xref: /linux/net/ipv4/igmp.c (revision 9e8ba5f3ec35cba4fd8a8bebda548c4db2651e40)
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 <asm/system.h>
77 #include <linux/types.h>
78 #include <linux/kernel.h>
79 #include <linux/jiffies.h>
80 #include <linux/string.h>
81 #include <linux/socket.h>
82 #include <linux/sockios.h>
83 #include <linux/in.h>
84 #include <linux/inet.h>
85 #include <linux/netdevice.h>
86 #include <linux/skbuff.h>
87 #include <linux/inetdevice.h>
88 #include <linux/igmp.h>
89 #include <linux/if_arp.h>
90 #include <linux/rtnetlink.h>
91 #include <linux/times.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_Unsolicited_Report_Interval	(10*HZ)
118 #define IGMP_Query_Response_Interval		(10*HZ)
119 #define IGMP_Unsolicited_Report_Count		2
120 
121 
122 #define IGMP_Initial_Report_Delay		(1)
123 
124 /* IGMP_Initial_Report_Delay is not from IGMP specs!
125  * IGMP specs require to report membership immediately after
126  * joining a group, but we delay the first report by a
127  * small interval. It seems more natural and still does not
128  * contradict to specs provided this delay is small enough.
129  */
130 
131 #define IGMP_V1_SEEN(in_dev) \
132 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
133 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
134 	 ((in_dev)->mr_v1_seen && \
135 	  time_before(jiffies, (in_dev)->mr_v1_seen)))
136 #define IGMP_V2_SEEN(in_dev) \
137 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
138 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
139 	 ((in_dev)->mr_v2_seen && \
140 	  time_before(jiffies, (in_dev)->mr_v2_seen)))
141 
142 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
144 static void igmpv3_clear_delrec(struct in_device *in_dev);
145 static int sf_setstate(struct ip_mc_list *pmc);
146 static void sf_markstate(struct ip_mc_list *pmc);
147 #endif
148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
150 			 int sfcount, __be32 *psfsrc, int delta);
151 
152 static void ip_ma_put(struct ip_mc_list *im)
153 {
154 	if (atomic_dec_and_test(&im->refcnt)) {
155 		in_dev_put(im->interface);
156 		kfree_rcu(im, rcu);
157 	}
158 }
159 
160 #define for_each_pmc_rcu(in_dev, pmc)				\
161 	for (pmc = rcu_dereference(in_dev->mc_list);		\
162 	     pmc != NULL;					\
163 	     pmc = rcu_dereference(pmc->next_rcu))
164 
165 #define for_each_pmc_rtnl(in_dev, pmc)				\
166 	for (pmc = rtnl_dereference(in_dev->mc_list);		\
167 	     pmc != NULL;					\
168 	     pmc = rtnl_dereference(pmc->next_rcu))
169 
170 #ifdef CONFIG_IP_MULTICAST
171 
172 /*
173  *	Timer management
174  */
175 
176 static void igmp_stop_timer(struct ip_mc_list *im)
177 {
178 	spin_lock_bh(&im->lock);
179 	if (del_timer(&im->timer))
180 		atomic_dec(&im->refcnt);
181 	im->tm_running = 0;
182 	im->reporter = 0;
183 	im->unsolicit_count = 0;
184 	spin_unlock_bh(&im->lock);
185 }
186 
187 /* It must be called with locked im->lock */
188 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
189 {
190 	int tv = net_random() % max_delay;
191 
192 	im->tm_running = 1;
193 	if (!mod_timer(&im->timer, jiffies+tv+2))
194 		atomic_inc(&im->refcnt);
195 }
196 
197 static void igmp_gq_start_timer(struct in_device *in_dev)
198 {
199 	int tv = net_random() % in_dev->mr_maxdelay;
200 
201 	in_dev->mr_gq_running = 1;
202 	if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
203 		in_dev_hold(in_dev);
204 }
205 
206 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
207 {
208 	int tv = net_random() % delay;
209 
210 	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
211 		in_dev_hold(in_dev);
212 }
213 
214 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
215 {
216 	spin_lock_bh(&im->lock);
217 	im->unsolicit_count = 0;
218 	if (del_timer(&im->timer)) {
219 		if ((long)(im->timer.expires-jiffies) < max_delay) {
220 			add_timer(&im->timer);
221 			im->tm_running = 1;
222 			spin_unlock_bh(&im->lock);
223 			return;
224 		}
225 		atomic_dec(&im->refcnt);
226 	}
227 	igmp_start_timer(im, max_delay);
228 	spin_unlock_bh(&im->lock);
229 }
230 
231 
232 /*
233  *	Send an IGMP report.
234  */
235 
236 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
237 
238 
239 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
240 	int gdeleted, int sdeleted)
241 {
242 	switch (type) {
243 	case IGMPV3_MODE_IS_INCLUDE:
244 	case IGMPV3_MODE_IS_EXCLUDE:
245 		if (gdeleted || sdeleted)
246 			return 0;
247 		if (!(pmc->gsquery && !psf->sf_gsresp)) {
248 			if (pmc->sfmode == MCAST_INCLUDE)
249 				return 1;
250 			/* don't include if this source is excluded
251 			 * in all filters
252 			 */
253 			if (psf->sf_count[MCAST_INCLUDE])
254 				return type == IGMPV3_MODE_IS_INCLUDE;
255 			return pmc->sfcount[MCAST_EXCLUDE] ==
256 				psf->sf_count[MCAST_EXCLUDE];
257 		}
258 		return 0;
259 	case IGMPV3_CHANGE_TO_INCLUDE:
260 		if (gdeleted || sdeleted)
261 			return 0;
262 		return psf->sf_count[MCAST_INCLUDE] != 0;
263 	case IGMPV3_CHANGE_TO_EXCLUDE:
264 		if (gdeleted || sdeleted)
265 			return 0;
266 		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
267 		    psf->sf_count[MCAST_INCLUDE])
268 			return 0;
269 		return pmc->sfcount[MCAST_EXCLUDE] ==
270 			psf->sf_count[MCAST_EXCLUDE];
271 	case IGMPV3_ALLOW_NEW_SOURCES:
272 		if (gdeleted || !psf->sf_crcount)
273 			return 0;
274 		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
275 	case IGMPV3_BLOCK_OLD_SOURCES:
276 		if (pmc->sfmode == MCAST_INCLUDE)
277 			return gdeleted || (psf->sf_crcount && sdeleted);
278 		return psf->sf_crcount && !gdeleted && !sdeleted;
279 	}
280 	return 0;
281 }
282 
283 static int
284 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
285 {
286 	struct ip_sf_list *psf;
287 	int scount = 0;
288 
289 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
290 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
291 			continue;
292 		scount++;
293 	}
294 	return scount;
295 }
296 
297 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
298 
299 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
300 {
301 	struct sk_buff *skb;
302 	struct rtable *rt;
303 	struct iphdr *pip;
304 	struct igmpv3_report *pig;
305 	struct net *net = dev_net(dev);
306 	struct flowi4 fl4;
307 	int hlen = LL_RESERVED_SPACE(dev);
308 	int tlen = dev->needed_tailroom;
309 
310 	while (1) {
311 		skb = alloc_skb(size + hlen + tlen,
312 				GFP_ATOMIC | __GFP_NOWARN);
313 		if (skb)
314 			break;
315 		size >>= 1;
316 		if (size < 256)
317 			return NULL;
318 	}
319 	igmp_skb_size(skb) = size;
320 
321 	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
322 				   0, 0,
323 				   IPPROTO_IGMP, 0, dev->ifindex);
324 	if (IS_ERR(rt)) {
325 		kfree_skb(skb);
326 		return NULL;
327 	}
328 
329 	skb_dst_set(skb, &rt->dst);
330 	skb->dev = dev;
331 
332 	skb_reserve(skb, hlen);
333 
334 	skb_reset_network_header(skb);
335 	pip = ip_hdr(skb);
336 	skb_put(skb, sizeof(struct iphdr) + 4);
337 
338 	pip->version  = 4;
339 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
340 	pip->tos      = 0xc0;
341 	pip->frag_off = htons(IP_DF);
342 	pip->ttl      = 1;
343 	pip->daddr    = fl4.daddr;
344 	pip->saddr    = fl4.saddr;
345 	pip->protocol = IPPROTO_IGMP;
346 	pip->tot_len  = 0;	/* filled in later */
347 	ip_select_ident(pip, &rt->dst, NULL);
348 	((u8*)&pip[1])[0] = IPOPT_RA;
349 	((u8*)&pip[1])[1] = 4;
350 	((u8*)&pip[1])[2] = 0;
351 	((u8*)&pip[1])[3] = 0;
352 
353 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
354 	skb_put(skb, sizeof(*pig));
355 	pig = igmpv3_report_hdr(skb);
356 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
357 	pig->resv1 = 0;
358 	pig->csum = 0;
359 	pig->resv2 = 0;
360 	pig->ngrec = 0;
361 	return skb;
362 }
363 
364 static int igmpv3_sendpack(struct sk_buff *skb)
365 {
366 	struct igmphdr *pig = igmp_hdr(skb);
367 	const int igmplen = skb->tail - skb->transport_header;
368 
369 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
370 
371 	return ip_local_out(skb);
372 }
373 
374 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
375 {
376 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
377 }
378 
379 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
380 	int type, struct igmpv3_grec **ppgr)
381 {
382 	struct net_device *dev = pmc->interface->dev;
383 	struct igmpv3_report *pih;
384 	struct igmpv3_grec *pgr;
385 
386 	if (!skb)
387 		skb = igmpv3_newpack(dev, dev->mtu);
388 	if (!skb)
389 		return NULL;
390 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
391 	pgr->grec_type = type;
392 	pgr->grec_auxwords = 0;
393 	pgr->grec_nsrcs = 0;
394 	pgr->grec_mca = pmc->multiaddr;
395 	pih = igmpv3_report_hdr(skb);
396 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
397 	*ppgr = pgr;
398 	return skb;
399 }
400 
401 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
402 	skb_tailroom(skb)) : 0)
403 
404 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
405 	int type, int gdeleted, int sdeleted)
406 {
407 	struct net_device *dev = pmc->interface->dev;
