xref: /openbmc/linux/net/ipv4/igmp.c (revision 5bd8e16d)
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 <linux/netfilter_ipv4.h>
101 #ifdef CONFIG_IP_MROUTE
102 #include <linux/mroute.h>
103 #endif
104 #ifdef CONFIG_PROC_FS
105 #include <linux/proc_fs.h>
106 #include <linux/seq_file.h>
107 #endif
108 
109 #define IP_MAX_MEMBERSHIPS	20
110 #define IP_MAX_MSF		10
111 
112 #ifdef CONFIG_IP_MULTICAST
113 /* Parameter names and values are taken from igmp-v2-06 draft */
114 
115 #define IGMP_V1_Router_Present_Timeout		(400*HZ)
116 #define IGMP_V2_Router_Present_Timeout		(400*HZ)
117 #define IGMP_V2_Unsolicited_Report_Interval	(10*HZ)
118 #define IGMP_V3_Unsolicited_Report_Interval	(1*HZ)
119 #define IGMP_Query_Response_Interval		(10*HZ)
120 #define IGMP_Unsolicited_Report_Count		2
121 
122 
123 #define IGMP_Initial_Report_Delay		(1)
124 
125 /* IGMP_Initial_Report_Delay is not from IGMP specs!
126  * IGMP specs require to report membership immediately after
127  * joining a group, but we delay the first report by a
128  * small interval. It seems more natural and still does not
129  * contradict to specs provided this delay is small enough.
130  */
131 
132 #define IGMP_V1_SEEN(in_dev) \
133 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
134 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
135 	 ((in_dev)->mr_v1_seen && \
136 	  time_before(jiffies, (in_dev)->mr_v1_seen)))
137 #define IGMP_V2_SEEN(in_dev) \
138 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
139 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
140 	 ((in_dev)->mr_v2_seen && \
141 	  time_before(jiffies, (in_dev)->mr_v2_seen)))
142 
143 static int unsolicited_report_interval(struct in_device *in_dev)
144 {
145 	int interval_ms, interval_jiffies;
146 
147 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
148 		interval_ms = IN_DEV_CONF_GET(
149 			in_dev,
150 			IGMPV2_UNSOLICITED_REPORT_INTERVAL);
151 	else /* v3 */
152 		interval_ms = IN_DEV_CONF_GET(
153 			in_dev,
154 			IGMPV3_UNSOLICITED_REPORT_INTERVAL);
155 
156 	interval_jiffies = msecs_to_jiffies(interval_ms);
157 
158 	/* _timer functions can't handle a delay of 0 jiffies so ensure
159 	 *  we always return a positive value.
160 	 */
161 	if (interval_jiffies <= 0)
162 		interval_jiffies = 1;
163 	return interval_jiffies;
164 }
165 
166 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
167 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
168 static void igmpv3_clear_delrec(struct in_device *in_dev);
169 static int sf_setstate(struct ip_mc_list *pmc);
170 static void sf_markstate(struct ip_mc_list *pmc);
171 #endif
172 static void ip_mc_clear_src(struct ip_mc_list *pmc);
173 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
174 			 int sfcount, __be32 *psfsrc, int delta);
175 
176 static void ip_ma_put(struct ip_mc_list *im)
177 {
178 	if (atomic_dec_and_test(&im->refcnt)) {
179 		in_dev_put(im->interface);
180 		kfree_rcu(im, rcu);
181 	}
182 }
183 
184 #define for_each_pmc_rcu(in_dev, pmc)				\
185 	for (pmc = rcu_dereference(in_dev->mc_list);		\
186 	     pmc != NULL;					\
187 	     pmc = rcu_dereference(pmc->next_rcu))
188 
189 #define for_each_pmc_rtnl(in_dev, pmc)				\
190 	for (pmc = rtnl_dereference(in_dev->mc_list);		\
191 	     pmc != NULL;					\
192 	     pmc = rtnl_dereference(pmc->next_rcu))
193 
194 #ifdef CONFIG_IP_MULTICAST
195 
196 /*
197  *	Timer management
198  */
199 
200 static void igmp_stop_timer(struct ip_mc_list *im)
201 {
202 	spin_lock_bh(&im->lock);
203 	if (del_timer(&im->timer))
204 		atomic_dec(&im->refcnt);
205 	im->tm_running = 0;
206 	im->reporter = 0;
207 	im->unsolicit_count = 0;
208 	spin_unlock_bh(&im->lock);
209 }
210 
211 /* It must be called with locked im->lock */
212 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
213 {
214 	int tv = net_random() % max_delay;
215 
216 	im->tm_running = 1;
217 	if (!mod_timer(&im->timer, jiffies+tv+2))
218 		atomic_inc(&im->refcnt);
219 }
220 
221 static void igmp_gq_start_timer(struct in_device *in_dev)
222 {
223 	int tv = net_random() % in_dev->mr_maxdelay;
224 
225 	in_dev->mr_gq_running = 1;
226 	if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
227 		in_dev_hold(in_dev);
228 }
229 
230 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
231 {
232 	int tv = net_random() % delay;
233 
234 	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
235 		in_dev_hold(in_dev);
236 }
237 
238 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
239 {
240 	spin_lock_bh(&im->lock);
241 	im->unsolicit_count = 0;
242 	if (del_timer(&im->timer)) {
243 		if ((long)(im->timer.expires-jiffies) < max_delay) {
244 			add_timer(&im->timer);
245 			im->tm_running = 1;
246 			spin_unlock_bh(&im->lock);
247 			return;
248 		}
249 		atomic_dec(&im->refcnt);
250 	}
251 	igmp_start_timer(im, max_delay);
252 	spin_unlock_bh(&im->lock);
253 }
254 
255 
256 /*
257  *	Send an IGMP report.
258  */
259 
260 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
261 
262 
263 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
264 	int gdeleted, int sdeleted)
265 {
266 	switch (type) {
267 	case IGMPV3_MODE_IS_INCLUDE:
268 	case IGMPV3_MODE_IS_EXCLUDE:
269 		if (gdeleted || sdeleted)
270 			return 0;
271 		if (!(pmc->gsquery && !psf->sf_gsresp)) {
272 			if (pmc->sfmode == MCAST_INCLUDE)
273 				return 1;
274 			/* don't include if this source is excluded
275 			 * in all filters
276 			 */
277 			if (psf->sf_count[MCAST_INCLUDE])
278 				return type == IGMPV3_MODE_IS_INCLUDE;
279 			return pmc->sfcount[MCAST_EXCLUDE] ==
280 				psf->sf_count[MCAST_EXCLUDE];
281 		}
282 		return 0;
283 	case IGMPV3_CHANGE_TO_INCLUDE:
284 		if (gdeleted || sdeleted)
285 			return 0;
286 		return psf->sf_count[MCAST_INCLUDE] != 0;
287 	case IGMPV3_CHANGE_TO_EXCLUDE:
288 		if (gdeleted || sdeleted)
289 			return 0;
290 		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
291 		    psf->sf_count[MCAST_INCLUDE])
292 			return 0;
293 		return pmc->sfcount[MCAST_EXCLUDE] ==
294 			psf->sf_count[MCAST_EXCLUDE];
295 	case IGMPV3_ALLOW_NEW_SOURCES:
296 		if (gdeleted || !psf->sf_crcount)
297 			return 0;
298 		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
299 	case IGMPV3_BLOCK_OLD_SOURCES:
300 		if (pmc->sfmode == MCAST_INCLUDE)
301 			return gdeleted || (psf->sf_crcount && sdeleted);
302 		return psf->sf_crcount && !gdeleted && !sdeleted;
303 	}
304 	return 0;
305 }
306 
307 static int
308 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
309 {
310 	struct ip_sf_list *psf;
311 	int scount = 0;
312 
313 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
314 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
315 			continue;
316 		scount++;
317 	}
318 	return scount;
319 }
320 
321 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
322 
323 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
324 {
325 	struct sk_buff *skb;
326 	struct rtable *rt;
327 	struct iphdr *pip;
328 	struct igmpv3_report *pig;
329 	struct net *net = dev_net(dev);
330 	struct flowi4 fl4;
331 	int hlen = LL_RESERVED_SPACE(dev);
332 	int tlen = dev->needed_tailroom;
333 
334 	while (1) {
335 		skb = alloc_skb(size + hlen + tlen,
336 				GFP_ATOMIC | __GFP_NOWARN);
337 		if (skb)
338 			break;
339 		size >>= 1;
340 		if (size < 256)
341 			return NULL;
342 	}
343 	skb->priority = TC_PRIO_CONTROL;
344 	igmp_skb_size(skb) = size;
345 
346 	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
347 				   0, 0,
348 				   IPPROTO_IGMP, 0, dev->ifindex);
349 	if (IS_ERR(rt)) {
350 		kfree_skb(skb);
351 		return NULL;
352 	}
353 
354 	skb_dst_set(skb, &rt->dst);
355 	skb->dev = dev;
356 
357 	skb_reserve(skb, hlen);
358 
359 	skb_reset_network_header(skb);
360 	pip = ip_hdr(skb);
361 	skb_put(skb, sizeof(struct iphdr) + 4);
362 
363 	pip->version  = 4;
364 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
365 	pip->tos      = 0xc0;
366 	pip->frag_off = htons(IP_DF);
367 	pip->ttl      = 1;
368 	pip->daddr    = fl4.daddr;
369 	pip->saddr    = fl4.saddr;
370 	pip->protocol = IPPROTO_IGMP;
371 	pip->tot_len  = 0;	/* filled in later */
372 	ip_select_ident(skb, &rt->dst, NULL);
373 	((u8 *)&pip[1])[0] = IPOPT_RA;
374 	((u8 *)&pip[1])[1] = 4;
375 	((u8 *)&pip[1])[2] = 0;
376 	((u8 *)&pip[1])[3] = 0;
377 
378 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
379 	skb_put(skb, sizeof(*pig));
380 	pig = igmpv3_report_hdr(skb);
381 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
382 	pig->resv1 = 0;
383 	pig->csum = 0;
384 	pig->resv2 = 0;
385 	pig->ngrec = 0;
386 	return skb;
387 }
388 
389 static int igmpv3_sendpack(struct sk_buff *skb)
390 {
391 	struct igmphdr *pig = igmp_hdr(skb);
392 	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
393 
394 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
395 
396 	return ip_local_out(skb);
397 }
398 
399 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
400 {
401 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
402 }
403 
404 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
405 	int type, struct igmpv3_grec **ppgr)
406 {
407 	struct net_device *dev = pmc->interface->dev;
408 	struct igmpv3_report *pih;
409 	struct igmpv3_grec *pgr;
410 
411 	if (!skb)
412 		skb = igmpv3_newpack(dev, dev->mtu);
413 	if (!skb)
414 		return NULL;
415 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
416 	pgr->grec_type = type;
417 	pgr->grec_auxwords = 0;
418 	pgr->grec_nsrcs = 0;
