[SCSI] libsas: replace event locks with atomic bitops
[linux-2.6.git] / net / ipv4 / igmp.c
blob450e5d21ed2af271574874479bd0c0d341bf6b38
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);
152 static void ip_ma_put(struct ip_mc_list *im)
154 if (atomic_dec_and_test(&im->refcnt)) {
155 in_dev_put(im->interface);
156 kfree_rcu(im, rcu);
160 #define for_each_pmc_rcu(in_dev, pmc) \
161 for (pmc = rcu_dereference(in_dev->mc_list); \
162 pmc != NULL; \
163 pmc = rcu_dereference(pmc->next_rcu))
165 #define for_each_pmc_rtnl(in_dev, pmc) \
166 for (pmc = rtnl_dereference(in_dev->mc_list); \
167 pmc != NULL; \
168 pmc = rtnl_dereference(pmc->next_rcu))
170 #ifdef CONFIG_IP_MULTICAST
173 * Timer management
176 static void igmp_stop_timer(struct ip_mc_list *im)
178 spin_lock_bh(&im->lock);
179 if (del_timer(&im->timer))
180 atomic_dec(&im->refcnt);
181 im->tm_running = 0;
182 im->reporter = 0;
183 im->unsolicit_count = 0;
184 spin_unlock_bh(&im->lock);
187 /* It must be called with locked im->lock */
188 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
190 int tv = net_random() % max_delay;
192 im->tm_running = 1;
193 if (!mod_timer(&im->timer, jiffies+tv+2))
194 atomic_inc(&im->refcnt);
197 static void igmp_gq_start_timer(struct in_device *in_dev)
199 int tv = net_random() % in_dev->mr_maxdelay;
201 in_dev->mr_gq_running = 1;
202 if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
203 in_dev_hold(in_dev);
206 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
208 int tv = net_random() % delay;
210 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
211 in_dev_hold(in_dev);
214 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
216 spin_lock_bh(&im->lock);
217 im->unsolicit_count = 0;
218 if (del_timer(&im->timer)) {
219 if ((long)(im->timer.expires-jiffies) < max_delay) {
220 add_timer(&im->timer);
221 im->tm_running = 1;
222 spin_unlock_bh(&im->lock);
223 return;
225 atomic_dec(&im->refcnt);
227 igmp_start_timer(im, max_delay);
228 spin_unlock_bh(&im->lock);
233 * Send an IGMP report.
236 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
239 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
240 int gdeleted, int sdeleted)
242 switch (type) {
243 case IGMPV3_MODE_IS_INCLUDE:
244 case IGMPV3_MODE_IS_EXCLUDE:
245 if (gdeleted || sdeleted)
246 return 0;
247 if (!(pmc->gsquery && !psf->sf_gsresp)) {
248 if (pmc->sfmode == MCAST_INCLUDE)
249 return 1;
250 /* don't include if this source is excluded
251 * in all filters
253 if (psf->sf_count[MCAST_INCLUDE])
254 return type == IGMPV3_MODE_IS_INCLUDE;
255 return pmc->sfcount[MCAST_EXCLUDE] ==
256 psf->sf_count[MCAST_EXCLUDE];
258 return 0;
259 case IGMPV3_CHANGE_TO_INCLUDE:
260 if (gdeleted || sdeleted)
261 return 0;
262 return psf->sf_count[MCAST_INCLUDE] != 0;
263 case IGMPV3_CHANGE_TO_EXCLUDE:
264 if (gdeleted || sdeleted)
265 return 0;
266 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
267 psf->sf_count[MCAST_INCLUDE])
268 return 0;
269 return pmc->sfcount[MCAST_EXCLUDE] ==
270 psf->sf_count[MCAST_EXCLUDE];
271 case IGMPV3_ALLOW_NEW_SOURCES:
272 if (gdeleted || !psf->sf_crcount)
273 return 0;
274 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
275 case IGMPV3_BLOCK_OLD_SOURCES:
276 if (pmc->sfmode == MCAST_INCLUDE)
277 return gdeleted || (psf->sf_crcount && sdeleted);
278 return psf->sf_crcount && !gdeleted && !sdeleted;
280 return 0;
283 static int
284 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
286 struct ip_sf_list *psf;
287 int scount = 0;
289 for (psf=pmc->sources; psf; psf=psf->sf_next) {
290 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
291 continue;
292 scount++;
294 return scount;
297 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
299 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
301 struct sk_buff *skb;
302 struct rtable *rt;
303 struct iphdr *pip;
304 struct igmpv3_report *pig;
305 struct net *net = dev_net(dev);
306 struct flowi4 fl4;
307 int hlen = LL_RESERVED_SPACE(dev);
308 int tlen = dev->needed_tailroom;
310 while (1) {
311 skb = alloc_skb(size + hlen + tlen,
312 GFP_ATOMIC | __GFP_NOWARN);
313 if (skb)
314 break;
315 size >>= 1;
316 if (size < 256)
317 return NULL;
319 igmp_skb_size(skb) = size;
321 rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
322 0, 0,
323 IPPROTO_IGMP, 0, dev->ifindex);
324 if (IS_ERR(rt)) {
325 kfree_skb(skb);
326 return NULL;
329 skb_dst_set(skb, &rt->dst);
330 skb->dev = dev;
332 skb_reserve(skb, hlen);
334 skb_reset_network_header(skb);
335 pip = ip_hdr(skb);
336 skb_put(skb, sizeof(struct iphdr) + 4);
338 pip->version = 4;
339 pip->ihl = (sizeof(struct iphdr)+4)>>2;
340 pip->tos = 0xc0;
341 pip->frag_off = htons(IP_DF);
342 pip->ttl = 1;
343 pip->daddr = fl4.daddr;
344 pip->saddr = fl4.saddr;
345 pip->protocol = IPPROTO_IGMP;
346 pip->tot_len = 0; /* filled in later */
347 ip_select_ident(pip, &rt->dst, NULL);
348 ((u8*)&pip[1])[0] = IPOPT_RA;
349 ((u8*)&pip[1])[1] = 4;
350 ((u8*)&pip[1])[2] = 0;
351 ((u8*)&pip[1])[3] = 0;
353 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
354 skb_put(skb, sizeof(*pig));
355 pig = igmpv3_report_hdr(skb);
356 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
357 pig->resv1 = 0;
358 pig->csum = 0;
359 pig->resv2 = 0;
360 pig->ngrec = 0;
361 return skb;
364 static int igmpv3_sendpack(struct sk_buff *skb)
366 struct igmphdr *pig = igmp_hdr(skb);
367 const int igmplen = skb->tail - skb->transport_header;
369 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
371 return ip_local_out(skb);
374 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
376 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
379 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
380 int type, struct igmpv3_grec **ppgr)
382 struct net_device *dev = pmc->interface->dev;
383 struct igmpv3_report *pih;
384 struct igmpv3_grec *pgr;
386 if (!skb)
387 skb = igmpv3_newpack(dev, dev->mtu);
388 if (!skb)
389 return NULL;
390 pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
391 pgr->grec_type = type;
392 pgr->grec_auxwords = 0;
393 pgr->grec_nsrcs = 0;
394 pgr->grec_mca = pmc->multiaddr;
395 pih = igmpv3_report_hdr(skb);
396 pih->ngrec = htons(ntohs(pih->ngrec)+1);
397 *ppgr = pgr;
398 return skb;
401 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
402 skb_tailroom(skb)) : 0)
404 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
405 int type, int gdeleted, int sdeleted)
407 struct net_device *dev = pmc->interface->dev;
408 struct igmpv3_report *pih;
409 struct igmpv3_grec *pgr = NULL;
410 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
411 int scount, stotal, first, isquery, truncate;
413 if (pmc->multiaddr == IGMP_ALL_HOSTS)
414 return skb;
416 isquery = type == IGMPV3_MODE_IS_INCLUDE ||
417 type == IGMPV3_MODE_IS_EXCLUDE;
418 truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
419 type == IGMPV3_CHANGE_TO_EXCLUDE;
421 stotal = scount = 0;
423 psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
425 if (!*psf_list)
426 goto empty_source;
428 pih = skb ? igmpv3_report_hdr(skb) : NULL;
430 /* EX and TO_EX get a fresh packet, if needed */
431 if (truncate) {
432 if (pih && pih->ngrec &&
433 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
434 if (skb)
435 igmpv3_sendpack(skb);
436 skb = igmpv3_newpack(dev, dev->mtu);
439 first = 1;
440 psf_prev = NULL;
441 for (psf=*psf_list; psf; psf=psf_next) {
442 __be32 *psrc;
444 psf_next = psf->sf_next;
446 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
