xref: /openbmc/linux/net/ipv4/igmp.c (revision 95e9fd10)
1 /*
2  *	Linux NET3:	Internet Group Management Protocol  [IGMP]
3  *
4  *	This code implements the IGMP protocol as defined in RFC1112. There has
5  *	been a further revision of this protocol since which is now supported.
6  *
7  *	If you have trouble with this module be careful what gcc you have used,
8  *	the older version didn't come out right using gcc 2.5.8, the newer one
9  *	seems to fall out with gcc 2.6.2.
10  *
11  *	Authors:
12  *		Alan Cox <alan@lxorguk.ukuu.org.uk>
13  *
14  *	This program is free software; you can redistribute it and/or
15  *	modify it under the terms of the GNU General Public License
16  *	as published by the Free Software Foundation; either version
17  *	2 of the License, or (at your option) any later version.
18  *
19  *	Fixes:
20  *
21  *		Alan Cox	:	Added lots of __inline__ to optimise
22  *					the memory usage of all the tiny little
23  *					functions.
24  *		Alan Cox	:	Dumped the header building experiment.
25  *		Alan Cox	:	Minor tweaks ready for multicast routing
26  *					and extended IGMP protocol.
27  *		Alan Cox	:	Removed a load of inline directives. Gcc 2.5.8
28  *					writes utterly bogus code otherwise (sigh)
29  *					fixed IGMP loopback to behave in the manner
30  *					desired by mrouted, fixed the fact it has been
31  *					broken since 1.3.6 and cleaned up a few minor
32  *					points.
33  *
34  *		Chih-Jen Chang	:	Tried to revise IGMP to Version 2
35  *		Tsu-Sheng Tsao		E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
36  *					The enhancements are mainly based on Steve Deering's
37  * 					ipmulti-3.5 source code.
38  *		Chih-Jen Chang	:	Added the igmp_get_mrouter_info and
39  *		Tsu-Sheng Tsao		igmp_set_mrouter_info to keep track of
40  *					the mrouted version on that device.
41  *		Chih-Jen Chang	:	Added the max_resp_time parameter to
42  *		Tsu-Sheng Tsao		igmp_heard_query(). Using this parameter
43  *					to identify the multicast router version
44  *					and do what the IGMP version 2 specified.
45  *		Chih-Jen Chang	:	Added a timer to revert to IGMP V2 router
46  *		Tsu-Sheng Tsao		if the specified time expired.
47  *		Alan Cox	:	Stop IGMP from 0.0.0.0 being accepted.
48  *		Alan Cox	:	Use GFP_ATOMIC in the right places.
49  *		Christian Daudt :	igmp timer wasn't set for local group
50  *					memberships but was being deleted,
51  *					which caused a "del_timer() called
52  *					from %p with timer not initialized\n"
53  *					message (960131).
54  *		Christian Daudt :	removed del_timer from
55  *					igmp_timer_expire function (960205).
56  *             Christian Daudt :       igmp_heard_report now only calls
57  *                                     igmp_timer_expire if tm->running is
58  *                                     true (960216).
59  *		Malcolm Beattie :	ttl comparison wrong in igmp_rcv made
60  *					igmp_heard_query never trigger. Expiry
61  *					miscalculation fixed in igmp_heard_query
62  *					and random() made to return unsigned to
63  *					prevent negative expiry times.
64  *		Alexey Kuznetsov:	Wrong group leaving behaviour, backport
65  *					fix from pending 2.1.x patches.
66  *		Alan Cox:		Forget to enable FDDI support earlier.
67  *		Alexey Kuznetsov:	Fixed leaving groups on device down.
68  *		Alexey Kuznetsov:	Accordance to igmp-v2-06 draft.
69  *		David L Stevens:	IGMPv3 support, with help from
70  *					Vinay Kulkarni
71  */
72 
73 #include <linux/module.h>
74 #include <linux/slab.h>
75 #include <asm/uaccess.h>
76 #include <linux/types.h>
77 #include <linux/kernel.h>
78 #include <linux/jiffies.h>
79 #include <linux/string.h>
80 #include <linux/socket.h>
81 #include <linux/sockios.h>
82 #include <linux/in.h>
83 #include <linux/inet.h>
84 #include <linux/netdevice.h>
85 #include <linux/skbuff.h>
86 #include <linux/inetdevice.h>
87 #include <linux/igmp.h>
88 #include <linux/if_arp.h>
89 #include <linux/rtnetlink.h>
90 #include <linux/times.h>
91 
92 #include <net/net_namespace.h>
93 #include <net/arp.h>
94 #include <net/ip.h>
95 #include <net/protocol.h>
96 #include <net/route.h>
97 #include <net/sock.h>
98 #include <net/checksum.h>
99 #include <linux/netfilter_ipv4.h>
100 #ifdef CONFIG_IP_MROUTE
101 #include <linux/mroute.h>
102 #endif
103 #ifdef CONFIG_PROC_FS
104 #include <linux/proc_fs.h>
105 #include <linux/seq_file.h>
106 #endif
107 
108 #define IP_MAX_MEMBERSHIPS	20
109 #define IP_MAX_MSF		10
110 
111 #ifdef CONFIG_IP_MULTICAST
112 /* Parameter names and values are taken from igmp-v2-06 draft */
113 
114 #define IGMP_V1_Router_Present_Timeout		(400*HZ)
115 #define IGMP_V2_Router_Present_Timeout		(400*HZ)
116 #define IGMP_Unsolicited_Report_Interval	(10*HZ)
117 #define IGMP_Query_Response_Interval		(10*HZ)
118 #define IGMP_Unsolicited_Report_Count		2
119 
120 
121 #define IGMP_Initial_Report_Delay		(1)
122 
123 /* IGMP_Initial_Report_Delay is not from IGMP specs!
124  * IGMP specs require to report membership immediately after
125  * joining a group, but we delay the first report by a
126  * small interval. It seems more natural and still does not
127  * contradict to specs provided this delay is small enough.
128  */
129 
130 #define IGMP_V1_SEEN(in_dev) \
131 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
132 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
133 	 ((in_dev)->mr_v1_seen && \
134 	  time_before(jiffies, (in_dev)->mr_v1_seen)))
135 #define IGMP_V2_SEEN(in_dev) \
136 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
137 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
138 	 ((in_dev)->mr_v2_seen && \
139 	  time_before(jiffies, (in_dev)->mr_v2_seen)))
140 
141 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
142 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
143 static void igmpv3_clear_delrec(struct in_device *in_dev);
144 static int sf_setstate(struct ip_mc_list *pmc);
145 static void sf_markstate(struct ip_mc_list *pmc);
146 #endif
147 static void ip_mc_clear_src(struct ip_mc_list *pmc);
148 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
149 			 int sfcount, __be32 *psfsrc, int delta);
150 
151 static void ip_ma_put(struct ip_mc_list *im)
152 {
153 	if (atomic_dec_and_test(&im->refcnt)) {
154 		in_dev_put(im->interface);
155 		kfree_rcu(im, rcu);
156 	}
157 }
158 
159 #define for_each_pmc_rcu(in_dev, pmc)				\
160 	for (pmc = rcu_dereference(in_dev->mc_list);		\
161 	     pmc != NULL;					\
162 	     pmc = rcu_dereference(pmc->next_rcu))
163 
164 #define for_each_pmc_rtnl(in_dev, pmc)				\
165 	for (pmc = rtnl_dereference(in_dev->mc_list);		\
166 	     pmc != NULL;					\
167 	     pmc = rtnl_dereference(pmc->next_rcu))
168 
169 #ifdef CONFIG_IP_MULTICAST
170 
171 /*
172  *	Timer management
173  */
174 
175 static void igmp_stop_timer(struct ip_mc_list *im)
176 {
177 	spin_lock_bh(&im->lock);
178 	if (del_timer(&im->timer))
179 		atomic_dec(&im->refcnt);
180 	im->tm_running = 0;
181 	im->reporter = 0;
182 	im->unsolicit_count = 0;
183 	spin_unlock_bh(&im->lock);
184 }
185 
186 /* It must be called with locked im->lock */
187 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
188 {
189 	int tv = net_random() % max_delay;
190 
191 	im->tm_running = 1;
192 	if (!mod_timer(&im->timer, jiffies+tv+2))
193 		atomic_inc(&im->refcnt);
194 }
195 
196 static void igmp_gq_start_timer(struct in_device *in_dev)
197 {
198 	int tv = net_random() % in_dev->mr_maxdelay;
199 
200 	in_dev->mr_gq_running = 1;
201 	if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
202 		in_dev_hold(in_dev);
203 }
204 
205 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
206 {
207 	int tv = net_random() % delay;
208 
209 	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
210 		in_dev_hold(in_dev);
211 }
212 
213 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
214 {
215 	spin_lock_bh(&im->lock);
216 	im->unsolicit_count = 0;
217 	if (del_timer(&im->timer)) {
218 		if ((long)(im->timer.expires-jiffies) < max_delay) {
219 			add_timer(&im->timer);
220 			im->tm_running = 1;
221 			spin_unlock_bh(&im->lock);
222 			return;
223 		}
224 		atomic_dec(&im->refcnt);
225 	}
226 	igmp_start_timer(im, max_delay);
227 	spin_unlock_bh(&im->lock);
228 }
229 
230 
231 /*
232  *	Send an IGMP report.