419 	pgr->grec_mca = pmc->multiaddr;
420 	pih = igmpv3_report_hdr(skb);
421 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
422 	*ppgr = pgr;
423 	return skb;
424 }
425 
426 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
427 	skb_tailroom(skb)) : 0)
428 
429 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
430 	int type, int gdeleted, int sdeleted)
431 {
432 	struct net_device *dev = pmc->interface->dev;
433 	struct igmpv3_report *pih;
434 	struct igmpv3_grec *pgr = NULL;
435 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
436 	int scount, stotal, first, isquery, truncate;
437 
438 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
439 		return skb;
440 
441 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
442 		  type == IGMPV3_MODE_IS_EXCLUDE;
443 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
444 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
445 
446 	stotal = scount = 0;
447 
448 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
449 
450 	if (!*psf_list)
451 		goto empty_source;
452 
453 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
454 
455 	/* EX and TO_EX get a fresh packet, if needed */
456 	if (truncate) {
457 		if (pih && pih->ngrec &&
458 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
459 			if (skb)
460 				igmpv3_sendpack(skb);
461 			skb = igmpv3_newpack(dev, dev->mtu);
462 		}
463 	}
464 	first = 1;
465 	psf_prev = NULL;
466 	for (psf=*psf_list; psf; psf=psf_next) {
467 		__be32 *psrc;
468 
469 		psf_next = psf->sf_next;
470 
471 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
472 			psf_prev = psf;
473 			continue;
474 		}
475 
476 		/* clear marks on query responses */
477 		if (isquery)
478 			psf->sf_gsresp = 0;
479 
480 		if (AVAILABLE(skb) < sizeof(__be32) +
481 		    first*sizeof(struct igmpv3_grec)) {
482 			if (truncate && !first)
483 				break;	 /* truncate these */
484 			if (pgr)
485 				pgr->grec_nsrcs = htons(scount);
486 			if (skb)
487 				igmpv3_sendpack(skb);
488 			skb = igmpv3_newpack(dev, dev->mtu);
489 			first = 1;
490 			scount = 0;
491 		}
492 		if (first) {
493 			skb = add_grhead(skb, pmc, type, &pgr);
494 			first = 0;
495 		}
496 		if (!skb)
497 			return NULL;
498 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
499 		*psrc = psf->sf_inaddr;
500 		scount++; stotal++;
501 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
502 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
503 			psf->sf_crcount--;
504 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
505 				if (psf_prev)
506 					psf_prev->sf_next = psf->sf_next;
507 				else
508 					*psf_list = psf->sf_next;
509 				kfree(psf);
510 				continue;
511 			}
512 		}
513 		psf_prev = psf;
514 	}
515 
516 empty_source:
517 	if (!stotal) {
518 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
519 		    type == IGMPV3_BLOCK_OLD_SOURCES)
520 			return skb;
521 		if (pmc->crcount || isquery) {
522 			/* make sure we have room for group header */
523 			if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
524 				igmpv3_sendpack(skb);
525 				skb = NULL; /* add_grhead will get a new one */
526 			}
527 			skb = add_grhead(skb, pmc, type, &pgr);
528 		}
529 	}
530 	if (pgr)
531 		pgr->grec_nsrcs = htons(scount);
532 
533 	if (isquery)
534 		pmc->gsquery = 0;	/* clear query state on report */
535 	return skb;
536 }
537 
538 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
539 {
540 	struct sk_buff *skb = NULL;
541 	int type;
542 
543 	if (!pmc) {
544 		rcu_read_lock();
545 		for_each_pmc_rcu(in_dev, pmc) {
546 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
547 				continue;
548 			spin_lock_bh(&pmc->lock);
549 			if (pmc->sfcount[MCAST_EXCLUDE])
550 				type = IGMPV3_MODE_IS_EXCLUDE;
551 			else
552 				type = IGMPV3_MODE_IS_INCLUDE;
553 			skb = add_grec(skb, pmc, type, 0, 0);
554 			spin_unlock_bh(&pmc->lock);
555 		}
556 		rcu_read_unlock();
557 	} else {
558 		spin_lock_bh(&pmc->lock);
559 		if (pmc->sfcount[MCAST_EXCLUDE])
560 			type = IGMPV3_MODE_IS_EXCLUDE;
561 		else
562 			type = IGMPV3_MODE_IS_INCLUDE;
563 		skb = add_grec(skb, pmc, type, 0, 0);
564 		spin_unlock_bh(&pmc->lock);
565 	}
566 	if (!skb)
567 		return 0;
568 	return igmpv3_sendpack(skb);
569 }
570 
571 /*
572  * remove zero-count source records from a source filter list
573  */
574 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
575 {
576 	struct ip_sf_list *psf_prev, *psf_next, *psf;
577 
578 	psf_prev = NULL;
579 	for (psf=*ppsf; psf; psf = psf_next) {
580 		psf_next = psf->sf_next;
581 		if (psf->sf_crcount == 0) {
582 			if (psf_prev)
583 				psf_prev->sf_next = psf->sf_next;
584 			else
585 				*ppsf = psf->sf_next;
586 			kfree(psf);
587 		} else
588 			psf_prev = psf;
589 	}
590 }
591 
592 static void igmpv3_send_cr(struct in_device *in_dev)
593 {
594 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
595 	struct sk_buff *skb = NULL;
596 	int type, dtype;
597 
598 	rcu_read_lock();
599 	spin_lock_bh(&in_dev->mc_tomb_lock);
600 
601 	/* deleted MCA's */
602 	pmc_prev = NULL;
603 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
604 		pmc_next = pmc->next;
605 		if (pmc->sfmode == MCAST_INCLUDE) {
606 			type = IGMPV3_BLOCK_OLD_SOURCES;
607 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
608 			skb = add_grec(skb, pmc, type, 1, 0);
609 			skb = add_grec(skb, pmc, dtype, 1, 1);
610 		}
611 		if (pmc->crcount) {
612 			if (pmc->sfmode == MCAST_EXCLUDE) {
613 				type = IGMPV3_CHANGE_TO_INCLUDE;
614 				skb = add_grec(skb, pmc, type, 1, 0);
615 			}
616 			pmc->crcount--;
617 			if (pmc->crcount == 0) {
618 				igmpv3_clear_zeros(&pmc->tomb);
619 				igmpv3_clear_zeros(&pmc->sources);
620 			}
621 		}
622 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
623 			if (pmc_prev)
624 				pmc_prev->next = pmc_next;
625 			else
626 				in_dev->mc_tomb = pmc_next;
627 			in_dev_put(pmc->interface);
628 			kfree(pmc);
629 		} else
630 			pmc_prev = pmc;
631 	}
632 	spin_unlock_bh(&in_dev->mc_tomb_lock);
633 
634 	/* change recs */
635 	for_each_pmc_rcu(in_dev, pmc) {
636 		spin_lock_bh(&pmc->lock);
637 		if (pmc->sfcount[MCAST_EXCLUDE]) {
638 			type = IGMPV3_BLOCK_OLD_SOURCES;
639 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
640 		} else {
641 			type = IGMPV3_ALLOW_NEW_SOURCES;
642 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
643 		}
644 		skb = add_grec(skb, pmc, type, 0, 0);
645 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
646 
647 		/* filter mode changes */
648 		if (pmc->crcount) {
649 			if (pmc->sfmode == MCAST_EXCLUDE)
650 				type = IGMPV3_CHANGE_TO_EXCLUDE;
651 			else
652 				type = IGMPV3_CHANGE_TO_INCLUDE;
653 			skb = add_grec(skb, pmc, type, 0, 0);
654 			pmc->crcount--;
655 		}
656 		spin_unlock_bh(&pmc->lock);
657 	}
658 	rcu_read_unlock();
659 
660 	if (!skb)
661 		return;
662 	(void) igmpv3_sendpack(skb);
663 }
664 
665 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
666 	int type)
667 {
668 	struct sk_buff *skb;
669 	struct iphdr *iph;
670 	struct igmphdr *ih;
671 	struct rtable *rt;
672 	struct net_device *dev = in_dev->dev;
673 	struct net *net = dev_net(dev);
674 	__be32	group = pmc ? pmc->multiaddr : 0;
675 	struct flowi4 fl4;
676 	__be32	dst;
677 	int hlen, tlen;
678 
679 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
680 		return igmpv3_send_report(in_dev, pmc);
681 	else if (type == IGMP_HOST_LEAVE_MESSAGE)
682 		dst = IGMP_ALL_ROUTER;
683 	else
684 		dst = group;
685 
686 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
687 				   0, 0,
688 				   IPPROTO_IGMP, 0, dev->ifindex);
689 	if (IS_ERR(rt))
690 		return -1;
691 
692 	hlen = LL_RESERVED_SPACE(dev);
693 	tlen = dev->needed_tailroom;
694 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
695 	if (skb == NULL) {
696 		ip_rt_put(rt);
697 		return -1;
698 	}
699 	skb->priority = TC_PRIO_CONTROL;
700 
701 	skb_dst_set(skb, &rt->dst);
702 
703 	skb_reserve(skb, hlen);
704 
705 	skb_reset_network_header(skb);
706 	iph = ip_hdr(skb);
707 	skb_put(skb, sizeof(struct iphdr) + 4);
708 
709 	iph->version  = 4;
710 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
711 	iph->tos      = 0xc0;
712 	iph->frag_off = htons(IP_DF);
713 	iph->ttl      = 1;
714 	iph->daddr    = dst;
715 	iph->saddr    = fl4.saddr;
716 	iph->protocol = IPPROTO_IGMP;
717 	ip_select_ident(skb, &rt->dst, NULL);
718 	((u8 *)&iph[1])[0] = IPOPT_RA;
719 	((u8 *)&iph[1])[1] = 4;
720 	((u8 *)&iph[1])[2] = 0;
721 	((u8 *)&iph[1])[3] = 0;
722 
723 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
724 	ih->type = type;
725 	ih->code = 0;
726 	ih->csum = 0;
727 	ih->group = group;
728 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
729 
730 	return ip_local_out(skb);
731 }
732 
733 static void igmp_gq_timer_expire(unsigned long data)
734 {
735 	struct in_device *in_dev = (struct in_device *)data;
736 
737 	in_dev->mr_gq_running = 0;
738 	igmpv3_send_report(in_dev, NULL);
739 	__in_dev_put(in_dev);
740 }
741 
742 static void igmp_ifc_timer_expire(unsigned long data)
743 {
744 	struct in_device *in_dev = (struct in_device *)data;
745 
746 	igmpv3_send_cr(in_dev);
747 	if (in_dev->mr_ifc_count) {
748 		in_dev->mr_ifc_count--;
749 		igmp_ifc_start_timer(in_dev,
750 				     unsolicited_report_interval(in_dev));
751 	}
752 	__in_dev_put(in_dev);
753 }
754 
755 static void igmp_ifc_event(struct in_device *in_dev)
756 {
757 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
