xref: /linux/net/ipv4/igmp.c (revision b85d45947951d23cb22d90caecf4c1eb81342c96)
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 
93 #include <net/net_namespace.h>
94 #include <net/arp.h>
95 #include <net/ip.h>
96 #include <net/protocol.h>
97 #include <net/route.h>
98 #include <net/sock.h>
99 #include <net/checksum.h>
100 #include <net/inet_common.h>
101 #include <linux/netfilter_ipv4.h>
102 #ifdef CONFIG_IP_MROUTE
103 #include <linux/mroute.h>
104 #endif
105 #ifdef CONFIG_PROC_FS
106 #include <linux/proc_fs.h>
107 #include <linux/seq_file.h>
108 #endif
109 
110 #define IP_MAX_MEMBERSHIPS	20
111 #define IP_MAX_MSF		10
112 
113 /* IGMP reports for link-local multicast groups are enabled by default */
114 int sysctl_igmp_llm_reports __read_mostly = 1;
115 
116 #ifdef CONFIG_IP_MULTICAST
117 /* Parameter names and values are taken from igmp-v2-06 draft */
118 
119 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT		(400*HZ)
120 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT		(400*HZ)
121 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL	(10*HZ)
122 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL	(1*HZ)
123 #define IGMP_QUERY_RESPONSE_INTERVAL		(10*HZ)
124 #define IGMP_QUERY_ROBUSTNESS_VARIABLE		2
125 
126 
127 #define IGMP_INITIAL_REPORT_DELAY		(1)
128 
129 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
130  * IGMP specs require to report membership immediately after
131  * joining a group, but we delay the first report by a
132  * small interval. It seems more natural and still does not
133  * contradict to specs provided this delay is small enough.
134  */
135 
136 #define IGMP_V1_SEEN(in_dev) \
137 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
138 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
139 	 ((in_dev)->mr_v1_seen && \
140 	  time_before(jiffies, (in_dev)->mr_v1_seen)))
141 #define IGMP_V2_SEEN(in_dev) \
142 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
143 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
144 	 ((in_dev)->mr_v2_seen && \
145 	  time_before(jiffies, (in_dev)->mr_v2_seen)))
146 
147 static int unsolicited_report_interval(struct in_device *in_dev)
148 {
149 	int interval_ms, interval_jiffies;
150 
151 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
152 		interval_ms = IN_DEV_CONF_GET(
153 			in_dev,
154 			IGMPV2_UNSOLICITED_REPORT_INTERVAL);
155 	else /* v3 */
156 		interval_ms = IN_DEV_CONF_GET(
157 			in_dev,
158 			IGMPV3_UNSOLICITED_REPORT_INTERVAL);
159 
160 	interval_jiffies = msecs_to_jiffies(interval_ms);
161 
162 	/* _timer functions can't handle a delay of 0 jiffies so ensure
163 	 *  we always return a positive value.
164 	 */
165 	if (interval_jiffies <= 0)
166 		interval_jiffies = 1;
167 	return interval_jiffies;
168 }
169 
170 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
171 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
172 static void igmpv3_clear_delrec(struct in_device *in_dev);
173 static int sf_setstate(struct ip_mc_list *pmc);
174 static void sf_markstate(struct ip_mc_list *pmc);
175 #endif
176 static void ip_mc_clear_src(struct ip_mc_list *pmc);
177 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
178 			 int sfcount, __be32 *psfsrc, int delta);
179 
180 static void ip_ma_put(struct ip_mc_list *im)
181 {
182 	if (atomic_dec_and_test(&im->refcnt)) {
183 		in_dev_put(im->interface);
184 		kfree_rcu(im, rcu);
185 	}
186 }
187 
188 #define for_each_pmc_rcu(in_dev, pmc)				\
189 	for (pmc = rcu_dereference(in_dev->mc_list);		\
190 	     pmc != NULL;					\
191 	     pmc = rcu_dereference(pmc->next_rcu))
192 
193 #define for_each_pmc_rtnl(in_dev, pmc)				\
194 	for (pmc = rtnl_dereference(in_dev->mc_list);		\
195 	     pmc != NULL;					\
196 	     pmc = rtnl_dereference(pmc->next_rcu))
197 
198 #ifdef CONFIG_IP_MULTICAST
199 
200 /*
201  *	Timer management
202  */
203 
204 static void igmp_stop_timer(struct ip_mc_list *im)
205 {
206 	spin_lock_bh(&im->lock);
207 	if (del_timer(&im->timer))
208 		atomic_dec(&im->refcnt);
209 	im->tm_running = 0;
210 	im->reporter = 0;
211 	im->unsolicit_count = 0;
212 	spin_unlock_bh(&im->lock);
213 }
214 
215 /* It must be called with locked im->lock */
216 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
217 {
218 	int tv = prandom_u32() % max_delay;
219 
220 	im->tm_running = 1;
221 	if (!mod_timer(&im->timer, jiffies+tv+2))
222 		atomic_inc(&im->refcnt);
223 }
224 
225 static void igmp_gq_start_timer(struct in_device *in_dev)
226 {
227 	int tv = prandom_u32() % in_dev->mr_maxdelay;
228 
229 	in_dev->mr_gq_running = 1;
230 	if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
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 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
326 {
327 	struct sk_buff *skb;
328 	struct rtable *rt;
329 	struct iphdr *pip;
330 	struct igmpv3_report *pig;
331 	struct net *net = dev_net(dev);
332 	struct flowi4 fl4;
333 	int hlen = LL_RESERVED_SPACE(dev);
334 	int tlen = dev->needed_tailroom;
335 	unsigned int size = mtu;
336 
337 	while (1) {
338 		skb = alloc_skb(size + hlen + tlen,
339 				GFP_ATOMIC | __GFP_NOWARN);
340 		if (skb)
341 			break;
342 		size >>= 1;
343 		if (size < 256)
344 			return NULL;
345 	}
346 	skb->priority = TC_PRIO_CONTROL;
347 
348 	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
349 				   0, 0,
350 				   IPPROTO_IGMP, 0, dev->ifindex);
351 	if (IS_ERR(rt)) {
352 		kfree_skb(skb);
353 		return NULL;
354 	}
355 
356 	skb_dst_set(skb, &rt->dst);
357 	skb->dev = dev;
358 
359 	skb->reserved_tailroom = skb_end_offset(skb) -
360 				 min(mtu, skb_end_offset(skb));
361 	skb_reserve(skb, hlen);
362 
363 	skb_reset_network_header(skb);
364 	pip = ip_hdr(skb);
365 	skb_put(skb, sizeof(struct iphdr) + 4);
366 
367 	pip->version  = 4;
368 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
369 	pip->tos      = 0xc0;
370 	pip->frag_off = htons(IP_DF);
371 	pip->ttl      = 1;
372 	pip->daddr    = fl4.daddr;
373 	pip->saddr    = fl4.saddr;
374 	pip->protocol = IPPROTO_IGMP;
375 	pip->tot_len  = 0;	/* filled in later */
376 	ip_select_ident(net, skb, NULL);
377 	((u8 *)&pip[1])[0] = IPOPT_RA;
378 	((u8 *)&pip[1])[1] = 4;
379 	((u8 *)&pip[1])[2] = 0;
380 	((u8 *)&pip[1])[3] = 0;
381 
382 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
383 	skb_put(skb, sizeof(*pig));
384 	pig = igmpv3_report_hdr(skb);
385 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
386 	pig->resv1 = 0;
387 	pig->csum = 0;
388 	pig->resv2 = 0;
389 	pig->ngrec = 0;
390 	return skb;
391 }
392 
393 static int igmpv3_sendpack(struct sk_buff *skb)
394 {
395 	struct igmphdr *pig = igmp_hdr(skb);
396 	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
397 
398 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
399 
400 	return ip_local_out(skb);
401 }
402 
403 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
404 {
405 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
406 }
407 
408 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
409 	int type, struct igmpv3_grec **ppgr)
410 {
411 	struct net_device *dev = pmc->interface->dev;
412 	struct igmpv3_report *pih;
413 	struct igmpv3_grec *pgr;
414 
415 	if (!skb)
416 		skb = igmpv3_newpack(dev, dev->mtu);
417 	if (!skb)
418 		return NULL;
419 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
420 	pgr->grec_type = type;
421 	pgr->grec_auxwords = 0;
422 	pgr->grec_nsrcs = 0;
423 	pgr->grec_mca = pmc->multiaddr;
424 	pih = igmpv3_report_hdr(skb);
425 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
426 	*ppgr = pgr;
427 	return skb;
428 }
429 
430 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
431 
432 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
433 	int type, int gdeleted, int sdeleted)
434 {
435 	struct net_device *dev = pmc->interface->dev;
436 	struct igmpv3_report *pih;
437 	struct igmpv3_grec *pgr = NULL;
438 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
439 	int scount, stotal, first, isquery, truncate;
440 
441 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
442 		return skb;
443 	if (ipv4_is_local_multicast(pmc->multiaddr) && !sysctl_igmp_llm_reports)
444 		return skb;
445 
446 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
447 		  type == IGMPV3_MODE_IS_EXCLUDE;
448 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
449 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
450 
451 	stotal = scount = 0;
452 
453 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
454 
455 	if (!*psf_list)
456 		goto empty_source;
457 
458 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
459 
460 	/* EX and TO_EX get a fresh packet, if needed */
461 	if (truncate) {
462 		if (pih && pih->ngrec &&
463 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
464 			if (skb)
465 				igmpv3_sendpack(skb);
466 			skb = igmpv3_newpack(dev, dev->mtu);
467 		}
468 	}
469 	first = 1;
470 	psf_prev = NULL;
471 	for (psf = *psf_list; psf; psf = psf_next) {
472 		__be32 *psrc;
473 
474 		psf_next = psf->sf_next;
475 
476 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
477 			psf_prev = psf;
478 			continue;
479 		}
480 
481 		/* clear marks on query responses */
482 		if (isquery)
483 			psf->sf_gsresp = 0;
484 
485 		if (AVAILABLE(skb) < sizeof(__be32) +
486 		    first*sizeof(struct igmpv3_grec)) {
487 			if (truncate && !first)
488 				break;	 /* truncate these */
489 			if (pgr)
490 				pgr->grec_nsrcs = htons(scount);
491 			if (skb)
492 				igmpv3_sendpack(skb);
493 			skb = igmpv3_newpack(dev, dev->mtu);
494 			first = 1;
495 			scount = 0;
496 		}
497 		if (first) {
498 			skb = add_grhead(skb, pmc, type, &pgr);
499 			first = 0;
500 		}
501 		if (!skb)
502 			return NULL;
503 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
504 		*psrc = psf->sf_inaddr;
505 		scount++; stotal++;
506 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
507 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
508 			psf->sf_crcount--;
509 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
510 				if (psf_prev)
511 					psf_prev->sf_next = psf->sf_next;
512 				else
513 					*psf_list = psf->sf_next;
514 				kfree(psf);
515 				continue;
516 			}
517 		}
518 		psf_prev = psf;
519 	}
520 
521 empty_source:
522 	if (!stotal) {
523 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
524 		    type == IGMPV3_BLOCK_OLD_SOURCES)
525 			return skb;
526 		if (pmc->crcount || isquery) {
527 			/* make sure we have room for group header */
528 			if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
529 				igmpv3_sendpack(skb);
530 				skb = NULL; /* add_grhead will get a new one */
531 			}
