xref: /openbmc/linux/net/ipv4/igmp.c (revision 425595a7)
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_reserve(skb, hlen);
360 	skb_tailroom_reserve(skb, mtu, tlen);
361 
362 	skb_reset_network_header(skb);
363 	pip = ip_hdr(skb);
364 	skb_put(skb, sizeof(struct iphdr) + 4);
365 
366 	pip->version  = 4;
367 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
368 	pip->tos      = 0xc0;
369 	pip->frag_off = htons(IP_DF);
370 	pip->ttl      = 1;
371 	pip->daddr    = fl4.daddr;
372 	pip->saddr    = fl4.saddr;
373 	pip->protocol = IPPROTO_IGMP;
374 	pip->tot_len  = 0;	/* filled in later */
375 	ip_select_ident(net, skb, NULL);
376 	((u8 *)&pip[1])[0] = IPOPT_RA;
377 	((u8 *)&pip[1])[1] = 4;
378 	((u8 *)&pip[1])[2] = 0;
379 	((u8 *)&pip[1])[3] = 0;
380 
381 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
382 	skb_put(skb, sizeof(*pig));
383 	pig = igmpv3_report_hdr(skb);
384 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
385 	pig->resv1 = 0;
386 	pig->csum = 0;
387 	pig->resv2 = 0;
388 	pig->ngrec = 0;
389 	return skb;
390 }
391 
392 static int igmpv3_sendpack(struct sk_buff *skb)
393 {
394 	struct igmphdr *pig = igmp_hdr(skb);
395 	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
396 
397 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
398 
399 	return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
400 }
401 
402 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
403 {
404 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
405 }
406 
407 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
408 	int type, struct igmpv3_grec **ppgr)
409 {
410 	struct net_device *dev = pmc->interface->dev;
411 	struct igmpv3_report *pih;
412 	struct igmpv3_grec *pgr;
413 
414 	if (!skb)
415 		skb = igmpv3_newpack(dev, dev->mtu);
416 	if (!skb)
417 		return NULL;
418 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
419 	pgr->grec_type = type;
420 	pgr->grec_auxwords = 0;
421 	pgr->grec_nsrcs = 0;
422 	pgr->grec_mca = pmc->multiaddr;
423 	pih = igmpv3_report_hdr(skb);
424 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
425 	*ppgr = pgr;
426 	return skb;
427 }
428 
429 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
430 
431 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
432 	int type, int gdeleted, int sdeleted)
433 {
434 	struct net_device *dev = pmc->interface->dev;
435 	struct igmpv3_report *pih;
436 	struct igmpv3_grec *pgr = NULL;
437 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
438 	int scount, stotal, first, isquery, truncate;
439 
440 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
441 		return skb;
442 	if (ipv4_is_local_multicast(pmc->multiaddr) && !sysctl_igmp_llm_reports)
443 		return skb;
444 
445 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
446 		  type == IGMPV3_MODE_IS_EXCLUDE;
447 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
448 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
449 
450 	stotal = scount = 0;
451 
452 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
453 
454 	if (!*psf_list)
455 		goto empty_source;
456 
457 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
458 
459 	/* EX and TO_EX get a fresh packet, if needed */
460 	if (truncate) {
461 		if (pih && pih->ngrec &&
462 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
463 			if (skb)
464 				igmpv3_sendpack(skb);
465 			skb = igmpv3_newpack(dev, dev->mtu);
466 		}
467 	}
468 	first = 1;
469 	psf_prev = NULL;
470 	for (psf = *psf_list; psf; psf = psf_next) {
471 		__be32 *psrc;
472 
473 		psf_next = psf->sf_next;
474 
475 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
476 			psf_prev = psf;
477 			continue;
478 		}
479 
480 		/* clear marks on query responses */
481 		if (isquery)
482 			psf->sf_gsresp = 0;
483 
484 		if (AVAILABLE(skb) < sizeof(__be32) +
485 		    first*sizeof(struct igmpv3_grec)) {
486 			if (truncate && !first)
487 				break;	 /* truncate these */
488 			if (pgr)
489 				pgr->grec_nsrcs = htons(scount);
490 			if (skb)
491 				igmpv3_sendpack(skb);
492 			skb = igmpv3_newpack(dev, dev->mtu);
493 			first = 1;
494 			scount = 0;
495 		}
496 		if (first) {
497 			skb = add_grhead(skb, pmc, type, &pgr);
498 			first = 0;
499 		}
500 		if (!skb)
501 			return NULL;
502 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
503 		*psrc = psf->sf_inaddr;
504 		scount++; stotal++;
505 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
506 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
507 			psf->sf_crcount--;
508 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
509 				if (psf_prev)
510 					psf_prev->sf_next = psf->sf_next;
511 				else
512 					*psf_list = psf->sf_next;
513 				kfree(psf);
514 				continue;
515 			}
516 		}
517 		psf_prev = psf;
518 	}
519 
520 empty_source:
521 	if (!stotal) {
522 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
523 		    type == IGMPV3_BLOCK_OLD_SOURCES)
524 			return skb;
525 		if (pmc->crcount || isquery) {
526 			/* make sure we have room for group header */
527 			if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
528 				igmpv3_sendpack(skb);
529 				skb = NULL; /* add_grhead will get a new one */
530 			}
531 			skb = add_grhead(skb, pmc, type, &pgr);
532 		}
533 	}
534 	if (pgr)
535 		pgr->grec_nsrcs = htons(scount);
536 
537 	if (isquery)
538 		pmc->gsquery = 0;	/* clear query state on report */
539 	return skb;
540 }
541 
542 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
543 {
544 	struct sk_buff *skb = NULL;
545 	int type;
546 
547 	if (!pmc) {
548 		rcu_read_lock();
549 		for_each_pmc_rcu(in_dev, pmc) {
550 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
551 				continue;
552 			if (ipv4_is_local_multicast(pmc->multiaddr) &&
553 			     !sysctl_igmp_llm_reports)
554 				continue;
555 			spin_lock_bh(&pmc->lock);
556 			if (pmc->sfcount[MCAST_EXCLUDE])
557 				type = IGMPV3_MODE_IS_EXCLUDE;
558 			else
559 				type = IGMPV3_MODE_IS_INCLUDE;
560 			skb = add_grec(skb, pmc, type, 0, 0);
561 			spin_unlock_bh(&pmc->lock);
562 		}
563 		rcu_read_unlock();
564 	} else {
565 		spin_lock_bh(&pmc->lock);
566 		if (pmc->sfcount[MCAST_EXCLUDE])
567 			type = IGMPV3_MODE_IS_EXCLUDE;
568 		else
569 			type = IGMPV3_MODE_IS_INCLUDE;
570 		skb = add_grec(skb, pmc, type, 0, 0);
571 		spin_unlock_bh(&pmc->lock);
572 	}
573 	if (!skb)
574 		return 0;
575 	return igmpv3_sendpack(skb);
576 }
577 
578 /*
579  * remove zero-count source records from a source filter list
580  */
581 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
582 {
583 	struct ip_sf_list *psf_prev, *psf_next, *psf;
584 
585 	psf_prev = NULL;
586 	for (psf = *ppsf; psf; psf = psf_next) {
587 		psf_next = psf->sf_next;
588 		if (psf->sf_crcount == 0) {
589 			if (psf_prev)
590 				psf_prev->sf_next = psf->sf_next;
591 			else
592 				*ppsf = psf->sf_next;
593 			kfree(psf);
594 		} else
595 			psf_prev = psf;
596 	}
597 }
598 
599 static void igmpv3_send_cr(struct in_device *in_dev)
600 {
601 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
602 	struct sk_buff *skb = NULL;
603 	int type, dtype;
604 
605 	rcu_read_lock();
606 	spin_lock_bh(&in_dev->mc_tomb_lock);
607 
608 	/* deleted MCA's */
609 	pmc_prev = NULL;
610 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
611 		pmc_next = pmc->next;
612 		if (pmc->sfmode == MCAST_INCLUDE) {
613 			type = IGMPV3_BLOCK_OLD_SOURCES;
614 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
615 			skb = add_grec(skb, pmc, type, 1, 0);
616 			skb = add_grec(skb, pmc, dtype, 1, 1);
617 		}
618 		if (pmc->crcount) {
619 			if (pmc->sfmode == MCAST_EXCLUDE) {
620 				type = IGMPV3_CHANGE_TO_INCLUDE;
621 				skb = add_grec(skb, pmc, type, 1, 0);
622 			}
623 			pmc->crcount--;
624 			if (pmc->crcount == 0) {
625 				igmpv3_clear_zeros(&pmc->tomb);
626 				igmpv3_clear_zeros(&pmc->sources);
627 			}
628 		}
629 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
630 			if (pmc_prev)
631 				pmc_prev->next = pmc_next;
632 			else
633 				in_dev->mc_tomb = pmc_next;