758 		return;
759 	in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
760 		IGMP_Unsolicited_Report_Count;
761 	igmp_ifc_start_timer(in_dev, 1);
762 }
763 
764 
765 static void igmp_timer_expire(unsigned long data)
766 {
767 	struct ip_mc_list *im=(struct ip_mc_list *)data;
768 	struct in_device *in_dev = im->interface;
769 
770 	spin_lock(&im->lock);
771 	im->tm_running = 0;
772 
773 	if (im->unsolicit_count) {
774 		im->unsolicit_count--;
775 		igmp_start_timer(im, unsolicited_report_interval(in_dev));
776 	}
777 	im->reporter = 1;
778 	spin_unlock(&im->lock);
779 
780 	if (IGMP_V1_SEEN(in_dev))
781 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
782 	else if (IGMP_V2_SEEN(in_dev))
783 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
784 	else
785 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
786 
787 	ip_ma_put(im);
788 }
789 
790 /* mark EXCLUDE-mode sources */
791 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
792 {
793 	struct ip_sf_list *psf;
794 	int i, scount;
795 
796 	scount = 0;
797 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
798 		if (scount == nsrcs)
799 			break;
800 		for (i=0; i<nsrcs; i++) {
801 			/* skip inactive filters */
802 			if (psf->sf_count[MCAST_INCLUDE] ||
803 			    pmc->sfcount[MCAST_EXCLUDE] !=
804 			    psf->sf_count[MCAST_EXCLUDE])
805 				break;
806 			if (srcs[i] == psf->sf_inaddr) {
807 				scount++;
808 				break;
809 			}
810 		}
811 	}
812 	pmc->gsquery = 0;
813 	if (scount == nsrcs)	/* all sources excluded */
814 		return 0;
815 	return 1;
816 }
817 
818 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
819 {
820 	struct ip_sf_list *psf;
821 	int i, scount;
822 
823 	if (pmc->sfmode == MCAST_EXCLUDE)
824 		return igmp_xmarksources(pmc, nsrcs, srcs);
825 
826 	/* mark INCLUDE-mode sources */
827 	scount = 0;
828 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
829 		if (scount == nsrcs)
830 			break;
831 		for (i=0; i<nsrcs; i++)
832 			if (srcs[i] == psf->sf_inaddr) {
833 				psf->sf_gsresp = 1;
834 				scount++;
835 				break;
836 			}
837 	}
838 	if (!scount) {
839 		pmc->gsquery = 0;
840 		return 0;
841 	}
842 	pmc->gsquery = 1;
843 	return 1;
844 }
845 
846 /* return true if packet was dropped */
847 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
848 {
849 	struct ip_mc_list *im;
850 
851 	/* Timers are only set for non-local groups */
852 
853 	if (group == IGMP_ALL_HOSTS)
854 		return false;
855 
856 	rcu_read_lock();
857 	for_each_pmc_rcu(in_dev, im) {
858 		if (im->multiaddr == group) {
859 			igmp_stop_timer(im);
860 			break;
861 		}
862 	}
863 	rcu_read_unlock();
864 	return false;
865 }
866 
867 /* return true if packet was dropped */
868 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
869 	int len)
870 {
871 	struct igmphdr 		*ih = igmp_hdr(skb);
872 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
873 	struct ip_mc_list	*im;
874 	__be32			group = ih->group;
875 	int			max_delay;
876 	int			mark = 0;
877 
878 
879 	if (len == 8) {
880 		if (ih->code == 0) {
881 			/* Alas, old v1 router presents here. */
882 
883 			max_delay = IGMP_Query_Response_Interval;
884 			in_dev->mr_v1_seen = jiffies +
885 				IGMP_V1_Router_Present_Timeout;
886 			group = 0;
887 		} else {
888 			/* v2 router present */
889 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
890 			in_dev->mr_v2_seen = jiffies +
891 				IGMP_V2_Router_Present_Timeout;
892 		}
893 		/* cancel the interface change timer */
894 		in_dev->mr_ifc_count = 0;
895 		if (del_timer(&in_dev->mr_ifc_timer))
896 			__in_dev_put(in_dev);
897 		/* clear deleted report items */
898 		igmpv3_clear_delrec(in_dev);
899 	} else if (len < 12) {
900 		return true;	/* ignore bogus packet; freed by caller */
901 	} else if (IGMP_V1_SEEN(in_dev)) {
902 		/* This is a v3 query with v1 queriers present */
903 		max_delay = IGMP_Query_Response_Interval;
904 		group = 0;
905 	} else if (IGMP_V2_SEEN(in_dev)) {
906 		/* this is a v3 query with v2 queriers present;
907 		 * Interpretation of the max_delay code is problematic here.
908 		 * A real v2 host would use ih_code directly, while v3 has a
909 		 * different encoding. We use the v3 encoding as more likely
910 		 * to be intended in a v3 query.
911 		 */
912 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
913 		if (!max_delay)
914 			max_delay = 1;	/* can't mod w/ 0 */
915 	} else { /* v3 */
916 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
917 			return true;
918 
919 		ih3 = igmpv3_query_hdr(skb);
920 		if (ih3->nsrcs) {
921 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
922 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
923 				return true;
924 			ih3 = igmpv3_query_hdr(skb);
925 		}
926 
927 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
928 		if (!max_delay)
929 			max_delay = 1;	/* can't mod w/ 0 */
930 		in_dev->mr_maxdelay = max_delay;
931 		if (ih3->qrv)
932 			in_dev->mr_qrv = ih3->qrv;
933 		if (!group) { /* general query */
934 			if (ih3->nsrcs)
935 				return false;	/* no sources allowed */
936 			igmp_gq_start_timer(in_dev);
937 			return false;
938 		}
939 		/* mark sources to include, if group & source-specific */
940 		mark = ih3->nsrcs != 0;
941 	}
942 
943 	/*
944 	 * - Start the timers in all of our membership records
945 	 *   that the query applies to for the interface on
946 	 *   which the query arrived excl. those that belong
947 	 *   to a "local" group (224.0.0.X)
948 	 * - For timers already running check if they need to
949 	 *   be reset.
950 	 * - Use the igmp->igmp_code field as the maximum
951 	 *   delay possible
952 	 */
953 	rcu_read_lock();
954 	for_each_pmc_rcu(in_dev, im) {
955 		int changed;
956 
957 		if (group && group != im->multiaddr)
958 			continue;
959 		if (im->multiaddr == IGMP_ALL_HOSTS)
960 			continue;
961 		spin_lock_bh(&im->lock);
962 		if (im->tm_running)
963 			im->gsquery = im->gsquery && mark;
964 		else
965 			im->gsquery = mark;
966 		changed = !im->gsquery ||
967 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
968 		spin_unlock_bh(&im->lock);
969 		if (changed)
970 			igmp_mod_timer(im, max_delay);
971 	}
972 	rcu_read_unlock();
973 	return false;
974 }
975 
976 /* called in rcu_read_lock() section */
977 int igmp_rcv(struct sk_buff *skb)
978 {
979 	/* This basically follows the spec line by line -- see RFC1112 */
980 	struct igmphdr *ih;
981 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
982 	int len = skb->len;
983 	bool dropped = true;
984 
985 	if (in_dev == NULL)
986 		goto drop;
987 
988 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
989 		goto drop;
990 
991 	switch (skb->ip_summed) {
992 	case CHECKSUM_COMPLETE:
993 		if (!csum_fold(skb->csum))
994 			break;
995 		/* fall through */
996 	case CHECKSUM_NONE:
997 		skb->csum = 0;
998 		if (__skb_checksum_complete(skb))
999 			goto drop;
1000 	}
1001 
1002 	ih = igmp_hdr(skb);
1003 	switch (ih->type) {
1004 	case IGMP_HOST_MEMBERSHIP_QUERY:
1005 		dropped = igmp_heard_query(in_dev, skb, len);
1006 		break;
1007 	case IGMP_HOST_MEMBERSHIP_REPORT:
1008 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1009 		/* Is it our report looped back? */
1010 		if (rt_is_output_route(skb_rtable(skb)))
1011 			break;
1012 		/* don't rely on MC router hearing unicast reports */
1013 		if (skb->pkt_type == PACKET_MULTICAST ||
1014 		    skb->pkt_type == PACKET_BROADCAST)
1015 			dropped = igmp_heard_report(in_dev, ih->group);
1016 		break;
1017 	case IGMP_PIM:
1018 #ifdef CONFIG_IP_PIMSM_V1
1019 		return pim_rcv_v1(skb);
1020 #endif
1021 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1022 	case IGMP_DVMRP:
1023 	case IGMP_TRACE:
1024 	case IGMP_HOST_LEAVE_MESSAGE:
1025 	case IGMP_MTRACE:
1026 	case IGMP_MTRACE_RESP:
1027 		break;
1028 	default:
1029 		break;
1030 	}
1031 
1032 drop:
1033 	if (dropped)
1034 		kfree_skb(skb);
1035 	else
1036 		consume_skb(skb);
1037 	return 0;
1038 }
1039 
1040 #endif
1041 
1042 
1043 /*
1044  *	Add a filter to a device
1045  */
1046 
1047 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1048 {
1049 	char buf[MAX_ADDR_LEN];
1050 	struct net_device *dev = in_dev->dev;
1051 
1052 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1053 	   We will get multicast token leakage, when IFF_MULTICAST
1054 	   is changed. This check should be done in ndo_set_rx_mode
1055 	   routine. Something sort of:
1056 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1057 	   --ANK
1058 	   */
1059 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1060 		dev_mc_add(dev, buf);
1061 }
1062 
1063 /*
1064  *	Remove a filter from a device
1065  */
1066 
1067 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1068 {
1069 	char buf[MAX_ADDR_LEN];
1070 	struct net_device *dev = in_dev->dev;
1071 
1072 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1073 		dev_mc_del(dev, buf);
1074 }
1075 
1076 #ifdef CONFIG_IP_MULTICAST
1077 /*
1078  * deleted ip_mc_list manipulation
1079  */
1080 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1081 {
1082 	struct ip_mc_list *pmc;
1083 
1084 	/* this is an "ip_mc_list" for convenience; only the fields below
1085 	 * are actually used. In particular, the refcnt and users are not
1086 	 * used for management of the delete list. Using the same structure
1087 	 * for deleted items allows change reports to use common code with
1088 	 * non-deleted or query-response MCA's.
