?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/speculation/mds: Fix documentation typo
[linux/fpc-iii.git] / net / ipv4 / igmp.c
blob502aae3e3ab8e9c263ab25a0e9dcf9e38d48a5f4
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 <linux/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 (refcount_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 refcount_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 refcount_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 refcount_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 = 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 = 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 = 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 net_device *dev = skb->dev;
1039 struct in_device *in_dev;
1040 int len = skb->len;
1041 bool dropped = true;
1042
1043 if (netif_is_l3_master(dev)) {
1044 dev = dev_get_by_index_rcu(dev_net(dev), IPCB(skb)->iif);
1045 if (!dev)
1046 goto drop;
1047 }
1048
1049 in_dev = __in_dev_get_rcu(dev);
1050 if (!in_dev)
1051 goto drop;
1052
1053 if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1054 goto drop;
1055
1056 if (skb_checksum_simple_validate(skb))
1057 goto drop;
1058
1059 ih = igmp_hdr(skb);
1060 switch (ih->type) {
1061 case IGMP_HOST_MEMBERSHIP_QUERY:
1062 dropped = igmp_heard_query(in_dev, skb, len);
1063 break;
1064 case IGMP_HOST_MEMBERSHIP_REPORT:
1065 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1066 /* Is it our report looped back? */
1067 if (rt_is_output_route(skb_rtable(skb)))
1068 break;
1069 /* don't rely on MC router hearing unicast reports */
1070 if (skb->pkt_type == PACKET_MULTICAST ||
1071 skb->pkt_type == PACKET_BROADCAST)
1072 dropped = igmp_heard_report(in_dev, ih->group);
1073 break;
1074 case IGMP_PIM:
1075 #ifdef CONFIG_IP_PIMSM_V1
1076 return pim_rcv_v1(skb);
1077 #endif
1078 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1079 case IGMP_DVMRP:
1080 case IGMP_TRACE:
1081 case IGMP_HOST_LEAVE_MESSAGE:
1082 case IGMP_MTRACE:
1083 case IGMP_MTRACE_RESP:
1084 break;
1085 default:
1086 break;
1087 }
1088
1089 drop:
1090 if (dropped)
1091 kfree_skb(skb);
1092 else
1093 consume_skb(skb);
1094 return 0;
1095 }
1096
1097 #endif
1098
1099
1100 /*
1101 * Add a filter to a device
1102 */
1103
1104 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1105 {
1106 char buf[MAX_ADDR_LEN];
1107 struct net_device *dev = in_dev->dev;
1108
1109 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1110 We will get multicast token leakage, when IFF_MULTICAST
1111 is changed. This check should be done in ndo_set_rx_mode
1112 routine. Something sort of:
1113 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1114 --ANK
1115 */
1116 if (arp_mc_map(addr, buf, dev, 0) == 0)
1117 dev_mc_add(dev, buf);
1118 }
1119
1120 /*
1121 * Remove a filter from a device
1122 */
1123
1124 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1125 {
1126 char buf[MAX_ADDR_LEN];
1127 struct net_device *dev = in_dev->dev;
1128
1129 if (arp_mc_map(addr, buf, dev, 0) == 0)
1130 dev_mc_del(dev, buf);
1131 }
1132
1133 #ifdef CONFIG_IP_MULTICAST
1134 /*
1135 * deleted ip_mc_list manipulation
1136 */
1137 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1138 {
1139 struct ip_mc_list *pmc;
1140 struct net *net = dev_net(in_dev->dev);
1141
1142 /* this is an "ip_mc_list" for convenience; only the fields below
1143 * are actually used. In particular, the refcnt and users are not
1144 * used for management of the delete list. Using the same structure
1145 * for deleted items allows change reports to use common code with
1146 * non-deleted or query-response MCA's.
1147 */
1148 pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1149 if (!pmc)
1150 return;
1151 spin_lock_init(&pmc->lock);
1152 spin_lock_bh(&im->lock);
1153 pmc->interface = im->interface;
1154 in_dev_hold(in_dev);
1155 pmc->multiaddr = im->multiaddr;
1156 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1157 pmc->sfmode = im->sfmode;
1158 if (pmc->sfmode == MCAST_INCLUDE) {
1159 struct ip_sf_list *psf;
1160
1161 pmc->tomb = im->tomb;
1162 pmc->sources = im->sources;
1163 im->tomb = im->sources = NULL;
1164 for (psf = pmc->sources; psf; psf = psf->sf_next)
1165 psf->sf_crcount = pmc->crcount;
1166 }
1167 spin_unlock_bh(&im->lock);
1168
1169 spin_lock_bh(&in_dev->mc_tomb_lock);
1170 pmc->next = in_dev->mc_tomb;
1171 in_dev->mc_tomb = pmc;
1172 spin_unlock_bh(&in_dev->mc_tomb_lock);
1173 }
1174
1175 /*
1176 * restore ip_mc_list deleted records
1177 */
1178 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1179 {
1180 struct ip_mc_list *pmc, *pmc_prev;
1181 struct ip_sf_list *psf;
1182 struct net *net = dev_net(in_dev->dev);
1183 __be32 multiaddr = im->multiaddr;
1184
1185 spin_lock_bh(&in_dev->mc_tomb_lock);
1186 pmc_prev = NULL;
1187 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1188 if (pmc->multiaddr == multiaddr)
1189 break;
1190 pmc_prev = pmc;
1191 }
1192 if (pmc) {
1193 if (pmc_prev)
1194 pmc_prev->next = pmc->next;
1195 else
1196 in_dev->mc_tomb = pmc->next;
1197 }
1198 spin_unlock_bh(&in_dev->mc_tomb_lock);
1199
1200 spin_lock_bh(&im->lock);
1201 if (pmc) {
1202 im->interface = pmc->interface;
1203 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1204 if (im->sfmode == MCAST_INCLUDE) {
1205 im->tomb = pmc->tomb;
1206 im->sources = pmc->sources;
1207 for (psf = im->sources; psf; psf = psf->sf_next)
1208 psf->sf_crcount = im->crcount;
1209 }
1210 in_dev_put(pmc->interface);
1211 kfree(pmc);
1212 }
1213 spin_unlock_bh(&im->lock);
1214 }
1215
1216 /*
1217 * flush ip_mc_list deleted records
1218 */
1219 static void igmpv3_clear_delrec(struct in_device *in_dev)
1220 {
1221 struct ip_mc_list *pmc, *nextpmc;
1222
1223 spin_lock_bh(&in_dev->mc_tomb_lock);
1224 pmc = in_dev->mc_tomb;
1225 in_dev->mc_tomb = NULL;
1226 spin_unlock_bh(&in_dev->mc_tomb_lock);
1227
1228 for (; pmc; pmc = nextpmc) {
1229 nextpmc = pmc->next;
1230 ip_mc_clear_src(pmc);
1231 in_dev_put(pmc->interface);
1232 kfree(pmc);
1233 }
1234 /* clear dead sources, too */
1235 rcu_read_lock();
1236 for_each_pmc_rcu(in_dev, pmc) {
1237 struct ip_sf_list *psf, *psf_next;
1238
1239 spin_lock_bh(&pmc->lock);
1240 psf = pmc->tomb;
1241 pmc->tomb = NULL;
1242 spin_unlock_bh(&pmc->lock);
1243 for (; psf; psf = psf_next) {
1244 psf_next = psf->sf_next;
1245 kfree(psf);
1246 }
1247 }
1248 rcu_read_unlock();
1249 }
1250 #endif
1251
1252 static void igmp_group_dropped(struct ip_mc_list *im)
1253 {
1254 struct in_device *in_dev = im->interface;
1255 #ifdef CONFIG_IP_MULTICAST
1256 struct net *net = dev_net(in_dev->dev);
1257 int reporter;
1258 #endif
1259
1260 if (im->loaded) {
1261 im->loaded = 0;
1262 ip_mc_filter_del(in_dev, im->multiaddr);
1263 }
1264
1265 #ifdef CONFIG_IP_MULTICAST
1266 if (im->multiaddr == IGMP_ALL_HOSTS)
1267 return;
1268 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1269 return;
1270
1271 reporter = im->reporter;
1272 igmp_stop_timer(im);
1273
1274 if (!in_dev->dead) {
1275 if (IGMP_V1_SEEN(in_dev))
1276 return;
1277 if (IGMP_V2_SEEN(in_dev)) {
1278 if (reporter)
1279 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1280 return;
1281 }
1282 /* IGMPv3 */
1283 igmpv3_add_delrec(in_dev, im);
