Linux Audio

Check our new training course

Loading...
v5.14.15
   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
 133static 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
 156static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im,
 157			      gfp_t gfp);
 158static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 159static void igmpv3_clear_delrec(struct in_device *in_dev);
 160static int sf_setstate(struct ip_mc_list *pmc);
 161static void sf_markstate(struct ip_mc_list *pmc);
 162#endif
 163static void ip_mc_clear_src(struct ip_mc_list *pmc);
 164static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 165			 int sfcount, __be32 *psfsrc, int delta);
 166
 167static 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
 185static 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
 202static 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 */
 214static 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 (!mod_timer(&im->timer, jiffies+tv+2))
 220		refcount_inc(&im->refcnt);
 221}
 222
 223static void igmp_gq_start_timer(struct in_device *in_dev)
 224{
 225	int tv = prandom_u32() % in_dev->mr_maxdelay;
 226	unsigned long exp = jiffies + tv + 2;
 227
 228	if (in_dev->mr_gq_running &&
 229	    time_after_eq(exp, (in_dev->mr_gq_timer).expires))
 230		return;
 231
 232	in_dev->mr_gq_running = 1;
 233	if (!mod_timer(&in_dev->mr_gq_timer, exp))
 234		in_dev_hold(in_dev);
 235}
 236
 237static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 238{
 239	int tv = prandom_u32() % delay;
 240
 241	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 242		in_dev_hold(in_dev);
 243}
 244
 245static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 246{
 247	spin_lock_bh(&im->lock);
 248	im->unsolicit_count = 0;
 249	if (del_timer(&im->timer)) {
 250		if ((long)(im->timer.expires-jiffies) < max_delay) {
 251			add_timer(&im->timer);
 252			im->tm_running = 1;
 253			spin_unlock_bh(&im->lock);
 254			return;
 255		}
 256		refcount_dec(&im->refcnt);
 257	}
 258	igmp_start_timer(im, max_delay);
 259	spin_unlock_bh(&im->lock);
 260}
 261
 262
 263/*
 264 *	Send an IGMP report.
 265 */
 266
 267#define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 268
 269
 270static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 271	int gdeleted, int sdeleted)
 272{
 273	switch (type) {
 274	case IGMPV3_MODE_IS_INCLUDE:
 275	case IGMPV3_MODE_IS_EXCLUDE:
 276		if (gdeleted || sdeleted)
 277			return 0;
 278		if (!(pmc->gsquery && !psf->sf_gsresp)) {
 279			if (pmc->sfmode == MCAST_INCLUDE)
 280				return 1;
 281			/* don't include if this source is excluded
 282			 * in all filters
 283			 */
 284			if (psf->sf_count[MCAST_INCLUDE])
 285				return type == IGMPV3_MODE_IS_INCLUDE;
 286			return pmc->sfcount[MCAST_EXCLUDE] ==
 287				psf->sf_count[MCAST_EXCLUDE];
 288		}
 289		return 0;
 290	case IGMPV3_CHANGE_TO_INCLUDE:
 291		if (gdeleted || sdeleted)
 292			return 0;
 293		return psf->sf_count[MCAST_INCLUDE] != 0;
 294	case IGMPV3_CHANGE_TO_EXCLUDE:
 295		if (gdeleted || sdeleted)
 296			return 0;
 297		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 298		    psf->sf_count[MCAST_INCLUDE])
 299			return 0;
 300		return pmc->sfcount[MCAST_EXCLUDE] ==
 301			psf->sf_count[MCAST_EXCLUDE];
 302	case IGMPV3_ALLOW_NEW_SOURCES:
 303		if (gdeleted || !psf->sf_crcount)
 304			return 0;
 305		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 306	case IGMPV3_BLOCK_OLD_SOURCES:
 307		if (pmc->sfmode == MCAST_INCLUDE)
 308			return gdeleted || (psf->sf_crcount && sdeleted);
 309		return psf->sf_crcount && !gdeleted && !sdeleted;
 310	}
 311	return 0;
 312}
 313
 314static int
 315igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 316{
 317	struct ip_sf_list *psf;
 318	int scount = 0;
 319
 320	for (psf = pmc->sources; psf; psf = psf->sf_next) {
 321		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 322			continue;
 323		scount++;
 324	}
 325	return scount;
 326}
 327
 328/* source address selection per RFC 3376 section 4.2.13 */
 329static __be32 igmpv3_get_srcaddr(struct net_device *dev,
 330				 const struct flowi4 *fl4)
 331{
 332	struct in_device *in_dev = __in_dev_get_rcu(dev);
 333	const struct in_ifaddr *ifa;
 334
 335	if (!in_dev)
 336		return htonl(INADDR_ANY);
 337
 338	in_dev_for_each_ifa_rcu(ifa, in_dev) {
 339		if (fl4->saddr == ifa->ifa_local)
 340			return fl4->saddr;
 341	}
 342
 343	return htonl(INADDR_ANY);
 344}
 345
 346static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
 347{
 348	struct sk_buff *skb;
 349	struct rtable *rt;
 350	struct iphdr *pip;
 351	struct igmpv3_report *pig;
 352	struct net *net = dev_net(dev);
 353	struct flowi4 fl4;
 354	int hlen = LL_RESERVED_SPACE(dev);
 355	int tlen = dev->needed_tailroom;
 356	unsigned int size = mtu;
 357
 358	while (1) {
 359		skb = alloc_skb(size + hlen + tlen,
 360				GFP_ATOMIC | __GFP_NOWARN);
 361		if (skb)
 362			break;
 363		size >>= 1;
 364		if (size < 256)
 365			return NULL;
 366	}
 367	skb->priority = TC_PRIO_CONTROL;
 368
 369	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 370				   0, 0,
 371				   IPPROTO_IGMP, 0, dev->ifindex);
 372	if (IS_ERR(rt)) {
 373		kfree_skb(skb);
 374		return NULL;
 375	}
 376
 377	skb_dst_set(skb, &rt->dst);
 378	skb->dev = dev;
 379
 380	skb_reserve(skb, hlen);
 381	skb_tailroom_reserve(skb, mtu, tlen);
 382
 383	skb_reset_network_header(skb);
 384	pip = ip_hdr(skb);
 385	skb_put(skb, sizeof(struct iphdr) + 4);
 386
 387	pip->version  = 4;
 388	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 389	pip->tos      = 0xc0;
 390	pip->frag_off = htons(IP_DF);
 391	pip->ttl      = 1;
 392	pip->daddr    = fl4.daddr;
 393
 394	rcu_read_lock();
 395	pip->saddr    = igmpv3_get_srcaddr(dev, &fl4);
 396	rcu_read_unlock();
 397
 398	pip->protocol = IPPROTO_IGMP;
 399	pip->tot_len  = 0;	/* filled in later */
 400	ip_select_ident(net, skb, NULL);
 401	((u8 *)&pip[1])[0] = IPOPT_RA;
 402	((u8 *)&pip[1])[1] = 4;
 403	((u8 *)&pip[1])[2] = 0;
 404	((u8 *)&pip[1])[3] = 0;
 405
 406	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 407	skb_put(skb, sizeof(*pig));
 408	pig = igmpv3_report_hdr(skb);
 409	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 410	pig->resv1 = 0;
 411	pig->csum = 0;
 412	pig->resv2 = 0;
 413	pig->ngrec = 0;
 414	return skb;
 415}
 416
 417static int igmpv3_sendpack(struct sk_buff *skb)
 418{
 419	struct igmphdr *pig = igmp_hdr(skb);
 420	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
 421
 422	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 423
 424	return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
 425}
 426
 427static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 428{
 429	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 430}
 431
 432static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 433	int type, struct igmpv3_grec **ppgr, unsigned int mtu)
 434{
 435	struct net_device *dev = pmc->interface->dev;
 436	struct igmpv3_report *pih;
 437	struct igmpv3_grec *pgr;
 438
 439	if (!skb) {
 440		skb = igmpv3_newpack(dev, mtu);
 441		if (!skb)
 442			return NULL;
 443	}
 444	pgr = skb_put(skb, sizeof(struct igmpv3_grec));
 445	pgr->grec_type = type;
 446	pgr->grec_auxwords = 0;
 447	pgr->grec_nsrcs = 0;
 448	pgr->grec_mca = pmc->multiaddr;
 449	pih = igmpv3_report_hdr(skb);
 450	pih->ngrec = htons(ntohs(pih->ngrec)+1);
 451	*ppgr = pgr;
 452	return skb;
 453}
 454
 455#define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
 456
 457static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 458	int type, int gdeleted, int sdeleted)
 459{
 460	struct net_device *dev = pmc->interface->dev;
 461	struct net *net = dev_net(dev);
 462	struct igmpv3_report *pih;
 463	struct igmpv3_grec *pgr = NULL;
 464	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 465	int scount, stotal, first, isquery, truncate;
 466	unsigned int mtu;
 467
 468	if (pmc->multiaddr == IGMP_ALL_HOSTS)
 469		return skb;
 470	if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
 471		return skb;
 472
 473	mtu = READ_ONCE(dev->mtu);
 474	if (mtu < IPV4_MIN_MTU)
 475		return skb;
 476
 477	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 478		  type == IGMPV3_MODE_IS_EXCLUDE;
 479	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 480		    type == IGMPV3_CHANGE_TO_EXCLUDE;
 481
 482	stotal = scount = 0;
 483
 484	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 485
 486	if (!*psf_list)
 487		goto empty_source;
 488
 489	pih = skb ? igmpv3_report_hdr(skb) : NULL;
 490
 491	/* EX and TO_EX get a fresh packet, if needed */
 492	if (truncate) {
 493		if (pih && pih->ngrec &&
 494		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 495			if (skb)
 496				igmpv3_sendpack(skb);
 497			skb = igmpv3_newpack(dev, mtu);
 498		}
 499	}
 500	first = 1;
 501	psf_prev = NULL;
 502	for (psf = *psf_list; psf; psf = psf_next) {
 503		__be32 *psrc;
 504
 505		psf_next = psf->sf_next;
 506
 507		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 508			psf_prev = psf;
 509			continue;
 510		}
 511
 512		/* Based on RFC3376 5.1. Should not send source-list change
 513		 * records when there is a filter mode change.
 514		 */
 515		if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) ||
 516		     (!gdeleted && pmc->crcount)) &&
 517		    (type == IGMPV3_ALLOW_NEW_SOURCES ||
 518		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount)
 519			goto decrease_sf_crcount;
 520
 521		/* clear marks on query responses */
 522		if (isquery)
 523			psf->sf_gsresp = 0;
 524
 525		if (AVAILABLE(skb) < sizeof(__be32) +
 526		    first*sizeof(struct igmpv3_grec)) {
 527			if (truncate && !first)
 528				break;	 /* truncate these */
 529			if (pgr)
 530				pgr->grec_nsrcs = htons(scount);
 531			if (skb)
 532				igmpv3_sendpack(skb);
 533			skb = igmpv3_newpack(dev, mtu);
 534			first = 1;
 535			scount = 0;
 536		}
 537		if (first) {
 538			skb = add_grhead(skb, pmc, type, &pgr, mtu);
 539			first = 0;
 540		}
 541		if (!skb)
 542			return NULL;
 543		psrc = skb_put(skb, sizeof(__be32));
 544		*psrc = psf->sf_inaddr;
 545		scount++; stotal++;
 546		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 547		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 548decrease_sf_crcount:
 549			psf->sf_crcount--;
 550			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 551				if (psf_prev)
 552					psf_prev->sf_next = psf->sf_next;
 553				else
 554					*psf_list = psf->sf_next;
 555				kfree(psf);
 556				continue;
 557			}
 558		}
 559		psf_prev = psf;
 560	}
 561
 562empty_source:
 563	if (!stotal) {
 564		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 565		    type == IGMPV3_BLOCK_OLD_SOURCES)
 566			return skb;
 567		if (pmc->crcount || isquery) {
 568			/* make sure we have room for group header */
 569			if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
 570				igmpv3_sendpack(skb);
 571				skb = NULL; /* add_grhead will get a new one */
 572			}
 573			skb = add_grhead(skb, pmc, type, &pgr, mtu);
 574		}
 575	}
 576	if (pgr)
 577		pgr->grec_nsrcs = htons(scount);
 578
 579	if (isquery)
 580		pmc->gsquery = 0;	/* clear query state on report */
 581	return skb;
 582}
 583
 584static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 585{
 586	struct sk_buff *skb = NULL;
 587	struct net *net = dev_net(in_dev->dev);
 588	int type;
 589
 590	if (!pmc) {
 591		rcu_read_lock();
 592		for_each_pmc_rcu(in_dev, pmc) {
 593			if (pmc->multiaddr == IGMP_ALL_HOSTS)
 594				continue;
 595			if (ipv4_is_local_multicast(pmc->multiaddr) &&
 596			     !net->ipv4.sysctl_igmp_llm_reports)
 597				continue;
 598			spin_lock_bh(&pmc->lock);
 599			if (pmc->sfcount[MCAST_EXCLUDE])
 600				type = IGMPV3_MODE_IS_EXCLUDE;
 601			else
 602				type = IGMPV3_MODE_IS_INCLUDE;
 603			skb = add_grec(skb, pmc, type, 0, 0);
 604			spin_unlock_bh(&pmc->lock);
 605		}
 606		rcu_read_unlock();
 607	} else {
 608		spin_lock_bh(&pmc->lock);
 609		if (pmc->sfcount[MCAST_EXCLUDE])
 610			type = IGMPV3_MODE_IS_EXCLUDE;
 611		else
 612			type = IGMPV3_MODE_IS_INCLUDE;
 613		skb = add_grec(skb, pmc, type, 0, 0);
 614		spin_unlock_bh(&pmc->lock);
 615	}
 616	if (!skb)
 617		return 0;
 618	return igmpv3_sendpack(skb);
 619}
 620
 621/*
 622 * remove zero-count source records from a source filter list
 623 */
 624static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 625{
 626	struct ip_sf_list *psf_prev, *psf_next, *psf;
 627
 628	psf_prev = NULL;
 629	for (psf = *ppsf; psf; psf = psf_next) {
 630		psf_next = psf->sf_next;
 631		if (psf->sf_crcount == 0) {
 632			if (psf_prev)
 633				psf_prev->sf_next = psf->sf_next;
 634			else
 635				*ppsf = psf->sf_next;
 636			kfree(psf);
 637		} else
 638			psf_prev = psf;
 639	}
 640}
 641
 642static void kfree_pmc(struct ip_mc_list *pmc)
 643{
 644	ip_sf_list_clear_all(pmc->sources);
 645	ip_sf_list_clear_all(pmc->tomb);
 646	kfree(pmc);
 647}
 648
 649static void igmpv3_send_cr(struct in_device *in_dev)
 650{
 651	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 652	struct sk_buff *skb = NULL;
 653	int type, dtype;
 654
 655	rcu_read_lock();
 656	spin_lock_bh(&in_dev->mc_tomb_lock);
 657
 658	/* deleted MCA's */
 659	pmc_prev = NULL;
 660	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
 661		pmc_next = pmc->next;
 662		if (pmc->sfmode == MCAST_INCLUDE) {
 663			type = IGMPV3_BLOCK_OLD_SOURCES;
 664			dtype = IGMPV3_BLOCK_OLD_SOURCES;
 665			skb = add_grec(skb, pmc, type, 1, 0);
 666			skb = add_grec(skb, pmc, dtype, 1, 1);
 667		}
 668		if (pmc->crcount) {
 669			if (pmc->sfmode == MCAST_EXCLUDE) {
 670				type = IGMPV3_CHANGE_TO_INCLUDE;
 671				skb = add_grec(skb, pmc, type, 1, 0);
 672			}
 673			pmc->crcount--;
 674			if (pmc->crcount == 0) {
 675				igmpv3_clear_zeros(&pmc->tomb);
 676				igmpv3_clear_zeros(&pmc->sources);
 677			}
 678		}
 679		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 680			if (pmc_prev)
 681				pmc_prev->next = pmc_next;
 682			else
 683				in_dev->mc_tomb = pmc_next;
 684			in_dev_put(pmc->interface);
 685			kfree_pmc(pmc);
 686		} else
 687			pmc_prev = pmc;
 688	}
 689	spin_unlock_bh(&in_dev->mc_tomb_lock);
 690
 691	/* change recs */
 692	for_each_pmc_rcu(in_dev, pmc) {
 693		spin_lock_bh(&pmc->lock);
 694		if (pmc->sfcount[MCAST_EXCLUDE]) {
 695			type = IGMPV3_BLOCK_OLD_SOURCES;
 696			dtype = IGMPV3_ALLOW_NEW_SOURCES;
 697		} else {
 698			type = IGMPV3_ALLOW_NEW_SOURCES;
 699			dtype = IGMPV3_BLOCK_OLD_SOURCES;
 700		}
 701		skb = add_grec(skb, pmc, type, 0, 0);
 702		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
 703
 704		/* filter mode changes */
 705		if (pmc->crcount) {
 706			if (pmc->sfmode == MCAST_EXCLUDE)
 707				type = IGMPV3_CHANGE_TO_EXCLUDE;
 708			else
 709				type = IGMPV3_CHANGE_TO_INCLUDE;
 710			skb = add_grec(skb, pmc, type, 0, 0);
 711			pmc->crcount--;
 712		}
 713		spin_unlock_bh(&pmc->lock);
 714	}
 715	rcu_read_unlock();
 716
 717	if (!skb)
 718		return;
 719	(void) igmpv3_sendpack(skb);
 720}
 721
 722static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 723	int type)
 724{
 725	struct sk_buff *skb;
 726	struct iphdr *iph;
 727	struct igmphdr *ih;
 728	struct rtable *rt;
 729	struct net_device *dev = in_dev->dev;
 730	struct net *net = dev_net(dev);
 731	__be32	group = pmc ? pmc->multiaddr : 0;
 732	struct flowi4 fl4;
 733	__be32	dst;
 734	int hlen, tlen;
 735
 736	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 737		return igmpv3_send_report(in_dev, pmc);
 738
 739	if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
 740		return 0;
 741
 742	if (type == IGMP_HOST_LEAVE_MESSAGE)
 743		dst = IGMP_ALL_ROUTER;
 744	else
 745		dst = group;
 746
 747	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 748				   0, 0,
 749				   IPPROTO_IGMP, 0, dev->ifindex);
 750	if (IS_ERR(rt))
 751		return -1;
 752
 753	hlen = LL_RESERVED_SPACE(dev);
 754	tlen = dev->needed_tailroom;
 755	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 756	if (!skb) {
 757		ip_rt_put(rt);
 758		return -1;
 759	}
 760	skb->priority = TC_PRIO_CONTROL;
 761
 762	skb_dst_set(skb, &rt->dst);
 763
 764	skb_reserve(skb, hlen);
 765
 766	skb_reset_network_header(skb);
 767	iph = ip_hdr(skb);
 768	skb_put(skb, sizeof(struct iphdr) + 4);
 769
 770	iph->version  = 4;
 771	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 772	iph->tos      = 0xc0;
 773	iph->frag_off = htons(IP_DF);
 774	iph->ttl      = 1;
 775	iph->daddr    = dst;
 776	iph->saddr    = fl4.saddr;
 777	iph->protocol = IPPROTO_IGMP;
 778	ip_select_ident(net, skb, NULL);
 779	((u8 *)&iph[1])[0] = IPOPT_RA;
 780	((u8 *)&iph[1])[1] = 4;
 781	((u8 *)&iph[1])[2] = 0;
 782	((u8 *)&iph[1])[3] = 0;
 783
 784	ih = skb_put(skb, sizeof(struct igmphdr));
 785	ih->type = type;
 786	ih->code = 0;
 787	ih->csum = 0;
 788	ih->group = group;
 789	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 790
 791	return ip_local_out(net, skb->sk, skb);
 792}
 793
 794static void igmp_gq_timer_expire(struct timer_list *t)
 795{
 796	struct in_device *in_dev = from_timer(in_dev, t, mr_gq_timer);
 797
 798	in_dev->mr_gq_running = 0;
 799	igmpv3_send_report(in_dev, NULL);
 800	in_dev_put(in_dev);
 801}
 802
 803static void igmp_ifc_timer_expire(struct timer_list *t)
 804{
 805	struct in_device *in_dev = from_timer(in_dev, t, mr_ifc_timer);
 806	u32 mr_ifc_count;
 807
 808	igmpv3_send_cr(in_dev);
 809restart:
 810	mr_ifc_count = READ_ONCE(in_dev->mr_ifc_count);
 811
 812	if (mr_ifc_count) {
 813		if (cmpxchg(&in_dev->mr_ifc_count,
 814			    mr_ifc_count,
 815			    mr_ifc_count - 1) != mr_ifc_count)
 816			goto restart;
 817		igmp_ifc_start_timer(in_dev,
 818				     unsolicited_report_interval(in_dev));
 819	}
 820	in_dev_put(in_dev);
 821}
 822
 823static void igmp_ifc_event(struct in_device *in_dev)
 824{
 825	struct net *net = dev_net(in_dev->dev);
 826	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 827		return;
 828	WRITE_ONCE(in_dev->mr_ifc_count, in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv);
 829	igmp_ifc_start_timer(in_dev, 1);
 830}
 831
 832
 833static void igmp_timer_expire(struct timer_list *t)
 834{
 835	struct ip_mc_list *im = from_timer(im, t, timer);
 836	struct in_device *in_dev = im->interface;
 837
 838	spin_lock(&im->lock);
 839	im->tm_running = 0;
 840
 841	if (im->unsolicit_count && --im->unsolicit_count)
 