1089 	 */
1090 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1091 	if (!pmc)
1092 		return;
1093 	spin_lock_bh(&im->lock);
1094 	pmc->interface = im->interface;
1095 	in_dev_hold(in_dev);
1096 	pmc->multiaddr = im->multiaddr;
1097 	pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1098 		IGMP_Unsolicited_Report_Count;
1099 	pmc->sfmode = im->sfmode;
1100 	if (pmc->sfmode == MCAST_INCLUDE) {
1101 		struct ip_sf_list *psf;
1102 
1103 		pmc->tomb = im->tomb;
1104 		pmc->sources = im->sources;
1105 		im->tomb = im->sources = NULL;
1106 		for (psf=pmc->sources; psf; psf=psf->sf_next)
1107 			psf->sf_crcount = pmc->crcount;
1108 	}
1109 	spin_unlock_bh(&im->lock);
1110 
1111 	spin_lock_bh(&in_dev->mc_tomb_lock);
1112 	pmc->next = in_dev->mc_tomb;
1113 	in_dev->mc_tomb = pmc;
1114 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1115 }
1116 
1117 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1118 {
1119 	struct ip_mc_list *pmc, *pmc_prev;
1120 	struct ip_sf_list *psf, *psf_next;
1121 
1122 	spin_lock_bh(&in_dev->mc_tomb_lock);
1123 	pmc_prev = NULL;
1124 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1125 		if (pmc->multiaddr == multiaddr)
1126 			break;
1127 		pmc_prev = pmc;
1128 	}
1129 	if (pmc) {
1130 		if (pmc_prev)
1131 			pmc_prev->next = pmc->next;
1132 		else
1133 			in_dev->mc_tomb = pmc->next;
1134 	}
1135 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1136 	if (pmc) {
1137 		for (psf=pmc->tomb; psf; psf=psf_next) {
1138 			psf_next = psf->sf_next;
1139 			kfree(psf);
1140 		}
1141 		in_dev_put(pmc->interface);
1142 		kfree(pmc);
1143 	}
1144 }
1145 
1146 static void igmpv3_clear_delrec(struct in_device *in_dev)
1147 {
1148 	struct ip_mc_list *pmc, *nextpmc;
1149 
1150 	spin_lock_bh(&in_dev->mc_tomb_lock);
1151 	pmc = in_dev->mc_tomb;
1152 	in_dev->mc_tomb = NULL;
1153 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1154 
1155 	for (; pmc; pmc = nextpmc) {
1156 		nextpmc = pmc->next;
1157 		ip_mc_clear_src(pmc);
1158 		in_dev_put(pmc->interface);
1159 		kfree(pmc);
1160 	}
1161 	/* clear dead sources, too */
1162 	rcu_read_lock();
1163 	for_each_pmc_rcu(in_dev, pmc) {
1164 		struct ip_sf_list *psf, *psf_next;
1165 
1166 		spin_lock_bh(&pmc->lock);
1167 		psf = pmc->tomb;
1168 		pmc->tomb = NULL;
1169 		spin_unlock_bh(&pmc->lock);
1170 		for (; psf; psf=psf_next) {
1171 			psf_next = psf->sf_next;
1172 			kfree(psf);
1173 		}
1174 	}
1175 	rcu_read_unlock();
1176 }
1177 #endif
1178 
1179 static void igmp_group_dropped(struct ip_mc_list *im)
1180 {
1181 	struct in_device *in_dev = im->interface;
1182 #ifdef CONFIG_IP_MULTICAST
1183 	int reporter;
1184 #endif
1185 
1186 	if (im->loaded) {
1187 		im->loaded = 0;
1188 		ip_mc_filter_del(in_dev, im->multiaddr);
1189 	}
1190 
1191 #ifdef CONFIG_IP_MULTICAST
1192 	if (im->multiaddr == IGMP_ALL_HOSTS)
1193 		return;
1194 
1195 	reporter = im->reporter;
1196 	igmp_stop_timer(im);
1197 
1198 	if (!in_dev->dead) {
1199 		if (IGMP_V1_SEEN(in_dev))
1200 			return;
1201 		if (IGMP_V2_SEEN(in_dev)) {
1202 			if (reporter)
1203 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1204 			return;
1205 		}
1206 		/* IGMPv3 */
1207 		igmpv3_add_delrec(in_dev, im);
1208 
1209 		igmp_ifc_event(in_dev);
1210 	}
1211 #endif
1212 }
1213 
1214 static void igmp_group_added(struct ip_mc_list *im)
1215 {
1216 	struct in_device *in_dev = im->interface;
1217 
1218 	if (im->loaded == 0) {
1219 		im->loaded = 1;
1220 		ip_mc_filter_add(in_dev, im->multiaddr);
1221 	}
1222 
1223 #ifdef CONFIG_IP_MULTICAST
1224 	if (im->multiaddr == IGMP_ALL_HOSTS)
1225 		return;
1226 
1227 	if (in_dev->dead)
1228 		return;
1229 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1230 		spin_lock_bh(&im->lock);
1231 		igmp_start_timer(im, IGMP_Initial_Report_Delay);
1232 		spin_unlock_bh(&im->lock);
1233 		return;
1234 	}
1235 	/* else, v3 */
1236 
1237 	im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1238 		IGMP_Unsolicited_Report_Count;
1239 	igmp_ifc_event(in_dev);
1240 #endif
1241 }
1242 
1243 
1244 /*
1245  *	Multicast list managers
1246  */
1247 
1248 static u32 ip_mc_hash(const struct ip_mc_list *im)
1249 {
1250 	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1251 }
1252 
1253 static void ip_mc_hash_add(struct in_device *in_dev,
1254 			   struct ip_mc_list *im)
1255 {
1256 	struct ip_mc_list __rcu **mc_hash;
1257 	u32 hash;
1258 
1259 	mc_hash = rtnl_dereference(in_dev->mc_hash);
1260 	if (mc_hash) {
1261 		hash = ip_mc_hash(im);
1262 		im->next_hash = mc_hash[hash];
1263 		rcu_assign_pointer(mc_hash[hash], im);
1264 		return;
1265 	}
1266 
1267 	/* do not use a hash table for small number of items */
1268 	if (in_dev->mc_count < 4)
1269 		return;
1270 
1271 	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1272 			  GFP_KERNEL);
1273 	if (!mc_hash)
1274 		return;
1275 
1276 	for_each_pmc_rtnl(in_dev, im) {
1277 		hash = ip_mc_hash(im);
1278 		im->next_hash = mc_hash[hash];
1279 		RCU_INIT_POINTER(mc_hash[hash], im);
1280 	}
1281 
1282 	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1283 }
1284 
1285 static void ip_mc_hash_remove(struct in_device *in_dev,
1286 			      struct ip_mc_list *im)
1287 {
1288 	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1289 	struct ip_mc_list *aux;
1290 
1291 	if (!mc_hash)
1292 		return;
1293 	mc_hash += ip_mc_hash(im);
1294 	while ((aux = rtnl_dereference(*mc_hash)) != im)
1295 		mc_hash = &aux->next_hash;
1296 	*mc_hash = im->next_hash;
1297 }
1298 
1299 
1300 /*
1301  *	A socket has joined a multicast group on device dev.
1302  */
1303 
1304 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1305 {
1306 	struct ip_mc_list *im;
1307 
1308 	ASSERT_RTNL();
1309 
1310 	for_each_pmc_rtnl(in_dev, im) {
1311 		if (im->multiaddr == addr) {
1312 			im->users++;
1313 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1314 			goto out;
1315 		}
1316 	}
1317 
1318 	im = kzalloc(sizeof(*im), GFP_KERNEL);
1319 	if (!im)
1320 		goto out;
1321 
1322 	im->users = 1;
1323 	im->interface = in_dev;
1324 	in_dev_hold(in_dev);
1325 	im->multiaddr = addr;
1326 	/* initial mode is (EX, empty) */
1327 	im->sfmode = MCAST_EXCLUDE;
1328 	im->sfcount[MCAST_EXCLUDE] = 1;
1329 	atomic_set(&im->refcnt, 1);
1330 	spin_lock_init(&im->lock);
1331 #ifdef CONFIG_IP_MULTICAST
1332 	setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1333 	im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1334 #endif
1335 
1336 	im->next_rcu = in_dev->mc_list;
1337 	in_dev->mc_count++;
1338 	rcu_assign_pointer(in_dev->mc_list, im);
1339 
1340 	ip_mc_hash_add(in_dev, im);
1341 
1342 #ifdef CONFIG_IP_MULTICAST
1343 	igmpv3_del_delrec(in_dev, im->multiaddr);
1344 #endif
1345 	igmp_group_added(im);
1346 	if (!in_dev->dead)
1347 		ip_rt_multicast_event(in_dev);
1348 out:
1349 	return;
1350 }
1351 EXPORT_SYMBOL(ip_mc_inc_group);
1352 
1353 /*
1354  *	Resend IGMP JOIN report; used by netdev notifier.
