xref: /linux/net/ipv4/igmp.c (revision 092e0e7e520a1fca03e13c9f2d157432a8657ff2)
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(im);
157 	}
158 }
159 
160 #ifdef CONFIG_IP_MULTICAST
161 
162 /*
163  *	Timer management
164  */
165 
166 static __inline__ void igmp_stop_timer(struct ip_mc_list *im)
167 {
168 	spin_lock_bh(&im->lock);
169 	if (del_timer(&im->timer))
170 		atomic_dec(&im->refcnt);
171 	im->tm_running = 0;
172 	im->reporter = 0;
173 	im->unsolicit_count = 0;
174 	spin_unlock_bh(&im->lock);
175 }
176 
177 /* It must be called with locked im->lock */
178 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
179 {
180 	int tv = net_random() % max_delay;
181 
182 	im->tm_running = 1;
183 	if (!mod_timer(&im->timer, jiffies+tv+2))
184 		atomic_inc(&im->refcnt);
185 }
186 
187 static void igmp_gq_start_timer(struct in_device *in_dev)
188 {
189 	int tv = net_random() % in_dev->mr_maxdelay;
190 
191 	in_dev->mr_gq_running = 1;
192 	if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
193 		in_dev_hold(in_dev);
194 }
195 
196 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
197 {
198 	int tv = net_random() % delay;
199 
200 	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
201 		in_dev_hold(in_dev);
202 }
203 
204 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
205 {
206 	spin_lock_bh(&im->lock);
207 	im->unsolicit_count = 0;
208 	if (del_timer(&im->timer)) {
209 		if ((long)(im->timer.expires-jiffies) < max_delay) {
210 			add_timer(&im->timer);
211 			im->tm_running = 1;
212 			spin_unlock_bh(&im->lock);
213 			return;
214 		}
215 		atomic_dec(&im->refcnt);
216 	}
217 	igmp_start_timer(im, max_delay);
218 	spin_unlock_bh(&im->lock);
219 }
220 
221 
222 /*
223  *	Send an IGMP report.
224  */
225 
226 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
227 
228 
229 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
230 	int gdeleted, int sdeleted)
231 {
232 	switch (type) {
233 	case IGMPV3_MODE_IS_INCLUDE:
234 	case IGMPV3_MODE_IS_EXCLUDE:
235 		if (gdeleted || sdeleted)
236 			return 0;
237 		if (!(pmc->gsquery && !psf->sf_gsresp)) {
238 			if (pmc->sfmode == MCAST_INCLUDE)
239 				return 1;
240 			/* don't include if this source is excluded
241 			 * in all filters
242 			 */
243 			if (psf->sf_count[MCAST_INCLUDE])
244 				return type == IGMPV3_MODE_IS_INCLUDE;
245 			return pmc->sfcount[MCAST_EXCLUDE] ==
246 				psf->sf_count[MCAST_EXCLUDE];
247 		}
248 		return 0;
249 	case IGMPV3_CHANGE_TO_INCLUDE:
250 		if (gdeleted || sdeleted)
251 			return 0;
252 		return psf->sf_count[MCAST_INCLUDE] != 0;
253 	case IGMPV3_CHANGE_TO_EXCLUDE:
254 		if (gdeleted || sdeleted)
255 			return 0;
256 		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
257 		    psf->sf_count[MCAST_INCLUDE])
258 			return 0;
259 		return pmc->sfcount[MCAST_EXCLUDE] ==
260 			psf->sf_count[MCAST_EXCLUDE];
261 	case IGMPV3_ALLOW_NEW_SOURCES:
262 		if (gdeleted || !psf->sf_crcount)
263 			return 0;
264 		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
265 	case IGMPV3_BLOCK_OLD_SOURCES:
266 		if (pmc->sfmode == MCAST_INCLUDE)
267 			return gdeleted || (psf->sf_crcount && sdeleted);
268 		return psf->sf_crcount && !gdeleted && !sdeleted;
269 	}
270 	return 0;
271 }
272 
273 static int
274 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
275 {
276 	struct ip_sf_list *psf;
277 	int scount = 0;
278 
279 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
280 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
281 			continue;
282 		scount++;
283 	}
284 	return scount;
285 }
286 
287 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
288 {
289 	struct sk_buff *skb;
290 	struct rtable *rt;
291 	struct iphdr *pip;
292 	struct igmpv3_report *pig;
293 	struct net *net = dev_net(dev);
294 
295 	skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
296 	if (skb == NULL)
297 		return NULL;
298 
299 	{
300 		struct flowi fl = { .oif = dev->ifindex,
301 				    .nl_u = { .ip4_u = {
302 				    .daddr = IGMPV3_ALL_MCR } },
303 				    .proto = IPPROTO_IGMP };
304 		if (ip_route_output_key(net, &rt, &fl)) {
305 			kfree_skb(skb);
306 			return NULL;
307 		}
308 	}
309 	if (rt->rt_src == 0) {
310 		kfree_skb(skb);
311 		ip_rt_put(rt);
312 		return NULL;
313 	}
314 
315 	skb_dst_set(skb, &rt->dst);
316 	skb->dev = dev;
317 
318 	skb_reserve(skb, LL_RESERVED_SPACE(dev));
319 
320 	skb_reset_network_header(skb);
321 	pip = ip_hdr(skb);
322 	skb_put(skb, sizeof(struct iphdr) + 4);
323 
324 	pip->version  = 4;
325 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
326 	pip->tos      = 0xc0;
327 	pip->frag_off = htons(IP_DF);
328 	pip->ttl      = 1;
329 	pip->daddr    = rt->rt_dst;
330 	pip->saddr    = rt->rt_src;
331 	pip->protocol = IPPROTO_IGMP;
332 	pip->tot_len  = 0;	/* filled in later */
333 	ip_select_ident(pip, &rt->dst, NULL);
334 	((u8*)&pip[1])[0] = IPOPT_RA;
335 	((u8*)&pip[1])[1] = 4;
336 	((u8*)&pip[1])[2] = 0;
337 	((u8*)&pip[1])[3] = 0;
338 
339 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
340 	skb_put(skb, sizeof(*pig));
341 	pig = igmpv3_report_hdr(skb);
342 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
343 	pig->resv1 = 0;
344 	pig->csum = 0;
345 	pig->resv2 = 0;
346 	pig->ngrec = 0;
347 	return skb;
348 }
349 
350 static int igmpv3_sendpack(struct sk_buff *skb)
351 {
352 	struct igmphdr *pig = igmp_hdr(skb);
353 	const int igmplen = skb->tail - skb->transport_header;
354 
355 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
356 
357 	return ip_local_out(skb);
358 }
359 
360 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
361 {
362 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
363 }
364 
365 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
366 	int type, struct igmpv3_grec **ppgr)
367 {
368 	struct net_device *dev = pmc->interface->dev;
369 	struct igmpv3_report *pih;
370 	struct igmpv3_grec *pgr;
371 
372 	if (!skb)
373 		skb = igmpv3_newpack(dev, dev->mtu);
374 	if (!skb)
375 		return NULL;
376 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
377 	pgr->grec_type = type;
378 	pgr->grec_auxwords = 0;
379 	pgr->grec_nsrcs = 0;
380 	pgr->grec_mca = pmc->multiaddr;
381 	pih = igmpv3_report_hdr(skb);
382 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
383 	*ppgr = pgr;
384 	return skb;
385 }
386 
387 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? (skb)->dev->mtu - (skb)->len : \
388 	skb_tailroom(skb)) : 0)
389 
390 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
391 	int type, int gdeleted, int sdeleted)
392 {
393 	struct net_device *dev = pmc->interface->dev;
394 	struct igmpv3_report *pih;
395 	struct igmpv3_grec *pgr = NULL;
396 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
397 	int scount, stotal, first, isquery, truncate;
398 
399 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
400 		return skb;
401 
402 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
403 		  type == IGMPV3_MODE_IS_EXCLUDE;
404 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
405 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
406 
407 	stotal = scount = 0;
408 
409 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
410 
411 	if (!*psf_list)
412 		goto empty_source;
413 
414 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
415 
416 	/* EX and TO_EX get a fresh packet, if needed */
417 	if (truncate) {
418 		if (pih && pih->ngrec &&
419 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
420 			if (skb)
421 				igmpv3_sendpack(skb);
422 			skb = igmpv3_newpack(dev, dev->mtu);
423 		}
424 	}
425 	first = 1;
426 	psf_prev = NULL;
427 	for (psf=*psf_list; psf; psf=psf_next) {
428 		__be32 *psrc;
429 
430 		psf_next = psf->sf_next;
431 
432 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
433 			psf_prev = psf;
434 			continue;
435 		}
436 
437 		/* clear marks on query responses */
438 		if (isquery)
439 			psf->sf_gsresp = 0;
440 
441 		if (AVAILABLE(skb) < sizeof(__be32) +
442 		    first*sizeof(struct igmpv3_grec)) {
443 			if (truncate && !first)
444 				break;	 /* truncate these */
445 			if (pgr)
446 				pgr->grec_nsrcs = htons(scount);
447 			if (skb)
448 				igmpv3_sendpack(skb);
449 			skb = igmpv3_newpack(dev, dev->mtu);
450 			first = 1;
451 			scount = 0;
452 		}
453 		if (first) {
454 			skb = add_grhead(skb, pmc, type, &pgr);
455 			first = 0;
456 		}
457 		if (!skb)
458 			return NULL;
459 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
460 		*psrc = psf->sf_inaddr;
461 		scount++; stotal++;
462 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
463 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
464 			psf->sf_crcount--;
465 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
466 				if (psf_prev)
467 					psf_prev->sf_next = psf->sf_next;
468 				else
469 					*psf_list = psf->sf_next;
470 				kfree(psf);
471 				continue;
472 			}
473 		}
474 		psf_prev = psf;
475 	}
476 
477 empty_source:
478 	if (!stotal) {
479 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
480 		    type == IGMPV3_BLOCK_OLD_SOURCES)
481 			return skb;
482 		if (pmc->crcount || isquery) {
483 			/* make sure we have room for group header */
484 			if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
485 				igmpv3_sendpack(skb);
486 				skb = NULL; /* add_grhead will get a new one */
487 			}
488 			skb = add_grhead(skb, pmc, type, &pgr);
489 		}
490 	}
491 	if (pgr)
492 		pgr->grec_nsrcs = htons(scount);
493 
494 	if (isquery)
495 		pmc->gsquery = 0;	/* clear query state on report */
496 	return skb;
497 }
498 
499 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
500 {
501 	struct sk_buff *skb = NULL;
502 	int type;
503 
504 	if (!pmc) {
505 		read_lock(&in_dev->mc_list_lock);
506 		for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
507 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
508 				continue;