634 			in_dev_put(pmc->interface);
635 			kfree(pmc);
636 		} else
637 			pmc_prev = pmc;
638 	}
639 	spin_unlock_bh(&in_dev->mc_tomb_lock);
640 
641 	/* change recs */
642 	for_each_pmc_rcu(in_dev, pmc) {
643 		spin_lock_bh(&pmc->lock);
644 		if (pmc->sfcount[MCAST_EXCLUDE]) {
645 			type = IGMPV3_BLOCK_OLD_SOURCES;
646 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
647 		} else {
648 			type = IGMPV3_ALLOW_NEW_SOURCES;
649 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
650 		}
651 		skb = add_grec(skb, pmc, type, 0, 0);
652 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
653 
654 		/* filter mode changes */
655 		if (pmc->crcount) {
656 			if (pmc->sfmode == MCAST_EXCLUDE)
657 				type = IGMPV3_CHANGE_TO_EXCLUDE;
658 			else
659 				type = IGMPV3_CHANGE_TO_INCLUDE;
660 			skb = add_grec(skb, pmc, type, 0, 0);
661 			pmc->crcount--;
662 		}
663 		spin_unlock_bh(&pmc->lock);
664 	}
665 	rcu_read_unlock();
666 
667 	if (!skb)
668 		return;
669 	(void) igmpv3_sendpack(skb);
670 }
671 
672 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
673 	int type)
674 {
675 	struct sk_buff *skb;
676 	struct iphdr *iph;
677 	struct igmphdr *ih;
678 	struct rtable *rt;
679 	struct net_device *dev = in_dev->dev;
680 	struct net *net = dev_net(dev);
681 	__be32	group = pmc ? pmc->multiaddr : 0;
682 	struct flowi4 fl4;
683 	__be32	dst;
684 	int hlen, tlen;
685 
686 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
687 		return igmpv3_send_report(in_dev, pmc);
688 
689 	if (ipv4_is_local_multicast(group) && !sysctl_igmp_llm_reports)
690 		return 0;
691 
692 	if (type == IGMP_HOST_LEAVE_MESSAGE)
693 		dst = IGMP_ALL_ROUTER;
694 	else
695 		dst = group;
696 
697 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
698 				   0, 0,
699 				   IPPROTO_IGMP, 0, dev->ifindex);
700 	if (IS_ERR(rt))
701 		return -1;
702 
703 	hlen = LL_RESERVED_SPACE(dev);
704 	tlen = dev->needed_tailroom;
705 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
706 	if (!skb) {
707 		ip_rt_put(rt);
708 		return -1;
709 	}
710 	skb->priority = TC_PRIO_CONTROL;
711 
712 	skb_dst_set(skb, &rt->dst);
713 
714 	skb_reserve(skb, hlen);
715 
716 	skb_reset_network_header(skb);
717 	iph = ip_hdr(skb);
718 	skb_put(skb, sizeof(struct iphdr) + 4);
719 
720 	iph->version  = 4;
721 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
722 	iph->tos      = 0xc0;
723 	iph->frag_off = htons(IP_DF);
724 	iph->ttl      = 1;
725 	iph->daddr    = dst;
726 	iph->saddr    = fl4.saddr;
727 	iph->protocol = IPPROTO_IGMP;
728 	ip_select_ident(net, skb, NULL);
729 	((u8 *)&iph[1])[0] = IPOPT_RA;
730 	((u8 *)&iph[1])[1] = 4;
731 	((u8 *)&iph[1])[2] = 0;
732 	((u8 *)&iph[1])[3] = 0;
733 
734 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
735 	ih->type = type;
736 	ih->code = 0;
737 	ih->csum = 0;
738 	ih->group = group;
739 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
740 
741 	return ip_local_out(net, skb->sk, skb);
742 }
743 
744 static void igmp_gq_timer_expire(unsigned long data)
745 {
746 	struct in_device *in_dev = (struct in_device *)data;
747 
748 	in_dev->mr_gq_running = 0;
749 	igmpv3_send_report(in_dev, NULL);
750 	in_dev_put(in_dev);
751 }
752 
753 static void igmp_ifc_timer_expire(unsigned long data)
754 {
755 	struct in_device *in_dev = (struct in_device *)data;
756 
757 	igmpv3_send_cr(in_dev);
758 	if (in_dev->mr_ifc_count) {
759 		in_dev->mr_ifc_count--;
760 		igmp_ifc_start_timer(in_dev,
761 				     unsolicited_report_interval(in_dev));
762 	}
763 	in_dev_put(in_dev);
764 }
765 
766 static void igmp_ifc_event(struct in_device *in_dev)
767 {
768 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
769 		return;
770 	in_dev->mr_ifc_count = in_dev->mr_qrv ?: sysctl_igmp_qrv;
771 	igmp_ifc_start_timer(in_dev, 1);
772 }
773 
774 
775 static void igmp_timer_expire(unsigned long data)
776 {
777 	struct ip_mc_list *im = (struct ip_mc_list *)data;
778 	struct in_device *in_dev = im->interface;
779 
780 	spin_lock(&im->lock);
781 	im->tm_running = 0;
782 
783 	if (im->unsolicit_count) {
784 		im->unsolicit_count--;
785 		igmp_start_timer(im, unsolicited_report_interval(in_dev));
786 	}
787 	im->reporter = 1;
788 	spin_unlock(&im->lock);
789 
790 	if (IGMP_V1_SEEN(in_dev))
791 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
792 	else if (IGMP_V2_SEEN(in_dev))
793 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
794 	else
795 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
796 
797 	ip_ma_put(im);
798 }
799 
800 /* mark EXCLUDE-mode sources */
801 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
802 {
803 	struct ip_sf_list *psf;
804 	int i, scount;
805 
806 	scount = 0;
807 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
808 		if (scount == nsrcs)
809 			break;
810 		for (i = 0; i < nsrcs; i++) {
811 			/* skip inactive filters */
812 			if (psf->sf_count[MCAST_INCLUDE] ||
813 			    pmc->sfcount[MCAST_EXCLUDE] !=
814 			    psf->sf_count[MCAST_EXCLUDE])
815 				break;
816 			if (srcs[i] == psf->sf_inaddr) {
817 				scount++;
818 				break;
819 			}
820 		}
821 	}
822 	pmc->gsquery = 0;
823 	if (scount == nsrcs)	/* all sources excluded */
824 		return 0;
825 	return 1;
826 }
827 
828 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
829 {
830 	struct ip_sf_list *psf;
831 	int i, scount;
832 
833 	if (pmc->sfmode == MCAST_EXCLUDE)
834 		return igmp_xmarksources(pmc, nsrcs, srcs);
835 
836 	/* mark INCLUDE-mode sources */
837 	scount = 0;
838 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
839 		if (scount == nsrcs)
840 			break;
841 		for (i = 0; i < nsrcs; i++)
842 			if (srcs[i] == psf->sf_inaddr) {
843 				psf->sf_gsresp = 1;
844 				scount++;
845 				break;
846 			}
847 	}
848 	if (!scount) {
849 		pmc->gsquery = 0;
850 		return 0;
851 	}
852 	pmc->gsquery = 1;
853 	return 1;
854 }
855 
856 /* return true if packet was dropped */
857 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
858 {
859 	struct ip_mc_list *im;
860 
861 	/* Timers are only set for non-local groups */
862 
863 	if (group == IGMP_ALL_HOSTS)
864 		return false;
865 	if (ipv4_is_local_multicast(group) && !sysctl_igmp_llm_reports)
866 		return false;
867 
868 	rcu_read_lock();
869 	for_each_pmc_rcu(in_dev, im) {
870 		if (im->multiaddr == group) {
871 			igmp_stop_timer(im);
872 			break;
873 		}
874 	}
875 	rcu_read_unlock();
876 	return false;
877 }
878 
879 /* return true if packet was dropped */
880 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
881 	int len)
882 {
883 	struct igmphdr 		*ih = igmp_hdr(skb);
884 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
885 	struct ip_mc_list	*im;
886 	__be32			group = ih->group;
887 	int			max_delay;
888 	int			mark = 0;
889 
890 
891 	if (len == 8) {
892 		if (ih->code == 0) {
893 			/* Alas, old v1 router presents here. */
894 
895 			max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
896 			in_dev->mr_v1_seen = jiffies +
897 				IGMP_V1_ROUTER_PRESENT_TIMEOUT;
898 			group = 0;
899 		} else {
900 			/* v2 router present */
901 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
902 			in_dev->mr_v2_seen = jiffies +
903 				IGMP_V2_ROUTER_PRESENT_TIMEOUT;
904 		}
905 		/* cancel the interface change timer */
906 		in_dev->mr_ifc_count = 0;
907 		if (del_timer(&in_dev->mr_ifc_timer))
908 			__in_dev_put(in_dev);
909 		/* clear deleted report items */
910 		igmpv3_clear_delrec(in_dev);
911 	} else if (len < 12) {
912 		return true;	/* ignore bogus packet; freed by caller */
913 	} else if (IGMP_V1_SEEN(in_dev)) {
914 		/* This is a v3 query with v1 queriers present */
915 		max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
916 		group = 0;
917 	} else if (IGMP_V2_SEEN(in_dev)) {
918 		/* this is a v3 query with v2 queriers present;
919 		 * Interpretation of the max_delay code is problematic here.
920 		 * A real v2 host would use ih_code directly, while v3 has a
921 		 * different encoding. We use the v3 encoding as more likely
922 		 * to be intended in a v3 query.
