vmxnet3: Fix inconsistent LRO state after initialization
[linux-2.6/linux-mips.git] / net / ipv4 / igmp.c
blob1fd3d9ce8398053588a45086ba813adfc7744de4
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 <asm/system.h>
77 #include <linux/types.h>
78 #include <linux/kernel.h>
79 #include <linux/jiffies.h>
80 #include <linux/string.h>
81 #include <linux/socket.h>
82 #include <linux/sockios.h>
83 #include <linux/in.h>
84 #include <linux/inet.h>
85 #include <linux/netdevice.h>
86 #include <linux/skbuff.h>
87 #include <linux/inetdevice.h>
88 #include <linux/igmp.h>
89 #include <linux/if_arp.h>
90 #include <linux/rtnetlink.h>
91 #include <linux/times.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_Unsolicited_Report_Interval (10*HZ)
118 #define IGMP_Query_Response_Interval (10*HZ)
119 #define IGMP_Unsolicited_Report_Count 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 void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
144 static void igmpv3_clear_delrec(struct in_device *in_dev);
145 static int sf_setstate(struct ip_mc_list *pmc);
146 static void sf_markstate(struct ip_mc_list *pmc);
147 #endif
148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
150 int sfcount, __be32 *psfsrc, int delta);
153 static void ip_mc_list_reclaim(struct rcu_head *head)
155 kfree(container_of(head, struct ip_mc_list, rcu));
158 static void ip_ma_put(struct ip_mc_list *im)
160 if (atomic_dec_and_test(&im->refcnt)) {
161 in_dev_put(im->interface);
162 call_rcu(&im->rcu, ip_mc_list_reclaim);
166 #define for_each_pmc_rcu(in_dev, pmc) \
167 for (pmc = rcu_dereference(in_dev->mc_list); \
168 pmc != NULL; \
169 pmc = rcu_dereference(pmc->next_rcu))
171 #define for_each_pmc_rtnl(in_dev, pmc) \
172 for (pmc = rtnl_dereference(in_dev->mc_list); \
173 pmc != NULL; \
174 pmc = rtnl_dereference(pmc->next_rcu))
176 #ifdef CONFIG_IP_MULTICAST
179 * Timer management
182 static void igmp_stop_timer(struct ip_mc_list *im)
184 spin_lock_bh(&im->lock);
185 if (del_timer(&im->timer))
186 atomic_dec(&im->refcnt);
187 im->tm_running = 0;
188 im->reporter = 0;
189 im->unsolicit_count = 0;
190 spin_unlock_bh(&im->lock);
193 /* It must be called with locked im->lock */
194 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
196 int tv = net_random() % max_delay;
198 im->tm_running = 1;
199 if (!mod_timer(&im->timer, jiffies+tv+2))
200 atomic_inc(&im->refcnt);
203 static void igmp_gq_start_timer(struct in_device *in_dev)
205 int tv = net_random() % in_dev->mr_maxdelay;
207 in_dev->mr_gq_running = 1;
208 if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
209 in_dev_hold(in_dev);
212 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
214 int tv = net_random() % delay;
216 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
217 in_dev_hold(in_dev);
220 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
222 spin_lock_bh(&im->lock);
223 im->unsolicit_count = 0;
224 if (del_timer(&im->timer)) {
225 if ((long)(im->timer.expires-jiffies) < max_delay) {
226 add_timer(&im->timer);
227 im->tm_running = 1;
228 spin_unlock_bh(&im->lock);
229 return;
231 atomic_dec(&im->refcnt);
233 igmp_start_timer(im, max_delay);
234 spin_unlock_bh(&im->lock);
239 * Send an IGMP report.
242 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
245 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
246 int gdeleted, int sdeleted)
248 switch (type) {
249 case IGMPV3_MODE_IS_INCLUDE:
250 case IGMPV3_MODE_IS_EXCLUDE:
251 if (gdeleted || sdeleted)
252 return 0;
253 if (!(pmc->gsquery && !psf->sf_gsresp)) {
254 if (pmc->sfmode == MCAST_INCLUDE)
255 return 1;
256 /* don't include if this source is excluded
257 * in all filters
259 if (psf->sf_count[MCAST_INCLUDE])
260 return type == IGMPV3_MODE_IS_INCLUDE;
261 return pmc->sfcount[MCAST_EXCLUDE] ==
262 psf->sf_count[MCAST_EXCLUDE];
264 return 0;
265 case IGMPV3_CHANGE_TO_INCLUDE:
266 if (gdeleted || sdeleted)
267 return 0;
268 return psf->sf_count[MCAST_INCLUDE] != 0;
269 case IGMPV3_CHANGE_TO_EXCLUDE:
270 if (gdeleted || sdeleted)
271 return 0;
272 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
273 psf->sf_count[MCAST_INCLUDE])
274 return 0;
275 return pmc->sfcount[MCAST_EXCLUDE] ==
276 psf->sf_count[MCAST_EXCLUDE];
277 case IGMPV3_ALLOW_NEW_SOURCES:
278 if (gdeleted || !psf->sf_crcount)
279 return 0;
280 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
281 case IGMPV3_BLOCK_OLD_SOURCES:
282 if (pmc->sfmode == MCAST_INCLUDE)
283 return gdeleted || (psf->sf_crcount && sdeleted);
284 return psf->sf_crcount && !gdeleted && !sdeleted;
286 return 0;
289 static int
290 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
292 struct ip_sf_list *psf;
293 int scount = 0;
295 for (psf=pmc->sources; psf; psf=psf->sf_next) {
296 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
297 continue;
298 scount++;
300 return scount;
303 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
305 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
307 struct sk_buff *skb;
308 struct rtable *rt;
309 struct iphdr *pip;
310 struct igmpv3_report *pig;
311 struct net *net = dev_net(dev);
313 while (1) {
314 skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev),
315 GFP_ATOMIC | __GFP_NOWARN);
316 if (skb)
317 break;
318 size >>= 1;
319 if (size < 256)
320 return NULL;
322 igmp_skb_size(skb) = size;
324 rt = ip_route_output_ports(net, NULL, IGMPV3_ALL_MCR, 0,
325 0, 0,
326 IPPROTO_IGMP, 0, dev->ifindex);
327 if (IS_ERR(rt)) {
328 kfree_skb(skb);
329 return NULL;
331 if (rt->rt_src == 0) {
332 kfree_skb(skb);
333 ip_rt_put(rt);
334 return NULL;
337 skb_dst_set(skb, &rt->dst);
338 skb->dev = dev;
340 skb_reserve(skb, LL_RESERVED_SPACE(dev));
342 skb_reset_network_header(skb);
343 pip = ip_hdr(skb);
344 skb_put(skb, sizeof(struct iphdr) + 4);
346 pip->version = 4;
347 pip->ihl = (sizeof(struct iphdr)+4)>>2;
348 pip->tos = 0xc0;
349 pip->frag_off = htons(IP_DF);
350 pip->ttl = 1;
351 pip->daddr = rt->rt_dst;
352 pip->saddr = rt->rt_src;
353 pip->protocol = IPPROTO_IGMP;
354 pip->tot_len = 0; /* filled in later */
355 ip_select_ident(pip, &rt->dst, NULL);
356 ((u8*)&pip[1])[0] = IPOPT_RA;
357 ((u8*)&pip[1])[1] = 4;
358 ((u8*)&pip[1])[2] = 0;
359 ((u8*)&pip[1])[3] = 0;
361 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
362 skb_put(skb, sizeof(*pig));
363 pig = igmpv3_report_hdr(skb);
364 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
365 pig->resv1 = 0;
366 pig->csum = 0;
367 pig->resv2 = 0;
368 pig->ngrec = 0;
369 return skb;
372 static int igmpv3_sendpack(struct sk_buff *skb)
374 struct igmphdr *pig = igmp_hdr(skb);
375 const int igmplen = skb->tail - skb->transport_header;
377 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
379 return ip_local_out(skb);
382 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
384 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
387 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
388 int type, struct igmpv3_grec **ppgr)
390 struct net_device *dev = pmc->interface->dev;
391 struct igmpv3_report *pih;
392 struct igmpv3_grec *pgr;
394 if (!skb)
395 skb = igmpv3_newpack(dev, dev->mtu);
396 if (!skb)
397 return NULL;
398 pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
399 pgr->grec_type = type;
400 pgr->grec_auxwords = 0;
401 pgr->grec_nsrcs = 0;
402 pgr->grec_mca = pmc->multiaddr;
403 pih = igmpv3_report_hdr(skb);
404 pih->ngrec = htons(ntohs(pih->ngrec)+1);
405 *ppgr = pgr;
406 return skb;
409 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
410 skb_tailroom(skb)) : 0)
412 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
413 int type, int gdeleted, int sdeleted)
415 struct net_device *dev = pmc->interface->dev;
416 struct igmpv3_report *pih;
417 struct igmpv3_grec *pgr = NULL;
418 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
419 int scount, stotal, first, isquery, truncate;