1355  */
1356 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1357 {
1358 #ifdef CONFIG_IP_MULTICAST
1359 	struct ip_mc_list *im;
1360 	int type;
1361 
1362 	ASSERT_RTNL();
1363 
1364 	for_each_pmc_rtnl(in_dev, im) {
1365 		if (im->multiaddr == IGMP_ALL_HOSTS)
1366 			continue;
1367 
1368 		/* a failover is happening and switches
1369 		 * must be notified immediately
1370 		 */
1371 		if (IGMP_V1_SEEN(in_dev))
1372 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1373 		else if (IGMP_V2_SEEN(in_dev))
1374 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1375 		else
1376 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1377 		igmp_send_report(in_dev, im, type);
1378 	}
1379 #endif
1380 }
1381 
1382 /*
1383  *	A socket has left a multicast group on device dev
1384  */
1385 
1386 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1387 {
1388 	struct ip_mc_list *i;
1389 	struct ip_mc_list __rcu **ip;
1390 
1391 	ASSERT_RTNL();
1392 
1393 	for (ip = &in_dev->mc_list;
1394 	     (i = rtnl_dereference(*ip)) != NULL;
1395 	     ip = &i->next_rcu) {
1396 		if (i->multiaddr == addr) {
1397 			if (--i->users == 0) {
1398 				ip_mc_hash_remove(in_dev, i);
1399 				*ip = i->next_rcu;
1400 				in_dev->mc_count--;
1401 				igmp_group_dropped(i);
1402 				ip_mc_clear_src(i);
1403 
1404 				if (!in_dev->dead)
1405 					ip_rt_multicast_event(in_dev);
1406 
1407 				ip_ma_put(i);
1408 				return;
1409 			}
1410 			break;
1411 		}
1412 	}
1413 }
1414 EXPORT_SYMBOL(ip_mc_dec_group);
1415 
1416 /* Device changing type */
1417 
1418 void ip_mc_unmap(struct in_device *in_dev)
1419 {
1420 	struct ip_mc_list *pmc;
1421 
1422 	ASSERT_RTNL();
1423 
1424 	for_each_pmc_rtnl(in_dev, pmc)
1425 		igmp_group_dropped(pmc);
1426 }
1427 
1428 void ip_mc_remap(struct in_device *in_dev)
1429 {
1430 	struct ip_mc_list *pmc;
1431 
1432 	ASSERT_RTNL();
1433 
1434 	for_each_pmc_rtnl(in_dev, pmc)
1435 		igmp_group_added(pmc);
1436 }
1437 
1438 /* Device going down */
1439 
1440 void ip_mc_down(struct in_device *in_dev)
1441 {
1442 	struct ip_mc_list *pmc;
1443 
1444 	ASSERT_RTNL();
1445 
1446 	for_each_pmc_rtnl(in_dev, pmc)
1447 		igmp_group_dropped(pmc);
1448 
1449 #ifdef CONFIG_IP_MULTICAST
1450 	in_dev->mr_ifc_count = 0;
1451 	if (del_timer(&in_dev->mr_ifc_timer))
1452 		__in_dev_put(in_dev);
1453 	in_dev->mr_gq_running = 0;
1454 	if (del_timer(&in_dev->mr_gq_timer))
1455 		__in_dev_put(in_dev);
1456 	igmpv3_clear_delrec(in_dev);
1457 #endif
1458 
1459 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1460 }
1461 
1462 void ip_mc_init_dev(struct in_device *in_dev)
1463 {
1464 	ASSERT_RTNL();
1465 
1466 #ifdef CONFIG_IP_MULTICAST
1467 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1468 			(unsigned long)in_dev);
1469 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1470 			(unsigned long)in_dev);
1471 	in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1472 #endif
1473 
1474 	spin_lock_init(&in_dev->mc_tomb_lock);
1475 }
1476 
1477 /* Device going up */
1478 
1479 void ip_mc_up(struct in_device *in_dev)
1480 {
1481 	struct ip_mc_list *pmc;
1482 
1483 	ASSERT_RTNL();
1484 
1485 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1486 
1487 	for_each_pmc_rtnl(in_dev, pmc)
1488 		igmp_group_added(pmc);
1489 }
1490 
1491 /*
1492  *	Device is about to be destroyed: clean up.
1493  */
1494 
1495 void ip_mc_destroy_dev(struct in_device *in_dev)
1496 {
1497 	struct ip_mc_list *i;
1498 
1499 	ASSERT_RTNL();
1500 
1501 	/* Deactivate timers */
1502 	ip_mc_down(in_dev);
1503 
1504 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1505 		in_dev->mc_list = i->next_rcu;
1506 		in_dev->mc_count--;
1507 
1508 		/* We've dropped the groups in ip_mc_down already */
1509 		ip_mc_clear_src(i);
1510 		ip_ma_put(i);
1511 	}
1512 }
1513 
1514 /* RTNL is locked */
1515 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1516 {
1517 	struct net_device *dev = NULL;
1518 	struct in_device *idev = NULL;
1519 
1520 	if (imr->imr_ifindex) {
1521 		idev = inetdev_by_index(net, imr->imr_ifindex);
1522 		return idev;
1523 	}
1524 	if (imr->imr_address.s_addr) {
1525 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1526 		if (!dev)
1527 			return NULL;
1528 	}
1529 
1530 	if (!dev) {
1531 		struct rtable *rt = ip_route_output(net,
1532 						    imr->imr_multiaddr.s_addr,
1533 						    0, 0, 0);
1534 		if (!IS_ERR(rt)) {
1535 			dev = rt->dst.dev;
1536 			ip_rt_put(rt);
1537 		}
1538 	}
1539 	if (dev) {
1540 		imr->imr_ifindex = dev->ifindex;
1541 		idev = __in_dev_get_rtnl(dev);
1542 	}
1543 	return idev;
1544 }
1545 
1546 /*
1547  *	Join a socket to a group
1548  */
1549 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1550 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1551 
1552 
1553 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1554 	__be32 *psfsrc)
1555 {
1556 	struct ip_sf_list *psf, *psf_prev;
1557 	int rv = 0;
1558 
1559 	psf_prev = NULL;
1560 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1561 		if (psf->sf_inaddr == *psfsrc)
1562 			break;
1563 		psf_prev = psf;
1564 	}
1565 	if (!psf || psf->sf_count[sfmode] == 0) {
1566 		/* source filter not found, or count wrong =>  bug */
1567 		return -ESRCH;
1568 	}
1569 	psf->sf_count[sfmode]--;
1570 	if (psf->sf_count[sfmode] == 0) {
1571 		ip_rt_multicast_event(pmc->interface);
1572 	}
1573 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1574 #ifdef CONFIG_IP_MULTICAST
1575 		struct in_device *in_dev = pmc->interface;
1576 #endif
1577 
1578 		/* no more filters for this source */
1579 		if (psf_prev)
1580 			psf_prev->sf_next = psf->sf_next;
1581 		else
1582 			pmc->sources = psf->sf_next;
1583 #ifdef CONFIG_IP_MULTICAST
1584 		if (psf->sf_oldin &&
1585 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1586 			psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1587 				IGMP_Unsolicited_Report_Count;
1588 			psf->sf_next = pmc->tomb;
1589 			pmc->tomb = psf;
1590 			rv = 1;
1591 		} else
1592 #endif
1593 			kfree(psf);
1594 	}
1595 	return rv;
1596 }
1597 
1598 #ifndef CONFIG_IP_MULTICAST
1599 #define igmp_ifc_event(x)	do { } while (0)
1600 #endif
1601 
1602 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1603 			 int sfcount, __be32 *psfsrc, int delta)
1604 {
1605 	struct ip_mc_list *pmc;
1606 	int	changerec = 0;
1607 	int	i, err;
1608 
1609 	if (!in_dev)
1610 		return -ENODEV;
1611 	rcu_read_lock();
1612 	for_each_pmc_rcu(in_dev, pmc) {
1613 		if (*pmca == pmc->multiaddr)
1614 			break;
1615 	}
1616 	if (!pmc) {
1617 		/* MCA not found?? bug */
1618 		rcu_read_unlock();
1619 		return -ESRCH;
1620 	}
1621 	spin_lock_bh(&pmc->lock);
1622 	rcu_read_unlock();
1623 #ifdef CONFIG_IP_MULTICAST
1624 	sf_markstate(pmc);
1625 #endif
1626 	if (!delta) {
1627 		err = -EINVAL;
1628 		if (!pmc->sfcount[sfmode])
1629 			goto out_unlock;
1630 		pmc->sfcount[sfmode]--;
1631 	}
1632 	err = 0;
1633 	for (i=0; i<sfcount; i++) {
1634 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1635 
1636 		changerec |= rv > 0;
1637 		if (!err && rv < 0)
1638 			err = rv;
1639 	}
1640 	if (pmc->sfmode == MCAST_EXCLUDE &&
1641 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1642 	    pmc->sfcount[MCAST_INCLUDE]) {
1643 #ifdef CONFIG_IP_MULTICAST
1644 		struct ip_sf_list *psf;
1645 #endif
1646 
1647 		/* filter mode change */
1648 		pmc->sfmode = MCAST_INCLUDE;
1649 #ifdef CONFIG_IP_MULTICAST
1650 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1651 			IGMP_Unsolicited_Report_Count;
1652 		in_dev->mr_ifc_count = pmc->crcount;
1653 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1654 			psf->sf_crcount = 0;
1655 		igmp_ifc_event(pmc->interface);
1656 	} else if (sf_setstate(pmc) || changerec) {
1657 		igmp_ifc_event(pmc->interface);
1658 #endif
1659 	}
1660 out_unlock:
1661 	spin_unlock_bh(&pmc->lock);
1662 	return err;
1663 }
1664 
1665 /*
1666  * Add multicast single-source filter to the interface list
1667  */
1668 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1669 	__be32 *psfsrc)
1670 {
1671 	struct ip_sf_list *psf, *psf_prev;
1672 
1673 	psf_prev = NULL;
1674 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1675 		if (psf->sf_inaddr == *psfsrc)
1676 			break;
1677 		psf_prev = psf;
1678 	}
1679 	if (!psf) {
1680 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1681 		if (!psf)
1682 			return -ENOBUFS;
1683 		psf->sf_inaddr = *psfsrc;
1684 		if (psf_prev) {
1685 			psf_prev->sf_next = psf;
1686 		} else
1687 			pmc->sources = psf;
1688 	}
1689 	psf->sf_count[sfmode]++;
1690 	if (psf->sf_count[sfmode] == 1) {
1691 		ip_rt_multicast_event(pmc->interface);
1692 	}
1693 	return 0;
1694 }
1695 
1696 #ifdef CONFIG_IP_MULTICAST
1697 static void sf_markstate(struct ip_mc_list *pmc)
1698 {
1699 	struct ip_sf_list *psf;
1700 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1701 
1702 	for (psf=pmc->sources; psf; psf=psf->sf_next)
1703 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1704 			psf->sf_oldin = mca_xcount ==
1705 				psf->sf_count[MCAST_EXCLUDE] &&
1706 				!psf->sf_count[MCAST_INCLUDE];
1707 		} else
1708 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1709 }
1710 
1711 static int sf_setstate(struct ip_mc_list *pmc)
1712 {
1713 	struct ip_sf_list *psf, *dpsf;
1714 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1715 	int qrv = pmc->interface->mr_qrv;