 842		igmp_start_timer(im, unsolicited_report_interval(in_dev));
 843
 844	im->reporter = 1;
 845	spin_unlock(&im->lock);
 846
 847	if (IGMP_V1_SEEN(in_dev))
 848		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 849	else if (IGMP_V2_SEEN(in_dev))
 850		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 851	else
 852		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 853
 854	ip_ma_put(im);
 855}
 856
 857/* mark EXCLUDE-mode sources */
 858static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 859{
 860	struct ip_sf_list *psf;
 861	int i, scount;
 862
 863	scount = 0;
 864	for (psf = pmc->sources; psf; psf = psf->sf_next) {
 865		if (scount == nsrcs)
 866			break;
 867		for (i = 0; i < nsrcs; i++) {
 868			/* skip inactive filters */
 869			if (psf->sf_count[MCAST_INCLUDE] ||
 870			    pmc->sfcount[MCAST_EXCLUDE] !=
 871			    psf->sf_count[MCAST_EXCLUDE])
 872				break;
 873			if (srcs[i] == psf->sf_inaddr) {
 874				scount++;
 875				break;
 876			}
 877		}
 878	}
 879	pmc->gsquery = 0;
 880	if (scount == nsrcs)	/* all sources excluded */
 881		return 0;
 882	return 1;
 883}
 884
 885static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 886{
 887	struct ip_sf_list *psf;
 888	int i, scount;
 889
 890	if (pmc->sfmode == MCAST_EXCLUDE)
 891		return igmp_xmarksources(pmc, nsrcs, srcs);
 892
 893	/* mark INCLUDE-mode sources */
 894	scount = 0;
 895	for (psf = pmc->sources; psf; psf = psf->sf_next) {
 896		if (scount == nsrcs)
 897			break;
 898		for (i = 0; i < nsrcs; i++)
 899			if (srcs[i] == psf->sf_inaddr) {
 900				psf->sf_gsresp = 1;
 901				scount++;
 902				break;
 903			}
 904	}
 905	if (!scount) {
 906		pmc->gsquery = 0;
 907		return 0;
 908	}
 909	pmc->gsquery = 1;
 910	return 1;
 911}
 912
 913/* return true if packet was dropped */
 914static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
 915{
 916	struct ip_mc_list *im;
 917	struct net *net = dev_net(in_dev->dev);
 918
 919	/* Timers are only set for non-local groups */
 920
 921	if (group == IGMP_ALL_HOSTS)
 922		return false;
 923	if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
 924		return false;
 925
 926	rcu_read_lock();
 927	for_each_pmc_rcu(in_dev, im) {
 928		if (im->multiaddr == group) {
 929			igmp_stop_timer(im);
 930			break;
 931		}
 932	}
 933	rcu_read_unlock();
 934	return false;
 935}
 936
 937/* return true if packet was dropped */
 938static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 939	int len)
 940{
 941	struct igmphdr 		*ih = igmp_hdr(skb);
 942	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 943	struct ip_mc_list	*im;
 944	__be32			group = ih->group;
 945	int			max_delay;
 946	int			mark = 0;
 947	struct net		*net = dev_net(in_dev->dev);
 948
 949
 950	if (len == 8) {
 951		if (ih->code == 0) {
 952			/* Alas, old v1 router presents here. */
 953
 954			max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 955			in_dev->mr_v1_seen = jiffies +
 956				(in_dev->mr_qrv * in_dev->mr_qi) +
 957				in_dev->mr_qri;
 958			group = 0;
 959		} else {
 960			/* v2 router present */
 961			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 962			in_dev->mr_v2_seen = jiffies +
 963				(in_dev->mr_qrv * in_dev->mr_qi) +
 964				in_dev->mr_qri;
 965		}
 966		/* cancel the interface change timer */
 967		WRITE_ONCE(in_dev->mr_ifc_count, 0);
 968		if (del_timer(&in_dev->mr_ifc_timer))
 969			__in_dev_put(in_dev);
 970		/* clear deleted report items */
 971		igmpv3_clear_delrec(in_dev);
 972	} else if (len < 12) {
 973		return true;	/* ignore bogus packet; freed by caller */
 974	} else if (IGMP_V1_SEEN(in_dev)) {
 975		/* This is a v3 query with v1 queriers present */
 976		max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 977		group = 0;
 978	} else if (IGMP_V2_SEEN(in_dev)) {
 979		/* this is a v3 query with v2 queriers present;
 980		 * Interpretation of the max_delay code is problematic here.
 981		 * A real v2 host would use ih_code directly, while v3 has a
 982		 * different encoding. We use the v3 encoding as more likely
 983		 * to be intended in a v3 query.
 984		 */
 985		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 986		if (!max_delay)
 987			max_delay = 1;	/* can't mod w/ 0 */
 988	} else { /* v3 */
 989		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 990			return true;
 991
 992		ih3 = igmpv3_query_hdr(skb);
 993		if (ih3->nsrcs) {
 994			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 995					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
 996				return true;
 997			ih3 = igmpv3_query_hdr(skb);
 998		}
 999
1000		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
1001		if (!max_delay)
1002			max_delay = 1;	/* can't mod w/ 0 */
1003		in_dev->mr_maxdelay = max_delay;
1004
1005		/* RFC3376, 4.1.6. QRV and 4.1.7. QQIC, when the most recently
1006		 * received value was zero, use the default or statically
1007		 * configured value.
1008		 */
1009		in_dev->mr_qrv = ih3->qrv ?: net->ipv4.sysctl_igmp_qrv;
1010		in_dev->mr_qi = IGMPV3_QQIC(ih3->qqic)*HZ ?: IGMP_QUERY_INTERVAL;
1011
1012		/* RFC3376, 8.3. Query Response Interval:
1013		 * The number of seconds represented by the [Query Response
1014		 * Interval] must be less than the [Query Interval].
1015		 */
1016		if (in_dev->mr_qri >= in_dev->mr_qi)
1017			in_dev->mr_qri = (in_dev->mr_qi/HZ - 1)*HZ;
1018
1019		if (!group) { /* general query */
1020			if (ih3->nsrcs)
1021				return true;	/* no sources allowed */
1022			igmp_gq_start_timer(in_dev);
1023			return false;
1024		}
1025		/* mark sources to include, if group & source-specific */
1026		mark = ih3->nsrcs != 0;
1027	}
1028
1029	/*
1030	 * - Start the timers in all of our membership records
1031	 *   that the query applies to for the interface on
1032	 *   which the query arrived excl. those that belong
1033	 *   to a "local" group (224.0.0.X)
1034	 * - For timers already running check if they need to
1035	 *   be reset.
1036	 * - Use the igmp->igmp_code field as the maximum
1037	 *   delay possible
1038	 */
1039	rcu_read_lock();
1040	for_each_pmc_rcu(in_dev, im) {
1041		int changed;
1042
1043		if (group && group != im->multiaddr)
1044			continue;
1045		if (im->multiaddr == IGMP_ALL_HOSTS)
1046			continue;
1047		if (ipv4_is_local_multicast(im->multiaddr) &&
1048		    !net->ipv4.sysctl_igmp_llm_reports)
1049			continue;
1050		spin_lock_bh(&im->lock);
1051		if (im->tm_running)
1052			im->gsquery = im->gsquery && mark;
1053		else
1054			im->gsquery = mark;
1055		changed = !im->gsquery ||
1056			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
1057		spin_unlock_bh(&im->lock);
1058		if (changed)
1059			igmp_mod_timer(im, max_delay);
1060	}
1061	rcu_read_unlock();
1062	return false;
1063}
1064
1065/* called in rcu_read_lock() section */
1066int igmp_rcv(struct sk_buff *skb)
1067{
1068	/* This basically follows the spec line by line -- see RFC1112 */
1069	struct igmphdr *ih;
1070	struct net_device *dev = skb->dev;
1071	struct in_device *in_dev;
1072	int len = skb->len;
1073	bool dropped = true;
1074
1075	if (netif_is_l3_master(dev)) {
1076		dev = dev_get_by_index_rcu(dev_net(dev), IPCB(skb)->iif);
1077		if (!dev)
1078			goto drop;
1079	}
1080
1081	in_dev = __in_dev_get_rcu(dev);
1082	if (!in_dev)
1083		goto drop;
1084
1085	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1086		goto drop;
1087
1088	if (skb_checksum_simple_validate(skb))
1089		goto drop;
1090
1091	ih = igmp_hdr(skb);
1092	switch (ih->type) {
1093	case IGMP_HOST_MEMBERSHIP_QUERY:
1094		dropped = igmp_heard_query(in_dev, skb, len);
1095		break;
1096	case IGMP_HOST_MEMBERSHIP_REPORT:
1097	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1098		/* Is it our report looped back? */
1099		if (rt_is_output_route(skb_rtable(skb)))
1100			break;
1101		/* don't rely on MC router hearing unicast reports */
1102		if (skb->pkt_type == PACKET_MULTICAST ||
1103		    skb->pkt_type == PACKET_BROADCAST)
1104			dropped = igmp_heard_report(in_dev, ih->group);
1105		break;
1106	case IGMP_PIM:
1107#ifdef CONFIG_IP_PIMSM_V1
1108		return pim_rcv_v1(skb);
1109#endif
1110	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1111	case IGMP_DVMRP:
1112	case IGMP_TRACE:
1113	case IGMP_HOST_LEAVE_MESSAGE:
1114	case IGMP_MTRACE:
1115	case IGMP_MTRACE_RESP:
1116		break;
1117	default:
1118		break;
1119	}
1120
1121drop:
1122	if (dropped)
1123		kfree_skb(skb);
1124	else
1125		consume_skb(skb);
1126	return 0;
1127}
1128
1129#endif
1130
1131
1132/*
1133 *	Add a filter to a device
1134 */
1135
1136static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1137{
1138	char buf[MAX_ADDR_LEN];
1139	struct net_device *dev = in_dev->dev;
1140
1141	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1142	   We will get multicast token leakage, when IFF_MULTICAST
1143	   is changed. This check should be done in ndo_set_rx_mode
1144	   routine. Something sort of:
1145	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1146	   --ANK
1147	   */
1148	if (arp_mc_map(addr, buf, dev, 0) == 0)
1149		dev_mc_add(dev, buf);
1150}
1151
1152/*
1153 *	Remove a filter from a device
1154 */
1155
1156static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1157{
1158	char buf[MAX_ADDR_LEN];
1159	struct net_device *dev = in_dev->dev;
1160
1161	if (arp_mc_map(addr, buf, dev, 0) == 0)
1162		dev_mc_del(dev, buf);
1163}
1164
1165#ifdef CONFIG_IP_MULTICAST
1166/*
1167 * deleted ip_mc_list manipulation
1168 */
1169static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im,
1170			      gfp_t gfp)
1171{
1172	struct ip_mc_list *pmc;
1173	struct net *net = dev_net(in_dev->dev);
1174
1175	/* this is an "ip_mc_list" for convenience; only the fields below
1176	 * are actually used. In particular, the refcnt and users are not
1177	 * used for management of the delete list. Using the same structure
1178	 * for deleted items allows change reports to use common code with
1179	 * non-deleted or query-response MCA's.
1180	 */
1181	pmc = kzalloc(sizeof(*pmc), gfp);
1182	if (!pmc)
1183		return;
1184	spin_lock_init(&pmc->lock);
1185	spin_lock_bh(&im->lock);
1186	pmc->interface = im->interface;
1187	in_dev_hold(in_dev);
1188	pmc->multiaddr = im->multiaddr;
1189	pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1190	pmc->sfmode = im->sfmode;
1191	if (pmc->sfmode == MCAST_INCLUDE) {
1192		struct ip_sf_list *psf;
1193
1194		pmc->tomb = im->tomb;
1195		pmc->sources = im->sources;
1196		im->tomb = im->sources = NULL;
1197		for (psf = pmc->sources; psf; psf = psf->sf_next)
1198			psf->sf_crcount = pmc->crcount;
1199	}
1200	spin_unlock_bh(&im->lock);
1201
1202	spin_lock_bh(&in_dev->mc_tomb_lock);
1203	pmc->next = in_dev->mc_tomb;
1204	in_dev->mc_tomb = pmc;
1205	spin_unlock_bh(&in_dev->mc_tomb_lock);
1206}
1207
1208/*
1209 * restore ip_mc_list deleted records
1210 */
1211static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1212{
1213	struct ip_mc_list *pmc, *pmc_prev;
1214	struct ip_sf_list *psf;
1215	struct net *net = dev_net(in_dev->dev);
1216	__be32 multiaddr = im->multiaddr;
1217
1218	spin_lock_bh(&in_dev->mc_tomb_lock);
1219	pmc_prev = NULL;
1220	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1221		if (pmc->multiaddr == multiaddr)
1222			break;
1223		pmc_prev = pmc;
1224	}
1225	if (pmc) {
1226		if (pmc_prev)
1227			pmc_prev->next = pmc->next;
1228		else
1229			in_dev->mc_tomb = pmc->next;
1230	}
1231	spin_unlock_bh(&in_dev->mc_tomb_lock);
1232
1233	spin_lock_bh(&im->lock);
1234	if (pmc) {
1235		im->interface = pmc->interface;
1236		if (im->sfmode == MCAST_INCLUDE) {
1237			swap(im->tomb, pmc->tomb);
1238			swap(im->sources, pmc->sources);
 
 
1239			for (psf = im->sources; psf; psf = psf->sf_next)
1240				psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1241		} else {
1242			im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1243		}
1244		in_dev_put(pmc->interface);
1245		kfree_pmc(pmc);
1246	}
1247	spin_unlock_bh(&im->lock);
1248}
1249
1250/*
1251 * flush ip_mc_list deleted records
1252 */
1253static void igmpv3_clear_delrec(struct in_device *in_dev)
1254{
1255	struct ip_mc_list *pmc, *nextpmc;
1256
1257	spin_lock_bh(&in_dev->mc_tomb_lock);
1258	pmc = in_dev->mc_tomb;
1259	in_dev->mc_tomb = NULL;
1260	spin_unlock_bh(&in_dev->mc_tomb_lock);
1261
1262	for (; pmc; pmc = nextpmc) {
1263		nextpmc = pmc->next;
1264		ip_mc_clear_src(pmc);
1265		in_dev_put(pmc->interface);
1266		kfree_pmc(pmc);
1267	}
1268	/* clear dead sources, too */
1269	rcu_read_lock();
1270	for_each_pmc_rcu(in_dev, pmc) {
1271		struct ip_sf_list *psf;
1272
1273		spin_lock_bh(&pmc->lock);
1274		psf = pmc->tomb;
1275		pmc->tomb = NULL;
1276		spin_unlock_bh(&pmc->lock);
1277		ip_sf_list_clear_all(psf);
 
 
 
1278	}
1279	rcu_read_unlock();
1280}
1281#endif
1282
1283static void __igmp_group_dropped(struct ip_mc_list *im, gfp_t gfp)
1284{
1285	struct in_device *in_dev = im->interface;
1286#ifdef CONFIG_IP_MULTICAST
1287	struct net *net = dev_net(in_dev->dev);
1288	int reporter;
1289#endif
1290
1291	if (im->loaded) {
1292		im->loaded = 0;
1293		ip_mc_filter_del(in_dev, im->multiaddr);
1294	}
1295
1296#ifdef CONFIG_IP_MULTICAST
1297	if (im->multiaddr == IGMP_ALL_HOSTS)
1298		return;
1299	if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1300		return;
1301
1302	reporter = im->reporter;
1303	igmp_stop_timer(im);
1304
1305	if (!in_dev->dead) {
1306		if (IGMP_V1_SEEN(in_dev))
1307			return;
1308		if (IGMP_V2_SEEN(in_dev)) {
1309			if (reporter)
1310				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1311			return;
1312		}
1313		/* IGMPv3 */
1314		igmpv3_add_delrec(in_dev, im, gfp);
1315
1316		igmp_ifc_event(in_dev);
1317	}
1318#endif
1319}
1320
1321static void igmp_group_dropped(struct ip_mc_list *im)
1322{
1323	__igmp_group_dropped(im, GFP_KERNEL);
1324}
1325
1326static void igmp_group_added(struct ip_mc_list *im)
1327{
1328	struct in_device *in_dev = im->interface;
1329#ifdef CONFIG_IP_MULTICAST
1330	struct net *net = dev_net(in_dev->dev);
1331#endif
1332
1333	if (im->loaded == 0) {
1334		im->loaded = 1;
1335		ip_mc_filter_add(in_dev, im->multiaddr);
1336	}
1337
1338#ifdef CONFIG_IP_MULTICAST
1339	if (im->multiaddr == IGMP_ALL_HOSTS)
1340		return;
1341	if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1342		return;
1343
1344	if (in_dev->dead)
1345		return;
1346
1347	im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1348	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1349		spin_lock_bh(&im->lock);
1350		igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1351		spin_unlock_bh(&im->lock);
1352		return;
1353	}
1354	/* else, v3 */
1355
1356	/* Based on RFC3376 5.1, for newly added INCLUDE SSM, we should
1357	 * not send filter-mode change record as the mode should be from
1358	 * IN() to IN(A).
1359	 */
1360	if (im->sfmode == MCAST_EXCLUDE)
1361		im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1362
1363	igmp_ifc_event(in_dev);
1364#endif
1365}
1366
1367
1368/*
1369 *	Multicast list managers
1370 */
1371
1372static u32 ip_mc_hash(const struct ip_mc_list *im)
1373{
1374	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1375}
1376
1377static void ip_mc_hash_add(struct in_device *in_dev,
1378			   struct ip_mc_list *im)
1379{
1380	struct ip_mc_list __rcu **mc_hash;
1381	u32 hash;
1382
1383	mc_hash = rtnl_dereference(in_dev->mc_hash);
1384	if (mc_hash) {
1385		hash = ip_mc_hash(im);
1386		im->next_hash = mc_hash[hash];
1387		rcu_assign_pointer(mc_hash[hash], im);
1388		return;
1389	}
1390
1391	/* do not use a hash table for small number of items */
1392	if (in_dev->mc_count < 4)
1393		return;
1394
1395	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1396			  GFP_KERNEL);
1397	if (!mc_hash)
1398		return;
1399
1400	for_each_pmc_rtnl(in_dev, im) {
1401		hash = ip_mc_hash(im);
1402		im->next_hash = mc_hash[hash];
1403		RCU_INIT_POINTER(mc_hash[hash], im);
1404	}
1405
1406	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1407}
1408
1409static void ip_mc_hash_remove(struct in_device *in_dev,
1410			      struct ip_mc_list *im)
1411{
1412	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1413	struct ip_mc_list *aux;
1414
1415	if (!mc_hash)
1416		return;
1417	mc_hash += ip_mc_hash(im);
1418	while ((aux = rtnl_dereference(*mc_hash)) != im)
1419		mc_hash = &aux->next_hash;
1420	*mc_hash = im->next_hash;
1421}
1422
1423
1424/*
1425 *	A socket has joined a multicast group on device dev.
1426 */
1427static void ____ip_mc_inc_group(struct in_device *in_dev, __be32 addr,
1428				unsigned int mode, gfp_t gfp)
1429{
1430	struct ip_mc_list *im;
 
 
 
1431
1432	ASSERT_RTNL();
1433
1434	for_each_pmc_rtnl(in_dev, im) {
1435		if (im->multiaddr == addr) {
1436			im->users++;
1437			ip_mc_add_src(in_dev, &addr, mode, 0, NULL, 0);
1438			goto out;
1439		}
1440	}
1441
1442	im = kzalloc(sizeof(*im), gfp);
1443	if (!im)
1444		goto out;
1445
1446	im->users = 1;
1447	im->interface = in_dev;
1448	in_dev_hold(in_dev);
1449	im->multiaddr = addr;
1450	/* initial mode is (EX, empty) */
1451	im->sfmode = mode;
1452	im->sfcount[mode] = 1;
1453	refcount_set(&im->refcnt, 1);
1454	spin_lock_init(&im->lock);
1455#ifdef CONFIG_IP_MULTICAST
1456	timer_setup(&im->timer, igmp_timer_expire, 0);
 
1457#endif
1458
1459	im->next_rcu = in_dev->mc_list;
1460	in_dev->mc_count++;
1461	rcu_assign_pointer(in_dev->mc_list, im);
1462
1463	ip_mc_hash_add(in_dev, im);
1464
1465#ifdef CONFIG_IP_MULTICAST
1466	igmpv3_del_delrec(in_dev, im);
1467#endif
1468	igmp_group_added(im);
1469	if (!in_dev->dead)
1470		ip_rt_multicast_event(in_dev);
1471out:
1472	return;
1473}
1474
1475void __ip_mc_inc_group(struct in_device *in_dev, __be32 addr, gfp_t gfp)
1476{
1477	____ip_mc_inc_group(in_dev, addr, MCAST_EXCLUDE, gfp);
1478}
1479EXPORT_SYMBOL(__ip_mc_inc_group);
1480
1481void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1482{
1483	__ip_mc_inc_group(in_dev, addr, GFP_KERNEL);
1484}
1485EXPORT_SYMBOL(ip_mc_inc_group);
1486
1487static int ip_mc_check_iphdr(struct sk_buff *skb)
1488{
1489	const struct iphdr *iph;
1490	unsigned int len;
1491	unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1492
1493	if (!pskb_may_pull(skb, offset))
1494		return -EINVAL;
1495
1496	iph = ip_hdr(skb);
1497
1498	if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1499		return -EINVAL;
1500
1501	offset += ip_hdrlen(skb) - sizeof(*iph);
1502
1503	if (!pskb_may_pull(skb, offset))
1504		return -EINVAL;
1505
1506	iph = ip_hdr(skb);
1507
1508	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1509		return -EINVAL;
1510
1511	len = skb_network_offset(skb) + ntohs(iph->tot_len);
1512	if (skb->len < len || len < offset)
1513		return -EINVAL;
1514
1515	skb_set_transport_header(skb, offset);
1516
1517	return 0;
1518}
1519
1520static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1521{
1522	unsigned int len = skb_transport_offset(skb);
1523
1524	len += sizeof(struct igmpv3_report);
1525
1526	return ip_mc_may_pull(skb, len) ? 0 : -EINVAL;
1527}
1528
1529static int ip_mc_check_igmp_query(struct sk_buff *skb)
1530{
1531	unsigned int transport_len = ip_transport_len(skb);
1532	unsigned int len;
 
 
 
1533
1534	/* IGMPv{1,2}? */
1535	if (transport_len != sizeof(struct igmphdr)) {
1536		/* or IGMPv3? */
1537		if (transport_len < sizeof(struct igmpv3_query))
1538			return -EINVAL;
1539
1540		len = skb_transport_offset(skb) + sizeof(struct igmpv3_query);
1541		if (!ip_mc_may_pull(skb, len))
1542			return -EINVAL;
1543	}
1544
1545	/* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1546	 * all-systems destination addresses (224.0.0.1) for general queries
1547	 */
1548	if (!igmp_hdr(skb)->group &&
1549	    ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1550		return -EINVAL;
1551
1552	return 0;
1553}
1554
1555static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1556{
1557	switch (igmp_hdr(skb)->type) {
1558	case IGMP_HOST_LEAVE_MESSAGE:
1559	case IGMP_HOST_MEMBERSHIP_REPORT:
1560	case IGMPV2_HOST_MEMBERSHIP_REPORT:
 
1561		return 0;
1562	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1563		return ip_mc_check_igmp_reportv3(skb);
1564	case IGMP_HOST_MEMBERSHIP_QUERY:
1565		return ip_mc_check_igmp_query(skb);
1566	default:
1567		return -ENOMSG;
1568	}
1569}
1570
1571static __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1572{
1573	return skb_checksum_simple_validate(skb);
1574}
1575
1576static int ip_mc_check_igmp_csum(struct sk_buff *skb)
 
1577{
1578	unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1579	unsigned int transport_len = ip_transport_len(skb);
1580	struct sk_buff *skb_chk;
 
 
 
1581
1582	if (!ip_mc_may_pull(skb, len))
1583		return -EINVAL;
1584
1585	skb_chk = skb_checksum_trimmed(skb, transport_len,
1586				       ip_mc_validate_checksum);
1587	if (!skb_chk)
1588		return -EINVAL;
 
 
 
1589
1590	if (skb_chk != skb)
 
 
 
 
 
 
 
1591		kfree_skb(skb_chk);
1592
1593	return 0;
 
 
 
 
 
 
1594}
1595
1596/**
1597 * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1598 * @skb: the skb to validate
 
1599 *
1600 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1601 * skb transport header accordingly and returns zero.
1602 *
1603 * -EINVAL: A broken packet was detected, i.e. it violates some internet
1604 *  standard
1605 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1606 * -ENOMEM: A memory allocation failure happened.
1607 *
 
 
 
 
 
 
 