408 	struct igmpv3_report *pih;
409 	struct igmpv3_grec *pgr = NULL;
410 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
411 	int scount, stotal, first, isquery, truncate;
412 
413 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
414 		return skb;
415 
416 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
417 		  type == IGMPV3_MODE_IS_EXCLUDE;
418 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
419 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
420 
421 	stotal = scount = 0;
422 
423 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
424 
425 	if (!*psf_list)
426 		goto empty_source;
427 
428 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
429 
430 	/* EX and TO_EX get a fresh packet, if needed */
431 	if (truncate) {
432 		if (pih && pih->ngrec &&
433 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
434 			if (skb)
435 				igmpv3_sendpack(skb);
436 			skb = igmpv3_newpack(dev, dev->mtu);
437 		}
438 	}
439 	first = 1;
440 	psf_prev = NULL;
441 	for (psf=*psf_list; psf; psf=psf_next) {
442 		__be32 *psrc;
443 
444 		psf_next = psf->sf_next;
445 
446 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
447 			psf_prev = psf;
448 			continue;
449 		}
450 
451 		/* clear marks on query responses */
452 		if (isquery)
453 			psf->sf_gsresp = 0;
454 
455 		if (AVAILABLE(skb) < sizeof(__be32) +
456 		    first*sizeof(struct igmpv3_grec)) {
457 			if (truncate && !first)
458 				break;	 /* truncate these */
459 			if (pgr)
460 				pgr->grec_nsrcs = htons(scount);
461 			if (skb)
462 				igmpv3_sendpack(skb);
463 			skb = igmpv3_newpack(dev, dev->mtu);
464 			first = 1;
465 			scount = 0;
466 		}
467 		if (first) {
468 			skb = add_grhead(skb, pmc, type, &pgr);
469 			first = 0;
470 		}
471 		if (!skb)
472 			return NULL;
473 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
474 		*psrc = psf->sf_inaddr;
475 		scount++; stotal++;
476 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
477 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
478 			psf->sf_crcount--;
479 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
480 				if (psf_prev)
481 					psf_prev->sf_next = psf->sf_next;
482 				else
483 					*psf_list = psf->sf_next;
484 				kfree(psf);
485 				continue;
486 			}
487 		}
488 		psf_prev = psf;
489 	}
490 
491 empty_source:
492 	if (!stotal) {
493 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
494 		    type == IGMPV3_BLOCK_OLD_SOURCES)
495 			return skb;
496 		if (pmc->crcount || isquery) {
497 			/* make sure we have room for group header */
498 			if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
499 				igmpv3_sendpack(skb);
500 				skb = NULL; /* add_grhead will get a new one */
501 			}
502 			skb = add_grhead(skb, pmc, type, &pgr);
503 		}
504 	}
505 	if (pgr)
506 		pgr->grec_nsrcs = htons(scount);
507 
508 	if (isquery)
509 		pmc->gsquery = 0;	/* clear query state on report */
510 	return skb;
511 }
512 
513 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
514 {
515 	struct sk_buff *skb = NULL;
516 	int type;
517 
518 	if (!pmc) {
519 		rcu_read_lock();
520 		for_each_pmc_rcu(in_dev, pmc) {
521 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
522 				continue;
523 			spin_lock_bh(&pmc->lock);
524 			if (pmc->sfcount[MCAST_EXCLUDE])
525 				type = IGMPV3_MODE_IS_EXCLUDE;
526 			else
527 				type = IGMPV3_MODE_IS_INCLUDE;
528 			skb = add_grec(skb, pmc, type, 0, 0);
529 			spin_unlock_bh(&pmc->lock);
530 		}
531 		rcu_read_unlock();
532 	} else {
533 		spin_lock_bh(&pmc->lock);
534 		if (pmc->sfcount[MCAST_EXCLUDE])
535 			type = IGMPV3_MODE_IS_EXCLUDE;
536 		else
537 			type = IGMPV3_MODE_IS_INCLUDE;
538 		skb = add_grec(skb, pmc, type, 0, 0);
539 		spin_unlock_bh(&pmc->lock);
540 	}
541 	if (!skb)
542 		return 0;
543 	return igmpv3_sendpack(skb);
544 }
545 
546 /*
547  * remove zero-count source records from a source filter list
548  */
549 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
550 {
551 	struct ip_sf_list *psf_prev, *psf_next, *psf;
552 
553 	psf_prev = NULL;
554 	for (psf=*ppsf; psf; psf = psf_next) {
555 		psf_next = psf->sf_next;
556 		if (psf->sf_crcount == 0) {
557 			if (psf_prev)
558 				psf_prev->sf_next = psf->sf_next;
559 			else
560 				*ppsf = psf->sf_next;
561 			kfree(psf);
562 		} else
563 			psf_prev = psf;
564 	}
565 }
566 
567 static void igmpv3_send_cr(struct in_device *in_dev)
568 {
569 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
570 	struct sk_buff *skb = NULL;
571 	int type, dtype;
572 
573 	rcu_read_lock();
574 	spin_lock_bh(&in_dev->mc_tomb_lock);
575 
576 	/* deleted MCA's */
577 	pmc_prev = NULL;
578 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
579 		pmc_next = pmc->next;
580 		if (pmc->sfmode == MCAST_INCLUDE) {
581 			type = IGMPV3_BLOCK_OLD_SOURCES;
582 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
583 			skb = add_grec(skb, pmc, type, 1, 0);
584 			skb = add_grec(skb, pmc, dtype, 1, 1);
585 		}
586 		if (pmc->crcount) {
587 			if (pmc->sfmode == MCAST_EXCLUDE) {
588 				type = IGMPV3_CHANGE_TO_INCLUDE;
589 				skb = add_grec(skb, pmc, type, 1, 0);
590 			}
591 			pmc->crcount--;
592 			if (pmc->crcount == 0) {
593 				igmpv3_clear_zeros(&pmc->tomb);
594 				igmpv3_clear_zeros(&pmc->sources);
595 			}
596 		}
597 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
598 			if (pmc_prev)
599 				pmc_prev->next = pmc_next;
600 			else
601 				in_dev->mc_tomb = pmc_next;
602 			in_dev_put(pmc->interface);
603 			kfree(pmc);
604 		} else
605 			pmc_prev = pmc;
606 	}
607 	spin_unlock_bh(&in_dev->mc_tomb_lock);
608 
609 	/* change recs */
610 	for_each_pmc_rcu(in_dev, pmc) {
611 		spin_lock_bh(&pmc->lock);
612 		if (pmc->sfcount[MCAST_EXCLUDE]) {
613 			type = IGMPV3_BLOCK_OLD_SOURCES;
614 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
615 		} else {
616 			type = IGMPV3_ALLOW_NEW_SOURCES;
617 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
618 		}
619 		skb = add_grec(skb, pmc, type, 0, 0);
620 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
621 
622 		/* filter mode changes */
623 		if (pmc->crcount) {
624 			if (pmc->sfmode == MCAST_EXCLUDE)
625 				type = IGMPV3_CHANGE_TO_EXCLUDE;
626 			else
627 				type = IGMPV3_CHANGE_TO_INCLUDE;
628 			skb = add_grec(skb, pmc, type, 0, 0);
629 			pmc->crcount--;
630 		}
631 		spin_unlock_bh(&pmc->lock);
632 	}
633 	rcu_read_unlock();
634 
635 	if (!skb)
636 		return;
637 	(void) igmpv3_sendpack(skb);
638 }
639 
640 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
641 	int type)
642 {
643 	struct sk_buff *skb;
644 	struct iphdr *iph;
645 	struct igmphdr *ih;
646 	struct rtable *rt;
647 	struct net_device *dev = in_dev->dev;
648 	struct net *net = dev_net(dev);
649 	__be32	group = pmc ? pmc->multiaddr : 0;
650 	struct flowi4 fl4;
651 	__be32	dst;
652 	int hlen, tlen;
653 
654 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
655 		return igmpv3_send_report(in_dev, pmc);
656 	else if (type == IGMP_HOST_LEAVE_MESSAGE)
657 		dst = IGMP_ALL_ROUTER;
658 	else
659 		dst = group;
660 
661 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
662 				   0, 0,
663 				   IPPROTO_IGMP, 0, dev->ifindex);
664 	if (IS_ERR(rt))
665 		return -1;
666 
667 	hlen = LL_RESERVED_SPACE(dev);
668 	tlen = dev->needed_tailroom;
669 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
670 	if (skb == NULL) {
671 		ip_rt_put(rt);
672 		return -1;
673 	}
674 
675 	skb_dst_set(skb, &rt->dst);
676 
677 	skb_reserve(skb, hlen);
678 
679 	skb_reset_network_header(skb);
680 	iph = ip_hdr(skb);
681 	skb_put(skb, sizeof(struct iphdr) + 4);
682 
683 	iph->version  = 4;
684 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
685 	iph->tos      = 0xc0;
686 	iph->frag_off = htons(IP_DF);
687 	iph->ttl      = 1;
688 	iph->daddr    = dst;
689 	iph->saddr    = fl4.saddr;
690 	iph->protocol = IPPROTO_IGMP;
691 	ip_select_ident(iph, &rt->dst, NULL);
692 	((u8*)&iph[1])[0] = IPOPT_RA;
693 	((u8*)&iph[1])[1] = 4;
694 	((u8*)&iph[1])[2] = 0;
695 	((u8*)&iph[1])[3] = 0;
696 
697 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
698 	ih->type = type;
699 	ih->code = 0;
700 	ih->csum = 0;
701 	ih->group = group;
702 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
703 
704 	return ip_local_out(skb);
705 }
706 
707 static void igmp_gq_timer_expire(unsigned long data)
708 {
709 	struct in_device *in_dev = (struct in_device *)data;
710 
711 	in_dev->mr_gq_running = 0;
712 	igmpv3_send_report(in_dev, NULL);
713 	__in_dev_put(in_dev);
714 }
715 
716 static void igmp_ifc_timer_expire(unsigned long data)
717 {
718 	struct in_device *in_dev = (struct in_device *)data;
719 
720 	igmpv3_send_cr(in_dev);
721 	if (in_dev->mr_ifc_count) {
722 		in_dev->mr_ifc_count--;
723 		igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
724 	}
725 	__in_dev_put(in_dev);
726 }
727 
728 static void igmp_ifc_event(struct in_device *in_dev)
729 {
730 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
731 		return;
732 	in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