532 			skb = add_grhead(skb, pmc, type, &pgr);
533 		}
534 	}
535 	if (pgr)
536 		pgr->grec_nsrcs = htons(scount);
537 
538 	if (isquery)
539 		pmc->gsquery = 0;	/* clear query state on report */
540 	return skb;
541 }
542 
543 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
544 {
545 	struct sk_buff *skb = NULL;
546 	int type;
547 
548 	if (!pmc) {
549 		rcu_read_lock();
550 		for_each_pmc_rcu(in_dev, pmc) {
551 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
552 				continue;
553 			if (ipv4_is_local_multicast(pmc->multiaddr) &&
554 			     !sysctl_igmp_llm_reports)
555 				continue;
556 			spin_lock_bh(&pmc->lock);
557 			if (pmc->sfcount[MCAST_EXCLUDE])
558 				type = IGMPV3_MODE_IS_EXCLUDE;
559 			else
560 				type = IGMPV3_MODE_IS_INCLUDE;
561 			skb = add_grec(skb, pmc, type, 0, 0);
562 			spin_unlock_bh(&pmc->lock);
563 		}
564 		rcu_read_unlock();
565 	} else {
566 		spin_lock_bh(&pmc->lock);
567 		if (pmc->sfcount[MCAST_EXCLUDE])
568 			type = IGMPV3_MODE_IS_EXCLUDE;
569 		else
570 			type = IGMPV3_MODE_IS_INCLUDE;
571 		skb = add_grec(skb, pmc, type, 0, 0);
572 		spin_unlock_bh(&pmc->lock);
573 	}
574 	if (!skb)
575 		return 0;
576 	return igmpv3_sendpack(skb);
577 }
578 
579 /*
580  * remove zero-count source records from a source filter list
581  */
582 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
583 {
584 	struct ip_sf_list *psf_prev, *psf_next, *psf;
585 
586 	psf_prev = NULL;
587 	for (psf = *ppsf; psf; psf = psf_next) {
588 		psf_next = psf->sf_next;
589 		if (psf->sf_crcount == 0) {
590 			if (psf_prev)
591 				psf_prev->sf_next = psf->sf_next;
592 			else
593 				*ppsf = psf->sf_next;
594 			kfree(psf);
595 		} else
596 			psf_prev = psf;
597 	}
598 }
599 
600 static void igmpv3_send_cr(struct in_device *in_dev)
601 {
602 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
603 	struct sk_buff *skb = NULL;
604 	int type, dtype;
605 
606 	rcu_read_lock();
607 	spin_lock_bh(&in_dev->mc_tomb_lock);
608 
609 	/* deleted MCA's */
610 	pmc_prev = NULL;
611 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
612 		pmc_next = pmc->next;
613 		if (pmc->sfmode == MCAST_INCLUDE) {
614 			type = IGMPV3_BLOCK_OLD_SOURCES;
615 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
616 			skb = add_grec(skb, pmc, type, 1, 0);
617 			skb = add_grec(skb, pmc, dtype, 1, 1);
618 		}
619 		if (pmc->crcount) {
620 			if (pmc->sfmode == MCAST_EXCLUDE) {
621 				type = IGMPV3_CHANGE_TO_INCLUDE;
622 				skb = add_grec(skb, pmc, type, 1, 0);
623 			}
624 			pmc->crcount--;
625 			if (pmc->crcount == 0) {
626 				igmpv3_clear_zeros(&pmc->tomb);
627 				igmpv3_clear_zeros(&pmc->sources);
628 			}
629 		}
630 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
631 			if (pmc_prev)
632 				pmc_prev->next = pmc_next;
633 			else
634 				in_dev->mc_tomb = pmc_next;
635 			in_dev_put(pmc->interface);
636 			kfree(pmc);
637 		} else
638 			pmc_prev = pmc;
639 	}
640 	spin_unlock_bh(&in_dev->mc_tomb_lock);
641 
642 	/* change recs */
643 	for_each_pmc_rcu(in_dev, pmc) {
644 		spin_lock_bh(&pmc->lock);
645 		if (pmc->sfcount[MCAST_EXCLUDE]) {
646 			type = IGMPV3_BLOCK_OLD_SOURCES;
647 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
648 		} else {
649 			type = IGMPV3_ALLOW_NEW_SOURCES;
650 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
651 		}
652 		skb = add_grec(skb, pmc, type, 0, 0);
653 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
654 
655 		/* filter mode changes */
656 		if (pmc->crcount) {
657 			if (pmc->sfmode == MCAST_EXCLUDE)
658 				type = IGMPV3_CHANGE_TO_EXCLUDE;
659 			else
660 				type = IGMPV3_CHANGE_TO_INCLUDE;
661 			skb = add_grec(skb, pmc, type, 0, 0);
662 			pmc->crcount--;
663 		}
664 		spin_unlock_bh(&pmc->lock);
665 	}
666 	rcu_read_unlock();
667 
668 	if (!skb)
669 		return;
670 	(void) igmpv3_sendpack(skb);
671 }
672 
673 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
674 	int type)
675 {
676 	struct sk_buff *skb;
677 	struct iphdr *iph;
678 	struct igmphdr *ih;
679 	struct rtable *rt;
680 	struct net_device *dev = in_dev->dev;
681 	struct net *net = dev_net(dev);
682 	__be32	group = pmc ? pmc->multiaddr : 0;
683 	struct flowi4 fl4;
684 	__be32	dst;
685 	int hlen, tlen;
686 
687 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
688 		return igmpv3_send_report(in_dev, pmc);
689 
690 	if (ipv4_is_local_multicast(group) && !sysctl_igmp_llm_reports)
691 		return 0;
692 
693 	if (type == IGMP_HOST_LEAVE_MESSAGE)
694 		dst = IGMP_ALL_ROUTER;
695 	else
696 		dst = group;
697 
698 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
699 				   0, 0,
700 				   IPPROTO_IGMP, 0, dev->ifindex);
701 	if (IS_ERR(rt))
702 		return -1;
703 
704 	hlen = LL_RESERVED_SPACE(dev);
705 	tlen = dev->needed_tailroom;
706 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
707 	if (!skb) {
708 		ip_rt_put(rt);
709 		return -1;
710 	}
711 	skb->priority = TC_PRIO_CONTROL;
712 
713 	skb_dst_set(skb, &rt->dst);
714 
715 	skb_reserve(skb, hlen);
716 
717 	skb_reset_network_header(skb);
718 	iph = ip_hdr(skb);
719 	skb_put(skb, sizeof(struct iphdr) + 4);
720 
721 	iph->version  = 4;
722 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
723 	iph->tos      = 0xc0;
724 	iph->frag_off = htons(IP_DF);
725 	iph->ttl      = 1;
726 	iph->daddr    = dst;
727 	iph->saddr    = fl4.saddr;
728 	iph->protocol = IPPROTO_IGMP;
729 	ip_select_ident(net, skb, NULL);
730 	((u8 *)&iph[1])[0] = IPOPT_RA;
731 	((u8 *)&iph[1])[1] = 4;
732 	((u8 *)&iph[1])[2] = 0;
733 	((u8 *)&iph[1])[3] = 0;
734 
735 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
736 	ih->type = type;
737 	ih->code = 0;
738 	ih->csum = 0;
739 	ih->group = group;
740 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
741 
742 	return ip_local_out(skb);
743 }
744 
745 static void igmp_gq_timer_expire(unsigned long data)
746 {
747 	struct in_device *in_dev = (struct in_device *)data;
748 
749 	in_dev->mr_gq_running = 0;
750 	igmpv3_send_report(in_dev, NULL);
751 	in_dev_put(in_dev);
752 }
753 
754 static void igmp_ifc_timer_expire(unsigned long data)
755 {
756 	struct in_device *in_dev = (struct in_device *)data;
757 
758 	igmpv3_send_cr(in_dev);
759 	if (in_dev->mr_ifc_count) {
760 		in_dev->mr_ifc_count--;
761 		igmp_ifc_start_timer(in_dev,
762 				     unsolicited_report_interval(in_dev));
763 	}
764 	in_dev_put(in_dev);
765 }
766 
767 static void igmp_ifc_event(struct in_device *in_dev)
768 {
769 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
770 		return;
771 	in_dev->mr_ifc_count = in_dev->mr_qrv ?: sysctl_igmp_qrv;
772 	igmp_ifc_start_timer(in_dev, 1);
773 }
774 
775 
776 static void igmp_timer_expire(unsigned long data)
777 {
778 	struct ip_mc_list *im = (struct ip_mc_list *)data;
779 	struct in_device *in_dev = im->interface;
780 
781 	spin_lock(&im->lock);
782 	im->tm_running = 0;
783 
784 	if (im->unsolicit_count) {
785 		im->unsolicit_count--;
786 		igmp_start_timer(im, unsolicited_report_interval(in_dev));
787 	}
788 	im->reporter = 1;
789 	spin_unlock(&im->lock);
790 
791 	if (IGMP_V1_SEEN(in_dev))
792 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
793 	else if (IGMP_V2_SEEN(in_dev))
794 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
795 	else
796 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
797 
798 	ip_ma_put(im);
799 }
800 
801 /* mark EXCLUDE-mode sources */
802 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
803 {
804 	struct ip_sf_list *psf;
805 	int i, scount;
806 
807 	scount = 0;
808 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
809 		if (scount == nsrcs)
810 			break;
811 		for (i = 0; i < nsrcs; i++) {
812 			/* skip inactive filters */
813 			if (psf->sf_count[MCAST_INCLUDE] ||
814 			    pmc->sfcount[MCAST_EXCLUDE] !=
815 			    psf->sf_count[MCAST_EXCLUDE])
816 				break;
817 			if (srcs[i] == psf->sf_inaddr) {
818 				scount++;
819 				break;
820 			}
821 		}
822 	}
823 	pmc->gsquery = 0;
824 	if (scount == nsrcs)	/* all sources excluded */
825 		return 0;
826 	return 1;
827 }
828 
829 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
830 {
831 	struct ip_sf_list *psf;
832 	int i, scount;
833 
834 	if (pmc->sfmode == MCAST_EXCLUDE)
835 		return igmp_xmarksources(pmc, nsrcs, srcs);
836 
837 	/* mark INCLUDE-mode sources */
838 	scount = 0;
839 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
840 		if (scount == nsrcs)
841 			break;
842 		for (i = 0; i < nsrcs; i++)
843 			if (srcs[i] == psf->sf_inaddr) {
844 				psf->sf_gsresp = 1;
845 				scount++;
846 				break;
847 			}
848 	}
849 	if (!scount) {
850 		pmc->gsquery = 0;
851 		return 0;
852 	}
853 	pmc->gsquery = 1;
854 	return 1;
855 }
856 
857 /* return true if packet was dropped */
858 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
859 {
860 	struct ip_mc_list *im;
861 
862 	/* Timers are only set for non-local groups */
863 
864 	if (group == IGMP_ALL_HOSTS)
865 		return false;
866 	if (ipv4_is_local_multicast(group) && !sysctl_igmp_llm_reports)
867 		return false;
868 
869 	rcu_read_lock();
870 	for_each_pmc_rcu(in_dev, im) {
871 		if (im->multiaddr == group) {
872 			igmp_stop_timer(im);
873 			break;
874 		}
875 	}
876 	rcu_read_unlock();
877 	return false;
878 }
879 
880 /* return true if packet was dropped */
881 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
882 	int len)
883 {
884 	struct igmphdr 		*ih = igmp_hdr(skb);
885 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
886 	struct ip_mc_list	*im;
887 	__be32			group = ih->group;
888 	int			max_delay;
889 	int			mark = 0;
890 
891 
892 	if (len == 8) {
893 		if (ih->code == 0) {
894 			/* Alas, old v1 router presents here. */
895 
896 			max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
897 			in_dev->mr_v1_seen = jiffies +
898 				IGMP_V1_ROUTER_PRESENT_TIMEOUT;
899 			group = 0;
900 		} else {
901 			/* v2 router present */
902 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
903 			in_dev->mr_v2_seen = jiffies +
904 				IGMP_V2_ROUTER_PRESENT_TIMEOUT;
905 		}
906 		/* cancel the interface change timer */
907 		in_dev->mr_ifc_count = 0;
908 		if (del_timer(&in_dev->mr_ifc_timer))
909 			__in_dev_put(in_dev);
910 		/* clear deleted report items */
911 		igmpv3_clear_delrec(in_dev);
912 	} else if (len < 12) {
913 		return true;	/* ignore bogus packet; freed by caller */
914 	} else if (IGMP_V1_SEEN(in_dev)) {
915 		/* This is a v3 query with v1 queriers present */
916 		max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
917 		group = 0;
918 	} else if (IGMP_V2_SEEN(in_dev)) {
919 		/* this is a v3 query with v2 queriers present;
920 		 * Interpretation of the max_delay code is problematic here.