447 psf_prev = psf;
448 continue;
451 /* clear marks on query responses */
452 if (isquery)
453 psf->sf_gsresp = 0;
455 if (AVAILABLE(skb) < sizeof(__be32) +
456 first*sizeof(struct igmpv3_grec)) {
457 if (truncate && !first)
458 break; /* truncate these */
459 if (pgr)
460 pgr->grec_nsrcs = htons(scount);
461 if (skb)
462 igmpv3_sendpack(skb);
463 skb = igmpv3_newpack(dev, dev->mtu);
464 first = 1;
465 scount = 0;
467 if (first) {
468 skb = add_grhead(skb, pmc, type, &pgr);
469 first = 0;
471 if (!skb)
472 return NULL;
473 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
474 *psrc = psf->sf_inaddr;
475 scount++; stotal++;
476 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
477 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
478 psf->sf_crcount--;
479 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
480 if (psf_prev)
481 psf_prev->sf_next = psf->sf_next;
482 else
483 *psf_list = psf->sf_next;
484 kfree(psf);
485 continue;
488 psf_prev = psf;
491 empty_source:
492 if (!stotal) {
493 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
494 type == IGMPV3_BLOCK_OLD_SOURCES)
495 return skb;
496 if (pmc->crcount || isquery) {
497 /* make sure we have room for group header */
498 if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
499 igmpv3_sendpack(skb);
500 skb = NULL; /* add_grhead will get a new one */
502 skb = add_grhead(skb, pmc, type, &pgr);
505 if (pgr)
506 pgr->grec_nsrcs = htons(scount);
508 if (isquery)
509 pmc->gsquery = 0; /* clear query state on report */
510 return skb;
513 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
515 struct sk_buff *skb = NULL;
516 int type;
518 if (!pmc) {
519 rcu_read_lock();
520 for_each_pmc_rcu(in_dev, pmc) {
521 if (pmc->multiaddr == IGMP_ALL_HOSTS)
522 continue;
523 spin_lock_bh(&pmc->lock);
524 if (pmc->sfcount[MCAST_EXCLUDE])
525 type = IGMPV3_MODE_IS_EXCLUDE;
526 else
527 type = IGMPV3_MODE_IS_INCLUDE;
528 skb = add_grec(skb, pmc, type, 0, 0);
529 spin_unlock_bh(&pmc->lock);
531 rcu_read_unlock();
532 } else {
533 spin_lock_bh(&pmc->lock);
534 if (pmc->sfcount[MCAST_EXCLUDE])
535 type = IGMPV3_MODE_IS_EXCLUDE;
536 else
537 type = IGMPV3_MODE_IS_INCLUDE;
538 skb = add_grec(skb, pmc, type, 0, 0);
539 spin_unlock_bh(&pmc->lock);
541 if (!skb)
542 return 0;
543 return igmpv3_sendpack(skb);
547 * remove zero-count source records from a source filter list
549 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
551 struct ip_sf_list *psf_prev, *psf_next, *psf;
553 psf_prev = NULL;
554 for (psf=*ppsf; psf; psf = psf_next) {
555 psf_next = psf->sf_next;
556 if (psf->sf_crcount == 0) {
557 if (psf_prev)
558 psf_prev->sf_next = psf->sf_next;
559 else
560 *ppsf = psf->sf_next;
561 kfree(psf);
562 } else
563 psf_prev = psf;
567 static void igmpv3_send_cr(struct in_device *in_dev)
569 struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
570 struct sk_buff *skb = NULL;
571 int type, dtype;
573 rcu_read_lock();
574 spin_lock_bh(&in_dev->mc_tomb_lock);
576 /* deleted MCA's */
577 pmc_prev = NULL;
578 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
579 pmc_next = pmc->next;
580 if (pmc->sfmode == MCAST_INCLUDE) {
581 type = IGMPV3_BLOCK_OLD_SOURCES;
582 dtype = IGMPV3_BLOCK_OLD_SOURCES;
583 skb = add_grec(skb, pmc, type, 1, 0);
584 skb = add_grec(skb, pmc, dtype, 1, 1);
586 if (pmc->crcount) {
587 if (pmc->sfmode == MCAST_EXCLUDE) {
588 type = IGMPV3_CHANGE_TO_INCLUDE;
589 skb = add_grec(skb, pmc, type, 1, 0);
591 pmc->crcount--;
592 if (pmc->crcount == 0) {
593 igmpv3_clear_zeros(&pmc->tomb);
594 igmpv3_clear_zeros(&pmc->sources);
597 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
598 if (pmc_prev)
599 pmc_prev->next = pmc_next;
600 else
601 in_dev->mc_tomb = pmc_next;
602 in_dev_put(pmc->interface);
603 kfree(pmc);
604 } else
605 pmc_prev = pmc;
607 spin_unlock_bh(&in_dev->mc_tomb_lock);
609 /* change recs */
610 for_each_pmc_rcu(in_dev, pmc) {
611 spin_lock_bh(&pmc->lock);
612 if (pmc->sfcount[MCAST_EXCLUDE]) {
613 type = IGMPV3_BLOCK_OLD_SOURCES;
614 dtype = IGMPV3_ALLOW_NEW_SOURCES;
615 } else {
616 type = IGMPV3_ALLOW_NEW_SOURCES;
617 dtype = IGMPV3_BLOCK_OLD_SOURCES;
619 skb = add_grec(skb, pmc, type, 0, 0);
620 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */
622 /* filter mode changes */
623 if (pmc->crcount) {
624 if (pmc->sfmode == MCAST_EXCLUDE)
625 type = IGMPV3_CHANGE_TO_EXCLUDE;
626 else
627 type = IGMPV3_CHANGE_TO_INCLUDE;
628 skb = add_grec(skb, pmc, type, 0, 0);
629 pmc->crcount--;
631 spin_unlock_bh(&pmc->lock);
633 rcu_read_unlock();
635 if (!skb)
636 return;
637 (void) igmpv3_sendpack(skb);
640 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
641 int type)
643 struct sk_buff *skb;
644 struct iphdr *iph;
645 struct igmphdr *ih;
646 struct rtable *rt;
647 struct net_device *dev = in_dev->dev;
648 struct net *net = dev_net(dev);
649 __be32 group = pmc ? pmc->multiaddr : 0;
650 struct flowi4 fl4;
651 __be32 dst;
652 int hlen, tlen;
654 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
655 return igmpv3_send_report(in_dev, pmc);
656 else if (type == IGMP_HOST_LEAVE_MESSAGE)
657 dst = IGMP_ALL_ROUTER;
658 else
659 dst = group;
661 rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
662 0, 0,
663 IPPROTO_IGMP, 0, dev->ifindex);
664 if (IS_ERR(rt))
665 return -1;
667 hlen = LL_RESERVED_SPACE(dev);
668 tlen = dev->needed_tailroom;
669 skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
670 if (skb == NULL) {
671 ip_rt_put(rt);
672 return -1;
675 skb_dst_set(skb, &rt->dst);
677 skb_reserve(skb, hlen);
679 skb_reset_network_header(skb);
680 iph = ip_hdr(skb);
681 skb_put(skb, sizeof(struct iphdr) + 4);
683 iph->version = 4;
684 iph->ihl = (sizeof(struct iphdr)+4)>>2;
685 iph->tos = 0xc0;
686 iph->frag_off = htons(IP_DF);
687 iph->ttl = 1;
688 iph->daddr = dst;
689 iph->saddr = fl4.saddr;
690 iph->protocol = IPPROTO_IGMP;
691 ip_select_ident(iph, &rt->dst, NULL);
692 ((u8*)&iph[1])[0] = IPOPT_RA;
693 ((u8*)&iph[1])[1] = 4;
694 ((u8*)&iph[1])[2] = 0;
695 ((u8*)&iph[1])[3] = 0;
697 ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
698 ih->type = type;
699 ih->code = 0;
700 ih->csum = 0;
701 ih->group = group;
702 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
704 return ip_local_out(skb);
707 static void igmp_gq_timer_expire(unsigned long data)
709 struct in_device *in_dev = (struct in_device *)data;
711 in_dev->mr_gq_running = 0;
712 igmpv3_send_report(in_dev, NULL);
713 __in_dev_put(in_dev);
716 static void igmp_ifc_timer_expire(unsigned long data)
718 struct in_device *in_dev = (struct in_device *)data;
720 igmpv3_send_cr(in_dev);
721 if (in_dev->mr_ifc_count) {
722 in_dev->mr_ifc_count--;
723 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
725 __in_dev_put(in_dev);
728 static void igmp_ifc_event(struct in_device *in_dev)
730 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
731 return;
732 in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
733 IGMP_Unsolicited_Report_Count;
734 igmp_ifc_start_timer(in_dev, 1);
738 static void igmp_timer_expire(unsigned long data)
740 struct ip_mc_list *im=(struct ip_mc_list *)data;
741 struct in_device *in_dev = im->interface;
743 spin_lock(&im->lock);
744 im->tm_running = 0;
746 if (im->unsolicit_count) {
747 im->unsolicit_count--;
748 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
750 im->reporter = 1;
751 spin_unlock(&im->lock);
753 if (IGMP_V1_SEEN(in_dev))
754 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