233  */
234 
235 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
236 
237 
238 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
239 	int gdeleted, int sdeleted)
240 {
241 	switch (type) {
242 	case IGMPV3_MODE_IS_INCLUDE:
243 	case IGMPV3_MODE_IS_EXCLUDE:
244 		if (gdeleted || sdeleted)
245 			return 0;
246 		if (!(pmc->gsquery && !psf->sf_gsresp)) {
247 			if (pmc->sfmode == MCAST_INCLUDE)
248 				return 1;
249 			/* don't include if this source is excluded
250 			 * in all filters
251 			 */
252 			if (psf->sf_count[MCAST_INCLUDE])
253 				return type == IGMPV3_MODE_IS_INCLUDE;
254 			return pmc->sfcount[MCAST_EXCLUDE] ==
255 				psf->sf_count[MCAST_EXCLUDE];
256 		}
257 		return 0;
258 	case IGMPV3_CHANGE_TO_INCLUDE:
259 		if (gdeleted || sdeleted)
260 			return 0;
261 		return psf->sf_count[MCAST_INCLUDE] != 0;
262 	case IGMPV3_CHANGE_TO_EXCLUDE:
263 		if (gdeleted || sdeleted)
264 			return 0;
265 		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
266 		    psf->sf_count[MCAST_INCLUDE])
267 			return 0;
268 		return pmc->sfcount[MCAST_EXCLUDE] ==
269 			psf->sf_count[MCAST_EXCLUDE];
270 	case IGMPV3_ALLOW_NEW_SOURCES:
271 		if (gdeleted || !psf->sf_crcount)
272 			return 0;
273 		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
274 	case IGMPV3_BLOCK_OLD_SOURCES:
275 		if (pmc->sfmode == MCAST_INCLUDE)
276 			return gdeleted || (psf->sf_crcount && sdeleted);
277 		return psf->sf_crcount && !gdeleted && !sdeleted;
278 	}
279 	return 0;
280 }
281 
282 static int
283 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
284 {
285 	struct ip_sf_list *psf;
286 	int scount = 0;
287 
288 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
289 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
290 			continue;
291 		scount++;
292 	}
293 	return scount;
294 }
295 
296 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
297 
298 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
299 {
300 	struct sk_buff *skb;
301 	struct rtable *rt;
302 	struct iphdr *pip;
303 	struct igmpv3_report *pig;
304 	struct net *net = dev_net(dev);
305 	struct flowi4 fl4;
306 	int hlen = LL_RESERVED_SPACE(dev);
307 	int tlen = dev->needed_tailroom;
308 
309 	while (1) {
310 		skb = alloc_skb(size + hlen + tlen,
311 				GFP_ATOMIC | __GFP_NOWARN);
312 		if (skb)
313 			break;
314 		size >>= 1;
315 		if (size < 256)
316 			return NULL;
317 	}
318 	igmp_skb_size(skb) = size;
319 
320 	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
321 				   0, 0,
322 				   IPPROTO_IGMP, 0, dev->ifindex);
323 	if (IS_ERR(rt)) {
324 		kfree_skb(skb);
325 		return NULL;
326 	}
327 
328 	skb_dst_set(skb, &rt->dst);
329 	skb->dev = dev;
330 
331 	skb_reserve(skb, hlen);
332 
333 	skb_reset_network_header(skb);
334 	pip = ip_hdr(skb);
335 	skb_put(skb, sizeof(struct iphdr) + 4);
336 
337 	pip->version  = 4;
338 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
339 	pip->tos      = 0xc0;
340 	pip->frag_off = htons(IP_DF);
341 	pip->ttl      = 1;
342 	pip->daddr    = fl4.daddr;
343 	pip->saddr    = fl4.saddr;
344 	pip->protocol = IPPROTO_IGMP;
345 	pip->tot_len  = 0;	/* filled in later */
346 	ip_select_ident(pip, &rt->dst, NULL);
347 	((u8 *)&pip[1])[0] = IPOPT_RA;
348 	((u8 *)&pip[1])[1] = 4;
349 	((u8 *)&pip[1])[2] = 0;
350 	((u8 *)&pip[1])[3] = 0;
351 
352 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
353 	skb_put(skb, sizeof(*pig));
354 	pig = igmpv3_report_hdr(skb);
355 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
356 	pig->resv1 = 0;
357 	pig->csum = 0;
358 	pig->resv2 = 0;
359 	pig->ngrec = 0;
360 	return skb;
361 }
362 
363 static int igmpv3_sendpack(struct sk_buff *skb)
364 {
365 	struct igmphdr *pig = igmp_hdr(skb);
366 	const int igmplen = skb->tail - skb->transport_header;
367 
368 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
369 
370 	return ip_local_out(skb);
371 }
372 
373 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
374 {
375 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
376 }
377 
378 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
379 	int type, struct igmpv3_grec **ppgr)
380 {
381 	struct net_device *dev = pmc->interface->dev;
382 	struct igmpv3_report *pih;
383 	struct igmpv3_grec *pgr;
384 
385 	if (!skb)
386 		skb = igmpv3_newpack(dev, dev->mtu);
387 	if (!skb)
388 		return NULL;
389 	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
390 	pgr->grec_type = type;
391 	pgr->grec_auxwords = 0;
392 	pgr->grec_nsrcs = 0;
393 	pgr->grec_mca = pmc->multiaddr;
394 	pih = igmpv3_report_hdr(skb);
395 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
396 	*ppgr = pgr;
397 	return skb;
398 }
399 
400 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
401 	skb_tailroom(skb)) : 0)
402 
403 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
404 	int type, int gdeleted, int sdeleted)
405 {
406 	struct net_device *dev = pmc->interface->dev;
407 	struct igmpv3_report *pih;
408 	struct igmpv3_grec *pgr = NULL;
409 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
410 	int scount, stotal, first, isquery, truncate;
411 
412 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
413 		return skb;
414 
415 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
416 		  type == IGMPV3_MODE_IS_EXCLUDE;
417 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
418 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
419 
420 	stotal = scount = 0;
421 
422 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
423 
424 	if (!*psf_list)
425 		goto empty_source;
426 
427 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
428 
429 	/* EX and TO_EX get a fresh packet, if needed */
430 	if (truncate) {
431 		if (pih && pih->ngrec &&
432 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
433 			if (skb)
434 				igmpv3_sendpack(skb);
435 			skb = igmpv3_newpack(dev, dev->mtu);
436 		}
437 	}
438 	first = 1;
439 	psf_prev = NULL;
440 	for (psf=*psf_list; psf; psf=psf_next) {
441 		__be32 *psrc;
442 
443 		psf_next = psf->sf_next;
444 
445 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
446 			psf_prev = psf;
447 			continue;
448 		}
449 
450 		/* clear marks on query responses */
451 		if (isquery)
452 			psf->sf_gsresp = 0;
453 
454 		if (AVAILABLE(skb) < sizeof(__be32) +
455 		    first*sizeof(struct igmpv3_grec)) {
456 			if (truncate && !first)
457 				break;	 /* truncate these */
458 			if (pgr)
459 				pgr->grec_nsrcs = htons(scount);
460 			if (skb)
461 				igmpv3_sendpack(skb);
462 			skb = igmpv3_newpack(dev, dev->mtu);
463 			first = 1;
464 			scount = 0;
465 		}
466 		if (first) {
467 			skb = add_grhead(skb, pmc, type, &pgr);
468 			first = 0;
469 		}
470 		if (!skb)
471 			return NULL;
472 		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
473 		*psrc = psf->sf_inaddr;
474 		scount++; stotal++;
475 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
476 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
477 			psf->sf_crcount--;
478 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
479 				if (psf_prev)
480 					psf_prev->sf_next = psf->sf_next;
481 				else
482 					*psf_list = psf->sf_next;
483 				kfree(psf);
484 				continue;
485 			}
486 		}
487 		psf_prev = psf;
488 	}
489 
490 empty_source:
491 	if (!stotal) {
492 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
493 		    type == IGMPV3_BLOCK_OLD_SOURCES)
494 			return skb;
495 		if (pmc->crcount || isquery) {
496 			/* make sure we have room for group header */
497 			if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
498 				igmpv3_sendpack(skb);
499 				skb = NULL; /* add_grhead will get a new one */
500 			}
501 			skb = add_grhead(skb, pmc, type, &pgr);
502 		}
503 	}
504 	if (pgr)
505 		pgr->grec_nsrcs = htons(scount);
506 
507 	if (isquery)
508 		pmc->gsquery = 0;	/* clear query state on report */
509 	return skb;
510 }
511 
512 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
513 {
514 	struct sk_buff *skb = NULL;
515 	int type;
516 
517 	if (!pmc) {
518 		rcu_read_lock();
519 		for_each_pmc_rcu(in_dev, pmc) {
520 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
521 				continue;
522 			spin_lock_bh(&pmc->lock);
523 			if (pmc->sfcount[MCAST_EXCLUDE])
524 				type = IGMPV3_MODE_IS_EXCLUDE;
525 			else
526 				type = IGMPV3_MODE_IS_INCLUDE;
527 			skb = add_grec(skb, pmc, type, 0, 0);
528 			spin_unlock_bh(&pmc->lock);
529 		}
530 		rcu_read_unlock();
531 	} else {
532 		spin_lock_bh(&pmc->lock);
533 		if (pmc->sfcount[MCAST_EXCLUDE])
534 			type = IGMPV3_MODE_IS_EXCLUDE;
535 		else
536 			type = IGMPV3_MODE_IS_INCLUDE;
537 		skb = add_grec(skb, pmc, type, 0, 0);
538 		spin_unlock_bh(&pmc->lock);
539 	}
540 	if (!skb)
541 		return 0;
542 	return igmpv3_sendpack(skb);
543 }
544 
545 /*
546  * remove zero-count source records from a source filter list
547  */
548 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
549 {
550 	struct ip_sf_list *psf_prev, *psf_next, *psf;
551 
552 	psf_prev = NULL;
553 	for (psf=*ppsf; psf; psf = psf_next) {
554 		psf_next = psf->sf_next;
555 		if (psf->sf_crcount == 0) {
556 			if (psf_prev)
557 				psf_prev->sf_next = psf->sf_next;
558 			else
559 				*ppsf = psf->sf_next;
560 			kfree(psf);
561 		} else
562 			psf_prev = psf;
563 	}
564 }
565 
566 static void igmpv3_send_cr(struct in_device *in_dev)
567 {
568 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
569 	struct sk_buff *skb = NULL;
570 	int type, dtype;
571 
572 	rcu_read_lock();
573 	spin_lock_bh(&in_dev->mc_tomb_lock);
574 
575 	/* deleted MCA's */
576 	pmc_prev = NULL;
577 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
578 		pmc_next = pmc->next;