509 			spin_lock_bh(&pmc->lock);
510 			if (pmc->sfcount[MCAST_EXCLUDE])
511 				type = IGMPV3_MODE_IS_EXCLUDE;
512 			else
513 				type = IGMPV3_MODE_IS_INCLUDE;
514 			skb = add_grec(skb, pmc, type, 0, 0);
515 			spin_unlock_bh(&pmc->lock);
516 		}
517 		read_unlock(&in_dev->mc_list_lock);
518 	} else {
519 		spin_lock_bh(&pmc->lock);
520 		if (pmc->sfcount[MCAST_EXCLUDE])
521 			type = IGMPV3_MODE_IS_EXCLUDE;
522 		else
523 			type = IGMPV3_MODE_IS_INCLUDE;
524 		skb = add_grec(skb, pmc, type, 0, 0);
525 		spin_unlock_bh(&pmc->lock);
526 	}
527 	if (!skb)
528 		return 0;
529 	return igmpv3_sendpack(skb);
530 }
531 
532 /*
533  * remove zero-count source records from a source filter list
534  */
535 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
536 {
537 	struct ip_sf_list *psf_prev, *psf_next, *psf;
538 
539 	psf_prev = NULL;
540 	for (psf=*ppsf; psf; psf = psf_next) {
541 		psf_next = psf->sf_next;
542 		if (psf->sf_crcount == 0) {
543 			if (psf_prev)
544 				psf_prev->sf_next = psf->sf_next;
545 			else
546 				*ppsf = psf->sf_next;
547 			kfree(psf);
548 		} else
549 			psf_prev = psf;
550 	}
551 }
552 
553 static void igmpv3_send_cr(struct in_device *in_dev)
554 {
555 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
556 	struct sk_buff *skb = NULL;
557 	int type, dtype;
558 
559 	read_lock(&in_dev->mc_list_lock);
560 	spin_lock_bh(&in_dev->mc_tomb_lock);
561 
562 	/* deleted MCA's */
563 	pmc_prev = NULL;
564 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
565 		pmc_next = pmc->next;
566 		if (pmc->sfmode == MCAST_INCLUDE) {
567 			type = IGMPV3_BLOCK_OLD_SOURCES;
568 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
569 			skb = add_grec(skb, pmc, type, 1, 0);
570 			skb = add_grec(skb, pmc, dtype, 1, 1);
571 		}
572 		if (pmc->crcount) {
573 			if (pmc->sfmode == MCAST_EXCLUDE) {
574 				type = IGMPV3_CHANGE_TO_INCLUDE;
575 				skb = add_grec(skb, pmc, type, 1, 0);
576 			}
577 			pmc->crcount--;
578 			if (pmc->crcount == 0) {
579 				igmpv3_clear_zeros(&pmc->tomb);
580 				igmpv3_clear_zeros(&pmc->sources);
581 			}
582 		}
583 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
584 			if (pmc_prev)
585 				pmc_prev->next = pmc_next;
586 			else
587 				in_dev->mc_tomb = pmc_next;
588 			in_dev_put(pmc->interface);
589 			kfree(pmc);
590 		} else
591 			pmc_prev = pmc;
592 	}
593 	spin_unlock_bh(&in_dev->mc_tomb_lock);
594 
595 	/* change recs */
596 	for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
597 		spin_lock_bh(&pmc->lock);
598 		if (pmc->sfcount[MCAST_EXCLUDE]) {
599 			type = IGMPV3_BLOCK_OLD_SOURCES;
600 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
601 		} else {
602 			type = IGMPV3_ALLOW_NEW_SOURCES;
603 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
604 		}
605 		skb = add_grec(skb, pmc, type, 0, 0);
606 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
607 
608 		/* filter mode changes */
609 		if (pmc->crcount) {
610 			if (pmc->sfmode == MCAST_EXCLUDE)
611 				type = IGMPV3_CHANGE_TO_EXCLUDE;
612 			else
613 				type = IGMPV3_CHANGE_TO_INCLUDE;
614 			skb = add_grec(skb, pmc, type, 0, 0);
615 			pmc->crcount--;
616 		}
617 		spin_unlock_bh(&pmc->lock);
618 	}
619 	read_unlock(&in_dev->mc_list_lock);
620 
621 	if (!skb)
622 		return;
623 	(void) igmpv3_sendpack(skb);
624 }
625 
626 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
627 	int type)
628 {
629 	struct sk_buff *skb;
630 	struct iphdr *iph;
631 	struct igmphdr *ih;
632 	struct rtable *rt;
633 	struct net_device *dev = in_dev->dev;
634 	struct net *net = dev_net(dev);
635 	__be32	group = pmc ? pmc->multiaddr : 0;
636 	__be32	dst;
637 
638 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
639 		return igmpv3_send_report(in_dev, pmc);
640 	else if (type == IGMP_HOST_LEAVE_MESSAGE)
641 		dst = IGMP_ALL_ROUTER;
642 	else
643 		dst = group;
644 
645 	{
646 		struct flowi fl = { .oif = dev->ifindex,
647 				    .nl_u = { .ip4_u = { .daddr = dst } },
648 				    .proto = IPPROTO_IGMP };
649 		if (ip_route_output_key(net, &rt, &fl))
650 			return -1;
651 	}
652 	if (rt->rt_src == 0) {
653 		ip_rt_put(rt);
654 		return -1;
655 	}
656 
657 	skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
658 	if (skb == NULL) {
659 		ip_rt_put(rt);
660 		return -1;
661 	}
662 
663 	skb_dst_set(skb, &rt->dst);
664 
665 	skb_reserve(skb, LL_RESERVED_SPACE(dev));
666 
667 	skb_reset_network_header(skb);
668 	iph = ip_hdr(skb);
669 	skb_put(skb, sizeof(struct iphdr) + 4);
670 
671 	iph->version  = 4;
672 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
673 	iph->tos      = 0xc0;
674 	iph->frag_off = htons(IP_DF);
675 	iph->ttl      = 1;
676 	iph->daddr    = dst;
677 	iph->saddr    = rt->rt_src;
678 	iph->protocol = IPPROTO_IGMP;
679 	ip_select_ident(iph, &rt->dst, NULL);
680 	((u8*)&iph[1])[0] = IPOPT_RA;
681 	((u8*)&iph[1])[1] = 4;
682 	((u8*)&iph[1])[2] = 0;
683 	((u8*)&iph[1])[3] = 0;
684 
685 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
686 	ih->type = type;
687 	ih->code = 0;
688 	ih->csum = 0;
689 	ih->group = group;
690 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
691 
692 	return ip_local_out(skb);
693 }
694 
695 static void igmp_gq_timer_expire(unsigned long data)
696 {
697 	struct in_device *in_dev = (struct in_device *)data;
698 
699 	in_dev->mr_gq_running = 0;
700 	igmpv3_send_report(in_dev, NULL);
701 	__in_dev_put(in_dev);
702 }
703 
704 static void igmp_ifc_timer_expire(unsigned long data)
705 {
706 	struct in_device *in_dev = (struct in_device *)data;
707 
708 	igmpv3_send_cr(in_dev);
709 	if (in_dev->mr_ifc_count) {
710 		in_dev->mr_ifc_count--;
711 		igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
712 	}
713 	__in_dev_put(in_dev);
714 }
715 
716 static void igmp_ifc_event(struct in_device *in_dev)
717 {
718 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
719 		return;
720 	in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
721 		IGMP_Unsolicited_Report_Count;
722 	igmp_ifc_start_timer(in_dev, 1);
723 }
724 
725 
726 static void igmp_timer_expire(unsigned long data)
727 {
728 	struct ip_mc_list *im=(struct ip_mc_list *)data;
729 	struct in_device *in_dev = im->interface;
730 
731 	spin_lock(&im->lock);
732 	im->tm_running = 0;
733 
734 	if (im->unsolicit_count) {
735 		im->unsolicit_count--;
736 		igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
737 	}
738 	im->reporter = 1;
739 	spin_unlock(&im->lock);
740 
741 	if (IGMP_V1_SEEN(in_dev))
742 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
743 	else if (IGMP_V2_SEEN(in_dev))
744 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
745 	else
746 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
747 
748 	ip_ma_put(im);
749 }
750 
751 /* mark EXCLUDE-mode sources */
752 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
753 {
754 	struct ip_sf_list *psf;
755 	int i, scount;
756 
757 	scount = 0;
758 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
759 		if (scount == nsrcs)
760 			break;
761 		for (i=0; i<nsrcs; i++) {
762 			/* skip inactive filters */
763 			if (pmc->sfcount[MCAST_INCLUDE] ||
764 			    pmc->sfcount[MCAST_EXCLUDE] !=
765 			    psf->sf_count[MCAST_EXCLUDE])
766 				continue;
767 			if (srcs[i] == psf->sf_inaddr) {
768 				scount++;
769 				break;
770 			}
771 		}
772 	}
773 	pmc->gsquery = 0;
774 	if (scount == nsrcs)	/* all sources excluded */
775 		return 0;
776 	return 1;
777 }
778 
779 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
780 {
781 	struct ip_sf_list *psf;
782 	int i, scount;
783 
784 	if (pmc->sfmode == MCAST_EXCLUDE)
785 		return igmp_xmarksources(pmc, nsrcs, srcs);
786 
787 	/* mark INCLUDE-mode sources */
788 	scount = 0;
789 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
790 		if (scount == nsrcs)
791 			break;
792 		for (i=0; i<nsrcs; i++)
793 			if (srcs[i] == psf->sf_inaddr) {
794 				psf->sf_gsresp = 1;
795 				scount++;
796 				break;
797 			}
798 	}
799 	if (!scount) {
800 		pmc->gsquery = 0;
801 		return 0;
802 	}
803 	pmc->gsquery = 1;
804 	return 1;
805 }
806 
807 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
808 {
809 	struct ip_mc_list *im;
810 
811 	/* Timers are only set for non-local groups */
812 
813 	if (group == IGMP_ALL_HOSTS)
814 		return;
815 
816 	read_lock(&in_dev->mc_list_lock);
817 	for (im=in_dev->mc_list; im!=NULL; im=im->next) {
818 		if (im->multiaddr == group) {
819 			igmp_stop_timer(im);
820 			break;
821 		}
822 	}
823 	read_unlock(&in_dev->mc_list_lock);
824 }
825 
826 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
827 	int len)
828 {
829 	struct igmphdr 		*ih = igmp_hdr(skb);
830 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
831 	struct ip_mc_list	*im;
832 	__be32			group = ih->group;
833 	int			max_delay;
834 	int			mark = 0;
835 
836 
837 	if (len == 8) {
838 		if (ih->code == 0) {
839 			/* Alas, old v1 router presents here. */
840 
841 			max_delay = IGMP_Query_Response_Interval;
842 			in_dev->mr_v1_seen = jiffies +
843 				IGMP_V1_Router_Present_Timeout;
844 			group = 0;
845 		} else {
846 			/* v2 router present */
847 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
848 			in_dev->mr_v2_seen = jiffies +
849 				IGMP_V2_Router_Present_Timeout;
850 		}
851 		/* cancel the interface change timer */
852 		in_dev->mr_ifc_count = 0;
853 		if (del_timer(&in_dev->mr_ifc_timer))
854 			__in_dev_put(in_dev);
855 		/* clear deleted report items */
856 		igmpv3_clear_delrec(in_dev);
857 	} else if (len < 12) {
858 		return;	/* ignore bogus packet; freed by caller */
859 	} else if (IGMP_V1_SEEN(in_dev)) {
860 		/* This is a v3 query with v1 queriers present */
861 		max_delay = IGMP_Query_Response_Interval;
862 		group = 0;
863 	} else if (IGMP_V2_SEEN(in_dev)) {
864 		/* this is a v3 query with v2 queriers present;
865 		 * Interpretation of the max_delay code is problematic here.