421 if (pmc->multiaddr == IGMP_ALL_HOSTS)
422 return skb;
424 isquery = type == IGMPV3_MODE_IS_INCLUDE ||
425 type == IGMPV3_MODE_IS_EXCLUDE;
426 truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
427 type == IGMPV3_CHANGE_TO_EXCLUDE;
429 stotal = scount = 0;
431 psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
433 if (!*psf_list)
434 goto empty_source;
436 pih = skb ? igmpv3_report_hdr(skb) : NULL;
438 /* EX and TO_EX get a fresh packet, if needed */
439 if (truncate) {
440 if (pih && pih->ngrec &&
441 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
442 if (skb)
443 igmpv3_sendpack(skb);
444 skb = igmpv3_newpack(dev, dev->mtu);
447 first = 1;
448 psf_prev = NULL;
449 for (psf=*psf_list; psf; psf=psf_next) {
450 __be32 *psrc;
452 psf_next = psf->sf_next;
454 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
455 psf_prev = psf;
456 continue;
459 /* clear marks on query responses */
460 if (isquery)
461 psf->sf_gsresp = 0;
463 if (AVAILABLE(skb) < sizeof(__be32) +
464 first*sizeof(struct igmpv3_grec)) {
465 if (truncate && !first)
466 break; /* truncate these */
467 if (pgr)
468 pgr->grec_nsrcs = htons(scount);
469 if (skb)
470 igmpv3_sendpack(skb);
471 skb = igmpv3_newpack(dev, dev->mtu);
472 first = 1;
473 scount = 0;
475 if (first) {
476 skb = add_grhead(skb, pmc, type, &pgr);
477 first = 0;
479 if (!skb)
480 return NULL;
481 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
482 *psrc = psf->sf_inaddr;
483 scount++; stotal++;
484 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
485 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
486 psf->sf_crcount--;
487 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
488 if (psf_prev)
489 psf_prev->sf_next = psf->sf_next;
490 else
491 *psf_list = psf->sf_next;
492 kfree(psf);
493 continue;
496 psf_prev = psf;
499 empty_source:
500 if (!stotal) {
501 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
502 type == IGMPV3_BLOCK_OLD_SOURCES)
503 return skb;
504 if (pmc->crcount || isquery) {
505 /* make sure we have room for group header */
506 if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
507 igmpv3_sendpack(skb);
508 skb = NULL; /* add_grhead will get a new one */
510 skb = add_grhead(skb, pmc, type, &pgr);
513 if (pgr)
514 pgr->grec_nsrcs = htons(scount);
516 if (isquery)
517 pmc->gsquery = 0; /* clear query state on report */
518 return skb;
521 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
523 struct sk_buff *skb = NULL;
524 int type;
526 if (!pmc) {
527 rcu_read_lock();
528 for_each_pmc_rcu(in_dev, pmc) {
529 if (pmc->multiaddr == IGMP_ALL_HOSTS)
530 continue;
531 spin_lock_bh(&pmc->lock);
532 if (pmc->sfcount[MCAST_EXCLUDE])
533 type = IGMPV3_MODE_IS_EXCLUDE;
534 else
535 type = IGMPV3_MODE_IS_INCLUDE;
536 skb = add_grec(skb, pmc, type, 0, 0);
537 spin_unlock_bh(&pmc->lock);
539 rcu_read_unlock();
540 } else {
541 spin_lock_bh(&pmc->lock);
542 if (pmc->sfcount[MCAST_EXCLUDE])
543 type = IGMPV3_MODE_IS_EXCLUDE;
544 else
545 type = IGMPV3_MODE_IS_INCLUDE;
546 skb = add_grec(skb, pmc, type, 0, 0);
547 spin_unlock_bh(&pmc->lock);
549 if (!skb)
550 return 0;
551 return igmpv3_sendpack(skb);
555 * remove zero-count source records from a source filter list
557 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
559 struct ip_sf_list *psf_prev, *psf_next, *psf;
561 psf_prev = NULL;
562 for (psf=*ppsf; psf; psf = psf_next) {
563 psf_next = psf->sf_next;
564 if (psf->sf_crcount == 0) {
565 if (psf_prev)
566 psf_prev->sf_next = psf->sf_next;
567 else
568 *ppsf = psf->sf_next;
569 kfree(psf);
570 } else
571 psf_prev = psf;
575 static void igmpv3_send_cr(struct in_device *in_dev)
577 struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
578 struct sk_buff *skb = NULL;
579 int type, dtype;
581 rcu_read_lock();
582 spin_lock_bh(&in_dev->mc_tomb_lock);
584 /* deleted MCA's */
585 pmc_prev = NULL;
586 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
587 pmc_next = pmc->next;
588 if (pmc->sfmode == MCAST_INCLUDE) {
589 type = IGMPV3_BLOCK_OLD_SOURCES;
590 dtype = IGMPV3_BLOCK_OLD_SOURCES;
591 skb = add_grec(skb, pmc, type, 1, 0);
592 skb = add_grec(skb, pmc, dtype, 1, 1);
594 if (pmc->crcount) {
595 if (pmc->sfmode == MCAST_EXCLUDE) {
596 type = IGMPV3_CHANGE_TO_INCLUDE;
597 skb = add_grec(skb, pmc, type, 1, 0);
599 pmc->crcount--;
600 if (pmc->crcount == 0) {
601 igmpv3_clear_zeros(&pmc->tomb);
602 igmpv3_clear_zeros(&pmc->sources);
605 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
606 if (pmc_prev)
607 pmc_prev->next = pmc_next;
608 else
609 in_dev->mc_tomb = pmc_next;
610 in_dev_put(pmc->interface);
611 kfree(pmc);
612 } else
613 pmc_prev = pmc;
615 spin_unlock_bh(&in_dev->mc_tomb_lock);
617 /* change recs */
618 for_each_pmc_rcu(in_dev, pmc) {
619 spin_lock_bh(&pmc->lock);
620 if (pmc->sfcount[MCAST_EXCLUDE]) {
621 type = IGMPV3_BLOCK_OLD_SOURCES;
622 dtype = IGMPV3_ALLOW_NEW_SOURCES;
623 } else {
624 type = IGMPV3_ALLOW_NEW_SOURCES;
625 dtype = IGMPV3_BLOCK_OLD_SOURCES;
627 skb = add_grec(skb, pmc, type, 0, 0);
628 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */
630 /* filter mode changes */
631 if (pmc->crcount) {
632 if (pmc->sfmode == MCAST_EXCLUDE)
633 type = IGMPV3_CHANGE_TO_EXCLUDE;
634 else
635 type = IGMPV3_CHANGE_TO_INCLUDE;
636 skb = add_grec(skb, pmc, type, 0, 0);
637 pmc->crcount--;
639 spin_unlock_bh(&pmc->lock);
641 rcu_read_unlock();
643 if (!skb)
644 return;
645 (void) igmpv3_sendpack(skb);
648 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
649 int type)
651 struct sk_buff *skb;
652 struct iphdr *iph;
653 struct igmphdr *ih;
654 struct rtable *rt;
655 struct net_device *dev = in_dev->dev;
656 struct net *net = dev_net(dev);
657 __be32 group = pmc ? pmc->multiaddr : 0;
658 __be32 dst;
660 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
661 return igmpv3_send_report(in_dev, pmc);
662 else if (type == IGMP_HOST_LEAVE_MESSAGE)
663 dst = IGMP_ALL_ROUTER;
664 else
665 dst = group;
667 rt = ip_route_output_ports(net, NULL, dst, 0,
668 0, 0,
669 IPPROTO_IGMP, 0, dev->ifindex);
670 if (IS_ERR(rt))
671 return -1;
673 if (rt->rt_src == 0) {
674 ip_rt_put(rt);
675 return -1;
678 skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
679 if (skb == NULL) {
680 ip_rt_put(rt);
681 return -1;
684 skb_dst_set(skb, &rt->dst);
686 skb_reserve(skb, LL_RESERVED_SPACE(dev));
688 skb_reset_network_header(skb);
689 iph = ip_hdr(skb);
690 skb_put(skb, sizeof(struct iphdr) + 4);
692 iph->version = 4;
693 iph->ihl = (sizeof(struct iphdr)+4)>>2;
694 iph->tos = 0xc0;
695 iph->frag_off = htons(IP_DF);
696 iph->ttl = 1;
697 iph->daddr = dst;
698 iph->saddr = rt->rt_src;
699 iph->protocol = IPPROTO_IGMP;
700 ip_select_ident(iph, &rt->dst, NULL);
701 ((u8*)&iph[1])[0] = IPOPT_RA;
702 ((u8*)&iph[1])[1] = 4;
703 ((u8*)&iph[1])[2] = 0;
704 ((u8*)&iph[1])[3] = 0;
706 ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
707 ih->type = type;
708 ih->code = 0;
709 ih->csum = 0;
710 ih->group = group;
711 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
713 return ip_local_out(skb);
716 static void igmp_gq_timer_expire(unsigned long data)
718 struct in_device *in_dev = (struct in_device *)data;
720 in_dev->mr_gq_running = 0;
721 igmpv3_send_report(in_dev, NULL);
722 __in_dev_put(in_dev);
725 static void igmp_ifc_timer_expire(unsigned long data)
727 struct in_device *in_dev = (struct in_device *)data;
729 igmpv3_send_cr(in_dev);
730 if (in_dev->mr_ifc_count) {
731 in_dev->mr_ifc_count--;
732 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
734 __in_dev_put(in_dev);
737 static void igmp_ifc_event(struct in_device *in_dev)
739 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
740 return;
741 in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
742 IGMP_Unsolicited_Report_Count;
743 igmp_ifc_start_timer(in_dev, 1);
747 static void igmp_timer_expire(unsigned long data)