579 		if (pmc->sfmode == MCAST_INCLUDE) {
580 			type = IGMPV3_BLOCK_OLD_SOURCES;
581 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
582 			skb = add_grec(skb, pmc, type, 1, 0);
583 			skb = add_grec(skb, pmc, dtype, 1, 1);
584 		}
585 		if (pmc->crcount) {
586 			if (pmc->sfmode == MCAST_EXCLUDE) {
587 				type = IGMPV3_CHANGE_TO_INCLUDE;
588 				skb = add_grec(skb, pmc, type, 1, 0);
589 			}
590 			pmc->crcount--;
591 			if (pmc->crcount == 0) {
592 				igmpv3_clear_zeros(&pmc->tomb);
593 				igmpv3_clear_zeros(&pmc->sources);
594 			}
595 		}
596 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
597 			if (pmc_prev)
598 				pmc_prev->next = pmc_next;
599 			else
600 				in_dev->mc_tomb = pmc_next;
601 			in_dev_put(pmc->interface);
602 			kfree(pmc);
603 		} else
604 			pmc_prev = pmc;
605 	}
606 	spin_unlock_bh(&in_dev->mc_tomb_lock);
607 
608 	/* change recs */
609 	for_each_pmc_rcu(in_dev, pmc) {
610 		spin_lock_bh(&pmc->lock);
611 		if (pmc->sfcount[MCAST_EXCLUDE]) {
612 			type = IGMPV3_BLOCK_OLD_SOURCES;
613 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
614 		} else {
615 			type = IGMPV3_ALLOW_NEW_SOURCES;
616 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
617 		}
618 		skb = add_grec(skb, pmc, type, 0, 0);
619 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
620 
621 		/* filter mode changes */
622 		if (pmc->crcount) {
623 			if (pmc->sfmode == MCAST_EXCLUDE)
624 				type = IGMPV3_CHANGE_TO_EXCLUDE;
625 			else
626 				type = IGMPV3_CHANGE_TO_INCLUDE;
627 			skb = add_grec(skb, pmc, type, 0, 0);
628 			pmc->crcount--;
629 		}
630 		spin_unlock_bh(&pmc->lock);
631 	}
632 	rcu_read_unlock();
633 
634 	if (!skb)
635 		return;
636 	(void) igmpv3_sendpack(skb);
637 }
638 
639 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
640 	int type)
641 {
642 	struct sk_buff *skb;
643 	struct iphdr *iph;
644 	struct igmphdr *ih;
645 	struct rtable *rt;
646 	struct net_device *dev = in_dev->dev;
647 	struct net *net = dev_net(dev);
648 	__be32	group = pmc ? pmc->multiaddr : 0;
649 	struct flowi4 fl4;
650 	__be32	dst;
651 	int hlen, tlen;
652 
653 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
654 		return igmpv3_send_report(in_dev, pmc);
655 	else if (type == IGMP_HOST_LEAVE_MESSAGE)
656 		dst = IGMP_ALL_ROUTER;
657 	else
658 		dst = group;
659 
660 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
661 				   0, 0,
662 				   IPPROTO_IGMP, 0, dev->ifindex);
663 	if (IS_ERR(rt))
664 		return -1;
665 
666 	hlen = LL_RESERVED_SPACE(dev);
667 	tlen = dev->needed_tailroom;
668 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
669 	if (skb == NULL) {
670 		ip_rt_put(rt);
671 		return -1;
672 	}
673 
674 	skb_dst_set(skb, &rt->dst);
675 
676 	skb_reserve(skb, hlen);
677 
678 	skb_reset_network_header(skb);
679 	iph = ip_hdr(skb);
680 	skb_put(skb, sizeof(struct iphdr) + 4);
681 
682 	iph->version  = 4;
683 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
684 	iph->tos      = 0xc0;
685 	iph->frag_off = htons(IP_DF);
686 	iph->ttl      = 1;
687 	iph->daddr    = dst;
688 	iph->saddr    = fl4.saddr;
689 	iph->protocol = IPPROTO_IGMP;
690 	ip_select_ident(iph, &rt->dst, NULL);
691 	((u8 *)&iph[1])[0] = IPOPT_RA;
692 	((u8 *)&iph[1])[1] = 4;
693 	((u8 *)&iph[1])[2] = 0;
694 	((u8 *)&iph[1])[3] = 0;
695 
696 	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
697 	ih->type = type;
698 	ih->code = 0;
699 	ih->csum = 0;
700 	ih->group = group;
701 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
702 
703 	return ip_local_out(skb);
704 }
705 
706 static void igmp_gq_timer_expire(unsigned long data)
707 {
708 	struct in_device *in_dev = (struct in_device *)data;
709 
710 	in_dev->mr_gq_running = 0;
711 	igmpv3_send_report(in_dev, NULL);
712 	__in_dev_put(in_dev);
713 }
714 
715 static void igmp_ifc_timer_expire(unsigned long data)
716 {
717 	struct in_device *in_dev = (struct in_device *)data;
718 
719 	igmpv3_send_cr(in_dev);
720 	if (in_dev->mr_ifc_count) {
721 		in_dev->mr_ifc_count--;
722 		igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
723 	}
724 	__in_dev_put(in_dev);
725 }
726 
727 static void igmp_ifc_event(struct in_device *in_dev)
728 {
729 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
730 		return;
731 	in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
732 		IGMP_Unsolicited_Report_Count;
733 	igmp_ifc_start_timer(in_dev, 1);
734 }
735 
736 
737 static void igmp_timer_expire(unsigned long data)
738 {
739 	struct ip_mc_list *im=(struct ip_mc_list *)data;
740 	struct in_device *in_dev = im->interface;
741 
742 	spin_lock(&im->lock);
743 	im->tm_running = 0;
744 
745 	if (im->unsolicit_count) {
746 		im->unsolicit_count--;
747 		igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
748 	}
749 	im->reporter = 1;
750 	spin_unlock(&im->lock);
751 
752 	if (IGMP_V1_SEEN(in_dev))
753 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
754 	else if (IGMP_V2_SEEN(in_dev))
755 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
756 	else
757 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
758 
759 	ip_ma_put(im);
760 }
761 
762 /* mark EXCLUDE-mode sources */
763 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
764 {
765 	struct ip_sf_list *psf;
766 	int i, scount;
767 
768 	scount = 0;
769 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
770 		if (scount == nsrcs)
771 			break;
772 		for (i=0; i<nsrcs; i++) {
773 			/* skip inactive filters */
774 			if (psf->sf_count[MCAST_INCLUDE] ||
775 			    pmc->sfcount[MCAST_EXCLUDE] !=
776 			    psf->sf_count[MCAST_EXCLUDE])
777 				break;
778 			if (srcs[i] == psf->sf_inaddr) {
779 				scount++;
780 				break;
781 			}
782 		}
783 	}
784 	pmc->gsquery = 0;
785 	if (scount == nsrcs)	/* all sources excluded */
786 		return 0;
787 	return 1;
788 }
789 
790 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
791 {
792 	struct ip_sf_list *psf;
793 	int i, scount;
794 
795 	if (pmc->sfmode == MCAST_EXCLUDE)
796 		return igmp_xmarksources(pmc, nsrcs, srcs);
797 
798 	/* mark INCLUDE-mode sources */
799 	scount = 0;
800 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
801 		if (scount == nsrcs)
802 			break;
803 		for (i=0; i<nsrcs; i++)
804 			if (srcs[i] == psf->sf_inaddr) {
805 				psf->sf_gsresp = 1;
806 				scount++;
807 				break;
808 			}
809 	}
810 	if (!scount) {
811 		pmc->gsquery = 0;
812 		return 0;
813 	}
814 	pmc->gsquery = 1;
815 	return 1;
816 }
817 
818 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
819 {
820 	struct ip_mc_list *im;
821 
822 	/* Timers are only set for non-local groups */
823 
824 	if (group == IGMP_ALL_HOSTS)
825 		return;
826 
827 	rcu_read_lock();
828 	for_each_pmc_rcu(in_dev, im) {
829 		if (im->multiaddr == group) {
830 			igmp_stop_timer(im);
831 			break;
832 		}
833 	}
834 	rcu_read_unlock();
835 }
836 
837 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
838 	int len)
839 {
840 	struct igmphdr 		*ih = igmp_hdr(skb);
841 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
842 	struct ip_mc_list	*im;
843 	__be32			group = ih->group;
844 	int			max_delay;
845 	int			mark = 0;
846 
847 
848 	if (len == 8) {
849 		if (ih->code == 0) {
850 			/* Alas, old v1 router presents here. */
851 
852 			max_delay = IGMP_Query_Response_Interval;
853 			in_dev->mr_v1_seen = jiffies +
854 				IGMP_V1_Router_Present_Timeout;
855 			group = 0;
856 		} else {
857 			/* v2 router present */
858 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
859 			in_dev->mr_v2_seen = jiffies +
860 				IGMP_V2_Router_Present_Timeout;
861 		}
862 		/* cancel the interface change timer */
863 		in_dev->mr_ifc_count = 0;
864 		if (del_timer(&in_dev->mr_ifc_timer))
865 			__in_dev_put(in_dev);
866 		/* clear deleted report items */
867 		igmpv3_clear_delrec(in_dev);
868 	} else if (len < 12) {
869 		return;	/* ignore bogus packet; freed by caller */
870 	} else if (IGMP_V1_SEEN(in_dev)) {
871 		/* This is a v3 query with v1 queriers present */
872 		max_delay = IGMP_Query_Response_Interval;
873 		group = 0;
874 	} else if (IGMP_V2_SEEN(in_dev)) {
875 		/* this is a v3 query with v2 queriers present;
876 		 * Interpretation of the max_delay code is problematic here.
877 		 * A real v2 host would use ih_code directly, while v3 has a
878 		 * different encoding. We use the v3 encoding as more likely
879 		 * to be intended in a v3 query.
880 		 */
881 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
882 		if (!max_delay)
883 			max_delay = 1;	/* can't mod w/ 0 */
884 	} else { /* v3 */
885 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
886 			return;
887 
888 		ih3 = igmpv3_query_hdr(skb);
889 		if (ih3->nsrcs) {
890 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
891 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
892 				return;
893 			ih3 = igmpv3_query_hdr(skb);
894 		}
895 
896 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
897 		if (!max_delay)
898 			max_delay = 1;	/* can't mod w/ 0 */
899 		in_dev->mr_maxdelay = max_delay;
900 		if (ih3->qrv)
901 			in_dev->mr_qrv = ih3->qrv;
902 		if (!group) { /* general query */
903 			if (ih3->nsrcs)
904 				return;	/* no sources allowed */
905 			igmp_gq_start_timer(in_dev);
906 			return;
907 		}
908 		/* mark sources to include, if group & source-specific */
909 		mark = ih3->nsrcs != 0;
910 	}
911 
912 	/*
913 	 * - Start the timers in all of our membership records
914 	 *   that the query applies to for the interface on
915 	 *   which the query arrived excl. those that belong
916 	 *   to a "local" group (224.0.0.X)
917 	 * - For timers already running check if they need to
918 	 *   be reset.