733 		IGMP_Unsolicited_Report_Count;
734 	igmp_ifc_start_timer(in_dev, 1);
735 }
736 
737 
738 static void igmp_timer_expire(unsigned long data)
739 {
740 	struct ip_mc_list *im=(struct ip_mc_list *)data;
741 	struct in_device *in_dev = im->interface;
742 
743 	spin_lock(&im->lock);
744 	im->tm_running = 0;
745 
746 	if (im->unsolicit_count) {
747 		im->unsolicit_count--;
748 		igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
749 	}
750 	im->reporter = 1;
751 	spin_unlock(&im->lock);
752 
753 	if (IGMP_V1_SEEN(in_dev))
754 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
755 	else if (IGMP_V2_SEEN(in_dev))
756 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
757 	else
758 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
759 
760 	ip_ma_put(im);
761 }
762 
763 /* mark EXCLUDE-mode sources */
764 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
765 {
766 	struct ip_sf_list *psf;
767 	int i, scount;
768 
769 	scount = 0;
770 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
771 		if (scount == nsrcs)
772 			break;
773 		for (i=0; i<nsrcs; i++) {
774 			/* skip inactive filters */
775 			if (psf->sf_count[MCAST_INCLUDE] ||
776 			    pmc->sfcount[MCAST_EXCLUDE] !=
777 			    psf->sf_count[MCAST_EXCLUDE])
778 				continue;
779 			if (srcs[i] == psf->sf_inaddr) {
780 				scount++;
781 				break;
782 			}
783 		}
784 	}
785 	pmc->gsquery = 0;
786 	if (scount == nsrcs)	/* all sources excluded */
787 		return 0;
788 	return 1;
789 }
790 
791 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
792 {
793 	struct ip_sf_list *psf;
794 	int i, scount;
795 
796 	if (pmc->sfmode == MCAST_EXCLUDE)
797 		return igmp_xmarksources(pmc, nsrcs, srcs);
798 
799 	/* mark INCLUDE-mode sources */
800 	scount = 0;
801 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
802 		if (scount == nsrcs)
803 			break;
804 		for (i=0; i<nsrcs; i++)
805 			if (srcs[i] == psf->sf_inaddr) {
806 				psf->sf_gsresp = 1;
807 				scount++;
808 				break;
809 			}
810 	}
811 	if (!scount) {
812 		pmc->gsquery = 0;
813 		return 0;
814 	}
815 	pmc->gsquery = 1;
816 	return 1;
817 }
818 
819 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
820 {
821 	struct ip_mc_list *im;
822 
823 	/* Timers are only set for non-local groups */
824 
825 	if (group == IGMP_ALL_HOSTS)
826 		return;
827 
828 	rcu_read_lock();
829 	for_each_pmc_rcu(in_dev, im) {
830 		if (im->multiaddr == group) {
831 			igmp_stop_timer(im);
832 			break;
833 		}
834 	}
835 	rcu_read_unlock();
836 }
837 
838 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
839 	int len)
840 {
841 	struct igmphdr 		*ih = igmp_hdr(skb);
842 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
843 	struct ip_mc_list	*im;
844 	__be32			group = ih->group;
845 	int			max_delay;
846 	int			mark = 0;
847 
848 
849 	if (len == 8) {
850 		if (ih->code == 0) {
851 			/* Alas, old v1 router presents here. */
852 
853 			max_delay = IGMP_Query_Response_Interval;
854 			in_dev->mr_v1_seen = jiffies +
855 				IGMP_V1_Router_Present_Timeout;
856 			group = 0;
857 		} else {
858 			/* v2 router present */
859 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
860 			in_dev->mr_v2_seen = jiffies +
861 				IGMP_V2_Router_Present_Timeout;
862 		}
863 		/* cancel the interface change timer */
864 		in_dev->mr_ifc_count = 0;
865 		if (del_timer(&in_dev->mr_ifc_timer))
866 			__in_dev_put(in_dev);
867 		/* clear deleted report items */
868 		igmpv3_clear_delrec(in_dev);
869 	} else if (len < 12) {
870 		return;	/* ignore bogus packet; freed by caller */
871 	} else if (IGMP_V1_SEEN(in_dev)) {
872 		/* This is a v3 query with v1 queriers present */
873 		max_delay = IGMP_Query_Response_Interval;
874 		group = 0;
875 	} else if (IGMP_V2_SEEN(in_dev)) {
876 		/* this is a v3 query with v2 queriers present;
877 		 * Interpretation of the max_delay code is problematic here.
878 		 * A real v2 host would use ih_code directly, while v3 has a
879 		 * different encoding. We use the v3 encoding as more likely
880 		 * to be intended in a v3 query.
881 		 */
882 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
883 	} else { /* v3 */
884 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
885 			return;
886 
887 		ih3 = igmpv3_query_hdr(skb);
888 		if (ih3->nsrcs) {
889 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
890 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
891 				return;
892 			ih3 = igmpv3_query_hdr(skb);
893 		}
894 
895 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
896 		if (!max_delay)
897 			max_delay = 1;	/* can't mod w/ 0 */
898 		in_dev->mr_maxdelay = max_delay;
899 		if (ih3->qrv)
900 			in_dev->mr_qrv = ih3->qrv;
901 		if (!group) { /* general query */
902 			if (ih3->nsrcs)
903 				return;	/* no sources allowed */
904 			igmp_gq_start_timer(in_dev);
905 			return;
906 		}
907 		/* mark sources to include, if group & source-specific */
908 		mark = ih3->nsrcs != 0;
909 	}
910 
911 	/*
912 	 * - Start the timers in all of our membership records
913 	 *   that the query applies to for the interface on
914 	 *   which the query arrived excl. those that belong
915 	 *   to a "local" group (224.0.0.X)
916 	 * - For timers already running check if they need to
917 	 *   be reset.
918 	 * - Use the igmp->igmp_code field as the maximum
919 	 *   delay possible
920 	 */
921 	rcu_read_lock();
922 	for_each_pmc_rcu(in_dev, im) {
923 		int changed;
924 
925 		if (group && group != im->multiaddr)
926 			continue;
927 		if (im->multiaddr == IGMP_ALL_HOSTS)
928 			continue;
929 		spin_lock_bh(&im->lock);
930 		if (im->tm_running)
931 			im->gsquery = im->gsquery && mark;
932 		else
933 			im->gsquery = mark;
934 		changed = !im->gsquery ||
935 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
936 		spin_unlock_bh(&im->lock);
937 		if (changed)
938 			igmp_mod_timer(im, max_delay);
939 	}
940 	rcu_read_unlock();
941 }
942 
943 /* called in rcu_read_lock() section */
944 int igmp_rcv(struct sk_buff *skb)
945 {
946 	/* This basically follows the spec line by line -- see RFC1112 */
947 	struct igmphdr *ih;
948 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
949 	int len = skb->len;
950 
951 	if (in_dev == NULL)
952 		goto drop;
953 
954 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
955 		goto drop;
956 
957 	switch (skb->ip_summed) {
958 	case CHECKSUM_COMPLETE:
959 		if (!csum_fold(skb->csum))
960 			break;
961 		/* fall through */
962 	case CHECKSUM_NONE:
963 		skb->csum = 0;
964 		if (__skb_checksum_complete(skb))
965 			goto drop;
966 	}
967 
968 	ih = igmp_hdr(skb);
969 	switch (ih->type) {
970 	case IGMP_HOST_MEMBERSHIP_QUERY:
971 		igmp_heard_query(in_dev, skb, len);
972 		break;
973 	case IGMP_HOST_MEMBERSHIP_REPORT:
974 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
975 		/* Is it our report looped back? */
976 		if (rt_is_output_route(skb_rtable(skb)))
977 			break;
978 		/* don't rely on MC router hearing unicast reports */
979 		if (skb->pkt_type == PACKET_MULTICAST ||
980 		    skb->pkt_type == PACKET_BROADCAST)
981 			igmp_heard_report(in_dev, ih->group);
982 		break;
983 	case IGMP_PIM:
984 #ifdef CONFIG_IP_PIMSM_V1
985 		return pim_rcv_v1(skb);
986 #endif
987 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
988 	case IGMP_DVMRP:
989 	case IGMP_TRACE:
990 	case IGMP_HOST_LEAVE_MESSAGE:
991 	case IGMP_MTRACE:
992 	case IGMP_MTRACE_RESP:
993 		break;
994 	default:
995 		break;
996 	}
997 
998 drop:
999 	kfree_skb(skb);
1000 	return 0;
1001 }
1002 
1003 #endif
1004 
1005 
1006 /*
1007  *	Add a filter to a device
1008  */
1009 
1010 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1011 {
1012 	char buf[MAX_ADDR_LEN];
1013 	struct net_device *dev = in_dev->dev;
1014 
1015 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1016 	   We will get multicast token leakage, when IFF_MULTICAST
1017 	   is changed. This check should be done in ndo_set_rx_mode
1018 	   routine. Something sort of:
1019 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1020 	   --ANK
1021 	   */
1022 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1023 		dev_mc_add(dev, buf);
1024 }
1025 
1026 /*
1027  *	Remove a filter from a device
1028  */
1029 
1030 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1031 {
1032 	char buf[MAX_ADDR_LEN];
1033 	struct net_device *dev = in_dev->dev;
1034 
1035 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1036 		dev_mc_del(dev, buf);
1037 }
1038 
1039 #ifdef CONFIG_IP_MULTICAST
1040 /*
1041  * deleted ip_mc_list manipulation
1042  */
1043 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1044 {
1045 	struct ip_mc_list *pmc;
1046 
1047 	/* this is an "ip_mc_list" for convenience; only the fields below
1048 	 * are actually used. In particular, the refcnt and users are not
1049 	 * used for management of the delete list. Using the same structure
1050 	 * for deleted items allows change reports to use common code with
1051 	 * non-deleted or query-response MCA's.
