?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/mm: Add TLB purge to free pmd/pte page interfaces
[linux/fpc-iii.git] / net / ipv4 / igmp.c
blobf2e6e874e4ec3e85cdc54d8b0c2a25b18130cde8
1 /*
2 * Linux NET3: Internet Group Management Protocol [IGMP]
3 *
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.
6 *
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.
10 *
11 * Authors:
12 * Alan Cox <alan@lxorguk.ukuu.org.uk>
13 *
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.
18 *
19 * Fixes:
20 *
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.
33 *
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
71 */
72
73 #include <linux/module.h>
74 #include <linux/slab.h>
75 #include <asm/uaccess.h>
76 #include <linux/types.h>
77 #include <linux/kernel.h>
78 #include <linux/jiffies.h>
79 #include <linux/string.h>
80 #include <linux/socket.h>
81 #include <linux/sockios.h>
82 #include <linux/in.h>
83 #include <linux/inet.h>
84 #include <linux/netdevice.h>
85 #include <linux/skbuff.h>
86 #include <linux/inetdevice.h>
87 #include <linux/igmp.h>
88 #include <linux/if_arp.h>
89 #include <linux/rtnetlink.h>
90 #include <linux/times.h>
91 #include <linux/pkt_sched.h>
92 #include <linux/byteorder/generic.h>
93
94 #include <net/net_namespace.h>
95 #include <net/arp.h>
96 #include <net/ip.h>
97 #include <net/protocol.h>
98 #include <net/route.h>
99 #include <net/sock.h>
100 #include <net/checksum.h>
101 #include <net/inet_common.h>
102 #include <linux/netfilter_ipv4.h>
103 #ifdef CONFIG_IP_MROUTE
104 #include <linux/mroute.h>
105 #endif
106 #ifdef CONFIG_PROC_FS
107 #include <linux/proc_fs.h>
108 #include <linux/seq_file.h>
109 #endif
110
111 #ifdef CONFIG_IP_MULTICAST
112 /* Parameter names and values are taken from igmp-v2-06 draft */
113
114 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT (400*HZ)
115 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT (400*HZ)
116 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL (10*HZ)
117 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL (1*HZ)
118 #define IGMP_QUERY_RESPONSE_INTERVAL (10*HZ)
119 #define IGMP_QUERY_ROBUSTNESS_VARIABLE 2
120
121
122 #define IGMP_INITIAL_REPORT_DELAY (1)
123
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.
129 */
130
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)))
141
142 static int unsolicited_report_interval(struct in_device *in_dev)
143 {
144 int interval_ms, interval_jiffies;
145
146 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
147 interval_ms = IN_DEV_CONF_GET(
148 in_dev,
149 IGMPV2_UNSOLICITED_REPORT_INTERVAL);
150 else /* v3 */
151 interval_ms = IN_DEV_CONF_GET(
152 in_dev,
153 IGMPV3_UNSOLICITED_REPORT_INTERVAL);
154
155 interval_jiffies = msecs_to_jiffies(interval_ms);
156
157 /* _timer functions can't handle a delay of 0 jiffies so ensure
158 * we always return a positive value.
159 */
160 if (interval_jiffies <= 0)
161 interval_jiffies = 1;
162 return interval_jiffies;
163 }
164
165 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
166 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im);
167 static void igmpv3_clear_delrec(struct in_device *in_dev);
168 static int sf_setstate(struct ip_mc_list *pmc);
169 static void sf_markstate(struct ip_mc_list *pmc);
170 #endif
171 static void ip_mc_clear_src(struct ip_mc_list *pmc);
172 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
173 int sfcount, __be32 *psfsrc, int delta);
174
175 static void ip_ma_put(struct ip_mc_list *im)
176 {
177 if (atomic_dec_and_test(&im->refcnt)) {
178 in_dev_put(im->interface);
179 kfree_rcu(im, rcu);
180 }
181 }
182
183 #define for_each_pmc_rcu(in_dev, pmc) \
184 for (pmc = rcu_dereference(in_dev->mc_list); \
185 pmc != NULL; \
186 pmc = rcu_dereference(pmc->next_rcu))
187
188 #define for_each_pmc_rtnl(in_dev, pmc) \
189 for (pmc = rtnl_dereference(in_dev->mc_list); \
190 pmc != NULL; \
191 pmc = rtnl_dereference(pmc->next_rcu))
192
193 #ifdef CONFIG_IP_MULTICAST
194
195 /*
196 * Timer management
197 */
198
199 static void igmp_stop_timer(struct ip_mc_list *im)
200 {
201 spin_lock_bh(&im->lock);
202 if (del_timer(&im->timer))
203 atomic_dec(&im->refcnt);
204 im->tm_running = 0;
205 im->reporter = 0;
206 im->unsolicit_count = 0;
207 spin_unlock_bh(&im->lock);
208 }
209
210 /* It must be called with locked im->lock */
211 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
212 {
213 int tv = prandom_u32() % max_delay;
214
215 im->tm_running = 1;
216 if (!mod_timer(&im->timer, jiffies+tv+2))
217 atomic_inc(&im->refcnt);
218 }
219
220 static void igmp_gq_start_timer(struct in_device *in_dev)
221 {
222 int tv = prandom_u32() % in_dev->mr_maxdelay;
223 unsigned long exp = jiffies + tv + 2;
224
225 if (in_dev->mr_gq_running &&
226 time_after_eq(exp, (in_dev->mr_gq_timer).expires))
227 return;
228
229 in_dev->mr_gq_running = 1;
230 if (!mod_timer(&in_dev->mr_gq_timer, exp))
231 in_dev_hold(in_dev);
232 }
233
234 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
235 {
236 int tv = prandom_u32() % delay;
237
238 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
239 in_dev_hold(in_dev);
240 }
241
242 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
243 {
244 spin_lock_bh(&im->lock);
245 im->unsolicit_count = 0;
246 if (del_timer(&im->timer)) {
247 if ((long)(im->timer.expires-jiffies) < max_delay) {
248 add_timer(&im->timer);
249 im->tm_running = 1;
250 spin_unlock_bh(&im->lock);
251 return;
252 }
253 atomic_dec(&im->refcnt);
254 }
255 igmp_start_timer(im, max_delay);
256 spin_unlock_bh(&im->lock);
257 }
258
259
260 /*
261 * Send an IGMP report.
262 */
263
264 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
265
266
267 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
268 int gdeleted, int sdeleted)
269 {
270 switch (type) {
271 case IGMPV3_MODE_IS_INCLUDE:
272 case IGMPV3_MODE_IS_EXCLUDE:
273 if (gdeleted || sdeleted)
274 return 0;
275 if (!(pmc->gsquery && !psf->sf_gsresp)) {
276 if (pmc->sfmode == MCAST_INCLUDE)
277 return 1;
278 /* don't include if this source is excluded
279 * in all filters
280 */
281 if (psf->sf_count[MCAST_INCLUDE])
282 return type == IGMPV3_MODE_IS_INCLUDE;
283 return pmc->sfcount[MCAST_EXCLUDE] ==
284 psf->sf_count[MCAST_EXCLUDE];
285 }
286 return 0;
287 case IGMPV3_CHANGE_TO_INCLUDE:
288 if (gdeleted || sdeleted)
289 return 0;
290 return psf->sf_count[MCAST_INCLUDE] != 0;
291 case IGMPV3_CHANGE_TO_EXCLUDE:
292 if (gdeleted || sdeleted)
293 return 0;
294 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
295 psf->sf_count[MCAST_INCLUDE])
296 return 0;
297 return pmc->sfcount[MCAST_EXCLUDE] ==
298 psf->sf_count[MCAST_EXCLUDE];
299 case IGMPV3_ALLOW_NEW_SOURCES:
300 if (gdeleted || !psf->sf_crcount)
301 return 0;
302 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
303 case IGMPV3_BLOCK_OLD_SOURCES:
304 if (pmc->sfmode == MCAST_INCLUDE)
305 return gdeleted || (psf->sf_crcount && sdeleted);
306 return psf->sf_crcount && !gdeleted && !sdeleted;
307 }
308 return 0;
309 }
310
311 static int
312 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
313 {
314 struct ip_sf_list *psf;
315 int scount = 0;
316
317 for (psf = pmc->sources; psf; psf = psf->sf_next) {
318 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
319 continue;
320 scount++;
321 }
322 return scount;
323 }
324
325 /* source address selection per RFC 3376 section 4.2.13 */
326 static __be32 igmpv3_get_srcaddr(struct net_device *dev,
327 const struct flowi4 *fl4)
328 {
329 struct in_device *in_dev = __in_dev_get_rcu(dev);
330
331 if (!in_dev)
332 return htonl(INADDR_ANY);
333
334 for_ifa(in_dev) {
335 if (fl4->saddr == ifa->ifa_local)
336 return fl4->saddr;
337 } endfor_ifa(in_dev);
338
339 return htonl(INADDR_ANY);
340 }
341
342 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
343 {
344 struct sk_buff *skb;
345 struct rtable *rt;
346 struct iphdr *pip;
347 struct igmpv3_report *pig;
348 struct net *net = dev_net(dev);
349 struct flowi4 fl4;
350 int hlen = LL_RESERVED_SPACE(dev);
351 int tlen = dev->needed_tailroom;
352 unsigned int size = mtu;
353
354 while (1) {
355 skb = alloc_skb(size + hlen + tlen,
356 GFP_ATOMIC | __GFP_NOWARN);
357 if (skb)
358 break;
359 size >>= 1;
360 if (size < 256)
361 return NULL;
362 }
363 skb->priority = TC_PRIO_CONTROL;
364
365 rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
366 0, 0,
367 IPPROTO_IGMP, 0, dev->ifindex);
368 if (IS_ERR(rt)) {
369 kfree_skb(skb);
370 return NULL;
371 }
372
373 skb_dst_set(skb, &rt->dst);
374 skb->dev = dev;
375
376 skb_reserve(skb, hlen);
377 skb_tailroom_reserve(skb, mtu, tlen);
378
379 skb_reset_network_header(skb);
380 pip = ip_hdr(skb);
381 skb_put(skb, sizeof(struct iphdr) + 4);
382
383 pip->version = 4;
384 pip->ihl = (sizeof(struct iphdr)+4)>>2;
385 pip->tos = 0xc0;
386 pip->frag_off = htons(IP_DF);
387 pip->ttl = 1;
388 pip->daddr = fl4.daddr;
389
390 rcu_read_lock();
391 pip->saddr = igmpv3_get_srcaddr(dev, &fl4);
392 rcu_read_unlock();
393
394 pip->protocol = IPPROTO_IGMP;
395 pip->tot_len = 0; /* filled in later */
396 ip_select_ident(net, skb, NULL);
397 ((u8 *)&pip[1])[0] = IPOPT_RA;
398 ((u8 *)&pip[1])[1] = 4;
399 ((u8 *)&pip[1])[2] = 0;
400 ((u8 *)&pip[1])[3] = 0;
401
402 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
403 skb_put(skb, sizeof(*pig));
404 pig = igmpv3_report_hdr(skb);
405 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
406 pig->resv1 = 0;
407 pig->csum = 0;
408 pig->resv2 = 0;
409 pig->ngrec = 0;
410 return skb;
411 }
412
413 static int igmpv3_sendpack(struct sk_buff *skb)
414 {
415 struct igmphdr *pig = igmp_hdr(skb);
416 const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
417
418 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
419
420 return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
421 }
422
423 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
424 {
425 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
426 }
427
428 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
429 int type, struct igmpv3_grec **ppgr, unsigned int mtu)
430 {
431 struct net_device *dev = pmc->interface->dev;
432 struct igmpv3_report *pih;
433 struct igmpv3_grec *pgr;
434
435 if (!skb) {
436 skb = igmpv3_newpack(dev, mtu);
437 if (!skb)
438 return NULL;
439 }
440 pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
441 pgr->grec_type = type;
442 pgr->grec_auxwords = 0;
443 pgr->grec_nsrcs = 0;
444 pgr->grec_mca = pmc->multiaddr;
445 pih = igmpv3_report_hdr(skb);
446 pih->ngrec = htons(ntohs(pih->ngrec)+1);
447 *ppgr = pgr;
448 return skb;
449 }
450
451 #define AVAILABLE(skb) ((skb) ? skb_availroom(skb) : 0)
452
453 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
454 int type, int gdeleted, int sdeleted)
455 {
456 struct net_device *dev = pmc->interface->dev;
457 struct net *net = dev_net(dev);
458 struct igmpv3_report *pih;
459 struct igmpv3_grec *pgr = NULL;
460 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
461 int scount, stotal, first, isquery, truncate;
462 unsigned int mtu;
463
464 if (pmc->multiaddr == IGMP_ALL_HOSTS)
465 return skb;
466 if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
467 return skb;
468
469 mtu = READ_ONCE(dev->mtu);
470 if (mtu < IPV4_MIN_MTU)
471 return skb;
472
473 isquery = type == IGMPV3_MODE_IS_INCLUDE ||
474 type == IGMPV3_MODE_IS_EXCLUDE;
475 truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
476 type == IGMPV3_CHANGE_TO_EXCLUDE;
477
478 stotal = scount = 0;
479
480 psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
481
482 if (!*psf_list)
483 goto empty_source;
484
485 pih = skb ? igmpv3_report_hdr(skb) : NULL;
486
487 /* EX and TO_EX get a fresh packet, if needed */
488 if (truncate) {
489 if (pih && pih->ngrec &&
490 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
491 if (skb)
492 igmpv3_sendpack(skb);
493 skb = igmpv3_newpack(dev, mtu);
494 }
495 }
496 first = 1;
497 psf_prev = NULL;
498 for (psf = *psf_list; psf; psf = psf_next) {
499 __be32 *psrc;
500
501 psf_next = psf->sf_next;
502
503 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
504 psf_prev = psf;
505 continue;
506 }
507
508 /* Based on RFC3376 5.1. Should not send source-list change
509 * records when there is a filter mode change.
