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_V2_UNSOLICITED_REPORT_INTERVAL (10*HZ) 111 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL (1*HZ) 112 #define IGMP_QUERY_INTERVAL (125*HZ) 113 #define IGMP_QUERY_RESPONSE_INTERVAL (10*HZ) 114 115 #define IGMP_INITIAL_REPORT_DELAY (1) 116 117 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs! 118 * IGMP specs require to report membership immediately after 119 * joining a group, but we delay the first report by a 120 * small interval. It seems more natural and still does not 121 * contradict to specs provided this delay is small enough. 122 */ 123 124 #define IGMP_V1_SEEN(in_dev) \ 125 (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \ 126 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \ 127 ((in_dev)->mr_v1_seen && \ 128 time_before(jiffies, (in_dev)->mr_v1_seen))) 129 #define IGMP_V2_SEEN(in_dev) \ 130 (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \ 131 IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \ 132 ((in_dev)->mr_v2_seen && \ 133 time_before(jiffies, (in_dev)->mr_v2_seen))) 134 135 static int unsolicited_report_interval(struct in_device *in_dev) 136 { 137 int interval_ms, interval_jiffies; 138 139 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) 140 interval_ms = IN_DEV_CONF_GET( 141 in_dev, 142 IGMPV2_UNSOLICITED_REPORT_INTERVAL); 143 else /* v3 */ 144 interval_ms = IN_DEV_CONF_GET( 145 in_dev, 146 IGMPV3_UNSOLICITED_REPORT_INTERVAL); 147 148 interval_jiffies = msecs_to_jiffies(interval_ms); 149 150 /* _timer functions can't handle a delay of 0 jiffies so ensure 151 * we always return a positive value. 152 */ 153 if (interval_jiffies <= 0) 154 interval_jiffies = 1; 155 return interval_jiffies; 156 } 157 158 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im, 159 gfp_t gfp); 160 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im); 161 static void igmpv3_clear_delrec(struct in_device *in_dev); 162 static int sf_setstate(struct ip_mc_list *pmc); 163 static void sf_markstate(struct ip_mc_list *pmc); 164 #endif 165 static void ip_mc_clear_src(struct ip_mc_list *pmc); 166 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode, 167 int sfcount, __be32 *psfsrc, int delta); 168 169 static void ip_ma_put(struct ip_mc_list *im) 170 { 171 if (refcount_dec_and_test(&im->refcnt)) { 172 in_dev_put(im->interface); 173 kfree_rcu(im, rcu); 174 } 175 } 176 177 #define for_each_pmc_rcu(in_dev, pmc) \ 178 for (pmc = rcu_dereference(in_dev->mc_list); \ 179 pmc != NULL; \ 180 pmc = rcu_dereference(pmc->next_rcu)) 181 182 #define for_each_pmc_rtnl(in_dev, pmc) \ 183 for (pmc = rtnl_dereference(in_dev->mc_list); \ 184 pmc != NULL; \ 185 pmc = rtnl_dereference(pmc->next_rcu)) 186 187 static void ip_sf_list_clear_all(struct ip_sf_list *psf) 188 { 189 struct ip_sf_list *next; 190 191 while (psf) { 192 next = psf->sf_next; 193 kfree(psf); 194 psf = next; 195 } 196 } 197 198 #ifdef CONFIG_IP_MULTICAST 199 200 /* 201 * Timer management 202 */ 203 204 static void igmp_stop_timer(struct ip_mc_list *im) 205 { 206 spin_lock_bh(&im->lock); 207 if (del_timer(&im->timer)) 208 refcount_dec(&im->refcnt); 209 im->tm_running = 0; 210 im->reporter = 0; 211 im->unsolicit_count = 0; 212 spin_unlock_bh(&im->lock); 213 } 214 215 /* It must be called with locked im->lock */ 216 static void igmp_start_timer(struct ip_mc_list *im, int max_delay) 217 { 218 int tv = prandom_u32() % max_delay; 219 220 im->tm_running = 1; 221 if (!mod_timer(&im->timer, jiffies+tv+2)) 222 refcount_inc(&im->refcnt); 223 } 224 225 static void igmp_gq_start_timer(struct in_device *in_dev) 226 { 227 int tv = prandom_u32() % in_dev->mr_maxdelay; 228 unsigned long exp = jiffies + tv + 2; 229 230 if (in_dev->mr_gq_running && 231 time_after_eq(exp, (in_dev->mr_gq_timer).expires)) 232 return; 233 234 in_dev->mr_gq_running = 1; 235 if (!mod_timer(&in_dev->mr_gq_timer, exp)) 236 in_dev_hold(in_dev); 237 } 238 239 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay) 240 { 241 int tv = prandom_u32() % delay; 242 243 if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2)) 244 in_dev_hold(in_dev); 245 } 246 247 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay) 248 { 249 spin_lock_bh(&im->lock); 250 im->unsolicit_count = 0; 251 if (del_timer(&im->timer)) { 252 if ((long)(im->timer.expires-jiffies) < max_delay) { 253 add_timer(&im->timer); 254 im->tm_running = 1; 255 spin_unlock_bh(&im->lock); 256 return; 257 } 258 refcount_dec(&im->refcnt); 259 } 260 igmp_start_timer(im, max_delay); 261 spin_unlock_bh(&im->lock); 262 } 263 264 265 /* 266 * Send an IGMP report. 267 */ 268 269 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4) 270 271 272 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type, 273 int gdeleted, int sdeleted) 274 { 275 switch (type) { 276 case IGMPV3_MODE_IS_INCLUDE: 277 case IGMPV3_MODE_IS_EXCLUDE: 278 if (gdeleted || sdeleted) 279 return 0; 280 if (!(pmc->gsquery && !psf->sf_gsresp)) { 281 if (pmc->sfmode == MCAST_INCLUDE) 282 return 1; 283 /* don't include if this source is excluded 284 * in all filters 285 */ 286 if (psf->sf_count[MCAST_INCLUDE]) 287 return type == IGMPV3_MODE_IS_INCLUDE; 288 return pmc->sfcount[MCAST_EXCLUDE] == 289 psf->sf_count[MCAST_EXCLUDE]; 290 } 291 return 0; 292 case IGMPV3_CHANGE_TO_INCLUDE: 293 if (gdeleted || sdeleted) 294 return 0; 295 return psf->sf_count[MCAST_INCLUDE] != 0; 296 case IGMPV3_CHANGE_TO_EXCLUDE: 297 if (gdeleted || sdeleted) 298 return 0; 299 if (pmc->sfcount[MCAST_EXCLUDE] == 0 || 300 psf->sf_count[MCAST_INCLUDE]) 301 return 0; 302 return pmc->sfcount[MCAST_EXCLUDE] == 303 psf->sf_count[MCAST_EXCLUDE]; 304 case IGMPV3_ALLOW_NEW_SOURCES: 305 if (gdeleted || !psf->sf_crcount) 306 return 0; 307 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted; 308 case IGMPV3_BLOCK_OLD_SOURCES: 309 if (pmc->sfmode == MCAST_INCLUDE) 310 return gdeleted || (psf->sf_crcount && sdeleted); 311 return psf->sf_crcount && !gdeleted && !sdeleted; 312 } 313 return 0; 314 } 315 316 static int 317 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted) 318 { 319 struct ip_sf_list *psf; 320 int scount = 0; 321 322 for (psf = pmc->sources; psf; psf = psf->sf_next) { 323 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) 324 continue; 325 scount++; 326 } 327 return scount; 328 } 329 330 /* source address selection per RFC 3376 section 4.2.13 */ 331 static __be32 igmpv3_get_srcaddr(struct net_device *dev, 332 const struct flowi4 *fl4) 333 { 334 struct in_device *in_dev = __in_dev_get_rcu(dev); 335 const struct in_ifaddr *ifa; 336 337 if (!in_dev) 338 return htonl(INADDR_ANY); 339 340 in_dev_for_each_ifa_rcu(ifa, in_dev) { 341 if (fl4->saddr == ifa->ifa_local) 342 return fl4->saddr; 343 } 344 345 return htonl(INADDR_ANY); 346 } 347 348 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu) 349 { 350 struct sk_buff *skb; 351 struct rtable *rt; 352 struct iphdr *pip; 353 struct igmpv3_report *pig; 354 struct net *net = dev_net(dev); 355 struct flowi4 fl4; 356 int hlen = LL_RESERVED_SPACE(dev); 357 int tlen = dev->needed_tailroom; 358 unsigned int size = mtu; 359 360 while (1) { 361 skb = alloc_skb(size + hlen + tlen, 362 GFP_ATOMIC | __GFP_NOWARN); 363 if (skb) 364 break; 365 size >>= 1; 366 if (size < 256) 367 return NULL; 368 } 369 skb->priority = TC_PRIO_CONTROL; 370 371 rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0, 372 0, 0, 373 IPPROTO_IGMP, 0, dev->ifindex); 374 if (IS_ERR(rt)) { 375 kfree_skb(skb); 376 return NULL; 377 } 378 379 skb_dst_set(skb, &rt->dst); 380 skb->dev = dev; 381 382 skb_reserve(skb, hlen); 383 skb_tailroom_reserve(skb, mtu, tlen); 384 385 skb_reset_network_header(skb); 386 pip = ip_hdr(skb); 387 skb_put(skb, sizeof(struct iphdr) + 4); 388 389 pip->version = 4; 390 pip->ihl = (sizeof(struct iphdr)+4)>>2; 391 pip->tos = 0xc0; 392 pip->frag_off = htons(IP_DF); 393 pip->ttl = 1; 394 pip->daddr = fl4.daddr; 395 396 rcu_read_lock(); 397 pip->saddr = igmpv3_get_srcaddr(dev, &fl4); 398 rcu_read_unlock(); 399 400 pip->protocol = IPPROTO_IGMP; 401 pip->tot_len = 0; /* filled in later */ 402 ip_select_ident(net, skb, NULL); 403 ((u8 *)&pip[1])[0] = IPOPT_RA; 404 ((u8 *)&pip[1])[1] = 4; 405 ((u8 *)&pip[1])[2] = 0; 406 ((u8 *)&pip[1])[3] = 0; 407 408 skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4; 409 skb_put(skb, sizeof(*pig)); 410 pig = igmpv3_report_hdr(skb); 411 pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT; 412 pig->resv1 = 0; 413 pig->csum = 0; 414 pig->resv2 = 0; 415 pig->ngrec = 0; 416 return skb; 417 } 418 419 static int igmpv3_sendpack(struct sk_buff *skb) 420 { 421 struct igmphdr *pig = igmp_hdr(skb); 422 const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb); 423 424 pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen); 425 426 return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb); 427 } 428 429 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel) 430 { 431 return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel); 432 } 433 434 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc, 435 int type, struct igmpv3_grec **ppgr, unsigned int mtu) 436 { 437 struct net_device *dev = pmc->interface->dev; 438 struct igmpv3_report *pih; 439 struct igmpv3_grec *pgr; 440 441 if (!skb) { 442 skb = igmpv3_newpack(dev, mtu); 443 if (!skb) 444 return NULL; 445 } 446 pgr = skb_put(skb, sizeof(struct igmpv3_grec)); 447 pgr->grec_type = type; 448 pgr->grec_auxwords = 0; 449 pgr->grec_nsrcs = 0; 450 pgr->grec_mca = pmc->multiaddr; 451 pih = igmpv3_report_hdr(skb); 452 pih->ngrec = htons(ntohs(pih->ngrec)+1); 453 *ppgr = pgr; 454 return skb; 455 } 456 457 #define AVAILABLE(skb) ((skb) ? skb_availroom(skb) : 0) 458 459 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc, 460 int type, int gdeleted, int sdeleted) 461 { 462 struct net_device *dev = pmc->interface->dev; 463 struct net *net = dev_net(dev); 464 struct igmpv3_report *pih; 465 struct igmpv3_grec *pgr = NULL; 466 struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list; 467 int scount, stotal, first, isquery, truncate; 468 unsigned int mtu; 469 470 if (pmc->multiaddr == IGMP_ALL_HOSTS) 471 return skb; 472 if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports) 473 return skb; 474 475 mtu = READ_ONCE(dev->mtu); 476 if (mtu < IPV4_MIN_MTU) 477 return skb; 478 479 isquery = type == IGMPV3_MODE_IS_INCLUDE || 480 type == IGMPV3_MODE_IS_EXCLUDE; 481 truncate = type == IGMPV3_MODE_IS_EXCLUDE || 482 type == IGMPV3_CHANGE_TO_EXCLUDE; 483 484 stotal = scount = 0; 485 486 psf_list = sdeleted ? &pmc->tomb : &pmc->sources; 487 488 if (!