749 struct ip_mc_list *im=(struct ip_mc_list *)data;
750 struct in_device *in_dev = im->interface;
752 spin_lock(&im->lock);
753 im->tm_running = 0;
755 if (im->unsolicit_count) {
756 im->unsolicit_count--;
757 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
759 im->reporter = 1;
760 spin_unlock(&im->lock);
762 if (IGMP_V1_SEEN(in_dev))
763 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
764 else if (IGMP_V2_SEEN(in_dev))
765 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
766 else
767 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
769 ip_ma_put(im);
772 /* mark EXCLUDE-mode sources */
773 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
775 struct ip_sf_list *psf;
776 int i, scount;
778 scount = 0;
779 for (psf=pmc->sources; psf; psf=psf->sf_next) {
780 if (scount == nsrcs)
781 break;
782 for (i=0; i<nsrcs; i++) {
783 /* skip inactive filters */
784 if (pmc->sfcount[MCAST_INCLUDE] ||
785 pmc->sfcount[MCAST_EXCLUDE] !=
786 psf->sf_count[MCAST_EXCLUDE])
787 continue;
788 if (srcs[i] == psf->sf_inaddr) {
789 scount++;
790 break;
794 pmc->gsquery = 0;
795 if (scount == nsrcs) /* all sources excluded */
796 return 0;
797 return 1;
800 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
802 struct ip_sf_list *psf;
803 int i, scount;
805 if (pmc->sfmode == MCAST_EXCLUDE)
806 return igmp_xmarksources(pmc, nsrcs, srcs);
808 /* mark INCLUDE-mode sources */
809 scount = 0;
810 for (psf=pmc->sources; psf; psf=psf->sf_next) {
811 if (scount == nsrcs)
812 break;
813 for (i=0; i<nsrcs; i++)
814 if (srcs[i] == psf->sf_inaddr) {
815 psf->sf_gsresp = 1;
816 scount++;
817 break;
820 if (!scount) {
821 pmc->gsquery = 0;
822 return 0;
824 pmc->gsquery = 1;
825 return 1;
828 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
830 struct ip_mc_list *im;
832 /* Timers are only set for non-local groups */
834 if (group == IGMP_ALL_HOSTS)
835 return;
837 rcu_read_lock();
838 for_each_pmc_rcu(in_dev, im) {
839 if (im->multiaddr == group) {
840 igmp_stop_timer(im);
841 break;
844 rcu_read_unlock();
847 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
848 int len)
850 struct igmphdr *ih = igmp_hdr(skb);
851 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
852 struct ip_mc_list *im;
853 __be32 group = ih->group;
854 int max_delay;
855 int mark = 0;
858 if (len == 8) {
859 if (ih->code == 0) {
860 /* Alas, old v1 router presents here. */
862 max_delay = IGMP_Query_Response_Interval;
863 in_dev->mr_v1_seen = jiffies +
864 IGMP_V1_Router_Present_Timeout;
865 group = 0;
866 } else {
867 /* v2 router present */
868 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
869 in_dev->mr_v2_seen = jiffies +
870 IGMP_V2_Router_Present_Timeout;
872 /* cancel the interface change timer */
873 in_dev->mr_ifc_count = 0;
874 if (del_timer(&in_dev->mr_ifc_timer))
875 __in_dev_put(in_dev);
876 /* clear deleted report items */
877 igmpv3_clear_delrec(in_dev);
878 } else if (len < 12) {
879 return; /* ignore bogus packet; freed by caller */
880 } else if (IGMP_V1_SEEN(in_dev)) {
881 /* This is a v3 query with v1 queriers present */
882 max_delay = IGMP_Query_Response_Interval;
883 group = 0;
884 } else if (IGMP_V2_SEEN(in_dev)) {
885 /* this is a v3 query with v2 queriers present;
886 * Interpretation of the max_delay code is problematic here.
887 * A real v2 host would use ih_code directly, while v3 has a
888 * different encoding. We use the v3 encoding as more likely
889 * to be intended in a v3 query.
891 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
892 } else { /* v3 */
893 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
894 return;
896 ih3 = igmpv3_query_hdr(skb);
897 if (ih3->nsrcs) {
898 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
899 + ntohs(ih3->nsrcs)*sizeof(__be32)))
900 return;
901 ih3 = igmpv3_query_hdr(skb);
904 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
905 if (!max_delay)
906 max_delay = 1; /* can't mod w/ 0 */
907 in_dev->mr_maxdelay = max_delay;
908 if (ih3->qrv)
909 in_dev->mr_qrv = ih3->qrv;
910 if (!group) { /* general query */
911 if (ih3->nsrcs)
912 return; /* no sources allowed */
913 igmp_gq_start_timer(in_dev);
914 return;
916 /* mark sources to include, if group & source-specific */
917 mark = ih3->nsrcs != 0;
921 * - Start the timers in all of our membership records
922 * that the query applies to for the interface on
923 * which the query arrived excl. those that belong
924 * to a "local" group (224.0.0.X)
925 * - For timers already running check if they need to
926 * be reset.
927 * - Use the igmp->igmp_code field as the maximum
928 * delay possible
930 rcu_read_lock();
931 for_each_pmc_rcu(in_dev, im) {
932 int changed;
934 if (group && group != im->multiaddr)
935 continue;
936 if (im->multiaddr == IGMP_ALL_HOSTS)
937 continue;
938 spin_lock_bh(&im->lock);
939 if (im->tm_running)
940 im->gsquery = im->gsquery && mark;
941 else
942 im->gsquery = mark;
943 changed = !im->gsquery ||
944 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
945 spin_unlock_bh(&im->lock);
946 if (changed)
947 igmp_mod_timer(im, max_delay);
949 rcu_read_unlock();
952 /* called in rcu_read_lock() section */
953 int igmp_rcv(struct sk_buff *skb)
955 /* This basically follows the spec line by line -- see RFC1112 */
956 struct igmphdr *ih;
957 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
958 int len = skb->len;
960 if (in_dev == NULL)
961 goto drop;
963 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
964 goto drop;
966 switch (skb->ip_summed) {
967 case CHECKSUM_COMPLETE:
968 if (!csum_fold(skb->csum))
969 break;
970 /* fall through */
971 case CHECKSUM_NONE:
972 skb->csum = 0;
973 if (__skb_checksum_complete(skb))
974 goto drop;
977 ih = igmp_hdr(skb);
978 switch (ih->type) {
979 case IGMP_HOST_MEMBERSHIP_QUERY:
980 igmp_heard_query(in_dev, skb, len);
981 break;
982 case IGMP_HOST_MEMBERSHIP_REPORT:
983 case IGMPV2_HOST_MEMBERSHIP_REPORT:
984 /* Is it our report looped back? */
985 if (rt_is_output_route(skb_rtable(skb)))
986 break;
987 /* don't rely on MC router hearing unicast reports */
988 if (skb->pkt_type == PACKET_MULTICAST ||
989 skb->pkt_type == PACKET_BROADCAST)
990 igmp_heard_report(in_dev, ih->group);
991 break;
992 case IGMP_PIM:
993 #ifdef CONFIG_IP_PIMSM_V1
994 return pim_rcv_v1(skb);
995 #endif
996 case IGMPV3_HOST_MEMBERSHIP_REPORT:
997 case IGMP_DVMRP:
998 case IGMP_TRACE:
999 case IGMP_HOST_LEAVE_MESSAGE:
1000 case IGMP_MTRACE:
1001 case IGMP_MTRACE_RESP:
1002 break;
1003 default:
1004 break;
1007 drop:
1008 kfree_skb(skb);
1009 return 0;
1012 #endif
1016 * Add a filter to a device
1019 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1021 char buf[MAX_ADDR_LEN];
1022 struct net_device *dev = in_dev->dev;
1024 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1025 We will get multicast token leakage, when IFF_MULTICAST
1026 is changed. This check should be done in dev->set_multicast_list
1027 routine. Something sort of:
1028 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1029 --ANK
1031 if (arp_mc_map(addr, buf, dev, 0) == 0)
1032 dev_mc_add(dev, buf);
1036 * Remove a filter from a device
1039 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1041 char buf[MAX_ADDR_LEN];
1042 struct net_device *dev = in_dev->dev;
1044 if (arp_mc_map(addr, buf, dev, 0) == 0)
1045 dev_mc_del(dev, buf);
1048 #ifdef CONFIG_IP_MULTICAST
1050 * deleted ip_mc_list manipulation
1052 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1054 struct ip_mc_list *pmc;
1056 /* this is an "ip_mc_list" for convenience; only the fields below
1057 * are actually used. In particular, the refcnt and users are not
1058 * used for management of the delete list. Using the same structure
1059 * for deleted items allows change reports to use common code with
1060 * non-deleted or query-response MCA's.