1716 	int new_in, rv;
1717 
1718 	rv = 0;
1719 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1720 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1721 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1722 				!psf->sf_count[MCAST_INCLUDE];
1723 		} else
1724 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1725 		if (new_in) {
1726 			if (!psf->sf_oldin) {
1727 				struct ip_sf_list *prev = NULL;
1728 
1729 				for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1730 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1731 						break;
1732 					prev = dpsf;
1733 				}
1734 				if (dpsf) {
1735 					if (prev)
1736 						prev->sf_next = dpsf->sf_next;
1737 					else
1738 						pmc->tomb = dpsf->sf_next;
1739 					kfree(dpsf);
1740 				}
1741 				psf->sf_crcount = qrv;
1742 				rv++;
1743 			}
1744 		} else if (psf->sf_oldin) {
1745 
1746 			psf->sf_crcount = 0;
1747 			/*
1748 			 * add or update "delete" records if an active filter
1749 			 * is now inactive
1750 			 */
1751 			for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1752 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1753 					break;
1754 			if (!dpsf) {
1755 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1756 				if (!dpsf)
1757 					continue;
1758 				*dpsf = *psf;
1759 				/* pmc->lock held by callers */
1760 				dpsf->sf_next = pmc->tomb;
1761 				pmc->tomb = dpsf;
1762 			}
1763 			dpsf->sf_crcount = qrv;
1764 			rv++;
1765 		}
1766 	}
1767 	return rv;
1768 }
1769 #endif
1770 
1771 /*
1772  * Add multicast source filter list to the interface list
1773  */
1774 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1775 			 int sfcount, __be32 *psfsrc, int delta)
1776 {
1777 	struct ip_mc_list *pmc;
1778 	int	isexclude;
1779 	int	i, err;
1780 
1781 	if (!in_dev)
1782 		return -ENODEV;
1783 	rcu_read_lock();
1784 	for_each_pmc_rcu(in_dev, pmc) {
1785 		if (*pmca == pmc->multiaddr)
1786 			break;
1787 	}
1788 	if (!pmc) {
1789 		/* MCA not found?? bug */
1790 		rcu_read_unlock();
1791 		return -ESRCH;
1792 	}
1793 	spin_lock_bh(&pmc->lock);
1794 	rcu_read_unlock();
1795 
1796 #ifdef CONFIG_IP_MULTICAST
1797 	sf_markstate(pmc);
1798 #endif
1799 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
1800 	if (!delta)
1801 		pmc->sfcount[sfmode]++;
1802 	err = 0;
1803 	for (i=0; i<sfcount; i++) {
1804 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1805 		if (err)
1806 			break;
1807 	}
1808 	if (err) {
1809 		int j;
1810 
1811 		if (!delta)
1812 			pmc->sfcount[sfmode]--;
1813 		for (j=0; j<i; j++)
1814 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1815 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1816 #ifdef CONFIG_IP_MULTICAST
1817 		struct ip_sf_list *psf;
1818 		in_dev = pmc->interface;
1819 #endif
1820 
1821 		/* filter mode change */
1822 		if (pmc->sfcount[MCAST_EXCLUDE])
1823 			pmc->sfmode = MCAST_EXCLUDE;
1824 		else if (pmc->sfcount[MCAST_INCLUDE])
1825 			pmc->sfmode = MCAST_INCLUDE;
1826 #ifdef CONFIG_IP_MULTICAST
1827 		/* else no filters; keep old mode for reports */
1828 
1829 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1830 			IGMP_Unsolicited_Report_Count;
1831 		in_dev->mr_ifc_count = pmc->crcount;
1832 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1833 			psf->sf_crcount = 0;
1834 		igmp_ifc_event(in_dev);
1835 	} else if (sf_setstate(pmc)) {
1836 		igmp_ifc_event(in_dev);
1837 #endif
1838 	}
1839 	spin_unlock_bh(&pmc->lock);
1840 	return err;
1841 }
1842 
1843 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1844 {
1845 	struct ip_sf_list *psf, *nextpsf;
1846 
1847 	for (psf=pmc->tomb; psf; psf=nextpsf) {
1848 		nextpsf = psf->sf_next;
1849 		kfree(psf);
1850 	}
1851 	pmc->tomb = NULL;
1852 	for (psf=pmc->sources; psf; psf=nextpsf) {
1853 		nextpsf = psf->sf_next;
1854 		kfree(psf);
1855 	}
1856 	pmc->sources = NULL;
1857 	pmc->sfmode = MCAST_EXCLUDE;
1858 	pmc->sfcount[MCAST_INCLUDE] = 0;
1859 	pmc->sfcount[MCAST_EXCLUDE] = 1;
1860 }
1861 
1862 
1863 /*
1864  * Join a multicast group
1865  */
1866 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1867 {
1868 	int err;
1869 	__be32 addr = imr->imr_multiaddr.s_addr;
1870 	struct ip_mc_socklist *iml = NULL, *i;
1871 	struct in_device *in_dev;
1872 	struct inet_sock *inet = inet_sk(sk);
1873 	struct net *net = sock_net(sk);
1874 	int ifindex;
1875 	int count = 0;
1876 
1877 	if (!ipv4_is_multicast(addr))
1878 		return -EINVAL;
1879 
1880 	rtnl_lock();
1881 
1882 	in_dev = ip_mc_find_dev(net, imr);
1883 
1884 	if (!in_dev) {
1885 		iml = NULL;
1886 		err = -ENODEV;
1887 		goto done;
1888 	}
1889 
1890 	err = -EADDRINUSE;
1891 	ifindex = imr->imr_ifindex;
1892 	for_each_pmc_rtnl(inet, i) {
1893 		if (i->multi.imr_multiaddr.s_addr == addr &&
1894 		    i->multi.imr_ifindex == ifindex)
1895 			goto done;
1896 		count++;
1897 	}
1898 	err = -ENOBUFS;
1899 	if (count >= sysctl_igmp_max_memberships)
1900 		goto done;
1901 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1902 	if (iml == NULL)
1903 		goto done;
1904 
1905 	memcpy(&iml->multi, imr, sizeof(*imr));
1906 	iml->next_rcu = inet->mc_list;
1907 	iml->sflist = NULL;
1908 	iml->sfmode = MCAST_EXCLUDE;
1909 	rcu_assign_pointer(inet->mc_list, iml);
1910 	ip_mc_inc_group(in_dev, addr);
1911 	err = 0;
1912 done:
1913 	rtnl_unlock();
1914 	return err;
1915 }
1916 EXPORT_SYMBOL(ip_mc_join_group);
1917 
1918 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1919 			   struct in_device *in_dev)
1920 {
1921 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1922 	int err;
1923 
1924 	if (psf == NULL) {
1925 		/* any-source empty exclude case */
1926 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1927 			iml->sfmode, 0, NULL, 0);
1928 	}
1929 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1930 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1931 	RCU_INIT_POINTER(iml->sflist, NULL);
1932 	/* decrease mem now to avoid the memleak warning */
1933 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1934 	kfree_rcu(psf, rcu);
1935 	return err;
1936 }
1937 
1938 /*
1939  *	Ask a socket to leave a group.
1940  */
1941 
1942 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1943 {
1944 	struct inet_sock *inet = inet_sk(sk);
1945 	struct ip_mc_socklist *iml;
1946 	struct ip_mc_socklist __rcu **imlp;
1947 	struct in_device *in_dev;
1948 	struct net *net = sock_net(sk);
1949 	__be32 group = imr->imr_multiaddr.s_addr;
1950 	u32 ifindex;
1951 	int ret = -EADDRNOTAVAIL;
1952 
1953 	rtnl_lock();
1954 	in_dev = ip_mc_find_dev(net, imr);
1955 	ifindex = imr->imr_ifindex;
1956 	for (imlp = &inet->mc_list;
1957 	     (iml = rtnl_dereference(*imlp)) != NULL;
1958 	     imlp = &iml->next_rcu) {
1959 		if (iml->multi.imr_multiaddr.s_addr != group)
1960 			continue;
1961 		if (ifindex) {
1962 			if (iml->multi.imr_ifindex != ifindex)
1963 				continue;
1964 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1965 				iml->multi.imr_address.s_addr)
1966 			continue;
1967 
1968 		(void) ip_mc_leave_src(sk, iml, in_dev);
1969 
1970 		*imlp = iml->next_rcu;
1971 
1972 		if (in_dev)
1973 			ip_mc_dec_group(in_dev, group);
1974 		rtnl_unlock();
1975 		/* decrease mem now to avoid the memleak warning */
1976 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1977 		kfree_rcu(iml, rcu);
1978 		return 0;
1979 	}
1980 	if (!in_dev)
1981 		ret = -ENODEV;
1982 	rtnl_unlock();
1983 	return ret;
1984 }
1985 EXPORT_SYMBOL(ip_mc_leave_group);
1986 
1987 int ip_mc_source(int add, int omode, struct sock *sk, struct
1988 	ip_mreq_source *mreqs, int ifindex)
1989 {
1990 	int err;
1991 	struct ip_mreqn imr;
1992 	__be32 addr = mreqs->imr_multiaddr;
1993 	struct ip_mc_socklist *pmc;
1994 	struct in_device *in_dev = NULL;
1995 	struct inet_sock *inet = inet_sk(sk);
1996 	struct ip_sf_socklist *psl;
1997 	struct net *net = sock_net(sk);
1998 	int leavegroup = 0;
1999 	int i, j, rv;
2000 
2001 	if (!ipv4_is_multicast(addr))
2002 		return -EINVAL;
2003 
2004 	rtnl_lock();
2005 
2006 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2007 	imr.imr_address.s_addr = mreqs->imr_interface;
2008 	imr.imr_ifindex = ifindex;
2009 	in_dev = ip_mc_find_dev(net, &imr);
2010 
2011 	if (!in_dev) {
2012 		err = -ENODEV;
2013 		goto done;
2014 	}
2015 	err = -EADDRNOTAVAIL;
2016 
2017 	for_each_pmc_rtnl(inet, pmc) {
2018 		if ((pmc->multi.imr_multiaddr.s_addr ==
2019 		     imr.imr_multiaddr.s_addr) &&
2020 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
2021 			break;
2022 	}
2023 	if (!pmc) {		/* must have a prior join */
2024 		err = -EINVAL;
2025 		goto done;
2026 	}
2027 	/* if a source filter was set, must be the same mode as before */
2028 	if (pmc->sflist) {
2029 		if (pmc->sfmode != omode) {
2030 			err = -EINVAL;
2031 			goto done;
2032 		}
2033 	} else if (pmc->sfmode != omode) {
2034 		/* allow mode switches for empty-set filters */
2035 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2036 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2037 			NULL, 0);
2038 		pmc->sfmode = omode;
2039 	}
2040 
2041 	psl = rtnl_dereference(pmc->sflist);
2042 	if (!add) {
2043 		if (!psl)
2044 			goto done;	/* err = -EADDRNOTAVAIL */
2045 		rv = !0;
2046 		for (i=0; i<psl->sl_count; i++) {
2047 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2048 				sizeof(__be32));
2049 			if (rv == 0)
2050 				break;
2051 		}