1284
1285 igmp_ifc_event(in_dev);
1286 }
1287 #endif
1288 }
1289
1290 static void igmp_group_added(struct ip_mc_list *im)
1291 {
1292 struct in_device *in_dev = im->interface;
1293 #ifdef CONFIG_IP_MULTICAST
1294 struct net *net = dev_net(in_dev->dev);
1295 #endif
1296
1297 if (im->loaded == 0) {
1298 im->loaded = 1;
1299 ip_mc_filter_add(in_dev, im->multiaddr);
1300 }
1301
1302 #ifdef CONFIG_IP_MULTICAST
1303 if (im->multiaddr == IGMP_ALL_HOSTS)
1304 return;
1305 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1306 return;
1307
1308 if (in_dev->dead)
1309 return;
1310 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1311 spin_lock_bh(&im->lock);
1312 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1313 spin_unlock_bh(&im->lock);
1314 return;
1315 }
1316 /* else, v3 */
1317
1318 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1319 igmp_ifc_event(in_dev);
1320 #endif
1321 }
1322
1323
1324 /*
1325 * Multicast list managers
1326 */
1327
1328 static u32 ip_mc_hash(const struct ip_mc_list *im)
1329 {
1330 return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1331 }
1332
1333 static void ip_mc_hash_add(struct in_device *in_dev,
1334 struct ip_mc_list *im)
1335 {
1336 struct ip_mc_list __rcu **mc_hash;
1337 u32 hash;
1338
1339 mc_hash = rtnl_dereference(in_dev->mc_hash);
1340 if (mc_hash) {
1341 hash = ip_mc_hash(im);
1342 im->next_hash = mc_hash[hash];
1343 rcu_assign_pointer(mc_hash[hash], im);
1344 return;
1345 }
1346
1347 /* do not use a hash table for small number of items */
1348 if (in_dev->mc_count < 4)
1349 return;
1350
1351 mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1352 GFP_KERNEL);
1353 if (!mc_hash)
1354 return;
1355
1356 for_each_pmc_rtnl(in_dev, im) {
1357 hash = ip_mc_hash(im);
1358 im->next_hash = mc_hash[hash];
1359 RCU_INIT_POINTER(mc_hash[hash], im);
1360 }
1361
1362 rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1363 }
1364
1365 static void ip_mc_hash_remove(struct in_device *in_dev,
1366 struct ip_mc_list *im)
1367 {
1368 struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1369 struct ip_mc_list *aux;
1370
1371 if (!mc_hash)
1372 return;
1373 mc_hash += ip_mc_hash(im);
1374 while ((aux = rtnl_dereference(*mc_hash)) != im)
1375 mc_hash = &aux->next_hash;
1376 *mc_hash = im->next_hash;
1377 }
1378
1379
1380 /*
1381 * A socket has joined a multicast group on device dev.
1382 */
1383
1384 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1385 {
1386 struct ip_mc_list *im;
1387 #ifdef CONFIG_IP_MULTICAST
1388 struct net *net = dev_net(in_dev->dev);
1389 #endif
1390
1391 ASSERT_RTNL();
1392
1393 for_each_pmc_rtnl(in_dev, im) {
1394 if (im->multiaddr == addr) {
1395 im->users++;
1396 ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1397 goto out;
1398 }
1399 }
1400
1401 im = kzalloc(sizeof(*im), GFP_KERNEL);
1402 if (!im)
1403 goto out;
1404
1405 im->users = 1;
1406 im->interface = in_dev;
1407 in_dev_hold(in_dev);
1408 im->multiaddr = addr;
1409 /* initial mode is (EX, empty) */
1410 im->sfmode = MCAST_EXCLUDE;
1411 im->sfcount[MCAST_EXCLUDE] = 1;
1412 refcount_set(&im->refcnt, 1);
1413 spin_lock_init(&im->lock);
1414 #ifdef CONFIG_IP_MULTICAST
1415 setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1416 im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1417 #endif
1418
1419 im->next_rcu = in_dev->mc_list;
1420 in_dev->mc_count++;
1421 rcu_assign_pointer(in_dev->mc_list, im);
1422
1423 ip_mc_hash_add(in_dev, im);
1424
1425 #ifdef CONFIG_IP_MULTICAST
1426 igmpv3_del_delrec(in_dev, im);
1427 #endif
1428 igmp_group_added(im);
1429 if (!in_dev->dead)
1430 ip_rt_multicast_event(in_dev);
1431 out:
1432 return;
1433 }
1434 EXPORT_SYMBOL(ip_mc_inc_group);
1435
1436 static int ip_mc_check_iphdr(struct sk_buff *skb)
1437 {
1438 const struct iphdr *iph;
1439 unsigned int len;
1440 unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1441
1442 if (!pskb_may_pull(skb, offset))
1443 return -EINVAL;
1444
1445 iph = ip_hdr(skb);
1446
1447 if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1448 return -EINVAL;
1449
1450 offset += ip_hdrlen(skb) - sizeof(*iph);
1451
1452 if (!pskb_may_pull(skb, offset))
1453 return -EINVAL;
1454
1455 iph = ip_hdr(skb);
1456
1457 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1458 return -EINVAL;
1459
1460 len = skb_network_offset(skb) + ntohs(iph->tot_len);
1461 if (skb->len < len || len < offset)
1462 return -EINVAL;
1463
1464 skb_set_transport_header(skb, offset);
1465
1466 return 0;
1467 }
1468
1469 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1470 {
1471 unsigned int len = skb_transport_offset(skb);
1472
1473 len += sizeof(struct igmpv3_report);
1474
1475 return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1476 }
1477
1478 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1479 {
1480 unsigned int len = skb_transport_offset(skb);
1481
1482 len += sizeof(struct igmphdr);
1483 if (skb->len < len)
1484 return -EINVAL;
1485
1486 /* IGMPv{1,2}? */
1487 if (skb->len != len) {
1488 /* or IGMPv3? */
1489 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1490 if (skb->len < len || !pskb_may_pull(skb, len))
1491 return -EINVAL;
1492 }
1493
1494 /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1495 * all-systems destination addresses (224.0.0.1) for general queries
1496 */
1497 if (!igmp_hdr(skb)->group &&
1498 ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1499 return -EINVAL;
1500
1501 return 0;
1502 }
1503
1504 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1505 {
1506 switch (igmp_hdr(skb)->type) {
1507 case IGMP_HOST_LEAVE_MESSAGE:
1508 case IGMP_HOST_MEMBERSHIP_REPORT:
1509 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1510 /* fall through */
1511 return 0;
1512 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1513 return ip_mc_check_igmp_reportv3(skb);
1514 case IGMP_HOST_MEMBERSHIP_QUERY:
1515 return ip_mc_check_igmp_query(skb);
1516 default:
1517 return -ENOMSG;
1518 }
1519 }
1520
1521 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1522 {
1523 return skb_checksum_simple_validate(skb);
1524 }
1525
1526 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1527
1528 {
1529 struct sk_buff *skb_chk;
1530 unsigned int transport_len;
1531 unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1532 int ret = -EINVAL;
1533
1534 transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1535
1536 skb_chk = skb_checksum_trimmed(skb, transport_len,
1537 ip_mc_validate_checksum);
1538 if (!skb_chk)
1539 goto err;
1540
1541 if (!pskb_may_pull(skb_chk, len))
1542 goto err;
1543
1544 ret = ip_mc_check_igmp_msg(skb_chk);
1545 if (ret)
1546 goto err;
1547
1548 if (skb_trimmed)
1549 *skb_trimmed = skb_chk;
1550 /* free now unneeded clone */
1551 else if (skb_chk != skb)
1552 kfree_skb(skb_chk);
1553
1554 ret = 0;
1555
1556 err:
1557 if (ret && skb_chk && skb_chk != skb)
1558 kfree_skb(skb_chk);
1559
1560 return ret;
1561 }
1562
1563 /**
1564 * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1565 * @skb: the skb to validate
1566 * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1567 *
1568 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1569 * skb transport header accordingly and returns zero.