1062 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1063 if (!pmc)
1064 return;
1065 spin_lock_bh(&im->lock);
1066 pmc->interface = im->interface;
1067 in_dev_hold(in_dev);
1068 pmc->multiaddr = im->multiaddr;
1069 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1070 IGMP_Unsolicited_Report_Count;
1071 pmc->sfmode = im->sfmode;
1072 if (pmc->sfmode == MCAST_INCLUDE) {
1073 struct ip_sf_list *psf;
1075 pmc->tomb = im->tomb;
1076 pmc->sources = im->sources;
1077 im->tomb = im->sources = NULL;
1078 for (psf=pmc->sources; psf; psf=psf->sf_next)
1079 psf->sf_crcount = pmc->crcount;
1081 spin_unlock_bh(&im->lock);
1083 spin_lock_bh(&in_dev->mc_tomb_lock);
1084 pmc->next = in_dev->mc_tomb;
1085 in_dev->mc_tomb = pmc;
1086 spin_unlock_bh(&in_dev->mc_tomb_lock);
1089 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1091 struct ip_mc_list *pmc, *pmc_prev;
1092 struct ip_sf_list *psf, *psf_next;
1094 spin_lock_bh(&in_dev->mc_tomb_lock);
1095 pmc_prev = NULL;
1096 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1097 if (pmc->multiaddr == multiaddr)
1098 break;
1099 pmc_prev = pmc;
1101 if (pmc) {
1102 if (pmc_prev)
1103 pmc_prev->next = pmc->next;
1104 else
1105 in_dev->mc_tomb = pmc->next;
1107 spin_unlock_bh(&in_dev->mc_tomb_lock);
1108 if (pmc) {
1109 for (psf=pmc->tomb; psf; psf=psf_next) {
1110 psf_next = psf->sf_next;
1111 kfree(psf);
1113 in_dev_put(pmc->interface);
1114 kfree(pmc);
1118 static void igmpv3_clear_delrec(struct in_device *in_dev)
1120 struct ip_mc_list *pmc, *nextpmc;
1122 spin_lock_bh(&in_dev->mc_tomb_lock);
1123 pmc = in_dev->mc_tomb;
1124 in_dev->mc_tomb = NULL;
1125 spin_unlock_bh(&in_dev->mc_tomb_lock);
1127 for (; pmc; pmc = nextpmc) {
1128 nextpmc = pmc->next;
1129 ip_mc_clear_src(pmc);
1130 in_dev_put(pmc->interface);
1131 kfree(pmc);
1133 /* clear dead sources, too */
1134 rcu_read_lock();
1135 for_each_pmc_rcu(in_dev, pmc) {
1136 struct ip_sf_list *psf, *psf_next;
1138 spin_lock_bh(&pmc->lock);
1139 psf = pmc->tomb;
1140 pmc->tomb = NULL;
1141 spin_unlock_bh(&pmc->lock);
1142 for (; psf; psf=psf_next) {
1143 psf_next = psf->sf_next;
1144 kfree(psf);
1147 rcu_read_unlock();
1149 #endif
1151 static void igmp_group_dropped(struct ip_mc_list *im)
1153 struct in_device *in_dev = im->interface;
1154 #ifdef CONFIG_IP_MULTICAST
1155 int reporter;
1156 #endif
1158 if (im->loaded) {
1159 im->loaded = 0;
1160 ip_mc_filter_del(in_dev, im->multiaddr);
1163 #ifdef CONFIG_IP_MULTICAST
1164 if (im->multiaddr == IGMP_ALL_HOSTS)
1165 return;
1167 reporter = im->reporter;
1168 igmp_stop_timer(im);
1170 if (!in_dev->dead) {
1171 if (IGMP_V1_SEEN(in_dev))
1172 goto done;
1173 if (IGMP_V2_SEEN(in_dev)) {
1174 if (reporter)
1175 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1176 goto done;
1178 /* IGMPv3 */
1179 igmpv3_add_delrec(in_dev, im);
1181 igmp_ifc_event(in_dev);
1183 done:
1184 #endif
1185 ip_mc_clear_src(im);
1188 static void igmp_group_added(struct ip_mc_list *im)
1190 struct in_device *in_dev = im->interface;
1192 if (im->loaded == 0) {
1193 im->loaded = 1;
1194 ip_mc_filter_add(in_dev, im->multiaddr);
1197 #ifdef CONFIG_IP_MULTICAST
1198 if (im->multiaddr == IGMP_ALL_HOSTS)
1199 return;
1201 if (in_dev->dead)
1202 return;
1203 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1204 spin_lock_bh(&im->lock);
1205 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1206 spin_unlock_bh(&im->lock);
1207 return;
1209 /* else, v3 */
1211 im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1212 IGMP_Unsolicited_Report_Count;
1213 igmp_ifc_event(in_dev);
1214 #endif
1219 * Multicast list managers
1224 * A socket has joined a multicast group on device dev.
1227 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1229 struct ip_mc_list *im;
1231 ASSERT_RTNL();
1233 for_each_pmc_rtnl(in_dev, im) {
1234 if (im->multiaddr == addr) {
1235 im->users++;
1236 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1237 goto out;
1241 im = kzalloc(sizeof(*im), GFP_KERNEL);
1242 if (!im)
1243 goto out;
1245 im->users = 1;
1246 im->interface = in_dev;
1247 in_dev_hold(in_dev);
1248 im->multiaddr = addr;
1249 /* initial mode is (EX, empty) */
1250 im->sfmode = MCAST_EXCLUDE;
1251 im->sfcount[MCAST_EXCLUDE] = 1;
1252 atomic_set(&im->refcnt, 1);
1253 spin_lock_init(&im->lock);
1254 #ifdef CONFIG_IP_MULTICAST
1255 setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1256 im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1257 #endif
1259 im->next_rcu = in_dev->mc_list;
1260 in_dev->mc_count++;
1261 rcu_assign_pointer(in_dev->mc_list, im);
1263 #ifdef CONFIG_IP_MULTICAST
1264 igmpv3_del_delrec(in_dev, im->multiaddr);
1265 #endif
1266 igmp_group_added(im);
1267 if (!in_dev->dead)
1268 ip_rt_multicast_event(in_dev);
1269 out:
1270 return;
1272 EXPORT_SYMBOL(ip_mc_inc_group);
1275 * Resend IGMP JOIN report; used for bonding.
1276 * Called with rcu_read_lock()
1278 void ip_mc_rejoin_groups(struct in_device *in_dev)
1280 #ifdef CONFIG_IP_MULTICAST
1281 struct ip_mc_list *im;
1282 int type;
1284 for_each_pmc_rcu(in_dev, im) {
1285 if (im->multiaddr == IGMP_ALL_HOSTS)
1286 continue;
1288 /* a failover is happening and switches
1289 * must be notified immediately
1291 if (IGMP_V1_SEEN(in_dev))
1292 type = IGMP_HOST_MEMBERSHIP_REPORT;
1293 else if (IGMP_V2_SEEN(in_dev))
1294 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1295 else
1296 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1297 igmp_send_report(in_dev, im, type);
1299 #endif
1301 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1304 * A socket has left a multicast group on device dev
1307 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1309 struct ip_mc_list *i;
1310 struct ip_mc_list __rcu **ip;
1312 ASSERT_RTNL();
1314 for (ip = &in_dev->mc_list;
1315 (i = rtnl_dereference(*ip)) != NULL;
1316 ip = &i->next_rcu) {
1317 if (i->multiaddr == addr) {
1318 if (--i->users == 0) {
1319 *ip = i->next_rcu;
1320 in_dev->mc_count--;
1321 igmp_group_dropped(i);
1323 if (!in_dev->dead)
1324 ip_rt_multicast_event(in_dev);
1326 ip_ma_put(i);
1327 return;
1329 break;
1333 EXPORT_SYMBOL(ip_mc_dec_group);
1335 /* Device changing type */
1337 void ip_mc_unmap(struct in_device *in_dev)
1339 struct ip_mc_list *pmc;
1341 ASSERT_RTNL();
1343 for_each_pmc_rtnl(in_dev, pmc)
1344 igmp_group_dropped(pmc);
1347 void ip_mc_remap(struct in_device *in_dev)
1349 struct ip_mc_list *pmc;
1351 ASSERT_RTNL();
1353 for_each_pmc_rtnl(in_dev, pmc)
1354 igmp_group_added(pmc);
1357 /* Device going down */
1359 void ip_mc_down(struct in_device *in_dev)
1361 struct ip_mc_list *pmc;
1363 ASSERT_RTNL();
1365 for_each_pmc_rtnl(in_dev, pmc)
1366 igmp_group_dropped(pmc);
1368 #ifdef CONFIG_IP_MULTICAST
1369 in_dev->mr_ifc_count = 0;
1370 if (del_timer(&in_dev->mr_ifc_timer))
1371 __in_dev_put(in_dev);
1372 in_dev->mr_gq_running = 0;
1373 if (del_timer(&in_dev->mr_gq_timer))
1374 __in_dev_put(in_dev);
1375 igmpv3_clear_delrec(in_dev);
1376 #endif
1378 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1381 void ip_mc_init_dev(struct in_device *in_dev)
1383 ASSERT_RTNL();
1385 in_dev->mc_tomb = NULL;
1386 #ifdef CONFIG_IP_MULTICAST
1387 in_dev->mr_gq_running = 0;
1388 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1389 (unsigned long)in_dev);
1390 in_dev->mr_ifc_count = 0;
1391 in_dev->mc_count = 0;
1392 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1393 (unsigned long)in_dev);
1394 in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1395 #endif
1397 spin_lock_init(&in_dev->mc_tomb_lock);
1400 /* Device going up */
1402 void ip_mc_up(struct in_device *in_dev)
1404 struct ip_mc_list *pmc;
1406 ASSERT_RTNL();
1408 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1410 for_each_pmc_rtnl(in_dev, pmc)
1411 igmp_group_added(pmc);
1415 * Device is about to be destroyed: clean up.