921 		 * A real v2 host would use ih_code directly, while v3 has a
922 		 * different encoding. We use the v3 encoding as more likely
923 		 * to be intended in a v3 query.
924 		 */
925 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
926 		if (!max_delay)
927 			max_delay = 1;	/* can't mod w/ 0 */
928 	} else { /* v3 */
929 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
930 			return true;
931 
932 		ih3 = igmpv3_query_hdr(skb);
933 		if (ih3->nsrcs) {
934 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
935 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
936 				return true;
937 			ih3 = igmpv3_query_hdr(skb);
938 		}
939 
940 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
941 		if (!max_delay)
942 			max_delay = 1;	/* can't mod w/ 0 */
943 		in_dev->mr_maxdelay = max_delay;
944 		if (ih3->qrv)
945 			in_dev->mr_qrv = ih3->qrv;
946 		if (!group) { /* general query */
947 			if (ih3->nsrcs)
948 				return true;	/* no sources allowed */
949 			igmp_gq_start_timer(in_dev);
950 			return false;
951 		}
952 		/* mark sources to include, if group & source-specific */
953 		mark = ih3->nsrcs != 0;
954 	}
955 
956 	/*
957 	 * - Start the timers in all of our membership records
958 	 *   that the query applies to for the interface on
959 	 *   which the query arrived excl. those that belong
960 	 *   to a "local" group (224.0.0.X)
961 	 * - For timers already running check if they need to
962 	 *   be reset.
963 	 * - Use the igmp->igmp_code field as the maximum
964 	 *   delay possible
965 	 */
966 	rcu_read_lock();
967 	for_each_pmc_rcu(in_dev, im) {
968 		int changed;
969 
970 		if (group && group != im->multiaddr)
971 			continue;
972 		if (im->multiaddr == IGMP_ALL_HOSTS)
973 			continue;
974 		if (ipv4_is_local_multicast(im->multiaddr) &&
975 		    !sysctl_igmp_llm_reports)
976 			continue;
977 		spin_lock_bh(&im->lock);
978 		if (im->tm_running)
979 			im->gsquery = im->gsquery && mark;
980 		else
981 			im->gsquery = mark;
982 		changed = !im->gsquery ||
983 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
984 		spin_unlock_bh(&im->lock);
985 		if (changed)
986 			igmp_mod_timer(im, max_delay);
987 	}
988 	rcu_read_unlock();
989 	return false;
990 }
991 
992 /* called in rcu_read_lock() section */
993 int igmp_rcv(struct sk_buff *skb)
994 {
995 	/* This basically follows the spec line by line -- see RFC1112 */
996 	struct igmphdr *ih;
997 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
998 	int len = skb->len;
999 	bool dropped = true;
1000 
1001 	if (!in_dev)
1002 		goto drop;
1003 
1004 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1005 		goto drop;
1006 
1007 	if (skb_checksum_simple_validate(skb))
1008 		goto drop;
1009 
1010 	ih = igmp_hdr(skb);
1011 	switch (ih->type) {
1012 	case IGMP_HOST_MEMBERSHIP_QUERY:
1013 		dropped = igmp_heard_query(in_dev, skb, len);
1014 		break;
1015 	case IGMP_HOST_MEMBERSHIP_REPORT:
1016 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1017 		/* Is it our report looped back? */
1018 		if (rt_is_output_route(skb_rtable(skb)))
1019 			break;
1020 		/* don't rely on MC router hearing unicast reports */
1021 		if (skb->pkt_type == PACKET_MULTICAST ||
1022 		    skb->pkt_type == PACKET_BROADCAST)
1023 			dropped = igmp_heard_report(in_dev, ih->group);
1024 		break;
1025 	case IGMP_PIM:
1026 #ifdef CONFIG_IP_PIMSM_V1
1027 		return pim_rcv_v1(skb);
1028 #endif
1029 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1030 	case IGMP_DVMRP:
1031 	case IGMP_TRACE:
1032 	case IGMP_HOST_LEAVE_MESSAGE:
1033 	case IGMP_MTRACE:
1034 	case IGMP_MTRACE_RESP:
1035 		break;
1036 	default:
1037 		break;
1038 	}
1039 
1040 drop:
1041 	if (dropped)
1042 		kfree_skb(skb);
1043 	else
1044 		consume_skb(skb);
1045 	return 0;
1046 }
1047 
1048 #endif
1049 
1050 
1051 /*
1052  *	Add a filter to a device
1053  */
1054 
1055 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1056 {
1057 	char buf[MAX_ADDR_LEN];
1058 	struct net_device *dev = in_dev->dev;
1059 
1060 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1061 	   We will get multicast token leakage, when IFF_MULTICAST
1062 	   is changed. This check should be done in ndo_set_rx_mode
1063 	   routine. Something sort of:
1064 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1065 	   --ANK
1066 	   */
1067 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1068 		dev_mc_add(dev, buf);
1069 }
1070 
1071 /*
1072  *	Remove a filter from a device
1073  */
1074 
1075 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1076 {
1077 	char buf[MAX_ADDR_LEN];
1078 	struct net_device *dev = in_dev->dev;
1079 
1080 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1081 		dev_mc_del(dev, buf);
1082 }
1083 
1084 #ifdef CONFIG_IP_MULTICAST
1085 /*
1086  * deleted ip_mc_list manipulation
1087  */
1088 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1089 {
1090 	struct ip_mc_list *pmc;
1091 
1092 	/* this is an "ip_mc_list" for convenience; only the fields below
1093 	 * are actually used. In particular, the refcnt and users are not
1094 	 * used for management of the delete list. Using the same structure
1095 	 * for deleted items allows change reports to use common code with
1096 	 * non-deleted or query-response MCA's.
1097 	 */
1098 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1099 	if (!pmc)
1100 		return;
1101 	spin_lock_bh(&im->lock);
1102 	pmc->interface = im->interface;
1103 	in_dev_hold(in_dev);
1104 	pmc->multiaddr = im->multiaddr;
1105 	pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1106 	pmc->sfmode = im->sfmode;
1107 	if (pmc->sfmode == MCAST_INCLUDE) {
1108 		struct ip_sf_list *psf;
1109 
1110 		pmc->tomb = im->tomb;
1111 		pmc->sources = im->sources;
1112 		im->tomb = im->sources = NULL;
1113 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1114 			psf->sf_crcount = pmc->crcount;
1115 	}
1116 	spin_unlock_bh(&im->lock);
1117 
1118 	spin_lock_bh(&in_dev->mc_tomb_lock);
1119 	pmc->next = in_dev->mc_tomb;
1120 	in_dev->mc_tomb = pmc;
1121 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1122 }
1123 
1124 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1125 {
1126 	struct ip_mc_list *pmc, *pmc_prev;
1127 	struct ip_sf_list *psf, *psf_next;
1128 
1129 	spin_lock_bh(&in_dev->mc_tomb_lock);
1130 	pmc_prev = NULL;
1131 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1132 		if (pmc->multiaddr == multiaddr)
1133 			break;
1134 		pmc_prev = pmc;
1135 	}
1136 	if (pmc) {
1137 		if (pmc_prev)
1138 			pmc_prev->next = pmc->next;
1139 		else
1140 			in_dev->mc_tomb = pmc->next;
1141 	}
1142 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1143 	if (pmc) {
1144 		for (psf = pmc->tomb; psf; psf = psf_next) {
1145 			psf_next = psf->sf_next;
1146 			kfree(psf);
1147 		}
1148 		in_dev_put(pmc->interface);
1149 		kfree(pmc);
1150 	}
1151 }
1152 
1153 static void igmpv3_clear_delrec(struct in_device *in_dev)
1154 {
1155 	struct ip_mc_list *pmc, *nextpmc;
1156 
1157 	spin_lock_bh(&in_dev->mc_tomb_lock);
1158 	pmc = in_dev->mc_tomb;
1159 	in_dev->mc_tomb = NULL;
1160 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1161 
1162 	for (; pmc; pmc = nextpmc) {
1163 		nextpmc = pmc->next;
1164 		ip_mc_clear_src(pmc);
1165 		in_dev_put(pmc->interface);
1166 		kfree(pmc);
1167 	}
1168 	/* clear dead sources, too */
1169 	rcu_read_lock();
1170 	for_each_pmc_rcu(in_dev, pmc) {
1171 		struct ip_sf_list *psf, *psf_next;
1172 
1173 		spin_lock_bh(&pmc->lock);
1174 		psf = pmc->tomb;
1175 		pmc->tomb = NULL;
1176 		spin_unlock_bh(&pmc->lock);
1177 		for (; psf; psf = psf_next) {
1178 			psf_next = psf->sf_next;
1179 			kfree(psf);
1180 		}
1181 	}
1182 	rcu_read_unlock();
1183 }
1184 #endif
1185 
1186 static void igmp_group_dropped(struct ip_mc_list *im)
1187 {
1188 	struct in_device *in_dev = im->interface;
1189 #ifdef CONFIG_IP_MULTICAST
1190 	int reporter;
1191 #endif
1192 
1193 	if (im->loaded) {
1194 		im->loaded = 0;
1195 		ip_mc_filter_del(in_dev, im->multiaddr);
1196 	}
1197 
1198 #ifdef CONFIG_IP_MULTICAST
1199 	if (im->multiaddr == IGMP_ALL_HOSTS)
1200 		return;
1201 	if (ipv4_is_local_multicast(im->multiaddr) && !sysctl_igmp_llm_reports)
1202 		return;
1203 
1204 	reporter = im->reporter;
1205 	igmp_stop_timer(im);
1206 
1207 	if (!in_dev->dead) {
1208 		if (IGMP_V1_SEEN(in_dev))
1209 			return;
1210 		if (IGMP_V2_SEEN(in_dev)) {
1211 			if (reporter)
1212 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1213 			return;
1214 		}
1215 		/* IGMPv3 */
1216 		igmpv3_add_delrec(in_dev, im);
1217 
1218 		igmp_ifc_event(in_dev);