1570 *
1571 * -EINVAL: A broken packet was detected, i.e. it violates some internet
1572 * standard
1573 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1574 * -ENOMEM: A memory allocation failure happened.
1575 *
1576 * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1577 * to NULL): After parsing an IGMP packet successfully it will point to
1578 * an skb which has its tail aligned to the IP packet end. This might
1579 * either be the originally provided skb or a trimmed, cloned version if
1580 * the skb frame had data beyond the IP packet. A cloned skb allows us
1581 * to leave the original skb and its full frame unchanged (which might be
1582 * desirable for layer 2 frame jugglers).
1583 *
1584 * Caller needs to set the skb network header and free any returned skb if it
1585 * differs from the provided skb.
1586 */
1587 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1588 {
1589 int ret = ip_mc_check_iphdr(skb);
1590
1591 if (ret < 0)
1592 return ret;
1593
1594 if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1595 return -ENOMSG;
1596
1597 return __ip_mc_check_igmp(skb, skb_trimmed);
1598 }
1599 EXPORT_SYMBOL(ip_mc_check_igmp);
1600
1601 /*
1602 * Resend IGMP JOIN report; used by netdev notifier.
1603 */
1604 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1605 {
1606 #ifdef CONFIG_IP_MULTICAST
1607 struct ip_mc_list *im;
1608 int type;
1609 struct net *net = dev_net(in_dev->dev);
1610
1611 ASSERT_RTNL();
1612
1613 for_each_pmc_rtnl(in_dev, im) {
1614 if (im->multiaddr == IGMP_ALL_HOSTS)
1615 continue;
1616 if (ipv4_is_local_multicast(im->multiaddr) &&
1617 !net->ipv4.sysctl_igmp_llm_reports)
1618 continue;
1619
1620 /* a failover is happening and switches
1621 * must be notified immediately
1622 */
1623 if (IGMP_V1_SEEN(in_dev))
1624 type = IGMP_HOST_MEMBERSHIP_REPORT;
1625 else if (IGMP_V2_SEEN(in_dev))
1626 type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1627 else
1628 type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1629 igmp_send_report(in_dev, im, type);
1630 }
1631 #endif
1632 }
1633
1634 /*
1635 * A socket has left a multicast group on device dev
1636 */
1637
1638 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1639 {
1640 struct ip_mc_list *i;
1641 struct ip_mc_list __rcu **ip;
1642
1643 ASSERT_RTNL();
1644
1645 for (ip = &in_dev->mc_list;
1646 (i = rtnl_dereference(*ip)) != NULL;
1647 ip = &i->next_rcu) {
1648 if (i->multiaddr == addr) {
1649 if (--i->users == 0) {
1650 ip_mc_hash_remove(in_dev, i);
1651 *ip = i->next_rcu;
1652 in_dev->mc_count--;
1653 igmp_group_dropped(i);
1654 ip_mc_clear_src(i);
1655
1656 if (!in_dev->dead)
1657 ip_rt_multicast_event(in_dev);
1658
1659 ip_ma_put(i);
1660 return;
1661 }
1662 break;
1663 }
1664 }
1665 }
1666 EXPORT_SYMBOL(ip_mc_dec_group);
1667
1668 /* Device changing type */
1669
1670 void ip_mc_unmap(struct in_device *in_dev)
1671 {
1672 struct ip_mc_list *pmc;
1673
1674 ASSERT_RTNL();
1675
1676 for_each_pmc_rtnl(in_dev, pmc)
1677 igmp_group_dropped(pmc);
1678 }
1679
1680 void ip_mc_remap(struct in_device *in_dev)
1681 {
1682 struct ip_mc_list *pmc;
1683
1684 ASSERT_RTNL();
1685
1686 for_each_pmc_rtnl(in_dev, pmc) {
1687 #ifdef CONFIG_IP_MULTICAST
1688 igmpv3_del_delrec(in_dev, pmc);
1689 #endif
1690 igmp_group_added(pmc);
1691 }
1692 }
1693
1694 /* Device going down */
1695
1696 void ip_mc_down(struct in_device *in_dev)
1697 {
1698 struct ip_mc_list *pmc;
1699
1700 ASSERT_RTNL();
1701
1702 for_each_pmc_rtnl(in_dev, pmc)
1703 igmp_group_dropped(pmc);
1704
1705 #ifdef CONFIG_IP_MULTICAST
1706 in_dev->mr_ifc_count = 0;
1707 if (del_timer(&in_dev->mr_ifc_timer))
1708 __in_dev_put(in_dev);
1709 in_dev->mr_gq_running = 0;
1710 if (del_timer(&in_dev->mr_gq_timer))
1711 __in_dev_put(in_dev);
1712 #endif
1713
1714 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1715 }
1716
1717 void ip_mc_init_dev(struct in_device *in_dev)
1718 {
1719 #ifdef CONFIG_IP_MULTICAST
1720 struct net *net = dev_net(in_dev->dev);
1721 #endif
1722 ASSERT_RTNL();
1723
1724 #ifdef CONFIG_IP_MULTICAST
1725 setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1726 (unsigned long)in_dev);
1727 setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1728 (unsigned long)in_dev);
1729 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1730 #endif
1731
1732 spin_lock_init(&in_dev->mc_tomb_lock);
1733 }
1734
1735 /* Device going up */
1736
1737 void ip_mc_up(struct in_device *in_dev)
1738 {
1739 struct ip_mc_list *pmc;
1740 #ifdef CONFIG_IP_MULTICAST
1741 struct net *net = dev_net(in_dev->dev);
1742 #endif
1743
1744 ASSERT_RTNL();
1745
1746 #ifdef CONFIG_IP_MULTICAST
1747 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1748 #endif
1749 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1750
1751 for_each_pmc_rtnl(in_dev, pmc) {
1752 #ifdef CONFIG_IP_MULTICAST
1753 igmpv3_del_delrec(in_dev, pmc);
1754 #endif
1755 igmp_group_added(pmc);
1756 }
1757 }
1758
1759 /*
1760 * Device is about to be destroyed: clean up.
