igmp: avoid two atomic ops in igmp_rcv()
[linux-rapidio-2.6.git] / net / ipv4 / igmp.c
blob3294f547c48113cba0858a6011aace1630074491
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(im);
160 #ifdef CONFIG_IP_MULTICAST
163 * Timer management
166 static __inline__ void igmp_stop_timer(struct ip_mc_list *im)
168 spin_lock_bh(&im->lock);
169 if (del_timer(&im->timer))
170 atomic_dec(&im->refcnt);
171 im->tm_running = 0;
172 im->reporter = 0;
173 im->unsolicit_count = 0;
174 spin_unlock_bh(&im->lock);
177 /* It must be called with locked im->lock */
178 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
180 int tv = net_random() % max_delay;
182 im->tm_running = 1;
183 if (!mod_timer(&im->timer, jiffies+tv+2))
184 atomic_inc(&im->refcnt);
187 static void igmp_gq_start_timer(struct in_device *in_dev)
189 int tv = net_random() % in_dev->mr_maxdelay;
191 in_dev->mr_gq_running = 1;
192 if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
193 in_dev_hold(in_dev);
196 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
198 int tv = net_random() % delay;
200 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
201 in_dev_hold(in_dev);
204 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
206 spin_lock_bh(&im->lock);
207 im->unsolicit_count = 0;
208 if (del_timer(&im->timer)) {
209 if ((long)(im->timer.expires-jiffies) < max_delay) {
210 add_timer(&im->timer);
211 im->tm_running = 1;
212 spin_unlock_bh(&im->lock);
213 return;
215 atomic_dec(&im->refcnt);
217 igmp_start_timer(im, max_delay);
218 spin_unlock_bh(&im->lock);
223 * Send an IGMP report.
226 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
229 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
230 int gdeleted, int sdeleted)
232 switch (type) {
233 case IGMPV3_MODE_IS_INCLUDE:
234 case IGMPV3_MODE_IS_EXCLUDE:
235 if (gdeleted || sdeleted)
236 return 0;
237 if (!(pmc->gsquery && !psf->sf_gsresp)) {
238 if (pmc->sfmode == MCAST_INCLUDE)
239 return 1;
240 /* don't include if this source is excluded
241 * in all filters
243 if (psf->sf_count[MCAST_INCLUDE])
244 return type == IGMPV3_MODE_IS_INCLUDE;
245 return pmc->sfcount[MCAST_EXCLUDE] ==
246 psf->sf_count[MCAST_EXCLUDE];
248 return 0;
249 case IGMPV3_CHANGE_TO_INCLUDE:
250 if (gdeleted || sdeleted)
251 return 0;
252 return psf->sf_count[MCAST_INCLUDE] != 0;
253 case IGMPV3_CHANGE_TO_EXCLUDE:
254 if (gdeleted || sdeleted)
255 return 0;
256 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
257 psf->sf_count[MCAST_INCLUDE])
258 return 0;
259 return pmc->sfcount[MCAST_EXCLUDE] ==
260 psf->sf_count[MCAST_EXCLUDE];
261 case IGMPV3_ALLOW_NEW_SOURCES:
262 if (gdeleted || !psf->sf_crcount)
263 return 0;
264 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
265 case IGMPV3_BLOCK_OLD_SOURCES:
266 if (pmc->sfmode == MCAST_INCLUDE)
267 return gdeleted || (psf->sf_crcount && sdeleted);
268 return psf->sf_crcount && !gdeleted && !sdeleted;
270 return 0;
273 static int
274 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
276 struct ip_sf_list *psf;
277 int scount = 0;
279 for (psf=pmc->sources; psf; psf=psf->sf_next) {
280 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
281 continue;
282 scount++;
284 return scount;
287 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
289 struct sk_buff *skb;
290 struct rtable *rt;
291 struct iphdr *pip;
292 struct igmpv3_report *pig;
293 struct net *net = dev_net(dev);
295 skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
296 if (skb == NULL)
297 return NULL;
300 struct flowi fl = { .oif = dev->ifindex,
301 .nl_u = { .ip4_u = {
302 .daddr = IGMPV3_ALL_MCR } },
303 .proto = IPPROTO_IGMP };
304 if (ip_route_output_key(net, &rt, &fl)) {
305 kfree_skb(skb);
306 return NULL;
309 if (rt->rt_src == 0) {
310 kfree_skb(skb);
311 ip_rt_put(rt);
312 return NULL;
315 skb_dst_set(skb, &rt->u.dst);
316 skb->dev = dev;
318 skb_reserve(skb, LL_RESERVED_SPACE(dev));
320 skb_reset_network_header(skb);
321 pip = ip_hdr(skb);
322 skb_put(skb, sizeof(struct iphdr) + 4);
324 pip->version = 4;
325 pip->ihl = (sizeof(struct iphdr)+4)>>2;
326 pip->tos = 0xc0;
327 pip->frag_off = htons(IP_DF);
328 pip->ttl = 1;
329 pip->daddr = rt->rt_dst;
330 pip->saddr = rt->rt_src;
331 pip->protocol = IPPROTO_IGMP;
332 pip->tot_len = 0; /* filled in later */
333 ip_select_ident(pip, &rt->u.dst, NULL);
334 ((u8*)&pip[1])[0] = IPOPT_RA;
335 ((u8*)&pip[1])[1] = 4;
336 ((u8*)&pip[1])[2] = 0;
337 ((u8*)&pip[1])[3] = 0;
339 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
340 skb_put(skb, sizeof(*pig));
341 pig = igmpv3_report_hdr(skb);
342 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
343 pig->resv1 = 0;
344 pig->csum = 0;
345 pig->resv2 = 0;
346 pig->ngrec = 0;
347 return skb;
350 static int igmpv3_sendpack(struct sk_buff *skb)
352 struct igmphdr *pig = igmp_hdr(skb);
353 const int igmplen = skb->tail - skb->transport_header;
355 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
357 return ip_local_out(skb);
360 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
362 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
365 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
366 int type, struct igmpv3_grec **ppgr)
368 struct net_device *dev = pmc->interface->dev;
369 struct igmpv3_report *pih;
370 struct igmpv3_grec *pgr;
372 if (!skb)
373 skb = igmpv3_newpack(dev, dev->mtu);
374 if (!skb)
375 return NULL;
376 pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
377 pgr->grec_type = type;
378 pgr->grec_auxwords = 0;
379 pgr->grec_nsrcs = 0;
380 pgr->grec_mca = pmc->multiaddr;
381 pih = igmpv3_report_hdr(skb);
382 pih->ngrec = htons(ntohs(pih->ngrec)+1);
383 *ppgr = pgr;
384 return skb;
387 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? (skb)->dev->mtu - (skb)->len : \
388 skb_tailroom(skb)) : 0)
390 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
391 int type, int gdeleted, int sdeleted)
393 struct net_device *dev = pmc->interface->dev;
394 struct igmpv3_report *pih;
395 struct igmpv3_grec *pgr = NULL;
396 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
397 int scount, stotal, first, isquery, truncate;
399 if (pmc->multiaddr == IGMP_ALL_HOSTS)
400 return skb;
402 isquery = type == IGMPV3_MODE_IS_INCLUDE ||
403 type == IGMPV3_MODE_IS_EXCLUDE;
404 truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
405 type == IGMPV3_CHANGE_TO_EXCLUDE;
407 stotal = scount = 0;
409 psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
411 if (!*psf_list)
412 goto empty_source;
414 pih = skb ? igmpv3_report_hdr(skb) : NULL;
416 /* EX and TO_EX get a fresh packet, if needed */
417 if (truncate) {
418 if (pih && pih->ngrec &&
419 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
420 if (skb)
421 igmpv3_sendpack(skb);
422 skb = igmpv3_newpack(dev, dev->mtu);
425 first = 1;
426 psf_prev = NULL;
427 for (psf=*psf_list; psf; psf=psf_next) {
428 __be32 *psrc;
430 psf_next = psf->sf_next;
432 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
433 psf_prev = psf;
434 continue;
437 /* clear marks on query responses */
438 if (isquery)
439 psf->sf_gsresp = 0;
441 if (AVAILABLE(skb) < sizeof(__be32) +
442 first*sizeof(struct igmpv3_grec)) {
443 if (truncate && !first)