866 		 * A real v2 host would use ih_code directly, while v3 has a
867 		 * different encoding. We use the v3 encoding as more likely
868 		 * to be intended in a v3 query.
869 		 */
870 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
871 	} else { /* v3 */
872 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
873 			return;
874 
875 		ih3 = igmpv3_query_hdr(skb);
876 		if (ih3->nsrcs) {
877 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
878 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
879 				return;
880 			ih3 = igmpv3_query_hdr(skb);
881 		}
882 
883 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
884 		if (!max_delay)
885 			max_delay = 1;	/* can't mod w/ 0 */
886 		in_dev->mr_maxdelay = max_delay;
887 		if (ih3->qrv)
888 			in_dev->mr_qrv = ih3->qrv;
889 		if (!group) { /* general query */
890 			if (ih3->nsrcs)
891 				return;	/* no sources allowed */
892 			igmp_gq_start_timer(in_dev);
893 			return;
894 		}
895 		/* mark sources to include, if group & source-specific */
896 		mark = ih3->nsrcs != 0;
897 	}
898 
899 	/*
900 	 * - Start the timers in all of our membership records
901 	 *   that the query applies to for the interface on
902 	 *   which the query arrived excl. those that belong
903 	 *   to a "local" group (224.0.0.X)
904 	 * - For timers already running check if they need to
905 	 *   be reset.
906 	 * - Use the igmp->igmp_code field as the maximum
907 	 *   delay possible
908 	 */
909 	read_lock(&in_dev->mc_list_lock);
910 	for (im=in_dev->mc_list; im!=NULL; im=im->next) {
911 		int changed;
912 
913 		if (group && group != im->multiaddr)
914 			continue;
915 		if (im->multiaddr == IGMP_ALL_HOSTS)
916 			continue;
917 		spin_lock_bh(&im->lock);
918 		if (im->tm_running)
919 			im->gsquery = im->gsquery && mark;
920 		else
921 			im->gsquery = mark;
922 		changed = !im->gsquery ||
923 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
924 		spin_unlock_bh(&im->lock);
925 		if (changed)
926 			igmp_mod_timer(im, max_delay);
927 	}
928 	read_unlock(&in_dev->mc_list_lock);
929 }
930 
931 /* called in rcu_read_lock() section */
932 int igmp_rcv(struct sk_buff *skb)
933 {
934 	/* This basically follows the spec line by line -- see RFC1112 */
935 	struct igmphdr *ih;
936 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
937 	int len = skb->len;
938 
939 	if (in_dev == NULL)
940 		goto drop;
941 
942 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
943 		goto drop;
944 
945 	switch (skb->ip_summed) {
946 	case CHECKSUM_COMPLETE:
947 		if (!csum_fold(skb->csum))
948 			break;
949 		/* fall through */
950 	case CHECKSUM_NONE:
951 		skb->csum = 0;
952 		if (__skb_checksum_complete(skb))
953 			goto drop;
954 	}
955 
956 	ih = igmp_hdr(skb);
957 	switch (ih->type) {
958 	case IGMP_HOST_MEMBERSHIP_QUERY:
959 		igmp_heard_query(in_dev, skb, len);
960 		break;
961 	case IGMP_HOST_MEMBERSHIP_REPORT:
962 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
963 		/* Is it our report looped back? */
964 		if (skb_rtable(skb)->fl.iif == 0)
965 			break;
966 		/* don't rely on MC router hearing unicast reports */
967 		if (skb->pkt_type == PACKET_MULTICAST ||
968 		    skb->pkt_type == PACKET_BROADCAST)
969 			igmp_heard_report(in_dev, ih->group);
970 		break;
971 	case IGMP_PIM:
972 #ifdef CONFIG_IP_PIMSM_V1
973 		return pim_rcv_v1(skb);
974 #endif
975 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
976 	case IGMP_DVMRP:
977 	case IGMP_TRACE:
978 	case IGMP_HOST_LEAVE_MESSAGE:
979 	case IGMP_MTRACE:
980 	case IGMP_MTRACE_RESP:
981 		break;
982 	default:
983 		break;
984 	}
985 
986 drop:
987 	kfree_skb(skb);
988 	return 0;
989 }
990 
991 #endif
992 
993 
994 /*
995  *	Add a filter to a device
996  */
997 
998 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
999 {
1000 	char buf[MAX_ADDR_LEN];
1001 	struct net_device *dev = in_dev->dev;
1002 
1003 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1004 	   We will get multicast token leakage, when IFF_MULTICAST
1005 	   is changed. This check should be done in dev->set_multicast_list
1006 	   routine. Something sort of:
1007 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1008 	   --ANK
1009 	   */
1010 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1011 		dev_mc_add(dev, buf);
1012 }
1013 
1014 /*
1015  *	Remove a filter from a device
1016  */
1017 
1018 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1019 {
1020 	char buf[MAX_ADDR_LEN];
1021 	struct net_device *dev = in_dev->dev;
1022 
1023 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1024 		dev_mc_del(dev, buf);
1025 }
1026 
1027 #ifdef CONFIG_IP_MULTICAST
1028 /*
1029  * deleted ip_mc_list manipulation
1030  */
1031 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1032 {
1033 	struct ip_mc_list *pmc;
1034 
1035 	/* this is an "ip_mc_list" for convenience; only the fields below
1036 	 * are actually used. In particular, the refcnt and users are not
1037 	 * used for management of the delete list. Using the same structure
1038 	 * for deleted items allows change reports to use common code with
1039 	 * non-deleted or query-response MCA's.