510 */
511 if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) ||
512 (!gdeleted && pmc->crcount)) &&
513 (type == IGMPV3_ALLOW_NEW_SOURCES ||
514 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount)
515 goto decrease_sf_crcount;
516
517 /* clear marks on query responses */
518 if (isquery)
519 psf->sf_gsresp = 0;
520
521 if (AVAILABLE(skb) < sizeof(__be32) +
522 first*sizeof(struct igmpv3_grec)) {
523 if (truncate && !first)
524 break; /* truncate these */
525 if (pgr)
526 pgr->grec_nsrcs = htons(scount);
527 if (skb)
528 igmpv3_sendpack(skb);
529 skb = igmpv3_newpack(dev, mtu);
530 first = 1;
531 scount = 0;
532 }
533 if (first) {
534 skb = add_grhead(skb, pmc, type, &pgr, mtu);
535 first = 0;
536 }
537 if (!skb)
538 return NULL;
539 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
540 *psrc = psf->sf_inaddr;
541 scount++; stotal++;
542 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
543 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
544 decrease_sf_crcount:
545 psf->sf_crcount--;
546 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
547 if (psf_prev)
548 psf_prev->sf_next = psf->sf_next;
549 else
550 *psf_list = psf->sf_next;
551 kfree(psf);
552 continue;
553 }
554 }
555 psf_prev = psf;
556 }
557
558 empty_source:
559 if (!stotal) {
560 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
561 type == IGMPV3_BLOCK_OLD_SOURCES)
562 return skb;
563 if (pmc->crcount || isquery) {
564 /* make sure we have room for group header */
565 if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
566 igmpv3_sendpack(skb);
567 skb = NULL; /* add_grhead will get a new one */
568 }
569 skb = add_grhead(skb, pmc, type, &pgr, mtu);
570 }
571 }
572 if (pgr)
573 pgr->grec_nsrcs = htons(scount);
574
575 if (isquery)
576 pmc->gsquery = 0; /* clear query state on report */
577 return skb;
578 }
579
580 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
581 {
582 struct sk_buff *skb = NULL;
583 struct net *net = dev_net(in_dev->dev);
584 int type;
585
586 if (!pmc) {
587 rcu_read_lock();
588 for_each_pmc_rcu(in_dev, pmc) {
589 if (pmc->multiaddr == IGMP_ALL_HOSTS)
590 continue;
591 if (ipv4_is_local_multicast(pmc->multiaddr) &&
592 !net->ipv4.sysctl_igmp_llm_reports)
593 continue;
594 spin_lock_bh(&pmc->lock);
595 if (pmc->sfcount[MCAST_EXCLUDE])
596 type = IGMPV3_MODE_IS_EXCLUDE;
597 else
598 type = IGMPV3_MODE_IS_INCLUDE;
599 skb = add_grec(skb, pmc, type, 0, 0);
600 spin_unlock_bh(&pmc->lock);
601 }
602 rcu_read_unlock();
603 } else {
604 spin_lock_bh(&pmc->lock);
605 if (pmc->sfcount[MCAST_EXCLUDE])
606 type = IGMPV3_MODE_IS_EXCLUDE;
607 else
608 type = IGMPV3_MODE_IS_INCLUDE;
609 skb = add_grec(skb, pmc, type, 0, 0);
610 spin_unlock_bh(&pmc->lock);
611 }
612 if (!skb)
613 return 0;
614 return igmpv3_sendpack(skb);
615 }
616
617 /*
618 * remove zero-count source records from a source filter list
619 */
620 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
621 {
622 struct ip_sf_list *psf_prev, *psf_next, *psf;
623
624 psf_prev = NULL;
625 for (psf = *ppsf; psf; psf = psf_next) {
626 psf_next = psf->sf_next;
627 if (psf->sf_crcount == 0) {
628 if (psf_prev)
629 psf_prev->sf_next = psf->sf_next;
630 else
631 *ppsf = psf->sf_next;
632 kfree(psf);
633 } else
634 psf_prev = psf;
635 }
636 }
637
638 static void igmpv3_send_cr(struct in_device *in_dev)
639 {
640 struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
641 struct sk_buff *skb = NULL;
642 int type, dtype;
643
644 rcu_read_lock();
645 spin_lock_bh(&in_dev->mc_tomb_lock);
646
647 /* deleted MCA's */
648 pmc_prev = NULL;
649 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
650 pmc_next = pmc->next;
651 if (pmc->sfmode == MCAST_INCLUDE) {
652 type = IGMPV3_BLOCK_OLD_SOURCES;
653 dtype = IGMPV3_BLOCK_OLD_SOURCES;
654 skb = add_grec(skb, pmc, type, 1, 0);
655 skb = add_grec(skb, pmc, dtype, 1, 1);
656 }
657 if (pmc->crcount) {
658 if (pmc->sfmode == MCAST_EXCLUDE) {
659 type = IGMPV3_CHANGE_TO_INCLUDE;
660 skb = add_grec(skb, pmc, type, 1, 0);
661 }
662 pmc->crcount--;
663 if (pmc->crcount == 0) {
664 igmpv3_clear_zeros(&pmc->tomb);
665 igmpv3_clear_zeros(&pmc->sources);
666 }
667 }
668 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
669 if (pmc_prev)
670 pmc_prev->next = pmc_next;
671 else
672 in_dev->mc_tomb = pmc_next;
673 in_dev_put(pmc->interface);
674 kfree(pmc);
675 } else
676 pmc_prev = pmc;
677 }
678 spin_unlock_bh(&in_dev->mc_tomb_lock);
679
680 /* change recs */
681 for_each_pmc_rcu(in_dev, pmc) {
682 spin_lock_bh(&pmc->lock);
683 if (pmc->sfcount[MCAST_EXCLUDE]) {
684 type = IGMPV3_BLOCK_OLD_SOURCES;
685 dtype = IGMPV3_ALLOW_NEW_SOURCES;
686 } else {
687 type = IGMPV3_ALLOW_NEW_SOURCES;
688 dtype = IGMPV3_BLOCK_OLD_SOURCES;
689 }
690 skb = add_grec(skb, pmc, type, 0, 0);
691 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */
692
693 /* filter mode changes */
694 if (pmc->crcount) {
695 if (pmc->sfmode == MCAST_EXCLUDE)
696 type = IGMPV3_CHANGE_TO_EXCLUDE;
697 else
698 type = IGMPV3_CHANGE_TO_INCLUDE;
699 skb = add_grec(skb, pmc, type, 0, 0);
700 pmc->crcount--;
701 }
702 spin_unlock_bh(&pmc->lock);
703 }
704 rcu_read_unlock();
705
706 if (!skb)
707 return;
708 (void) igmpv3_sendpack(skb);
709 }
710
711 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
712 int type)
713 {
714 struct sk_buff *skb;
715 struct iphdr *iph;
716 struct igmphdr *ih;
717 struct rtable *rt;
718 struct net_device *dev = in_dev->dev;
719 struct net *net = dev_net(dev);
720 __be32 group = pmc ? pmc->multiaddr : 0;
721 struct flowi4 fl4;
722 __be32 dst;
723 int hlen, tlen;
724
725 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
726 return igmpv3_send_report(in_dev, pmc);
727
728 if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
729 return 0;
730
731 if (type == IGMP_HOST_LEAVE_MESSAGE)
732 dst = IGMP_ALL_ROUTER;
733 else
734 dst = group;
735
736 rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
737 0, 0,
738 IPPROTO_IGMP, 0, dev->ifindex);
739 if (IS_ERR(rt))
740 return -1;
741
742 hlen = LL_RESERVED_SPACE(dev);
743 tlen = dev->needed_tailroom;
744 skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
745 if (!skb) {
746 ip_rt_put(rt);
747 return -1;
748 }
749 skb->priority = TC_PRIO_CONTROL;
750
751 skb_dst_set(skb, &rt->dst);
752
753 skb_reserve(skb, hlen);
754
755 skb_reset_network_header(skb);
756 iph = ip_hdr(skb);
757 skb_put(skb, sizeof(struct iphdr) + 4);
758
759 iph->version = 4;
760 iph->ihl = (sizeof(struct iphdr)+4)>>2;
761 iph->tos = 0xc0;
762 iph->frag_off = htons(IP_DF);
763 iph->ttl = 1;
764 iph->daddr = dst;
765 iph->saddr = fl4.saddr;
766 iph->protocol = IPPROTO_IGMP;
767 ip_select_ident(net, skb, NULL);
768 ((u8 *)&iph[1])[0] = IPOPT_RA;
769 ((u8 *)&iph[1])[1] = 4;
770 ((u8 *)&iph[1])[2] = 0;
771 ((u8 *)&iph[1])[3] = 0;
772
773 ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
774 ih->type = type;
775 ih->code = 0;
776 ih->csum = 0;
777 ih->group = group;
778 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
779
780 return ip_local_out(net, skb->sk, skb);
781 }
782
783 static void igmp_gq_timer_expire(unsigned long data)
784 {
785 struct in_device *in_dev = (struct in_device *)data;
786
787 in_dev->mr_gq_running = 0;
788 igmpv3_send_report(in_dev, NULL);
789 in_dev_put(in_dev);
790 }
791
792 static void igmp_ifc_timer_expire(unsigned long data)
793 {
794 struct in_device *in_dev = (struct in_device *)data;
795
796 igmpv3_send_cr(in_dev);
797 if (in_dev->mr_ifc_count) {
798 in_dev->mr_ifc_count--;
799 igmp_ifc_start_timer(in_dev,
800 unsolicited_report_interval(in_dev));
801 }
802 in_dev_put(in_dev);
803 }
804
805 static void igmp_ifc_event(struct in_device *in_dev)
806 {
807 struct net *net = dev_net(in_dev->dev);
808 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
809 return;
810 in_dev->mr_ifc_count = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
811 igmp_ifc_start_timer(in_dev, 1);
812 }
813
814
815 static void igmp_timer_expire(unsigned long data)
816 {
817 struct ip_mc_list *im = (struct ip_mc_list *)data;
818 struct in_device *in_dev = im->interface;
819
820 spin_lock(&im->lock);
821 im->tm_running = 0;
822
823 if (im->unsolicit_count) {
824 im->unsolicit_count--;
825 igmp_start_timer(im, unsolicited_report_interval(in_dev));
826 }
827 im->reporter = 1;
828 spin_unlock(&im->lock);
829
830 if (IGMP_V1_SEEN(in_dev))
831 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