*psf_list) 489 goto empty_source; 490 491 pih = skb ? igmpv3_report_hdr(skb) : NULL; 492 493 /* EX and TO_EX get a fresh packet, if needed */ 494 if (truncate) { 495 if (pih && pih->ngrec && 496 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) { 497 if (skb) 498 igmpv3_sendpack(skb); 499 skb = igmpv3_newpack(dev, mtu); 500 } 501 } 502 first = 1; 503 psf_prev = NULL; 504 for (psf = *psf_list; psf; psf = psf_next) { 505 __be32 *psrc; 506 507 psf_next = psf->sf_next; 508 509 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) { 510 psf_prev = psf; 511 continue; 512 } 513 514 /* Based on RFC3376 5.1. Should not send source-list change 515 * records when there is a filter mode change. 516 */ 517 if (((gdeleted && pmc->sfmode == MCAST_EXCLUDE) || 518 (!gdeleted && pmc->crcount)) && 519 (type == IGMPV3_ALLOW_NEW_SOURCES || 520 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) 521 goto decrease_sf_crcount; 522 523 /* clear marks on query responses */ 524 if (isquery) 525 psf->sf_gsresp = 0; 526 527 if (AVAILABLE(skb) < sizeof(__be32) + 528 first*sizeof(struct igmpv3_grec)) { 529 if (truncate && !first) 530 break; /* truncate these */ 531 if (pgr) 532 pgr->grec_nsrcs = htons(scount); 533 if (skb) 534 igmpv3_sendpack(skb); 535 skb = igmpv3_newpack(dev, mtu); 536 first = 1; 537 scount = 0; 538 } 539 if (first) { 540 skb = add_grhead(skb, pmc, type, &pgr, mtu); 541 first = 0; 542 } 543 if (!skb) 544 return NULL; 545 psrc = skb_put(skb, sizeof(__be32)); 546 *psrc = psf->sf_inaddr; 547 scount++; stotal++; 548 if ((type == IGMPV3_ALLOW_NEW_SOURCES || 549 type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) { 550 decrease_sf_crcount: 551 psf->sf_crcount--; 552 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) { 553 if (psf_prev) 554 psf_prev->sf_next = psf->sf_next; 555 else 556 *psf_list = psf->sf_next; 557 kfree(psf); 558 continue; 559 } 560 } 561 psf_prev = psf; 562 } 563 564 empty_source: 565 if (!stotal) { 566 if (type == IGMPV3_ALLOW_NEW_SOURCES || 567 type == IGMPV3_BLOCK_OLD_SOURCES) 568 return skb; 569 if (pmc->crcount || isquery) { 570 /* make sure we have room for group header */ 571 if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) { 572 igmpv3_sendpack(skb); 573 skb = NULL; /* add_grhead will get a new one */ 574 } 575 skb = add_grhead(skb, pmc, type, &pgr, mtu); 576 } 577 } 578 if (pgr) 579 pgr->grec_nsrcs = htons(scount); 580 581 if (isquery) 582 pmc->gsquery = 0; /* clear query state on report */ 583 return skb; 584 } 585 586 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc) 587 { 588 struct sk_buff *skb = NULL; 589 struct net *net = dev_net(in_dev->dev); 590 int type; 591 592 if (!pmc) { 593 rcu_read_lock(); 594 for_each_pmc_rcu(in_dev, pmc) { 595 if (pmc->multiaddr == IGMP_ALL_HOSTS) 596 continue; 597 if (ipv4_is_local_multicast(pmc->multiaddr) && 598 !net->ipv4.sysctl_igmp_llm_reports) 599 continue; 600 spin_lock_bh(&pmc->lock); 601 if (pmc->sfcount[MCAST_EXCLUDE]) 602 type = IGMPV3_MODE_IS_EXCLUDE; 603 else 604 type = IGMPV3_MODE_IS_INCLUDE; 605 skb = add_grec(skb, pmc, type, 0, 0); 606 spin_unlock_bh(&pmc->lock); 607 } 608 rcu_read_unlock(); 609 } else { 610 spin_lock_bh(&pmc->lock); 611 if (pmc->sfcount[MCAST_EXCLUDE]) 612 type = IGMPV3_MODE_IS_EXCLUDE; 613 else 614 type = IGMPV3_MODE_IS_INCLUDE; 615 skb = add_grec(skb, pmc, type, 0, 0); 616 spin_unlock_bh(&pmc->lock); 617 } 618 if (!skb) 619 return 0; 620 return igmpv3_sendpack(skb); 621 } 622 623 /* 624 * remove zero-count source records from a source filter list 625 */ 626 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf) 627 { 628 struct ip_sf_list *psf_prev, *psf_next, *psf; 629 630 psf_prev = NULL; 631 for (psf = *ppsf; psf; psf = psf_next) { 632 psf_next = psf->sf_next; 633 if (psf->sf_crcount == 0) { 634 if (psf_prev) 635 psf_prev->sf_next = psf->sf_next; 636 else 637 *ppsf = psf->sf_next; 638 kfree(psf); 639 } else 640 psf_prev = psf; 641 } 642 } 643 644 static void kfree_pmc(struct ip_mc_list *pmc) 645 { 646 ip_sf_list_clear_all(pmc->sources); 647 ip_sf_list_clear_all(pmc->tomb); 648 kfree(pmc); 649 } 650 651 static void igmpv3_send_cr(struct in_device *in_dev) 652 { 653 struct ip_mc_list *pmc, *pmc_prev, *pmc_next; 654 struct sk_buff *skb = NULL; 655 int type, dtype; 656 657 rcu_read_lock(); 658 spin_lock_bh(&in_dev->mc_tomb_lock); 659 660 /* deleted MCA's */ 661 pmc_prev = NULL; 662 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) { 663 pmc_next = pmc->next; 664 if (pmc->sfmode == MCAST_INCLUDE) { 665 type = IGMPV3_BLOCK_OLD_SOURCES; 666 dtype = IGMPV3_BLOCK_OLD_SOURCES; 667 skb = add_grec(skb, pmc, type, 1, 0); 668 skb = add_grec(skb, pmc, dtype, 1, 1); 669 } 670 if (pmc->crcount) { 671 if (pmc->sfmode == MCAST_EXCLUDE) { 672 type = IGMPV3_CHANGE_TO_INCLUDE; 673 skb = add_grec(skb, pmc, type, 1, 0); 674 } 675 pmc->crcount--; 676 if (pmc->crcount == 0) { 677 igmpv3_clear_zeros(&pmc->tomb); 678 igmpv3_clear_zeros(&pmc->sources); 679 } 680 } 681 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) { 682 if (pmc_prev) 683 pmc_prev->next = pmc_next; 684 else 685 in_dev->mc_tomb = pmc_next; 686 in_dev_put(pmc->interface); 687 kfree_pmc(pmc); 688 } else 689 pmc_prev = pmc; 690 } 691 spin_unlock_bh(&in_dev->mc_tomb_lock); 692 693 /* change recs */ 694 for_each_pmc_rcu(in_dev, pmc) { 695 spin_lock_bh(&pmc->lock); 696 if (pmc->sfcount[MCAST_EXCLUDE]) { 697 type = IGMPV3_BLOCK_OLD_SOURCES; 698 dtype = IGMPV3_ALLOW_NEW_SOURCES; 699 } else { 700 type = IGMPV3_ALLOW_NEW_SOURCES; 701 dtype = IGMPV3_BLOCK_OLD_SOURCES; 702 } 703 skb = add_grec(skb, pmc, type, 0, 0); 704 skb = add_grec(skb, pmc, dtype, 0, 1); /* deleted sources */ 705 706 /* filter mode changes */ 707 if (pmc->crcount) { 708 if (pmc->sfmode == MCAST_EXCLUDE) 709 type = IGMPV3_CHANGE_TO_EXCLUDE; 710 else 711 type = IGMPV3_CHANGE_TO_INCLUDE; 712 skb = add_grec(skb, pmc, type, 0, 0); 713 pmc->crcount--; 714 } 715 spin_unlock_bh(&pmc->lock); 716 } 717 rcu_read_unlock(); 718 719 if (!skb) 720 return; 721 (void) igmpv3_sendpack(skb); 722 } 723 724 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc, 725 int type) 726 { 727 struct sk_buff *skb; 728 struct iphdr *iph; 729 struct igmphdr *ih; 730 struct rtable *rt; 731 struct net_device *dev = in_dev->dev; 732 struct net *net = dev_net(dev); 733 __be32 group = pmc ? pmc->multiaddr : 0; 734 struct flowi4 fl4; 735 __be32 dst; 736 int hlen, tlen; 737 738 if (type == IGMPV3_HOST_MEMBERSHIP_REPORT) 739 return igmpv3_send_report(in_dev, pmc); 740 741 if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports) 742 return 0; 743 744 if (type == IGMP_HOST_LEAVE_MESSAGE) 745 dst = IGMP_ALL_ROUTER; 746 else 747 dst = group; 748 749 rt = ip_route_output_ports(net, &fl4, NULL, dst, 0, 750 0, 0, 751 IPPROTO_IGMP, 0, dev->ifindex); 752 if (IS_ERR(rt)) 753 return -1; 754 755 hlen = LL_RESERVED_SPACE(dev); 756 tlen = dev->needed_tailroom; 757 skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC); 758 if (!skb) { 759 ip_rt_put(rt); 760 return -1; 761 } 762 skb->priority = TC_PRIO_CONTROL; 763 764 skb_dst_set(skb, &rt->dst); 765 766 skb_reserve(skb, hlen); 767 768 skb_reset_network_header(skb); 769 iph = ip_hdr(skb); 770 skb_put(skb, sizeof(struct iphdr) + 4); 771 772 iph->version = 4; 773 iph->ihl = (sizeof(struct iphdr)+4)>>2; 774 iph->tos = 0xc0; 775 iph->frag_off = htons(IP_DF); 776 iph->ttl = 1; 777 iph->daddr = dst; 778 iph->saddr = fl4.saddr; 779 iph->protocol = IPPROTO_IGMP; 780 ip_select_ident(net, skb, NULL); 781 ((u8 *)&iph[1])[0] = IPOPT_RA; 782 ((u8 *)&iph[1])[1] = 4; 783 ((u8 *)&iph[1])[2] = 0; 784 ((u8 *)&iph[1])[3] = 0; 785 786 ih = skb_put(skb, sizeof(struct igmphdr)); 787 ih->type = type; 788 ih->code = 0; 789 ih->csum = 0; 790 ih->group = group; 791 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr)); 792 793 return ip_local_out(net, skb->sk, skb); 794 } 795 796 static void igmp_gq_timer_expire(struct timer_list *t) 797 { 798 struct in_device *in_dev = from_timer(in_dev, t, mr_gq_timer); 799 800 in_dev->mr_gq_running = 0; 801 igmpv3_send_report(in_dev, NULL); 802 in_dev_put(in_dev); 803 } 804 805 static void igmp_ifc_timer_expire(struct timer_list *t) 806 { 807 struct in_device *in_dev = from_timer(in_dev, t, mr_ifc_timer); 808 809 igmpv3_send_cr(in_dev); 810 if (in_dev->mr_ifc_count) { 811 in_dev->mr_ifc_count--; 812 igmp_ifc_start_timer(in_dev, 813 unsolicited_report_interval(in_dev)); 814 } 815 in_dev_put(in_dev); 816 } 817 818 static void igmp_ifc_event(struct in_device *in_dev) 819 { 820 struct net *net = dev_net(in_dev->dev); 821 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) 822 return; 823 in_dev->mr_ifc_count = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 824 igmp_ifc_start_timer(in_dev, 1); 825 } 826 827 828 static void igmp_timer_expire(struct timer_list *t) 829 { 830 struct ip_mc_list *im = from_timer(im, t, timer); 831 struct in_device *in_dev = im->interface; 832 833 spin_lock(&im->lock); 834 im->tm_running = 0; 835 836 if (im->unsolicit_count && --im->unsolicit_count) 837 igmp_start_timer(im, unsolicited_report_interval(in_dev)); 838 839 im->reporter = 1; 840 spin_unlock(&im->lock); 841 842 if (IGMP_V1_SEEN(in_dev)) 843 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT); 844 else if (IGMP_V2_SEEN(in_dev)) 