 
1608 * Caller needs to set the skb network header and free any returned skb if it
1609 * differs from the provided skb.
1610 */
1611int ip_mc_check_igmp(struct sk_buff *skb)
1612{
1613	int ret = ip_mc_check_iphdr(skb);
1614
1615	if (ret < 0)
1616		return ret;
1617
1618	if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1619		return -ENOMSG;
1620
1621	ret = ip_mc_check_igmp_csum(skb);
1622	if (ret < 0)
1623		return ret;
1624
1625	return ip_mc_check_igmp_msg(skb);
1626}
1627EXPORT_SYMBOL(ip_mc_check_igmp);
1628
1629/*
1630 *	Resend IGMP JOIN report; used by netdev notifier.
1631 */
1632static void ip_mc_rejoin_groups(struct in_device *in_dev)
1633{
1634#ifdef CONFIG_IP_MULTICAST
1635	struct ip_mc_list *im;
1636	int type;
1637	struct net *net = dev_net(in_dev->dev);
1638
1639	ASSERT_RTNL();
1640
1641	for_each_pmc_rtnl(in_dev, im) {
1642		if (im->multiaddr == IGMP_ALL_HOSTS)
1643			continue;
1644		if (ipv4_is_local_multicast(im->multiaddr) &&
1645		    !net->ipv4.sysctl_igmp_llm_reports)
1646			continue;
1647
1648		/* a failover is happening and switches
1649		 * must be notified immediately
1650		 */
1651		if (IGMP_V1_SEEN(in_dev))
1652			type = IGMP_HOST_MEMBERSHIP_REPORT;
1653		else if (IGMP_V2_SEEN(in_dev))
1654			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1655		else
1656			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1657		igmp_send_report(in_dev, im, type);
1658	}
1659#endif
1660}
1661
1662/*
1663 *	A socket has left a multicast group on device dev
1664 */
1665
1666void __ip_mc_dec_group(struct in_device *in_dev, __be32 addr, gfp_t gfp)
1667{
1668	struct ip_mc_list *i;
1669	struct ip_mc_list __rcu **ip;
1670
1671	ASSERT_RTNL();
1672
1673	for (ip = &in_dev->mc_list;
1674	     (i = rtnl_dereference(*ip)) != NULL;
1675	     ip = &i->next_rcu) {
1676		if (i->multiaddr == addr) {
1677			if (--i->users == 0) {
1678				ip_mc_hash_remove(in_dev, i);
1679				*ip = i->next_rcu;
1680				in_dev->mc_count--;
1681				__igmp_group_dropped(i, gfp);
1682				ip_mc_clear_src(i);
1683
1684				if (!in_dev->dead)
1685					ip_rt_multicast_event(in_dev);
1686
1687				ip_ma_put(i);
1688				return;
1689			}
1690			break;
1691		}
1692	}
1693}
1694EXPORT_SYMBOL(__ip_mc_dec_group);
1695
1696/* Device changing type */
1697
1698void ip_mc_unmap(struct in_device *in_dev)
1699{
1700	struct ip_mc_list *pmc;
1701
1702	ASSERT_RTNL();
1703
1704	for_each_pmc_rtnl(in_dev, pmc)
1705		igmp_group_dropped(pmc);
1706}
1707
1708void ip_mc_remap(struct in_device *in_dev)
1709{
1710	struct ip_mc_list *pmc;
1711
1712	ASSERT_RTNL();
1713
1714	for_each_pmc_rtnl(in_dev, pmc) {
1715#ifdef CONFIG_IP_MULTICAST
1716		igmpv3_del_delrec(in_dev, pmc);
1717#endif
1718		igmp_group_added(pmc);
1719	}
1720}
1721
1722/* Device going down */
1723
1724void ip_mc_down(struct in_device *in_dev)
1725{
1726	struct ip_mc_list *pmc;
1727
1728	ASSERT_RTNL();
1729
1730	for_each_pmc_rtnl(in_dev, pmc)
1731		igmp_group_dropped(pmc);
1732
1733#ifdef CONFIG_IP_MULTICAST
1734	WRITE_ONCE(in_dev->mr_ifc_count, 0);
1735	if (del_timer(&in_dev->mr_ifc_timer))
1736		__in_dev_put(in_dev);
1737	in_dev->mr_gq_running = 0;
1738	if (del_timer(&in_dev->mr_gq_timer))
1739		__in_dev_put(in_dev);
1740#endif
1741
1742	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1743}
1744
1745#ifdef CONFIG_IP_MULTICAST
1746static void ip_mc_reset(struct in_device *in_dev)
1747{
 
1748	struct net *net = dev_net(in_dev->dev);
1749
1750	in_dev->mr_qi = IGMP_QUERY_INTERVAL;
1751	in_dev->mr_qri = IGMP_QUERY_RESPONSE_INTERVAL;
1752	in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1753}
1754#else
1755static void ip_mc_reset(struct in_device *in_dev)
1756{
1757}
1758#endif
1759
1760void ip_mc_init_dev(struct in_device *in_dev)
1761{
1762	ASSERT_RTNL();
1763
1764#ifdef CONFIG_IP_MULTICAST
1765	timer_setup(&in_dev->mr_gq_timer, igmp_gq_timer_expire, 0);
1766	timer_setup(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire, 0);
 
 
 
1767#endif
1768	ip_mc_reset(in_dev);
1769
1770	spin_lock_init(&in_dev->mc_tomb_lock);
1771}
1772
1773/* Device going up */
1774
1775void ip_mc_up(struct in_device *in_dev)
1776{
1777	struct ip_mc_list *pmc;
 
 
 
1778
1779	ASSERT_RTNL();
1780
1781	ip_mc_reset(in_dev);
 
 
1782	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1783
1784	for_each_pmc_rtnl(in_dev, pmc) {
1785#ifdef CONFIG_IP_MULTICAST
1786		igmpv3_del_delrec(in_dev, pmc);
1787#endif
1788		igmp_group_added(pmc);
1789	}
1790}
1791
1792/*
1793 *	Device is about to be destroyed: clean up.
1794 */
1795
1796void ip_mc_destroy_dev(struct in_device *in_dev)
1797{
1798	struct ip_mc_list *i;
1799
1800	ASSERT_RTNL();
1801
1802	/* Deactivate timers */
1803	ip_mc_down(in_dev);
1804#ifdef CONFIG_IP_MULTICAST
1805	igmpv3_clear_delrec(in_dev);
1806#endif
1807
1808	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1809		in_dev->mc_list = i->next_rcu;
1810		in_dev->mc_count--;
1811		ip_mc_clear_src(i);
1812		ip_ma_put(i);
1813	}
1814}
1815
1816/* RTNL is locked */
1817static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1818{
1819	struct net_device *dev = NULL;
1820	struct in_device *idev = NULL;
1821
1822	if (imr->imr_ifindex) {
1823		idev = inetdev_by_index(net, imr->imr_ifindex);
1824		return idev;
1825	}
1826	if (imr->imr_address.s_addr) {
1827		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1828		if (!dev)
1829			return NULL;
1830	}
1831
1832	if (!dev) {
1833		struct rtable *rt = ip_route_output(net,
1834						    imr->imr_multiaddr.s_addr,
1835						    0, 0, 0);
1836		if (!IS_ERR(rt)) {
1837			dev = rt->dst.dev;
1838			ip_rt_put(rt);
1839		}
1840	}
1841	if (dev) {
1842		imr->imr_ifindex = dev->ifindex;
1843		idev = __in_dev_get_rtnl(dev);
1844	}
1845	return idev;
1846}
1847
1848/*
1849 *	Join a socket to a group
1850 */
1851
1852static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1853	__be32 *psfsrc)
1854{
1855	struct ip_sf_list *psf, *psf_prev;
1856	int rv = 0;
1857
1858	psf_prev = NULL;
1859	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1860		if (psf->sf_inaddr == *psfsrc)
1861			break;
1862		psf_prev = psf;
1863	}
1864	if (!psf || psf->sf_count[sfmode] == 0) {
1865		/* source filter not found, or count wrong =>  bug */
1866		return -ESRCH;
1867	}
1868	psf->sf_count[sfmode]--;
1869	if (psf->sf_count[sfmode] == 0) {
1870		ip_rt_multicast_event(pmc->interface);
1871	}
1872	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1873#ifdef CONFIG_IP_MULTICAST
1874		struct in_device *in_dev = pmc->interface;
1875		struct net *net = dev_net(in_dev->dev);
1876#endif
1877
1878		/* no more filters for this source */
1879		if (psf_prev)
1880			psf_prev->sf_next = psf->sf_next;
1881		else
1882			pmc->sources = psf->sf_next;
1883#ifdef CONFIG_IP_MULTICAST
1884		if (psf->sf_oldin &&
1885		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1886			psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1887			psf->sf_next = pmc->tomb;
1888			pmc->tomb = psf;
1889			rv = 1;
1890		} else
1891#endif
1892			kfree(psf);
1893	}
1894	return rv;
1895}
1896
1897#ifndef CONFIG_IP_MULTICAST
1898#define igmp_ifc_event(x)	do { } while (0)
1899#endif
1900
1901static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1902			 int sfcount, __be32 *psfsrc, int delta)
1903{
1904	struct ip_mc_list *pmc;
1905	int	changerec = 0;
1906	int	i, err;
1907
1908	if (!in_dev)
1909		return -ENODEV;
1910	rcu_read_lock();
1911	for_each_pmc_rcu(in_dev, pmc) {
1912		if (*pmca == pmc->multiaddr)
1913			break;
1914	}
1915	if (!pmc) {
1916		/* MCA not found?? bug */
1917		rcu_read_unlock();
1918		return -ESRCH;
1919	}
1920	spin_lock_bh(&pmc->lock);
1921	rcu_read_unlock();
1922#ifdef CONFIG_IP_MULTICAST
1923	sf_markstate(pmc);
1924#endif
1925	if (!delta) {
1926		err = -EINVAL;
1927		if (!pmc->sfcount[sfmode])
1928			goto out_unlock;
1929		pmc->sfcount[sfmode]--;
1930	}
1931	err = 0;
1932	for (i = 0; i < sfcount; i++) {
1933		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1934
1935		changerec |= rv > 0;
1936		if (!err && rv < 0)
1937			err = rv;
1938	}
1939	if (pmc->sfmode == MCAST_EXCLUDE &&
1940	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1941	    pmc->sfcount[MCAST_INCLUDE]) {
1942#ifdef CONFIG_IP_MULTICAST
1943		struct ip_sf_list *psf;
1944		struct net *net = dev_net(in_dev->dev);
1945#endif
1946
1947		/* filter mode change */
1948		pmc->sfmode = MCAST_INCLUDE;
1949#ifdef CONFIG_IP_MULTICAST
1950		pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1951		WRITE_ONCE(in_dev->mr_ifc_count, pmc->crcount);
1952		for (psf = pmc->sources; psf; psf = psf->sf_next)
1953			psf->sf_crcount = 0;
1954		igmp_ifc_event(pmc->interface);
1955	} else if (sf_setstate(pmc) || changerec) {
1956		igmp_ifc_event(pmc->interface);
1957#endif
1958	}
1959out_unlock:
1960	spin_unlock_bh(&pmc->lock);
1961	return err;
1962}
1963
1964/*
1965 * Add multicast single-source filter to the interface list
1966 */
1967static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1968	__be32 *psfsrc)
1969{
1970	struct ip_sf_list *psf, *psf_prev;
1971
1972	psf_prev = NULL;
1973	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1974		if (psf->sf_inaddr == *psfsrc)
1975			break;
1976		psf_prev = psf;
1977	}
1978	if (!psf) {
1979		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1980		if (!psf)
1981			return -ENOBUFS;
1982		psf->sf_inaddr = *psfsrc;
1983		if (psf_prev) {
1984			psf_prev->sf_next = psf;
1985		} else
1986			pmc->sources = psf;
1987	}
1988	psf->sf_count[sfmode]++;
1989	if (psf->sf_count[sfmode] == 1) {
1990		ip_rt_multicast_event(pmc->interface);
1991	}
1992	return 0;
1993}
1994
1995#ifdef CONFIG_IP_MULTICAST
1996static void sf_markstate(struct ip_mc_list *pmc)
1997{
1998	struct ip_sf_list *psf;
1999	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
2000
2001	for (psf = pmc->sources; psf; psf = psf->sf_next)
2002		if (pmc->sfcount[MCAST_EXCLUDE]) {
2003			psf->sf_oldin = mca_xcount ==
2004				psf->sf_count[MCAST_EXCLUDE] &&
2005				!psf->sf_count[MCAST_INCLUDE];
2006		} else
2007			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2008}
2009
2010static int sf_setstate(struct ip_mc_list *pmc)
2011{
2012	struct ip_sf_list *psf, *dpsf;
2013	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
2014	int qrv = pmc->interface->mr_qrv;
2015	int new_in, rv;
2016
2017	rv = 0;
2018	for (psf = pmc->sources; psf; psf = psf->sf_next) {
2019		if (pmc->sfcount[MCAST_EXCLUDE]) {
2020			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2021				!psf->sf_count[MCAST_INCLUDE];
2022		} else
2023			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2024		if (new_in) {
2025			if (!psf->sf_oldin) {
2026				struct ip_sf_list *prev = NULL;
2027
2028				for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
2029					if (dpsf->sf_inaddr == psf->sf_inaddr)
2030						break;
2031					prev = dpsf;
2032				}
2033				if (dpsf) {
2034					if (prev)
2035						prev->sf_next = dpsf->sf_next;
2036					else
2037						pmc->tomb = dpsf->sf_next;
2038					kfree(dpsf);
2039				}
2040				psf->sf_crcount = qrv;
2041				rv++;
2042			}
2043		} else if (psf->sf_oldin) {
2044
2045			psf->sf_crcount = 0;
2046			/*
2047			 * add or update "delete" records if an active filter
2048			 * is now inactive
2049			 */
2050			for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
2051				if (dpsf->sf_inaddr == psf->sf_inaddr)
2052					break;
2053			if (!dpsf) {
2054				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2055				if (!dpsf)
2056					continue;
2057				*dpsf = *psf;
2058				/* pmc->lock held by callers */
2059				dpsf->sf_next = pmc->tomb;
2060				pmc->tomb = dpsf;
2061			}
2062			dpsf->sf_crcount = qrv;
2063			rv++;
2064		}
2065	}
2066	return rv;
2067}
2068#endif
2069
2070/*
2071 * Add multicast source filter list to the interface list
2072 */
2073static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2074			 int sfcount, __be32 *psfsrc, int delta)
2075{
2076	struct ip_mc_list *pmc;
2077	int	isexclude;
2078	int	i, err;
2079
2080	if (!in_dev)
2081		return -ENODEV;
2082	rcu_read_lock();
2083	for_each_pmc_rcu(in_dev, pmc) {
2084		if (*pmca == pmc->multiaddr)
2085			break;
2086	}
2087	if (!pmc) {
2088		/* MCA not found?? bug */
2089		rcu_read_unlock();
2090		return -ESRCH;
2091	}
2092	spin_lock_bh(&pmc->lock);
2093	rcu_read_unlock();
2094
2095#ifdef CONFIG_IP_MULTICAST
2096	sf_markstate(pmc);
2097#endif
2098	isexclude = pmc->sfmode == MCAST_EXCLUDE;
2099	if (!delta)
2100		pmc->sfcount[sfmode]++;
2101	err = 0;
2102	for (i = 0; i < sfcount; i++) {
2103		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2104		if (err)
2105			break;
2106	}
2107	if (err) {
2108		int j;
2109
2110		if (!delta)
2111			pmc->sfcount[sfmode]--;
2112		for (j = 0; j < i; j++)
2113			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2114	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2115#ifdef CONFIG_IP_MULTICAST
2116		struct ip_sf_list *psf;
2117		struct net *net = dev_net(pmc->interface->dev);
2118		in_dev = pmc->interface;
2119#endif
2120
2121		/* filter mode change */
2122		if (pmc->sfcount[MCAST_EXCLUDE])
2123			pmc->sfmode = MCAST_EXCLUDE;
2124		else if (pmc->sfcount[MCAST_INCLUDE])
2125			pmc->sfmode = MCAST_INCLUDE;
2126#ifdef CONFIG_IP_MULTICAST
2127		/* else no filters; keep old mode for reports */
2128
2129		pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2130		WRITE_ONCE(in_dev->mr_ifc_count, pmc->crcount);
2131		for (psf = pmc->sources; psf; psf = psf->sf_next)
2132			psf->sf_crcount = 0;
2133		igmp_ifc_event(in_dev);
2134	} else if (sf_setstate(pmc)) {
2135		igmp_ifc_event(in_dev);
2136#endif
2137	}
2138	spin_unlock_bh(&pmc->lock);
2139	return err;
2140}
2141
2142static void ip_mc_clear_src(struct ip_mc_list *pmc)
2143{
2144	struct ip_sf_list *tomb, *sources;
2145
2146	spin_lock_bh(&pmc->lock);
2147	tomb = pmc->tomb;
 
 
2148	pmc->tomb = NULL;
2149	sources = pmc->sources;
 
 
 
2150	pmc->sources = NULL;
2151	pmc->sfmode = MCAST_EXCLUDE;
2152	pmc->sfcount[MCAST_INCLUDE] = 0;
2153	pmc->sfcount[MCAST_EXCLUDE] = 1;
2154	spin_unlock_bh(&pmc->lock);
2155
2156	ip_sf_list_clear_all(tomb);
2157	ip_sf_list_clear_all(sources);
2158}
2159
2160/* Join a multicast group
2161 */
2162static int __ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr,
2163			      unsigned int mode)
2164{
2165	__be32 addr = imr->imr_multiaddr.s_addr;
2166	struct ip_mc_socklist *iml, *i;
2167	struct in_device *in_dev;
2168	struct inet_sock *inet = inet_sk(sk);
2169	struct net *net = sock_net(sk);
2170	int ifindex;
2171	int count = 0;
2172	int err;
2173
2174	ASSERT_RTNL();
2175
2176	if (!ipv4_is_multicast(addr))
2177		return -EINVAL;
2178
2179	in_dev = ip_mc_find_dev(net, imr);
2180
2181	if (!in_dev) {
2182		err = -ENODEV;
2183		goto done;
2184	}
2185
2186	err = -EADDRINUSE;
2187	ifindex = imr->imr_ifindex;
2188	for_each_pmc_rtnl(inet, i) {
2189		if (i->multi.imr_multiaddr.s_addr == addr &&
2190		    i->multi.imr_ifindex == ifindex)
2191			goto done;
2192		count++;
2193	}
2194	err = -ENOBUFS;
2195	if (count >= net->ipv4.sysctl_igmp_max_memberships)
2196		goto done;
2197	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2198	if (!iml)
2199		goto done;
2200
2201	memcpy(&iml->multi, imr, sizeof(*imr));
2202	iml->next_rcu = inet->mc_list;
2203	iml->sflist = NULL;
2204	iml->sfmode = mode;
2205	rcu_assign_pointer(inet->mc_list, iml);
2206	____ip_mc_inc_group(in_dev, addr, mode, GFP_KERNEL);
2207	err = 0;
2208done:
2209	return err;
2210}
2211
2212/* Join ASM (Any-Source Multicast) group
2213 */
2214int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2215{
2216	return __ip_mc_join_group(sk, imr, MCAST_EXCLUDE);
2217}
2218EXPORT_SYMBOL(ip_mc_join_group);
2219
2220/* Join SSM (Source-Specific Multicast) group
2221 */
2222int ip_mc_join_group_ssm(struct sock *sk, struct ip_mreqn *imr,
2223			 unsigned int mode)
2224{
2225	return __ip_mc_join_group(sk, imr, mode);
2226}
2227
2228static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2229			   struct in_device *in_dev)
2230{
2231	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2232	int err;
2233
2234	if (!psf) {
2235		/* any-source empty exclude case */
2236		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2237			iml->sfmode, 0, NULL, 0);
2238	}
2239	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2240			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2241	RCU_INIT_POINTER(iml->sflist, NULL);
2242	/* decrease mem now to avoid the memleak warning */
2243	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2244	kfree_rcu(psf, rcu);
2245	return err;
2246}
2247
2248int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2249{
2250	struct inet_sock *inet = inet_sk(sk);
2251	struct ip_mc_socklist *iml;
2252	struct ip_mc_socklist __rcu **imlp;
2253	struct in_device *in_dev;
2254	struct net *net = sock_net(sk);
2255	__be32 group = imr->imr_multiaddr.s_addr;
2256	u32 ifindex;
2257	int ret = -EADDRNOTAVAIL;
2258
2259	ASSERT_RTNL();
2260
2261	in_dev = ip_mc_find_dev(net, imr);
2262	if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2263		ret = -ENODEV;
2264		goto out;
2265	}
2266	ifindex = imr->imr_ifindex;
2267	for (imlp = &inet->mc_list;
2268	     (iml = rtnl_dereference(*imlp)) != NULL;
2269	     imlp = &iml->next_rcu) {
2270		if (iml->multi.imr_multiaddr.s_addr != group)
2271			continue;
2272		if (ifindex) {
2273			if (iml->multi.imr_ifindex != ifindex)
2274				continue;
2275		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2276				iml->multi.imr_address.s_addr)
2277			continue;
2278
2279		(void) ip_mc_leave_src(sk, iml, in_dev);
2280
2281		*imlp = iml->next_rcu;
2282
2283		if (in_dev)
2284			ip_mc_dec_group(in_dev, group);
2285
2286		/* decrease mem now to avoid the memleak warning */
2287		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2288		kfree_rcu(iml, rcu);
2289		return 0;
2290	}
2291out:
2292	return ret;
2293}
2294EXPORT_SYMBOL(ip_mc_leave_group);
2295
2296int ip_mc_source(int add, int omode, struct sock *sk, struct
2297	ip_mreq_source *mreqs, int ifindex)
2298{
2299	int err;
2300	struct ip_mreqn imr;
2301	__be32 addr = mreqs->imr_multiaddr;
2302	struct ip_mc_socklist *pmc;
2303	struct in_device *in_dev = NULL;
2304	struct inet_sock *inet = inet_sk(sk);
2305	struct ip_sf_socklist *psl;
2306	struct net *net = sock_net(sk);
2307	int leavegroup = 0;
2308	int i, j, rv;
2309
2310	if (!ipv4_is_multicast(addr))
2311		return -EINVAL;
2312
2313	ASSERT_RTNL();
2314
2315	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2316	imr.imr_address.s_addr = mreqs->imr_interface;
2317	imr.imr_ifindex = ifindex;
2318	in_dev = ip_mc_find_dev(net, &imr);
2319
2320	if (!in_dev) {
2321		err = -ENODEV;
2322		goto done;
2323	}
2324	err = -EADDRNOTAVAIL;
2325
2326	for_each_pmc_rtnl(inet, pmc) {
2327		if ((pmc->multi.imr_multiaddr.s_addr ==
2328		     imr.imr_multiaddr.s_addr) &&
2329		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
2330			break;
2331	}
2332	if (!pmc) {		/* must have a prior join */
2333		err = -EINVAL;
2334		goto done;
2335	}
2336	/* if a source filter was set, must be the same mode as before */
2337	if (pmc->sflist) {
2338		if (pmc->sfmode != omode) {
2339			err = -EINVAL;
2340			goto done;
2341		}
2342	} else if (pmc->sfmode != omode) {
2343		/* allow mode switches for empty-set filters */
2344		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2345		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2346			NULL, 0);
2347		pmc->sfmode = omode;
2348	}
2349
2350	psl = rtnl_dereference(pmc->sflist);
2351	if (!add) {
2352		if (!psl)
2353			goto done;	/* err = -EADDRNOTAVAIL */
2354		rv = !0;
2355		for (i = 0; i < psl->sl_count; i++) {
2356			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2357				sizeof(__be32));
2358			if (rv == 0)
2359				break;
2360		}
2361		if (rv)		/* source not found */
2362			goto done;	/* err = -EADDRNOTAVAIL */
2363
2364		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2365		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2366			leavegroup = 1;
2367			goto done;
2368		}
2369
2370		/* update the interface filter */
2371		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2372			&mreqs->imr_sourceaddr, 1);
2373
2374		for (j = i+1; j < psl->sl_count; j++)
2375			psl->sl_addr[j-1] = psl->sl_addr[j];
2376		psl->sl_count--;
2377		err = 0;
2378		goto done;
2379	}
2380	/* else, add a new source to the filter */
2381
2382	if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2383		err = -ENOBUFS;
2384		goto done;
2385	}
2386	if (!psl || psl->sl_count == psl->sl_max) {
2387		struct ip_sf_socklist *newpsl;
2388		int count = IP_SFBLOCK;
2389
2390		if (psl)
2391			count += psl->sl_max;
2392		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2393		if (!newpsl) {
2394			err = -ENOBUFS;
2395			goto done;
2396		}
2397		newpsl->sl_max = count;
2398		newpsl->sl_count = count - IP_SFBLOCK;
2399		if (psl) {
2400			for (i = 0; i < psl->sl_count; i++)
2401				newpsl->sl_addr[i] = psl->sl_addr[i];
2402			/* decrease mem now to avoid the memleak warning */
2403			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2404			kfree_rcu(psl, rcu);
2405		}
2406		rcu_assign_pointer(pmc->sflist, newpsl);
2407		psl = newpsl;
2408	}
2409	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2410	for (i = 0; i < psl->sl_count; i++) {
2411		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2412			sizeof(__be32));
2413		if (rv == 0)
2414			break;
2415	}
2416	if (rv == 0)		/* address already there is an error */
2417		goto done;
2418	for (j = psl->sl_count-1; j >= i; j--)
2419		psl->sl_addr[j+1] = psl->sl_addr[j];
2420	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2421	psl->sl_count++;
2422	err = 0;
2423	/* update the interface list */
2424	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2425		&mreqs->imr_sourceaddr, 1);
2426done:
2427	if (leavegroup)
2428		err = ip_mc_leave_group(sk, &imr);
2429	return err;
2430}
2431
2432int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2433{
2434	int err = 0;
2435	struct ip_mreqn	imr;
2436	__be32 addr = msf->imsf_multiaddr;
2437	struct ip_mc_socklist *pmc;
2438	struct in_device *in_dev;
2439	struct inet_sock *inet = inet_sk(sk);
2440	struct ip_sf_socklist *newpsl, *psl;
2441	struct net *net = sock_net(sk);
2442	int leavegroup = 0;
2443
2444	if (!ipv4_is_multicast(addr))
2445		return -EINVAL;
2446	if (msf->imsf_fmode != MCAST_INCLUDE &&
2447	    msf->imsf_fmode != MCAST_EXCLUDE)
2448		return -EINVAL;
2449
2450	ASSERT_RTNL();
2451
2452	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2453	imr.imr_address.s_addr = msf->imsf_interface;
2454	imr.imr_ifindex = ifindex;
2455	in_dev = ip_mc_find_dev(net, &imr);
2456
2457	if (!in_dev) {
2458		err = -ENODEV;
2459		goto done;
2460	}
2461
2462	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2463	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2464		leavegroup = 1;
2465		goto done;
2466	}
2467
2468	for_each_pmc_rtnl(inet, pmc) {
2469		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2470		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2471			break;
2472	}
2473	if (!pmc) {		/* must have a prior join */
2474		err = -EINVAL;
2475		goto done;
2476	}
2477	if (msf->imsf_numsrc) {
2478		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2479							   GFP_KERNEL);
2480		if (!newpsl) {
2481			err = -ENOBUFS;
2482			goto done;
2483		}
2484		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2485		memcpy(newpsl->sl_addr, msf->imsf_slist,
2486			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2487		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2488			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2489		if (err) {
2490			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2491			goto done;
2492		}
2493	} else {
2494		newpsl = NULL;
2495		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2496				     msf->imsf_fmode, 0, NULL, 0);
2497	}
2498	psl = rtnl_dereference(pmc->sflist);
2499	if (psl) {
2500		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2501			psl->sl_count, psl->sl_addr, 0);
2502		/* decrease mem now to avoid the memleak warning */
2503		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2504		kfree_rcu(psl, rcu);
2505	} else
2506		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2507			0, NULL, 0);
2508	rcu_assign_pointer(pmc->sflist, newpsl);
2509	pmc->sfmode = msf->imsf_fmode;
2510	err = 0;
2511done:
2512	if (leavegroup)
2513		err = ip_mc_leave_group(sk, &imr);
2514	return err;
2515}
2516
2517int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2518	struct ip_msfilter __user *optval, int __user *optlen)
2519{
2520	int err, len, count, copycount;
2521	struct ip_mreqn	imr;
2522	__be32 addr = msf->imsf_multiaddr;
2523	struct ip_mc_socklist *pmc;
2524	struct in_device *in_dev;
2525	struct inet_sock *inet = inet_sk(sk);
2526	struct ip_sf_socklist *psl;
2527	struct net *net = sock_net(sk);
2528
2529	ASSERT_RTNL();
2530
2531	if (!ipv4_is_multicast(addr))
2532		return -EINVAL;
2533
2534	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2535	imr.imr_address.s_addr = msf->imsf_interface;
2536	imr.imr_ifindex = 0;
2537	in_dev = ip_mc_find_dev(net, &imr);
2538
2539	if (!in_dev) {
2540		err = -ENODEV;
2541		goto done;
2542	}
2543	err = -EADDRNOTAVAIL;
2544
2545	for_each_pmc_rtnl(inet, pmc) {
2546		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2547		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2548			break;
2549	}
2550	if (!pmc)		/* must have a prior join */
2551		goto done;
2552	msf->imsf_fmode = pmc->sfmode;
2553	psl = rtnl_dereference(pmc->sflist);
2554	if (!psl) {
2555		len = 0;
2556		count = 0;
2557	} else {
2558		count = psl->sl_count;
2559	}
2560	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2561	len = copycount * sizeof(psl->sl_addr[0]);
2562	msf->imsf_numsrc = count;
2563	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2564	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2565		return -EFAULT;
2566	}
2567	if (len &&
2568	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2569		return -EFAULT;
2570	return 0;
2571done:
2572	return err;
2573}
2574
2575int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2576	struct sockaddr_storage __user *p)
2577{
2578	int i, count, copycount;
2579	struct sockaddr_in *psin;
2580	__be32 addr;
2581	struct ip_mc_socklist *pmc;
2582	struct inet_sock *inet = inet_sk(sk);
2583	struct ip_sf_socklist *psl;
2584
2585	ASSERT_RTNL();
2586
2587	psin = (struct sockaddr_in *)&gsf->gf_group;
2588	if (psin->sin_family != AF_INET)
2589		return -EINVAL;
2590	addr = psin->sin_addr.s_addr;
2591	if (!ipv4_is_multicast(addr))
2592		return -EINVAL;
2593
 