1418 void ip_mc_destroy_dev(struct in_device *in_dev)
1420 struct ip_mc_list *i;
1422 ASSERT_RTNL();
1424 /* Deactivate timers */
1425 ip_mc_down(in_dev);
1427 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1428 in_dev->mc_list = i->next_rcu;
1429 in_dev->mc_count--;
1431 igmp_group_dropped(i);
1432 ip_ma_put(i);
1436 /* RTNL is locked */
1437 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1439 struct net_device *dev = NULL;
1440 struct in_device *idev = NULL;
1442 if (imr->imr_ifindex) {
1443 idev = inetdev_by_index(net, imr->imr_ifindex);
1444 return idev;
1446 if (imr->imr_address.s_addr) {
1447 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1448 if (!dev)
1449 return NULL;
1452 if (!dev) {
1453 struct rtable *rt = ip_route_output(net,
1454 imr->imr_multiaddr.s_addr,
1455 0, 0, 0);
1456 if (!IS_ERR(rt)) {
1457 dev = rt->dst.dev;
1458 ip_rt_put(rt);
1461 if (dev) {
1462 imr->imr_ifindex = dev->ifindex;
1463 idev = __in_dev_get_rtnl(dev);
1465 return idev;
1469 * Join a socket to a group
1471 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1472 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1475 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1476 __be32 *psfsrc)
1478 struct ip_sf_list *psf, *psf_prev;
1479 int rv = 0;
1481 psf_prev = NULL;
1482 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1483 if (psf->sf_inaddr == *psfsrc)
1484 break;
1485 psf_prev = psf;
1487 if (!psf || psf->sf_count[sfmode] == 0) {
1488 /* source filter not found, or count wrong => bug */
1489 return -ESRCH;
1491 psf->sf_count[sfmode]--;
1492 if (psf->sf_count[sfmode] == 0) {
1493 ip_rt_multicast_event(pmc->interface);
1495 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1496 #ifdef CONFIG_IP_MULTICAST
1497 struct in_device *in_dev = pmc->interface;
1498 #endif
1500 /* no more filters for this source */
1501 if (psf_prev)
1502 psf_prev->sf_next = psf->sf_next;
1503 else
1504 pmc->sources = psf->sf_next;
1505 #ifdef CONFIG_IP_MULTICAST
1506 if (psf->sf_oldin &&
1507 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1508 psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1509 IGMP_Unsolicited_Report_Count;
1510 psf->sf_next = pmc->tomb;
1511 pmc->tomb = psf;
1512 rv = 1;
1513 } else
1514 #endif
1515 kfree(psf);
1517 return rv;
1520 #ifndef CONFIG_IP_MULTICAST
1521 #define igmp_ifc_event(x) do { } while (0)
1522 #endif
1524 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1525 int sfcount, __be32 *psfsrc, int delta)
1527 struct ip_mc_list *pmc;
1528 int changerec = 0;
1529 int i, err;
1531 if (!in_dev)
1532 return -ENODEV;
1533 rcu_read_lock();
1534 for_each_pmc_rcu(in_dev, pmc) {
1535 if (*pmca == pmc->multiaddr)
1536 break;
1538 if (!pmc) {
1539 /* MCA not found?? bug */
1540 rcu_read_unlock();
1541 return -ESRCH;
1543 spin_lock_bh(&pmc->lock);
1544 rcu_read_unlock();
1545 #ifdef CONFIG_IP_MULTICAST
1546 sf_markstate(pmc);
1547 #endif
1548 if (!delta) {
1549 err = -EINVAL;
1550 if (!pmc->sfcount[sfmode])
1551 goto out_unlock;
1552 pmc->sfcount[sfmode]--;
1554 err = 0;
1555 for (i=0; i<sfcount; i++) {
1556 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1558 changerec |= rv > 0;
1559 if (!err && rv < 0)
1560 err = rv;
1562 if (pmc->sfmode == MCAST_EXCLUDE &&
1563 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1564 pmc->sfcount[MCAST_INCLUDE]) {
1565 #ifdef CONFIG_IP_MULTICAST
1566 struct ip_sf_list *psf;
1567 #endif
1569 /* filter mode change */
1570 pmc->sfmode = MCAST_INCLUDE;
1571 #ifdef CONFIG_IP_MULTICAST
1572 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1573 IGMP_Unsolicited_Report_Count;
1574 in_dev->mr_ifc_count = pmc->crcount;
1575 for (psf=pmc->sources; psf; psf = psf->sf_next)
1576 psf->sf_crcount = 0;
1577 igmp_ifc_event(pmc->interface);
1578 } else if (sf_setstate(pmc) || changerec) {
1579 igmp_ifc_event(pmc->interface);
1580 #endif
1582 out_unlock:
1583 spin_unlock_bh(&pmc->lock);
1584 return err;
1588 * Add multicast single-source filter to the interface list
1590 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1591 __be32 *psfsrc, int delta)
1593 struct ip_sf_list *psf, *psf_prev;
1595 psf_prev = NULL;
1596 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1597 if (psf->sf_inaddr == *psfsrc)
1598 break;
1599 psf_prev = psf;
1601 if (!psf) {
1602 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1603 if (!psf)
1604 return -ENOBUFS;
1605 psf->sf_inaddr = *psfsrc;
1606 if (psf_prev) {
1607 psf_prev->sf_next = psf;
1608 } else
1609 pmc->sources = psf;
1611 psf->sf_count[sfmode]++;
1612 if (psf->sf_count[sfmode] == 1) {
1613 ip_rt_multicast_event(pmc->interface);
1615 return 0;
1618 #ifdef CONFIG_IP_MULTICAST
1619 static void sf_markstate(struct ip_mc_list *pmc)
1621 struct ip_sf_list *psf;
1622 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1624 for (psf=pmc->sources; psf; psf=psf->sf_next)
1625 if (pmc->sfcount[MCAST_EXCLUDE]) {
1626 psf->sf_oldin = mca_xcount ==
1627 psf->sf_count[MCAST_EXCLUDE] &&
1628 !psf->sf_count[MCAST_INCLUDE];
1629 } else
1630 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1633 static int sf_setstate(struct ip_mc_list *pmc)
1635 struct ip_sf_list *psf, *dpsf;
1636 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1637 int qrv = pmc->interface->mr_qrv;
1638 int new_in, rv;
1640 rv = 0;
1641 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1642 if (pmc->sfcount[MCAST_EXCLUDE]) {
1643 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1644 !psf->sf_count[MCAST_INCLUDE];
1645 } else
1646 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1647 if (new_in) {
1648 if (!psf->sf_oldin) {
1649 struct ip_sf_list *prev = NULL;
1651 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1652 if (dpsf->sf_inaddr == psf->sf_inaddr)
1653 break;
1654 prev = dpsf;
1656 if (dpsf) {
1657 if (prev)
1658 prev->sf_next = dpsf->sf_next;
1659 else
1660 pmc->tomb = dpsf->sf_next;
1661 kfree(dpsf);
1663 psf->sf_crcount = qrv;
1664 rv++;
1666 } else if (psf->sf_oldin) {
1668 psf->sf_crcount = 0;
1670 * add or update "delete" records if an active filter
1671 * is now inactive
1673 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1674 if (dpsf->sf_inaddr == psf->sf_inaddr)
1675 break;
1676 if (!dpsf) {
1677 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1678 if (!dpsf)
1679 continue;
1680 *dpsf = *psf;
1681 /* pmc->lock held by callers */
1682 dpsf->sf_next = pmc->tomb;
1683 pmc->tomb = dpsf;
1685 dpsf->sf_crcount = qrv;
1686 rv++;
1689 return rv;
1691 #endif
1694 * Add multicast source filter list to the interface list
1696 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1697 int sfcount, __be32 *psfsrc, int delta)
1699 struct ip_mc_list *pmc;
1700 int isexclude;
1701 int i, err;
1703 if (!in_dev)
1704 return -ENODEV;
1705 rcu_read_lock();
1706 for_each_pmc_rcu(in_dev, pmc) {
1707 if (*pmca == pmc->multiaddr)
1708 break;
1710 if (!pmc) {
1711 /* MCA not found?? bug */
1712 rcu_read_unlock();
1713 return -ESRCH;
1715 spin_lock_bh(&pmc->lock);
1716 rcu_read_unlock();
1718 #ifdef CONFIG_IP_MULTICAST
1719 sf_markstate(pmc);
1720 #endif
1721 isexclude = pmc->sfmode == MCAST_EXCLUDE;
1722 if (!delta)
1723 pmc->sfcount[sfmode]++;
1724 err = 0;
1725 for (i=0; i<sfcount; i++) {
1726 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1727 if (err)
1728 break;
1730 if (err) {
1731 int j;
1733 pmc->sfcount[sfmode]--;
1734 for (j=0; j<i; j++)
1735 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1736 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1737 #ifdef CONFIG_IP_MULTICAST