845 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT); 846 else 847 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT); 848 849 ip_ma_put(im); 850 } 851 852 /* mark EXCLUDE-mode sources */ 853 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs) 854 { 855 struct ip_sf_list *psf; 856 int i, scount; 857 858 scount = 0; 859 for (psf = pmc->sources; psf; psf = psf->sf_next) { 860 if (scount == nsrcs) 861 break; 862 for (i = 0; i < nsrcs; i++) { 863 /* skip inactive filters */ 864 if (psf->sf_count[MCAST_INCLUDE] || 865 pmc->sfcount[MCAST_EXCLUDE] != 866 psf->sf_count[MCAST_EXCLUDE]) 867 break; 868 if (srcs[i] == psf->sf_inaddr) { 869 scount++; 870 break; 871 } 872 } 873 } 874 pmc->gsquery = 0; 875 if (scount == nsrcs) /* all sources excluded */ 876 return 0; 877 return 1; 878 } 879 880 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs) 881 { 882 struct ip_sf_list *psf; 883 int i, scount; 884 885 if (pmc->sfmode == MCAST_EXCLUDE) 886 return igmp_xmarksources(pmc, nsrcs, srcs); 887 888 /* mark INCLUDE-mode sources */ 889 scount = 0; 890 for (psf = pmc->sources; psf; psf = psf->sf_next) { 891 if (scount == nsrcs) 892 break; 893 for (i = 0; i < nsrcs; i++) 894 if (srcs[i] == psf->sf_inaddr) { 895 psf->sf_gsresp = 1; 896 scount++; 897 break; 898 } 899 } 900 if (!scount) { 901 pmc->gsquery = 0; 902 return 0; 903 } 904 pmc->gsquery = 1; 905 return 1; 906 } 907 908 /* return true if packet was dropped */ 909 static bool igmp_heard_report(struct in_device *in_dev, __be32 group) 910 { 911 struct ip_mc_list *im; 912 struct net *net = dev_net(in_dev->dev); 913 914 /* Timers are only set for non-local groups */ 915 916 if (group == IGMP_ALL_HOSTS) 917 return false; 918 if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports) 919 return false; 920 921 rcu_read_lock(); 922 for_each_pmc_rcu(in_dev, im) { 923 if (im->multiaddr == group) { 924 igmp_stop_timer(im); 925 break; 926 } 927 } 928 rcu_read_unlock(); 929 return false; 930 } 931 932 /* return true if packet was dropped */ 933 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb, 934 int len) 935 { 936 struct igmphdr *ih = igmp_hdr(skb); 937 struct igmpv3_query *ih3 = igmpv3_query_hdr(skb); 938 struct ip_mc_list *im; 939 __be32 group = ih->group; 940 int max_delay; 941 int mark = 0; 942 struct net *net = dev_net(in_dev->dev); 943 944 945 if (len == 8) { 946 if (ih->code == 0) { 947 /* Alas, old v1 router presents here. */ 948 949 max_delay = IGMP_QUERY_RESPONSE_INTERVAL; 950 in_dev->mr_v1_seen = jiffies + 951 (in_dev->mr_qrv * in_dev->mr_qi) + 952 in_dev->mr_qri; 953 group = 0; 954 } else { 955 /* v2 router present */ 956 max_delay = ih->code*(HZ/IGMP_TIMER_SCALE); 957 in_dev->mr_v2_seen = jiffies + 958 (in_dev->mr_qrv * in_dev->mr_qi) + 959 in_dev->mr_qri; 960 } 961 /* cancel the interface change timer */ 962 in_dev->mr_ifc_count = 0; 963 if (del_timer(&in_dev->mr_ifc_timer)) 964 __in_dev_put(in_dev); 965 /* clear deleted report items */ 966 igmpv3_clear_delrec(in_dev); 967 } else if (len < 12) { 968 return true; /* ignore bogus packet; freed by caller */ 969 } else if (IGMP_V1_SEEN(in_dev)) { 970 /* This is a v3 query with v1 queriers present */ 971 max_delay = IGMP_QUERY_RESPONSE_INTERVAL; 972 group = 0; 973 } else if (IGMP_V2_SEEN(in_dev)) { 974 /* this is a v3 query with v2 queriers present; 975 * Interpretation of the max_delay code is problematic here. 976 * A real v2 host would use ih_code directly, while v3 has a 977 * different encoding. We use the v3 encoding as more likely 978 * to be intended in a v3 query. 979 */ 980 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE); 981 if (!max_delay) 982 max_delay = 1; /* can't mod w/ 0 */ 983 } else { /* v3 */ 984 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) 985 return true; 986 987 ih3 = igmpv3_query_hdr(skb); 988 if (ih3->nsrcs) { 989 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query) 990 + ntohs(ih3->nsrcs)*sizeof(__be32))) 991 return true; 992 ih3 = igmpv3_query_hdr(skb); 993 } 994 995 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE); 996 if (!max_delay) 997 max_delay = 1; /* can't mod w/ 0 */ 998 in_dev->mr_maxdelay = max_delay; 999 1000 /* RFC3376, 4.1.6. QRV and 4.1.7. QQIC, when the most recently 1001 * received value was zero, use the default or statically 1002 * configured value. 1003 */ 1004 in_dev->mr_qrv = ih3->qrv ?: net->ipv4.sysctl_igmp_qrv; 1005 in_dev->mr_qi = IGMPV3_QQIC(ih3->qqic)*HZ ?: IGMP_QUERY_INTERVAL; 1006 1007 /* RFC3376, 8.3. Query Response Interval: 1008 * The number of seconds represented by the [Query Response 1009 * Interval] must be less than the [Query Interval]. 1010 */ 1011 if (in_dev->mr_qri >= in_dev->mr_qi) 1012 in_dev->mr_qri = (in_dev->mr_qi/HZ - 1)*HZ; 1013 1014 if (!group) { /* general query */ 1015 if (ih3->nsrcs) 1016 return true; /* no sources allowed */ 1017 igmp_gq_start_timer(in_dev); 1018 return false; 1019 } 1020 /* mark sources to include, if group & source-specific */ 1021 mark = ih3->nsrcs != 0; 1022 } 1023 1024 /* 1025 * - Start the timers in all of our membership records 1026 * that the query applies to for the interface on 1027 * which the query arrived excl. those that belong 1028 * to a "local" group (224.0.0.X) 1029 * - For timers already running check if they need to 1030 * be reset. 1031 * - Use the igmp->igmp_code field as the maximum 1032 * delay possible 1033 */ 1034 rcu_read_lock(); 1035 for_each_pmc_rcu(in_dev, im) { 1036 int changed; 1037 1038 if (group && group != im->multiaddr) 1039 continue; 1040 if (im->multiaddr == IGMP_ALL_HOSTS) 1041 continue; 1042 if (ipv4_is_local_multicast(im->multiaddr) && 1043 !net->ipv4.sysctl_igmp_llm_reports) 1044 continue; 1045 spin_lock_bh(&im->lock); 1046 if (im->tm_running) 1047 im->gsquery = im->gsquery && mark; 1048 else 1049 im->gsquery = mark; 1050 changed = !im->gsquery || 1051 igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs); 1052 spin_unlock_bh(&im->lock); 1053 if (changed) 1054 igmp_mod_timer(im, max_delay); 1055 } 1056 rcu_read_unlock(); 1057 return false; 1058 } 1059 1060 /* called in rcu_read_lock() section */ 1061 int igmp_rcv(struct sk_buff *skb) 1062 { 1063 /* This basically follows the spec line by line -- see RFC1112 */ 1064 struct igmphdr *ih; 1065 struct net_device *dev = skb->dev; 1066 struct in_device *in_dev; 1067 int len = skb->len; 1068 bool dropped = true; 1069 1070 if (netif_is_l3_master(dev)) { 1071 dev = dev_get_by_index_rcu(dev_net(dev), IPCB(skb)->iif); 1072 if (!dev) 1073 goto drop; 1074 } 1075 1076 in_dev = __in_dev_get_rcu(dev); 1077 if (!in_dev) 1078 goto drop; 1079 1080 if (!pskb_may_pull(skb, sizeof(struct igmphdr))) 1081 goto drop; 1082 1083 if (skb_checksum_simple_validate(skb)) 1084 goto drop; 1085 1086 ih = igmp_hdr(skb); 1087 switch (ih->type) { 1088 case IGMP_HOST_MEMBERSHIP_QUERY: 1089 dropped = igmp_heard_query(in_dev, skb, len); 1090 break; 1091 case IGMP_HOST_MEMBERSHIP_REPORT: 1092 case IGMPV2_HOST_MEMBERSHIP_REPORT: 1093 /* Is it our report looped back? */ 1094 if (rt_is_output_route(skb_rtable(skb))) 1095 break; 1096 /* don't rely on MC router hearing unicast reports */ 1097 if (skb->pkt_type == PACKET_MULTICAST || 1098 skb->pkt_type == PACKET_BROADCAST) 1099 dropped = igmp_heard_report(in_dev, ih->group); 1100 break; 1101 case IGMP_PIM: 1102 #ifdef CONFIG_IP_PIMSM_V1 1103 return pim_rcv_v1(skb); 1104 #endif 1105 case IGMPV3_HOST_MEMBERSHIP_REPORT: 1106 case IGMP_DVMRP: 1107 case IGMP_TRACE: 1108 case IGMP_HOST_LEAVE_MESSAGE: 1109 case IGMP_MTRACE: 1110 case IGMP_MTRACE_RESP: 1111 break; 1112 default: 1113 break; 1114 } 1115 1116 drop: 1117 if (dropped) 1118 kfree_skb(skb); 1119 else 1120 consume_skb(skb); 1121 return 0; 1122 } 1123 1124 #endif 1125 1126 1127 /* 1128 * Add a filter to a device 1129 */ 1130 1131 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr) 1132 { 1133 char buf[MAX_ADDR_LEN]; 1134 struct net_device *dev = in_dev->dev; 1135 1136 /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG. 1137 We will get multicast token leakage, when IFF_MULTICAST 1138 is changed. This check should be done in ndo_set_rx_mode 1139 routine. Something sort of: 1140 if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; } 1141 --ANK 1142 */ 1143 if (arp_mc_map(addr, buf, dev, 0) == 0) 1144 dev_mc_add(dev, buf); 1145 } 1146 1147 /* 1148 * Remove a filter from a device 1149 */ 1150 1151 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr) 1152 { 1153 char buf[MAX_ADDR_LEN]; 1154 struct net_device *dev = in_dev->dev; 1155 1156 if (arp_mc_map(addr, buf, dev, 0) == 0) 1157 dev_mc_del(dev, buf); 1158 } 1159 1160 #ifdef CONFIG_IP_MULTICAST 1161 /* 1162 * deleted ip_mc_list manipulation 1163 */ 1164 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im, 1165 gfp_t gfp) 1166 { 1167 struct ip_mc_list *pmc; 1168 struct net *net = dev_net(in_dev->dev); 1169 1170 /* this is an "ip_mc_list" for convenience; only the fields below 1171 * are actually used. In particular, the refcnt and users are not 1172 * used for management of the delete list. Using the same structure 1173 * for deleted items allows change reports to use common code with 1174 * non-deleted or query-response MCA's. 1175 */ 1176 pmc = kzalloc(sizeof(*pmc), gfp); 1177 if (!pmc) 1178 return; 1179 spin_lock_init(&pmc->lock); 1180 spin_lock_bh(&im->lock); 1181 pmc->interface = im->interface; 1182 in_dev_hold(in_dev); 1183 pmc->multiaddr = im->multiaddr; 1184 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 1185 pmc->sfmode = im->sfmode; 1186 if (pmc->sfmode == MCAST_INCLUDE) { 1187 struct ip_sf_list *psf; 1188 1189 pmc->tomb = im->tomb; 1190 pmc->sources = im->sources; 1191 im->tomb = im->sources = NULL; 1192 for (psf = pmc->sources; psf; psf = psf->sf_next) 1193 psf->sf_crcount = pmc->crcount; 1194 } 1195 spin_unlock_bh(&im->lock); 1196 1197 spin_lock_bh(&in_dev->mc_tomb_lock); 1198 pmc->next = in_dev->mc_tomb; 1199 in_dev->mc_tomb = pmc; 