 
2594	for_each_pmc_rtnl(inet, pmc) {
2595		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2596		    pmc->multi.imr_ifindex == gsf->gf_interface)
2597			break;
2598	}
2599	if (!pmc)		/* must have a prior join */
2600		return -EADDRNOTAVAIL;
2601	gsf->gf_fmode = pmc->sfmode;
2602	psl = rtnl_dereference(pmc->sflist);
2603	count = psl ? psl->sl_count : 0;
2604	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2605	gsf->gf_numsrc = count;
2606	for (i = 0; i < copycount; i++, p++) {
 
 
 
 
2607		struct sockaddr_storage ss;
2608
2609		psin = (struct sockaddr_in *)&ss;
2610		memset(&ss, 0, sizeof(ss));
2611		psin->sin_family = AF_INET;
2612		psin->sin_addr.s_addr = psl->sl_addr[i];
2613		if (copy_to_user(p, &ss, sizeof(ss)))
2614			return -EFAULT;
2615	}
2616	return 0;
 
 
2617}
2618
2619/*
2620 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2621 */
2622int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr,
2623		   int dif, int sdif)
2624{
2625	struct inet_sock *inet = inet_sk(sk);
2626	struct ip_mc_socklist *pmc;
2627	struct ip_sf_socklist *psl;
2628	int i;
2629	int ret;
2630
2631	ret = 1;
2632	if (!ipv4_is_multicast(loc_addr))
2633		goto out;
2634
2635	rcu_read_lock();
2636	for_each_pmc_rcu(inet, pmc) {
2637		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2638		    (pmc->multi.imr_ifindex == dif ||
2639		     (sdif && pmc->multi.imr_ifindex == sdif)))
2640			break;
2641	}
2642	ret = inet->mc_all;
2643	if (!pmc)
2644		goto unlock;
2645	psl = rcu_dereference(pmc->sflist);
2646	ret = (pmc->sfmode == MCAST_EXCLUDE);
2647	if (!psl)
2648		goto unlock;
2649
2650	for (i = 0; i < psl->sl_count; i++) {
2651		if (psl->sl_addr[i] == rmt_addr)
2652			break;
2653	}
2654	ret = 0;
2655	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2656		goto unlock;
2657	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2658		goto unlock;
2659	ret = 1;
2660unlock:
2661	rcu_read_unlock();
2662out:
2663	return ret;
2664}
2665
2666/*
2667 *	A socket is closing.
2668 */
2669
2670void ip_mc_drop_socket(struct sock *sk)
2671{
2672	struct inet_sock *inet = inet_sk(sk);
2673	struct ip_mc_socklist *iml;
2674	struct net *net = sock_net(sk);
2675
2676	if (!inet->mc_list)
2677		return;
2678
2679	rtnl_lock();
2680	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2681		struct in_device *in_dev;
2682
2683		inet->mc_list = iml->next_rcu;
2684		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2685		(void) ip_mc_leave_src(sk, iml, in_dev);
2686		if (in_dev)
2687			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2688		/* decrease mem now to avoid the memleak warning */
2689		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2690		kfree_rcu(iml, rcu);
2691	}
2692	rtnl_unlock();
2693}
2694
2695/* called with rcu_read_lock() */
2696int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2697{
2698	struct ip_mc_list *im;
2699	struct ip_mc_list __rcu **mc_hash;
2700	struct ip_sf_list *psf;
2701	int rv = 0;
2702
2703	mc_hash = rcu_dereference(in_dev->mc_hash);
2704	if (mc_hash) {
2705		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2706
2707		for (im = rcu_dereference(mc_hash[hash]);
2708		     im != NULL;
2709		     im = rcu_dereference(im->next_hash)) {
2710			if (im->multiaddr == mc_addr)
2711				break;
2712		}
2713	} else {
2714		for_each_pmc_rcu(in_dev, im) {
2715			if (im->multiaddr == mc_addr)
2716				break;
2717		}
2718	}
2719	if (im && proto == IPPROTO_IGMP) {
2720		rv = 1;
2721	} else if (im) {
2722		if (src_addr) {
2723			spin_lock_bh(&im->lock);
2724			for (psf = im->sources; psf; psf = psf->sf_next) {
2725				if (psf->sf_inaddr == src_addr)
2726					break;
2727			}
2728			if (psf)
2729				rv = psf->sf_count[MCAST_INCLUDE] ||
2730					psf->sf_count[MCAST_EXCLUDE] !=
2731					im->sfcount[MCAST_EXCLUDE];
2732			else
2733				rv = im->sfcount[MCAST_EXCLUDE] != 0;
2734			spin_unlock_bh(&im->lock);
2735		} else
2736			rv = 1; /* unspecified source; tentatively allow */
2737	}
2738	return rv;
2739}
2740
2741#if defined(CONFIG_PROC_FS)
2742struct igmp_mc_iter_state {
2743	struct seq_net_private p;
2744	struct net_device *dev;
2745	struct in_device *in_dev;
2746};
2747
2748#define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2749
2750static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2751{
2752	struct net *net = seq_file_net(seq);
2753	struct ip_mc_list *im = NULL;
2754	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2755
2756	state->in_dev = NULL;
2757	for_each_netdev_rcu(net, state->dev) {
2758		struct in_device *in_dev;
2759
2760		in_dev = __in_dev_get_rcu(state->dev);
2761		if (!in_dev)
2762			continue;
2763		im = rcu_dereference(in_dev->mc_list);
2764		if (im) {
2765			state->in_dev = in_dev;
2766			break;
2767		}
2768	}
2769	return im;
2770}
2771
2772static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2773{
2774	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2775
2776	im = rcu_dereference(im->next_rcu);
2777	while (!im) {
2778		state->dev = next_net_device_rcu(state->dev);
2779		if (!state->dev) {
2780			state->in_dev = NULL;
2781			break;
2782		}
2783		state->in_dev = __in_dev_get_rcu(state->dev);
2784		if (!state->in_dev)
2785			continue;
2786		im = rcu_dereference(state->in_dev->mc_list);
2787	}
2788	return im;
2789}
2790
2791static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2792{
2793	struct ip_mc_list *im = igmp_mc_get_first(seq);
2794	if (im)
2795		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2796			--pos;
2797	return pos ? NULL : im;
2798}
2799
2800static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2801	__acquires(rcu)
2802{
2803	rcu_read_lock();
2804	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2805}
2806
2807static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2808{
2809	struct ip_mc_list *im;
2810	if (v == SEQ_START_TOKEN)
2811		im = igmp_mc_get_first(seq);
2812	else
2813		im = igmp_mc_get_next(seq, v);
2814	++*pos;
2815	return im;
2816}
2817
2818static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2819	__releases(rcu)
2820{
2821	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2822
2823	state->in_dev = NULL;
2824	state->dev = NULL;
2825	rcu_read_unlock();
2826}
2827
2828static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2829{
2830	if (v == SEQ_START_TOKEN)
2831		seq_puts(seq,
2832			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2833	else {
2834		struct ip_mc_list *im = (struct ip_mc_list *)v;
2835		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2836		char   *querier;
2837		long delta;
2838
2839#ifdef CONFIG_IP_MULTICAST
2840		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2841			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2842			  "V3";
2843#else
2844		querier = "NONE";
2845#endif
2846
2847		if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2848			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2849				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2850		}
2851
2852		delta = im->timer.expires - jiffies;
2853		seq_printf(seq,
2854			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2855			   im->multiaddr, im->users,
2856			   im->tm_running,
2857			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2858			   im->reporter);
2859	}
2860	return 0;
2861}
2862
2863static const struct seq_operations igmp_mc_seq_ops = {
2864	.start	=	igmp_mc_seq_start,
2865	.next	=	igmp_mc_seq_next,
2866	.stop	=	igmp_mc_seq_stop,
2867	.show	=	igmp_mc_seq_show,
2868};
2869
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2870struct igmp_mcf_iter_state {
2871	struct seq_net_private p;
2872	struct net_device *dev;
2873	struct in_device *idev;
2874	struct ip_mc_list *im;
2875};
2876
2877#define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2878
2879static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2880{
2881	struct net *net = seq_file_net(seq);
2882	struct ip_sf_list *psf = NULL;
2883	struct ip_mc_list *im = NULL;
2884	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2885
2886	state->idev = NULL;
2887	state->im = NULL;
2888	for_each_netdev_rcu(net, state->dev) {
2889		struct in_device *idev;
2890		idev = __in_dev_get_rcu(state->dev);
2891		if (unlikely(!idev))
2892			continue;
2893		im = rcu_dereference(idev->mc_list);
2894		if (likely(im)) {
2895			spin_lock_bh(&im->lock);
2896			psf = im->sources;
2897			if (likely(psf)) {
2898				state->im = im;
2899				state->idev = idev;
2900				break;
2901			}
2902			spin_unlock_bh(&im->lock);
2903		}
2904	}
2905	return psf;
2906}
2907
2908static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2909{
2910	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2911
2912	psf = psf->sf_next;
2913	while (!psf) {
2914		spin_unlock_bh(&state->im->lock);
2915		state->im = state->im->next;
2916		while (!state->im) {
2917			state->dev = next_net_device_rcu(state->dev);
2918			if (!state->dev) {
2919				state->idev = NULL;
2920				goto out;
2921			}
2922			state->idev = __in_dev_get_rcu(state->dev);
2923			if (!state->idev)
2924				continue;
2925			state->im = rcu_dereference(state->idev->mc_list);
2926		}
2927		if (!state->im)
2928			break;
2929		spin_lock_bh(&state->im->lock);
2930		psf = state->im->sources;
2931	}
2932out:
2933	return psf;
2934}
2935
2936static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2937{
2938	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2939	if (psf)
2940		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2941			--pos;
2942	return pos ? NULL : psf;
2943}
2944
2945static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2946	__acquires(rcu)
2947{
2948	rcu_read_lock();
2949	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2950}
2951
2952static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2953{
2954	struct ip_sf_list *psf;
2955	if (v == SEQ_START_TOKEN)
2956		psf = igmp_mcf_get_first(seq);
2957	else
2958		psf = igmp_mcf_get_next(seq, v);
2959	++*pos;
2960	return psf;
2961}
2962
2963static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2964	__releases(rcu)
2965{
2966	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2967	if (likely(state->im)) {
2968		spin_unlock_bh(&state->im->lock);
2969		state->im = NULL;
2970	}
2971	state->idev = NULL;
2972	state->dev = NULL;
2973	rcu_read_unlock();
2974}
2975
2976static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2977{
2978	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2979	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2980
2981	if (v == SEQ_START_TOKEN) {
2982		seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2983	} else {
2984		seq_printf(seq,
2985			   "%3d %6.6s 0x%08x "
2986			   "0x%08x %6lu %6lu\n",
2987			   state->dev->ifindex, state->dev->name,
2988			   ntohl(state->im->multiaddr),
2989			   ntohl(psf->sf_inaddr),
2990			   psf->sf_count[MCAST_INCLUDE],
2991			   psf->sf_count[MCAST_EXCLUDE]);
2992	}
2993	return 0;
2994}
2995
2996static const struct seq_operations igmp_mcf_seq_ops = {
2997	.start	=	igmp_mcf_seq_start,
2998	.next	=	igmp_mcf_seq_next,
2999	.stop	=	igmp_mcf_seq_stop,
3000	.show	=	igmp_mcf_seq_show,
3001};
3002
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3003static int __net_init igmp_net_init(struct net *net)
3004{
3005	struct proc_dir_entry *pde;
3006	int err;
3007
3008	pde = proc_create_net("igmp", 0444, net->proc_net, &igmp_mc_seq_ops,
3009			sizeof(struct igmp_mc_iter_state));
3010	if (!pde)
3011		goto out_igmp;
3012	pde = proc_create_net("mcfilter", 0444, net->proc_net,
3013			&igmp_mcf_seq_ops, sizeof(struct igmp_mcf_iter_state));
3014	if (!pde)
3015		goto out_mcfilter;
3016	err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
3017				   SOCK_DGRAM, 0, net);
3018	if (err < 0) {
3019		pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
3020		       err);
3021		goto out_sock;
3022	}
3023
 
 
 
 
 
 
3024	return 0;
3025
3026out_sock:
3027	remove_proc_entry("mcfilter", net->proc_net);
3028out_mcfilter:
3029	remove_proc_entry("igmp", net->proc_net);
3030out_igmp:
3031	return -ENOMEM;
3032}
3033
3034static void __net_exit igmp_net_exit(struct net *net)
3035{
3036	remove_proc_entry("mcfilter", net->proc_net);
3037	remove_proc_entry("igmp", net->proc_net);
3038	inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
3039}
3040
3041static struct pernet_operations igmp_net_ops = {
3042	.init = igmp_net_init,
3043	.exit = igmp_net_exit,
3044};
3045#endif
3046
3047static int igmp_netdev_event(struct notifier_block *this,
3048			     unsigned long event, void *ptr)
3049{
3050	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3051	struct in_device *in_dev;
3052
3053	switch (event) {
3054	case NETDEV_RESEND_IGMP:
3055		in_dev = __in_dev_get_rtnl(dev);
3056		if (in_dev)
3057			ip_mc_rejoin_groups(in_dev);
3058		break;
3059	default:
3060		break;
3061	}
3062	return NOTIFY_DONE;
3063}
3064
3065static struct notifier_block igmp_notifier = {
3066	.notifier_call = igmp_netdev_event,
3067};
3068
3069int __init igmp_mc_init(void)
3070{
3071#if defined(CONFIG_PROC_FS)
3072	int err;
3073
3074	err = register_pernet_subsys(&igmp_net_ops);
3075	if (err)
3076		return err;
3077	err = register_netdevice_notifier(&igmp_notifier);
3078	if (err)
3079		goto reg_notif_fail;
3080	return 0;
3081
3082reg_notif_fail:
3083	unregister_pernet_subsys(&igmp_net_ops);
3084	return err;
3085#else
3086	return register_netdevice_notifier(&igmp_notifier);
3087#endif
3088}
v4.10.11
 
   1/*
   2 *	Linux NET3:	Internet Group Management Protocol  [IGMP]
   3 *
   4 *	This code implements the IGMP protocol as defined in RFC1112. There has
   5 *	been a further revision of this protocol since which is now supported.
   6 *
   7 *	If you have trouble with this module be careful what gcc you have used,
   8 *	the older version didn't come out right using gcc 2.5.8, the newer one
   9 *	seems to fall out with gcc 2.6.2.
  10 *
  11 *	Authors:
  12 *		Alan Cox <alan@lxorguk.ukuu.org.uk>
  13 *
  14 *	This program is free software; you can redistribute it and/or
  15 *	modify it under the terms of the GNU General Public License
  16 *	as published by the Free Software Foundation; either version
  17 *	2 of the License, or (at your option) any later version.
  18 *
  19 *	Fixes:
  20 *
  21 *		Alan Cox	:	Added lots of __inline__ to optimise
  22 *					the memory usage of all the tiny little
  23 *					functions.
  24 *		Alan Cox	:	Dumped the header building experiment.
  25 *		Alan Cox	:	Minor tweaks ready for multicast routing
  26 *					and extended IGMP protocol.
  27 *		Alan Cox	:	Removed a load of inline directives. Gcc 2.5.8
  28 *					writes utterly bogus code otherwise (sigh)
  29 *					fixed IGMP loopback to behave in the manner
  30 *					desired by mrouted, fixed the fact it has been
  31 *					broken since 1.3.6 and cleaned up a few minor
  32 *					points.
  33 *
  34 *		Chih-Jen Chang	:	Tried to revise IGMP to Version 2
  35 *		Tsu-Sheng Tsao		E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
  36 *					The enhancements are mainly based on Steve Deering's
  37 * 					ipmulti-3.5 source code.
  38 *		Chih-Jen Chang	:	Added the igmp_get_mrouter_info and
  39 *		Tsu-Sheng Tsao		igmp_set_mrouter_info to keep track of
  40 *					the mrouted version on that device.
  41 *		Chih-Jen Chang	:	Added the max_resp_time parameter to
  42 *		Tsu-Sheng Tsao		igmp_heard_query(). Using this parameter
  43 *					to identify the multicast router version
  44 *					and do what the IGMP version 2 specified.
  45 *		Chih-Jen Chang	:	Added a timer to revert to IGMP V2 router
  46 *		Tsu-Sheng Tsao		if the specified time expired.
  47 *		Alan Cox	:	Stop IGMP from 0.0.0.0 being accepted.
  48 *		Alan Cox	:	Use GFP_ATOMIC in the right places.
  49 *		Christian Daudt :	igmp timer wasn't set for local group
  50 *					memberships but was being deleted,
  51 *					which caused a "del_timer() called
  52 *					from %p with timer not initialized\n"
  53 *					message (960131).
  54 *		Christian Daudt :	removed del_timer from
  55 *					igmp_timer_expire function (960205).
  56 *             Christian Daudt :       igmp_heard_report now only calls
  57 *                                     igmp_timer_expire if tm->running is
  58 *                                     true (960216).
  59 *		Malcolm Beattie :	ttl comparison wrong in igmp_rcv made
  60 *					igmp_heard_query never trigger. Expiry
  61 *					miscalculation fixed in igmp_heard_query
  62 *					and random() made to return unsigned to
  63 *					prevent negative expiry times.
  64 *		Alexey Kuznetsov:	Wrong group leaving behaviour, backport
  65 *					fix from pending 2.1.x patches.
  66 *		Alan Cox:		Forget to enable FDDI support earlier.
  67 *		Alexey Kuznetsov:	Fixed leaving groups on device down.
  68 *		Alexey Kuznetsov:	Accordance to igmp-v2-06 draft.
  69 *		David L Stevens:	IGMPv3 support, with help from
  70 *					Vinay Kulkarni
  71 */
  72
  73#include <linux/module.h>
  74#include <linux/slab.h>
  75#include <linux/uaccess.h>
  76#include <linux/types.h>
  77#include <linux/kernel.h>
  78#include <linux/jiffies.h>
  79#include <linux/string.h>
  80#include <linux/socket.h>
  81#include <linux/sockios.h>
  82#include <linux/in.h>
  83#include <linux/inet.h>
  84#include <linux/netdevice.h>
  85#include <linux/skbuff.h>
  86#include <linux/inetdevice.h>
  87#include <linux/igmp.h>
  88#include <linux/if_arp.h>
  89#include <linux/rtnetlink.h>
  90#include <linux/times.h>
  91#include <linux/pkt_sched.h>
 
