST
1121 is changed. This check should be done in ndo_set_rx_mode
1122 routine. Something sort of:
1123 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1126 if (arp_mc_map(addr, buf, dev, 0) == 0)
1127 dev_mc_add(dev, buf);
1131 * Remove a filter from a device
1134 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1136 char buf[MAX_ADDR_LEN];
1137 struct net_device *dev = in_dev->dev;
1139 if (arp_mc_map(addr, buf, dev, 0) == 0)
1140 dev_mc_del(dev, buf);
1143 #ifdef CONFIG_IP_MULTICAST
1145 * deleted ip_mc_list manipulation
1147 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1149 struct ip_mc_list *pmc;
1150 struct net *net = dev_net(in_dev->dev);
1152 /* this is an "ip_mc_list" for convenience; only the fields below
1153 * are actually used. In particular, the refcnt and users are not
1154 * used for management of the delete list. Using the same structure
1155 * for deleted items allows change reports to use common code with
1156 * non-deleted or query-response MCA's.
1158 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1161 spin_lock_init(&pmc->lock);
1162 spin_lock_bh(&im->lock);
1163 pmc->interface = im->interface;
1164 in_dev_hold(in_dev);
1165 pmc->multiaddr = im->multiaddr;
1166 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1167 pmc->sfmode = im->sfmode;
1168 if (pmc->sfmode == MCAST_INCLUDE) {
1169 struct ip_sf_list *psf;
1171 pmc->tomb = im->tomb;
1172 pmc->sources = im->sources;
1173 im->tomb = im->sources = NULL;
1174 for (psf = pmc->sources; psf; psf = psf->sf_next)
1175 psf->sf_crcount = pmc->crcount;
1177 spin_unlock_bh(&im->lock);
1179 spin_lock_bh(&in_dev->mc_tomb_lock);
1180 pmc->next = in_dev->mc_tomb;
1181 in_dev->mc_tomb = pmc;
1182 spin_unlock_bh(&in_dev->mc_tomb_lock);
1186 * restore ip_mc_list deleted records
1188 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1190 struct ip_mc_list *pmc, *pmc_prev;
1191 struct ip_sf_list *psf;
1192 struct net *net = dev_net(in_dev->dev);
1193 __be32 multiaddr = im->multiaddr;
1195 spin_lock_bh(&in_dev->mc_tomb_lock);
1197 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1198 if (pmc->multiaddr == multiaddr)
1204 pmc_prev->next = pmc->next;
1206 in_dev->mc_tomb = pmc->next;
1208 spin_unlock_bh(&in_dev->mc_tomb_lock);
1210 spin_lock_bh(&im->lock);
1212 im->interface = pmc->interface;
1213 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1214 if (im->sfmode == MCAST_INCLUDE) {
1215 swap(im->tomb, pmc->tomb);
1216 swap(im->sources, pmc->sources);
1217 for (psf = im->sources; psf; psf = psf->sf_next)
1218 psf->sf_crcount = im->crcount;
1220 in_dev_put(pmc->interface);
1223 spin_unlock_bh(&im->lock);
1227 * flush ip_mc_list deleted records
1229 static void igmpv3_clear_delrec(struct in_device *in_dev)
1231 struct ip_mc_list *pmc, *nextpmc;
1233 spin_lock_bh(&in_dev->mc_tomb_lock);
1234 pmc = in_dev->mc_tomb;
1235 in_dev->mc_tomb = NULL;
1236 spin_unlock_bh(&in_dev->mc_tomb_lock);
1238 for (; pmc; pmc = nextpmc) {
1239 nextpmc = pmc->next;
1240 ip_mc_clear_src(pmc);
1241 in_dev_put(pmc->interface);
1244 /* clear dead sources, too */
1246 for_each_pmc_rcu(in_dev, pmc) {
1247 struct ip_sf_list *psf;
1249 spin_lock_bh(&pmc->lock);
1252 spin_unlock_bh(&pmc->lock);
1253 ip_sf_list_clear_all(psf);
1259 static void igmp_group_dropped(struct ip_mc_list *im)
1261 struct in_device *in_dev = im->interface;
1262 #ifdef CONFIG_IP_MULTICAST
1263 struct net *net = dev_net(in_dev->dev);
1269 ip_mc_filter_del(in_dev, im->multiaddr);
1272 #ifdef CONFIG_IP_MULTICAST
1273 if (im->multiaddr == IGMP_ALL_HOSTS)
1275 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1278 reporter = im->reporter;
1279 igmp_stop_timer(im);
1281 if (!in_dev->dead) {
1282 if (IGMP_V1_SEEN(in_dev))
1284 if (IGMP_V2_SEEN(in_dev)) {
1286 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1290 igmpv3_add_delrec(in_dev, im);
1292 igmp_ifc_event(in_dev);
1297 static void igmp_group_added(struct ip_mc_list *im)
1299 struct in_device *in_dev = im->interface;
1300 #ifdef CONFIG_IP_MULTICAST
1301 struct net *net = dev_net(in_dev->dev);
1304 if (im->loaded == 0) {
1306 ip_mc_filter_add(in_dev, im->multiaddr);
1309 #ifdef CONFIG_IP_MULTICAST
1310 if (im->multiaddr == IGMP_ALL_HOSTS)
1312 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1317 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1318 spin_lock_bh(&im->lock);
1319 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1320 spin_unlock_bh(&im->lock);
1325 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1326 igmp_ifc_event(in_dev);
1332 * Multicast list managers
1335 static u32 ip_mc_hash(const struct ip_mc_list *im)
1337 return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1340 static void ip_mc_hash_add(struct in_device *in_dev,
1341 struct ip_mc_list *im)
