• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Linux NET3:	Internet Group Management Protocol  [IGMP]
4  *
5  *	This code implements the IGMP protocol as defined in RFC1112. There has
6  *	been a further revision of this protocol since which is now supported.
7  *
8  *	If you have trouble with this module be careful what gcc you have used,
9  *	the older version didn't come out right using gcc 2.5.8, the newer one
10  *	seems to fall out with gcc 2.6.2.
11  *
12  *	Authors:
13  *		Alan Cox <alan@lxorguk.ukuu.org.uk>
14  *
15  *	Fixes:
16  *
17  *		Alan Cox	:	Added lots of __inline__ to optimise
18  *					the memory usage of all the tiny little
19  *					functions.
20  *		Alan Cox	:	Dumped the header building experiment.
21  *		Alan Cox	:	Minor tweaks ready for multicast routing
22  *					and extended IGMP protocol.
23  *		Alan Cox	:	Removed a load of inline directives. Gcc 2.5.8
24  *					writes utterly bogus code otherwise (sigh)
25  *					fixed IGMP loopback to behave in the manner
26  *					desired by mrouted, fixed the fact it has been
27  *					broken since 1.3.6 and cleaned up a few minor
28  *					points.
29  *
30  *		Chih-Jen Chang	:	Tried to revise IGMP to Version 2
31  *		Tsu-Sheng Tsao		E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
32  *					The enhancements are mainly based on Steve Deering's
33  * 					ipmulti-3.5 source code.
34  *		Chih-Jen Chang	:	Added the igmp_get_mrouter_info and
35  *		Tsu-Sheng Tsao		igmp_set_mrouter_info to keep track of
36  *					the mrouted version on that device.
37  *		Chih-Jen Chang	:	Added the max_resp_time parameter to
38  *		Tsu-Sheng Tsao		igmp_heard_query(). Using this parameter
39  *					to identify the multicast router version
40  *					and do what the IGMP version 2 specified.
41  *		Chih-Jen Chang	:	Added a timer to revert to IGMP V2 router
42  *		Tsu-Sheng Tsao		if the specified time expired.
43  *		Alan Cox	:	Stop IGMP from 0.0.0.0 being accepted.
44  *		Alan Cox	:	Use GFP_ATOMIC in the right places.
45  *		Christian Daudt :	igmp timer wasn't set for local group
46  *					memberships but was being deleted,
47  *					which caused a "del_timer() called
48  *					from %p with timer not initialized\n"
49  *					message (960131).
50  *		Christian Daudt :	removed del_timer from
51  *					igmp_timer_expire function (960205).
52  *             Christian Daudt :       igmp_heard_report now only calls
53  *                                     igmp_timer_expire if tm->running is
54  *                                     true (960216).
55  *		Malcolm Beattie :	ttl comparison wrong in igmp_rcv made
56  *					igmp_heard_query never trigger. Expiry
57  *					miscalculation fixed in igmp_heard_query
58  *					and random() made to return unsigned to
59  *					prevent negative expiry times.
60  *		Alexey Kuznetsov:	Wrong group leaving behaviour, backport
61  *					fix from pending 2.1.x patches.
62  *		Alan Cox:		Forget to enable FDDI support earlier.
63  *		Alexey Kuznetsov:	Fixed leaving groups on device down.
64  *		Alexey Kuznetsov:	Accordance to igmp-v2-06 draft.
65  *		David L Stevens:	IGMPv3 support, with help from
66  *					Vinay Kulkarni
67  */
68 
69 #include <linux/module.h>
70 #include <linux/slab.h>
71 #include <linux/uaccess.h>
72 #include <linux/types.h>
73 #include <linux/kernel.h>
74 #include <linux/jiffies.h>
75 #include <linux/string.h>
76 #include <linux/socket.h>
77 #include <linux/sockios.h>
78 #include <linux/in.h>
79 #include <linux/inet.h>
80 #include <linux/netdevice.h>
81 #include <linux/skbuff.h>
82 #include <linux/inetdevice.h>
83 #include <linux/igmp.h>
84 #include <linux/if_arp.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/times.h>
87 #include <linux/pkt_sched.h>
88 #include <linux/byteorder/generic.h>
89 
90 #include <net/net_namespace.h>
91 #include <net/arp.h>
92 #include <net/ip.h>
93 #include <net/protocol.h>
94 #include <net/route.h>
95 #include <net/sock.h>
96 #include <net/checksum.h>
97 #include <net/inet_common.h>
98 #include <linux/netfilter_ipv4.h>
99 #ifdef CONFIG_IP_MROUTE
100 #include <linux/mroute.h>
101 #endif
102 #ifdef CONFIG_PROC_FS
103 #include <linux/proc_fs.h>
104 #include <linux/seq_file.h>
105 #endif
106 
107 #ifdef CONFIG_IP_MULTICAST
108 /* Parameter names and values are taken from igmp-v2-06 draft */
109 
110 #define IGMP_QUERY_INTERVAL			(125*HZ)
111 #define IGMP_QUERY_RESPONSE_INTERVAL		(10*HZ)
112 
113 #define IGMP_INITIAL_REPORT_DELAY		(1)
114 
115 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
116  * IGMP specs require to report membership immediately after
117  * joining a group, but we delay the first report by a
118  * small interval. It seems more natural and still does not
119  * contradict to specs provided this delay is small enough.
120  */
121 
122 #define IGMP_V1_SEEN(in_dev) \
123 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
124 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
125 	 ((in_dev)->mr_v1_seen && \
126 	  time_before(jiffies, (in_dev)->mr_v1_seen)))
127 #define IGMP_V2_SEEN(in_dev) \
128 	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
129 	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
130 	 ((in_dev)->mr_v2_seen && \
131 	  time_before(jiffies, (in_dev)->mr_v2_seen)))
132 
unsolicited_report_interval(struct in_device * in_dev)133 static int unsolicited_report_interval(struct in_device *in_dev)
134 {
135 	int interval_ms, interval_jiffies;
136 
137 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
138 		interval_ms = IN_DEV_CONF_GET(
139 			in_dev,
140 			IGMPV2_UNSOLICITED_REPORT_INTERVAL);
141 	else /* v3 */
142 		interval_ms = IN_DEV_CONF_GET(
143 			in_dev,
144 			IGMPV3_UNSOLICITED_REPORT_INTERVAL);
145 
146 	interval_jiffies = msecs_to_jiffies(interval_ms);
147 
148 	/* _timer functions can't handle a delay of 0 jiffies so ensure
149 	 *  we always return a positive value.
150 	 */
151 	if (interval_jiffies <= 0)
152 		interval_jiffies = 1;
153 	return interval_jiffies;
154 }
155 
156 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im,
157 			      gfp_t gfp);
158 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im);
159 static void igmpv3_clear_delrec(struct in_device *in_dev);
160 static int sf_setstate(struct ip_mc_list *pmc);
161 static void sf_markstate(struct ip_mc_list *pmc);
162 #endif
163 static void ip_mc_clear_src(struct ip_mc_list *pmc);
164 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
165 			 int sfcount, __be32 *psfsrc, int delta);
166 
ip_ma_put(struct ip_mc_list * im)167 static void ip_ma_put(struct ip_mc_list *im)
168 {
169 	if (refcount_dec_and_test(&im->refcnt)) {
170 		in_dev_put(im->interface);
171 		kfree_rcu(im, rcu);
172 	}
173 }
174 
175 #define for_each_pmc_rcu(in_dev, pmc)				\
176 	for (pmc = rcu_dereference(in_dev->mc_list);		\
177 	     pmc != NULL;					\
178 	     pmc = rcu_dereference(pmc->next_rcu))
179 
180 #define for_each_pmc_rtnl(in_dev, pmc)				\
181 	for (pmc = rtnl_dereference(in_dev->mc_list);		\
182 	     pmc != NULL;					\
183 	     pmc = rtnl_dereference(pmc->next_rcu))
184 
ip_sf_list_clear_all(struct ip_sf_list * psf)185 static void ip_sf_list_clear_all(struct ip_sf_list *psf)
186 {
187 	struct ip_sf_list *next;
188 
189 	while (psf) {
190 		next = psf->sf_next;
191 		kfree(psf);
192 		psf = next;
193 	}
194 }
195 
196 #ifdef CONFIG_IP_MULTICAST
197 
198 /*
199  *	Timer management
200  */
201 
igmp_stop_timer(struct ip_mc_list * im)202 static void igmp_stop_timer(struct ip_mc_list *im)
203 {
204 	spin_lock_bh(&im->lock);
205 	if (del_timer(&im->timer))
206 		refcount_dec(&im->refcnt);
207 	im->tm_running = 0;
208 	im->reporter = 0;
209 	im->unsolicit_count = 0;
210 	spin_unlock_bh(&im->lock);
211 }
212 
213 /* It must be called with locked im->lock */
igmp_start_timer(struct ip_mc_list * im,int max_delay)214 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
215 {
216 	int tv = prandom_u32() % max_delay;
217 
218 	im->tm_running = 1;
219 	if (refcount_inc_not_zero(&im->refcnt)) {
220 		if (mod_timer(&im->timer, jiffies + tv + 2))
221 			ip_ma_put(im);
222 	}
223 }
224 
igmp_gq_start_timer(struct in_device * in_dev)225 static void igmp_gq_start_timer(struct in_device *in_dev)
226 {
227 	int tv = prandom_u32() % in_dev->mr_maxdelay;
228 	unsigned long exp = jiffies + tv + 2;
229 
230 	if (in_dev->mr_gq_running &&
231 	    time_after_eq(exp, (in_dev->mr_gq_timer).expires))
232 		return;
233 
234 	in_dev->mr_gq_running = 1;
235 	if (!mod_timer(&in_dev->mr_gq_timer, exp))
236 		in_dev_hold(in_dev);
237 }
238 
igmp_ifc_start_timer(struct in_device * in_dev,int delay)239 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
240 {
241 	int tv = prandom_u32() % delay;
242 
243 	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
244 		in_dev_hold(in_dev);
245 }
246 
igmp_mod_timer(struct ip_mc_list * im,int max_delay)247 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
248 {
249 	spin_lock_bh(&im->lock);
250 	im->unsolicit_count = 0;
251 	if (del_timer(&im->timer)) {
252 		if ((long)(im->timer.expires-jiffies) < max_delay) {
253 			add_timer(&im->timer);
254 			im->tm_running = 1;
255 			spin_unlock_bh(&im->lock);
256 			return;
257 		}
258 		refcount_dec(&im->refcnt);
259 	}
260 	igmp_start_timer(im, max_delay);
261 	spin_unlock_bh(&im->lock);
262 }
263 
264 
265 /*
266  *	Send an IGMP report.