755 else if (IGMP_V2_SEEN(in_dev))
756 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
757 else
758 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
760 ip_ma_put(im);
763 /* mark EXCLUDE-mode sources */
764 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
766 struct ip_sf_list *psf;
767 int i, scount;
769 scount = 0;
770 for (psf=pmc->sources; psf; psf=psf->sf_next) {
771 if (scount == nsrcs)
772 break;
773 for (i=0; i<nsrcs; i++) {
774 /* skip inactive filters */
775 if (psf->sf_count[MCAST_INCLUDE] ||
776 pmc->sfcount[MCAST_EXCLUDE] !=
777 psf->sf_count[MCAST_EXCLUDE])
778 continue;
779 if (srcs[i] == psf->sf_inaddr) {
780 scount++;
781 break;
785 pmc->gsquery = 0;
786 if (scount == nsrcs) /* all sources excluded */
787 return 0;
788 return 1;
791 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
793 struct ip_sf_list *psf;
794 int i, scount;
796 if (pmc->sfmode == MCAST_EXCLUDE)
797 return igmp_xmarksources(pmc, nsrcs, srcs);
799 /* mark INCLUDE-mode sources */
800 scount = 0;
801 for (psf=pmc->sources; psf; psf=psf->sf_next) {
802 if (scount == nsrcs)
803 break;
804 for (i=0; i<nsrcs; i++)
805 if (srcs[i] == psf->sf_inaddr) {
806 psf->sf_gsresp = 1;
807 scount++;
808 break;
811 if (!scount) {
812 pmc->gsquery = 0;
813 return 0;
815 pmc->gsquery = 1;
816 return 1;
819 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
821 struct ip_mc_list *im;
823 /* Timers are only set for non-local groups */
825 if (group == IGMP_ALL_HOSTS)
826 return;
828 rcu_read_lock();
829 for_each_pmc_rcu(in_dev, im) {
830 if (im->multiaddr == group) {
831 igmp_stop_timer(im);
832 break;
835 rcu_read_unlock();
838 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
839 int len)
841 struct igmphdr *ih = igmp_hdr(skb);
842 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
843 struct ip_mc_list *im;
844 __be32 group = ih->group;
845 int max_delay;
846 int mark = 0;
849 if (len == 8) {
850 if (ih->code == 0) {
851 /* Alas, old v1 router presents here. */
853 max_delay = IGMP_Query_Response_Interval;
854 in_dev->mr_v1_seen = jiffies +
855 IGMP_V1_Router_Present_Timeout;
856 group = 0;
857 } else {
858 /* v2 router present */
859 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
860 in_dev->mr_v2_seen = jiffies +
861 IGMP_V2_Router_Present_Timeout;
863 /* cancel the interface change timer */
864 in_dev->mr_ifc_count = 0;
865 if (del_timer(&in_dev->mr_ifc_timer))
866 __in_dev_put(in_dev);
867 /* clear deleted report items */
868 igmpv3_clear_delrec(in_dev);
869 } else if (len < 12) {
870 return; /* ignore bogus packet; freed by caller */
871 } else if (IGMP_V1_SEEN(in_dev)) {
872 /* This is a v3 query with v1 queriers present */
873 max_delay = IGMP_Query_Response_Interval;
874 group = 0;
875 } else if (IGMP_V2_SEEN(in_dev)) {
876 /* this is a v3 query with v2 queriers present;
877 * Interpretation of the max_delay code is problematic here.
878 * A real v2 host would use ih_code directly, while v3 has a
879 * different encoding. We use the v3 encoding as more likely
880 * to be intended in a v3 query.
882 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
883 if (!max_delay)
884 max_delay = 1; /* can't mod w/ 0 */
885 } else { /* v3 */
886 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
887 return;
889 ih3 = igmpv3_query_hdr(skb);
890 if (ih3->nsrcs) {
891 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
892 + ntohs(ih3->nsrcs)*sizeof(__be32)))
893 return;
894 ih3 = igmpv3_query_hdr(skb);
897 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
898 if (!max_delay)
899 max_delay = 1; /* can't mod w/ 0 */
900 in_dev->mr_maxdelay = max_delay;
901 if (ih3->qrv)
902 in_dev->mr_qrv = ih3->qrv;
903 if (!group) { /* general query */
904 if (ih3->nsrcs)
905 return; /* no sources allowed */
906 igmp_gq_start_timer(in_dev);
907 return;
909 /* mark sources to include, if group & source-specific */
910 mark = ih3->nsrcs != 0;
914 * - Start the timers in all of our membership records
915 * that the query applies to for the interface on
916 * which the query arrived excl. those that belong
917 * to a "local" group (224.0.0.X)
918 * - For timers already running check if they need to
919 * be reset.
920 * - Use the igmp->igmp_code field as the maximum
921 * delay possible
923 rcu_read_lock();
924 for_each_pmc_rcu(in_dev, im) {
925 int changed;
927 if (group && group != im->multiaddr)
928 continue;
929 if (im->multiaddr == IGMP_ALL_HOSTS)
930 continue;
931 spin_lock_bh(&im->lock);
932 if (im->tm_running)
933 im->gsquery = im->gsquery && mark;
934 else
935 im->gsquery = mark;
936 changed = !im->gsquery ||
937 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
938 spin_unlock_bh(&im->lock);
939 if (changed)
940 igmp_mod_timer(im, max_delay);
942 rcu_read_unlock();
945 /* called in rcu_read_lock() section */
946 int igmp_rcv(struct sk_buff *skb)
948 /* This basically follows the spec line by line -- see RFC1112 */
949 struct igmphdr *ih;
950 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
951 int len = skb->len;
953 if (in_dev == NULL)
954 goto drop;
956 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
957 goto drop;
959 switch (skb->ip_summed) {
960 case CHECKSUM_COMPLETE:
961 if (!csum_fold(skb->csum))
962 break;
963 /* fall through */
964 case CHECKSUM_NONE:
965 skb->csum = 0;
966 if (__skb_checksum_complete(skb))
967 goto drop;
970 ih = igmp_hdr(skb);
971 switch (ih->type) {
972 case IGMP_HOST_MEMBERSHIP_QUERY:
973 igmp_heard_query(in_dev, skb, len);
974 break;
975 case IGMP_HOST_MEMBERSHIP_REPORT:
976 case IGMPV2_HOST_MEMBERSHIP_REPORT:
977 /* Is it our report looped back? */
978 if (rt_is_output_route(skb_rtable(skb)))
979 break;
980 /* don't rely on MC router hearing unicast reports */
981 if (skb->pkt_type == PACKET_MULTICAST ||
982 skb->pkt_type == PACKET_BROADCAST)
983 igmp_heard_report(in_dev, ih->group);
984 break;
985 case IGMP_PIM:
986 #ifdef CONFIG_IP_PIMSM_V1
987 return pim_rcv_v1(skb);
988 #endif
989 case IGMPV3_HOST_MEMBERSHIP_REPORT:
990 case IGMP_DVMRP:
991 case IGMP_TRACE:
992 case IGMP_HOST_LEAVE_MESSAGE:
993 case IGMP_MTRACE:
994 case IGMP_MTRACE_RESP:
995 break;
996 default:
997 break;
1000 drop:
1001 kfree_skb(skb);
1002 return 0;
1005 #endif
1009 * Add a filter to a device
1012 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1014 char buf[MAX_ADDR_LEN];
1015 struct net_device *dev = in_dev->dev;
1017 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1018 We will get multicast token leakage, when IFF_MULTICAST
1019 is changed. This check should be done in ndo_set_rx_mode
1020 routine. Something sort of:
1021 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1022 --ANK
1024 if (arp_mc_map(addr, buf, dev, 0) == 0)
1025 dev_mc_add(dev, buf);
1029 * Remove a filter from a device
1032 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1034 char buf[MAX_ADDR_LEN];
1035 struct net_device *dev = in_dev->dev;
1037 if (arp_mc_map(addr, buf, dev, 0) == 0)
1038 dev_mc_del(dev, buf);
1041 #ifdef CONFIG_IP_MULTICAST
1043 * deleted ip_mc_list manipulation
1045 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1047 struct ip_mc_list *pmc;
1049 /* this is an "ip_mc_list" for convenience; only the fields below
1050 * are actually used. In particular, the refcnt and users are not
1051 * used for management of the delete list. Using the same structure
1052 * for deleted items allows change reports to use common code with
1053 * non-deleted or query-response MCA's.