923 		 */
924 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
925 		if (!max_delay)
926 			max_delay = 1;	/* can't mod w/ 0 */
927 	} else { /* v3 */
928 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
929 			return true;
930 
931 		ih3 = igmpv3_query_hdr(skb);
932 		if (ih3->nsrcs) {
933 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
934 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
935 				return true;
936 			ih3 = igmpv3_query_hdr(skb);
937 		}
938 
939 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
940 		if (!max_delay)
941 			max_delay = 1;	/* can't mod w/ 0 */
942 		in_dev->mr_maxdelay = max_delay;
943 		if (ih3->qrv)
944 			in_dev->mr_qrv = ih3->qrv;
945 		if (!group) { /* general query */
946 			if (ih3->nsrcs)
947 				return true;	/* no sources allowed */
948 			igmp_gq_start_timer(in_dev);
949 			return false;
950 		}
951 		/* mark sources to include, if group & source-specific */
952 		mark = ih3->nsrcs != 0;
953 	}
954 
955 	/*
956 	 * - Start the timers in all of our membership records
957 	 *   that the query applies to for the interface on
958 	 *   which the query arrived excl. those that belong
959 	 *   to a "local" group (224.0.0.X)
960 	 * - For timers already running check if they need to
961 	 *   be reset.
962 	 * - Use the igmp->igmp_code field as the maximum
963 	 *   delay possible
964 	 */
965 	rcu_read_lock();
966 	for_each_pmc_rcu(in_dev, im) {
967 		int changed;
968 
969 		if (group && group != im->multiaddr)
970 			continue;
971 		if (im->multiaddr == IGMP_ALL_HOSTS)
972 			continue;
973 		if (ipv4_is_local_multicast(im->multiaddr) &&
974 		    !sysctl_igmp_llm_reports)
975 			continue;
976 		spin_lock_bh(&im->lock);
977 		if (im->tm_running)
978 			im->gsquery = im->gsquery && mark;
979 		else
980 			im->gsquery = mark;
981 		changed = !im->gsquery ||
982 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
983 		spin_unlock_bh(&im->lock);
984 		if (changed)
985 			igmp_mod_timer(im, max_delay);
986 	}
987 	rcu_read_unlock();
988 	return false;
989 }
990 
991 /* called in rcu_read_lock() section */
992 int igmp_rcv(struct sk_buff *skb)
993 {
994 	/* This basically follows the spec line by line -- see RFC1112 */
995 	struct igmphdr *ih;
996 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
997 	int len = skb->len;
998 	bool dropped = true;
999 
1000 	if (!in_dev)
1001 		goto drop;
1002 
1003 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1004 		goto drop;
1005 
1006 	if (skb_checksum_simple_validate(skb))
1007 		goto drop;
1008 
1009 	ih = igmp_hdr(skb);
1010 	switch (ih->type) {
1011 	case IGMP_HOST_MEMBERSHIP_QUERY:
1012 		dropped = igmp_heard_query(in_dev, skb, len);
1013 		break;
1014 	case IGMP_HOST_MEMBERSHIP_REPORT:
1015 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1016 		/* Is it our report looped back? */
1017 		if (rt_is_output_route(skb_rtable(skb)))
1018 			break;
1019 		/* don't rely on MC router hearing unicast reports */
1020 		if (skb->pkt_type == PACKET_MULTICAST ||
1021 		    skb->pkt_type == PACKET_BROADCAST)
1022 			dropped = igmp_heard_report(in_dev, ih->group);
1023 		break;
1024 	case IGMP_PIM:
1025 #ifdef CONFIG_IP_PIMSM_V1
1026 		return pim_rcv_v1(skb);
1027 #endif
1028 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1029 	case IGMP_DVMRP:
1030 	case IGMP_TRACE:
1031 	case IGMP_HOST_LEAVE_MESSAGE:
1032 	case IGMP_MTRACE:
1033 	case IGMP_MTRACE_RESP:
1034 		break;
1035 	default:
1036 		break;
1037 	}
1038 
1039 drop:
1040 	if (dropped)
1041 		kfree_skb(skb);
1042 	else
1043 		consume_skb(skb);
1044 	return 0;
1045 }
1046 
1047 #endif
1048 
1049 
1050 /*
1051  *	Add a filter to a device
1052  */
1053 
1054 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1055 {
1056 	char buf[MAX_ADDR_LEN];
1057 	struct net_device *dev = in_dev->dev;
1058 
1059 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1060 	   We will get multicast token leakage, when IFF_MULTICAST
1061 	   is changed. This check should be done in ndo_set_rx_mode
1062 	   routine. Something sort of:
1063 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1064 	   --ANK
1065 	   */
1066 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1067 		dev_mc_add(dev, buf);
1068 }
1069 
1070 /*
1071  *	Remove a filter from a device
1072  */
1073 
1074 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1075 {
1076 	char buf[MAX_ADDR_LEN];
1077 	struct net_device *dev = in_dev->dev;
1078 
1079 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1080 		dev_mc_del(dev, buf);
1081 }
1082 
1083 #ifdef CONFIG_IP_MULTICAST
1084 /*
1085  * deleted ip_mc_list manipulation
1086  */
1087 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1088 {
1089 	struct ip_mc_list *pmc;
1090 
1091 	/* this is an "ip_mc_list" for convenience; only the fields below
1092 	 * are actually used. In particular, the refcnt and users are not
1093 	 * used for management of the delete list. Using the same structure
1094 	 * for deleted items allows change reports to use common code with
1095 	 * non-deleted or query-response MCA's.
1096 	 */
1097 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1098 	if (!pmc)
1099 		return;
1100 	spin_lock_bh(&im->lock);
1101 	pmc->interface = im->interface;
1102 	in_dev_hold(in_dev);
1103 	pmc->multiaddr = im->multiaddr;
1104 	pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1105 	pmc->sfmode = im->sfmode;
1106 	if (pmc->sfmode == MCAST_INCLUDE) {
1107 		struct ip_sf_list *psf;
1108 
1109 		pmc->tomb = im->tomb;
1110 		pmc->sources = im->sources;
1111 		im->tomb = im->sources = NULL;
1112 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1113 			psf->sf_crcount = pmc->crcount;
1114 	}
1115 	spin_unlock_bh(&im->lock);
1116 
1117 	spin_lock_bh(&in_dev->mc_tomb_lock);
1118 	pmc->next = in_dev->mc_tomb;
1119 	in_dev->mc_tomb = pmc;
1120 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1121 }
1122 
1123 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1124 {
1125 	struct ip_mc_list *pmc, *pmc_prev;
1126 	struct ip_sf_list *psf, *psf_next;
1127 
1128 	spin_lock_bh(&in_dev->mc_tomb_lock);
1129 	pmc_prev = NULL;
1130 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1131 		if (pmc->multiaddr == multiaddr)
1132 			break;
1133 		pmc_prev = pmc;
1134 	}
1135 	if (pmc) {
1136 		if (pmc_prev)
1137 			pmc_prev->next = pmc->next;
1138 		else
1139 			in_dev->mc_tomb = pmc->next;
1140 	}
1141 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1142 	if (pmc) {
1143 		for (psf = pmc->tomb; psf; psf = psf_next) {
1144 			psf_next = psf->sf_next;
1145 			kfree(psf);
1146 		}
1147 		in_dev_put(pmc->interface);
1148 		kfree(pmc);
1149 	}
1150 }
1151 
1152 static void igmpv3_clear_delrec(struct in_device *in_dev)
1153 {
1154 	struct ip_mc_list *pmc, *nextpmc;
1155 
1156 	spin_lock_bh(&in_dev->mc_tomb_lock);
1157 	pmc = in_dev->mc_tomb;
1158 	in_dev->mc_tomb = NULL;
1159 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1160 
1161 	for (; pmc; pmc = nextpmc) {
1162 		nextpmc = pmc->next;
1163 		ip_mc_clear_src(pmc);
1164 		in_dev_put(pmc->interface);
1165 		kfree(pmc);
1166 	}
1167 	/* clear dead sources, too */
1168 	rcu_read_lock();
1169 	for_each_pmc_rcu(in_dev, pmc) {
1170 		struct ip_sf_list *psf, *psf_next;
1171 
1172 		spin_lock_bh(&pmc->lock);
1173 		psf = pmc->tomb;
1174 		pmc->tomb = NULL;
1175 		spin_unlock_bh(&pmc->lock);
1176 		for (; psf; psf = psf_next) {
1177 			psf_next = psf->sf_next;
1178 			kfree(psf);
1179 		}
1180 	}
1181 	rcu_read_unlock();
1182 }
1183 #endif
1184 
1185 static void igmp_group_dropped(struct ip_mc_list *im)
1186 {
1187 	struct in_device *in_dev = im->interface;
1188 #ifdef CONFIG_IP_MULTICAST
1189 	int reporter;
1190 #endif
1191 
1192 	if (im->loaded) {
1193 		im->loaded = 0;
1194 		ip_mc_filter_del(in_dev, im->multiaddr);
1195 	}
1196 
1197 #ifdef CONFIG_IP_MULTICAST
1198 	if (im->multiaddr == IGMP_ALL_HOSTS)
1199 		return;
1200 	if (ipv4_is_local_multicast(im->multiaddr) && !sysctl_igmp_llm_reports)
1201 		return;
1202 
1203 	reporter = im->reporter;
1204 	igmp_stop_timer(im);
1205 
1206 	if (!in_dev->dead) {
1207 		if (IGMP_V1_SEEN(in_dev))
1208 			return;
1209 		if (IGMP_V2_SEEN(in_dev)) {