444 break; /* truncate these */
445 if (pgr)
446 pgr->grec_nsrcs = htons(scount);
447 if (skb)
448 igmpv3_sendpack(skb);
449 skb = igmpv3_newpack(dev, dev->mtu);
450 first = 1;
451 scount = 0;
453 if (first) {
454 skb = add_grhead(skb, pmc, type, &pgr);
455 first = 0;
457 if (!skb)
458 return NULL;
459 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
460 *psrc = psf->sf_inaddr;
461 scount++; stotal++;
462 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
463 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
464 psf->sf_crcount--;
465 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
466 if (psf_prev)
467 psf_prev->sf_next = psf->sf_next;
468 else
469 *psf_list = psf->sf_next;
470 kfree(psf);
471 continue;
474 psf_prev = psf;
477 empty_source:
478 if (!stotal) {
479 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
480 type == IGMPV3_BLOCK_OLD_SOURCES)
481 return skb;
482 if (pmc->crcount || isquery) {
483 /* make sure we have room for group header */
484 if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
485 igmpv3_sendpack(skb);
486 skb = NULL; /* add_grhead will get a new one */
488 skb = add_grhead(skb, pmc, type, &pgr);
491 if (pgr)
492 pgr->grec_nsrcs = htons(scount);
494 if (isquery)
495 pmc->gsquery = 0; /* clear query state on report */
496 return skb;
499 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
501 struct sk_buff *skb = NULL;
502 int type;
504 if (!pmc) {
505 read_lock(&in_dev->mc_list_lock);
506 for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
507 if (pmc->multiaddr == IGMP_ALL_HOSTS)
508 continue;
509 spin_lock_bh(&pmc->lock);
510 if (pmc->sfcount[MCAST_EXCLUDE])
511 type = IGMPV3_MODE_IS_EXCLUDE;
512 else
513 type = IGMPV3_MODE_IS_INCLUDE;
514 skb = add_grec(skb, pmc, type, 0, 0);
515 spin_unlock_bh(&pmc->lock);
517 read_unlock(&in_dev->mc_list_lock);
518 } else {
519 spin_lock_bh(&pmc->lock);
520 if (pmc->sfcount[MCAST_EXCLUDE])
521 type = IGMPV3_MODE_IS_EXCLUDE;
522 else
523 type = IGMPV3_MODE_IS_INCLUDE;
524 skb = add_grec(skb, pmc, type, 0, 0);
525 spin_unlock_bh(&pmc->lock);
527 if (!skb)
528 return 0;
529 return igmpv3_sendpack(skb);
533 * remove zero-count source records from a source filter list
535 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
537 struct ip_sf_list *psf_prev, *psf_next, *psf;
539 psf_prev = NULL;
540 for (psf=*ppsf; psf; psf = psf_next) {
541 psf_next = psf->sf_next;
542 if (psf->sf_crcount == 0) {
543 if (psf_prev)
544 psf_prev->sf_next = psf->sf_next;
545 else
546 *ppsf = psf->sf_next;
547 kfree(psf);
548 } else
549 psf_prev = psf;
553 static void igmpv3_send_cr(struct in_device *in_dev)
555 struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
556 struct sk_buff *skb = NULL;
557 int type, dtype;
559 read_lock(&in_dev->mc_list_lock);
560 spin_lock_bh(&in_dev->mc_tomb_lock);
562 /* deleted MCA's */
563 pmc_prev = NULL;
564 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
565 pmc_next = pmc->next;
566 if (pmc->sfmode == MCAST_INCLUDE) {
567 type = IGMPV3_BLOCK_OLD_SOURCES;
568 dtype = IGMPV3_BLOCK_OLD_SOURCES;
569 skb = add_grec(skb, pmc, type, 1, 0);
570 skb = add_grec(skb, pmc, dtype, 1, 1);
572 if (pmc->crcount) {
573 if (pmc->sfmode == MCAST_EXCLUDE) {
574 type = IGMPV3_CHANGE_TO_INCLUDE;
575 skb = add_grec(skb, pmc, type, 1, 0);
577 pmc->crcount--;
578 if (pmc->crcount == 0) {
579 igmpv3_clear_zeros(&pmc->tomb);
580 igmpv3_clear_zeros(&pmc->sources);
583 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
584 if (pmc_prev)
585 pmc_prev->next = pmc_next;
586 else
587 in_dev->mc_tomb = pmc_next;
588 in_dev_put(pmc->interface);
589 kfree(pmc);
590 } else
591 pmc_prev = pmc;
593 spin_unlock_bh(&in_dev->mc_tomb_lock);
595 /* change recs */
596 for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
597 spin_lock_bh(&pmc->lock);
598 if (pmc->sfcount[MCAST_EXCLUDE]) {
599 type = IGMPV3_BLOCK_OLD_SOURCES;
600 dtype = IGMPV3_ALLOW_NEW_SOURCES;
601 } else {
602 type = IGMPV3_ALLOW_NEW_SOURCES;
603 dtype = IGMPV3_BLOCK_OLD_SOURCES;
605 skb = add_grec(skb, pmc, type, 0, 0);
606 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */
608 /* filter mode changes */
609 if (pmc->crcount) {
610 if (pmc->sfmode == MCAST_EXCLUDE)
611 type = IGMPV3_CHANGE_TO_EXCLUDE;
612 else
613 type = IGMPV3_CHANGE_TO_INCLUDE;
614 skb = add_grec(skb, pmc, type, 0, 0);
615 pmc->crcount--;
617 spin_unlock_bh(&pmc->lock);
619 read_unlock(&in_dev->mc_list_lock);
621 if (!skb)
622 return;
623 (void) igmpv3_sendpack(skb);
626 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
627 int type)
629 struct sk_buff *skb;
630 struct iphdr *iph;
631 struct igmphdr *ih;
632 struct rtable *rt;
633 struct net_device *dev = in_dev->dev;
634 struct net *net = dev_net(dev);
635 __be32 group = pmc ? pmc->multiaddr : 0;
636 __be32 dst;
638 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
639 return igmpv3_send_report(in_dev, pmc);
640 else if (type == IGMP_HOST_LEAVE_MESSAGE)
641 dst = IGMP_ALL_ROUTER;
642 else
643 dst = group;
646 struct flowi fl = { .oif = dev->ifindex,
647 .nl_u = { .ip4_u = { .daddr = dst } },
648 .proto = IPPROTO_IGMP };
649 if (ip_route_output_key(net, &rt, &fl))
650 return -1;
652 if (rt->rt_src == 0) {
653 ip_rt_put(rt);
654 return -1;
657 skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
658 if (skb == NULL) {
659 ip_rt_put(rt);
660 return -1;
663 skb_dst_set(skb, &rt->u.dst);
665 skb_reserve(skb, LL_RESERVED_SPACE(dev));
667 skb_reset_network_header(skb);
668 iph = ip_hdr(skb);
669 skb_put(skb, sizeof(struct iphdr) + 4);
671 iph->version = 4;
672 iph->ihl = (sizeof(struct iphdr)+4)>>2;
673 iph->tos = 0xc0;
674 iph->frag_off = htons(IP_DF);
675 iph->ttl = 1;
676 iph->daddr = dst;
677 iph->saddr = rt->rt_src;
678 iph->protocol = IPPROTO_IGMP;
679 ip_select_ident(iph, &rt->u.dst, NULL);
680 ((u8*)&iph[1])[0] = IPOPT_RA;
681 ((u8*)&iph[1])[1] = 4;
682 ((u8*)&iph[1])[2] = 0;
683 ((u8*)&iph[1])[3] = 0;
685 ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
686 ih->type = type;
687 ih->code = 0;
688 ih->csum = 0;
689 ih->group = group;
690 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
692 return ip_local_out(skb);
695 static void igmp_gq_timer_expire(unsigned long data)
697 struct in_device *in_dev = (struct in_device *)data;
699 in_dev->mr_gq_running = 0;
700 igmpv3_send_report(in_dev, NULL);
701 __in_dev_put(in_dev);
704 static void igmp_ifc_timer_expire(unsigned long data)
706 struct in_device *in_dev = (struct in_device *)data;
708 igmpv3_send_cr(in_dev);
709 if (in_dev->mr_ifc_count) {
710 in_dev->mr_ifc_count--;
711 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
713 __in_dev_put(in_dev);
716 static void igmp_ifc_event(struct in_device *in_dev)
718 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
719 return;
720 in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
721 IGMP_Unsolicited_Report_Count;
722 igmp_ifc_start_timer(in_dev, 1);
726 static void igmp_timer_expire(unsigned long data)
728 struct ip_mc_list *im=(struct ip_mc_list *)data;
729 struct in_device *in_dev = im->interface;
731 spin_lock(&im->lock);
732 im->tm_running = 0;
734 if (im->unsolicit_count) {
735 im->unsolicit_count--;