1343 struct ip_mc_list __rcu **mc_hash;
1346 mc_hash = rtnl_dereference(in_dev->mc_hash);
1348 hash = ip_mc_hash(im);
1349 im->next_hash = mc_hash[hash];
1350 rcu_assign_pointer(mc_hash[hash], im);
1354 /* do not use a hash table for small number of items */
1355 if (in_dev->mc_count < 4)
1358 mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1363 for_each_pmc_rtnl(in_dev, im) {
1364 hash = ip_mc_hash(im);
1365 im->next_hash = mc_hash[hash];
1366 RCU_INIT_POINTER(mc_hash[hash], im);
1369 rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1372 static void ip_mc_hash_remove(struct in_device *in_dev,
1373 struct ip_mc_list *im)
1375 struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1376 struct ip_mc_list *aux;
1380 mc_hash += ip_mc_hash(im);
1381 while ((aux = rtnl_dereference(*mc_hash)) != im)
1382 mc_hash = &aux->next_hash;
1383 *mc_hash = im->next_hash;
1388 * A socket has joined a multicast group on device dev.
1391 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1393 struct ip_mc_list *im;
1394 #ifdef CONFIG_IP_MULTICAST
1395 struct net *net = dev_net(in_dev->dev);
1400 for_each_pmc_rtnl(in_dev, im) {
1401 if (im->multiaddr == addr) {
1403 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1408 im = kzalloc(sizeof(*im), GFP_KERNEL);
1413 im->interface = in_dev;
1414 in_dev_hold(in_dev);
1415 im->multiaddr = addr;
1416 /* initial mode is (EX, empty) */
1417 im->sfmode = MCAST_EXCLUDE;
1418 im->sfcount[MCAST_EXCLUDE] = 1;
1419 atomic_set(&im->refcnt, 1);
1420 spin_lock_init(&im->lock);
1421 #ifdef CONFIG_IP_MULTICAST
1422 setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1423 im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1426 im->next_rcu = in_dev->mc_list;
1428 rcu_assign_pointer(in_dev->mc_list, im);
1430 ip_mc_hash_add(in_dev, im);
1432 #ifdef CONFIG_IP_MULTICAST
1433 igmpv3_del_delrec(in_dev, im);
1435 igmp_group_added(im);
1437 ip_rt_multicast_event(in_dev);
1441 EXPORT_SYMBOL(ip_mc_inc_group);
1443 static int ip_mc_check_iphdr(struct sk_buff *skb)
1445 const struct iphdr *iph;
1447 unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1449 if (!pskb_may_pull(skb, offset))
1454 if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1457 offset += ip_hdrlen(skb) - sizeof(*iph);
1459 if (!pskb_may_pull(skb, offset))
1464 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1467 len = skb_network_offset(skb) + ntohs(iph->tot_len);
1468 if (skb->len < len || len < offset)
1471 skb_set_transport_header(skb, offset);
1476 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1478 unsigned int len = skb_transport_offset(skb);
1480 len += sizeof(struct igmpv3_report);
1482 return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1485 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1487 unsigned int len = skb_transport_offset(skb);
1489 len += sizeof(struct igmphdr);
1494 if (skb->len != len) {
1496 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1497 if (skb->len < len || !pskb_may_pull(skb, len))
1501 /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1502 * all-systems destination addresses (224.0.0.1) for general queries
1504 if (!igmp_hdr(skb)->group &&
1505 ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1511 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1513 switch (igmp_hdr(skb)->type) {
1514 case IGMP_HOST_LEAVE_MESSAGE:
1515 case IGMP_HOST_MEMBERSHIP_REPORT:
1516 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1519 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1520 return ip_mc_check_igmp_reportv3(skb);
1521 case IGMP_HOST_MEMBERSHIP_QUERY:
1522 return ip_mc_check_igmp_query(skb);
1528 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1530 return skb_checksum_simple_validate(skb);
1533 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1536 struct sk_buff *skb_chk;
1537 unsigned int transport_len;
1538 unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1541 transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1543 skb_chk = skb_checksum_trimmed(skb, transport_len,
1544 ip_mc_validate_checksum);
1548 if (!pskb_may_pull(skb_chk, len))
1551 ret = ip_mc_check_igmp_msg(skb_chk);
1556 *skb_trimmed = skb_chk;
1557 /* free now unneeded clone */
1558 else if (skb_chk != skb)
1564 if (ret && skb_chk && skb_chk != skb)
1571 * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1572 * @skb: the skb to validate
1573 * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1575 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1576 * skb transport header accordingly and returns zero.
1578 * -EINVAL: A broken packet was detected, i.e. it violates some internet
1580 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1581 * -ENOMEM: A memory allocation failure happened.