267  */
268 
269 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
270 
271 
is_in(struct ip_mc_list * pmc,struct ip_sf_list * psf,int type,int gdeleted,int sdeleted)272 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
273 	int gdeleted, int sdeleted)
274 {
275 	switch (type) {
276 	case IGMPV3_MODE_IS_INCLUDE:
277 	case IGMPV3_MODE_IS_EXCLUDE:
278 		if (gdeleted || sdeleted)
279 			return 0;
280 		if (!(pmc->gsquery && !psf->sf_gsresp)) {
281 			if (pmc->sfmode == MCAST_INCLUDE)
282 				return 1;
283 			/* don't include if this source is excluded
284 			 * in all filters
285 			 */
286 			if (psf->sf_count[MCAST_INCLUDE])
287 				return type == IGMPV3_MODE_IS_INCLUDE;
288 			return pmc->sfcount[MCAST_EXCLUDE] ==
289 				psf->sf_count[MCAST_EXCLUDE];
290 		}
291 		return 0;
292 	case IGMPV3_CHANGE_TO_INCLUDE:
293 		if (gdeleted || sdeleted)
294 			return 0;
295 		return psf->sf_count[MCAST_INCLUDE] != 0;
296 	case IGMPV3_CHANGE_TO_EXCLUDE:
297 		if (gdeleted || sdeleted)
298 			return 0;
299 		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
300 		    psf->sf_count[MCAST_INCLUDE])
301 			return 0;
302 		return pmc->sfcount[MCAST_EXCLUDE] ==
303 			psf->sf_count[MCAST_EXCLUDE];
304 	case IGMPV3_ALLOW_NEW_SOURCES:
305 		if (gdeleted || !psf->sf_crcount)
306 			return 0;
307 		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
308 	case IGMPV3_BLOCK_OLD_SOURCES:
309 		if (pmc->sfmode == MCAST_INCLUDE)
310 			return gdeleted || (psf->sf_crcount && sdeleted);
311 		return psf->sf_crcount && !gdeleted && !sdeleted;
312 	}
313 	return 0;
314 }
315 
316 static int
igmp_scount(struct ip_mc_list * pmc,int type,int gdeleted,int sdeleted)317 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
318 {
319 	struct ip_sf_list *psf;
320 	int scount = 0;
321 
322 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
323 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
324 			continue;
325 		scount++;
326 	}
327 	return scount;
328 }
329 
330 /* source address selection per RFC 3376 section 4.2.13 */
igmpv3_get_srcaddr(struct net_device * dev,const struct flowi4 * fl4)331 static __be32 igmpv3_get_srcaddr(struct net_device *dev,
332 				 const struct flowi4 *fl4)
333 {
334 	struct in_device *in_dev = __in_dev_get_rcu(dev);
335 	const struct in_ifaddr *ifa;
336 
337 	if (!in_dev)
338 		return htonl(INADDR_ANY);
339 
340 	in_dev_for_each_ifa_rcu(ifa, in_dev) {
341 		if (fl4->saddr == ifa->ifa_local)
342 			return fl4->saddr;
343 	}
344 
345 	return htonl(INADDR_ANY);
346 }
347 
igmpv3_newpack(struct net_device * dev,unsigned int mtu)348 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
349 {
350 	struct sk_buff *skb;
351 	struct rtable *rt;
352 	struct iphdr *pip;
353 	struct igmpv3_report *pig;
354 	struct net *net = dev_net(dev);
355 	struct flowi4 fl4;
356 	int hlen = LL_RESERVED_SPACE(dev);
357 	int tlen = dev->needed_tailroom;
358 	unsigned int size = mtu;
359 
360 	while (1) {
361 		skb = alloc_skb(size + hlen + tlen,
362 				GFP_ATOMIC | __GFP_NOWARN);
363 		if (skb)
364 			break;
365 		size >>= 1;
366 		if (size < 256)
367 			return NULL;
368 	}
369 	skb->priority = TC_PRIO_CONTROL;
370 
371 	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
372 				   0, 0,
373 				   IPPROTO_IGMP, 0, dev->ifindex);
374 	if (IS_ERR(rt)) {
375 		kfree_skb(skb);
376 		return NULL;
377 	}
378 
379 	skb_dst_set(skb, &rt->dst);
380 	skb->dev = dev;
381 
382 	skb_reserve(skb, hlen);
383 	skb_tailroom_reserve(skb, mtu, tlen);
384 
385 	skb_reset_network_header(skb);
386 	pip = ip_hdr(skb);
387 	skb_put(skb, sizeof(struct iphdr) + 4);
388 
389 	pip->version  = 4;
390 	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
391 	pip->tos      = 0xc0;
392 	pip->frag_off = htons(IP_DF);
393 	pip->ttl      = 1;
394 	pip->daddr    = fl4.daddr;
395 
396 	rcu_read_lock();
397 	pip->saddr    = igmpv3_get_srcaddr(dev, &fl4);
398 	rcu_read_unlock();
399 
400 	pip->protocol = IPPROTO_IGMP;
401 	pip->tot_len  = 0;	/* filled in later */
402 	ip_select_ident(net, skb, NULL);
403 	((u8 *)&pip[1])[0] = IPOPT_RA;
404 	((u8 *)&pip[1])[1] = 4;
405 	((u8 *)&pip[1])[2] = 0;
406 	((u8 *)&pip[1])[3] = 0;
407 
408 	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
409 	skb_put(skb, sizeof(*pig));
410 	pig = igmpv3_report_hdr(skb);
411 	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
412 	pig->resv1 = 0;
413 	pig->csum = 0;
414 	pig->resv2 = 0;
415 	pig->ngrec = 0;
416 	return skb;
417 }
418 
igmpv3_sendpack(struct sk_buff * skb)419 static int igmpv3_sendpack(struct sk_buff *skb)
420 {
421 	struct igmphdr *pig = igmp_hdr(skb);
422 	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
423 
424 	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
425 
426 	return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
427 }
428 
grec_size(struct ip_mc_list * pmc,int type,int gdel,int sdel)429 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
430 {
431 	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
432 }
433 
add_grhead(struct sk_buff * skb,struct ip_mc_list * pmc,int type,struct igmpv3_grec ** ppgr,unsigned int mtu)434 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
435 	int type, struct igmpv3_grec **ppgr, unsigned int mtu)
436 {
437 	struct net_device *dev = pmc->interface->dev;
438 	struct igmpv3_report *pih;
439 	struct igmpv3_grec *pgr;
440 
441 	if (!skb) {
442 		skb = igmpv3_newpack(dev, mtu);
443 		if (!skb)
444 			return NULL;
445 	}
446 	pgr = skb_put(skb, sizeof(struct igmpv3_grec));
447 	pgr->grec_type = type;
448 	pgr->grec_auxwords = 0;
449 	pgr->grec_nsrcs = 0;
450 	pgr->grec_mca = pmc->multiaddr;
451 	pih = igmpv3_report_hdr(skb);
452 	pih->ngrec = htons(ntohs(pih->ngrec)+1);
453 	*ppgr = pgr;
454 	return skb;
455 }
456 
457 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
458 
add_grec(struct sk_buff * skb,struct ip_mc_list * pmc,int type,int gdeleted,int sdeleted)459 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
460 	int type, int gdeleted, int sdeleted)
461 {
462 	struct net_device *dev = pmc->interface->dev;
463 	struct net *net = dev_net(dev);
464 	struct igmpv3_report *pih;
465 	struct igmpv3_grec *pgr = NULL;
466 	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
467 	int scount, stotal, first, isquery, truncate;
468 	unsigned int mtu;
469 
470 	if (pmc->multiaddr == IGMP_ALL_HOSTS)
471 		return skb;
472 	if (ipv4_is_local_multicast(pmc->multiaddr) &&
473 	    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
474 		return skb;
475 
476 	mtu = READ_ONCE(dev->mtu);
477 	if (mtu < IPV4_MIN_MTU)
478 		return skb;
479 
480 	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
481 		  type == IGMPV3_MODE_IS_EXCLUDE;
482 	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
483 		    type == IGMPV3_CHANGE_TO_EXCLUDE;
484 
485 	stotal = scount = 0;
486 
487 	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
488 
489 	if (!*psf_list)
490 		goto empty_source;
491 
492 	pih = skb ? igmpv3_report_hdr(skb) : NULL;
493 
494 	/* EX and TO_EX get a fresh packet, if needed */
495 	if (truncate) {
496 		if (pih && pih->ngrec &&
497 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
498 			if (skb)
499 				igmpv3_sendpack(skb);
500 			skb = igmpv3_newpack(dev, mtu);
501 		}
502 	}
503 	first = 1;
504 	psf_prev = NULL;
505 	for (psf = *psf_list; psf; psf = psf_next) {
506 		__be32 *psrc;
507 
508 		psf_next = psf->sf_next;
509 
510 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
511 			psf_prev = psf;
512 			continue;
513 		}
514 
515 		/* Based on RFC3376 5.1. Should not send source-list change
516 		 * records when there is a filter mode change.
517 		 */
518 		if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) ||
519 		     (!gdeleted && pmc->crcount)) &&
520 		    (type == IGMPV3_ALLOW_NEW_SOURCES ||
521 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount)
522 			goto decrease_sf_crcount;
523 
524 		/* clear marks on query responses */
525 		if (isquery)
526 			psf->sf_gsresp = 0;
527 
528 		if (AVAILABLE(skb) < sizeof(__be32) +
529 		    first*sizeof(struct igmpv3_grec)) {
530 			if (truncate && !first)
531 				break;	 /* truncate these */
532 			if (pgr)
533 				pgr->grec_nsrcs = htons(scount);
534 			if (skb)
535 				igmpv3_sendpack(skb);
536 			skb = igmpv3_newpack(dev, mtu);
537 			first = 1;
538 			scount = 0;
539 		}
540 		if (first) {
541 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
542 			first = 0;
543 		}
544 		if (!skb)
545 			return NULL;
546 		psrc = skb_put(skb, sizeof(__be32));
547 		*psrc = psf->sf_inaddr;
548 		scount++; stotal++;
549 		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
550 		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
551 decrease_sf_crcount:
552 			psf->sf_crcount--;
553 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
554 				if (psf_prev)
555 					psf_prev->sf_next = psf->sf_next;
556 				else
557 					*psf_list = psf->sf_next;
558 				kfree(psf);
559 				continue;
560 			}
561 		}
562 		psf_prev = psf;
563 	}
564 
565 empty_source:
566 	if (!stotal) {
567 		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
568 		    type == IGMPV3_BLOCK_OLD_SOURCES)
569 			return skb;
570 		if (pmc->crcount || isquery) {
571 			/* make sure we have room for group header */
572 			if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
573 				igmpv3_sendpack(skb);
574 				skb = NULL; /* add_grhead will get a new one */
575 			}
576 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
577 		}
578 	}
579 	if (pgr)
580 		pgr->grec_nsrcs = htons(scount);
581 
582 	if (isquery)
583 		pmc->gsquery = 0;	/* clear query state on report */
584 	return skb;
585 }
586 
igmpv3_send_report(struct in_device * in_dev,struct ip_mc_list * pmc)587 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
588 {
589 	struct sk_buff *skb = NULL;
590 	struct net *net = dev_net(in_dev->dev);
591 	int type;
592 
593 	if (!pmc) {
594 		rcu_read_lock();
595 		for_each_pmc_rcu(in_dev, pmc) {
596 			if (pmc->multiaddr == IGMP_ALL_HOSTS)
597 				continue;
598 			if (ipv4_is_local_multicast(pmc->multiaddr) &&
599 			    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
600 				continue;
601 			spin_lock_bh(&pmc->lock);
602 			if (pmc->sfcount[MCAST_EXCLUDE])
603 				type = IGMPV3_MODE_IS_EXCLUDE;
604 			else
605 				type = IGMPV3_MODE_IS_INCLUDE;
606 			skb = add_grec(skb, pmc, type, 0, 0);
607 			spin_unlock_bh(&pmc->lock);
608 		}
609 		rcu_read_unlock();
610 	} else {
611 		spin_lock_bh(&pmc->lock);
612 		if (pmc->sfcount[MCAST_EXCLUDE])
613 			type = IGMPV3_MODE_IS_EXCLUDE;
614 		else
615 			type = IGMPV3_MODE_IS_INCLUDE;
616 		skb = add_grec(skb, pmc, type, 0, 0);
617 		spin_unlock_bh(&pmc->lock);
618 	}
619 	if (!skb)
620 		return 0;
621 	return igmpv3_sendpack(skb);
622 }
623 
624 /*
625  * remove zero-count source records from a source filter list
626  */
igmpv3_clear_zeros(struct ip_sf_list ** ppsf)627 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
628 {
629 	struct ip_sf_list *psf_prev, *psf_next, *psf;
630 
631 	psf_prev = NULL;
632 	for (psf = *ppsf; psf; psf = psf_next) {
633 		psf_next = psf->sf_next;
634 		if (psf->sf_crcount == 0) {
635 			if (psf_prev)
636 				psf_prev->sf_next = psf->sf_next;
637 			else
638 				*ppsf = psf->sf_next;
639 			kfree(psf);
640 		} else
641 			psf_prev = psf;
642 	}
643 }
644 
kfree_pmc(struct ip_mc_list * pmc)645 static void kfree_pmc(struct ip_mc_list *pmc)
646 {
647 	ip_sf_list_clear_all(pmc->sources);
648 	ip_sf_list_clear_all(pmc->tomb);
649 	kfree(pmc);
650 }
651 
igmpv3_send_cr(struct in_device * in_dev)652 static void igmpv3_send_cr(struct in_device *in_dev)
653 {
654 	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
655 	struct sk_buff *skb = NULL;
656 	int type, dtype;
657 
658 	rcu_read_lock();
659 	spin_lock_bh(&in_dev->mc_tomb_lock);
660 
661 	/* deleted MCA's */
662 	pmc_prev = NULL;
663 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
664 		pmc_next = pmc->next;
665 		if (pmc->sfmode == MCAST_INCLUDE) {
666 			type = IGMPV3_BLOCK_OLD_SOURCES;
667 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
668 			skb = add_grec(skb, pmc, type, 1, 0);
669 			skb = add_grec(skb, pmc, dtype, 1, 1);
670 		}
671 		if (pmc->crcount) {
672 			if (pmc->sfmode == MCAST_EXCLUDE) {
673 				type = IGMPV3_CHANGE_TO_INCLUDE;
674 				skb = add_grec(skb, pmc, type, 1, 0);
675 			}
676 			pmc->crcount--;
677 			if (pmc->crcount == 0) {
678 				igmpv3_clear_zeros(&pmc->tomb);
679 				igmpv3_clear_zeros(&pmc->sources);
680 			}
681 		}
682 		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
683 			if (pmc_prev)
684 				pmc_prev->next = pmc_next;
685 			else
686 				in_dev->mc_tomb = pmc_next;
687 			in_dev_put(pmc->interface);
688 			kfree_pmc(pmc);
689 		} else
690 			pmc_prev = pmc;
691 	}
692 	spin_unlock_bh(&in_dev->mc_tomb_lock);
693 
694 	/* change recs */
695 	for_each_pmc_rcu(in_dev, pmc) {
696 		spin_lock_bh(&pmc->lock);
697 		if (pmc->sfcount[MCAST_EXCLUDE]) {
698 			type = IGMPV3_BLOCK_OLD_SOURCES;
699 			dtype = IGMPV3_ALLOW_NEW_SOURCES;
700 		} else {
701 			type = IGMPV3_ALLOW_NEW_SOURCES;
702 			dtype = IGMPV3_BLOCK_OLD_SOURCES;
703 		}
704 		skb = add_grec(skb, pmc, type, 0, 0);
705 		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
706 
707 		/* filter mode changes */
708 		if (pmc->crcount) {
709 			if (pmc->sfmode == MCAST_EXCLUDE)
710 				type = IGMPV3_CHANGE_TO_EXCLUDE;
711 			else
712 				type = IGMPV3_CHANGE_TO_INCLUDE;
713 			skb = add_grec(skb, pmc, type, 0, 0);
714 			pmc->crcount--;
715 		}
716 		spin_unlock_bh(&pmc->lock);
717 	}
718 	rcu_read_unlock();
719 
720 	if (!skb)
721 		return;
722 	(void) igmpv3_sendpack(skb);
723 }
724 
igmp_send_report(struct in_device * in_dev,struct ip_mc_list * pmc,int type)725 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
726 	int type)
727 {
728 	struct sk_buff *skb;
729 	struct iphdr *iph;
730 	struct igmphdr *ih;
731 	struct rtable *rt;
732 	struct net_device *dev = in_dev->dev;
