xref: /openbmc/linux/net/ipv4/igmp.c (revision 6dfcd296)
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 #ifdef CONFIG_IP_MULTICAST
111 /* Parameter names and values are taken from igmp-v2-06 draft */
112 
113 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT		(400*HZ)
114 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT		(400*HZ)
115 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL	(10*HZ)
116 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL	(1*HZ)
117 #define IGMP_QUERY_RESPONSE_INTERVAL		(10*HZ)
118 #define IGMP_QUERY_ROBUSTNESS_VARIABLE		2
119 
120 
121 #define IGMP_INITIAL_REPORT_DELAY		(1)
122 
123 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
124  * IGMP specs require to report membership immediately after
125  * joining a group, but we delay the first report by a
126  * small interval. It seems more natural and still does not
127  * contradict to specs provided this delay is small enough.
128  */
129 
130 #define IGMP_V1_SEEN(in_dev) \
131 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
132 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
133 	 ((in_dev)->mr_v1_seen && \
134 	  time_before(jiffies, (in_dev)->mr_v1_seen)))
135 #define IGMP_V2_SEEN(in_dev) \
136 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
137 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
138 	 ((in_dev)->mr_v2_seen && \
139 	  time_before(jiffies, (in_dev)->mr_v2_seen)))
140 
141 static int unsolicited_report_interval(struct in_device *in_dev)
142 {
143 	int interval_ms, interval_jiffies;
144 
145 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
146 		interval_ms = IN_DEV_CONF_GET(
147 			in_dev,
148 			IGMPV2_UNSOLICITED_REPORT_INTERVAL);
149 	else /* v3 */
150 		interval_ms = IN_DEV_CONF_GET(
151 			in_dev,
152 			IGMPV3_UNSOLICITED_REPORT_INTERVAL);
153 
154 	interval_jiffies = msecs_to_jiffies(interval_ms);
155 
156 	/* _timer functions can't handle a delay of 0 jiffies so ensure
157 	 *  we always return a positive value.
158 	 */
159 	if (interval_jiffies <= 0)
160 		interval_jiffies = 1;
161 	return interval_jiffies;
162 }
163 
164 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
165 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
166 static void igmpv3_clear_delrec(struct in_device *in_dev);
167 static int sf_setstate(struct ip_mc_list *pmc);
168 static void sf_markstate(struct ip_mc_list *pmc);
169 #endif
170 static void ip_mc_clear_src(struct ip_mc_list *pmc);
171 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
172 			 int sfcount, __be32 *psfsrc, int delta);
173 
174 static void ip_ma_put(struct ip_mc_list *im)
175 {
176 	if (atomic_dec_and_test(&im->refcnt)) {
177 		in_dev_put(im->interface);
178 		kfree_rcu(im, rcu);
179 	}
180 }
181 
182 #define for_each_pmc_rcu(in_dev, pmc)				\
183 	for (pmc = rcu_dereference(in_dev->mc_list);		\
184 	     pmc != NULL;					\
185 	     pmc = rcu_dereference(pmc->next_rcu))
186 
187 #define for_each_pmc_rtnl(in_dev, pmc)				\
188 	for (pmc = rtnl_dereference(in_dev->mc_list);		\
189 	     pmc != NULL;					\
190 	     pmc = rtnl_dereference(pmc->next_rcu))
191 
192 #ifdef CONFIG_IP_MULTICAST
193 
194 /*
195  *	Timer management
196  */
197 
198 static void igmp_stop_timer(struct ip_mc_list *im)
199 {
200 	spin_lock_bh(&im->lock);
201 	if (del_timer(&im->timer))
202 		atomic_dec(&im->refcnt);
203 	im->tm_running = 0;
204 	im->reporter = 0;
205 	im->unsolicit_count = 0;
206 	spin_unlock_bh(&im->lock);
207 }
208 
209 /* It must be called with locked im->lock */
210 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
211 {
212 	int tv = prandom_u32() % max_delay;
213 
214 	im->tm_running = 1;
215 	if (!mod_timer(&im->timer, jiffies+tv+2))
216 		atomic_inc(&im->refcnt);
217 }
218 
219 static void igmp_gq_start_timer(struct in_device *in_dev)
220 {
221 	int tv = prandom_u32() % in_dev->mr_maxdelay;
222 
223 	in_dev->mr_gq_running = 1;
224 	if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
225 		in_dev_hold(in_dev);
226 }
227 
228 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
229 {
230 	int tv = prandom_u32() % delay;
231 
232 	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
233 		in_dev_hold(in_dev);
234 }
235 
236 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
237 {
238 	spin_lock_bh(&im->lock);
239 	im->unsolicit_count = 0;
240 	if (del_timer(&im->timer)) {
241 		if ((long)(im->timer.expires-jiffies) < max_delay) {
242 			add_timer(&im->timer);
243 			im->tm_running = 1;
244 			spin_unlock_bh(&im->lock);
245 			return;
246 		}
247 		atomic_dec(&im->refcnt);
248 	}
249 	igmp_start_timer(im, max_delay);
250 	spin_unlock_bh(&im->lock);
251 }
252 
253 
254 /*
255  *	Send an IGMP report.
256  */
257 
258 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
259 
260 
261 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
262 	int gdeleted, int sdeleted)
263 {
264 	switch (type) {
265 	case IGMPV3_MODE_IS_INCLUDE:
266 	case IGMPV3_MODE_IS_EXCLUDE:
267 		if (gdeleted || sdeleted)
268 			return 0;
269 		if (!(pmc->gsquery && !psf->sf_gsresp)) {
270 			if (pmc->sfmode == MCAST_INCLUDE)
271 				return 1;
272 			/* don't include if this source is excluded
273 			 * in all filters
274 			 */
275 			if (psf->sf_count[MCAST_INCLUDE])
276 				return type == IGMPV3_MODE_IS_INCLUDE;
277 			return pmc->sfcount[MCAST_EXCLUDE] ==
278 				psf->sf_count[MCAST_EXCLUDE];
279 		}
280 		return 0;
281 	case IGMPV3_CHANGE_TO_INCLUDE:
282 		if (gdeleted || sdeleted)
283 			return 0;
284 		return psf->sf_count[MCAST_INCLUDE] != 0;
285 	case IGMPV3_CHANGE_TO_EXCLUDE:
286 		if (gdeleted || sdeleted)
287 			return 0;
288 		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
289 		    psf->sf_count[MCAST_INCLUDE])
290 			return 0;
291 		return pmc->sfcount[MCAST_EXCLUDE] ==
292 			psf->sf_count[MCAST_EXCLUDE];
293 	case IGMPV3_ALLOW_NEW_SOURCES:
294 		if (gdeleted || !psf->sf_crcount)
295 			return 0;
296 		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
297 	case IGMPV3_BLOCK_OLD_SOURCES:
298 		if (pmc->sfmode == MCAST_INCLUDE)
299 			return gdeleted || (psf->sf_crcount && sdeleted);
300 		return psf->sf_crcount && !gdeleted && !sdeleted;
301 	}
302 	return 0;
303 }
304 
305 static int
306 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
307 {
308 	struct ip_sf_list *psf;
309 	int scount = 0;
310 
311 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
312 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
313 			continue;
314 		scount++;
315 	}
316 	return scount;
317 }
318 
319 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
320 {
321 	struct sk_buff *skb;
322 	struct rtable *rt;
323 	struct iphdr *pip;
324 	struct igmpv3_report *pig;
325 	struct net *net = dev_net(dev);
326 	struct flowi4 fl4;
327 	int hlen = LL_RESERVED_SPACE(dev);
328 	int tlen = dev->needed_tailroom;
329 	unsigned int size = mtu;
330 
331 	while (1) {
332 		skb = alloc_skb(size + hlen + tlen,
333 				GFP_ATOMIC | __GFP_NOWARN);
334 		if (skb)
335 			break;
336 		size >>= 1;
337 		if (size < 256)
338 			return NULL;
339 	}
340 	skb->priority = TC_PRIO_CONTROL;
341 
342 	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
343 				   0, 0,
344 				   IPPROTO_IGMP, 0, dev->ifindex);
345 	if (IS_ERR(rt)) {
346 		kfree_skb(skb);
347 		return NULL;
348 	}
349 
350 	skb_dst_set(skb, &rt->dst);
351 	skb->dev = dev;
352 
353 	skb_reserve(skb, hlen);
354 	skb_tailroom_reserve(skb, mtu, tlen);
355 
356 	skb_reset_network_header(skb);
357 	pip = ip_hdr(skb);
358 	skb_put(skb, sizeof(struct iphdr) + 4);
359 
360 	pip->version  = 4;
361 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
362 	pip->tos      = 0xc0;
363 	pip->frag_off = htons(IP_DF);
364 	pip->ttl      = 1;
365 	pip->daddr    = fl4.daddr;
366 	pip->saddr    = fl4.saddr;
367 	pip->protocol = IPPROTO_IGMP;
368 	pip->tot_len  = 0;	/* filled in later */
369 	ip_select_ident(net, skb, NULL);
370 	((u8 *)&pip[1])[0] = IPOPT_RA;
371 	((u8 *)&pip[1])[1] = 4;
372 	((u8 *)&pip[1])[2] = 0;
373 	((u8 *)&pip[1])[3] = 0;
374 
375 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
376 	skb_put(skb, sizeof(*pig));
377 	pig = igmpv3_report_hdr(skb);
378 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
379 	pig->resv1 = 0;
380 	pig->csum = 0;
381 	pig->resv2 = 0;
382 	pig->ngrec = 0;
383 	return skb;
384 }
385 
386 static int igmpv3_sendpack(struct sk_buff *skb)
387 {
388 	struct igmphdr *pig = igmp_hdr(skb);
389 	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
390 
391 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
392 
393 	return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
394 }
395 
396 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
397 {
398 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
399 }
400 
401 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
402 	int type, struct igmpv3_grec **ppgr)
403 {
404 	struct net_device *dev = pmc->interface->dev;
405 	struct igmpv3_report *pih;
406 	struct igmpv3_grec *pgr;
407 
408 	if (!skb)
409 		skb = igmpv3_newpack(dev, dev->mtu);
410 	if (!skb)
411 		return NULL;
412 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
413 	pgr->grec_type = type;
414 	pgr->grec_auxwords = 0;
415 	pgr->grec_nsrcs = 0;
416 	pgr->grec_mca = pmc->multiaddr;
417 	pih = igmpv3_report_hdr(skb);
418 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
419 	*ppgr = pgr;
420 	return skb;
421 }
422 
423 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
424 
425 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
426 	int type, int gdeleted, int sdeleted)
427 {
428 	struct net_device *dev = pmc->interface->dev;
429 	struct net *net = dev_net(dev);
430 	struct igmpv3_report *pih;
431 	struct igmpv3_grec *pgr = NULL;
432 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
433 	int scount, stotal, first, isquery, truncate;
434 
435 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
436 		return skb;
437 	if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
438 		return skb;
439 
440 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
441 		  type == IGMPV3_MODE_IS_EXCLUDE;
442 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
443 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
444 
445 	stotal = scount = 0;
446 
447 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
448 
449 	if (!*psf_list)
450 		goto empty_source;
451 
452 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
453 
454 	/* EX and TO_EX get a fresh packet, if needed */
455 	if (truncate) {
456 		if (pih && pih->ngrec &&
457 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
458 			if (skb)
459 				igmpv3_sendpack(skb);
460 			skb = igmpv3_newpack(dev, dev->mtu);
461 		}
462 	}
463 	first = 1;
464 	psf_prev = NULL;
465 	for (psf = *psf_list; psf; psf = psf_next) {
466 		__be32 *psrc;
467 
468 		psf_next = psf->sf_next;
469 
470 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
471 			psf_prev = psf;
472 			continue;
473 		}
474 
475 		/* Based on RFC3376 5.1. Should not send source-list change
476 		 * records when there is a filter mode change.