1040 	 */
1041 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1042 	if (!pmc)
1043 		return;
1044 	spin_lock_bh(&im->lock);
1045 	pmc->interface = im->interface;
1046 	in_dev_hold(in_dev);
1047 	pmc->multiaddr = im->multiaddr;
1048 	pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1049 		IGMP_Unsolicited_Report_Count;
1050 	pmc->sfmode = im->sfmode;
1051 	if (pmc->sfmode == MCAST_INCLUDE) {
1052 		struct ip_sf_list *psf;
1053 
1054 		pmc->tomb = im->tomb;
1055 		pmc->sources = im->sources;
1056 		im->tomb = im->sources = NULL;
1057 		for (psf=pmc->sources; psf; psf=psf->sf_next)
1058 			psf->sf_crcount = pmc->crcount;
1059 	}
1060 	spin_unlock_bh(&im->lock);
1061 
1062 	spin_lock_bh(&in_dev->mc_tomb_lock);
1063 	pmc->next = in_dev->mc_tomb;
1064 	in_dev->mc_tomb = pmc;
1065 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1066 }
1067 
1068 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1069 {
1070 	struct ip_mc_list *pmc, *pmc_prev;
1071 	struct ip_sf_list *psf, *psf_next;
1072 
1073 	spin_lock_bh(&in_dev->mc_tomb_lock);
1074 	pmc_prev = NULL;
1075 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1076 		if (pmc->multiaddr == multiaddr)
1077 			break;
1078 		pmc_prev = pmc;
1079 	}
1080 	if (pmc) {
1081 		if (pmc_prev)
1082 			pmc_prev->next = pmc->next;
1083 		else
1084 			in_dev->mc_tomb = pmc->next;
1085 	}
1086 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1087 	if (pmc) {
1088 		for (psf=pmc->tomb; psf; psf=psf_next) {
1089 			psf_next = psf->sf_next;
1090 			kfree(psf);
1091 		}
1092 		in_dev_put(pmc->interface);
1093 		kfree(pmc);
1094 	}
1095 }
1096 
1097 static void igmpv3_clear_delrec(struct in_device *in_dev)
1098 {
1099 	struct ip_mc_list *pmc, *nextpmc;
1100 
1101 	spin_lock_bh(&in_dev->mc_tomb_lock);
1102 	pmc = in_dev->mc_tomb;
1103 	in_dev->mc_tomb = NULL;
1104 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1105 
1106 	for (; pmc; pmc = nextpmc) {
1107 		nextpmc = pmc->next;
1108 		ip_mc_clear_src(pmc);
1109 		in_dev_put(pmc->interface);
1110 		kfree(pmc);
1111 	}
1112 	/* clear dead sources, too */
1113 	read_lock(&in_dev->mc_list_lock);
1114 	for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
1115 		struct ip_sf_list *psf, *psf_next;
1116 
1117 		spin_lock_bh(&pmc->lock);
1118 		psf = pmc->tomb;
1119 		pmc->tomb = NULL;
1120 		spin_unlock_bh(&pmc->lock);
1121 		for (; psf; psf=psf_next) {
1122 			psf_next = psf->sf_next;
1123 			kfree(psf);
1124 		}
1125 	}
1126 	read_unlock(&in_dev->mc_list_lock);
1127 }
1128 #endif
1129 
1130 static void igmp_group_dropped(struct ip_mc_list *im)
1131 {
1132 	struct in_device *in_dev = im->interface;
1133 #ifdef CONFIG_IP_MULTICAST
1134 	int reporter;
1135 #endif
1136 
1137 	if (im->loaded) {
1138 		im->loaded = 0;
1139 		ip_mc_filter_del(in_dev, im->multiaddr);
1140 	}
1141 
1142 #ifdef CONFIG_IP_MULTICAST
1143 	if (im->multiaddr == IGMP_ALL_HOSTS)
1144 		return;
1145 
1146 	reporter = im->reporter;
1147 	igmp_stop_timer(im);
1148 
1149 	if (!in_dev->dead) {
1150 		if (IGMP_V1_SEEN(in_dev))
1151 			goto done;
1152 		if (IGMP_V2_SEEN(in_dev)) {
1153 			if (reporter)
1154 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1155 			goto done;
1156 		}
1157 		/* IGMPv3 */
1158 		igmpv3_add_delrec(in_dev, im);
1159 
1160 		igmp_ifc_event(in_dev);
1161 	}
1162 done:
1163 #endif
1164 	ip_mc_clear_src(im);
1165 }
1166 
1167 static void igmp_group_added(struct ip_mc_list *im)
1168 {
1169 	struct in_device *in_dev = im->interface;
1170 
1171 	if (im->loaded == 0) {
1172 		im->loaded = 1;
1173 		ip_mc_filter_add(in_dev, im->multiaddr);
1174 	}
1175 
1176 #ifdef CONFIG_IP_MULTICAST
1177 	if (im->multiaddr == IGMP_ALL_HOSTS)
1178 		return;
1179 
1180 	if (in_dev->dead)
1181 		return;
1182 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1183 		spin_lock_bh(&im->lock);
1184 		igmp_start_timer(im, IGMP_Initial_Report_Delay);
1185 		spin_unlock_bh(&im->lock);
1186 		return;
1187 	}
1188 	/* else, v3 */
1189 
1190 	im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1191 		IGMP_Unsolicited_Report_Count;
1192 	igmp_ifc_event(in_dev);
1193 #endif
1194 }
1195 
1196 
1197 /*
1198  *	Multicast list managers
1199  */
1200 
1201 
1202 /*
1203  *	A socket has joined a multicast group on device dev.
1204  */
1205 
1206 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1207 {
1208 	struct ip_mc_list *im;
1209 
1210 	ASSERT_RTNL();
1211 
1212 	for (im=in_dev->mc_list; im; im=im->next) {
1213 		if (im->multiaddr == addr) {
1214 			im->users++;
1215 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1216 			goto out;
1217 		}
1218 	}
1219 
1220 	im = kmalloc(sizeof(*im), GFP_KERNEL);
1221 	if (!im)
1222 		goto out;
1223 
1224 	im->users = 1;
1225 	im->interface = in_dev;
1226 	in_dev_hold(in_dev);
1227 	im->multiaddr = addr;
1228 	/* initial mode is (EX, empty) */
1229 	im->sfmode = MCAST_EXCLUDE;
1230 	im->sfcount[MCAST_INCLUDE] = 0;
1231 	im->sfcount[MCAST_EXCLUDE] = 1;
1232 	im->sources = NULL;
1233 	im->tomb = NULL;
1234 	im->crcount = 0;
1235 	atomic_set(&im->refcnt, 1);
1236 	spin_lock_init(&im->lock);
1237 #ifdef CONFIG_IP_MULTICAST
1238 	im->tm_running = 0;
1239 	setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1240 	im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1241 	im->reporter = 0;
1242 	im->gsquery = 0;
1243 #endif
1244 	im->loaded = 0;
1245 	write_lock_bh(&in_dev->mc_list_lock);
1246 	im->next = in_dev->mc_list;
1247 	in_dev->mc_list = im;
1248 	in_dev->mc_count++;
1249 	write_unlock_bh(&in_dev->mc_list_lock);
1250 #ifdef CONFIG_IP_MULTICAST
1251 	igmpv3_del_delrec(in_dev, im->multiaddr);
1252 #endif
1253 	igmp_group_added(im);
1254 	if (!in_dev->dead)
1255 		ip_rt_multicast_event(in_dev);
1256 out:
1257 	return;
1258 }
1259 EXPORT_SYMBOL(ip_mc_inc_group);
1260 
1261 /*
1262  *	Resend IGMP JOIN report; used for bonding.
1263  */
1264 void ip_mc_rejoin_group(struct ip_mc_list *im)
1265 {
1266 #ifdef CONFIG_IP_MULTICAST
1267 	struct in_device *in_dev = im->interface;
1268 
1269 	if (im->multiaddr == IGMP_ALL_HOSTS)
1270 		return;
1271 
1272 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1273 		igmp_mod_timer(im, IGMP_Initial_Report_Delay);
1274 		return;
1275 	}
1276 	/* else, v3 */
1277 	im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1278 		IGMP_Unsolicited_Report_Count;
1279 	igmp_ifc_event(in_dev);
1280 #endif
1281 }
1282 EXPORT_SYMBOL(ip_mc_rejoin_group);
1283 
1284 /*
1285  *	A socket has left a multicast group on device dev
1286  */
1287 
1288 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1289 {
1290 	struct ip_mc_list *i, **ip;
1291 
1292 	ASSERT_RTNL();
1293 
1294 	for (ip=&in_dev->mc_list; (i=*ip)!=NULL; ip=&i->next) {
1295 		if (i->multiaddr == addr) {
1296 			if (--i->users == 0) {
1297 				write_lock_bh(&in_dev->mc_list_lock);
1298 				*ip = i->next;
1299 				in_dev->mc_count--;
1300 				write_unlock_bh(&in_dev->mc_list_lock);
1301 				igmp_group_dropped(i);
1302 
1303 				if (!in_dev->dead)
1304 					ip_rt_multicast_event(in_dev);
1305 
1306 				ip_ma_put(i);
1307 				return;
1308 			}
1309 			break;
1310 		}
1311 	}
1312 }
1313 EXPORT_SYMBOL(ip_mc_dec_group);
1314 
1315 /* Device changing type */
1316 
1317 void ip_mc_unmap(struct in_device *in_dev)
1318 {
1319 	struct ip_mc_list *i;
1320 
1321 	ASSERT_RTNL();
1322 
1323 	for (i = in_dev->mc_list; i; i = i->next)
1324 		igmp_group_dropped(i);
1325 }
1326 
1327 void ip_mc_remap(struct in_device *in_dev)
1328 {
1329 	struct ip_mc_list *i;
1330 
1331 	ASSERT_RTNL();
1332 
1333 	for (i = in_dev->mc_list; i; i = i->next)
1334 		igmp_group_added(i);
1335 }
1336 
1337 /* Device going down */
1338 
1339 void ip_mc_down(struct in_device *in_dev)
1340 {
1341 	struct ip_mc_list *i;
1342 
1343 	ASSERT_RTNL();
1344 
1345 	for (i=in_dev->mc_list; i; i=i->next)
1346 		igmp_group_dropped(i);
1347 
1348 #ifdef CONFIG_IP_MULTICAST
1349 	in_dev->mr_ifc_count = 0;
1350 	if (del_timer(&in_dev->mr_ifc_timer))
1351 		__in_dev_put(in_dev);
1352 	in_dev->mr_gq_running = 0;
1353 	if (del_timer(&in_dev->mr_gq_timer))
1354 		__in_dev_put(in_dev);
1355 	igmpv3_clear_delrec(in_dev);
1356 #endif
1357 
1358 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1359 }
1360 
1361 void ip_mc_init_dev(struct in_device *in_dev)
1362 {
1363 	ASSERT_RTNL();
1364 
1365 	in_dev->mc_tomb = NULL;
1366 #ifdef CONFIG_IP_MULTICAST
1367 	in_dev->mr_gq_running = 0;
1368 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1369 			(unsigned long)in_dev);
1370 	in_dev->mr_ifc_count = 0;
1371 	in_dev->mc_count     = 0;
1372 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1373 			(unsigned long)in_dev);
1374 	in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1375 #endif
1376 
1377 	rwlock_init(&in_dev->mc_list_lock);
1378 	spin_lock_init(&in_dev->mc_tomb_lock);
1379 }
1380 
1381 /* Device going up */
1382 
1383 void ip_mc_up(struct in_device *in_dev)
1384 {
1385 	struct ip_mc_list *i;
1386 
1387 	ASSERT_RTNL();
1388 
1389 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1390 
1391 	for (i=in_dev->mc_list; i; i=i->next)
1392 		igmp_group_added(i);
1393 }
1394 
1395 /*
1396  *	Device is about to be destroyed: clean up.