736 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
738 im->reporter = 1;
739 spin_unlock(&im->lock);
741 if (IGMP_V1_SEEN(in_dev))
742 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
743 else if (IGMP_V2_SEEN(in_dev))
744 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
745 else
746 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
748 ip_ma_put(im);
751 /* mark EXCLUDE-mode sources */
752 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
754 struct ip_sf_list *psf;
755 int i, scount;
757 scount = 0;
758 for (psf=pmc->sources; psf; psf=psf->sf_next) {
759 if (scount == nsrcs)
760 break;
761 for (i=0; i<nsrcs; i++) {
762 /* skip inactive filters */
763 if (pmc->sfcount[MCAST_INCLUDE] ||
764 pmc->sfcount[MCAST_EXCLUDE] !=
765 psf->sf_count[MCAST_EXCLUDE])
766 continue;
767 if (srcs[i] == psf->sf_inaddr) {
768 scount++;
769 break;
773 pmc->gsquery = 0;
774 if (scount == nsrcs) /* all sources excluded */
775 return 0;
776 return 1;
779 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
781 struct ip_sf_list *psf;
782 int i, scount;
784 if (pmc->sfmode == MCAST_EXCLUDE)
785 return igmp_xmarksources(pmc, nsrcs, srcs);
787 /* mark INCLUDE-mode sources */
788 scount = 0;
789 for (psf=pmc->sources; psf; psf=psf->sf_next) {
790 if (scount == nsrcs)
791 break;
792 for (i=0; i<nsrcs; i++)
793 if (srcs[i] == psf->sf_inaddr) {
794 psf->sf_gsresp = 1;
795 scount++;
796 break;
799 if (!scount) {
800 pmc->gsquery = 0;
801 return 0;
803 pmc->gsquery = 1;
804 return 1;
807 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
809 struct ip_mc_list *im;
811 /* Timers are only set for non-local groups */
813 if (group == IGMP_ALL_HOSTS)
814 return;
816 read_lock(&in_dev->mc_list_lock);
817 for (im=in_dev->mc_list; im!=NULL; im=im->next) {
818 if (im->multiaddr == group) {
819 igmp_stop_timer(im);
820 break;
823 read_unlock(&in_dev->mc_list_lock);
826 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
827 int len)
829 struct igmphdr *ih = igmp_hdr(skb);
830 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
831 struct ip_mc_list *im;
832 __be32 group = ih->group;
833 int max_delay;
834 int mark = 0;
837 if (len == 8) {
838 if (ih->code == 0) {
839 /* Alas, old v1 router presents here. */
841 max_delay = IGMP_Query_Response_Interval;
842 in_dev->mr_v1_seen = jiffies +
843 IGMP_V1_Router_Present_Timeout;
844 group = 0;
845 } else {
846 /* v2 router present */
847 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
848 in_dev->mr_v2_seen = jiffies +
849 IGMP_V2_Router_Present_Timeout;
851 /* cancel the interface change timer */
852 in_dev->mr_ifc_count = 0;
853 if (del_timer(&in_dev->mr_ifc_timer))
854 __in_dev_put(in_dev);
855 /* clear deleted report items */
856 igmpv3_clear_delrec(in_dev);
857 } else if (len < 12) {
858 return; /* ignore bogus packet; freed by caller */
859 } else { /* v3 */
860 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
861 return;
863 ih3 = igmpv3_query_hdr(skb);
864 if (ih3->nsrcs) {
865 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
866 + ntohs(ih3->nsrcs)*sizeof(__be32)))
867 return;
868 ih3 = igmpv3_query_hdr(skb);
871 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
872 if (!max_delay)
873 max_delay = 1; /* can't mod w/ 0 */
874 in_dev->mr_maxdelay = max_delay;
875 if (ih3->qrv)
876 in_dev->mr_qrv = ih3->qrv;
877 if (!group) { /* general query */
878 if (ih3->nsrcs)
879 return; /* no sources allowed */
880 igmp_gq_start_timer(in_dev);
881 return;
883 /* mark sources to include, if group & source-specific */
884 mark = ih3->nsrcs != 0;
888 * - Start the timers in all of our membership records
889 * that the query applies to for the interface on
890 * which the query arrived excl. those that belong
891 * to a "local" group (224.0.0.X)
892 * - For timers already running check if they need to
893 * be reset.
894 * - Use the igmp->igmp_code field as the maximum
895 * delay possible
897 read_lock(&in_dev->mc_list_lock);
898 for (im=in_dev->mc_list; im!=NULL; im=im->next) {
899 int changed;
901 if (group && group != im->multiaddr)
902 continue;
903 if (im->multiaddr == IGMP_ALL_HOSTS)
904 continue;
905 spin_lock_bh(&im->lock);
906 if (im->tm_running)
907 im->gsquery = im->gsquery && mark;
908 else
909 im->gsquery = mark;
910 changed = !im->gsquery ||
911 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
912 spin_unlock_bh(&im->lock);
913 if (changed)
914 igmp_mod_timer(im, max_delay);
916 read_unlock(&in_dev->mc_list_lock);
919 /* called in rcu_read_lock() section */
920 int igmp_rcv(struct sk_buff *skb)
922 /* This basically follows the spec line by line -- see RFC1112 */
923 struct igmphdr *ih;
924 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
925 int len = skb->len;
927 if (in_dev == NULL)
928 goto drop;
930 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
931 goto drop;
933 switch (skb->ip_summed) {
934 case CHECKSUM_COMPLETE:
935 if (!csum_fold(skb->csum))
936 break;
937 /* fall through */
938 case CHECKSUM_NONE:
939 skb->csum = 0;
940 if (__skb_checksum_complete(skb))
941 goto drop;
944 ih = igmp_hdr(skb);
945 switch (ih->type) {
946 case IGMP_HOST_MEMBERSHIP_QUERY:
947 igmp_heard_query(in_dev, skb, len);
948 break;
949 case IGMP_HOST_MEMBERSHIP_REPORT:
950 case IGMPV2_HOST_MEMBERSHIP_REPORT:
951 /* Is it our report looped back? */
952 if (skb_rtable(skb)->fl.iif == 0)
953 break;
954 /* don't rely on MC router hearing unicast reports */
955 if (skb->pkt_type == PACKET_MULTICAST ||
956 skb->pkt_type == PACKET_BROADCAST)
957 igmp_heard_report(in_dev, ih->group);
958 break;
959 case IGMP_PIM:
960 #ifdef CONFIG_IP_PIMSM_V1
961 return pim_rcv_v1(skb);
962 #endif
963 case IGMPV3_HOST_MEMBERSHIP_REPORT:
964 case IGMP_DVMRP:
965 case IGMP_TRACE:
966 case IGMP_HOST_LEAVE_MESSAGE:
967 case IGMP_MTRACE:
968 case IGMP_MTRACE_RESP:
969 break;
970 default:
971 break;
974 drop:
975 kfree_skb(skb);
976 return 0;
979 #endif
983 * Add a filter to a device
986 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
988 char buf[MAX_ADDR_LEN];
989 struct net_device *dev = in_dev->dev;
991 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
992 We will get multicast token leakage, when IFF_MULTICAST
993 is changed. This check should be done in dev->set_multicast_list
994 routine. Something sort of:
995 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
996 --ANK
998 if (arp_mc_map(addr, buf, dev, 0) == 0)
999 dev_mc_add(dev, buf);
1003 * Remove a filter from a device
1006 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1008 char buf[MAX_ADDR_LEN];
1009 struct net_device *dev = in_dev->dev;
1011 if (arp_mc_map(addr, buf, dev, 0) == 0)
1012 dev_mc_del(dev, buf);
1015 #ifdef CONFIG_IP_MULTICAST
1017 * deleted ip_mc_list manipulation
1019 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1021 struct ip_mc_list *pmc;
1023 /* this is an "ip_mc_list" for convenience; only the fields below
1024 * are actually used. In particular, the refcnt and users are not
1025 * used for management of the delete list. Using the same structure
1026 * for deleted items allows change reports to use common code with
1027 * non-deleted or query-response MCA's.