832 else if (IGMP_V2_SEEN(in_dev))
833 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
834 else
835 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
836
837 ip_ma_put(im);
838 }
839
840 /* mark EXCLUDE-mode sources */
841 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
842 {
843 struct ip_sf_list *psf;
844 int i, scount;
845
846 scount = 0;
847 for (psf = pmc->sources; psf; psf = psf->sf_next) {
848 if (scount == nsrcs)
849 break;
850 for (i = 0; i < nsrcs; i++) {
851 /* skip inactive filters */
852 if (psf->sf_count[MCAST_INCLUDE] ||
853 pmc->sfcount[MCAST_EXCLUDE] !=
854 psf->sf_count[MCAST_EXCLUDE])
855 break;
856 if (srcs[i] == psf->sf_inaddr) {
857 scount++;
858 break;
859 }
860 }
861 }
862 pmc->gsquery = 0;
863 if (scount == nsrcs) /* all sources excluded */
864 return 0;
865 return 1;
866 }
867
868 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
869 {
870 struct ip_sf_list *psf;
871 int i, scount;
872
873 if (pmc->sfmode == MCAST_EXCLUDE)
874 return igmp_xmarksources(pmc, nsrcs, srcs);
875
876 /* mark INCLUDE-mode sources */
877 scount = 0;
878 for (psf = pmc->sources; psf; psf = psf->sf_next) {
879 if (scount == nsrcs)
880 break;
881 for (i = 0; i < nsrcs; i++)
882 if (srcs[i] == psf->sf_inaddr) {
883 psf->sf_gsresp = 1;
884 scount++;
885 break;
886 }
887 }
888 if (!scount) {
889 pmc->gsquery = 0;
890 return 0;
891 }
892 pmc->gsquery = 1;
893 return 1;
894 }
895
896 /* return true if packet was dropped */
897 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
898 {
899 struct ip_mc_list *im;
900 struct net *net = dev_net(in_dev->dev);
901
902 /* Timers are only set for non-local groups */
903
904 if (group == IGMP_ALL_HOSTS)
905 return false;
906 if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
907 return false;
908
909 rcu_read_lock();
910 for_each_pmc_rcu(in_dev, im) {
911 if (im->multiaddr == group) {
912 igmp_stop_timer(im);
913 break;
914 }
915 }
916 rcu_read_unlock();
917 return false;
918 }
919
920 /* return true if packet was dropped */
921 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
922 int len)
923 {
924 struct igmphdr *ih = igmp_hdr(skb);
925 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
926 struct ip_mc_list *im;
927 __be32 group = ih->group;
928 int max_delay;
929 int mark = 0;
930 struct net *net = dev_net(in_dev->dev);
931
932
933 if (len == 8) {
934 if (ih->code == 0) {
935 /* Alas, old v1 router presents here. */
936
937 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
938 in_dev->mr_v1_seen = jiffies +
939 IGMP_V1_ROUTER_PRESENT_TIMEOUT;
940 group = 0;
941 } else {
942 /* v2 router present */
943 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
944 in_dev->mr_v2_seen = jiffies +
945 IGMP_V2_ROUTER_PRESENT_TIMEOUT;
946 }
947 /* cancel the interface change timer */
948 in_dev->mr_ifc_count = 0;
949 if (del_timer(&in_dev->mr_ifc_timer))
950 __in_dev_put(in_dev);
951 /* clear deleted report items */
952 igmpv3_clear_delrec(in_dev);
953 } else if (len < 12) {
954 return true; /* ignore bogus packet; freed by caller */
955 } else if (IGMP_V1_SEEN(in_dev)) {
956 /* This is a v3 query with v1 queriers present */
957 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
958 group = 0;
959 } else if (IGMP_V2_SEEN(in_dev)) {
960 /* this is a v3 query with v2 queriers present;
961 * Interpretation of the max_delay code is problematic here.
962 * A real v2 host would use ih_code directly, while v3 has a
963 * different encoding. We use the v3 encoding as more likely
964 * to be intended in a v3 query.
965 */
966 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
967 if (!max_delay)
968 max_delay = 1; /* can't mod w/ 0 */
969 } else { /* v3 */
970 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
971 return true;
972
973 ih3 = igmpv3_query_hdr(skb);
974 if (ih3->nsrcs) {
975 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
976 + ntohs(ih3->nsrcs)*sizeof(__be32)))
977 return true;
978 ih3 = igmpv3_query_hdr(skb);
979 }
980
981 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
982 if (!max_delay)
983 max_delay = 1; /* can't mod w/ 0 */
984 in_dev->mr_maxdelay = max_delay;
985 if (ih3->qrv)
986 in_dev->mr_qrv = ih3->qrv;
987 if (!group) { /* general query */
988 if (ih3->nsrcs)
989 return true; /* no sources allowed */
990 igmp_gq_start_timer(in_dev);
991 return false;
992 }
993 /* mark sources to include, if group & source-specific */
994 mark = ih3->nsrcs != 0;
995 }
996
997 /*
998 * - Start the timers in all of our membership records
999 * that the query applies to for the interface on
1000 * which the query arrived excl. those that belong
1001 * to a "local" group (224.0.0.X)
1002 * - For timers already running check if they need to
1003 * be reset.
1004 * - Use the igmp->igmp_code field as the maximum
1005 * delay possible
1006 */
1007 rcu_read_lock();
1008 for_each_pmc_rcu(in_dev, im) {
1009 int changed;
1010
1011 if (group && group != im->multiaddr)
1012 continue;
1013 if (im->multiaddr == IGMP_ALL_HOSTS)
1014 continue;
1015 if (ipv4_is_local_multicast(im->multiaddr) &&
1016 !net->ipv4.sysctl_igmp_llm_reports)
1017 continue;
1018 spin_lock_bh(&im->lock);
1019 if (im->tm_running)
1020 im->gsquery = im->gsquery && mark;
1021 else
1022 im->gsquery = mark;
1023 changed = !im->gsquery ||
1024 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
1025 spin_unlock_bh(&im->lock);
1026 if (changed)
1027 igmp_mod_timer(im, max_delay);
1028 }
1029 rcu_read_unlock();
1030 return false;
1031 }
1032
1033 /* called in rcu_read_lock() section */
1034 int igmp_rcv(struct sk_buff *skb)
1035 {
1036 /* This basically follows the spec line by line -- see RFC1112 */
1037 struct igmphdr *ih;
1038 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1039 int len = skb->len;
1040 bool dropped = true;
1041
1042 if (!in_dev)
1043 goto drop;
1044
1045 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1046 goto drop;
1047
1048 if (skb_checksum_simple_validate(skb))
1049 goto drop;
1050
1051 ih = igmp_hdr(skb);
1052 switch (ih->type) {
1053 case IGMP_HOST_MEMBERSHIP_QUERY:
1054 dropped = igmp_heard_query(in_dev, skb, len);
1055 break;
1056 case IGMP_HOST_MEMBERSHIP_REPORT:
1057 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1058 /* Is it our report looped back? */
1059 if (rt_is_output_route(skb_rtable(skb)))
1060 break;
1061 /* don't rely on MC router hearing unicast reports */
1062 if (skb->pkt_type == PACKET_MULTICAST ||
1063 skb->pkt_type == PACKET_BROADCAST)
1064 dropped = igmp_heard_report(in_dev, ih->group);
1065 break;
1066 case IGMP_PIM:
1067 #ifdef CONFIG_IP_PIMSM_V1
1068 return pim_rcv_v1(skb);
1069 #endif
1070 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1071 case IGMP_DVMRP:
1072 case IGMP_TRACE:
1073 case IGMP_HOST_LEAVE_MESSAGE:
1074 case IGMP_MTRACE:
1075 case IGMP_MTRACE_RESP:
1076 break;
1077 default:
1078 break;
1079 }
1080
1081 drop:
1082 if (dropped)
1083 kfree_skb(skb);
1084 else
1085 consume_skb(skb);
1086 return 0;
1087 }
1088
1089 #endif
1090
1091
1092 /*
1093 * Add a filter to a device
1094 */
1095
1096 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1097 {
1098 char buf[MAX_ADDR_LEN];
1099 struct net_device *dev = in_dev->dev;
1100
1101 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1102 We will get multicast token leakage, when IFF_MULTICAST
1103 is changed. This check should be done in ndo_set_rx_mode
1104 routine. Something sort of:
1105 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1106 --ANK
1107 */
1108 if (arp_mc_map(addr, buf, dev, 0) == 0)
1109 dev_mc_add(dev, buf);
1110 }
1111
1112 /*
1113 * Remove a filter from a device
1114 */
1115
1116 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1117 {
1118 char buf[MAX_ADDR_LEN];
1119 struct net_device *dev = in_dev->dev;
1120
1121 if (arp_mc_map(addr, buf, dev, 0) == 0)
1122 dev_mc_del(dev, buf);
1123 }
1124
1125 #ifdef CONFIG_IP_MULTICAST
1126 /*
1127 * deleted ip_mc_list manipulation
1128 */
1129 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1130 {
1131 struct ip_mc_list *pmc;
1132 struct net *net = dev_net(in_dev->dev);
1133
1134 /* this is an "ip_mc_list" for convenience; only the fields below
1135 * are actually used. In particular, the refcnt and users are not
1136 * used for management of the delete list. Using the same structure
1137 * for deleted items allows change reports to use common code with
1138 * non-deleted or query-response MCA's.