1583 * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1584 * to NULL): After parsing an IGMP packet successfully it will point to
1585 * an skb which has its tail aligned to the IP packet end. This might
1586 * either be the originally provided skb or a trimmed, cloned version if
1587 * the skb frame had data beyond the IP packet. A cloned skb allows us
1588 * to leave the original skb and its full frame unchanged (which might be
1589 * desirable for layer 2 frame jugglers).
1591 * Caller needs to set the skb network header and free any returned skb if it
1592 * differs from the provided skb.
1594 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1596 int ret = ip_mc_check_iphdr(skb);
1601 if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1604 return __ip_mc_check_igmp(skb, skb_trimmed);
1606 EXPORT_SYMBOL(ip_mc_check_igmp);
1609 * Resend IGMP JOIN report; used by netdev notifier.
1611 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1613 #ifdef CONFIG_IP_MULTICAST
1614 struct ip_mc_list *im;
1616 struct net *net = dev_net(in_dev->dev);
1620 for_each_pmc_rtnl(in_dev, im) {
1621 if (im->multiaddr == IGMP_ALL_HOSTS)
1623 if (ipv4_is_local_multicast(im->multiaddr) &&
1624 !net->ipv4.sysctl_igmp_llm_reports)
1627 /* a failover is happening and switches
1628 * must be notified immediately
1630 if (IGMP_V1_SEEN(in_dev))
1631 type = IGMP_HOST_MEMBERSHIP_REPORT;
1632 else if (IGMP_V2_SEEN(in_dev))
1633 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1635 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1636 igmp_send_report(in_dev, im, type);
1642 * A socket has left a multicast group on device dev
1645 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1647 struct ip_mc_list *i;
1648 struct ip_mc_list __rcu **ip;
1652 for (ip = &in_dev->mc_list;
1653 (i = rtnl_dereference(*ip)) != NULL;
1654 ip = &i->next_rcu) {
1655 if (i->multiaddr == addr) {
1656 if (--i->users == 0) {
1657 ip_mc_hash_remove(in_dev, i);
1660 igmp_group_dropped(i);
1664 ip_rt_multicast_event(in_dev);
1673 EXPORT_SYMBOL(ip_mc_dec_group);
1675 /* Device changing type */
1677 void ip_mc_unmap(struct in_device *in_dev)
1679 struct ip_mc_list *pmc;
1683 for_each_pmc_rtnl(in_dev, pmc)
1684 igmp_group_dropped(pmc);
1687 void ip_mc_remap(struct in_device *in_dev)
1689 struct ip_mc_list *pmc;
1693 for_each_pmc_rtnl(in_dev, pmc) {
1694 #ifdef CONFIG_IP_MULTICAST
1695 igmpv3_del_delrec(in_dev, pmc);
1697 igmp_group_added(pmc);
1701 /* Device going down */
1703 void ip_mc_down(struct in_device *in_dev)
1705 struct ip_mc_list *pmc;
1709 for_each_pmc_rtnl(in_dev, pmc)
1710 igmp_group_dropped(pmc);
1712 #ifdef CONFIG_IP_MULTICAST
1713 in_dev->mr_ifc_count = 0;
1714 if (del_timer(&in_dev->mr_ifc_timer))
1715 __in_dev_put(in_dev);
1716 in_dev->mr_gq_running = 0;
1717 if (del_timer(&in_dev->mr_gq_timer))
1718 __in_dev_put(in_dev);
1721 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1724 void ip_mc_init_dev(struct in_device *in_dev)
1726 #ifdef CONFIG_IP_MULTICAST
1727 struct net *net = dev_net(in_dev->dev);
1731 #ifdef CONFIG_IP_MULTICAST
1732 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1733 (unsigned long)in_dev);
1734 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1735 (unsigned long)in_dev);
1736 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1739 spin_lock_init(&in_dev->mc_tomb_lock);
1742 /* Device going up */
1744 void ip_mc_up(struct in_device *in_dev)
1746 struct ip_mc_list *pmc;
1747 #ifdef CONFIG_IP_MULTICAST
1748 struct net *net = dev_net(in_dev->dev);
1753 #ifdef CONFIG_IP_MULTICAST
1754 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1756 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1758 for_each_pmc_rtnl(in_dev, pmc) {
1759 #ifdef CONFIG_IP_MULTICAST
1760 igmpv3_del_delrec(in_dev, pmc);
1762 igmp_group_added(pmc);
1767 * Device is about to be destroyed: clean up.