1200 spin_unlock_bh(&in_dev->mc_tomb_lock); 1201 } 1202 1203 /* 1204 * restore ip_mc_list deleted records 1205 */ 1206 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im) 1207 { 1208 struct ip_mc_list *pmc, *pmc_prev; 1209 struct ip_sf_list *psf; 1210 struct net *net = dev_net(in_dev->dev); 1211 __be32 multiaddr = im->multiaddr; 1212 1213 spin_lock_bh(&in_dev->mc_tomb_lock); 1214 pmc_prev = NULL; 1215 for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) { 1216 if (pmc->multiaddr == multiaddr) 1217 break; 1218 pmc_prev = pmc; 1219 } 1220 if (pmc) { 1221 if (pmc_prev) 1222 pmc_prev->next = pmc->next; 1223 else 1224 in_dev->mc_tomb = pmc->next; 1225 } 1226 spin_unlock_bh(&in_dev->mc_tomb_lock); 1227 1228 spin_lock_bh(&im->lock); 1229 if (pmc) { 1230 im->interface = pmc->interface; 1231 if (im->sfmode == MCAST_INCLUDE) { 1232 im->tomb = pmc->tomb; 1233 pmc->tomb = NULL; 1234 1235 im->sources = pmc->sources; 1236 pmc->sources = NULL; 1237 1238 for (psf = im->sources; psf; psf = psf->sf_next) 1239 psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 1240 } else { 1241 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 1242 } 1243 in_dev_put(pmc->interface); 1244 kfree_pmc(pmc); 1245 } 1246 spin_unlock_bh(&im->lock); 1247 } 1248 1249 /* 1250 * flush ip_mc_list deleted records 1251 */ 1252 static void igmpv3_clear_delrec(struct in_device *in_dev) 1253 { 1254 struct ip_mc_list *pmc, *nextpmc; 1255 1256 spin_lock_bh(&in_dev->mc_tomb_lock); 1257 pmc = in_dev->mc_tomb; 1258 in_dev->mc_tomb = NULL; 1259 spin_unlock_bh(&in_dev->mc_tomb_lock); 1260 1261 for (; pmc; pmc = nextpmc) { 1262 nextpmc = pmc->next; 1263 ip_mc_clear_src(pmc); 1264 in_dev_put(pmc->interface); 1265 kfree_pmc(pmc); 1266 } 1267 /* clear dead sources, too */ 1268 rcu_read_lock(); 1269 for_each_pmc_rcu(in_dev, pmc) { 1270 struct ip_sf_list *psf; 1271 1272 spin_lock_bh(&pmc->lock); 1273 psf = pmc->tomb; 1274 pmc->tomb = NULL; 1275 spin_unlock_bh(&pmc->lock); 1276 ip_sf_list_clear_all(psf); 1277 } 1278 rcu_read_unlock(); 1279 } 1280 #endif 1281 1282 static void __igmp_group_dropped(struct ip_mc_list *im, gfp_t gfp) 1283 { 1284 struct in_device *in_dev = im->interface; 1285 #ifdef CONFIG_IP_MULTICAST 1286 struct net *net = dev_net(in_dev->dev); 1287 int reporter; 1288 #endif 1289 1290 if (im->loaded) { 1291 im->loaded = 0; 1292 ip_mc_filter_del(in_dev, im->multiaddr); 1293 } 1294 1295 #ifdef CONFIG_IP_MULTICAST 1296 if (im->multiaddr == IGMP_ALL_HOSTS) 1297 return; 1298 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports) 1299 return; 1300 1301 reporter = im->reporter; 1302 igmp_stop_timer(im); 1303 1304 if (!in_dev->dead) { 1305 if (IGMP_V1_SEEN(in_dev)) 1306 return; 1307 if (IGMP_V2_SEEN(in_dev)) { 1308 if (reporter) 1309 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE); 1310 return; 1311 } 1312 /* IGMPv3 */ 1313 igmpv3_add_delrec(in_dev, im, gfp); 1314 1315 igmp_ifc_event(in_dev); 1316 } 1317 #endif 1318 } 1319 1320 static void igmp_group_dropped(struct ip_mc_list *im) 1321 { 1322 __igmp_group_dropped(im, GFP_KERNEL); 1323 } 1324 1325 static void igmp_group_added(struct ip_mc_list *im) 1326 { 1327 struct in_device *in_dev = im->interface; 1328 #ifdef CONFIG_IP_MULTICAST 1329 struct net *net = dev_net(in_dev->dev); 1330 #endif 1331 1332 if (im->loaded == 0) { 1333 im->loaded = 1; 1334 ip_mc_filter_add(in_dev, im->multiaddr); 1335 } 1336 1337 #ifdef CONFIG_IP_MULTICAST 1338 if (im->multiaddr == IGMP_ALL_HOSTS) 1339 return; 1340 if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports) 1341 return; 1342 1343 if (in_dev->dead) 1344 return; 1345 1346 im->unsolicit_count = net->ipv4.sysctl_igmp_qrv; 1347 if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) { 1348 spin_lock_bh(&im->lock); 1349 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY); 1350 spin_unlock_bh(&im->lock); 1351 return; 1352 } 1353 /* else, v3 */ 1354 1355 /* Based on RFC3376 5.1, for newly added INCLUDE SSM, we should 1356 * not send filter-mode change record as the mode should be from 1357 * IN() to IN(A). 1358 */ 1359 if (im->sfmode == MCAST_EXCLUDE) 1360 im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 1361 1362 igmp_ifc_event(in_dev); 1363 #endif 1364 } 1365 1366 1367 /* 1368 * Multicast list managers 1369 */ 1370 1371 static u32 ip_mc_hash(const struct ip_mc_list *im) 1372 { 1373 return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG); 1374 } 1375 1376 static void ip_mc_hash_add(struct in_device *in_dev, 1377 struct ip_mc_list *im) 1378 { 1379 struct ip_mc_list __rcu **mc_hash; 1380 u32 hash; 1381 1382 mc_hash = rtnl_dereference(in_dev->mc_hash); 1383 if (mc_hash) { 1384 hash = ip_mc_hash(im); 1385 im->next_hash = mc_hash[hash]; 1386 rcu_assign_pointer(mc_hash[hash], im); 1387 return; 1388 } 1389 1390 /* do not use a hash table for small number of items */ 1391 if (in_dev->mc_count < 4) 1392 return; 1393 1394 mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG, 1395 GFP_KERNEL); 1396 if (!mc_hash) 1397 return; 1398 1399 for_each_pmc_rtnl(in_dev, im) { 1400 hash = ip_mc_hash(im); 1401 im->next_hash = mc_hash[hash]; 1402 RCU_INIT_POINTER(mc_hash[hash], im); 1403 } 1404 1405 rcu_assign_pointer(in_dev->mc_hash, mc_hash); 1406 } 1407 1408 static void ip_mc_hash_remove(struct in_device *in_dev, 1409 struct ip_mc_list *im) 1410 { 1411 struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash); 1412 struct ip_mc_list *aux; 1413 1414 if (!mc_hash) 1415 return; 1416 mc_hash += ip_mc_hash(im); 1417 while ((aux = rtnl_dereference(*mc_hash)) != im) 1418 mc_hash = &aux->next_hash; 1419 *mc_hash = im->next_hash; 1420 } 1421 1422 1423 /* 1424 * A socket has joined a multicast group on device dev. 1425 */ 1426 static void ____ip_mc_inc_group(struct in_device *in_dev, __be32 addr, 1427 unsigned int mode, gfp_t gfp) 1428 { 1429 struct ip_mc_list *im; 1430 1431 ASSERT_RTNL(); 1432 1433 for_each_pmc_rtnl(in_dev, im) { 1434 if (im->multiaddr == addr) { 1435 im->users++; 1436 ip_mc_add_src(in_dev, &addr, mode, 0, NULL, 0); 1437 goto out; 1438 } 1439 } 1440 1441 im = kzalloc(sizeof(*im), gfp); 1442 if (!im) 1443 goto out; 1444 1445 im->users = 1; 1446 im->interface = in_dev; 1447 in_dev_hold(in_dev); 1448 im->multiaddr = addr; 1449 /* initial mode is (EX, empty) */ 1450 im->sfmode = mode; 1451 im->sfcount[mode] = 1; 1452 refcount_set(&im->refcnt, 1); 1453 spin_lock_init(&im->lock); 1454 #ifdef CONFIG_IP_MULTICAST 1455 timer_setup(&im->timer, igmp_timer_expire, 0); 1456 #endif 1457 1458 im->next_rcu = in_dev->mc_list; 1459 in_dev->mc_count++; 1460 rcu_assign_pointer(in_dev->mc_list, im); 1461 1462 ip_mc_hash_add(in_dev, im); 1463 1464 #ifdef CONFIG_IP_MULTICAST 1465 igmpv3_del_delrec(in_dev, im); 1466 #endif 1467 igmp_group_added(im); 1468 if (!in_dev->dead) 1469 ip_rt_multicast_event(in_dev); 1470 out: 1471 return; 1472 } 1473 1474 void __ip_mc_inc_group(struct in_device *in_dev, __be32 addr, gfp_t gfp) 1475 { 1476 ____ip_mc_inc_group(in_dev, addr, MCAST_EXCLUDE, gfp); 1477 } 1478 EXPORT_SYMBOL(__ip_mc_inc_group); 1479 1480 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr) 1481 { 1482 __ip_mc_inc_group(in_dev, addr, MCAST_EXCLUDE); 1483 } 1484 EXPORT_SYMBOL(ip_mc_inc_group); 1485 1486 static int ip_mc_check_iphdr(struct sk_buff *skb) 1487 { 1488 const struct iphdr *iph; 1489 unsigned int len; 1490 unsigned int offset = skb_network_offset(skb) + sizeof(*iph); 1491 1492 if (!pskb_may_pull(skb, offset)) 1493 return -EINVAL; 1494 1495 iph = ip_hdr(skb); 1496 1497 if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph)) 1498 return -EINVAL; 1499 1500 offset += ip_hdrlen(skb) - sizeof(*iph); 1501 1502 if (!pskb_may_pull(skb, offset)) 1503 return -EINVAL; 1504 1505 iph = ip_hdr(skb); 1506 1507 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) 1508 return -EINVAL; 1509 1510 len = skb_network_offset(skb) + ntohs(iph->tot_len); 1511 if (skb->len < len || len < offset) 1512 return -EINVAL; 1513 1514 skb_set_transport_header(skb, offset); 1515 1516 return 0; 1517 } 1518 1519 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb) 1520 { 1521 unsigned int len = skb_transport_offset(skb); 1522 1523 len += sizeof(struct igmpv3_report); 1524 1525 return ip_mc_may_pull(skb, len) ? 0 : -EINVAL; 1526 } 1527 1528 static int ip_mc_check_igmp_query(struct sk_buff *skb) 1529 { 1530 unsigned int transport_len = ip_transport_len(skb); 1531 unsigned int len; 1532 1533 /* IGMPv{1,2}? */ 1534 if (transport_len != sizeof(struct igmphdr)) { 1535 /* or IGMPv3? */ 1536 if (transport_len < sizeof(struct igmpv3_query)) 1537 return -EINVAL; 1538 1539 len = skb_transport_offset(skb) + sizeof(struct igmpv3_query); 1540 if (!ip_mc_may_pull(skb, len)) 1541 return -EINVAL; 1542 } 1543 1544 /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer 1545 * all-systems destination addresses (224.0.0.1) for general queries 1546 */ 1547 if (!igmp_hdr(skb)->group && 1548 ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP)) 1549 return -EINVAL; 1550 1551 return 0; 1552 } 1553 1554 static int ip_mc_check_igmp_msg(struct sk_buff *skb) 1555 { 1556 switch (igmp_hdr(skb)->type) { 1557 case IGMP_HOST_LEAVE_MESSAGE: 1558 case IGMP_HOST_MEMBERSHIP_REPORT: 1559 case IGMPV2_HOST_MEMBERSHIP_REPORT: 1560 return 0; 1561 case IGMPV3_HOST_MEMBERSHIP_REPORT: 1562 return ip_mc_check_igmp_reportv3(skb); 1563 case IGMP_HOST_MEMBERSHIP_QUERY: 1564 return ip_mc_check_igmp_query(skb); 1565 default: 1566 return -ENOMSG; 1567 } 1568 } 1569 1570 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb) 1571 { 1572 return skb_checksum_simple_validate(skb); 1573 } 1574 1575 static int ip_mc_check_igmp_csum(struct sk_buff *skb) 1576 { 1577 unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr); 1578 unsigned int transport_len = ip_transport_len(skb); 1579 struct sk_buff *skb_chk; 1580 1581 if (!ip_mc_may_pull(skb, len)) 1582 return -EINVAL; 1583 1584 skb_chk = skb_checksum_trimmed(skb, transport_len, 1585 ip_mc_validate_checksum); 1586 if (!skb_chk) 1587 return -EINVAL; 1588 1589 if (skb_chk != skb) 1590 kfree_skb(skb_chk); 1591 1592 return 0; 1593 } 1594 1595 /** 1596 * ip_mc_check_igmp - checks whether this is a sane IGMP packet 1597 * @skb: the skb to validate 1598 * 1599 * Checks whether an IPv4 packet is a valid IGMP packet. If so sets 1600 * skb transport header accordingly and returns zero. 1601 * 1602 * -EINVAL: A broken packet was detected, i.e. it violates some internet 1603 * standard 1604 * -ENOMSG: IP header validation succeeded but it is not an IGMP packet. 