477 		 */
478 		if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) ||
479 		     (!gdeleted && pmc->crcount)) &&
480 		    (type == IGMPV3_ALLOW_NEW_SOURCES ||
481 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount)
482 			goto decrease_sf_crcount;
483 
484 		/* clear marks on query responses */
485 		if (isquery)
486 			psf->sf_gsresp = 0;
487 
488 		if (AVAILABLE(skb) < sizeof(__be32) +
489 		    first*sizeof(struct igmpv3_grec)) {
490 			if (truncate && !first)
491 				break;	 /* truncate these */
492 			if (pgr)
493 				pgr->grec_nsrcs = htons(scount);
494 			if (skb)
495 				igmpv3_sendpack(skb);
496 			skb = igmpv3_newpack(dev, dev->mtu);
497 			first = 1;
498 			scount = 0;
499 		}
500 		if (first) {
501 			skb = add_grhead(skb, pmc, type, &pgr);
502 			first = 0;
503 		}
504 		if (!skb)
505 			return NULL;
506 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
507 		*psrc = psf->sf_inaddr;
508 		scount++; stotal++;
509 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
510 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
511 decrease_sf_crcount:
512 			psf->sf_crcount--;
513 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
514 				if (psf_prev)
515 					psf_prev->sf_next = psf->sf_next;
516 				else
517 					*psf_list = psf->sf_next;
518 				kfree(psf);
519 				continue;
520 			}
521 		}
522 		psf_prev = psf;
523 	}
524 
525 empty_source:
526 	if (!stotal) {
527 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
528 		    type == IGMPV3_BLOCK_OLD_SOURCES)
529 			return skb;
530 		if (pmc->crcount || isquery) {
531 			/* make sure we have room for group header */
532 			if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
533 				igmpv3_sendpack(skb);
534 				skb = NULL; /* add_grhead will get a new one */
535 			}
536 			skb = add_grhead(skb, pmc, type, &pgr);
537 		}
538 	}
539 	if (pgr)
540 		pgr->grec_nsrcs = htons(scount);
541 
542 	if (isquery)
543 		pmc->gsquery = 0;	/* clear query state on report */
544 	return skb;
545 }
546 
547 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
548 {
549 	struct sk_buff *skb = NULL;
550 	struct net *net = dev_net(in_dev->dev);
551 	int type;
552 
553 	if (!pmc) {
554 		rcu_read_lock();
555 		for_each_pmc_rcu(in_dev, pmc) {
556 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
557 				continue;
558 			if (ipv4_is_local_multicast(pmc->multiaddr) &&
559 			     !net->ipv4.sysctl_igmp_llm_reports)
560 				continue;
561 			spin_lock_bh(&pmc->lock);
562 			if (pmc->sfcount[MCAST_EXCLUDE])
563 				type = IGMPV3_MODE_IS_EXCLUDE;
564 			else
565 				type = IGMPV3_MODE_IS_INCLUDE;
566 			skb = add_grec(skb, pmc, type, 0, 0);
567 			spin_unlock_bh(&pmc->lock);
568 		}
569 		rcu_read_unlock();
570 	} else {
571 		spin_lock_bh(&pmc->lock);
572 		if (pmc->sfcount[MCAST_EXCLUDE])
573 			type = IGMPV3_MODE_IS_EXCLUDE;
574 		else
575 			type = IGMPV3_MODE_IS_INCLUDE;
576 		skb = add_grec(skb, pmc, type, 0, 0);
577 		spin_unlock_bh(&pmc->lock);
578 	}
579 	if (!skb)
580 		return 0;
581 	return igmpv3_sendpack(skb);
582 }
583 
584 /*
585  * remove zero-count source records from a source filter list
586  */
587 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
588 {
589 	struct ip_sf_list *psf_prev, *psf_next, *psf;
590 
591 	psf_prev = NULL;
592 	for (psf = *ppsf; psf; psf = psf_next) {
593 		psf_next = psf->sf_next;
594 		if (psf->sf_crcount == 0) {
595 			if (psf_prev)
596 				psf_prev->sf_next = psf->sf_next;
597 			else
598 				*ppsf = psf->sf_next;
599 			kfree(psf);
600 		} else
601 			psf_prev = psf;
602 	}
603 }
604 
605 static void igmpv3_send_cr(struct in_device *in_dev)
606 {
607 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
608 	struct sk_buff *skb = NULL;
609 	int type, dtype;
610 
611 	rcu_read_lock();
612 	spin_lock_bh(&in_dev->mc_tomb_lock);
613 
614 	/* deleted MCA's */
615 	pmc_prev = NULL;
616 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
617 		pmc_next = pmc->next;
618 		if (pmc->sfmode == MCAST_INCLUDE) {
619 			type = IGMPV3_BLOCK_OLD_SOURCES;
620 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
621 			skb = add_grec(skb, pmc, type, 1, 0);
622 			skb = add_grec(skb, pmc, dtype, 1, 1);
623 		}
624 		if (pmc->crcount) {
625 			if (pmc->sfmode == MCAST_EXCLUDE) {
626 				type = IGMPV3_CHANGE_TO_INCLUDE;
627 				skb = add_grec(skb, pmc, type, 1, 0);
628 			}
629 			pmc->crcount--;
630 			if (pmc->crcount == 0) {
631 				igmpv3_clear_zeros(&pmc->tomb);
632 				igmpv3_clear_zeros(&pmc->sources);
633 			}
634 		}
635 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
636 			if (pmc_prev)
637 				pmc_prev->next = pmc_next;
638 			else
639 				in_dev->mc_tomb = pmc_next;
640 			in_dev_put(pmc->interface);
641 			kfree(pmc);
642 		} else
643 			pmc_prev = pmc;
644 	}
645 	spin_unlock_bh(&in_dev->mc_tomb_lock);
646 
647 	/* change recs */
648 	for_each_pmc_rcu(in_dev, pmc) {
649 		spin_lock_bh(&pmc->lock);
650 		if (pmc->sfcount[MCAST_EXCLUDE]) {
651 			type = IGMPV3_BLOCK_OLD_SOURCES;
652 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
653 		} else {
654 			type = IGMPV3_ALLOW_NEW_SOURCES;
655 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
656 		}
657 		skb = add_grec(skb, pmc, type, 0, 0);
658 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
659 
660 		/* filter mode changes */
661 		if (pmc->crcount) {
662 			if (pmc->sfmode == MCAST_EXCLUDE)
663 				type = IGMPV3_CHANGE_TO_EXCLUDE;
664 			else
665 				type = IGMPV3_CHANGE_TO_INCLUDE;
666 			skb = add_grec(skb, pmc, type, 0, 0);
667 			pmc->crcount--;
668 		}
669 		spin_unlock_bh(&pmc->lock);
670 	}
671 	rcu_read_unlock();
672 
673 	if (!skb)
674 		return;
675 	(void) igmpv3_sendpack(skb);
676 }
677 
678 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
679 	int type)
680 {
681 	struct sk_buff *skb;
682 	struct iphdr *iph;
683 	struct igmphdr *ih;
684 	struct rtable *rt;
685 	struct net_device *dev = in_dev->dev;
686 	struct net *net = dev_net(dev);
687 	__be32	group = pmc ? pmc->multiaddr : 0;
688 	struct flowi4 fl4;
689 	__be32	dst;
690 	int hlen, tlen;
691 
692 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
693 		return igmpv3_send_report(in_dev, pmc);
694 
695 	if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
696 		return 0;
697 
698 	if (type == IGMP_HOST_LEAVE_MESSAGE)
699 		dst = IGMP_ALL_ROUTER;
700 	else
701 		dst = group;
702 
703 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
704 				   0, 0,
705 				   IPPROTO_IGMP, 0, dev->ifindex);
706 	if (IS_ERR(rt))
707 		return -1;
708 
709 	hlen = LL_RESERVED_SPACE(dev);
710 	tlen = dev->needed_tailroom;
711 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
712 	if (!skb) {
713 		ip_rt_put(rt);
714 		return -1;
715 	}
716 	skb->priority = TC_PRIO_CONTROL;
717 
718 	skb_dst_set(skb, &rt->dst);
719 
720 	skb_reserve(skb, hlen);
721 
722 	skb_reset_network_header(skb);
723 	iph = ip_hdr(skb);
724 	skb_put(skb, sizeof(struct iphdr) + 4);
725 
726 	iph->version  = 4;
727 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
728 	iph->tos      = 0xc0;
729 	iph->frag_off = htons(IP_DF);
730 	iph->ttl      = 1;
731 	iph->daddr    = dst;
732 	iph->saddr    = fl4.saddr;
733 	iph->protocol = IPPROTO_IGMP;
734 	ip_select_ident(net, skb, NULL);
735 	((u8 *)&iph[1])[0] = IPOPT_RA;
736 	((u8 *)&iph[1])[1] = 4;
737 	((u8 *)&iph[1])[2] = 0;
738 	((u8 *)&iph[1])[3] = 0;
739 
740 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
741 	ih->type = type;
742 	ih->code = 0;
743 	ih->csum = 0;
744 	ih->group = group;
745 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
746 
747 	return ip_local_out(net, skb->sk, skb);
748 }
749 
750 static void igmp_gq_timer_expire(unsigned long data)
751 {
752 	struct in_device *in_dev = (struct in_device *)data;
753 
754 	in_dev->mr_gq_running = 0;
755 	igmpv3_send_report(in_dev, NULL);
756 	in_dev_put(in_dev);
757 }
758 
759 static void igmp_ifc_timer_expire(unsigned long data)
760 {
761 	struct in_device *in_dev = (struct in_device *)data;
762 
763 	igmpv3_send_cr(in_dev);
764 	if (in_dev->mr_ifc_count) {
765 		in_dev->mr_ifc_count--;
766 		igmp_ifc_start_timer(in_dev,
767 				     unsolicited_report_interval(in_dev));
768 	}
769 	in_dev_put(in_dev);
770 }
771 
772 static void igmp_ifc_event(struct in_device *in_dev)
773 {
774 	struct net *net = dev_net(in_dev->dev);
775 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
776 		return;
777 	in_dev->mr_ifc_count = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
778 	igmp_ifc_start_timer(in_dev, 1);
779 }
780 
781 
782 static void igmp_timer_expire(unsigned long data)
783 {
784 	struct ip_mc_list *im = (struct ip_mc_list *)data;
785 	struct in_device *in_dev = im->interface;
786 
787 	spin_lock(&im->lock);
788 	im->tm_running = 0;
789 
790 	if (im->unsolicit_count) {
791 		im->unsolicit_count--;
792 		igmp_start_timer(im, unsolicited_report_interval(in_dev));
793 	}
794 	im->reporter = 1;
795 	spin_unlock(&im->lock);
796 
797 	if (IGMP_V1_SEEN(in_dev))
