Linux 3.12.28
[linux/fpc-iii.git] / net / ipv4 / igmp.c
blob94d40cc793220171d98830010c6f8cdd2b4c783d
1 /*
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.
11 * Authors:
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.
19 * Fixes:
21 * Alan Cox : Added lots of __inline__ to optimise
22 * the memory usage of all the tiny little
23 * functions.
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
32 * points.
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"
53 * message (960131).
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
58 * true (960216).
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
70 * Vinay Kulkarni
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>
82 #include <linux/in.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>
93 #include <net/net_namespace.h>
94 #include <net/arp.h>
95 #include <net/ip.h>
96 #include <net/protocol.h>
97 #include <net/route.h>
98 #include <net/sock.h>
99 #include <net/checksum.h>
100 #include <linux/netfilter_ipv4.h>
101 #ifdef CONFIG_IP_MROUTE
102 #include <linux/mroute.h>
103 #endif
104 #ifdef CONFIG_PROC_FS
105 #include <linux/proc_fs.h>
106 #include <linux/seq_file.h>
107 #endif
109 #define IP_MAX_MEMBERSHIPS 20
110 #define IP_MAX_MSF 10
112 #ifdef CONFIG_IP_MULTICAST
113 /* Parameter names and values are taken from igmp-v2-06 draft */
115 #define IGMP_V1_Router_Present_Timeout (400*HZ)
116 #define IGMP_V2_Router_Present_Timeout (400*HZ)
117 #define IGMP_V2_Unsolicited_Report_Interval (10*HZ)
118 #define IGMP_V3_Unsolicited_Report_Interval (1*HZ)
119 #define IGMP_Query_Response_Interval (10*HZ)
120 #define IGMP_Unsolicited_Report_Count 2
123 #define IGMP_Initial_Report_Delay (1)
125 /* IGMP_Initial_Report_Delay is not from IGMP specs!
126 * IGMP specs require to report membership immediately after
127 * joining a group, but we delay the first report by a
128 * small interval. It seems more natural and still does not
129 * contradict to specs provided this delay is small enough.
132 #define IGMP_V1_SEEN(in_dev) \
133 (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
134 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
135 ((in_dev)->mr_v1_seen && \
136 time_before(jiffies, (in_dev)->mr_v1_seen)))
137 #define IGMP_V2_SEEN(in_dev) \
138 (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
139 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
140 ((in_dev)->mr_v2_seen && \
141 time_before(jiffies, (in_dev)->mr_v2_seen)))
143 static int unsolicited_report_interval(struct in_device *in_dev)
145 int interval_ms, interval_jiffies;
147 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
148 interval_ms = IN_DEV_CONF_GET(
149 in_dev,
150 IGMPV2_UNSOLICITED_REPORT_INTERVAL);
151 else /* v3 */
152 interval_ms = IN_DEV_CONF_GET(
153 in_dev,
154 IGMPV3_UNSOLICITED_REPORT_INTERVAL);
156 interval_jiffies = msecs_to_jiffies(interval_ms);
158 /* _timer functions can't handle a delay of 0 jiffies so ensure
159 * we always return a positive value.
161 if (interval_jiffies <= 0)
162 interval_jiffies = 1;
163 return interval_jiffies;
166 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
167 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
168 static void igmpv3_clear_delrec(struct in_device *in_dev);
169 static int sf_setstate(struct ip_mc_list *pmc);
170 static void sf_markstate(struct ip_mc_list *pmc);
171 #endif
172 static void ip_mc_clear_src(struct ip_mc_list *pmc);
173 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
174 int sfcount, __be32 *psfsrc, int delta);
176 static void ip_ma_put(struct ip_mc_list *im)
178 if (atomic_dec_and_test(&im->refcnt)) {
179 in_dev_put(im->interface);
180 kfree_rcu(im, rcu);
184 #define for_each_pmc_rcu(in_dev, pmc) \
185 for (pmc = rcu_dereference(in_dev->mc_list); \
186 pmc != NULL; \
187 pmc = rcu_dereference(pmc->next_rcu))
189 #define for_each_pmc_rtnl(in_dev, pmc) \
190 for (pmc = rtnl_dereference(in_dev->mc_list); \
191 pmc != NULL; \
192 pmc = rtnl_dereference(pmc->next_rcu))
194 #ifdef CONFIG_IP_MULTICAST
197 * Timer management
200 static void igmp_stop_timer(struct ip_mc_list *im)
202 spin_lock_bh(&im->lock);
203 if (del_timer(&im->timer))
204 atomic_dec(&im->refcnt);
205 im->tm_running = 0;
206 im->reporter = 0;
207 im->unsolicit_count = 0;
208 spin_unlock_bh(&im->lock);
211 /* It must be called with locked im->lock */
212 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
214 int tv = net_random() % max_delay;
216 im->tm_running = 1;
217 if (!mod_timer(&im->timer, jiffies+tv+2))
218 atomic_inc(&im->refcnt);
221 static void igmp_gq_start_timer(struct in_device *in_dev)
223 int tv = net_random() % in_dev->mr_maxdelay;
225 in_dev->mr_gq_running = 1;
226 if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
227 in_dev_hold(in_dev);
230 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
232 int tv = net_random() % delay;
234 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
235 in_dev_hold(in_dev);
238 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
240 spin_lock_bh(&im->lock);
241 im->unsolicit_count = 0;
242 if (del_timer(&im->timer)) {
243 if ((long)(im->timer.expires-jiffies) < max_delay) {
244 add_timer(&im->timer);
245 im->tm_running = 1;
246 spin_unlock_bh(&im->lock);
247 return;
249 atomic_dec(&im->refcnt);
251 igmp_start_timer(im, max_delay);
252 spin_unlock_bh(&im->lock);
257 * Send an IGMP report.
260 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
263 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
264 int gdeleted, int sdeleted)
266 switch (type) {
267 case IGMPV3_MODE_IS_INCLUDE:
268 case IGMPV3_MODE_IS_EXCLUDE:
269 if (gdeleted || sdeleted)
270 return 0;
271 if (!(pmc->gsquery && !psf->sf_gsresp)) {
272 if (pmc->sfmode == MCAST_INCLUDE)
273 return 1;
274 /* don't include if this source is excluded
275 * in all filters
277 if (psf->sf_count[MCAST_INCLUDE])
278 return type == IGMPV3_MODE_IS_INCLUDE;
279 return pmc->sfcount[MCAST_EXCLUDE] ==
280 psf->sf_count[MCAST_EXCLUDE];
282 return 0;
283 case IGMPV3_CHANGE_TO_INCLUDE:
284 if (gdeleted || sdeleted)
285 return 0;
286 return psf->sf_count[MCAST_INCLUDE] != 0;
287 case IGMPV3_CHANGE_TO_EXCLUDE:
288 if (gdeleted || sdeleted)
289 return 0;
290 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
291 psf->sf_count[MCAST_INCLUDE])
292 return 0;
293 return pmc->sfcount[MCAST_EXCLUDE] ==
294 psf->sf_count[MCAST_EXCLUDE];
295 case IGMPV3_ALLOW_NEW_SOURCES:
296 if (gdeleted || !psf->sf_crcount)
297 return 0;
298 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
299 case IGMPV3_BLOCK_OLD_SOURCES:
300 if (pmc->sfmode == MCAST_INCLUDE)
301 return gdeleted || (psf->sf_crcount && sdeleted);
302 return psf->sf_crcount && !gdeleted && !sdeleted;
304 return 0;
307 static int
308 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
310 struct ip_sf_list *psf;
311 int scount = 0;
313 for (psf=pmc->sources; psf; psf=psf->sf_next) {
314 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
315 continue;
316 scount++;
318 return scount;
321 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
323 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
325 struct sk_buff *skb;
326 struct rtable *rt;
327 struct iphdr *pip;
328 struct igmpv3_report *pig;
329 struct net *net = dev_net(dev);
330 struct flowi4 fl4;
331 int hlen = LL_RESERVED_SPACE(dev);
332 int tlen = dev->needed_tailroom;
334 while (1) {
335 skb = alloc_skb(size + hlen + tlen,
336 GFP_ATOMIC | __GFP_NOWARN);
337 if (skb)
338 break;
339 size >>= 1;
340 if (size < 256)
341 return NULL;
343 skb->priority = TC_PRIO_CONTROL;
344 igmp_skb_size(skb) = size;
346 rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
347 0, 0,
348 IPPROTO_IGMP, 0, dev->ifindex);
349 if (IS_ERR(rt)) {
350 kfree_skb(skb);
351 return NULL;
354 skb_dst_set(skb, &rt->dst);
355 skb->dev = dev;
357 skb_reserve(skb, hlen);
359 skb_reset_network_header(skb);
360 pip = ip_hdr(skb);
361 skb_put(skb, sizeof(struct iphdr) + 4);
363 pip->version = 4;
364 pip->ihl = (sizeof(struct iphdr)+4)>>2;
365 pip->tos = 0xc0;
366 pip->frag_off = htons(IP_DF);
367 pip->ttl = 1;
368 pip->daddr = fl4.daddr;
369 pip->saddr = fl4.saddr;
370 pip->protocol = IPPROTO_IGMP;
371 pip->tot_len = 0; /* filled in later */
372 ip_select_ident(skb, NULL);
373 ((u8 *)&pip[1])[0] = IPOPT_RA;
374 ((u8 *)&pip[1])[1] = 4;
375 ((u8 *)&pip[1])[2] = 0;
376 ((u8 *)&pip[1])[3] = 0;
378 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
379 skb_put(skb, sizeof(*pig));
380 pig = igmpv3_report_hdr(skb);
381 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
382 pig->resv1 = 0;
383 pig->csum = 0;
384 pig->resv2 = 0;
385 pig->ngrec = 0;
386 return skb;
389 static int igmpv3_sendpack(struct sk_buff *skb)