1770 void ip_mc_destroy_dev(struct in_device *in_dev)
1772 struct ip_mc_list *i;
1776 /* Deactivate timers */
1778 #ifdef CONFIG_IP_MULTICAST
1779 igmpv3_clear_delrec(in_dev);
1782 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1783 in_dev->mc_list = i->next_rcu;
1790 /* RTNL is locked */
1791 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1793 struct net_device *dev = NULL;
1794 struct in_device *idev = NULL;
1796 if (imr->imr_ifindex) {
1797 idev = inetdev_by_index(net, imr->imr_ifindex);
1800 if (imr->imr_address.s_addr) {
1801 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1807 struct rtable *rt = ip_route_output(net,
1808 imr->imr_multiaddr.s_addr,
1816 imr->imr_ifindex = dev->ifindex;
1817 idev = __in_dev_get_rtnl(dev);
1823 * Join a socket to a group
1826 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1829 struct ip_sf_list *psf, *psf_prev;
1833 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1834 if (psf->sf_inaddr == *psfsrc)
1838 if (!psf || psf->sf_count[sfmode] == 0) {
1839 /* source filter not found, or count wrong => bug */
1842 psf->sf_count[sfmode]--;
1843 if (psf->sf_count[sfmode] == 0) {
1844 ip_rt_multicast_event(pmc->interface);
1846 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1847 #ifdef CONFIG_IP_MULTICAST
1848 struct in_device *in_dev = pmc->interface;
1849 struct net *net = dev_net(in_dev->dev);
1852 /* no more filters for this source */
1854 psf_prev->sf_next = psf->sf_next;
1856 pmc->sources = psf->sf_next;
1857 #ifdef CONFIG_IP_MULTICAST
1858 if (psf->sf_oldin &&
1859 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1860 psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1861 psf->sf_next = pmc->tomb;
1871 #ifndef CONFIG_IP_MULTICAST
1872 #define igmp_ifc_event(x) do { } while (0)
1875 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1876 int sfcount, __be32 *psfsrc, int delta)
1878 struct ip_mc_list *pmc;
1885 for_each_pmc_rcu(in_dev, pmc) {
1886 if (*pmca == pmc->multiaddr)
1890 /* MCA not found?? bug */
1894 spin_lock_bh(&pmc->lock);
1896 #ifdef CONFIG_IP_MULTICAST
1901 if (!pmc->sfcount[sfmode])
1903 pmc->sfcount[sfmode]--;
1906 for (i = 0; i < sfcount; i++) {
1907 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1909 changerec |= rv > 0;
1913 if (pmc->sfmode == MCAST_EXCLUDE &&
1914 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1915 pmc->sfcount[MCAST_INCLUDE]) {
1916 #ifdef CONFIG_IP_MULTICAST
1917 struct ip_sf_list *psf;
1918 struct net *net = dev_net(in_dev->dev);
1921 /* filter mode change */
1922 pmc->sfmode = MCAST_INCLUDE;
1923 #ifdef CONFIG_IP_MULTICAST
1924 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1925 in_dev->mr_ifc_count = pmc->crcount;
1926 for (psf = pmc->sources; psf; psf = psf->sf_next)
1927 psf->sf_crcount = 0;
1928 igmp_ifc_event(pmc->interface);
1929 } else if (sf_setstate(pmc) || changerec) {
1930 igmp_ifc_event(pmc->interface);
1934 spin_unlock_bh(&pmc->lock);
1939 * Add multicast single-source filter to the interface list
1941 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1944 struct ip_sf_list *psf, *psf_prev;
1947 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1948 if (psf->sf_inaddr == *psfsrc)
1953 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1956 psf->sf_inaddr = *psfsrc;
1958 psf_prev->sf_next = psf;
1962 psf->sf_count[sfmode]++;
1963 if (psf->sf_count[sfmode] == 1) {
1964 ip_rt_multicast_event(pmc->interface);
1969 #ifdef CONFIG_IP_MULTICAST
1970 static void sf_markstate(struct ip_mc_list *pmc)
1972 struct ip_sf_list *psf;
1973 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1975 for (psf = pmc->sources; psf; psf = psf->sf_next)
1976 if (pmc->sfcount[MCAST_EXCLUDE]) {
1977 psf->sf_oldin = mca_xcount ==
1978 psf->sf_count[MCAST_EXCLUDE] &&
1979 !psf->sf_count[MCAST_INCLUDE];
1981 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1984 static int sf_setstate(struct ip_mc_list *pmc)
1986 struct ip_sf_list *psf, *dpsf;
1987 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1988 int qrv = pmc->interface->mr_qrv;
1992 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1993 if (pmc->sfcount[MCAST_EXCLUDE]) {
1994 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1995 !psf->sf_count[MCAST_INCLUDE];
1997 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1999 if (!psf->sf_oldin) {
2000 struct ip_sf_list *prev = NULL;
2002 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
2003 if (dpsf->sf_inaddr == psf->sf_inaddr)
2009 prev->sf_next = dpsf->sf_next;
2011 pmc->tomb = dpsf->sf_next;
2014 psf->sf_crcount = qrv;
2017 } else if (psf->sf_oldin) {
2019 psf->sf_crcount = 0;
2021 * add or update "delete" records if an active filter
2024 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
2025 if (dpsf->sf_inaddr == psf->sf_inaddr)
2028 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2032 /* pmc->lock held by callers */
2033 dpsf->sf_next = pmc->tomb;
2036 dpsf->sf_crcount = qrv;
2045 * Add multicast source filter list to the interface list
2047 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2048 int sfcount, __be32 *psfsrc, int delta)
2050 struct ip_mc_list *pmc;
2057 for_each_pmc_rcu(in_dev, pmc) {
2058 if (*pmca == pmc->multiaddr)