798 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
799 	else if (IGMP_V2_SEEN(in_dev))
800 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
801 	else
802 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
803 
804 	ip_ma_put(im);
805 }
806 
807 /* mark EXCLUDE-mode sources */
808 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
809 {
810 	struct ip_sf_list *psf;
811 	int i, scount;
812 
813 	scount = 0;
814 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
815 		if (scount == nsrcs)
816 			break;
817 		for (i = 0; i < nsrcs; i++) {
818 			/* skip inactive filters */
819 			if (psf->sf_count[MCAST_INCLUDE] ||
820 			    pmc->sfcount[MCAST_EXCLUDE] !=
821 			    psf->sf_count[MCAST_EXCLUDE])
822 				break;
823 			if (srcs[i] == psf->sf_inaddr) {
824 				scount++;
825 				break;
826 			}
827 		}
828 	}
829 	pmc->gsquery = 0;
830 	if (scount == nsrcs)	/* all sources excluded */
831 		return 0;
832 	return 1;
833 }
834 
835 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
836 {
837 	struct ip_sf_list *psf;
838 	int i, scount;
839 
840 	if (pmc->sfmode == MCAST_EXCLUDE)
841 		return igmp_xmarksources(pmc, nsrcs, srcs);
842 
843 	/* mark INCLUDE-mode sources */
844 	scount = 0;
845 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
846 		if (scount == nsrcs)
847 			break;
848 		for (i = 0; i < nsrcs; i++)
849 			if (srcs[i] == psf->sf_inaddr) {
850 				psf->sf_gsresp = 1;
851 				scount++;
852 				break;
853 			}
854 	}
855 	if (!scount) {
856 		pmc->gsquery = 0;
857 		return 0;
858 	}
859 	pmc->gsquery = 1;
860 	return 1;
861 }
862 
863 /* return true if packet was dropped */
864 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
865 {
866 	struct ip_mc_list *im;
867 	struct net *net = dev_net(in_dev->dev);
868 
869 	/* Timers are only set for non-local groups */
870 
871 	if (group == IGMP_ALL_HOSTS)
872 		return false;
873 	if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
874 		return false;
875 
876 	rcu_read_lock();
877 	for_each_pmc_rcu(in_dev, im) {
878 		if (im->multiaddr == group) {
879 			igmp_stop_timer(im);
880 			break;
881 		}
882 	}
883 	rcu_read_unlock();
884 	return false;
885 }
886 
887 /* return true if packet was dropped */
888 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
889 	int len)
890 {
891 	struct igmphdr 		*ih = igmp_hdr(skb);
892 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
893 	struct ip_mc_list	*im;
894 	__be32			group = ih->group;
895 	int			max_delay;
896 	int			mark = 0;
897 	struct net		*net = dev_net(in_dev->dev);
898 
899 
900 	if (len == 8) {
901 		if (ih->code == 0) {
902 			/* Alas, old v1 router presents here. */
903 
904 			max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
905 			in_dev->mr_v1_seen = jiffies +
906 				IGMP_V1_ROUTER_PRESENT_TIMEOUT;
907 			group = 0;
908 		} else {
909 			/* v2 router present */
910 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
911 			in_dev->mr_v2_seen = jiffies +
912 				IGMP_V2_ROUTER_PRESENT_TIMEOUT;
913 		}
914 		/* cancel the interface change timer */
915 		in_dev->mr_ifc_count = 0;
916 		if (del_timer(&in_dev->mr_ifc_timer))
917 			__in_dev_put(in_dev);
918 		/* clear deleted report items */
919 		igmpv3_clear_delrec(in_dev);
920 	} else if (len < 12) {
921 		return true;	/* ignore bogus packet; freed by caller */
922 	} else if (IGMP_V1_SEEN(in_dev)) {
923 		/* This is a v3 query with v1 queriers present */
924 		max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
925 		group = 0;
926 	} else if (IGMP_V2_SEEN(in_dev)) {
927 		/* this is a v3 query with v2 queriers present;
928 		 * Interpretation of the max_delay code is problematic here.
929 		 * A real v2 host would use ih_code directly, while v3 has a
930 		 * different encoding. We use the v3 encoding as more likely
931 		 * to be intended in a v3 query.
932 		 */
933 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
934 		if (!max_delay)
935 			max_delay = 1;	/* can't mod w/ 0 */
936 	} else { /* v3 */
937 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
938 			return true;
939 
940 		ih3 = igmpv3_query_hdr(skb);
941 		if (ih3->nsrcs) {
942 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
943 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
944 				return true;
945 			ih3 = igmpv3_query_hdr(skb);
946 		}
947 
948 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
949 		if (!max_delay)
950 			max_delay = 1;	/* can't mod w/ 0 */
951 		in_dev->mr_maxdelay = max_delay;
952 		if (ih3->qrv)
953 			in_dev->mr_qrv = ih3->qrv;
954 		if (!group) { /* general query */
955 			if (ih3->nsrcs)
956 				return true;	/* no sources allowed */
957 			igmp_gq_start_timer(in_dev);
958 			return false;
959 		}
960 		/* mark sources to include, if group & source-specific */
961 		mark = ih3->nsrcs != 0;
962 	}
963 
964 	/*
965 	 * - Start the timers in all of our membership records
966 	 *   that the query applies to for the interface on
967 	 *   which the query arrived excl. those that belong
968 	 *   to a "local" group (224.0.0.X)
969 	 * - For timers already running check if they need to
970 	 *   be reset.
971 	 * - Use the igmp->igmp_code field as the maximum
972 	 *   delay possible
973 	 */
974 	rcu_read_lock();
975 	for_each_pmc_rcu(in_dev, im) {
976 		int changed;
977 
978 		if (group && group != im->multiaddr)
979 			continue;
980 		if (im->multiaddr == IGMP_ALL_HOSTS)
981 			continue;
982 		if (ipv4_is_local_multicast(im->multiaddr) &&
983 		    !net->ipv4.sysctl_igmp_llm_reports)
984 			continue;
985 		spin_lock_bh(&im->lock);
986 		if (im->tm_running)
987 			im->gsquery = im->gsquery && mark;
988 		else
989 			im->gsquery = mark;
990 		changed = !im->gsquery ||
991 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
992 		spin_unlock_bh(&im->lock);
993 		if (changed)
994 			igmp_mod_timer(im, max_delay);
995 	}
996 	rcu_read_unlock();
997 	return false;
998 }
999 
1000 /* called in rcu_read_lock() section */
1001 int igmp_rcv(struct sk_buff *skb)
1002 {
1003 	/* This basically follows the spec line by line -- see RFC1112 */
1004 	struct igmphdr *ih;
1005 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1006 	int len = skb->len;
1007 	bool dropped = true;
1008 
1009 	if (!in_dev)
1010 		goto drop;
1011 
1012 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1013 		goto drop;
1014 
1015 	if (skb_checksum_simple_validate(skb))
1016 		goto drop;
1017 
1018 	ih = igmp_hdr(skb);
1019 	switch (ih->type) {
1020 	case IGMP_HOST_MEMBERSHIP_QUERY:
1021 		dropped = igmp_heard_query(in_dev, skb, len);
1022 		break;
1023 	case IGMP_HOST_MEMBERSHIP_REPORT:
1024 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1025 		/* Is it our report looped back? */
1026 		if (rt_is_output_route(skb_rtable(skb)))
1027 			break;
1028 		/* don't rely on MC router hearing unicast reports */
1029 		if (skb->pkt_type == PACKET_MULTICAST ||
1030 		    skb->pkt_type == PACKET_BROADCAST)
1031 			dropped = igmp_heard_report(in_dev, ih->group);
1032 		break;
1033 	case IGMP_PIM:
1034 #ifdef CONFIG_IP_PIMSM_V1
1035 		return pim_rcv_v1(skb);
1036 #endif
1037 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1038 	case IGMP_DVMRP:
1039 	case IGMP_TRACE:
1040 	case IGMP_HOST_LEAVE_MESSAGE:
1041 	case IGMP_MTRACE:
1042 	case IGMP_MTRACE_RESP:
1043 		break;
1044 	default:
1045 		break;
1046 	}
1047 
1048 drop:
1049 	if (dropped)
1050 		kfree_skb(skb);
1051 	else
1052 		consume_skb(skb);
1053 	return 0;
1054 }
1055 
1056 #endif
1057 
1058 
1059 /*
1060  *	Add a filter to a device
1061  */
1062 
1063 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1064 {
1065 	char buf[MAX_ADDR_LEN];
1066 	struct net_device *dev = in_dev->dev;
1067 
1068 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1069 	   We will get multicast token leakage, when IFF_MULTICAST
1070 	   is changed. This check should be done in ndo_set_rx_mode
1071 	   routine. Something sort of:
1072 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1073 	   --ANK
1074 	   */
1075 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1076 		dev_mc_add(dev, buf);
1077 }
1078 
1079 /*
1080  *	Remove a filter from a device
1081  */
1082 
1083 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1084 {
1085 	char buf[MAX_ADDR_LEN];
1086 	struct net_device *dev = in_dev->dev;
1087 
1088 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1089 		dev_mc_del(dev, buf);
1090 }
1091 
1092 #ifdef CONFIG_IP_MULTICAST
1093 /*
1094  * deleted ip_mc_list manipulation
1095  */
1096 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1097 {
1098 	struct ip_mc_list *pmc;
1099 	struct net *net = dev_net(in_dev->dev);
1100 
1101 	/* this is an "ip_mc_list" for convenience; only the fields below
1102 	 * are actually used. In particular, the refcnt and users are not
1103 	 * used for management of the delete list. Using the same structure
1104 	 * for deleted items allows change reports to use common code with
1105 	 * non-deleted or query-response MCA's.