1738 struct ip_sf_list *psf;
1739 in_dev = pmc->interface;
1740 #endif
1742 /* filter mode change */
1743 if (pmc->sfcount[MCAST_EXCLUDE])
1744 pmc->sfmode = MCAST_EXCLUDE;
1745 else if (pmc->sfcount[MCAST_INCLUDE])
1746 pmc->sfmode = MCAST_INCLUDE;
1747 #ifdef CONFIG_IP_MULTICAST
1748 /* else no filters; keep old mode for reports */
1750 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1751 IGMP_Unsolicited_Report_Count;
1752 in_dev->mr_ifc_count = pmc->crcount;
1753 for (psf=pmc->sources; psf; psf = psf->sf_next)
1754 psf->sf_crcount = 0;
1755 igmp_ifc_event(in_dev);
1756 } else if (sf_setstate(pmc)) {
1757 igmp_ifc_event(in_dev);
1758 #endif
1760 spin_unlock_bh(&pmc->lock);
1761 return err;
1764 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1766 struct ip_sf_list *psf, *nextpsf;
1768 for (psf=pmc->tomb; psf; psf=nextpsf) {
1769 nextpsf = psf->sf_next;
1770 kfree(psf);
1772 pmc->tomb = NULL;
1773 for (psf=pmc->sources; psf; psf=nextpsf) {
1774 nextpsf = psf->sf_next;
1775 kfree(psf);
1777 pmc->sources = NULL;
1778 pmc->sfmode = MCAST_EXCLUDE;
1779 pmc->sfcount[MCAST_INCLUDE] = 0;
1780 pmc->sfcount[MCAST_EXCLUDE] = 1;
1785 * Join a multicast group
1787 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1789 int err;
1790 __be32 addr = imr->imr_multiaddr.s_addr;
1791 struct ip_mc_socklist *iml = NULL, *i;
1792 struct in_device *in_dev;
1793 struct inet_sock *inet = inet_sk(sk);
1794 struct net *net = sock_net(sk);
1795 int ifindex;
1796 int count = 0;
1798 if (!ipv4_is_multicast(addr))
1799 return -EINVAL;
1801 rtnl_lock();
1803 in_dev = ip_mc_find_dev(net, imr);
1805 if (!in_dev) {
1806 iml = NULL;
1807 err = -ENODEV;
1808 goto done;
1811 err = -EADDRINUSE;
1812 ifindex = imr->imr_ifindex;
1813 for_each_pmc_rtnl(inet, i) {
1814 if (i->multi.imr_multiaddr.s_addr == addr &&
1815 i->multi.imr_ifindex == ifindex)
1816 goto done;
1817 count++;
1819 err = -ENOBUFS;
1820 if (count >= sysctl_igmp_max_memberships)
1821 goto done;
1822 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1823 if (iml == NULL)
1824 goto done;
1826 memcpy(&iml->multi, imr, sizeof(*imr));
1827 iml->next_rcu = inet->mc_list;
1828 iml->sflist = NULL;
1829 iml->sfmode = MCAST_EXCLUDE;
1830 rcu_assign_pointer(inet->mc_list, iml);
1831 ip_mc_inc_group(in_dev, addr);
1832 err = 0;
1833 done:
1834 rtnl_unlock();
1835 return err;
1837 EXPORT_SYMBOL(ip_mc_join_group);
1839 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1841 kfree(container_of(rp, struct ip_sf_socklist, rcu));
1842 /* sk_omem_alloc should have been decreased by the caller*/
1845 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1846 struct in_device *in_dev)
1848 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1849 int err;
1851 if (psf == NULL) {
1852 /* any-source empty exclude case */
1853 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1854 iml->sfmode, 0, NULL, 0);
1856 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1857 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1858 rcu_assign_pointer(iml->sflist, NULL);
1859 /* decrease mem now to avoid the memleak warning */
1860 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1861 call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1862 return err;
1866 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1868 kfree(container_of(rp, struct ip_mc_socklist, rcu));
1869 /* sk_omem_alloc should have been decreased by the caller*/
1874 * Ask a socket to leave a group.
1877 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1879 struct inet_sock *inet = inet_sk(sk);
1880 struct ip_mc_socklist *iml;
1881 struct ip_mc_socklist __rcu **imlp;
1882 struct in_device *in_dev;
1883 struct net *net = sock_net(sk);
1884 __be32 group = imr->imr_multiaddr.s_addr;
1885 u32 ifindex;
1886 int ret = -EADDRNOTAVAIL;
1888 rtnl_lock();
1889 in_dev = ip_mc_find_dev(net, imr);
1890 ifindex = imr->imr_ifindex;
1891 for (imlp = &inet->mc_list;
1892 (iml = rtnl_dereference(*imlp)) != NULL;
1893 imlp = &iml->next_rcu) {
1894 if (iml->multi.imr_multiaddr.s_addr != group)
1895 continue;
1896 if (ifindex) {
1897 if (iml->multi.imr_ifindex != ifindex)
1898 continue;
1899 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1900 iml->multi.imr_address.s_addr)
1901 continue;
1903 (void) ip_mc_leave_src(sk, iml, in_dev);
1905 *imlp = iml->next_rcu;
1907 if (in_dev)
1908 ip_mc_dec_group(in_dev, group);
1909 rtnl_unlock();
1910 /* decrease mem now to avoid the memleak warning */
1911 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1912 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1913 return 0;
1915 if (!in_dev)
1916 ret = -ENODEV;
1917 rtnl_unlock();
1918 return ret;
1921 int ip_mc_source(int add, int omode, struct sock *sk, struct
1922 ip_mreq_source *mreqs, int ifindex)
1924 int err;
1925 struct ip_mreqn imr;
1926 __be32 addr = mreqs->imr_multiaddr;
1927 struct ip_mc_socklist *pmc;
1928 struct in_device *in_dev = NULL;
1929 struct inet_sock *inet = inet_sk(sk);
1930 struct ip_sf_socklist *psl;
1931 struct net *net = sock_net(sk);
1932 int leavegroup = 0;
1933 int i, j, rv;
1935 if (!ipv4_is_multicast(addr))
1936 return -EINVAL;
1938 rtnl_lock();
1940 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1941 imr.imr_address.s_addr = mreqs->imr_interface;
1942 imr.imr_ifindex = ifindex;
1943 in_dev = ip_mc_find_dev(net, &imr);
1945 if (!in_dev) {
1946 err = -ENODEV;
1947 goto done;
1949 err = -EADDRNOTAVAIL;
1951 for_each_pmc_rtnl(inet, pmc) {
1952 if ((pmc->multi.imr_multiaddr.s_addr ==
1953 imr.imr_multiaddr.s_addr) &&
1954 (pmc->multi.imr_ifindex == imr.imr_ifindex))
1955 break;
1957 if (!pmc) { /* must have a prior join */
1958 err = -EINVAL;
1959 goto done;
1961 /* if a source filter was set, must be the same mode as before */
1962 if (pmc->sflist) {
1963 if (pmc->sfmode != omode) {
1964 err = -EINVAL;
1965 goto done;
1967 } else if (pmc->sfmode != omode) {
1968 /* allow mode switches for empty-set filters */
1969 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1970 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1971 NULL, 0);
1972 pmc->sfmode = omode;
1975 psl = rtnl_dereference(pmc->sflist);
1976 if (!add) {
1977 if (!psl)
1978 goto done; /* err = -EADDRNOTAVAIL */
1979 rv = !0;
1980 for (i=0; i<psl->sl_count; i++) {
1981 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1982 sizeof(__be32));
1983 if (rv == 0)
1984 break;
1986 if (rv) /* source not found */
1987 goto done; /* err = -EADDRNOTAVAIL */
1989 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1990 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1991 leavegroup = 1;
1992 goto done;
1995 /* update the interface filter */
1996 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1997 &mreqs->imr_sourceaddr, 1);
1999 for (j=i+1; j<psl->sl_count; j++)
2000 psl->sl_addr[j-1] = psl->sl_addr[j];
2001 psl->sl_count--;
2002 err = 0;
2003 goto done;
2005 /* else, add a new source to the filter */
2007 if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2008 err = -ENOBUFS;
2009 goto done;
2011 if (!psl || psl->sl_count == psl->sl_max) {
2012 struct ip_sf_socklist *newpsl;
2013 int count = IP_SFBLOCK;
2015 if (psl)
2016 count += psl->sl_max;
2017 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2018 if (!newpsl) {
2019 err = -ENOBUFS;
2020 goto done;
2022 newpsl->sl_max = count;
2023 newpsl->sl_count = count - IP_SFBLOCK;
2024 if (psl) {
2025 for (i=0; i<psl->sl_count; i++)