1029 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1030 if (!pmc)
1031 return;
1032 spin_lock_bh(&im->lock);
1033 pmc->interface = im->interface;
1034 in_dev_hold(in_dev);
1035 pmc->multiaddr = im->multiaddr;
1036 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1037 IGMP_Unsolicited_Report_Count;
1038 pmc->sfmode = im->sfmode;
1039 if (pmc->sfmode == MCAST_INCLUDE) {
1040 struct ip_sf_list *psf;
1042 pmc->tomb = im->tomb;
1043 pmc->sources = im->sources;
1044 im->tomb = im->sources = NULL;
1045 for (psf=pmc->sources; psf; psf=psf->sf_next)
1046 psf->sf_crcount = pmc->crcount;
1048 spin_unlock_bh(&im->lock);
1050 spin_lock_bh(&in_dev->mc_tomb_lock);
1051 pmc->next = in_dev->mc_tomb;
1052 in_dev->mc_tomb = pmc;
1053 spin_unlock_bh(&in_dev->mc_tomb_lock);
1056 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1058 struct ip_mc_list *pmc, *pmc_prev;
1059 struct ip_sf_list *psf, *psf_next;
1061 spin_lock_bh(&in_dev->mc_tomb_lock);
1062 pmc_prev = NULL;
1063 for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1064 if (pmc->multiaddr == multiaddr)
1065 break;
1066 pmc_prev = pmc;
1068 if (pmc) {
1069 if (pmc_prev)
1070 pmc_prev->next = pmc->next;
1071 else
1072 in_dev->mc_tomb = pmc->next;
1074 spin_unlock_bh(&in_dev->mc_tomb_lock);
1075 if (pmc) {
1076 for (psf=pmc->tomb; psf; psf=psf_next) {
1077 psf_next = psf->sf_next;
1078 kfree(psf);
1080 in_dev_put(pmc->interface);
1081 kfree(pmc);
1085 static void igmpv3_clear_delrec(struct in_device *in_dev)
1087 struct ip_mc_list *pmc, *nextpmc;
1089 spin_lock_bh(&in_dev->mc_tomb_lock);
1090 pmc = in_dev->mc_tomb;
1091 in_dev->mc_tomb = NULL;
1092 spin_unlock_bh(&in_dev->mc_tomb_lock);
1094 for (; pmc; pmc = nextpmc) {
1095 nextpmc = pmc->next;
1096 ip_mc_clear_src(pmc);
1097 in_dev_put(pmc->interface);
1098 kfree(pmc);
1100 /* clear dead sources, too */
1101 read_lock(&in_dev->mc_list_lock);
1102 for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
1103 struct ip_sf_list *psf, *psf_next;
1105 spin_lock_bh(&pmc->lock);
1106 psf = pmc->tomb;
1107 pmc->tomb = NULL;
1108 spin_unlock_bh(&pmc->lock);
1109 for (; psf; psf=psf_next) {
1110 psf_next = psf->sf_next;
1111 kfree(psf);
1114 read_unlock(&in_dev->mc_list_lock);
1116 #endif
1118 static void igmp_group_dropped(struct ip_mc_list *im)
1120 struct in_device *in_dev = im->interface;
1121 #ifdef CONFIG_IP_MULTICAST
1122 int reporter;
1123 #endif
1125 if (im->loaded) {
1126 im->loaded = 0;
1127 ip_mc_filter_del(in_dev, im->multiaddr);
1130 #ifdef CONFIG_IP_MULTICAST
1131 if (im->multiaddr == IGMP_ALL_HOSTS)
1132 return;
1134 reporter = im->reporter;
1135 igmp_stop_timer(im);
1137 if (!in_dev->dead) {
1138 if (IGMP_V1_SEEN(in_dev))
1139 goto done;
1140 if (IGMP_V2_SEEN(in_dev)) {
1141 if (reporter)
1142 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1143 goto done;
1145 /* IGMPv3 */
1146 igmpv3_add_delrec(in_dev, im);
1148 igmp_ifc_event(in_dev);
1150 done:
1151 #endif
1152 ip_mc_clear_src(im);
1155 static void igmp_group_added(struct ip_mc_list *im)
1157 struct in_device *in_dev = im->interface;
1159 if (im->loaded == 0) {
1160 im->loaded = 1;
1161 ip_mc_filter_add(in_dev, im->multiaddr);
1164 #ifdef CONFIG_IP_MULTICAST
1165 if (im->multiaddr == IGMP_ALL_HOSTS)
1166 return;
1168 if (in_dev->dead)
1169 return;
1170 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1171 spin_lock_bh(&im->lock);
1172 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1173 spin_unlock_bh(&im->lock);
1174 return;
1176 /* else, v3 */
1178 im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1179 IGMP_Unsolicited_Report_Count;
1180 igmp_ifc_event(in_dev);
1181 #endif
1186 * Multicast list managers
1191 * A socket has joined a multicast group on device dev.
1194 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1196 struct ip_mc_list *im;
1198 ASSERT_RTNL();
1200 for (im=in_dev->mc_list; im; im=im->next) {
1201 if (im->multiaddr == addr) {
1202 im->users++;
1203 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1204 goto out;
1208 im = kmalloc(sizeof(*im), GFP_KERNEL);
1209 if (!im)
1210 goto out;
1212 im->users = 1;
1213 im->interface = in_dev;
1214 in_dev_hold(in_dev);
1215 im->multiaddr = addr;
1216 /* initial mode is (EX, empty) */
1217 im->sfmode = MCAST_EXCLUDE;
1218 im->sfcount[MCAST_INCLUDE] = 0;
1219 im->sfcount[MCAST_EXCLUDE] = 1;
1220 im->sources = NULL;
1221 im->tomb = NULL;
1222 im->crcount = 0;
1223 atomic_set(&im->refcnt, 1);
1224 spin_lock_init(&im->lock);
1225 #ifdef CONFIG_IP_MULTICAST
1226 im->tm_running = 0;
1227 setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1228 im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1229 im->reporter = 0;
1230 im->gsquery = 0;
1231 #endif
1232 im->loaded = 0;
1233 write_lock_bh(&in_dev->mc_list_lock);
1234 im->next = in_dev->mc_list;
1235 in_dev->mc_list = im;
1236 in_dev->mc_count++;
1237 write_unlock_bh(&in_dev->mc_list_lock);
1238 #ifdef CONFIG_IP_MULTICAST
1239 igmpv3_del_delrec(in_dev, im->multiaddr);
1240 #endif
1241 igmp_group_added(im);
1242 if (!in_dev->dead)
1243 ip_rt_multicast_event(in_dev);
1244 out:
1245 return;
1249 * Resend IGMP JOIN report; used for bonding.
1251 void ip_mc_rejoin_group(struct ip_mc_list *im)
1253 #ifdef CONFIG_IP_MULTICAST
1254 struct in_device *in_dev = im->interface;
1256 if (im->multiaddr == IGMP_ALL_HOSTS)
1257 return;
1259 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1260 igmp_mod_timer(im, IGMP_Initial_Report_Delay);
1261 return;
1263 /* else, v3 */
1264 im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1265 IGMP_Unsolicited_Report_Count;
1266 igmp_ifc_event(in_dev);
1267 #endif
1271 * A socket has left a multicast group on device dev
1274 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1276 struct ip_mc_list *i, **ip;
1278 ASSERT_RTNL();
1280 for (ip=&in_dev->mc_list; (i=*ip)!=NULL; ip=&i->next) {
1281 if (i->multiaddr == addr) {
1282 if (--i->users == 0) {
1283 write_lock_bh(&in_dev->mc_list_lock);
1284 *ip = i->next;
1285 in_dev->mc_count--;
1286 write_unlock_bh(&in_dev->mc_list_lock);
1287 igmp_group_dropped(i);
1289 if (!in_dev->dead)
1290 ip_rt_multicast_event(in_dev);
1292 ip_ma_put(i);
1293 return;
1295 break;
1300 /* Device changing type */
1302 void ip_mc_unmap(struct in_device *in_dev)
1304 struct ip_mc_list *i;
1306 ASSERT_RTNL();
1308 for (i = in_dev->mc_list; i; i = i->next)
1309 igmp_group_dropped(i);
1312 void ip_mc_remap(struct in_device *in_dev)
1314 struct ip_mc_list *i;
1316 ASSERT_RTNL();
1318 for (i = in_dev->mc_list; i; i = i->next)
1319 igmp_group_added(i);
1322 /* Device going down */
1324 void ip_mc_down(struct in_device *in_dev)
1326 struct ip_mc_list *i;
1328 ASSERT_RTNL();
1330 for (i=in_dev->mc_list; i; i=i->next)
1331 igmp_group_dropped(i);
1333 #ifdef CONFIG_IP_MULTICAST
1334 in_dev->mr_ifc_count = 0;
1335 if (del_timer(&in_dev->mr_ifc_timer))
1336 __in_dev_put(in_dev);
1337 in_dev->mr_gq_running = 0;
1338 if (del_timer(&in_dev->mr_gq_timer))
1339 __in_dev_put(in_dev);
1340 igmpv3_clear_delrec(in_dev);
1341 #endif
1343 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1346 void ip_mc_init_dev(struct in_device *in_dev)
1348 ASSERT_RTNL();
1350 in_dev->mc_tomb = NULL;
1351 #ifdef CONFIG_IP_MULTICAST
1352 in_dev->mr_gq_running = 0;
1353 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1354 (unsigned long)in_dev);
1355 in_dev->mr_ifc_count = 0;
1356 in_dev->mc_count = 0;
1357 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1358 (unsigned long)in_dev);
1359 in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1360 #endif
1362 rwlock_init(&in_dev->mc_list_lock);
1363 spin_lock_init(&in_dev->mc_tomb_lock);
1366 /* Device going up */
1368 void ip_mc_up(struct in_device *in_dev)
1370 struct ip_mc_list *i;
1372 ASSERT_RTNL();
1374 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1376 for (i=in_dev->mc_list; i; i=i->next)
1377 igmp_group_added(i);
1381 * Device is about to be destroyed: clean up.