919 	 * - Use the igmp->igmp_code field as the maximum
920 	 *   delay possible
921 	 */
922 	rcu_read_lock();
923 	for_each_pmc_rcu(in_dev, im) {
924 		int changed;
925 
926 		if (group && group != im->multiaddr)
927 			continue;
928 		if (im->multiaddr == IGMP_ALL_HOSTS)
929 			continue;
930 		spin_lock_bh(&im->lock);
931 		if (im->tm_running)
932 			im->gsquery = im->gsquery && mark;
933 		else
934 			im->gsquery = mark;
935 		changed = !im->gsquery ||
936 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
937 		spin_unlock_bh(&im->lock);
938 		if (changed)
939 			igmp_mod_timer(im, max_delay);
940 	}
941 	rcu_read_unlock();
942 }
943 
944 /* called in rcu_read_lock() section */
945 int igmp_rcv(struct sk_buff *skb)
946 {
947 	/* This basically follows the spec line by line -- see RFC1112 */
948 	struct igmphdr *ih;
949 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
950 	int len = skb->len;
951 
952 	if (in_dev == NULL)
953 		goto drop;
954 
955 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
956 		goto drop;
957 
958 	switch (skb->ip_summed) {
959 	case CHECKSUM_COMPLETE:
960 		if (!csum_fold(skb->csum))
961 			break;
962 		/* fall through */
963 	case CHECKSUM_NONE:
964 		skb->csum = 0;
965 		if (__skb_checksum_complete(skb))
966 			goto drop;
967 	}
968 
969 	ih = igmp_hdr(skb);
970 	switch (ih->type) {
971 	case IGMP_HOST_MEMBERSHIP_QUERY:
972 		igmp_heard_query(in_dev, skb, len);
973 		break;
974 	case IGMP_HOST_MEMBERSHIP_REPORT:
975 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
976 		/* Is it our report looped back? */
977 		if (rt_is_output_route(skb_rtable(skb)))
978 			break;
979 		/* don't rely on MC router hearing unicast reports */
980 		if (skb->pkt_type == PACKET_MULTICAST ||
981 		    skb->pkt_type == PACKET_BROADCAST)
982 			igmp_heard_report(in_dev, ih->group);
983 		break;
984 	case IGMP_PIM:
985 #ifdef CONFIG_IP_PIMSM_V1
986 		return pim_rcv_v1(skb);
987 #endif
988 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
989 	case IGMP_DVMRP:
990 	case IGMP_TRACE:
991 	case IGMP_HOST_LEAVE_MESSAGE:
992 	case IGMP_MTRACE:
993 	case IGMP_MTRACE_RESP:
994 		break;
995 	default:
996 		break;
997 	}
998 
999 drop:
1000 	kfree_skb(skb);
1001 	return 0;
1002 }
1003 
1004 #endif
1005 
1006 
1007 /*
1008  *	Add a filter to a device
1009  */
1010 
1011 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1012 {
1013 	char buf[MAX_ADDR_LEN];
1014 	struct net_device *dev = in_dev->dev;
1015 
1016 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1017 	   We will get multicast token leakage, when IFF_MULTICAST
1018 	   is changed. This check should be done in ndo_set_rx_mode
1019 	   routine. Something sort of:
1020 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1021 	   --ANK
1022 	   */
1023 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1024 		dev_mc_add(dev, buf);
1025 }
1026 
1027 /*
1028  *	Remove a filter from a device
1029  */
1030 
1031 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1032 {
1033 	char buf[MAX_ADDR_LEN];
1034 	struct net_device *dev = in_dev->dev;
1035 
1036 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1037 		dev_mc_del(dev, buf);
1038 }
1039 
1040 #ifdef CONFIG_IP_MULTICAST
1041 /*
1042  * deleted ip_mc_list manipulation
1043  */
1044 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1045 {
1046 	struct ip_mc_list *pmc;
1047 
1048 	/* this is an "ip_mc_list" for convenience; only the fields below
1049 	 * are actually used. In particular, the refcnt and users are not
1050 	 * used for management of the delete list. Using the same structure
1051 	 * for deleted items allows change reports to use common code with
1052 	 * non-deleted or query-response MCA's.
1053 	 */
1054 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1055 	if (!pmc)
1056 		return;
1057 	spin_lock_bh(&im->lock);
1058 	pmc->interface = im->interface;
1059 	in_dev_hold(in_dev);
1060 	pmc->multiaddr = im->multiaddr;
1061 	pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1062 		IGMP_Unsolicited_Report_Count;
1063 	pmc->sfmode = im->sfmode;
1064 	if (pmc->sfmode == MCAST_INCLUDE) {
1065 		struct ip_sf_list *psf;
1066 
1067 		pmc->tomb = im->tomb;
1068 		pmc->sources = im->sources;
1069 		im->tomb = im->sources = NULL;
1070 		for (psf=pmc->sources; psf; psf=psf->sf_next)
1071 			psf->sf_crcount = pmc->crcount;
1072 	}
1073 	spin_unlock_bh(&im->lock);
1074 
1075 	spin_lock_bh(&in_dev->mc_tomb_lock);
1076 	pmc->next = in_dev->mc_tomb;
1077 	in_dev->mc_tomb = pmc;
1078 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1079 }
1080 
1081 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1082 {
1083 	struct ip_mc_list *pmc, *pmc_prev;
1084 	struct ip_sf_list *psf, *psf_next;
1085 
1086 	spin_lock_bh(&in_dev->mc_tomb_lock);
1087 	pmc_prev = NULL;
1088 	for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1089 		if (pmc->multiaddr == multiaddr)
1090 			break;
1091 		pmc_prev = pmc;
1092 	}
1093 	if (pmc) {
1094 		if (pmc_prev)
1095 			pmc_prev->next = pmc->next;
1096 		else
1097 			in_dev->mc_tomb = pmc->next;
1098 	}
1099 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1100 	if (pmc) {
1101 		for (psf=pmc->tomb; psf; psf=psf_next) {
1102 			psf_next = psf->sf_next;
1103 			kfree(psf);
1104 		}
1105 		in_dev_put(pmc->interface);
1106 		kfree(pmc);
1107 	}
1108 }
1109 
1110 static void igmpv3_clear_delrec(struct in_device *in_dev)
1111 {
1112 	struct ip_mc_list *pmc, *nextpmc;
1113 
1114 	spin_lock_bh(&in_dev->mc_tomb_lock);
1115 	pmc = in_dev->mc_tomb;
1116 	in_dev->mc_tomb = NULL;
1117 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1118 
1119 	for (; pmc; pmc = nextpmc) {
1120 		nextpmc = pmc->next;
1121 		ip_mc_clear_src(pmc);
1122 		in_dev_put(pmc->interface);
1123 		kfree(pmc);
1124 	}
1125 	/* clear dead sources, too */
1126 	rcu_read_lock();
1127 	for_each_pmc_rcu(in_dev, pmc) {
1128 		struct ip_sf_list *psf, *psf_next;
1129 
1130 		spin_lock_bh(&pmc->lock);
1131 		psf = pmc->tomb;
1132 		pmc->tomb = NULL;
1133 		spin_unlock_bh(&pmc->lock);
1134 		for (; psf; psf=psf_next) {
1135 			psf_next = psf->sf_next;
1136 			kfree(psf);
1137 		}
1138 	}
1139 	rcu_read_unlock();
1140 }
1141 #endif
1142 
1143 static void igmp_group_dropped(struct ip_mc_list *im)
1144 {
1145 	struct in_device *in_dev = im->interface;
1146 #ifdef CONFIG_IP_MULTICAST
1147 	int reporter;
1148 #endif
1149 
1150 	if (im->loaded) {
1151 		im->loaded = 0;
1152 		ip_mc_filter_del(in_dev, im->multiaddr);
1153 	}
1154 
1155 #ifdef CONFIG_IP_MULTICAST
1156 	if (im->multiaddr == IGMP_ALL_HOSTS)
1157 		return;
1158 
1159 	reporter = im->reporter;
1160 	igmp_stop_timer(im);
1161 
1162 	if (!in_dev->dead) {
1163 		if (IGMP_V1_SEEN(in_dev))
1164 			return;
1165 		if (IGMP_V2_SEEN(in_dev)) {
1166 			if (reporter)
1167 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1168 			return;
1169 		}
1170 		/* IGMPv3 */
1171 		igmpv3_add_delrec(in_dev, im);
1172 
1173 		igmp_ifc_event(in_dev);
1174 	}
1175 #endif
1176 }
1177 
1178 static void igmp_group_added(struct ip_mc_list *im)
1179 {
1180 	struct in_device *in_dev = im->interface;
1181 
1182 	if (im->loaded == 0) {
1183 		im->loaded = 1;
1184 		ip_mc_filter_add(in_dev, im->multiaddr);
1185 	}
1186 
1187 #ifdef CONFIG_IP_MULTICAST
1188 	if (im->multiaddr == IGMP_ALL_HOSTS)
1189 		return;
1190 
1191 	if (in_dev->dead)
1192 		return;
1193 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1194 		spin_lock_bh(&im->lock);
1195 		igmp_start_timer(im, IGMP_Initial_Report_Delay);
1196 		spin_unlock_bh(&im->lock);
1197 		return;
1198 	}
1199 	/* else, v3 */
1200 
1201 	im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1202 		IGMP_Unsolicited_Report_Count;
1203 	igmp_ifc_event(in_dev);
1204 #endif
1205 }
1206 
1207 
1208 /*
1209  *	Multicast list managers
1210  */
1211 
1212 
1213 /*
1214  *	A socket has joined a multicast group on device dev.
1215  */
1216 
1217 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1218 {
1219 	struct ip_mc_list *im;
1220 
1221 	ASSERT_RTNL();
1222 
1223 	for_each_pmc_rtnl(in_dev, im) {
1224 		if (im->multiaddr == addr) {
1225 			im->users++;
1226 			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1227 			goto out;
1228 		}
1229 	}
1230 
1231 	im = kzalloc(sizeof(*im), GFP_KERNEL);
1232 	if (!im)
1233 		goto out;
1234 
1235 	im->users = 1;
1236 	im->interface = in_dev;
1237 	in_dev_hold(in_dev);
1238 	im->multiaddr = addr;
1239 	/* initial mode is (EX, empty) */
1240 	im->sfmode = MCAST_EXCLUDE;
1241 	im->sfcount[MCAST_EXCLUDE] = 1;
1242 	atomic_set(&im->refcnt, 1);
1243 	spin_lock_init(&im->lock);
1244 #ifdef CONFIG_IP_MULTICAST
1245 	setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1246 	im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1247 #endif
1248 
1249 	im->next_rcu = in_dev->mc_list;
1250 	in_dev->mc_count++;
1251 	rcu_assign_pointer(in_dev->mc_list, im);
1252 
1253 #ifdef CONFIG_IP_MULTICAST
1254 	igmpv3_del_delrec(in_dev, im->multiaddr);
1255 #endif
1256 	igmp_group_added(im);
1257 	if (!in_dev->dead)
1258 		ip_rt_multicast_event(in_dev);
1259 out:
1260 	return;
1261 }
1262 EXPORT_SYMBOL(ip_mc_inc_group);
1263 
1264 /*
1265  *	Resend IGMP JOIN report; used for bonding.