1139 */
1140 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1141 if (!pmc)
1142 return;
1143 spin_lock_init(&pmc->lock);
1144 spin_lock_bh(&im->lock);
1145 pmc->interface = im->interface;
1146 in_dev_hold(in_dev);
1147 pmc->multiaddr = im->multiaddr;
1148 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1149 pmc->sfmode = im->sfmode;
1150 if (pmc->sfmode == MCAST_INCLUDE) {
1151 struct ip_sf_list *psf;
1152
1153 pmc->tomb = im->tomb;
1154 pmc->sources = im->sources;
1155 im->tomb = im->sources = NULL;
1156 for (psf = pmc->sources; psf; psf = psf->sf_next)
1157 psf->sf_crcount = pmc->crcount;
1158 }
1159 spin_unlock_bh(&im->lock);
1160
1161 spin_lock_bh(&in_dev->mc_tomb_lock);
1162 pmc->next = in_dev->mc_tomb;
1163 in_dev->mc_tomb = pmc;
1164 spin_unlock_bh(&in_dev->mc_tomb_lock);
1165 }
1166
1167 /*
1168 * restore ip_mc_list deleted records
1169 */
1170 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1171 {
1172 struct ip_mc_list *pmc, *pmc_prev;
1173 struct ip_sf_list *psf;
1174 struct net *net = dev_net(in_dev->dev);
1175 __be32 multiaddr = im->multiaddr;
1176
1177 spin_lock_bh(&in_dev->mc_tomb_lock);
1178 pmc_prev = NULL;
1179 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1180 if (pmc->multiaddr == multiaddr)
1181 break;
1182 pmc_prev = pmc;
1183 }
1184 if (pmc) {
1185 if (pmc_prev)
1186 pmc_prev->next = pmc->next;
1187 else
1188 in_dev->mc_tomb = pmc->next;
1189 }
1190 spin_unlock_bh(&in_dev->mc_tomb_lock);
1191
1192 spin_lock_bh(&im->lock);
1193 if (pmc) {
1194 im->interface = pmc->interface;
1195 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1196 if (im->sfmode == MCAST_INCLUDE) {
1197 im->tomb = pmc->tomb;
1198 im->sources = pmc->sources;
1199 for (psf = im->sources; psf; psf = psf->sf_next)
1200 psf->sf_crcount = im->crcount;
1201 }
1202 in_dev_put(pmc->interface);
1203 kfree(pmc);
1204 }
1205 spin_unlock_bh(&im->lock);
1206 }
1207
1208 /*
1209 * flush ip_mc_list deleted records
1210 */
1211 static void igmpv3_clear_delrec(struct in_device *in_dev)
1212 {
1213 struct ip_mc_list *pmc, *nextpmc;
1214
1215 spin_lock_bh(&in_dev->mc_tomb_lock);
1216 pmc = in_dev->mc_tomb;
1217 in_dev->mc_tomb = NULL;
1218 spin_unlock_bh(&in_dev->mc_tomb_lock);
1219
1220 for (; pmc; pmc = nextpmc) {
1221 nextpmc = pmc->next;
1222 ip_mc_clear_src(pmc);
1223 in_dev_put(pmc->interface);
1224 kfree(pmc);
1225 }
1226 /* clear dead sources, too */
1227 rcu_read_lock();
1228 for_each_pmc_rcu(in_dev, pmc) {
1229 struct ip_sf_list *psf, *psf_next;
1230
1231 spin_lock_bh(&pmc->lock);
1232 psf = pmc->tomb;
1233 pmc->tomb = NULL;
1234 spin_unlock_bh(&pmc->lock);
1235 for (; psf; psf = psf_next) {
1236 psf_next = psf->sf_next;
1237 kfree(psf);
1238 }
1239 }
1240 rcu_read_unlock();
1241 }
1242 #endif
1243
1244 static void igmp_group_dropped(struct ip_mc_list *im)
1245 {
1246 struct in_device *in_dev = im->interface;
1247 #ifdef CONFIG_IP_MULTICAST
1248 struct net *net = dev_net(in_dev->dev);
1249 int reporter;
1250 #endif
1251
1252 if (im->loaded) {
1253 im->loaded = 0;
1254 ip_mc_filter_del(in_dev, im->multiaddr);
1255 }
1256
1257 #ifdef CONFIG_IP_MULTICAST
1258 if (im->multiaddr == IGMP_ALL_HOSTS)
1259 return;
1260 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1261 return;
1262
1263 reporter = im->reporter;
1264 igmp_stop_timer(im);
1265
1266 if (!in_dev->dead) {
1267 if (IGMP_V1_SEEN(in_dev))
1268 return;
1269 if (IGMP_V2_SEEN(in_dev)) {
1270 if (reporter)
1271 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1272 return;
1273 }
1274 /* IGMPv3 */
1275 igmpv3_add_delrec(in_dev, im);
1276
1277 igmp_ifc_event(in_dev);
1278 }
1279 #endif
1280 }
1281
1282 static void igmp_group_added(struct ip_mc_list *im)
1283 {
1284 struct in_device *in_dev = im->interface;
1285 #ifdef CONFIG_IP_MULTICAST
1286 struct net *net = dev_net(in_dev->dev);
1287 #endif
1288
1289 if (im->loaded == 0) {
1290 im->loaded = 1;
1291 ip_mc_filter_add(in_dev, im->multiaddr);
1292 }
1293
1294 #ifdef CONFIG_IP_MULTICAST
1295 if (im->multiaddr == IGMP_ALL_HOSTS)
1296 return;
1297 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1298 return;
1299
1300 if (in_dev->dead)
1301 return;
1302 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1303 spin_lock_bh(&im->lock);
1304 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1305 spin_unlock_bh(&im->lock);
1306 return;
1307 }
1308 /* else, v3 */
1309
1310 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1311 igmp_ifc_event(in_dev);
1312 #endif
1313 }
1314
1315
1316 /*
1317 * Multicast list managers
1318 */
1319
1320 static u32 ip_mc_hash(const struct ip_mc_list *im)
1321 {
1322 return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1323 }
1324
1325 static void ip_mc_hash_add(struct in_device *in_dev,
1326 struct ip_mc_list *im)
1327 {
1328 struct ip_mc_list __rcu **mc_hash;
1329 u32 hash;
1330
1331 mc_hash = rtnl_dereference(in_dev->mc_hash);
1332 if (mc_hash) {
1333 hash = ip_mc_hash(im);
1334 im->next_hash = mc_hash[hash];
1335 rcu_assign_pointer(mc_hash[hash], im);
1336 return;
1337 }
1338
1339 /* do not use a hash table for small number of items */
1340 if (in_dev->mc_count < 4)
1341 return;
1342
1343 mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1344 GFP_KERNEL);
1345 if (!mc_hash)
1346 return;
1347
1348 for_each_pmc_rtnl(in_dev, im) {
1349 hash = ip_mc_hash(im);
1350 im->next_hash = mc_hash[hash];
1351 RCU_INIT_POINTER(mc_hash[hash], im);
1352 }
1353
1354 rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1355 }
1356
1357 static void ip_mc_hash_remove(struct in_device *in_dev,
1358 struct ip_mc_list *im)
1359 {
1360 struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1361 struct ip_mc_list *aux;
1362
1363 if (!mc_hash)
1364 return;
1365 mc_hash += ip_mc_hash(im);
1366 while ((aux = rtnl_dereference(*mc_hash)) != im)
1367 mc_hash = &aux->next_hash;
1368 *mc_hash = im->next_hash;
1369 }
1370
1371
1372 /*
1373 * A socket has joined a multicast group on device dev.
1374 */
1375
1376 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1377 {
1378 struct ip_mc_list *im;
1379 #ifdef CONFIG_IP_MULTICAST
1380 struct net *net = dev_net(in_dev->dev);
1381 #endif
1382
1383 ASSERT_RTNL();
1384
1385 for_each_pmc_rtnl(in_dev, im) {
1386 if (im->multiaddr == addr) {
1387 im->users++;
1388 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1389 goto out;
1390 }
1391 }
1392
1393 im = kzalloc(sizeof(*im), GFP_KERNEL);
1394 if (!im)
1395 goto out;
1396
1397 im->users = 1;
1398 im->interface = in_dev;
1399 in_dev_hold(in_dev);
1400 im->multiaddr = addr;
1401 /* initial mode is (EX, empty) */
1402 im->sfmode = MCAST_EXCLUDE;
1403 im->sfcount[MCAST_EXCLUDE] = 1;
1404 atomic_set(&im->refcnt, 1);
1405 spin_lock_init(&im->lock);
1406 #ifdef CONFIG_IP_MULTICAST
1407 setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1408 im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1409 #endif
1410
1411 im->next_rcu = in_dev->mc_list;
1412 in_dev->mc_count++;
1413 rcu_assign_pointer(in_dev->mc_list, im);
1414
1415 ip_mc_hash_add(in_dev, im);
1416
1417 #ifdef CONFIG_IP_MULTICAST
1418 igmpv3_del_delrec(in_dev, im);
1419 #endif
1420 igmp_group_added(im);
1421 if (!in_dev->dead)
1422 ip_rt_multicast_event(in_dev);
1423 out:
1424 return;
1425 }
1426 EXPORT_SYMBOL(ip_mc_inc_group);
1427
1428 static int ip_mc_check_iphdr(struct sk_buff *skb)
1429 {
1430 const struct iphdr *iph;
1431 unsigned int len;
1432 unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1433
1434 if (!pskb_may_pull(skb, offset))
1435 return -EINVAL;
1436
1437 iph = ip_hdr(skb);
1438
1439 if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1440 return -EINVAL;
1441
1442 offset += ip_hdrlen(skb) - sizeof(*iph);
1443
1444 if (!pskb_may_pull(skb, offset))
1445 return -EINVAL;
1446
1447 iph = ip_hdr(skb);
1448
1449 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1450 return -EINVAL;
1451
1452 len = skb_network_offset(skb) + ntohs(iph->tot_len);
1453 if (skb->len < len || len < offset)
1454 return -EINVAL;
1455
1456 skb_set_transport_header(skb, offset);
1457
1458 return 0;
1459 }
1460
1461 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1462 {
1463 unsigned int len = skb_transport_offset(skb);
1464
1465 len += sizeof(struct igmpv3_report);
1466
1467 return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1468 }
1469
1470 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1471 {
1472 unsigned int len = skb_transport_offset(skb);
1473
1474 len += sizeof(struct igmphdr);
1475 if (skb->len < len)
1476 return -EINVAL;
1477
1478 /* IGMPv{1,2}? */
1479 if (skb->len != len) {
1480 /* or IGMPv3? */
1481 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1482 if (skb->len < len || !pskb_may_pull(skb, len))
1483 return -EINVAL;
1484 }
1485
1486 /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1487 * all-systems destination addresses (224.0.0.1) for general queries
1488 */
1489 if (!igmp_hdr(skb)->group &&
1490 ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1491 return -EINVAL;
1492
1493 return 0;
1494 }
1495
1496 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1497 {
1498 switch (igmp_hdr(skb)->type) {
1499 case IGMP_HOST_LEAVE_MESSAGE:
1500 case IGMP_HOST_MEMBERSHIP_REPORT:
1501 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1502 /* fall through */
1503 return 0;
1504 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1505 return ip_mc_check_igmp_reportv3(skb);
1506 case IGMP_HOST_MEMBERSHIP_QUERY:
1507 return ip_mc_check_igmp_query(skb);
1508 default:
1509 return -ENOMSG;
1510 }
1511 }
1512
1513 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1514 {
1515 return skb_checksum_simple_validate(skb);
1516 }
1517
1518 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1519
1520 {
1521 struct sk_buff *skb_chk;
1522 unsigned int transport_len;
1523 unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1524 int ret = -EINVAL;
1525
1526 transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1527
1528 skb_chk = skb_checksum_trimmed(skb, transport_len,
1529 ip_mc_validate_checksum);
1530 if (!skb_chk)
1531 goto err;
1532
1533 if (!pskb_may_pull(skb_chk, len))
1534 goto err;
1535
1536 ret = ip_mc_check_igmp_msg(skb_chk);
1537 if (ret)
1538 goto err;
1539
1540 if (skb_trimmed)
1541 *skb_trimmed = skb_chk;
1542 /* free now unneeded clone */
1543 else if (skb_chk != skb)
1544 kfree_skb(skb_chk);
1545
1546 ret = 0;
1547
1548 err:
1549 if (ret && skb_chk && skb_chk != skb)
1550 kfree_skb(skb_chk);
1551
1552 return ret;
1553 }
1554
1555 /**
1556 * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1557 * @skb: the skb to validate
1558 * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1559 *
1560 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1561 * skb transport header accordingly and returns zero.