2026 newpsl->sl_addr[i] = psl->sl_addr[i];
2027 /* decrease mem now to avoid the memleak warning */
2028 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2029 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2031 rcu_assign_pointer(pmc->sflist, newpsl);
2032 psl = newpsl;
2034 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2035 for (i=0; i<psl->sl_count; i++) {
2036 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2037 sizeof(__be32));
2038 if (rv == 0)
2039 break;
2041 if (rv == 0) /* address already there is an error */
2042 goto done;
2043 for (j=psl->sl_count-1; j>=i; j--)
2044 psl->sl_addr[j+1] = psl->sl_addr[j];
2045 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2046 psl->sl_count++;
2047 err = 0;
2048 /* update the interface list */
2049 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2050 &mreqs->imr_sourceaddr, 1);
2051 done:
2052 rtnl_unlock();
2053 if (leavegroup)
2054 return ip_mc_leave_group(sk, &imr);
2055 return err;
2058 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2060 int err = 0;
2061 struct ip_mreqn imr;
2062 __be32 addr = msf->imsf_multiaddr;
2063 struct ip_mc_socklist *pmc;
2064 struct in_device *in_dev;
2065 struct inet_sock *inet = inet_sk(sk);
2066 struct ip_sf_socklist *newpsl, *psl;
2067 struct net *net = sock_net(sk);
2068 int leavegroup = 0;
2070 if (!ipv4_is_multicast(addr))
2071 return -EINVAL;
2072 if (msf->imsf_fmode != MCAST_INCLUDE &&
2073 msf->imsf_fmode != MCAST_EXCLUDE)
2074 return -EINVAL;
2076 rtnl_lock();
2078 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2079 imr.imr_address.s_addr = msf->imsf_interface;
2080 imr.imr_ifindex = ifindex;
2081 in_dev = ip_mc_find_dev(net, &imr);
2083 if (!in_dev) {
2084 err = -ENODEV;
2085 goto done;
2088 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2089 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2090 leavegroup = 1;
2091 goto done;
2094 for_each_pmc_rtnl(inet, pmc) {
2095 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2096 pmc->multi.imr_ifindex == imr.imr_ifindex)
2097 break;
2099 if (!pmc) { /* must have a prior join */
2100 err = -EINVAL;
2101 goto done;
2103 if (msf->imsf_numsrc) {
2104 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2105 GFP_KERNEL);
2106 if (!newpsl) {
2107 err = -ENOBUFS;
2108 goto done;
2110 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2111 memcpy(newpsl->sl_addr, msf->imsf_slist,
2112 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2113 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2114 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2115 if (err) {
2116 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2117 goto done;
2119 } else {
2120 newpsl = NULL;
2121 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2122 msf->imsf_fmode, 0, NULL, 0);
2124 psl = rtnl_dereference(pmc->sflist);
2125 if (psl) {
2126 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2127 psl->sl_count, psl->sl_addr, 0);
2128 /* decrease mem now to avoid the memleak warning */
2129 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2130 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2131 } else
2132 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2133 0, NULL, 0);
2134 rcu_assign_pointer(pmc->sflist, newpsl);
2135 pmc->sfmode = msf->imsf_fmode;
2136 err = 0;
2137 done:
2138 rtnl_unlock();
2139 if (leavegroup)
2140 err = ip_mc_leave_group(sk, &imr);
2141 return err;
2144 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2145 struct ip_msfilter __user *optval, int __user *optlen)
2147 int err, len, count, copycount;
2148 struct ip_mreqn imr;
2149 __be32 addr = msf->imsf_multiaddr;
2150 struct ip_mc_socklist *pmc;
2151 struct in_device *in_dev;
2152 struct inet_sock *inet = inet_sk(sk);
2153 struct ip_sf_socklist *psl;
2154 struct net *net = sock_net(sk);
2156 if (!ipv4_is_multicast(addr))
2157 return -EINVAL;
2159 rtnl_lock();
2161 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2162 imr.imr_address.s_addr = msf->imsf_interface;
2163 imr.imr_ifindex = 0;
2164 in_dev = ip_mc_find_dev(net, &imr);
2166 if (!in_dev) {
2167 err = -ENODEV;
2168 goto done;
2170 err = -EADDRNOTAVAIL;
2172 for_each_pmc_rtnl(inet, pmc) {
2173 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2174 pmc->multi.imr_ifindex == imr.imr_ifindex)
2175 break;
2177 if (!pmc) /* must have a prior join */
2178 goto done;
2179 msf->imsf_fmode = pmc->sfmode;
2180 psl = rtnl_dereference(pmc->sflist);
2181 rtnl_unlock();
2182 if (!psl) {
2183 len = 0;
2184 count = 0;
2185 } else {
2186 count = psl->sl_count;
2188 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2189 len = copycount * sizeof(psl->sl_addr[0]);
2190 msf->imsf_numsrc = count;
2191 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2192 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2193 return -EFAULT;
2195 if (len &&
2196 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2197 return -EFAULT;
2198 return 0;
2199 done:
2200 rtnl_unlock();
2201 return err;
2204 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2205 struct group_filter __user *optval, int __user *optlen)
2207 int err, i, count, copycount;
2208 struct sockaddr_in *psin;
2209 __be32 addr;
2210 struct ip_mc_socklist *pmc;
2211 struct inet_sock *inet = inet_sk(sk);
2212 struct ip_sf_socklist *psl;
2214 psin = (struct sockaddr_in *)&gsf->gf_group;
2215 if (psin->sin_family != AF_INET)
2216 return -EINVAL;
2217 addr = psin->sin_addr.s_addr;
2218 if (!ipv4_is_multicast(addr))
2219 return -EINVAL;
2221 rtnl_lock();
2223 err = -EADDRNOTAVAIL;
2225 for_each_pmc_rtnl(inet, pmc) {
2226 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2227 pmc->multi.imr_ifindex == gsf->gf_interface)
2228 break;
2230 if (!pmc) /* must have a prior join */
2231 goto done;
2232 gsf->gf_fmode = pmc->sfmode;
2233 psl = rtnl_dereference(pmc->sflist);
2234 rtnl_unlock();
2235 count = psl ? psl->sl_count : 0;
2236 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2237 gsf->gf_numsrc = count;
2238 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2239 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2240 return -EFAULT;
2242 for (i=0; i<copycount; i++) {
2243 struct sockaddr_storage ss;
2245 psin = (struct sockaddr_in *)&ss;
2246 memset(&ss, 0, sizeof(ss));
2247 psin->sin_family = AF_INET;
2248 psin->sin_addr.s_addr = psl->sl_addr[i];
2249 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2250 return -EFAULT;
2252 return 0;
2253 done:
2254 rtnl_unlock();
2255 return err;
2259 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2261 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2263 struct inet_sock *inet = inet_sk(sk);
2264 struct ip_mc_socklist *pmc;
2265 struct ip_sf_socklist *psl;
2266 int i;
2267 int ret;
2269 ret = 1;
2270 if (!ipv4_is_multicast(loc_addr))
2271 goto out;
2273 rcu_read_lock();
2274 for_each_pmc_rcu(inet, pmc) {
2275 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2276 pmc->multi.imr_ifindex == dif)
2277 break;
2279 ret = inet->mc_all;
2280 if (!pmc)
2281 goto unlock;
2282 psl = rcu_dereference(pmc->sflist);
2283 ret = (pmc->sfmode == MCAST_EXCLUDE);
2284 if (!psl)
2285 goto unlock;
2287 for (i=0; i<psl->sl_count; i++) {
2288 if (psl->sl_addr[i] == rmt_addr)
2289 break;
2291 ret = 0;
2292 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2293 goto unlock;
2294 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2295 goto unlock;
2296 ret = 1;
2297 unlock:
2298 rcu_read_unlock();
2299 out:
2300 return ret;
2304 * A socket is closing.