1106 	 */
1107 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1108 	if (!pmc)
1109 		return;
1110 	spin_lock_bh(&im->lock);
1111 	pmc->interface = im->interface;
1112 	in_dev_hold(in_dev);
1113 	pmc->multiaddr = im->multiaddr;
1114 	pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1115 	pmc->sfmode = im->sfmode;
1116 	if (pmc->sfmode == MCAST_INCLUDE) {
1117 		struct ip_sf_list *psf;
1118 
1119 		pmc->tomb = im->tomb;
1120 		pmc->sources = im->sources;
1121 		im->tomb = im->sources = NULL;
1122 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1123 			psf->sf_crcount = pmc->crcount;
1124 	}
1125 	spin_unlock_bh(&im->lock);
1126 
1127 	spin_lock_bh(&in_dev->mc_tomb_lock);
1128 	pmc->next = in_dev->mc_tomb;
1129 	in_dev->mc_tomb = pmc;
1130 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1131 }
1132 
1133 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1134 {
1135 	struct ip_mc_list *pmc, *pmc_prev;
1136 	struct ip_sf_list *psf, *psf_next;
1137 
1138 	spin_lock_bh(&in_dev->mc_tomb_lock);
1139 	pmc_prev = NULL;
1140 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1141 		if (pmc->multiaddr == multiaddr)
1142 			break;
1143 		pmc_prev = pmc;
1144 	}
1145 	if (pmc) {
1146 		if (pmc_prev)
1147 			pmc_prev->next = pmc->next;
1148 		else
1149 			in_dev->mc_tomb = pmc->next;
1150 	}
1151 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1152 	if (pmc) {
1153 		for (psf = pmc->tomb; psf; psf = psf_next) {
1154 			psf_next = psf->sf_next;
1155 			kfree(psf);
1156 		}
1157 		in_dev_put(pmc->interface);
1158 		kfree(pmc);
1159 	}
1160 }
1161 
1162 static void igmpv3_clear_delrec(struct in_device *in_dev)
1163 {
1164 	struct ip_mc_list *pmc, *nextpmc;
1165 
1166 	spin_lock_bh(&in_dev->mc_tomb_lock);
1167 	pmc = in_dev->mc_tomb;
1168 	in_dev->mc_tomb = NULL;
1169 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1170 
1171 	for (; pmc; pmc = nextpmc) {
1172 		nextpmc = pmc->next;
1173 		ip_mc_clear_src(pmc);
1174 		in_dev_put(pmc->interface);
1175 		kfree(pmc);
1176 	}
1177 	/* clear dead sources, too */
1178 	rcu_read_lock();
1179 	for_each_pmc_rcu(in_dev, pmc) {
1180 		struct ip_sf_list *psf, *psf_next;
1181 
1182 		spin_lock_bh(&pmc->lock);
1183 		psf = pmc->tomb;
1184 		pmc->tomb = NULL;
1185 		spin_unlock_bh(&pmc->lock);
1186 		for (; psf; psf = psf_next) {
1187 			psf_next = psf->sf_next;
1188 			kfree(psf);
1189 		}
1190 	}
1191 	rcu_read_unlock();
1192 }
1193 #endif
1194 
1195 static void igmp_group_dropped(struct ip_mc_list *im)
1196 {
1197 	struct in_device *in_dev = im->interface;
1198 #ifdef CONFIG_IP_MULTICAST
1199 	struct net *net = dev_net(in_dev->dev);
1200 	int reporter;
1201 #endif
1202 
1203 	if (im->loaded) {
1204 		im->loaded = 0;
1205 		ip_mc_filter_del(in_dev, im->multiaddr);
1206 	}
1207 
1208 #ifdef CONFIG_IP_MULTICAST
1209 	if (im->multiaddr == IGMP_ALL_HOSTS)
1210 		return;
1211 	if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1212 		return;
1213 
1214 	reporter = im->reporter;
1215 	igmp_stop_timer(im);
1216 
1217 	if (!in_dev->dead) {
1218 		if (IGMP_V1_SEEN(in_dev))
1219 			return;
1220 		if (IGMP_V2_SEEN(in_dev)) {
1221 			if (reporter)
1222 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1223 			return;
1224 		}
1225 		/* IGMPv3 */
1226 		igmpv3_add_delrec(in_dev, im);
1227 
1228 		igmp_ifc_event(in_dev);
1229 	}
1230 #endif
1231 }
1232 
1233 static void igmp_group_added(struct ip_mc_list *im)
1234 {
1235 	struct in_device *in_dev = im->interface;
1236 #ifdef CONFIG_IP_MULTICAST
1237 	struct net *net = dev_net(in_dev->dev);
1238 #endif
1239 
1240 	if (im->loaded == 0) {
1241 		im->loaded = 1;
1242 		ip_mc_filter_add(in_dev, im->multiaddr);
1243 	}
1244 
1245 #ifdef CONFIG_IP_MULTICAST
1246 	if (im->multiaddr == IGMP_ALL_HOSTS)
1247 		return;
1248 	if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1249 		return;
1250 
1251 	if (in_dev->dead)
1252 		return;
1253 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1254 		spin_lock_bh(&im->lock);
1255 		igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1256 		spin_unlock_bh(&im->lock);
1257 		return;
1258 	}
1259 	/* else, v3 */
1260 
1261 	im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1262 	igmp_ifc_event(in_dev);
1263 #endif
1264 }
1265 
1266 
1267 /*
1268  *	Multicast list managers
1269  */
1270 
1271 static u32 ip_mc_hash(const struct ip_mc_list *im)
1272 {
1273 	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1274 }
1275 
1276 static void ip_mc_hash_add(struct in_device *in_dev,
1277 			   struct ip_mc_list *im)
1278 {
1279 	struct ip_mc_list __rcu **mc_hash;
1280 	u32 hash;
1281 
1282 	mc_hash = rtnl_dereference(in_dev->mc_hash);
1283 	if (mc_hash) {
1284 		hash = ip_mc_hash(im);
1285 		im->next_hash = mc_hash[hash];
1286 		rcu_assign_pointer(mc_hash[hash], im);
1287 		return;
1288 	}
1289 
1290 	/* do not use a hash table for small number of items */
1291 	if (in_dev->mc_count < 4)
1292 		return;
1293 
1294 	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1295 			  GFP_KERNEL);
1296 	if (!mc_hash)
1297 		return;
1298 
1299 	for_each_pmc_rtnl(in_dev, im) {
1300 		hash = ip_mc_hash(im);
1301 		im->next_hash = mc_hash[hash];
1302 		RCU_INIT_POINTER(mc_hash[hash], im);
1303 	}
1304 
1305 	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1306 }
1307 
1308 static void ip_mc_hash_remove(struct in_device *in_dev,
1309 			      struct ip_mc_list *im)
1310 {
1311 	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1312 	struct ip_mc_list *aux;
1313 
1314 	if (!mc_hash)
1315 		return;
1316 	mc_hash += ip_mc_hash(im);
1317 	while ((aux = rtnl_dereference(*mc_hash)) != im)
1318 		mc_hash = &aux->next_hash;
1319 	*mc_hash = im->next_hash;
1320 }
1321 
1322 
1323 /*
1324  *	A socket has joined a multicast group on device dev.
1325  */
1326 
1327 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1328 {
1329 	struct ip_mc_list *im;
1330 #ifdef CONFIG_IP_MULTICAST
1331 	struct net *net = dev_net(in_dev->dev);
1332 #endif
1333 
1334 	ASSERT_RTNL();
1335 
1336 	for_each_pmc_rtnl(in_dev, im) {
1337 		if (im->multiaddr == addr) {
1338 			im->users++;
1339 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1340 			goto out;
1341 		}
1342 	}
1343 
1344 	im = kzalloc(sizeof(*im), GFP_KERNEL);
1345 	if (!im)
1346 		goto out;
1347 
1348 	im->users = 1;
1349 	im->interface = in_dev;
1350 	in_dev_hold(in_dev);
1351 	im->multiaddr = addr;
1352 	/* initial mode is (EX, empty) */
1353 	im->sfmode = MCAST_EXCLUDE;
1354 	im->sfcount[MCAST_EXCLUDE] = 1;
1355 	atomic_set(&im->refcnt, 1);
1356 	spin_lock_init(&im->lock);
1357 #ifdef CONFIG_IP_MULTICAST
1358 	setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1359 	im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1360 #endif
1361 
1362 	im->next_rcu = in_dev->mc_list;
1363 	in_dev->mc_count++;
1364 	rcu_assign_pointer(in_dev->mc_list, im);
1365 
1366 	ip_mc_hash_add(in_dev, im);
1367 
1368 #ifdef CONFIG_IP_MULTICAST
1369 	igmpv3_del_delrec(in_dev, im->multiaddr);
1370 #endif
1371 	igmp_group_added(im);
1372 	if (!in_dev->dead)
1373 		ip_rt_multicast_event(in_dev);
1374 out:
1375 	return;
1376 }
1377 EXPORT_SYMBOL(ip_mc_inc_group);
1378 
1379 static int ip_mc_check_iphdr(struct sk_buff *skb)
1380 {
1381 	const struct iphdr *iph;
1382 	unsigned int len;
1383 	unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1384 
1385 	if (!pskb_may_pull(skb, offset))
1386 		return -EINVAL;
1387 
1388 	iph = ip_hdr(skb);
1389 
1390 	if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1391 		return -EINVAL;
1392 
1393 	offset += ip_hdrlen(skb) - sizeof(*iph);
1394 
1395 	if (!pskb_may_pull(skb, offset))
1396 		return -EINVAL;
1397 
1398 	iph = ip_hdr(skb);
1399 
1400 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1401 		return -EINVAL;
1402 
1403 	len = skb_network_offset(skb) + ntohs(iph->tot_len);
1404 	if (skb->len < len || len < offset)
1405 		return -EINVAL;
1406 
1407 	skb_set_transport_header(skb, offset);
1408 
1409 	return 0;
1410 }
1411 
1412 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1413 {
1414 	unsigned int len = skb_transport_offset(skb);
1415 
1416 	len += sizeof(struct igmpv3_report);
1417 
1418 	return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1419 }
1420 
1421 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1422 {
1423 	unsigned int len = skb_transport_offset(skb);
1424 
1425 	len += sizeof(struct igmphdr);
1426 	if (skb->len < len)
1427 		return -EINVAL;
1428 
1429 	/* IGMPv{1,2}? */
1430 	if (skb->len != len) {
1431 		/* or IGMPv3? */
1432 		len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1433 		if (skb->len < len || !pskb_may_pull(skb, len))
1434 			return -EINVAL;
1435 	}
1436 
1437 	/* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1438 	 * all-systems destination addresses (224.0.0.1) for general queries
1439 	 */
1440 	if (!igmp_hdr(skb)->group &&
1441 	    ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1442 		return -EINVAL;
1443 
1444 	return 0;
1445 }
1446 
1447 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1448 {
1449 	switch (igmp_hdr(skb)->type) {
1450 	case IGMP_HOST_LEAVE_MESSAGE:
1451 	case IGMP_HOST_MEMBERSHIP_REPORT:
1452 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1453 		/* fall through */
1454 		return 0;
1455 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1456 		return ip_mc_check_igmp_reportv3(skb);
1457 	case IGMP_HOST_MEMBERSHIP_QUERY:
1458 		return ip_mc_check_igmp_query(skb);
1459 	default:
1460 		return -ENOMSG;
1461 	}
1462 }
1463 
1464 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1465 {
1466 	return skb_checksum_simple_validate(skb);
1467 }
1468 
1469 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1470 
1471 {
1472 	struct sk_buff *skb_chk;
1473 	unsigned int transport_len;
1474 	unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1475 	int ret = -EINVAL;
1476 
1477 	transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1478 
1479 	skb_chk = skb_checksum_trimmed(skb, transport_len,
1480 				       ip_mc_validate_checksum);
1481 	if (!skb_chk)
1482 		goto err;
1483 
1484 	if (!pskb_may_pull(skb_chk, len))
1485 		goto err;
1486 
1487 	ret = ip_mc_check_igmp_msg(skb_chk);
1488 	if (ret)
1489 		goto err;
1490 
1491 	if (skb_trimmed)
1492 		*skb_trimmed = skb_chk;
1493 	/* free now unneeded clone */
1494 	else if (skb_chk != skb)
1495 		kfree_skb(skb_chk);
1496 
1497 	ret = 0;
1498 
1499 err:
1500 	if (ret && skb_chk && skb_chk != skb)
1501 		kfree_skb(skb_chk);
1502 
1503 	return ret;
1504 }
1505 
1506 /**
1507  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1508  * @skb: the skb to validate
1509  * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1510  *
1511  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1512  * skb transport header accordingly and returns zero.
1513  *
1514  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1515  *  standard
1516  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1517  * -ENOMEM: A memory allocation failure happened.
1518  *
1519  * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1520  * to NULL): After parsing an IGMP packet successfully it will point to
1521  * an skb which has its tail aligned to the IP packet end. This might
1522  * either be the originally provided skb or a trimmed, cloned version if
1523  * the skb frame had data beyond the IP packet. A cloned skb allows us
1524  * to leave the original skb and its full frame unchanged (which might be
1525  * desirable for layer 2 frame jugglers).
1526  *
1527  * Caller needs to set the skb network header and free any returned skb if it
1528  * differs from the provided skb.
1529  */
1530 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1531 {
1532 	int ret = ip_mc_check_iphdr(skb);
1533 
1534 	if (ret < 0)
1535 		return ret;
1536 
1537 	if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1538 		return -ENOMSG;
1539 
1540 	return __ip_mc_check_igmp(skb, skb_trimmed);
1541 }
1542 EXPORT_SYMBOL(ip_mc_check_igmp);
1543 
1544 /*
1545  *	Resend IGMP JOIN report; used by netdev notifier.