1562 *
1563 * -EINVAL: A broken packet was detected, i.e. it violates some internet
1564 * standard
1565 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1566 * -ENOMEM: A memory allocation failure happened.
1567 *
1568 * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1569 * to NULL): After parsing an IGMP packet successfully it will point to
1570 * an skb which has its tail aligned to the IP packet end. This might
1571 * either be the originally provided skb or a trimmed, cloned version if
1572 * the skb frame had data beyond the IP packet. A cloned skb allows us
1573 * to leave the original skb and its full frame unchanged (which might be
1574 * desirable for layer 2 frame jugglers).
1575 *
1576 * Caller needs to set the skb network header and free any returned skb if it
1577 * differs from the provided skb.
1578 */
1579 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1580 {
1581 int ret = ip_mc_check_iphdr(skb);
1582
1583 if (ret < 0)
1584 return ret;
1585
1586 if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1587 return -ENOMSG;
1588
1589 return __ip_mc_check_igmp(skb, skb_trimmed);
1590 }
1591 EXPORT_SYMBOL(ip_mc_check_igmp);
1592
1593 /*
1594 * Resend IGMP JOIN report; used by netdev notifier.
1595 */
1596 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1597 {
1598 #ifdef CONFIG_IP_MULTICAST
1599 struct ip_mc_list *im;
1600 int type;
1601 struct net *net = dev_net(in_dev->dev);
1602
1603 ASSERT_RTNL();
1604
1605 for_each_pmc_rtnl(in_dev, im) {
1606 if (im->multiaddr == IGMP_ALL_HOSTS)
1607 continue;
1608 if (ipv4_is_local_multicast(im->multiaddr) &&
1609 !net->ipv4.sysctl_igmp_llm_reports)
1610 continue;
1611
1612 /* a failover is happening and switches
1613 * must be notified immediately
1614 */
1615 if (IGMP_V1_SEEN(in_dev))
1616 type = IGMP_HOST_MEMBERSHIP_REPORT;
1617 else if (IGMP_V2_SEEN(in_dev))
1618 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1619 else
1620 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1621 igmp_send_report(in_dev, im, type);
1622 }
1623 #endif
1624 }
1625
1626 /*
1627 * A socket has left a multicast group on device dev
1628 */
1629
1630 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1631 {
1632 struct ip_mc_list *i;
1633 struct ip_mc_list __rcu **ip;
1634
1635 ASSERT_RTNL();
1636
1637 for (ip = &in_dev->mc_list;
1638 (i = rtnl_dereference(*ip)) != NULL;
1639 ip = &i->next_rcu) {
1640 if (i->multiaddr == addr) {
1641 if (--i->users == 0) {
1642 ip_mc_hash_remove(in_dev, i);
1643 *ip = i->next_rcu;
1644 in_dev->mc_count--;
1645 igmp_group_dropped(i);
1646 ip_mc_clear_src(i);
1647
1648 if (!in_dev->dead)
1649 ip_rt_multicast_event(in_dev);
1650
1651 ip_ma_put(i);
1652 return;
1653 }
1654 break;
1655 }
1656 }
1657 }
1658 EXPORT_SYMBOL(ip_mc_dec_group);
1659
1660 /* Device changing type */
1661
1662 void ip_mc_unmap(struct in_device *in_dev)
1663 {
1664 struct ip_mc_list *pmc;
1665
1666 ASSERT_RTNL();
1667
1668 for_each_pmc_rtnl(in_dev, pmc)
1669 igmp_group_dropped(pmc);
1670 }
1671
1672 void ip_mc_remap(struct in_device *in_dev)
1673 {
1674 struct ip_mc_list *pmc;
1675
1676 ASSERT_RTNL();
1677
1678 for_each_pmc_rtnl(in_dev, pmc) {
1679 #ifdef CONFIG_IP_MULTICAST
1680 igmpv3_del_delrec(in_dev, pmc);
1681 #endif
1682 igmp_group_added(pmc);
1683 }
1684 }
1685
1686 /* Device going down */
1687
1688 void ip_mc_down(struct in_device *in_dev)
1689 {
1690 struct ip_mc_list *pmc;
1691
1692 ASSERT_RTNL();
1693
1694 for_each_pmc_rtnl(in_dev, pmc)
1695 igmp_group_dropped(pmc);
1696
1697 #ifdef CONFIG_IP_MULTICAST
1698 in_dev->mr_ifc_count = 0;
1699 if (del_timer(&in_dev->mr_ifc_timer))
1700 __in_dev_put(in_dev);
1701 in_dev->mr_gq_running = 0;
1702 if (del_timer(&in_dev->mr_gq_timer))
1703 __in_dev_put(in_dev);
1704 #endif
1705
1706 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1707 }
1708
1709 void ip_mc_init_dev(struct in_device *in_dev)
1710 {
1711 #ifdef CONFIG_IP_MULTICAST
1712 struct net *net = dev_net(in_dev->dev);
1713 #endif
1714 ASSERT_RTNL();
1715
1716 #ifdef CONFIG_IP_MULTICAST
1717 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1718 (unsigned long)in_dev);
1719 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1720 (unsigned long)in_dev);
1721 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1722 #endif
1723
1724 spin_lock_init(&in_dev->mc_tomb_lock);
1725 }
1726
1727 /* Device going up */
1728
1729 void ip_mc_up(struct in_device *in_dev)
1730 {
1731 struct ip_mc_list *pmc;
1732 #ifdef CONFIG_IP_MULTICAST
1733 struct net *net = dev_net(in_dev->dev);
1734 #endif
1735
1736 ASSERT_RTNL();
1737
1738 #ifdef CONFIG_IP_MULTICAST
1739 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1740 #endif
1741 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1742
1743 for_each_pmc_rtnl(in_dev, pmc) {
1744 #ifdef CONFIG_IP_MULTICAST
1745 igmpv3_del_delrec(in_dev, pmc);
1746 #endif
1747 igmp_group_added(pmc);
1748 }
1749 }
1750
1751 /*
1752 * Device is about to be destroyed: clean up.
1753 */
1754
1755 void ip_mc_destroy_dev(struct in_device *in_dev)
1756 {
1757 struct ip_mc_list *i;
1758
1759 ASSERT_RTNL();
1760
1761 /* Deactivate timers */
1762 ip_mc_down(in_dev);
1763 #ifdef CONFIG_IP_MULTICAST
1764 igmpv3_clear_delrec(in_dev);
1765 #endif
1766
1767 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1768 in_dev->mc_list = i->next_rcu;
1769 in_dev->mc_count--;
1770 ip_ma_put(i);
1771 }
1772 }
1773
1774 /* RTNL is locked */
1775 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1776 {
1777 struct net_device *dev = NULL;
1778 struct in_device *idev = NULL;
1779
1780 if (imr->imr_ifindex) {
1781 idev = inetdev_by_index(net, imr->imr_ifindex);
1782 return idev;
1783 }
1784 if (imr->imr_address.s_addr) {
1785 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1786 if (!dev)
1787 return NULL;
1788 }
1789
1790 if (!dev) {
1791 struct rtable *rt = ip_route_output(net,
1792 imr->imr_multiaddr.s_addr,
1793 0, 0, 0);
1794 if (!IS_ERR(rt)) {
1795 dev = rt->dst.dev;
1796 ip_rt_put(rt);
1797 }
1798 }
1799 if (dev) {
1800 imr->imr_ifindex = dev->ifindex;
1801 idev = __in_dev_get_rtnl(dev);
1802 }
1803 return idev;
1804 }
1805
1806 /*
1807 * Join a socket to a group
1808 */
1809
1810 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1811 __be32 *psfsrc)
1812 {
1813 struct ip_sf_list *psf, *psf_prev;
1814 int rv = 0;
1815
1816 psf_prev = NULL;
1817 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1818 if (psf->sf_inaddr == *psfsrc)
1819 break;
1820 psf_prev = psf;
1821 }
1822 if (!psf || psf->sf_count[sfmode] == 0) {
1823 /* source filter not found, or count wrong => bug */
1824 return -ESRCH;
1825 }
1826 psf->sf_count[sfmode]--;
1827 if (psf->sf_count[sfmode] == 0) {
1828 ip_rt_multicast_event(pmc->interface);
1829 }
1830 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1831 #ifdef CONFIG_IP_MULTICAST
1832 struct in_device *in_dev = pmc->interface;
1833 struct net *net = dev_net(in_dev->dev);
1834 #endif
1835
1836 /* no more filters for this source */
1837 if (psf_prev)
1838 psf_prev->sf_next = psf->sf_next;
1839 else
1840 pmc->sources = psf->sf_next;
1841 #ifdef CONFIG_IP_MULTICAST
1842 if (psf->sf_oldin &&
1843 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1844 psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1845 psf->sf_next = pmc->tomb;
1846 pmc->tomb = psf;
1847 rv = 1;
1848 } else
1849 #endif
1850 kfree(psf);
1851 }
1852 return rv;
1853 }
1854
1855 #ifndef CONFIG_IP_MULTICAST
1856 #define igmp_ifc_event(x) do { } while (0)
1857 #endif
1858
1859 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1860 int sfcount, __be32 *psfsrc, int delta)
1861 {
1862 struct ip_mc_list *pmc;
1863 int changerec = 0;
1864 int i, err;
1865
1866 if (!in_dev)
1867 return -ENODEV;
1868 rcu_read_lock();
1869 for_each_pmc_rcu(in_dev, pmc) {
1870 if (*pmca == pmc->multiaddr)
1871 break;
1872 }
1873 if (!pmc) {
1874 /* MCA not found?? bug */
1875 rcu_read_unlock();
1876 return -ESRCH;
1877 }
1878 spin_lock_bh(&pmc->lock);
1879 rcu_read_unlock();
1880 #ifdef CONFIG_IP_MULTICAST
1881 sf_markstate(pmc);
1882 #endif
1883 if (!delta) {
1884 err = -EINVAL;
1885 if (!pmc->sfcount[sfmode])
1886 goto out_unlock;
1887 pmc->sfcount[sfmode]--;
1888 }
1889 err = 0;
1890 for (i = 0; i < sfcount; i++) {
1891 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1892
1893 changerec |= rv > 0;
1894 if (!err && rv < 0)
1895 err = rv;
1896 }
1897 if (pmc->sfmode == MCAST_EXCLUDE &&
1898 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1899 pmc->sfcount[MCAST_INCLUDE]) {
1900 #ifdef CONFIG_IP_MULTICAST
1901 struct ip_sf_list *psf;
1902 struct net *net = dev_net(in_dev->dev);
1903 #endif
1904
1905 /* filter mode change */
1906 pmc->sfmode = MCAST_INCLUDE;
1907 #ifdef CONFIG_IP_MULTICAST
1908 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1909 in_dev->mr_ifc_count = pmc->crcount;
1910 for (psf = pmc->sources; psf; psf = psf->sf_next)
1911 psf->sf_crcount = 0;
1912 igmp_ifc_event(pmc->interface);
1913 } else if (sf_setstate(pmc) || changerec) {
1914 igmp_ifc_event(pmc->interface);
1915 #endif
1916 }
1917 out_unlock:
1918 spin_unlock_bh(&pmc->lock);
1919 return err;
1920 }
1921
1922 /*
1923 * Add multicast single-source filter to the interface list
1924 */
1925 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1926 __be32 *psfsrc)
1927 {
1928 struct ip_sf_list *psf, *psf_prev;
1929
1930 psf_prev = NULL;