1266  *	Called with rcu_read_lock()
1267  */
1268 void ip_mc_rejoin_groups(struct in_device *in_dev)
1269 {
1270 #ifdef CONFIG_IP_MULTICAST
1271 	struct ip_mc_list *im;
1272 	int type;
1273 
1274 	for_each_pmc_rcu(in_dev, im) {
1275 		if (im->multiaddr == IGMP_ALL_HOSTS)
1276 			continue;
1277 
1278 		/* a failover is happening and switches
1279 		 * must be notified immediately
1280 		 */
1281 		if (IGMP_V1_SEEN(in_dev))
1282 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1283 		else if (IGMP_V2_SEEN(in_dev))
1284 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1285 		else
1286 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1287 		igmp_send_report(in_dev, im, type);
1288 	}
1289 #endif
1290 }
1291 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1292 
1293 /*
1294  *	A socket has left a multicast group on device dev
1295  */
1296 
1297 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1298 {
1299 	struct ip_mc_list *i;
1300 	struct ip_mc_list __rcu **ip;
1301 
1302 	ASSERT_RTNL();
1303 
1304 	for (ip = &in_dev->mc_list;
1305 	     (i = rtnl_dereference(*ip)) != NULL;
1306 	     ip = &i->next_rcu) {
1307 		if (i->multiaddr == addr) {
1308 			if (--i->users == 0) {
1309 				*ip = i->next_rcu;
1310 				in_dev->mc_count--;
1311 				igmp_group_dropped(i);
1312 				ip_mc_clear_src(i);
1313 
1314 				if (!in_dev->dead)
1315 					ip_rt_multicast_event(in_dev);
1316 
1317 				ip_ma_put(i);
1318 				return;
1319 			}
1320 			break;
1321 		}
1322 	}
1323 }
1324 EXPORT_SYMBOL(ip_mc_dec_group);
1325 
1326 /* Device changing type */
1327 
1328 void ip_mc_unmap(struct in_device *in_dev)
1329 {
1330 	struct ip_mc_list *pmc;
1331 
1332 	ASSERT_RTNL();
1333 
1334 	for_each_pmc_rtnl(in_dev, pmc)
1335 		igmp_group_dropped(pmc);
1336 }
1337 
1338 void ip_mc_remap(struct in_device *in_dev)
1339 {
1340 	struct ip_mc_list *pmc;
1341 
1342 	ASSERT_RTNL();
1343 
1344 	for_each_pmc_rtnl(in_dev, pmc)
1345 		igmp_group_added(pmc);
1346 }
1347 
1348 /* Device going down */
1349 
1350 void ip_mc_down(struct in_device *in_dev)
1351 {
1352 	struct ip_mc_list *pmc;
1353 
1354 	ASSERT_RTNL();
1355 
1356 	for_each_pmc_rtnl(in_dev, pmc)
1357 		igmp_group_dropped(pmc);
1358 
1359 #ifdef CONFIG_IP_MULTICAST
1360 	in_dev->mr_ifc_count = 0;
1361 	if (del_timer(&in_dev->mr_ifc_timer))
1362 		__in_dev_put(in_dev);
1363 	in_dev->mr_gq_running = 0;
1364 	if (del_timer(&in_dev->mr_gq_timer))
1365 		__in_dev_put(in_dev);
1366 	igmpv3_clear_delrec(in_dev);
1367 #endif
1368 
1369 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1370 }
1371 
1372 void ip_mc_init_dev(struct in_device *in_dev)
1373 {
1374 	ASSERT_RTNL();
1375 
1376 	in_dev->mc_tomb = NULL;
1377 #ifdef CONFIG_IP_MULTICAST
1378 	in_dev->mr_gq_running = 0;
1379 	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1380 			(unsigned long)in_dev);
1381 	in_dev->mr_ifc_count = 0;
1382 	in_dev->mc_count     = 0;
1383 	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1384 			(unsigned long)in_dev);
1385 	in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1386 #endif
1387 
1388 	spin_lock_init(&in_dev->mc_tomb_lock);
1389 }
1390 
1391 /* Device going up */
1392 
1393 void ip_mc_up(struct in_device *in_dev)
1394 {
1395 	struct ip_mc_list *pmc;
1396 
1397 	ASSERT_RTNL();
1398 
1399 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1400 
1401 	for_each_pmc_rtnl(in_dev, pmc)
1402 		igmp_group_added(pmc);
1403 }
1404 
1405 /*
1406  *	Device is about to be destroyed: clean up.
1407  */
1408 
1409 void ip_mc_destroy_dev(struct in_device *in_dev)
1410 {
1411 	struct ip_mc_list *i;
1412 
1413 	ASSERT_RTNL();
1414 
1415 	/* Deactivate timers */
1416 	ip_mc_down(in_dev);
1417 
1418 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1419 		in_dev->mc_list = i->next_rcu;
1420 		in_dev->mc_count--;
1421 
1422 		/* We've dropped the groups in ip_mc_down already */
1423 		ip_mc_clear_src(i);
1424 		ip_ma_put(i);
1425 	}
1426 }
1427 
1428 /* RTNL is locked */
1429 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1430 {
1431 	struct net_device *dev = NULL;
1432 	struct in_device *idev = NULL;
1433 
1434 	if (imr->imr_ifindex) {
1435 		idev = inetdev_by_index(net, imr->imr_ifindex);
1436 		return idev;
1437 	}
1438 	if (imr->imr_address.s_addr) {
1439 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1440 		if (!dev)
1441 			return NULL;
1442 	}
1443 
1444 	if (!dev) {
1445 		struct rtable *rt = ip_route_output(net,
1446 						    imr->imr_multiaddr.s_addr,
1447 						    0, 0, 0);
1448 		if (!IS_ERR(rt)) {
1449 			dev = rt->dst.dev;
1450 			ip_rt_put(rt);
1451 		}
1452 	}
1453 	if (dev) {
1454 		imr->imr_ifindex = dev->ifindex;
1455 		idev = __in_dev_get_rtnl(dev);
1456 	}
1457 	return idev;
1458 }
1459 
1460 /*
1461  *	Join a socket to a group
1462  */
1463 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1464 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1465 
1466 
1467 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1468 	__be32 *psfsrc)
1469 {
1470 	struct ip_sf_list *psf, *psf_prev;
1471 	int rv = 0;
1472 
1473 	psf_prev = NULL;
1474 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1475 		if (psf->sf_inaddr == *psfsrc)
1476 			break;
1477 		psf_prev = psf;
1478 	}
1479 	if (!psf || psf->sf_count[sfmode] == 0) {
1480 		/* source filter not found, or count wrong =>  bug */
1481 		return -ESRCH;
1482 	}
1483 	psf->sf_count[sfmode]--;
1484 	if (psf->sf_count[sfmode] == 0) {
1485 		ip_rt_multicast_event(pmc->interface);
1486 	}
1487 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1488 #ifdef CONFIG_IP_MULTICAST
1489 		struct in_device *in_dev = pmc->interface;
1490 #endif
1491 
1492 		/* no more filters for this source */
1493 		if (psf_prev)
1494 			psf_prev->sf_next = psf->sf_next;
1495 		else
1496 			pmc->sources = psf->sf_next;
1497 #ifdef CONFIG_IP_MULTICAST
1498 		if (psf->sf_oldin &&
1499 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1500 			psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1501 				IGMP_Unsolicited_Report_Count;
1502 			psf->sf_next = pmc->tomb;
1503 			pmc->tomb = psf;
1504 			rv = 1;
1505 		} else
1506 #endif
1507 			kfree(psf);
1508 	}
1509 	return rv;
1510 }
1511 
1512 #ifndef CONFIG_IP_MULTICAST
1513 #define igmp_ifc_event(x)	do { } while (0)
1514 #endif
1515 
1516 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1517 			 int sfcount, __be32 *psfsrc, int delta)
1518 {
1519 	struct ip_mc_list *pmc;
1520 	int	changerec = 0;
1521 	int	i, err;
1522 
1523 	if (!in_dev)
1524 		return -ENODEV;
1525 	rcu_read_lock();
1526 	for_each_pmc_rcu(in_dev, pmc) {
1527 		if (*pmca == pmc->multiaddr)
1528 			break;
1529 	}
1530 	if (!pmc) {
1531 		/* MCA not found?? bug */
1532 		rcu_read_unlock();
1533 		return -ESRCH;
1534 	}
1535 	spin_lock_bh(&pmc->lock);
1536 	rcu_read_unlock();
1537 #ifdef CONFIG_IP_MULTICAST
1538 	sf_markstate(pmc);
1539 #endif
1540 	if (!delta) {
1541 		err = -EINVAL;
1542 		if (!pmc->sfcount[sfmode])
1543 			goto out_unlock;
1544 		pmc->sfcount[sfmode]--;
1545 	}
1546 	err = 0;
1547 	for (i=0; i<sfcount; i++) {
1548 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1549 
1550 		changerec |= rv > 0;
1551 		if (!err && rv < 0)
1552 			err = rv;
1553 	}
1554 	if (pmc->sfmode == MCAST_EXCLUDE &&
1555 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1556 	    pmc->sfcount[MCAST_INCLUDE]) {
1557 #ifdef CONFIG_IP_MULTICAST
1558 		struct ip_sf_list *psf;
1559 #endif
1560 
1561 		/* filter mode change */
1562 		pmc->sfmode = MCAST_INCLUDE;
1563 #ifdef CONFIG_IP_MULTICAST
1564 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1565 			IGMP_Unsolicited_Report_Count;
1566 		in_dev->mr_ifc_count = pmc->crcount;
1567 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1568 			psf->sf_crcount = 0;
1569 		igmp_ifc_event(pmc->interface);
1570 	} else if (sf_setstate(pmc) || changerec) {
1571 		igmp_ifc_event(pmc->interface);
1572 #endif
1573 	}
1574 out_unlock:
1575 	spin_unlock_bh(&pmc->lock);
1576 	return err;
1577 }
1578 
1579 /*
1580  * Add multicast single-source filter to the interface list
1581  */
1582 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1583 	__be32 *psfsrc)
1584 {
1585 	struct ip_sf_list *psf, *psf_prev;
1586 
1587 	psf_prev = NULL;
1588 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1589 		if (psf->sf_inaddr == *psfsrc)
1590 			break;
1591 		psf_prev = psf;
1592 	}
1593 	if (!psf) {
1594 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1595 		if (!psf)
1596 			return -ENOBUFS;
1597 		psf->sf_inaddr = *psfsrc;
1598 		if (psf_prev) {
1599 			psf_prev->sf_next = psf;
1600 		} else
1601 			pmc->sources = psf;
1602 	}
1603 	psf->sf_count[sfmode]++;
1604 	if (psf->sf_count[sfmode] == 1) {
1605 		ip_rt_multicast_event(pmc->interface);
1606 	}
1607 	return 0;
1608 }
1609 
1610 #ifdef CONFIG_IP_MULTICAST
1611 static void sf_markstate(struct ip_mc_list *pmc)
1612 {
1613 	struct ip_sf_list *psf;
1614 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1615 
1616 	for (psf=pmc->sources; psf; psf=psf->sf_next)
1617 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1618 			psf->sf_oldin = mca_xcount ==
1619 				psf->sf_count[MCAST_EXCLUDE] &&
1620 				!psf->sf_count[MCAST_INCLUDE];
1621 		} else
1622 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1623 }
1624 
1625 static int sf_setstate(struct ip_mc_list *pmc)
1626 {