1384 void ip_mc_destroy_dev(struct in_device *in_dev)
1386 struct ip_mc_list *i;
1388 ASSERT_RTNL();
1390 /* Deactivate timers */
1391 ip_mc_down(in_dev);
1393 write_lock_bh(&in_dev->mc_list_lock);
1394 while ((i = in_dev->mc_list) != NULL) {
1395 in_dev->mc_list = i->next;
1396 in_dev->mc_count--;
1397 write_unlock_bh(&in_dev->mc_list_lock);
1398 igmp_group_dropped(i);
1399 ip_ma_put(i);
1401 write_lock_bh(&in_dev->mc_list_lock);
1403 write_unlock_bh(&in_dev->mc_list_lock);
1406 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1408 struct flowi fl = { .nl_u = { .ip4_u =
1409 { .daddr = imr->imr_multiaddr.s_addr } } };
1410 struct rtable *rt;
1411 struct net_device *dev = NULL;
1412 struct in_device *idev = NULL;
1414 if (imr->imr_ifindex) {
1415 idev = inetdev_by_index(net, imr->imr_ifindex);
1416 if (idev)
1417 __in_dev_put(idev);
1418 return idev;
1420 if (imr->imr_address.s_addr) {
1421 dev = ip_dev_find(net, imr->imr_address.s_addr);
1422 if (!dev)
1423 return NULL;
1424 dev_put(dev);
1427 if (!dev && !ip_route_output_key(net, &rt, &fl)) {
1428 dev = rt->u.dst.dev;
1429 ip_rt_put(rt);
1431 if (dev) {
1432 imr->imr_ifindex = dev->ifindex;
1433 idev = __in_dev_get_rtnl(dev);
1435 return idev;
1439 * Join a socket to a group
1441 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1442 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1445 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1446 __be32 *psfsrc)
1448 struct ip_sf_list *psf, *psf_prev;
1449 int rv = 0;
1451 psf_prev = NULL;
1452 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1453 if (psf->sf_inaddr == *psfsrc)
1454 break;
1455 psf_prev = psf;
1457 if (!psf || psf->sf_count[sfmode] == 0) {
1458 /* source filter not found, or count wrong => bug */
1459 return -ESRCH;
1461 psf->sf_count[sfmode]--;
1462 if (psf->sf_count[sfmode] == 0) {
1463 ip_rt_multicast_event(pmc->interface);
1465 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1466 #ifdef CONFIG_IP_MULTICAST
1467 struct in_device *in_dev = pmc->interface;
1468 #endif
1470 /* no more filters for this source */
1471 if (psf_prev)
1472 psf_prev->sf_next = psf->sf_next;
1473 else
1474 pmc->sources = psf->sf_next;
1475 #ifdef CONFIG_IP_MULTICAST
1476 if (psf->sf_oldin &&
1477 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1478 psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1479 IGMP_Unsolicited_Report_Count;
1480 psf->sf_next = pmc->tomb;
1481 pmc->tomb = psf;
1482 rv = 1;
1483 } else
1484 #endif
1485 kfree(psf);
1487 return rv;
1490 #ifndef CONFIG_IP_MULTICAST
1491 #define igmp_ifc_event(x) do { } while (0)
1492 #endif
1494 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1495 int sfcount, __be32 *psfsrc, int delta)
1497 struct ip_mc_list *pmc;
1498 int changerec = 0;
1499 int i, err;
1501 if (!in_dev)
1502 return -ENODEV;
1503 read_lock(&in_dev->mc_list_lock);
1504 for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
1505 if (*pmca == pmc->multiaddr)
1506 break;
1508 if (!pmc) {
1509 /* MCA not found?? bug */
1510 read_unlock(&in_dev->mc_list_lock);
1511 return -ESRCH;
1513 spin_lock_bh(&pmc->lock);
1514 read_unlock(&in_dev->mc_list_lock);
1515 #ifdef CONFIG_IP_MULTICAST
1516 sf_markstate(pmc);
1517 #endif
1518 if (!delta) {
1519 err = -EINVAL;
1520 if (!pmc->sfcount[sfmode])
1521 goto out_unlock;
1522 pmc->sfcount[sfmode]--;
1524 err = 0;
1525 for (i=0; i<sfcount; i++) {
1526 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1528 changerec |= rv > 0;
1529 if (!err && rv < 0)
1530 err = rv;
1532 if (pmc->sfmode == MCAST_EXCLUDE &&
1533 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1534 pmc->sfcount[MCAST_INCLUDE]) {
1535 #ifdef CONFIG_IP_MULTICAST
1536 struct ip_sf_list *psf;
1537 #endif
1539 /* filter mode change */
1540 pmc->sfmode = MCAST_INCLUDE;
1541 #ifdef CONFIG_IP_MULTICAST
1542 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1543 IGMP_Unsolicited_Report_Count;
1544 in_dev->mr_ifc_count = pmc->crcount;
1545 for (psf=pmc->sources; psf; psf = psf->sf_next)
1546 psf->sf_crcount = 0;
1547 igmp_ifc_event(pmc->interface);
1548 } else if (sf_setstate(pmc) || changerec) {
1549 igmp_ifc_event(pmc->interface);
1550 #endif
1552 out_unlock:
1553 spin_unlock_bh(&pmc->lock);
1554 return err;
1558 * Add multicast single-source filter to the interface list
1560 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1561 __be32 *psfsrc, int delta)
1563 struct ip_sf_list *psf, *psf_prev;
1565 psf_prev = NULL;
1566 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1567 if (psf->sf_inaddr == *psfsrc)
1568 break;
1569 psf_prev = psf;
1571 if (!psf) {
1572 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1573 if (!psf)
1574 return -ENOBUFS;
1575 psf->sf_inaddr = *psfsrc;
1576 if (psf_prev) {
1577 psf_prev->sf_next = psf;
1578 } else
1579 pmc->sources = psf;
1581 psf->sf_count[sfmode]++;
1582 if (psf->sf_count[sfmode] == 1) {
1583 ip_rt_multicast_event(pmc->interface);
1585 return 0;
1588 #ifdef CONFIG_IP_MULTICAST
1589 static void sf_markstate(struct ip_mc_list *pmc)
1591 struct ip_sf_list *psf;
1592 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1594 for (psf=pmc->sources; psf; psf=psf->sf_next)
1595 if (pmc->sfcount[MCAST_EXCLUDE]) {
1596 psf->sf_oldin = mca_xcount ==
1597 psf->sf_count[MCAST_EXCLUDE] &&
1598 !psf->sf_count[MCAST_INCLUDE];
1599 } else
1600 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1603 static int sf_setstate(struct ip_mc_list *pmc)
1605 struct ip_sf_list *psf, *dpsf;
1606 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1607 int qrv = pmc->interface->mr_qrv;
1608 int new_in, rv;
1610 rv = 0;
1611 for (psf=pmc->sources; psf; psf=psf->sf_next) {
1612 if (pmc->sfcount[MCAST_EXCLUDE]) {
1613 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1614 !psf->sf_count[MCAST_INCLUDE];
1615 } else
1616 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1617 if (new_in) {
1618 if (!psf->sf_oldin) {
1619 struct ip_sf_list *prev = NULL;
1621 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1622 if (dpsf->sf_inaddr == psf->sf_inaddr)
1623 break;
1624 prev = dpsf;
1626 if (dpsf) {
1627 if (prev)
1628 prev->sf_next = dpsf->sf_next;
1629 else
1630 pmc->tomb = dpsf->sf_next;
1631 kfree(dpsf);
1633 psf->sf_crcount = qrv;
1634 rv++;
1636 } else if (psf->sf_oldin) {
1638 psf->sf_crcount = 0;
1640 * add or update "delete" records if an active filter
1641 * is now inactive
1643 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1644 if (dpsf->sf_inaddr == psf->sf_inaddr)
1645 break;
1646 if (!dpsf) {
1647 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1648 if (!dpsf)
1649 continue;
1650 *dpsf = *psf;
1651 /* pmc->lock held by callers */
1652 dpsf->sf_next = pmc->tomb;
1653 pmc->tomb = dpsf;
1655 dpsf->sf_crcount = qrv;
1656 rv++;
1659 return rv;
1661 #endif
1664 * Add multicast source filter list to the interface list
1666 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1667 int sfcount, __be32 *psfsrc, int delta)
1669 struct ip_mc_list *pmc;
1670 int isexclude;
1671 int i, err;
1673 if (!in_dev)
1674 return -ENODEV;
1675 read_lock(&in_dev->mc_list_lock);
1676 for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
1677 if (*pmca == pmc->multiaddr)
1678 break;
1680 if (!pmc) {
1681 /* MCA not found?? bug */
1682 read_unlock(&in_dev->mc_list_lock);
1683 return -ESRCH;
1685 spin_lock_bh(&pmc->lock);
1686 read_unlock(&in_dev->mc_list_lock);
1688 #ifdef CONFIG_IP_MULTICAST
1689 sf_markstate(pmc);
1690 #endif
1691 isexclude = pmc->sfmode == MCAST_EXCLUDE;
1692 if (!delta)
1693 pmc->sfcount[sfmode]++;
1694 err = 0;
1695 for (i=0; i<sfcount; i++) {
1696 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1697 if (err)