  92
  93#include <net/net_namespace.h>
  94#include <net/arp.h>
  95#include <net/ip.h>
  96#include <net/protocol.h>
  97#include <net/route.h>
  98#include <net/sock.h>
  99#include <net/checksum.h>
 100#include <net/inet_common.h>
 101#include <linux/netfilter_ipv4.h>
 102#ifdef CONFIG_IP_MROUTE
 103#include <linux/mroute.h>
 104#endif
 105#ifdef CONFIG_PROC_FS
 106#include <linux/proc_fs.h>
 107#include <linux/seq_file.h>
 108#endif
 109
 110#ifdef CONFIG_IP_MULTICAST
 111/* Parameter names and values are taken from igmp-v2-06 draft */
 112
 113#define IGMP_V1_ROUTER_PRESENT_TIMEOUT		(400*HZ)
 114#define IGMP_V2_ROUTER_PRESENT_TIMEOUT		(400*HZ)
 115#define IGMP_V2_UNSOLICITED_REPORT_INTERVAL	(10*HZ)
 116#define IGMP_V3_UNSOLICITED_REPORT_INTERVAL	(1*HZ)
 117#define IGMP_QUERY_RESPONSE_INTERVAL		(10*HZ)
 118#define IGMP_QUERY_ROBUSTNESS_VARIABLE		2
 119
 120
 121#define IGMP_INITIAL_REPORT_DELAY		(1)
 122
 123/* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
 124 * IGMP specs require to report membership immediately after
 125 * joining a group, but we delay the first report by a
 126 * small interval. It seems more natural and still does not
 127 * contradict to specs provided this delay is small enough.
 128 */
 129
 130#define IGMP_V1_SEEN(in_dev) \
 131	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 132	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 133	 ((in_dev)->mr_v1_seen && \
 134	  time_before(jiffies, (in_dev)->mr_v1_seen)))
 135#define IGMP_V2_SEEN(in_dev) \
 136	(IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 137	 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 138	 ((in_dev)->mr_v2_seen && \
 139	  time_before(jiffies, (in_dev)->mr_v2_seen)))
 140
 141static int unsolicited_report_interval(struct in_device *in_dev)
 142{
 143	int interval_ms, interval_jiffies;
 144
 145	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 146		interval_ms = IN_DEV_CONF_GET(
 147			in_dev,
 148			IGMPV2_UNSOLICITED_REPORT_INTERVAL);
 149	else /* v3 */
 150		interval_ms = IN_DEV_CONF_GET(
 151			in_dev,
 152			IGMPV3_UNSOLICITED_REPORT_INTERVAL);
 153
 154	interval_jiffies = msecs_to_jiffies(interval_ms);
 155
 156	/* _timer functions can't handle a delay of 0 jiffies so ensure
 157	 *  we always return a positive value.
 158	 */
 159	if (interval_jiffies <= 0)
 160		interval_jiffies = 1;
 161	return interval_jiffies;
 162}
 163
 164static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 
 165static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 166static void igmpv3_clear_delrec(struct in_device *in_dev);
 167static int sf_setstate(struct ip_mc_list *pmc);
 168static void sf_markstate(struct ip_mc_list *pmc);
 169#endif
 170static void ip_mc_clear_src(struct ip_mc_list *pmc);
 171static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 172			 int sfcount, __be32 *psfsrc, int delta);
 173
 174static void ip_ma_put(struct ip_mc_list *im)
 175{
 176	if (atomic_dec_and_test(&im->refcnt)) {
 177		in_dev_put(im->interface);
 178		kfree_rcu(im, rcu);
 179	}
 180}
 181
 182#define for_each_pmc_rcu(in_dev, pmc)				\
 183	for (pmc = rcu_dereference(in_dev->mc_list);		\
 184	     pmc != NULL;					\
 185	     pmc = rcu_dereference(pmc->next_rcu))
 186
 187#define for_each_pmc_rtnl(in_dev, pmc)				\
 188	for (pmc = rtnl_dereference(in_dev->mc_list);		\
 189	     pmc != NULL;					\
 190	     pmc = rtnl_dereference(pmc->next_rcu))
 191
 
 
 
 
 
 
 
 
 
 
 
 192#ifdef CONFIG_IP_MULTICAST
 193
 194/*
 195 *	Timer management
 196 */
 197
 198static void igmp_stop_timer(struct ip_mc_list *im)
 199{
 200	spin_lock_bh(&im->lock);
 201	if (del_timer(&im->timer))
 202		atomic_dec(&im->refcnt);
 203	im->tm_running = 0;
 204	im->reporter = 0;
 205	im->unsolicit_count = 0;
 206	spin_unlock_bh(&im->lock);
 207}
 208
 209/* It must be called with locked im->lock */
 210static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 211{
 212	int tv = prandom_u32() % max_delay;
 213
 214	im->tm_running = 1;
 215	if (!mod_timer(&im->timer, jiffies+tv+2))
 216		atomic_inc(&im->refcnt);
 217}
 218
 219static void igmp_gq_start_timer(struct in_device *in_dev)
 220{
 221	int tv = prandom_u32() % in_dev->mr_maxdelay;
 222	unsigned long exp = jiffies + tv + 2;
 223
 224	if (in_dev->mr_gq_running &&
 225	    time_after_eq(exp, (in_dev->mr_gq_timer).expires))
 226		return;
 227
 228	in_dev->mr_gq_running = 1;
 229	if (!mod_timer(&in_dev->mr_gq_timer, exp))
 230		in_dev_hold(in_dev);
 231}
 232
 233static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 234{
 235	int tv = prandom_u32() % delay;
 236
 237	if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 238		in_dev_hold(in_dev);
 239}
 240
 241static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 242{
 243	spin_lock_bh(&im->lock);
 244	im->unsolicit_count = 0;
 245	if (del_timer(&im->timer)) {
 246		if ((long)(im->timer.expires-jiffies) < max_delay) {
 247			add_timer(&im->timer);
 248			im->tm_running = 1;
 249			spin_unlock_bh(&im->lock);
 250			return;
 251		}
 252		atomic_dec(&im->refcnt);
 253	}
 254	igmp_start_timer(im, max_delay);
 255	spin_unlock_bh(&im->lock);
 256}
 257
 258
 259/*
 260 *	Send an IGMP report.
 261 */
 262
 263#define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 264
 265
 266static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 267	int gdeleted, int sdeleted)
 268{
 269	switch (type) {
 270	case IGMPV3_MODE_IS_INCLUDE:
 271	case IGMPV3_MODE_IS_EXCLUDE:
 272		if (gdeleted || sdeleted)
 273			return 0;
 274		if (!(pmc->gsquery && !psf->sf_gsresp)) {
 275			if (pmc->sfmode == MCAST_INCLUDE)
 276				return 1;
 277			/* don't include if this source is excluded
 278			 * in all filters
 279			 */
 280			if (psf->sf_count[MCAST_INCLUDE])
 281				return type == IGMPV3_MODE_IS_INCLUDE;
 282			return pmc->sfcount[MCAST_EXCLUDE] ==
 283				psf->sf_count[MCAST_EXCLUDE];
 284		}
 285		return 0;
 286	case IGMPV3_CHANGE_TO_INCLUDE:
 287		if (gdeleted || sdeleted)
 288			return 0;
 289		return psf->sf_count[MCAST_INCLUDE] != 0;
 290	case IGMPV3_CHANGE_TO_EXCLUDE:
 291		if (gdeleted || sdeleted)
 292			return 0;
 293		if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 294		    psf->sf_count[MCAST_INCLUDE])
 295			return 0;
 296		return pmc->sfcount[MCAST_EXCLUDE] ==
 297			psf->sf_count[MCAST_EXCLUDE];
 298	case IGMPV3_ALLOW_NEW_SOURCES:
 299		if (gdeleted || !psf->sf_crcount)
 300			return 0;
 301		return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 302	case IGMPV3_BLOCK_OLD_SOURCES:
 303		if (pmc->sfmode == MCAST_INCLUDE)
 304			return gdeleted || (psf->sf_crcount && sdeleted);
 305		return psf->sf_crcount && !gdeleted && !sdeleted;
 306	}
 307	return 0;
 308}
 309
 310static int
 311igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 312{
 313	struct ip_sf_list *psf;
 314	int scount = 0;
 315
 316	for (psf = pmc->sources; psf; psf = psf->sf_next) {
 317		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 318			continue;
 319		scount++;
 320	}
 321	return scount;
 322}
 323
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 324static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
 325{
 326	struct sk_buff *skb;
 327	struct rtable *rt;
 328	struct iphdr *pip;
 329	struct igmpv3_report *pig;
 330	struct net *net = dev_net(dev);
 331	struct flowi4 fl4;
 332	int hlen = LL_RESERVED_SPACE(dev);
 333	int tlen = dev->needed_tailroom;
 334	unsigned int size = mtu;
 335
 336	while (1) {
 337		skb = alloc_skb(size + hlen + tlen,
 338				GFP_ATOMIC | __GFP_NOWARN);
 339		if (skb)
 340			break;
 341		size >>= 1;
 342		if (size < 256)
 343			return NULL;
 344	}
 345	skb->priority = TC_PRIO_CONTROL;
 346
 347	rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 348				   0, 0,
 349				   IPPROTO_IGMP, 0, dev->ifindex);
 350	if (IS_ERR(rt)) {
 351		kfree_skb(skb);
 352		return NULL;
 353	}
 354
 355	skb_dst_set(skb, &rt->dst);
 356	skb->dev = dev;
 357
 358	skb_reserve(skb, hlen);
 359	skb_tailroom_reserve(skb, mtu, tlen);
 360
 361	skb_reset_network_header(skb);
 362	pip = ip_hdr(skb);
 363	skb_put(skb, sizeof(struct iphdr) + 4);
 364
 365	pip->version  = 4;
 366	pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 367	pip->tos      = 0xc0;
 368	pip->frag_off = htons(IP_DF);
 369	pip->ttl      = 1;
 370	pip->daddr    = fl4.daddr;
 371	pip->saddr    = fl4.saddr;
 
 
 
 
 372	pip->protocol = IPPROTO_IGMP;
 373	pip->tot_len  = 0;	/* filled in later */
 374	ip_select_ident(net, skb, NULL);
 375	((u8 *)&pip[1])[0] = IPOPT_RA;
 376	((u8 *)&pip[1])[1] = 4;
 377	((u8 *)&pip[1])[2] = 0;
 378	((u8 *)&pip[1])[3] = 0;
 379
 380	skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 381	skb_put(skb, sizeof(*pig));
 382	pig = igmpv3_report_hdr(skb);
 383	pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 384	pig->resv1 = 0;
 385	pig->csum = 0;
 386	pig->resv2 = 0;
 387	pig->ngrec = 0;
 388	return skb;
 389}
 390
 391static int igmpv3_sendpack(struct sk_buff *skb)
 392{
 393	struct igmphdr *pig = igmp_hdr(skb);
 394	const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
 395
 396	pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 397
 398	return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
 399}
 400
 401static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 402{
 403	return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 404}
 405
 406static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 407	int type, struct igmpv3_grec **ppgr)
 408{
 409	struct net_device *dev = pmc->interface->dev;
 410	struct igmpv3_report *pih;
 411	struct igmpv3_grec *pgr;
 412
 413	if (!skb)
 414		skb = igmpv3_newpack(dev, dev->mtu);
 415	if (!skb)
 416		return NULL;
 417	pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 
 418	pgr->grec_type = type;
 419	pgr->grec_auxwords = 0;
 420	pgr->grec_nsrcs = 0;
 421	pgr->grec_mca = pmc->multiaddr;
 422	pih = igmpv3_report_hdr(skb);
 423	pih->ngrec = htons(ntohs(pih->ngrec)+1);
 424	*ppgr = pgr;
 425	return skb;
 426}
 427
 428#define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
 429
 430static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 431	int type, int gdeleted, int sdeleted)
 432{
 433	struct net_device *dev = pmc->interface->dev;
 434	struct net *net = dev_net(dev);
 435	struct igmpv3_report *pih;
 436	struct igmpv3_grec *pgr = NULL;
 437	struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 438	int scount, stotal, first, isquery, truncate;
 
 439
 440	if (pmc->multiaddr == IGMP_ALL_HOSTS)
 441		return skb;
 442	if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
 443		return skb;
 444
 
 
 
 
 445	isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 446		  type == IGMPV3_MODE_IS_EXCLUDE;
 447	truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 448		    type == IGMPV3_CHANGE_TO_EXCLUDE;
 449
 450	stotal = scount = 0;
 451
 452	psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 453
 454	if (!*psf_list)
 455		goto empty_source;
 456
 457	pih = skb ? igmpv3_report_hdr(skb) : NULL;
 458
 459	/* EX and TO_EX get a fresh packet, if needed */
 460	if (truncate) {
 461		if (pih && pih->ngrec &&
 462		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 463			if (skb)
 464				igmpv3_sendpack(skb);
 465			skb = igmpv3_newpack(dev, dev->mtu);
 466		}
 467	}
 468	first = 1;
 469	psf_prev = NULL;
 470	for (psf = *psf_list; psf; psf = psf_next) {
 471		__be32 *psrc;
 472
 473		psf_next = psf->sf_next;
 474
 475		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 476			psf_prev = psf;
 477			continue;
 478		}
 479
 480		/* Based on RFC3376 5.1. Should not send source-list change
 481		 * records when there is a filter mode change.
 482		 */
 483		if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) ||
 484		     (!gdeleted && pmc->crcount)) &&
 485		    (type == IGMPV3_ALLOW_NEW_SOURCES ||
 486		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount)
 487			goto decrease_sf_crcount;
 488
 489		/* clear marks on query responses */
 490		if (isquery)
 491			psf->sf_gsresp = 0;
 492
 493		if (AVAILABLE(skb) < sizeof(__be32) +
 494		    first*sizeof(struct igmpv3_grec)) {
 495			if (truncate && !first)
 496				break;	 /* truncate these */
 497			if (pgr)
 498				pgr->grec_nsrcs = htons(scount);
 499			if (skb)
 500				igmpv3_sendpack(skb);
 501			skb = igmpv3_newpack(dev, dev->mtu);
 502			first = 1;
 503			scount = 0;
 504		}
 505		if (first) {
 506			skb = add_grhead(skb, pmc, type, &pgr);
 507			first = 0;
 508		}
 509		if (!skb)
 510			return NULL;
 511		psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 512		*psrc = psf->sf_inaddr;
 513		scount++; stotal++;
 514		if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 515		     type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 516decrease_sf_crcount:
 517			psf->sf_crcount--;
 518			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 519				if (psf_prev)
 520					psf_prev->sf_next = psf->sf_next;
 521				else
 522					*psf_list = psf->sf_next;
 523				kfree(psf);
 524				continue;
 525			}
 526		}
 527		psf_prev = psf;
 528	}
 529
 530empty_source:
 531	if (!stotal) {
 532		if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 533		    type == IGMPV3_BLOCK_OLD_SOURCES)
 534			return skb;
 535		if (pmc->crcount || isquery) {
 536			/* make sure we have room for group header */
 537			if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
 538				igmpv3_sendpack(skb);
 539				skb = NULL; /* add_grhead will get a new one */
 540			}
 541			skb = add_grhead(skb, pmc, type, &pgr);
 542		}
 543	}
 544	if (pgr)
 545		pgr->grec_nsrcs = htons(scount);
 546
 547	if (isquery)
 548		pmc->gsquery = 0;	/* clear query state on report */
 549	return skb;
 550}
 551
 552static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 553{
 554	struct sk_buff *skb = NULL;
 555	struct net *net = dev_net(in_dev->dev);
 556	int type;
 557
 558	if (!pmc) {
 559		rcu_read_lock();
 560		for_each_pmc_rcu(in_dev, pmc) {
 561			if (pmc->multiaddr == IGMP_ALL_HOSTS)
 562				continue;
 563			if (ipv4_is_local_multicast(pmc->multiaddr) &&
 564			     !net->ipv4.sysctl_igmp_llm_reports)
 565				continue;
 566			spin_lock_bh(&pmc->lock);
 567			if (pmc->sfcount[MCAST_EXCLUDE])
 568				type = IGMPV3_MODE_IS_EXCLUDE;
 569			else
 570				type = IGMPV3_MODE_IS_INCLUDE;
 571			skb = add_grec(skb, pmc, type, 0, 0);
 572			spin_unlock_bh(&pmc->lock);
 573		}
 574		rcu_read_unlock();
 575	} else {
 576		spin_lock_bh(&pmc->lock);
 577		if (pmc->sfcount[MCAST_EXCLUDE])
 578			type = IGMPV3_MODE_IS_EXCLUDE;
 579		else
 580			type = IGMPV3_MODE_IS_INCLUDE;
 581		skb = add_grec(skb, pmc, type, 0, 0);
 582		spin_unlock_bh(&pmc->lock);
 583	}
 584	if (!skb)
 585		return 0;
 586	return igmpv3_sendpack(skb);
 587}
 588
 589/*
 590 * remove zero-count source records from a source filter list
 591 */
 592static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 593{
 594	struct ip_sf_list *psf_prev, *psf_next, *psf;
 595
 596	psf_prev = NULL;
 597	for (psf = *ppsf; psf; psf = psf_next) {
 598		psf_next = psf->sf_next;
 599		if (psf->sf_crcount == 0) {
 600			if (psf_prev)
 601				psf_prev->sf_next = psf->sf_next;
 602			else
 603				*ppsf = psf->sf_next;
 604			kfree(psf);
 605		} else
 606			psf_prev = psf;
 607	}
 608}
 609
 
 
 
 
 
 
 
 610static void igmpv3_send_cr(struct in_device *in_dev)
 611{
 612	struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 613	struct sk_buff *skb = NULL;
 614	int type, dtype;
 615
 616	rcu_read_lock();
 617	spin_lock_bh(&in_dev->mc_tomb_lock);
 618
 619	/* deleted MCA's */
 620	pmc_prev = NULL;
 621	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
 622		pmc_next = pmc->next;
 623		if (pmc->sfmode == MCAST_INCLUDE) {
 624			type = IGMPV3_BLOCK_OLD_SOURCES;
 625			dtype = IGMPV3_BLOCK_OLD_SOURCES;
 626			skb = add_grec(skb, pmc, type, 1, 0);
 627			skb = add_grec(skb, pmc, dtype, 1, 1);
 628		}
 629		if (pmc->crcount) {
 630			if (pmc->sfmode == MCAST_EXCLUDE) {
 631				type = IGMPV3_CHANGE_TO_INCLUDE;
 632				skb = add_grec(skb, pmc, type, 1, 0);
 633			}
 634			pmc->crcount--;
 635			if (pmc->crcount == 0) {
 636				igmpv3_clear_zeros(&pmc->tomb);
 637				igmpv3_clear_zeros(&pmc->sources);
 638			}
 639		}
 640		if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 641			if (pmc_prev)
 642				pmc_prev->next = pmc_next;
 643			else
 644				in_dev->mc_tomb = pmc_next;
 645			in_dev_put(pmc->interface);
 646			kfree(pmc);
 647		} else
 648			pmc_prev = pmc;
 649	}
 650	spin_unlock_bh(&in_dev->mc_tomb_lock);
 651
 652	/* change recs */
 653	for_each_pmc_rcu(in_dev, pmc) {
 654		spin_lock_bh(&pmc->lock);
 655		if (pmc->sfcount[MCAST_EXCLUDE]) {
 656			type = IGMPV3_BLOCK_OLD_SOURCES;
 657			dtype = IGMPV3_ALLOW_NEW_SOURCES;
 658		} else {
 659			type = IGMPV3_ALLOW_NEW_SOURCES;
 660			dtype = IGMPV3_BLOCK_OLD_SOURCES;
 661		}
 662		skb = add_grec(skb, pmc, type, 0, 0);
 663		skb = add_grec(skb, pmc, dtype, 0, 1);	/* deleted sources */
 664
 665		/* filter mode changes */
 666		if (pmc->crcount) {
 667			if (pmc->sfmode == MCAST_EXCLUDE)
 668				type = IGMPV3_CHANGE_TO_EXCLUDE;
 669			else
 670				type = IGMPV3_CHANGE_TO_INCLUDE;
 671			skb = add_grec(skb, pmc, type, 0, 0);
 672			pmc->crcount--;
 673		}
 674		spin_unlock_bh(&pmc->lock);
 675	}
 676	rcu_read_unlock();
 677
 678	if (!skb)
 679		return;
 680	(void) igmpv3_sendpack(skb);
 681}
 682
 683static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 684	int type)
 685{
 686	struct sk_buff *skb;
 687	struct iphdr *iph;
 688	struct igmphdr *ih;
 689	struct rtable *rt;
 690	struct net_device *dev = in_dev->dev;
 691	struct net *net = dev_net(dev);
 692	__be32	group = pmc ? pmc->multiaddr : 0;
 693	struct flowi4 fl4;
 694	__be32	dst;
 695	int hlen, tlen;
 696
 697	if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 698		return igmpv3_send_report(in_dev, pmc);
 699
 700	if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
 701		return 0;
 702
 703	if (type == IGMP_HOST_LEAVE_MESSAGE)
 704		dst = IGMP_ALL_ROUTER;
 705	else
 706		dst = group;
 707
 708	rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 709				   0, 0,
 710				   IPPROTO_IGMP, 0, dev->ifindex);
 711	if (IS_ERR(rt))
 712		return -1;
 713
 714	hlen = LL_RESERVED_SPACE(dev);
 715	tlen = dev->needed_tailroom;
 716	skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 717	if (!skb) {
 718		ip_rt_put(rt);
 719		return -1;
 720	}
 721	skb->priority = TC_PRIO_CONTROL;
 722
 723	skb_dst_set(skb, &rt->dst);
 724
 725	skb_reserve(skb, hlen);
 726
 727	skb_reset_network_header(skb);
 728	iph = ip_hdr(skb);
 729	skb_put(skb, sizeof(struct iphdr) + 4);
 730
 731	iph->version  = 4;
 732	iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 733	iph->tos      = 0xc0;
 734	iph->frag_off = htons(IP_DF);
 735	iph->ttl      = 1;
 736	iph->daddr    = dst;
 737	iph->saddr    = fl4.saddr;
 738	iph->protocol = IPPROTO_IGMP;
 739	ip_select_ident(net, skb, NULL);
 740	((u8 *)&iph[1])[0] = IPOPT_RA;
 741	((u8 *)&iph[1])[1] = 4;
 742	((u8 *)&iph[1])[2] = 0;
 743	((u8 *)&iph[1])[3] = 0;
 744
 745	ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 746	ih->type = type;
 747	ih->code = 0;
 748	ih->csum = 0;
 749	ih->group = group;
 750	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 751
 752	return ip_local_out(net, skb->sk, skb);
 753}
 754
 755static void igmp_gq_timer_expire(unsigned long data)
 756{
 757	struct in_device *in_dev = (struct in_device *)data;
 758
 759	in_dev->mr_gq_running = 0;
 760	igmpv3_send_report(in_dev, NULL);
 761	in_dev_put(in_dev);
 762}
 763
 764static void igmp_ifc_timer_expire(unsigned long data)
 765{
 766	struct in_device *in_dev = (struct in_device *)data;
 