1219 	}
1220 #endif
1221 }
1222 
1223 static void igmp_group_added(struct ip_mc_list *im)
1224 {
1225 	struct in_device *in_dev = im->interface;
1226 
1227 	if (im->loaded == 0) {
1228 		im->loaded = 1;
1229 		ip_mc_filter_add(in_dev, im->multiaddr);
1230 	}
1231 
1232 #ifdef CONFIG_IP_MULTICAST
1233 	if (im->multiaddr == IGMP_ALL_HOSTS)
1234 		return;
1235 	if (ipv4_is_local_multicast(im->multiaddr) && !sysctl_igmp_llm_reports)
1236 		return;
1237 
1238 	if (in_dev->dead)
1239 		return;
1240 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1241 		spin_lock_bh(&im->lock);
1242 		igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1243 		spin_unlock_bh(&im->lock);
1244 		return;
1245 	}
1246 	/* else, v3 */
1247 
1248 	im->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1249 	igmp_ifc_event(in_dev);
1250 #endif
1251 }
1252 
1253 
1254 /*
1255  *	Multicast list managers
1256  */
1257 
1258 static u32 ip_mc_hash(const struct ip_mc_list *im)
1259 {
1260 	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1261 }
1262 
1263 static void ip_mc_hash_add(struct in_device *in_dev,
1264 			   struct ip_mc_list *im)
1265 {
1266 	struct ip_mc_list __rcu **mc_hash;
1267 	u32 hash;
1268 
1269 	mc_hash = rtnl_dereference(in_dev->mc_hash);
1270 	if (mc_hash) {
1271 		hash = ip_mc_hash(im);
1272 		im->next_hash = mc_hash[hash];
1273 		rcu_assign_pointer(mc_hash[hash], im);
1274 		return;
1275 	}
1276 
1277 	/* do not use a hash table for small number of items */
1278 	if (in_dev->mc_count < 4)
1279 		return;
1280 
1281 	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1282 			  GFP_KERNEL);
1283 	if (!mc_hash)
1284 		return;
1285 
1286 	for_each_pmc_rtnl(in_dev, im) {
1287 		hash = ip_mc_hash(im);
1288 		im->next_hash = mc_hash[hash];
1289 		RCU_INIT_POINTER(mc_hash[hash], im);
1290 	}
1291 
1292 	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1293 }
1294 
1295 static void ip_mc_hash_remove(struct in_device *in_dev,
1296 			      struct ip_mc_list *im)
1297 {
1298 	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1299 	struct ip_mc_list *aux;
1300 
1301 	if (!mc_hash)
1302 		return;
1303 	mc_hash += ip_mc_hash(im);
1304 	while ((aux = rtnl_dereference(*mc_hash)) != im)
1305 		mc_hash = &aux->next_hash;
1306 	*mc_hash = im->next_hash;
1307 }
1308 
1309 
1310 /*
1311  *	A socket has joined a multicast group on device dev.
1312  */
1313 
1314 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1315 {
1316 	struct ip_mc_list *im;
1317 
1318 	ASSERT_RTNL();
1319 
1320 	for_each_pmc_rtnl(in_dev, im) {
1321 		if (im->multiaddr == addr) {
1322 			im->users++;
1323 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1324 			goto out;
1325 		}
1326 	}
1327 
1328 	im = kzalloc(sizeof(*im), GFP_KERNEL);
1329 	if (!im)
1330 		goto out;
1331 
1332 	im->users = 1;
1333 	im->interface = in_dev;
1334 	in_dev_hold(in_dev);
1335 	im->multiaddr = addr;
1336 	/* initial mode is (EX, empty) */
1337 	im->sfmode = MCAST_EXCLUDE;
1338 	im->sfcount[MCAST_EXCLUDE] = 1;
1339 	atomic_set(&im->refcnt, 1);
1340 	spin_lock_init(&im->lock);
1341 #ifdef CONFIG_IP_MULTICAST
1342 	setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1343 	im->unsolicit_count = sysctl_igmp_qrv;
1344 #endif
1345 
1346 	im->next_rcu = in_dev->mc_list;
1347 	in_dev->mc_count++;
1348 	rcu_assign_pointer(in_dev->mc_list, im);
1349 
1350 	ip_mc_hash_add(in_dev, im);
1351 
1352 #ifdef CONFIG_IP_MULTICAST
1353 	igmpv3_del_delrec(in_dev, im->multiaddr);
1354 #endif
1355 	igmp_group_added(im);
1356 	if (!in_dev->dead)
1357 		ip_rt_multicast_event(in_dev);
1358 out:
1359 	return;
1360 }
1361 EXPORT_SYMBOL(ip_mc_inc_group);
1362 
1363 static int ip_mc_check_iphdr(struct sk_buff *skb)
1364 {
1365 	const struct iphdr *iph;
1366 	unsigned int len;
1367 	unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1368 
1369 	if (!pskb_may_pull(skb, offset))
1370 		return -EINVAL;
1371 
1372 	iph = ip_hdr(skb);
1373 
1374 	if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1375 		return -EINVAL;
1376 
1377 	offset += ip_hdrlen(skb) - sizeof(*iph);
1378 
1379 	if (!pskb_may_pull(skb, offset))
1380 		return -EINVAL;
1381 
1382 	iph = ip_hdr(skb);
1383 
1384 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1385 		return -EINVAL;
1386 
1387 	len = skb_network_offset(skb) + ntohs(iph->tot_len);
1388 	if (skb->len < len || len < offset)
1389 		return -EINVAL;
1390 
1391 	skb_set_transport_header(skb, offset);
1392 
1393 	return 0;
1394 }
1395 
1396 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1397 {
1398 	unsigned int len = skb_transport_offset(skb);
1399 
1400 	len += sizeof(struct igmpv3_report);
1401 
1402 	return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1403 }
1404 
1405 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1406 {
1407 	unsigned int len = skb_transport_offset(skb);
1408 
1409 	len += sizeof(struct igmphdr);
1410 	if (skb->len < len)
1411 		return -EINVAL;
1412 
1413 	/* IGMPv{1,2}? */
1414 	if (skb->len != len) {
1415 		/* or IGMPv3? */
1416 		len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1417 		if (skb->len < len || !pskb_may_pull(skb, len))
1418 			return -EINVAL;
1419 	}
1420 
1421 	/* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1422 	 * all-systems destination addresses (224.0.0.1) for general queries
1423 	 */
1424 	if (!igmp_hdr(skb)->group &&
1425 	    ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1426 		return -EINVAL;
1427 
1428 	return 0;
1429 }
1430 
1431 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1432 {
1433 	switch (igmp_hdr(skb)->type) {
1434 	case IGMP_HOST_LEAVE_MESSAGE:
1435 	case IGMP_HOST_MEMBERSHIP_REPORT:
1436 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1437 		/* fall through */
1438 		return 0;
1439 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1440 		return ip_mc_check_igmp_reportv3(skb);
1441 	case IGMP_HOST_MEMBERSHIP_QUERY:
1442 		return ip_mc_check_igmp_query(skb);
1443 	default:
1444 		return -ENOMSG;
1445 	}
1446 }
1447 
1448 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1449 {
1450 	return skb_checksum_simple_validate(skb);
1451 }
1452 
1453 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1454 
1455 {
1456 	struct sk_buff *skb_chk;
1457 	unsigned int transport_len;
1458 	unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1459 	int ret = -EINVAL;
1460 
1461 	transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1462 
1463 	skb_chk = skb_checksum_trimmed(skb, transport_len,
1464 				       ip_mc_validate_checksum);
1465 	if (!skb_chk)
1466 		goto err;
1467 
1468 	if (!pskb_may_pull(skb_chk, len))
1469 		goto err;
1470 
1471 	ret = ip_mc_check_igmp_msg(skb_chk);
1472 	if (ret)
1473 		goto err;
1474 
1475 	if (skb_trimmed)
1476 		*skb_trimmed = skb_chk;
1477 	/* free now unneeded clone */
1478 	else if (skb_chk != skb)
1479 		kfree_skb(skb_chk);
1480 
1481 	ret = 0;
1482 
1483 err:
1484 	if (ret && skb_chk && skb_chk != skb)
1485 		kfree_skb(skb_chk);
1486 
1487 	return ret;
1488 }
1489 
1490 /**
1491  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1492  * @skb: the skb to validate
1493  * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1494  *
1495  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1496  * skb transport header accordingly and returns zero.
1497  *
1498  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1499  *  standard
1500  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1501  * -ENOMEM: A memory allocation failure happened.
1502  *
1503  * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1504  * to NULL): After parsing an IGMP packet successfully it will point to
1505  * an skb which has its tail aligned to the IP packet end. This might
1506  * either be the originally provided skb or a trimmed, cloned version if
1507  * the skb frame had data beyond the IP packet. A cloned skb allows us
1508  * to leave the original skb and its full frame unchanged (which might be
1509  * desirable for layer 2 frame jugglers).
1510  *
1511  * Caller needs to set the skb network header and free any returned skb if it
1512  * differs from the provided skb.
1513  */
1514 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1515 {
1516 	int ret = ip_mc_check_iphdr(skb);
1517 
1518 	if (ret < 0)
1519 		return ret;
1520 
1521 	if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1522 		return -ENOMSG;
1523 
1524 	return __ip_mc_check_igmp(skb, skb_trimmed);
1525 }
1526 EXPORT_SYMBOL(ip_mc_check_igmp);
1527 
1528 /*
1529  *	Resend IGMP JOIN report; used by netdev notifier.