1055 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1056 if (!pmc)
1057 return;
1058 spin_lock_bh(&im->lock);
1059 pmc->interface = im->interface;
1060 in_dev_hold(in_dev);
1061 pmc->multiaddr = im->multiaddr;
1062 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1063 IGMP_Unsolicited_Report_Count;
1064 pmc->sfmode = im->sfmode;
1065 if (pmc->sfmode == MCAST_INCLUDE) {
1066 struct ip_sf_list *psf;
1068 pmc->tomb = im->tomb;
1069 pmc->sources = im->sources;
1070 im->tomb = im->sources = NULL;
1071 for (psf=pmc->sources; psf; psf=psf->sf_next)
1072 psf->sf_crcount = pmc->crcount;
1074 spin_unlock_bh(&im->lock);
1076 spin_lock_bh(&in_dev->mc_tomb_lock);
1077 pmc->next = in_dev->mc_tomb;
1078 in_dev->mc_tomb = pmc;
1079 spin_unlock_bh(&in_dev->mc_tomb_lock);
1082 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1084 struct ip_mc_list *pmc, *pmc_prev;
1085 struct ip_sf_list *psf, *psf_next;
1087 spin_lock_bh(&in_dev->mc_tomb_lock);
1088 pmc_prev = NULL;
1089 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1090 if (pmc->multiaddr == multiaddr)
1091 break;
1092 pmc_prev = pmc;
1094 if (pmc) {
1095 if (pmc_prev)
1096 pmc_prev->next = pmc->next;
1097 else
1098 in_dev->mc_tomb = pmc->next;
1100 spin_unlock_bh(&in_dev->mc_tomb_lock);
1101 if (pmc) {
1102 for (psf=pmc->tomb; psf; psf=psf_next) {
1103 psf_next = psf->sf_next;
1104 kfree(psf);
1106 in_dev_put(pmc->interface);
1107 kfree(pmc);
1111 static void igmpv3_clear_delrec(struct in_device *in_dev)
1113 struct ip_mc_list *pmc, *nextpmc;
1115 spin_lock_bh(&in_dev->mc_tomb_lock);
1116 pmc = in_dev->mc_tomb;
1117 in_dev->mc_tomb = NULL;
1118 spin_unlock_bh(&in_dev->mc_tomb_lock);
1120 for (; pmc; pmc = nextpmc) {
1121 nextpmc = pmc->next;
1122 ip_mc_clear_src(pmc);
1123 in_dev_put(pmc->interface);
1124 kfree(pmc);
1126 /* clear dead sources, too */
1127 rcu_read_lock();
1128 for_each_pmc_rcu(in_dev, pmc) {
1129 struct ip_sf_list *psf, *psf_next;
1131 spin_lock_bh(&pmc->lock);
1132 psf = pmc->tomb;
1133 pmc->tomb = NULL;
1134 spin_unlock_bh(&pmc->lock);
1135 for (; psf; psf=psf_next) {
1136 psf_next = psf->sf_next;
1137 kfree(psf);
1140 rcu_read_unlock();
1142 #endif
1144 static void igmp_group_dropped(struct ip_mc_list *im)
1146 struct in_device *in_dev = im->interface;
1147 #ifdef CONFIG_IP_MULTICAST
1148 int reporter;
1149 #endif
1151 if (im->loaded) {
1152 im->loaded = 0;
1153 ip_mc_filter_del(in_dev, im->multiaddr);
1156 #ifdef CONFIG_IP_MULTICAST
1157 if (im->multiaddr == IGMP_ALL_HOSTS)
1158 return;
1160 reporter = im->reporter;
1161 igmp_stop_timer(im);
1163 if (!in_dev->dead) {
1164 if (IGMP_V1_SEEN(in_dev))
1165 return;
1166 if (IGMP_V2_SEEN(in_dev)) {
1167 if (reporter)
1168 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1169 return;
1171 /* IGMPv3 */
1172 igmpv3_add_delrec(in_dev, im);
1174 igmp_ifc_event(in_dev);
1176 #endif
1179 static void igmp_group_added(struct ip_mc_list *im)
1181 struct in_device *in_dev = im->interface;
1183 if (im->loaded == 0) {
1184 im->loaded = 1;
1185 ip_mc_filter_add(in_dev, im->multiaddr);
1188 #ifdef CONFIG_IP_MULTICAST
1189 if (im->multiaddr == IGMP_ALL_HOSTS)
1190 return;
1192 if (in_dev->dead)
1193 return;
1194 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1195 spin_lock_bh(&im->lock);
1196 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1197 spin_unlock_bh(&im->lock);
1198 return;
1200 /* else, v3 */
1202 im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1203 IGMP_Unsolicited_Report_Count;
1204 igmp_ifc_event(in_dev);
1205 #endif
1210 * Multicast list managers
1215 * A socket has joined a multicast group on device dev.
1218 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1220 struct ip_mc_list *im;
1222 ASSERT_RTNL();
1224 for_each_pmc_rtnl(in_dev, im) {
1225 if (im->multiaddr == addr) {
1226 im->users++;
1227 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1228 goto out;
1232 im = kzalloc(sizeof(*im), GFP_KERNEL);
1233 if (!im)
1234 goto out;
1236 im->users = 1;
1237 im->interface = in_dev;
1238 in_dev_hold(in_dev);
1239 im->multiaddr = addr;
1240 /* initial mode is (EX, empty) */
1241 im->sfmode = MCAST_EXCLUDE;
1242 im->sfcount[MCAST_EXCLUDE] = 1;
1243 atomic_set(&im->refcnt, 1);
1244 spin_lock_init(&im->lock);
1245 #ifdef CONFIG_IP_MULTICAST
1246 setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1247 im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1248 #endif
1250 im->next_rcu = in_dev->mc_list;
1251 in_dev->mc_count++;
1252 rcu_assign_pointer(in_dev->mc_list, im);
1254 #ifdef CONFIG_IP_MULTICAST
1255 igmpv3_del_delrec(in_dev, im->multiaddr);
1256 #endif
1257 igmp_group_added(im);
1258 if (!in_dev->dead)
1259 ip_rt_multicast_event(in_dev);
1260 out:
1261 return;
1263 EXPORT_SYMBOL(ip_mc_inc_group);
1266 * Resend IGMP JOIN report; used for bonding.
1267 * Called with rcu_read_lock()
1269 void ip_mc_rejoin_groups(struct in_device *in_dev)
1271 #ifdef CONFIG_IP_MULTICAST
1272 struct ip_mc_list *im;
1273 int type;
1275 for_each_pmc_rcu(in_dev, im) {
1276 if (im->multiaddr == IGMP_ALL_HOSTS)
1277 continue;
1279 /* a failover is happening and switches
1280 * must be notified immediately
1282 if (IGMP_V1_SEEN(in_dev))
1283 type = IGMP_HOST_MEMBERSHIP_REPORT;
1284 else if (IGMP_V2_SEEN(in_dev))
1285 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1286 else
1287 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1288 igmp_send_report(in_dev, im, type);
1290 #endif
1292 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1295 * A socket has left a multicast group on device dev
1298 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1300 struct ip_mc_list *i;
1301 struct ip_mc_list __rcu **ip;
1303 ASSERT_RTNL();
1305 for (ip = &in_dev->mc_list;
1306 (i = rtnl_dereference(*ip)) != NULL;
1307 ip = &i->next_rcu) {
1308 if (i->multiaddr == addr) {
1309 if (--i->users == 0) {
1310 *ip = i->next_rcu;
1311 in_dev->mc_count--;
1312 igmp_group_dropped(i);
1313 ip_mc_clear_src(i);
1315 if (!in_dev->dead)
1316 ip_rt_multicast_event(in_dev);
1318 ip_ma_put(i);
1319 return;
1321 break;
1325 EXPORT_SYMBOL(ip_mc_dec_group);
1327 /* Device changing type */
1329 void ip_mc_unmap(struct in_device *in_dev)
1331 struct ip_mc_list *pmc;
1333 ASSERT_RTNL();
1335 for_each_pmc_rtnl(in_dev, pmc)
1336 igmp_group_dropped(pmc);
1339 void ip_mc_remap(struct in_device *in_dev)
1341 struct ip_mc_list *pmc;
1343 ASSERT_RTNL();
1345 for_each_pmc_rtnl(in_dev, pmc)
1346 igmp_group_added(pmc);
1349 /* Device going down */
1351 void ip_mc_down(struct in_device *in_dev)
1353 struct ip_mc_list *pmc;
1355 ASSERT_RTNL();
1357 for_each_pmc_rtnl(in_dev, pmc)
1358 igmp_group_dropped(pmc);
1360 #ifdef CONFIG_IP_MULTICAST
1361 in_dev->mr_ifc_count = 0;
1362 if (del_timer(&in_dev->mr_ifc_timer))
1363 __in_dev_put(in_dev);
1364 in_dev->mr_gq_running = 0;
1365 if (del_timer(&in_dev->mr_gq_timer))
1366 __in_dev_put(in_dev);
1367 igmpv3_clear_delrec(in_dev);
1368 #endif
1370 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1373 void ip_mc_init_dev(struct in_device *in_dev)
1375 ASSERT_RTNL();
1377 in_dev->mc_tomb = NULL;
1378 #ifdef CONFIG_IP_MULTICAST
1379 in_dev->mr_gq_running = 0;
1380 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1381 (unsigned long)in_dev);
1382 in_dev->mr_ifc_count = 0;
1383 in_dev->mc_count = 0;
1384 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1385 (unsigned long)in_dev);
1386 in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1387 #endif
1389 spin_lock_init(&in_dev->mc_tomb_lock);
1392 /* Device going up */
1394 void ip_mc_up(struct in_device *in_dev)
1396 struct ip_mc_list *pmc;
1398 ASSERT_RTNL();
1400 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1402 for_each_pmc_rtnl(in_dev, pmc)
1403 igmp_group_added(pmc);
1407 * Device is about to be destroyed: clean up.