2052 		if (rv)		/* source not found */
2053 			goto done;	/* err = -EADDRNOTAVAIL */
2054 
2055 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2056 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2057 			leavegroup = 1;
2058 			goto done;
2059 		}
2060 
2061 		/* update the interface filter */
2062 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2063 			&mreqs->imr_sourceaddr, 1);
2064 
2065 		for (j=i+1; j<psl->sl_count; j++)
2066 			psl->sl_addr[j-1] = psl->sl_addr[j];
2067 		psl->sl_count--;
2068 		err = 0;
2069 		goto done;
2070 	}
2071 	/* else, add a new source to the filter */
2072 
2073 	if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2074 		err = -ENOBUFS;
2075 		goto done;
2076 	}
2077 	if (!psl || psl->sl_count == psl->sl_max) {
2078 		struct ip_sf_socklist *newpsl;
2079 		int count = IP_SFBLOCK;
2080 
2081 		if (psl)
2082 			count += psl->sl_max;
2083 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2084 		if (!newpsl) {
2085 			err = -ENOBUFS;
2086 			goto done;
2087 		}
2088 		newpsl->sl_max = count;
2089 		newpsl->sl_count = count - IP_SFBLOCK;
2090 		if (psl) {
2091 			for (i=0; i<psl->sl_count; i++)
2092 				newpsl->sl_addr[i] = psl->sl_addr[i];
2093 			/* decrease mem now to avoid the memleak warning */
2094 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2095 			kfree_rcu(psl, rcu);
2096 		}
2097 		rcu_assign_pointer(pmc->sflist, newpsl);
2098 		psl = newpsl;
2099 	}
2100 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2101 	for (i=0; i<psl->sl_count; i++) {
2102 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2103 			sizeof(__be32));
2104 		if (rv == 0)
2105 			break;
2106 	}
2107 	if (rv == 0)		/* address already there is an error */
2108 		goto done;
2109 	for (j=psl->sl_count-1; j>=i; j--)
2110 		psl->sl_addr[j+1] = psl->sl_addr[j];
2111 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2112 	psl->sl_count++;
2113 	err = 0;
2114 	/* update the interface list */
2115 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2116 		&mreqs->imr_sourceaddr, 1);
2117 done:
2118 	rtnl_unlock();
2119 	if (leavegroup)
2120 		return ip_mc_leave_group(sk, &imr);
2121 	return err;
2122 }
2123 
2124 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2125 {
2126 	int err = 0;
2127 	struct ip_mreqn	imr;
2128 	__be32 addr = msf->imsf_multiaddr;
2129 	struct ip_mc_socklist *pmc;
2130 	struct in_device *in_dev;
2131 	struct inet_sock *inet = inet_sk(sk);
2132 	struct ip_sf_socklist *newpsl, *psl;
2133 	struct net *net = sock_net(sk);
2134 	int leavegroup = 0;
2135 
2136 	if (!ipv4_is_multicast(addr))
2137 		return -EINVAL;
2138 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2139 	    msf->imsf_fmode != MCAST_EXCLUDE)
2140 		return -EINVAL;
2141 
2142 	rtnl_lock();
2143 
2144 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2145 	imr.imr_address.s_addr = msf->imsf_interface;
2146 	imr.imr_ifindex = ifindex;
2147 	in_dev = ip_mc_find_dev(net, &imr);
2148 
2149 	if (!in_dev) {
2150 		err = -ENODEV;
2151 		goto done;
2152 	}
2153 
2154 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2155 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2156 		leavegroup = 1;
2157 		goto done;
2158 	}
2159 
2160 	for_each_pmc_rtnl(inet, pmc) {
2161 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2162 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2163 			break;
2164 	}
2165 	if (!pmc) {		/* must have a prior join */
2166 		err = -EINVAL;
2167 		goto done;
2168 	}
2169 	if (msf->imsf_numsrc) {
2170 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2171 							   GFP_KERNEL);
2172 		if (!newpsl) {
2173 			err = -ENOBUFS;
2174 			goto done;
2175 		}
2176 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2177 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2178 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2179 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2180 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2181 		if (err) {
2182 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2183 			goto done;
2184 		}
2185 	} else {
2186 		newpsl = NULL;
2187 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2188 				     msf->imsf_fmode, 0, NULL, 0);
2189 	}
2190 	psl = rtnl_dereference(pmc->sflist);
2191 	if (psl) {
2192 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2193 			psl->sl_count, psl->sl_addr, 0);
2194 		/* decrease mem now to avoid the memleak warning */
2195 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2196 		kfree_rcu(psl, rcu);
2197 	} else
2198 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2199 			0, NULL, 0);
2200 	rcu_assign_pointer(pmc->sflist, newpsl);
2201 	pmc->sfmode = msf->imsf_fmode;
2202 	err = 0;
2203 done:
2204 	rtnl_unlock();
2205 	if (leavegroup)
2206 		err = ip_mc_leave_group(sk, &imr);
2207 	return err;
2208 }
2209 
2210 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2211 	struct ip_msfilter __user *optval, int __user *optlen)
2212 {
2213 	int err, len, count, copycount;
2214 	struct ip_mreqn	imr;
2215 	__be32 addr = msf->imsf_multiaddr;
2216 	struct ip_mc_socklist *pmc;
2217 	struct in_device *in_dev;
2218 	struct inet_sock *inet = inet_sk(sk);
2219 	struct ip_sf_socklist *psl;
2220 	struct net *net = sock_net(sk);
2221 
2222 	if (!ipv4_is_multicast(addr))
2223 		return -EINVAL;
2224 
2225 	rtnl_lock();
2226 
2227 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2228 	imr.imr_address.s_addr = msf->imsf_interface;
2229 	imr.imr_ifindex = 0;
2230 	in_dev = ip_mc_find_dev(net, &imr);
2231 
2232 	if (!in_dev) {
2233 		err = -ENODEV;
2234 		goto done;
2235 	}
2236 	err = -EADDRNOTAVAIL;
2237 
2238 	for_each_pmc_rtnl(inet, pmc) {
2239 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2240 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2241 			break;
2242 	}
2243 	if (!pmc)		/* must have a prior join */
2244 		goto done;
2245 	msf->imsf_fmode = pmc->sfmode;
2246 	psl = rtnl_dereference(pmc->sflist);
2247 	rtnl_unlock();
2248 	if (!psl) {
2249 		len = 0;
2250 		count = 0;
2251 	} else {
2252 		count = psl->sl_count;
2253 	}
2254 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2255 	len = copycount * sizeof(psl->sl_addr[0]);
2256 	msf->imsf_numsrc = count;
2257 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2258 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2259 		return -EFAULT;
2260 	}
2261 	if (len &&
2262 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2263 		return -EFAULT;
2264 	return 0;
2265 done:
2266 	rtnl_unlock();
2267 	return err;
2268 }
2269 
2270 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2271 	struct group_filter __user *optval, int __user *optlen)
2272 {
2273 	int err, i, count, copycount;
2274 	struct sockaddr_in *psin;
2275 	__be32 addr;
2276 	struct ip_mc_socklist *pmc;
2277 	struct inet_sock *inet = inet_sk(sk);
2278 	struct ip_sf_socklist *psl;
2279 
2280 	psin = (struct sockaddr_in *)&gsf->gf_group;
2281 	if (psin->sin_family != AF_INET)
2282 		return -EINVAL;
2283 	addr = psin->sin_addr.s_addr;
2284 	if (!ipv4_is_multicast(addr))
2285 		return -EINVAL;
2286 
2287 	rtnl_lock();
2288 
2289 	err = -EADDRNOTAVAIL;
2290 
2291 	for_each_pmc_rtnl(inet, pmc) {
2292 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2293 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2294 			break;
2295 	}
2296 	if (!pmc)		/* must have a prior join */
2297 		goto done;
2298 	gsf->gf_fmode = pmc->sfmode;
2299 	psl = rtnl_dereference(pmc->sflist);
2300 	rtnl_unlock();
2301 	count = psl ? psl->sl_count : 0;
2302 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2303 	gsf->gf_numsrc = count;
2304 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2305 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2306 		return -EFAULT;
2307 	}
2308 	for (i=0; i<copycount; i++) {
2309 		struct sockaddr_storage ss;
2310 
2311 		psin = (struct sockaddr_in *)&ss;
2312 		memset(&ss, 0, sizeof(ss));
2313 		psin->sin_family = AF_INET;
2314 		psin->sin_addr.s_addr = psl->sl_addr[i];
2315 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2316 			return -EFAULT;
2317 	}
2318 	return 0;
2319 done:
2320 	rtnl_unlock();
2321 	return err;
2322 }
2323 
2324 /*
2325  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2326  */
2327 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2328 {
2329 	struct inet_sock *inet = inet_sk(sk);
2330 	struct ip_mc_socklist *pmc;
2331 	struct ip_sf_socklist *psl;
2332 	int i;
2333 	int ret;
2334 
2335 	ret = 1;
2336 	if (!ipv4_is_multicast(loc_addr))
2337 		goto out;
2338 
2339 	rcu_read_lock();
2340 	for_each_pmc_rcu(inet, pmc) {
2341 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2342 		    pmc->multi.imr_ifindex == dif)
2343 			break;
2344 	}
2345 	ret = inet->mc_all;
2346 	if (!pmc)
2347 		goto unlock;
2348 	psl = rcu_dereference(pmc->sflist);
2349 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2350 	if (!psl)
2351 		goto unlock;
2352 
2353 	for (i=0; i<psl->sl_count; i++) {
2354 		if (psl->sl_addr[i] == rmt_addr)
2355 			break;
2356 	}
2357 	ret = 0;
2358 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2359 		goto unlock;
2360 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2361 		goto unlock;
2362 	ret = 1;
2363 unlock:
2364 	rcu_read_unlock();
2365 out:
2366 	return ret;
2367 }
2368 
2369 /*
2370  *	A socket is closing.