733 	struct net *net = dev_net(dev);
734 	__be32	group = pmc ? pmc->multiaddr : 0;
735 	struct flowi4 fl4;
736 	__be32	dst;
737 	int hlen, tlen;
738 
739 	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
740 		return igmpv3_send_report(in_dev, pmc);
741 
742 	if (ipv4_is_local_multicast(group) &&
743 	    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
744 		return 0;
745 
746 	if (type == IGMP_HOST_LEAVE_MESSAGE)
747 		dst = IGMP_ALL_ROUTER;
748 	else
749 		dst = group;
750 
751 	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
752 				   0, 0,
753 				   IPPROTO_IGMP, 0, dev->ifindex);
754 	if (IS_ERR(rt))
755 		return -1;
756 
757 	hlen = LL_RESERVED_SPACE(dev);
758 	tlen = dev->needed_tailroom;
759 	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
760 	if (!skb) {
761 		ip_rt_put(rt);
762 		return -1;
763 	}
764 	skb->priority = TC_PRIO_CONTROL;
765 
766 	skb_dst_set(skb, &rt->dst);
767 
768 	skb_reserve(skb, hlen);
769 
770 	skb_reset_network_header(skb);
771 	iph = ip_hdr(skb);
772 	skb_put(skb, sizeof(struct iphdr) + 4);
773 
774 	iph->version  = 4;
775 	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
776 	iph->tos      = 0xc0;
777 	iph->frag_off = htons(IP_DF);
778 	iph->ttl      = 1;
779 	iph->daddr    = dst;
780 	iph->saddr    = fl4.saddr;
781 	iph->protocol = IPPROTO_IGMP;
782 	ip_select_ident(net, skb, NULL);
783 	((u8 *)&iph[1])[0] = IPOPT_RA;
784 	((u8 *)&iph[1])[1] = 4;
785 	((u8 *)&iph[1])[2] = 0;
786 	((u8 *)&iph[1])[3] = 0;
787 
788 	ih = skb_put(skb, sizeof(struct igmphdr));
789 	ih->type = type;
790 	ih->code = 0;
791 	ih->csum = 0;
792 	ih->group = group;
793 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
794 
795 	return ip_local_out(net, skb->sk, skb);
796 }
797 
igmp_gq_timer_expire(struct timer_list * t)798 static void igmp_gq_timer_expire(struct timer_list *t)
799 {
800 	struct in_device *in_dev = from_timer(in_dev, t, mr_gq_timer);
801 
802 	in_dev->mr_gq_running = 0;
803 	igmpv3_send_report(in_dev, NULL);
804 	in_dev_put(in_dev);
805 }
806 
igmp_ifc_timer_expire(struct timer_list * t)807 static void igmp_ifc_timer_expire(struct timer_list *t)
808 {
809 	struct in_device *in_dev = from_timer(in_dev, t, mr_ifc_timer);
810 	u32 mr_ifc_count;
811 
812 	igmpv3_send_cr(in_dev);
813 restart:
814 	mr_ifc_count = READ_ONCE(in_dev->mr_ifc_count);
815 
816 	if (mr_ifc_count) {
817 		if (cmpxchg(&in_dev->mr_ifc_count,
818 			    mr_ifc_count,
819 			    mr_ifc_count - 1) != mr_ifc_count)
820 			goto restart;
821 		igmp_ifc_start_timer(in_dev,
822 				     unsolicited_report_interval(in_dev));
823 	}
824 	in_dev_put(in_dev);
825 }
826 
igmp_ifc_event(struct in_device * in_dev)827 static void igmp_ifc_event(struct in_device *in_dev)
828 {
829 	struct net *net = dev_net(in_dev->dev);
830 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
831 		return;
832 	WRITE_ONCE(in_dev->mr_ifc_count, in_dev->mr_qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv));
833 	igmp_ifc_start_timer(in_dev, 1);
834 }
835 
836 
igmp_timer_expire(struct timer_list * t)837 static void igmp_timer_expire(struct timer_list *t)
838 {
839 	struct ip_mc_list *im = from_timer(im, t, timer);
840 	struct in_device *in_dev = im->interface;
841 
842 	spin_lock(&im->lock);
843 	im->tm_running = 0;
844 
845 	if (im->unsolicit_count && --im->unsolicit_count)
846 		igmp_start_timer(im, unsolicited_report_interval(in_dev));
847 
848 	im->reporter = 1;
849 	spin_unlock(&im->lock);
850 
851 	if (IGMP_V1_SEEN(in_dev))
852 		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
853 	else if (IGMP_V2_SEEN(in_dev))
854 		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
855 	else
856 		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
857 
858 	ip_ma_put(im);
859 }
860 
861 /* mark EXCLUDE-mode sources */
igmp_xmarksources(struct ip_mc_list * pmc,int nsrcs,__be32 * srcs)862 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
863 {
864 	struct ip_sf_list *psf;
865 	int i, scount;
866 
867 	scount = 0;
868 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
869 		if (scount == nsrcs)
870 			break;
871 		for (i = 0; i < nsrcs; i++) {
872 			/* skip inactive filters */
873 			if (psf->sf_count[MCAST_INCLUDE] ||
874 			    pmc->sfcount[MCAST_EXCLUDE] !=
875 			    psf->sf_count[MCAST_EXCLUDE])
876 				break;
877 			if (srcs[i] == psf->sf_inaddr) {
878 				scount++;
879 				break;
880 			}
881 		}
882 	}
883 	pmc->gsquery = 0;
884 	if (scount == nsrcs)	/* all sources excluded */
885 		return 0;
886 	return 1;
887 }
888 
igmp_marksources(struct ip_mc_list * pmc,int nsrcs,__be32 * srcs)889 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
890 {
891 	struct ip_sf_list *psf;
892 	int i, scount;
893 
894 	if (pmc->sfmode == MCAST_EXCLUDE)
895 		return igmp_xmarksources(pmc, nsrcs, srcs);
896 
897 	/* mark INCLUDE-mode sources */
898 	scount = 0;
899 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
900 		if (scount == nsrcs)
901 			break;
902 		for (i = 0; i < nsrcs; i++)
903 			if (srcs[i] == psf->sf_inaddr) {
904 				psf->sf_gsresp = 1;
905 				scount++;
906 				break;
907 			}
908 	}
909 	if (!scount) {
910 		pmc->gsquery = 0;
911 		return 0;
912 	}
913 	pmc->gsquery = 1;
914 	return 1;
915 }
916 
917 /* return true if packet was dropped */
igmp_heard_report(struct in_device * in_dev,__be32 group)918 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
919 {
920 	struct ip_mc_list *im;
921 	struct net *net = dev_net(in_dev->dev);
922 
923 	/* Timers are only set for non-local groups */
924 
925 	if (group == IGMP_ALL_HOSTS)
926 		return false;
927 	if (ipv4_is_local_multicast(group) &&
928 	    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
929 		return false;
930 
931 	rcu_read_lock();
932 	for_each_pmc_rcu(in_dev, im) {
933 		if (im->multiaddr == group) {
934 			igmp_stop_timer(im);
935 			break;
936 		}
937 	}
938 	rcu_read_unlock();
939 	return false;
940 }
941 
942 /* return true if packet was dropped */
igmp_heard_query(struct in_device * in_dev,struct sk_buff * skb,int len)943 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
944 	int len)
945 {
946 	struct igmphdr 		*ih = igmp_hdr(skb);
947 	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
948 	struct ip_mc_list	*im;
949 	__be32			group = ih->group;
950 	int			max_delay;
951 	int			mark = 0;
952 	struct net		*net = dev_net(in_dev->dev);
953 
954 
955 	if (len == 8) {
956 		if (ih->code == 0) {
957 			/* Alas, old v1 router presents here. */
958 
959 			max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
960 			in_dev->mr_v1_seen = jiffies +
961 				(in_dev->mr_qrv * in_dev->mr_qi) +
962 				in_dev->mr_qri;
963 			group = 0;
964 		} else {
965 			/* v2 router present */
966 			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
967 			in_dev->mr_v2_seen = jiffies +
968 				(in_dev->mr_qrv * in_dev->mr_qi) +
969 				in_dev->mr_qri;
970 		}
971 		/* cancel the interface change timer */
972 		WRITE_ONCE(in_dev->mr_ifc_count, 0);
973 		if (del_timer(&in_dev->mr_ifc_timer))
974 			__in_dev_put(in_dev);
975 		/* clear deleted report items */
976 		igmpv3_clear_delrec(in_dev);
977 	} else if (len < 12) {
978 		return true;	/* ignore bogus packet; freed by caller */
979 	} else if (IGMP_V1_SEEN(in_dev)) {
980 		/* This is a v3 query with v1 queriers present */
981 		max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
982 		group = 0;
983 	} else if (IGMP_V2_SEEN(in_dev)) {
984 		/* this is a v3 query with v2 queriers present;
985 		 * Interpretation of the max_delay code is problematic here.
986 		 * A real v2 host would use ih_code directly, while v3 has a
987 		 * different encoding. We use the v3 encoding as more likely
988 		 * to be intended in a v3 query.
989 		 */
990 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
991 		if (!max_delay)
992 			max_delay = 1;	/* can't mod w/ 0 */
993 	} else { /* v3 */
994 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
995 			return true;
996 
997 		ih3 = igmpv3_query_hdr(skb);
998 		if (ih3->nsrcs) {
999 			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
1000 					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
1001 				return true;
1002 			ih3 = igmpv3_query_hdr(skb);
1003 		}
1004 
1005 		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
1006 		if (!max_delay)
1007 			max_delay = 1;	/* can't mod w/ 0 */
1008 		in_dev->mr_maxdelay = max_delay;
1009 
1010 		/* RFC3376, 4.1.6. QRV and 4.1.7. QQIC, when the most recently
1011 		 * received value was zero, use the default or statically
1012 		 * configured value.
1013 		 */
1014 		in_dev->mr_qrv = ih3->qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1015 		in_dev->mr_qi = IGMPV3_QQIC(ih3->qqic)*HZ ?: IGMP_QUERY_INTERVAL;
1016 
1017 		/* RFC3376, 8.3. Query Response Interval:
1018 		 * The number of seconds represented by the [Query Response
1019 		 * Interval] must be less than the [Query Interval].
1020 		 */
1021 		if (in_dev->mr_qri >= in_dev->mr_qi)
1022 			in_dev->mr_qri = (in_dev->mr_qi/HZ - 1)*HZ;
1023 
1024 		if (!group) { /* general query */
1025 			if (ih3->nsrcs)
1026 				return true;	/* no sources allowed */
1027 			igmp_gq_start_timer(in_dev);
1028 			return false;
1029 		}
1030 		/* mark sources to include, if group & source-specific */
1031 		mark = ih3->nsrcs != 0;
1032 	}
1033 
1034 	/*
1035 	 * - Start the timers in all of our membership records
1036 	 *   that the query applies to for the interface on
1037 	 *   which the query arrived excl. those that belong
1038 	 *   to a "local" group (224.0.0.X)
1039 	 * - For timers already running check if they need to
1040 	 *   be reset.
1041 	 * - Use the igmp->igmp_code field as the maximum
1042 	 *   delay possible
1043 	 */
1044 	rcu_read_lock();
1045 	for_each_pmc_rcu(in_dev, im) {
1046 		int changed;
1047 
1048 		if (group && group != im->multiaddr)
1049 			continue;
1050 		if (im->multiaddr == IGMP_ALL_HOSTS)
1051 			continue;
1052 		if (ipv4_is_local_multicast(im->multiaddr) &&
1053 		    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
1054 			continue;
1055 		spin_lock_bh(&im->lock);
1056 		if (im->tm_running)
1057 			im->gsquery = im->gsquery && mark;
1058 		else
1059 			im->gsquery = mark;
1060 		changed = !im->gsquery ||
1061 			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
1062 		spin_unlock_bh(&im->lock);
1063 		if (changed)
1064 			igmp_mod_timer(im, max_delay);
1065 	}
1066 	rcu_read_unlock();
1067 	return false;
1068 }
1069 
1070 /* called in rcu_read_lock() section */
igmp_rcv(struct sk_buff * skb)1071 int igmp_rcv(struct sk_buff *skb)
1072 {
1073 	/* This basically follows the spec line by line -- see RFC1112 */
1074 	struct igmphdr *ih;
1075 	struct net_device *dev = skb->dev;
1076 	struct in_device *in_dev;
1077 	int len = skb->len;
1078 	bool dropped = true;
1079 
1080 	if (netif_is_l3_master(dev)) {
1081 		dev = dev_get_by_index_rcu(dev_net(dev), IPCB(skb)->iif);
1082 		if (!dev)
1083 			goto drop;
1084 	}
1085 
1086 	in_dev = __in_dev_get_rcu(dev);
1087 	if (!in_dev)
1088 		goto drop;
1089 
1090 	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1091 		goto drop;
1092 
1093 	if (skb_checksum_simple_validate(skb))
1094 		goto drop;
1095 
1096 	ih = igmp_hdr(skb);
1097 	switch (ih->type) {
1098 	case IGMP_HOST_MEMBERSHIP_QUERY:
1099 		dropped = igmp_heard_query(in_dev, skb, len);
1100 		break;
1101 	case IGMP_HOST_MEMBERSHIP_REPORT:
1102 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1103 		/* Is it our report looped back? */
1104 		if (rt_is_output_route(skb_rtable(skb)))
1105 			break;
1106 		/* don't rely on MC router hearing unicast reports */
1107 		if (skb->pkt_type == PACKET_MULTICAST ||
1108 		    skb->pkt_type == PACKET_BROADCAST)
1109 			dropped = igmp_heard_report(in_dev, ih->group);
1110 		break;
1111 	case IGMP_PIM:
1112 #ifdef CONFIG_IP_PIMSM_V1
1113 		return pim_rcv_v1(skb);
1114 #endif
1115 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1116 	case IGMP_DVMRP:
1117 	case IGMP_TRACE:
1118 	case IGMP_HOST_LEAVE_MESSAGE:
1119 	case IGMP_MTRACE:
1120 	case IGMP_MTRACE_RESP:
1121 		break;
1122 	default:
1123 		break;
1124 	}
1125 
1126 drop:
1127 	if (dropped)
1128 		kfree_skb(skb);
1129 	else
1130 		consume_skb(skb);
1131 	return 0;
1132 }
1133 
1134 #endif
1135 
1136 
1137 /*
1138  *	Add a filter to a device
1139  */
1140 
ip_mc_filter_add(struct in_device * in_dev,__be32 addr)1141 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1142 {
1143 	char buf[MAX_ADDR_LEN];
1144 	struct net_device *dev = in_dev->dev;
1145 
1146 	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1147 	   We will get multicast token leakage, when IFF_MULTICAST
1148 	   is changed. This check should be done in ndo_set_rx_mode
1149 	   routine. Something sort of:
1150 	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1151 	   --ANK
1152 	   */
1153 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1154 		dev_mc_add(dev, buf);
1155 }
1156 
1157 /*
1158  *	Remove a filter from a device
1159  */
1160 
ip_mc_filter_del(struct in_device * in_dev,__be32 addr)1161 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1162 {
1163 	char buf[MAX_ADDR_LEN];
1164 	struct net_device *dev = in_dev->dev;
1165 
1166 	if (arp_mc_map(addr, buf, dev, 0) == 0)
1167 		dev_mc_del(dev, buf);
1168 }
1169 
1170 #ifdef CONFIG_IP_MULTICAST
1171 /*
1172  * deleted ip_mc_list manipulation
1173  */
igmpv3_add_delrec(struct in_device * in_dev,struct ip_mc_list * im,gfp_t gfp)1174 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im,
1175 			      gfp_t gfp)
1176 {
1177 	struct ip_mc_list *pmc;
1178 	struct net *net = dev_net(in_dev->dev);
1179 
1180 	/* this is an "ip_mc_list" for convenience; only the fields below
1181 	 * are actually used. In particular, the refcnt and users are not
1182 	 * used for management of the delete list. Using the same structure
1183 	 * for deleted items allows change reports to use common code with
1184 	 * non-deleted or query-response MCA's.