1397  */
1398 
1399 void ip_mc_destroy_dev(struct in_device *in_dev)
1400 {
1401 	struct ip_mc_list *i;
1402 
1403 	ASSERT_RTNL();
1404 
1405 	/* Deactivate timers */
1406 	ip_mc_down(in_dev);
1407 
1408 	write_lock_bh(&in_dev->mc_list_lock);
1409 	while ((i = in_dev->mc_list) != NULL) {
1410 		in_dev->mc_list = i->next;
1411 		in_dev->mc_count--;
1412 		write_unlock_bh(&in_dev->mc_list_lock);
1413 		igmp_group_dropped(i);
1414 		ip_ma_put(i);
1415 
1416 		write_lock_bh(&in_dev->mc_list_lock);
1417 	}
1418 	write_unlock_bh(&in_dev->mc_list_lock);
1419 }
1420 
1421 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1422 {
1423 	struct flowi fl = { .nl_u = { .ip4_u =
1424 				      { .daddr = imr->imr_multiaddr.s_addr } } };
1425 	struct rtable *rt;
1426 	struct net_device *dev = NULL;
1427 	struct in_device *idev = NULL;
1428 
1429 	if (imr->imr_ifindex) {
1430 		idev = inetdev_by_index(net, imr->imr_ifindex);
1431 		if (idev)
1432 			__in_dev_put(idev);
1433 		return idev;
1434 	}
1435 	if (imr->imr_address.s_addr) {
1436 		dev = ip_dev_find(net, imr->imr_address.s_addr);
1437 		if (!dev)
1438 			return NULL;
1439 		dev_put(dev);
1440 	}
1441 
1442 	if (!dev && !ip_route_output_key(net, &rt, &fl)) {
1443 		dev = rt->dst.dev;
1444 		ip_rt_put(rt);
1445 	}
1446 	if (dev) {
1447 		imr->imr_ifindex = dev->ifindex;
1448 		idev = __in_dev_get_rtnl(dev);
1449 	}
1450 	return idev;
1451 }
1452 
1453 /*
1454  *	Join a socket to a group
1455  */
1456 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1457 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1458 
1459 
1460 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1461 	__be32 *psfsrc)
1462 {
1463 	struct ip_sf_list *psf, *psf_prev;
1464 	int rv = 0;
1465 
1466 	psf_prev = NULL;
1467 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1468 		if (psf->sf_inaddr == *psfsrc)
1469 			break;
1470 		psf_prev = psf;
1471 	}
1472 	if (!psf || psf->sf_count[sfmode] == 0) {
1473 		/* source filter not found, or count wrong =>  bug */
1474 		return -ESRCH;
1475 	}
1476 	psf->sf_count[sfmode]--;
1477 	if (psf->sf_count[sfmode] == 0) {
1478 		ip_rt_multicast_event(pmc->interface);
1479 	}
1480 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1481 #ifdef CONFIG_IP_MULTICAST
1482 		struct in_device *in_dev = pmc->interface;
1483 #endif
1484 
1485 		/* no more filters for this source */
1486 		if (psf_prev)
1487 			psf_prev->sf_next = psf->sf_next;
1488 		else
1489 			pmc->sources = psf->sf_next;
1490 #ifdef CONFIG_IP_MULTICAST
1491 		if (psf->sf_oldin &&
1492 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1493 			psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1494 				IGMP_Unsolicited_Report_Count;
1495 			psf->sf_next = pmc->tomb;
1496 			pmc->tomb = psf;
1497 			rv = 1;
1498 		} else
1499 #endif
1500 			kfree(psf);
1501 	}
1502 	return rv;
1503 }
1504 
1505 #ifndef CONFIG_IP_MULTICAST
1506 #define igmp_ifc_event(x)	do { } while (0)
1507 #endif
1508 
1509 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1510 			 int sfcount, __be32 *psfsrc, int delta)
1511 {
1512 	struct ip_mc_list *pmc;
1513 	int	changerec = 0;
1514 	int	i, err;
1515 
1516 	if (!in_dev)
1517 		return -ENODEV;
1518 	read_lock(&in_dev->mc_list_lock);
1519 	for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
1520 		if (*pmca == pmc->multiaddr)
1521 			break;
1522 	}
1523 	if (!pmc) {
1524 		/* MCA not found?? bug */
1525 		read_unlock(&in_dev->mc_list_lock);
1526 		return -ESRCH;
1527 	}
1528 	spin_lock_bh(&pmc->lock);
1529 	read_unlock(&in_dev->mc_list_lock);
1530 #ifdef CONFIG_IP_MULTICAST
1531 	sf_markstate(pmc);
1532 #endif
1533 	if (!delta) {
1534 		err = -EINVAL;
1535 		if (!pmc->sfcount[sfmode])
1536 			goto out_unlock;
1537 		pmc->sfcount[sfmode]--;
1538 	}
1539 	err = 0;
1540 	for (i=0; i<sfcount; i++) {
1541 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1542 
1543 		changerec |= rv > 0;
1544 		if (!err && rv < 0)
1545 			err = rv;
1546 	}
1547 	if (pmc->sfmode == MCAST_EXCLUDE &&
1548 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1549 	    pmc->sfcount[MCAST_INCLUDE]) {
1550 #ifdef CONFIG_IP_MULTICAST
1551 		struct ip_sf_list *psf;
1552 #endif
1553 
1554 		/* filter mode change */
1555 		pmc->sfmode = MCAST_INCLUDE;
1556 #ifdef CONFIG_IP_MULTICAST
1557 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1558 			IGMP_Unsolicited_Report_Count;
1559 		in_dev->mr_ifc_count = pmc->crcount;
1560 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1561 			psf->sf_crcount = 0;
1562 		igmp_ifc_event(pmc->interface);
1563 	} else if (sf_setstate(pmc) || changerec) {
1564 		igmp_ifc_event(pmc->interface);
1565 #endif
1566 	}
1567 out_unlock:
1568 	spin_unlock_bh(&pmc->lock);
1569 	return err;
1570 }
1571 
1572 /*
1573  * Add multicast single-source filter to the interface list
1574  */
1575 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1576 	__be32 *psfsrc, int delta)
1577 {
1578 	struct ip_sf_list *psf, *psf_prev;
1579 
1580 	psf_prev = NULL;
1581 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1582 		if (psf->sf_inaddr == *psfsrc)
1583 			break;
1584 		psf_prev = psf;
1585 	}
1586 	if (!psf) {
1587 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1588 		if (!psf)
1589 			return -ENOBUFS;
1590 		psf->sf_inaddr = *psfsrc;
1591 		if (psf_prev) {
1592 			psf_prev->sf_next = psf;
1593 		} else
1594 			pmc->sources = psf;
1595 	}
1596 	psf->sf_count[sfmode]++;
1597 	if (psf->sf_count[sfmode] == 1) {
1598 		ip_rt_multicast_event(pmc->interface);
1599 	}
1600 	return 0;
1601 }
1602 
1603 #ifdef CONFIG_IP_MULTICAST
1604 static void sf_markstate(struct ip_mc_list *pmc)
1605 {
1606 	struct ip_sf_list *psf;
1607 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1608 
1609 	for (psf=pmc->sources; psf; psf=psf->sf_next)
1610 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1611 			psf->sf_oldin = mca_xcount ==
1612 				psf->sf_count[MCAST_EXCLUDE] &&
1613 				!psf->sf_count[MCAST_INCLUDE];
1614 		} else
1615 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1616 }
1617 
1618 static int sf_setstate(struct ip_mc_list *pmc)
1619 {
1620 	struct ip_sf_list *psf, *dpsf;
1621 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1622 	int qrv = pmc->interface->mr_qrv;
1623 	int new_in, rv;
1624 
1625 	rv = 0;
1626 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1627 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1628 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1629 				!psf->sf_count[MCAST_INCLUDE];
1630 		} else
1631 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1632 		if (new_in) {
1633 			if (!psf->sf_oldin) {
1634 				struct ip_sf_list *prev = NULL;
1635 
1636 				for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1637 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1638 						break;
1639 					prev = dpsf;
1640 				}
1641 				if (dpsf) {
1642 					if (prev)
1643 						prev->sf_next = dpsf->sf_next;
1644 					else
1645 						pmc->tomb = dpsf->sf_next;
1646 					kfree(dpsf);
1647 				}
1648 				psf->sf_crcount = qrv;
1649 				rv++;
1650 			}
1651 		} else if (psf->sf_oldin) {
1652 
1653 			psf->sf_crcount = 0;
1654 			/*
1655 			 * add or update "delete" records if an active filter
1656 			 * is now inactive
1657 			 */
1658 			for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1659 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1660 					break;
1661 			if (!dpsf) {
1662 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1663 				if (!dpsf)
1664 					continue;
1665 				*dpsf = *psf;
1666 				/* pmc->lock held by callers */
1667 				dpsf->sf_next = pmc->tomb;
1668 				pmc->tomb = dpsf;
1669 			}
1670 			dpsf->sf_crcount = qrv;
1671 			rv++;
1672 		}
1673 	}
1674 	return rv;
1675 }
1676 #endif
1677 
1678 /*
1679  * Add multicast source filter list to the interface list
1680  */
1681 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1682 			 int sfcount, __be32 *psfsrc, int delta)
1683 {
1684 	struct ip_mc_list *pmc;
1685 	int	isexclude;
1686 	int	i, err;
1687 
1688 	if (!in_dev)
1689 		return -ENODEV;
1690 	read_lock(&in_dev->mc_list_lock);
1691 	for (pmc=in_dev->mc_list; pmc; pmc=pmc->next) {
1692 		if (*pmca == pmc->multiaddr)
1693 			break;
1694 	}
1695 	if (!pmc) {
1696 		/* MCA not found?? bug */
1697 		read_unlock(&in_dev->mc_list_lock);
1698 		return -ESRCH;
1699 	}
1700 	spin_lock_bh(&pmc->lock);
1701 	read_unlock(&in_dev->mc_list_lock);
1702 
1703 #ifdef CONFIG_IP_MULTICAST
1704 	sf_markstate(pmc);