391 struct igmphdr *pig = igmp_hdr(skb);
392 const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
394 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
396 return ip_local_out(skb);
399 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
401 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
404 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
405 int type, struct igmpv3_grec **ppgr)
407 struct net_device *dev = pmc->interface->dev;
408 struct igmpv3_report *pih;
409 struct igmpv3_grec *pgr;
411 if (!skb)
412 skb = igmpv3_newpack(dev, dev->mtu);
413 if (!skb)
414 return NULL;
415 pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
416 pgr->grec_type = type;
417 pgr->grec_auxwords = 0;
418 pgr->grec_nsrcs = 0;
419 pgr->grec_mca = pmc->multiaddr;
420 pih = igmpv3_report_hdr(skb);
421 pih->ngrec = htons(ntohs(pih->ngrec)+1);
422 *ppgr = pgr;
423 return skb;
426 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
427 skb_tailroom(skb)) : 0)
429 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
430 int type, int gdeleted, int sdeleted)
432 struct net_device *dev = pmc->interface->dev;
433 struct igmpv3_report *pih;
434 struct igmpv3_grec *pgr = NULL;
435 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
436 int scount, stotal, first, isquery, truncate;
438 if (pmc->multiaddr == IGMP_ALL_HOSTS)
439 return skb;
441 isquery = type == IGMPV3_MODE_IS_INCLUDE ||
442 type == IGMPV3_MODE_IS_EXCLUDE;
443 truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
444 type == IGMPV3_CHANGE_TO_EXCLUDE;
446 stotal = scount = 0;
448 psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
450 if (!*psf_list)
451 goto empty_source;
453 pih = skb ? igmpv3_report_hdr(skb) : NULL;
455 /* EX and TO_EX get a fresh packet, if needed */
456 if (truncate) {
457 if (pih && pih->ngrec &&
458 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
459 if (skb)
460 igmpv3_sendpack(skb);
461 skb = igmpv3_newpack(dev, dev->mtu);
464 first = 1;
465 psf_prev = NULL;
466 for (psf=*psf_list; psf; psf=psf_next) {
467 __be32 *psrc;
469 psf_next = psf->sf_next;
471 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
472 psf_prev = psf;
473 continue;
476 /* clear marks on query responses */
477 if (isquery)
478 psf->sf_gsresp = 0;
480 if (AVAILABLE(skb) < sizeof(__be32) +
481 first*sizeof(struct igmpv3_grec)) {
482 if (truncate && !first)
483 break; /* truncate these */
484 if (pgr)
485 pgr->grec_nsrcs = htons(scount);
486 if (skb)
487 igmpv3_sendpack(skb);
488 skb = igmpv3_newpack(dev, dev->mtu);
489 first = 1;
490 scount = 0;
492 if (first) {
493 skb = add_grhead(skb, pmc, type, &pgr);
494 first = 0;
496 if (!skb)
497 return NULL;
498 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
499 *psrc = psf->sf_inaddr;
500 scount++; stotal++;
501 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
502 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
503 psf->sf_crcount--;
504 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
505 if (psf_prev)
506 psf_prev->sf_next = psf->sf_next;
507 else
508 *psf_list = psf->sf_next;
509 kfree(psf);
510 continue;
513 psf_prev = psf;
516 empty_source:
517 if (!stotal) {
518 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
519 type == IGMPV3_BLOCK_OLD_SOURCES)
520 return skb;
521 if (pmc->crcount || isquery) {
522 /* make sure we have room for group header */
523 if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
524 igmpv3_sendpack(skb);
525 skb = NULL; /* add_grhead will get a new one */
527 skb = add_grhead(skb, pmc, type, &pgr);
530 if (pgr)
531 pgr->grec_nsrcs = htons(scount);
533 if (isquery)
534 pmc->gsquery = 0; /* clear query state on report */
535 return skb;
538 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
540 struct sk_buff *skb = NULL;
541 int type;
543 if (!pmc) {
544 rcu_read_lock();
545 for_each_pmc_rcu(in_dev, pmc) {
546 if (pmc->multiaddr == IGMP_ALL_HOSTS)
547 continue;
548 spin_lock_bh(&pmc->lock);
549 if (pmc->sfcount[MCAST_EXCLUDE])
550 type = IGMPV3_MODE_IS_EXCLUDE;
551 else
552 type = IGMPV3_MODE_IS_INCLUDE;
553 skb = add_grec(skb, pmc, type, 0, 0);
554 spin_unlock_bh(&pmc->lock);
556 rcu_read_unlock();
557 } else {
558 spin_lock_bh(&pmc->lock);
559 if (pmc->sfcount[MCAST_EXCLUDE])
560 type = IGMPV3_MODE_IS_EXCLUDE;
561 else
562 type = IGMPV3_MODE_IS_INCLUDE;
563 skb = add_grec(skb, pmc, type, 0, 0);
564 spin_unlock_bh(&pmc->lock);
566 if (!skb)
567 return 0;
568 return igmpv3_sendpack(skb);
572 * remove zero-count source records from a source filter list
574 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
576 struct ip_sf_list *psf_prev, *psf_next, *psf;
578 psf_prev = NULL;
579 for (psf=*ppsf; psf; psf = psf_next) {
580 psf_next = psf->sf_next;
581 if (psf->sf_crcount == 0) {
582 if (psf_prev)
583 psf_prev->sf_next = psf->sf_next;
584 else
585 *ppsf = psf->sf_next;
586 kfree(psf);
587 } else
588 psf_prev = psf;
592 static void igmpv3_send_cr(struct in_device *in_dev)
594 struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
595 struct sk_buff *skb = NULL;
596 int type, dtype;
598 rcu_read_lock();
599 spin_lock_bh(&in_dev->mc_tomb_lock);
601 /* deleted MCA's */
602 pmc_prev = NULL;
603 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
604 pmc_next = pmc->next;
605 if (pmc->sfmode == MCAST_INCLUDE) {
606 type = IGMPV3_BLOCK_OLD_SOURCES;
607 dtype = IGMPV3_BLOCK_OLD_SOURCES;
608 skb = add_grec(skb, pmc, type, 1, 0);
609 skb = add_grec(skb, pmc, dtype, 1, 1);
611 if (pmc->crcount) {
612 if (pmc->sfmode == MCAST_EXCLUDE) {
613 type = IGMPV3_CHANGE_TO_INCLUDE;
614 skb = add_grec(skb, pmc, type, 1, 0);
616 pmc->crcount--;
617 if (pmc->crcount == 0) {
618 igmpv3_clear_zeros(&pmc->tomb);
619 igmpv3_clear_zeros(&pmc->sources);
622 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
623 if (pmc_prev)
624 pmc_prev->next = pmc_next;
625 else
626 in_dev->mc_tomb = pmc_next;
627 in_dev_put(pmc->interface);
628 kfree(pmc);
629 } else
630 pmc_prev = pmc;
632 spin_unlock_bh(&in_dev->mc_tomb_lock);
634 /* change recs */
635 for_each_pmc_rcu(in_dev, pmc) {
636 spin_lock_bh(&pmc->lock);
637 if (pmc->sfcount[MCAST_EXCLUDE]) {
638 type = IGMPV3_BLOCK_OLD_SOURCES;
639 dtype = IGMPV3_ALLOW_NEW_SOURCES;
640 } else {
641 type = IGMPV3_ALLOW_NEW_SOURCES;
642 dtype = IGMPV3_BLOCK_OLD_SOURCES;
644 skb = add_grec(skb, pmc, type, 0, 0);
645 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */
647 /* filter mode changes */
648 if (pmc->crcount) {
649 if (pmc->sfmode == MCAST_EXCLUDE)
650 type = IGMPV3_CHANGE_TO_EXCLUDE;
651 else
652 type = IGMPV3_CHANGE_TO_INCLUDE;
653 skb = add_grec(skb, pmc, type, 0, 0);
654 pmc->crcount--;
656 spin_unlock_bh(&pmc->lock);
658 rcu_read_unlock();
660 if (!skb)
661 return;
662 (void) igmpv3_sendpack(skb);
665 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
666 int type)
668 struct sk_buff *skb;
669 struct iphdr *iph;
670 struct igmphdr *ih;
671 struct rtable *rt;
672 struct net_device *dev = in_dev->dev;
673 struct net *net = dev_net(dev);
674 __be32 group = pmc ? pmc->multiaddr : 0;
675 struct flowi4 fl4;
676 __be32 dst;
677 int hlen, tlen;
679 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
680 return igmpv3_send_report(in_dev, pmc);
681 else if (type == IGMP_HOST_LEAVE_MESSAGE)
682 dst = IGMP_ALL_ROUTER;
683 else
684 dst = group;
686 rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
687 0, 0,
688 IPPROTO_IGMP, 0, dev->ifindex);
689 if (IS_ERR(rt))
690 return -1;
692 hlen = LL_RESERVED_SPACE(dev);
693 tlen = dev->needed_tailroom;
694 skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
695 if (skb == NULL) {
696 ip_rt_put(rt);
697 return -1;
699 skb->priority = TC_PRIO_CONTROL;
701 skb_dst_set(skb, &rt->dst);
703 skb_reserve(skb, hlen);
705 skb_reset_network_header(skb);
706 iph = ip_hdr(skb);
707 skb_put(skb, sizeof(struct iphdr) + 4);
709 iph->version = 4;
710 iph->ihl = (sizeof(struct iphdr)+4)>>2;
711 iph->tos = 0xc0;
712 iph->frag_off = htons(IP_DF);
713 iph->ttl = 1;
714 iph->daddr = dst;
715 iph->saddr = fl4.saddr;
716 iph->protocol = IPPROTO_IGMP;
717 ip_select_ident(skb, NULL);
718 ((u8 *)&iph[1])[0] = IPOPT_RA;
719 ((u8 *)&iph[1])[1] = 4;
720 ((u8 *)&iph[1])[2] = 0;
721 ((u8 *)&iph[1])[3] = 0;
723 ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
724 ih->type = type;
725 ih->code = 0;
726 ih->csum = 0;
727 ih->group = group;
728 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
730 return ip_local_out(skb);
733 static void igmp_gq_timer_expire(unsigned long data)
735 struct in_device *in_dev = (struct in_device *)data;
737 in_dev->mr_gq_running = 0;
738 igmpv3_send_report(in_dev, NULL);
739 in_dev_put(in_dev);
742 static void igmp_ifc_timer_expire(unsigned long data)
744 struct in_device *in_dev = (struct in_device *)data;