1605 * -ENOMEM: A memory allocation failure happened. 1606 * 1607 * Caller needs to set the skb network header and free any returned skb if it 1608 * differs from the provided skb. 1609 */ 1610 int ip_mc_check_igmp(struct sk_buff *skb) 1611 { 1612 int ret = ip_mc_check_iphdr(skb); 1613 1614 if (ret < 0) 1615 return ret; 1616 1617 if (ip_hdr(skb)->protocol != IPPROTO_IGMP) 1618 return -ENOMSG; 1619 1620 ret = ip_mc_check_igmp_csum(skb); 1621 if (ret < 0) 1622 return ret; 1623 1624 return ip_mc_check_igmp_msg(skb); 1625 } 1626 EXPORT_SYMBOL(ip_mc_check_igmp); 1627 1628 /* 1629 * Resend IGMP JOIN report; used by netdev notifier. 1630 */ 1631 static void ip_mc_rejoin_groups(struct in_device *in_dev) 1632 { 1633 #ifdef CONFIG_IP_MULTICAST 1634 struct ip_mc_list *im; 1635 int type; 1636 struct net *net = dev_net(in_dev->dev); 1637 1638 ASSERT_RTNL(); 1639 1640 for_each_pmc_rtnl(in_dev, im) { 1641 if (im->multiaddr == IGMP_ALL_HOSTS) 1642 continue; 1643 if (ipv4_is_local_multicast(im->multiaddr) && 1644 !net->ipv4.sysctl_igmp_llm_reports) 1645 continue; 1646 1647 /* a failover is happening and switches 1648 * must be notified immediately 1649 */ 1650 if (IGMP_V1_SEEN(in_dev)) 1651 type = IGMP_HOST_MEMBERSHIP_REPORT; 1652 else if (IGMP_V2_SEEN(in_dev)) 1653 type = IGMPV2_HOST_MEMBERSHIP_REPORT; 1654 else 1655 type = IGMPV3_HOST_MEMBERSHIP_REPORT; 1656 igmp_send_report(in_dev, im, type); 1657 } 1658 #endif 1659 } 1660 1661 /* 1662 * A socket has left a multicast group on device dev 1663 */ 1664 1665 void __ip_mc_dec_group(struct in_device *in_dev, __be32 addr, gfp_t gfp) 1666 { 1667 struct ip_mc_list *i; 1668 struct ip_mc_list __rcu **ip; 1669 1670 ASSERT_RTNL(); 1671 1672 for (ip = &in_dev->mc_list; 1673 (i = rtnl_dereference(*ip)) != NULL; 1674 ip = &i->next_rcu) { 1675 if (i->multiaddr == addr) { 1676 if (--i->users == 0) { 1677 ip_mc_hash_remove(in_dev, i); 1678 *ip = i->next_rcu; 1679 in_dev->mc_count--; 1680 __igmp_group_dropped(i, gfp); 1681 ip_mc_clear_src(i); 1682 1683 if (!in_dev->dead) 1684 ip_rt_multicast_event(in_dev); 1685 1686 ip_ma_put(i); 1687 return; 1688 } 1689 break; 1690 } 1691 } 1692 } 1693 EXPORT_SYMBOL(__ip_mc_dec_group); 1694 1695 /* Device changing type */ 1696 1697 void ip_mc_unmap(struct in_device *in_dev) 1698 { 1699 struct ip_mc_list *pmc; 1700 1701 ASSERT_RTNL(); 1702 1703 for_each_pmc_rtnl(in_dev, pmc) 1704 igmp_group_dropped(pmc); 1705 } 1706 1707 void ip_mc_remap(struct in_device *in_dev) 1708 { 1709 struct ip_mc_list *pmc; 1710 1711 ASSERT_RTNL(); 1712 1713 for_each_pmc_rtnl(in_dev, pmc) { 1714 #ifdef CONFIG_IP_MULTICAST 1715 igmpv3_del_delrec(in_dev, pmc); 1716 #endif 1717 igmp_group_added(pmc); 1718 } 1719 } 1720 1721 /* Device going down */ 1722 1723 void ip_mc_down(struct in_device *in_dev) 1724 { 1725 struct ip_mc_list *pmc; 1726 1727 ASSERT_RTNL(); 1728 1729 for_each_pmc_rtnl(in_dev, pmc) 1730 igmp_group_dropped(pmc); 1731 1732 #ifdef CONFIG_IP_MULTICAST 1733 in_dev->mr_ifc_count = 0; 1734 if (del_timer(&in_dev->mr_ifc_timer)) 1735 __in_dev_put(in_dev); 1736 in_dev->mr_gq_running = 0; 1737 if (del_timer(&in_dev->mr_gq_timer)) 1738 __in_dev_put(in_dev); 1739 #endif 1740 1741 ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS); 1742 } 1743 1744 #ifdef CONFIG_IP_MULTICAST 1745 static void ip_mc_reset(struct in_device *in_dev) 1746 { 1747 struct net *net = dev_net(in_dev->dev); 1748 1749 in_dev->mr_qi = IGMP_QUERY_INTERVAL; 1750 in_dev->mr_qri = IGMP_QUERY_RESPONSE_INTERVAL; 1751 in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv; 1752 } 1753 #else 1754 static void ip_mc_reset(struct in_device *in_dev) 1755 { 1756 } 1757 #endif 1758 1759 void ip_mc_init_dev(struct in_device *in_dev) 1760 { 1761 ASSERT_RTNL(); 1762 1763 #ifdef CONFIG_IP_MULTICAST 1764 timer_setup(&in_dev->mr_gq_timer, igmp_gq_timer_expire, 0); 1765 timer_setup(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire, 0); 1766 #endif 1767 ip_mc_reset(in_dev); 1768 1769 spin_lock_init(&in_dev->mc_tomb_lock); 1770 } 1771 1772 /* Device going up */ 1773 1774 void ip_mc_up(struct in_device *in_dev) 1775 { 1776 struct ip_mc_list *pmc; 1777 1778 ASSERT_RTNL(); 1779 1780 ip_mc_reset(in_dev); 1781 ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS); 1782 1783 for_each_pmc_rtnl(in_dev, pmc) { 1784 #ifdef CONFIG_IP_MULTICAST 1785 igmpv3_del_delrec(in_dev, pmc); 1786 #endif 1787 igmp_group_added(pmc); 1788 } 1789 } 1790 1791 /* 1792 * Device is about to be destroyed: clean up. 1793 */ 1794 1795 void ip_mc_destroy_dev(struct in_device *in_dev) 1796 { 1797 struct ip_mc_list *i; 1798 1799 ASSERT_RTNL(); 1800 1801 /* Deactivate timers */ 1802 ip_mc_down(in_dev); 1803 #ifdef CONFIG_IP_MULTICAST 1804 igmpv3_clear_delrec(in_dev); 1805 #endif 1806 1807 while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) { 1808 in_dev->mc_list = i->next_rcu; 1809 in_dev->mc_count--; 1810 ip_ma_put(i); 1811 } 1812 } 1813 1814 /* RTNL is locked */ 1815 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr) 1816 { 1817 struct net_device *dev = NULL; 1818 struct in_device *idev = NULL; 1819 1820 if (imr->imr_ifindex) { 1821 idev = inetdev_by_index(net, imr->imr_ifindex); 1822 return idev; 1823 } 1824 if (imr->imr_address.s_addr) { 1825 dev = __ip_dev_find(net, imr->imr_address.s_addr, false); 1826 if (!dev) 1827 return NULL; 1828 } 1829 1830 if (!dev) { 1831 struct rtable *rt = ip_route_output(net, 1832 imr->imr_multiaddr.s_addr, 1833 0, 0, 0); 1834 if (!IS_ERR(rt)) { 1835 dev = rt->dst.dev; 1836 ip_rt_put(rt); 1837 } 1838 } 1839 if (dev) { 1840 imr->imr_ifindex = dev->ifindex; 1841 idev = __in_dev_get_rtnl(dev); 1842 } 1843 return idev; 1844 } 1845 1846 /* 1847 * Join a socket to a group 1848 */ 1849 1850 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode, 1851 __be32 *psfsrc) 1852 { 1853 struct ip_sf_list *psf, *psf_prev; 1854 int rv = 0; 1855 1856 psf_prev = NULL; 1857 for (psf = pmc->sources; psf; psf = psf->sf_next) { 1858 if (psf->sf_inaddr == *psfsrc) 1859 break; 1860 psf_prev = psf; 1861 } 1862 if (!psf || psf->sf_count[sfmode] == 0) { 1863 /* source filter not found, or count wrong => bug */ 1864 return -ESRCH; 1865 } 1866 psf->sf_count[sfmode]--; 1867 if (psf->sf_count[sfmode] == 0) { 1868 ip_rt_multicast_event(pmc->interface); 1869 } 1870 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) { 1871 #ifdef CONFIG_IP_MULTICAST 1872 struct in_device *in_dev = pmc->interface; 1873 struct net *net = dev_net(in_dev->dev); 1874 #endif 1875 1876 /* no more filters for this source */ 1877 if (psf_prev) 1878 psf_prev->sf_next = psf->sf_next; 1879 else 1880 pmc->sources = psf->sf_next; 1881 #ifdef CONFIG_IP_MULTICAST 1882 if (psf->sf_oldin && 1883 !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) { 1884 psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 1885 psf->sf_next = pmc->tomb; 1886 pmc->tomb = psf; 1887 rv = 1; 1888 } else 1889 #endif 1890 kfree(psf); 1891 } 1892 return rv; 1893 } 1894 1895 #ifndef CONFIG_IP_MULTICAST 1896 #define igmp_ifc_event(x) do { } while (0) 1897 #endif 1898 1899 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode, 1900 int sfcount, __be32 *psfsrc, int delta) 1901 { 1902 struct ip_mc_list *pmc; 1903 int changerec = 0; 1904 int i, err; 1905 1906 if (!in_dev) 1907 return -ENODEV; 1908 rcu_read_lock(); 1909 for_each_pmc_rcu(in_dev, pmc) { 1910 if (*pmca == pmc->multiaddr) 1911 break; 1912 } 1913 if (!pmc) { 1914 /* MCA not found?? bug */ 1915 rcu_read_unlock(); 1916 return -ESRCH; 1917 } 1918 spin_lock_bh(&pmc->lock); 1919 rcu_read_unlock(); 1920 #ifdef CONFIG_IP_MULTICAST 1921 sf_markstate(pmc); 1922 #endif 1923 if (!delta) { 1924 err = -EINVAL; 1925 if (!pmc->sfcount[sfmode]) 1926 goto out_unlock; 1927 pmc->sfcount[sfmode]--; 1928 } 1929 err = 0; 1930 for (i = 0; i < sfcount; i++) { 1931 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]); 1932 1933 changerec |= rv > 0; 1934 if (!err && rv < 0) 1935 err = rv; 1936 } 1937 if (pmc->sfmode == MCAST_EXCLUDE && 1938 pmc->sfcount[MCAST_EXCLUDE] == 0 && 1939 pmc->sfcount[MCAST_INCLUDE]) { 1940 #ifdef CONFIG_IP_MULTICAST 1941 struct ip_sf_list *psf; 1942 struct net *net = dev_net(in_dev->dev); 1943 #endif 1944 1945 /* filter mode change */ 1946 pmc->sfmode = MCAST_INCLUDE; 1947 #ifdef CONFIG_IP_MULTICAST 1948 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 1949 in_dev->mr_ifc_count = pmc->crcount; 1950 for (psf = pmc->sources; psf; psf = psf->sf_next) 1951 psf->sf_crcount = 0; 1952 igmp_ifc_event(pmc->interface); 1953 } else if (sf_setstate(pmc) || changerec) { 1954 igmp_ifc_event(pmc->interface); 1955 #endif 1956 } 1957 out_unlock: 1958 spin_unlock_bh(&pmc->lock); 1959 return err; 1960 } 1961 1962 /* 1963 * Add multicast single-source filter to the interface list 1964 */ 1965 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode, 1966 __be32 *psfsrc) 1967 { 1968 struct ip_sf_list *psf, *psf_prev; 1969 1970 psf_prev = NULL; 1971 for (psf = pmc->sources; psf; psf = psf->sf_next) { 1972 if (psf->sf_inaddr == *psfsrc) 1973 break; 1974 psf_prev = psf; 1975 } 1976 if (!psf) { 1977 psf = kzalloc(sizeof(*psf), GFP_ATOMIC); 1978 if (!psf) 1979 return -ENOBUFS; 1980 psf->sf_inaddr = *psfsrc; 1981 if (psf_prev) { 1982 psf_prev->sf_next = psf; 1983 } else 1984 pmc->sources = psf; 1985 } 1986 psf->sf_count[sfmode]++; 1987 if (psf->sf_count[sfmode] == 1) { 