1761 */
1762
1763 void ip_mc_destroy_dev(struct in_device *in_dev)
1764 {
1765 struct ip_mc_list *i;
1766
1767 ASSERT_RTNL();
1768
1769 /* Deactivate timers */
1770 ip_mc_down(in_dev);
1771 #ifdef CONFIG_IP_MULTICAST
1772 igmpv3_clear_delrec(in_dev);
1773 #endif
1774
1775 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1776 in_dev->mc_list = i->next_rcu;
1777 in_dev->mc_count--;
1778 ip_ma_put(i);
1779 }
1780 }
1781
1782 /* RTNL is locked */
1783 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1784 {
1785 struct net_device *dev = NULL;
1786 struct in_device *idev = NULL;
1787
1788 if (imr->imr_ifindex) {
1789 idev = inetdev_by_index(net, imr->imr_ifindex);
1790 return idev;
1791 }
1792 if (imr->imr_address.s_addr) {
1793 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1794 if (!dev)
1795 return NULL;
1796 }
1797
1798 if (!dev) {
1799 struct rtable *rt = ip_route_output(net,
1800 imr->imr_multiaddr.s_addr,
1801 0, 0, 0);
1802 if (!IS_ERR(rt)) {
1803 dev = rt->dst.dev;
1804 ip_rt_put(rt);
1805 }
1806 }
1807 if (dev) {
1808 imr->imr_ifindex = dev->ifindex;
1809 idev = __in_dev_get_rtnl(dev);
1810 }
1811 return idev;
1812 }
1813
1814 /*
1815 * Join a socket to a group
1816 */
1817
1818 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1819 __be32 *psfsrc)
1820 {
1821 struct ip_sf_list *psf, *psf_prev;
1822 int rv = 0;
1823
1824 psf_prev = NULL;
1825 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1826 if (psf->sf_inaddr == *psfsrc)
1827 break;
1828 psf_prev = psf;
1829 }
1830 if (!psf || psf->sf_count[sfmode] == 0) {
1831 /* source filter not found, or count wrong => bug */
1832 return -ESRCH;
1833 }
1834 psf->sf_count[sfmode]--;
1835 if (psf->sf_count[sfmode] == 0) {
1836 ip_rt_multicast_event(pmc->interface);
1837 }
1838 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1839 #ifdef CONFIG_IP_MULTICAST
1840 struct in_device *in_dev = pmc->interface;
1841 struct net *net = dev_net(in_dev->dev);
1842 #endif
1843
1844 /* no more filters for this source */
1845 if (psf_prev)
1846 psf_prev->sf_next = psf->sf_next;
1847 else
1848 pmc->sources = psf->sf_next;
1849 #ifdef CONFIG_IP_MULTICAST
1850 if (psf->sf_oldin &&
1851 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1852 psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1853 psf->sf_next = pmc->tomb;
1854 pmc->tomb = psf;
1855 rv = 1;
1856 } else
1857 #endif
1858 kfree(psf);
1859 }
1860 return rv;
1861 }
1862
1863 #ifndef CONFIG_IP_MULTICAST
1864 #define igmp_ifc_event(x) do { } while (0)
1865 #endif
1866
1867 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1868 int sfcount, __be32 *psfsrc, int delta)
1869 {
1870 struct ip_mc_list *pmc;
1871 int changerec = 0;
1872 int i, err;
1873
1874 if (!in_dev)
1875 return -ENODEV;
1876 rcu_read_lock();
1877 for_each_pmc_rcu(in_dev, pmc) {
1878 if (*pmca == pmc->multiaddr)
1879 break;
1880 }
1881 if (!pmc) {
1882 /* MCA not found?? bug */
1883 rcu_read_unlock();
1884 return -ESRCH;
1885 }
1886 spin_lock_bh(&pmc->lock);
1887 rcu_read_unlock();
1888 #ifdef CONFIG_IP_MULTICAST
1889 sf_markstate(pmc);
1890 #endif
1891 if (!delta) {
1892 err = -EINVAL;
1893 if (!pmc->sfcount[sfmode])
1894 goto out_unlock;
1895 pmc->sfcount[sfmode]--;
1896 }
1897 err = 0;
1898 for (i = 0; i < sfcount; i++) {
1899 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1900
1901 changerec |= rv > 0;
1902 if (!err && rv < 0)
1903 err = rv;
1904 }
1905 if (pmc->sfmode == MCAST_EXCLUDE &&
1906 pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1907 pmc->sfcount[MCAST_INCLUDE]) {
1908 #ifdef CONFIG_IP_MULTICAST
1909 struct ip_sf_list *psf;
1910 struct net *net = dev_net(in_dev->dev);
1911 #endif
1912
1913 /* filter mode change */
1914 pmc->sfmode = MCAST_INCLUDE;
1915 #ifdef CONFIG_IP_MULTICAST
1916 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1917 in_dev->mr_ifc_count = pmc->crcount;
1918 for (psf = pmc->sources; psf; psf = psf->sf_next)
1919 psf->sf_crcount = 0;
1920 igmp_ifc_event(pmc->interface);
1921 } else if (sf_setstate(pmc) || changerec) {
1922 igmp_ifc_event(pmc->interface);
1923 #endif
1924 }
1925 out_unlock:
1926 spin_unlock_bh(&pmc->lock);
1927 return err;
1928 }
1929
1930 /*
1931 * Add multicast single-source filter to the interface list
1932 */
1933 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1934 __be32 *psfsrc)
1935 {
1936 struct ip_sf_list *psf, *psf_prev;
1937
1938 psf_prev = NULL;
1939 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1940 if (psf->sf_inaddr == *psfsrc)
1941 break;
1942 psf_prev = psf;
1943 }
1944 if (!psf) {
1945 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1946 if (!psf)
1947 return -ENOBUFS;
1948 psf->sf_inaddr = *psfsrc;
1949 if (psf_prev) {
1950 psf_prev->sf_next = psf;
1951 } else
1952 pmc->sources = psf;
1953 }
1954 psf->sf_count[sfmode]++;
1955 if (psf->sf_count[sfmode] == 1) {
1956 ip_rt_multicast_event(pmc->interface);
1957 }
1958 return 0;
1959 }
1960
1961 #ifdef CONFIG_IP_MULTICAST
1962 static void sf_markstate(struct ip_mc_list *pmc)
1963 {
1964 struct ip_sf_list *psf;
1965 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1966
1967 for (psf = pmc->sources; psf; psf = psf->sf_next)
1968 if (pmc->sfcount[MCAST_EXCLUDE]) {
1969 psf->sf_oldin = mca_xcount ==
1970 psf->sf_count[MCAST_EXCLUDE] &&
1971 !psf->sf_count[MCAST_INCLUDE];
1972 } else
1973 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1974 }
1975
1976 static int sf_setstate(struct ip_mc_list *pmc)
1977 {
1978 struct ip_sf_list *psf, *dpsf;
1979 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1980 int qrv = pmc->interface->mr_qrv;
1981 int new_in, rv;
1982
1983 rv = 0;
1984 for (psf = pmc->sources; psf; psf = psf->sf_next) {
1985 if (pmc->sfcount[MCAST_EXCLUDE]) {
1986 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1987 !psf->sf_count[MCAST_INCLUDE];
1988 } else
1989 new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1990 if (new_in) {
1991 if (!psf->sf_oldin) {
1992 struct ip_sf_list *prev = NULL;
1993
1994 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1995 if (dpsf->sf_inaddr == psf->sf_inaddr)
1996 break;
1997 prev = dpsf;
1998 }
1999 if (dpsf) {
2000 if (prev)
2001 prev->sf_next = dpsf->sf_next;
2002 else
2003 pmc->tomb = dpsf->sf_next;
2004 kfree(dpsf);
2005 }
2006 psf->sf_crcount = qrv;
2007 rv++;
2008 }
2009 } else if (psf->sf_oldin) {
2010
2011 psf->sf_crcount = 0;
2012 /*
2013 * add or update "delete" records if an active filter
2014 * is now inactive
2015 */
2016 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
2017 if (dpsf->sf_inaddr == psf->sf_inaddr)
2018 break;
2019 if (!dpsf) {
2020 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2021 if (!dpsf)
2022 continue;
2023 *dpsf = *psf;
2024 /* pmc->lock held by callers */
2025 dpsf->sf_next = pmc->tomb;
2026 pmc->tomb = dpsf;
2027 }
2028 dpsf->sf_crcount = qrv;
2029 rv++;
2030 }
2031 }
2032 return rv;
2033 }
2034 #endif
2035
2036 /*
2037 * Add multicast source filter list to the interface list
2038 */