1052 	 */
1053 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1054 	if (!pmc)
1055 		return;
1056 	spin_lock_bh(&im->lock);
1057 	pmc->interface = im->interface;
1058 	in_dev_hold(in_dev);
1059 	pmc->multiaddr = im->multiaddr;
1060 	pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1061 		IGMP_Unsolicited_Report_Count;
1062 	pmc->sfmode = im->sfmode;
1063 	if (pmc->sfmode == MCAST_INCLUDE) {
1064 		struct ip_sf_list *psf;
1065 
1066 		pmc->tomb = im->tomb;
1067 		pmc->sources = im->sources;
1068 		im->tomb = im->sources = NULL;
1069 		for (psf=pmc->sources; psf; psf=psf->sf_next)
1070 			psf->sf_crcount = pmc->crcount;
1071 	}
1072 	spin_unlock_bh(&im->lock);
1073 
1074 	spin_lock_bh(&in_dev->mc_tomb_lock);
1075 	pmc->next = in_dev->mc_tomb;
1076 	in_dev->mc_tomb = pmc;
1077 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1078 }
1079 
1080 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1081 {
1082 	struct ip_mc_list *pmc, *pmc_prev;
1083 	struct ip_sf_list *psf, *psf_next;
1084 
1085 	spin_lock_bh(&in_dev->mc_tomb_lock);
1086 	pmc_prev = NULL;
1087 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1088 		if (pmc->multiaddr == multiaddr)
1089 			break;
1090 		pmc_prev = pmc;
1091 	}
1092 	if (pmc) {
1093 		if (pmc_prev)
1094 			pmc_prev->next = pmc->next;
1095 		else
1096 			in_dev->mc_tomb = pmc->next;
1097 	}
1098 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1099 	if (pmc) {
1100 		for (psf=pmc->tomb; psf; psf=psf_next) {
1101 			psf_next = psf->sf_next;
1102 			kfree(psf);
1103 		}
1104 		in_dev_put(pmc->interface);
1105 		kfree(pmc);
1106 	}
1107 }
1108 
1109 static void igmpv3_clear_delrec(struct in_device *in_dev)
1110 {
1111 	struct ip_mc_list *pmc, *nextpmc;
1112 
1113 	spin_lock_bh(&in_dev->mc_tomb_lock);
1114 	pmc = in_dev->mc_tomb;
1115 	in_dev->mc_tomb = NULL;
1116 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1117 
1118 	for (; pmc; pmc = nextpmc) {
1119 		nextpmc = pmc->next;
1120 		ip_mc_clear_src(pmc);
1121 		in_dev_put(pmc->interface);
1122 		kfree(pmc);
1123 	}
1124 	/* clear dead sources, too */
1125 	rcu_read_lock();
1126 	for_each_pmc_rcu(in_dev, pmc) {
1127 		struct ip_sf_list *psf, *psf_next;
1128 
1129 		spin_lock_bh(&pmc->lock);
1130 		psf = pmc->tomb;
1131 		pmc->tomb = NULL;
1132 		spin_unlock_bh(&pmc->lock);
1133 		for (; psf; psf=psf_next) {
1134 			psf_next = psf->sf_next;
1135 			kfree(psf);
1136 		}
1137 	}
1138 	rcu_read_unlock();
1139 }
1140 #endif
1141 
1142 static void igmp_group_dropped(struct ip_mc_list *im)
1143 {
1144 	struct in_device *in_dev = im->interface;
1145 #ifdef CONFIG_IP_MULTICAST
1146 	int reporter;
1147 #endif
1148 
1149 	if (im->loaded) {
1150 		im->loaded = 0;
1151 		ip_mc_filter_del(in_dev, im->multiaddr);
1152 	}
1153 
1154 #ifdef CONFIG_IP_MULTICAST
1155 	if (im->multiaddr == IGMP_ALL_HOSTS)
1156 		return;
1157 
1158 	reporter = im->reporter;
1159 	igmp_stop_timer(im);
1160 
1161 	if (!in_dev->dead) {
1162 		if (IGMP_V1_SEEN(in_dev))
1163 			return;
1164 		if (IGMP_V2_SEEN(in_dev)) {
1165 			if (reporter)
1166 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1167 			return;
1168 		}
1169 		/* IGMPv3 */
1170 		igmpv3_add_delrec(in_dev, im);
1171 
1172 		igmp_ifc_event(in_dev);
1173 	}
1174 #endif
1175 }
1176 
1177 static void igmp_group_added(struct ip_mc_list *im)
1178 {
1179 	struct in_device *in_dev = im->interface;
1180 
1181 	if (im->loaded == 0) {
1182 		im->loaded = 1;
1183 		ip_mc_filter_add(in_dev, im->multiaddr);
1184 	}
1185 
1186 #ifdef CONFIG_IP_MULTICAST
1187 	if (im->multiaddr == IGMP_ALL_HOSTS)
1188 		return;
1189 
1190 	if (in_dev->dead)
1191 		return;
1192 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1193 		spin_lock_bh(&im->lock);
1194 		igmp_start_timer(im, IGMP_Initial_Report_Delay);
1195 		spin_unlock_bh(&im->lock);
1196 		return;
1197 	}
1198 	/* else, v3 */
1199 
1200 	im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1201 		IGMP_Unsolicited_Report_Count;
1202 	igmp_ifc_event(in_dev);
1203 #endif
1204 }
1205 
1206 
1207 /*
1208  *	Multicast list managers
1209  */
1210 
1211 
1212 /*
1213  *	A socket has joined a multicast group on device dev.
1214  */
1215 
1216 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1217 {
1218 	struct ip_mc_list *im;
1219 
1220 	ASSERT_RTNL();
1221 
1222 	for_each_pmc_rtnl(in_dev, im) {
1223 		if (im->multiaddr == addr) {
1224 			im->users++;
1225 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1226 			goto out;
1227 		}
1228 	}
1229 
1230 	im = kzalloc(sizeof(*im), GFP_KERNEL);
1231 	if (!im)
1232 		goto out;
1233 
1234 	im->users = 1;
1235 	im->interface = in_dev;
1236 	in_dev_hold(in_dev);
1237 	im->multiaddr = addr;
1238 	/* initial mode is (EX, empty) */
1239 	im->sfmode = MCAST_EXCLUDE;
1240 	im->sfcount[MCAST_EXCLUDE] = 1;
1241 	atomic_set(&im->refcnt, 1);
1242 	spin_lock_init(&im->lock);
1243 #ifdef CONFIG_IP_MULTICAST
1244 	setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1245 	im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1246 #endif
1247 
1248 	im->next_rcu = in_dev->mc_list;
1249 	in_dev->mc_count++;
1250 	RCU_INIT_POINTER(in_dev->mc_list, im);
1251 
1252 #ifdef CONFIG_IP_MULTICAST
1253 	igmpv3_del_delrec(in_dev, im->multiaddr);
1254 #endif
1255 	igmp_group_added(im);
1256 	if (!in_dev->dead)
1257 		ip_rt_multicast_event(in_dev);
1258 out:
1259 	return;
1260 }
1261 EXPORT_SYMBOL(ip_mc_inc_group);
1262 
1263 /*
1264  *	Resend IGMP JOIN report; used for bonding.
1265  *	Called with rcu_read_lock()
1266  */
1267 void ip_mc_rejoin_groups(struct in_device *in_dev)
1268 {
1269 #ifdef CONFIG_IP_MULTICAST
1270 	struct ip_mc_list *im;
1271 	int type;
1272 
1273 	for_each_pmc_rcu(in_dev, im) {
1274 		if (im->multiaddr == IGMP_ALL_HOSTS)
1275 			continue;
1276 
1277 		/* a failover is happening and switches
1278 		 * must be notified immediately
1279 		 */
1280 		if (IGMP_V1_SEEN(in_dev))
1281 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1282 		else if (IGMP_V2_SEEN(in_dev))
1283 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1284 		else
1285 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1286 		igmp_send_report(in_dev, im, type);
1287 	}
1288 #endif
1289 }
1290 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1291 
1292 /*
1293  *	A socket has left a multicast group on device dev
1294  */
1295 
1296 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1297 {
1298 	struct ip_mc_list *i;
1299 	struct ip_mc_list __rcu **ip;
1300 
1301 	ASSERT_RTNL();
1302 
1303 	for (ip = &in_dev->mc_list;
1304 	     (i = rtnl_dereference(*ip)) != NULL;
1305 	     ip = &i->next_rcu) {
1306 		if (i->multiaddr == addr) {
1307 			if (--i->users == 0) {
1308 				*ip = i->next_rcu;
1309 				in_dev->mc_count--;
1310 				igmp_group_dropped(i);
1311 				ip_mc_clear_src(i);
1312 
1313 				if (!in_dev->dead)
1314 					ip_rt_multicast_event(in_dev);
1315 
1316 				ip_ma_put(i);
1317 				return;
1318 			}
1319 			break;
1320 		}
1321 	}
1322 }
1323 EXPORT_SYMBOL(ip_mc_dec_group);
1324 
1325 /* Device changing type */
1326 
1327 void ip_mc_unmap(struct in_device *in_dev)
1328 {
1329 	struct ip_mc_list *pmc;
1330 
1331 	ASSERT_RTNL();
1332 
1333 	for_each_pmc_rtnl(in_dev, pmc)
1334 		igmp_group_dropped(pmc);
1335 }
1336 
1337 void ip_mc_remap(struct in_device *in_dev)
1338 {
1339 	struct ip_mc_list *pmc;
1340 
1341 	ASSERT_RTNL();
1342 
1343 	for_each_pmc_rtnl(in_dev, pmc)
1344 		igmp_group_added(pmc);
1345 }
1346 
1347 /* Device going down */
1348 
1349 void ip_mc_down(struct in_device *in_dev)
1350 {
1351 	struct ip_mc_list *pmc;
1352 
1353 	ASSERT_RTNL();
1354 
1355 	for_each_pmc_rtnl(in_dev, pmc)
1356 		igmp_group_dropped(pmc);
1357 
1358 #ifdef CONFIG_IP_MULTICAST
1359 	in_dev->mr_ifc_count = 0;
1360 	if (del_timer(&in_dev->mr_ifc_timer))
1361 		__in_dev_put(in_dev);
1362 	in_dev->mr_gq_running = 0;
1363 	if (del_timer(&in_dev->mr_gq_timer))
1364 		__in_dev_put(in_dev);
1365 	igmpv3_clear_delrec(in_dev);
1366 #endif
1367 
1368 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1369 }
1370 
1371 void ip_mc_init_dev(struct in_device *in_dev)
1372 {
1373 	ASSERT_RTNL();
1374 
1375 	in_dev->mc_tomb = NULL;
1376 #ifdef CONFIG_IP_MULTICAST
1377 	in_dev->mr_gq_running = 0;
1378 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1379 			(unsigned long)in_dev);
1380 	in_dev->mr_ifc_count = 0;
1381 	in_dev->mc_count     = 0;
1382 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1383 			(unsigned long)in_dev);
1384 	in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1385 #endif
1386 
1387 	spin_lock_init(&in_dev->mc_tomb_lock);
1388 }
1389 
1390 /* Device going up */
1391 
1392 void ip_mc_up(struct in_device *in_dev)
1393 {
1394 	struct ip_mc_list *pmc;
1395 
1396 	ASSERT_RTNL();
1397 
1398 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1399 
1400 	for_each_pmc_rtnl(in_dev, pmc)
1401 		igmp_group_added(pmc);
1402 }
1403 
1404 /*
1405  *	Device is about to be destroyed: clean up.