1185 	 */
1186 	pmc = kzalloc(sizeof(*pmc), gfp);
1187 	if (!pmc)
1188 		return;
1189 	spin_lock_init(&pmc->lock);
1190 	spin_lock_bh(&im->lock);
1191 	pmc->interface = im->interface;
1192 	in_dev_hold(in_dev);
1193 	pmc->multiaddr = im->multiaddr;
1194 	pmc->crcount = in_dev->mr_qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1195 	pmc->sfmode = im->sfmode;
1196 	if (pmc->sfmode == MCAST_INCLUDE) {
1197 		struct ip_sf_list *psf;
1198 
1199 		pmc->tomb = im->tomb;
1200 		pmc->sources = im->sources;
1201 		im->tomb = im->sources = NULL;
1202 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1203 			psf->sf_crcount = pmc->crcount;
1204 	}
1205 	spin_unlock_bh(&im->lock);
1206 
1207 	spin_lock_bh(&in_dev->mc_tomb_lock);
1208 	pmc->next = in_dev->mc_tomb;
1209 	in_dev->mc_tomb = pmc;
1210 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1211 }
1212 
1213 /*
1214  * restore ip_mc_list deleted records
1215  */
igmpv3_del_delrec(struct in_device * in_dev,struct ip_mc_list * im)1216 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1217 {
1218 	struct ip_mc_list *pmc, *pmc_prev;
1219 	struct ip_sf_list *psf;
1220 	struct net *net = dev_net(in_dev->dev);
1221 	__be32 multiaddr = im->multiaddr;
1222 
1223 	spin_lock_bh(&in_dev->mc_tomb_lock);
1224 	pmc_prev = NULL;
1225 	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1226 		if (pmc->multiaddr == multiaddr)
1227 			break;
1228 		pmc_prev = pmc;
1229 	}
1230 	if (pmc) {
1231 		if (pmc_prev)
1232 			pmc_prev->next = pmc->next;
1233 		else
1234 			in_dev->mc_tomb = pmc->next;
1235 	}
1236 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1237 
1238 	spin_lock_bh(&im->lock);
1239 	if (pmc) {
1240 		im->interface = pmc->interface;
1241 		if (im->sfmode == MCAST_INCLUDE) {
1242 			swap(im->tomb, pmc->tomb);
1243 			swap(im->sources, pmc->sources);
1244 			for (psf = im->sources; psf; psf = psf->sf_next)
1245 				psf->sf_crcount = in_dev->mr_qrv ?:
1246 					READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1247 		} else {
1248 			im->crcount = in_dev->mr_qrv ?:
1249 				READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1250 		}
1251 		in_dev_put(pmc->interface);
1252 		kfree_pmc(pmc);
1253 	}
1254 	spin_unlock_bh(&im->lock);
1255 }
1256 
1257 /*
1258  * flush ip_mc_list deleted records
1259  */
igmpv3_clear_delrec(struct in_device * in_dev)1260 static void igmpv3_clear_delrec(struct in_device *in_dev)
1261 {
1262 	struct ip_mc_list *pmc, *nextpmc;
1263 
1264 	spin_lock_bh(&in_dev->mc_tomb_lock);
1265 	pmc = in_dev->mc_tomb;
1266 	in_dev->mc_tomb = NULL;
1267 	spin_unlock_bh(&in_dev->mc_tomb_lock);
1268 
1269 	for (; pmc; pmc = nextpmc) {
1270 		nextpmc = pmc->next;
1271 		ip_mc_clear_src(pmc);
1272 		in_dev_put(pmc->interface);
1273 		kfree_pmc(pmc);
1274 	}
1275 	/* clear dead sources, too */
1276 	rcu_read_lock();
1277 	for_each_pmc_rcu(in_dev, pmc) {
1278 		struct ip_sf_list *psf;
1279 
1280 		spin_lock_bh(&pmc->lock);
1281 		psf = pmc->tomb;
1282 		pmc->tomb = NULL;
1283 		spin_unlock_bh(&pmc->lock);
1284 		ip_sf_list_clear_all(psf);
1285 	}
1286 	rcu_read_unlock();
1287 }
1288 #endif
1289 
__igmp_group_dropped(struct ip_mc_list * im,gfp_t gfp)1290 static void __igmp_group_dropped(struct ip_mc_list *im, gfp_t gfp)
1291 {
1292 	struct in_device *in_dev = im->interface;
1293 #ifdef CONFIG_IP_MULTICAST
1294 	struct net *net = dev_net(in_dev->dev);
1295 	int reporter;
1296 #endif
1297 
1298 	if (im->loaded) {
1299 		im->loaded = 0;
1300 		ip_mc_filter_del(in_dev, im->multiaddr);
1301 	}
1302 
1303 #ifdef CONFIG_IP_MULTICAST
1304 	if (im->multiaddr == IGMP_ALL_HOSTS)
1305 		return;
1306 	if (ipv4_is_local_multicast(im->multiaddr) &&
1307 	    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
1308 		return;
1309 
1310 	reporter = im->reporter;
1311 	igmp_stop_timer(im);
1312 
1313 	if (!in_dev->dead) {
1314 		if (IGMP_V1_SEEN(in_dev))
1315 			return;
1316 		if (IGMP_V2_SEEN(in_dev)) {
1317 			if (reporter)
1318 				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1319 			return;
1320 		}
1321 		/* IGMPv3 */
1322 		igmpv3_add_delrec(in_dev, im, gfp);
1323 
1324 		igmp_ifc_event(in_dev);
1325 	}
1326 #endif
1327 }
1328 
igmp_group_dropped(struct ip_mc_list * im)1329 static void igmp_group_dropped(struct ip_mc_list *im)
1330 {
1331 	__igmp_group_dropped(im, GFP_KERNEL);
1332 }
1333 
igmp_group_added(struct ip_mc_list * im)1334 static void igmp_group_added(struct ip_mc_list *im)
1335 {
1336 	struct in_device *in_dev = im->interface;
1337 #ifdef CONFIG_IP_MULTICAST
1338 	struct net *net = dev_net(in_dev->dev);
1339 #endif
1340 
1341 	if (im->loaded == 0) {
1342 		im->loaded = 1;
1343 		ip_mc_filter_add(in_dev, im->multiaddr);
1344 	}
1345 
1346 #ifdef CONFIG_IP_MULTICAST
1347 	if (im->multiaddr == IGMP_ALL_HOSTS)
1348 		return;
1349 	if (ipv4_is_local_multicast(im->multiaddr) &&
1350 	    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
1351 		return;
1352 
1353 	if (in_dev->dead)
1354 		return;
1355 
1356 	im->unsolicit_count = READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1357 	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1358 		spin_lock_bh(&im->lock);
1359 		igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1360 		spin_unlock_bh(&im->lock);
1361 		return;
1362 	}
1363 	/* else, v3 */
1364 
1365 	/* Based on RFC3376 5.1, for newly added INCLUDE SSM, we should
1366 	 * not send filter-mode change record as the mode should be from
1367 	 * IN() to IN(A).
1368 	 */
1369 	if (im->sfmode == MCAST_EXCLUDE)
1370 		im->crcount = in_dev->mr_qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1371 
1372 	igmp_ifc_event(in_dev);
1373 #endif
1374 }
1375 
1376 
1377 /*
1378  *	Multicast list managers
1379  */
1380 
ip_mc_hash(const struct ip_mc_list * im)1381 static u32 ip_mc_hash(const struct ip_mc_list *im)
1382 {
1383 	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1384 }
1385 
ip_mc_hash_add(struct in_device * in_dev,struct ip_mc_list * im)1386 static void ip_mc_hash_add(struct in_device *in_dev,
1387 			   struct ip_mc_list *im)
1388 {
1389 	struct ip_mc_list __rcu **mc_hash;
1390 	u32 hash;
1391 
1392 	mc_hash = rtnl_dereference(in_dev->mc_hash);
1393 	if (mc_hash) {
1394 		hash = ip_mc_hash(im);
1395 		im->next_hash = mc_hash[hash];
1396 		rcu_assign_pointer(mc_hash[hash], im);
1397 		return;
1398 	}
1399 
1400 	/* do not use a hash table for small number of items */
1401 	if (in_dev->mc_count < 4)
1402 		return;
1403 
1404 	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1405 			  GFP_KERNEL);
1406 	if (!mc_hash)
1407 		return;
1408 
1409 	for_each_pmc_rtnl(in_dev, im) {
1410 		hash = ip_mc_hash(im);
1411 		im->next_hash = mc_hash[hash];
1412 		RCU_INIT_POINTER(mc_hash[hash], im);
1413 	}
1414 
1415 	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1416 }
1417 
ip_mc_hash_remove(struct in_device * in_dev,struct ip_mc_list * im)1418 static void ip_mc_hash_remove(struct in_device *in_dev,
1419 			      struct ip_mc_list *im)
1420 {
1421 	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1422 	struct ip_mc_list *aux;
1423 
1424 	if (!mc_hash)
1425 		return;
1426 	mc_hash += ip_mc_hash(im);
1427 	while ((aux = rtnl_dereference(*mc_hash)) != im)
1428 		mc_hash = &aux->next_hash;
1429 	*mc_hash = im->next_hash;
1430 }
1431 
1432 
1433 /*
1434  *	A socket has joined a multicast group on device dev.