746 igmpv3_send_cr(in_dev);
747 if (in_dev->mr_ifc_count) {
748 in_dev->mr_ifc_count--;
749 igmp_ifc_start_timer(in_dev,
750 unsolicited_report_interval(in_dev));
752 in_dev_put(in_dev);
755 static void igmp_ifc_event(struct in_device *in_dev)
757 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
758 return;
759 in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
760 IGMP_Unsolicited_Report_Count;
761 igmp_ifc_start_timer(in_dev, 1);
765 static void igmp_timer_expire(unsigned long data)
767 struct ip_mc_list *im=(struct ip_mc_list *)data;
768 struct in_device *in_dev = im->interface;
770 spin_lock(&im->lock);
771 im->tm_running = 0;
773 if (im->unsolicit_count) {
774 im->unsolicit_count--;
775 igmp_start_timer(im, unsolicited_report_interval(in_dev));
777 im->reporter = 1;
778 spin_unlock(&im->lock);
780 if (IGMP_V1_SEEN(in_dev))
781 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
782 else if (IGMP_V2_SEEN(in_dev))
783 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
784 else
785 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
787 ip_ma_put(im);
790 /* mark EXCLUDE-mode sources */
791 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
793 struct ip_sf_list *psf;
794 int i, scount;
796 scount = 0;
797 for (psf=pmc->sources; psf; psf=psf->sf_next) {
798 if (scount == nsrcs)
799 break;
800 for (i=0; i<nsrcs; i++) {
801 /* skip inactive filters */
802 if (psf->sf_count[MCAST_INCLUDE] ||
803 pmc->sfcount[MCAST_EXCLUDE] !=
804 psf->sf_count[MCAST_EXCLUDE])
805 break;
806 if (srcs[i] == psf->sf_inaddr) {
807 scount++;
808 break;
812 pmc->gsquery = 0;
813 if (scount == nsrcs) /* all sources excluded */
814 return 0;
815 return 1;
818 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
820 struct ip_sf_list *psf;
821 int i, scount;
823 if (pmc->sfmode == MCAST_EXCLUDE)
824 return igmp_xmarksources(pmc, nsrcs, srcs);
826 /* mark INCLUDE-mode sources */
827 scount = 0;
828 for (psf=pmc->sources; psf; psf=psf->sf_next) {
829 if (scount == nsrcs)
830 break;
831 for (i=0; i<nsrcs; i++)
832 if (srcs[i] == psf->sf_inaddr) {
833 psf->sf_gsresp = 1;
834 scount++;
835 break;
838 if (!scount) {
839 pmc->gsquery = 0;
840 return 0;
842 pmc->gsquery = 1;
843 return 1;
846 /* return true if packet was dropped */
847 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
849 struct ip_mc_list *im;
851 /* Timers are only set for non-local groups */
853 if (group == IGMP_ALL_HOSTS)
854 return false;
856 rcu_read_lock();
857 for_each_pmc_rcu(in_dev, im) {
858 if (im->multiaddr == group) {
859 igmp_stop_timer(im);
860 break;
863 rcu_read_unlock();
864 return false;
867 /* return true if packet was dropped */
868 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
869 int len)
871 struct igmphdr *ih = igmp_hdr(skb);
872 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
873 struct ip_mc_list *im;
874 __be32 group = ih->group;
875 int max_delay;
876 int mark = 0;
879 if (len == 8) {
880 if (ih->code == 0) {
881 /* Alas, old v1 router presents here. */
883 max_delay = IGMP_Query_Response_Interval;
884 in_dev->mr_v1_seen = jiffies +
885 IGMP_V1_Router_Present_Timeout;
886 group = 0;
887 } else {
888 /* v2 router present */
889 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
890 in_dev->mr_v2_seen = jiffies +
891 IGMP_V2_Router_Present_Timeout;
893 /* cancel the interface change timer */
894 in_dev->mr_ifc_count = 0;
895 if (del_timer(&in_dev->mr_ifc_timer))
896 __in_dev_put(in_dev);
897 /* clear deleted report items */
898 igmpv3_clear_delrec(in_dev);
899 } else if (len < 12) {
900 return true; /* ignore bogus packet; freed by caller */
901 } else if (IGMP_V1_SEEN(in_dev)) {
902 /* This is a v3 query with v1 queriers present */
903 max_delay = IGMP_Query_Response_Interval;
904 group = 0;
905 } else if (IGMP_V2_SEEN(in_dev)) {
906 /* this is a v3 query with v2 queriers present;
907 * Interpretation of the max_delay code is problematic here.
908 * A real v2 host would use ih_code directly, while v3 has a
909 * different encoding. We use the v3 encoding as more likely
910 * to be intended in a v3 query.
912 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
913 if (!max_delay)
914 max_delay = 1; /* can't mod w/ 0 */
915 } else { /* v3 */
916 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
917 return true;
919 ih3 = igmpv3_query_hdr(skb);
920 if (ih3->nsrcs) {
921 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
922 + ntohs(ih3->nsrcs)*sizeof(__be32)))
923 return true;
924 ih3 = igmpv3_query_hdr(skb);
927 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
928 if (!max_delay)
929 max_delay = 1; /* can't mod w/ 0 */
930 in_dev->mr_maxdelay = max_delay;
931 if (ih3->qrv)
932 in_dev->mr_qrv = ih3->qrv;
933 if (!group) { /* general query */
934 if (ih3->nsrcs)
935 return false; /* no sources allowed */
936 igmp_gq_start_timer(in_dev);
937 return false;
939 /* mark sources to include, if group & source-specific */
940 mark = ih3->nsrcs != 0;
944 * - Start the timers in all of our membership records
945 * that the query applies to for the interface on
946 * which the query arrived excl. those that belong
947 * to a "local" group (224.0.0.X)
948 * - For timers already running check if they need to
949 * be reset.
950 * - Use the igmp->igmp_code field as the maximum
951 * delay possible
953 rcu_read_lock();
954 for_each_pmc_rcu(in_dev, im) {
955 int changed;
957 if (group && group != im->multiaddr)
958 continue;
959 if (im->multiaddr == IGMP_ALL_HOSTS)
960 continue;
961 spin_lock_bh(&im->lock);
962 if (im->tm_running)
963 im->gsquery = im->gsquery && mark;
964 else
965 im->gsquery = mark;
966 changed = !im->gsquery ||
967 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
968 spin_unlock_bh(&im->lock);
969 if (changed)
970 igmp_mod_timer(im, max_delay);
972 rcu_read_unlock();
973 return false;
976 /* called in rcu_read_lock() section */
977 int igmp_rcv(struct sk_buff *skb)
979 /* This basically follows the spec line by line -- see RFC1112 */
980 struct igmphdr *ih;
981 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
982 int len = skb->len;
983 bool dropped = true;
985 if (in_dev == NULL)
986 goto drop;
988 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
989 goto drop;
991 switch (skb->ip_summed) {
992 case CHECKSUM_COMPLETE:
993 if (!csum_fold(skb->csum))
994 break;
995 /* fall through */
996 case CHECKSUM_NONE:
997 skb->csum = 0;
998 if (__skb_checksum_complete(skb))
999 goto drop;
1002 ih = igmp_hdr(skb);
1003 switch (ih->type) {
1004 case IGMP_HOST_MEMBERSHIP_QUERY:
1005 dropped = igmp_heard_query(in_dev, skb, len);
1006 break;
1007 case IGMP_HOST_MEMBERSHIP_REPORT:
1008 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1009 /* Is it our report looped back? */
1010 if (rt_is_output_route(skb_rtable(skb)))
1011 break;
1012 /* don't rely on MC router hearing unicast reports */
1013 if (skb->pkt_type == PACKET_MULTICAST ||
1014 skb->pkt_type == PACKET_BROADCAST)
1015 dropped = igmp_heard_report(in_dev, ih->group);
1016 break;
1017 case IGMP_PIM:
1018 #ifdef CONFIG_IP_PIMSM_V1
1019 return pim_rcv_v1(skb);
1020 #endif
1021 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1022 case IGMP_DVMRP:
1023 case IGMP_TRACE:
1024 case IGMP_HOST_LEAVE_MESSAGE:
1025 case IGMP_MTRACE:
1026 case IGMP_MTRACE_RESP:
1027 break;
1028 default:
1029 break;
1032 drop:
1033 if (dropped)
1034 kfree_skb(skb);
1035 else
1036 consume_skb(skb);
1037 return 0;
1040 #endif
1044 * Add a filter to a device
1047 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1049 char buf[MAX_ADDR_LEN];
1050 struct net_device *dev = in_dev->dev;
1052 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1053 We will get multicast token leakage, when IFF_MULTICAST
1054 is changed. This check should be done in ndo_set_rx_mode
1055 routine. Something sort of:
1056 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1057 --ANK
1059 if (arp_mc_map(addr, buf, dev, 0) == 0)
1060 dev_mc_add(dev, buf);
1064 * Remove a filter from a device
1067 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1069 char buf[MAX_ADDR_LEN];
1070 struct net_device *dev = in_dev->dev;
1072 if (arp_mc_map(addr, buf, dev, 0) == 0)
1073 dev_mc_del(dev, buf);
1076 #ifdef CONFIG_IP_MULTICAST
1078 * deleted ip_mc_list manipulation
1080 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1082 struct ip_mc_list *pmc;
1084 /* this is an "ip_mc_list" for convenience; only the fields below
1085 * are actually used. In particular, the refcnt and users are not
1086 * used for management of the delete list. Using the same structure
1087 * for deleted items allows change reports to use common code with
1088 * non-deleted or query-response MCA's.