1406  */
1407 
1408 void ip_mc_destroy_dev(struct in_device *in_dev)
1409 {
1410 	struct ip_mc_list *i;
1411 
1412 	ASSERT_RTNL();
1413 
1414 	/* Deactivate timers */
1415 	ip_mc_down(in_dev);
1416 
1417 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1418 		in_dev->mc_list = i->next_rcu;
1419 		in_dev->mc_count--;
1420 
1421 		/* We've dropped the groups in ip_mc_down already */
1422 		ip_mc_clear_src(i);
1423 		ip_ma_put(i);
1424 	}
1425 }
1426 
1427 /* RTNL is locked */
1428 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1429 {
1430 	struct net_device *dev = NULL;
1431 	struct in_device *idev = NULL;
1432 
1433 	if (imr->imr_ifindex) {
1434 		idev = inetdev_by_index(net, imr->imr_ifindex);
1435 		return idev;
1436 	}
1437 	if (imr->imr_address.s_addr) {
1438 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1439 		if (!dev)
1440 			return NULL;
1441 	}
1442 
1443 	if (!dev) {
1444 		struct rtable *rt = ip_route_output(net,
1445 						    imr->imr_multiaddr.s_addr,
1446 						    0, 0, 0);
1447 		if (!IS_ERR(rt)) {
1448 			dev = rt->dst.dev;
1449 			ip_rt_put(rt);
1450 		}
1451 	}
1452 	if (dev) {
1453 		imr->imr_ifindex = dev->ifindex;
1454 		idev = __in_dev_get_rtnl(dev);
1455 	}
1456 	return idev;
1457 }
1458 
1459 /*
1460  *	Join a socket to a group
1461  */
1462 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1463 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1464 
1465 
1466 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1467 	__be32 *psfsrc)
1468 {
1469 	struct ip_sf_list *psf, *psf_prev;
1470 	int rv = 0;
1471 
1472 	psf_prev = NULL;
1473 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1474 		if (psf->sf_inaddr == *psfsrc)
1475 			break;
1476 		psf_prev = psf;
1477 	}
1478 	if (!psf || psf->sf_count[sfmode] == 0) {
1479 		/* source filter not found, or count wrong =>  bug */
1480 		return -ESRCH;
1481 	}
1482 	psf->sf_count[sfmode]--;
1483 	if (psf->sf_count[sfmode] == 0) {
1484 		ip_rt_multicast_event(pmc->interface);
1485 	}
1486 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1487 #ifdef CONFIG_IP_MULTICAST
1488 		struct in_device *in_dev = pmc->interface;
1489 #endif
1490 
1491 		/* no more filters for this source */
1492 		if (psf_prev)
1493 			psf_prev->sf_next = psf->sf_next;
1494 		else
1495 			pmc->sources = psf->sf_next;
1496 #ifdef CONFIG_IP_MULTICAST
1497 		if (psf->sf_oldin &&
1498 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1499 			psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1500 				IGMP_Unsolicited_Report_Count;
1501 			psf->sf_next = pmc->tomb;
1502 			pmc->tomb = psf;
1503 			rv = 1;
1504 		} else
1505 #endif
1506 			kfree(psf);
1507 	}
1508 	return rv;
1509 }
1510 
1511 #ifndef CONFIG_IP_MULTICAST
1512 #define igmp_ifc_event(x)	do { } while (0)
1513 #endif
1514 
1515 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1516 			 int sfcount, __be32 *psfsrc, int delta)
1517 {
1518 	struct ip_mc_list *pmc;
1519 	int	changerec = 0;
1520 	int	i, err;
1521 
1522 	if (!in_dev)
1523 		return -ENODEV;
1524 	rcu_read_lock();
1525 	for_each_pmc_rcu(in_dev, pmc) {
1526 		if (*pmca == pmc->multiaddr)
1527 			break;
1528 	}
1529 	if (!pmc) {
1530 		/* MCA not found?? bug */
1531 		rcu_read_unlock();
1532 		return -ESRCH;
1533 	}
1534 	spin_lock_bh(&pmc->lock);
1535 	rcu_read_unlock();
1536 #ifdef CONFIG_IP_MULTICAST
1537 	sf_markstate(pmc);
1538 #endif
1539 	if (!delta) {
1540 		err = -EINVAL;
1541 		if (!pmc->sfcount[sfmode])
1542 			goto out_unlock;
1543 		pmc->sfcount[sfmode]--;
1544 	}
1545 	err = 0;
1546 	for (i=0; i<sfcount; i++) {
1547 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1548 
1549 		changerec |= rv > 0;
1550 		if (!err && rv < 0)
1551 			err = rv;
1552 	}
1553 	if (pmc->sfmode == MCAST_EXCLUDE &&
1554 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1555 	    pmc->sfcount[MCAST_INCLUDE]) {
1556 #ifdef CONFIG_IP_MULTICAST
1557 		struct ip_sf_list *psf;
1558 #endif
1559 
1560 		/* filter mode change */
1561 		pmc->sfmode = MCAST_INCLUDE;
1562 #ifdef CONFIG_IP_MULTICAST
1563 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1564 			IGMP_Unsolicited_Report_Count;
1565 		in_dev->mr_ifc_count = pmc->crcount;
1566 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1567 			psf->sf_crcount = 0;
1568 		igmp_ifc_event(pmc->interface);
1569 	} else if (sf_setstate(pmc) || changerec) {
1570 		igmp_ifc_event(pmc->interface);
1571 #endif
1572 	}
1573 out_unlock:
1574 	spin_unlock_bh(&pmc->lock);
1575 	return err;
1576 }
1577 
1578 /*
1579  * Add multicast single-source filter to the interface list
1580  */
1581 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1582 	__be32 *psfsrc)
1583 {
1584 	struct ip_sf_list *psf, *psf_prev;
1585 
1586 	psf_prev = NULL;
1587 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1588 		if (psf->sf_inaddr == *psfsrc)
1589 			break;
1590 		psf_prev = psf;
1591 	}
1592 	if (!psf) {
1593 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1594 		if (!psf)
1595 			return -ENOBUFS;
1596 		psf->sf_inaddr = *psfsrc;
1597 		if (psf_prev) {
1598 			psf_prev->sf_next = psf;
1599 		} else
1600 			pmc->sources = psf;
1601 	}
1602 	psf->sf_count[sfmode]++;
1603 	if (psf->sf_count[sfmode] == 1) {
1604 		ip_rt_multicast_event(pmc->interface);
1605 	}
1606 	return 0;
1607 }
1608 
1609 #ifdef CONFIG_IP_MULTICAST
1610 static void sf_markstate(struct ip_mc_list *pmc)
1611 {
1612 	struct ip_sf_list *psf;
1613 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1614 
1615 	for (psf=pmc->sources; psf; psf=psf->sf_next)
1616 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1617 			psf->sf_oldin = mca_xcount ==
1618 				psf->sf_count[MCAST_EXCLUDE] &&
1619 				!psf->sf_count[MCAST_INCLUDE];
1620 		} else
1621 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1622 }
1623 
1624 static int sf_setstate(struct ip_mc_list *pmc)
1625 {
1626 	struct ip_sf_list *psf, *dpsf;
1627 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1628 	int qrv = pmc->interface->mr_qrv;
1629 	int new_in, rv;
1630 
1631 	rv = 0;
1632 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1633 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1634 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1635 				!psf->sf_count[MCAST_INCLUDE];
1636 		} else
1637 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1638 		if (new_in) {
1639 			if (!psf->sf_oldin) {
1640 				struct ip_sf_list *prev = NULL;
1641 
1642 				for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1643 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1644 						break;
1645 					prev = dpsf;
1646 				}
1647 				if (dpsf) {
1648 					if (prev)
1649 						prev->sf_next = dpsf->sf_next;
1650 					else
1651 						pmc->tomb = dpsf->sf_next;
1652 					kfree(dpsf);
1653 				}
1654 				psf->sf_crcount = qrv;
1655 				rv++;
1656 			}
1657 		} else if (psf->sf_oldin) {
1658 
1659 			psf->sf_crcount = 0;
1660 			/*
1661 			 * add or update "delete" records if an active filter
1662 			 * is now inactive
1663 			 */
1664 			for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1665 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1666 					break;
1667 			if (!dpsf) {
1668 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1669 				if (!dpsf)
1670 					continue;
1671 				*dpsf = *psf;
1672 				/* pmc->lock held by callers */
1673 				dpsf->sf_next = pmc->tomb;
1674 				pmc->tomb = dpsf;
1675 			}
1676 			dpsf->sf_crcount = qrv;
1677 			rv++;
1678 		}
1679 	}
1680 	return rv;
1681 }
1682 #endif
1683 
1684 /*
1685  * Add multicast source filter list to the interface list
1686  */
1687 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1688 			 int sfcount, __be32 *psfsrc, int delta)
1689 {
1690 	struct ip_mc_list *pmc;
1691 	int	isexclude;
1692 	int	i, err;
1693 
1694 	if (!in_dev)
1695 		return -ENODEV;
1696 	rcu_read_lock();
1697 	for_each_pmc_rcu(in_dev, pmc) {
1698 		if (*pmca == pmc->multiaddr)
1699 			break;
1700 	}
1701 	if (!pmc) {
1702 		/* MCA not found?? bug */
1703 		rcu_read_unlock();
1704 		return -ESRCH;
1705 	}
1706 	spin_lock_bh(&pmc->lock);
1707 	rcu_read_unlock();
1708 
1709 #ifdef CONFIG_IP_MULTICAST
1710 	sf_markstate(pmc);
1711 #endif