1705 #endif
1706 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
1707 	if (!delta)
1708 		pmc->sfcount[sfmode]++;
1709 	err = 0;
1710 	for (i=0; i<sfcount; i++) {
1711 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1712 		if (err)
1713 			break;
1714 	}
1715 	if (err) {
1716 		int j;
1717 
1718 		pmc->sfcount[sfmode]--;
1719 		for (j=0; j<i; j++)
1720 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1721 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1722 #ifdef CONFIG_IP_MULTICAST
1723 		struct ip_sf_list *psf;
1724 		in_dev = pmc->interface;
1725 #endif
1726 
1727 		/* filter mode change */
1728 		if (pmc->sfcount[MCAST_EXCLUDE])
1729 			pmc->sfmode = MCAST_EXCLUDE;
1730 		else if (pmc->sfcount[MCAST_INCLUDE])
1731 			pmc->sfmode = MCAST_INCLUDE;
1732 #ifdef CONFIG_IP_MULTICAST
1733 		/* else no filters; keep old mode for reports */
1734 
1735 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1736 			IGMP_Unsolicited_Report_Count;
1737 		in_dev->mr_ifc_count = pmc->crcount;
1738 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1739 			psf->sf_crcount = 0;
1740 		igmp_ifc_event(in_dev);
1741 	} else if (sf_setstate(pmc)) {
1742 		igmp_ifc_event(in_dev);
1743 #endif
1744 	}
1745 	spin_unlock_bh(&pmc->lock);
1746 	return err;
1747 }
1748 
1749 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1750 {
1751 	struct ip_sf_list *psf, *nextpsf;
1752 
1753 	for (psf=pmc->tomb; psf; psf=nextpsf) {
1754 		nextpsf = psf->sf_next;
1755 		kfree(psf);
1756 	}
1757 	pmc->tomb = NULL;
1758 	for (psf=pmc->sources; psf; psf=nextpsf) {
1759 		nextpsf = psf->sf_next;
1760 		kfree(psf);
1761 	}
1762 	pmc->sources = NULL;
1763 	pmc->sfmode = MCAST_EXCLUDE;
1764 	pmc->sfcount[MCAST_INCLUDE] = 0;
1765 	pmc->sfcount[MCAST_EXCLUDE] = 1;
1766 }
1767 
1768 
1769 /*
1770  * Join a multicast group
1771  */
1772 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1773 {
1774 	int err;
1775 	__be32 addr = imr->imr_multiaddr.s_addr;
1776 	struct ip_mc_socklist *iml = NULL, *i;
1777 	struct in_device *in_dev;
1778 	struct inet_sock *inet = inet_sk(sk);
1779 	struct net *net = sock_net(sk);
1780 	int ifindex;
1781 	int count = 0;
1782 
1783 	if (!ipv4_is_multicast(addr))
1784 		return -EINVAL;
1785 
1786 	rtnl_lock();
1787 
1788 	in_dev = ip_mc_find_dev(net, imr);
1789 
1790 	if (!in_dev) {
1791 		iml = NULL;
1792 		err = -ENODEV;
1793 		goto done;
1794 	}
1795 
1796 	err = -EADDRINUSE;
1797 	ifindex = imr->imr_ifindex;
1798 	for (i = inet->mc_list; i; i = i->next) {
1799 		if (i->multi.imr_multiaddr.s_addr == addr &&
1800 		    i->multi.imr_ifindex == ifindex)
1801 			goto done;
1802 		count++;
1803 	}
1804 	err = -ENOBUFS;
1805 	if (count >= sysctl_igmp_max_memberships)
1806 		goto done;
1807 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1808 	if (iml == NULL)
1809 		goto done;
1810 
1811 	memcpy(&iml->multi, imr, sizeof(*imr));
1812 	iml->next = inet->mc_list;
1813 	iml->sflist = NULL;
1814 	iml->sfmode = MCAST_EXCLUDE;
1815 	rcu_assign_pointer(inet->mc_list, iml);
1816 	ip_mc_inc_group(in_dev, addr);
1817 	err = 0;
1818 done:
1819 	rtnl_unlock();
1820 	return err;
1821 }
1822 EXPORT_SYMBOL(ip_mc_join_group);
1823 
1824 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1825 {
1826 	struct ip_sf_socklist *psf;
1827 
1828 	psf = container_of(rp, struct ip_sf_socklist, rcu);
1829 	/* sk_omem_alloc should have been decreased by the caller*/
1830 	kfree(psf);
1831 }
1832 
1833 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1834 			   struct in_device *in_dev)
1835 {
1836 	struct ip_sf_socklist *psf = iml->sflist;
1837 	int err;
1838 
1839 	if (psf == NULL) {
1840 		/* any-source empty exclude case */
1841 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1842 			iml->sfmode, 0, NULL, 0);
1843 	}
1844 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1845 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1846 	rcu_assign_pointer(iml->sflist, NULL);
1847 	/* decrease mem now to avoid the memleak warning */
1848 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1849 	call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1850 	return err;
1851 }
1852 
1853 
1854 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1855 {
1856 	struct ip_mc_socklist *iml;
1857 
1858 	iml = container_of(rp, struct ip_mc_socklist, rcu);
1859 	/* sk_omem_alloc should have been decreased by the caller*/
1860 	kfree(iml);
1861 }
1862 
1863 
1864 /*
1865  *	Ask a socket to leave a group.
1866  */
1867 
1868 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1869 {
1870 	struct inet_sock *inet = inet_sk(sk);
1871 	struct ip_mc_socklist *iml, **imlp;
1872 	struct in_device *in_dev;
1873 	struct net *net = sock_net(sk);
1874 	__be32 group = imr->imr_multiaddr.s_addr;
1875 	u32 ifindex;
1876 	int ret = -EADDRNOTAVAIL;
1877 
1878 	rtnl_lock();
1879 	in_dev = ip_mc_find_dev(net, imr);
1880 	ifindex = imr->imr_ifindex;
1881 	for (imlp = &inet->mc_list; (iml = *imlp) != NULL; imlp = &iml->next) {
1882 		if (iml->multi.imr_multiaddr.s_addr != group)
1883 			continue;
1884 		if (ifindex) {
1885 			if (iml->multi.imr_ifindex != ifindex)
1886 				continue;
1887 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1888 				iml->multi.imr_address.s_addr)
1889 			continue;
1890 
1891 		(void) ip_mc_leave_src(sk, iml, in_dev);
1892 
1893 		rcu_assign_pointer(*imlp, iml->next);
1894 
1895 		if (in_dev)
1896 			ip_mc_dec_group(in_dev, group);
1897 		rtnl_unlock();
1898 		/* decrease mem now to avoid the memleak warning */
1899 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1900 		call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1901 		return 0;
1902 	}
1903 	if (!in_dev)
1904 		ret = -ENODEV;
1905 	rtnl_unlock();
1906 	return ret;
1907 }
1908 
1909 int ip_mc_source(int add, int omode, struct sock *sk, struct
1910 	ip_mreq_source *mreqs, int ifindex)
1911 {
1912 	int err;
1913 	struct ip_mreqn imr;
1914 	__be32 addr = mreqs->imr_multiaddr;
1915 	struct ip_mc_socklist *pmc;
1916 	struct in_device *in_dev = NULL;
1917 	struct inet_sock *inet = inet_sk(sk);
1918 	struct ip_sf_socklist *psl;
1919 	struct net *net = sock_net(sk);
1920 	int leavegroup = 0;
1921 	int i, j, rv;
1922 
1923 	if (!ipv4_is_multicast(addr))
1924 		return -EINVAL;
1925 
1926 	rtnl_lock();
1927 
1928 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1929 	imr.imr_address.s_addr = mreqs->imr_interface;
1930 	imr.imr_ifindex = ifindex;
1931 	in_dev = ip_mc_find_dev(net, &imr);
1932 
1933 	if (!in_dev) {
1934 		err = -ENODEV;
1935 		goto done;
1936 	}
1937 	err = -EADDRNOTAVAIL;
1938 
1939 	for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
1940 		if ((pmc->multi.imr_multiaddr.s_addr ==
1941 		     imr.imr_multiaddr.s_addr) &&
1942 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
1943 			break;
1944 	}
1945 	if (!pmc) {		/* must have a prior join */
1946 		err = -EINVAL;
1947 		goto done;
1948 	}
1949 	/* if a source filter was set, must be the same mode as before */
1950 	if (pmc->sflist) {
1951 		if (pmc->sfmode != omode) {
1952 			err = -EINVAL;
1953 			goto done;
1954 		}
1955 	} else if (pmc->sfmode != omode) {
1956 		/* allow mode switches for empty-set filters */
1957 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1958 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1959 			NULL, 0);
1960 		pmc->sfmode = omode;
1961 	}
1962 
1963 	psl = pmc->sflist;
1964 	if (!add) {
1965 		if (!psl)
1966 			goto done;	/* err = -EADDRNOTAVAIL */
1967 		rv = !0;
1968 		for (i=0; i<psl->sl_count; i++) {
1969 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1970 				sizeof(__be32));
1971 			if (rv == 0)
1972 				break;
1973 		}
1974 		if (rv)		/* source not found */
1975 			goto done;	/* err = -EADDRNOTAVAIL */
1976 
1977 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
1978 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1979 			leavegroup = 1;
1980 			goto done;
1981 		}
1982 
1983 		/* update the interface filter */
1984 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1985 			&mreqs->imr_sourceaddr, 1);
1986 
1987 		for (j=i+1; j<psl->sl_count; j++)
1988 			psl->sl_addr[j-1] = psl->sl_addr[j];
1989 		psl->sl_count--;
1990 		err = 0;
1991 		goto done;
1992 	}
1993 	/* else, add a new source to the filter */
1994 
1995 	if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1996 		err = -ENOBUFS;
1997 		goto done;
1998 	}
1999 	if (!psl || psl->sl_count == psl->sl_max) {
2000 		struct ip_sf_socklist *newpsl;