1090 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1091 if (!pmc)
1092 return;
1093 spin_lock_bh(&im->lock);
1094 pmc->interface = im->interface;
1095 in_dev_hold(in_dev);
1096 pmc->multiaddr = im->multiaddr;
1097 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1098 IGMP_Unsolicited_Report_Count;
1099 pmc->sfmode = im->sfmode;
1100 if (pmc->sfmode == MCAST_INCLUDE) {
1101 struct ip_sf_list *psf;
1103 pmc->tomb = im->tomb;
1104 pmc->sources = im->sources;
1105 im->tomb = im->sources = NULL;
1106 for (psf=pmc->sources; psf; psf=psf->sf_next)
1107 psf->sf_crcount = pmc->crcount;
1109 spin_unlock_bh(&im->lock);
1111 spin_lock_bh(&in_dev->mc_tomb_lock);
1112 pmc->next = in_dev->mc_tomb;
1113 in_dev->mc_tomb = pmc;
1114 spin_unlock_bh(&in_dev->mc_tomb_lock);
1117 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1119 struct ip_mc_list *pmc, *pmc_prev;
1120 struct ip_sf_list *psf, *psf_next;
1122 spin_lock_bh(&in_dev->mc_tomb_lock);
1123 pmc_prev = NULL;
1124 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1125 if (pmc->multiaddr == multiaddr)
1126 break;
1127 pmc_prev = pmc;
1129 if (pmc) {
1130 if (pmc_prev)
1131 pmc_prev->next = pmc->next;
1132 else
1133 in_dev->mc_tomb = pmc->next;
1135 spin_unlock_bh(&in_dev->mc_tomb_lock);
1136 if (pmc) {
1137 for (psf=pmc->tomb; psf; psf=psf_next) {
1138 psf_next = psf->sf_next;
1139 kfree(psf);
1141 in_dev_put(pmc->interface);
1142 kfree(pmc);
1146 static void igmpv3_clear_delrec(struct in_device *in_dev)
1148 struct ip_mc_list *pmc, *nextpmc;
1150 spin_lock_bh(&in_dev->mc_tomb_lock);
1151 pmc = in_dev->mc_tomb;
1152 in_dev->mc_tomb = NULL;
1153 spin_unlock_bh(&in_dev->mc_tomb_lock);
1155 for (; pmc; pmc = nextpmc) {
1156 nextpmc = pmc->next;
1157 ip_mc_clear_src(pmc);
1158 in_dev_put(pmc->interface);
1159 kfree(pmc);
1161 /* clear dead sources, too */
1162 rcu_read_lock();
1163 for_each_pmc_rcu(in_dev, pmc) {
1164 struct ip_sf_list *psf, *psf_next;
1166 spin_lock_bh(&pmc->lock);
1167 psf = pmc->tomb;
1168 pmc->tomb = NULL;
1169 spin_unlock_bh(&pmc->lock);
1170 for (; psf; psf=psf_next) {
1171 psf_next = psf->sf_next;
1172 kfree(psf);
1175 rcu_read_unlock();
1177 #endif
1179 static void igmp_group_dropped(struct ip_mc_list *im)
1181 struct in_device *in_dev = im->interface;
1182 #ifdef CONFIG_IP_MULTICAST
1183 int reporter;
1184 #endif
1186 if (im->loaded) {
1187 im->loaded = 0;
1188 ip_mc_filter_del(in_dev, im->multiaddr);
1191 #ifdef CONFIG_IP_MULTICAST
1192 if (im->multiaddr == IGMP_ALL_HOSTS)
1193 return;
1195 reporter = im->reporter;
1196 igmp_stop_timer(im);
1198 if (!in_dev->dead) {
1199 if (IGMP_V1_SEEN(in_dev))
1200 return;
1201 if (IGMP_V2_SEEN(in_dev)) {
1202 if (reporter)
1203 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1204 return;
1206 /* IGMPv3 */
1207 igmpv3_add_delrec(in_dev, im);
1209 igmp_ifc_event(in_dev);
1211 #endif
1214 static void igmp_group_added(struct ip_mc_list *im)
1216 struct in_device *in_dev = im->interface;
1218 if (im->loaded == 0) {
1219 im->loaded = 1;
1220 ip_mc_filter_add(in_dev, im->multiaddr);
1223 #ifdef CONFIG_IP_MULTICAST
1224 if (im->multiaddr == IGMP_ALL_HOSTS)
1225 return;
1227 if (in_dev->dead)
1228 return;
1229 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1230 spin_lock_bh(&im->lock);
1231 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1232 spin_unlock_bh(&im->lock);
1233 return;
1235 /* else, v3 */
1237 im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1238 IGMP_Unsolicited_Report_Count;
1239 igmp_ifc_event(in_dev);
1240 #endif
1245 * Multicast list managers
1248 static u32 ip_mc_hash(const struct ip_mc_list *im)
1250 return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1253 static void ip_mc_hash_add(struct in_device *in_dev,
1254 struct ip_mc_list *im)
1256 struct ip_mc_list __rcu **mc_hash;
1257 u32 hash;
1259 mc_hash = rtnl_dereference(in_dev->mc_hash);
1260 if (mc_hash) {
1261 hash = ip_mc_hash(im);
1262 im->next_hash = mc_hash[hash];
1263 rcu_assign_pointer(mc_hash[hash], im);
1264 return;
1267 /* do not use a hash table for small number of items */
1268 if (in_dev->mc_count < 4)
1269 return;
1271 mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1272 GFP_KERNEL);
1273 if (!mc_hash)
1274 return;
1276 for_each_pmc_rtnl(in_dev, im) {
1277 hash = ip_mc_hash(im);
1278 im->next_hash = mc_hash[hash];
1279 RCU_INIT_POINTER(mc_hash[hash], im);
1282 rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1285 static void ip_mc_hash_remove(struct in_device *in_dev,
1286 struct ip_mc_list *im)
1288 struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1289 struct ip_mc_list *aux;
1291 if (!mc_hash)
1292 return;
1293 mc_hash += ip_mc_hash(im);
1294 while ((aux = rtnl_dereference(*mc_hash)) != im)
1295 mc_hash = &aux->next_hash;
1296 *mc_hash = im->next_hash;
1301 * A socket has joined a multicast group on device dev.
1304 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1306 struct ip_mc_list *im;
1308 ASSERT_RTNL();
1310 for_each_pmc_rtnl(in_dev, im) {
1311 if (im->multiaddr == addr) {
1312 im->users++;
1313 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1314 goto out;
1318 im = kzalloc(sizeof(*im), GFP_KERNEL);
1319 if (!im)
1320 goto out;
1322 im->users = 1;
1323 im->interface = in_dev;
1324 in_dev_hold(in_dev);
1325 im->multiaddr = addr;
1326 /* initial mode is (EX, empty) */
1327 im->sfmode = MCAST_EXCLUDE;
1328 im->sfcount[MCAST_EXCLUDE] = 1;
1329 atomic_set(&im->refcnt, 1);
1330 spin_lock_init(&im->lock);
1331 #ifdef CONFIG_IP_MULTICAST
1332 setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1333 im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1334 #endif
1336 im->next_rcu = in_dev->mc_list;
1337 in_dev->mc_count++;
1338 rcu_assign_pointer(in_dev->mc_list, im);
1340 ip_mc_hash_add(in_dev, im);
1342 #ifdef CONFIG_IP_MULTICAST
1343 igmpv3_del_delrec(in_dev, im->multiaddr);
1344 #endif
1345 igmp_group_added(im);
1346 if (!in_dev->dead)
1347 ip_rt_multicast_event(in_dev);
1348 out:
1349 return;
1351 EXPORT_SYMBOL(ip_mc_inc_group);
1354 * Resend IGMP JOIN report; used by netdev notifier.