2039 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2040 int sfcount, __be32 *psfsrc, int delta)
2041 {
2042 struct ip_mc_list *pmc;
2043 int isexclude;
2044 int i, err;
2045
2046 if (!in_dev)
2047 return -ENODEV;
2048 rcu_read_lock();
2049 for_each_pmc_rcu(in_dev, pmc) {
2050 if (*pmca == pmc->multiaddr)
2051 break;
2052 }
2053 if (!pmc) {
2054 /* MCA not found?? bug */
2055 rcu_read_unlock();
2056 return -ESRCH;
2057 }
2058 spin_lock_bh(&pmc->lock);
2059 rcu_read_unlock();
2060
2061 #ifdef CONFIG_IP_MULTICAST
2062 sf_markstate(pmc);
2063 #endif
2064 isexclude = pmc->sfmode == MCAST_EXCLUDE;
2065 if (!delta)
2066 pmc->sfcount[sfmode]++;
2067 err = 0;
2068 for (i = 0; i < sfcount; i++) {
2069 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2070 if (err)
2071 break;
2072 }
2073 if (err) {
2074 int j;
2075
2076 if (!delta)
2077 pmc->sfcount[sfmode]--;
2078 for (j = 0; j < i; j++)
2079 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2080 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2081 #ifdef CONFIG_IP_MULTICAST
2082 struct ip_sf_list *psf;
2083 struct net *net = dev_net(pmc->interface->dev);
2084 in_dev = pmc->interface;
2085 #endif
2086
2087 /* filter mode change */
2088 if (pmc->sfcount[MCAST_EXCLUDE])
2089 pmc->sfmode = MCAST_EXCLUDE;
2090 else if (pmc->sfcount[MCAST_INCLUDE])
2091 pmc->sfmode = MCAST_INCLUDE;
2092 #ifdef CONFIG_IP_MULTICAST
2093 /* else no filters; keep old mode for reports */
2094
2095 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2096 in_dev->mr_ifc_count = pmc->crcount;
2097 for (psf = pmc->sources; psf; psf = psf->sf_next)
2098 psf->sf_crcount = 0;
2099 igmp_ifc_event(in_dev);
2100 } else if (sf_setstate(pmc)) {
2101 igmp_ifc_event(in_dev);
2102 #endif
2103 }
2104 spin_unlock_bh(&pmc->lock);
2105 return err;
2106 }
2107
2108 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2109 {
2110 struct ip_sf_list *psf, *nextpsf, *tomb, *sources;
2111
2112 spin_lock_bh(&pmc->lock);
2113 tomb = pmc->tomb;
2114 pmc->tomb = NULL;
2115 sources = pmc->sources;
2116 pmc->sources = NULL;
2117 pmc->sfmode = MCAST_EXCLUDE;
2118 pmc->sfcount[MCAST_INCLUDE] = 0;
2119 pmc->sfcount[MCAST_EXCLUDE] = 1;
2120 spin_unlock_bh(&pmc->lock);
2121
2122 for (psf = tomb; psf; psf = nextpsf) {
2123 nextpsf = psf->sf_next;
2124 kfree(psf);
2125 }
2126 for (psf = sources; psf; psf = nextpsf) {
2127 nextpsf = psf->sf_next;
2128 kfree(psf);
2129 }
2130 }
2131
2132 /* Join a multicast group
2133 */
2134
2135 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2136 {
2137 __be32 addr = imr->imr_multiaddr.s_addr;
2138 struct ip_mc_socklist *iml, *i;
2139 struct in_device *in_dev;
2140 struct inet_sock *inet = inet_sk(sk);
2141 struct net *net = sock_net(sk);
2142 int ifindex;
2143 int count = 0;
2144 int err;
2145
2146 ASSERT_RTNL();
2147
2148 if (!ipv4_is_multicast(addr))
2149 return -EINVAL;
2150
2151 in_dev = ip_mc_find_dev(net, imr);
2152
2153 if (!in_dev) {
2154 err = -ENODEV;
2155 goto done;
2156 }
2157
2158 err = -EADDRINUSE;
2159 ifindex = imr->imr_ifindex;
2160 for_each_pmc_rtnl(inet, i) {
2161 if (i->multi.imr_multiaddr.s_addr == addr &&
2162 i->multi.imr_ifindex == ifindex)
2163 goto done;
2164 count++;
2165 }
2166 err = -ENOBUFS;
2167 if (count >= net->ipv4.sysctl_igmp_max_memberships)
2168 goto done;
2169 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2170 if (!iml)
2171 goto done;
2172
2173 memcpy(&iml->multi, imr, sizeof(*imr));
2174 iml->next_rcu = inet->mc_list;
2175 iml->sflist = NULL;
2176 iml->sfmode = MCAST_EXCLUDE;
2177 rcu_assign_pointer(inet->mc_list, iml);
2178 ip_mc_inc_group(in_dev, addr);
2179 err = 0;
2180 done:
2181 return err;
2182 }
2183 EXPORT_SYMBOL(ip_mc_join_group);
2184
2185 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2186 struct in_device *in_dev)
2187 {
2188 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2189 int err;
2190
2191 if (!psf) {
2192 /* any-source empty exclude case */
2193 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2194 iml->sfmode, 0, NULL, 0);
2195 }
2196 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2197 iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2198 RCU_INIT_POINTER(iml->sflist, NULL);
2199 /* decrease mem now to avoid the memleak warning */
2200 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2201 kfree_rcu(psf, rcu);
2202 return err;
2203 }
2204
2205 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2206 {
2207 struct inet_sock *inet = inet_sk(sk);
2208 struct ip_mc_socklist *iml;
2209 struct ip_mc_socklist __rcu **imlp;
2210 struct in_device *in_dev;
2211 struct net *net = sock_net(sk);
2212 __be32 group = imr->imr_multiaddr.s_addr;
2213 u32 ifindex;
2214 int ret = -EADDRNOTAVAIL;
2215
2216 ASSERT_RTNL();
2217
2218 in_dev = ip_mc_find_dev(net, imr);
2219 if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2220 ret = -ENODEV;
2221 goto out;
2222 }
2223 ifindex = imr->imr_ifindex;
2224 for (imlp = &inet->mc_list;
2225 (iml = rtnl_dereference(*imlp)) != NULL;
2226 imlp = &iml->next_rcu) {
2227 if (iml->multi.imr_multiaddr.s_addr != group)
2228 continue;
2229 if (ifindex) {
2230 if (iml->multi.imr_ifindex != ifindex)
2231 continue;
2232 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2233 iml->multi.imr_address.s_addr)
2234 continue;
2235
2236 (void) ip_mc_leave_src(sk, iml, in_dev);
2237
2238 *imlp = iml->next_rcu;
2239
2240 if (in_dev)
2241 ip_mc_dec_group(in_dev, group);
2242
2243 /* decrease mem now to avoid the memleak warning */
2244 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2245 kfree_rcu(iml, rcu);
2246 return 0;
2247 }
2248 out:
2249 return ret;
2250 }
2251 EXPORT_SYMBOL(ip_mc_leave_group);
2252
2253 int ip_mc_source(int add, int omode, struct sock *sk, struct
2254 ip_mreq_source *mreqs, int ifindex)
2255 {
2256 int err;
2257 struct ip_mreqn imr;
2258 __be32 addr = mreqs->imr_multiaddr;
2259 struct ip_mc_socklist *pmc;
2260 struct in_device *in_dev = NULL;
2261 struct inet_sock *inet = inet_sk(sk);
2262 struct ip_sf_socklist *psl;
2263 struct net *net = sock_net(sk);
2264 int leavegroup = 0;
2265 int i, j, rv;
2266
2267 if (!ipv4_is_multicast(addr))
2268 return -EINVAL;
2269
2270 ASSERT_RTNL();
2271
2272 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2273 imr.imr_address.s_addr = mreqs->imr_interface;
2274 imr.imr_ifindex = ifindex;
2275 in_dev = ip_mc_find_dev(net, &imr);
2276
2277 if (!in_dev) {
2278 err = -ENODEV;
2279 goto done;
2280 }
2281 err = -EADDRNOTAVAIL;
2282
2283 for_each_pmc_rtnl(inet, pmc) {
2284 if ((pmc->multi.imr_multiaddr.s_addr ==
2285 imr.imr_multiaddr.s_addr) &&
2286 (pmc->multi.imr_ifindex == imr.imr_ifindex))
2287 break;
2288 }
2289 if (!pmc) { /* must have a prior join */
2290 err = -EINVAL;
2291 goto done;
2292 }
2293 /* if a source filter was set, must be the same mode as before */
2294 if (pmc->sflist) {
2295 if (pmc->sfmode != omode) {
2296 err = -EINVAL;
2297 goto done;
2298 }
2299 } else if (pmc->sfmode != omode) {
2300 /* allow mode switches for empty-set filters */