2307 void ip_mc_drop_socket(struct sock *sk)
2309 struct inet_sock *inet = inet_sk(sk);
2310 struct ip_mc_socklist *iml;
2311 struct net *net = sock_net(sk);
2313 if (inet->mc_list == NULL)
2314 return;
2316 rtnl_lock();
2317 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2318 struct in_device *in_dev;
2320 inet->mc_list = iml->next_rcu;
2321 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2322 (void) ip_mc_leave_src(sk, iml, in_dev);
2323 if (in_dev != NULL)
2324 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2325 /* decrease mem now to avoid the memleak warning */
2326 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2327 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2329 rtnl_unlock();
2332 /* called with rcu_read_lock() */
2333 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2335 struct ip_mc_list *im;
2336 struct ip_sf_list *psf;
2337 int rv = 0;
2339 for_each_pmc_rcu(in_dev, im) {
2340 if (im->multiaddr == mc_addr)
2341 break;
2343 if (im && proto == IPPROTO_IGMP) {
2344 rv = 1;
2345 } else if (im) {
2346 if (src_addr) {
2347 for (psf=im->sources; psf; psf=psf->sf_next) {
2348 if (psf->sf_inaddr == src_addr)
2349 break;
2351 if (psf)
2352 rv = psf->sf_count[MCAST_INCLUDE] ||
2353 psf->sf_count[MCAST_EXCLUDE] !=
2354 im->sfcount[MCAST_EXCLUDE];
2355 else
2356 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2357 } else
2358 rv = 1; /* unspecified source; tentatively allow */
2360 return rv;
2363 #if defined(CONFIG_PROC_FS)
2364 struct igmp_mc_iter_state {
2365 struct seq_net_private p;
2366 struct net_device *dev;
2367 struct in_device *in_dev;
2370 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2372 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2374 struct net *net = seq_file_net(seq);
2375 struct ip_mc_list *im = NULL;
2376 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2378 state->in_dev = NULL;
2379 for_each_netdev_rcu(net, state->dev) {
2380 struct in_device *in_dev;
2382 in_dev = __in_dev_get_rcu(state->dev);
2383 if (!in_dev)
2384 continue;
2385 im = rcu_dereference(in_dev->mc_list);
2386 if (im) {
2387 state->in_dev = in_dev;
2388 break;
2391 return im;
2394 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2396 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2398 im = rcu_dereference(im->next_rcu);
2399 while (!im) {
2400 state->dev = next_net_device_rcu(state->dev);
2401 if (!state->dev) {
2402 state->in_dev = NULL;
2403 break;
2405 state->in_dev = __in_dev_get_rcu(state->dev);
2406 if (!state->in_dev)
2407 continue;
2408 im = rcu_dereference(state->in_dev->mc_list);
2410 return im;
2413 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2415 struct ip_mc_list *im = igmp_mc_get_first(seq);
2416 if (im)
2417 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2418 --pos;
2419 return pos ? NULL : im;
2422 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2423 __acquires(rcu)
2425 rcu_read_lock();
2426 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2429 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2431 struct ip_mc_list *im;
2432 if (v == SEQ_START_TOKEN)
2433 im = igmp_mc_get_first(seq);
2434 else
2435 im = igmp_mc_get_next(seq, v);
2436 ++*pos;
2437 return im;
2440 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2441 __releases(rcu)
2443 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2445 state->in_dev = NULL;
2446 state->dev = NULL;
2447 rcu_read_unlock();
2450 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2452 if (v == SEQ_START_TOKEN)
2453 seq_puts(seq,
2454 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2455 else {
2456 struct ip_mc_list *im = (struct ip_mc_list *)v;
2457 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2458 char *querier;
2459 #ifdef CONFIG_IP_MULTICAST
2460 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2461 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2462 "V3";
2463 #else
2464 querier = "NONE";
2465 #endif
2467 if (rcu_dereference(state->in_dev->mc_list) == im) {
2468 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2469 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2472 seq_printf(seq,
2473 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2474 im->multiaddr, im->users,
2475 im->tm_running, im->tm_running ?
2476 jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2477 im->reporter);
2479 return 0;
2482 static const struct seq_operations igmp_mc_seq_ops = {
2483 .start = igmp_mc_seq_start,
2484 .next = igmp_mc_seq_next,
2485 .stop = igmp_mc_seq_stop,
2486 .show = igmp_mc_seq_show,
2489 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2491 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2492 sizeof(struct igmp_mc_iter_state));
2495 static const struct file_operations igmp_mc_seq_fops = {
2496 .owner = THIS_MODULE,
2497 .open = igmp_mc_seq_open,
2498 .read = seq_read,
2499 .llseek = seq_lseek,
2500 .release = seq_release_net,
2503 struct igmp_mcf_iter_state {
2504 struct seq_net_private p;
2505 struct net_device *dev;
2506 struct in_device *idev;
2507 struct ip_mc_list *im;
2510 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2512 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2514 struct net *net = seq_file_net(seq);
2515 struct ip_sf_list *psf = NULL;
2516 struct ip_mc_list *im = NULL;
2517 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2519 state->idev = NULL;
2520 state->im = NULL;
2521 for_each_netdev_rcu(net, state->dev) {
2522 struct in_device *idev;
2523 idev = __in_dev_get_rcu(state->dev);
2524 if (unlikely(idev == NULL))
2525 continue;
2526 im = rcu_dereference(idev->mc_list);
2527 if (likely(im != NULL)) {
2528 spin_lock_bh(&im->lock);
2529 psf = im->sources;
2530 if (likely(psf != NULL)) {
2531 state->im = im;
2532 state->idev = idev;
2533 break;
2535 spin_unlock_bh(&im->lock);
2538 return psf;
2541 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2543 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2545 psf = psf->sf_next;
2546 while (!psf) {
2547 spin_unlock_bh(&state->im->lock);
2548 state->im = state->im->next;
2549 while (!state->im) {
2550 state->dev = next_net_device_rcu(state->dev);
2551 if (!state->dev) {
2552 state->idev = NULL;
2553 goto out;
2555 state->idev = __in_dev_get_rcu(state->dev);
2556 if (!state->idev)
2557 continue;
2558 state->im = rcu_dereference(state->idev->mc_list);
2560 if (!state->im)
2561 break;
2562 spin_lock_bh(&state->im->lock);
2563 psf = state->im->sources;
2565 out:
2566 return psf;
2569 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2571 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2572 if (psf)
2573 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2574 --pos;
2575 return pos ? NULL : psf;
2578 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2579 __acquires(rcu)
2581 rcu_read_lock();
2582 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2585 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2587 struct ip_sf_list *psf;
2588 if (v == SEQ_START_TOKEN)
2589 psf = igmp_mcf_get_first(seq);
2590 else
2591 psf = igmp_mcf_get_next(seq, v);
2592 ++*pos;
2593 return psf;
2596 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2597 __releases(rcu)
2599 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2600 if (likely(state->im != NULL)) {
2601 spin_unlock_bh(&state->im->lock);
2602 state->im = NULL;
2604 state->idev = NULL;
2605 state->dev = NULL;
2606 rcu_read_unlock();
2609 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2611 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2612 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2614 if (v == SEQ_START_TOKEN) {
2615 seq_printf(seq,
2616 "%3s %6s "
2617 "%10s %10s %6s %6s\n", "Idx",
2618 "Device", "MCA",
2619 "SRC", "INC", "EXC");
2620 } else {
2621 seq_printf(seq,
2622 "%3d %6.6s 0x%08x "
2623 "0x%08x %6lu %6lu\n",
2624 state->dev->ifindex, state->dev->name,
2625 ntohl(state->im->multiaddr),
2626 ntohl(psf->sf_inaddr),
2627 psf->sf_count[MCAST_INCLUDE],
2628 psf->sf_count[MCAST_EXCLUDE]);
2630 return 0;
2633 static const struct seq_operations igmp_mcf_seq_ops = {
2634 .start = igmp_mcf_seq_start,
2635 .next = igmp_mcf_seq_next,
2636 .stop = igmp_mcf_seq_stop,
2637 .show = igmp_mcf_seq_show,
2640 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2642 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2643 sizeof(struct igmp_mcf_iter_state));
2646 static const struct file_operations igmp_mcf_seq_fops = {
2647 .owner = THIS_MODULE,
2648 .open = igmp_mcf_seq_open,
2649 .read = seq_read,
2650 .llseek = seq_lseek,
2651 .release = seq_release_net,
2654 static int __net_init igmp_net_init(struct net *net)
2656 struct proc_dir_entry *pde;
2658 pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2659 if (!pde)
2660 goto out_igmp;
2661 pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2662 if (!pde)
2663 goto out_mcfilter;
2664 return 0;
2666 out_mcfilter:
2667 proc_net_remove(net, "igmp");
2668 out_igmp:
2669 return -ENOMEM;
2672 static void __net_exit igmp_net_exit(struct net *net)
2674 proc_net_remove(net, "mcfilter");
2675 proc_net_remove(net, "igmp");
2678 static struct pernet_operations igmp_net_ops = {
2679 .init = igmp_net_init,
2680 .exit = igmp_net_exit,
2683 int __init igmp_mc_proc_init(void)
2685 return register_pernet_subsys(&igmp_net_ops);
2687 #endif