1410 void ip_mc_destroy_dev(struct in_device *in_dev)
1412 struct ip_mc_list *i;
1414 ASSERT_RTNL();
1416 /* Deactivate timers */
1417 ip_mc_down(in_dev);
1419 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1420 in_dev->mc_list = i->next_rcu;
1421 in_dev->mc_count--;
1423 /* We've dropped the groups in ip_mc_down already */
1424 ip_mc_clear_src(i);
1425 ip_ma_put(i);
1429 /* RTNL is locked */
1430 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1432 struct net_device *dev = NULL;
1433 struct in_device *idev = NULL;
1435 if (imr->imr_ifindex) {
1436 idev = inetdev_by_index(net, imr->imr_ifindex);
1437 return idev;
1439 if (imr->imr_address.s_addr) {
1440 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1441 if (!dev)
1442 return NULL;
1445 if (!dev) {
1446 struct rtable *rt = ip_route_output(net,
1447 imr->imr_multiaddr.s_addr,
1448 0, 0, 0);
1449 if (!IS_ERR(rt)) {
1450 dev = rt->dst.dev;
1451 ip_rt_put(rt);
1454 if (dev) {
1455 imr->imr_ifindex = dev->ifindex;
1456 idev = __in_dev_get_rtnl(dev);
1458 return idev;
1462 * Join a socket to a group
1464 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1465 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1468 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1469 __be32 *psfsrc)
1471 struct ip_sf_list *psf, *psf_prev;
1472 int rv = 0;
1474 psf_prev = NULL;
1475 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1476 if (psf->sf_inaddr == *psfsrc)
1477 break;
1478 psf_prev = psf;
1480 if (!psf || psf->sf_count[sfmode] == 0) {
1481 /* source filter not found, or count wrong => bug */
1482 return -ESRCH;
1484 psf->sf_count[sfmode]--;
1485 if (psf->sf_count[sfmode] == 0) {
1486 ip_rt_multicast_event(pmc->interface);
1488 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1489 #ifdef CONFIG_IP_MULTICAST
1490 struct in_device *in_dev = pmc->interface;
1491 #endif
1493 /* no more filters for this source */
1494 if (psf_prev)
1495 psf_prev->sf_next = psf->sf_next;
1496 else
1497 pmc->sources = psf->sf_next;
1498 #ifdef CONFIG_IP_MULTICAST
1499 if (psf->sf_oldin &&
1500 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1501 psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1502 IGMP_Unsolicited_Report_Count;
1503 psf->sf_next = pmc->tomb;
1504 pmc->tomb = psf;
1505 rv = 1;
1506 } else
1507 #endif
1508 kfree(psf);
1510 return rv;
1513 #ifndef CONFIG_IP_MULTICAST
1514 #define igmp_ifc_event(x) do { } while (0)
1515 #endif
1517 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1518 int sfcount, __be32 *psfsrc, int delta)
1520 struct ip_mc_list *pmc;
1521 int changerec = 0;
1522 int i, err;
1524 if (!in_dev)
1525 return -ENODEV;
1526 rcu_read_lock();
1527 for_each_pmc_rcu(in_dev, pmc) {
1528 if (*pmca == pmc->multiaddr)
1529 break;
1531 if (!pmc) {
1532 /* MCA not found?? bug */
1533 rcu_read_unlock();
1534 return -ESRCH;
1536 spin_lock_bh(&pmc->lock);
1537 rcu_read_unlock();
1538 #ifdef CONFIG_IP_MULTICAST
1539 sf_markstate(pmc);
1540 #endif
1541 if (!delta) {
1542 err = -EINVAL;
1543 if (!pmc->sfcount[sfmode])
1544 goto out_unlock;
1545 pmc->sfcount[sfmode]--;
1547 err = 0;
1548 for (i=0; i<sfcount; i++) {
1549 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1551 changerec |= rv > 0;
1552 if (!err && rv < 0)
1553 err = rv;
1555 if (pmc->sfmode == MCAST_EXCLUDE &&
1556 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1557 pmc->sfcount[MCAST_INCLUDE]) {
1558 #ifdef CONFIG_IP_MULTICAST
1559 struct ip_sf_list *psf;
1560 #endif
1562 /* filter mode change */
1563 pmc->sfmode = MCAST_INCLUDE;
1564 #ifdef CONFIG_IP_MULTICAST
1565 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1566 IGMP_Unsolicited_Report_Count;
1567 in_dev->mr_ifc_count = pmc->crcount;
1568 for (psf=pmc->sources; psf; psf = psf->sf_next)
1569 psf->sf_crcount = 0;
1570 igmp_ifc_event(pmc->interface);
1571 } else if (sf_setstate(pmc) || changerec) {
1572 igmp_ifc_event(pmc->interface);
1573 #endif
1575 out_unlock:
1576 spin_unlock_bh(&pmc->lock);
1577 return err;
1581 * Add multicast single-source filter to the interface list
1583 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1584 __be32 *psfsrc)
1586 struct ip_sf_list *psf, *psf_prev;
1588 psf_prev = NULL;
1589 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1590 if (psf->sf_inaddr == *psfsrc)
1591 break;
1592 psf_prev = psf;
1594 if (!psf) {
1595 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1596 if (!psf)
1597 return -ENOBUFS;
1598 psf->sf_inaddr = *psfsrc;
1599 if (psf_prev) {
1600 psf_prev->sf_next = psf;
1601 } else
1602 pmc->sources = psf;
1604 psf->sf_count[sfmode]++;
1605 if (psf->sf_count[sfmode] == 1) {
1606 ip_rt_multicast_event(pmc->interface);
1608 return 0;
1611 #ifdef CONFIG_IP_MULTICAST
1612 static void sf_markstate(struct ip_mc_list *pmc)
1614 struct ip_sf_list *psf;
1615 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1617 for (psf=pmc->sources; psf; psf=psf->sf_next)
1618 if (pmc->sfcount[MCAST_EXCLUDE]) {
1619 psf->sf_oldin = mca_xcount ==
1620 psf->sf_count[MCAST_EXCLUDE] &&
1621 !psf->sf_count[MCAST_INCLUDE];
1622 } else
1623 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1626 static int sf_setstate(struct ip_mc_list *pmc)
1628 struct ip_sf_list *psf, *dpsf;
1629 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1630 int qrv = pmc->interface->mr_qrv;
1631 int new_in, rv;
1633 rv = 0;
1634 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1635 if (pmc->sfcount[MCAST_EXCLUDE]) {
1636 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1637 !psf->sf_count[MCAST_INCLUDE];
1638 } else
1639 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1640 if (new_in) {
1641 if (!psf->sf_oldin) {
1642 struct ip_sf_list *prev = NULL;
1644 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1645 if (dpsf->sf_inaddr == psf->sf_inaddr)
1646 break;
1647 prev = dpsf;
1649 if (dpsf) {
1650 if (prev)
1651 prev->sf_next = dpsf->sf_next;
1652 else
1653 pmc->tomb = dpsf->sf_next;
1654 kfree(dpsf);
1656 psf->sf_crcount = qrv;
1657 rv++;
1659 } else if (psf->sf_oldin) {
1661 psf->sf_crcount = 0;
1663 * add or update "delete" records if an active filter
1664 * is now inactive
1666 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1667 if (dpsf->sf_inaddr == psf->sf_inaddr)
1668 break;
1669 if (!dpsf) {
1670 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1671 if (!dpsf)
1672 continue;
1673 *dpsf = *psf;
1674 /* pmc->lock held by callers */
1675 dpsf->sf_next = pmc->tomb;
1676 pmc->tomb = dpsf;
1678 dpsf->sf_crcount = qrv;
1679 rv++;
1682 return rv;
1684 #endif
1687 * Add multicast source filter list to the interface list
1689 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1690 int sfcount, __be32 *psfsrc, int delta)
1692 struct ip_mc_list *pmc;
1693 int isexclude;
1694 int i, err;
1696 if (!in_dev)
1697 return -ENODEV;
1698 rcu_read_lock();
1699 for_each_pmc_rcu(in_dev, pmc) {
1700 if (*pmca == pmc->multiaddr)
1701 break;
1703 if (!pmc) {
1704 /* MCA not found?? bug */
1705 rcu_read_unlock();
1706 return -ESRCH;
1708 spin_lock_bh(&pmc->lock);
1709 rcu_read_unlock();
1711 #ifdef CONFIG_IP_MULTICAST
1712 sf_markstate(pmc);
1713 #endif
1714 isexclude = pmc->sfmode == MCAST_EXCLUDE;
1715 if (!delta)
1716 pmc->sfcount[sfmode]++;
1717 err = 0;
1718 for (i=0; i<sfcount; i++) {
1719 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1720 if (err)