1698 break;
1700 if (err) {
1701 int j;
1703 pmc->sfcount[sfmode]--;
1704 for (j=0; j<i; j++)
1705 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1706 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1707 #ifdef CONFIG_IP_MULTICAST
1708 struct ip_sf_list *psf;
1709 in_dev = pmc->interface;
1710 #endif
1712 /* filter mode change */
1713 if (pmc->sfcount[MCAST_EXCLUDE])
1714 pmc->sfmode = MCAST_EXCLUDE;
1715 else if (pmc->sfcount[MCAST_INCLUDE])
1716 pmc->sfmode = MCAST_INCLUDE;
1717 #ifdef CONFIG_IP_MULTICAST
1718 /* else no filters; keep old mode for reports */
1720 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1721 IGMP_Unsolicited_Report_Count;
1722 in_dev->mr_ifc_count = pmc->crcount;
1723 for (psf=pmc->sources; psf; psf = psf->sf_next)
1724 psf->sf_crcount = 0;
1725 igmp_ifc_event(in_dev);
1726 } else if (sf_setstate(pmc)) {
1727 igmp_ifc_event(in_dev);
1728 #endif
1730 spin_unlock_bh(&pmc->lock);
1731 return err;
1734 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1736 struct ip_sf_list *psf, *nextpsf;
1738 for (psf=pmc->tomb; psf; psf=nextpsf) {
1739 nextpsf = psf->sf_next;
1740 kfree(psf);
1742 pmc->tomb = NULL;
1743 for (psf=pmc->sources; psf; psf=nextpsf) {
1744 nextpsf = psf->sf_next;
1745 kfree(psf);
1747 pmc->sources = NULL;
1748 pmc->sfmode = MCAST_EXCLUDE;
1749 pmc->sfcount[MCAST_INCLUDE] = 0;
1750 pmc->sfcount[MCAST_EXCLUDE] = 1;
1755 * Join a multicast group
1757 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1759 int err;
1760 __be32 addr = imr->imr_multiaddr.s_addr;
1761 struct ip_mc_socklist *iml = NULL, *i;
1762 struct in_device *in_dev;
1763 struct inet_sock *inet = inet_sk(sk);
1764 struct net *net = sock_net(sk);
1765 int ifindex;
1766 int count = 0;
1768 if (!ipv4_is_multicast(addr))
1769 return -EINVAL;
1771 rtnl_lock();
1773 in_dev = ip_mc_find_dev(net, imr);
1775 if (!in_dev) {
1776 iml = NULL;
1777 err = -ENODEV;
1778 goto done;
1781 err = -EADDRINUSE;
1782 ifindex = imr->imr_ifindex;
1783 for (i = inet->mc_list; i; i = i->next) {
1784 if (i->multi.imr_multiaddr.s_addr == addr &&
1785 i->multi.imr_ifindex == ifindex)
1786 goto done;
1787 count++;
1789 err = -ENOBUFS;
1790 if (count >= sysctl_igmp_max_memberships)
1791 goto done;
1792 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1793 if (iml == NULL)
1794 goto done;
1796 memcpy(&iml->multi, imr, sizeof(*imr));
1797 iml->next = inet->mc_list;
1798 iml->sflist = NULL;
1799 iml->sfmode = MCAST_EXCLUDE;
1800 rcu_assign_pointer(inet->mc_list, iml);
1801 ip_mc_inc_group(in_dev, addr);
1802 err = 0;
1803 done:
1804 rtnl_unlock();
1805 return err;
1808 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1810 struct ip_sf_socklist *psf;
1812 psf = container_of(rp, struct ip_sf_socklist, rcu);
1813 /* sk_omem_alloc should have been decreased by the caller*/
1814 kfree(psf);
1817 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1818 struct in_device *in_dev)
1820 struct ip_sf_socklist *psf = iml->sflist;
1821 int err;
1823 if (psf == NULL) {
1824 /* any-source empty exclude case */
1825 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1826 iml->sfmode, 0, NULL, 0);
1828 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1829 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1830 rcu_assign_pointer(iml->sflist, NULL);
1831 /* decrease mem now to avoid the memleak warning */
1832 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1833 call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1834 return err;
1838 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1840 struct ip_mc_socklist *iml;
1842 iml = container_of(rp, struct ip_mc_socklist, rcu);
1843 /* sk_omem_alloc should have been decreased by the caller*/
1844 kfree(iml);
1849 * Ask a socket to leave a group.
1852 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1854 struct inet_sock *inet = inet_sk(sk);
1855 struct ip_mc_socklist *iml, **imlp;
1856 struct in_device *in_dev;
1857 struct net *net = sock_net(sk);
1858 __be32 group = imr->imr_multiaddr.s_addr;
1859 u32 ifindex;
1860 int ret = -EADDRNOTAVAIL;
1862 rtnl_lock();
1863 in_dev = ip_mc_find_dev(net, imr);
1864 ifindex = imr->imr_ifindex;
1865 for (imlp = &inet->mc_list; (iml = *imlp) != NULL; imlp = &iml->next) {
1866 if (iml->multi.imr_multiaddr.s_addr != group)
1867 continue;
1868 if (ifindex) {
1869 if (iml->multi.imr_ifindex != ifindex)
1870 continue;
1871 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1872 iml->multi.imr_address.s_addr)
1873 continue;
1875 (void) ip_mc_leave_src(sk, iml, in_dev);
1877 rcu_assign_pointer(*imlp, iml->next);
1879 if (in_dev)
1880 ip_mc_dec_group(in_dev, group);
1881 rtnl_unlock();
1882 /* decrease mem now to avoid the memleak warning */
1883 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1884 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1885 return 0;
1887 if (!in_dev)
1888 ret = -ENODEV;
1889 rtnl_unlock();
1890 return ret;
1893 int ip_mc_source(int add, int omode, struct sock *sk, struct
1894 ip_mreq_source *mreqs, int ifindex)
1896 int err;
1897 struct ip_mreqn imr;
1898 __be32 addr = mreqs->imr_multiaddr;
1899 struct ip_mc_socklist *pmc;
1900 struct in_device *in_dev = NULL;
1901 struct inet_sock *inet = inet_sk(sk);
1902 struct ip_sf_socklist *psl;
1903 struct net *net = sock_net(sk);
1904 int leavegroup = 0;
1905 int i, j, rv;
1907 if (!ipv4_is_multicast(addr))
1908 return -EINVAL;
1910 rtnl_lock();
1912 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1913 imr.imr_address.s_addr = mreqs->imr_interface;
1914 imr.imr_ifindex = ifindex;
1915 in_dev = ip_mc_find_dev(net, &imr);
1917 if (!in_dev) {
1918 err = -ENODEV;
1919 goto done;
1921 err = -EADDRNOTAVAIL;
1923 for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
1924 if ((pmc->multi.imr_multiaddr.s_addr ==
1925 imr.imr_multiaddr.s_addr) &&
1926 (pmc->multi.imr_ifindex == imr.imr_ifindex))
1927 break;
1929 if (!pmc) { /* must have a prior join */
1930 err = -EINVAL;
1931 goto done;
1933 /* if a source filter was set, must be the same mode as before */
1934 if (pmc->sflist) {
1935 if (pmc->sfmode != omode) {
1936 err = -EINVAL;
1937 goto done;
1939 } else if (pmc->sfmode != omode) {
1940 /* allow mode switches for empty-set filters */
1941 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1942 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1943 NULL, 0);
1944 pmc->sfmode = omode;
1947 psl = pmc->sflist;
1948 if (!add) {
1949 if (!psl)
1950 goto done; /* err = -EADDRNOTAVAIL */
1951 rv = !0;
1952 for (i=0; i<psl->sl_count; i++) {
1953 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1954 sizeof(__be32));
1955 if (rv == 0)
1956 break;
1958 if (rv) /* source not found */
1959 goto done; /* err = -EADDRNOTAVAIL */
1961 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1962 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1963 leavegroup = 1;
1964 goto done;
1967 /* update the interface filter */
1968 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1969 &mreqs->imr_sourceaddr, 1);
1971 for (j=i+1; j<psl->sl_count; j++)
1972 psl->sl_addr[j-1] = psl->sl_addr[j];
1973 psl->sl_count--;
1974 err = 0;
1975 goto done;
1977 /* else, add a new source to the filter */
1979 if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1980 err = -ENOBUFS;
1981 goto done;
1983 if (!psl || psl->sl_count == psl->sl_max) {
1984 struct ip_sf_socklist *newpsl;
1985 int count = IP_SFBLOCK;
1987 if (psl)
1988 count += psl->sl_max;
1989 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
1990 if (!newpsl) {
1991 err = -ENOBUFS;
1992 goto done;
1994 newpsl->sl_max = count;