2001 		int count = IP_SFBLOCK;
2002 
2003 		if (psl)
2004 			count += psl->sl_max;
2005 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2006 		if (!newpsl) {
2007 			err = -ENOBUFS;
2008 			goto done;
2009 		}
2010 		newpsl->sl_max = count;
2011 		newpsl->sl_count = count - IP_SFBLOCK;
2012 		if (psl) {
2013 			for (i=0; i<psl->sl_count; i++)
2014 				newpsl->sl_addr[i] = psl->sl_addr[i];
2015 			/* decrease mem now to avoid the memleak warning */
2016 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2017 			call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2018 		}
2019 		rcu_assign_pointer(pmc->sflist, newpsl);
2020 		psl = newpsl;
2021 	}
2022 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2023 	for (i=0; i<psl->sl_count; i++) {
2024 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2025 			sizeof(__be32));
2026 		if (rv == 0)
2027 			break;
2028 	}
2029 	if (rv == 0)		/* address already there is an error */
2030 		goto done;
2031 	for (j=psl->sl_count-1; j>=i; j--)
2032 		psl->sl_addr[j+1] = psl->sl_addr[j];
2033 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2034 	psl->sl_count++;
2035 	err = 0;
2036 	/* update the interface list */
2037 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2038 		&mreqs->imr_sourceaddr, 1);
2039 done:
2040 	rtnl_unlock();
2041 	if (leavegroup)
2042 		return ip_mc_leave_group(sk, &imr);
2043 	return err;
2044 }
2045 
2046 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2047 {
2048 	int err = 0;
2049 	struct ip_mreqn	imr;
2050 	__be32 addr = msf->imsf_multiaddr;
2051 	struct ip_mc_socklist *pmc;
2052 	struct in_device *in_dev;
2053 	struct inet_sock *inet = inet_sk(sk);
2054 	struct ip_sf_socklist *newpsl, *psl;
2055 	struct net *net = sock_net(sk);
2056 	int leavegroup = 0;
2057 
2058 	if (!ipv4_is_multicast(addr))
2059 		return -EINVAL;
2060 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2061 	    msf->imsf_fmode != MCAST_EXCLUDE)
2062 		return -EINVAL;
2063 
2064 	rtnl_lock();
2065 
2066 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2067 	imr.imr_address.s_addr = msf->imsf_interface;
2068 	imr.imr_ifindex = ifindex;
2069 	in_dev = ip_mc_find_dev(net, &imr);
2070 
2071 	if (!in_dev) {
2072 		err = -ENODEV;
2073 		goto done;
2074 	}
2075 
2076 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2077 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2078 		leavegroup = 1;
2079 		goto done;
2080 	}
2081 
2082 	for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
2083 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2084 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2085 			break;
2086 	}
2087 	if (!pmc) {		/* must have a prior join */
2088 		err = -EINVAL;
2089 		goto done;
2090 	}
2091 	if (msf->imsf_numsrc) {
2092 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2093 							   GFP_KERNEL);
2094 		if (!newpsl) {
2095 			err = -ENOBUFS;
2096 			goto done;
2097 		}
2098 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2099 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2100 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2101 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2102 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2103 		if (err) {
2104 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2105 			goto done;
2106 		}
2107 	} else {
2108 		newpsl = NULL;
2109 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2110 				     msf->imsf_fmode, 0, NULL, 0);
2111 	}
2112 	psl = pmc->sflist;
2113 	if (psl) {
2114 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2115 			psl->sl_count, psl->sl_addr, 0);
2116 		/* decrease mem now to avoid the memleak warning */
2117 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2118 		call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2119 	} else
2120 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2121 			0, NULL, 0);
2122 	rcu_assign_pointer(pmc->sflist, newpsl);
2123 	pmc->sfmode = msf->imsf_fmode;
2124 	err = 0;
2125 done:
2126 	rtnl_unlock();
2127 	if (leavegroup)
2128 		err = ip_mc_leave_group(sk, &imr);
2129 	return err;
2130 }
2131 
2132 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2133 	struct ip_msfilter __user *optval, int __user *optlen)
2134 {
2135 	int err, len, count, copycount;
2136 	struct ip_mreqn	imr;
2137 	__be32 addr = msf->imsf_multiaddr;
2138 	struct ip_mc_socklist *pmc;
2139 	struct in_device *in_dev;
2140 	struct inet_sock *inet = inet_sk(sk);
2141 	struct ip_sf_socklist *psl;
2142 	struct net *net = sock_net(sk);
2143 
2144 	if (!ipv4_is_multicast(addr))
2145 		return -EINVAL;
2146 
2147 	rtnl_lock();
2148 
2149 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2150 	imr.imr_address.s_addr = msf->imsf_interface;
2151 	imr.imr_ifindex = 0;
2152 	in_dev = ip_mc_find_dev(net, &imr);
2153 
2154 	if (!in_dev) {
2155 		err = -ENODEV;
2156 		goto done;
2157 	}
2158 	err = -EADDRNOTAVAIL;
2159 
2160 	for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
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 		goto done;
2167 	msf->imsf_fmode = pmc->sfmode;
2168 	psl = pmc->sflist;
2169 	rtnl_unlock();
2170 	if (!psl) {
2171 		len = 0;
2172 		count = 0;
2173 	} else {
2174 		count = psl->sl_count;
2175 	}
2176 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2177 	len = copycount * sizeof(psl->sl_addr[0]);
2178 	msf->imsf_numsrc = count;
2179 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2180 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2181 		return -EFAULT;
2182 	}
2183 	if (len &&
2184 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2185 		return -EFAULT;
2186 	return 0;
2187 done:
2188 	rtnl_unlock();
2189 	return err;
2190 }
2191 
2192 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2193 	struct group_filter __user *optval, int __user *optlen)
2194 {
2195 	int err, i, count, copycount;
2196 	struct sockaddr_in *psin;
2197 	__be32 addr;
2198 	struct ip_mc_socklist *pmc;
2199 	struct inet_sock *inet = inet_sk(sk);
2200 	struct ip_sf_socklist *psl;
2201 
2202 	psin = (struct sockaddr_in *)&gsf->gf_group;
2203 	if (psin->sin_family != AF_INET)
2204 		return -EINVAL;
2205 	addr = psin->sin_addr.s_addr;
2206 	if (!ipv4_is_multicast(addr))
2207 		return -EINVAL;
2208 
2209 	rtnl_lock();
2210 
2211 	err = -EADDRNOTAVAIL;
2212 
2213 	for (pmc=inet->mc_list; pmc; pmc=pmc->next) {
2214 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2215 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2216 			break;
2217 	}
2218 	if (!pmc)		/* must have a prior join */
2219 		goto done;
2220 	gsf->gf_fmode = pmc->sfmode;
2221 	psl = pmc->sflist;
2222 	rtnl_unlock();
2223 	count = psl ? psl->sl_count : 0;
2224 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2225 	gsf->gf_numsrc = count;
2226 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2227 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2228 		return -EFAULT;
2229 	}
2230 	for (i=0; i<copycount; i++) {
2231 		struct sockaddr_storage ss;
2232 
2233 		psin = (struct sockaddr_in *)&ss;
2234 		memset(&ss, 0, sizeof(ss));
2235 		psin->sin_family = AF_INET;
2236 		psin->sin_addr.s_addr = psl->sl_addr[i];
2237 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2238 			return -EFAULT;
2239 	}
2240 	return 0;
2241 done:
2242 	rtnl_unlock();
2243 	return err;
2244 }
2245 
2246 /*
2247  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2248  */
2249 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2250 {
2251 	struct inet_sock *inet = inet_sk(sk);
2252 	struct ip_mc_socklist *pmc;
2253 	struct ip_sf_socklist *psl;
2254 	int i;
2255 	int ret;
2256 
2257 	ret = 1;
2258 	if (!ipv4_is_multicast(loc_addr))
2259 		goto out;
2260 
2261 	rcu_read_lock();
2262 	for (pmc=rcu_dereference(inet->mc_list); pmc; pmc=rcu_dereference(pmc->next)) {
2263 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2264 		    pmc->multi.imr_ifindex == dif)
2265 			break;
2266 	}
2267 	ret = inet->mc_all;
2268 	if (!pmc)
2269 		goto unlock;
2270 	psl = pmc->sflist;
2271 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2272 	if (!psl)
2273 		goto unlock;
2274 
2275 	for (i=0; i<psl->sl_count; i++) {
2276 		if (psl->sl_addr[i] == rmt_addr)
2277 			break;
2278 	}
2279 	ret = 0;
2280 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2281 		goto unlock;
2282 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2283 		goto unlock;
2284 	ret = 1;
2285 unlock:
2286 	rcu_read_unlock();
2287 out:
2288 	return ret;
2289 }
2290 
2291 /*
2292  *	A socket is closing.