1721 break;
1723 if (err) {
1724 int j;
1726 if (!delta)
1727 pmc->sfcount[sfmode]--;
1728 for (j=0; j<i; j++)
1729 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1730 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1731 #ifdef CONFIG_IP_MULTICAST
1732 struct ip_sf_list *psf;
1733 in_dev = pmc->interface;
1734 #endif
1736 /* filter mode change */
1737 if (pmc->sfcount[MCAST_EXCLUDE])
1738 pmc->sfmode = MCAST_EXCLUDE;
1739 else if (pmc->sfcount[MCAST_INCLUDE])
1740 pmc->sfmode = MCAST_INCLUDE;
1741 #ifdef CONFIG_IP_MULTICAST
1742 /* else no filters; keep old mode for reports */
1744 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1745 IGMP_Unsolicited_Report_Count;
1746 in_dev->mr_ifc_count = pmc->crcount;
1747 for (psf=pmc->sources; psf; psf = psf->sf_next)
1748 psf->sf_crcount = 0;
1749 igmp_ifc_event(in_dev);
1750 } else if (sf_setstate(pmc)) {
1751 igmp_ifc_event(in_dev);
1752 #endif
1754 spin_unlock_bh(&pmc->lock);
1755 return err;
1758 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1760 struct ip_sf_list *psf, *nextpsf;
1762 for (psf=pmc->tomb; psf; psf=nextpsf) {
1763 nextpsf = psf->sf_next;
1764 kfree(psf);
1766 pmc->tomb = NULL;
1767 for (psf=pmc->sources; psf; psf=nextpsf) {
1768 nextpsf = psf->sf_next;
1769 kfree(psf);
1771 pmc->sources = NULL;
1772 pmc->sfmode = MCAST_EXCLUDE;
1773 pmc->sfcount[MCAST_INCLUDE] = 0;
1774 pmc->sfcount[MCAST_EXCLUDE] = 1;
1779 * Join a multicast group
1781 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1783 int err;
1784 __be32 addr = imr->imr_multiaddr.s_addr;
1785 struct ip_mc_socklist *iml = NULL, *i;
1786 struct in_device *in_dev;
1787 struct inet_sock *inet = inet_sk(sk);
1788 struct net *net = sock_net(sk);
1789 int ifindex;
1790 int count = 0;
1792 if (!ipv4_is_multicast(addr))
1793 return -EINVAL;
1795 rtnl_lock();
1797 in_dev = ip_mc_find_dev(net, imr);
1799 if (!in_dev) {
1800 iml = NULL;
1801 err = -ENODEV;
1802 goto done;
1805 err = -EADDRINUSE;
1806 ifindex = imr->imr_ifindex;
1807 for_each_pmc_rtnl(inet, i) {
1808 if (i->multi.imr_multiaddr.s_addr == addr &&
1809 i->multi.imr_ifindex == ifindex)
1810 goto done;
1811 count++;
1813 err = -ENOBUFS;
1814 if (count >= sysctl_igmp_max_memberships)
1815 goto done;
1816 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1817 if (iml == NULL)
1818 goto done;
1820 memcpy(&iml->multi, imr, sizeof(*imr));
1821 iml->next_rcu = inet->mc_list;
1822 iml->sflist = NULL;
1823 iml->sfmode = MCAST_EXCLUDE;
1824 rcu_assign_pointer(inet->mc_list, iml);
1825 ip_mc_inc_group(in_dev, addr);
1826 err = 0;
1827 done:
1828 rtnl_unlock();
1829 return err;
1831 EXPORT_SYMBOL(ip_mc_join_group);
1833 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1834 struct in_device *in_dev)
1836 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1837 int err;
1839 if (psf == NULL) {
1840 /* any-source empty exclude case */
1841 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1842 iml->sfmode, 0, NULL, 0);
1844 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1845 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1846 RCU_INIT_POINTER(iml->sflist, NULL);
1847 /* decrease mem now to avoid the memleak warning */
1848 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1849 kfree_rcu(psf, rcu);
1850 return err;
1854 * Ask a socket to leave a group.
1857 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1859 struct inet_sock *inet = inet_sk(sk);
1860 struct ip_mc_socklist *iml;
1861 struct ip_mc_socklist __rcu **imlp;
1862 struct in_device *in_dev;
1863 struct net *net = sock_net(sk);
1864 __be32 group = imr->imr_multiaddr.s_addr;
1865 u32 ifindex;
1866 int ret = -EADDRNOTAVAIL;
1868 rtnl_lock();
1869 in_dev = ip_mc_find_dev(net, imr);
1870 ifindex = imr->imr_ifindex;
1871 for (imlp = &inet->mc_list;
1872 (iml = rtnl_dereference(*imlp)) != NULL;
1873 imlp = &iml->next_rcu) {
1874 if (iml->multi.imr_multiaddr.s_addr != group)
1875 continue;
1876 if (ifindex) {
1877 if (iml->multi.imr_ifindex != ifindex)
1878 continue;
1879 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1880 iml->multi.imr_address.s_addr)
1881 continue;
1883 (void) ip_mc_leave_src(sk, iml, in_dev);
1885 *imlp = iml->next_rcu;
1887 if (in_dev)
1888 ip_mc_dec_group(in_dev, group);
1889 rtnl_unlock();
1890 /* decrease mem now to avoid the memleak warning */
1891 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1892 kfree_rcu(iml, rcu);
1893 return 0;
1895 if (!in_dev)
1896 ret = -ENODEV;
1897 rtnl_unlock();
1898 return ret;
1901 int ip_mc_source(int add, int omode, struct sock *sk, struct
1902 ip_mreq_source *mreqs, int ifindex)
1904 int err;
1905 struct ip_mreqn imr;
1906 __be32 addr = mreqs->imr_multiaddr;
1907 struct ip_mc_socklist *pmc;
1908 struct in_device *in_dev = NULL;
1909 struct inet_sock *inet = inet_sk(sk);
1910 struct ip_sf_socklist *psl;
1911 struct net *net = sock_net(sk);
1912 int leavegroup = 0;
1913 int i, j, rv;
1915 if (!ipv4_is_multicast(addr))
1916 return -EINVAL;
1918 rtnl_lock();
1920 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1921 imr.imr_address.s_addr = mreqs->imr_interface;
1922 imr.imr_ifindex = ifindex;
1923 in_dev = ip_mc_find_dev(net, &imr);
1925 if (!in_dev) {
1926 err = -ENODEV;
1927 goto done;
1929 err = -EADDRNOTAVAIL;
1931 for_each_pmc_rtnl(inet, pmc) {
1932 if ((pmc->multi.imr_multiaddr.s_addr ==
1933 imr.imr_multiaddr.s_addr) &&
1934 (pmc->multi.imr_ifindex == imr.imr_ifindex))
1935 break;
1937 if (!pmc) { /* must have a prior join */
1938 err = -EINVAL;
1939 goto done;
1941 /* if a source filter was set, must be the same mode as before */
1942 if (pmc->sflist) {
1943 if (pmc->sfmode != omode) {
1944 err = -EINVAL;
1945 goto done;
1947 } else if (pmc->sfmode != omode) {
1948 /* allow mode switches for empty-set filters */
1949 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1950 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1951 NULL, 0);
1952 pmc->sfmode = omode;
1955 psl = rtnl_dereference(pmc->sflist);
1956 if (!add) {
1957 if (!psl)
1958 goto done; /* err = -EADDRNOTAVAIL */
1959 rv = !0;
1960 for (i=0; i<psl->sl_count; i++) {
1961 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1962 sizeof(__be32));
1963 if (rv == 0)
1964 break;
1966 if (rv) /* source not found */
1967 goto done; /* err = -EADDRNOTAVAIL */
1969 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1970 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1971 leavegroup = 1;
1972 goto done;
1975 /* update the interface filter */
1976 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1977 &mreqs->imr_sourceaddr, 1);
1979 for (j=i+1; j<psl->sl_count; j++)
1980 psl->sl_addr[j-1] = psl->sl_addr[j];
1981 psl->sl_count--;
1982 err = 0;
1983 goto done;
1985 /* else, add a new source to the filter */
1987 if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1988 err = -ENOBUFS;
1989 goto done;
1991 if (!psl || psl->sl_count == psl->sl_max) {
1992 struct ip_sf_socklist *newpsl;
1993 int count = IP_SFBLOCK;
1995 if (psl)
1996 count += psl->sl_max;
1997 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
1998 if (!newpsl) {
1999 err = -ENOBUFS;
2000 goto done;
2002 newpsl->sl_max = count;
2003 newpsl->sl_count = count - IP_SFBLOCK;
2004 if (psl) {
2005 for (i=0; i<psl->sl_count; i++)
2006 newpsl->sl_addr[i] = psl->sl_addr[i];
2007 /* decrease mem now to avoid the memleak warning */