2062 /* MCA not found?? bug */
2066 spin_lock_bh(&pmc->lock);
2069 #ifdef CONFIG_IP_MULTICAST
2072 isexclude = pmc->sfmode == MCAST_EXCLUDE;
2074 pmc->sfcount[sfmode]++;
2076 for (i = 0; i < sfcount; i++) {
2077 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2085 pmc->sfcount[sfmode]--;
2086 for (j = 0; j < i; j++)
2087 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2088 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2089 #ifdef CONFIG_IP_MULTICAST
2090 struct ip_sf_list *psf;
2091 struct net *net = dev_net(pmc->interface->dev);
2092 in_dev = pmc->interface;
2095 /* filter mode change */
2096 if (pmc->sfcount[MCAST_EXCLUDE])
2097 pmc->sfmode = MCAST_EXCLUDE;
2098 else if (pmc->sfcount[MCAST_INCLUDE])
2099 pmc->sfmode = MCAST_INCLUDE;
2100 #ifdef CONFIG_IP_MULTICAST
2101 /* else no filters; keep old mode for reports */
2103 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2104 in_dev->mr_ifc_count = pmc->crcount;
2105 for (psf = pmc->sources; psf; psf = psf->sf_next)
2106 psf->sf_crcount = 0;
2107 igmp_ifc_event(in_dev);
2108 } else if (sf_setstate(pmc)) {
2109 igmp_ifc_event(in_dev);
2112 spin_unlock_bh(&pmc->lock);
2116 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2118 struct ip_sf_list *tomb, *sources;
2120 spin_lock_bh(&pmc->lock);
2123 sources = pmc->sources;
2124 pmc->sources = NULL;
2125 pmc->sfmode = MCAST_EXCLUDE;
2126 pmc->sfcount[MCAST_INCLUDE] = 0;
2127 pmc->sfcount[MCAST_EXCLUDE] = 1;
2128 spin_unlock_bh(&pmc->lock);
2130 ip_sf_list_clear_all(tomb);
2131 ip_sf_list_clear_all(sources);
2134 /* Join a multicast group
2137 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2139 __be32 addr = imr->imr_multiaddr.s_addr;
2140 struct ip_mc_socklist *iml, *i;
2141 struct in_device *in_dev;
2142 struct inet_sock *inet = inet_sk(sk);
2143 struct net *net = sock_net(sk);
2150 if (!ipv4_is_multicast(addr))
2153 in_dev = ip_mc_find_dev(net, imr);
2161 ifindex = imr->imr_ifindex;
2162 for_each_pmc_rtnl(inet, i) {
2163 if (i->multi.imr_multiaddr.s_addr == addr &&
2164 i->multi.imr_ifindex == ifindex)
2169 if (count >= net->ipv4.sysctl_igmp_max_memberships)
2171 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2175 memcpy(&iml->multi, imr, sizeof(*imr));
2176 iml->next_rcu = inet->mc_list;
2178 iml->sfmode = MCAST_EXCLUDE;
2179 rcu_assign_pointer(inet->mc_list, iml);
2180 ip_mc_inc_group(in_dev, addr);
2185 EXPORT_SYMBOL(ip_mc_join_group);
2187 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2188 struct in_device *in_dev)
2190 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2194 /* any-source empty exclude case */
2195 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2196 iml->sfmode, 0, NULL, 0);
2198 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2199 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2200 RCU_INIT_POINTER(iml->sflist, NULL);
2201 /* decrease mem now to avoid the memleak warning */
2202 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2203 kfree_rcu(psf, rcu);
2207 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2209 struct inet_sock *inet = inet_sk(sk);
2210 struct ip_mc_socklist *iml;
2211 struct ip_mc_socklist __rcu **imlp;
2212 struct in_device *in_dev;
2213 struct net *net = sock_net(sk);
2214 __be32 group = imr->imr_multiaddr.s_addr;
2216 int ret = -EADDRNOTAVAIL;
2220 in_dev = ip_mc_find_dev(net, imr);
2221 if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2225 ifindex = imr->imr_ifindex;
2226 for (imlp = &inet->mc_list;
2227 (iml = rtnl_dereference(*imlp)) != NULL;
2228 imlp = &iml->next_rcu) {
2229 if (iml->multi.imr_multiaddr.s_addr != group)
2232 if (iml->multi.imr_ifindex != ifindex)
2234 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2235 iml->multi.imr_address.s_addr)
2238 (void) ip_mc_leave_src(sk, iml, in_dev);
2240 *imlp = iml->next_rcu;
2243 ip_mc_dec_group(in_dev, group);
2245 /* decrease mem now to avoid the memleak warning */
2246 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2247 kfree_rcu(iml, rcu);
2253 EXPORT_SYMBOL(ip_mc_leave_group);
2255 int ip_mc_source(int add, int omode, struct sock *sk, struct
2256 ip_mreq_source *mreqs, int ifindex)
2259 struct ip_mreqn imr;
2260 __be32 addr = mreqs->imr_multiaddr;
2261 struct ip_mc_socklist *pmc;
2262 struct in_device *in_dev = NULL;
2263 struct inet_sock *inet = inet_sk(sk);
2264 struct ip_sf_socklist *psl;
2265 struct net *net = sock_net(sk);
2269 if (!ipv4_is_multicast(addr))
2274 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2275 imr.imr_address.s_addr = mreqs->imr_interface;
2276 imr.imr_ifindex = ifindex;
2277 in_dev = ip_mc_find_dev(net, &imr);
2283 err = -EADDRNOTAVAIL;
2285 for_each_pmc_rtnl(inet, pmc) {
2286 if ((pmc->multi.imr_multiaddr.s_addr ==
2287 imr.imr_multiaddr.s_addr) &&
2288 (pmc->multi.imr_ifindex == imr.imr_ifindex))
2291 if (!pmc) { /* must have a prior join */
2295 /* if a source filter was set, must be the same mode as before */
2297 if (pmc->sfmode != omode) {
2301 } else if (pmc->sfmode != omode) {
2302 /* allow mode switches for empty-set filters */
2303 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2304 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2306 pmc->sfmode = omode;