1435  */
____ip_mc_inc_group(struct in_device * in_dev,__be32 addr,unsigned int mode,gfp_t gfp)1436 static void ____ip_mc_inc_group(struct in_device *in_dev, __be32 addr,
1437 				unsigned int mode, gfp_t gfp)
1438 {
1439 	struct ip_mc_list *im;
1440 
1441 	ASSERT_RTNL();
1442 
1443 	for_each_pmc_rtnl(in_dev, im) {
1444 		if (im->multiaddr == addr) {
1445 			im->users++;
1446 			ip_mc_add_src(in_dev, &addr, mode, 0, NULL, 0);
1447 			goto out;
1448 		}
1449 	}
1450 
1451 	im = kzalloc(sizeof(*im), gfp);
1452 	if (!im)
1453 		goto out;
1454 
1455 	im->users = 1;
1456 	im->interface = in_dev;
1457 	in_dev_hold(in_dev);
1458 	im->multiaddr = addr;
1459 	/* initial mode is (EX, empty) */
1460 	im->sfmode = mode;
1461 	im->sfcount[mode] = 1;
1462 	refcount_set(&im->refcnt, 1);
1463 	spin_lock_init(&im->lock);
1464 #ifdef CONFIG_IP_MULTICAST
1465 	timer_setup(&im->timer, igmp_timer_expire, 0);
1466 #endif
1467 
1468 	im->next_rcu = in_dev->mc_list;
1469 	in_dev->mc_count++;
1470 	rcu_assign_pointer(in_dev->mc_list, im);
1471 
1472 	ip_mc_hash_add(in_dev, im);
1473 
1474 #ifdef CONFIG_IP_MULTICAST
1475 	igmpv3_del_delrec(in_dev, im);
1476 #endif
1477 	igmp_group_added(im);
1478 	if (!in_dev->dead)
1479 		ip_rt_multicast_event(in_dev);
1480 out:
1481 	return;
1482 }
1483 
__ip_mc_inc_group(struct in_device * in_dev,__be32 addr,gfp_t gfp)1484 void __ip_mc_inc_group(struct in_device *in_dev, __be32 addr, gfp_t gfp)
1485 {
1486 	____ip_mc_inc_group(in_dev, addr, MCAST_EXCLUDE, gfp);
1487 }
1488 EXPORT_SYMBOL(__ip_mc_inc_group);
1489 
ip_mc_inc_group(struct in_device * in_dev,__be32 addr)1490 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1491 {
1492 	__ip_mc_inc_group(in_dev, addr, GFP_KERNEL);
1493 }
1494 EXPORT_SYMBOL(ip_mc_inc_group);
1495 
ip_mc_check_iphdr(struct sk_buff * skb)1496 static int ip_mc_check_iphdr(struct sk_buff *skb)
1497 {
1498 	const struct iphdr *iph;
1499 	unsigned int len;
1500 	unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1501 
1502 	if (!pskb_may_pull(skb, offset))
1503 		return -EINVAL;
1504 
1505 	iph = ip_hdr(skb);
1506 
1507 	if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1508 		return -EINVAL;
1509 
1510 	offset += ip_hdrlen(skb) - sizeof(*iph);
1511 
1512 	if (!pskb_may_pull(skb, offset))
1513 		return -EINVAL;
1514 
1515 	iph = ip_hdr(skb);
1516 
1517 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1518 		return -EINVAL;
1519 
1520 	len = skb_network_offset(skb) + ntohs(iph->tot_len);
1521 	if (skb->len < len || len < offset)
1522 		return -EINVAL;
1523 
1524 	skb_set_transport_header(skb, offset);
1525 
1526 	return 0;
1527 }
1528 
ip_mc_check_igmp_reportv3(struct sk_buff * skb)1529 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1530 {
1531 	unsigned int len = skb_transport_offset(skb);
1532 
1533 	len += sizeof(struct igmpv3_report);
1534 
1535 	return ip_mc_may_pull(skb, len) ? 0 : -EINVAL;
1536 }
1537 
ip_mc_check_igmp_query(struct sk_buff * skb)1538 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1539 {
1540 	unsigned int transport_len = ip_transport_len(skb);
1541 	unsigned int len;
1542 
1543 	/* IGMPv{1,2}? */
1544 	if (transport_len != sizeof(struct igmphdr)) {
1545 		/* or IGMPv3? */
1546 		if (transport_len < sizeof(struct igmpv3_query))
1547 			return -EINVAL;
1548 
1549 		len = skb_transport_offset(skb) + sizeof(struct igmpv3_query);
1550 		if (!ip_mc_may_pull(skb, len))
1551 			return -EINVAL;
1552 	}
1553 
1554 	/* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1555 	 * all-systems destination addresses (224.0.0.1) for general queries
1556 	 */
1557 	if (!igmp_hdr(skb)->group &&
1558 	    ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1559 		return -EINVAL;
1560 
1561 	return 0;
1562 }
1563 
ip_mc_check_igmp_msg(struct sk_buff * skb)1564 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1565 {
1566 	switch (igmp_hdr(skb)->type) {
1567 	case IGMP_HOST_LEAVE_MESSAGE:
1568 	case IGMP_HOST_MEMBERSHIP_REPORT:
1569 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1570 		return 0;
1571 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1572 		return ip_mc_check_igmp_reportv3(skb);
1573 	case IGMP_HOST_MEMBERSHIP_QUERY:
1574 		return ip_mc_check_igmp_query(skb);
1575 	default:
1576 		return -ENOMSG;
1577 	}
1578 }
1579 
ip_mc_validate_checksum(struct sk_buff * skb)1580 static __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1581 {
1582 	return skb_checksum_simple_validate(skb);
1583 }
1584 
ip_mc_check_igmp_csum(struct sk_buff * skb)1585 static int ip_mc_check_igmp_csum(struct sk_buff *skb)
1586 {
1587 	unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1588 	unsigned int transport_len = ip_transport_len(skb);
1589 	struct sk_buff *skb_chk;
1590 
1591 	if (!ip_mc_may_pull(skb, len))
1592 		return -EINVAL;
1593 
1594 	skb_chk = skb_checksum_trimmed(skb, transport_len,
1595 				       ip_mc_validate_checksum);
1596 	if (!skb_chk)
1597 		return -EINVAL;
1598 
1599 	if (skb_chk != skb)
1600 		kfree_skb(skb_chk);
1601 
1602 	return 0;
1603 }
1604 
1605 /**
1606  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1607  * @skb: the skb to validate
1608  *
1609  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1610  * skb transport header accordingly and returns zero.
1611  *
1612  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1613  *  standard
1614  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1615  * -ENOMEM: A memory allocation failure happened.
1616  *
1617  * Caller needs to set the skb network header and free any returned skb if it
1618  * differs from the provided skb.
1619  */
ip_mc_check_igmp(struct sk_buff * skb)1620 int ip_mc_check_igmp(struct sk_buff *skb)
1621 {
1622 	int ret = ip_mc_check_iphdr(skb);
1623 
1624 	if (ret < 0)
1625 		return ret;
1626 
1627 	if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1628 		return -ENOMSG;
1629 
1630 	ret = ip_mc_check_igmp_csum(skb);
1631 	if (ret < 0)
1632 		return ret;
1633 
1634 	return ip_mc_check_igmp_msg(skb);
1635 }
1636 EXPORT_SYMBOL(ip_mc_check_igmp);
1637 
1638 /*
1639  *	Resend IGMP JOIN report; used by netdev notifier.
1640  */
ip_mc_rejoin_groups(struct in_device * in_dev)1641 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1642 {
1643 #ifdef CONFIG_IP_MULTICAST
1644 	struct ip_mc_list *im;
1645 	int type;
1646 	struct net *net = dev_net(in_dev->dev);
1647 
1648 	ASSERT_RTNL();
1649 
1650 	for_each_pmc_rtnl(in_dev, im) {
1651 		if (im->multiaddr == IGMP_ALL_HOSTS)
1652 			continue;
1653 		if (ipv4_is_local_multicast(im->multiaddr) &&
1654 		    !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))
1655 			continue;
1656 
1657 		/* a failover is happening and switches
1658 		 * must be notified immediately
1659 		 */
1660 		if (IGMP_V1_SEEN(in_dev))
1661 			type = IGMP_HOST_MEMBERSHIP_REPORT;
1662 		else if (IGMP_V2_SEEN(in_dev))
1663 			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1664 		else
1665 			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1666 		igmp_send_report(in_dev, im, type);
1667 	}
1668 #endif
1669 }
1670 
1671 /*
1672  *	A socket has left a multicast group on device dev
1673  */
1674 
__ip_mc_dec_group(struct in_device * in_dev,__be32 addr,gfp_t gfp)1675 void __ip_mc_dec_group(struct in_device *in_dev, __be32 addr, gfp_t gfp)
1676 {
1677 	struct ip_mc_list *i;
1678 	struct ip_mc_list __rcu **ip;
1679 
1680 	ASSERT_RTNL();
1681 
1682 	for (ip = &in_dev->mc_list;
1683 	     (i = rtnl_dereference(*ip)) != NULL;
1684 	     ip = &i->next_rcu) {
1685 		if (i->multiaddr == addr) {
1686 			if (--i->users == 0) {
1687 				ip_mc_hash_remove(in_dev, i);
1688 				*ip = i->next_rcu;
1689 				in_dev->mc_count--;
1690 				__igmp_group_dropped(i, gfp);
1691 				ip_mc_clear_src(i);
1692 
1693 				if (!in_dev->dead)
1694 					ip_rt_multicast_event(in_dev);
1695 
1696 				ip_ma_put(i);
1697 				return;
1698 			}
1699 			break;
1700 		}
1701 	}
1702 }
1703 EXPORT_SYMBOL(__ip_mc_dec_group);
1704 
1705 /* Device changing type */
1706 
ip_mc_unmap(struct in_device * in_dev)1707 void ip_mc_unmap(struct in_device *in_dev)
1708 {
1709 	struct ip_mc_list *pmc;
1710 
1711 	ASSERT_RTNL();
1712 
1713 	for_each_pmc_rtnl(in_dev, pmc)
1714 		igmp_group_dropped(pmc);
1715 }
1716 
ip_mc_remap(struct in_device * in_dev)1717 void ip_mc_remap(struct in_device *in_dev)
1718 {
1719 	struct ip_mc_list *pmc;
1720 
1721 	ASSERT_RTNL();
1722 
1723 	for_each_pmc_rtnl(in_dev, pmc) {
1724 #ifdef CONFIG_IP_MULTICAST
1725 		igmpv3_del_delrec(in_dev, pmc);
1726 #endif
1727 		igmp_group_added(pmc);
1728 	}
1729 }
1730 
1731 /* Device going down */
1732 
ip_mc_down(struct in_device * in_dev)1733 void ip_mc_down(struct in_device *in_dev)
1734 {
1735 	struct ip_mc_list *pmc;
1736 
1737 	ASSERT_RTNL();
1738 
1739 	for_each_pmc_rtnl(in_dev, pmc)
1740 		igmp_group_dropped(pmc);
1741 
1742 #ifdef CONFIG_IP_MULTICAST
1743 	WRITE_ONCE(in_dev->mr_ifc_count, 0);
1744 	if (del_timer(&in_dev->mr_ifc_timer))
1745 		__in_dev_put(in_dev);
1746 	in_dev->mr_gq_running = 0;
1747 	if (del_timer(&in_dev->mr_gq_timer))
1748 		__in_dev_put(in_dev);
1749 #endif
1750 
1751 	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1752 }
1753 
1754 #ifdef CONFIG_IP_MULTICAST
ip_mc_reset(struct in_device * in_dev)1755 static void ip_mc_reset(struct in_device *in_dev)
1756 {
1757 	struct net *net = dev_net(in_dev->dev);
1758 
1759 	in_dev->mr_qi = IGMP_QUERY_INTERVAL;
1760 	in_dev->mr_qri = IGMP_QUERY_RESPONSE_INTERVAL;
1761 	in_dev->mr_qrv = READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1762 }
1763 #else
ip_mc_reset(struct in_device * in_dev)1764 static void ip_mc_reset(struct in_device *in_dev)
1765 {
1766 }
1767 #endif
1768 
ip_mc_init_dev(struct in_device * in_dev)1769 void ip_mc_init_dev(struct in_device *in_dev)
1770 {
1771 	ASSERT_RTNL();
1772 
1773 #ifdef CONFIG_IP_MULTICAST
1774 	timer_setup(&in_dev->mr_gq_timer, igmp_gq_timer_expire, 0);
1775 	timer_setup(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire, 0);
1776 #endif
1777 	ip_mc_reset(in_dev);
1778 
1779 	spin_lock_init(&in_dev->mc_tomb_lock);
1780 }
1781 
1782 /* Device going up */
1783 
ip_mc_up(struct in_device * in_dev)1784 void ip_mc_up(struct in_device *in_dev)
1785 {
1786 	struct ip_mc_list *pmc;
1787 
1788 	ASSERT_RTNL();
1789 
1790 	ip_mc_reset(in_dev);
1791 	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1792 
1793 	for_each_pmc_rtnl(in_dev, pmc) {
1794 #ifdef CONFIG_IP_MULTICAST
1795 		igmpv3_del_delrec(in_dev, pmc);
1796 #endif
1797 		igmp_group_added(pmc);
1798 	}
1799 }
1800 
1801 /*
1802  *	Device is about to be destroyed: clean up.