2008 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2009 kfree_rcu(psl, rcu);
2011 rcu_assign_pointer(pmc->sflist, newpsl);
2012 psl = newpsl;
2014 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2015 for (i=0; i<psl->sl_count; i++) {
2016 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2017 sizeof(__be32));
2018 if (rv == 0)
2019 break;
2021 if (rv == 0) /* address already there is an error */
2022 goto done;
2023 for (j=psl->sl_count-1; j>=i; j--)
2024 psl->sl_addr[j+1] = psl->sl_addr[j];
2025 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2026 psl->sl_count++;
2027 err = 0;
2028 /* update the interface list */
2029 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2030 &mreqs->imr_sourceaddr, 1);
2031 done:
2032 rtnl_unlock();
2033 if (leavegroup)
2034 return ip_mc_leave_group(sk, &imr);
2035 return err;
2038 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2040 int err = 0;
2041 struct ip_mreqn imr;
2042 __be32 addr = msf->imsf_multiaddr;
2043 struct ip_mc_socklist *pmc;
2044 struct in_device *in_dev;
2045 struct inet_sock *inet = inet_sk(sk);
2046 struct ip_sf_socklist *newpsl, *psl;
2047 struct net *net = sock_net(sk);
2048 int leavegroup = 0;
2050 if (!ipv4_is_multicast(addr))
2051 return -EINVAL;
2052 if (msf->imsf_fmode != MCAST_INCLUDE &&
2053 msf->imsf_fmode != MCAST_EXCLUDE)
2054 return -EINVAL;
2056 rtnl_lock();
2058 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2059 imr.imr_address.s_addr = msf->imsf_interface;
2060 imr.imr_ifindex = ifindex;
2061 in_dev = ip_mc_find_dev(net, &imr);
2063 if (!in_dev) {
2064 err = -ENODEV;
2065 goto done;
2068 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2069 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2070 leavegroup = 1;
2071 goto done;
2074 for_each_pmc_rtnl(inet, pmc) {
2075 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2076 pmc->multi.imr_ifindex == imr.imr_ifindex)
2077 break;
2079 if (!pmc) { /* must have a prior join */
2080 err = -EINVAL;
2081 goto done;
2083 if (msf->imsf_numsrc) {
2084 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2085 GFP_KERNEL);
2086 if (!newpsl) {
2087 err = -ENOBUFS;
2088 goto done;
2090 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2091 memcpy(newpsl->sl_addr, msf->imsf_slist,
2092 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2093 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2094 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2095 if (err) {
2096 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2097 goto done;
2099 } else {
2100 newpsl = NULL;
2101 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2102 msf->imsf_fmode, 0, NULL, 0);
2104 psl = rtnl_dereference(pmc->sflist);
2105 if (psl) {
2106 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2107 psl->sl_count, psl->sl_addr, 0);
2108 /* decrease mem now to avoid the memleak warning */
2109 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2110 kfree_rcu(psl, rcu);
2111 } else
2112 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2113 0, NULL, 0);
2114 rcu_assign_pointer(pmc->sflist, newpsl);
2115 pmc->sfmode = msf->imsf_fmode;
2116 err = 0;
2117 done:
2118 rtnl_unlock();
2119 if (leavegroup)
2120 err = ip_mc_leave_group(sk, &imr);
2121 return err;
2124 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2125 struct ip_msfilter __user *optval, int __user *optlen)
2127 int err, len, count, copycount;
2128 struct ip_mreqn imr;
2129 __be32 addr = msf->imsf_multiaddr;
2130 struct ip_mc_socklist *pmc;
2131 struct in_device *in_dev;
2132 struct inet_sock *inet = inet_sk(sk);
2133 struct ip_sf_socklist *psl;
2134 struct net *net = sock_net(sk);
2136 if (!ipv4_is_multicast(addr))
2137 return -EINVAL;
2139 rtnl_lock();
2141 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2142 imr.imr_address.s_addr = msf->imsf_interface;
2143 imr.imr_ifindex = 0;
2144 in_dev = ip_mc_find_dev(net, &imr);
2146 if (!in_dev) {
2147 err = -ENODEV;
2148 goto done;
2150 err = -EADDRNOTAVAIL;
2152 for_each_pmc_rtnl(inet, pmc) {
2153 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2154 pmc->multi.imr_ifindex == imr.imr_ifindex)
2155 break;
2157 if (!pmc) /* must have a prior join */
2158 goto done;
2159 msf->imsf_fmode = pmc->sfmode;
2160 psl = rtnl_dereference(pmc->sflist);
2161 rtnl_unlock();
2162 if (!psl) {
2163 len = 0;
2164 count = 0;
2165 } else {
2166 count = psl->sl_count;
2168 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2169 len = copycount * sizeof(psl->sl_addr[0]);
2170 msf->imsf_numsrc = count;
2171 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2172 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2173 return -EFAULT;
2175 if (len &&
2176 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2177 return -EFAULT;
2178 return 0;
2179 done:
2180 rtnl_unlock();
2181 return err;
2184 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2185 struct group_filter __user *optval, int __user *optlen)
2187 int err, i, count, copycount;
2188 struct sockaddr_in *psin;
2189 __be32 addr;
2190 struct ip_mc_socklist *pmc;
2191 struct inet_sock *inet = inet_sk(sk);
2192 struct ip_sf_socklist *psl;
2194 psin = (struct sockaddr_in *)&gsf->gf_group;
2195 if (psin->sin_family != AF_INET)
2196 return -EINVAL;
2197 addr = psin->sin_addr.s_addr;
2198 if (!ipv4_is_multicast(addr))
2199 return -EINVAL;
2201 rtnl_lock();
2203 err = -EADDRNOTAVAIL;
2205 for_each_pmc_rtnl(inet, pmc) {
2206 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2207 pmc->multi.imr_ifindex == gsf->gf_interface)
2208 break;
2210 if (!pmc) /* must have a prior join */
2211 goto done;
2212 gsf->gf_fmode = pmc->sfmode;
2213 psl = rtnl_dereference(pmc->sflist);
2214 rtnl_unlock();
2215 count = psl ? psl->sl_count : 0;
2216 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2217 gsf->gf_numsrc = count;
2218 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2219 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2220 return -EFAULT;
2222 for (i=0; i<copycount; i++) {
2223 struct sockaddr_storage ss;
2225 psin = (struct sockaddr_in *)&ss;
2226 memset(&ss, 0, sizeof(ss));
2227 psin->sin_family = AF_INET;
2228 psin->sin_addr.s_addr = psl->sl_addr[i];
2229 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2230 return -EFAULT;
2232 return 0;
2233 done:
2234 rtnl_unlock();
2235 return err;
2239 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2241 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2243 struct inet_sock *inet = inet_sk(sk);
2244 struct ip_mc_socklist *pmc;
2245 struct ip_sf_socklist *psl;
2246 int i;
2247 int ret;
2249 ret = 1;
2250 if (!ipv4_is_multicast(loc_addr))
2251 goto out;
2253 rcu_read_lock();
2254 for_each_pmc_rcu(inet, pmc) {
2255 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2256 pmc->multi.imr_ifindex == dif)
2257 break;
2259 ret = inet->mc_all;
2260 if (!pmc)
2261 goto unlock;
2262 psl = rcu_dereference(pmc->sflist);
2263 ret = (pmc->sfmode == MCAST_EXCLUDE);
2264 if (!psl)
2265 goto unlock;
2267 for (i=0; i<psl->sl_count; i++) {
2268 if (psl->sl_addr[i] == rmt_addr)
2269 break;
2271 ret = 0;
2272 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2273 goto unlock;
2274 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2275 goto unlock;
2276 ret = 1;
2277 unlock:
2278 rcu_read_unlock();
2279 out:
2280 return ret;
2284 * A socket is closing.