1210 			if (reporter)
1211 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1212 			return;
1213 		}
1214 		/* IGMPv3 */
1215 		igmpv3_add_delrec(in_dev, im);
1216 
1217 		igmp_ifc_event(in_dev);
1218 	}
1219 #endif
1220 }
1221 
1222 static void igmp_group_added(struct ip_mc_list *im)
1223 {
1224 	struct in_device *in_dev = im->interface;
1225 
1226 	if (im->loaded == 0) {
1227 		im->loaded = 1;
1228 		ip_mc_filter_add(in_dev, im->multiaddr);
1229 	}
1230 
1231 #ifdef CONFIG_IP_MULTICAST
1232 	if (im->multiaddr == IGMP_ALL_HOSTS)
1233 		return;
1234 	if (ipv4_is_local_multicast(im->multiaddr) && !sysctl_igmp_llm_reports)
1235 		return;
1236 
1237 	if (in_dev->dead)
1238 		return;
1239 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1240 		spin_lock_bh(&im->lock);
1241 		igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1242 		spin_unlock_bh(&im->lock);
1243 		return;
1244 	}
1245 	/* else, v3 */
1246 
1247 	im->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1248 	igmp_ifc_event(in_dev);
1249 #endif
1250 }
1251 
1252 
1253 /*
1254  *	Multicast list managers
1255  */
1256 
1257 static u32 ip_mc_hash(const struct ip_mc_list *im)
1258 {
1259 	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1260 }
1261 
1262 static void ip_mc_hash_add(struct in_device *in_dev,
1263 			   struct ip_mc_list *im)
1264 {
1265 	struct ip_mc_list __rcu **mc_hash;
1266 	u32 hash;
1267 
1268 	mc_hash = rtnl_dereference(in_dev->mc_hash);
1269 	if (mc_hash) {
1270 		hash = ip_mc_hash(im);
1271 		im->next_hash = mc_hash[hash];
1272 		rcu_assign_pointer(mc_hash[hash], im);
1273 		return;
1274 	}
1275 
1276 	/* do not use a hash table for small number of items */
1277 	if (in_dev->mc_count < 4)
1278 		return;
1279 
1280 	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1281 			  GFP_KERNEL);
1282 	if (!mc_hash)
1283 		return;
1284 
1285 	for_each_pmc_rtnl(in_dev, im) {
1286 		hash = ip_mc_hash(im);
1287 		im->next_hash = mc_hash[hash];
1288 		RCU_INIT_POINTER(mc_hash[hash], im);
1289 	}
1290 
1291 	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1292 }
1293 
1294 static void ip_mc_hash_remove(struct in_device *in_dev,
1295 			      struct ip_mc_list *im)
1296 {
1297 	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1298 	struct ip_mc_list *aux;
1299 
1300 	if (!mc_hash)
1301 		return;
1302 	mc_hash += ip_mc_hash(im);
1303 	while ((aux = rtnl_dereference(*mc_hash)) != im)
1304 		mc_hash = &aux->next_hash;
1305 	*mc_hash = im->next_hash;
1306 }
1307 
1308 
1309 /*
1310  *	A socket has joined a multicast group on device dev.
1311  */
1312 
1313 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1314 {
1315 	struct ip_mc_list *im;
1316 
1317 	ASSERT_RTNL();
1318 
1319 	for_each_pmc_rtnl(in_dev, im) {
1320 		if (im->multiaddr == addr) {
1321 			im->users++;
1322 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1323 			goto out;
1324 		}
1325 	}
1326 
1327 	im = kzalloc(sizeof(*im), GFP_KERNEL);
1328 	if (!im)
1329 		goto out;
1330 
1331 	im->users = 1;
1332 	im->interface = in_dev;
1333 	in_dev_hold(in_dev);
1334 	im->multiaddr = addr;
1335 	/* initial mode is (EX, empty) */
1336 	im->sfmode = MCAST_EXCLUDE;
1337 	im->sfcount[MCAST_EXCLUDE] = 1;
1338 	atomic_set(&im->refcnt, 1);
1339 	spin_lock_init(&im->lock);
1340 #ifdef CONFIG_IP_MULTICAST
1341 	setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1342 	im->unsolicit_count = sysctl_igmp_qrv;
1343 #endif
1344 
1345 	im->next_rcu = in_dev->mc_list;
1346 	in_dev->mc_count++;
1347 	rcu_assign_pointer(in_dev->mc_list, im);
1348 
1349 	ip_mc_hash_add(in_dev, im);
1350 
1351 #ifdef CONFIG_IP_MULTICAST
1352 	igmpv3_del_delrec(in_dev, im->multiaddr);
1353 #endif
1354 	igmp_group_added(im);
1355 	if (!in_dev->dead)
1356 		ip_rt_multicast_event(in_dev);
1357 out:
1358 	return;
1359 }
1360 EXPORT_SYMBOL(ip_mc_inc_group);
1361 
1362 static int ip_mc_check_iphdr(struct sk_buff *skb)
1363 {
1364 	const struct iphdr *iph;
1365 	unsigned int len;
1366 	unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1367 
1368 	if (!pskb_may_pull(skb, offset))
1369 		return -EINVAL;
1370 
1371 	iph = ip_hdr(skb);
1372 
1373 	if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1374 		return -EINVAL;
1375 
1376 	offset += ip_hdrlen(skb) - sizeof(*iph);
1377 
1378 	if (!pskb_may_pull(skb, offset))
1379 		return -EINVAL;
1380 
1381 	iph = ip_hdr(skb);
1382 
1383 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1384 		return -EINVAL;
1385 
1386 	len = skb_network_offset(skb) + ntohs(iph->tot_len);
1387 	if (skb->len < len || len < offset)
1388 		return -EINVAL;
1389 
1390 	skb_set_transport_header(skb, offset);
1391 
1392 	return 0;
1393 }
1394 
1395 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1396 {
1397 	unsigned int len = skb_transport_offset(skb);
1398 
1399 	len += sizeof(struct igmpv3_report);
1400 
1401 	return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1402 }
1403 
1404 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1405 {
1406 	unsigned int len = skb_transport_offset(skb);
1407 
1408 	len += sizeof(struct igmphdr);
1409 	if (skb->len < len)
1410 		return -EINVAL;
1411 
1412 	/* IGMPv{1,2}? */
1413 	if (skb->len != len) {
1414 		/* or IGMPv3? */
1415 		len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1416 		if (skb->len < len || !pskb_may_pull(skb, len))
1417 			return -EINVAL;
1418 	}
1419 
1420 	/* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1421 	 * all-systems destination addresses (224.0.0.1) for general queries
1422 	 */
1423 	if (!igmp_hdr(skb)->group &&
1424 	    ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1425 		return -EINVAL;
1426 
1427 	return 0;
1428 }
1429 
1430 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1431 {
1432 	switch (igmp_hdr(skb)->type) {
1433 	case IGMP_HOST_LEAVE_MESSAGE:
1434 	case IGMP_HOST_MEMBERSHIP_REPORT:
1435 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1436 		/* fall through */
1437 		return 0;
1438 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1439 		return ip_mc_check_igmp_reportv3(skb);
1440 	case IGMP_HOST_MEMBERSHIP_QUERY:
1441 		return ip_mc_check_igmp_query(skb);
1442 	default:
1443 		return -ENOMSG;
1444 	}
1445 }
1446 
1447 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1448 {
1449 	return skb_checksum_simple_validate(skb);
1450 }
1451 
1452 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1453 
1454 {
1455 	struct sk_buff *skb_chk;
1456 	unsigned int transport_len;
1457 	unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1458 	int ret = -EINVAL;
1459 
1460 	transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1461 
1462 	skb_chk = skb_checksum_trimmed(skb, transport_len,
1463 				       ip_mc_validate_checksum);
1464 	if (!skb_chk)
1465 		goto err;
1466 
1467 	if (!pskb_may_pull(skb_chk, len))
1468 		goto err;
1469 
1470 	ret = ip_mc_check_igmp_msg(skb_chk);
1471 	if (ret)
1472 		goto err;
1473 
1474 	if (skb_trimmed)
1475 		*skb_trimmed = skb_chk;
1476 	/* free now unneeded clone */
1477 	else if (skb_chk != skb)
1478 		kfree_skb(skb_chk);
1479 
1480 	ret = 0;
1481 
1482 err:
1483 	if (ret && skb_chk && skb_chk != skb)
1484 		kfree_skb(skb_chk);
1485 
1486 	return ret;
1487 }
1488 
1489 /**
1490  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1491  * @skb: the skb to validate
1492  * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1493  *
1494  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1495  * skb transport header accordingly and returns zero.
1496  *
1497  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1498  *  standard
1499  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1500  * -ENOMEM: A memory allocation failure happened.
1501  *
1502  * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1503  * to NULL): After parsing an IGMP packet successfully it will point to
1504  * an skb which has its tail aligned to the IP packet end. This might
1505  * either be the originally provided skb or a trimmed, cloned version if
1506  * the skb frame had data beyond the IP packet. A cloned skb allows us
1507  * to leave the original skb and its full frame unchanged (which might be
1508  * desirable for layer 2 frame jugglers).