1546  */
1547 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1548 {
1549 #ifdef CONFIG_IP_MULTICAST
1550 	struct ip_mc_list *im;
1551 	int type;
1552 	struct net *net = dev_net(in_dev->dev);
1553 
1554 	ASSERT_RTNL();
1555 
1556 	for_each_pmc_rtnl(in_dev, im) {
1557 		if (im->multiaddr == IGMP_ALL_HOSTS)
1558 			continue;
1559 		if (ipv4_is_local_multicast(im->multiaddr) &&
1560 		    !net->ipv4.sysctl_igmp_llm_reports)
1561 			continue;
1562 
1563 		/* a failover is happening and switches
1564 		 * must be notified immediately
1565 		 */
1566 		if (IGMP_V1_SEEN(in_dev))
1567 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1568 		else if (IGMP_V2_SEEN(in_dev))
1569 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1570 		else
1571 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1572 		igmp_send_report(in_dev, im, type);
1573 	}
1574 #endif
1575 }
1576 
1577 /*
1578  *	A socket has left a multicast group on device dev
1579  */
1580 
1581 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1582 {
1583 	struct ip_mc_list *i;
1584 	struct ip_mc_list __rcu **ip;
1585 
1586 	ASSERT_RTNL();
1587 
1588 	for (ip = &in_dev->mc_list;
1589 	     (i = rtnl_dereference(*ip)) != NULL;
1590 	     ip = &i->next_rcu) {
1591 		if (i->multiaddr == addr) {
1592 			if (--i->users == 0) {
1593 				ip_mc_hash_remove(in_dev, i);
1594 				*ip = i->next_rcu;
1595 				in_dev->mc_count--;
1596 				igmp_group_dropped(i);
1597 				ip_mc_clear_src(i);
1598 
1599 				if (!in_dev->dead)
1600 					ip_rt_multicast_event(in_dev);
1601 
1602 				ip_ma_put(i);
1603 				return;
1604 			}
1605 			break;
1606 		}
1607 	}
1608 }
1609 EXPORT_SYMBOL(ip_mc_dec_group);
1610 
1611 /* Device changing type */
1612 
1613 void ip_mc_unmap(struct in_device *in_dev)
1614 {
1615 	struct ip_mc_list *pmc;
1616 
1617 	ASSERT_RTNL();
1618 
1619 	for_each_pmc_rtnl(in_dev, pmc)
1620 		igmp_group_dropped(pmc);
1621 }
1622 
1623 void ip_mc_remap(struct in_device *in_dev)
1624 {
1625 	struct ip_mc_list *pmc;
1626 
1627 	ASSERT_RTNL();
1628 
1629 	for_each_pmc_rtnl(in_dev, pmc)
1630 		igmp_group_added(pmc);
1631 }
1632 
1633 /* Device going down */
1634 
1635 void ip_mc_down(struct in_device *in_dev)
1636 {
1637 	struct ip_mc_list *pmc;
1638 
1639 	ASSERT_RTNL();
1640 
1641 	for_each_pmc_rtnl(in_dev, pmc)
1642 		igmp_group_dropped(pmc);
1643 
1644 #ifdef CONFIG_IP_MULTICAST
1645 	in_dev->mr_ifc_count = 0;
1646 	if (del_timer(&in_dev->mr_ifc_timer))
1647 		__in_dev_put(in_dev);
1648 	in_dev->mr_gq_running = 0;
1649 	if (del_timer(&in_dev->mr_gq_timer))
1650 		__in_dev_put(in_dev);
1651 	igmpv3_clear_delrec(in_dev);
1652 #endif
1653 
1654 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1655 }
1656 
1657 void ip_mc_init_dev(struct in_device *in_dev)
1658 {
1659 #ifdef CONFIG_IP_MULTICAST
1660 	struct net *net = dev_net(in_dev->dev);
1661 #endif
1662 	ASSERT_RTNL();
1663 
1664 #ifdef CONFIG_IP_MULTICAST
1665 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1666 			(unsigned long)in_dev);
1667 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1668 			(unsigned long)in_dev);
1669 	in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1670 #endif
1671 
1672 	spin_lock_init(&in_dev->mc_tomb_lock);
1673 }
1674 
1675 /* Device going up */
1676 
1677 void ip_mc_up(struct in_device *in_dev)
1678 {
1679 	struct ip_mc_list *pmc;
1680 #ifdef CONFIG_IP_MULTICAST
1681 	struct net *net = dev_net(in_dev->dev);
1682 #endif
1683 
1684 	ASSERT_RTNL();
1685 
1686 #ifdef CONFIG_IP_MULTICAST
1687 	in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1688 #endif
1689 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1690 
1691 	for_each_pmc_rtnl(in_dev, pmc)
1692 		igmp_group_added(pmc);
1693 }
1694 
1695 /*
1696  *	Device is about to be destroyed: clean up.
1697  */
1698 
1699 void ip_mc_destroy_dev(struct in_device *in_dev)
1700 {
1701 	struct ip_mc_list *i;
1702 
1703 	ASSERT_RTNL();
1704 
1705 	/* Deactivate timers */
1706 	ip_mc_down(in_dev);
1707 
1708 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1709 		in_dev->mc_list = i->next_rcu;
1710 		in_dev->mc_count--;
1711 
1712 		/* We've dropped the groups in ip_mc_down already */
1713 		ip_mc_clear_src(i);
1714 		ip_ma_put(i);
1715 	}
1716 }
1717 
1718 /* RTNL is locked */
1719 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1720 {
1721 	struct net_device *dev = NULL;
1722 	struct in_device *idev = NULL;
1723 
1724 	if (imr->imr_ifindex) {
1725 		idev = inetdev_by_index(net, imr->imr_ifindex);
1726 		return idev;
1727 	}
1728 	if (imr->imr_address.s_addr) {
1729 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1730 		if (!dev)
1731 			return NULL;
1732 	}
1733 
1734 	if (!dev) {
1735 		struct rtable *rt = ip_route_output(net,
1736 						    imr->imr_multiaddr.s_addr,
1737 						    0, 0, 0);
1738 		if (!IS_ERR(rt)) {
1739 			dev = rt->dst.dev;
1740 			ip_rt_put(rt);
1741 		}
1742 	}
1743 	if (dev) {
1744 		imr->imr_ifindex = dev->ifindex;
1745 		idev = __in_dev_get_rtnl(dev);
1746 	}
1747 	return idev;
1748 }
1749 
1750 /*
1751  *	Join a socket to a group
1752  */
1753 
1754 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1755 	__be32 *psfsrc)
1756 {
1757 	struct ip_sf_list *psf, *psf_prev;
1758 	int rv = 0;
1759 
1760 	psf_prev = NULL;
1761 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1762 		if (psf->sf_inaddr == *psfsrc)
1763 			break;
1764 		psf_prev = psf;
1765 	}
1766 	if (!psf || psf->sf_count[sfmode] == 0) {
1767 		/* source filter not found, or count wrong =>  bug */
1768 		return -ESRCH;
1769 	}
1770 	psf->sf_count[sfmode]--;
1771 	if (psf->sf_count[sfmode] == 0) {
1772 		ip_rt_multicast_event(pmc->interface);
1773 	}
1774 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1775 #ifdef CONFIG_IP_MULTICAST
1776 		struct in_device *in_dev = pmc->interface;
1777 		struct net *net = dev_net(in_dev->dev);
1778 #endif
1779 
1780 		/* no more filters for this source */
1781 		if (psf_prev)
1782 			psf_prev->sf_next = psf->sf_next;
1783 		else
1784 			pmc->sources = psf->sf_next;
1785 #ifdef CONFIG_IP_MULTICAST
1786 		if (psf->sf_oldin &&
1787 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1788 			psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1789 			psf->sf_next = pmc->tomb;
1790 			pmc->tomb = psf;
1791 			rv = 1;
1792 		} else
1793 #endif
1794 			kfree(psf);
1795 	}
1796 	return rv;
1797 }
1798 
1799 #ifndef CONFIG_IP_MULTICAST
1800 #define igmp_ifc_event(x)	do { } while (0)
1801 #endif
1802 
1803 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1804 			 int sfcount, __be32 *psfsrc, int delta)
1805 {
1806 	struct ip_mc_list *pmc;
1807 	int	changerec = 0;
1808 	int	i, err;
1809 
1810 	if (!in_dev)
1811 		return -ENODEV;
1812 	rcu_read_lock();
1813 	for_each_pmc_rcu(in_dev, pmc) {
1814 		if (*pmca == pmc->multiaddr)
1815 			break;
1816 	}
1817 	if (!pmc) {
1818 		/* MCA not found?? bug */
1819 		rcu_read_unlock();
1820 		return -ESRCH;
1821 	}
1822 	spin_lock_bh(&pmc->lock);
1823 	rcu_read_unlock();
1824 #ifdef CONFIG_IP_MULTICAST
1825 	sf_markstate(pmc);
1826 #endif
1827 	if (!delta) {
1828 		err = -EINVAL;
1829 		if (!pmc->sfcount[sfmode])
1830 			goto out_unlock;
1831 		pmc->sfcount[sfmode]--;
1832 	}
1833 	err = 0;
1834 	for (i = 0; i < sfcount; i++) {
1835 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1836 
1837 		changerec |= rv > 0;
1838 		if (!err && rv < 0)
1839 			err = rv;
1840 	}
1841 	if (pmc->sfmode == MCAST_EXCLUDE &&
1842 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1843 	    pmc->sfcount[MCAST_INCLUDE]) {
1844 #ifdef CONFIG_IP_MULTICAST
1845 		struct ip_sf_list *psf;
1846 		struct net *net = dev_net(in_dev->dev);
1847 #endif
1848 
1849 		/* filter mode change */
1850 		pmc->sfmode = MCAST_INCLUDE;
1851 #ifdef CONFIG_IP_MULTICAST
1852 		pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1853 		in_dev->mr_ifc_count = pmc->crcount;
1854 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1855 			psf->sf_crcount = 0;
1856 		igmp_ifc_event(pmc->interface);
1857 	} else if (sf_setstate(pmc) || changerec) {
1858 		igmp_ifc_event(pmc->interface);
1859 #endif
1860 	}
1861 out_unlock:
1862 	spin_unlock_bh(&pmc->lock);
1863 	return err;
1864 }
1865 
1866 /*
1867  * Add multicast single-source filter to the interface list
1868  */
1869 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1870 	__be32 *psfsrc)
1871 {
1872 	struct ip_sf_list *psf, *psf_prev;
1873 
1874 	psf_prev = NULL;
1875 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1876 		if (psf->sf_inaddr == *psfsrc)
1877 			break;
1878 		psf_prev = psf;
1879 	}
1880 	if (!psf) {
1881 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1882 		if (!psf)
1883 			return -ENOBUFS;
1884 		psf->sf_inaddr = *psfsrc;
1885 		if (psf_prev) {
1886 			psf_prev->sf_next = psf;
1887 		} else
1888 			pmc->sources = psf;
1889 	}
1890 	psf->sf_count[sfmode]++;
1891 	if (psf->sf_count[sfmode] == 1) {
1892 		ip_rt_multicast_event(pmc->interface);
1893 	}
1894 	return 0;
1895 }
1896 
1897 #ifdef CONFIG_IP_MULTICAST
1898 static void sf_markstate(struct ip_mc_list *pmc)
1899 {
1900 	struct ip_sf_list *psf;
1901 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1902 
1903 	for (psf = pmc->sources; psf; psf = psf->sf_next)
1904 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1905 			psf->sf_oldin = mca_xcount ==
1906 				psf->sf_count[MCAST_EXCLUDE] &&
1907 				!psf->sf_count[MCAST_INCLUDE];