1356 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1358 #ifdef CONFIG_IP_MULTICAST
1359 struct ip_mc_list *im;
1360 int type;
1362 ASSERT_RTNL();
1364 for_each_pmc_rtnl(in_dev, im) {
1365 if (im->multiaddr == IGMP_ALL_HOSTS)
1366 continue;
1368 /* a failover is happening and switches
1369 * must be notified immediately
1371 if (IGMP_V1_SEEN(in_dev))
1372 type = IGMP_HOST_MEMBERSHIP_REPORT;
1373 else if (IGMP_V2_SEEN(in_dev))
1374 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1375 else
1376 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1377 igmp_send_report(in_dev, im, type);
1379 #endif
1383 * A socket has left a multicast group on device dev
1386 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1388 struct ip_mc_list *i;
1389 struct ip_mc_list __rcu **ip;
1391 ASSERT_RTNL();
1393 for (ip = &in_dev->mc_list;
1394 (i = rtnl_dereference(*ip)) != NULL;
1395 ip = &i->next_rcu) {
1396 if (i->multiaddr == addr) {
1397 if (--i->users == 0) {
1398 ip_mc_hash_remove(in_dev, i);
1399 *ip = i->next_rcu;
1400 in_dev->mc_count--;
1401 igmp_group_dropped(i);
1402 ip_mc_clear_src(i);
1404 if (!in_dev->dead)
1405 ip_rt_multicast_event(in_dev);
1407 ip_ma_put(i);
1408 return;
1410 break;
1414 EXPORT_SYMBOL(ip_mc_dec_group);
1416 /* Device changing type */
1418 void ip_mc_unmap(struct in_device *in_dev)
1420 struct ip_mc_list *pmc;
1422 ASSERT_RTNL();
1424 for_each_pmc_rtnl(in_dev, pmc)
1425 igmp_group_dropped(pmc);
1428 void ip_mc_remap(struct in_device *in_dev)
1430 struct ip_mc_list *pmc;
1432 ASSERT_RTNL();
1434 for_each_pmc_rtnl(in_dev, pmc)
1435 igmp_group_added(pmc);
1438 /* Device going down */
1440 void ip_mc_down(struct in_device *in_dev)
1442 struct ip_mc_list *pmc;
1444 ASSERT_RTNL();
1446 for_each_pmc_rtnl(in_dev, pmc)
1447 igmp_group_dropped(pmc);
1449 #ifdef CONFIG_IP_MULTICAST
1450 in_dev->mr_ifc_count = 0;
1451 if (del_timer(&in_dev->mr_ifc_timer))
1452 __in_dev_put(in_dev);
1453 in_dev->mr_gq_running = 0;
1454 if (del_timer(&in_dev->mr_gq_timer))
1455 __in_dev_put(in_dev);
1456 igmpv3_clear_delrec(in_dev);
1457 #endif
1459 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1462 void ip_mc_init_dev(struct in_device *in_dev)
1464 ASSERT_RTNL();
1466 #ifdef CONFIG_IP_MULTICAST
1467 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1468 (unsigned long)in_dev);
1469 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1470 (unsigned long)in_dev);
1471 in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1472 #endif
1474 spin_lock_init(&in_dev->mc_tomb_lock);
1477 /* Device going up */
1479 void ip_mc_up(struct in_device *in_dev)
1481 struct ip_mc_list *pmc;
1483 ASSERT_RTNL();
1485 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1487 for_each_pmc_rtnl(in_dev, pmc)
1488 igmp_group_added(pmc);
1492 * Device is about to be destroyed: clean up.
1495 void ip_mc_destroy_dev(struct in_device *in_dev)
1497 struct ip_mc_list *i;
1499 ASSERT_RTNL();
1501 /* Deactivate timers */
1502 ip_mc_down(in_dev);
1504 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1505 in_dev->mc_list = i->next_rcu;
1506 in_dev->mc_count--;
1508 /* We've dropped the groups in ip_mc_down already */
1509 ip_mc_clear_src(i);
1510 ip_ma_put(i);
1514 /* RTNL is locked */
1515 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1517 struct net_device *dev = NULL;
1518 struct in_device *idev = NULL;
1520 if (imr->imr_ifindex) {
1521 idev = inetdev_by_index(net, imr->imr_ifindex);
1522 return idev;
1524 if (imr->imr_address.s_addr) {
1525 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1526 if (!dev)
1527 return NULL;
1530 if (!dev) {
1531 struct rtable *rt = ip_route_output(net,
1532 imr->imr_multiaddr.s_addr,
1533 0, 0, 0);
1534 if (!IS_ERR(rt)) {
1535 dev = rt->dst.dev;
1536 ip_rt_put(rt);
1539 if (dev) {
1540 imr->imr_ifindex = dev->ifindex;
1541 idev = __in_dev_get_rtnl(dev);
1543 return idev;
1547 * Join a socket to a group
1549 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1550 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1553 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1554 __be32 *psfsrc)
1556 struct ip_sf_list *psf, *psf_prev;
1557 int rv = 0;
1559 psf_prev = NULL;
1560 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1561 if (psf->sf_inaddr == *psfsrc)
1562 break;
1563 psf_prev = psf;
1565 if (!psf || psf->sf_count[sfmode] == 0) {
1566 /* source filter not found, or count wrong => bug */
1567 return -ESRCH;
1569 psf->sf_count[sfmode]--;
1570 if (psf->sf_count[sfmode] == 0) {
1571 ip_rt_multicast_event(pmc->interface);
1573 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1574 #ifdef CONFIG_IP_MULTICAST
1575 struct in_device *in_dev = pmc->interface;
1576 #endif
1578 /* no more filters for this source */
1579 if (psf_prev)
1580 psf_prev->sf_next = psf->sf_next;
1581 else
1582 pmc->sources = psf->sf_next;
1583 #ifdef CONFIG_IP_MULTICAST
1584 if (psf->sf_oldin &&
1585 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1586 psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1587 IGMP_Unsolicited_Report_Count;
1588 psf->sf_next = pmc->tomb;
1589 pmc->tomb = psf;
1590 rv = 1;
1591 } else
1592 #endif
1593 kfree(psf);
1595 return rv;
1598 #ifndef CONFIG_IP_MULTICAST
1599 #define igmp_ifc_event(x) do { } while (0)
1600 #endif
1602 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1603 int sfcount, __be32 *psfsrc, int delta)
1605 struct ip_mc_list *pmc;
1606 int changerec = 0;
1607 int i, err;
1609 if (!in_dev)
1610 return -ENODEV;
1611 rcu_read_lock();
1612 for_each_pmc_rcu(in_dev, pmc) {
1613 if (*pmca == pmc->multiaddr)
1614 break;
1616 if (!pmc) {
1617 /* MCA not found?? bug */
1618 rcu_read_unlock();
1619 return -ESRCH;
1621 spin_lock_bh(&pmc->lock);
1622 rcu_read_unlock();
1623 #ifdef CONFIG_IP_MULTICAST
1624 sf_markstate(pmc);
1625 #endif
1626 if (!delta) {
1627 err = -EINVAL;
1628 if (!pmc->sfcount[sfmode])
1629 goto out_unlock;
1630 pmc->sfcount[sfmode]--;
1632 err = 0;
1633 for (i=0; i<sfcount; i++) {
1634 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1636 changerec |= rv > 0;
1637 if (!err && rv < 0)
1638 err = rv;
1640 if (pmc->sfmode == MCAST_EXCLUDE &&
1641 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1642 pmc->sfcount[MCAST_INCLUDE]) {
1643 #ifdef CONFIG_IP_MULTICAST
1644 struct ip_sf_list *psf;
1645 #endif
1647 /* filter mode change */
1648 pmc->sfmode = MCAST_INCLUDE;
1649 #ifdef CONFIG_IP_MULTICAST
1650 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1651 IGMP_Unsolicited_Report_Count;
1652 in_dev->mr_ifc_count = pmc->crcount;
1653 for (psf=pmc->sources; psf; psf = psf->sf_next)
1654 psf->sf_crcount = 0;
1655 igmp_ifc_event(pmc->interface);
1656 } else if (sf_setstate(pmc) || changerec) {
1657 igmp_ifc_event(pmc->interface);
1658 #endif
1660 out_unlock:
1661 spin_unlock_bh(&pmc->lock);
1662 return err;
1666 * Add multicast single-source filter to the interface list
1668 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1669 __be32 *psfsrc)
1671 struct ip_sf_list *psf, *psf_prev;
1673 psf_prev = NULL;
1674 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1675 if (psf->sf_inaddr == *psfsrc)
1676 break;
1677 psf_prev = psf;
1679 if (!psf) {
1680 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1681 if (!psf)
1682 return -ENOBUFS;
1683 psf->sf_inaddr = *psfsrc;
1684 if (psf_prev) {
1685 psf_prev->sf_next = psf;
1686 } else
1687 pmc->sources = psf;
1689 psf->sf_count[sfmode]++;
1690 if (psf->sf_count[sfmode] == 1) {
1691 ip_rt_multicast_event(pmc->interface);
1693 return 0;
1696 #ifdef CONFIG_IP_MULTICAST
1697 static void sf_markstate(struct ip_mc_list *pmc)
1699 struct ip_sf_list *psf;
1700 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1702 for (psf=pmc->sources; psf; psf=psf->sf_next)
1703 if (pmc->sfcount[MCAST_EXCLUDE]) {
1704 psf->sf_oldin = mca_xcount ==
1705 psf->sf_count[MCAST_EXCLUDE] &&
1706 !psf->sf_count[MCAST_INCLUDE];
1707 } else
1708 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1711 static int sf_setstate(struct ip_mc_list *pmc)
1713 struct ip_sf_list *psf, *dpsf;
1714 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1715 int qrv = pmc->interface->mr_qrv;
1716 int new_in, rv;
1718 rv = 0;
1719 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1720 if (pmc->sfcount[MCAST_EXCLUDE]) {