2293  */
2294 
2295 void ip_mc_drop_socket(struct sock *sk)
2296 {
2297 	struct inet_sock *inet = inet_sk(sk);
2298 	struct ip_mc_socklist *iml;
2299 	struct net *net = sock_net(sk);
2300 
2301 	if (inet->mc_list == NULL)
2302 		return;
2303 
2304 	rtnl_lock();
2305 	while ((iml = inet->mc_list) != NULL) {
2306 		struct in_device *in_dev;
2307 		rcu_assign_pointer(inet->mc_list, iml->next);
2308 
2309 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2310 		(void) ip_mc_leave_src(sk, iml, in_dev);
2311 		if (in_dev != NULL) {
2312 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2313 			in_dev_put(in_dev);
2314 		}
2315 		/* decrease mem now to avoid the memleak warning */
2316 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2317 		call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2318 	}
2319 	rtnl_unlock();
2320 }
2321 
2322 int ip_check_mc(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2323 {
2324 	struct ip_mc_list *im;
2325 	struct ip_sf_list *psf;
2326 	int rv = 0;
2327 
2328 	read_lock(&in_dev->mc_list_lock);
2329 	for (im=in_dev->mc_list; im; im=im->next) {
2330 		if (im->multiaddr == mc_addr)
2331 			break;
2332 	}
2333 	if (im && proto == IPPROTO_IGMP) {
2334 		rv = 1;
2335 	} else if (im) {
2336 		if (src_addr) {
2337 			for (psf=im->sources; psf; psf=psf->sf_next) {
2338 				if (psf->sf_inaddr == src_addr)
2339 					break;
2340 			}
2341 			if (psf)
2342 				rv = psf->sf_count[MCAST_INCLUDE] ||
2343 					psf->sf_count[MCAST_EXCLUDE] !=
2344 					im->sfcount[MCAST_EXCLUDE];
2345 			else
2346 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2347 		} else
2348 			rv = 1; /* unspecified source; tentatively allow */
2349 	}
2350 	read_unlock(&in_dev->mc_list_lock);
2351 	return rv;
2352 }
2353 
2354 #if defined(CONFIG_PROC_FS)
2355 struct igmp_mc_iter_state {
2356 	struct seq_net_private p;
2357 	struct net_device *dev;
2358 	struct in_device *in_dev;
2359 };
2360 
2361 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2362 
2363 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2364 {
2365 	struct net *net = seq_file_net(seq);
2366 	struct ip_mc_list *im = NULL;
2367 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2368 
2369 	state->in_dev = NULL;
2370 	for_each_netdev_rcu(net, state->dev) {
2371 		struct in_device *in_dev;
2372 
2373 		in_dev = __in_dev_get_rcu(state->dev);
2374 		if (!in_dev)
2375 			continue;
2376 		read_lock(&in_dev->mc_list_lock);
2377 		im = in_dev->mc_list;
2378 		if (im) {
2379 			state->in_dev = in_dev;
2380 			break;
2381 		}
2382 		read_unlock(&in_dev->mc_list_lock);
2383 	}
2384 	return im;
2385 }
2386 
2387 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2388 {
2389 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2390 	im = im->next;
2391 	while (!im) {
2392 		if (likely(state->in_dev != NULL))
2393 			read_unlock(&state->in_dev->mc_list_lock);
2394 
2395 		state->dev = next_net_device_rcu(state->dev);
2396 		if (!state->dev) {
2397 			state->in_dev = NULL;
2398 			break;
2399 		}
2400 		state->in_dev = __in_dev_get_rcu(state->dev);
2401 		if (!state->in_dev)
2402 			continue;
2403 		read_lock(&state->in_dev->mc_list_lock);
2404 		im = state->in_dev->mc_list;
2405 	}
2406 	return im;
2407 }
2408 
2409 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2410 {
2411 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2412 	if (im)
2413 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2414 			--pos;
2415 	return pos ? NULL : im;
2416 }
2417 
2418 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2419 	__acquires(rcu)
2420 {
2421 	rcu_read_lock();
2422 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2423 }
2424 
2425 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2426 {
2427 	struct ip_mc_list *im;
2428 	if (v == SEQ_START_TOKEN)
2429 		im = igmp_mc_get_first(seq);
2430 	else
2431 		im = igmp_mc_get_next(seq, v);
2432 	++*pos;
2433 	return im;
2434 }
2435 
2436 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2437 	__releases(rcu)
2438 {
2439 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2440 	if (likely(state->in_dev != NULL)) {
2441 		read_unlock(&state->in_dev->mc_list_lock);
2442 		state->in_dev = NULL;
2443 	}
2444 	state->dev = NULL;
2445 	rcu_read_unlock();
2446 }
2447 
2448 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2449 {
2450 	if (v == SEQ_START_TOKEN)
2451 		seq_puts(seq,
2452 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2453 	else {
2454 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2455 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2456 		char   *querier;
2457 #ifdef CONFIG_IP_MULTICAST
2458 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2459 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2460 			  "V3";
2461 #else
2462 		querier = "NONE";
2463 #endif
2464 
2465 		if (state->in_dev->mc_list == im) {
2466 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2467 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2468 		}
2469 
2470 		seq_printf(seq,
2471 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2472 			   im->multiaddr, im->users,
2473 			   im->tm_running, im->tm_running ?
2474 			   jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2475 			   im->reporter);
2476 	}
2477 	return 0;
2478 }
2479 
2480 static const struct seq_operations igmp_mc_seq_ops = {
2481 	.start	=	igmp_mc_seq_start,
2482 	.next	=	igmp_mc_seq_next,
2483 	.stop	=	igmp_mc_seq_stop,
2484 	.show	=	igmp_mc_seq_show,
2485 };
2486 
2487 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2488 {
2489 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2490 			sizeof(struct igmp_mc_iter_state));
2491 }
2492 
2493 static const struct file_operations igmp_mc_seq_fops = {
2494 	.owner		=	THIS_MODULE,
2495 	.open		=	igmp_mc_seq_open,
2496 	.read		=	seq_read,
2497 	.llseek		=	seq_lseek,
2498 	.release	=	seq_release_net,
2499 };
2500 
2501 struct igmp_mcf_iter_state {
2502 	struct seq_net_private p;
2503 	struct net_device *dev;
2504 	struct in_device *idev;
2505 	struct ip_mc_list *im;
2506 };
2507 
2508 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2509 
2510 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2511 {
2512 	struct net *net = seq_file_net(seq);
2513 	struct ip_sf_list *psf = NULL;
2514 	struct ip_mc_list *im = NULL;
2515 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2516 
2517 	state->idev = NULL;
2518 	state->im = NULL;
2519 	for_each_netdev_rcu(net, state->dev) {
2520 		struct in_device *idev;
2521 		idev = __in_dev_get_rcu(state->dev);
2522 		if (unlikely(idev == NULL))
2523 			continue;
2524 		read_lock(&idev->mc_list_lock);
2525 		im = idev->mc_list;
2526 		if (likely(im != NULL)) {
2527 			spin_lock_bh(&im->lock);
2528 			psf = im->sources;
2529 			if (likely(psf != NULL)) {
2530 				state->im = im;
2531 				state->idev = idev;
2532 				break;
2533 			}
2534 			spin_unlock_bh(&im->lock);
2535 		}
2536 		read_unlock(&idev->mc_list_lock);
2537 	}
2538 	return psf;
2539 }
2540 
2541 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2542 {
2543 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2544 
2545 	psf = psf->sf_next;
2546 	while (!psf) {
2547 		spin_unlock_bh(&state->im->lock);
2548 		state->im = state->im->next;
2549 		while (!state->im) {
2550 			if (likely(state->idev != NULL))
2551 				read_unlock(&state->idev->mc_list_lock);
2552 
2553 			state->dev = next_net_device_rcu(state->dev);
2554 			if (!state->dev) {
2555 				state->idev = NULL;
2556 				goto out;
2557 			}
2558 			state->idev = __in_dev_get_rcu(state->dev);
2559 			if (!state->idev)
2560 				continue;
2561 			read_lock(&state->idev->mc_list_lock);
2562 			state->im = state->idev->mc_list;
2563 		}
2564 		if (!state->im)
2565 			break;
2566 		spin_lock_bh(&state->im->lock);
2567 		psf = state->im->sources;
2568 	}
2569 out:
2570 	return psf;
2571 }
2572 
2573 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2574 {
2575 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2576 	if (psf)
2577 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2578 			--pos;
2579 	return pos ? NULL : psf;
2580 }
2581 
2582 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2583 	__acquires(rcu)
2584 {
2585 	rcu_read_lock();
2586 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2587 }
2588 
2589 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2590 {
2591 	struct ip_sf_list *psf;
2592 	if (v == SEQ_START_TOKEN)
2593 		psf = igmp_mcf_get_first(seq);
2594 	else
2595 		psf = igmp_mcf_get_next(seq, v);
2596 	++*pos;
2597 	return psf;
2598 }
2599 
2600 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2601 	__releases(rcu)
2602 {
2603 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2604 	if (likely(state->im != NULL)) {
2605 		spin_unlock_bh(&state->im->lock);
2606 		state->im = NULL;
2607 	}
2608 	if (likely(state->idev != NULL)) {
2609 		read_unlock(&state->idev->mc_list_lock);
2610 		state->idev = NULL;
2611 	}
2612 	state->dev = NULL;
2613 	rcu_read_unlock();
2614 }
2615 
2616 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2617 {
2618 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2619 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2620 
2621 	if (v == SEQ_START_TOKEN) {
2622 		seq_printf(seq,
2623 			   "%3s %6s "
2624 			   "%10s %10s %6s %6s\n", "Idx",
2625 			   "Device", "MCA",
2626 			   "SRC", "INC", "EXC");
2627 	} else {
2628 		seq_printf(seq,
2629 			   "%3d %6.6s 0x%08x "
2630 			   "0x%08x %6lu %6lu\n",
2631 			   state->dev->ifindex, state->dev->name,
2632 			   ntohl(state->im->multiaddr),
2633 			   ntohl(psf->sf_inaddr),
2634 			   psf->sf_count[MCAST_INCLUDE],
2635 			   psf->sf_count[MCAST_EXCLUDE]);
2636 	}
2637 	return 0;
2638 }
2639 
2640 static const struct seq_operations igmp_mcf_seq_ops = {
2641 	.start	=	igmp_mcf_seq_start,
2642 	.next	=	igmp_mcf_seq_next,
2643 	.stop	=	igmp_mcf_seq_stop,
2644 	.show	=	igmp_mcf_seq_show,
2645 };
2646 
2647 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2648 {
2649 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2650 			sizeof(struct igmp_mcf_iter_state));
2651 }
2652 
2653 static const struct file_operations igmp_mcf_seq_fops = {
2654 	.owner		=	THIS_MODULE,
2655 	.open		=	igmp_mcf_seq_open,
2656 	.read		=	seq_read,
2657 	.llseek		=	seq_lseek,
2658 	.release	=	seq_release_net,
2659 };
2660 
2661 static int __net_init igmp_net_init(struct net *net)
2662 {
2663 	struct proc_dir_entry *pde;
2664 
2665 	pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2666 	if (!pde)
2667 		goto out_igmp;
2668 	pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2669 	if (!pde)
2670 		goto out_mcfilter;
2671 	return 0;
2672 
2673 out_mcfilter:
2674 	proc_net_remove(net, "igmp");
2675 out_igmp:
2676 	return -ENOMEM;
2677 }
2678 
2679 static void __net_exit igmp_net_exit(struct net *net)
2680 {
2681 	proc_net_remove(net, "mcfilter");
2682 	proc_net_remove(net, "igmp");
2683 }
2684 
2685 static struct pernet_operations igmp_net_ops = {
2686 	.init = igmp_net_init,
2687 	.exit = igmp_net_exit,
2688 };
2689 
2690 int __init igmp_mc_proc_init(void)
2691 {
2692 	return register_pernet_subsys(&igmp_net_ops);
2693 }
2694 #endif
2695