2309 psl = rtnl_dereference(pmc->sflist);
2312 goto done; /* err = -EADDRNOTAVAIL */
2314 for (i = 0; i < psl->sl_count; i++) {
2315 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2320 if (rv) /* source not found */
2321 goto done; /* err = -EADDRNOTAVAIL */
2323 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2324 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2329 /* update the interface filter */
2330 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2331 &mreqs->imr_sourceaddr, 1);
2333 for (j = i+1; j < psl->sl_count; j++)
2334 psl->sl_addr[j-1] = psl->sl_addr[j];
2339 /* else, add a new source to the filter */
2341 if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2345 if (!psl || psl->sl_count == psl->sl_max) {
2346 struct ip_sf_socklist *newpsl;
2347 int count = IP_SFBLOCK;
2350 count += psl->sl_max;
2351 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2356 newpsl->sl_max = count;
2357 newpsl->sl_count = count - IP_SFBLOCK;
2359 for (i = 0; i < psl->sl_count; i++)
2360 newpsl->sl_addr[i] = psl->sl_addr[i];
2361 /* decrease mem now to avoid the memleak warning */
2362 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2363 kfree_rcu(psl, rcu);
2365 rcu_assign_pointer(pmc->sflist, newpsl);
2368 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2369 for (i = 0; i < psl->sl_count; i++) {
2370 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2375 if (rv == 0) /* address already there is an error */
2377 for (j = psl->sl_count-1; j >= i; j--)
2378 psl->sl_addr[j+1] = psl->sl_addr[j];
2379 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2382 /* update the interface list */
2383 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2384 &mreqs->imr_sourceaddr, 1);
2387 err = ip_mc_leave_group(sk, &imr);
2391 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2394 struct ip_mreqn imr;
2395 __be32 addr = msf->imsf_multiaddr;
2396 struct ip_mc_socklist *pmc;
2397 struct in_device *in_dev;
2398 struct inet_sock *inet = inet_sk(sk);
2399 struct ip_sf_socklist *newpsl, *psl;
2400 struct net *net = sock_net(sk);
2403 if (!ipv4_is_multicast(addr))
2405 if (msf->imsf_fmode != MCAST_INCLUDE &&
2406 msf->imsf_fmode != MCAST_EXCLUDE)
2411 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2412 imr.imr_address.s_addr = msf->imsf_interface;
2413 imr.imr_ifindex = ifindex;
2414 in_dev = ip_mc_find_dev(net, &imr);
2421 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2422 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2427 for_each_pmc_rtnl(inet, pmc) {
2428 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2429 pmc->multi.imr_ifindex == imr.imr_ifindex)
2432 if (!pmc) { /* must have a prior join */
2436 if (msf->imsf_numsrc) {
2437 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2443 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2444 memcpy(newpsl->sl_addr, msf->imsf_slist,
2445 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2446 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2447 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2449 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2454 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2455 msf->imsf_fmode, 0, NULL, 0);
2457 psl = rtnl_dereference(pmc->sflist);
2459 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2460 psl->sl_count, psl->sl_addr, 0);
2461 /* decrease mem now to avoid the memleak warning */
2462 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2463 kfree_rcu(psl, rcu);
2465 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2467 rcu_assign_pointer(pmc->sflist, newpsl);
2468 pmc->sfmode = msf->imsf_fmode;
2472 err = ip_mc_leave_group(sk, &imr);
2476 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2477 struct ip_msfilter __user *optval, int __user *optlen)
2479 int err, len, count, copycount;
2480 struct ip_mreqn imr;
2481 __be32 addr = msf->imsf_multiaddr;
2482 struct ip_mc_socklist *pmc;
2483 struct in_device *in_dev;
2484 struct inet_sock *inet = inet_sk(sk);
2485 struct ip_sf_socklist *psl;
2486 struct net *net = sock_net(sk);
2490 if (!ipv4_is_multicast(addr))
2493 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2494 imr.imr_address.s_addr = msf->imsf_interface;
2495 imr.imr_ifindex = 0;
2496 in_dev = ip_mc_find_dev(net, &imr);
2502 err = -EADDRNOTAVAIL;
2504 for_each_pmc_rtnl(inet, pmc) {
2505 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2506 pmc->multi.imr_ifindex == imr.imr_ifindex)
2509 if (!pmc) /* must have a prior join */
2511 msf->imsf_fmode = pmc->sfmode;
2512 psl = rtnl_dereference(pmc->sflist);
2517 count = psl->sl_count;
2519 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2520 len = copycount * sizeof(psl->sl_addr[0]);
2521 msf->imsf_numsrc = count;
2522 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2523 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2527 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2534 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2535 struct group_filter __user *optval, int __user *optlen)
2537 int err, i, count, copycount;
2538 struct sockaddr_in *psin;
2540 struct ip_mc_socklist *pmc;
2541 struct inet_sock *inet = inet_sk(sk);
2542 struct ip_sf_socklist *psl;