1530  */
1531 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1532 {
1533 #ifdef CONFIG_IP_MULTICAST
1534 	struct ip_mc_list *im;
1535 	int type;
1536 
1537 	ASSERT_RTNL();
1538 
1539 	for_each_pmc_rtnl(in_dev, im) {
1540 		if (im->multiaddr == IGMP_ALL_HOSTS)
1541 			continue;
1542 		if (ipv4_is_local_multicast(im->multiaddr) &&
1543 		    !sysctl_igmp_llm_reports)
1544 			continue;
1545 
1546 		/* a failover is happening and switches
1547 		 * must be notified immediately
1548 		 */
1549 		if (IGMP_V1_SEEN(in_dev))
1550 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1551 		else if (IGMP_V2_SEEN(in_dev))
1552 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1553 		else
1554 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1555 		igmp_send_report(in_dev, im, type);
1556 	}
1557 #endif
1558 }
1559 
1560 /*
1561  *	A socket has left a multicast group on device dev
1562  */
1563 
1564 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1565 {
1566 	struct ip_mc_list *i;
1567 	struct ip_mc_list __rcu **ip;
1568 
1569 	ASSERT_RTNL();
1570 
1571 	for (ip = &in_dev->mc_list;
1572 	     (i = rtnl_dereference(*ip)) != NULL;
1573 	     ip = &i->next_rcu) {
1574 		if (i->multiaddr == addr) {
1575 			if (--i->users == 0) {
1576 				ip_mc_hash_remove(in_dev, i);
1577 				*ip = i->next_rcu;
1578 				in_dev->mc_count--;
1579 				igmp_group_dropped(i);
1580 				ip_mc_clear_src(i);
1581 
1582 				if (!in_dev->dead)
1583 					ip_rt_multicast_event(in_dev);
1584 
1585 				ip_ma_put(i);
1586 				return;
1587 			}
1588 			break;
1589 		}
1590 	}
1591 }
1592 EXPORT_SYMBOL(ip_mc_dec_group);
1593 
1594 /* Device changing type */
1595 
1596 void ip_mc_unmap(struct in_device *in_dev)
1597 {
1598 	struct ip_mc_list *pmc;
1599 
1600 	ASSERT_RTNL();
1601 
1602 	for_each_pmc_rtnl(in_dev, pmc)
1603 		igmp_group_dropped(pmc);
1604 }
1605 
1606 void ip_mc_remap(struct in_device *in_dev)
1607 {
1608 	struct ip_mc_list *pmc;
1609 
1610 	ASSERT_RTNL();
1611 
1612 	for_each_pmc_rtnl(in_dev, pmc)
1613 		igmp_group_added(pmc);
1614 }
1615 
1616 /* Device going down */
1617 
1618 void ip_mc_down(struct in_device *in_dev)
1619 {
1620 	struct ip_mc_list *pmc;
1621 
1622 	ASSERT_RTNL();
1623 
1624 	for_each_pmc_rtnl(in_dev, pmc)
1625 		igmp_group_dropped(pmc);
1626 
1627 #ifdef CONFIG_IP_MULTICAST
1628 	in_dev->mr_ifc_count = 0;
1629 	if (del_timer(&in_dev->mr_ifc_timer))
1630 		__in_dev_put(in_dev);
1631 	in_dev->mr_gq_running = 0;
1632 	if (del_timer(&in_dev->mr_gq_timer))
1633 		__in_dev_put(in_dev);
1634 	igmpv3_clear_delrec(in_dev);
1635 #endif
1636 
1637 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1638 }
1639 
1640 void ip_mc_init_dev(struct in_device *in_dev)
1641 {
1642 	ASSERT_RTNL();
1643 
1644 #ifdef CONFIG_IP_MULTICAST
1645 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1646 			(unsigned long)in_dev);
1647 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1648 			(unsigned long)in_dev);
1649 	in_dev->mr_qrv = sysctl_igmp_qrv;
1650 #endif
1651 
1652 	spin_lock_init(&in_dev->mc_tomb_lock);
1653 }
1654 
1655 /* Device going up */
1656 
1657 void ip_mc_up(struct in_device *in_dev)
1658 {
1659 	struct ip_mc_list *pmc;
1660 
1661 	ASSERT_RTNL();
1662 
1663 #ifdef CONFIG_IP_MULTICAST
1664 	in_dev->mr_qrv = sysctl_igmp_qrv;
1665 #endif
1666 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1667 
1668 	for_each_pmc_rtnl(in_dev, pmc)
1669 		igmp_group_added(pmc);
1670 }
1671 
1672 /*
1673  *	Device is about to be destroyed: clean up.
1674  */
1675 
1676 void ip_mc_destroy_dev(struct in_device *in_dev)
1677 {
1678 	struct ip_mc_list *i;
1679 
1680 	ASSERT_RTNL();
1681 
1682 	/* Deactivate timers */
1683 	ip_mc_down(in_dev);
1684 
1685 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1686 		in_dev->mc_list = i->next_rcu;
1687 		in_dev->mc_count--;
1688 
1689 		/* We've dropped the groups in ip_mc_down already */
1690 		ip_mc_clear_src(i);
1691 		ip_ma_put(i);
1692 	}
1693 }
1694 
1695 /* RTNL is locked */
1696 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1697 {
1698 	struct net_device *dev = NULL;
1699 	struct in_device *idev = NULL;
1700 
1701 	if (imr->imr_ifindex) {
1702 		idev = inetdev_by_index(net, imr->imr_ifindex);
1703 		return idev;
1704 	}
1705 	if (imr->imr_address.s_addr) {
1706 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1707 		if (!dev)
1708 			return NULL;
1709 	}
1710 
1711 	if (!dev) {
1712 		struct rtable *rt = ip_route_output(net,
1713 						    imr->imr_multiaddr.s_addr,
1714 						    0, 0, 0);
1715 		if (!IS_ERR(rt)) {
1716 			dev = rt->dst.dev;
1717 			ip_rt_put(rt);
1718 		}
1719 	}
1720 	if (dev) {
1721 		imr->imr_ifindex = dev->ifindex;
1722 		idev = __in_dev_get_rtnl(dev);
1723 	}
1724 	return idev;
1725 }
1726 
1727 /*
1728  *	Join a socket to a group
1729  */
1730 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1731 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1732 #ifdef CONFIG_IP_MULTICAST
1733 int sysctl_igmp_qrv __read_mostly = IGMP_QUERY_ROBUSTNESS_VARIABLE;
1734 #endif
1735 
1736 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1737 	__be32 *psfsrc)
1738 {
1739 	struct ip_sf_list *psf, *psf_prev;
1740 	int rv = 0;
1741 
1742 	psf_prev = NULL;
1743 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1744 		if (psf->sf_inaddr == *psfsrc)
1745 			break;
1746 		psf_prev = psf;
1747 	}
1748 	if (!psf || psf->sf_count[sfmode] == 0) {
1749 		/* source filter not found, or count wrong =>  bug */
1750 		return -ESRCH;
1751 	}
1752 	psf->sf_count[sfmode]--;
1753 	if (psf->sf_count[sfmode] == 0) {
1754 		ip_rt_multicast_event(pmc->interface);
1755 	}
1756 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1757 #ifdef CONFIG_IP_MULTICAST
1758 		struct in_device *in_dev = pmc->interface;
1759 #endif
1760 
1761 		/* no more filters for this source */
1762 		if (psf_prev)
1763 			psf_prev->sf_next = psf->sf_next;
1764 		else
1765 			pmc->sources = psf->sf_next;
1766 #ifdef CONFIG_IP_MULTICAST
1767 		if (psf->sf_oldin &&
1768 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1769 			psf->sf_crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1770 			psf->sf_next = pmc->tomb;
1771 			pmc->tomb = psf;
1772 			rv = 1;
1773 		} else
1774 #endif
1775 			kfree(psf);
1776 	}
1777 	return rv;
1778 }
1779 
1780 #ifndef CONFIG_IP_MULTICAST
1781 #define igmp_ifc_event(x)	do { } while (0)
1782 #endif
1783 
1784 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1785 			 int sfcount, __be32 *psfsrc, int delta)
1786 {
1787 	struct ip_mc_list *pmc;
1788 	int	changerec = 0;
1789 	int	i, err;
1790 
1791 	if (!in_dev)
1792 		return -ENODEV;
1793 	rcu_read_lock();
1794 	for_each_pmc_rcu(in_dev, pmc) {
1795 		if (*pmca == pmc->multiaddr)
1796 			break;
1797 	}
1798 	if (!pmc) {
1799 		/* MCA not found?? bug */
1800 		rcu_read_unlock();
1801 		return -ESRCH;
1802 	}
1803 	spin_lock_bh(&pmc->lock);
1804 	rcu_read_unlock();
1805 #ifdef CONFIG_IP_MULTICAST
1806 	sf_markstate(pmc);
1807 #endif
1808 	if (!delta) {
1809 		err = -EINVAL;
1810 		if (!pmc->sfcount[sfmode])
1811 			goto out_unlock;
1812 		pmc->sfcount[sfmode]--;
1813 	}
1814 	err = 0;
1815 	for (i = 0; i < sfcount; i++) {
1816 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1817 
1818 		changerec |= rv > 0;
1819 		if (!err && rv < 0)
1820 			err = rv;
1821 	}
1822 	if (pmc->sfmode == MCAST_EXCLUDE &&
1823 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1824 	    pmc->sfcount[MCAST_INCLUDE]) {
1825 #ifdef CONFIG_IP_MULTICAST
1826 		struct ip_sf_list *psf;
1827 #endif
1828 
1829 		/* filter mode change */
1830 		pmc->sfmode = MCAST_INCLUDE;
1831 #ifdef CONFIG_IP_MULTICAST
1832 		pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1833 		in_dev->mr_ifc_count = pmc->crcount;
1834 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1835 			psf->sf_crcount = 0;
1836 		igmp_ifc_event(pmc->interface);
1837 	} else if (sf_setstate(pmc) || changerec) {
1838 		igmp_ifc_event(pmc->interface);
1839 #endif
1840 	}
1841 out_unlock:
1842 	spin_unlock_bh(&pmc->lock);
1843 	return err;
1844 }
1845 
1846 /*
1847  * Add multicast single-source filter to the interface list
1848  */
1849 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1850 	__be32 *psfsrc)
1851 {
1852 	struct ip_sf_list *psf, *psf_prev;
1853 
1854 	psf_prev = NULL;
1855 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1856 		if (psf->sf_inaddr == *psfsrc)
1857 			break;
1858 		psf_prev = psf;
1859 	}
1860 	if (!psf) {
1861 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1862 		if (!psf)
1863 			return -ENOBUFS;
1864 		psf->sf_inaddr = *psfsrc;
1865 		if (psf_prev) {
1866 			psf_prev->sf_next = psf;
1867 		} else
1868 			pmc->sources = psf;
1869 	}
1870 	psf->sf_count[sfmode]++;
1871 	if (psf->sf_count[sfmode] == 1) {
1872 		ip_rt_multicast_event(pmc->interface);
1873 	}
1874 	return 0;
1875 }
1876 
1877 #ifdef CONFIG_IP_MULTICAST
1878 static void sf_markstate(struct ip_mc_list *pmc)
1879 {
1880 	struct ip_sf_list *psf;
1881 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1882 
1883 	for (psf = pmc->sources; psf; psf = psf->sf_next)
1884 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1885 			psf->sf_oldin = mca_xcount ==
1886 				psf->sf_count[MCAST_EXCLUDE] &&
1887 				!psf->sf_count[MCAST_INCLUDE];
1888 		} else
1889 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1890 }