1509  *
1510  * Caller needs to set the skb network header and free any returned skb if it
1511  * differs from the provided skb.
1512  */
1513 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1514 {
1515 	int ret = ip_mc_check_iphdr(skb);
1516 
1517 	if (ret < 0)
1518 		return ret;
1519 
1520 	if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1521 		return -ENOMSG;
1522 
1523 	return __ip_mc_check_igmp(skb, skb_trimmed);
1524 }
1525 EXPORT_SYMBOL(ip_mc_check_igmp);
1526 
1527 /*
1528  *	Resend IGMP JOIN report; used by netdev notifier.
1529  */
1530 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1531 {
1532 #ifdef CONFIG_IP_MULTICAST
1533 	struct ip_mc_list *im;
1534 	int type;
1535 
1536 	ASSERT_RTNL();
1537 
1538 	for_each_pmc_rtnl(in_dev, im) {
1539 		if (im->multiaddr == IGMP_ALL_HOSTS)
1540 			continue;
1541 		if (ipv4_is_local_multicast(im->multiaddr) &&
1542 		    !sysctl_igmp_llm_reports)
1543 			continue;
1544 
1545 		/* a failover is happening and switches
1546 		 * must be notified immediately
1547 		 */
1548 		if (IGMP_V1_SEEN(in_dev))
1549 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1550 		else if (IGMP_V2_SEEN(in_dev))
1551 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1552 		else
1553 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1554 		igmp_send_report(in_dev, im, type);
1555 	}
1556 #endif
1557 }
1558 
1559 /*
1560  *	A socket has left a multicast group on device dev
1561  */
1562 
1563 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1564 {
1565 	struct ip_mc_list *i;
1566 	struct ip_mc_list __rcu **ip;
1567 
1568 	ASSERT_RTNL();
1569 
1570 	for (ip = &in_dev->mc_list;
1571 	     (i = rtnl_dereference(*ip)) != NULL;
1572 	     ip = &i->next_rcu) {
1573 		if (i->multiaddr == addr) {
1574 			if (--i->users == 0) {
1575 				ip_mc_hash_remove(in_dev, i);
1576 				*ip = i->next_rcu;
1577 				in_dev->mc_count--;
1578 				igmp_group_dropped(i);
1579 				ip_mc_clear_src(i);
1580 
1581 				if (!in_dev->dead)
1582 					ip_rt_multicast_event(in_dev);
1583 
1584 				ip_ma_put(i);
1585 				return;
1586 			}
1587 			break;
1588 		}
1589 	}
1590 }
1591 EXPORT_SYMBOL(ip_mc_dec_group);
1592 
1593 /* Device changing type */
1594 
1595 void ip_mc_unmap(struct in_device *in_dev)
1596 {
1597 	struct ip_mc_list *pmc;
1598 
1599 	ASSERT_RTNL();
1600 
1601 	for_each_pmc_rtnl(in_dev, pmc)
1602 		igmp_group_dropped(pmc);
1603 }
1604 
1605 void ip_mc_remap(struct in_device *in_dev)
1606 {
1607 	struct ip_mc_list *pmc;
1608 
1609 	ASSERT_RTNL();
1610 
1611 	for_each_pmc_rtnl(in_dev, pmc)
1612 		igmp_group_added(pmc);
1613 }
1614 
1615 /* Device going down */
1616 
1617 void ip_mc_down(struct in_device *in_dev)
1618 {
1619 	struct ip_mc_list *pmc;
1620 
1621 	ASSERT_RTNL();
1622 
1623 	for_each_pmc_rtnl(in_dev, pmc)
1624 		igmp_group_dropped(pmc);
1625 
1626 #ifdef CONFIG_IP_MULTICAST
1627 	in_dev->mr_ifc_count = 0;
1628 	if (del_timer(&in_dev->mr_ifc_timer))
1629 		__in_dev_put(in_dev);
1630 	in_dev->mr_gq_running = 0;
1631 	if (del_timer(&in_dev->mr_gq_timer))
1632 		__in_dev_put(in_dev);
1633 	igmpv3_clear_delrec(in_dev);
1634 #endif
1635 
1636 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1637 }
1638 
1639 void ip_mc_init_dev(struct in_device *in_dev)
1640 {
1641 	ASSERT_RTNL();
1642 
1643 #ifdef CONFIG_IP_MULTICAST
1644 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1645 			(unsigned long)in_dev);
1646 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1647 			(unsigned long)in_dev);
1648 	in_dev->mr_qrv = sysctl_igmp_qrv;
1649 #endif
1650 
1651 	spin_lock_init(&in_dev->mc_tomb_lock);
1652 }
1653 
1654 /* Device going up */
1655 
1656 void ip_mc_up(struct in_device *in_dev)
1657 {
1658 	struct ip_mc_list *pmc;
1659 
1660 	ASSERT_RTNL();
1661 
1662 #ifdef CONFIG_IP_MULTICAST
1663 	in_dev->mr_qrv = sysctl_igmp_qrv;
1664 #endif
1665 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1666 
1667 	for_each_pmc_rtnl(in_dev, pmc)
1668 		igmp_group_added(pmc);
1669 }
1670 
1671 /*
1672  *	Device is about to be destroyed: clean up.
1673  */
1674 
1675 void ip_mc_destroy_dev(struct in_device *in_dev)
1676 {
1677 	struct ip_mc_list *i;
1678 
1679 	ASSERT_RTNL();
1680 
1681 	/* Deactivate timers */
1682 	ip_mc_down(in_dev);
1683 
1684 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1685 		in_dev->mc_list = i->next_rcu;
1686 		in_dev->mc_count--;
1687 
1688 		/* We've dropped the groups in ip_mc_down already */
1689 		ip_mc_clear_src(i);
1690 		ip_ma_put(i);
1691 	}
1692 }
1693 
1694 /* RTNL is locked */
1695 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1696 {
1697 	struct net_device *dev = NULL;
1698 	struct in_device *idev = NULL;
1699 
1700 	if (imr->imr_ifindex) {
1701 		idev = inetdev_by_index(net, imr->imr_ifindex);
1702 		return idev;
1703 	}
1704 	if (imr->imr_address.s_addr) {
1705 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1706 		if (!dev)
1707 			return NULL;
1708 	}
1709 
1710 	if (!dev) {
1711 		struct rtable *rt = ip_route_output(net,
1712 						    imr->imr_multiaddr.s_addr,
1713 						    0, 0, 0);
1714 		if (!IS_ERR(rt)) {
1715 			dev = rt->dst.dev;
1716 			ip_rt_put(rt);
1717 		}
1718 	}
1719 	if (dev) {
1720 		imr->imr_ifindex = dev->ifindex;
1721 		idev = __in_dev_get_rtnl(dev);
1722 	}
1723 	return idev;
1724 }
1725 
1726 /*
1727  *	Join a socket to a group
1728  */
1729 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1730 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1731 #ifdef CONFIG_IP_MULTICAST
1732 int sysctl_igmp_qrv __read_mostly = IGMP_QUERY_ROBUSTNESS_VARIABLE;
1733 #endif
1734 
1735 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1736 	__be32 *psfsrc)
1737 {
1738 	struct ip_sf_list *psf, *psf_prev;
1739 	int rv = 0;
1740 
1741 	psf_prev = NULL;
1742 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1743 		if (psf->sf_inaddr == *psfsrc)
1744 			break;
1745 		psf_prev = psf;
1746 	}
1747 	if (!psf || psf->sf_count[sfmode] == 0) {
1748 		/* source filter not found, or count wrong =>  bug */
1749 		return -ESRCH;
1750 	}
1751 	psf->sf_count[sfmode]--;
1752 	if (psf->sf_count[sfmode] == 0) {
1753 		ip_rt_multicast_event(pmc->interface);
1754 	}
1755 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1756 #ifdef CONFIG_IP_MULTICAST
1757 		struct in_device *in_dev = pmc->interface;
1758 #endif
1759 
1760 		/* no more filters for this source */
1761 		if (psf_prev)
1762 			psf_prev->sf_next = psf->sf_next;
1763 		else
1764 			pmc->sources = psf->sf_next;
1765 #ifdef CONFIG_IP_MULTICAST
1766 		if (psf->sf_oldin &&
1767 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1768 			psf->sf_crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1769 			psf->sf_next = pmc->tomb;
1770 			pmc->tomb = psf;
1771 			rv = 1;
1772 		} else
1773 #endif
1774 			kfree(psf);
1775 	}
1776 	return rv;
1777 }
1778 
1779 #ifndef CONFIG_IP_MULTICAST
1780 #define igmp_ifc_event(x)	do { } while (0)
1781 #endif
1782 
1783 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1784 			 int sfcount, __be32 *psfsrc, int delta)
1785 {
1786 	struct ip_mc_list *pmc;
1787 	int	changerec = 0;
1788 	int	i, err;
1789 
1790 	if (!in_dev)
1791 		return -ENODEV;
1792 	rcu_read_lock();
1793 	for_each_pmc_rcu(in_dev, pmc) {
1794 		if (*pmca == pmc->multiaddr)
1795 			break;
1796 	}
1797 	if (!pmc) {
1798 		/* MCA not found?? bug */
1799 		rcu_read_unlock();
1800 		return -ESRCH;
1801 	}
1802 	spin_lock_bh(&pmc->lock);
1803 	rcu_read_unlock();
1804 #ifdef CONFIG_IP_MULTICAST
1805 	sf_markstate(pmc);
1806 #endif
1807 	if (!delta) {
1808 		err = -EINVAL;
1809 		if (!pmc->sfcount[sfmode])
1810 			goto out_unlock;
1811 		pmc->sfcount[sfmode]--;
1812 	}
1813 	err = 0;
1814 	for (i = 0; i < sfcount; i++) {
1815 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1816 
1817 		changerec |= rv > 0;
1818 		if (!err && rv < 0)
1819 			err = rv;
1820 	}
1821 	if (pmc->sfmode == MCAST_EXCLUDE &&
1822 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1823 	    pmc->sfcount[MCAST_INCLUDE]) {
1824 #ifdef CONFIG_IP_MULTICAST
1825 		struct ip_sf_list *psf;
1826 #endif
1827 
1828 		/* filter mode change */
1829 		pmc->sfmode = MCAST_INCLUDE;