1721 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1722 !psf->sf_count[MCAST_INCLUDE];
1723 } else
1724 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1725 if (new_in) {
1726 if (!psf->sf_oldin) {
1727 struct ip_sf_list *prev = NULL;
1729 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1730 if (dpsf->sf_inaddr == psf->sf_inaddr)
1731 break;
1732 prev = dpsf;
1734 if (dpsf) {
1735 if (prev)
1736 prev->sf_next = dpsf->sf_next;
1737 else
1738 pmc->tomb = dpsf->sf_next;
1739 kfree(dpsf);
1741 psf->sf_crcount = qrv;
1742 rv++;
1744 } else if (psf->sf_oldin) {
1746 psf->sf_crcount = 0;
1748 * add or update "delete" records if an active filter
1749 * is now inactive
1751 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1752 if (dpsf->sf_inaddr == psf->sf_inaddr)
1753 break;
1754 if (!dpsf) {
1755 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1756 if (!dpsf)
1757 continue;
1758 *dpsf = *psf;
1759 /* pmc->lock held by callers */
1760 dpsf->sf_next = pmc->tomb;
1761 pmc->tomb = dpsf;
1763 dpsf->sf_crcount = qrv;
1764 rv++;
1767 return rv;
1769 #endif
1772 * Add multicast source filter list to the interface list
1774 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1775 int sfcount, __be32 *psfsrc, int delta)
1777 struct ip_mc_list *pmc;
1778 int isexclude;
1779 int i, err;
1781 if (!in_dev)
1782 return -ENODEV;
1783 rcu_read_lock();
1784 for_each_pmc_rcu(in_dev, pmc) {
1785 if (*pmca == pmc->multiaddr)
1786 break;
1788 if (!pmc) {
1789 /* MCA not found?? bug */
1790 rcu_read_unlock();
1791 return -ESRCH;
1793 spin_lock_bh(&pmc->lock);
1794 rcu_read_unlock();
1796 #ifdef CONFIG_IP_MULTICAST
1797 sf_markstate(pmc);
1798 #endif
1799 isexclude = pmc->sfmode == MCAST_EXCLUDE;
1800 if (!delta)
1801 pmc->sfcount[sfmode]++;
1802 err = 0;
1803 for (i=0; i<sfcount; i++) {
1804 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1805 if (err)
1806 break;
1808 if (err) {
1809 int j;
1811 if (!delta)
1812 pmc->sfcount[sfmode]--;
1813 for (j=0; j<i; j++)
1814 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1815 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1816 #ifdef CONFIG_IP_MULTICAST
1817 struct ip_sf_list *psf;
1818 in_dev = pmc->interface;
1819 #endif
1821 /* filter mode change */
1822 if (pmc->sfcount[MCAST_EXCLUDE])
1823 pmc->sfmode = MCAST_EXCLUDE;
1824 else if (pmc->sfcount[MCAST_INCLUDE])
1825 pmc->sfmode = MCAST_INCLUDE;
1826 #ifdef CONFIG_IP_MULTICAST
1827 /* else no filters; keep old mode for reports */
1829 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1830 IGMP_Unsolicited_Report_Count;
1831 in_dev->mr_ifc_count = pmc->crcount;
1832 for (psf=pmc->sources; psf; psf = psf->sf_next)
1833 psf->sf_crcount = 0;
1834 igmp_ifc_event(in_dev);
1835 } else if (sf_setstate(pmc)) {
1836 igmp_ifc_event(in_dev);
1837 #endif
1839 spin_unlock_bh(&pmc->lock);
1840 return err;
1843 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1845 struct ip_sf_list *psf, *nextpsf;
1847 for (psf=pmc->tomb; psf; psf=nextpsf) {
1848 nextpsf = psf->sf_next;
1849 kfree(psf);
1851 pmc->tomb = NULL;
1852 for (psf=pmc->sources; psf; psf=nextpsf) {
1853 nextpsf = psf->sf_next;
1854 kfree(psf);
1856 pmc->sources = NULL;
1857 pmc->sfmode = MCAST_EXCLUDE;
1858 pmc->sfcount[MCAST_INCLUDE] = 0;
1859 pmc->sfcount[MCAST_EXCLUDE] = 1;
1864 * Join a multicast group
1866 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1868 int err;
1869 __be32 addr = imr->imr_multiaddr.s_addr;
1870 struct ip_mc_socklist *iml = NULL, *i;
1871 struct in_device *in_dev;
1872 struct inet_sock *inet = inet_sk(sk);
1873 struct net *net = sock_net(sk);
1874 int ifindex;
1875 int count = 0;
1877 if (!ipv4_is_multicast(addr))
1878 return -EINVAL;
1880 rtnl_lock();
1882 in_dev = ip_mc_find_dev(net, imr);
1884 if (!in_dev) {
1885 iml = NULL;
1886 err = -ENODEV;
1887 goto done;
1890 err = -EADDRINUSE;
1891 ifindex = imr->imr_ifindex;
1892 for_each_pmc_rtnl(inet, i) {
1893 if (i->multi.imr_multiaddr.s_addr == addr &&
1894 i->multi.imr_ifindex == ifindex)
1895 goto done;
1896 count++;
1898 err = -ENOBUFS;
1899 if (count >= sysctl_igmp_max_memberships)
1900 goto done;
1901 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1902 if (iml == NULL)
1903 goto done;
1905 memcpy(&iml->multi, imr, sizeof(*imr));
1906 iml->next_rcu = inet->mc_list;
1907 iml->sflist = NULL;
1908 iml->sfmode = MCAST_EXCLUDE;
1909 rcu_assign_pointer(inet->mc_list, iml);
1910 ip_mc_inc_group(in_dev, addr);
1911 err = 0;
1912 done:
1913 rtnl_unlock();
1914 return err;
1916 EXPORT_SYMBOL(ip_mc_join_group);
1918 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1919 struct in_device *in_dev)
1921 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1922 int err;
1924 if (psf == NULL) {
1925 /* any-source empty exclude case */
1926 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1927 iml->sfmode, 0, NULL, 0);
1929 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1930 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1931 RCU_INIT_POINTER(iml->sflist, NULL);
1932 /* decrease mem now to avoid the memleak warning */
1933 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1934 kfree_rcu(psf, rcu);
1935 return err;
1939 * Ask a socket to leave a group.
1942 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1944 struct inet_sock *inet = inet_sk(sk);
1945 struct ip_mc_socklist *iml;
1946 struct ip_mc_socklist __rcu **imlp;
1947 struct in_device *in_dev;
1948 struct net *net = sock_net(sk);
1949 __be32 group = imr->imr_multiaddr.s_addr;
1950 u32 ifindex;
1951 int ret = -EADDRNOTAVAIL;
1953 rtnl_lock();
1954 in_dev = ip_mc_find_dev(net, imr);
1955 if (!in_dev) {
1956 ret = -ENODEV;
1957 goto out;
1959 ifindex = imr->imr_ifindex;
1960 for (imlp = &inet->mc_list;
1961 (iml = rtnl_dereference(*imlp)) != NULL;
1962 imlp = &iml->next_rcu) {
1963 if (iml->multi.imr_multiaddr.s_addr != group)
1964 continue;
1965 if (ifindex) {
1966 if (iml->multi.imr_ifindex != ifindex)
1967 continue;
1968 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1969 iml->multi.imr_address.s_addr)
1970 continue;
1972 (void) ip_mc_leave_src(sk, iml, in_dev);
1974 *imlp = iml->next_rcu;
1976 ip_mc_dec_group(in_dev, group);
1977 rtnl_unlock();
1978 /* decrease mem now to avoid the memleak warning */
1979 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1980 kfree_rcu(iml, rcu);
1981 return 0;
1983 out:
1984 rtnl_unlock();
1985 return ret;
1987 EXPORT_SYMBOL(ip_mc_leave_group);
1989 int ip_mc_source(int add, int omode, struct sock *sk, struct
1990 ip_mreq_source *mreqs, int ifindex)
1992 int err;
1993 struct ip_mreqn imr;
1994 __be32 addr = mreqs->imr_multiaddr;
1995 struct ip_mc_socklist *pmc;
1996 struct in_device *in_dev = NULL;
1997 struct inet_sock *inet = inet_sk(sk);
1998 struct ip_sf_socklist *psl;
1999 struct net *net = sock_net(sk);
2000 int leavegroup = 0;
2001 int i, j, rv;
2003 if (!ipv4_is_multicast(addr))
2004 return -EINVAL;
2006 rtnl_lock();
2008 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2009 imr.imr_address.s_addr = mreqs->imr_interface;
2010 imr.imr_ifindex = ifindex;
2011 in_dev = ip_mc_find_dev(net, &imr);
2013 if (!in_dev) {
2014 err = -ENODEV;
2015 goto done;
2017 err = -EADDRNOTAVAIL;
2019 for_each_pmc_rtnl(inet, pmc) {
2020 if ((pmc->multi.imr_multiaddr.s_addr ==
2021 imr.imr_multiaddr.s_addr) &&
2022 (pmc->multi.imr_ifindex == imr.imr_ifindex))
2023 break;
2025 if (!pmc) { /* must have a prior join */
2026 err = -EINVAL;
2027 goto done;
2029 /* if a source filter was set, must be the same mode as before */
2030 if (pmc->sflist) {
2031 if (pmc->sfmode != omode) {
2032 err = -EINVAL;
2033 goto done;
2035 } else if (pmc->sfmode != omode) {
2036 /* allow mode switches for empty-set filters */
2037 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2038 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2039 NULL, 0);
2040 pmc->sfmode = omode;
2043 psl = rtnl_dereference(pmc->sflist);
2044 if (!add) {
2045 if (!psl)
2046 goto done; /* err = -EADDRNOTAVAIL */
2047 rv = !0;
2048 for (i=0; i<psl->sl_count; i++) {
2049 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2050 sizeof(__be32));
2051 if (rv == 0)
2052 break;
2054 if (rv) /* source not found */
2055 goto done; /* err = -EADDRNOTAVAIL */
2057 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2058 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2059 leavegroup = 1;