1891 
1892 static int sf_setstate(struct ip_mc_list *pmc)
1893 {
1894 	struct ip_sf_list *psf, *dpsf;
1895 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1896 	int qrv = pmc->interface->mr_qrv;
1897 	int new_in, rv;
1898 
1899 	rv = 0;
1900 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1901 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1902 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1903 				!psf->sf_count[MCAST_INCLUDE];
1904 		} else
1905 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1906 		if (new_in) {
1907 			if (!psf->sf_oldin) {
1908 				struct ip_sf_list *prev = NULL;
1909 
1910 				for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1911 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1912 						break;
1913 					prev = dpsf;
1914 				}
1915 				if (dpsf) {
1916 					if (prev)
1917 						prev->sf_next = dpsf->sf_next;
1918 					else
1919 						pmc->tomb = dpsf->sf_next;
1920 					kfree(dpsf);
1921 				}
1922 				psf->sf_crcount = qrv;
1923 				rv++;
1924 			}
1925 		} else if (psf->sf_oldin) {
1926 
1927 			psf->sf_crcount = 0;
1928 			/*
1929 			 * add or update "delete" records if an active filter
1930 			 * is now inactive
1931 			 */
1932 			for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1933 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1934 					break;
1935 			if (!dpsf) {
1936 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1937 				if (!dpsf)
1938 					continue;
1939 				*dpsf = *psf;
1940 				/* pmc->lock held by callers */
1941 				dpsf->sf_next = pmc->tomb;
1942 				pmc->tomb = dpsf;
1943 			}
1944 			dpsf->sf_crcount = qrv;
1945 			rv++;
1946 		}
1947 	}
1948 	return rv;
1949 }
1950 #endif
1951 
1952 /*
1953  * Add multicast source filter list to the interface list
1954  */
1955 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1956 			 int sfcount, __be32 *psfsrc, int delta)
1957 {
1958 	struct ip_mc_list *pmc;
1959 	int	isexclude;
1960 	int	i, err;
1961 
1962 	if (!in_dev)
1963 		return -ENODEV;
1964 	rcu_read_lock();
1965 	for_each_pmc_rcu(in_dev, pmc) {
1966 		if (*pmca == pmc->multiaddr)
1967 			break;
1968 	}
1969 	if (!pmc) {
1970 		/* MCA not found?? bug */
1971 		rcu_read_unlock();
1972 		return -ESRCH;
1973 	}
1974 	spin_lock_bh(&pmc->lock);
1975 	rcu_read_unlock();
1976 
1977 #ifdef CONFIG_IP_MULTICAST
1978 	sf_markstate(pmc);
1979 #endif
1980 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
1981 	if (!delta)
1982 		pmc->sfcount[sfmode]++;
1983 	err = 0;
1984 	for (i = 0; i < sfcount; i++) {
1985 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1986 		if (err)
1987 			break;
1988 	}
1989 	if (err) {
1990 		int j;
1991 
1992 		if (!delta)
1993 			pmc->sfcount[sfmode]--;
1994 		for (j = 0; j < i; j++)
1995 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1996 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1997 #ifdef CONFIG_IP_MULTICAST
1998 		struct ip_sf_list *psf;
1999 		in_dev = pmc->interface;
2000 #endif
2001 
2002 		/* filter mode change */
2003 		if (pmc->sfcount[MCAST_EXCLUDE])
2004 			pmc->sfmode = MCAST_EXCLUDE;
2005 		else if (pmc->sfcount[MCAST_INCLUDE])
2006 			pmc->sfmode = MCAST_INCLUDE;
2007 #ifdef CONFIG_IP_MULTICAST
2008 		/* else no filters; keep old mode for reports */
2009 
2010 		pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
2011 		in_dev->mr_ifc_count = pmc->crcount;
2012 		for (psf = pmc->sources; psf; psf = psf->sf_next)
2013 			psf->sf_crcount = 0;
2014 		igmp_ifc_event(in_dev);
2015 	} else if (sf_setstate(pmc)) {
2016 		igmp_ifc_event(in_dev);
2017 #endif
2018 	}
2019 	spin_unlock_bh(&pmc->lock);
2020 	return err;
2021 }
2022 
2023 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2024 {
2025 	struct ip_sf_list *psf, *nextpsf;
2026 
2027 	for (psf = pmc->tomb; psf; psf = nextpsf) {
2028 		nextpsf = psf->sf_next;
2029 		kfree(psf);
2030 	}
2031 	pmc->tomb = NULL;
2032 	for (psf = pmc->sources; psf; psf = nextpsf) {
2033 		nextpsf = psf->sf_next;
2034 		kfree(psf);
2035 	}
2036 	pmc->sources = NULL;
2037 	pmc->sfmode = MCAST_EXCLUDE;
2038 	pmc->sfcount[MCAST_INCLUDE] = 0;
2039 	pmc->sfcount[MCAST_EXCLUDE] = 1;
2040 }
2041 
2042 /* Join a multicast group
2043  */
2044 
2045 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2046 {
2047 	__be32 addr = imr->imr_multiaddr.s_addr;
2048 	struct ip_mc_socklist *iml, *i;
2049 	struct in_device *in_dev;
2050 	struct inet_sock *inet = inet_sk(sk);
2051 	struct net *net = sock_net(sk);
2052 	int ifindex;
2053 	int count = 0;
2054 	int err;
2055 
2056 	ASSERT_RTNL();
2057 
2058 	if (!ipv4_is_multicast(addr))
2059 		return -EINVAL;
2060 
2061 	in_dev = ip_mc_find_dev(net, imr);
2062 
2063 	if (!in_dev) {
2064 		err = -ENODEV;
2065 		goto done;
2066 	}
2067 
2068 	err = -EADDRINUSE;
2069 	ifindex = imr->imr_ifindex;
2070 	for_each_pmc_rtnl(inet, i) {
2071 		if (i->multi.imr_multiaddr.s_addr == addr &&
2072 		    i->multi.imr_ifindex == ifindex)
2073 			goto done;
2074 		count++;
2075 	}
2076 	err = -ENOBUFS;
2077 	if (count >= sysctl_igmp_max_memberships)
2078 		goto done;
2079 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2080 	if (!iml)
2081 		goto done;
2082 
2083 	memcpy(&iml->multi, imr, sizeof(*imr));
2084 	iml->next_rcu = inet->mc_list;
2085 	iml->sflist = NULL;
2086 	iml->sfmode = MCAST_EXCLUDE;
2087 	rcu_assign_pointer(inet->mc_list, iml);
2088 	ip_mc_inc_group(in_dev, addr);
2089 	err = 0;
2090 done:
2091 	return err;
2092 }
2093 EXPORT_SYMBOL(ip_mc_join_group);
2094 
2095 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2096 			   struct in_device *in_dev)
2097 {
2098 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2099 	int err;
2100 
2101 	if (!psf) {
2102 		/* any-source empty exclude case */
2103 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2104 			iml->sfmode, 0, NULL, 0);
2105 	}
2106 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2107 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2108 	RCU_INIT_POINTER(iml->sflist, NULL);
2109 	/* decrease mem now to avoid the memleak warning */
2110 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2111 	kfree_rcu(psf, rcu);
2112 	return err;
2113 }
2114 
2115 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2116 {
2117 	struct inet_sock *inet = inet_sk(sk);
2118 	struct ip_mc_socklist *iml;
2119 	struct ip_mc_socklist __rcu **imlp;
2120 	struct in_device *in_dev;
2121 	struct net *net = sock_net(sk);
2122 	__be32 group = imr->imr_multiaddr.s_addr;
2123 	u32 ifindex;
2124 	int ret = -EADDRNOTAVAIL;
2125 
2126 	ASSERT_RTNL();
2127 
2128 	in_dev = ip_mc_find_dev(net, imr);
2129 	if (!in_dev) {
2130 		ret = -ENODEV;
2131 		goto out;
2132 	}
2133 	ifindex = imr->imr_ifindex;
2134 	for (imlp = &inet->mc_list;
2135 	     (iml = rtnl_dereference(*imlp)) != NULL;
2136 	     imlp = &iml->next_rcu) {
2137 		if (iml->multi.imr_multiaddr.s_addr != group)
2138 			continue;
2139 		if (ifindex) {
2140 			if (iml->multi.imr_ifindex != ifindex)
2141 				continue;
2142 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2143 				iml->multi.imr_address.s_addr)
2144 			continue;
2145 
2146 		(void) ip_mc_leave_src(sk, iml, in_dev);
2147 
2148 		*imlp = iml->next_rcu;
2149 
2150 		ip_mc_dec_group(in_dev, group);
2151 
2152 		/* decrease mem now to avoid the memleak warning */
2153 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2154 		kfree_rcu(iml, rcu);
2155 		return 0;
2156 	}
2157 out:
2158 	return ret;
2159 }
2160 EXPORT_SYMBOL(ip_mc_leave_group);
2161 
2162 int ip_mc_source(int add, int omode, struct sock *sk, struct
2163 	ip_mreq_source *mreqs, int ifindex)
2164 {
2165 	int err;
2166 	struct ip_mreqn imr;
2167 	__be32 addr = mreqs->imr_multiaddr;
2168 	struct ip_mc_socklist *pmc;
2169 	struct in_device *in_dev = NULL;
2170 	struct inet_sock *inet = inet_sk(sk);
2171 	struct ip_sf_socklist *psl;
2172 	struct net *net = sock_net(sk);
2173 	int leavegroup = 0;
2174 	int i, j, rv;
2175 
2176 	if (!ipv4_is_multicast(addr))
2177 		return -EINVAL;
2178 
2179 	ASSERT_RTNL();
2180 
2181 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2182 	imr.imr_address.s_addr = mreqs->imr_interface;
2183 	imr.imr_ifindex = ifindex;
2184 	in_dev = ip_mc_find_dev(net, &imr);
2185 
2186 	if (!in_dev) {
2187 		err = -ENODEV;
2188 		goto done;
2189 	}
2190 	err = -EADDRNOTAVAIL;
2191 
2192 	for_each_pmc_rtnl(inet, pmc) {
2193 		if ((pmc->multi.imr_multiaddr.s_addr ==
2194 		     imr.imr_multiaddr.s_addr) &&
2195 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
2196 			break;
2197 	}
2198 	if (!pmc) {		/* must have a prior join */
2199 		err = -EINVAL;
2200 		goto done;
2201 	}
2202 	/* if a source filter was set, must be the same mode as before */
2203 	if (pmc->sflist) {
2204 		if (pmc->sfmode != omode) {
2205 			err = -EINVAL;
2206 			goto done;
2207 		}
2208 	} else if (pmc->sfmode != omode) {
2209 		/* allow mode switches for empty-set filters */
2210 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2211 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2212 			NULL, 0);
2213 		pmc->sfmode = omode;
2214 	}
2215 
2216 	psl = rtnl_dereference(pmc->sflist);
2217 	if (!add) {
2218 		if (!psl)
2219 			goto done;	/* err = -EADDRNOTAVAIL */
2220 		rv = !0;
2221 		for (i = 0; i < psl->sl_count; i++) {
2222 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2223 				sizeof(__be32));
2224 			if (rv == 0)