2546 psin = (struct sockaddr_in *)&gsf->gf_group;
2547 if (psin->sin_family != AF_INET)
2549 addr = psin->sin_addr.s_addr;
2550 if (!ipv4_is_multicast(addr))
2553 err = -EADDRNOTAVAIL;
2555 for_each_pmc_rtnl(inet, pmc) {
2556 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2557 pmc->multi.imr_ifindex == gsf->gf_interface)
2560 if (!pmc) /* must have a prior join */
2562 gsf->gf_fmode = pmc->sfmode;
2563 psl = rtnl_dereference(pmc->sflist);
2564 count = psl ? psl->sl_count : 0;
2565 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2566 gsf->gf_numsrc = count;
2567 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2568 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2571 for (i = 0; i < copycount; i++) {
2572 struct sockaddr_storage ss;
2574 psin = (struct sockaddr_in *)&ss;
2575 memset(&ss, 0, sizeof(ss));
2576 psin->sin_family = AF_INET;
2577 psin->sin_addr.s_addr = psl->sl_addr[i];
2578 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2587 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2589 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2591 struct inet_sock *inet = inet_sk(sk);
2592 struct ip_mc_socklist *pmc;
2593 struct ip_sf_socklist *psl;
2598 if (!ipv4_is_multicast(loc_addr))
2602 for_each_pmc_rcu(inet, pmc) {
2603 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2604 pmc->multi.imr_ifindex == dif)
2610 psl = rcu_dereference(pmc->sflist);
2611 ret = (pmc->sfmode == MCAST_EXCLUDE);
2615 for (i = 0; i < psl->sl_count; i++) {
2616 if (psl->sl_addr[i] == rmt_addr)
2620 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2622 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2632 * A socket is closing.
2635 void ip_mc_drop_socket(struct sock *sk)
2637 struct inet_sock *inet = inet_sk(sk);
2638 struct ip_mc_socklist *iml;
2639 struct net *net = sock_net(sk);
2645 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2646 struct in_device *in_dev;
2648 inet->mc_list = iml->next_rcu;
2649 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2650 (void) ip_mc_leave_src(sk, iml, in_dev);
2652 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2653 /* decrease mem now to avoid the memleak warning */
2654 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2655 kfree_rcu(iml, rcu);
2660 /* called with rcu_read_lock() */
2661 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2663 struct ip_mc_list *im;
2664 struct ip_mc_list __rcu **mc_hash;
2665 struct ip_sf_list *psf;
2668 mc_hash = rcu_dereference(in_dev->mc_hash);
2670 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2672 for (im = rcu_dereference(mc_hash[hash]);
2674 im = rcu_dereference(im->next_hash)) {
2675 if (im->multiaddr == mc_addr)
2679 for_each_pmc_rcu(in_dev, im) {
2680 if (im->multiaddr == mc_addr)
2684 if (im && proto == IPPROTO_IGMP) {
2688 spin_lock_bh(&im->lock);
2689 for (psf = im->sources; psf; psf = psf->sf_next) {
2690 if (psf->sf_inaddr == src_addr)
2694 rv = psf->sf_count[MCAST_INCLUDE] ||
2695 psf->sf_count[MCAST_EXCLUDE] !=
2696 im->sfcount[MCAST_EXCLUDE];
2698 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2699 spin_unlock_bh(&im->lock);
2701 rv = 1; /* unspecified source; tentatively allow */
2706 #if defined(CONFIG_PROC_FS)
2707 struct igmp_mc_iter_state {
2708 struct seq_net_private p;
2709 struct net_device *dev;
2710 struct in_device *in_dev;
2713 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2715 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2717 struct net *net = seq_file_net(seq);
2718 struct ip_mc_list *im = NULL;
2719 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2721 state->in_dev = NULL;
2722 for_each_netdev_rcu(net, state->dev) {
2723 struct in_device *in_dev;
2725 in_dev = __in_dev_get_rcu(state->dev);
2728 im = rcu_dereference(in_dev->mc_list);
2730 state->in_dev = in_dev;
2737 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2739 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2741 im = rcu_dereference(im->next_rcu);
2743 state->dev = next_net_device_rcu(state->dev);
2745 state->in_dev = NULL;
2748 state->in_dev = __in_dev_get_rcu(state->dev);
2751 im = rcu_dereference(state->in_dev->mc_list);
2756 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2758 struct ip_mc_list *im = igmp_mc_get_first(seq);
2760 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2762 return pos ? NULL : im;
2765 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2769 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2772 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2774 struct ip_mc_list *im;
2775 if (v == SEQ_START_TOKEN)
2776 im = igmp_mc_get_first(seq);
2778 im = igmp_mc_get_next(seq, v);
2783 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2786 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2788 state->in_dev = NULL;
2793 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2795 if (v == SEQ_START_TOKEN)
2797 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2799 struct ip_mc_list *im = (struct ip_mc_list *)v;
2800 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2804 #ifdef CONFIG_IP_MULTICAST