2287 void ip_mc_drop_socket(struct sock *sk)
2289 struct inet_sock *inet = inet_sk(sk);
2290 struct ip_mc_socklist *iml;
2291 struct net *net = sock_net(sk);
2293 if (inet->mc_list == NULL)
2294 return;
2296 rtnl_lock();
2297 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2298 struct in_device *in_dev;
2300 inet->mc_list = iml->next_rcu;
2301 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2302 (void) ip_mc_leave_src(sk, iml, in_dev);
2303 if (in_dev != NULL)
2304 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2305 /* decrease mem now to avoid the memleak warning */
2306 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2307 kfree_rcu(iml, rcu);
2309 rtnl_unlock();
2312 /* called with rcu_read_lock() */
2313 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2315 struct ip_mc_list *im;
2316 struct ip_sf_list *psf;
2317 int rv = 0;
2319 for_each_pmc_rcu(in_dev, im) {
2320 if (im->multiaddr == mc_addr)
2321 break;
2323 if (im && proto == IPPROTO_IGMP) {
2324 rv = 1;
2325 } else if (im) {
2326 if (src_addr) {
2327 for (psf=im->sources; psf; psf=psf->sf_next) {
2328 if (psf->sf_inaddr == src_addr)
2329 break;
2331 if (psf)
2332 rv = psf->sf_count[MCAST_INCLUDE] ||
2333 psf->sf_count[MCAST_EXCLUDE] !=
2334 im->sfcount[MCAST_EXCLUDE];
2335 else
2336 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2337 } else
2338 rv = 1; /* unspecified source; tentatively allow */
2340 return rv;
2343 #if defined(CONFIG_PROC_FS)
2344 struct igmp_mc_iter_state {
2345 struct seq_net_private p;
2346 struct net_device *dev;
2347 struct in_device *in_dev;
2350 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2352 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2354 struct net *net = seq_file_net(seq);
2355 struct ip_mc_list *im = NULL;
2356 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2358 state->in_dev = NULL;
2359 for_each_netdev_rcu(net, state->dev) {
2360 struct in_device *in_dev;
2362 in_dev = __in_dev_get_rcu(state->dev);
2363 if (!in_dev)
2364 continue;
2365 im = rcu_dereference(in_dev->mc_list);
2366 if (im) {
2367 state->in_dev = in_dev;
2368 break;
2371 return im;
2374 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2376 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2378 im = rcu_dereference(im->next_rcu);
2379 while (!im) {
2380 state->dev = next_net_device_rcu(state->dev);
2381 if (!state->dev) {
2382 state->in_dev = NULL;
2383 break;
2385 state->in_dev = __in_dev_get_rcu(state->dev);
2386 if (!state->in_dev)
2387 continue;
2388 im = rcu_dereference(state->in_dev->mc_list);
2390 return im;
2393 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2395 struct ip_mc_list *im = igmp_mc_get_first(seq);
2396 if (im)
2397 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2398 --pos;
2399 return pos ? NULL : im;
2402 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2403 __acquires(rcu)
2405 rcu_read_lock();
2406 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2409 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2411 struct ip_mc_list *im;
2412 if (v == SEQ_START_TOKEN)
2413 im = igmp_mc_get_first(seq);
2414 else
2415 im = igmp_mc_get_next(seq, v);
2416 ++*pos;
2417 return im;
2420 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2421 __releases(rcu)
2423 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2425 state->in_dev = NULL;
2426 state->dev = NULL;
2427 rcu_read_unlock();
2430 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2432 if (v == SEQ_START_TOKEN)
2433 seq_puts(seq,
2434 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2435 else {
2436 struct ip_mc_list *im = (struct ip_mc_list *)v;
2437 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2438 char *querier;
2439 #ifdef CONFIG_IP_MULTICAST
2440 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2441 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2442 "V3";
2443 #else
2444 querier = "NONE";
2445 #endif
2447 if (rcu_dereference(state->in_dev->mc_list) == im) {
2448 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2449 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2452 seq_printf(seq,
2453 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2454 im->multiaddr, im->users,
2455 im->tm_running, im->tm_running ?
2456 jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2457 im->reporter);
2459 return 0;
2462 static const struct seq_operations igmp_mc_seq_ops = {
2463 .start = igmp_mc_seq_start,
2464 .next = igmp_mc_seq_next,
2465 .stop = igmp_mc_seq_stop,
2466 .show = igmp_mc_seq_show,
2469 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2471 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2472 sizeof(struct igmp_mc_iter_state));
2475 static const struct file_operations igmp_mc_seq_fops = {
2476 .owner = THIS_MODULE,
2477 .open = igmp_mc_seq_open,
2478 .read = seq_read,
2479 .llseek = seq_lseek,
2480 .release = seq_release_net,
2483 struct igmp_mcf_iter_state {
2484 struct seq_net_private p;
2485 struct net_device *dev;
2486 struct in_device *idev;
2487 struct ip_mc_list *im;
2490 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2492 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2494 struct net *net = seq_file_net(seq);
2495 struct ip_sf_list *psf = NULL;
2496 struct ip_mc_list *im = NULL;
2497 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2499 state->idev = NULL;
2500 state->im = NULL;
2501 for_each_netdev_rcu(net, state->dev) {
2502 struct in_device *idev;
2503 idev = __in_dev_get_rcu(state->dev);
2504 if (unlikely(idev == NULL))
2505 continue;
2506 im = rcu_dereference(idev->mc_list);
2507 if (likely(im != NULL)) {
2508 spin_lock_bh(&im->lock);
2509 psf = im->sources;
2510 if (likely(psf != NULL)) {
2511 state->im = im;
2512 state->idev = idev;
2513 break;
2515 spin_unlock_bh(&im->lock);
2518 return psf;
2521 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2523 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2525 psf = psf->sf_next;
2526 while (!psf) {
2527 spin_unlock_bh(&state->im->lock);
2528 state->im = state->im->next;
2529 while (!state->im) {
2530 state->dev = next_net_device_rcu(state->dev);
2531 if (!state->dev) {
2532 state->idev = NULL;
2533 goto out;
2535 state->idev = __in_dev_get_rcu(state->dev);
2536 if (!state->idev)
2537 continue;
2538 state->im = rcu_dereference(state->idev->mc_list);
2540 if (!state->im)
2541 break;
2542 spin_lock_bh(&state->im->lock);
2543 psf = state->im->sources;
2545 out:
2546 return psf;
2549 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2551 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2552 if (psf)
2553 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2554 --pos;
2555 return pos ? NULL : psf;
2558 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2559 __acquires(rcu)
2561 rcu_read_lock();
2562 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2565 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2567 struct ip_sf_list *psf;
2568 if (v == SEQ_START_TOKEN)
2569 psf = igmp_mcf_get_first(seq);
2570 else
2571 psf = igmp_mcf_get_next(seq, v);
2572 ++*pos;
2573 return psf;
2576 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2577 __releases(rcu)
2579 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2580 if (likely(state->im != NULL)) {
2581 spin_unlock_bh(&state->im->lock);
2582 state->im = NULL;
2584 state->idev = NULL;
2585 state->dev = NULL;
2586 rcu_read_unlock();
2589 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2591 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2592 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2594 if (v == SEQ_START_TOKEN) {
2595 seq_printf(seq,
2596 "%3s %6s "
2597 "%10s %10s %6s %6s\n", "Idx",
2598 "Device", "MCA",
2599 "SRC", "INC", "EXC");
2600 } else {
2601 seq_printf(seq,
2602 "%3d %6.6s 0x%08x "
2603 "0x%08x %6lu %6lu\n",
2604 state->dev->ifindex, state->dev->name,
2605 ntohl(state->im->multiaddr),
2606 ntohl(psf->sf_inaddr),
2607 psf->sf_count[MCAST_INCLUDE],
2608 psf->sf_count[MCAST_EXCLUDE]);
2610 return 0;
2613 static const struct seq_operations igmp_mcf_seq_ops = {
2614 .start = igmp_mcf_seq_start,
2615 .next = igmp_mcf_seq_next,
2616 .stop = igmp_mcf_seq_stop,
2617 .show = igmp_mcf_seq_show,
2620 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2622 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2623 sizeof(struct igmp_mcf_iter_state));
2626 static const struct file_operations igmp_mcf_seq_fops = {
2627 .owner = THIS_MODULE,
2628 .open = igmp_mcf_seq_open,
2629 .read = seq_read,
2630 .llseek = seq_lseek,
2631 .release = seq_release_net,
2634 static int __net_init igmp_net_init(struct net *net)
2636 struct proc_dir_entry *pde;
2638 pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2639 if (!pde)
2640 goto out_igmp;
2641 pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2642 if (!pde)
2643 goto out_mcfilter;
2644 return 0;
2646 out_mcfilter:
2647 proc_net_remove(net, "igmp");
2648 out_igmp:
2649 return -ENOMEM;
2652 static void __net_exit igmp_net_exit(struct net *net)
2654 proc_net_remove(net, "mcfilter");
2655 proc_net_remove(net, "igmp");
2658 static struct pernet_operations igmp_net_ops = {
2659 .init = igmp_net_init,
2660 .exit = igmp_net_exit,
2663 int __init igmp_mc_proc_init(void)
2665 return register_pernet_subsys(&igmp_net_ops);
2667 #endif