1995 newpsl->sl_count = count - IP_SFBLOCK;
1996 if (psl) {
1997 for (i=0; i<psl->sl_count; i++)
1998 newpsl->sl_addr[i] = psl->sl_addr[i];
1999 /* decrease mem now to avoid the memleak warning */
2000 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2001 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2003 rcu_assign_pointer(pmc->sflist, newpsl);
2004 psl = newpsl;
2006 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2007 for (i=0; i<psl->sl_count; i++) {
2008 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2009 sizeof(__be32));
2010 if (rv == 0)
2011 break;
2013 if (rv == 0) /* address already there is an error */
2014 goto done;
2015 for (j=psl->sl_count-1; j>=i; j--)
2016 psl->sl_addr[j+1] = psl->sl_addr[j];
2017 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2018 psl->sl_count++;
2019 err = 0;
2020 /* update the interface list */
2021 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2022 &mreqs->imr_sourceaddr, 1);
2023 done:
2024 rtnl_unlock();
2025 if (leavegroup)
2026 return ip_mc_leave_group(sk, &imr);
2027 return err;
2030 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2032 int err = 0;
2033 struct ip_mreqn imr;
2034 __be32 addr = msf->imsf_multiaddr;
2035 struct ip_mc_socklist *pmc;
2036 struct in_device *in_dev;
2037 struct inet_sock *inet = inet_sk(sk);
2038 struct ip_sf_socklist *newpsl, *psl;
2039 struct net *net = sock_net(sk);
2040 int leavegroup = 0;
2042 if (!ipv4_is_multicast(addr))
2043 return -EINVAL;
2044 if (msf->imsf_fmode != MCAST_INCLUDE &&
2045 msf->imsf_fmode != MCAST_EXCLUDE)
2046 return -EINVAL;
2048 rtnl_lock();
2050 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2051 imr.imr_address.s_addr = msf->imsf_interface;
2052 imr.imr_ifindex = ifindex;
2053 in_dev = ip_mc_find_dev(net, &imr);
2055 if (!in_dev) {
2056 err = -ENODEV;
2057 goto done;
2060 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2061 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2062 leavegroup = 1;
2063 goto done;
2066 for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
2067 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2068 pmc->multi.imr_ifindex == imr.imr_ifindex)
2069 break;
2071 if (!pmc) { /* must have a prior join */
2072 err = -EINVAL;
2073 goto done;
2075 if (msf->imsf_numsrc) {
2076 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2077 GFP_KERNEL);
2078 if (!newpsl) {
2079 err = -ENOBUFS;
2080 goto done;
2082 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2083 memcpy(newpsl->sl_addr, msf->imsf_slist,
2084 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2085 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2086 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2087 if (err) {
2088 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2089 goto done;
2091 } else {
2092 newpsl = NULL;
2093 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2094 msf->imsf_fmode, 0, NULL, 0);
2096 psl = pmc->sflist;
2097 if (psl) {
2098 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2099 psl->sl_count, psl->sl_addr, 0);
2100 /* decrease mem now to avoid the memleak warning */
2101 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2102 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2103 } else
2104 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2105 0, NULL, 0);
2106 rcu_assign_pointer(pmc->sflist, newpsl);
2107 pmc->sfmode = msf->imsf_fmode;
2108 err = 0;
2109 done:
2110 rtnl_unlock();
2111 if (leavegroup)
2112 err = ip_mc_leave_group(sk, &imr);
2113 return err;
2116 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2117 struct ip_msfilter __user *optval, int __user *optlen)
2119 int err, len, count, copycount;
2120 struct ip_mreqn imr;
2121 __be32 addr = msf->imsf_multiaddr;
2122 struct ip_mc_socklist *pmc;
2123 struct in_device *in_dev;
2124 struct inet_sock *inet = inet_sk(sk);
2125 struct ip_sf_socklist *psl;
2126 struct net *net = sock_net(sk);
2128 if (!ipv4_is_multicast(addr))
2129 return -EINVAL;
2131 rtnl_lock();
2133 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2134 imr.imr_address.s_addr = msf->imsf_interface;
2135 imr.imr_ifindex = 0;
2136 in_dev = ip_mc_find_dev(net, &imr);
2138 if (!in_dev) {
2139 err = -ENODEV;
2140 goto done;
2142 err = -EADDRNOTAVAIL;
2144 for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
2145 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2146 pmc->multi.imr_ifindex == imr.imr_ifindex)
2147 break;
2149 if (!pmc) /* must have a prior join */
2150 goto done;
2151 msf->imsf_fmode = pmc->sfmode;
2152 psl = pmc->sflist;
2153 rtnl_unlock();
2154 if (!psl) {
2155 len = 0;
2156 count = 0;
2157 } else {
2158 count = psl->sl_count;
2160 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2161 len = copycount * sizeof(psl->sl_addr[0]);
2162 msf->imsf_numsrc = count;
2163 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2164 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2165 return -EFAULT;
2167 if (len &&
2168 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2169 return -EFAULT;
2170 return 0;
2171 done:
2172 rtnl_unlock();
2173 return err;
2176 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2177 struct group_filter __user *optval, int __user *optlen)
2179 int err, i, count, copycount;
2180 struct sockaddr_in *psin;
2181 __be32 addr;
2182 struct ip_mc_socklist *pmc;
2183 struct inet_sock *inet = inet_sk(sk);
2184 struct ip_sf_socklist *psl;
2186 psin = (struct sockaddr_in *)&gsf->gf_group;
2187 if (psin->sin_family != AF_INET)
2188 return -EINVAL;
2189 addr = psin->sin_addr.s_addr;
2190 if (!ipv4_is_multicast(addr))
2191 return -EINVAL;
2193 rtnl_lock();
2195 err = -EADDRNOTAVAIL;
2197 for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
2198 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2199 pmc->multi.imr_ifindex == gsf->gf_interface)
2200 break;
2202 if (!pmc) /* must have a prior join */
2203 goto done;
2204 gsf->gf_fmode = pmc->sfmode;
2205 psl = pmc->sflist;
2206 rtnl_unlock();
2207 count = psl ? psl->sl_count : 0;
2208 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2209 gsf->gf_numsrc = count;
2210 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2211 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2212 return -EFAULT;
2214 for (i=0; i<copycount; i++) {
2215 struct sockaddr_storage ss;
2217 psin = (struct sockaddr_in *)&ss;
2218 memset(&ss, 0, sizeof(ss));
2219 psin->sin_family = AF_INET;
2220 psin->sin_addr.s_addr = psl->sl_addr[i];
2221 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2222 return -EFAULT;
2224 return 0;
2225 done:
2226 rtnl_unlock();
2227 return err;
2231 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2233 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2235 struct inet_sock *inet = inet_sk(sk);
2236 struct ip_mc_socklist *pmc;
2237 struct ip_sf_socklist *psl;
2238 int i;
2239 int ret;
2241 ret = 1;
2242 if (!ipv4_is_multicast(loc_addr))
2243 goto out;
2245 rcu_read_lock();
2246 for (pmc=rcu_dereference(inet->mc_list); pmc; pmc=rcu_dereference(pmc->next)) {
2247 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2248 pmc->multi.imr_ifindex == dif)
2249 break;
2251 ret = inet->mc_all;
2252 if (!pmc)
2253 goto unlock;
2254 psl = pmc->sflist;
2255 ret = (pmc->sfmode == MCAST_EXCLUDE);
2256 if (!psl)
2257 goto unlock;
2259 for (i=0; i<psl->sl_count; i++) {
2260 if (psl->sl_addr[i] == rmt_addr)
2261 break;
2263 ret = 0;
2264 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2265 goto unlock;
2266 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2267 goto unlock;
2268 ret = 1;
2269 unlock:
2270 rcu_read_unlock();
2271 out:
2272 return ret;
2276 * A socket is closing.