1931 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1932 if (psf->sf_inaddr == *psfsrc)
1933 break;
1934 psf_prev = psf;
1935 }
1936 if (!psf) {
1937 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1938 if (!psf)
1939 return -ENOBUFS;
1940 psf->sf_inaddr = *psfsrc;
1941 if (psf_prev) {
1942 psf_prev->sf_next = psf;
1943 } else
1944 pmc->sources = psf;
1945 }
1946 psf->sf_count[sfmode]++;
1947 if (psf->sf_count[sfmode] == 1) {
1948 ip_rt_multicast_event(pmc->interface);
1949 }
1950 return 0;
1951 }
1952
1953 #ifdef CONFIG_IP_MULTICAST
1954 static void sf_markstate(struct ip_mc_list *pmc)
1955 {
1956 struct ip_sf_list *psf;
1957 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1958
1959 for (psf = pmc->sources; psf; psf = psf->sf_next)
1960 if (pmc->sfcount[MCAST_EXCLUDE]) {
1961 psf->sf_oldin = mca_xcount ==
1962 psf->sf_count[MCAST_EXCLUDE] &&
1963 !psf->sf_count[MCAST_INCLUDE];
1964 } else
1965 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1966 }
1967
1968 static int sf_setstate(struct ip_mc_list *pmc)
1969 {
1970 struct ip_sf_list *psf, *dpsf;
1971 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1972 int qrv = pmc->interface->mr_qrv;
1973 int new_in, rv;
1974
1975 rv = 0;
1976 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1977 if (pmc->sfcount[MCAST_EXCLUDE]) {
1978 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1979 !psf->sf_count[MCAST_INCLUDE];
1980 } else
1981 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1982 if (new_in) {
1983 if (!psf->sf_oldin) {
1984 struct ip_sf_list *prev = NULL;
1985
1986 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1987 if (dpsf->sf_inaddr == psf->sf_inaddr)
1988 break;
1989 prev = dpsf;
1990 }
1991 if (dpsf) {
1992 if (prev)
1993 prev->sf_next = dpsf->sf_next;
1994 else
1995 pmc->tomb = dpsf->sf_next;
1996 kfree(dpsf);
1997 }
1998 psf->sf_crcount = qrv;
1999 rv++;
2000 }
2001 } else if (psf->sf_oldin) {
2002
2003 psf->sf_crcount = 0;
2004 /*
2005 * add or update "delete" records if an active filter
2006 * is now inactive
2007 */
2008 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
2009 if (dpsf->sf_inaddr == psf->sf_inaddr)
2010 break;
2011 if (!dpsf) {
2012 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2013 if (!dpsf)
2014 continue;
2015 *dpsf = *psf;
2016 /* pmc->lock held by callers */
2017 dpsf->sf_next = pmc->tomb;
2018 pmc->tomb = dpsf;
2019 }
2020 dpsf->sf_crcount = qrv;
2021 rv++;
2022 }
2023 }
2024 return rv;
2025 }
2026 #endif
2027
2028 /*
2029 * Add multicast source filter list to the interface list
2030 */
2031 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2032 int sfcount, __be32 *psfsrc, int delta)
2033 {
2034 struct ip_mc_list *pmc;
2035 int isexclude;
2036 int i, err;
2037
2038 if (!in_dev)
2039 return -ENODEV;
2040 rcu_read_lock();
2041 for_each_pmc_rcu(in_dev, pmc) {
2042 if (*pmca == pmc->multiaddr)
2043 break;
2044 }
2045 if (!pmc) {
2046 /* MCA not found?? bug */
2047 rcu_read_unlock();
2048 return -ESRCH;
2049 }
2050 spin_lock_bh(&pmc->lock);
2051 rcu_read_unlock();
2052
2053 #ifdef CONFIG_IP_MULTICAST
2054 sf_markstate(pmc);
2055 #endif
2056 isexclude = pmc->sfmode == MCAST_EXCLUDE;
2057 if (!delta)
2058 pmc->sfcount[sfmode]++;
2059 err = 0;
2060 for (i = 0; i < sfcount; i++) {
2061 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2062 if (err)
2063 break;
2064 }
2065 if (err) {
2066 int j;
2067
2068 if (!delta)
2069 pmc->sfcount[sfmode]--;
2070 for (j = 0; j < i; j++)
2071 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2072 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2073 #ifdef CONFIG_IP_MULTICAST
2074 struct ip_sf_list *psf;
2075 struct net *net = dev_net(pmc->interface->dev);
2076 in_dev = pmc->interface;
2077 #endif
2078
2079 /* filter mode change */
2080 if (pmc->sfcount[MCAST_EXCLUDE])
2081 pmc->sfmode = MCAST_EXCLUDE;
2082 else if (pmc->sfcount[MCAST_INCLUDE])
2083 pmc->sfmode = MCAST_INCLUDE;
2084 #ifdef CONFIG_IP_MULTICAST
2085 /* else no filters; keep old mode for reports */
2086
2087 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2088 in_dev->mr_ifc_count = pmc->crcount;
2089 for (psf = pmc->sources; psf; psf = psf->sf_next)
2090 psf->sf_crcount = 0;
2091 igmp_ifc_event(in_dev);
2092 } else if (sf_setstate(pmc)) {
2093 igmp_ifc_event(in_dev);
2094 #endif
2095 }
2096 spin_unlock_bh(&pmc->lock);
2097 return err;
2098 }
2099
2100 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2101 {
2102 struct ip_sf_list *psf, *nextpsf, *tomb, *sources;
2103
2104 spin_lock_bh(&pmc->lock);
2105 tomb = pmc->tomb;
2106 pmc->tomb = NULL;
2107 sources = pmc->sources;
2108 pmc->sources = NULL;
2109 pmc->sfmode = MCAST_EXCLUDE;
2110 pmc->sfcount[MCAST_INCLUDE] = 0;
2111 pmc->sfcount[MCAST_EXCLUDE] = 1;
2112 spin_unlock_bh(&pmc->lock);
2113
2114 for (psf = tomb; psf; psf = nextpsf) {
2115 nextpsf = psf->sf_next;
2116 kfree(psf);
2117 }
2118 for (psf = sources; psf; psf = nextpsf) {
2119 nextpsf = psf->sf_next;
2120 kfree(psf);
2121 }
2122 }
2123
2124 /* Join a multicast group
2125 */
2126
2127 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2128 {
2129 __be32 addr = imr->imr_multiaddr.s_addr;
2130 struct ip_mc_socklist *iml, *i;
2131 struct in_device *in_dev;
2132 struct inet_sock *inet = inet_sk(sk);
2133 struct net *net = sock_net(sk);
2134 int ifindex;
2135 int count = 0;
2136 int err;
2137
2138 ASSERT_RTNL();
2139
2140 if (!ipv4_is_multicast(addr))
2141 return -EINVAL;
2142
2143 in_dev = ip_mc_find_dev(net, imr);
2144
2145 if (!in_dev) {
2146 err = -ENODEV;
2147 goto done;
2148 }
2149
2150 err = -EADDRINUSE;
2151 ifindex = imr->imr_ifindex;
2152 for_each_pmc_rtnl(inet, i) {
2153 if (i->multi.imr_multiaddr.s_addr == addr &&
2154 i->multi.imr_ifindex == ifindex)
2155 goto done;
2156 count++;
2157 }
2158 err = -ENOBUFS;
2159 if (count >= net->ipv4.sysctl_igmp_max_memberships)
2160 goto done;
2161 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2162 if (!iml)
2163 goto done;
2164
2165 memcpy(&iml->multi, imr, sizeof(*imr));
2166 iml->next_rcu = inet->mc_list;
2167 iml->sflist = NULL;
2168 iml->sfmode = MCAST_EXCLUDE;
2169 rcu_assign_pointer(inet->mc_list, iml);
2170 ip_mc_inc_group(in_dev, addr);
2171 err = 0;
2172 done:
2173 return err;
2174 }
2175 EXPORT_SYMBOL(ip_mc_join_group);
2176
2177 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2178 struct in_device *in_dev)
2179 {
2180 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2181 int err;
2182
2183 if (!psf) {
2184 /* any-source empty exclude case */
2185 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2186 iml->sfmode, 0, NULL, 0);
2187 }
2188 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2189 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2190 RCU_INIT_POINTER(iml->sflist, NULL);
2191 /* decrease mem now to avoid the memleak warning */
2192 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2193 kfree_rcu(psf, rcu);
2194 return err;
2195 }
2196
2197 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2198 {
2199 struct inet_sock *inet = inet_sk(sk);
2200 struct ip_mc_socklist *iml;
2201 struct ip_mc_socklist __rcu **imlp;
2202 struct in_device *in_dev;
2203 struct net *net = sock_net(sk);
2204 __be32 group = imr->imr_multiaddr.s_addr;
2205 u32 ifindex;
2206 int ret = -EADDRNOTAVAIL;
2207
2208 ASSERT_RTNL();
2209
2210 in_dev = ip_mc_find_dev(net, imr);
2211 if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2212 ret = -ENODEV;
2213 goto out;
2214 }
2215 ifindex = imr->imr_ifindex;
2216 for (imlp = &inet->mc_list;
2217 (iml = rtnl_dereference(*imlp)) != NULL;
2218 imlp = &iml->next_rcu) {
2219 if (iml->multi.imr_multiaddr.s_addr != group)
2220 continue;
2221 if (ifindex) {
2222 if (iml->multi.imr_ifindex != ifindex)
2223 continue;
2224 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2225 iml->multi.imr_address.s_addr)
2226 continue;
2227
2228 (void) ip_mc_leave_src(sk, iml, in_dev);
2229
2230 *imlp = iml->next_rcu;
2231
2232 if (in_dev)
2233 ip_mc_dec_group(in_dev, group);
2234
2235 /* decrease mem now to avoid the memleak warning */
2236 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2237 kfree_rcu(iml, rcu);
2238 return 0;
2239 }
2240 out:
2241 return ret;
2242 }
2243 EXPORT_SYMBOL(ip_mc_leave_group);
2244
2245 int ip_mc_source(int add, int omode, struct sock *sk, struct
2246 ip_mreq_source *mreqs, int ifindex)
2247 {
2248 int err;
2249 struct ip_mreqn imr;
2250 __be32 addr = mreqs->imr_multiaddr;
2251 struct ip_mc_socklist *pmc;
2252 struct in_device *in_dev = NULL;
2253 struct inet_sock *inet = inet_sk(sk);
2254 struct ip_sf_socklist *psl;
2255 struct net *net = sock_net(sk);
2256 int leavegroup = 0;
2257 int i, j, rv;
2258
2259 if (!ipv4_is_multicast(addr))
2260 return -EINVAL;
2261
2262 ASSERT_RTNL();
2263
2264 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2265 imr.imr_address.s_addr = mreqs->imr_interface;
2266 imr.imr_ifindex = ifindex;
2267 in_dev = ip_mc_find_dev(net, &imr);
2268
2269 if (!in_dev) {
2270 err = -ENODEV;
2271 goto done;
2272 }
2273 err = -EADDRNOTAVAIL;
2274
2275 for_each_pmc_rtnl(inet, pmc) {
2276 if ((pmc->multi.imr_multiaddr.s_addr ==
2277 imr.imr_multiaddr.s_addr) &&
2278 (pmc->multi.imr_ifindex == imr.imr_ifindex))
2279 break;
2280 }
2281 if (!pmc) { /* must have a prior join */
2282 err = -EINVAL;
2283 goto done;
2284 }
2285 /* if a source filter was set, must be the same mode as before */
2286 if (pmc->sflist) {