1908 		} else
1909 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1910 }
1911 
1912 static int sf_setstate(struct ip_mc_list *pmc)
1913 {
1914 	struct ip_sf_list *psf, *dpsf;
1915 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1916 	int qrv = pmc->interface->mr_qrv;
1917 	int new_in, rv;
1918 
1919 	rv = 0;
1920 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1921 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1922 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1923 				!psf->sf_count[MCAST_INCLUDE];
1924 		} else
1925 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1926 		if (new_in) {
1927 			if (!psf->sf_oldin) {
1928 				struct ip_sf_list *prev = NULL;
1929 
1930 				for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1931 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1932 						break;
1933 					prev = dpsf;
1934 				}
1935 				if (dpsf) {
1936 					if (prev)
1937 						prev->sf_next = dpsf->sf_next;
1938 					else
1939 						pmc->tomb = dpsf->sf_next;
1940 					kfree(dpsf);
1941 				}
1942 				psf->sf_crcount = qrv;
1943 				rv++;
1944 			}
1945 		} else if (psf->sf_oldin) {
1946 
1947 			psf->sf_crcount = 0;
1948 			/*
1949 			 * add or update "delete" records if an active filter
1950 			 * is now inactive
1951 			 */
1952 			for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1953 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1954 					break;
1955 			if (!dpsf) {
1956 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1957 				if (!dpsf)
1958 					continue;
1959 				*dpsf = *psf;
1960 				/* pmc->lock held by callers */
1961 				dpsf->sf_next = pmc->tomb;
1962 				pmc->tomb = dpsf;
1963 			}
1964 			dpsf->sf_crcount = qrv;
1965 			rv++;
1966 		}
1967 	}
1968 	return rv;
1969 }
1970 #endif
1971 
1972 /*
1973  * Add multicast source filter list to the interface list
1974  */
1975 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1976 			 int sfcount, __be32 *psfsrc, int delta)
1977 {
1978 	struct ip_mc_list *pmc;
1979 	int	isexclude;
1980 	int	i, err;
1981 
1982 	if (!in_dev)
1983 		return -ENODEV;
1984 	rcu_read_lock();
1985 	for_each_pmc_rcu(in_dev, pmc) {
1986 		if (*pmca == pmc->multiaddr)
1987 			break;
1988 	}
1989 	if (!pmc) {
1990 		/* MCA not found?? bug */
1991 		rcu_read_unlock();
1992 		return -ESRCH;
1993 	}
1994 	spin_lock_bh(&pmc->lock);
1995 	rcu_read_unlock();
1996 
1997 #ifdef CONFIG_IP_MULTICAST
1998 	sf_markstate(pmc);
1999 #endif
2000 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
2001 	if (!delta)
2002 		pmc->sfcount[sfmode]++;
2003 	err = 0;
2004 	for (i = 0; i < sfcount; i++) {
2005 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2006 		if (err)
2007 			break;
2008 	}
2009 	if (err) {
2010 		int j;
2011 
2012 		if (!delta)
2013 			pmc->sfcount[sfmode]--;
2014 		for (j = 0; j < i; j++)
2015 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2016 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2017 #ifdef CONFIG_IP_MULTICAST
2018 		struct ip_sf_list *psf;
2019 		struct net *net = dev_net(pmc->interface->dev);
2020 		in_dev = pmc->interface;
2021 #endif
2022 
2023 		/* filter mode change */
2024 		if (pmc->sfcount[MCAST_EXCLUDE])
2025 			pmc->sfmode = MCAST_EXCLUDE;
2026 		else if (pmc->sfcount[MCAST_INCLUDE])
2027 			pmc->sfmode = MCAST_INCLUDE;
2028 #ifdef CONFIG_IP_MULTICAST
2029 		/* else no filters; keep old mode for reports */
2030 
2031 		pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2032 		in_dev->mr_ifc_count = pmc->crcount;
2033 		for (psf = pmc->sources; psf; psf = psf->sf_next)
2034 			psf->sf_crcount = 0;
2035 		igmp_ifc_event(in_dev);
2036 	} else if (sf_setstate(pmc)) {
2037 		igmp_ifc_event(in_dev);
2038 #endif
2039 	}
2040 	spin_unlock_bh(&pmc->lock);
2041 	return err;
2042 }
2043 
2044 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2045 {
2046 	struct ip_sf_list *psf, *nextpsf;
2047 
2048 	for (psf = pmc->tomb; psf; psf = nextpsf) {
2049 		nextpsf = psf->sf_next;
2050 		kfree(psf);
2051 	}
2052 	pmc->tomb = NULL;
2053 	for (psf = pmc->sources; psf; psf = nextpsf) {
2054 		nextpsf = psf->sf_next;
2055 		kfree(psf);
2056 	}
2057 	pmc->sources = NULL;
2058 	pmc->sfmode = MCAST_EXCLUDE;
2059 	pmc->sfcount[MCAST_INCLUDE] = 0;
2060 	pmc->sfcount[MCAST_EXCLUDE] = 1;
2061 }
2062 
2063 /* Join a multicast group
2064  */
2065 
2066 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2067 {
2068 	__be32 addr = imr->imr_multiaddr.s_addr;
2069 	struct ip_mc_socklist *iml, *i;
2070 	struct in_device *in_dev;
2071 	struct inet_sock *inet = inet_sk(sk);
2072 	struct net *net = sock_net(sk);
2073 	int ifindex;
2074 	int count = 0;
2075 	int err;
2076 
2077 	ASSERT_RTNL();
2078 
2079 	if (!ipv4_is_multicast(addr))
2080 		return -EINVAL;
2081 
2082 	in_dev = ip_mc_find_dev(net, imr);
2083 
2084 	if (!in_dev) {
2085 		err = -ENODEV;
2086 		goto done;
2087 	}
2088 
2089 	err = -EADDRINUSE;
2090 	ifindex = imr->imr_ifindex;
2091 	for_each_pmc_rtnl(inet, i) {
2092 		if (i->multi.imr_multiaddr.s_addr == addr &&
2093 		    i->multi.imr_ifindex == ifindex)
2094 			goto done;
2095 		count++;
2096 	}
2097 	err = -ENOBUFS;
2098 	if (count >= net->ipv4.sysctl_igmp_max_memberships)
2099 		goto done;
2100 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2101 	if (!iml)
2102 		goto done;
2103 
2104 	memcpy(&iml->multi, imr, sizeof(*imr));
2105 	iml->next_rcu = inet->mc_list;
2106 	iml->sflist = NULL;
2107 	iml->sfmode = MCAST_EXCLUDE;
2108 	rcu_assign_pointer(inet->mc_list, iml);
2109 	ip_mc_inc_group(in_dev, addr);
2110 	err = 0;
2111 done:
2112 	return err;
2113 }
2114 EXPORT_SYMBOL(ip_mc_join_group);
2115 
2116 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2117 			   struct in_device *in_dev)
2118 {
2119 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2120 	int err;
2121 
2122 	if (!psf) {
2123 		/* any-source empty exclude case */
2124 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2125 			iml->sfmode, 0, NULL, 0);
2126 	}
2127 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2128 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2129 	RCU_INIT_POINTER(iml->sflist, NULL);
2130 	/* decrease mem now to avoid the memleak warning */
2131 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2132 	kfree_rcu(psf, rcu);
2133 	return err;
2134 }
2135 
2136 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2137 {
2138 	struct inet_sock *inet = inet_sk(sk);
2139 	struct ip_mc_socklist *iml;
2140 	struct ip_mc_socklist __rcu **imlp;
2141 	struct in_device *in_dev;
2142 	struct net *net = sock_net(sk);
2143 	__be32 group = imr->imr_multiaddr.s_addr;
2144 	u32 ifindex;
2145 	int ret = -EADDRNOTAVAIL;
2146 
2147 	ASSERT_RTNL();
2148 
2149 	in_dev = ip_mc_find_dev(net, imr);
2150 	if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2151 		ret = -ENODEV;
2152 		goto out;
2153 	}
2154 	ifindex = imr->imr_ifindex;
2155 	for (imlp = &inet->mc_list;
2156 	     (iml = rtnl_dereference(*imlp)) != NULL;
2157 	     imlp = &iml->next_rcu) {
2158 		if (iml->multi.imr_multiaddr.s_addr != group)
2159 			continue;
2160 		if (ifindex) {
2161 			if (iml->multi.imr_ifindex != ifindex)
2162 				continue;
2163 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2164 				iml->multi.imr_address.s_addr)
2165 			continue;
2166 
2167 		(void) ip_mc_leave_src(sk, iml, in_dev);
2168 
2169 		*imlp = iml->next_rcu;
2170 
2171 		if (in_dev)
2172 			ip_mc_dec_group(in_dev, group);
2173 
2174 		/* decrease mem now to avoid the memleak warning */
2175 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2176 		kfree_rcu(iml, rcu);
2177 		return 0;
2178 	}
2179 out:
2180 	return ret;
2181 }
2182 EXPORT_SYMBOL(ip_mc_leave_group);
2183 
2184 int ip_mc_source(int add, int omode, struct sock *sk, struct
2185 	ip_mreq_source *mreqs, int ifindex)
2186 {
2187 	int err;
2188 	struct ip_mreqn imr;
2189 	__be32 addr = mreqs->imr_multiaddr;
2190 	struct ip_mc_socklist *pmc;
2191 	struct in_device *in_dev = NULL;
2192 	struct inet_sock *inet = inet_sk(sk);
2193 	struct ip_sf_socklist *psl;
2194 	struct net *net = sock_net(sk);
2195 	int leavegroup = 0;
2196 	int i, j, rv;
2197 
2198 	if (!ipv4_is_multicast(addr))
2199 		return -EINVAL;
2200 
2201 	ASSERT_RTNL();
2202 
2203 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2204 	imr.imr_address.s_addr = mreqs->imr_interface;
2205 	imr.imr_ifindex = ifindex;
2206 	in_dev = ip_mc_find_dev(net, &imr);
2207 
2208 	if (!in_dev) {
2209 		err = -ENODEV;
2210 		goto done;
2211 	}
2212 	err = -EADDRNOTAVAIL;
2213 
2214 	for_each_pmc_rtnl(inet, pmc) {
2215 		if ((pmc->multi.imr_multiaddr.s_addr ==
2216 		     imr.imr_multiaddr.s_addr) &&
2217 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
2218 			break;
2219 	}
2220 	if (!pmc) {		/* must have a prior join */
2221 		err = -EINVAL;
2222 		goto done;
2223 	}
2224 	/* if a source filter was set, must be the same mode as before */
2225 	if (pmc->sflist) {
2226 		if (pmc->sfmode != omode) {
2227 			err = -EINVAL;
2228 			goto done;
2229 		}
2230 	} else if (pmc->sfmode != omode) {
2231 		/* allow mode switches for empty-set filters */
2232 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2233 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2234 			NULL, 0);
2235 		pmc->sfmode = omode;
2236 	}
2237 
2238 	psl = rtnl_dereference(pmc->sflist);
2239 	if (!add) {
2240 		if (!psl)
2241 			goto done;	/* err = -EADDRNOTAVAIL */
2242 		rv = !0;