 767
 768	igmpv3_send_cr(in_dev);
 769	if (in_dev->mr_ifc_count) {
 770		in_dev->mr_ifc_count--;
 
 
 
 
 
 
 771		igmp_ifc_start_timer(in_dev,
 772				     unsolicited_report_interval(in_dev));
 773	}
 774	in_dev_put(in_dev);
 775}
 776
 777static void igmp_ifc_event(struct in_device *in_dev)
 778{
 779	struct net *net = dev_net(in_dev->dev);
 780	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 781		return;
 782	in_dev->mr_ifc_count = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
 783	igmp_ifc_start_timer(in_dev, 1);
 784}
 785
 786
 787static void igmp_timer_expire(unsigned long data)
 788{
 789	struct ip_mc_list *im = (struct ip_mc_list *)data;
 790	struct in_device *in_dev = im->interface;
 791
 792	spin_lock(&im->lock);
 793	im->tm_running = 0;
 794
 795	if (im->unsolicit_count) {
 796		im->unsolicit_count--;
 797		igmp_start_timer(im, unsolicited_report_interval(in_dev));
 798	}
 799	im->reporter = 1;
 800	spin_unlock(&im->lock);
 801
 802	if (IGMP_V1_SEEN(in_dev))
 803		igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 804	else if (IGMP_V2_SEEN(in_dev))
 805		igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 806	else
 807		igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 808
 809	ip_ma_put(im);
 810}
 811
 812/* mark EXCLUDE-mode sources */
 813static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 814{
 815	struct ip_sf_list *psf;
 816	int i, scount;
 817
 818	scount = 0;
 819	for (psf = pmc->sources; psf; psf = psf->sf_next) {
 820		if (scount == nsrcs)
 821			break;
 822		for (i = 0; i < nsrcs; i++) {
 823			/* skip inactive filters */
 824			if (psf->sf_count[MCAST_INCLUDE] ||
 825			    pmc->sfcount[MCAST_EXCLUDE] !=
 826			    psf->sf_count[MCAST_EXCLUDE])
 827				break;
 828			if (srcs[i] == psf->sf_inaddr) {
 829				scount++;
 830				break;
 831			}
 832		}
 833	}
 834	pmc->gsquery = 0;
 835	if (scount == nsrcs)	/* all sources excluded */
 836		return 0;
 837	return 1;
 838}
 839
 840static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 841{
 842	struct ip_sf_list *psf;
 843	int i, scount;
 844
 845	if (pmc->sfmode == MCAST_EXCLUDE)
 846		return igmp_xmarksources(pmc, nsrcs, srcs);
 847
 848	/* mark INCLUDE-mode sources */
 849	scount = 0;
 850	for (psf = pmc->sources; psf; psf = psf->sf_next) {
 851		if (scount == nsrcs)
 852			break;
 853		for (i = 0; i < nsrcs; i++)
 854			if (srcs[i] == psf->sf_inaddr) {
 855				psf->sf_gsresp = 1;
 856				scount++;
 857				break;
 858			}
 859	}
 860	if (!scount) {
 861		pmc->gsquery = 0;
 862		return 0;
 863	}
 864	pmc->gsquery = 1;
 865	return 1;
 866}
 867
 868/* return true if packet was dropped */
 869static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
 870{
 871	struct ip_mc_list *im;
 872	struct net *net = dev_net(in_dev->dev);
 873
 874	/* Timers are only set for non-local groups */
 875
 876	if (group == IGMP_ALL_HOSTS)
 877		return false;
 878	if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
 879		return false;
 880
 881	rcu_read_lock();
 882	for_each_pmc_rcu(in_dev, im) {
 883		if (im->multiaddr == group) {
 884			igmp_stop_timer(im);
 885			break;
 886		}
 887	}
 888	rcu_read_unlock();
 889	return false;
 890}
 891
 892/* return true if packet was dropped */
 893static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 894	int len)
 895{
 896	struct igmphdr 		*ih = igmp_hdr(skb);
 897	struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 898	struct ip_mc_list	*im;
 899	__be32			group = ih->group;
 900	int			max_delay;
 901	int			mark = 0;
 902	struct net		*net = dev_net(in_dev->dev);
 903
 904
 905	if (len == 8) {
 906		if (ih->code == 0) {
 907			/* Alas, old v1 router presents here. */
 908
 909			max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 910			in_dev->mr_v1_seen = jiffies +
 911				IGMP_V1_ROUTER_PRESENT_TIMEOUT;
 
 912			group = 0;
 913		} else {
 914			/* v2 router present */
 915			max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 916			in_dev->mr_v2_seen = jiffies +
 917				IGMP_V2_ROUTER_PRESENT_TIMEOUT;
 
 918		}
 919		/* cancel the interface change timer */
 920		in_dev->mr_ifc_count = 0;
 921		if (del_timer(&in_dev->mr_ifc_timer))
 922			__in_dev_put(in_dev);
 923		/* clear deleted report items */
 924		igmpv3_clear_delrec(in_dev);
 925	} else if (len < 12) {
 926		return true;	/* ignore bogus packet; freed by caller */
 927	} else if (IGMP_V1_SEEN(in_dev)) {
 928		/* This is a v3 query with v1 queriers present */
 929		max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 930		group = 0;
 931	} else if (IGMP_V2_SEEN(in_dev)) {
 932		/* this is a v3 query with v2 queriers present;
 933		 * Interpretation of the max_delay code is problematic here.
 934		 * A real v2 host would use ih_code directly, while v3 has a
 935		 * different encoding. We use the v3 encoding as more likely
 936		 * to be intended in a v3 query.
 937		 */
 938		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 939		if (!max_delay)
 940			max_delay = 1;	/* can't mod w/ 0 */
 941	} else { /* v3 */
 942		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 943			return true;
 944
 945		ih3 = igmpv3_query_hdr(skb);
 946		if (ih3->nsrcs) {
 947			if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 948					   + ntohs(ih3->nsrcs)*sizeof(__be32)))
 949				return true;
 950			ih3 = igmpv3_query_hdr(skb);
 951		}
 952
 953		max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 954		if (!max_delay)
 955			max_delay = 1;	/* can't mod w/ 0 */
 956		in_dev->mr_maxdelay = max_delay;
 957		if (ih3->qrv)
 958			in_dev->mr_qrv = ih3->qrv;
 
 
 
 
 
 
 
 
 
 
 
 
 
 959		if (!group) { /* general query */
 960			if (ih3->nsrcs)
 961				return true;	/* no sources allowed */
 962			igmp_gq_start_timer(in_dev);
 963			return false;
 964		}
 965		/* mark sources to include, if group & source-specific */
 966		mark = ih3->nsrcs != 0;
 967	}
 968
 969	/*
 970	 * - Start the timers in all of our membership records
 971	 *   that the query applies to for the interface on
 972	 *   which the query arrived excl. those that belong
 973	 *   to a "local" group (224.0.0.X)
 974	 * - For timers already running check if they need to
 975	 *   be reset.
 976	 * - Use the igmp->igmp_code field as the maximum
 977	 *   delay possible
 978	 */
 979	rcu_read_lock();
 980	for_each_pmc_rcu(in_dev, im) {
 981		int changed;
 982
 983		if (group && group != im->multiaddr)
 984			continue;
 985		if (im->multiaddr == IGMP_ALL_HOSTS)
 986			continue;
 987		if (ipv4_is_local_multicast(im->multiaddr) &&
 988		    !net->ipv4.sysctl_igmp_llm_reports)
 989			continue;
 990		spin_lock_bh(&im->lock);
 991		if (im->tm_running)
 992			im->gsquery = im->gsquery && mark;
 993		else
 994			im->gsquery = mark;
 995		changed = !im->gsquery ||
 996			igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 997		spin_unlock_bh(&im->lock);
 998		if (changed)
 999			igmp_mod_timer(im, max_delay);
1000	}
1001	rcu_read_unlock();
1002	return false;
1003}
1004
1005/* called in rcu_read_lock() section */
1006int igmp_rcv(struct sk_buff *skb)
1007{
1008	/* This basically follows the spec line by line -- see RFC1112 */
1009	struct igmphdr *ih;
1010	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 
1011	int len = skb->len;
1012	bool dropped = true;
1013
 
 
 
 
 
 
 
1014	if (!in_dev)
1015		goto drop;
1016
1017	if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1018		goto drop;
1019
1020	if (skb_checksum_simple_validate(skb))
1021		goto drop;
1022
1023	ih = igmp_hdr(skb);
1024	switch (ih->type) {
1025	case IGMP_HOST_MEMBERSHIP_QUERY:
1026		dropped = igmp_heard_query(in_dev, skb, len);
1027		break;
1028	case IGMP_HOST_MEMBERSHIP_REPORT:
1029	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1030		/* Is it our report looped back? */
1031		if (rt_is_output_route(skb_rtable(skb)))
1032			break;
1033		/* don't rely on MC router hearing unicast reports */
1034		if (skb->pkt_type == PACKET_MULTICAST ||
1035		    skb->pkt_type == PACKET_BROADCAST)
1036			dropped = igmp_heard_report(in_dev, ih->group);
1037		break;
1038	case IGMP_PIM:
1039#ifdef CONFIG_IP_PIMSM_V1
1040		return pim_rcv_v1(skb);
1041#endif
1042	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1043	case IGMP_DVMRP:
1044	case IGMP_TRACE:
1045	case IGMP_HOST_LEAVE_MESSAGE:
1046	case IGMP_MTRACE:
1047	case IGMP_MTRACE_RESP:
1048		break;
1049	default:
1050		break;
1051	}
1052
1053drop:
1054	if (dropped)
1055		kfree_skb(skb);
1056	else
1057		consume_skb(skb);
1058	return 0;
1059}
1060
1061#endif
1062
1063
1064/*
1065 *	Add a filter to a device
1066 */
1067
1068static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1069{
1070	char buf[MAX_ADDR_LEN];
1071	struct net_device *dev = in_dev->dev;
1072
1073	/* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1074	   We will get multicast token leakage, when IFF_MULTICAST
1075	   is changed. This check should be done in ndo_set_rx_mode
1076	   routine. Something sort of:
1077	   if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1078	   --ANK
1079	   */
1080	if (arp_mc_map(addr, buf, dev, 0) == 0)
1081		dev_mc_add(dev, buf);
1082}
1083
1084/*
1085 *	Remove a filter from a device
1086 */
1087
1088static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1089{
1090	char buf[MAX_ADDR_LEN];
1091	struct net_device *dev = in_dev->dev;
1092
1093	if (arp_mc_map(addr, buf, dev, 0) == 0)
1094		dev_mc_del(dev, buf);
1095}
1096
1097#ifdef CONFIG_IP_MULTICAST
1098/*
1099 * deleted ip_mc_list manipulation
1100 */
1101static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
 
1102{
1103	struct ip_mc_list *pmc;
1104	struct net *net = dev_net(in_dev->dev);
1105
1106	/* this is an "ip_mc_list" for convenience; only the fields below
1107	 * are actually used. In particular, the refcnt and users are not
1108	 * used for management of the delete list. Using the same structure
1109	 * for deleted items allows change reports to use common code with
1110	 * non-deleted or query-response MCA's.
1111	 */
1112	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1113	if (!pmc)
1114		return;
 
1115	spin_lock_bh(&im->lock);
1116	pmc->interface = im->interface;
1117	in_dev_hold(in_dev);
1118	pmc->multiaddr = im->multiaddr;
1119	pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1120	pmc->sfmode = im->sfmode;
1121	if (pmc->sfmode == MCAST_INCLUDE) {
1122		struct ip_sf_list *psf;
1123
1124		pmc->tomb = im->tomb;
1125		pmc->sources = im->sources;
1126		im->tomb = im->sources = NULL;
1127		for (psf = pmc->sources; psf; psf = psf->sf_next)
1128			psf->sf_crcount = pmc->crcount;
1129	}
1130	spin_unlock_bh(&im->lock);
1131
1132	spin_lock_bh(&in_dev->mc_tomb_lock);
1133	pmc->next = in_dev->mc_tomb;
1134	in_dev->mc_tomb = pmc;
1135	spin_unlock_bh(&in_dev->mc_tomb_lock);
1136}
1137
1138/*
1139 * restore ip_mc_list deleted records
1140 */
1141static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1142{
1143	struct ip_mc_list *pmc, *pmc_prev;
1144	struct ip_sf_list *psf;
1145	struct net *net = dev_net(in_dev->dev);
1146	__be32 multiaddr = im->multiaddr;
1147
1148	spin_lock_bh(&in_dev->mc_tomb_lock);
1149	pmc_prev = NULL;
1150	for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1151		if (pmc->multiaddr == multiaddr)
1152			break;
1153		pmc_prev = pmc;
1154	}
1155	if (pmc) {
1156		if (pmc_prev)
1157			pmc_prev->next = pmc->next;
1158		else
1159			in_dev->mc_tomb = pmc->next;
1160	}
1161	spin_unlock_bh(&in_dev->mc_tomb_lock);
1162
1163	spin_lock_bh(&im->lock);
1164	if (pmc) {
1165		im->interface = pmc->interface;
1166		im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1167		im->sfmode = pmc->sfmode;
1168		if (pmc->sfmode == MCAST_INCLUDE) {
1169			im->tomb = pmc->tomb;
1170			im->sources = pmc->sources;
1171			for (psf = im->sources; psf; psf = psf->sf_next)
1172				psf->sf_crcount = im->crcount;
 
 
1173		}
1174		in_dev_put(pmc->interface);
1175		kfree(pmc);
1176	}
1177	spin_unlock_bh(&im->lock);
1178}
1179
1180/*
1181 * flush ip_mc_list deleted records
1182 */
1183static void igmpv3_clear_delrec(struct in_device *in_dev)
1184{
1185	struct ip_mc_list *pmc, *nextpmc;
1186
1187	spin_lock_bh(&in_dev->mc_tomb_lock);
1188	pmc = in_dev->mc_tomb;
1189	in_dev->mc_tomb = NULL;
1190	spin_unlock_bh(&in_dev->mc_tomb_lock);
1191
1192	for (; pmc; pmc = nextpmc) {
1193		nextpmc = pmc->next;
1194		ip_mc_clear_src(pmc);
1195		in_dev_put(pmc->interface);
1196		kfree(pmc);
1197	}
1198	/* clear dead sources, too */
1199	rcu_read_lock();
1200	for_each_pmc_rcu(in_dev, pmc) {
1201		struct ip_sf_list *psf, *psf_next;
1202
1203		spin_lock_bh(&pmc->lock);
1204		psf = pmc->tomb;
1205		pmc->tomb = NULL;
1206		spin_unlock_bh(&pmc->lock);
1207		for (; psf; psf = psf_next) {
1208			psf_next = psf->sf_next;
1209			kfree(psf);
1210		}
1211	}
1212	rcu_read_unlock();
1213}
1214#endif
1215
1216static void igmp_group_dropped(struct ip_mc_list *im)
1217{
1218	struct in_device *in_dev = im->interface;
1219#ifdef CONFIG_IP_MULTICAST
1220	struct net *net = dev_net(in_dev->dev);
1221	int reporter;
1222#endif
1223
1224	if (im->loaded) {
1225		im->loaded = 0;
1226		ip_mc_filter_del(in_dev, im->multiaddr);
1227	}
1228
1229#ifdef CONFIG_IP_MULTICAST
1230	if (im->multiaddr == IGMP_ALL_HOSTS)
1231		return;
1232	if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1233		return;
1234
1235	reporter = im->reporter;
1236	igmp_stop_timer(im);
1237
1238	if (!in_dev->dead) {
1239		if (IGMP_V1_SEEN(in_dev))
1240			return;
1241		if (IGMP_V2_SEEN(in_dev)) {
1242			if (reporter)
1243				igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1244			return;
1245		}
1246		/* IGMPv3 */
1247		igmpv3_add_delrec(in_dev, im);
1248
1249		igmp_ifc_event(in_dev);
1250	}
1251#endif
1252}
1253
 
 
 
 
 
1254static void igmp_group_added(struct ip_mc_list *im)
1255{
1256	struct in_device *in_dev = im->interface;
1257#ifdef CONFIG_IP_MULTICAST
1258	struct net *net = dev_net(in_dev->dev);
1259#endif
1260
1261	if (im->loaded == 0) {
1262		im->loaded = 1;
1263		ip_mc_filter_add(in_dev, im->multiaddr);
1264	}
1265
1266#ifdef CONFIG_IP_MULTICAST
1267	if (im->multiaddr == IGMP_ALL_HOSTS)
1268		return;
1269	if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1270		return;
1271
1272	if (in_dev->dead)
1273		return;
 
 
1274	if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1275		spin_lock_bh(&im->lock);
1276		igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1277		spin_unlock_bh(&im->lock);
1278		return;
1279	}
1280	/* else, v3 */
1281
1282	im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
 
 
 
 
 
 
1283	igmp_ifc_event(in_dev);
1284#endif
1285}
1286
1287
1288/*
1289 *	Multicast list managers
1290 */
1291
1292static u32 ip_mc_hash(const struct ip_mc_list *im)
1293{
1294	return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1295}
1296
1297static void ip_mc_hash_add(struct in_device *in_dev,
1298			   struct ip_mc_list *im)
1299{
1300	struct ip_mc_list __rcu **mc_hash;
1301	u32 hash;
1302
1303	mc_hash = rtnl_dereference(in_dev->mc_hash);
1304	if (mc_hash) {
1305		hash = ip_mc_hash(im);
1306		im->next_hash = mc_hash[hash];
1307		rcu_assign_pointer(mc_hash[hash], im);
1308		return;
1309	}
1310
1311	/* do not use a hash table for small number of items */
1312	if (in_dev->mc_count < 4)
1313		return;
1314
1315	mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1316			  GFP_KERNEL);
1317	if (!mc_hash)
1318		return;
1319
1320	for_each_pmc_rtnl(in_dev, im) {
1321		hash = ip_mc_hash(im);
1322		im->next_hash = mc_hash[hash];
1323		RCU_INIT_POINTER(mc_hash[hash], im);
1324	}
1325
1326	rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1327}
1328
1329static void ip_mc_hash_remove(struct in_device *in_dev,
1330			      struct ip_mc_list *im)
1331{
1332	struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1333	struct ip_mc_list *aux;
1334
1335	if (!mc_hash)
1336		return;
1337	mc_hash += ip_mc_hash(im);
1338	while ((aux = rtnl_dereference(*mc_hash)) != im)
1339		mc_hash = &aux->next_hash;
1340	*mc_hash = im->next_hash;
1341}
1342
1343
1344/*
1345 *	A socket has joined a multicast group on device dev.
1346 */
1347
1348void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1349{
1350	struct ip_mc_list *im;
1351#ifdef CONFIG_IP_MULTICAST
1352	struct net *net = dev_net(in_dev->dev);
1353#endif
1354
1355	ASSERT_RTNL();
1356
1357	for_each_pmc_rtnl(in_dev, im) {
1358		if (im->multiaddr == addr) {
1359			im->users++;
1360			ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1361			goto out;
1362		}
1363	}
1364
1365	im = kzalloc(sizeof(*im), GFP_KERNEL);
1366	if (!im)
1367		goto out;
1368
1369	im->users = 1;
1370	im->interface = in_dev;
1371	in_dev_hold(in_dev);
1372	im->multiaddr = addr;
1373	/* initial mode is (EX, empty) */
1374	im->sfmode = MCAST_EXCLUDE;
1375	im->sfcount[MCAST_EXCLUDE] = 1;
1376	atomic_set(&im->refcnt, 1);
1377	spin_lock_init(&im->lock);
1378#ifdef CONFIG_IP_MULTICAST
1379	setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1380	im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1381#endif
1382
1383	im->next_rcu = in_dev->mc_list;
1384	in_dev->mc_count++;
1385	rcu_assign_pointer(in_dev->mc_list, im);
1386
1387	ip_mc_hash_add(in_dev, im);
1388
1389#ifdef CONFIG_IP_MULTICAST
1390	igmpv3_del_delrec(in_dev, im);
1391#endif
1392	igmp_group_added(im);
1393	if (!in_dev->dead)
1394		ip_rt_multicast_event(in_dev);
1395out:
1396	return;
1397}
 
 
 
 
 
 
 
 
 
 
 
1398EXPORT_SYMBOL(ip_mc_inc_group);
1399
1400static int ip_mc_check_iphdr(struct sk_buff *skb)
1401{
1402	const struct iphdr *iph;
1403	unsigned int len;
1404	unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1405
1406	if (!pskb_may_pull(skb, offset))
1407		return -EINVAL;
1408
1409	iph = ip_hdr(skb);
1410
1411	if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1412		return -EINVAL;
1413
1414	offset += ip_hdrlen(skb) - sizeof(*iph);
1415
1416	if (!pskb_may_pull(skb, offset))
1417		return -EINVAL;
1418
1419	iph = ip_hdr(skb);
1420
1421	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1422		return -EINVAL;
1423
1424	len = skb_network_offset(skb) + ntohs(iph->tot_len);
1425	if (skb->len < len || len < offset)
1426		return -EINVAL;
1427
1428	skb_set_transport_header(skb, offset);
1429
1430	return 0;
1431}
1432
1433static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1434{
1435	unsigned int len = skb_transport_offset(skb);
1436
1437	len += sizeof(struct igmpv3_report);
1438
1439	return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1440}
1441
1442static int ip_mc_check_igmp_query(struct sk_buff *skb)
1443{
1444	unsigned int len = skb_transport_offset(skb);
1445
1446	len += sizeof(struct igmphdr);
1447	if (skb->len < len)
1448		return -EINVAL;
1449
1450	/* IGMPv{1,2}? */
1451	if (skb->len != len) {
1452		/* or IGMPv3? */
1453		len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1454		if (skb->len < len || !pskb_may_pull(skb, len))
 
 
 
1455			return -EINVAL;
1456	}
1457
1458	/* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1459	 * all-systems destination addresses (224.0.0.1) for general queries
1460	 */
1461	if (!igmp_hdr(skb)->group &&
1462	    ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1463		return -EINVAL;
1464
1465	return 0;
1466}
1467
1468static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1469{
1470	switch (igmp_hdr(skb)->type) {
1471	case IGMP_HOST_LEAVE_MESSAGE:
1472	case IGMP_HOST_MEMBERSHIP_REPORT:
1473	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1474		/* fall through */
1475		return 0;
1476	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1477		return ip_mc_check_igmp_reportv3(skb);
1478	case IGMP_HOST_MEMBERSHIP_QUERY:
1479		return ip_mc_check_igmp_query(skb);
1480	default:
1481		return -ENOMSG;
1482	}
1483}
1484
1485static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1486{
1487	return skb_checksum_simple_validate(skb);
1488}
1489
1490static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1491
1492{
 