1830 #ifdef CONFIG_IP_MULTICAST
1831 		pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1832 		in_dev->mr_ifc_count = pmc->crcount;
1833 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1834 			psf->sf_crcount = 0;
1835 		igmp_ifc_event(pmc->interface);
1836 	} else if (sf_setstate(pmc) || changerec) {
1837 		igmp_ifc_event(pmc->interface);
1838 #endif
1839 	}
1840 out_unlock:
1841 	spin_unlock_bh(&pmc->lock);
1842 	return err;
1843 }
1844 
1845 /*
1846  * Add multicast single-source filter to the interface list
1847  */
1848 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1849 	__be32 *psfsrc)
1850 {
1851 	struct ip_sf_list *psf, *psf_prev;
1852 
1853 	psf_prev = NULL;
1854 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1855 		if (psf->sf_inaddr == *psfsrc)
1856 			break;
1857 		psf_prev = psf;
1858 	}
1859 	if (!psf) {
1860 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1861 		if (!psf)
1862 			return -ENOBUFS;
1863 		psf->sf_inaddr = *psfsrc;
1864 		if (psf_prev) {
1865 			psf_prev->sf_next = psf;
1866 		} else
1867 			pmc->sources = psf;
1868 	}
1869 	psf->sf_count[sfmode]++;
1870 	if (psf->sf_count[sfmode] == 1) {
1871 		ip_rt_multicast_event(pmc->interface);
1872 	}
1873 	return 0;
1874 }
1875 
1876 #ifdef CONFIG_IP_MULTICAST
1877 static void sf_markstate(struct ip_mc_list *pmc)
1878 {
1879 	struct ip_sf_list *psf;
1880 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1881 
1882 	for (psf = pmc->sources; psf; psf = psf->sf_next)
1883 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1884 			psf->sf_oldin = mca_xcount ==
1885 				psf->sf_count[MCAST_EXCLUDE] &&
1886 				!psf->sf_count[MCAST_INCLUDE];
1887 		} else
1888 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1889 }
1890 
1891 static int sf_setstate(struct ip_mc_list *pmc)
1892 {
1893 	struct ip_sf_list *psf, *dpsf;
1894 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1895 	int qrv = pmc->interface->mr_qrv;
1896 	int new_in, rv;
1897 
1898 	rv = 0;
1899 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1900 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1901 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1902 				!psf->sf_count[MCAST_INCLUDE];
1903 		} else
1904 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1905 		if (new_in) {
1906 			if (!psf->sf_oldin) {
1907 				struct ip_sf_list *prev = NULL;
1908 
1909 				for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1910 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1911 						break;
1912 					prev = dpsf;
1913 				}
1914 				if (dpsf) {
1915 					if (prev)
1916 						prev->sf_next = dpsf->sf_next;
1917 					else
1918 						pmc->tomb = dpsf->sf_next;
1919 					kfree(dpsf);
1920 				}
1921 				psf->sf_crcount = qrv;
1922 				rv++;
1923 			}
1924 		} else if (psf->sf_oldin) {
1925 
1926 			psf->sf_crcount = 0;
1927 			/*
1928 			 * add or update "delete" records if an active filter
1929 			 * is now inactive
1930 			 */
1931 			for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1932 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1933 					break;
1934 			if (!dpsf) {
1935 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1936 				if (!dpsf)
1937 					continue;
1938 				*dpsf = *psf;
1939 				/* pmc->lock held by callers */
1940 				dpsf->sf_next = pmc->tomb;
1941 				pmc->tomb = dpsf;
1942 			}
1943 			dpsf->sf_crcount = qrv;
1944 			rv++;
1945 		}
1946 	}
1947 	return rv;
1948 }
1949 #endif
1950 
1951 /*
1952  * Add multicast source filter list to the interface list
1953  */
1954 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1955 			 int sfcount, __be32 *psfsrc, int delta)
1956 {
1957 	struct ip_mc_list *pmc;
1958 	int	isexclude;
1959 	int	i, err;
1960 
1961 	if (!in_dev)
1962 		return -ENODEV;
1963 	rcu_read_lock();
1964 	for_each_pmc_rcu(in_dev, pmc) {
1965 		if (*pmca == pmc->multiaddr)
1966 			break;
1967 	}
1968 	if (!pmc) {
1969 		/* MCA not found?? bug */
1970 		rcu_read_unlock();
1971 		return -ESRCH;
1972 	}
1973 	spin_lock_bh(&pmc->lock);
1974 	rcu_read_unlock();
1975 
1976 #ifdef CONFIG_IP_MULTICAST
1977 	sf_markstate(pmc);
1978 #endif
1979 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
1980 	if (!delta)
1981 		pmc->sfcount[sfmode]++;
1982 	err = 0;
1983 	for (i = 0; i < sfcount; i++) {
1984 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1985 		if (err)
1986 			break;
1987 	}
1988 	if (err) {
1989 		int j;
1990 
1991 		if (!delta)
1992 			pmc->sfcount[sfmode]--;
1993 		for (j = 0; j < i; j++)
1994 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1995 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1996 #ifdef CONFIG_IP_MULTICAST
1997 		struct ip_sf_list *psf;
1998 		in_dev = pmc->interface;
1999 #endif
2000 
2001 		/* filter mode change */
2002 		if (pmc->sfcount[MCAST_EXCLUDE])
2003 			pmc->sfmode = MCAST_EXCLUDE;
2004 		else if (pmc->sfcount[MCAST_INCLUDE])
2005 			pmc->sfmode = MCAST_INCLUDE;
2006 #ifdef CONFIG_IP_MULTICAST
2007 		/* else no filters; keep old mode for reports */
2008 
2009 		pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
2010 		in_dev->mr_ifc_count = pmc->crcount;
2011 		for (psf = pmc->sources; psf; psf = psf->sf_next)
2012 			psf->sf_crcount = 0;
2013 		igmp_ifc_event(in_dev);
2014 	} else if (sf_setstate(pmc)) {
2015 		igmp_ifc_event(in_dev);
2016 #endif
2017 	}
2018 	spin_unlock_bh(&pmc->lock);
2019 	return err;
2020 }
2021 
2022 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2023 {
2024 	struct ip_sf_list *psf, *nextpsf;
2025 
2026 	for (psf = pmc->tomb; psf; psf = nextpsf) {
2027 		nextpsf = psf->sf_next;
2028 		kfree(psf);
2029 	}
2030 	pmc->tomb = NULL;
2031 	for (psf = pmc->sources; psf; psf = nextpsf) {
2032 		nextpsf = psf->sf_next;
2033 		kfree(psf);
2034 	}
2035 	pmc->sources = NULL;
2036 	pmc->sfmode = MCAST_EXCLUDE;
2037 	pmc->sfcount[MCAST_INCLUDE] = 0;
2038 	pmc->sfcount[MCAST_EXCLUDE] = 1;
2039 }
2040 
2041 /* Join a multicast group
2042  */
2043 
2044 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2045 {
2046 	__be32 addr = imr->imr_multiaddr.s_addr;
2047 	struct ip_mc_socklist *iml, *i;
2048 	struct in_device *in_dev;
2049 	struct inet_sock *inet = inet_sk(sk);
2050 	struct net *net = sock_net(sk);
2051 	int ifindex;
2052 	int count = 0;
2053 	int err;
2054 
2055 	ASSERT_RTNL();
2056 
2057 	if (!ipv4_is_multicast(addr))
2058 		return -EINVAL;
2059 
2060 	in_dev = ip_mc_find_dev(net, imr);
2061 
2062 	if (!in_dev) {
2063 		err = -ENODEV;
2064 		goto done;
2065 	}
2066 
2067 	err = -EADDRINUSE;
2068 	ifindex = imr->imr_ifindex;
2069 	for_each_pmc_rtnl(inet, i) {
2070 		if (i->multi.imr_multiaddr.s_addr == addr &&
2071 		    i->multi.imr_ifindex == ifindex)
2072 			goto done;
2073 		count++;
2074 	}
2075 	err = -ENOBUFS;
2076 	if (count >= sysctl_igmp_max_memberships)
2077 		goto done;
2078 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2079 	if (!iml)
2080 		goto done;
2081 
2082 	memcpy(&iml->multi, imr, sizeof(*imr));
2083 	iml->next_rcu = inet->mc_list;
2084 	iml->sflist = NULL;
2085 	iml->sfmode = MCAST_EXCLUDE;
2086 	rcu_assign_pointer(inet->mc_list, iml);
2087 	ip_mc_inc_group(in_dev, addr);
2088 	err = 0;
2089 done:
2090 	return err;
2091 }
2092 EXPORT_SYMBOL(ip_mc_join_group);
2093 
2094 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2095 			   struct in_device *in_dev)
2096 {
2097 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2098 	int err;
2099 
2100 	if (!psf) {
2101 		/* any-source empty exclude case */
2102 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2103 			iml->sfmode, 0, NULL, 0);
2104 	}
2105 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2106 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2107 	RCU_INIT_POINTER(iml->sflist, NULL);
2108 	/* decrease mem now to avoid the memleak warning */
2109 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2110 	kfree_rcu(psf, rcu);
2111 	return err;
2112 }
2113 
2114 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2115 {
2116 	struct inet_sock *inet = inet_sk(sk);
2117 	struct ip_mc_socklist *iml;
2118 	struct ip_mc_socklist __rcu **imlp;
2119 	struct in_device *in_dev;
2120 	struct net *net = sock_net(sk);