1988 ip_rt_multicast_event(pmc->interface); 1989 } 1990 return 0; 1991 } 1992 1993 #ifdef CONFIG_IP_MULTICAST 1994 static void sf_markstate(struct ip_mc_list *pmc) 1995 { 1996 struct ip_sf_list *psf; 1997 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE]; 1998 1999 for (psf = pmc->sources; psf; psf = psf->sf_next) 2000 if (pmc->sfcount[MCAST_EXCLUDE]) { 2001 psf->sf_oldin = mca_xcount == 2002 psf->sf_count[MCAST_EXCLUDE] && 2003 !psf->sf_count[MCAST_INCLUDE]; 2004 } else 2005 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0; 2006 } 2007 2008 static int sf_setstate(struct ip_mc_list *pmc) 2009 { 2010 struct ip_sf_list *psf, *dpsf; 2011 int mca_xcount = pmc->sfcount[MCAST_EXCLUDE]; 2012 int qrv = pmc->interface->mr_qrv; 2013 int new_in, rv; 2014 2015 rv = 0; 2016 for (psf = pmc->sources; psf; psf = psf->sf_next) { 2017 if (pmc->sfcount[MCAST_EXCLUDE]) { 2018 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] && 2019 !psf->sf_count[MCAST_INCLUDE]; 2020 } else 2021 new_in = psf->sf_count[MCAST_INCLUDE] != 0; 2022 if (new_in) { 2023 if (!psf->sf_oldin) { 2024 struct ip_sf_list *prev = NULL; 2025 2026 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) { 2027 if (dpsf->sf_inaddr == psf->sf_inaddr) 2028 break; 2029 prev = dpsf; 2030 } 2031 if (dpsf) { 2032 if (prev) 2033 prev->sf_next = dpsf->sf_next; 2034 else 2035 pmc->tomb = dpsf->sf_next; 2036 kfree(dpsf); 2037 } 2038 psf->sf_crcount = qrv; 2039 rv++; 2040 } 2041 } else if (psf->sf_oldin) { 2042 2043 psf->sf_crcount = 0; 2044 /* 2045 * add or update "delete" records if an active filter 2046 * is now inactive 2047 */ 2048 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) 2049 if (dpsf->sf_inaddr == psf->sf_inaddr) 2050 break; 2051 if (!dpsf) { 2052 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC); 2053 if (!dpsf) 2054 continue; 2055 *dpsf = *psf; 2056 /* pmc->lock held by callers */ 2057 dpsf->sf_next = pmc->tomb; 2058 pmc->tomb = dpsf; 2059 } 2060 dpsf->sf_crcount = qrv; 2061 rv++; 2062 } 2063 } 2064 return rv; 2065 } 2066 #endif 2067 2068 /* 2069 * Add multicast source filter list to the interface list 2070 */ 2071 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode, 2072 int sfcount, __be32 *psfsrc, int delta) 2073 { 2074 struct ip_mc_list *pmc; 2075 int isexclude; 2076 int i, err; 2077 2078 if (!in_dev) 2079 return -ENODEV; 2080 rcu_read_lock(); 2081 for_each_pmc_rcu(in_dev, pmc) { 2082 if (*pmca == pmc->multiaddr) 2083 break; 2084 } 2085 if (!pmc) { 2086 /* MCA not found?? bug */ 2087 rcu_read_unlock(); 2088 return -ESRCH; 2089 } 2090 spin_lock_bh(&pmc->lock); 2091 rcu_read_unlock(); 2092 2093 #ifdef CONFIG_IP_MULTICAST 2094 sf_markstate(pmc); 2095 #endif 2096 isexclude = pmc->sfmode == MCAST_EXCLUDE; 2097 if (!delta) 2098 pmc->sfcount[sfmode]++; 2099 err = 0; 2100 for (i = 0; i < sfcount; i++) { 2101 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]); 2102 if (err) 2103 break; 2104 } 2105 if (err) { 2106 int j; 2107 2108 if (!delta) 2109 pmc->sfcount[sfmode]--; 2110 for (j = 0; j < i; j++) 2111 (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]); 2112 } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) { 2113 #ifdef CONFIG_IP_MULTICAST 2114 struct ip_sf_list *psf; 2115 struct net *net = dev_net(pmc->interface->dev); 2116 in_dev = pmc->interface; 2117 #endif 2118 2119 /* filter mode change */ 2120 if (pmc->sfcount[MCAST_EXCLUDE]) 2121 pmc->sfmode = MCAST_EXCLUDE; 2122 else if (pmc->sfcount[MCAST_INCLUDE]) 2123 pmc->sfmode = MCAST_INCLUDE; 2124 #ifdef CONFIG_IP_MULTICAST 2125 /* else no filters; keep old mode for reports */ 2126 2127 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv; 2128 in_dev->mr_ifc_count = pmc->crcount; 2129 for (psf = pmc->sources; psf; psf = psf->sf_next) 2130 psf->sf_crcount = 0; 2131 igmp_ifc_event(in_dev); 2132 } else if (sf_setstate(pmc)) { 2133 igmp_ifc_event(in_dev); 2134 #endif 2135 } 2136 spin_unlock_bh(&pmc->lock); 2137 return err; 2138 } 2139 2140 static void ip_mc_clear_src(struct ip_mc_list *pmc) 2141 { 2142 struct ip_sf_list *tomb, *sources; 2143 2144 spin_lock_bh(&pmc->lock); 2145 tomb = pmc->tomb; 2146 pmc->tomb = NULL; 2147 sources = pmc->sources; 2148 pmc->sources = NULL; 2149 pmc->sfmode = MCAST_EXCLUDE; 2150 pmc->sfcount[MCAST_INCLUDE] = 0; 2151 pmc->sfcount[MCAST_EXCLUDE] = 1; 2152 spin_unlock_bh(&pmc->lock); 2153 2154 ip_sf_list_clear_all(tomb); 2155 ip_sf_list_clear_all(sources); 2156 } 2157 2158 /* Join a multicast group 2159 */ 2160 static int __ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr, 2161 unsigned int mode) 2162 { 2163 __be32 addr = imr->imr_multiaddr.s_addr; 2164 struct ip_mc_socklist *iml, *i; 2165 struct in_device *in_dev; 2166 struct inet_sock *inet = inet_sk(sk); 2167 struct net *net = sock_net(sk); 2168 int ifindex; 2169 int count = 0; 2170 int err; 2171 2172 ASSERT_RTNL(); 2173 2174 if (!ipv4_is_multicast(addr)) 2175 return -EINVAL; 2176 2177 in_dev = ip_mc_find_dev(net, imr); 2178 2179 if (!in_dev) { 2180 err = -ENODEV; 2181 goto done; 2182 } 2183 2184 err = -EADDRINUSE; 2185 ifindex = imr->imr_ifindex; 2186 for_each_pmc_rtnl(inet, i) { 2187 if (i->multi.imr_multiaddr.s_addr == addr && 2188 i->multi.imr_ifindex == ifindex) 2189 goto done; 2190 count++; 2191 } 2192 err = -ENOBUFS; 2193 if (count >= net->ipv4.sysctl_igmp_max_memberships) 2194 goto done; 2195 iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL); 2196 if (!iml) 2197 goto done; 2198 2199 memcpy(&iml->multi, imr, sizeof(*imr)); 2200 iml->next_rcu = inet->mc_list; 2201 iml->sflist = NULL; 2202 iml->sfmode = mode; 2203 rcu_assign_pointer(inet->mc_list, iml); 2204 __ip_mc_inc_group(in_dev, addr, mode); 2205 err = 0; 2206 done: 2207 return err; 2208 } 2209 2210 /* Join ASM (Any-Source Multicast) group 2211 */ 2212 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr) 2213 { 2214 return __ip_mc_join_group(sk, imr, MCAST_EXCLUDE); 2215 } 2216 EXPORT_SYMBOL(ip_mc_join_group); 2217 2218 /* Join SSM (Source-Specific Multicast) group 2219 */ 2220 int ip_mc_join_group_ssm(struct sock *sk, struct ip_mreqn *imr, 2221 unsigned int mode) 2222 { 2223 return __ip_mc_join_group(sk, imr, mode); 2224 } 2225 2226 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml, 2227 struct in_device *in_dev) 2228 { 2229 struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist); 2230 int err; 2231 2232 if (!psf) { 2233 /* any-source empty exclude case */ 2234 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr, 2235 iml->sfmode, 0, NULL, 0); 2236 } 2237 err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr, 2238 iml->sfmode, psf->sl_count, psf->sl_addr, 0); 2239 RCU_INIT_POINTER(iml->sflist, NULL); 2240 /* decrease mem now to avoid the memleak warning */ 2241 atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc); 2242 kfree_rcu(psf, rcu); 2243 return err; 2244 } 2245 2246 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr) 2247 { 2248 struct inet_sock *inet = inet_sk(sk); 2249 struct ip_mc_socklist *iml; 2250 struct ip_mc_socklist __rcu **imlp; 2251 struct in_device *in_dev; 2252 struct net *net = sock_net(sk); 2253 __be32 group = imr->imr_multiaddr.s_addr; 2254 u32 ifindex; 2255 int ret = -EADDRNOTAVAIL; 2256 2257 ASSERT_RTNL(); 2258 2259 in_dev = ip_mc_find_dev(net, imr); 2260 if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) { 2261 ret = -ENODEV; 2262 goto out; 2263 } 2264 ifindex = imr->imr_ifindex; 2265 for (imlp = &inet->mc_list; 2266 (iml = rtnl_dereference(*imlp)) != NULL; 2267 imlp = &iml->next_rcu) { 2268 if (iml->multi.imr_multiaddr.s_addr != group) 2269 continue; 2270 if (ifindex) { 2271 if (iml->multi.imr_ifindex != ifindex) 2272 continue; 2273 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr != 2274 iml->multi.imr_address.s_addr) 2275 continue; 2276 2277 (void) ip_mc_leave_src(sk, iml, in_dev); 2278 2279 *imlp = iml->next_rcu; 2280 2281 if (in_dev) 2282 ip_mc_dec_group(in_dev, group); 2283 2284 /* decrease mem now to avoid the memleak warning */ 2285 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc); 2286 kfree_rcu(iml, rcu); 2287 return 0; 2288 } 2289 out: 2290 return ret; 2291 } 2292 EXPORT_SYMBOL(ip_mc_leave_group); 2293 2294 int ip_mc_source(int add, int omode, struct sock *sk, struct 2295 ip_mreq_source *mreqs, int ifindex) 2296 { 2297 int err; 2298 struct ip_mreqn imr; 2299 __be32 addr = mreqs->imr_multiaddr; 2300 struct ip_mc_socklist *pmc; 2301 struct in_device *in_dev = NULL; 2302 struct inet_sock *inet = inet_sk(sk); 2303 struct ip_sf_socklist *psl; 2304 struct net *net = sock_net(sk); 2305 int leavegroup = 0; 2306 int i, j, rv; 2307 2308 if (!ipv4_is_multicast(addr)) 2309 return -EINVAL; 2310 2311 ASSERT_RTNL(); 2312 2313 imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr; 2314 imr.imr_address.s_addr = mreqs->imr_interface; 2315 imr.imr_ifindex = ifindex; 2316 in_dev = ip_mc_find_dev(net, &imr); 2317 2318 if (!in_dev) { 2319 err = -ENODEV; 2320 goto done; 2321 } 2322 err = -EADDRNOTAVAIL; 2323 2324 for_each_pmc_rtnl(inet, pmc) { 2325 if ((pmc->multi.imr_multiaddr.s_addr == 2326 imr.imr_multiaddr.s_addr) && 2327 (pmc->multi.imr_ifindex == imr.imr_ifindex)) 2328 break; 2329 } 2330 if (!pmc) { /* must have a prior join */ 2331 err = -EINVAL; 2332 goto done; 2333 } 2334 /* if a source filter was set, must be the same mode as before */ 2335 if (pmc->sflist) { 2336 if (pmc->sfmode != omode) { 2337 err = -EINVAL; 2338 goto done; 2339 } 2340 } else if (pmc->sfmode != omode) { 2341 /* allow mode switches for empty-set filters */ 2342 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0); 2343 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0, 2344 NULL, 0); 2345 pmc->sfmode = omode; 2346 } 2347 2348 psl = rtnl_dereference(pmc->sflist); 2349 if (!add) { 2350 if (!psl) 2351 goto done; /* err = -EADDRNOTAVAIL */ 2352 rv = !0; 2353 for (i = 0; i < psl->sl_count; i++) { 2354 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr, 