2805 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2806 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2812 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2813 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2814 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2817 delta = im->timer.expires - jiffies;
2819 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2820 im->multiaddr, im->users,
2822 im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2828 static const struct seq_operations igmp_mc_seq_ops = {
2829 .start = igmp_mc_seq_start,
2830 .next = igmp_mc_seq_next,
2831 .stop = igmp_mc_seq_stop,
2832 .show = igmp_mc_seq_show,
2835 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2837 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2838 sizeof(struct igmp_mc_iter_state));
2841 static const struct file_operations igmp_mc_seq_fops = {
2842 .owner = THIS_MODULE,
2843 .open = igmp_mc_seq_open,
2845 .llseek = seq_lseek,
2846 .release = seq_release_net,
2849 struct igmp_mcf_iter_state {
2850 struct seq_net_private p;
2851 struct net_device *dev;
2852 struct in_device *idev;
2853 struct ip_mc_list *im;
2856 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2858 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2860 struct net *net = seq_file_net(seq);
2861 struct ip_sf_list *psf = NULL;
2862 struct ip_mc_list *im = NULL;
2863 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2867 for_each_netdev_rcu(net, state->dev) {
2868 struct in_device *idev;
2869 idev = __in_dev_get_rcu(state->dev);
2870 if (unlikely(!idev))
2872 im = rcu_dereference(idev->mc_list);
2874 spin_lock_bh(&im->lock);
2881 spin_unlock_bh(&im->lock);
2887 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2889 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2893 spin_unlock_bh(&state->im->lock);
2894 state->im = state->im->next;
2895 while (!state->im) {
2896 state->dev = next_net_device_rcu(state->dev);
2901 state->idev = __in_dev_get_rcu(state->dev);
2904 state->im = rcu_dereference(state->idev->mc_list);
2908 spin_lock_bh(&state->im->lock);
2909 psf = state->im->sources;
2915 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2917 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2919 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2921 return pos ? NULL : psf;
2924 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2928 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2931 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2933 struct ip_sf_list *psf;
2934 if (v == SEQ_START_TOKEN)
2935 psf = igmp_mcf_get_first(seq);
2937 psf = igmp_mcf_get_next(seq, v);
2942 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2945 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2946 if (likely(state->im)) {
2947 spin_unlock_bh(&state->im->lock);
2955 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2957 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2958 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2960 if (v == SEQ_START_TOKEN) {
2961 seq_puts(seq, "Idx Device MCA SRC INC EXC\n");
2965 "0x%08x %6lu %6lu\n",
2966 state->dev->ifindex, state->dev->name,
2967 ntohl(state->im->multiaddr),
2968 ntohl(psf->sf_inaddr),
2969 psf->sf_count[MCAST_INCLUDE],
2970 psf->sf_count[MCAST_EXCLUDE]);
2975 static const struct seq_operations igmp_mcf_seq_ops = {
2976 .start = igmp_mcf_seq_start,
2977 .next = igmp_mcf_seq_next,
2978 .stop = igmp_mcf_seq_stop,
2979 .show = igmp_mcf_seq_show,
2982 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2984 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2985 sizeof(struct igmp_mcf_iter_state));
2988 static const struct file_operations igmp_mcf_seq_fops = {
2989 .owner = THIS_MODULE,
2990 .open = igmp_mcf_seq_open,
2992 .llseek = seq_lseek,
2993 .release = seq_release_net,
2996 static int __net_init igmp_net_init(struct net *net)
2998 struct proc_dir_entry *pde;
3001 pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
3004 pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
3005 &igmp_mcf_seq_fops);
3008 err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
3009 SOCK_DGRAM, 0, net);
3011 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
3019 remove_proc_entry("mcfilter", net->proc_net);
3021 remove_proc_entry("igmp", net->proc_net);
3026 static void __net_exit igmp_net_exit(struct net *net)
3028 remove_proc_entry("mcfilter", net->proc_net);
3029 remove_proc_entry("igmp", net->proc_net);
3030 inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
3033 static struct pernet_operations igmp_net_ops = {
3034 .init = igmp_net_init,
3035 .exit = igmp_net_exit,
3039 static int igmp_netdev_event(struct notifier_block *this,
3040 unsigned long event, void *ptr)
3042 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3043 struct in_device *in_dev;
3046 case NETDEV_RESEND_IGMP:
3047 in_dev = __in_dev_get_rtnl(dev);
3049 ip_mc_rejoin_groups(in_dev);
3057 static struct notifier_block igmp_notifier = {
3058 .notifier_call = igmp_netdev_event,
3061 int __init igmp_mc_init(void)
3063 #if defined(CONFIG_PROC_FS)
3066 err = register_pernet_subsys(&igmp_net_ops);
3069 err = register_netdevice_notifier(&igmp_notifier);
3071 goto reg_notif_fail;
3075 unregister_pernet_subsys(&igmp_net_ops);
3078 return register_netdevice_notifier(&igmp_notifier);