2121 	__be32 group = imr->imr_multiaddr.s_addr;
2122 	u32 ifindex;
2123 	int ret = -EADDRNOTAVAIL;
2124 
2125 	ASSERT_RTNL();
2126 
2127 	in_dev = ip_mc_find_dev(net, imr);
2128 	if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2129 		ret = -ENODEV;
2130 		goto out;
2131 	}
2132 	ifindex = imr->imr_ifindex;
2133 	for (imlp = &inet->mc_list;
2134 	     (iml = rtnl_dereference(*imlp)) != NULL;
2135 	     imlp = &iml->next_rcu) {
2136 		if (iml->multi.imr_multiaddr.s_addr != group)
2137 			continue;
2138 		if (ifindex) {
2139 			if (iml->multi.imr_ifindex != ifindex)
2140 				continue;
2141 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2142 				iml->multi.imr_address.s_addr)
2143 			continue;
2144 
2145 		(void) ip_mc_leave_src(sk, iml, in_dev);
2146 
2147 		*imlp = iml->next_rcu;
2148 
2149 		if (in_dev)
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 	ASSERT_RTNL();
2396 
2397 	if (!ipv4_is_multicast(addr))
2398 		return -EINVAL;
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 	if (!psl) {
2421 		len = 0;
2422 		count = 0;
2423 	} else {
2424 		count = psl->sl_count;
2425 	}
2426 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2427 	len = copycount * sizeof(psl->sl_addr[0]);
2428 	msf->imsf_numsrc = count;
2429 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2430 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2431 		return -EFAULT;
2432 	}
2433 	if (len &&
2434 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2435 		return -EFAULT;
2436 	return 0;
2437 done:
2438 	return err;
2439 }
2440 
2441 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2442 	struct group_filter __user *optval, int __user *optlen)
2443 {
2444 	int err, i, count, copycount;
2445 	struct sockaddr_in *psin;
2446 	__be32 addr;
2447 	struct ip_mc_socklist *pmc;
2448 	struct inet_sock *inet = inet_sk(sk);
2449 	struct ip_sf_socklist *psl;
2450 
2451 	ASSERT_RTNL();
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 	err = -EADDRNOTAVAIL;
2461 
2462 	for_each_pmc_rtnl(inet, pmc) {
2463 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2464 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2465 			break;
2466 	}
2467 	if (!pmc)		/* must have a prior join */
2468 		goto done;
2469 	gsf->gf_fmode = pmc->sfmode;
2470 	psl = rtnl_dereference(pmc->sflist);
2471 	count = psl ? psl->sl_count : 0;
2472 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2473 	gsf->gf_numsrc = count;
2474 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2475 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2476 		return -EFAULT;
2477 	}
2478 	for (i = 0; i < copycount; i++) {
2479 		struct sockaddr_storage ss;
2480 
2481 		psin = (struct sockaddr_in *)&ss;
2482 		memset(&ss, 0, sizeof(ss));
2483 		psin->sin_family = AF_INET;
2484 		psin->sin_addr.s_addr = psl->sl_addr[i];
2485 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2486 			return -EFAULT;
2487 	}
2488 	return 0;
2489 done:
2490 	return err;
2491 }
2492 
2493 /*
2494  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2495  */
2496 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2497 {
2498 	struct inet_sock *inet = inet_sk(sk);
2499 	struct ip_mc_socklist *pmc;
2500 	struct ip_sf_socklist *psl;
2501 	int i;
2502 	int ret;
2503 
2504 	ret = 1;
2505 	if (!ipv4_is_multicast(loc_addr))
2506 		goto out;
2507 
2508 	rcu_read_lock();
2509 	for_each_pmc_rcu(inet, pmc) {
2510 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2511 		    pmc->multi.imr_ifindex == dif)
2512 			break;
2513 	}
2514 	ret = inet->mc_all;
2515 	if (!pmc)
2516 		goto unlock;
2517 	psl = rcu_dereference(pmc->sflist);
2518 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2519 	if (!psl)
2520 		goto unlock;
2521 
2522 	for (i = 0; i < psl->sl_count; i++) {
2523 		if (psl->sl_addr[i] == rmt_addr)
2524 			break;
2525 	}
2526 	ret = 0;
2527 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2528 		goto unlock;
2529 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2530 		goto unlock;
2531 	ret = 1;
2532 unlock:
2533 	rcu_read_unlock();
2534 out:
2535 	return ret;
2536 }
2537 
2538 /*
2539  *	A socket is closing.
2540  */
2541 
2542 void ip_mc_drop_socket(struct sock *sk)
2543 {
2544 	struct inet_sock *inet = inet_sk(sk);
2545 	struct ip_mc_socklist *iml;
2546 	struct net *net = sock_net(sk);
2547 
2548 	if (!inet->mc_list)
2549 		return;
2550 
2551 	rtnl_lock();
2552 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2553 		struct in_device *in_dev;
2554 
2555 		inet->mc_list = iml->next_rcu;
2556 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2557 		(void) ip_mc_leave_src(sk, iml, in_dev);
2558 		if (in_dev)
2559 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2560 		/* decrease mem now to avoid the memleak warning */
2561 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2562 		kfree_rcu(iml, rcu);
2563 	}
2564 	rtnl_unlock();
2565 }
2566 
2567 /* called with rcu_read_lock() */
2568 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2569 {
2570 	struct ip_mc_list *im;
2571 	struct ip_mc_list __rcu **mc_hash;
2572 	struct ip_sf_list *psf;
2573 	int rv = 0;
2574 
2575 	mc_hash = rcu_dereference(in_dev->mc_hash);
2576 	if (mc_hash) {
2577 		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2578 
2579 		for (im = rcu_dereference(mc_hash[hash]);
2580 		     im != NULL;
2581 		     im = rcu_dereference(im->next_hash)) {
2582 			if (im->multiaddr == mc_addr)
2583 				break;
2584 		}
2585 	} else {
2586 		for_each_pmc_rcu(in_dev, im) {
2587 			if (im->multiaddr == mc_addr)
2588 				break;
2589 		}
2590 	}
2591 	if (im && proto == IPPROTO_IGMP) {
2592 		rv = 1;
2593 	} else if (im) {
2594 		if (src_addr) {
2595 			for (psf = im->sources; psf; psf = psf->sf_next) {
2596 				if (psf->sf_inaddr == src_addr)
2597 					break;
2598 			}
2599 			if (psf)
2600 				rv = psf->sf_count[MCAST_INCLUDE] ||
2601 					psf->sf_count[MCAST_EXCLUDE] !=
2602 					im->sfcount[MCAST_EXCLUDE];
2603 			else
2604 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2605 		} else
2606 			rv = 1; /* unspecified source; tentatively allow */
2607 	}
2608 	return rv;
2609 }
2610 
2611 #if defined(CONFIG_PROC_FS)
2612 struct igmp_mc_iter_state {
2613 	struct seq_net_private p;
2614 	struct net_device *dev;
2615 	struct in_device *in_dev;
2616 };
2617 
2618 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2619 
2620 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2621 {
2622 	struct net *net = seq_file_net(seq);
2623 	struct ip_mc_list *im = NULL;
2624 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2625 
2626 	state->in_dev = NULL;
2627 	for_each_netdev_rcu(net, state->dev) {
2628 		struct in_device *in_dev;
2629 
2630 		in_dev = __in_dev_get_rcu(state->dev);
2631 		if (!in_dev)
2632 			continue;
2633 		im = rcu_dereference(in_dev->mc_list);
2634 		if (im) {
2635 			state->in_dev = in_dev;
2636 			break;
2637 		}
2638 	}
2639 	return im;
2640 }
2641 
2642 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2643 {
2644 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2645 
2646 	im = rcu_dereference(im->next_rcu);
2647 	while (!im) {
2648 		state->dev = next_net_device_rcu(state->dev);
2649 		if (!state->dev) {
2650 			state->in_dev = NULL;
2651 			break;
2652 		}
2653 		state->in_dev = __in_dev_get_rcu(state->dev);
2654 		if (!state->in_dev)
2655 			continue;
2656 		im = rcu_dereference(state->in_dev->mc_list);
2657 	}
2658 	return im;
2659 }
2660 
2661 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2662 {
2663 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2664 	if (im)
2665 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2666 			--pos;
2667 	return pos ? NULL : im;
2668 }
2669 
2670 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2671 	__acquires(rcu)
2672 {
2673 	rcu_read_lock();
2674 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2675 }
2676 
2677 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2678 {
2679 	struct ip_mc_list *im;
2680 	if (v == SEQ_START_TOKEN)
2681 		im = igmp_mc_get_first(seq);
2682 	else
2683 		im = igmp_mc_get_next(seq, v);
2684 	++*pos;
2685 	return im;
2686 }
2687 