2371  */
2372 
2373 void ip_mc_drop_socket(struct sock *sk)
2374 {
2375 	struct inet_sock *inet = inet_sk(sk);
2376 	struct ip_mc_socklist *iml;
2377 	struct net *net = sock_net(sk);
2378 
2379 	if (inet->mc_list == NULL)
2380 		return;
2381 
2382 	rtnl_lock();
2383 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2384 		struct in_device *in_dev;
2385 
2386 		inet->mc_list = iml->next_rcu;
2387 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2388 		(void) ip_mc_leave_src(sk, iml, in_dev);
2389 		if (in_dev != NULL)
2390 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2391 		/* decrease mem now to avoid the memleak warning */
2392 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2393 		kfree_rcu(iml, rcu);
2394 	}
2395 	rtnl_unlock();
2396 }
2397 
2398 /* called with rcu_read_lock() */
2399 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2400 {
2401 	struct ip_mc_list *im;
2402 	struct ip_mc_list __rcu **mc_hash;
2403 	struct ip_sf_list *psf;
2404 	int rv = 0;
2405 
2406 	mc_hash = rcu_dereference(in_dev->mc_hash);
2407 	if (mc_hash) {
2408 		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2409 
2410 		for (im = rcu_dereference(mc_hash[hash]);
2411 		     im != NULL;
2412 		     im = rcu_dereference(im->next_hash)) {
2413 			if (im->multiaddr == mc_addr)
2414 				break;
2415 		}
2416 	} else {
2417 		for_each_pmc_rcu(in_dev, im) {
2418 			if (im->multiaddr == mc_addr)
2419 				break;
2420 		}
2421 	}
2422 	if (im && proto == IPPROTO_IGMP) {
2423 		rv = 1;
2424 	} else if (im) {
2425 		if (src_addr) {
2426 			for (psf=im->sources; psf; psf=psf->sf_next) {
2427 				if (psf->sf_inaddr == src_addr)
2428 					break;
2429 			}
2430 			if (psf)
2431 				rv = psf->sf_count[MCAST_INCLUDE] ||
2432 					psf->sf_count[MCAST_EXCLUDE] !=
2433 					im->sfcount[MCAST_EXCLUDE];
2434 			else
2435 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2436 		} else
2437 			rv = 1; /* unspecified source; tentatively allow */
2438 	}
2439 	return rv;
2440 }
2441 
2442 #if defined(CONFIG_PROC_FS)
2443 struct igmp_mc_iter_state {
2444 	struct seq_net_private p;
2445 	struct net_device *dev;
2446 	struct in_device *in_dev;
2447 };
2448 
2449 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2450 
2451 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2452 {
2453 	struct net *net = seq_file_net(seq);
2454 	struct ip_mc_list *im = NULL;
2455 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2456 
2457 	state->in_dev = NULL;
2458 	for_each_netdev_rcu(net, state->dev) {
2459 		struct in_device *in_dev;
2460 
2461 		in_dev = __in_dev_get_rcu(state->dev);
2462 		if (!in_dev)
2463 			continue;
2464 		im = rcu_dereference(in_dev->mc_list);
2465 		if (im) {
2466 			state->in_dev = in_dev;
2467 			break;
2468 		}
2469 	}
2470 	return im;
2471 }
2472 
2473 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2474 {
2475 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2476 
2477 	im = rcu_dereference(im->next_rcu);
2478 	while (!im) {
2479 		state->dev = next_net_device_rcu(state->dev);
2480 		if (!state->dev) {
2481 			state->in_dev = NULL;
2482 			break;
2483 		}
2484 		state->in_dev = __in_dev_get_rcu(state->dev);
2485 		if (!state->in_dev)
2486 			continue;
2487 		im = rcu_dereference(state->in_dev->mc_list);
2488 	}
2489 	return im;
2490 }
2491 
2492 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2493 {
2494 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2495 	if (im)
2496 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2497 			--pos;
2498 	return pos ? NULL : im;
2499 }
2500 
2501 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2502 	__acquires(rcu)
2503 {
2504 	rcu_read_lock();
2505 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2506 }
2507 
2508 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2509 {
2510 	struct ip_mc_list *im;
2511 	if (v == SEQ_START_TOKEN)
2512 		im = igmp_mc_get_first(seq);
2513 	else
2514 		im = igmp_mc_get_next(seq, v);
2515 	++*pos;
2516 	return im;
2517 }
2518 
2519 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2520 	__releases(rcu)
2521 {
2522 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2523 
2524 	state->in_dev = NULL;
2525 	state->dev = NULL;
2526 	rcu_read_unlock();
2527 }
2528 
2529 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2530 {
2531 	if (v == SEQ_START_TOKEN)
2532 		seq_puts(seq,
2533 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2534 	else {
2535 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2536 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2537 		char   *querier;
2538 		long delta;
2539 
2540 #ifdef CONFIG_IP_MULTICAST
2541 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2542 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2543 			  "V3";
2544 #else
2545 		querier = "NONE";
2546 #endif
2547 
2548 		if (rcu_dereference(state->in_dev->mc_list) == im) {
2549 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2550 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2551 		}
2552 
2553 		delta = im->timer.expires - jiffies;
2554 		seq_printf(seq,
2555 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2556 			   im->multiaddr, im->users,
2557 			   im->tm_running,
2558 			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2559 			   im->reporter);
2560 	}
2561 	return 0;
2562 }
2563 
2564 static const struct seq_operations igmp_mc_seq_ops = {
2565 	.start	=	igmp_mc_seq_start,
2566 	.next	=	igmp_mc_seq_next,
2567 	.stop	=	igmp_mc_seq_stop,
2568 	.show	=	igmp_mc_seq_show,
2569 };
2570 
2571 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2572 {
2573 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2574 			sizeof(struct igmp_mc_iter_state));
2575 }
2576 
2577 static const struct file_operations igmp_mc_seq_fops = {
2578 	.owner		=	THIS_MODULE,
2579 	.open		=	igmp_mc_seq_open,
2580 	.read		=	seq_read,
2581 	.llseek		=	seq_lseek,
2582 	.release	=	seq_release_net,
2583 };
2584 
2585 struct igmp_mcf_iter_state {
2586 	struct seq_net_private p;
2587 	struct net_device *dev;
2588 	struct in_device *idev;
2589 	struct ip_mc_list *im;
2590 };
2591 
2592 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2593 
2594 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2595 {
2596 	struct net *net = seq_file_net(seq);
2597 	struct ip_sf_list *psf = NULL;
2598 	struct ip_mc_list *im = NULL;
2599 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2600 
2601 	state->idev = NULL;
2602 	state->im = NULL;
2603 	for_each_netdev_rcu(net, state->dev) {
2604 		struct in_device *idev;
2605 		idev = __in_dev_get_rcu(state->dev);
2606 		if (unlikely(idev == NULL))
2607 			continue;
2608 		im = rcu_dereference(idev->mc_list);
2609 		if (likely(im != NULL)) {
2610 			spin_lock_bh(&im->lock);
2611 			psf = im->sources;
2612 			if (likely(psf != NULL)) {
2613 				state->im = im;
2614 				state->idev = idev;
2615 				break;
2616 			}
2617 			spin_unlock_bh(&im->lock);
2618 		}
2619 	}
2620 	return psf;
2621 }
2622 
2623 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2624 {
2625 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2626 
2627 	psf = psf->sf_next;
2628 	while (!psf) {
2629 		spin_unlock_bh(&state->im->lock);
2630 		state->im = state->im->next;
2631 		while (!state->im) {
2632 			state->dev = next_net_device_rcu(state->dev);
2633 			if (!state->dev) {
2634 				state->idev = NULL;
2635 				goto out;
2636 			}
2637 			state->idev = __in_dev_get_rcu(state->dev);
2638 			if (!state->idev)
2639 				continue;
2640 			state->im = rcu_dereference(state->idev->mc_list);
2641 		}
2642 		if (!state->im)
2643 			break;
2644 		spin_lock_bh(&state->im->lock);
2645 		psf = state->im->sources;
2646 	}
2647 out:
2648 	return psf;
2649 }
2650 
2651 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2652 {
2653 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2654 	if (psf)
2655 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2656 			--pos;
2657 	return pos ? NULL : psf;
2658 }
2659 
2660 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2661 	__acquires(rcu)
2662 {
2663 	rcu_read_lock();
2664 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2665 }
2666 
2667 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2668 {
2669 	struct ip_sf_list *psf;
2670 	if (v == SEQ_START_TOKEN)
2671 		psf = igmp_mcf_get_first(seq);
2672 	else
2673 		psf = igmp_mcf_get_next(seq, v);
2674 	++*pos;
2675 	return psf;
2676 }
2677 
2678 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2679 	__releases(rcu)
2680 {
2681 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2682 	if (likely(state->im != NULL)) {
2683 		spin_unlock_bh(&state->im->lock);
2684 		state->im = NULL;
2685 	}
2686 	state->idev = NULL;
2687 	state->dev = NULL;
2688 	rcu_read_unlock();
2689 }
2690 
2691 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2692 {
2693 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2694 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2695 
2696 	if (v == SEQ_START_TOKEN) {
2697 		seq_printf(seq,
2698 			   "%3s %6s "
2699 			   "%10s %10s %6s %6s\n", "Idx",
2700 			   "Device", "MCA",
2701 			   "SRC", "INC", "EXC");
2702 	} else {
2703 		seq_printf(seq,
2704 			   "%3d %6.6s 0x%08x "
2705 			   "0x%08x %6lu %6lu\n",
2706 			   state->dev->ifindex, state->dev->name,
2707 			   ntohl(state->im->multiaddr),
2708 			   ntohl(psf->sf_inaddr),
2709 			   psf->sf_count[MCAST_INCLUDE],
2710 			   psf->sf_count[MCAST_EXCLUDE]);
2711 	}
2712 	return 0;
2713 }
2714 
2715 static const struct seq_operations igmp_mcf_seq_ops = {
2716 	.start	=	igmp_mcf_seq_start,
2717 	.next	=	igmp_mcf_seq_next,
2718 	.stop	=	igmp_mcf_seq_stop,
2719 	.show	=	igmp_mcf_seq_show,
2720 };
2721 
2722 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2723 {
2724 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2725 			sizeof(struct igmp_mcf_iter_state));
2726 }
2727 
2728 static const struct file_operations igmp_mcf_seq_fops = {
2729 	.owner		=	THIS_MODULE,
2730 	.open		=	igmp_mcf_seq_open,
2731 	.read		=	seq_read,
2732 	.llseek		=	seq_lseek,
2733 	.release	=	seq_release_net,
2734 };
2735 
2736 static int __net_init igmp_net_init(struct net *net)
2737 {
2738 	struct proc_dir_entry *pde;
2739 
2740 	pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2741 	if (!pde)
2742 		goto out_igmp;
2743 	pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2744 			  &igmp_mcf_seq_fops);
2745 	if (!pde)
2746 		goto out_mcfilter;
2747 	return 0;
2748 
2749 out_mcfilter:
2750 	remove_proc_entry("igmp", net->proc_net);
2751 out_igmp:
2752 	return -ENOMEM;
2753 }
2754 
2755 static void __net_exit igmp_net_exit(struct net *net)
2756 {
2757 	remove_proc_entry("mcfilter", net->proc_net);
2758 	remove_proc_entry("igmp", net->proc_net);
2759 }
2760 
2761 static struct pernet_operations igmp_net_ops = {
2762 	.init = igmp_net_init,
2763 	.exit = igmp_net_exit,
2764 };
2765 
2766 static int igmp_netdev_event(struct notifier_block *this,
2767 			     unsigned long event, void *ptr)
2768 {
2769 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2770 	struct in_device *in_dev;
2771 
2772 	switch (event) {
2773 	case NETDEV_RESEND_IGMP:
2774 		in_dev = __in_dev_get_rtnl(dev);
2775 		if (in_dev)
2776 			ip_mc_rejoin_groups(in_dev);
2777 		break;
2778 	default:
2779 		break;
2780 	}
2781 	return NOTIFY_DONE;
2782 }
2783 
2784 static struct notifier_block igmp_notifier = {
2785 	.notifier_call = igmp_netdev_event,
2786 };
2787 
2788 int __init igmp_mc_proc_init(void)
2789 {
2790 	int err;
2791 
2792 	err = register_pernet_subsys(&igmp_net_ops);
2793 	if (err)
2794 		return err;
2795 	err = register_netdevice_notifier(&igmp_notifier);
2796 	if (err)
2797 		goto reg_notif_fail;
2798 	return 0;
2799 
2800 reg_notif_fail:
2801 	unregister_pernet_subsys(&igmp_net_ops);
2802 	return err;
2803 }
2804 #endif
2805