1627 	struct ip_sf_list *psf, *dpsf;
1628 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1629 	int qrv = pmc->interface->mr_qrv;
1630 	int new_in, rv;
1631 
1632 	rv = 0;
1633 	for (psf=pmc->sources; psf; psf=psf->sf_next) {
1634 		if (pmc->sfcount[MCAST_EXCLUDE]) {
1635 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1636 				!psf->sf_count[MCAST_INCLUDE];
1637 		} else
1638 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1639 		if (new_in) {
1640 			if (!psf->sf_oldin) {
1641 				struct ip_sf_list *prev = NULL;
1642 
1643 				for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1644 					if (dpsf->sf_inaddr == psf->sf_inaddr)
1645 						break;
1646 					prev = dpsf;
1647 				}
1648 				if (dpsf) {
1649 					if (prev)
1650 						prev->sf_next = dpsf->sf_next;
1651 					else
1652 						pmc->tomb = dpsf->sf_next;
1653 					kfree(dpsf);
1654 				}
1655 				psf->sf_crcount = qrv;
1656 				rv++;
1657 			}
1658 		} else if (psf->sf_oldin) {
1659 
1660 			psf->sf_crcount = 0;
1661 			/*
1662 			 * add or update "delete" records if an active filter
1663 			 * is now inactive
1664 			 */
1665 			for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1666 				if (dpsf->sf_inaddr == psf->sf_inaddr)
1667 					break;
1668 			if (!dpsf) {
1669 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1670 				if (!dpsf)
1671 					continue;
1672 				*dpsf = *psf;
1673 				/* pmc->lock held by callers */
1674 				dpsf->sf_next = pmc->tomb;
1675 				pmc->tomb = dpsf;
1676 			}
1677 			dpsf->sf_crcount = qrv;
1678 			rv++;
1679 		}
1680 	}
1681 	return rv;
1682 }
1683 #endif
1684 
1685 /*
1686  * Add multicast source filter list to the interface list
1687  */
1688 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1689 			 int sfcount, __be32 *psfsrc, int delta)
1690 {
1691 	struct ip_mc_list *pmc;
1692 	int	isexclude;
1693 	int	i, err;
1694 
1695 	if (!in_dev)
1696 		return -ENODEV;
1697 	rcu_read_lock();
1698 	for_each_pmc_rcu(in_dev, pmc) {
1699 		if (*pmca == pmc->multiaddr)
1700 			break;
1701 	}
1702 	if (!pmc) {
1703 		/* MCA not found?? bug */
1704 		rcu_read_unlock();
1705 		return -ESRCH;
1706 	}
1707 	spin_lock_bh(&pmc->lock);
1708 	rcu_read_unlock();
1709 
1710 #ifdef CONFIG_IP_MULTICAST
1711 	sf_markstate(pmc);
1712 #endif
1713 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
1714 	if (!delta)
1715 		pmc->sfcount[sfmode]++;
1716 	err = 0;
1717 	for (i=0; i<sfcount; i++) {
1718 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1719 		if (err)
1720 			break;
1721 	}
1722 	if (err) {
1723 		int j;
1724 
1725 		if (!delta)
1726 			pmc->sfcount[sfmode]--;
1727 		for (j=0; j<i; j++)
1728 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1729 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1730 #ifdef CONFIG_IP_MULTICAST
1731 		struct ip_sf_list *psf;
1732 		in_dev = pmc->interface;
1733 #endif
1734 
1735 		/* filter mode change */
1736 		if (pmc->sfcount[MCAST_EXCLUDE])
1737 			pmc->sfmode = MCAST_EXCLUDE;
1738 		else if (pmc->sfcount[MCAST_INCLUDE])
1739 			pmc->sfmode = MCAST_INCLUDE;
1740 #ifdef CONFIG_IP_MULTICAST
1741 		/* else no filters; keep old mode for reports */
1742 
1743 		pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1744 			IGMP_Unsolicited_Report_Count;
1745 		in_dev->mr_ifc_count = pmc->crcount;
1746 		for (psf=pmc->sources; psf; psf = psf->sf_next)
1747 			psf->sf_crcount = 0;
1748 		igmp_ifc_event(in_dev);
1749 	} else if (sf_setstate(pmc)) {
1750 		igmp_ifc_event(in_dev);
1751 #endif
1752 	}
1753 	spin_unlock_bh(&pmc->lock);
1754 	return err;
1755 }
1756 
1757 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1758 {
1759 	struct ip_sf_list *psf, *nextpsf;
1760 
1761 	for (psf=pmc->tomb; psf; psf=nextpsf) {
1762 		nextpsf = psf->sf_next;
1763 		kfree(psf);
1764 	}
1765 	pmc->tomb = NULL;
1766 	for (psf=pmc->sources; psf; psf=nextpsf) {
1767 		nextpsf = psf->sf_next;
1768 		kfree(psf);
1769 	}
1770 	pmc->sources = NULL;
1771 	pmc->sfmode = MCAST_EXCLUDE;
1772 	pmc->sfcount[MCAST_INCLUDE] = 0;
1773 	pmc->sfcount[MCAST_EXCLUDE] = 1;
1774 }
1775 
1776 
1777 /*
1778  * Join a multicast group
1779  */
1780 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1781 {
1782 	int err;
1783 	__be32 addr = imr->imr_multiaddr.s_addr;
1784 	struct ip_mc_socklist *iml = NULL, *i;
1785 	struct in_device *in_dev;
1786 	struct inet_sock *inet = inet_sk(sk);
1787 	struct net *net = sock_net(sk);
1788 	int ifindex;
1789 	int count = 0;
1790 
1791 	if (!ipv4_is_multicast(addr))
1792 		return -EINVAL;
1793 
1794 	rtnl_lock();
1795 
1796 	in_dev = ip_mc_find_dev(net, imr);
1797 
1798 	if (!in_dev) {
1799 		iml = NULL;
1800 		err = -ENODEV;
1801 		goto done;
1802 	}
1803 
1804 	err = -EADDRINUSE;
1805 	ifindex = imr->imr_ifindex;
1806 	for_each_pmc_rtnl(inet, i) {
1807 		if (i->multi.imr_multiaddr.s_addr == addr &&
1808 		    i->multi.imr_ifindex == ifindex)
1809 			goto done;
1810 		count++;
1811 	}
1812 	err = -ENOBUFS;
1813 	if (count >= sysctl_igmp_max_memberships)
1814 		goto done;
1815 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1816 	if (iml == NULL)
1817 		goto done;
1818 
1819 	memcpy(&iml->multi, imr, sizeof(*imr));
1820 	iml->next_rcu = inet->mc_list;
1821 	iml->sflist = NULL;
1822 	iml->sfmode = MCAST_EXCLUDE;
1823 	rcu_assign_pointer(inet->mc_list, iml);
1824 	ip_mc_inc_group(in_dev, addr);
1825 	err = 0;
1826 done:
1827 	rtnl_unlock();
1828 	return err;
1829 }
1830 EXPORT_SYMBOL(ip_mc_join_group);
1831 
1832 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1833 			   struct in_device *in_dev)
1834 {
1835 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1836 	int err;
1837 
1838 	if (psf == NULL) {
1839 		/* any-source empty exclude case */
1840 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1841 			iml->sfmode, 0, NULL, 0);
1842 	}
1843 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1844 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1845 	RCU_INIT_POINTER(iml->sflist, NULL);
1846 	/* decrease mem now to avoid the memleak warning */
1847 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1848 	kfree_rcu(psf, rcu);
1849 	return err;
1850 }
1851 
1852 /*
1853  *	Ask a socket to leave a group.
1854  */
1855 
1856 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1857 {
1858 	struct inet_sock *inet = inet_sk(sk);
1859 	struct ip_mc_socklist *iml;
1860 	struct ip_mc_socklist __rcu **imlp;
1861 	struct in_device *in_dev;
1862 	struct net *net = sock_net(sk);
1863 	__be32 group = imr->imr_multiaddr.s_addr;
1864 	u32 ifindex;
1865 	int ret = -EADDRNOTAVAIL;
1866 
1867 	rtnl_lock();
1868 	in_dev = ip_mc_find_dev(net, imr);
1869 	ifindex = imr->imr_ifindex;
1870 	for (imlp = &inet->mc_list;
1871 	     (iml = rtnl_dereference(*imlp)) != NULL;
1872 	     imlp = &iml->next_rcu) {
1873 		if (iml->multi.imr_multiaddr.s_addr != group)
1874 			continue;
1875 		if (ifindex) {
1876 			if (iml->multi.imr_ifindex != ifindex)
1877 				continue;
1878 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1879 				iml->multi.imr_address.s_addr)
1880 			continue;
1881 
1882 		(void) ip_mc_leave_src(sk, iml, in_dev);
1883 
1884 		*imlp = iml->next_rcu;
1885 
1886 		if (in_dev)
1887 			ip_mc_dec_group(in_dev, group);
1888 		rtnl_unlock();
1889 		/* decrease mem now to avoid the memleak warning */
1890 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1891 		kfree_rcu(iml, rcu);
1892 		return 0;
1893 	}
1894 	if (!in_dev)
1895 		ret = -ENODEV;
1896 	rtnl_unlock();
1897 	return ret;
1898 }
1899 
1900 int ip_mc_source(int add, int omode, struct sock *sk, struct
1901 	ip_mreq_source *mreqs, int ifindex)
1902 {
1903 	int err;
1904 	struct ip_mreqn imr;
1905 	__be32 addr = mreqs->imr_multiaddr;
1906 	struct ip_mc_socklist *pmc;
1907 	struct in_device *in_dev = NULL;
1908 	struct inet_sock *inet = inet_sk(sk);
1909 	struct ip_sf_socklist *psl;
1910 	struct net *net = sock_net(sk);
1911 	int leavegroup = 0;
1912 	int i, j, rv;
1913 
1914 	if (!ipv4_is_multicast(addr))
1915 		return -EINVAL;
1916 
1917 	rtnl_lock();
1918 
1919 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1920 	imr.imr_address.s_addr = mreqs->imr_interface;
1921 	imr.imr_ifindex = ifindex;
1922 	in_dev = ip_mc_find_dev(net, &imr);
1923 
1924 	if (!in_dev) {
1925 		err = -ENODEV;
1926 		goto done;
1927 	}
1928 	err = -EADDRNOTAVAIL;
1929 
1930 	for_each_pmc_rtnl(inet, pmc) {
1931 		if ((pmc->multi.imr_multiaddr.s_addr ==
1932 		     imr.imr_multiaddr.s_addr) &&
1933 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
1934 			break;
1935 	}
1936 	if (!pmc) {		/* must have a prior join */
1937 		err = -EINVAL;
1938 		goto done;
1939 	}
1940 	/* if a source filter was set, must be the same mode as before */
1941 	if (pmc->sflist) {
1942 		if (pmc->sfmode != omode) {
1943 			err = -EINVAL;
1944 			goto done;
1945 		}
1946 	} else if (pmc->sfmode != omode) {
1947 		/* allow mode switches for empty-set filters */
1948 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1949 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1950 			NULL, 0);
1951 		pmc->sfmode = omode;
1952 	}
1953 
1954 	psl = rtnl_dereference(pmc->sflist);
1955 	if (!add) {
1956 		if (!psl)
1957 			goto done;	/* err = -EADDRNOTAVAIL */
1958 		rv = !0;
1959 		for (i=0; i<psl->sl_count; i++) {
1960 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1961 				sizeof(__be32));
1962 			if (rv == 0)