2060 goto done;
2063 /* update the interface filter */
2064 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2065 &mreqs->imr_sourceaddr, 1);
2067 for (j=i+1; j<psl->sl_count; j++)
2068 psl->sl_addr[j-1] = psl->sl_addr[j];
2069 psl->sl_count--;
2070 err = 0;
2071 goto done;
2073 /* else, add a new source to the filter */
2075 if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2076 err = -ENOBUFS;
2077 goto done;
2079 if (!psl || psl->sl_count == psl->sl_max) {
2080 struct ip_sf_socklist *newpsl;
2081 int count = IP_SFBLOCK;
2083 if (psl)
2084 count += psl->sl_max;
2085 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2086 if (!newpsl) {
2087 err = -ENOBUFS;
2088 goto done;
2090 newpsl->sl_max = count;
2091 newpsl->sl_count = count - IP_SFBLOCK;
2092 if (psl) {
2093 for (i=0; i<psl->sl_count; i++)
2094 newpsl->sl_addr[i] = psl->sl_addr[i];
2095 /* decrease mem now to avoid the memleak warning */
2096 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2097 kfree_rcu(psl, rcu);
2099 rcu_assign_pointer(pmc->sflist, newpsl);
2100 psl = newpsl;
2102 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2103 for (i=0; i<psl->sl_count; i++) {
2104 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2105 sizeof(__be32));
2106 if (rv == 0)
2107 break;
2109 if (rv == 0) /* address already there is an error */
2110 goto done;
2111 for (j=psl->sl_count-1; j>=i; j--)
2112 psl->sl_addr[j+1] = psl->sl_addr[j];
2113 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2114 psl->sl_count++;
2115 err = 0;
2116 /* update the interface list */
2117 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2118 &mreqs->imr_sourceaddr, 1);
2119 done:
2120 rtnl_unlock();
2121 if (leavegroup)
2122 return ip_mc_leave_group(sk, &imr);
2123 return err;
2126 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2128 int err = 0;
2129 struct ip_mreqn imr;
2130 __be32 addr = msf->imsf_multiaddr;
2131 struct ip_mc_socklist *pmc;
2132 struct in_device *in_dev;
2133 struct inet_sock *inet = inet_sk(sk);
2134 struct ip_sf_socklist *newpsl, *psl;
2135 struct net *net = sock_net(sk);
2136 int leavegroup = 0;
2138 if (!ipv4_is_multicast(addr))
2139 return -EINVAL;
2140 if (msf->imsf_fmode != MCAST_INCLUDE &&
2141 msf->imsf_fmode != MCAST_EXCLUDE)
2142 return -EINVAL;
2144 rtnl_lock();
2146 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2147 imr.imr_address.s_addr = msf->imsf_interface;
2148 imr.imr_ifindex = ifindex;
2149 in_dev = ip_mc_find_dev(net, &imr);
2151 if (!in_dev) {
2152 err = -ENODEV;
2153 goto done;
2156 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2157 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2158 leavegroup = 1;
2159 goto done;
2162 for_each_pmc_rtnl(inet, pmc) {
2163 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2164 pmc->multi.imr_ifindex == imr.imr_ifindex)
2165 break;
2167 if (!pmc) { /* must have a prior join */
2168 err = -EINVAL;
2169 goto done;
2171 if (msf->imsf_numsrc) {
2172 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2173 GFP_KERNEL);
2174 if (!newpsl) {
2175 err = -ENOBUFS;
2176 goto done;
2178 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2179 memcpy(newpsl->sl_addr, msf->imsf_slist,
2180 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2181 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2182 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2183 if (err) {
2184 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2185 goto done;
2187 } else {
2188 newpsl = NULL;
2189 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2190 msf->imsf_fmode, 0, NULL, 0);
2192 psl = rtnl_dereference(pmc->sflist);
2193 if (psl) {
2194 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2195 psl->sl_count, psl->sl_addr, 0);
2196 /* decrease mem now to avoid the memleak warning */
2197 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2198 kfree_rcu(psl, rcu);
2199 } else
2200 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2201 0, NULL, 0);
2202 rcu_assign_pointer(pmc->sflist, newpsl);
2203 pmc->sfmode = msf->imsf_fmode;
2204 err = 0;
2205 done:
2206 rtnl_unlock();
2207 if (leavegroup)
2208 err = ip_mc_leave_group(sk, &imr);
2209 return err;
2212 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2213 struct ip_msfilter __user *optval, int __user *optlen)
2215 int err, len, count, copycount;
2216 struct ip_mreqn imr;
2217 __be32 addr = msf->imsf_multiaddr;
2218 struct ip_mc_socklist *pmc;
2219 struct in_device *in_dev;
2220 struct inet_sock *inet = inet_sk(sk);
2221 struct ip_sf_socklist *psl;
2222 struct net *net = sock_net(sk);
2224 if (!ipv4_is_multicast(addr))
2225 return -EINVAL;
2227 rtnl_lock();
2229 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2230 imr.imr_address.s_addr = msf->imsf_interface;
2231 imr.imr_ifindex = 0;
2232 in_dev = ip_mc_find_dev(net, &imr);
2234 if (!in_dev) {
2235 err = -ENODEV;
2236 goto done;
2238 err = -EADDRNOTAVAIL;
2240 for_each_pmc_rtnl(inet, pmc) {
2241 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2242 pmc->multi.imr_ifindex == imr.imr_ifindex)
2243 break;
2245 if (!pmc) /* must have a prior join */
2246 goto done;
2247 msf->imsf_fmode = pmc->sfmode;
2248 psl = rtnl_dereference(pmc->sflist);
2249 rtnl_unlock();
2250 if (!psl) {
2251 len = 0;
2252 count = 0;
2253 } else {
2254 count = psl->sl_count;
2256 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2257 len = copycount * sizeof(psl->sl_addr[0]);
2258 msf->imsf_numsrc = count;
2259 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2260 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2261 return -EFAULT;
2263 if (len &&
2264 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2265 return -EFAULT;
2266 return 0;
2267 done:
2268 rtnl_unlock();
2269 return err;
2272 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2273 struct group_filter __user *optval, int __user *optlen)
2275 int err, i, count, copycount;
2276 struct sockaddr_in *psin;
2277 __be32 addr;
2278 struct ip_mc_socklist *pmc;
2279 struct inet_sock *inet = inet_sk(sk);
2280 struct ip_sf_socklist *psl;
2282 psin = (struct sockaddr_in *)&gsf->gf_group;
2283 if (psin->sin_family != AF_INET)
2284 return -EINVAL;
2285 addr = psin->sin_addr.s_addr;
2286 if (!ipv4_is_multicast(addr))
2287 return -EINVAL;
2289 rtnl_lock();
2291 err = -EADDRNOTAVAIL;
2293 for_each_pmc_rtnl(inet, pmc) {
2294 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2295 pmc->multi.imr_ifindex == gsf->gf_interface)
2296 break;
2298 if (!pmc) /* must have a prior join */
2299 goto done;
2300 gsf->gf_fmode = pmc->sfmode;
2301 psl = rtnl_dereference(pmc->sflist);
2302 rtnl_unlock();
2303 count = psl ? psl->sl_count : 0;
2304 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2305 gsf->gf_numsrc = count;
2306 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2307 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2308 return -EFAULT;
2310 for (i=0; i<copycount; i++) {
2311 struct sockaddr_storage ss;
2313 psin = (struct sockaddr_in *)&ss;
2314 memset(&ss, 0, sizeof(ss));
2315 psin->sin_family = AF_INET;
2316 psin->sin_addr.s_addr = psl->sl_addr[i];
2317 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2318 return -EFAULT;
2320 return 0;
2321 done:
2322 rtnl_unlock();
2323 return err;
2327 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2329 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2331 struct inet_sock *inet = inet_sk(sk);
2332 struct ip_mc_socklist *pmc;
2333 struct ip_sf_socklist *psl;
2334 int i;
2335 int ret;
2337 ret = 1;
2338 if (!ipv4_is_multicast(loc_addr))
2339 goto out;
2341 rcu_read_lock();
2342 for_each_pmc_rcu(inet, pmc) {
2343 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2344 pmc->multi.imr_ifindex == dif)
2345 break;
2347 ret = inet->mc_all;
2348 if (!pmc)
2349 goto unlock;
2350 psl = rcu_dereference(pmc->sflist);
2351 ret = (pmc->sfmode == MCAST_EXCLUDE);
2352 if (!psl)
2353 goto unlock;
2355 for (i=0; i<psl->sl_count; i++) {
2356 if (psl->sl_addr[i] == rmt_addr)
2357 break;
2359 ret = 0;
2360 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2361 goto unlock;
2362 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2363 goto unlock;
2364 ret = 1;
2365 unlock:
2366 rcu_read_unlock();
2367 out:
2368 return ret;
2372 * A socket is closing.