2243 		for (i = 0; i < psl->sl_count; i++) {
2244 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2245 				sizeof(__be32));
2246 			if (rv == 0)
2247 				break;
2248 		}
2249 		if (rv)		/* source not found */
2250 			goto done;	/* err = -EADDRNOTAVAIL */
2251 
2252 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2253 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2254 			leavegroup = 1;
2255 			goto done;
2256 		}
2257 
2258 		/* update the interface filter */
2259 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2260 			&mreqs->imr_sourceaddr, 1);
2261 
2262 		for (j = i+1; j < psl->sl_count; j++)
2263 			psl->sl_addr[j-1] = psl->sl_addr[j];
2264 		psl->sl_count--;
2265 		err = 0;
2266 		goto done;
2267 	}
2268 	/* else, add a new source to the filter */
2269 
2270 	if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2271 		err = -ENOBUFS;
2272 		goto done;
2273 	}
2274 	if (!psl || psl->sl_count == psl->sl_max) {
2275 		struct ip_sf_socklist *newpsl;
2276 		int count = IP_SFBLOCK;
2277 
2278 		if (psl)
2279 			count += psl->sl_max;
2280 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2281 		if (!newpsl) {
2282 			err = -ENOBUFS;
2283 			goto done;
2284 		}
2285 		newpsl->sl_max = count;
2286 		newpsl->sl_count = count - IP_SFBLOCK;
2287 		if (psl) {
2288 			for (i = 0; i < psl->sl_count; i++)
2289 				newpsl->sl_addr[i] = psl->sl_addr[i];
2290 			/* decrease mem now to avoid the memleak warning */
2291 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2292 			kfree_rcu(psl, rcu);
2293 		}
2294 		rcu_assign_pointer(pmc->sflist, newpsl);
2295 		psl = newpsl;
2296 	}
2297 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2298 	for (i = 0; i < psl->sl_count; i++) {
2299 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2300 			sizeof(__be32));
2301 		if (rv == 0)
2302 			break;
2303 	}
2304 	if (rv == 0)		/* address already there is an error */
2305 		goto done;
2306 	for (j = psl->sl_count-1; j >= i; j--)
2307 		psl->sl_addr[j+1] = psl->sl_addr[j];
2308 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2309 	psl->sl_count++;
2310 	err = 0;
2311 	/* update the interface list */
2312 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2313 		&mreqs->imr_sourceaddr, 1);
2314 done:
2315 	if (leavegroup)
2316 		err = ip_mc_leave_group(sk, &imr);
2317 	return err;
2318 }
2319 
2320 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2321 {
2322 	int err = 0;
2323 	struct ip_mreqn	imr;
2324 	__be32 addr = msf->imsf_multiaddr;
2325 	struct ip_mc_socklist *pmc;
2326 	struct in_device *in_dev;
2327 	struct inet_sock *inet = inet_sk(sk);
2328 	struct ip_sf_socklist *newpsl, *psl;
2329 	struct net *net = sock_net(sk);
2330 	int leavegroup = 0;
2331 
2332 	if (!ipv4_is_multicast(addr))
2333 		return -EINVAL;
2334 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2335 	    msf->imsf_fmode != MCAST_EXCLUDE)
2336 		return -EINVAL;
2337 
2338 	ASSERT_RTNL();
2339 
2340 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2341 	imr.imr_address.s_addr = msf->imsf_interface;
2342 	imr.imr_ifindex = ifindex;
2343 	in_dev = ip_mc_find_dev(net, &imr);
2344 
2345 	if (!in_dev) {
2346 		err = -ENODEV;
2347 		goto done;
2348 	}
2349 
2350 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2351 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2352 		leavegroup = 1;
2353 		goto done;
2354 	}
2355 
2356 	for_each_pmc_rtnl(inet, pmc) {
2357 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2358 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2359 			break;
2360 	}
2361 	if (!pmc) {		/* must have a prior join */
2362 		err = -EINVAL;
2363 		goto done;
2364 	}
2365 	if (msf->imsf_numsrc) {
2366 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2367 							   GFP_KERNEL);
2368 		if (!newpsl) {
2369 			err = -ENOBUFS;
2370 			goto done;
2371 		}
2372 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2373 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2374 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2375 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2376 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2377 		if (err) {
2378 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2379 			goto done;
2380 		}
2381 	} else {
2382 		newpsl = NULL;
2383 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2384 				     msf->imsf_fmode, 0, NULL, 0);
2385 	}
2386 	psl = rtnl_dereference(pmc->sflist);
2387 	if (psl) {
2388 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2389 			psl->sl_count, psl->sl_addr, 0);
2390 		/* decrease mem now to avoid the memleak warning */
2391 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2392 		kfree_rcu(psl, rcu);
2393 	} else
2394 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2395 			0, NULL, 0);
2396 	rcu_assign_pointer(pmc->sflist, newpsl);
2397 	pmc->sfmode = msf->imsf_fmode;
2398 	err = 0;
2399 done:
2400 	if (leavegroup)
2401 		err = ip_mc_leave_group(sk, &imr);
2402 	return err;
2403 }
2404 
2405 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2406 	struct ip_msfilter __user *optval, int __user *optlen)
2407 {
2408 	int err, len, count, copycount;
2409 	struct ip_mreqn	imr;
2410 	__be32 addr = msf->imsf_multiaddr;
2411 	struct ip_mc_socklist *pmc;
2412 	struct in_device *in_dev;
2413 	struct inet_sock *inet = inet_sk(sk);
2414 	struct ip_sf_socklist *psl;
2415 	struct net *net = sock_net(sk);
2416 
2417 	ASSERT_RTNL();
2418 
2419 	if (!ipv4_is_multicast(addr))
2420 		return -EINVAL;
2421 
2422 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2423 	imr.imr_address.s_addr = msf->imsf_interface;
2424 	imr.imr_ifindex = 0;
2425 	in_dev = ip_mc_find_dev(net, &imr);
2426 
2427 	if (!in_dev) {
2428 		err = -ENODEV;
2429 		goto done;
2430 	}
2431 	err = -EADDRNOTAVAIL;
2432 
2433 	for_each_pmc_rtnl(inet, pmc) {
2434 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2435 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2436 			break;
2437 	}
2438 	if (!pmc)		/* must have a prior join */
2439 		goto done;
2440 	msf->imsf_fmode = pmc->sfmode;
2441 	psl = rtnl_dereference(pmc->sflist);
2442 	if (!psl) {
2443 		len = 0;
2444 		count = 0;
2445 	} else {
2446 		count = psl->sl_count;
2447 	}
2448 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2449 	len = copycount * sizeof(psl->sl_addr[0]);
2450 	msf->imsf_numsrc = count;
2451 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2452 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2453 		return -EFAULT;
2454 	}
2455 	if (len &&
2456 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2457 		return -EFAULT;
2458 	return 0;
2459 done:
2460 	return err;
2461 }
2462 
2463 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2464 	struct group_filter __user *optval, int __user *optlen)
2465 {
2466 	int err, i, count, copycount;
2467 	struct sockaddr_in *psin;
2468 	__be32 addr;
2469 	struct ip_mc_socklist *pmc;
2470 	struct inet_sock *inet = inet_sk(sk);
2471 	struct ip_sf_socklist *psl;
2472 
2473 	ASSERT_RTNL();
2474 
2475 	psin = (struct sockaddr_in *)&gsf->gf_group;
2476 	if (psin->sin_family != AF_INET)
2477 		return -EINVAL;
2478 	addr = psin->sin_addr.s_addr;
2479 	if (!ipv4_is_multicast(addr))
2480 		return -EINVAL;
2481 
2482 	err = -EADDRNOTAVAIL;
2483 
2484 	for_each_pmc_rtnl(inet, pmc) {
2485 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2486 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2487 			break;
2488 	}
2489 	if (!pmc)		/* must have a prior join */
2490 		goto done;
2491 	gsf->gf_fmode = pmc->sfmode;
2492 	psl = rtnl_dereference(pmc->sflist);
2493 	count = psl ? psl->sl_count : 0;
2494 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2495 	gsf->gf_numsrc = count;
2496 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2497 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2498 		return -EFAULT;
2499 	}
2500 	for (i = 0; i < copycount; i++) {
2501 		struct sockaddr_storage ss;
2502 
2503 		psin = (struct sockaddr_in *)&ss;
2504 		memset(&ss, 0, sizeof(ss));
2505 		psin->sin_family = AF_INET;
2506 		psin->sin_addr.s_addr = psl->sl_addr[i];
2507 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2508 			return -EFAULT;
2509 	}
2510 	return 0;
2511 done:
2512 	return err;
2513 }
2514 
2515 /*
2516  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2517  */
2518 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2519 {
2520 	struct inet_sock *inet = inet_sk(sk);
2521 	struct ip_mc_socklist *pmc;
2522 	struct ip_sf_socklist *psl;
2523 	int i;
2524 	int ret;
2525 
2526 	ret = 1;
2527 	if (!ipv4_is_multicast(loc_addr))
2528 		goto out;
2529 
2530 	rcu_read_lock();
2531 	for_each_pmc_rcu(inet, pmc) {
2532 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2533 		    pmc->multi.imr_ifindex == dif)
2534 			break;
2535 	}
2536 	ret = inet->mc_all;
2537 	if (!pmc)
2538 		goto unlock;
2539 	psl = rcu_dereference(pmc->sflist);
2540 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2541 	if (!psl)
2542 		goto unlock;
2543 
2544 	for (i = 0; i < psl->sl_count; i++) {
2545 		if (psl->sl_addr[i] == rmt_addr)
2546 			break;
2547 	}
2548 	ret = 0;
2549 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2550 		goto unlock;
2551 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2552 		goto unlock;
2553 	ret = 1;
2554 unlock:
2555 	rcu_read_unlock();
2556 out:
2557 	return ret;
2558 }
2559 
2560 /*
2561  *	A socket is closing.