 
1493	struct sk_buff *skb_chk;
1494	unsigned int transport_len;
1495	unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1496	int ret = -EINVAL;
1497
1498	transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
 
1499
1500	skb_chk = skb_checksum_trimmed(skb, transport_len,
1501				       ip_mc_validate_checksum);
1502	if (!skb_chk)
1503		goto err;
1504
1505	if (!pskb_may_pull(skb_chk, len))
1506		goto err;
1507
1508	ret = ip_mc_check_igmp_msg(skb_chk);
1509	if (ret)
1510		goto err;
1511
1512	if (skb_trimmed)
1513		*skb_trimmed = skb_chk;
1514	/* free now unneeded clone */
1515	else if (skb_chk != skb)
1516		kfree_skb(skb_chk);
1517
1518	ret = 0;
1519
1520err:
1521	if (ret && skb_chk && skb_chk != skb)
1522		kfree_skb(skb_chk);
1523
1524	return ret;
1525}
1526
1527/**
1528 * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1529 * @skb: the skb to validate
1530 * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1531 *
1532 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1533 * skb transport header accordingly and returns zero.
1534 *
1535 * -EINVAL: A broken packet was detected, i.e. it violates some internet
1536 *  standard
1537 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1538 * -ENOMEM: A memory allocation failure happened.
1539 *
1540 * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1541 * to NULL): After parsing an IGMP packet successfully it will point to
1542 * an skb which has its tail aligned to the IP packet end. This might
1543 * either be the originally provided skb or a trimmed, cloned version if
1544 * the skb frame had data beyond the IP packet. A cloned skb allows us
1545 * to leave the original skb and its full frame unchanged (which might be
1546 * desirable for layer 2 frame jugglers).
1547 *
1548 * Caller needs to set the skb network header and free any returned skb if it
1549 * differs from the provided skb.
1550 */
1551int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1552{
1553	int ret = ip_mc_check_iphdr(skb);
1554
1555	if (ret < 0)
1556		return ret;
1557
1558	if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1559		return -ENOMSG;
1560
1561	return __ip_mc_check_igmp(skb, skb_trimmed);
 
 
 
 
1562}
1563EXPORT_SYMBOL(ip_mc_check_igmp);
1564
1565/*
1566 *	Resend IGMP JOIN report; used by netdev notifier.
1567 */
1568static void ip_mc_rejoin_groups(struct in_device *in_dev)
1569{
1570#ifdef CONFIG_IP_MULTICAST
1571	struct ip_mc_list *im;
1572	int type;
1573	struct net *net = dev_net(in_dev->dev);
1574
1575	ASSERT_RTNL();
1576
1577	for_each_pmc_rtnl(in_dev, im) {
1578		if (im->multiaddr == IGMP_ALL_HOSTS)
1579			continue;
1580		if (ipv4_is_local_multicast(im->multiaddr) &&
1581		    !net->ipv4.sysctl_igmp_llm_reports)
1582			continue;
1583
1584		/* a failover is happening and switches
1585		 * must be notified immediately
1586		 */
1587		if (IGMP_V1_SEEN(in_dev))
1588			type = IGMP_HOST_MEMBERSHIP_REPORT;
1589		else if (IGMP_V2_SEEN(in_dev))
1590			type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1591		else
1592			type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1593		igmp_send_report(in_dev, im, type);
1594	}
1595#endif
1596}
1597
1598/*
1599 *	A socket has left a multicast group on device dev
1600 */
1601
1602void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1603{
1604	struct ip_mc_list *i;
1605	struct ip_mc_list __rcu **ip;
1606
1607	ASSERT_RTNL();
1608
1609	for (ip = &in_dev->mc_list;
1610	     (i = rtnl_dereference(*ip)) != NULL;
1611	     ip = &i->next_rcu) {
1612		if (i->multiaddr == addr) {
1613			if (--i->users == 0) {
1614				ip_mc_hash_remove(in_dev, i);
1615				*ip = i->next_rcu;
1616				in_dev->mc_count--;
1617				igmp_group_dropped(i);
1618				ip_mc_clear_src(i);
1619
1620				if (!in_dev->dead)
1621					ip_rt_multicast_event(in_dev);
1622
1623				ip_ma_put(i);
1624				return;
1625			}
1626			break;
1627		}
1628	}
1629}
1630EXPORT_SYMBOL(ip_mc_dec_group);
1631
1632/* Device changing type */
1633
1634void ip_mc_unmap(struct in_device *in_dev)
1635{
1636	struct ip_mc_list *pmc;
1637
1638	ASSERT_RTNL();
1639
1640	for_each_pmc_rtnl(in_dev, pmc)
1641		igmp_group_dropped(pmc);
1642}
1643
1644void ip_mc_remap(struct in_device *in_dev)
1645{
1646	struct ip_mc_list *pmc;
1647
1648	ASSERT_RTNL();
1649
1650	for_each_pmc_rtnl(in_dev, pmc) {
1651#ifdef CONFIG_IP_MULTICAST
1652		igmpv3_del_delrec(in_dev, pmc);
1653#endif
1654		igmp_group_added(pmc);
1655	}
1656}
1657
1658/* Device going down */
1659
1660void ip_mc_down(struct in_device *in_dev)
1661{
1662	struct ip_mc_list *pmc;
1663
1664	ASSERT_RTNL();
1665
1666	for_each_pmc_rtnl(in_dev, pmc)
1667		igmp_group_dropped(pmc);
1668
1669#ifdef CONFIG_IP_MULTICAST
1670	in_dev->mr_ifc_count = 0;
1671	if (del_timer(&in_dev->mr_ifc_timer))
1672		__in_dev_put(in_dev);
1673	in_dev->mr_gq_running = 0;
1674	if (del_timer(&in_dev->mr_gq_timer))
1675		__in_dev_put(in_dev);
1676#endif
1677
1678	ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1679}
1680
1681void ip_mc_init_dev(struct in_device *in_dev)
 
1682{
1683#ifdef CONFIG_IP_MULTICAST
1684	struct net *net = dev_net(in_dev->dev);
 
 
 
 
 
 
 
 
 
1685#endif
 
 
 
1686	ASSERT_RTNL();
1687
1688#ifdef CONFIG_IP_MULTICAST
1689	setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1690			(unsigned long)in_dev);
1691	setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1692			(unsigned long)in_dev);
1693	in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1694#endif
 
1695
1696	spin_lock_init(&in_dev->mc_tomb_lock);
1697}
1698
1699/* Device going up */
1700
1701void ip_mc_up(struct in_device *in_dev)
1702{
1703	struct ip_mc_list *pmc;
1704#ifdef CONFIG_IP_MULTICAST
1705	struct net *net = dev_net(in_dev->dev);
1706#endif
1707
1708	ASSERT_RTNL();
1709
1710#ifdef CONFIG_IP_MULTICAST
1711	in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1712#endif
1713	ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1714
1715	for_each_pmc_rtnl(in_dev, pmc) {
1716#ifdef CONFIG_IP_MULTICAST
1717		igmpv3_del_delrec(in_dev, pmc);
1718#endif
1719		igmp_group_added(pmc);
1720	}
1721}
1722
1723/*
1724 *	Device is about to be destroyed: clean up.
1725 */
1726
1727void ip_mc_destroy_dev(struct in_device *in_dev)
1728{
1729	struct ip_mc_list *i;
1730
1731	ASSERT_RTNL();
1732
1733	/* Deactivate timers */
1734	ip_mc_down(in_dev);
1735#ifdef CONFIG_IP_MULTICAST
1736	igmpv3_clear_delrec(in_dev);
1737#endif
1738
1739	while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1740		in_dev->mc_list = i->next_rcu;
1741		in_dev->mc_count--;
 
1742		ip_ma_put(i);
1743	}
1744}
1745
1746/* RTNL is locked */
1747static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1748{
1749	struct net_device *dev = NULL;
1750	struct in_device *idev = NULL;
1751
1752	if (imr->imr_ifindex) {
1753		idev = inetdev_by_index(net, imr->imr_ifindex);
1754		return idev;
1755	}
1756	if (imr->imr_address.s_addr) {
1757		dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1758		if (!dev)
1759			return NULL;
1760	}
1761
1762	if (!dev) {
1763		struct rtable *rt = ip_route_output(net,
1764						    imr->imr_multiaddr.s_addr,
1765						    0, 0, 0);
1766		if (!IS_ERR(rt)) {
1767			dev = rt->dst.dev;
1768			ip_rt_put(rt);
1769		}
1770	}
1771	if (dev) {
1772		imr->imr_ifindex = dev->ifindex;
1773		idev = __in_dev_get_rtnl(dev);
1774	}
1775	return idev;
1776}
1777
1778/*
1779 *	Join a socket to a group
1780 */
1781
1782static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1783	__be32 *psfsrc)
1784{
1785	struct ip_sf_list *psf, *psf_prev;
1786	int rv = 0;
1787
1788	psf_prev = NULL;
1789	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1790		if (psf->sf_inaddr == *psfsrc)
1791			break;
1792		psf_prev = psf;
1793	}
1794	if (!psf || psf->sf_count[sfmode] == 0) {
1795		/* source filter not found, or count wrong =>  bug */
1796		return -ESRCH;
1797	}
1798	psf->sf_count[sfmode]--;
1799	if (psf->sf_count[sfmode] == 0) {
1800		ip_rt_multicast_event(pmc->interface);
1801	}
1802	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1803#ifdef CONFIG_IP_MULTICAST
1804		struct in_device *in_dev = pmc->interface;
1805		struct net *net = dev_net(in_dev->dev);
1806#endif
1807
1808		/* no more filters for this source */
1809		if (psf_prev)
1810			psf_prev->sf_next = psf->sf_next;
1811		else
1812			pmc->sources = psf->sf_next;
1813#ifdef CONFIG_IP_MULTICAST
1814		if (psf->sf_oldin &&
1815		    !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1816			psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1817			psf->sf_next = pmc->tomb;
1818			pmc->tomb = psf;
1819			rv = 1;
1820		} else
1821#endif
1822			kfree(psf);
1823	}
1824	return rv;
1825}
1826
1827#ifndef CONFIG_IP_MULTICAST
1828#define igmp_ifc_event(x)	do { } while (0)
1829#endif
1830
1831static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1832			 int sfcount, __be32 *psfsrc, int delta)
1833{
1834	struct ip_mc_list *pmc;
1835	int	changerec = 0;
1836	int	i, err;
1837
1838	if (!in_dev)
1839		return -ENODEV;
1840	rcu_read_lock();
1841	for_each_pmc_rcu(in_dev, pmc) {
1842		if (*pmca == pmc->multiaddr)
1843			break;
1844	}
1845	if (!pmc) {
1846		/* MCA not found?? bug */
1847		rcu_read_unlock();
1848		return -ESRCH;
1849	}
1850	spin_lock_bh(&pmc->lock);
1851	rcu_read_unlock();
1852#ifdef CONFIG_IP_MULTICAST
1853	sf_markstate(pmc);
1854#endif
1855	if (!delta) {
1856		err = -EINVAL;
1857		if (!pmc->sfcount[sfmode])
1858			goto out_unlock;
1859		pmc->sfcount[sfmode]--;
1860	}
1861	err = 0;
1862	for (i = 0; i < sfcount; i++) {
1863		int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1864
1865		changerec |= rv > 0;
1866		if (!err && rv < 0)
1867			err = rv;
1868	}
1869	if (pmc->sfmode == MCAST_EXCLUDE &&
1870	    pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1871	    pmc->sfcount[MCAST_INCLUDE]) {
1872#ifdef CONFIG_IP_MULTICAST
1873		struct ip_sf_list *psf;
1874		struct net *net = dev_net(in_dev->dev);
1875#endif
1876
1877		/* filter mode change */
1878		pmc->sfmode = MCAST_INCLUDE;
1879#ifdef CONFIG_IP_MULTICAST
1880		pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1881		in_dev->mr_ifc_count = pmc->crcount;
1882		for (psf = pmc->sources; psf; psf = psf->sf_next)
1883			psf->sf_crcount = 0;
1884		igmp_ifc_event(pmc->interface);
1885	} else if (sf_setstate(pmc) || changerec) {
1886		igmp_ifc_event(pmc->interface);
1887#endif
1888	}
1889out_unlock:
1890	spin_unlock_bh(&pmc->lock);
1891	return err;
1892}
1893
1894/*
1895 * Add multicast single-source filter to the interface list
1896 */
1897static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1898	__be32 *psfsrc)
1899{
1900	struct ip_sf_list *psf, *psf_prev;
1901
1902	psf_prev = NULL;
1903	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1904		if (psf->sf_inaddr == *psfsrc)
1905			break;
1906		psf_prev = psf;
1907	}
1908	if (!psf) {
1909		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1910		if (!psf)
1911			return -ENOBUFS;
1912		psf->sf_inaddr = *psfsrc;
1913		if (psf_prev) {
1914			psf_prev->sf_next = psf;
1915		} else
1916			pmc->sources = psf;
1917	}
1918	psf->sf_count[sfmode]++;
1919	if (psf->sf_count[sfmode] == 1) {
1920		ip_rt_multicast_event(pmc->interface);
1921	}
1922	return 0;
1923}
1924
1925#ifdef CONFIG_IP_MULTICAST
1926static void sf_markstate(struct ip_mc_list *pmc)
1927{
1928	struct ip_sf_list *psf;
1929	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1930
1931	for (psf = pmc->sources; psf; psf = psf->sf_next)
1932		if (pmc->sfcount[MCAST_EXCLUDE]) {
1933			psf->sf_oldin = mca_xcount ==
1934				psf->sf_count[MCAST_EXCLUDE] &&
1935				!psf->sf_count[MCAST_INCLUDE];
1936		} else
1937			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1938}
1939
1940static int sf_setstate(struct ip_mc_list *pmc)
1941{
1942	struct ip_sf_list *psf, *dpsf;
1943	int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1944	int qrv = pmc->interface->mr_qrv;
1945	int new_in, rv;
1946
1947	rv = 0;
1948	for (psf = pmc->sources; psf; psf = psf->sf_next) {
1949		if (pmc->sfcount[MCAST_EXCLUDE]) {
1950			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1951				!psf->sf_count[MCAST_INCLUDE];
1952		} else
1953			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1954		if (new_in) {
1955			if (!psf->sf_oldin) {
1956				struct ip_sf_list *prev = NULL;
1957
1958				for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1959					if (dpsf->sf_inaddr == psf->sf_inaddr)
1960						break;
1961					prev = dpsf;
1962				}
1963				if (dpsf) {
1964					if (prev)
1965						prev->sf_next = dpsf->sf_next;
1966					else
1967						pmc->tomb = dpsf->sf_next;
1968					kfree(dpsf);
1969				}
1970				psf->sf_crcount = qrv;
1971				rv++;
1972			}
1973		} else if (psf->sf_oldin) {
1974
1975			psf->sf_crcount = 0;
1976			/*
1977			 * add or update "delete" records if an active filter
1978			 * is now inactive
1979			 */
1980			for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1981				if (dpsf->sf_inaddr == psf->sf_inaddr)
1982					break;
1983			if (!dpsf) {
1984				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1985				if (!dpsf)
1986					continue;
1987				*dpsf = *psf;
1988				/* pmc->lock held by callers */
1989				dpsf->sf_next = pmc->tomb;
1990				pmc->tomb = dpsf;
1991			}
1992			dpsf->sf_crcount = qrv;
1993			rv++;
1994		}
1995	}
1996	return rv;
1997}
1998#endif
1999
2000/*
2001 * Add multicast source filter list to the interface list
2002 */
2003static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
2004			 int sfcount, __be32 *psfsrc, int delta)
2005{
2006	struct ip_mc_list *pmc;
2007	int	isexclude;
2008	int	i, err;
2009
2010	if (!in_dev)
2011		return -ENODEV;
2012	rcu_read_lock();
2013	for_each_pmc_rcu(in_dev, pmc) {
2014		if (*pmca == pmc->multiaddr)
2015			break;
2016	}
2017	if (!pmc) {
2018		/* MCA not found?? bug */
2019		rcu_read_unlock();
2020		return -ESRCH;
2021	}
2022	spin_lock_bh(&pmc->lock);
2023	rcu_read_unlock();
2024
2025#ifdef CONFIG_IP_MULTICAST
2026	sf_markstate(pmc);
2027#endif
2028	isexclude = pmc->sfmode == MCAST_EXCLUDE;
2029	if (!delta)
2030		pmc->sfcount[sfmode]++;
2031	err = 0;
2032	for (i = 0; i < sfcount; i++) {
2033		err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2034		if (err)
2035			break;
2036	}
2037	if (err) {
2038		int j;
2039
2040		if (!delta)
2041			pmc->sfcount[sfmode]--;
2042		for (j = 0; j < i; j++)
2043			(void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2044	} else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2045#ifdef CONFIG_IP_MULTICAST
2046		struct ip_sf_list *psf;
2047		struct net *net = dev_net(pmc->interface->dev);
2048		in_dev = pmc->interface;
2049#endif
2050
2051		/* filter mode change */
2052		if (pmc->sfcount[MCAST_EXCLUDE])
2053			pmc->sfmode = MCAST_EXCLUDE;
2054		else if (pmc->sfcount[MCAST_INCLUDE])
2055			pmc->sfmode = MCAST_INCLUDE;
2056#ifdef CONFIG_IP_MULTICAST
2057		/* else no filters; keep old mode for reports */
2058
2059		pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2060		in_dev->mr_ifc_count = pmc->crcount;
2061		for (psf = pmc->sources; psf; psf = psf->sf_next)
2062			psf->sf_crcount = 0;
2063		igmp_ifc_event(in_dev);
2064	} else if (sf_setstate(pmc)) {
2065		igmp_ifc_event(in_dev);
2066#endif
2067	}
2068	spin_unlock_bh(&pmc->lock);
2069	return err;
2070}
2071
2072static void ip_mc_clear_src(struct ip_mc_list *pmc)
2073{
2074	struct ip_sf_list *psf, *nextpsf;
2075
2076	for (psf = pmc->tomb; psf; psf = nextpsf) {
2077		nextpsf = psf->sf_next;
2078		kfree(psf);
2079	}
2080	pmc->tomb = NULL;
2081	for (psf = pmc->sources; psf; psf = nextpsf) {
2082		nextpsf = psf->sf_next;
2083		kfree(psf);
2084	}
2085	pmc->sources = NULL;
2086	pmc->sfmode = MCAST_EXCLUDE;
2087	pmc->sfcount[MCAST_INCLUDE] = 0;
2088	pmc->sfcount[MCAST_EXCLUDE] = 1;
 
 
 
 
2089}
2090
2091/* Join a multicast group
2092 */
2093
2094int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2095{
2096	__be32 addr = imr->imr_multiaddr.s_addr;
2097	struct ip_mc_socklist *iml, *i;
2098	struct in_device *in_dev;
2099	struct inet_sock *inet = inet_sk(sk);
2100	struct net *net = sock_net(sk);
2101	int ifindex;
2102	int count = 0;
2103	int err;
2104
2105	ASSERT_RTNL();
2106
2107	if (!ipv4_is_multicast(addr))
2108		return -EINVAL;
2109
2110	in_dev = ip_mc_find_dev(net, imr);
2111
2112	if (!in_dev) {
2113		err = -ENODEV;
2114		goto done;
2115	}
2116
2117	err = -EADDRINUSE;
2118	ifindex = imr->imr_ifindex;
2119	for_each_pmc_rtnl(inet, i) {
2120		if (i->multi.imr_multiaddr.s_addr == addr &&
2121		    i->multi.imr_ifindex == ifindex)
2122			goto done;
2123		count++;
2124	}
2125	err = -ENOBUFS;
2126	if (count >= net->ipv4.sysctl_igmp_max_memberships)
2127		goto done;
2128	iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2129	if (!iml)
2130		goto done;
2131
2132	memcpy(&iml->multi, imr, sizeof(*imr));
2133	iml->next_rcu = inet->mc_list;
2134	iml->sflist = NULL;
2135	iml->sfmode = MCAST_EXCLUDE;
2136	rcu_assign_pointer(inet->mc_list, iml);
2137	ip_mc_inc_group(in_dev, addr);
2138	err = 0;
2139done:
2140	return err;
2141}
 
 
 
 
 
 
 
2142EXPORT_SYMBOL(ip_mc_join_group);
2143
 
 
 
 
 
 
 