2688 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2689 	__releases(rcu)
2690 {
2691 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2692 
2693 	state->in_dev = NULL;
2694 	state->dev = NULL;
2695 	rcu_read_unlock();
2696 }
2697 
2698 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2699 {
2700 	if (v == SEQ_START_TOKEN)
2701 		seq_puts(seq,
2702 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2703 	else {
2704 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2705 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2706 		char   *querier;
2707 		long delta;
2708 
2709 #ifdef CONFIG_IP_MULTICAST
2710 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2711 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2712 			  "V3";
2713 #else
2714 		querier = "NONE";
2715 #endif
2716 
2717 		if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2718 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2719 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2720 		}
2721 
2722 		delta = im->timer.expires - jiffies;
2723 		seq_printf(seq,
2724 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2725 			   im->multiaddr, im->users,
2726 			   im->tm_running,
2727 			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2728 			   im->reporter);
2729 	}
2730 	return 0;
2731 }
2732 
2733 static const struct seq_operations igmp_mc_seq_ops = {
2734 	.start	=	igmp_mc_seq_start,
2735 	.next	=	igmp_mc_seq_next,
2736 	.stop	=	igmp_mc_seq_stop,
2737 	.show	=	igmp_mc_seq_show,
2738 };
2739 
2740 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2741 {
2742 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2743 			sizeof(struct igmp_mc_iter_state));
2744 }
2745 
2746 static const struct file_operations igmp_mc_seq_fops = {
2747 	.owner		=	THIS_MODULE,
2748 	.open		=	igmp_mc_seq_open,
2749 	.read		=	seq_read,
2750 	.llseek		=	seq_lseek,
2751 	.release	=	seq_release_net,
2752 };
2753 
2754 struct igmp_mcf_iter_state {
2755 	struct seq_net_private p;
2756 	struct net_device *dev;
2757 	struct in_device *idev;
2758 	struct ip_mc_list *im;
2759 };
2760 
2761 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2762 
2763 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2764 {
2765 	struct net *net = seq_file_net(seq);
2766 	struct ip_sf_list *psf = NULL;
2767 	struct ip_mc_list *im = NULL;
2768 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2769 
2770 	state->idev = NULL;
2771 	state->im = NULL;
2772 	for_each_netdev_rcu(net, state->dev) {
2773 		struct in_device *idev;
2774 		idev = __in_dev_get_rcu(state->dev);
2775 		if (unlikely(!idev))
2776 			continue;
2777 		im = rcu_dereference(idev->mc_list);
2778 		if (likely(im)) {
2779 			spin_lock_bh(&im->lock);
2780 			psf = im->sources;
2781 			if (likely(psf)) {
2782 				state->im = im;
2783 				state->idev = idev;
2784 				break;
2785 			}
2786 			spin_unlock_bh(&im->lock);
2787 		}
2788 	}
2789 	return psf;
2790 }
2791 
2792 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2793 {
2794 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2795 
2796 	psf = psf->sf_next;
2797 	while (!psf) {
2798 		spin_unlock_bh(&state->im->lock);
2799 		state->im = state->im->next;
2800 		while (!state->im) {
2801 			state->dev = next_net_device_rcu(state->dev);
2802 			if (!state->dev) {
2803 				state->idev = NULL;
2804 				goto out;
2805 			}
2806 			state->idev = __in_dev_get_rcu(state->dev);
2807 			if (!state->idev)
2808 				continue;
2809 			state->im = rcu_dereference(state->idev->mc_list);
2810 		}
2811 		if (!state->im)
2812 			break;
2813 		spin_lock_bh(&state->im->lock);
2814 		psf = state->im->sources;
2815 	}
2816 out:
2817 	return psf;
2818 }
2819 
2820 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2821 {
2822 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2823 	if (psf)
2824 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2825 			--pos;
2826 	return pos ? NULL : psf;
2827 }
2828 
2829 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2830 	__acquires(rcu)
2831 {
2832 	rcu_read_lock();
2833 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2834 }
2835 
2836 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2837 {
2838 	struct ip_sf_list *psf;
2839 	if (v == SEQ_START_TOKEN)
2840 		psf = igmp_mcf_get_first(seq);
2841 	else
2842 		psf = igmp_mcf_get_next(seq, v);
2843 	++*pos;
2844 	return psf;
2845 }
2846 
2847 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2848 	__releases(rcu)
2849 {
2850 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2851 	if (likely(state->im)) {
2852 		spin_unlock_bh(&state->im->lock);
2853 		state->im = NULL;
2854 	}
2855 	state->idev = NULL;
2856 	state->dev = NULL;
2857 	rcu_read_unlock();
2858 }
2859 
2860 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2861 {
2862 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2863 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2864 
2865 	if (v == SEQ_START_TOKEN) {
2866 		seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2867 	} else {
2868 		seq_printf(seq,
2869 			   "%3d %6.6s 0x%08x "
2870 			   "0x%08x %6lu %6lu\n",
2871 			   state->dev->ifindex, state->dev->name,
2872 			   ntohl(state->im->multiaddr),
2873 			   ntohl(psf->sf_inaddr),
2874 			   psf->sf_count[MCAST_INCLUDE],
2875 			   psf->sf_count[MCAST_EXCLUDE]);
2876 	}
2877 	return 0;
2878 }
2879 
2880 static const struct seq_operations igmp_mcf_seq_ops = {
2881 	.start	=	igmp_mcf_seq_start,
2882 	.next	=	igmp_mcf_seq_next,
2883 	.stop	=	igmp_mcf_seq_stop,
2884 	.show	=	igmp_mcf_seq_show,
2885 };
2886 
2887 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2888 {
2889 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2890 			sizeof(struct igmp_mcf_iter_state));
2891 }
2892 
2893 static const struct file_operations igmp_mcf_seq_fops = {
2894 	.owner		=	THIS_MODULE,
2895 	.open		=	igmp_mcf_seq_open,
2896 	.read		=	seq_read,
2897 	.llseek		=	seq_lseek,
2898 	.release	=	seq_release_net,
2899 };
2900 
2901 static int __net_init igmp_net_init(struct net *net)
2902 {
2903 	struct proc_dir_entry *pde;
2904 	int err;
2905 
2906 	pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2907 	if (!pde)
2908 		goto out_igmp;
2909 	pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2910 			  &igmp_mcf_seq_fops);
2911 	if (!pde)
2912 		goto out_mcfilter;
2913 	err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2914 				   SOCK_DGRAM, 0, net);
2915 	if (err < 0) {
2916 		pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2917 		       err);
2918 		goto out_sock;
2919 	}
2920 
2921 	return 0;
2922 
2923 out_sock:
2924 	remove_proc_entry("mcfilter", net->proc_net);
2925 out_mcfilter:
2926 	remove_proc_entry("igmp", net->proc_net);
2927 out_igmp:
2928 	return -ENOMEM;
2929 }
2930 
2931 static void __net_exit igmp_net_exit(struct net *net)
2932 {
2933 	remove_proc_entry("mcfilter", net->proc_net);
2934 	remove_proc_entry("igmp", net->proc_net);
2935 	inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2936 }
2937 
2938 static struct pernet_operations igmp_net_ops = {
2939 	.init = igmp_net_init,
2940 	.exit = igmp_net_exit,
2941 };
2942 #endif
2943 
2944 static int igmp_netdev_event(struct notifier_block *this,
2945 			     unsigned long event, void *ptr)
2946 {
2947 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2948 	struct in_device *in_dev;
2949 
2950 	switch (event) {
2951 	case NETDEV_RESEND_IGMP:
2952 		in_dev = __in_dev_get_rtnl(dev);
2953 		if (in_dev)
2954 			ip_mc_rejoin_groups(in_dev);
2955 		break;
2956 	default:
2957 		break;
2958 	}
2959 	return NOTIFY_DONE;
2960 }
2961 
2962 static struct notifier_block igmp_notifier = {
2963 	.notifier_call = igmp_netdev_event,
2964 };
2965 
2966 int __init igmp_mc_init(void)
2967 {
2968 #if defined(CONFIG_PROC_FS)
2969 	int err;
2970 
2971 	err = register_pernet_subsys(&igmp_net_ops);
2972 	if (err)
2973 		return err;
2974 	err = register_netdevice_notifier(&igmp_notifier);
2975 	if (err)
2976 		goto reg_notif_fail;
2977 	return 0;
2978 
2979 reg_notif_fail:
2980 	unregister_pernet_subsys(&igmp_net_ops);
2981 	return err;
2982 #else
2983 	return register_netdevice_notifier(&igmp_notifier);
2984 #endif
2985 }
2986