2562  */
2563 
2564 void ip_mc_drop_socket(struct sock *sk)
2565 {
2566 	struct inet_sock *inet = inet_sk(sk);
2567 	struct ip_mc_socklist *iml;
2568 	struct net *net = sock_net(sk);
2569 
2570 	if (!inet->mc_list)
2571 		return;
2572 
2573 	rtnl_lock();
2574 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2575 		struct in_device *in_dev;
2576 
2577 		inet->mc_list = iml->next_rcu;
2578 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2579 		(void) ip_mc_leave_src(sk, iml, in_dev);
2580 		if (in_dev)
2581 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2582 		/* decrease mem now to avoid the memleak warning */
2583 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2584 		kfree_rcu(iml, rcu);
2585 	}
2586 	rtnl_unlock();
2587 }
2588 
2589 /* called with rcu_read_lock() */
2590 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2591 {
2592 	struct ip_mc_list *im;
2593 	struct ip_mc_list __rcu **mc_hash;
2594 	struct ip_sf_list *psf;
2595 	int rv = 0;
2596 
2597 	mc_hash = rcu_dereference(in_dev->mc_hash);
2598 	if (mc_hash) {
2599 		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2600 
2601 		for (im = rcu_dereference(mc_hash[hash]);
2602 		     im != NULL;
2603 		     im = rcu_dereference(im->next_hash)) {
2604 			if (im->multiaddr == mc_addr)
2605 				break;
2606 		}
2607 	} else {
2608 		for_each_pmc_rcu(in_dev, im) {
2609 			if (im->multiaddr == mc_addr)
2610 				break;
2611 		}
2612 	}
2613 	if (im && proto == IPPROTO_IGMP) {
2614 		rv = 1;
2615 	} else if (im) {
2616 		if (src_addr) {
2617 			for (psf = im->sources; psf; psf = psf->sf_next) {
2618 				if (psf->sf_inaddr == src_addr)
2619 					break;
2620 			}
2621 			if (psf)
2622 				rv = psf->sf_count[MCAST_INCLUDE] ||
2623 					psf->sf_count[MCAST_EXCLUDE] !=
2624 					im->sfcount[MCAST_EXCLUDE];
2625 			else
2626 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2627 		} else
2628 			rv = 1; /* unspecified source; tentatively allow */
2629 	}
2630 	return rv;
2631 }
2632 
2633 #if defined(CONFIG_PROC_FS)
2634 struct igmp_mc_iter_state {
2635 	struct seq_net_private p;
2636 	struct net_device *dev;
2637 	struct in_device *in_dev;
2638 };
2639 
2640 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2641 
2642 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2643 {
2644 	struct net *net = seq_file_net(seq);
2645 	struct ip_mc_list *im = NULL;
2646 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2647 
2648 	state->in_dev = NULL;
2649 	for_each_netdev_rcu(net, state->dev) {
2650 		struct in_device *in_dev;
2651 
2652 		in_dev = __in_dev_get_rcu(state->dev);
2653 		if (!in_dev)
2654 			continue;
2655 		im = rcu_dereference(in_dev->mc_list);
2656 		if (im) {
2657 			state->in_dev = in_dev;
2658 			break;
2659 		}
2660 	}
2661 	return im;
2662 }
2663 
2664 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2665 {
2666 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2667 
2668 	im = rcu_dereference(im->next_rcu);
2669 	while (!im) {
2670 		state->dev = next_net_device_rcu(state->dev);
2671 		if (!state->dev) {
2672 			state->in_dev = NULL;
2673 			break;
2674 		}
2675 		state->in_dev = __in_dev_get_rcu(state->dev);
2676 		if (!state->in_dev)
2677 			continue;
2678 		im = rcu_dereference(state->in_dev->mc_list);
2679 	}
2680 	return im;
2681 }
2682 
2683 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2684 {
2685 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2686 	if (im)
2687 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2688 			--pos;
2689 	return pos ? NULL : im;
2690 }
2691 
2692 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2693 	__acquires(rcu)
2694 {
2695 	rcu_read_lock();
2696 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2697 }
2698 
2699 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2700 {
2701 	struct ip_mc_list *im;
2702 	if (v == SEQ_START_TOKEN)
2703 		im = igmp_mc_get_first(seq);
2704 	else
2705 		im = igmp_mc_get_next(seq, v);
2706 	++*pos;
2707 	return im;
2708 }
2709 
2710 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2711 	__releases(rcu)
2712 {
2713 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2714 
2715 	state->in_dev = NULL;
2716 	state->dev = NULL;
2717 	rcu_read_unlock();
2718 }
2719 
2720 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2721 {
2722 	if (v == SEQ_START_TOKEN)
2723 		seq_puts(seq,
2724 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2725 	else {
2726 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2727 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2728 		char   *querier;
2729 		long delta;
2730 
2731 #ifdef CONFIG_IP_MULTICAST
2732 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2733 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2734 			  "V3";
2735 #else
2736 		querier = "NONE";
2737 #endif
2738 
2739 		if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2740 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2741 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2742 		}
2743 
2744 		delta = im->timer.expires - jiffies;
2745 		seq_printf(seq,
2746 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2747 			   im->multiaddr, im->users,
2748 			   im->tm_running,
2749 			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2750 			   im->reporter);
2751 	}
2752 	return 0;
2753 }
2754 
2755 static const struct seq_operations igmp_mc_seq_ops = {
2756 	.start	=	igmp_mc_seq_start,
2757 	.next	=	igmp_mc_seq_next,
2758 	.stop	=	igmp_mc_seq_stop,
2759 	.show	=	igmp_mc_seq_show,
2760 };
2761 
2762 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2763 {
2764 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2765 			sizeof(struct igmp_mc_iter_state));
2766 }
2767 
2768 static const struct file_operations igmp_mc_seq_fops = {
2769 	.owner		=	THIS_MODULE,
2770 	.open		=	igmp_mc_seq_open,
2771 	.read		=	seq_read,
2772 	.llseek		=	seq_lseek,
2773 	.release	=	seq_release_net,
2774 };
2775 
2776 struct igmp_mcf_iter_state {
2777 	struct seq_net_private p;
2778 	struct net_device *dev;
2779 	struct in_device *idev;
2780 	struct ip_mc_list *im;
2781 };
2782 
2783 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2784 
2785 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2786 {
2787 	struct net *net = seq_file_net(seq);
2788 	struct ip_sf_list *psf = NULL;
2789 	struct ip_mc_list *im = NULL;
2790 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2791 
2792 	state->idev = NULL;
2793 	state->im = NULL;
2794 	for_each_netdev_rcu(net, state->dev) {
2795 		struct in_device *idev;
2796 		idev = __in_dev_get_rcu(state->dev);
2797 		if (unlikely(!idev))
2798 			continue;
2799 		im = rcu_dereference(idev->mc_list);
2800 		if (likely(im)) {
2801 			spin_lock_bh(&im->lock);
2802 			psf = im->sources;
2803 			if (likely(psf)) {
2804 				state->im = im;
2805 				state->idev = idev;
2806 				break;
2807 			}
2808 			spin_unlock_bh(&im->lock);
2809 		}
2810 	}
2811 	return psf;
2812 }
2813 
2814 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2815 {
2816 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2817 
2818 	psf = psf->sf_next;
2819 	while (!psf) {
2820 		spin_unlock_bh(&state->im->lock);
2821 		state->im = state->im->next;
2822 		while (!state->im) {
2823 			state->dev = next_net_device_rcu(state->dev);
2824 			if (!state->dev) {
2825 				state->idev = NULL;
2826 				goto out;
2827 			}
2828 			state->idev = __in_dev_get_rcu(state->dev);
2829 			if (!state->idev)
2830 				continue;
2831 			state->im = rcu_dereference(state->idev->mc_list);
2832 		}
2833 		if (!state->im)
2834 			break;
2835 		spin_lock_bh(&state->im->lock);
2836 		psf = state->im->sources;
2837 	}
2838 out:
2839 	return psf;
2840 }
2841 
2842 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2843 {
2844 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2845 	if (psf)
2846 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2847 			--pos;
2848 	return pos ? NULL : psf;
2849 }
2850 
2851 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2852 	__acquires(rcu)
2853 {
2854 	rcu_read_lock();
2855 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2856 }
2857 
2858 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2859 {
2860 	struct ip_sf_list *psf;
2861 	if (v == SEQ_START_TOKEN)
2862 		psf = igmp_mcf_get_first(seq);
2863 	else
2864 		psf = igmp_mcf_get_next(seq, v);
2865 	++*pos;
2866 	return psf;
2867 }
2868 
2869 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2870 	__releases(rcu)
2871 {
2872 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2873 	if (likely(state->im)) {
2874 		spin_unlock_bh(&state->im->lock);
2875 		state->im = NULL;
2876 	}
2877 	state->idev = NULL;
2878 	state->dev = NULL;
2879 	rcu_read_unlock();
2880 }
2881 
2882 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2883 {
2884 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2885 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2886 
2887 	if (v == SEQ_START_TOKEN) {
2888 		seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2889 	} else {
2890 		seq_printf(seq,
2891 			   "%3d %6.6s 0x%08x "
2892 			   "0x%08x %6lu %6lu\n",
2893 			   state->dev->ifindex, state->dev->name,
2894 			   ntohl(state->im->multiaddr),
2895 			   ntohl(psf->sf_inaddr),
2896 			   psf->sf_count[MCAST_INCLUDE],
2897 			   psf->sf_count[MCAST_EXCLUDE]);
2898 	}
2899 	return 0;
2900 }
2901 
2902 static const struct seq_operations igmp_mcf_seq_ops = {
2903 	.start	=	igmp_mcf_seq_start,
2904 	.next	=	igmp_mcf_seq_next,
2905 	.stop	=	igmp_mcf_seq_stop,
2906 	.show	=	igmp_mcf_seq_show,
2907 };
2908 
2909 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2910 {
2911 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2912 			sizeof(struct igmp_mcf_iter_state));
2913 }
2914 
2915 static const struct file_operations igmp_mcf_seq_fops = {
2916 	.owner		=	THIS_MODULE,
2917 	.open		=	igmp_mcf_seq_open,
2918 	.read		=	seq_read,
2919 	.llseek		=	seq_lseek,
2920 	.release	=	seq_release_net,
2921 };
2922 
2923 static int __net_init igmp_net_init(struct net *net)
2924 {
2925 	struct proc_dir_entry *pde;
2926 	int err;
2927 
2928 	pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2929 	if (!pde)
2930 		goto out_igmp;
2931 	pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2932 			  &igmp_mcf_seq_fops);
2933 	if (!pde)
2934 		goto out_mcfilter;
2935 	err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2936 				   SOCK_DGRAM, 0, net);
2937 	if (err < 0) {
2938 		pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2939 		       err);
2940 		goto out_sock;
2941 	}
2942 
2943 	/* Sysctl initialization */
2944 	net->ipv4.sysctl_igmp_max_memberships = 20;
2945 	net->ipv4.sysctl_igmp_max_msf = 10;
2946 	/* IGMP reports for link-local multicast groups are enabled by default */
2947 	net->ipv4.sysctl_igmp_llm_reports = 1;
2948 	net->ipv4.sysctl_igmp_qrv = 2;
2949 	return 0;
2950 
2951 out_sock:
2952 	remove_proc_entry("mcfilter", net->proc_net);
2953 out_mcfilter:
2954 	remove_proc_entry("igmp", net->proc_net);
2955 out_igmp:
2956 	return -ENOMEM;
2957 }
2958 
2959 static void __net_exit igmp_net_exit(struct net *net)
2960 {
2961 	remove_proc_entry("mcfilter", net->proc_net);
2962 	remove_proc_entry("igmp", net->proc_net);
2963 	inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2964 }
2965 
2966 static struct pernet_operations igmp_net_ops = {
2967 	.init = igmp_net_init,
2968 	.exit = igmp_net_exit,
2969 };
2970 #endif
2971 
2972 static int igmp_netdev_event(struct notifier_block *this,
2973 			     unsigned long event, void *ptr)
2974 {
2975 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2976 	struct in_device *in_dev;
2977 
2978 	switch (event) {
2979 	case NETDEV_RESEND_IGMP:
2980 		in_dev = __in_dev_get_rtnl(dev);
2981 		if (in_dev)
2982 			ip_mc_rejoin_groups(in_dev);
2983 		break;
2984 	default:
2985 		break;
2986 	}
2987 	return NOTIFY_DONE;
2988 }
2989 
2990 static struct notifier_block igmp_notifier = {
2991 	.notifier_call = igmp_netdev_event,
2992 };
2993 
2994 int __init igmp_mc_init(void)
2995 {
2996 #if defined(CONFIG_PROC_FS)
2997 	int err;
2998 
2999 	err = register_pernet_subsys(&igmp_net_ops);
3000 	if (err)
3001 		return err;
3002 	err = register_netdevice_notifier(&igmp_notifier);
3003 	if (err)
3004 		goto reg_notif_fail;
3005 	return 0;
3006 
3007 reg_notif_fail:
3008 	unregister_pernet_subsys(&igmp_net_ops);
3009 	return err;
3010 #else
3011 	return register_netdevice_notifier(&igmp_notifier);
3012 #endif
3013 }
3014