 
2144static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2145			   struct in_device *in_dev)
2146{
2147	struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2148	int err;
2149
2150	if (!psf) {
2151		/* any-source empty exclude case */
2152		return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2153			iml->sfmode, 0, NULL, 0);
2154	}
2155	err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2156			iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2157	RCU_INIT_POINTER(iml->sflist, NULL);
2158	/* decrease mem now to avoid the memleak warning */
2159	atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2160	kfree_rcu(psf, rcu);
2161	return err;
2162}
2163
2164int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2165{
2166	struct inet_sock *inet = inet_sk(sk);
2167	struct ip_mc_socklist *iml;
2168	struct ip_mc_socklist __rcu **imlp;
2169	struct in_device *in_dev;
2170	struct net *net = sock_net(sk);
2171	__be32 group = imr->imr_multiaddr.s_addr;
2172	u32 ifindex;
2173	int ret = -EADDRNOTAVAIL;
2174
2175	ASSERT_RTNL();
2176
2177	in_dev = ip_mc_find_dev(net, imr);
2178	if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2179		ret = -ENODEV;
2180		goto out;
2181	}
2182	ifindex = imr->imr_ifindex;
2183	for (imlp = &inet->mc_list;
2184	     (iml = rtnl_dereference(*imlp)) != NULL;
2185	     imlp = &iml->next_rcu) {
2186		if (iml->multi.imr_multiaddr.s_addr != group)
2187			continue;
2188		if (ifindex) {
2189			if (iml->multi.imr_ifindex != ifindex)
2190				continue;
2191		} else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2192				iml->multi.imr_address.s_addr)
2193			continue;
2194
2195		(void) ip_mc_leave_src(sk, iml, in_dev);
2196
2197		*imlp = iml->next_rcu;
2198
2199		if (in_dev)
2200			ip_mc_dec_group(in_dev, group);
2201
2202		/* decrease mem now to avoid the memleak warning */
2203		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2204		kfree_rcu(iml, rcu);
2205		return 0;
2206	}
2207out:
2208	return ret;
2209}
2210EXPORT_SYMBOL(ip_mc_leave_group);
2211
2212int ip_mc_source(int add, int omode, struct sock *sk, struct
2213	ip_mreq_source *mreqs, int ifindex)
2214{
2215	int err;
2216	struct ip_mreqn imr;
2217	__be32 addr = mreqs->imr_multiaddr;
2218	struct ip_mc_socklist *pmc;
2219	struct in_device *in_dev = NULL;
2220	struct inet_sock *inet = inet_sk(sk);
2221	struct ip_sf_socklist *psl;
2222	struct net *net = sock_net(sk);
2223	int leavegroup = 0;
2224	int i, j, rv;
2225
2226	if (!ipv4_is_multicast(addr))
2227		return -EINVAL;
2228
2229	ASSERT_RTNL();
2230
2231	imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2232	imr.imr_address.s_addr = mreqs->imr_interface;
2233	imr.imr_ifindex = ifindex;
2234	in_dev = ip_mc_find_dev(net, &imr);
2235
2236	if (!in_dev) {
2237		err = -ENODEV;
2238		goto done;
2239	}
2240	err = -EADDRNOTAVAIL;
2241
2242	for_each_pmc_rtnl(inet, pmc) {
2243		if ((pmc->multi.imr_multiaddr.s_addr ==
2244		     imr.imr_multiaddr.s_addr) &&
2245		    (pmc->multi.imr_ifindex == imr.imr_ifindex))
2246			break;
2247	}
2248	if (!pmc) {		/* must have a prior join */
2249		err = -EINVAL;
2250		goto done;
2251	}
2252	/* if a source filter was set, must be the same mode as before */
2253	if (pmc->sflist) {
2254		if (pmc->sfmode != omode) {
2255			err = -EINVAL;
2256			goto done;
2257		}
2258	} else if (pmc->sfmode != omode) {
2259		/* allow mode switches for empty-set filters */
2260		ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2261		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2262			NULL, 0);
2263		pmc->sfmode = omode;
2264	}
2265
2266	psl = rtnl_dereference(pmc->sflist);
2267	if (!add) {
2268		if (!psl)
2269			goto done;	/* err = -EADDRNOTAVAIL */
2270		rv = !0;
2271		for (i = 0; i < psl->sl_count; i++) {
2272			rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2273				sizeof(__be32));
2274			if (rv == 0)
2275				break;
2276		}
2277		if (rv)		/* source not found */
2278			goto done;	/* err = -EADDRNOTAVAIL */
2279
2280		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2281		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2282			leavegroup = 1;
2283			goto done;
2284		}
2285
2286		/* update the interface filter */
2287		ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2288			&mreqs->imr_sourceaddr, 1);
2289
2290		for (j = i+1; j < psl->sl_count; j++)
2291			psl->sl_addr[j-1] = psl->sl_addr[j];
2292		psl->sl_count--;
2293		err = 0;
2294		goto done;
2295	}
2296	/* else, add a new source to the filter */
2297
2298	if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2299		err = -ENOBUFS;
2300		goto done;
2301	}
2302	if (!psl || psl->sl_count == psl->sl_max) {
2303		struct ip_sf_socklist *newpsl;
2304		int count = IP_SFBLOCK;
2305
2306		if (psl)
2307			count += psl->sl_max;
2308		newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2309		if (!newpsl) {
2310			err = -ENOBUFS;
2311			goto done;
2312		}
2313		newpsl->sl_max = count;
2314		newpsl->sl_count = count - IP_SFBLOCK;
2315		if (psl) {
2316			for (i = 0; i < psl->sl_count; i++)
2317				newpsl->sl_addr[i] = psl->sl_addr[i];
2318			/* decrease mem now to avoid the memleak warning */
2319			atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2320			kfree_rcu(psl, rcu);
2321		}
2322		rcu_assign_pointer(pmc->sflist, newpsl);
2323		psl = newpsl;
2324	}
2325	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
2326	for (i = 0; i < psl->sl_count; i++) {
2327		rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2328			sizeof(__be32));
2329		if (rv == 0)
2330			break;
2331	}
2332	if (rv == 0)		/* address already there is an error */
2333		goto done;
2334	for (j = psl->sl_count-1; j >= i; j--)
2335		psl->sl_addr[j+1] = psl->sl_addr[j];
2336	psl->sl_addr[i] = mreqs->imr_sourceaddr;
2337	psl->sl_count++;
2338	err = 0;
2339	/* update the interface list */
2340	ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2341		&mreqs->imr_sourceaddr, 1);
2342done:
2343	if (leavegroup)
2344		err = ip_mc_leave_group(sk, &imr);
2345	return err;
2346}
2347
2348int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2349{
2350	int err = 0;
2351	struct ip_mreqn	imr;
2352	__be32 addr = msf->imsf_multiaddr;
2353	struct ip_mc_socklist *pmc;
2354	struct in_device *in_dev;
2355	struct inet_sock *inet = inet_sk(sk);
2356	struct ip_sf_socklist *newpsl, *psl;
2357	struct net *net = sock_net(sk);
2358	int leavegroup = 0;
2359
2360	if (!ipv4_is_multicast(addr))
2361		return -EINVAL;
2362	if (msf->imsf_fmode != MCAST_INCLUDE &&
2363	    msf->imsf_fmode != MCAST_EXCLUDE)
2364		return -EINVAL;
2365
2366	ASSERT_RTNL();
2367
2368	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2369	imr.imr_address.s_addr = msf->imsf_interface;
2370	imr.imr_ifindex = ifindex;
2371	in_dev = ip_mc_find_dev(net, &imr);
2372
2373	if (!in_dev) {
2374		err = -ENODEV;
2375		goto done;
2376	}
2377
2378	/* special case - (INCLUDE, empty) == LEAVE_GROUP */
2379	if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2380		leavegroup = 1;
2381		goto done;
2382	}
2383
2384	for_each_pmc_rtnl(inet, pmc) {
2385		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2386		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2387			break;
2388	}
2389	if (!pmc) {		/* must have a prior join */
2390		err = -EINVAL;
2391		goto done;
2392	}
2393	if (msf->imsf_numsrc) {
2394		newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2395							   GFP_KERNEL);
2396		if (!newpsl) {
2397			err = -ENOBUFS;
2398			goto done;
2399		}
2400		newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2401		memcpy(newpsl->sl_addr, msf->imsf_slist,
2402			msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2403		err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2404			msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2405		if (err) {
2406			sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2407			goto done;
2408		}
2409	} else {
2410		newpsl = NULL;
2411		(void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2412				     msf->imsf_fmode, 0, NULL, 0);
2413	}
2414	psl = rtnl_dereference(pmc->sflist);
2415	if (psl) {
2416		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2417			psl->sl_count, psl->sl_addr, 0);
2418		/* decrease mem now to avoid the memleak warning */
2419		atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2420		kfree_rcu(psl, rcu);
2421	} else
2422		(void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2423			0, NULL, 0);
2424	rcu_assign_pointer(pmc->sflist, newpsl);
2425	pmc->sfmode = msf->imsf_fmode;
2426	err = 0;
2427done:
2428	if (leavegroup)
2429		err = ip_mc_leave_group(sk, &imr);
2430	return err;
2431}
2432
2433int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2434	struct ip_msfilter __user *optval, int __user *optlen)
2435{
2436	int err, len, count, copycount;
2437	struct ip_mreqn	imr;
2438	__be32 addr = msf->imsf_multiaddr;
2439	struct ip_mc_socklist *pmc;
2440	struct in_device *in_dev;
2441	struct inet_sock *inet = inet_sk(sk);
2442	struct ip_sf_socklist *psl;
2443	struct net *net = sock_net(sk);
2444
2445	ASSERT_RTNL();
2446
2447	if (!ipv4_is_multicast(addr))
2448		return -EINVAL;
2449
2450	imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2451	imr.imr_address.s_addr = msf->imsf_interface;
2452	imr.imr_ifindex = 0;
2453	in_dev = ip_mc_find_dev(net, &imr);
2454
2455	if (!in_dev) {
2456		err = -ENODEV;
2457		goto done;
2458	}
2459	err = -EADDRNOTAVAIL;
2460
2461	for_each_pmc_rtnl(inet, pmc) {
2462		if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2463		    pmc->multi.imr_ifindex == imr.imr_ifindex)
2464			break;
2465	}
2466	if (!pmc)		/* must have a prior join */
2467		goto done;
2468	msf->imsf_fmode = pmc->sfmode;
2469	psl = rtnl_dereference(pmc->sflist);
2470	if (!psl) {
2471		len = 0;
2472		count = 0;
2473	} else {
2474		count = psl->sl_count;
2475	}
2476	copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2477	len = copycount * sizeof(psl->sl_addr[0]);
2478	msf->imsf_numsrc = count;
2479	if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2480	    copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2481		return -EFAULT;
2482	}
2483	if (len &&
2484	    copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2485		return -EFAULT;
2486	return 0;
2487done:
2488	return err;
2489}
2490
2491int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2492	struct group_filter __user *optval, int __user *optlen)
2493{
2494	int err, i, count, copycount;
2495	struct sockaddr_in *psin;
2496	__be32 addr;
2497	struct ip_mc_socklist *pmc;
2498	struct inet_sock *inet = inet_sk(sk);
2499	struct ip_sf_socklist *psl;
2500
2501	ASSERT_RTNL();
2502
2503	psin = (struct sockaddr_in *)&gsf->gf_group;
2504	if (psin->sin_family != AF_INET)
2505		return -EINVAL;
2506	addr = psin->sin_addr.s_addr;
2507	if (!ipv4_is_multicast(addr))
2508		return -EINVAL;
2509
2510	err = -EADDRNOTAVAIL;
2511
2512	for_each_pmc_rtnl(inet, pmc) {
2513		if (pmc->multi.imr_multiaddr.s_addr == addr &&
2514		    pmc->multi.imr_ifindex == gsf->gf_interface)
2515			break;
2516	}
2517	if (!pmc)		/* must have a prior join */
2518		goto done;
2519	gsf->gf_fmode = pmc->sfmode;
2520	psl = rtnl_dereference(pmc->sflist);
2521	count = psl ? psl->sl_count : 0;
2522	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2523	gsf->gf_numsrc = count;
2524	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2525	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2526		return -EFAULT;
2527	}
2528	for (i = 0; i < copycount; i++) {
2529		struct sockaddr_storage ss;
2530
2531		psin = (struct sockaddr_in *)&ss;
2532		memset(&ss, 0, sizeof(ss));
2533		psin->sin_family = AF_INET;
2534		psin->sin_addr.s_addr = psl->sl_addr[i];
2535		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2536			return -EFAULT;
2537	}
2538	return 0;
2539done:
2540	return err;
2541}
2542
2543/*
2544 * check if a multicast source filter allows delivery for a given <src,dst,intf>
2545 */
2546int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
 
2547{
2548	struct inet_sock *inet = inet_sk(sk);
2549	struct ip_mc_socklist *pmc;
2550	struct ip_sf_socklist *psl;
2551	int i;
2552	int ret;
2553
2554	ret = 1;
2555	if (!ipv4_is_multicast(loc_addr))
2556		goto out;
2557
2558	rcu_read_lock();
2559	for_each_pmc_rcu(inet, pmc) {
2560		if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2561		    pmc->multi.imr_ifindex == dif)
 
2562			break;
2563	}
2564	ret = inet->mc_all;
2565	if (!pmc)
2566		goto unlock;
2567	psl = rcu_dereference(pmc->sflist);
2568	ret = (pmc->sfmode == MCAST_EXCLUDE);
2569	if (!psl)
2570		goto unlock;
2571
2572	for (i = 0; i < psl->sl_count; i++) {
2573		if (psl->sl_addr[i] == rmt_addr)
2574			break;
2575	}
2576	ret = 0;
2577	if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2578		goto unlock;
2579	if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2580		goto unlock;
2581	ret = 1;
2582unlock:
2583	rcu_read_unlock();
2584out:
2585	return ret;
2586}
2587
2588/*
2589 *	A socket is closing.
2590 */
2591
2592void ip_mc_drop_socket(struct sock *sk)
2593{
2594	struct inet_sock *inet = inet_sk(sk);
2595	struct ip_mc_socklist *iml;
2596	struct net *net = sock_net(sk);
2597
2598	if (!inet->mc_list)
2599		return;
2600
2601	rtnl_lock();
2602	while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2603		struct in_device *in_dev;
2604
2605		inet->mc_list = iml->next_rcu;
2606		in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2607		(void) ip_mc_leave_src(sk, iml, in_dev);
2608		if (in_dev)
2609			ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2610		/* decrease mem now to avoid the memleak warning */
2611		atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2612		kfree_rcu(iml, rcu);
2613	}
2614	rtnl_unlock();
2615}
2616
2617/* called with rcu_read_lock() */
2618int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2619{
2620	struct ip_mc_list *im;
2621	struct ip_mc_list __rcu **mc_hash;
2622	struct ip_sf_list *psf;
2623	int rv = 0;
2624
2625	mc_hash = rcu_dereference(in_dev->mc_hash);
2626	if (mc_hash) {
2627		u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2628
2629		for (im = rcu_dereference(mc_hash[hash]);
2630		     im != NULL;
2631		     im = rcu_dereference(im->next_hash)) {
2632			if (im->multiaddr == mc_addr)
2633				break;
2634		}
2635	} else {
2636		for_each_pmc_rcu(in_dev, im) {
2637			if (im->multiaddr == mc_addr)
2638				break;
2639		}
2640	}
2641	if (im && proto == IPPROTO_IGMP) {
2642		rv = 1;
2643	} else if (im) {
2644		if (src_addr) {
 
2645			for (psf = im->sources; psf; psf = psf->sf_next) {
2646				if (psf->sf_inaddr == src_addr)
2647					break;
2648			}
2649			if (psf)
2650				rv = psf->sf_count[MCAST_INCLUDE] ||
2651					psf->sf_count[MCAST_EXCLUDE] !=
2652					im->sfcount[MCAST_EXCLUDE];
2653			else
2654				rv = im->sfcount[MCAST_EXCLUDE] != 0;
 
2655		} else
2656			rv = 1; /* unspecified source; tentatively allow */
2657	}
2658	return rv;
2659}
2660
2661#if defined(CONFIG_PROC_FS)
2662struct igmp_mc_iter_state {
2663	struct seq_net_private p;
2664	struct net_device *dev;
2665	struct in_device *in_dev;
2666};
2667
2668#define	igmp_mc_seq_private(seq)	((struct igmp_mc_iter_state *)(seq)->private)
2669
2670static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2671{
2672	struct net *net = seq_file_net(seq);
2673	struct ip_mc_list *im = NULL;
2674	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2675
2676	state->in_dev = NULL;
2677	for_each_netdev_rcu(net, state->dev) {
2678		struct in_device *in_dev;
2679
2680		in_dev = __in_dev_get_rcu(state->dev);
2681		if (!in_dev)
2682			continue;
2683		im = rcu_dereference(in_dev->mc_list);
2684		if (im) {
2685			state->in_dev = in_dev;
2686			break;
2687		}
2688	}
2689	return im;
2690}
2691
2692static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2693{
2694	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2695
2696	im = rcu_dereference(im->next_rcu);
2697	while (!im) {
2698		state->dev = next_net_device_rcu(state->dev);
2699		if (!state->dev) {
2700			state->in_dev = NULL;
2701			break;
2702		}
2703		state->in_dev = __in_dev_get_rcu(state->dev);
2704		if (!state->in_dev)
2705			continue;
2706		im = rcu_dereference(state->in_dev->mc_list);
2707	}
2708	return im;
2709}
2710
2711static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2712{
2713	struct ip_mc_list *im = igmp_mc_get_first(seq);
2714	if (im)
2715		while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2716			--pos;
2717	return pos ? NULL : im;
2718}
2719
2720static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2721	__acquires(rcu)
2722{
2723	rcu_read_lock();
2724	return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2725}
2726
2727static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2728{
2729	struct ip_mc_list *im;
2730	if (v == SEQ_START_TOKEN)
2731		im = igmp_mc_get_first(seq);
2732	else
2733		im = igmp_mc_get_next(seq, v);
2734	++*pos;
2735	return im;
2736}
2737
2738static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2739	__releases(rcu)
2740{
2741	struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2742
2743	state->in_dev = NULL;
2744	state->dev = NULL;
2745	rcu_read_unlock();
2746}
2747
2748static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2749{
2750	if (v == SEQ_START_TOKEN)
2751		seq_puts(seq,
2752			 "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2753	else {
2754		struct ip_mc_list *im = (struct ip_mc_list *)v;
2755		struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2756		char   *querier;
2757		long delta;
2758
2759#ifdef CONFIG_IP_MULTICAST
2760		querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2761			  IGMP_V2_SEEN(state->in_dev) ? "V2" :
2762			  "V3";
2763#else
2764		querier = "NONE";
2765#endif
2766
2767		if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2768			seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2769				   state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2770		}
2771
2772		delta = im->timer.expires - jiffies;
2773		seq_printf(seq,
2774			   "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2775			   im->multiaddr, im->users,
2776			   im->tm_running,
2777			   im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2778			   im->reporter);
2779	}
2780	return 0;
2781}
2782
2783static const struct seq_operations igmp_mc_seq_ops = {
2784	.start	=	igmp_mc_seq_start,
2785	.next	=	igmp_mc_seq_next,
2786	.stop	=	igmp_mc_seq_stop,
2787	.show	=	igmp_mc_seq_show,
2788};
2789
2790static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2791{
2792	return seq_open_net(inode, file, &igmp_mc_seq_ops,
2793			sizeof(struct igmp_mc_iter_state));
2794}
2795
2796static const struct file_operations igmp_mc_seq_fops = {
2797	.owner		=	THIS_MODULE,
2798	.open		=	igmp_mc_seq_open,
2799	.read		=	seq_read,
2800	.llseek		=	seq_lseek,
2801	.release	=	seq_release_net,
2802};
2803
2804struct igmp_mcf_iter_state {
2805	struct seq_net_private p;
2806	struct net_device *dev;
2807	struct in_device *idev;
2808	struct ip_mc_list *im;
2809};
2810
2811#define igmp_mcf_seq_private(seq)	((struct igmp_mcf_iter_state *)(seq)->private)
2812
2813static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2814{
2815	struct net *net = seq_file_net(seq);
2816	struct ip_sf_list *psf = NULL;
2817	struct ip_mc_list *im = NULL;
2818	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2819
2820	state->idev = NULL;
2821	state->im = NULL;
2822	for_each_netdev_rcu(net, state->dev) {
2823		struct in_device *idev;
2824		idev = __in_dev_get_rcu(state->dev);
2825		if (unlikely(!idev))
2826			continue;
2827		im = rcu_dereference(idev->mc_list);
2828		if (likely(im)) {
2829			spin_lock_bh(&im->lock);
2830			psf = im->sources;
2831			if (likely(psf)) {
2832				state->im = im;
2833				state->idev = idev;
2834				break;
2835			}
2836			spin_unlock_bh(&im->lock);
2837		}
2838	}
2839	return psf;
2840}
2841
2842static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2843{
2844	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2845
2846	psf = psf->sf_next;
2847	while (!psf) {
2848		spin_unlock_bh(&state->im->lock);
2849		state->im = state->im->next;
2850		while (!state->im) {
2851			state->dev = next_net_device_rcu(state->dev);
2852			if (!state->dev) {
2853				state->idev = NULL;
2854				goto out;
2855			}
2856			state->idev = __in_dev_get_rcu(state->dev);
2857			if (!state->idev)
2858				continue;
2859			state->im = rcu_dereference(state->idev->mc_list);
2860		}
2861		if (!state->im)
2862			break;
2863		spin_lock_bh(&state->im->lock);
2864		psf = state->im->sources;
2865	}
2866out:
2867	return psf;
2868}
2869
2870static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2871{
2872	struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2873	if (psf)
2874		while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2875			--pos;
2876	return pos ? NULL : psf;
2877}
2878
2879static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2880	__acquires(rcu)
2881{
2882	rcu_read_lock();
2883	return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2884}
2885
2886static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2887{
2888	struct ip_sf_list *psf;
2889	if (v == SEQ_START_TOKEN)
2890		psf = igmp_mcf_get_first(seq);
2891	else
2892		psf = igmp_mcf_get_next(seq, v);
2893	++*pos;
2894	return psf;
2895}
2896
2897static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2898	__releases(rcu)
2899{
2900	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2901	if (likely(state->im)) {
2902		spin_unlock_bh(&state->im->lock);
2903		state->im = NULL;
2904	}
2905	state->idev = NULL;
2906	state->dev = NULL;
2907	rcu_read_unlock();
2908}
2909
2910static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2911{
2912	struct ip_sf_list *psf = (struct ip_sf_list *)v;
2913	struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2914
2915	if (v == SEQ_START_TOKEN) {
2916		seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2917	} else {
2918		seq_printf(seq,
2919			   "%3d %6.6s 0x%08x "
2920			   "0x%08x %6lu %6lu\n",
2921			   state->dev->ifindex, state->dev->name,
2922			   ntohl(state->im->multiaddr),
2923			   ntohl(psf->sf_inaddr),
2924			   psf->sf_count[MCAST_INCLUDE],
2925			   psf->sf_count[MCAST_EXCLUDE]);
2926	}
2927	return 0;
2928}
2929
2930static const struct seq_operations igmp_mcf_seq_ops = {
2931	.start	=	igmp_mcf_seq_start,
2932	.next	=	igmp_mcf_seq_next,
2933	.stop	=	igmp_mcf_seq_stop,
2934	.show	=	igmp_mcf_seq_show,
2935};
2936
2937static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2938{
2939	return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2940			sizeof(struct igmp_mcf_iter_state));
2941}
2942
2943static const struct file_operations igmp_mcf_seq_fops = {
2944	.owner		=	THIS_MODULE,
2945	.open		=	igmp_mcf_seq_open,
2946	.read		=	seq_read,
2947	.llseek		=	seq_lseek,
2948	.release	=	seq_release_net,
2949};
2950
2951static int __net_init igmp_net_init(struct net *net)
2952{
2953	struct proc_dir_entry *pde;
2954	int err;
2955
2956	pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
 
2957	if (!pde)
2958		goto out_igmp;
2959	pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2960			  &igmp_mcf_seq_fops);
2961	if (!pde)
2962		goto out_mcfilter;
2963	err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2964				   SOCK_DGRAM, 0, net);
2965	if (err < 0) {
2966		pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2967		       err);
2968		goto out_sock;
2969	}
2970
2971	/* Sysctl initialization */
2972	net->ipv4.sysctl_igmp_max_memberships = 20;
2973	net->ipv4.sysctl_igmp_max_msf = 10;
2974	/* IGMP reports for link-local multicast groups are enabled by default */
2975	net->ipv4.sysctl_igmp_llm_reports = 1;
2976	net->ipv4.sysctl_igmp_qrv = 2;
2977	return 0;
2978
2979out_sock:
2980	remove_proc_entry("mcfilter", net->proc_net);
2981out_mcfilter:
2982	remove_proc_entry("igmp", net->proc_net);
2983out_igmp:
2984	return -ENOMEM;
2985}
2986
2987static void __net_exit igmp_net_exit(struct net *net)
2988{
2989	remove_proc_entry("mcfilter", net->proc_net);
2990	remove_proc_entry("igmp", net->proc_net);
2991	inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2992}
2993
2994static struct pernet_operations igmp_net_ops = {
2995	.init = igmp_net_init,
2996	.exit = igmp_net_exit,
2997};
2998#endif
2999
3000static int igmp_netdev_event(struct notifier_block *this,
3001			     unsigned long event, void *ptr)
3002{
3003	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3004	struct in_device *in_dev;
3005
3006	switch (event) {
3007	case NETDEV_RESEND_IGMP:
3008		in_dev = __in_dev_get_rtnl(dev);
3009		if (in_dev)
3010			ip_mc_rejoin_groups(in_dev);
3011		break;
3012	default:
3013		break;
3014	}
3015	return NOTIFY_DONE;
3016}
3017
3018static struct notifier_block igmp_notifier = {
3019	.notifier_call = igmp_netdev_event,
3020};
3021
3022int __init igmp_mc_init(void)
3023{
3024#if defined(CONFIG_PROC_FS)
3025	int err;
3026
3027	err = register_pernet_subsys(&igmp_net_ops);
3028	if (err)
3029		return err;
3030	err = register_netdevice_notifier(&igmp_notifier);
3031	if (err)
3032		goto reg_notif_fail;
3033	return 0;
3034
3035reg_notif_fail:
3036	unregister_pernet_subsys(&igmp_net_ops);
3037	return err;
3038#else
3039	return register_netdevice_notifier(&igmp_notifier);
3040#endif
3041}