2287 if (pmc->sfmode != omode) {
2288 err = -EINVAL;
2289 goto done;
2290 }
2291 } else if (pmc->sfmode != omode) {
2292 /* allow mode switches for empty-set filters */
2293 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2294 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2295 NULL, 0);
2296 pmc->sfmode = omode;
2297 }
2298
2299 psl = rtnl_dereference(pmc->sflist);
2300 if (!add) {
2301 if (!psl)
2302 goto done; /* err = -EADDRNOTAVAIL */
2303 rv = !0;
2304 for (i = 0; i < psl->sl_count; i++) {
2305 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2306 sizeof(__be32));
2307 if (rv == 0)
2308 break;
2309 }
2310 if (rv) /* source not found */
2311 goto done; /* err = -EADDRNOTAVAIL */
2312
2313 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2314 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2315 leavegroup = 1;
2316 goto done;
2317 }
2318
2319 /* update the interface filter */
2320 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2321 &mreqs->imr_sourceaddr, 1);
2322
2323 for (j = i+1; j < psl->sl_count; j++)
2324 psl->sl_addr[j-1] = psl->sl_addr[j];
2325 psl->sl_count--;
2326 err = 0;
2327 goto done;
2328 }
2329 /* else, add a new source to the filter */
2330
2331 if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2332 err = -ENOBUFS;
2333 goto done;
2334 }
2335 if (!psl || psl->sl_count == psl->sl_max) {
2336 struct ip_sf_socklist *newpsl;
2337 int count = IP_SFBLOCK;
2338
2339 if (psl)
2340 count += psl->sl_max;
2341 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2342 if (!newpsl) {
2343 err = -ENOBUFS;
2344 goto done;
2345 }
2346 newpsl->sl_max = count;
2347 newpsl->sl_count = count - IP_SFBLOCK;
2348 if (psl) {
2349 for (i = 0; i < psl->sl_count; i++)
2350 newpsl->sl_addr[i] = psl->sl_addr[i];
2351 /* decrease mem now to avoid the memleak warning */
2352 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2353 kfree_rcu(psl, rcu);
2354 }
2355 rcu_assign_pointer(pmc->sflist, newpsl);
2356 psl = newpsl;
2357 }
2358 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2359 for (i = 0; i < psl->sl_count; i++) {
2360 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2361 sizeof(__be32));
2362 if (rv == 0)
2363 break;
2364 }
2365 if (rv == 0) /* address already there is an error */
2366 goto done;
2367 for (j = psl->sl_count-1; j >= i; j--)
2368 psl->sl_addr[j+1] = psl->sl_addr[j];
2369 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2370 psl->sl_count++;
2371 err = 0;
2372 /* update the interface list */
2373 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2374 &mreqs->imr_sourceaddr, 1);
2375 done:
2376 if (leavegroup)
2377 err = ip_mc_leave_group(sk, &imr);
2378 return err;
2379 }
2380
2381 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2382 {
2383 int err = 0;
2384 struct ip_mreqn imr;
2385 __be32 addr = msf->imsf_multiaddr;
2386 struct ip_mc_socklist *pmc;
2387 struct in_device *in_dev;
2388 struct inet_sock *inet = inet_sk(sk);
2389 struct ip_sf_socklist *newpsl, *psl;
2390 struct net *net = sock_net(sk);
2391 int leavegroup = 0;
2392
2393 if (!ipv4_is_multicast(addr))
2394 return -EINVAL;
2395 if (msf->imsf_fmode != MCAST_INCLUDE &&
2396 msf->imsf_fmode != MCAST_EXCLUDE)
2397 return -EINVAL;
2398
2399 ASSERT_RTNL();
2400
2401 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2402 imr.imr_address.s_addr = msf->imsf_interface;
2403 imr.imr_ifindex = ifindex;
2404 in_dev = ip_mc_find_dev(net, &imr);
2405
2406 if (!in_dev) {
2407 err = -ENODEV;
2408 goto done;
2409 }
2410
2411 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2412 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2413 leavegroup = 1;
2414 goto done;
2415 }
2416
2417 for_each_pmc_rtnl(inet, pmc) {
2418 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2419 pmc->multi.imr_ifindex == imr.imr_ifindex)
2420 break;
2421 }
2422 if (!pmc) { /* must have a prior join */
2423 err = -EINVAL;
2424 goto done;
2425 }
2426 if (msf->imsf_numsrc) {
2427 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2428 GFP_KERNEL);
2429 if (!newpsl) {
2430 err = -ENOBUFS;
2431 goto done;
2432 }
2433 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2434 memcpy(newpsl->sl_addr, msf->imsf_slist,
2435 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2436 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2437 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2438 if (err) {
2439 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2440 goto done;
2441 }
2442 } else {
2443 newpsl = NULL;
2444 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2445 msf->imsf_fmode, 0, NULL, 0);
2446 }
2447 psl = rtnl_dereference(pmc->sflist);
2448 if (psl) {
2449 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2450 psl->sl_count, psl->sl_addr, 0);
2451 /* decrease mem now to avoid the memleak warning */
2452 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2453 kfree_rcu(psl, rcu);
2454 } else
2455 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2456 0, NULL, 0);
2457 rcu_assign_pointer(pmc->sflist, newpsl);
2458 pmc->sfmode = msf->imsf_fmode;
2459 err = 0;
2460 done:
2461 if (leavegroup)
2462 err = ip_mc_leave_group(sk, &imr);
2463 return err;
2464 }
2465
2466 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2467 struct ip_msfilter __user *optval, int __user *optlen)
2468 {
2469 int err, len, count, copycount;
2470 struct ip_mreqn imr;
2471 __be32 addr = msf->imsf_multiaddr;
2472 struct ip_mc_socklist *pmc;
2473 struct in_device *in_dev;
2474 struct inet_sock *inet = inet_sk(sk);
2475 struct ip_sf_socklist *psl;
2476 struct net *net = sock_net(sk);
2477
2478 ASSERT_RTNL();
2479
2480 if (!ipv4_is_multicast(addr))
2481 return -EINVAL;
2482
2483 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2484 imr.imr_address.s_addr = msf->imsf_interface;
2485 imr.imr_ifindex = 0;
2486 in_dev = ip_mc_find_dev(net, &imr);
2487
2488 if (!in_dev) {
2489 err = -ENODEV;
2490 goto done;
2491 }
2492 err = -EADDRNOTAVAIL;
2493
2494 for_each_pmc_rtnl(inet, pmc) {
2495 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2496 pmc->multi.imr_ifindex == imr.imr_ifindex)
2497 break;
2498 }
2499 if (!pmc) /* must have a prior join */
2500 goto done;
2501 msf->imsf_fmode = pmc->sfmode;
2502 psl = rtnl_dereference(pmc->sflist);
2503 if (!psl) {
2504 len = 0;
2505 count = 0;
2506 } else {
2507 count = psl->sl_count;
2508 }
2509 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2510 len = copycount * sizeof(psl->sl_addr[0]);
2511 msf->imsf_numsrc = count;
2512 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2513 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2514 return -EFAULT;
2515 }
2516 if (len &&
2517 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2518 return -EFAULT;
2519 return 0;
2520 done:
2521 return err;
2522 }
2523
2524 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2525 struct group_filter __user *optval, int __user *optlen)
2526 {
2527 int err, i, count, copycount;
2528 struct sockaddr_in *psin;
2529 __be32 addr;
2530 struct ip_mc_socklist *pmc;
2531 struct inet_sock *inet = inet_sk(sk);
2532 struct ip_sf_socklist *psl;
2533
2534 ASSERT_RTNL();
2535
2536 psin = (struct sockaddr_in *)&gsf->gf_group;
2537 if (psin->sin_family != AF_INET)
2538 return -EINVAL;
2539 addr = psin->sin_addr.s_addr;
2540 if (!ipv4_is_multicast(addr))
2541 return -EINVAL;
2542
2543 err = -EADDRNOTAVAIL;
2544
2545 for_each_pmc_rtnl(inet, pmc) {
2546 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2547 pmc->multi.imr_ifindex == gsf->gf_interface)
2548 break;
2549 }
2550 if (!pmc) /* must have a prior join */
2551 goto done;
2552 gsf->gf_fmode = pmc->sfmode;
2553 psl = rtnl_dereference(pmc->sflist);
2554 count = psl ? psl->sl_count : 0;
2555 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2556 gsf->gf_numsrc = count;
2557 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2558 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2559 return -EFAULT;
2560 }
2561 for (i = 0; i < copycount; i++) {
2562 struct sockaddr_storage ss;
2563
2564 psin = (struct sockaddr_in *)&ss;
2565 memset(&ss, 0, sizeof(ss));
2566 psin->sin_family = AF_INET;
2567 psin->sin_addr.s_addr = psl->sl_addr[i];
2568 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2569 return -EFAULT;
2570 }
2571 return 0;
2572 done:
2573 return err;
2574 }
2575
2576 /*
2577 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2578 */
2579 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2580 {
2581 struct inet_sock *inet = inet_sk(sk);
2582 struct ip_mc_socklist *pmc;
2583 struct ip_sf_socklist *psl;
2584 int i;
2585 int ret;
2586
2587 ret = 1;
2588 if (!ipv4_is_multicast(loc_addr))
2589 goto out;
2590
2591 rcu_read_lock();
2592 for_each_pmc_rcu(inet, pmc) {
2593 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2594 pmc->multi.imr_ifindex == dif)
2595 break;
2596 }
2597 ret = inet->mc_all;
2598 if (!pmc)
2599 goto unlock;
2600 psl = rcu_dereference(pmc->sflist);
2601 ret = (pmc->sfmode == MCAST_EXCLUDE);
2602 if (!psl)
2603 goto unlock;
2604
2605 for (i = 0; i < psl->sl_count; i++) {
2606 if (psl->sl_addr[i] == rmt_addr)
2607 break;
2608 }
2609 ret = 0;
2610 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2611 goto unlock;
2612 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2613 goto unlock;
2614 ret = 1;
2615 unlock:
2616 rcu_read_unlock();
2617 out:
2618 return ret;
2619 }
2620
2621 /*
2622 * A socket is closing.