2355 sizeof(__be32)); 2356 if (rv == 0) 2357 break; 2358 } 2359 if (rv) /* source not found */ 2360 goto done; /* err = -EADDRNOTAVAIL */ 2361 2362 /* special case - (INCLUDE, empty) == LEAVE_GROUP */ 2363 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) { 2364 leavegroup = 1; 2365 goto done; 2366 } 2367 2368 /* update the interface filter */ 2369 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1, 2370 &mreqs->imr_sourceaddr, 1); 2371 2372 for (j = i+1; j < psl->sl_count; j++) 2373 psl->sl_addr[j-1] = psl->sl_addr[j]; 2374 psl->sl_count--; 2375 err = 0; 2376 goto done; 2377 } 2378 /* else, add a new source to the filter */ 2379 2380 if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) { 2381 err = -ENOBUFS; 2382 goto done; 2383 } 2384 if (!psl || psl->sl_count == psl->sl_max) { 2385 struct ip_sf_socklist *newpsl; 2386 int count = IP_SFBLOCK; 2387 2388 if (psl) 2389 count += psl->sl_max; 2390 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL); 2391 if (!newpsl) { 2392 err = -ENOBUFS; 2393 goto done; 2394 } 2395 newpsl->sl_max = count; 2396 newpsl->sl_count = count - IP_SFBLOCK; 2397 if (psl) { 2398 for (i = 0; i < psl->sl_count; i++) 2399 newpsl->sl_addr[i] = psl->sl_addr[i]; 2400 /* decrease mem now to avoid the memleak warning */ 2401 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc); 2402 kfree_rcu(psl, rcu); 2403 } 2404 rcu_assign_pointer(pmc->sflist, newpsl); 2405 psl = newpsl; 2406 } 2407 rv = 1; /* > 0 for insert logic below if sl_count is 0 */ 2408 for (i = 0; i < psl->sl_count; i++) { 2409 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr, 2410 sizeof(__be32)); 2411 if (rv == 0) 2412 break; 2413 } 2414 if (rv == 0) /* address already there is an error */ 2415 goto done; 2416 for (j = psl->sl_count-1; j >= i; j--) 2417 psl->sl_addr[j+1] = psl->sl_addr[j]; 2418 psl->sl_addr[i] = mreqs->imr_sourceaddr; 2419 psl->sl_count++; 2420 err = 0; 2421 /* update the interface list */ 2422 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1, 2423 &mreqs->imr_sourceaddr, 1); 2424 done: 2425 if (leavegroup) 2426 err = ip_mc_leave_group(sk, &imr); 2427 return err; 2428 } 2429 2430 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex) 2431 { 2432 int err = 0; 2433 struct ip_mreqn imr; 2434 __be32 addr = msf->imsf_multiaddr; 2435 struct ip_mc_socklist *pmc; 2436 struct in_device *in_dev; 2437 struct inet_sock *inet = inet_sk(sk); 2438 struct ip_sf_socklist *newpsl, *psl; 2439 struct net *net = sock_net(sk); 2440 int leavegroup = 0; 2441 2442 if (!ipv4_is_multicast(addr)) 2443 return -EINVAL; 2444 if (msf->imsf_fmode != MCAST_INCLUDE && 2445 msf->imsf_fmode != MCAST_EXCLUDE) 2446 return -EINVAL; 2447 2448 ASSERT_RTNL(); 2449 2450 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr; 2451 imr.imr_address.s_addr = msf->imsf_interface; 2452 imr.imr_ifindex = ifindex; 2453 in_dev = ip_mc_find_dev(net, &imr); 2454 2455 if (!in_dev) { 2456 err = -ENODEV; 2457 goto done; 2458 } 2459 2460 /* special case - (INCLUDE, empty) == LEAVE_GROUP */ 2461 if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) { 2462 leavegroup = 1; 2463 goto done; 2464 } 2465 2466 for_each_pmc_rtnl(inet, pmc) { 2467 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr && 2468 pmc->multi.imr_ifindex == imr.imr_ifindex) 2469 break; 2470 } 2471 if (!pmc) { /* must have a prior join */ 2472 err = -EINVAL; 2473 goto done; 2474 } 2475 if (msf->imsf_numsrc) { 2476 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc), 2477 GFP_KERNEL); 2478 if (!newpsl) { 2479 err = -ENOBUFS; 2480 goto done; 2481 } 2482 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc; 2483 memcpy(newpsl->sl_addr, msf->imsf_slist, 2484 msf->imsf_numsrc * sizeof(msf->imsf_slist[0])); 2485 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr, 2486 msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0); 2487 if (err) { 2488 sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max)); 2489 goto done; 2490 } 2491 } else { 2492 newpsl = NULL; 2493 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr, 2494 msf->imsf_fmode, 0, NULL, 0); 2495 } 2496 psl = rtnl_dereference(pmc->sflist); 2497 if (psl) { 2498 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode, 2499 psl->sl_count, psl->sl_addr, 0); 2500 /* decrease mem now to avoid the memleak warning */ 2501 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc); 2502 kfree_rcu(psl, rcu); 2503 } else 2504 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode, 2505 0, NULL, 0); 2506 rcu_assign_pointer(pmc->sflist, newpsl); 2507 pmc->sfmode = msf->imsf_fmode; 2508 err = 0; 2509 done: 2510 if (leavegroup) 2511 err = ip_mc_leave_group(sk, &imr); 2512 return err; 2513 } 2514 2515 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf, 2516 struct ip_msfilter __user *optval, int __user *optlen) 2517 { 2518 int err, len, count, copycount; 2519 struct ip_mreqn imr; 2520 __be32 addr = msf->imsf_multiaddr; 2521 struct ip_mc_socklist *pmc; 2522 struct in_device *in_dev; 2523 struct inet_sock *inet = inet_sk(sk); 2524 struct ip_sf_socklist *psl; 2525 struct net *net = sock_net(sk); 2526 2527 ASSERT_RTNL(); 2528 2529 if (!ipv4_is_multicast(addr)) 2530 return -EINVAL; 2531 2532 imr.imr_multiaddr.s_addr = msf->imsf_multiaddr; 2533 imr.imr_address.s_addr = msf->imsf_interface; 2534 imr.imr_ifindex = 0; 2535 in_dev = ip_mc_find_dev(net, &imr); 2536 2537 if (!in_dev) { 2538 err = -ENODEV; 2539 goto done; 2540 } 2541 err = -EADDRNOTAVAIL; 2542 2543 for_each_pmc_rtnl(inet, pmc) { 2544 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr && 2545 pmc->multi.imr_ifindex == imr.imr_ifindex) 2546 break; 2547 } 2548 if (!pmc) /* must have a prior join */ 2549 goto done; 2550 msf->imsf_fmode = pmc->sfmode; 2551 psl = rtnl_dereference(pmc->sflist); 2552 if (!psl) { 2553 len = 0; 2554 count = 0; 2555 } else { 2556 count = psl->sl_count; 2557 } 2558 copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc; 2559 len = copycount * sizeof(psl->sl_addr[0]); 2560 msf->imsf_numsrc = count; 2561 if (put_user(IP_MSFILTER_SIZE(copycount), optlen) || 2562 copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) { 2563 return -EFAULT; 2564 } 2565 if (len && 2566 copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len)) 2567 return -EFAULT; 2568 return 0; 2569 done: 2570 return err; 2571 } 2572 2573 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf, 2574 struct group_filter __user *optval, int __user *optlen) 2575 { 2576 int err, i, count, copycount; 2577 struct sockaddr_in *psin; 2578 __be32 addr; 2579 struct ip_mc_socklist *pmc; 2580 struct inet_sock *inet = inet_sk(sk); 2581 struct ip_sf_socklist *psl; 2582 2583 ASSERT_RTNL(); 2584 2585 psin = (struct sockaddr_in *)&gsf->gf_group; 2586 if (psin->sin_family != AF_INET) 2587 return -EINVAL; 2588 addr = psin->sin_addr.s_addr; 2589 if (!ipv4_is_multicast(addr)) 2590 return -EINVAL; 2591 2592 err = -EADDRNOTAVAIL; 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 goto done; 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 if (put_user(GROUP_FILTER_SIZE(copycount), optlen) || 2607 copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) { 2608 return -EFAULT; 2609 } 2610 for (i = 0; i < copycount; i++) { 2611 struct sockaddr_storage ss; 2612 2613 psin = (struct sockaddr_in *)&ss; 2614 memset(&ss, 0, sizeof(ss)); 2615 psin->sin_family = AF_INET; 2616 psin->sin_addr.s_addr = psl->sl_addr[i]; 2617 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss))) 2618 return -EFAULT; 2619 } 2620 return 0; 2621 done: 2622 return err; 2623 } 2624 2625 /* 2626 * check if a multicast source filter allows delivery for a given <src,dst,intf> 2627 */ 2628 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, 2629 int dif, int sdif) 2630 { 2631 struct inet_sock *inet = inet_sk(sk); 2632 struct ip_mc_socklist *pmc; 2633 struct ip_sf_socklist *psl; 2634 int i; 2635 int ret; 2636 2637 ret = 1; 2638 if (!ipv4_is_multicast(loc_addr)) 2639 goto out; 2640 2641 rcu_read_lock(); 2642 for_each_pmc_rcu(inet, pmc) { 2643 if (pmc->multi.imr_multiaddr.s_addr == loc_addr && 2644 (pmc->multi.imr_ifindex == dif || 2645 (sdif && pmc->multi.imr_ifindex == sdif))) 2646 break; 2647 } 2648 ret = inet->mc_all; 2649 if (!pmc) 2650 goto unlock; 2651 psl = rcu_dereference(pmc->sflist); 2652 ret = (pmc->sfmode == MCAST_EXCLUDE); 2653 if (!psl) 2654 goto unlock; 2655 2656 for (i = 0; i < psl->sl_count; i++) { 2657 if (psl->sl_addr[i] == rmt_addr) 2658 break; 2659 } 2660 ret = 0; 2661 if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count) 2662 goto unlock; 2663 if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count) 2664 goto unlock; 2665 ret = 1; 2666 unlock: 2667 rcu_read_unlock(); 2668 out: 2669 return ret; 2670 } 2671 2672 /* 2673 * A socket is closing. 