1712 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
1713 	if (!delta)
1714 		pmc->sfcount[sfmode]++;
1715 	err = 0;
1716 	for (i=0; i<sfcount; i++) {
1717 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1718 		if (err)
1719 			break;
1720 	}
1721 	if (err) {
1722 		int j;
1723 
1724 		if (!delta)
1725 			pmc->sfcount[sfmode]--;
1726 		for (j=0; j<i; j++)
1727 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1728 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1729 #ifdef CONFIG_IP_MULTICAST
1730 		struct ip_sf_list *psf;
1731 		in_dev = pmc->interface;
1732 #endif
1733 
1734 		/* filter mode change */
1735 		if (pmc->sfcount[MCAST_EXCLUDE])
1736 			pmc->sfmode = MCAST_EXCLUDE;
1737 		else if (pmc->sfcount[MCAST_INCLUDE])
1738 			pmc->sfmode = MCAST_INCLUDE;
1739 #ifdef CONFIG_IP_MULTICAST
1740 		/* else no filters; keep old mode for reports */
1741 
1742 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1743 			IGMP_Unsolicited_Report_Count;
1744 		in_dev->mr_ifc_count = pmc->crcount;
1745 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1746 			psf->sf_crcount = 0;
1747 		igmp_ifc_event(in_dev);
1748 	} else if (sf_setstate(pmc)) {
1749 		igmp_ifc_event(in_dev);
1750 #endif
1751 	}
1752 	spin_unlock_bh(&pmc->lock);
1753 	return err;
1754 }
1755 
1756 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1757 {
1758 	struct ip_sf_list *psf, *nextpsf;
1759 
1760 	for (psf=pmc->tomb; psf; psf=nextpsf) {
1761 		nextpsf = psf->sf_next;
1762 		kfree(psf);
1763 	}
1764 	pmc->tomb = NULL;
1765 	for (psf=pmc->sources; psf; psf=nextpsf) {
1766 		nextpsf = psf->sf_next;
1767 		kfree(psf);
1768 	}
1769 	pmc->sources = NULL;
1770 	pmc->sfmode = MCAST_EXCLUDE;
1771 	pmc->sfcount[MCAST_INCLUDE] = 0;
1772 	pmc->sfcount[MCAST_EXCLUDE] = 1;
1773 }
1774 
1775 
1776 /*
1777  * Join a multicast group
1778  */
1779 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1780 {
1781 	int err;
1782 	__be32 addr = imr->imr_multiaddr.s_addr;
1783 	struct ip_mc_socklist *iml = NULL, *i;
1784 	struct in_device *in_dev;
1785 	struct inet_sock *inet = inet_sk(sk);
1786 	struct net *net = sock_net(sk);
1787 	int ifindex;
1788 	int count = 0;
1789 
1790 	if (!ipv4_is_multicast(addr))
1791 		return -EINVAL;
1792 
1793 	rtnl_lock();
1794 
1795 	in_dev = ip_mc_find_dev(net, imr);
1796 
1797 	if (!in_dev) {
1798 		iml = NULL;
1799 		err = -ENODEV;
1800 		goto done;
1801 	}
1802 
1803 	err = -EADDRINUSE;
1804 	ifindex = imr->imr_ifindex;
1805 	for_each_pmc_rtnl(inet, i) {
1806 		if (i->multi.imr_multiaddr.s_addr == addr &&
1807 		    i->multi.imr_ifindex == ifindex)
1808 			goto done;
1809 		count++;
1810 	}
1811 	err = -ENOBUFS;
1812 	if (count >= sysctl_igmp_max_memberships)
1813 		goto done;
1814 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1815 	if (iml == NULL)
1816 		goto done;
1817 
1818 	memcpy(&iml->multi, imr, sizeof(*imr));
1819 	iml->next_rcu = inet->mc_list;
1820 	iml->sflist = NULL;
1821 	iml->sfmode = MCAST_EXCLUDE;
1822 	RCU_INIT_POINTER(inet->mc_list, iml);
1823 	ip_mc_inc_group(in_dev, addr);
1824 	err = 0;
1825 done:
1826 	rtnl_unlock();
1827 	return err;
1828 }
1829 EXPORT_SYMBOL(ip_mc_join_group);
1830 
1831 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1832 			   struct in_device *in_dev)
1833 {
1834 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1835 	int err;
1836 
1837 	if (psf == NULL) {
1838 		/* any-source empty exclude case */
1839 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1840 			iml->sfmode, 0, NULL, 0);
1841 	}
1842 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1843 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1844 	RCU_INIT_POINTER(iml->sflist, NULL);
1845 	/* decrease mem now to avoid the memleak warning */
1846 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1847 	kfree_rcu(psf, rcu);
1848 	return err;
1849 }
1850 
1851 /*
1852  *	Ask a socket to leave a group.
1853  */
1854 
1855 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1856 {
1857 	struct inet_sock *inet = inet_sk(sk);
1858 	struct ip_mc_socklist *iml;
1859 	struct ip_mc_socklist __rcu **imlp;
1860 	struct in_device *in_dev;
1861 	struct net *net = sock_net(sk);
1862 	__be32 group = imr->imr_multiaddr.s_addr;
1863 	u32 ifindex;
1864 	int ret = -EADDRNOTAVAIL;
1865 
1866 	rtnl_lock();
1867 	in_dev = ip_mc_find_dev(net, imr);
1868 	ifindex = imr->imr_ifindex;
1869 	for (imlp = &inet->mc_list;
1870 	     (iml = rtnl_dereference(*imlp)) != NULL;
1871 	     imlp = &iml->next_rcu) {
1872 		if (iml->multi.imr_multiaddr.s_addr != group)
1873 			continue;
1874 		if (ifindex) {
1875 			if (iml->multi.imr_ifindex != ifindex)
1876 				continue;
1877 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1878 				iml->multi.imr_address.s_addr)
1879 			continue;
1880 
1881 		(void) ip_mc_leave_src(sk, iml, in_dev);
1882 
1883 		*imlp = iml->next_rcu;
1884 
1885 		if (in_dev)
1886 			ip_mc_dec_group(in_dev, group);
1887 		rtnl_unlock();
1888 		/* decrease mem now to avoid the memleak warning */
1889 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1890 		kfree_rcu(iml, rcu);
1891 		return 0;
1892 	}
1893 	if (!in_dev)
1894 		ret = -ENODEV;
1895 	rtnl_unlock();
1896 	return ret;
1897 }
1898 
1899 int ip_mc_source(int add, int omode, struct sock *sk, struct
1900 	ip_mreq_source *mreqs, int ifindex)
1901 {
1902 	int err;
1903 	struct ip_mreqn imr;
1904 	__be32 addr = mreqs->imr_multiaddr;
1905 	struct ip_mc_socklist *pmc;
1906 	struct in_device *in_dev = NULL;
1907 	struct inet_sock *inet = inet_sk(sk);
1908 	struct ip_sf_socklist *psl;
1909 	struct net *net = sock_net(sk);
1910 	int leavegroup = 0;
1911 	int i, j, rv;
1912 
1913 	if (!ipv4_is_multicast(addr))
1914 		return -EINVAL;
1915 
1916 	rtnl_lock();
1917 
1918 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1919 	imr.imr_address.s_addr = mreqs->imr_interface;
1920 	imr.imr_ifindex = ifindex;
1921 	in_dev = ip_mc_find_dev(net, &imr);
1922 
1923 	if (!in_dev) {
1924 		err = -ENODEV;
1925 		goto done;
1926 	}
1927 	err = -EADDRNOTAVAIL;
1928 
1929 	for_each_pmc_rtnl(inet, pmc) {
1930 		if ((pmc->multi.imr_multiaddr.s_addr ==
1931 		     imr.imr_multiaddr.s_addr) &&
1932 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
1933 			break;
1934 	}
1935 	if (!pmc) {		/* must have a prior join */
1936 		err = -EINVAL;
1937 		goto done;
1938 	}
1939 	/* if a source filter was set, must be the same mode as before */
1940 	if (pmc->sflist) {
1941 		if (pmc->sfmode != omode) {
1942 			err = -EINVAL;
1943 			goto done;
1944 		}
1945 	} else if (pmc->sfmode != omode) {
1946 		/* allow mode switches for empty-set filters */
1947 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1948 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1949 			NULL, 0);
1950 		pmc->sfmode = omode;
1951 	}
1952 
1953 	psl = rtnl_dereference(pmc->sflist);
1954 	if (!add) {
1955 		if (!psl)
1956 			goto done;	/* err = -EADDRNOTAVAIL */
1957 		rv = !0;
1958 		for (i=0; i<psl->sl_count; i++) {
1959 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1960 				sizeof(__be32));
1961 			if (rv == 0)
1962 				break;
1963 		}
1964 		if (rv)		/* source not found */
1965 			goto done;	/* err = -EADDRNOTAVAIL */
1966 
1967 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
1968 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1969 			leavegroup = 1;
1970 			goto done;
1971 		}
1972 
1973 		/* update the interface filter */
1974 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1975 			&mreqs->imr_sourceaddr, 1);
1976 
1977 		for (j=i+1; j<psl->sl_count; j++)
1978 			psl->sl_addr[j-1] = psl->sl_addr[j];
1979 		psl->sl_count--;
1980 		err = 0;
1981 		goto done;
1982 	}
1983 	/* else, add a new source to the filter */
1984 
1985 	if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1986 		err = -ENOBUFS;
1987 		goto done;
1988 	}
1989 	if (!psl || psl->sl_count == psl->sl_max) {
1990 		struct ip_sf_socklist *newpsl;
1991 		int count = IP_SFBLOCK;
1992 
1993 		if (psl)
1994 			count += psl->sl_max;
1995 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
1996 		if (!newpsl) {
1997 			err = -ENOBUFS;
1998 			goto done;
1999 		}
2000 		newpsl->sl_max = count;
2001 		newpsl->sl_count = count - IP_SFBLOCK;
2002 		if (psl) {