2279 void ip_mc_drop_socket(struct sock *sk)
2281 struct inet_sock *inet = inet_sk(sk);
2282 struct ip_mc_socklist *iml;
2283 struct net *net = sock_net(sk);
2285 if (inet->mc_list == NULL)
2286 return;
2288 rtnl_lock();
2289 while ((iml = inet->mc_list) != NULL) {
2290 struct in_device *in_dev;
2291 rcu_assign_pointer(inet->mc_list, iml->next);
2293 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2294 (void) ip_mc_leave_src(sk, iml, in_dev);
2295 if (in_dev != NULL) {
2296 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2297 in_dev_put(in_dev);
2299 /* decrease mem now to avoid the memleak warning */
2300 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2301 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2303 rtnl_unlock();
2306 int ip_check_mc(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2308 struct ip_mc_list *im;
2309 struct ip_sf_list *psf;
2310 int rv = 0;
2312 read_lock(&in_dev->mc_list_lock);
2313 for (im=in_dev->mc_list; im; im=im->next) {
2314 if (im->multiaddr == mc_addr)
2315 break;
2317 if (im && proto == IPPROTO_IGMP) {
2318 rv = 1;
2319 } else if (im) {
2320 if (src_addr) {
2321 for (psf=im->sources; psf; psf=psf->sf_next) {
2322 if (psf->sf_inaddr == src_addr)
2323 break;
2325 if (psf)
2326 rv = psf->sf_count[MCAST_INCLUDE] ||
2327 psf->sf_count[MCAST_EXCLUDE] !=
2328 im->sfcount[MCAST_EXCLUDE];
2329 else
2330 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2331 } else
2332 rv = 1; /* unspecified source; tentatively allow */
2334 read_unlock(&in_dev->mc_list_lock);
2335 return rv;
2338 #if defined(CONFIG_PROC_FS)
2339 struct igmp_mc_iter_state {
2340 struct seq_net_private p;
2341 struct net_device *dev;
2342 struct in_device *in_dev;
2345 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2347 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2349 struct net *net = seq_file_net(seq);
2350 struct ip_mc_list *im = NULL;
2351 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2353 state->in_dev = NULL;
2354 for_each_netdev_rcu(net, state->dev) {
2355 struct in_device *in_dev;
2357 in_dev = __in_dev_get_rcu(state->dev);
2358 if (!in_dev)
2359 continue;
2360 read_lock(&in_dev->mc_list_lock);
2361 im = in_dev->mc_list;
2362 if (im) {
2363 state->in_dev = in_dev;
2364 break;
2366 read_unlock(&in_dev->mc_list_lock);
2368 return im;
2371 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2373 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2374 im = im->next;
2375 while (!im) {
2376 if (likely(state->in_dev != NULL))
2377 read_unlock(&state->in_dev->mc_list_lock);
2379 state->dev = next_net_device_rcu(state->dev);
2380 if (!state->dev) {
2381 state->in_dev = NULL;
2382 break;
2384 state->in_dev = __in_dev_get_rcu(state->dev);
2385 if (!state->in_dev)
2386 continue;
2387 read_lock(&state->in_dev->mc_list_lock);
2388 im = 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);
2424 if (likely(state->in_dev != NULL)) {
2425 read_unlock(&state->in_dev->mc_list_lock);
2426 state->in_dev = NULL;
2428 state->dev = NULL;
2429 rcu_read_unlock();
2432 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2434 if (v == SEQ_START_TOKEN)
2435 seq_puts(seq,
2436 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2437 else {
2438 struct ip_mc_list *im = (struct ip_mc_list *)v;
2439 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2440 char *querier;
2441 #ifdef CONFIG_IP_MULTICAST
2442 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2443 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2444 "V3";
2445 #else
2446 querier = "NONE";
2447 #endif
2449 if (state->in_dev->mc_list == im) {
2450 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2451 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2454 seq_printf(seq,
2455 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2456 im->multiaddr, im->users,
2457 im->tm_running, im->tm_running ?
2458 jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2459 im->reporter);
2461 return 0;
2464 static const struct seq_operations igmp_mc_seq_ops = {
2465 .start = igmp_mc_seq_start,
2466 .next = igmp_mc_seq_next,
2467 .stop = igmp_mc_seq_stop,
2468 .show = igmp_mc_seq_show,
2471 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2473 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2474 sizeof(struct igmp_mc_iter_state));
2477 static const struct file_operations igmp_mc_seq_fops = {
2478 .owner = THIS_MODULE,
2479 .open = igmp_mc_seq_open,
2480 .read = seq_read,
2481 .llseek = seq_lseek,
2482 .release = seq_release_net,
2485 struct igmp_mcf_iter_state {
2486 struct seq_net_private p;
2487 struct net_device *dev;
2488 struct in_device *idev;
2489 struct ip_mc_list *im;
2492 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2494 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2496 struct net *net = seq_file_net(seq);
2497 struct ip_sf_list *psf = NULL;
2498 struct ip_mc_list *im = NULL;
2499 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2501 state->idev = NULL;
2502 state->im = NULL;
2503 for_each_netdev_rcu(net, state->dev) {
2504 struct in_device *idev;
2505 idev = __in_dev_get_rcu(state->dev);
2506 if (unlikely(idev == NULL))
2507 continue;
2508 read_lock(&idev->mc_list_lock);
2509 im = idev->mc_list;
2510 if (likely(im != NULL)) {
2511 spin_lock_bh(&im->lock);
2512 psf = im->sources;
2513 if (likely(psf != NULL)) {
2514 state->im = im;
2515 state->idev = idev;
2516 break;
2518 spin_unlock_bh(&im->lock);
2520 read_unlock(&idev->mc_list_lock);
2522 return psf;
2525 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2527 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2529 psf = psf->sf_next;
2530 while (!psf) {
2531 spin_unlock_bh(&state->im->lock);
2532 state->im = state->im->next;
2533 while (!state->im) {
2534 if (likely(state->idev != NULL))
2535 read_unlock(&state->idev->mc_list_lock);
2537 state->dev = next_net_device_rcu(state->dev);
2538 if (!state->dev) {
2539 state->idev = NULL;
2540 goto out;
2542 state->idev = __in_dev_get_rcu(state->dev);
2543 if (!state->idev)
2544 continue;
2545 read_lock(&state->idev->mc_list_lock);
2546 state->im = state->idev->mc_list;
2548 if (!state->im)
2549 break;
2550 spin_lock_bh(&state->im->lock);
2551 psf = state->im->sources;
2553 out:
2554 return psf;
2557 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2559 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2560 if (psf)
2561 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2562 --pos;
2563 return pos ? NULL : psf;
2566 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2567 __acquires(rcu)
2569 rcu_read_lock();
2570 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2573 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2575 struct ip_sf_list *psf;
2576 if (v == SEQ_START_TOKEN)
2577 psf = igmp_mcf_get_first(seq);
2578 else
2579 psf = igmp_mcf_get_next(seq, v);
2580 ++*pos;
2581 return psf;
2584 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2585 __releases(rcu)
2587 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2588 if (likely(state->im != NULL)) {
2589 spin_unlock_bh(&state->im->lock);
2590 state->im = NULL;
2592 if (likely(state->idev != NULL)) {
2593 read_unlock(&state->idev->mc_list_lock);
2594 state->idev = NULL;
2596 state->dev = NULL;
2597 rcu_read_unlock();
2600 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2602 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2603 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2605 if (v == SEQ_START_TOKEN) {
2606 seq_printf(seq,
2607 "%3s %6s "
2608 "%10s %10s %6s %6s\n", "Idx",
2609 "Device", "MCA",
2610 "SRC", "INC", "EXC");
2611 } else {
2612 seq_printf(seq,
2613 "%3d %6.6s 0x%08x "
2614 "0x%08x %6lu %6lu\n",
2615 state->dev->ifindex, state->dev->name,
2616 ntohl(state->im->multiaddr),
2617 ntohl(psf->sf_inaddr),
2618 psf->sf_count[MCAST_INCLUDE],
2619 psf->sf_count[MCAST_EXCLUDE]);
2621 return 0;
2624 static const struct seq_operations igmp_mcf_seq_ops = {
2625 .start = igmp_mcf_seq_start,
2626 .next = igmp_mcf_seq_next,
2627 .stop = igmp_mcf_seq_stop,
2628 .show = igmp_mcf_seq_show,
2631 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2633 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2634 sizeof(struct igmp_mcf_iter_state));
2637 static const struct file_operations igmp_mcf_seq_fops = {
2638 .owner = THIS_MODULE,
2639 .open = igmp_mcf_seq_open,
2640 .read = seq_read,
2641 .llseek = seq_lseek,
2642 .release = seq_release_net,
2645 static int __net_init igmp_net_init(struct net *net)
2647 struct proc_dir_entry *pde;
2649 pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2650 if (!pde)
2651 goto out_igmp;
2652 pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2653 if (!pde)
2654 goto out_mcfilter;
2655 return 0;
2657 out_mcfilter:
2658 proc_net_remove(net, "igmp");
2659 out_igmp:
2660 return -ENOMEM;
2663 static void __net_exit igmp_net_exit(struct net *net)
2665 proc_net_remove(net, "mcfilter");
2666 proc_net_remove(net, "igmp");
2669 static struct pernet_operations igmp_net_ops = {
2670 .init = igmp_net_init,
2671 .exit = igmp_net_exit,
2674 int __init igmp_mc_proc_init(void)
2676 return register_pernet_subsys(&igmp_net_ops);
2678 #endif
2680 EXPORT_SYMBOL(ip_mc_dec_group);
2681 EXPORT_SYMBOL(ip_mc_inc_group);
2682 EXPORT_SYMBOL(ip_mc_join_group);
2683 EXPORT_SYMBOL(ip_mc_rejoin_group);