1963 				break;
1964 		}
1965 		if (rv)		/* source not found */
1966 			goto done;	/* err = -EADDRNOTAVAIL */
1967 
1968 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
1969 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1970 			leavegroup = 1;
1971 			goto done;
1972 		}
1973 
1974 		/* update the interface filter */
1975 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1976 			&mreqs->imr_sourceaddr, 1);
1977 
1978 		for (j=i+1; j<psl->sl_count; j++)
1979 			psl->sl_addr[j-1] = psl->sl_addr[j];
1980 		psl->sl_count--;
1981 		err = 0;
1982 		goto done;
1983 	}
1984 	/* else, add a new source to the filter */
1985 
1986 	if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1987 		err = -ENOBUFS;
1988 		goto done;
1989 	}
1990 	if (!psl || psl->sl_count == psl->sl_max) {
1991 		struct ip_sf_socklist *newpsl;
1992 		int count = IP_SFBLOCK;
1993 
1994 		if (psl)
1995 			count += psl->sl_max;
1996 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
1997 		if (!newpsl) {
1998 			err = -ENOBUFS;
1999 			goto done;
2000 		}
2001 		newpsl->sl_max = count;
2002 		newpsl->sl_count = count - IP_SFBLOCK;
2003 		if (psl) {
2004 			for (i=0; i<psl->sl_count; i++)
2005 				newpsl->sl_addr[i] = psl->sl_addr[i];
2006 			/* decrease mem now to avoid the memleak warning */
2007 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2008 			kfree_rcu(psl, rcu);
2009 		}
2010 		rcu_assign_pointer(pmc->sflist, newpsl);
2011 		psl = newpsl;
2012 	}
2013 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2014 	for (i=0; i<psl->sl_count; i++) {
2015 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2016 			sizeof(__be32));
2017 		if (rv == 0)
2018 			break;
2019 	}
2020 	if (rv == 0)		/* address already there is an error */
2021 		goto done;
2022 	for (j=psl->sl_count-1; j>=i; j--)
2023 		psl->sl_addr[j+1] = psl->sl_addr[j];
2024 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2025 	psl->sl_count++;
2026 	err = 0;
2027 	/* update the interface list */
2028 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2029 		&mreqs->imr_sourceaddr, 1);
2030 done:
2031 	rtnl_unlock();
2032 	if (leavegroup)
2033 		return ip_mc_leave_group(sk, &imr);
2034 	return err;
2035 }
2036 
2037 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2038 {
2039 	int err = 0;
2040 	struct ip_mreqn	imr;
2041 	__be32 addr = msf->imsf_multiaddr;
2042 	struct ip_mc_socklist *pmc;
2043 	struct in_device *in_dev;
2044 	struct inet_sock *inet = inet_sk(sk);
2045 	struct ip_sf_socklist *newpsl, *psl;
2046 	struct net *net = sock_net(sk);
2047 	int leavegroup = 0;
2048 
2049 	if (!ipv4_is_multicast(addr))
2050 		return -EINVAL;
2051 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2052 	    msf->imsf_fmode != MCAST_EXCLUDE)
2053 		return -EINVAL;
2054 
2055 	rtnl_lock();
2056 
2057 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2058 	imr.imr_address.s_addr = msf->imsf_interface;
2059 	imr.imr_ifindex = ifindex;
2060 	in_dev = ip_mc_find_dev(net, &imr);
2061 
2062 	if (!in_dev) {
2063 		err = -ENODEV;
2064 		goto done;
2065 	}
2066 
2067 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2068 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2069 		leavegroup = 1;
2070 		goto done;
2071 	}
2072 
2073 	for_each_pmc_rtnl(inet, pmc) {
2074 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2075 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2076 			break;
2077 	}
2078 	if (!pmc) {		/* must have a prior join */
2079 		err = -EINVAL;
2080 		goto done;
2081 	}
2082 	if (msf->imsf_numsrc) {
2083 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2084 							   GFP_KERNEL);
2085 		if (!newpsl) {
2086 			err = -ENOBUFS;
2087 			goto done;
2088 		}
2089 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2090 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2091 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2092 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2093 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2094 		if (err) {
2095 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2096 			goto done;
2097 		}
2098 	} else {
2099 		newpsl = NULL;
2100 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2101 				     msf->imsf_fmode, 0, NULL, 0);
2102 	}
2103 	psl = rtnl_dereference(pmc->sflist);
2104 	if (psl) {
2105 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2106 			psl->sl_count, psl->sl_addr, 0);
2107 		/* decrease mem now to avoid the memleak warning */
2108 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2109 		kfree_rcu(psl, rcu);
2110 	} else
2111 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2112 			0, NULL, 0);
2113 	rcu_assign_pointer(pmc->sflist, newpsl);
2114 	pmc->sfmode = msf->imsf_fmode;
2115 	err = 0;
2116 done:
2117 	rtnl_unlock();
2118 	if (leavegroup)
2119 		err = ip_mc_leave_group(sk, &imr);
2120 	return err;
2121 }
2122 
2123 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2124 	struct ip_msfilter __user *optval, int __user *optlen)
2125 {
2126 	int err, len, count, copycount;
2127 	struct ip_mreqn	imr;
2128 	__be32 addr = msf->imsf_multiaddr;
2129 	struct ip_mc_socklist *pmc;
2130 	struct in_device *in_dev;
2131 	struct inet_sock *inet = inet_sk(sk);
2132 	struct ip_sf_socklist *psl;
2133 	struct net *net = sock_net(sk);
2134 
2135 	if (!ipv4_is_multicast(addr))
2136 		return -EINVAL;
2137 
2138 	rtnl_lock();
2139 
2140 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2141 	imr.imr_address.s_addr = msf->imsf_interface;
2142 	imr.imr_ifindex = 0;
2143 	in_dev = ip_mc_find_dev(net, &imr);
2144 
2145 	if (!in_dev) {
2146 		err = -ENODEV;
2147 		goto done;
2148 	}
2149 	err = -EADDRNOTAVAIL;
2150 
2151 	for_each_pmc_rtnl(inet, pmc) {
2152 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2153 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2154 			break;
2155 	}
2156 	if (!pmc)		/* must have a prior join */
2157 		goto done;
2158 	msf->imsf_fmode = pmc->sfmode;
2159 	psl = rtnl_dereference(pmc->sflist);
2160 	rtnl_unlock();
2161 	if (!psl) {
2162 		len = 0;
2163 		count = 0;
2164 	} else {
2165 		count = psl->sl_count;
2166 	}
2167 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2168 	len = copycount * sizeof(psl->sl_addr[0]);
2169 	msf->imsf_numsrc = count;
2170 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2171 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2172 		return -EFAULT;
2173 	}
2174 	if (len &&
2175 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2176 		return -EFAULT;
2177 	return 0;
2178 done:
2179 	rtnl_unlock();
2180 	return err;
2181 }
2182 
2183 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2184 	struct group_filter __user *optval, int __user *optlen)
2185 {
2186 	int err, i, count, copycount;
2187 	struct sockaddr_in *psin;
2188 	__be32 addr;
2189 	struct ip_mc_socklist *pmc;
2190 	struct inet_sock *inet = inet_sk(sk);
2191 	struct ip_sf_socklist *psl;
2192 
2193 	psin = (struct sockaddr_in *)&gsf->gf_group;
2194 	if (psin->sin_family != AF_INET)
2195 		return -EINVAL;
2196 	addr = psin->sin_addr.s_addr;
2197 	if (!ipv4_is_multicast(addr))
2198 		return -EINVAL;
2199 
2200 	rtnl_lock();
2201 
2202 	err = -EADDRNOTAVAIL;
2203 
2204 	for_each_pmc_rtnl(inet, pmc) {
2205 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2206 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2207 			break;
2208 	}
2209 	if (!pmc)		/* must have a prior join */
2210 		goto done;
2211 	gsf->gf_fmode = pmc->sfmode;
2212 	psl = rtnl_dereference(pmc->sflist);
2213 	rtnl_unlock();
2214 	count = psl ? psl->sl_count : 0;
2215 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2216 	gsf->gf_numsrc = count;
2217 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2218 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2219 		return -EFAULT;
2220 	}
2221 	for (i=0; i<copycount; i++) {
2222 		struct sockaddr_storage ss;
2223 
2224 		psin = (struct sockaddr_in *)&ss;
2225 		memset(&ss, 0, sizeof(ss));
2226 		psin->sin_family = AF_INET;
2227 		psin->sin_addr.s_addr = psl->sl_addr[i];
2228 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2229 			return -EFAULT;
2230 	}
2231 	return 0;
2232 done:
2233 	rtnl_unlock();
2234 	return err;
2235 }
2236 
2237 /*
2238  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2239  */
2240 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2241 {
2242 	struct inet_sock *inet = inet_sk(sk);
2243 	struct ip_mc_socklist *pmc;
2244 	struct ip_sf_socklist *psl;
2245 	int i;
2246 	int ret;
2247 
2248 	ret = 1;
2249 	if (!ipv4_is_multicast(loc_addr))
2250 		goto out;
2251 
2252 	rcu_read_lock();
2253 	for_each_pmc_rcu(inet, pmc) {
2254 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2255 		    pmc->multi.imr_ifindex == dif)
2256 			break;
2257 	}
2258 	ret = inet->mc_all;
2259 	if (!pmc)
2260 		goto unlock;
2261 	psl = rcu_dereference(pmc->sflist);
2262 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2263 	if (!psl)
2264 		goto unlock;
2265 
2266 	for (i=0; i<psl->sl_count; i++) {
2267 		if (psl->sl_addr[i] == rmt_addr)
2268 			break;
2269 	}
2270 	ret = 0;
2271 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2272 		goto unlock;
2273 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2274 		goto unlock;
2275 	ret = 1;
2276 unlock:
2277 	rcu_read_unlock();
2278 out:
2279 	return ret;
2280 }
2281 
2282 /*
2283  *	A socket is closing.