2003 			for (i=0; i<psl->sl_count; i++)
2004 				newpsl->sl_addr[i] = psl->sl_addr[i];
2005 			/* decrease mem now to avoid the memleak warning */
2006 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2007 			kfree_rcu(psl, rcu);
2008 		}
2009 		RCU_INIT_POINTER(pmc->sflist, newpsl);
2010 		psl = newpsl;
2011 	}
2012 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2013 	for (i=0; i<psl->sl_count; i++) {
2014 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2015 			sizeof(__be32));
2016 		if (rv == 0)
2017 			break;
2018 	}
2019 	if (rv == 0)		/* address already there is an error */
2020 		goto done;
2021 	for (j=psl->sl_count-1; j>=i; j--)
2022 		psl->sl_addr[j+1] = psl->sl_addr[j];
2023 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2024 	psl->sl_count++;
2025 	err = 0;
2026 	/* update the interface list */
2027 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2028 		&mreqs->imr_sourceaddr, 1);
2029 done:
2030 	rtnl_unlock();
2031 	if (leavegroup)
2032 		return ip_mc_leave_group(sk, &imr);
2033 	return err;
2034 }
2035 
2036 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2037 {
2038 	int err = 0;
2039 	struct ip_mreqn	imr;
2040 	__be32 addr = msf->imsf_multiaddr;
2041 	struct ip_mc_socklist *pmc;
2042 	struct in_device *in_dev;
2043 	struct inet_sock *inet = inet_sk(sk);
2044 	struct ip_sf_socklist *newpsl, *psl;
2045 	struct net *net = sock_net(sk);
2046 	int leavegroup = 0;
2047 
2048 	if (!ipv4_is_multicast(addr))
2049 		return -EINVAL;
2050 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2051 	    msf->imsf_fmode != MCAST_EXCLUDE)
2052 		return -EINVAL;
2053 
2054 	rtnl_lock();
2055 
2056 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2057 	imr.imr_address.s_addr = msf->imsf_interface;
2058 	imr.imr_ifindex = ifindex;
2059 	in_dev = ip_mc_find_dev(net, &imr);
2060 
2061 	if (!in_dev) {
2062 		err = -ENODEV;
2063 		goto done;
2064 	}
2065 
2066 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2067 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2068 		leavegroup = 1;
2069 		goto done;
2070 	}
2071 
2072 	for_each_pmc_rtnl(inet, pmc) {
2073 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2074 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2075 			break;
2076 	}
2077 	if (!pmc) {		/* must have a prior join */
2078 		err = -EINVAL;
2079 		goto done;
2080 	}
2081 	if (msf->imsf_numsrc) {
2082 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2083 							   GFP_KERNEL);
2084 		if (!newpsl) {
2085 			err = -ENOBUFS;
2086 			goto done;
2087 		}
2088 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2089 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2090 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2091 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2092 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2093 		if (err) {
2094 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2095 			goto done;
2096 		}
2097 	} else {
2098 		newpsl = NULL;
2099 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2100 				     msf->imsf_fmode, 0, NULL, 0);
2101 	}
2102 	psl = rtnl_dereference(pmc->sflist);
2103 	if (psl) {
2104 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2105 			psl->sl_count, psl->sl_addr, 0);
2106 		/* decrease mem now to avoid the memleak warning */
2107 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2108 		kfree_rcu(psl, rcu);
2109 	} else
2110 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2111 			0, NULL, 0);
2112 	RCU_INIT_POINTER(pmc->sflist, newpsl);
2113 	pmc->sfmode = msf->imsf_fmode;
2114 	err = 0;
2115 done:
2116 	rtnl_unlock();
2117 	if (leavegroup)
2118 		err = ip_mc_leave_group(sk, &imr);
2119 	return err;
2120 }
2121 
2122 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2123 	struct ip_msfilter __user *optval, int __user *optlen)
2124 {
2125 	int err, len, count, copycount;
2126 	struct ip_mreqn	imr;
2127 	__be32 addr = msf->imsf_multiaddr;
2128 	struct ip_mc_socklist *pmc;
2129 	struct in_device *in_dev;
2130 	struct inet_sock *inet = inet_sk(sk);
2131 	struct ip_sf_socklist *psl;
2132 	struct net *net = sock_net(sk);
2133 
2134 	if (!ipv4_is_multicast(addr))
2135 		return -EINVAL;
2136 
2137 	rtnl_lock();
2138 
2139 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2140 	imr.imr_address.s_addr = msf->imsf_interface;
2141 	imr.imr_ifindex = 0;
2142 	in_dev = ip_mc_find_dev(net, &imr);
2143 
2144 	if (!in_dev) {
2145 		err = -ENODEV;
2146 		goto done;
2147 	}
2148 	err = -EADDRNOTAVAIL;
2149 
2150 	for_each_pmc_rtnl(inet, pmc) {
2151 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2152 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2153 			break;
2154 	}
2155 	if (!pmc)		/* must have a prior join */
2156 		goto done;
2157 	msf->imsf_fmode = pmc->sfmode;
2158 	psl = rtnl_dereference(pmc->sflist);
2159 	rtnl_unlock();
2160 	if (!psl) {
2161 		len = 0;
2162 		count = 0;
2163 	} else {
2164 		count = psl->sl_count;
2165 	}
2166 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2167 	len = copycount * sizeof(psl->sl_addr[0]);
2168 	msf->imsf_numsrc = count;
2169 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2170 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2171 		return -EFAULT;
2172 	}
2173 	if (len &&
2174 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2175 		return -EFAULT;
2176 	return 0;
2177 done:
2178 	rtnl_unlock();
2179 	return err;
2180 }
2181 
2182 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2183 	struct group_filter __user *optval, int __user *optlen)
2184 {
2185 	int err, i, count, copycount;
2186 	struct sockaddr_in *psin;
2187 	__be32 addr;
2188 	struct ip_mc_socklist *pmc;
2189 	struct inet_sock *inet = inet_sk(sk);
2190 	struct ip_sf_socklist *psl;
2191 
2192 	psin = (struct sockaddr_in *)&gsf->gf_group;
2193 	if (psin->sin_family != AF_INET)
2194 		return -EINVAL;
2195 	addr = psin->sin_addr.s_addr;
2196 	if (!ipv4_is_multicast(addr))
2197 		return -EINVAL;
2198 
2199 	rtnl_lock();
2200 
2201 	err = -EADDRNOTAVAIL;
2202 
2203 	for_each_pmc_rtnl(inet, pmc) {
2204 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2205 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2206 			break;
2207 	}
2208 	if (!pmc)		/* must have a prior join */
2209 		goto done;
2210 	gsf->gf_fmode = pmc->sfmode;
2211 	psl = rtnl_dereference(pmc->sflist);
2212 	rtnl_unlock();
2213 	count = psl ? psl->sl_count : 0;
2214 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2215 	gsf->gf_numsrc = count;
2216 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2217 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2218 		return -EFAULT;
2219 	}
2220 	for (i=0; i<copycount; i++) {
2221 		struct sockaddr_storage ss;
2222 
2223 		psin = (struct sockaddr_in *)&ss;
2224 		memset(&ss, 0, sizeof(ss));
2225 		psin->sin_family = AF_INET;
2226 		psin->sin_addr.s_addr = psl->sl_addr[i];
2227 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2228 			return -EFAULT;
2229 	}
2230 	return 0;
2231 done:
2232 	rtnl_unlock();
2233 	return err;
2234 }
2235 
2236 /*
2237  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2238  */
2239 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2240 {
2241 	struct inet_sock *inet = inet_sk(sk);
2242 	struct ip_mc_socklist *pmc;
2243 	struct ip_sf_socklist *psl;
2244 	int i;
2245 	int ret;
2246 
2247 	ret = 1;
2248 	if (!ipv4_is_multicast(loc_addr))
2249 		goto out;
2250 
2251 	rcu_read_lock();
2252 	for_each_pmc_rcu(inet, pmc) {
2253 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2254 		    pmc->multi.imr_ifindex == dif)
2255 			break;
2256 	}
2257 	ret = inet->mc_all;
2258 	if (!pmc)
2259 		goto unlock;
2260 	psl = rcu_dereference(pmc->sflist);
2261 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2262 	if (!psl)
2263 		goto unlock;
2264 
2265 	for (i=0; i<psl->sl_count; i++) {
2266 		if (psl->sl_addr[i] == rmt_addr)
2267 			break;
2268 	}
2269 	ret = 0;
2270 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2271 		goto unlock;
2272 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2273 		goto unlock;
2274 	ret = 1;
2275 unlock:
2276 	rcu_read_unlock();
2277 out:
2278 	return ret;
2279 }
2280 
2281 /*
2282  *	A socket is closing.