1803  */
1804 
ip_mc_destroy_dev(struct in_device * in_dev)1805 void ip_mc_destroy_dev(struct in_device *in_dev)
1806 {
1807 	struct ip_mc_list *i;
1808 
1809 	ASSERT_RTNL();
1810 
1811 	/* Deactivate timers */
1812 	ip_mc_down(in_dev);
1813 #ifdef CONFIG_IP_MULTICAST
1814 	igmpv3_clear_delrec(in_dev);
1815 #endif
1816 
1817 	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1818 		in_dev->mc_list = i->next_rcu;
1819 		in_dev->mc_count--;
1820 		ip_mc_clear_src(i);
1821 		ip_ma_put(i);
1822 	}
1823 }
1824 
1825 /* RTNL is locked */
ip_mc_find_dev(struct net * net,struct ip_mreqn * imr)1826 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1827 {
1828 	struct net_device *dev = NULL;
1829 	struct in_device *idev = NULL;
1830 
1831 	if (imr->imr_ifindex) {
1832 		idev = inetdev_by_index(net, imr->imr_ifindex);
1833 		return idev;
1834 	}
1835 	if (imr->imr_address.s_addr) {
1836 		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1837 		if (!dev)
1838 			return NULL;
1839 	}
1840 
1841 	if (!dev) {
1842 		struct rtable *rt = ip_route_output(net,
1843 						    imr->imr_multiaddr.s_addr,
1844 						    0, 0, 0);
1845 		if (!IS_ERR(rt)) {
1846 			dev = rt->dst.dev;
1847 			ip_rt_put(rt);
1848 		}
1849 	}
1850 	if (dev) {
1851 		imr->imr_ifindex = dev->ifindex;
1852 		idev = __in_dev_get_rtnl(dev);
1853 	}
1854 	return idev;
1855 }
1856 
1857 /*
1858  *	Join a socket to a group
1859  */
1860 
ip_mc_del1_src(struct ip_mc_list * pmc,int sfmode,__be32 * psfsrc)1861 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1862 	__be32 *psfsrc)
1863 {
1864 	struct ip_sf_list *psf, *psf_prev;
1865 	int rv = 0;
1866 
1867 	psf_prev = NULL;
1868 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1869 		if (psf->sf_inaddr == *psfsrc)
1870 			break;
1871 		psf_prev = psf;
1872 	}
1873 	if (!psf || psf->sf_count[sfmode] == 0) {
1874 		/* source filter not found, or count wrong =>  bug */
1875 		return -ESRCH;
1876 	}
1877 	psf->sf_count[sfmode]--;
1878 	if (psf->sf_count[sfmode] == 0) {
1879 		ip_rt_multicast_event(pmc->interface);
1880 	}
1881 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1882 #ifdef CONFIG_IP_MULTICAST
1883 		struct in_device *in_dev = pmc->interface;
1884 		struct net *net = dev_net(in_dev->dev);
1885 #endif
1886 
1887 		/* no more filters for this source */
1888 		if (psf_prev)
1889 			psf_prev->sf_next = psf->sf_next;
1890 		else
1891 			pmc->sources = psf->sf_next;
1892 #ifdef CONFIG_IP_MULTICAST
1893 		if (psf->sf_oldin &&
1894 		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1895 			psf->sf_crcount = in_dev->mr_qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1896 			psf->sf_next = pmc->tomb;
1897 			pmc->tomb = psf;
1898 			rv = 1;
1899 		} else
1900 #endif
1901 			kfree(psf);
1902 	}
1903 	return rv;
1904 }
1905 
1906 #ifndef CONFIG_IP_MULTICAST
1907 #define igmp_ifc_event(x)	do { } while (0)
1908 #endif
1909 
ip_mc_del_src(struct in_device * in_dev,__be32 * pmca,int sfmode,int sfcount,__be32 * psfsrc,int delta)1910 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1911 			 int sfcount, __be32 *psfsrc, int delta)
1912 {
1913 	struct ip_mc_list *pmc;
1914 	int	changerec = 0;
1915 	int	i, err;
1916 
1917 	if (!in_dev)
1918 		return -ENODEV;
1919 	rcu_read_lock();
1920 	for_each_pmc_rcu(in_dev, pmc) {
1921 		if (*pmca == pmc->multiaddr)
1922 			break;
1923 	}
1924 	if (!pmc) {
1925 		/* MCA not found?? bug */
1926 		rcu_read_unlock();
1927 		return -ESRCH;
1928 	}
1929 	spin_lock_bh(&pmc->lock);
1930 	rcu_read_unlock();
1931 #ifdef CONFIG_IP_MULTICAST
1932 	sf_markstate(pmc);
1933 #endif
1934 	if (!delta) {
1935 		err = -EINVAL;
1936 		if (!pmc->sfcount[sfmode])
1937 			goto out_unlock;
1938 		pmc->sfcount[sfmode]--;
1939 	}
1940 	err = 0;
1941 	for (i = 0; i < sfcount; i++) {
1942 		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1943 
1944 		changerec |= rv > 0;
1945 		if (!err && rv < 0)
1946 			err = rv;
1947 	}
1948 	if (pmc->sfmode == MCAST_EXCLUDE &&
1949 	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1950 	    pmc->sfcount[MCAST_INCLUDE]) {
1951 #ifdef CONFIG_IP_MULTICAST
1952 		struct ip_sf_list *psf;
1953 		struct net *net = dev_net(in_dev->dev);
1954 #endif
1955 
1956 		/* filter mode change */
1957 		pmc->sfmode = MCAST_INCLUDE;
1958 #ifdef CONFIG_IP_MULTICAST
1959 		pmc->crcount = in_dev->mr_qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);
1960 		WRITE_ONCE(in_dev->mr_ifc_count, pmc->crcount);
1961 		for (psf = pmc->sources; psf; psf = psf->sf_next)
1962 			psf->sf_crcount = 0;
1963 		igmp_ifc_event(pmc->interface);
1964 	} else if (sf_setstate(pmc) || changerec) {
1965 		igmp_ifc_event(pmc->interface);
1966 #endif
1967 	}
1968 out_unlock:
1969 	spin_unlock_bh(&pmc->lock);
1970 	return err;
1971 }
1972 
1973 /*
1974  * Add multicast single-source filter to the interface list
1975  */
ip_mc_add1_src(struct ip_mc_list * pmc,int sfmode,__be32 * psfsrc)1976 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1977 	__be32 *psfsrc)
1978 {
1979 	struct ip_sf_list *psf, *psf_prev;
1980 
1981 	psf_prev = NULL;
1982 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1983 		if (psf->sf_inaddr == *psfsrc)
1984 			break;
1985 		psf_prev = psf;
1986 	}
1987 	if (!psf) {
1988 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1989 		if (!psf)
1990 			return -ENOBUFS;
1991 		psf->sf_inaddr = *psfsrc;
1992 		if (psf_prev) {
1993 			psf_prev->sf_next = psf;
1994 		} else
1995 			pmc->sources = psf;
1996 	}
1997 	psf->sf_count[sfmode]++;
1998 	if (psf->sf_count[sfmode] == 1) {
1999 		ip_rt_multicast_event(pmc->interface);
2000 	}
2001 	return 0;
2002 }
2003 
2004 #ifdef CONFIG_IP_MULTICAST
sf_markstate(struct ip_mc_list * pmc)2005 static void sf_markstate(struct ip_mc_list *pmc)
2006 {
2007 	struct ip_sf_list *psf;
2008 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
2009 
2010 	for (psf = pmc->sources; psf; psf = psf->sf_next)
2011 		if (pmc->sfcount[MCAST_EXCLUDE]) {
2012 			psf->sf_oldin = mca_xcount ==
2013 				psf->sf_count[MCAST_EXCLUDE] &&
2014 				!psf->sf_count[MCAST_INCLUDE];
2015 		} else
2016 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2017 }
2018 
sf_setstate(struct ip_mc_list * pmc)2019 static int sf_setstate(struct ip_mc_list *pmc)
2020 {
2021 	struct ip_sf_list *psf, *dpsf;
2022 	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
2023 	int qrv = pmc->interface->mr_qrv;
2024 	int new_in, rv;
2025 
2026 	rv = 0;
2027 	for (psf = pmc->sources; psf; psf = psf->sf_next) {
2028 		if (pmc->sfcount[MCAST_EXCLUDE]) {
2029 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2030 				!psf->sf_count[MCAST_INCLUDE];
2031 		} else
2032 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2033 		if (new_in) {
2034 			if (!psf->sf_oldin) {
2035 				struct ip_sf_list *prev = NULL;
2036 
2037 				for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
2038 					if (dpsf->sf_inaddr == psf->sf_inaddr)
2039 						break;
2040 					prev = dpsf;
2041 				}
2042 				if (dpsf) {
2043 					if (prev)
2044 						prev->sf_next = dpsf->sf_next;
2045 					else
2046 						pmc->tomb = dpsf->sf_next;
2047 					kfree(dpsf);
2048 				}
2049 				psf->sf_crcount = qrv;
2050 				rv++;
2051 			}
2052 		} else if (psf->sf_oldin) {
2053 
2054 			psf->sf_crcount = 0;
2055 			/*
2056 			 * add or update "delete" records if an active filter
2057 			 * is now inactive
2058 			 */
2059 			for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
2060 				if (dpsf->sf_inaddr == psf->sf_inaddr)
2061 					break;
2062 			if (!dpsf) {
2063 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2064 				if (!dpsf)
2065 					continue;
2066 				*dpsf = *psf;
2067 				/* pmc->lock held by callers */
2068 				dpsf->sf_next = pmc->tomb;
2069 				pmc->tomb = dpsf;
2070 			}
2071 			dpsf->sf_crcount = qrv;
2072 			rv++;
2073 		}
2074 	}
2075 	return rv;
2076 }
2077 #endif
2078 
2079 /*
2080  * Add multicast source filter list to the interface list
2081  */
ip_mc_add_src(struct in_device * in_dev,__be32 * pmca,int sfmode,int sfcount,__be32 * psfsrc,int delta)2082 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2083 			 int sfcount, __be32 *psfsrc, int delta)
2084 {
2085 	struct ip_mc_list *pmc;
2086 	int	isexclude;
2087 	int	i, err;
2088 
2089 	if (!in_dev)
2090 		return -ENODEV;
2091 	rcu_read_lock();
2092 	for_each_pmc_rcu(in_dev, pmc) {
2093 		if (*pmca == pmc->multiaddr)
2094 			break;
2095 	}
2096 	if (!pmc) {
2097 		/* MCA not found?? bug */
2098 		rcu_read_unlock();
2099 		return -ESRCH;
2100 	}
2101 	spin_lock_bh(&pmc->lock);
2102 	rcu_read_unlock();
2103 
2104 #ifdef CONFIG_IP_MULTICAST
2105 	sf_markstate(pmc);
2106 #endif
2107 	isexclude = pmc->sfmode == MCAST_EXCLUDE;
2108 	if (!delta)
2109 		pmc->sfcount[sfmode]++;
2110 	err = 0;
2111 	for (i = 0; i < sfcount; i++) {
2112 		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2113 		if (err)
2114 			break;
2115 	}
2116 	if (err) {
2117 		int j;
2118 
2119 		if (!delta)
2120 			pmc->sfcount[sfmode]--;
2121 		for (j = 0; j < i; j++)
2122 			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2123 	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2124 #ifdef CONFIG_IP_MULTICAST
2125 		struct ip_sf_list *psf;
2126 		struct net *net = dev_net(pmc->interface->dev);
2127 		in_dev = pmc->interface;
2128 #endif
2129 
2130 		/* filter mode change */
2131 		if (pmc->sfcount[MCAST_EXCLUDE])
2132 			pmc->sfmode = MCAST_EXCLUDE;
2133 		else if (pmc->sfcount[MCAST_INCLUDE])
2134 			pmc->sfmode = MCAST_INCLUDE;
2135 #ifdef CONFIG_IP_MULTICAST
2136 		/* else no filters; keep old mode for reports */
2137 
2138 		pmc->crcount = in_dev->mr_qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);
2139 		WRITE_ONCE(in_dev->mr_ifc_count, pmc->crcount);
2140 		for (psf = pmc->sources; psf; psf = psf->sf_next)
2141 			psf->sf_crcount = 0;
2142 		igmp_ifc_event(in_dev);
2143 	} else if (sf_setstate(pmc)) {
2144 		igmp_ifc_event(in_dev);
2145 #endif
2146 	}
2147 	spin_unlock_bh(&pmc->lock);
2148 	return err;
2149 }
2150 
ip_mc_clear_src(struct ip_mc_list * pmc)2151 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2152 {
2153 	struct ip_sf_list *tomb, *sources;
2154 
2155 	spin_lock_bh(&pmc->lock);
2156 	tomb = pmc->tomb;
2157 	pmc->tomb = NULL;
2158 	sources = pmc->sources;
2159 	pmc->sources = NULL;
2160 	pmc->sfmode = MCAST_EXCLUDE;
2161 	pmc->sfcount[MCAST_INCLUDE] = 0;
2162 	pmc->sfcount[MCAST_EXCLUDE] = 1;
2163 	spin_unlock_bh(&pmc->lock);
2164 
2165 	ip_sf_list_clear_all(tomb);
2166 	ip_sf_list_clear_all(sources);
2167 }
2168 
2169 /* Join a multicast group
2170  */
__ip_mc_join_group(struct sock * sk,struct ip_mreqn * imr,unsigned int mode)2171 static int __ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr,
2172 			      unsigned int mode)
2173 {
2174 	__be32 addr = imr->imr_multiaddr.s_addr;
2175 	struct ip_mc_socklist *iml, *i;
2176 	struct in_device *in_dev;
2177 	struct inet_sock *inet = inet_sk(sk);
2178 	struct net *net = sock_net(sk);
2179 	int ifindex;
2180 	int count = 0;
2181 	int err;
2182 
2183 	ASSERT_RTNL();
2184 
2185 	if (!ipv4_is_multicast(addr))
2186 		return -EINVAL;
2187 
2188 	in_dev = ip_mc_find_dev(net, imr);
2189 
2190 	if (!in_dev) {
2191 		err = -ENODEV;
2192 		goto done;
2193 	}
2194 
2195 	err = -EADDRINUSE;
2196 	ifindex = imr->imr_ifindex;
2197 	for_each_pmc_rtnl(inet, i) {
2198 		if (i->multi.imr_multiaddr.s_addr == addr &&
2199 		    i->multi.imr_ifindex == ifindex)
2200 			goto done;
2201 		count++;
2202 	}
2203 	err = -ENOBUFS;
2204 	if (count >= READ_ONCE(net->ipv4.sysctl_igmp_max_memberships))
2205 		goto done;
2206 	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2207 	if (!iml)
2208 		goto done;
2209 
2210 	memcpy(&iml->multi, imr, sizeof(*imr));
2211 	iml->next_rcu = inet->mc_list;
2212 	iml->sflist = NULL;
2213 	iml->sfmode = mode;
2214 	rcu_assign_pointer(inet->mc_list, iml);
2215 	____ip_mc_inc_group(in_dev, addr, mode, GFP_KERNEL);
2216 	err = 0;
2217 done:
2218 	return err;
2219 }
2220 
2221 /* Join ASM (Any-Source Multicast) group
2222  */
ip_mc_join_group(struct sock * sk,struct ip_mreqn * imr)2223 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2224 {
2225 	return __ip_mc_join_group(sk, imr, MCAST_EXCLUDE);
2226 }
2227 EXPORT_SYMBOL(ip_mc_join_group);
2228 
2229 /* Join SSM (Source-Specific Multicast) group
2230  */
ip_mc_join_group_ssm(struct sock * sk,struct ip_mreqn * imr,unsigned int mode)2231 int ip_mc_join_group_ssm(struct sock *sk, struct ip_mreqn *imr,
2232 			 unsigned int mode)
2233 {
2234 	return __ip_mc_join_group(sk, imr, mode);
2235 }
2236 
ip_mc_leave_src(struct sock * sk,struct ip_mc_socklist * iml,struct in_device * in_dev)2237 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2238 			   struct in_device *in_dev)
2239 {
2240 	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2241 	int err;
2242 
2243 	if (!psf) {
2244 		/* any-source empty exclude case */
2245 		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2246 			iml->sfmode, 0, NULL, 0);
2247 	}
2248 	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2249 			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2250 	RCU_INIT_POINTER(iml->sflist, NULL);