2301 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2302 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2303 NULL, 0);
2304 pmc->sfmode = omode;
2305 }
2306
2307 psl = rtnl_dereference(pmc->sflist);
2308 if (!add) {
2309 if (!psl)
2310 goto done; /* err = -EADDRNOTAVAIL */
2311 rv = !0;
2312 for (i = 0; i < psl->sl_count; i++) {
2313 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2314 sizeof(__be32));
2315 if (rv == 0)
2316 break;
2317 }
2318 if (rv) /* source not found */
2319 goto done; /* err = -EADDRNOTAVAIL */
2320
2321 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2322 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2323 leavegroup = 1;
2324 goto done;
2325 }
2326
2327 /* update the interface filter */
2328 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2329 &mreqs->imr_sourceaddr, 1);
2330
2331 for (j = i+1; j < psl->sl_count; j++)
2332 psl->sl_addr[j-1] = psl->sl_addr[j];
2333 psl->sl_count--;
2334 err = 0;
2335 goto done;
2336 }
2337 /* else, add a new source to the filter */
2338
2339 if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2340 err = -ENOBUFS;
2341 goto done;
2342 }
2343 if (!psl || psl->sl_count == psl->sl_max) {
2344 struct ip_sf_socklist *newpsl;
2345 int count = IP_SFBLOCK;
2346
2347 if (psl)
2348 count += psl->sl_max;
2349 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2350 if (!newpsl) {
2351 err = -ENOBUFS;
2352 goto done;
2353 }
2354 newpsl->sl_max = count;
2355 newpsl->sl_count = count - IP_SFBLOCK;
2356 if (psl) {
2357 for (i = 0; i < psl->sl_count; i++)
2358 newpsl->sl_addr[i] = psl->sl_addr[i];
2359 /* decrease mem now to avoid the memleak warning */
2360 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2361 kfree_rcu(psl, rcu);
2362 }
2363 rcu_assign_pointer(pmc->sflist, newpsl);
2364 psl = newpsl;
2365 }
2366 rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2367 for (i = 0; i < psl->sl_count; i++) {
2368 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2369 sizeof(__be32));
2370 if (rv == 0)
2371 break;
2372 }
2373 if (rv == 0) /* address already there is an error */
2374 goto done;
2375 for (j = psl->sl_count-1; j >= i; j--)
2376 psl->sl_addr[j+1] = psl->sl_addr[j];
2377 psl->sl_addr[i] = mreqs->imr_sourceaddr;
2378 psl->sl_count++;
2379 err = 0;
2380 /* update the interface list */
2381 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2382 &mreqs->imr_sourceaddr, 1);
2383 done:
2384 if (leavegroup)
2385 err = ip_mc_leave_group(sk, &imr);
2386 return err;
2387 }
2388
2389 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2390 {
2391 int err = 0;
2392 struct ip_mreqn imr;
2393 __be32 addr = msf->imsf_multiaddr;
2394 struct ip_mc_socklist *pmc;
2395 struct in_device *in_dev;
2396 struct inet_sock *inet = inet_sk(sk);
2397 struct ip_sf_socklist *newpsl, *psl;
2398 struct net *net = sock_net(sk);
2399 int leavegroup = 0;
2400
2401 if (!ipv4_is_multicast(addr))
2402 return -EINVAL;
2403 if (msf->imsf_fmode != MCAST_INCLUDE &&
2404 msf->imsf_fmode != MCAST_EXCLUDE)
2405 return -EINVAL;
2406
2407 ASSERT_RTNL();
2408
2409 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2410 imr.imr_address.s_addr = msf->imsf_interface;
2411 imr.imr_ifindex = ifindex;
2412 in_dev = ip_mc_find_dev(net, &imr);
2413
2414 if (!in_dev) {
2415 err = -ENODEV;
2416 goto done;
2417 }
2418
2419 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2420 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2421 leavegroup = 1;
2422 goto done;
2423 }
2424
2425 for_each_pmc_rtnl(inet, pmc) {
2426 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2427 pmc->multi.imr_ifindex == imr.imr_ifindex)
2428 break;
2429 }
2430 if (!pmc) { /* must have a prior join */
2431 err = -EINVAL;
2432 goto done;
2433 }
2434 if (msf->imsf_numsrc) {
2435 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2436 GFP_KERNEL);
2437 if (!newpsl) {
2438 err = -ENOBUFS;
2439 goto done;
2440 }
2441 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2442 memcpy(newpsl->sl_addr, msf->imsf_slist,
2443 msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2444 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2445 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2446 if (err) {
2447 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2448 goto done;
2449 }
2450 } else {
2451 newpsl = NULL;
2452 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2453 msf->imsf_fmode, 0, NULL, 0);
2454 }
2455 psl = rtnl_dereference(pmc->sflist);
2456 if (psl) {
2457 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2458 psl->sl_count, psl->sl_addr, 0);
2459 /* decrease mem now to avoid the memleak warning */
2460 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2461 kfree_rcu(psl, rcu);
2462 } else
2463 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2464 0, NULL, 0);
2465 rcu_assign_pointer(pmc->sflist, newpsl);
2466 pmc->sfmode = msf->imsf_fmode;
2467 err = 0;
2468 done:
2469 if (leavegroup)
2470 err = ip_mc_leave_group(sk, &imr);
2471 return err;
2472 }
2473
2474 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2475 struct ip_msfilter __user *optval, int __user *optlen)
2476 {
2477 int err, len, count, copycount;
2478 struct ip_mreqn imr;
2479 __be32 addr = msf->imsf_multiaddr;
2480 struct ip_mc_socklist *pmc;
2481 struct in_device *in_dev;
2482 struct inet_sock *inet = inet_sk(sk);
2483 struct ip_sf_socklist *psl;
2484 struct net *net = sock_net(sk);
2485
2486 ASSERT_RTNL();
2487
2488 if (!ipv4_is_multicast(addr))
2489 return -EINVAL;
2490
2491 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2492 imr.imr_address.s_addr = msf->imsf_interface;
2493 imr.imr_ifindex = 0;
2494 in_dev = ip_mc_find_dev(net, &imr);
2495
2496 if (!in_dev) {
2497 err = -ENODEV;
2498 goto done;
2499 }
2500 err = -EADDRNOTAVAIL;
2501
2502 for_each_pmc_rtnl(inet, pmc) {
2503 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2504 pmc->multi.imr_ifindex == imr.imr_ifindex)
2505 break;
2506 }
2507 if (!pmc) /* must have a prior join */
2508 goto done;
2509 msf->imsf_fmode = pmc->sfmode;
2510 psl = rtnl_dereference(pmc->sflist);
2511 if (!psl) {
2512 len = 0;
2513 count = 0;
2514 } else {
2515 count = psl->sl_count;
2516 }
2517 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2518 len = copycount * sizeof(psl->sl_addr[0]);
2519 msf->imsf_numsrc = count;
2520 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2521 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2522 return -EFAULT;
2523 }
2524 if (len &&
2525 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2526 return -EFAULT;
2527 return 0;
2528 done:
2529 return err;
2530 }
2531
2532 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2533 struct group_filter __user *optval, int __user *optlen)
2534 {
2535 int err, i, count, copycount;
2536 struct sockaddr_in *psin;
2537 __be32 addr;
2538 struct ip_mc_socklist *pmc;
2539 struct inet_sock *inet = inet_sk(sk);
2540 struct ip_sf_socklist *psl;
2541
2542 ASSERT_RTNL();
2543