2283  */
2284 
2285 void ip_mc_drop_socket(struct sock *sk)
2286 {
2287 	struct inet_sock *inet = inet_sk(sk);
2288 	struct ip_mc_socklist *iml;
2289 	struct net *net = sock_net(sk);
2290 
2291 	if (inet->mc_list == NULL)
2292 		return;
2293 
2294 	rtnl_lock();
2295 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2296 		struct in_device *in_dev;
2297 
2298 		inet->mc_list = iml->next_rcu;
2299 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2300 		(void) ip_mc_leave_src(sk, iml, in_dev);
2301 		if (in_dev != NULL)
2302 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2303 		/* decrease mem now to avoid the memleak warning */
2304 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2305 		kfree_rcu(iml, rcu);
2306 	}
2307 	rtnl_unlock();
2308 }
2309 
2310 /* called with rcu_read_lock() */
2311 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2312 {
2313 	struct ip_mc_list *im;
2314 	struct ip_sf_list *psf;
2315 	int rv = 0;
2316 
2317 	for_each_pmc_rcu(in_dev, im) {
2318 		if (im->multiaddr == mc_addr)
2319 			break;
2320 	}
2321 	if (im && proto == IPPROTO_IGMP) {
2322 		rv = 1;
2323 	} else if (im) {
2324 		if (src_addr) {
2325 			for (psf=im->sources; psf; psf=psf->sf_next) {
2326 				if (psf->sf_inaddr == src_addr)
2327 					break;
2328 			}
2329 			if (psf)
2330 				rv = psf->sf_count[MCAST_INCLUDE] ||
2331 					psf->sf_count[MCAST_EXCLUDE] !=
2332 					im->sfcount[MCAST_EXCLUDE];
2333 			else
2334 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2335 		} else
2336 			rv = 1; /* unspecified source; tentatively allow */
2337 	}
2338 	return rv;
2339 }
2340 
2341 #if defined(CONFIG_PROC_FS)
2342 struct igmp_mc_iter_state {
2343 	struct seq_net_private p;
2344 	struct net_device *dev;
2345 	struct in_device *in_dev;
2346 };
2347 
2348 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2349 
2350 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2351 {
2352 	struct net *net = seq_file_net(seq);
2353 	struct ip_mc_list *im = NULL;
2354 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2355 
2356 	state->in_dev = NULL;
2357 	for_each_netdev_rcu(net, state->dev) {
2358 		struct in_device *in_dev;
2359 
2360 		in_dev = __in_dev_get_rcu(state->dev);
2361 		if (!in_dev)
2362 			continue;
2363 		im = rcu_dereference(in_dev->mc_list);
2364 		if (im) {
2365 			state->in_dev = in_dev;
2366 			break;
2367 		}
2368 	}
2369 	return im;
2370 }
2371 
2372 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2373 {
2374 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2375 
2376 	im = rcu_dereference(im->next_rcu);
2377 	while (!im) {
2378 		state->dev = next_net_device_rcu(state->dev);
2379 		if (!state->dev) {
2380 			state->in_dev = NULL;
2381 			break;
2382 		}
2383 		state->in_dev = __in_dev_get_rcu(state->dev);
2384 		if (!state->in_dev)
2385 			continue;
2386 		im = rcu_dereference(state->in_dev->mc_list);
2387 	}
2388 	return im;
2389 }
2390 
2391 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2392 {
2393 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2394 	if (im)
2395 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2396 			--pos;
2397 	return pos ? NULL : im;
2398 }
2399 
2400 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2401 	__acquires(rcu)
2402 {
2403 	rcu_read_lock();
2404 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2405 }
2406 
2407 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2408 {
2409 	struct ip_mc_list *im;
2410 	if (v == SEQ_START_TOKEN)
2411 		im = igmp_mc_get_first(seq);
2412 	else
2413 		im = igmp_mc_get_next(seq, v);
2414 	++*pos;
2415 	return im;
2416 }
2417 
2418 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2419 	__releases(rcu)
2420 {
2421 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2422 
2423 	state->in_dev = NULL;
2424 	state->dev = NULL;
2425 	rcu_read_unlock();
2426 }
2427 
2428 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2429 {
2430 	if (v == SEQ_START_TOKEN)
2431 		seq_puts(seq,
2432 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2433 	else {
2434 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2435 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2436 		char   *querier;
2437 #ifdef CONFIG_IP_MULTICAST
2438 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2439 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2440 			  "V3";
2441 #else
2442 		querier = "NONE";
2443 #endif
2444 
2445 		if (rcu_dereference(state->in_dev->mc_list) == im) {
2446 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2447 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2448 		}
2449 
2450 		seq_printf(seq,
2451 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2452 			   im->multiaddr, im->users,
2453 			   im->tm_running, im->tm_running ?
2454 			   jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2455 			   im->reporter);
2456 	}
2457 	return 0;
2458 }
2459 
2460 static const struct seq_operations igmp_mc_seq_ops = {
2461 	.start	=	igmp_mc_seq_start,
2462 	.next	=	igmp_mc_seq_next,
2463 	.stop	=	igmp_mc_seq_stop,
2464 	.show	=	igmp_mc_seq_show,
2465 };
2466 
2467 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2468 {
2469 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2470 			sizeof(struct igmp_mc_iter_state));
2471 }
2472 
2473 static const struct file_operations igmp_mc_seq_fops = {
2474 	.owner		=	THIS_MODULE,
2475 	.open		=	igmp_mc_seq_open,
2476 	.read		=	seq_read,
2477 	.llseek		=	seq_lseek,
2478 	.release	=	seq_release_net,
2479 };
2480 
2481 struct igmp_mcf_iter_state {
2482 	struct seq_net_private p;
2483 	struct net_device *dev;
2484 	struct in_device *idev;
2485 	struct ip_mc_list *im;
2486 };
2487 
2488 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2489 
2490 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2491 {
2492 	struct net *net = seq_file_net(seq);
2493 	struct ip_sf_list *psf = NULL;
2494 	struct ip_mc_list *im = NULL;
2495 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2496 
2497 	state->idev = NULL;
2498 	state->im = NULL;
2499 	for_each_netdev_rcu(net, state->dev) {
2500 		struct in_device *idev;
2501 		idev = __in_dev_get_rcu(state->dev);
2502 		if (unlikely(idev == NULL))
2503 			continue;
2504 		im = rcu_dereference(idev->mc_list);
2505 		if (likely(im != NULL)) {
2506 			spin_lock_bh(&im->lock);
2507 			psf = im->sources;
2508 			if (likely(psf != NULL)) {
2509 				state->im = im;
2510 				state->idev = idev;
2511 				break;
2512 			}
2513 			spin_unlock_bh(&im->lock);
2514 		}
2515 	}
2516 	return psf;
2517 }
2518 
2519 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2520 {
2521 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2522 
2523 	psf = psf->sf_next;
2524 	while (!psf) {
2525 		spin_unlock_bh(&state->im->lock);
2526 		state->im = state->im->next;
2527 		while (!state->im) {
2528 			state->dev = next_net_device_rcu(state->dev);
2529 			if (!state->dev) {
2530 				state->idev = NULL;
2531 				goto out;
2532 			}
2533 			state->idev = __in_dev_get_rcu(state->dev);
2534 			if (!state->idev)
2535 				continue;
2536 			state->im = rcu_dereference(state->idev->mc_list);
2537 		}
2538 		if (!state->im)
2539 			break;
2540 		spin_lock_bh(&state->im->lock);
2541 		psf = state->im->sources;
2542 	}
2543 out:
2544 	return psf;
2545 }
2546 
2547 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2548 {
2549 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2550 	if (psf)
2551 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2552 			--pos;
2553 	return pos ? NULL : psf;
2554 }
2555 
2556 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2557 	__acquires(rcu)
2558 {
2559 	rcu_read_lock();
2560 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2561 }
2562 
2563 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2564 {
2565 	struct ip_sf_list *psf;
2566 	if (v == SEQ_START_TOKEN)
2567 		psf = igmp_mcf_get_first(seq);
2568 	else
2569 		psf = igmp_mcf_get_next(seq, v);
2570 	++*pos;
2571 	return psf;
2572 }
2573 
2574 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2575 	__releases(rcu)
2576 {
2577 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2578 	if (likely(state->im != NULL)) {
2579 		spin_unlock_bh(&state->im->lock);
2580 		state->im = NULL;
2581 	}
2582 	state->idev = NULL;
2583 	state->dev = NULL;
2584 	rcu_read_unlock();
2585 }
2586 
2587 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2588 {
2589 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2590 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2591 
2592 	if (v == SEQ_START_TOKEN) {
2593 		seq_printf(seq,
2594 			   "%3s %6s "
2595 			   "%10s %10s %6s %6s\n", "Idx",
2596 			   "Device", "MCA",
2597 			   "SRC", "INC", "EXC");
2598 	} else {
2599 		seq_printf(seq,
2600 			   "%3d %6.6s 0x%08x "
2601 			   "0x%08x %6lu %6lu\n",
2602 			   state->dev->ifindex, state->dev->name,
2603 			   ntohl(state->im->multiaddr),
2604 			   ntohl(psf->sf_inaddr),
2605 			   psf->sf_count[MCAST_INCLUDE],
2606 			   psf->sf_count[MCAST_EXCLUDE]);
2607 	}
2608 	return 0;
2609 }
2610 
2611 static const struct seq_operations igmp_mcf_seq_ops = {
2612 	.start	=	igmp_mcf_seq_start,
2613 	.next	=	igmp_mcf_seq_next,
2614 	.stop	=	igmp_mcf_seq_stop,
2615 	.show	=	igmp_mcf_seq_show,
2616 };
2617 
2618 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2619 {
2620 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2621 			sizeof(struct igmp_mcf_iter_state));
2622 }
2623 
2624 static const struct file_operations igmp_mcf_seq_fops = {
2625 	.owner		=	THIS_MODULE,
2626 	.open		=	igmp_mcf_seq_open,
2627 	.read		=	seq_read,
2628 	.llseek		=	seq_lseek,
2629 	.release	=	seq_release_net,
2630 };
2631 
2632 static int __net_init igmp_net_init(struct net *net)
2633 {
2634 	struct proc_dir_entry *pde;
2635 
2636 	pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2637 	if (!pde)
2638 		goto out_igmp;
2639 	pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2640 	if (!pde)
2641 		goto out_mcfilter;
2642 	return 0;
2643 
2644 out_mcfilter:
2645 	proc_net_remove(net, "igmp");
2646 out_igmp:
2647 	return -ENOMEM;
2648 }
2649 
2650 static void __net_exit igmp_net_exit(struct net *net)
2651 {
2652 	proc_net_remove(net, "mcfilter");
2653 	proc_net_remove(net, "igmp");
2654 }
2655 
2656 static struct pernet_operations igmp_net_ops = {
2657 	.init = igmp_net_init,
2658 	.exit = igmp_net_exit,
2659 };
2660 
2661 int __init igmp_mc_proc_init(void)
2662 {
2663 	return register_pernet_subsys(&igmp_net_ops);
2664 }
2665 #endif
2666