2225 				break;
2226 		}
2227 		if (rv)		/* source not found */
2228 			goto done;	/* err = -EADDRNOTAVAIL */
2229 
2230 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2231 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2232 			leavegroup = 1;
2233 			goto done;
2234 		}
2235 
2236 		/* update the interface filter */
2237 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2238 			&mreqs->imr_sourceaddr, 1);
2239 
2240 		for (j = i+1; j < psl->sl_count; j++)
2241 			psl->sl_addr[j-1] = psl->sl_addr[j];
2242 		psl->sl_count--;
2243 		err = 0;
2244 		goto done;
2245 	}
2246 	/* else, add a new source to the filter */
2247 
2248 	if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2249 		err = -ENOBUFS;
2250 		goto done;
2251 	}
2252 	if (!psl || psl->sl_count == psl->sl_max) {
2253 		struct ip_sf_socklist *newpsl;
2254 		int count = IP_SFBLOCK;
2255 
2256 		if (psl)
2257 			count += psl->sl_max;
2258 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2259 		if (!newpsl) {
2260 			err = -ENOBUFS;
2261 			goto done;
2262 		}
2263 		newpsl->sl_max = count;
2264 		newpsl->sl_count = count - IP_SFBLOCK;
2265 		if (psl) {
2266 			for (i = 0; i < psl->sl_count; i++)
2267 				newpsl->sl_addr[i] = psl->sl_addr[i];
2268 			/* decrease mem now to avoid the memleak warning */
2269 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2270 			kfree_rcu(psl, rcu);
2271 		}
2272 		rcu_assign_pointer(pmc->sflist, newpsl);
2273 		psl = newpsl;
2274 	}
2275 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2276 	for (i = 0; i < psl->sl_count; i++) {
2277 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2278 			sizeof(__be32));
2279 		if (rv == 0)
2280 			break;
2281 	}
2282 	if (rv == 0)		/* address already there is an error */
2283 		goto done;
2284 	for (j = psl->sl_count-1; j >= i; j--)
2285 		psl->sl_addr[j+1] = psl->sl_addr[j];
2286 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2287 	psl->sl_count++;
2288 	err = 0;
2289 	/* update the interface list */
2290 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2291 		&mreqs->imr_sourceaddr, 1);
2292 done:
2293 	if (leavegroup)
2294 		err = ip_mc_leave_group(sk, &imr);
2295 	return err;
2296 }
2297 
2298 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2299 {
2300 	int err = 0;
2301 	struct ip_mreqn	imr;
2302 	__be32 addr = msf->imsf_multiaddr;
2303 	struct ip_mc_socklist *pmc;
2304 	struct in_device *in_dev;
2305 	struct inet_sock *inet = inet_sk(sk);
2306 	struct ip_sf_socklist *newpsl, *psl;
2307 	struct net *net = sock_net(sk);
2308 	int leavegroup = 0;
2309 
2310 	if (!ipv4_is_multicast(addr))
2311 		return -EINVAL;
2312 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2313 	    msf->imsf_fmode != MCAST_EXCLUDE)
2314 		return -EINVAL;
2315 
2316 	ASSERT_RTNL();
2317 
2318 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2319 	imr.imr_address.s_addr = msf->imsf_interface;
2320 	imr.imr_ifindex = ifindex;
2321 	in_dev = ip_mc_find_dev(net, &imr);
2322 
2323 	if (!in_dev) {
2324 		err = -ENODEV;
2325 		goto done;
2326 	}
2327 
2328 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2329 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2330 		leavegroup = 1;
2331 		goto done;
2332 	}
2333 
2334 	for_each_pmc_rtnl(inet, pmc) {
2335 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2336 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2337 			break;
2338 	}
2339 	if (!pmc) {		/* must have a prior join */
2340 		err = -EINVAL;
2341 		goto done;
2342 	}
2343 	if (msf->imsf_numsrc) {
2344 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2345 							   GFP_KERNEL);
2346 		if (!newpsl) {
2347 			err = -ENOBUFS;
2348 			goto done;
2349 		}
2350 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2351 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2352 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2353 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2354 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2355 		if (err) {
2356 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2357 			goto done;
2358 		}
2359 	} else {
2360 		newpsl = NULL;
2361 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2362 				     msf->imsf_fmode, 0, NULL, 0);
2363 	}
2364 	psl = rtnl_dereference(pmc->sflist);
2365 	if (psl) {
2366 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2367 			psl->sl_count, psl->sl_addr, 0);
2368 		/* decrease mem now to avoid the memleak warning */
2369 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2370 		kfree_rcu(psl, rcu);
2371 	} else
2372 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2373 			0, NULL, 0);
2374 	rcu_assign_pointer(pmc->sflist, newpsl);
2375 	pmc->sfmode = msf->imsf_fmode;
2376 	err = 0;
2377 done:
2378 	if (leavegroup)
2379 		err = ip_mc_leave_group(sk, &imr);
2380 	return err;
2381 }
2382 
2383 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2384 	struct ip_msfilter __user *optval, int __user *optlen)
2385 {
2386 	int err, len, count, copycount;
2387 	struct ip_mreqn	imr;
2388 	__be32 addr = msf->imsf_multiaddr;
2389 	struct ip_mc_socklist *pmc;
2390 	struct in_device *in_dev;
2391 	struct inet_sock *inet = inet_sk(sk);
2392 	struct ip_sf_socklist *psl;
2393 	struct net *net = sock_net(sk);
2394 
2395 	if (!ipv4_is_multicast(addr))
2396 		return -EINVAL;
2397 
2398 	rtnl_lock();
2399 
2400 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2401 	imr.imr_address.s_addr = msf->imsf_interface;
2402 	imr.imr_ifindex = 0;
2403 	in_dev = ip_mc_find_dev(net, &imr);
2404 
2405 	if (!in_dev) {
2406 		err = -ENODEV;
2407 		goto done;
2408 	}
2409 	err = -EADDRNOTAVAIL;
2410 
2411 	for_each_pmc_rtnl(inet, pmc) {
2412 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2413 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2414 			break;
2415 	}
2416 	if (!pmc)		/* must have a prior join */
2417 		goto done;
2418 	msf->imsf_fmode = pmc->sfmode;
2419 	psl = rtnl_dereference(pmc->sflist);
2420 	rtnl_unlock();
2421 	if (!psl) {
2422 		len = 0;
2423 		count = 0;
2424 	} else {
2425 		count = psl->sl_count;
2426 	}
2427 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2428 	len = copycount * sizeof(psl->sl_addr[0]);
2429 	msf->imsf_numsrc = count;
2430 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2431 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2432 		return -EFAULT;
2433 	}
2434 	if (len &&
2435 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2436 		return -EFAULT;
2437 	return 0;
2438 done:
2439 	rtnl_unlock();
2440 	return err;
2441 }
2442 
2443 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2444 	struct group_filter __user *optval, int __user *optlen)
2445 {
2446 	int err, i, count, copycount;
2447 	struct sockaddr_in *psin;
2448 	__be32 addr;
2449 	struct ip_mc_socklist *pmc;
2450 	struct inet_sock *inet = inet_sk(sk);
2451 	struct ip_sf_socklist *psl;
2452 
2453 	psin = (struct sockaddr_in *)&gsf->gf_group;
2454 	if (psin->sin_family != AF_INET)
2455 		return -EINVAL;
2456 	addr = psin->sin_addr.s_addr;
2457 	if (!ipv4_is_multicast(addr))
2458 		return -EINVAL;
2459 
2460 	rtnl_lock();
2461 
2462 	err = -EADDRNOTAVAIL;
2463 
2464 	for_each_pmc_rtnl(inet, pmc) {
2465 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2466 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2467 			break;
2468 	}
2469 	if (!pmc)		/* must have a prior join */
2470 		goto done;
2471 	gsf->gf_fmode = pmc->sfmode;
2472 	psl = rtnl_dereference(pmc->sflist);
2473 	rtnl_unlock();
2474 	count = psl ? psl->sl_count : 0;
2475 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2476 	gsf->gf_numsrc = count;
2477 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2478 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2479 		return -EFAULT;
2480 	}
2481 	for (i = 0; i < copycount; i++) {
2482 		struct sockaddr_storage ss;
2483 
2484 		psin = (struct sockaddr_in *)&ss;
2485 		memset(&ss, 0, sizeof(ss));
2486 		psin->sin_family = AF_INET;
2487 		psin->sin_addr.s_addr = psl->sl_addr[i];
2488 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2489 			return -EFAULT;
2490 	}
2491 	return 0;
2492 done:
2493 	rtnl_unlock();
2494 	return err;
2495 }
2496 
2497 /*
2498  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2499  */
2500 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2501 {
2502 	struct inet_sock *inet = inet_sk(sk);
2503 	struct ip_mc_socklist *pmc;
2504 	struct ip_sf_socklist *psl;
2505 	int i;
2506 	int ret;
2507 
2508 	ret = 1;
2509 	if (!ipv4_is_multicast(loc_addr))
2510 		goto out;
2511 
2512 	rcu_read_lock();
2513 	for_each_pmc_rcu(inet, pmc) {
2514 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2515 		    pmc->multi.imr_ifindex == dif)
2516 			break;
2517 	}
2518 	ret = inet->mc_all;
2519 	if (!pmc)
2520 		goto unlock;
2521 	psl = rcu_dereference(pmc->sflist);
2522 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2523 	if (!psl)
2524 		goto unlock;
2525 
2526 	for (i = 0; i < psl->sl_count; i++) {
2527 		if (psl->sl_addr[i] == rmt_addr)
2528 			break;
2529 	}
2530 	ret = 0;
2531 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2532 		goto unlock;
2533 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2534 		goto unlock;
2535 	ret = 1;
2536 unlock:
2537 	rcu_read_unlock();
2538 out:
2539 	return ret;
2540 }
2541 
2542 /*
2543  *	A socket is closing.