2251 	/* decrease mem now to avoid the memleak warning */
2252 	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2253 	kfree_rcu(psf, rcu);
2254 	return err;
2255 }
2256 
ip_mc_leave_group(struct sock * sk,struct ip_mreqn * imr)2257 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2258 {
2259 	struct inet_sock *inet = inet_sk(sk);
2260 	struct ip_mc_socklist *iml;
2261 	struct ip_mc_socklist __rcu **imlp;
2262 	struct in_device *in_dev;
2263 	struct net *net = sock_net(sk);
2264 	__be32 group = imr->imr_multiaddr.s_addr;
2265 	u32 ifindex;
2266 	int ret = -EADDRNOTAVAIL;
2267 
2268 	ASSERT_RTNL();
2269 
2270 	in_dev = ip_mc_find_dev(net, imr);
2271 	if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2272 		ret = -ENODEV;
2273 		goto out;
2274 	}
2275 	ifindex = imr->imr_ifindex;
2276 	for (imlp = &inet->mc_list;
2277 	     (iml = rtnl_dereference(*imlp)) != NULL;
2278 	     imlp = &iml->next_rcu) {
2279 		if (iml->multi.imr_multiaddr.s_addr != group)
2280 			continue;
2281 		if (ifindex) {
2282 			if (iml->multi.imr_ifindex != ifindex)
2283 				continue;
2284 		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2285 				iml->multi.imr_address.s_addr)
2286 			continue;
2287 
2288 		(void) ip_mc_leave_src(sk, iml, in_dev);
2289 
2290 		*imlp = iml->next_rcu;
2291 
2292 		if (in_dev)
2293 			ip_mc_dec_group(in_dev, group);
2294 
2295 		/* decrease mem now to avoid the memleak warning */
2296 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2297 		kfree_rcu(iml, rcu);
2298 		return 0;
2299 	}
2300 out:
2301 	return ret;
2302 }
2303 EXPORT_SYMBOL(ip_mc_leave_group);
2304 
ip_mc_source(int add,int omode,struct sock * sk,struct ip_mreq_source * mreqs,int ifindex)2305 int ip_mc_source(int add, int omode, struct sock *sk, struct
2306 	ip_mreq_source *mreqs, int ifindex)
2307 {
2308 	int err;
2309 	struct ip_mreqn imr;
2310 	__be32 addr = mreqs->imr_multiaddr;
2311 	struct ip_mc_socklist *pmc;
2312 	struct in_device *in_dev = NULL;
2313 	struct inet_sock *inet = inet_sk(sk);
2314 	struct ip_sf_socklist *psl;
2315 	struct net *net = sock_net(sk);
2316 	int leavegroup = 0;
2317 	int i, j, rv;
2318 
2319 	if (!ipv4_is_multicast(addr))
2320 		return -EINVAL;
2321 
2322 	ASSERT_RTNL();
2323 
2324 	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2325 	imr.imr_address.s_addr = mreqs->imr_interface;
2326 	imr.imr_ifindex = ifindex;
2327 	in_dev = ip_mc_find_dev(net, &imr);
2328 
2329 	if (!in_dev) {
2330 		err = -ENODEV;
2331 		goto done;
2332 	}
2333 	err = -EADDRNOTAVAIL;
2334 
2335 	for_each_pmc_rtnl(inet, pmc) {
2336 		if ((pmc->multi.imr_multiaddr.s_addr ==
2337 		     imr.imr_multiaddr.s_addr) &&
2338 		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
2339 			break;
2340 	}
2341 	if (!pmc) {		/* must have a prior join */
2342 		err = -EINVAL;
2343 		goto done;
2344 	}
2345 	/* if a source filter was set, must be the same mode as before */
2346 	if (pmc->sflist) {
2347 		if (pmc->sfmode != omode) {
2348 			err = -EINVAL;
2349 			goto done;
2350 		}
2351 	} else if (pmc->sfmode != omode) {
2352 		/* allow mode switches for empty-set filters */
2353 		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2354 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2355 			NULL, 0);
2356 		pmc->sfmode = omode;
2357 	}
2358 
2359 	psl = rtnl_dereference(pmc->sflist);
2360 	if (!add) {
2361 		if (!psl)
2362 			goto done;	/* err = -EADDRNOTAVAIL */
2363 		rv = !0;
2364 		for (i = 0; i < psl->sl_count; i++) {
2365 			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2366 				sizeof(__be32));
2367 			if (rv == 0)
2368 				break;
2369 		}
2370 		if (rv)		/* source not found */
2371 			goto done;	/* err = -EADDRNOTAVAIL */
2372 
2373 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2374 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2375 			leavegroup = 1;
2376 			goto done;
2377 		}
2378 
2379 		/* update the interface filter */
2380 		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2381 			&mreqs->imr_sourceaddr, 1);
2382 
2383 		for (j = i+1; j < psl->sl_count; j++)
2384 			psl->sl_addr[j-1] = psl->sl_addr[j];
2385 		psl->sl_count--;
2386 		err = 0;
2387 		goto done;
2388 	}
2389 	/* else, add a new source to the filter */
2390 
2391 	if (psl && psl->sl_count >= READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
2392 		err = -ENOBUFS;
2393 		goto done;
2394 	}
2395 	if (!psl || psl->sl_count == psl->sl_max) {
2396 		struct ip_sf_socklist *newpsl;
2397 		int count = IP_SFBLOCK;
2398 
2399 		if (psl)
2400 			count += psl->sl_max;
2401 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2402 		if (!newpsl) {
2403 			err = -ENOBUFS;
2404 			goto done;
2405 		}
2406 		newpsl->sl_max = count;
2407 		newpsl->sl_count = count - IP_SFBLOCK;
2408 		if (psl) {
2409 			for (i = 0; i < psl->sl_count; i++)
2410 				newpsl->sl_addr[i] = psl->sl_addr[i];
2411 			/* decrease mem now to avoid the memleak warning */
2412 			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2413 		}
2414 		rcu_assign_pointer(pmc->sflist, newpsl);
2415 		if (psl)
2416 			kfree_rcu(psl, rcu);
2417 		psl = newpsl;
2418 	}
2419 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2420 	for (i = 0; i < psl->sl_count; i++) {
2421 		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2422 			sizeof(__be32));
2423 		if (rv == 0)
2424 			break;
2425 	}
2426 	if (rv == 0)		/* address already there is an error */
2427 		goto done;
2428 	for (j = psl->sl_count-1; j >= i; j--)
2429 		psl->sl_addr[j+1] = psl->sl_addr[j];
2430 	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2431 	psl->sl_count++;
2432 	err = 0;
2433 	/* update the interface list */
2434 	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2435 		&mreqs->imr_sourceaddr, 1);
2436 done:
2437 	if (leavegroup)
2438 		err = ip_mc_leave_group(sk, &imr);
2439 	return err;
2440 }
2441 
ip_mc_msfilter(struct sock * sk,struct ip_msfilter * msf,int ifindex)2442 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2443 {
2444 	int err = 0;
2445 	struct ip_mreqn	imr;
2446 	__be32 addr = msf->imsf_multiaddr;
2447 	struct ip_mc_socklist *pmc;
2448 	struct in_device *in_dev;
2449 	struct inet_sock *inet = inet_sk(sk);
2450 	struct ip_sf_socklist *newpsl, *psl;
2451 	struct net *net = sock_net(sk);
2452 	int leavegroup = 0;
2453 
2454 	if (!ipv4_is_multicast(addr))
2455 		return -EINVAL;
2456 	if (msf->imsf_fmode != MCAST_INCLUDE &&
2457 	    msf->imsf_fmode != MCAST_EXCLUDE)
2458 		return -EINVAL;
2459 
2460 	ASSERT_RTNL();
2461 
2462 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2463 	imr.imr_address.s_addr = msf->imsf_interface;
2464 	imr.imr_ifindex = ifindex;
2465 	in_dev = ip_mc_find_dev(net, &imr);
2466 
2467 	if (!in_dev) {
2468 		err = -ENODEV;
2469 		goto done;
2470 	}
2471 
2472 	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2473 	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2474 		leavegroup = 1;
2475 		goto done;
2476 	}
2477 
2478 	for_each_pmc_rtnl(inet, pmc) {
2479 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2480 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2481 			break;
2482 	}
2483 	if (!pmc) {		/* must have a prior join */
2484 		err = -EINVAL;
2485 		goto done;
2486 	}
2487 	if (msf->imsf_numsrc) {
2488 		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2489 							   GFP_KERNEL);
2490 		if (!newpsl) {
2491 			err = -ENOBUFS;
2492 			goto done;
2493 		}
2494 		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2495 		memcpy(newpsl->sl_addr, msf->imsf_slist,
2496 			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2497 		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2498 			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2499 		if (err) {
2500 			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2501 			goto done;
2502 		}
2503 	} else {
2504 		newpsl = NULL;
2505 		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2506 				     msf->imsf_fmode, 0, NULL, 0);
2507 	}
2508 	psl = rtnl_dereference(pmc->sflist);
2509 	if (psl) {
2510 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2511 			psl->sl_count, psl->sl_addr, 0);
2512 		/* decrease mem now to avoid the memleak warning */
2513 		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2514 	} else {
2515 		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2516 			0, NULL, 0);
2517 	}
2518 	rcu_assign_pointer(pmc->sflist, newpsl);
2519 	if (psl)
2520 		kfree_rcu(psl, rcu);
2521 	pmc->sfmode = msf->imsf_fmode;
2522 	err = 0;
2523 done:
2524 	if (leavegroup)
2525 		err = ip_mc_leave_group(sk, &imr);
2526 	return err;
2527 }
2528 
ip_mc_msfget(struct sock * sk,struct ip_msfilter * msf,struct ip_msfilter __user * optval,int __user * optlen)2529 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2530 	struct ip_msfilter __user *optval, int __user *optlen)
2531 {
2532 	int err, len, count, copycount;
2533 	struct ip_mreqn	imr;
2534 	__be32 addr = msf->imsf_multiaddr;
2535 	struct ip_mc_socklist *pmc;
2536 	struct in_device *in_dev;
2537 	struct inet_sock *inet = inet_sk(sk);
2538 	struct ip_sf_socklist *psl;
2539 	struct net *net = sock_net(sk);
2540 
2541 	ASSERT_RTNL();
2542 
2543 	if (!ipv4_is_multicast(addr))
2544 		return -EINVAL;
2545 
2546 	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2547 	imr.imr_address.s_addr = msf->imsf_interface;
2548 	imr.imr_ifindex = 0;
2549 	in_dev = ip_mc_find_dev(net, &imr);
2550 
2551 	if (!in_dev) {
2552 		err = -ENODEV;
2553 		goto done;
2554 	}
2555 	err = -EADDRNOTAVAIL;
2556 
2557 	for_each_pmc_rtnl(inet, pmc) {
2558 		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2559 		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2560 			break;
2561 	}
2562 	if (!pmc)		/* must have a prior join */
2563 		goto done;
2564 	msf->imsf_fmode = pmc->sfmode;
2565 	psl = rtnl_dereference(pmc->sflist);
2566 	if (!psl) {
2567 		len = 0;
2568 		count = 0;
2569 	} else {
2570 		count = psl->sl_count;
2571 	}
2572 	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2573 	len = copycount * sizeof(psl->sl_addr[0]);
2574 	msf->imsf_numsrc = count;
2575 	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2576 	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2577 		return -EFAULT;
2578 	}
2579 	if (len &&
2580 	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2581 		return -EFAULT;
2582 	return 0;
2583 done:
2584 	return err;
2585 }
2586 
ip_mc_gsfget(struct sock * sk,struct group_filter * gsf,struct sockaddr_storage __user * p)2587 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2588 	struct sockaddr_storage __user *p)
2589 {
2590 	int i, count, copycount;
2591 	struct sockaddr_in *psin;
2592 	__be32 addr;
2593 	struct ip_mc_socklist *pmc;
2594 	struct inet_sock *inet = inet_sk(sk);
2595 	struct ip_sf_socklist *psl;
2596 
2597 	ASSERT_RTNL();
2598 
2599 	psin = (struct sockaddr_in *)&gsf->gf_group;
2600 	if (psin->sin_family != AF_INET)
2601 		return -EINVAL;
2602 	addr = psin->sin_addr.s_addr;
2603 	if (!ipv4_is_multicast(addr))
2604 		return -EINVAL;
2605 
2606 	for_each_pmc_rtnl(inet, pmc) {
2607 		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2608 		    pmc->multi.imr_ifindex == gsf->gf_interface)
2609 			break;
2610 	}
2611 	if (!pmc)		/* must have a prior join */
2612 		return -EADDRNOTAVAIL;
2613 	gsf->gf_fmode = pmc->sfmode;
2614 	psl = rtnl_dereference(pmc->sflist);
2615 	count = psl ? psl->sl_count : 0;
2616 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2617 	gsf->gf_numsrc = count;
2618 	for (i = 0; i < copycount; i++, p++) {
2619 		struct sockaddr_storage ss;
2620 
2621 		psin = (struct sockaddr_in *)&ss;
2622 		memset(&ss, 0, sizeof(ss));
2623 		psin->sin_family = AF_INET;
2624 		psin->sin_addr.s_addr = psl->sl_addr[i];
2625 		if (copy_to_user(p, &ss, sizeof(ss)))
2626 			return -EFAULT;
2627 	}
2628 	return 0;
2629 }
2630 
2631 /*
2632  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2633  */
ip_mc_sf_allow(struct sock * sk,__be32 loc_addr,__be32 rmt_addr,int dif,int sdif)2634 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr,
2635 		   int dif, int sdif)
2636 {
2637 	struct inet_sock *inet = inet_sk(sk);
2638 	struct ip_mc_socklist *pmc;
2639 	struct ip_sf_socklist *psl;
2640 	int i;
2641 	int ret;
2642 
2643 	ret = 1;
2644 	if (!ipv4_is_multicast(loc_addr))
2645 		goto out;
2646 
2647 	rcu_read_lock();
2648 	for_each_pmc_rcu(inet, pmc) {
2649 		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2650 		    (pmc->multi.imr_ifindex == dif ||
2651 		     (sdif && pmc->multi.imr_ifindex == sdif)))
2652 			break;
2653 	}
2654 	ret = inet->mc_all;
2655 	if (!pmc)
2656 		goto unlock;
2657 	psl = rcu_dereference(pmc->sflist);
2658 	ret = (pmc->sfmode == MCAST_EXCLUDE);
2659 	if (!psl)
2660 		goto unlock;
2661 
2662 	for (i = 0; i < psl->sl_count; i++) {
2663 		if (psl->sl_addr[i] == rmt_addr)
2664 			break;
2665 	}
2666 	ret = 0;
2667 	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2668 		goto unlock;
2669 	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2670 		goto unlock;
2671 	ret = 1;
2672 unlock:
2673 	rcu_read_unlock();
2674 out:
2675 	return ret;
2676 }
2677 
2678 /*
2679  *	A socket is closing.