2544 psin = (struct sockaddr_in *)&gsf->gf_group;
2545 if (psin->sin_family != AF_INET)
2546 return -EINVAL;
2547 addr = psin->sin_addr.s_addr;
2548 if (!ipv4_is_multicast(addr))
2549 return -EINVAL;
2550
2551 err = -EADDRNOTAVAIL;
2552
2553 for_each_pmc_rtnl(inet, pmc) {
2554 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2555 pmc->multi.imr_ifindex == gsf->gf_interface)
2556 break;
2557 }
2558 if (!pmc) /* must have a prior join */
2559 goto done;
2560 gsf->gf_fmode = pmc->sfmode;
2561 psl = rtnl_dereference(pmc->sflist);
2562 count = psl ? psl->sl_count : 0;
2563 copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2564 gsf->gf_numsrc = count;
2565 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2566 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2567 return -EFAULT;
2568 }
2569 for (i = 0; i < copycount; i++) {
2570 struct sockaddr_storage ss;
2571
2572 psin = (struct sockaddr_in *)&ss;
2573 memset(&ss, 0, sizeof(ss));
2574 psin->sin_family = AF_INET;
2575 psin->sin_addr.s_addr = psl->sl_addr[i];
2576 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2577 return -EFAULT;
2578 }
2579 return 0;
2580 done:
2581 return err;
2582 }
2583
2584 /*
2585 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2586 */
2587 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr,
2588 int dif, int sdif)
2589 {
2590 struct inet_sock *inet = inet_sk(sk);
2591 struct ip_mc_socklist *pmc;
2592 struct ip_sf_socklist *psl;
2593 int i;
2594 int ret;
2595
2596 ret = 1;
2597 if (!ipv4_is_multicast(loc_addr))
2598 goto out;
2599
2600 rcu_read_lock();
2601 for_each_pmc_rcu(inet, pmc) {
2602 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2603 (pmc->multi.imr_ifindex == dif ||
2604 (sdif && pmc->multi.imr_ifindex == sdif)))
2605 break;
2606 }
2607 ret = inet->mc_all;
2608 if (!pmc)
2609 goto unlock;
2610 psl = rcu_dereference(pmc->sflist);
2611 ret = (pmc->sfmode == MCAST_EXCLUDE);
2612 if (!psl)
2613 goto unlock;
2614
2615 for (i = 0; i < psl->sl_count; i++) {
2616 if (psl->sl_addr[i] == rmt_addr)
2617 break;
2618 }
2619 ret = 0;
2620 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2621 goto unlock;
2622 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2623 goto unlock;
2624 ret = 1;
2625 unlock:
2626 rcu_read_unlock();
2627 out:
2628 return ret;
2629 }
2630
2631 /*
2632 * A socket is closing.
2633 */
2634
2635 void ip_mc_drop_socket(struct sock *sk)
2636 {
2637 struct inet_sock *inet = inet_sk(sk);
2638 struct ip_mc_socklist *iml;
2639 struct net *net = sock_net(sk);
2640
2641 if (!inet->mc_list)
2642 return;
2643
2644 rtnl_lock();
2645 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2646 struct in_device *in_dev;
2647
2648 inet->mc_list = iml->next_rcu;
2649 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2650 (void) ip_mc_leave_src(sk, iml, in_dev);
2651 if (in_dev)
2652 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2653 /* decrease mem now to avoid the memleak warning */
2654 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2655 kfree_rcu(iml, rcu);
2656 }
2657 rtnl_unlock();
2658 }
2659
2660 /* called with rcu_read_lock() */
2661 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2662 {
2663 struct ip_mc_list *im;
2664 struct ip_mc_list __rcu **mc_hash;
2665 struct ip_sf_list *psf;
2666 int rv = 0;
2667
2668 mc_hash = rcu_dereference(in_dev->mc_hash);
2669 if (mc_hash) {
2670 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2671
2672 for (im = rcu_dereference(mc_hash[hash]);
2673 im != NULL;
2674 im = rcu_dereference(im->next_hash)) {
2675 if (im->multiaddr == mc_addr)
2676 break;
2677 }
2678 } else {
2679 for_each_pmc_rcu(in_dev, im) {
2680 if (im->multiaddr == mc_addr)
2681 break;
2682 }
2683 }
2684 if (im && proto == IPPROTO_IGMP) {
2685 rv = 1;
2686 } else if (im) {
2687 if (src_addr) {
2688 for (psf = im->sources; psf; psf = psf->sf_next) {
2689 if (psf->sf_inaddr == src_addr)
2690 break;
2691 }
2692 if (psf)
2693 rv = psf->sf_count[MCAST_INCLUDE] ||
2694 psf->sf_count[MCAST_EXCLUDE] !=
2695 im->sfcount[MCAST_EXCLUDE];
2696 else
2697 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2698 } else
2699 rv = 1; /* unspecified source; tentatively allow */
2700 }
2701 return rv;
2702 }
2703
2704 #if defined(CONFIG_PROC_FS)
2705 struct igmp_mc_iter_state {
2706 struct seq_net_private p;
2707 struct net_device *dev;
2708 struct in_device *in_dev;
2709 };
2710
2711 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private)
2712
2713 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2714 {
2715 struct net *net = seq_file_net(seq);
2716 struct ip_mc_list *im = NULL;
2717 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2718
2719 state->in_dev = NULL;
2720 for_each_netdev_rcu(net, state->dev) {
2721 struct in_device *in_dev;
2722
2723 in_dev = __in_dev_get_rcu(state->dev);
2724 if (!in_dev)
2725 continue;
2726 im = rcu_dereference(in_dev->mc_list);
2727 if (im) {
2728 state->in_dev = in_dev;
2729 break;
2730 }
2731 }
2732 return im;
2733 }
2734
2735 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2736 {
2737 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2738
2739 im = rcu_dereference(im->next_rcu);
2740 while (!im) {
2741 state->dev = next_net_device_rcu(state->dev);
2742 if (!state->dev) {
2743 state->in_dev = NULL;
2744 break;
2745 }
2746 state->in_dev = __in_dev_get_rcu(state->dev);
2747 if (!state->in_dev)
2748 continue;
2749 im = rcu_dereference(state->in_dev->mc_list);
2750 }
2751 return im;
2752 }
2753
2754 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2755 {
2756 struct ip_mc_list *im = igmp_mc_get_first(seq);
2757 if (im)
2758 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2759 --pos;
2760 return pos ? NULL : im;
2761 }
2762
2763 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2764 __acquires(rcu)
2765 {
2766 rcu_read_lock();
2767 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2768 }
2769
2770 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2771 {
2772 struct ip_mc_list *im;
2773 if (v == SEQ_START_TOKEN)
2774 im = igmp_mc_get_first(seq);
2775 else
2776 im = igmp_mc_get_next(seq, v);
2777 ++*pos;
2778 return im;
2779 }
2780
2781 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2782 __releases(rcu)
2783 {
2784 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2785
2786 state->in_dev = NULL;
2787 state->dev = NULL;
2788 rcu_read_unlock();
2789 }
2790
2791 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2792 {
2793 if (v == SEQ_START_TOKEN)
2794 seq_puts(seq,
2795 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n");
2796 else {
2797 struct ip_mc_list *im = (struct ip_mc_list *)v;
2798 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2799 char *querier;
2800 long delta;
2801
2802 #ifdef CONFIG_IP_MULTICAST
2803 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2804 IGMP_V2_SEEN(state->in_dev) ? "V2" :
2805 "V3";
2806 #else
2807 querier = "NONE";
2808 #endif
2809