2674 */ 2675 2676 void ip_mc_drop_socket(struct sock *sk) 2677 { 2678 struct inet_sock *inet = inet_sk(sk); 2679 struct ip_mc_socklist *iml; 2680 struct net *net = sock_net(sk); 2681 2682 if (!inet->mc_list) 2683 return; 2684 2685 rtnl_lock(); 2686 while ((iml = rtnl_dereference(inet->mc_list)) != NULL) { 2687 struct in_device *in_dev; 2688 2689 inet->mc_list = iml->next_rcu; 2690 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex); 2691 (void) ip_mc_leave_src(sk, iml, in_dev); 2692 if (in_dev) 2693 ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr); 2694 /* decrease mem now to avoid the memleak warning */ 2695 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc); 2696 kfree_rcu(iml, rcu); 2697 } 2698 rtnl_unlock(); 2699 } 2700 2701 /* called with rcu_read_lock() */ 2702 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto) 2703 { 2704 struct ip_mc_list *im; 2705 struct ip_mc_list __rcu **mc_hash; 2706 struct ip_sf_list *psf; 2707 int rv = 0; 2708 2709 mc_hash = rcu_dereference(in_dev->mc_hash); 2710 if (mc_hash) { 2711 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG); 2712 2713 for (im = rcu_dereference(mc_hash[hash]); 2714 im != NULL; 2715 im = rcu_dereference(im->next_hash)) { 2716 if (im->multiaddr == mc_addr) 2717 break; 2718 } 2719 } else { 2720 for_each_pmc_rcu(in_dev, im) { 2721 if (im->multiaddr == mc_addr) 2722 break; 2723 } 2724 } 2725 if (im && proto == IPPROTO_IGMP) { 2726 rv = 1; 2727 } else if (im) { 2728 if (src_addr) { 2729 for (psf = im->sources; psf; psf = psf->sf_next) { 2730 if (psf->sf_inaddr == src_addr) 2731 break; 2732 } 2733 if (psf) 2734 rv = psf->sf_count[MCAST_INCLUDE] || 2735 psf->sf_count[MCAST_EXCLUDE] != 2736 im->sfcount[MCAST_EXCLUDE]; 2737 else 2738 rv = im->sfcount[MCAST_EXCLUDE] != 0; 2739 } else 2740 rv = 1; /* unspecified source; tentatively allow */ 2741 } 2742 return rv; 2743 } 2744 2745 #if defined(CONFIG_PROC_FS) 2746 struct igmp_mc_iter_state { 2747 struct seq_net_private p; 2748 struct net_device *dev; 2749 struct in_device *in_dev; 2750 }; 2751 2752 #define igmp_mc_seq_private(seq) ((struct igmp_mc_iter_state *)(seq)->private) 2753 2754 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq) 2755 { 2756 struct net *net = seq_file_net(seq); 2757 struct ip_mc_list *im = NULL; 2758 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq); 2759 2760 state->in_dev = NULL; 2761 for_each_netdev_rcu(net, state->dev) { 2762 struct in_device *in_dev; 2763 2764 in_dev = __in_dev_get_rcu(state->dev); 2765 if (!in_dev) 2766 continue; 2767 im = rcu_dereference(in_dev->mc_list); 2768 if (im) { 2769 state->in_dev = in_dev; 2770 break; 2771 } 2772 } 2773 return im; 2774 } 2775 2776 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im) 2777 { 2778 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq); 2779 2780 im = rcu_dereference(im->next_rcu); 2781 while (!im) { 2782 state->dev = next_net_device_rcu(state->dev); 2783 if (!state->dev) { 2784 state->in_dev = NULL; 2785 break; 2786 } 2787 state->in_dev = __in_dev_get_rcu(state->dev); 2788 if (!state->in_dev) 2789 continue; 2790 im = rcu_dereference(state->in_dev->mc_list); 2791 } 2792 return im; 2793 } 2794 2795 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos) 2796 { 2797 struct ip_mc_list *im = igmp_mc_get_first(seq); 2798 if (im) 2799 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL) 2800 --pos; 2801 return pos ? NULL : im; 2802 } 2803 2804 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos) 2805 __acquires(rcu) 2806 { 2807 rcu_read_lock(); 2808 return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 2809 } 2810 2811 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2812 { 2813 struct ip_mc_list *im; 2814 if (v == SEQ_START_TOKEN) 2815 im = igmp_mc_get_first(seq); 2816 else 2817 im = igmp_mc_get_next(seq, v); 2818 ++*pos; 2819 return im; 2820 } 2821 2822 static void igmp_mc_seq_stop(struct seq_file *seq, void *v) 2823 __releases(rcu) 2824 { 2825 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq); 2826 2827 state->in_dev = NULL; 2828 state->dev = NULL; 2829 rcu_read_unlock(); 2830 } 2831 2832 static int igmp_mc_seq_show(struct seq_file *seq, void *v) 2833 { 2834 if (v == SEQ_START_TOKEN) 2835 seq_puts(seq, 2836 "Idx\tDevice : Count Querier\tGroup Users Timer\tReporter\n"); 2837 else { 2838 struct ip_mc_list *im = (struct ip_mc_list *)v; 2839 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq); 2840 char *querier; 2841 long delta; 2842 2843 #ifdef CONFIG_IP_MULTICAST 2844 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" : 2845 IGMP_V2_SEEN(state->in_dev) ? "V2" : 2846 "V3"; 2847 #else 2848 querier = "NONE"; 2849 #endif 2850 2851 if (rcu_access_pointer(state->in_dev->mc_list) == im) { 2852 seq_printf(seq, "%d\t%-10s: %5d %7s\n", 2853 state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier); 2854 } 2855 2856 delta = im->timer.expires - jiffies; 2857 seq_printf(seq, 2858 "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n", 2859 im->multiaddr, im->users, 2860 im->tm_running, 2861 im->tm_running ? jiffies_delta_to_clock_t(delta) : 0, 2862 im->reporter); 2863 } 2864 return 0; 2865 } 2866 2867 static const struct seq_operations igmp_mc_seq_ops = { 2868 .start = igmp_mc_seq_start, 2869 .next = igmp_mc_seq_next, 2870 .stop = igmp_mc_seq_stop, 2871 .show = igmp_mc_seq_show, 2872 }; 2873 2874 struct igmp_mcf_iter_state { 2875 struct seq_net_private p; 2876 struct net_device *dev; 2877 struct in_device *idev; 2878 struct ip_mc_list *im; 2879 }; 2880 2881 #define igmp_mcf_seq_private(seq) ((struct igmp_mcf_iter_state *)(seq)->private) 2882 2883 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq) 2884 { 2885 struct net *net = seq_file_net(seq); 2886 struct ip_sf_list *psf = NULL; 2887 struct ip_mc_list *im = NULL; 2888 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq); 2889 2890 state->idev = NULL; 2891 state->im = NULL; 2892 for_each_netdev_rcu(net, state->dev) { 2893 struct in_device *idev; 2894 idev = __in_dev_get_rcu(state->dev); 2895 if (unlikely(!idev)) 2896 continue; 2897 im = rcu_dereference(idev->mc_list); 2898 if (likely(im)) { 2899 spin_lock_bh(&im->lock); 2900 psf = im->sources; 2901 if (likely(psf)) { 2902 state->im = im; 2903 state->idev = idev; 2904 break; 2905 } 2906 spin_unlock_bh(&im->lock); 2907 } 2908 } 2909 return psf; 2910 } 2911 2912 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf) 2913 { 2914 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq); 2915 2916 psf = psf->sf_next; 2917 while (!psf) { 2918 spin_unlock_bh(&state->im->lock); 2919 state->im = state->im->next; 2920 while (!state->im) { 2921 state->dev = next_net_device_rcu(state->dev); 2922 if (!state->dev) { 2923 state->idev = NULL; 2924 goto out; 2925 } 2926 state->idev = __in_dev_get_rcu(state->dev); 2927 if (!state->idev) 2928 continue; 2929 state->im = rcu_dereference(state->idev->mc_list); 2930 } 2931 if (!state->im) 2932 break; 2933 spin_lock_bh(&state->im->lock); 2934 psf = state->im->sources; 2935 } 2936 out: 2937 return psf; 2938 } 2939 2940 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos) 2941 { 2942 struct ip_sf_list *psf = igmp_mcf_get_first(seq); 2943 if (psf) 2944 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL) 2945 --pos; 2946 return pos ? NULL : psf; 2947 } 2948 2949 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos) 2950 __acquires(rcu) 2951 { 2952 rcu_read_lock(); 2953 return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 2954 } 2955 2956 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2957 { 2958 struct ip_sf_list *psf; 2959 if (v == SEQ_START_TOKEN) 2960 psf = igmp_mcf_get_first(seq); 2961 else 2962 psf = igmp_mcf_get_next(seq, v); 2963 ++*pos; 2964 return psf; 2965 } 2966 2967 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v) 2968 __releases(rcu) 2969 { 2970 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq); 2971 if (likely(state->im)) { 2972 spin_unlock_bh(&state->im->lock); 2973 state->im = NULL; 2974 } 2975 state->idev = NULL; 2976 state->dev = NULL; 2977 rcu_read_unlock(); 2978 } 2979 2980 static int igmp_mcf_seq_show(struct seq_file *seq, void *v) 2981 { 2982 struct ip_sf_list *psf = (struct ip_sf_list *)v; 2983 struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq); 2984 2985 if (v == SEQ_START_TOKEN) { 2986 seq_puts(seq, "Idx Device MCA SRC INC EXC\n"); 2987 } else { 2988 seq_printf(seq, 2989 "%3d %6.6s 0x%08x " 2990 "0x%08x %6lu %6lu\n", 2991 state->dev->ifindex, state->dev->name, 2992 ntohl(state->im->multiaddr), 2993 ntohl(psf->sf_inaddr), 2994 psf->sf_count[MCAST_INCLUDE], 2995 psf->sf_count[MCAST_EXCLUDE]); 2996 } 2997 return 0; 2998 } 2999 3000 static const struct seq_operations igmp_mcf_seq_ops = { 3001 .start = igmp_mcf_seq_start, 3002 .next = igmp_mcf_seq_next, 3003 .stop = igmp_mcf_seq_stop, 3004 .show = igmp_mcf_seq_show, 3005 }; 3006 3007 static int __net_init igmp_net_init(struct net *net) 3008 { 3009 struct proc_dir_entry *pde; 3010 int err; 3011 3012 pde = proc_create_net("igmp", 0444, net->proc_net, &igmp_mc_seq_ops, 3013 sizeof(struct igmp_mc_iter_state)); 3014 if (!pde) 3015 goto out_igmp; 3016 pde = proc_create_net("mcfilter", 0444, net->proc_net, 3017 &igmp_mcf_seq_ops, sizeof(struct igmp_mcf_iter_state)); 3018 if (!pde) 3019 goto out_mcfilter; 3020 err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET, 3021 SOCK_DGRAM, 0, net); 3022 if (err < 0) { 3023 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n", 3024 err); 3025 goto out_sock; 3026 } 3027 3028 return 0; 3029 3030 out_sock: 3031 remove_proc_entry("mcfilter", net->proc_net); 3032 out_mcfilter: 3033 remove_proc_entry("igmp", net->proc_net); 3034 out_igmp: 3035 return -ENOMEM; 3036 } 3037 3038 static void __net_exit igmp_net_exit(struct net *net) 3039 { 3040 remove_proc_entry("mcfilter", net->proc_net); 3041 remove_proc_entry("igmp", net->proc_net); 3042 inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk); 3043 } 3044 3045 static struct pernet_operations igmp_net_ops = { 3046 .init = igmp_net_init, 3047 .exit = igmp_net_exit, 3048 }; 3049 #endif 3050 3051 static int igmp_netdev_event(struct notifier_block *this, 3052 unsigned long event, void *ptr) 3053 { 3054 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 3055 struct in_device *in_dev; 3056 3057 switch (event) { 3058 case NETDEV_RESEND_IGMP: 3059 in_dev = __in_dev_get_rtnl(dev); 3060 if (in_dev) 3061 ip_mc_rejoin_groups(in_dev); 3062 break; 3063 default: 3064 break; 3065 } 3066 return NOTIFY_DONE; 3067 } 3068 3069 static struct notifier_block igmp_notifier = { 3070 .notifier_call = igmp_netdev_event, 3071 }; 3072 3073 int __init igmp_mc_init(void) 3074 { 3075 #if defined(CONFIG_PROC_FS) 3076 int err; 3077 3078 err = register_pernet_subsys(&igmp_net_ops); 3079 if (err) 3080 return err; 3081 err = register_netdevice_notifier(&igmp_notifier); 3082 if (err) 3083 goto reg_notif_fail; 3084 return 0; 3085 3086 reg_notif_fail: 3087 unregister_pernet_subsys(&igmp_net_ops); 3088 return err; 3089 #else 3090 return register_netdevice_notifier(&igmp_notifier); 3091 #endif 3092 } 3093