2284  */
2285 
2286 void ip_mc_drop_socket(struct sock *sk)
2287 {
2288 	struct inet_sock *inet = inet_sk(sk);
2289 	struct ip_mc_socklist *iml;
2290 	struct net *net = sock_net(sk);
2291 
2292 	if (inet->mc_list == NULL)
2293 		return;
2294 
2295 	rtnl_lock();
2296 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2297 		struct in_device *in_dev;
2298 
2299 		inet->mc_list = iml->next_rcu;
2300 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2301 		(void) ip_mc_leave_src(sk, iml, in_dev);
2302 		if (in_dev != NULL)
2303 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2304 		/* decrease mem now to avoid the memleak warning */
2305 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2306 		kfree_rcu(iml, rcu);
2307 	}
2308 	rtnl_unlock();
2309 }
2310 
2311 /* called with rcu_read_lock() */
2312 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2313 {
2314 	struct ip_mc_list *im;
2315 	struct ip_sf_list *psf;
2316 	int rv = 0;
2317 
2318 	for_each_pmc_rcu(in_dev, im) {
2319 		if (im->multiaddr == mc_addr)
2320 			break;
2321 	}
2322 	if (im && proto == IPPROTO_IGMP) {
2323 		rv = 1;
2324 	} else if (im) {
2325 		if (src_addr) {
2326 			for (psf=im->sources; psf; psf=psf->sf_next) {
2327 				if (psf->sf_inaddr == src_addr)
2328 					break;
2329 			}
2330 			if (psf)
2331 				rv = psf->sf_count[MCAST_INCLUDE] ||
2332 					psf->sf_count[MCAST_EXCLUDE] !=
2333 					im->sfcount[MCAST_EXCLUDE];
2334 			else
2335 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2336 		} else
2337 			rv = 1; /* unspecified source; tentatively allow */
2338 	}
2339 	return rv;
2340 }
2341 
2342 #if defined(CONFIG_PROC_FS)
2343 struct igmp_mc_iter_state {
2344 	struct seq_net_private p;
2345 	struct net_device *dev;
2346 	struct in_device *in_dev;
2347 };
2348 
2349 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2350 
2351 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2352 {
2353 	struct net *net = seq_file_net(seq);
2354 	struct ip_mc_list *im = NULL;
2355 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2356 
2357 	state->in_dev = NULL;
2358 	for_each_netdev_rcu(net, state->dev) {
2359 		struct in_device *in_dev;
2360 
2361 		in_dev = __in_dev_get_rcu(state->dev);
2362 		if (!in_dev)
2363 			continue;
2364 		im = rcu_dereference(in_dev->mc_list);
2365 		if (im) {
2366 			state->in_dev = in_dev;
2367 			break;
2368 		}
2369 	}
2370 	return im;
2371 }
2372 
2373 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2374 {
2375 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2376 
2377 	im = rcu_dereference(im->next_rcu);
2378 	while (!im) {
2379 		state->dev = next_net_device_rcu(state->dev);
2380 		if (!state->dev) {
2381 			state->in_dev = NULL;
2382 			break;
2383 		}
2384 		state->in_dev = __in_dev_get_rcu(state->dev);
2385 		if (!state->in_dev)
2386 			continue;
2387 		im = rcu_dereference(state->in_dev->mc_list);
2388 	}
2389 	return im;
2390 }
2391 
2392 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2393 {
2394 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2395 	if (im)
2396 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2397 			--pos;
2398 	return pos ? NULL : im;
2399 }
2400 
2401 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2402 	__acquires(rcu)
2403 {
2404 	rcu_read_lock();
2405 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2406 }
2407 
2408 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2409 {
2410 	struct ip_mc_list *im;
2411 	if (v == SEQ_START_TOKEN)
2412 		im = igmp_mc_get_first(seq);
2413 	else
2414 		im = igmp_mc_get_next(seq, v);
2415 	++*pos;
2416 	return im;
2417 }
2418 
2419 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2420 	__releases(rcu)
2421 {
2422 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2423 
2424 	state->in_dev = NULL;
2425 	state->dev = NULL;
2426 	rcu_read_unlock();
2427 }
2428 
2429 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2430 {
2431 	if (v == SEQ_START_TOKEN)
2432 		seq_puts(seq,
2433 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2434 	else {
2435 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2436 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2437 		char   *querier;
2438 #ifdef CONFIG_IP_MULTICAST
2439 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2440 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2441 			  "V3";
2442 #else
2443 		querier = "NONE";
2444 #endif
2445 
2446 		if (rcu_dereference(state->in_dev->mc_list) == im) {
2447 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2448 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2449 		}
2450 
2451 		seq_printf(seq,
2452 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2453 			   im->multiaddr, im->users,
2454 			   im->tm_running, im->tm_running ?
2455 			   jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2456 			   im->reporter);
2457 	}
2458 	return 0;
2459 }
2460 
2461 static const struct seq_operations igmp_mc_seq_ops = {
2462 	.start	=	igmp_mc_seq_start,
2463 	.next	=	igmp_mc_seq_next,
2464 	.stop	=	igmp_mc_seq_stop,
2465 	.show	=	igmp_mc_seq_show,
2466 };
2467 
2468 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2469 {
2470 	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2471 			sizeof(struct igmp_mc_iter_state));
2472 }
2473 
2474 static const struct file_operations igmp_mc_seq_fops = {
2475 	.owner		=	THIS_MODULE,
2476 	.open		=	igmp_mc_seq_open,
2477 	.read		=	seq_read,
2478 	.llseek		=	seq_lseek,
2479 	.release	=	seq_release_net,
2480 };
2481 
2482 struct igmp_mcf_iter_state {
2483 	struct seq_net_private p;
2484 	struct net_device *dev;
2485 	struct in_device *idev;
2486 	struct ip_mc_list *im;
2487 };
2488 
2489 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2490 
2491 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2492 {
2493 	struct net *net = seq_file_net(seq);
2494 	struct ip_sf_list *psf = NULL;
2495 	struct ip_mc_list *im = NULL;
2496 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2497 
2498 	state->idev = NULL;
2499 	state->im = NULL;
2500 	for_each_netdev_rcu(net, state->dev) {
2501 		struct in_device *idev;
2502 		idev = __in_dev_get_rcu(state->dev);
2503 		if (unlikely(idev == NULL))
2504 			continue;
2505 		im = rcu_dereference(idev->mc_list);
2506 		if (likely(im != NULL)) {
2507 			spin_lock_bh(&im->lock);
2508 			psf = im->sources;
2509 			if (likely(psf != NULL)) {
2510 				state->im = im;
2511 				state->idev = idev;
2512 				break;
2513 			}
2514 			spin_unlock_bh(&im->lock);
2515 		}
2516 	}
2517 	return psf;
2518 }
2519 
2520 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2521 {
2522 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2523 
2524 	psf = psf->sf_next;
2525 	while (!psf) {
2526 		spin_unlock_bh(&state->im->lock);
2527 		state->im = state->im->next;
2528 		while (!state->im) {
2529 			state->dev = next_net_device_rcu(state->dev);
2530 			if (!state->dev) {
2531 				state->idev = NULL;
2532 				goto out;
2533 			}
2534 			state->idev = __in_dev_get_rcu(state->dev);
2535 			if (!state->idev)
2536 				continue;
2537 			state->im = rcu_dereference(state->idev->mc_list);
2538 		}
2539 		if (!state->im)
2540 			break;
2541 		spin_lock_bh(&state->im->lock);
2542 		psf = state->im->sources;
2543 	}
2544 out:
2545 	return psf;
2546 }
2547 
2548 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2549 {
2550 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2551 	if (psf)
2552 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2553 			--pos;
2554 	return pos ? NULL : psf;
2555 }
2556 
2557 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2558 	__acquires(rcu)
2559 {
2560 	rcu_read_lock();
2561 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2562 }
2563 
2564 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2565 {
2566 	struct ip_sf_list *psf;
2567 	if (v == SEQ_START_TOKEN)
2568 		psf = igmp_mcf_get_first(seq);
2569 	else
2570 		psf = igmp_mcf_get_next(seq, v);
2571 	++*pos;
2572 	return psf;
2573 }
2574 
2575 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2576 	__releases(rcu)
2577 {
2578 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2579 	if (likely(state->im != NULL)) {
2580 		spin_unlock_bh(&state->im->lock);
2581 		state->im = NULL;
2582 	}
2583 	state->idev = NULL;
2584 	state->dev = NULL;
2585 	rcu_read_unlock();
2586 }
2587 
2588 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2589 {
2590 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2591 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2592 
2593 	if (v == SEQ_START_TOKEN) {
2594 		seq_printf(seq,
2595 			   "%3s %6s "
2596 			   "%10s %10s %6s %6s\n", "Idx",
2597 			   "Device", "MCA",
2598 			   "SRC", "INC", "EXC");
2599 	} else {
2600 		seq_printf(seq,
2601 			   "%3d %6.6s 0x%08x "
2602 			   "0x%08x %6lu %6lu\n",
2603 			   state->dev->ifindex, state->dev->name,
2604 			   ntohl(state->im->multiaddr),
2605 			   ntohl(psf->sf_inaddr),
2606 			   psf->sf_count[MCAST_INCLUDE],
2607 			   psf->sf_count[MCAST_EXCLUDE]);
2608 	}
2609 	return 0;
2610 }
2611 
2612 static const struct seq_operations igmp_mcf_seq_ops = {
2613 	.start	=	igmp_mcf_seq_start,
2614 	.next	=	igmp_mcf_seq_next,
2615 	.stop	=	igmp_mcf_seq_stop,
2616 	.show	=	igmp_mcf_seq_show,
2617 };
2618 
2619 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2620 {
2621 	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2622 			sizeof(struct igmp_mcf_iter_state));
2623 }
2624 
2625 static const struct file_operations igmp_mcf_seq_fops = {
2626 	.owner		=	THIS_MODULE,
2627 	.open		=	igmp_mcf_seq_open,
2628 	.read		=	seq_read,
2629 	.llseek		=	seq_lseek,
2630 	.release	=	seq_release_net,
2631 };
2632 
2633 static int __net_init igmp_net_init(struct net *net)
2634 {
2635 	struct proc_dir_entry *pde;
2636 
2637 	pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2638 	if (!pde)
2639 		goto out_igmp;
2640 	pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2641 	if (!pde)
2642 		goto out_mcfilter;
2643 	return 0;
2644 
2645 out_mcfilter:
2646 	proc_net_remove(net, "igmp");
2647 out_igmp:
2648 	return -ENOMEM;
2649 }
2650 
2651 static void __net_exit igmp_net_exit(struct net *net)
2652 {
2653 	proc_net_remove(net, "mcfilter");
2654 	proc_net_remove(net, "igmp");
2655 }
2656 
2657 static struct pernet_operations igmp_net_ops = {
2658 	.init = igmp_net_init,
2659 	.exit = igmp_net_exit,
2660 };
2661 
2662 int __init igmp_mc_proc_init(void)
2663 {
2664 	return register_pernet_subsys(&igmp_net_ops);
2665 }
2666 #endif
2667