2810 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2811 seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2812 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2813 }
2814
2815 delta = im->timer.expires - jiffies;
2816 seq_printf(seq,
2817 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2818 im->multiaddr, im->users,
2819 im->tm_running,
2820 im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2821 im->reporter);
2822 }
2823 return 0;
2824 }
2825
2826 static const struct seq_operations igmp_mc_seq_ops = {
2827 .start = igmp_mc_seq_start,
2828 .next = igmp_mc_seq_next,
2829 .stop = igmp_mc_seq_stop,
2830 .show = igmp_mc_seq_show,
2831 };
2832
2833 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2834 {
2835 return seq_open_net(inode, file, &igmp_mc_seq_ops,
2836 sizeof(struct igmp_mc_iter_state));
2837 }
2838
2839 static const struct file_operations igmp_mc_seq_fops = {
2840 .owner = THIS_MODULE,
2841 .open = igmp_mc_seq_open,
2842 .read = seq_read,
2843 .llseek = seq_lseek,
2844 .release = seq_release_net,
2845 };
2846
2847 struct igmp_mcf_iter_state {
2848 struct seq_net_private p;
2849 struct net_device *dev;
2850 struct in_device *idev;
2851 struct ip_mc_list *im;
2852 };
2853
2854 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private)
2855
2856 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2857 {
2858 struct net *net = seq_file_net(seq);
2859 struct ip_sf_list *psf = NULL;
2860 struct ip_mc_list *im = NULL;
2861 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2862
2863 state->idev = NULL;
2864 state->im = NULL;
2865 for_each_netdev_rcu(net, state->dev) {
2866 struct in_device *idev;
2867 idev = __in_dev_get_rcu(state->dev);
2868 if (unlikely(!idev))
2869 continue;
2870 im = rcu_dereference(idev->mc_list);
2871 if (likely(im)) {
2872 spin_lock_bh(&im->lock);
2873 psf = im->sources;
2874 if (likely(psf)) {
2875 state->im = im;
2876 state->idev = idev;
2877 break;
2878 }
2879 spin_unlock_bh(&im->lock);
2880 }
2881 }
2882 return psf;
2883 }
2884
2885 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2886 {
2887 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2888
2889 psf = psf->sf_next;
2890 while (!psf) {
2891 spin_unlock_bh(&state->im->lock);
2892 state->im = state->im->next;
2893 while (!state->im) {
2894 state->dev = next_net_device_rcu(state->dev);
2895 if (!state->dev) {
2896 state->idev = NULL;
2897 goto out;
2898 }
2899 state->idev = __in_dev_get_rcu(state->dev);
2900 if (!state->idev)
2901 continue;
2902 state->im = rcu_dereference(state->idev->mc_list);
2903 }
2904 if (!state->im)
2905 break;
2906 spin_lock_bh(&state->im->lock);
2907 psf = state->im->sources;
2908 }
2909 out:
2910 return psf;
2911 }
2912
2913 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2914 {
2915 struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2916 if (psf)
2917 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2918 --pos;
2919 return pos ? NULL : psf;
2920 }
2921
2922 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2923 __acquires(rcu)
2924 {
2925 rcu_read_lock();
2926 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2927 }
2928
2929 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2930 {
2931 struct ip_sf_list *psf;
2932 if (v == SEQ_START_TOKEN)
2933 psf = igmp_mcf_get_first(seq);
2934 else
2935 psf = igmp_mcf_get_next(seq, v);
2936 ++*pos;
2937 return psf;
2938 }
2939
2940 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2941 __releases(rcu)
2942 {
2943 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2944 if (likely(state->im)) {
2945 spin_unlock_bh(&state->im->lock);
2946 state->im = NULL;
2947 }
2948 state->idev = NULL;
2949 state->dev = NULL;
2950 rcu_read_unlock();
2951 }
2952
2953 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2954 {
2955 struct ip_sf_list *psf = (struct ip_sf_list *)v;
2956 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2957
2958 if (v == SEQ_START_TOKEN) {
2959 seq_puts(seq, "Idx Device MCA SRC INC EXC\n");
2960 } else {
2961 seq_printf(seq,
2962 "%3d %6.6s 0x%08x "
2963 "0x%08x %6lu %6lu\n",
2964 state->dev->ifindex, state->dev->name,
2965 ntohl(state->im->multiaddr),
2966 ntohl(psf->sf_inaddr),
2967 psf->sf_count[MCAST_INCLUDE],
2968 psf->sf_count[MCAST_EXCLUDE]);
2969 }
2970 return 0;
2971 }
2972
2973 static const struct seq_operations igmp_mcf_seq_ops = {
2974 .start = igmp_mcf_seq_start,
2975 .next = igmp_mcf_seq_next,
2976 .stop = igmp_mcf_seq_stop,
2977 .show = igmp_mcf_seq_show,
2978 };
2979
2980 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2981 {
2982 return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2983 sizeof(struct igmp_mcf_iter_state));
2984 }
2985
2986 static const struct file_operations igmp_mcf_seq_fops = {
2987 .owner = THIS_MODULE,
2988 .open = igmp_mcf_seq_open,
2989 .read = seq_read,
2990 .llseek = seq_lseek,
2991 .release = seq_release_net,
2992 };
2993
2994 static int __net_init igmp_net_init(struct net *net)
2995 {
2996 struct proc_dir_entry *pde;
2997 int err;
2998
2999 pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
3000 if (!pde)
3001 goto out_igmp;
3002 pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
3003 &igmp_mcf_seq_fops);
3004 if (!pde)
3005 goto out_mcfilter;
3006 err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
3007 SOCK_DGRAM, 0, net);
3008 if (err < 0) {
3009 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
3010 err);
3011 goto out_sock;
3012 }
3013
3014 return 0;
3015
3016 out_sock:
3017 remove_proc_entry("mcfilter", net->proc_net);
3018 out_mcfilter:
3019 remove_proc_entry("igmp", net->proc_net);
3020 out_igmp:
3021 return -ENOMEM;
3022 }
3023
3024 static void __net_exit igmp_net_exit(struct net *net)
3025 {
3026 remove_proc_entry("mcfilter", net->proc_net);
3027 remove_proc_entry("igmp", net->proc_net);
3028 inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
3029 }
3030
3031 static struct pernet_operations igmp_net_ops = {
3032 .init = igmp_net_init,
3033 .exit = igmp_net_exit,
3034 };
3035 #endif
3036
3037 static int igmp_netdev_event(struct notifier_block *this,
3038 unsigned long event, void *ptr)
3039 {
3040 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3041 struct in_device *in_dev;
3042
3043 switch (event) {
3044 case NETDEV_RESEND_IGMP:
3045 in_dev = __in_dev_get_rtnl(dev);
3046 if (in_dev)
3047 ip_mc_rejoin_groups(in_dev);
3048 break;
3049 default:
3050 break;
3051 }
3052 return NOTIFY_DONE;
3053 }
3054
3055 static struct notifier_block igmp_notifier = {
3056 .notifier_call = igmp_netdev_event,
3057 };
3058
3059 int __init igmp_mc_init(void)
3060 {
3061 #if defined(CONFIG_PROC_FS)
3062 int err;
3063
3064 err = register_pernet_subsys(&igmp_net_ops);
3065 if (err)
3066 return err;
3067 err = register_netdevice_notifier(&igmp_notifier);
3068 if (err)
3069 goto reg_notif_fail;
3070 return 0;
3071
3072 reg_notif_fail:
3073 unregister_pernet_subsys(&igmp_net_ops);
3074 return err;
3075 #else
3076 return register_netdevice_notifier(&igmp_notifier);
3077 #endif
3078 }
Linux with added FPC-III support