2544  */
2545 
2546 void ip_mc_drop_socket(struct sock *sk)
2547 {
2548 	struct inet_sock *inet = inet_sk(sk);
2549 	struct ip_mc_socklist *iml;
2550 	struct net *net = sock_net(sk);
2551 
2552 	if (!inet->mc_list)
2553 		return;
2554 
2555 	rtnl_lock();
2556 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2557 		struct in_device *in_dev;
2558 
2559 		inet->mc_list = iml->next_rcu;
2560 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2561 		(void) ip_mc_leave_src(sk, iml, in_dev);
2562 		if (in_dev)
2563 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2564 		/* decrease mem now to avoid the memleak warning */
2565 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2566 		kfree_rcu(iml, rcu);
2567 	}
2568 	rtnl_unlock();
2569 }
2570 
2571 /* called with rcu_read_lock() */
2572 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2573 {
2574 	struct ip_mc_list *im;
2575 	struct ip_mc_list __rcu **mc_hash;
2576 	struct ip_sf_list *psf;
2577 	int rv = 0;
2578 
2579 	mc_hash = rcu_dereference(in_dev->mc_hash);
2580 	if (mc_hash) {
2581 		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2582 
2583 		for (im = rcu_dereference(mc_hash[hash]);
2584 		     im != NULL;
2585 		     im = rcu_dereference(im->next_hash)) {
2586 			if (im->multiaddr == mc_addr)
2587 				break;
2588 		}
2589 	} else {
2590 		for_each_pmc_rcu(in_dev, im) {
2591 			if (im->multiaddr == mc_addr)
2592 				break;
2593 		}
2594 	}
2595 	if (im && proto == IPPROTO_IGMP) {
2596 		rv = 1;
2597 	} else if (im) {
2598 		if (src_addr) {
2599 			for (psf = im->sources; psf; psf = psf->sf_next) {
2600 				if (psf->sf_inaddr == src_addr)
2601 					break;
2602 			}
2603 			if (psf)
2604 				rv = psf->sf_count[MCAST_INCLUDE] ||
2605 					psf->sf_count[MCAST_EXCLUDE] !=
2606 					im->sfcount[MCAST_EXCLUDE];
2607 			else
2608 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2609 		} else
2610 			rv = 1; /* unspecified source; tentatively allow */
2611 	}
2612 	return rv;
2613 }
2614 
2615 #if defined(CONFIG_PROC_FS)
2616 struct igmp_mc_iter_state {
2617 	struct seq_net_private p;
2618 	struct net_device *dev;
2619 	struct in_device *in_dev;
2620 };
2621 
2622 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2623 
2624 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2625 {
2626 	struct net *net = seq_file_net(seq);
2627 	struct ip_mc_list *im = NULL;
2628 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2629 
2630 	state->in_dev = NULL;
2631 	for_each_netdev_rcu(net, state->dev) {
2632 		struct in_device *in_dev;
2633 
2634 		in_dev = __in_dev_get_rcu(state->dev);
2635 		if (!in_dev)
2636 			continue;
2637 		im = rcu_dereference(in_dev->mc_list);
2638 		if (im) {
2639 			state->in_dev = in_dev;
2640 			break;
2641 		}
2642 	}
2643 	return im;
2644 }
2645 
2646 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2647 {
2648 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2649 
2650 	im = rcu_dereference(im->next_rcu);
2651 	while (!im) {
2652 		state->dev = next_net_device_rcu(state->dev);
2653 		if (!state->dev) {
2654 			state->in_dev = NULL;
2655 			break;
2656 		}
2657 		state->in_dev = __in_dev_get_rcu(state->dev);
2658 		if (!state->in_dev)
2659 			continue;
2660 		im = rcu_dereference(state->in_dev->mc_list);
2661 	}
2662 	return im;
2663 }
2664 
2665 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2666 {
2667 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2668 	if (im)
2669 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2670 			--pos;
2671 	return pos ? NULL : im;
2672 }
2673 
2674 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2675 	__acquires(rcu)
2676 {
2677 	rcu_read_lock();
2678 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2679 }
2680 
2681 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2682 {
2683 	struct ip_mc_list *im;
2684 	if (v == SEQ_START_TOKEN)
2685 		im = igmp_mc_get_first(seq);
2686 	else
2687 		im = igmp_mc_get_next(seq, v);
2688 	++*pos;
2689 	return im;
2690 }
2691 
2692 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2693 	__releases(rcu)
2694 {
2695 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2696 
2697 	state->in_dev = NULL;
2698 	state->dev = NULL;
2699 	rcu_read_unlock();
2700 }
2701 
2702 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2703 {
2704 	if (v == SEQ_START_TOKEN)
2705 		seq_puts(seq,
2706 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2707 	else {
2708 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2709 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2710 		char   *querier;
2711 		long delta;
2712 
2713 #ifdef CONFIG_IP_MULTICAST
2714 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2715 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2716 			  "V3";
2717 #else
2718 		querier = "NONE";
2719 #endif
2720 
2721 		if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2722 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2723 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2724 		}
2725 
2726 		delta = im->timer.expires - jiffies;
2727 		seq_printf(seq,
2728 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2729 			   im->multiaddr, im->users,
2730 			   im->tm_running,
2731 			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2732 			   im->reporter);
2733 	}
2734 	return 0;
2735 }
2736 
2737 static const struct seq_operations igmp_mc_seq_ops = {
2738 	.start	=	igmp_mc_seq_start,
2739 	.next	=	igmp_mc_seq_next,
2740 	.stop	=	igmp_mc_seq_stop,
2741 	.show	=	igmp_mc_seq_show,
2742 };
2743 
2744 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2745 {
2746 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2747 			sizeof(struct igmp_mc_iter_state));
2748 }
2749 
2750 static const struct file_operations igmp_mc_seq_fops = {
2751 	.owner		=	THIS_MODULE,
2752 	.open		=	igmp_mc_seq_open,
2753 	.read		=	seq_read,
2754 	.llseek		=	seq_lseek,
2755 	.release	=	seq_release_net,
2756 };
2757 
2758 struct igmp_mcf_iter_state {
2759 	struct seq_net_private p;
2760 	struct net_device *dev;
2761 	struct in_device *idev;
2762 	struct ip_mc_list *im;
2763 };
2764 
2765 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2766 
2767 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2768 {
2769 	struct net *net = seq_file_net(seq);
2770 	struct ip_sf_list *psf = NULL;
2771 	struct ip_mc_list *im = NULL;
2772 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2773 
2774 	state->idev = NULL;
2775 	state->im = NULL;
2776 	for_each_netdev_rcu(net, state->dev) {
2777 		struct in_device *idev;
2778 		idev = __in_dev_get_rcu(state->dev);
2779 		if (unlikely(!idev))
2780 			continue;
2781 		im = rcu_dereference(idev->mc_list);
2782 		if (likely(im)) {
2783 			spin_lock_bh(&im->lock);
2784 			psf = im->sources;
2785 			if (likely(psf)) {
2786 				state->im = im;
2787 				state->idev = idev;
2788 				break;
2789 			}
2790 			spin_unlock_bh(&im->lock);
2791 		}
2792 	}
2793 	return psf;
2794 }
2795 
2796 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2797 {
2798 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2799 
2800 	psf = psf->sf_next;
2801 	while (!psf) {
2802 		spin_unlock_bh(&state->im->lock);
2803 		state->im = state->im->next;
2804 		while (!state->im) {
2805 			state->dev = next_net_device_rcu(state->dev);
2806 			if (!state->dev) {
2807 				state->idev = NULL;
2808 				goto out;
2809 			}
2810 			state->idev = __in_dev_get_rcu(state->dev);
2811 			if (!state->idev)
2812 				continue;
2813 			state->im = rcu_dereference(state->idev->mc_list);
2814 		}
2815 		if (!state->im)
2816 			break;
2817 		spin_lock_bh(&state->im->lock);
2818 		psf = state->im->sources;
2819 	}
2820 out:
2821 	return psf;
2822 }
2823 
2824 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2825 {
2826 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2827 	if (psf)
2828 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2829 			--pos;
2830 	return pos ? NULL : psf;
2831 }
2832 
2833 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2834 	__acquires(rcu)
2835 {
2836 	rcu_read_lock();
2837 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2838 }
2839 
2840 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2841 {
2842 	struct ip_sf_list *psf;
2843 	if (v == SEQ_START_TOKEN)
2844 		psf = igmp_mcf_get_first(seq);
2845 	else
2846 		psf = igmp_mcf_get_next(seq, v);
2847 	++*pos;
2848 	return psf;
2849 }
2850 
2851 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2852 	__releases(rcu)
2853 {
2854 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2855 	if (likely(state->im)) {
2856 		spin_unlock_bh(&state->im->lock);
2857 		state->im = NULL;
2858 	}
2859 	state->idev = NULL;
2860 	state->dev = NULL;
2861 	rcu_read_unlock();
2862 }
2863 
2864 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2865 {
2866 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2867 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2868 
2869 	if (v == SEQ_START_TOKEN) {
2870 		seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2871 	} else {
2872 		seq_printf(seq,
2873 			   "%3d %6.6s 0x%08x "
2874 			   "0x%08x %6lu %6lu\n",
2875 			   state->dev->ifindex, state->dev->name,
2876 			   ntohl(state->im->multiaddr),
2877 			   ntohl(psf->sf_inaddr),
2878 			   psf->sf_count[MCAST_INCLUDE],
2879 			   psf->sf_count[MCAST_EXCLUDE]);
2880 	}
2881 	return 0;
2882 }
2883 
2884 static const struct seq_operations igmp_mcf_seq_ops = {
2885 	.start	=	igmp_mcf_seq_start,
2886 	.next	=	igmp_mcf_seq_next,
2887 	.stop	=	igmp_mcf_seq_stop,
2888 	.show	=	igmp_mcf_seq_show,
2889 };
2890 
2891 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2892 {
2893 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2894 			sizeof(struct igmp_mcf_iter_state));
2895 }
2896 
2897 static const struct file_operations igmp_mcf_seq_fops = {
2898 	.owner		=	THIS_MODULE,
2899 	.open		=	igmp_mcf_seq_open,
2900 	.read		=	seq_read,
2901 	.llseek		=	seq_lseek,
2902 	.release	=	seq_release_net,
2903 };
2904 
2905 static int __net_init igmp_net_init(struct net *net)
2906 {
2907 	struct proc_dir_entry *pde;
2908 	int err;
2909 
2910 	pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2911 	if (!pde)
2912 		goto out_igmp;
2913 	pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2914 			  &igmp_mcf_seq_fops);
2915 	if (!pde)
2916 		goto out_mcfilter;
2917 	err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2918 				   SOCK_DGRAM, 0, net);
2919 	if (err < 0) {
2920 		pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2921 		       err);
2922 		goto out_sock;
2923 	}
2924 
2925 	return 0;
2926 
2927 out_sock:
2928 	remove_proc_entry("mcfilter", net->proc_net);
2929 out_mcfilter:
2930 	remove_proc_entry("igmp", net->proc_net);
2931 out_igmp:
2932 	return -ENOMEM;
2933 }
2934 
2935 static void __net_exit igmp_net_exit(struct net *net)
2936 {
2937 	remove_proc_entry("mcfilter", net->proc_net);
2938 	remove_proc_entry("igmp", net->proc_net);
2939 	inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2940 }
2941 
2942 static struct pernet_operations igmp_net_ops = {
2943 	.init = igmp_net_init,
2944 	.exit = igmp_net_exit,
2945 };
2946 #endif
2947 
2948 static int igmp_netdev_event(struct notifier_block *this,
2949 			     unsigned long event, void *ptr)
2950 {
2951 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2952 	struct in_device *in_dev;
2953 
2954 	switch (event) {
2955 	case NETDEV_RESEND_IGMP:
2956 		in_dev = __in_dev_get_rtnl(dev);
2957 		if (in_dev)
2958 			ip_mc_rejoin_groups(in_dev);
2959 		break;
2960 	default:
2961 		break;
2962 	}
2963 	return NOTIFY_DONE;
2964 }
2965 
2966 static struct notifier_block igmp_notifier = {
2967 	.notifier_call = igmp_netdev_event,
2968 };
2969 
2970 int __init igmp_mc_init(void)
2971 {
2972 #if defined(CONFIG_PROC_FS)
2973 	int err;
2974 
2975 	err = register_pernet_subsys(&igmp_net_ops);
2976 	if (err)
2977 		return err;
2978 	err = register_netdevice_notifier(&igmp_notifier);
2979 	if (err)
2980 		goto reg_notif_fail;
2981 	return 0;
2982 
2983 reg_notif_fail:
2984 	unregister_pernet_subsys(&igmp_net_ops);
2985 	return err;
2986 #else
2987 	return register_netdevice_notifier(&igmp_notifier);
2988 #endif
2989 }
2990