2375 void ip_mc_drop_socket(struct sock *sk)
2377 struct inet_sock *inet = inet_sk(sk);
2378 struct ip_mc_socklist *iml;
2379 struct net *net = sock_net(sk);
2381 if (inet->mc_list == NULL)
2382 return;
2384 rtnl_lock();
2385 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2386 struct in_device *in_dev;
2388 inet->mc_list = iml->next_rcu;
2389 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2390 (void) ip_mc_leave_src(sk, iml, in_dev);
2391 if (in_dev != NULL)
2392 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2393 /* decrease mem now to avoid the memleak warning */
2394 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2395 kfree_rcu(iml, rcu);
2397 rtnl_unlock();
2400 /* called with rcu_read_lock() */
2401 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2403 struct ip_mc_list *im;
2404 struct ip_mc_list __rcu **mc_hash;
2405 struct ip_sf_list *psf;
2406 int rv = 0;
2408 mc_hash = rcu_dereference(in_dev->mc_hash);
2409 if (mc_hash) {
2410 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2412 for (im = rcu_dereference(mc_hash[hash]);
2413 im != NULL;
2414 im = rcu_dereference(im->next_hash)) {
2415 if (im->multiaddr == mc_addr)
2416 break;
2418 } else {
2419 for_each_pmc_rcu(in_dev, im) {
2420 if (im->multiaddr == mc_addr)
2421 break;
2424 if (im && proto == IPPROTO_IGMP) {
2425 rv = 1;
2426 } else if (im) {
2427 if (src_addr) {
2428 for (psf=im->sources; psf; psf=psf->sf_next) {
2429 if (psf->sf_inaddr == src_addr)
2430 break;
2432 if (psf)
2433 rv = psf->sf_count[MCAST_INCLUDE] ||
2434 psf->sf_count[MCAST_EXCLUDE] !=
2435 im->sfcount[MCAST_EXCLUDE];
2436 else
2437 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2438 } else
2439 rv = 1; /* unspecified source; tentatively allow */
2441 return rv;
2444 #if defined(CONFIG_PROC_FS)
2445 struct igmp_mc_iter_state {
2446 struct seq_net_private p;
2447 struct net_device *dev;
2448 struct in_device *in_dev;
2451 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2453 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2455 struct net *net = seq_file_net(seq);
2456 struct ip_mc_list *im = NULL;
2457 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2459 state->in_dev = NULL;
2460 for_each_netdev_rcu(net, state->dev) {
2461 struct in_device *in_dev;
2463 in_dev = __in_dev_get_rcu(state->dev);
2464 if (!in_dev)
2465 continue;
2466 im = rcu_dereference(in_dev->mc_list);
2467 if (im) {
2468 state->in_dev = in_dev;
2469 break;
2472 return im;
2475 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2477 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2479 im = rcu_dereference(im->next_rcu);
2480 while (!im) {
2481 state->dev = next_net_device_rcu(state->dev);
2482 if (!state->dev) {
2483 state->in_dev = NULL;
2484 break;
2486 state->in_dev = __in_dev_get_rcu(state->dev);
2487 if (!state->in_dev)
2488 continue;
2489 im = rcu_dereference(state->in_dev->mc_list);
2491 return im;
2494 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2496 struct ip_mc_list *im = igmp_mc_get_first(seq);
2497 if (im)
2498 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2499 --pos;
2500 return pos ? NULL : im;
2503 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2504 __acquires(rcu)
2506 rcu_read_lock();
2507 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2510 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2512 struct ip_mc_list *im;
2513 if (v == SEQ_START_TOKEN)
2514 im = igmp_mc_get_first(seq);
2515 else
2516 im = igmp_mc_get_next(seq, v);
2517 ++*pos;
2518 return im;
2521 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2522 __releases(rcu)
2524 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2526 state->in_dev = NULL;
2527 state->dev = NULL;
2528 rcu_read_unlock();
2531 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2533 if (v == SEQ_START_TOKEN)
2534 seq_puts(seq,
2535 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2536 else {
2537 struct ip_mc_list *im = (struct ip_mc_list *)v;
2538 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2539 char *querier;
2540 long delta;
2542 #ifdef CONFIG_IP_MULTICAST
2543 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2544 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2545 "V3";
2546 #else
2547 querier = "NONE";
2548 #endif
2550 if (rcu_dereference(state->in_dev->mc_list) == im) {
2551 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2552 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2555 delta = im->timer.expires - jiffies;
2556 seq_printf(seq,
2557 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2558 im->multiaddr, im->users,
2559 im->tm_running,
2560 im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2561 im->reporter);
2563 return 0;
2566 static const struct seq_operations igmp_mc_seq_ops = {
2567 .start = igmp_mc_seq_start,
2568 .next = igmp_mc_seq_next,
2569 .stop = igmp_mc_seq_stop,
2570 .show = igmp_mc_seq_show,
2573 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2575 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2576 sizeof(struct igmp_mc_iter_state));
2579 static const struct file_operations igmp_mc_seq_fops = {
2580 .owner = THIS_MODULE,
2581 .open = igmp_mc_seq_open,
2582 .read = seq_read,
2583 .llseek = seq_lseek,
2584 .release = seq_release_net,
2587 struct igmp_mcf_iter_state {
2588 struct seq_net_private p;
2589 struct net_device *dev;
2590 struct in_device *idev;
2591 struct ip_mc_list *im;
2594 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2596 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2598 struct net *net = seq_file_net(seq);
2599 struct ip_sf_list *psf = NULL;
2600 struct ip_mc_list *im = NULL;
2601 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2603 state->idev = NULL;
2604 state->im = NULL;
2605 for_each_netdev_rcu(net, state->dev) {
2606 struct in_device *idev;
2607 idev = __in_dev_get_rcu(state->dev);
2608 if (unlikely(idev == NULL))
2609 continue;
2610 im = rcu_dereference(idev->mc_list);
2611 if (likely(im != NULL)) {
2612 spin_lock_bh(&im->lock);
2613 psf = im->sources;
2614 if (likely(psf != NULL)) {
2615 state->im = im;
2616 state->idev = idev;
2617 break;
2619 spin_unlock_bh(&im->lock);
2622 return psf;
2625 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2627 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2629 psf = psf->sf_next;
2630 while (!psf) {
2631 spin_unlock_bh(&state->im->lock);
2632 state->im = state->im->next;
2633 while (!state->im) {
2634 state->dev = next_net_device_rcu(state->dev);
2635 if (!state->dev) {
2636 state->idev = NULL;
2637 goto out;
2639 state->idev = __in_dev_get_rcu(state->dev);
2640 if (!state->idev)
2641 continue;
2642 state->im = rcu_dereference(state->idev->mc_list);
2644 if (!state->im)
2645 break;
2646 spin_lock_bh(&state->im->lock);
2647 psf = state->im->sources;
2649 out:
2650 return psf;
2653 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2655 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2656 if (psf)
2657 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2658 --pos;
2659 return pos ? NULL : psf;
2662 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2663 __acquires(rcu)
2665 rcu_read_lock();
2666 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2669 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2671 struct ip_sf_list *psf;
2672 if (v == SEQ_START_TOKEN)
2673 psf = igmp_mcf_get_first(seq);
2674 else
2675 psf = igmp_mcf_get_next(seq, v);
2676 ++*pos;
2677 return psf;
2680 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2681 __releases(rcu)
2683 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2684 if (likely(state->im != NULL)) {
2685 spin_unlock_bh(&state->im->lock);
2686 state->im = NULL;
2688 state->idev = NULL;
2689 state->dev = NULL;
2690 rcu_read_unlock();
2693 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2695 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2696 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2698 if (v == SEQ_START_TOKEN) {
2699 seq_printf(seq,
2700 "%3s %6s "
2701 "%10s %10s %6s %6s\n", "Idx",
2702 "Device", "MCA",
2703 "SRC", "INC", "EXC");
2704 } else {
2705 seq_printf(seq,
2706 "%3d %6.6s 0x%08x "
2707 "0x%08x %6lu %6lu\n",
2708 state->dev->ifindex, state->dev->name,
2709 ntohl(state->im->multiaddr),
2710 ntohl(psf->sf_inaddr),
2711 psf->sf_count[MCAST_INCLUDE],
2712 psf->sf_count[MCAST_EXCLUDE]);
2714 return 0;
2717 static const struct seq_operations igmp_mcf_seq_ops = {
2718 .start = igmp_mcf_seq_start,
2719 .next = igmp_mcf_seq_next,
2720 .stop = igmp_mcf_seq_stop,
2721 .show = igmp_mcf_seq_show,
2724 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2726 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2727 sizeof(struct igmp_mcf_iter_state));
2730 static const struct file_operations igmp_mcf_seq_fops = {
2731 .owner = THIS_MODULE,
2732 .open = igmp_mcf_seq_open,
2733 .read = seq_read,
2734 .llseek = seq_lseek,
2735 .release = seq_release_net,
2738 static int __net_init igmp_net_init(struct net *net)
2740 struct proc_dir_entry *pde;
2742 pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2743 if (!pde)
2744 goto out_igmp;
2745 pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2746 &igmp_mcf_seq_fops);
2747 if (!pde)
2748 goto out_mcfilter;
2749 return 0;
2751 out_mcfilter:
2752 remove_proc_entry("igmp", net->proc_net);
2753 out_igmp:
2754 return -ENOMEM;
2757 static void __net_exit igmp_net_exit(struct net *net)
2759 remove_proc_entry("mcfilter", net->proc_net);
2760 remove_proc_entry("igmp", net->proc_net);
2763 static struct pernet_operations igmp_net_ops = {
2764 .init = igmp_net_init,
2765 .exit = igmp_net_exit,
2768 static int igmp_netdev_event(struct notifier_block *this,
2769 unsigned long event, void *ptr)
2771 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2772 struct in_device *in_dev;
2774 switch (event) {
2775 case NETDEV_RESEND_IGMP:
2776 in_dev = __in_dev_get_rtnl(dev);
2777 if (in_dev)
2778 ip_mc_rejoin_groups(in_dev);
2779 break;
2780 default:
2781 break;
2783 return NOTIFY_DONE;
2786 static struct notifier_block igmp_notifier = {
2787 .notifier_call = igmp_netdev_event,
2790 int __init igmp_mc_proc_init(void)
2792 int err;
2794 err = register_pernet_subsys(&igmp_net_ops);
2795 if (err)
2796 return err;
2797 err = register_netdevice_notifier(&igmp_notifier);
2798 if (err)
2799 goto reg_notif_fail;
2800 return 0;
2802 reg_notif_fail:
2803 unregister_pernet_subsys(&igmp_net_ops);
2804 return err;
2806 #endif