2623 */
2624
2625 void ip_mc_drop_socket(struct sock *sk)
2626 {
2627 struct inet_sock *inet = inet_sk(sk);
2628 struct ip_mc_socklist *iml;
2629 struct net *net = sock_net(sk);
2630
2631 if (!inet->mc_list)
2632 return;
2633
2634 rtnl_lock();
2635 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2636 struct in_device *in_dev;
2637
2638 inet->mc_list = iml->next_rcu;
2639 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2640 (void) ip_mc_leave_src(sk, iml, in_dev);
2641 if (in_dev)
2642 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2643 /* decrease mem now to avoid the memleak warning */
2644 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2645 kfree_rcu(iml, rcu);
2646 }
2647 rtnl_unlock();
2648 }
2649
2650 /* called with rcu_read_lock() */
2651 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2652 {
2653 struct ip_mc_list *im;
2654 struct ip_mc_list __rcu **mc_hash;
2655 struct ip_sf_list *psf;
2656 int rv = 0;
2657
2658 mc_hash = rcu_dereference(in_dev->mc_hash);
2659 if (mc_hash) {
2660 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2661
2662 for (im = rcu_dereference(mc_hash[hash]);
2663 im != NULL;
2664 im = rcu_dereference(im->next_hash)) {
2665 if (im->multiaddr == mc_addr)
2666 break;
2667 }
2668 } else {
2669 for_each_pmc_rcu(in_dev, im) {
2670 if (im->multiaddr == mc_addr)
2671 break;
2672 }
2673 }
2674 if (im && proto == IPPROTO_IGMP) {
2675 rv = 1;
2676 } else if (im) {
2677 if (src_addr) {
2678 for (psf = im->sources; psf; psf = psf->sf_next) {
2679 if (psf->sf_inaddr == src_addr)
2680 break;
2681 }
2682 if (psf)
2683 rv = psf->sf_count[MCAST_INCLUDE] ||
2684 psf->sf_count[MCAST_EXCLUDE] !=
2685 im->sfcount[MCAST_EXCLUDE];
2686 else
2687 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2688 } else
2689 rv = 1; /* unspecified source; tentatively allow */
2690 }
2691 return rv;
2692 }
2693
2694 #if defined(CONFIG_PROC_FS)
2695 struct igmp_mc_iter_state {
2696 struct seq_net_private p;
2697 struct net_device *dev;
2698 struct in_device *in_dev;
2699 };
2700
2701 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2702
2703 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2704 {
2705 struct net *net = seq_file_net(seq);
2706 struct ip_mc_list *im = NULL;
2707 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2708
2709 state->in_dev = NULL;
2710 for_each_netdev_rcu(net, state->dev) {
2711 struct in_device *in_dev;
2712
2713 in_dev = __in_dev_get_rcu(state->dev);
2714 if (!in_dev)
2715 continue;
2716 im = rcu_dereference(in_dev->mc_list);
2717 if (im) {
2718 state->in_dev = in_dev;
2719 break;
2720 }
2721 }
2722 return im;
2723 }
2724
2725 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2726 {
2727 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2728
2729 im = rcu_dereference(im->next_rcu);
2730 while (!im) {
2731 state->dev = next_net_device_rcu(state->dev);
2732 if (!state->dev) {
2733 state->in_dev = NULL;
2734 break;
2735 }
2736 state->in_dev = __in_dev_get_rcu(state->dev);
2737 if (!state->in_dev)
2738 continue;
2739 im = rcu_dereference(state->in_dev->mc_list);
2740 }
2741 return im;
2742 }
2743
2744 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2745 {
2746 struct ip_mc_list *im = igmp_mc_get_first(seq);
2747 if (im)
2748 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2749 --pos;
2750 return pos ? NULL : im;
2751 }
2752
2753 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2754 __acquires(rcu)
2755 {
2756 rcu_read_lock();
2757 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2758 }
2759
2760 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2761 {
2762 struct ip_mc_list *im;
2763 if (v == SEQ_START_TOKEN)
2764 im = igmp_mc_get_first(seq);
2765 else
2766 im = igmp_mc_get_next(seq, v);
2767 ++*pos;
2768 return im;
2769 }
2770
2771 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2772 __releases(rcu)
2773 {
2774 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2775
2776 state->in_dev = NULL;
2777 state->dev = NULL;
2778 rcu_read_unlock();
2779 }
2780
2781 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2782 {
2783 if (v == SEQ_START_TOKEN)
2784 seq_puts(seq,
2785 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2786 else {
2787 struct ip_mc_list *im = (struct ip_mc_list *)v;
2788 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2789 char *querier;
2790 long delta;
2791
2792 #ifdef CONFIG_IP_MULTICAST
2793 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2794 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2795 "V3";
2796 #else
2797 querier = "NONE";
2798 #endif
2799
2800 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2801 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2802 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2803 }
2804
2805 delta = im->timer.expires - jiffies;
2806 seq_printf(seq,
2807 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2808 im->multiaddr, im->users,
2809 im->tm_running,
2810 im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2811 im->reporter);
2812 }
2813 return 0;
2814 }
2815
2816 static const struct seq_operations igmp_mc_seq_ops = {
2817 .start = igmp_mc_seq_start,
2818 .next = igmp_mc_seq_next,
2819 .stop = igmp_mc_seq_stop,
2820 .show = igmp_mc_seq_show,
2821 };
2822
2823 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2824 {
2825 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2826 sizeof(struct igmp_mc_iter_state));
2827 }
2828
2829 static const struct file_operations igmp_mc_seq_fops = {
2830 .owner = THIS_MODULE,
2831 .open = igmp_mc_seq_open,
2832 .read = seq_read,
2833 .llseek = seq_lseek,
2834 .release = seq_release_net,
2835 };
2836
2837 struct igmp_mcf_iter_state {
2838 struct seq_net_private p;
2839 struct net_device *dev;
2840 struct in_device *idev;
2841 struct ip_mc_list *im;
2842 };
2843
2844 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2845
2846 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2847 {
2848 struct net *net = seq_file_net(seq);
2849 struct ip_sf_list *psf = NULL;
2850 struct ip_mc_list *im = NULL;
2851 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2852
2853 state->idev = NULL;
2854 state->im = NULL;
2855 for_each_netdev_rcu(net, state->dev) {
2856 struct in_device *idev;
2857 idev = __in_dev_get_rcu(state->dev);
2858 if (unlikely(!idev))
2859 continue;
2860 im = rcu_dereference(idev->mc_list);
2861 if (likely(im)) {
2862 spin_lock_bh(&im->lock);
2863 psf = im->sources;
2864 if (likely(psf)) {
2865 state->im = im;
2866 state->idev = idev;
2867 break;
2868 }
2869 spin_unlock_bh(&im->lock);
2870 }
2871 }
2872 return psf;
2873 }
2874
2875 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2876 {
2877 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2878
2879 psf = psf->sf_next;
2880 while (!psf) {
2881 spin_unlock_bh(&state->im->lock);
2882 state->im = state->im->next;
2883 while (!state->im) {
2884 state->dev = next_net_device_rcu(state->dev);
2885 if (!state->dev) {
2886 state->idev = NULL;
2887 goto out;
2888 }
2889 state->idev = __in_dev_get_rcu(state->dev);
2890 if (!state->idev)
2891 continue;
2892 state->im = rcu_dereference(state->idev->mc_list);
2893 }
2894 if (!state->im)
2895 break;
2896 spin_lock_bh(&state->im->lock);
2897 psf = state->im->sources;
2898 }
2899 out:
2900 return psf;
2901 }
2902
2903 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2904 {
2905 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2906 if (psf)
2907 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2908 --pos;
2909 return pos ? NULL : psf;
2910 }
2911
2912 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2913 __acquires(rcu)
2914 {
2915 rcu_read_lock();
2916 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2917 }
2918
2919 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2920 {
2921 struct ip_sf_list *psf;
2922 if (v == SEQ_START_TOKEN)
2923 psf = igmp_mcf_get_first(seq);
2924 else
2925 psf = igmp_mcf_get_next(seq, v);
2926 ++*pos;
2927 return psf;
2928 }
2929
2930 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2931 __releases(rcu)
2932 {
2933 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2934 if (likely(state->im)) {
2935 spin_unlock_bh(&state->im->lock);
2936 state->im = NULL;
2937 }
2938 state->idev = NULL;
2939 state->dev = NULL;
2940 rcu_read_unlock();
2941 }
2942
2943 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2944 {
2945 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2946 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2947
2948 if (v == SEQ_START_TOKEN) {
2949 seq_puts(seq, "Idx Device MCA SRC INC EXC\n");
2950 } else {
2951 seq_printf(seq,
2952 "%3d %6.6s 0x%08x "
2953 "0x%08x %6lu %6lu\n",
2954 state->dev->ifindex, state->dev->name,
2955 ntohl(state->im->multiaddr),
2956 ntohl(psf->sf_inaddr),
2957 psf->sf_count[MCAST_INCLUDE],
2958 psf->sf_count[MCAST_EXCLUDE]);
2959 }
2960 return 0;
2961 }
2962
2963 static const struct seq_operations igmp_mcf_seq_ops = {
2964 .start = igmp_mcf_seq_start,
2965 .next = igmp_mcf_seq_next,
2966 .stop = igmp_mcf_seq_stop,
2967 .show = igmp_mcf_seq_show,
2968 };
2969
2970 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2971 {
2972 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2973 sizeof(struct igmp_mcf_iter_state));
2974 }
2975
2976 static const struct file_operations igmp_mcf_seq_fops = {
2977 .owner = THIS_MODULE,
2978 .open = igmp_mcf_seq_open,
2979 .read = seq_read,
2980 .llseek = seq_lseek,
2981 .release = seq_release_net,
2982 };
2983
2984 static int __net_init igmp_net_init(struct net *net)
2985 {
2986 struct proc_dir_entry *pde;
2987 int err;
2988
2989 pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2990 if (!pde)
2991 goto out_igmp;
2992 pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2993 &igmp_mcf_seq_fops);
2994 if (!pde)
2995 goto out_mcfilter;
2996 err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2997 SOCK_DGRAM, 0, net);
2998 if (err < 0) {
2999 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
3000 err);
3001 goto out_sock;
3002 }
3003
3004 return 0;
3005
3006 out_sock:
3007 remove_proc_entry("mcfilter", net->proc_net);
3008 out_mcfilter:
3009 remove_proc_entry("igmp", net->proc_net);
3010 out_igmp:
3011 return -ENOMEM;
3012 }
3013
3014 static void __net_exit igmp_net_exit(struct net *net)
3015 {
3016 remove_proc_entry("mcfilter", net->proc_net);
3017 remove_proc_entry("igmp", net->proc_net);
3018 inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
3019 }
3020
3021 static struct pernet_operations igmp_net_ops = {
3022 .init = igmp_net_init,
3023 .exit = igmp_net_exit,
3024 };
3025 #endif
3026
3027 static int igmp_netdev_event(struct notifier_block *this,
3028 unsigned long event, void *ptr)
3029 {
3030 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3031 struct in_device *in_dev;
3032
3033 switch (event) {
3034 case NETDEV_RESEND_IGMP:
3035 in_dev = __in_dev_get_rtnl(dev);
3036 if (in_dev)
3037 ip_mc_rejoin_groups(in_dev);
3038 break;
3039 default:
3040 break;
3041 }
3042 return NOTIFY_DONE;
3043 }
3044
3045 static struct notifier_block igmp_notifier = {
3046 .notifier_call = igmp_netdev_event,
3047 };
3048
3049 int __init igmp_mc_init(void)
3050 {
3051 #if defined(CONFIG_PROC_FS)
3052 int err;
3053
3054 err = register_pernet_subsys(&igmp_net_ops);
3055 if (err)
3056 return err;
3057 err = register_netdevice_notifier(&igmp_notifier);
3058 if (err)
3059 goto reg_notif_fail;
3060 return 0;
3061
3062 reg_notif_fail:
3063 unregister_pernet_subsys(&igmp_net_ops);
3064 return err;
3065 #else
3066 return register_netdevice_notifier(&igmp_notifier);
3067 #endif
3068 }
Linux with added FPC-III support