2680  */
2681 
ip_mc_drop_socket(struct sock * sk)2682 void ip_mc_drop_socket(struct sock *sk)
2683 {
2684 	struct inet_sock *inet = inet_sk(sk);
2685 	struct ip_mc_socklist *iml;
2686 	struct net *net = sock_net(sk);
2687 
2688 	if (!inet->mc_list)
2689 		return;
2690 
2691 	rtnl_lock();
2692 	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2693 		struct in_device *in_dev;
2694 
2695 		inet->mc_list = iml->next_rcu;
2696 		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2697 		(void) ip_mc_leave_src(sk, iml, in_dev);
2698 		if (in_dev)
2699 			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2700 		/* decrease mem now to avoid the memleak warning */
2701 		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2702 		kfree_rcu(iml, rcu);
2703 	}
2704 	rtnl_unlock();
2705 }
2706 
2707 /* called with rcu_read_lock() */
ip_check_mc_rcu(struct in_device * in_dev,__be32 mc_addr,__be32 src_addr,u8 proto)2708 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2709 {
2710 	struct ip_mc_list *im;
2711 	struct ip_mc_list __rcu **mc_hash;
2712 	struct ip_sf_list *psf;
2713 	int rv = 0;
2714 
2715 	mc_hash = rcu_dereference(in_dev->mc_hash);
2716 	if (mc_hash) {
2717 		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2718 
2719 		for (im = rcu_dereference(mc_hash[hash]);
2720 		     im != NULL;
2721 		     im = rcu_dereference(im->next_hash)) {
2722 			if (im->multiaddr == mc_addr)
2723 				break;
2724 		}
2725 	} else {
2726 		for_each_pmc_rcu(in_dev, im) {
2727 			if (im->multiaddr == mc_addr)
2728 				break;
2729 		}
2730 	}
2731 	if (im && proto == IPPROTO_IGMP) {
2732 		rv = 1;
2733 	} else if (im) {
2734 		if (src_addr) {
2735 			spin_lock_bh(&im->lock);
2736 			for (psf = im->sources; psf; psf = psf->sf_next) {
2737 				if (psf->sf_inaddr == src_addr)
2738 					break;
2739 			}
2740 			if (psf)
2741 				rv = psf->sf_count[MCAST_INCLUDE] ||
2742 					psf->sf_count[MCAST_EXCLUDE] !=
2743 					im->sfcount[MCAST_EXCLUDE];
2744 			else
2745 				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2746 			spin_unlock_bh(&im->lock);
2747 		} else
2748 			rv = 1; /* unspecified source; tentatively allow */
2749 	}
2750 	return rv;
2751 }
2752 
2753 #if defined(CONFIG_PROC_FS)
2754 struct igmp_mc_iter_state {
2755 	struct seq_net_private p;
2756 	struct net_device *dev;
2757 	struct in_device *in_dev;
2758 };
2759 
2760 #define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2761 
igmp_mc_get_first(struct seq_file * seq)2762 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2763 {
2764 	struct net *net = seq_file_net(seq);
2765 	struct ip_mc_list *im = NULL;
2766 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2767 
2768 	state->in_dev = NULL;
2769 	for_each_netdev_rcu(net, state->dev) {
2770 		struct in_device *in_dev;
2771 
2772 		in_dev = __in_dev_get_rcu(state->dev);
2773 		if (!in_dev)
2774 			continue;
2775 		im = rcu_dereference(in_dev->mc_list);
2776 		if (im) {
2777 			state->in_dev = in_dev;
2778 			break;
2779 		}
2780 	}
2781 	return im;
2782 }
2783 
igmp_mc_get_next(struct seq_file * seq,struct ip_mc_list * im)2784 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2785 {
2786 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2787 
2788 	im = rcu_dereference(im->next_rcu);
2789 	while (!im) {
2790 		state->dev = next_net_device_rcu(state->dev);
2791 		if (!state->dev) {
2792 			state->in_dev = NULL;
2793 			break;
2794 		}
2795 		state->in_dev = __in_dev_get_rcu(state->dev);
2796 		if (!state->in_dev)
2797 			continue;
2798 		im = rcu_dereference(state->in_dev->mc_list);
2799 	}
2800 	return im;
2801 }
2802 
igmp_mc_get_idx(struct seq_file * seq,loff_t pos)2803 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2804 {
2805 	struct ip_mc_list *im = igmp_mc_get_first(seq);
2806 	if (im)
2807 		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2808 			--pos;
2809 	return pos ? NULL : im;
2810 }
2811 
igmp_mc_seq_start(struct seq_file * seq,loff_t * pos)2812 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2813 	__acquires(rcu)
2814 {
2815 	rcu_read_lock();
2816 	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2817 }
2818 
igmp_mc_seq_next(struct seq_file * seq,void * v,loff_t * pos)2819 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2820 {
2821 	struct ip_mc_list *im;
2822 	if (v == SEQ_START_TOKEN)
2823 		im = igmp_mc_get_first(seq);
2824 	else
2825 		im = igmp_mc_get_next(seq, v);
2826 	++*pos;
2827 	return im;
2828 }
2829 
igmp_mc_seq_stop(struct seq_file * seq,void * v)2830 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2831 	__releases(rcu)
2832 {
2833 	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2834 
2835 	state->in_dev = NULL;
2836 	state->dev = NULL;
2837 	rcu_read_unlock();
2838 }
2839 
igmp_mc_seq_show(struct seq_file * seq,void * v)2840 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2841 {
2842 	if (v == SEQ_START_TOKEN)
2843 		seq_puts(seq,
2844 			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2845 	else {
2846 		struct ip_mc_list *im = (struct ip_mc_list *)v;
2847 		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2848 		char   *querier;
2849 		long delta;
2850 
2851 #ifdef CONFIG_IP_MULTICAST
2852 		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2853 			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2854 			  "V3";
2855 #else
2856 		querier = "NONE";
2857 #endif
2858 
2859 		if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2860 			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2861 				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2862 		}
2863 
2864 		delta = im->timer.expires - jiffies;
2865 		seq_printf(seq,
2866 			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2867 			   im->multiaddr, im->users,
2868 			   im->tm_running,
2869 			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2870 			   im->reporter);
2871 	}
2872 	return 0;
2873 }
2874 
2875 static const struct seq_operations igmp_mc_seq_ops = {
2876 	.start	=	igmp_mc_seq_start,
2877 	.next	=	igmp_mc_seq_next,
2878 	.stop	=	igmp_mc_seq_stop,
2879 	.show	=	igmp_mc_seq_show,
2880 };
2881 
2882 struct igmp_mcf_iter_state {
2883 	struct seq_net_private p;
2884 	struct net_device *dev;
2885 	struct in_device *idev;
2886 	struct ip_mc_list *im;
2887 };
2888 
2889 #define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2890 
igmp_mcf_get_first(struct seq_file * seq)2891 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2892 {
2893 	struct net *net = seq_file_net(seq);
2894 	struct ip_sf_list *psf = NULL;
2895 	struct ip_mc_list *im = NULL;
2896 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2897 
2898 	state->idev = NULL;
2899 	state->im = NULL;
2900 	for_each_netdev_rcu(net, state->dev) {
2901 		struct in_device *idev;
2902 		idev = __in_dev_get_rcu(state->dev);
2903 		if (unlikely(!idev))
2904 			continue;
2905 		im = rcu_dereference(idev->mc_list);
2906 		if (likely(im)) {
2907 			spin_lock_bh(&im->lock);
2908 			psf = im->sources;
2909 			if (likely(psf)) {
2910 				state->im = im;
2911 				state->idev = idev;
2912 				break;
2913 			}
2914 			spin_unlock_bh(&im->lock);
2915 		}
2916 	}
2917 	return psf;
2918 }
2919 
igmp_mcf_get_next(struct seq_file * seq,struct ip_sf_list * psf)2920 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2921 {
2922 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2923 
2924 	psf = psf->sf_next;
2925 	while (!psf) {
2926 		spin_unlock_bh(&state->im->lock);
2927 		state->im = state->im->next;
2928 		while (!state->im) {
2929 			state->dev = next_net_device_rcu(state->dev);
2930 			if (!state->dev) {
2931 				state->idev = NULL;
2932 				goto out;
2933 			}
2934 			state->idev = __in_dev_get_rcu(state->dev);
2935 			if (!state->idev)
2936 				continue;
2937 			state->im = rcu_dereference(state->idev->mc_list);
2938 		}
2939 		if (!state->im)
2940 			break;
2941 		spin_lock_bh(&state->im->lock);
2942 		psf = state->im->sources;
2943 	}
2944 out:
2945 	return psf;
2946 }
2947 
igmp_mcf_get_idx(struct seq_file * seq,loff_t pos)2948 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2949 {
2950 	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2951 	if (psf)
2952 		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2953 			--pos;
2954 	return pos ? NULL : psf;
2955 }
2956 
igmp_mcf_seq_start(struct seq_file * seq,loff_t * pos)2957 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2958 	__acquires(rcu)
2959 {
2960 	rcu_read_lock();
2961 	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2962 }
2963 
igmp_mcf_seq_next(struct seq_file * seq,void * v,loff_t * pos)2964 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2965 {
2966 	struct ip_sf_list *psf;
2967 	if (v == SEQ_START_TOKEN)
2968 		psf = igmp_mcf_get_first(seq);
2969 	else
2970 		psf = igmp_mcf_get_next(seq, v);
2971 	++*pos;
2972 	return psf;
2973 }
2974 
igmp_mcf_seq_stop(struct seq_file * seq,void * v)2975 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2976 	__releases(rcu)
2977 {
2978 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2979 	if (likely(state->im)) {
2980 		spin_unlock_bh(&state->im->lock);
2981 		state->im = NULL;
2982 	}
2983 	state->idev = NULL;
2984 	state->dev = NULL;
2985 	rcu_read_unlock();
2986 }
2987 
igmp_mcf_seq_show(struct seq_file * seq,void * v)2988 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2989 {
2990 	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2991 	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2992 
2993 	if (v == SEQ_START_TOKEN) {
2994 		seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2995 	} else {
2996 		seq_printf(seq,
2997 			   "%3d %6.6s 0x%08x "
2998 			   "0x%08x %6lu %6lu\n",
2999 			   state->dev->ifindex, state->dev->name,
3000 			   ntohl(state->im->multiaddr),
3001 			   ntohl(psf->sf_inaddr),
3002 			   psf->sf_count[MCAST_INCLUDE],
3003 			   psf->sf_count[MCAST_EXCLUDE]);
3004 	}
3005 	return 0;
3006 }
3007 
3008 static const struct seq_operations igmp_mcf_seq_ops = {
3009 	.start	=	igmp_mcf_seq_start,
3010 	.next	=	igmp_mcf_seq_next,
3011 	.stop	=	igmp_mcf_seq_stop,
3012 	.show	=	igmp_mcf_seq_show,
3013 };
3014 
igmp_net_init(struct net * net)3015 static int __net_init igmp_net_init(struct net *net)
3016 {
3017 	struct proc_dir_entry *pde;
3018 	int err;
3019 
3020 	pde = proc_create_net("igmp", 0444, net->proc_net, &igmp_mc_seq_ops,
3021 			sizeof(struct igmp_mc_iter_state));
3022 	if (!pde)
3023 		goto out_igmp;
3024 	pde = proc_create_net("mcfilter", 0444, net->proc_net,
3025 			&igmp_mcf_seq_ops, sizeof(struct igmp_mcf_iter_state));
3026 	if (!pde)
3027 		goto out_mcfilter;
3028 	err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
3029 				   SOCK_DGRAM, 0, net);
3030 	if (err < 0) {
3031 		pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
3032 		       err);
3033 		goto out_sock;
3034 	}
3035 
3036 	return 0;
3037 
3038 out_sock:
3039 	remove_proc_entry("mcfilter", net->proc_net);
3040 out_mcfilter:
3041 	remove_proc_entry("igmp", net->proc_net);
3042 out_igmp:
3043 	return -ENOMEM;
3044 }
3045 
igmp_net_exit(struct net * net)3046 static void __net_exit igmp_net_exit(struct net *net)
3047 {
3048 	remove_proc_entry("mcfilter", net->proc_net);
3049 	remove_proc_entry("igmp", net->proc_net);
3050 	inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
3051 }
3052 
3053 static struct pernet_operations igmp_net_ops = {
3054 	.init = igmp_net_init,
3055 	.exit = igmp_net_exit,
3056 };
3057 #endif
3058 
igmp_netdev_event(struct notifier_block * this,unsigned long event,void * ptr)3059 static int igmp_netdev_event(struct notifier_block *this,
3060 			     unsigned long event, void *ptr)
3061 {
3062 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3063 	struct in_device *in_dev;
3064 
3065 	switch (event) {
3066 	case NETDEV_RESEND_IGMP:
3067 		in_dev = __in_dev_get_rtnl(dev);
3068 		if (in_dev)
3069 			ip_mc_rejoin_groups(in_dev);
3070 		break;
3071 	default:
3072 		break;
3073 	}
3074 	return NOTIFY_DONE;
3075 }
3076 
3077 static struct notifier_block igmp_notifier = {
3078 	.notifier_call = igmp_netdev_event,
3079 };
3080 
igmp_mc_init(void)3081 int __init igmp_mc_init(void)
3082 {
3083 #if defined(CONFIG_PROC_FS)
3084 	int err;
3085 
3086 	err = register_pernet_subsys(&igmp_net_ops);
3087 	if (err)
3088 		return err;
3089 	err = register_netdevice_notifier(&igmp_notifier);
3090 	if (err)
3091 		goto reg_notif_fail;
3092 	return 0;
3093 
3094 reg_notif_fail:
3095 	unregister_pernet_subsys(&igmp_net_ops);
3096 	return err;
3097 #else
3098 	return register_netdevice_notifier(&igmp_notifier);
3099 #endif
3100 }
3101