1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Multicast support for IPv6 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c 10 */ 11 12 /* Changes: 13 * 14 * yoshfuji : fix format of router-alert option 15 * YOSHIFUJI Hideaki @USAGI: 16 * Fixed source address for MLD message based on 17 * <draft-ietf-magma-mld-source-05.txt>. 18 * YOSHIFUJI Hideaki @USAGI: 19 * - Ignore Queries for invalid addresses. 20 * - MLD for link-local addresses. 21 * David L Stevens <dlstevens@us.ibm.com>: 22 * - MLDv2 support 23 */ 24 25 #include <linux/module.h> 26 #include <linux/errno.h> 27 #include <linux/types.h> 28 #include <linux/string.h> 29 #include <linux/socket.h> 30 #include <linux/sockios.h> 31 #include <linux/jiffies.h> 32 #include <linux/net.h> 33 #include <linux/in.h> 34 #include <linux/in6.h> 35 #include <linux/netdevice.h> 36 #include <linux/if_addr.h> 37 #include <linux/if_arp.h> 38 #include <linux/route.h> 39 #include <linux/rtnetlink.h> 40 #include <linux/init.h> 41 #include <linux/proc_fs.h> 42 #include <linux/seq_file.h> 43 #include <linux/slab.h> 44 #include <linux/pkt_sched.h> 45 #include <net/mld.h> 46 #include <linux/workqueue.h> 47 48 #include <linux/netfilter.h> 49 #include <linux/netfilter_ipv6.h> 50 51 #include <net/net_namespace.h> 52 #include <net/netlink.h> 53 #include <net/sock.h> 54 #include <net/snmp.h> 55 56 #include <net/ipv6.h> 57 #include <net/protocol.h> 58 #include <net/if_inet6.h> 59 #include <net/ndisc.h> 60 #include <net/addrconf.h> 61 #include <net/ip6_route.h> 62 #include <net/inet_common.h> 63 64 #include <net/ip6_checksum.h> 65 66 /* Ensure that we have struct in6_addr aligned on 32bit word. */ 67 static int __mld2_query_bugs[] __attribute__((__unused__)) = { 68 BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4), 69 BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4), 70 BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4) 71 }; 72 73 static struct workqueue_struct *mld_wq; 74 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT; 75 76 static void igmp6_join_group(struct ifmcaddr6 *ma); 77 static void igmp6_leave_group(struct ifmcaddr6 *ma); 78 static void mld_mca_work(struct work_struct *work); 79 80 static void mld_ifc_event(struct inet6_dev *idev); 81 static bool mld_in_v1_mode(const struct inet6_dev *idev); 82 static int sf_setstate(struct ifmcaddr6 *pmc); 83 static void sf_markstate(struct ifmcaddr6 *pmc); 84 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc); 85 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca, 86 int sfmode, int sfcount, const struct in6_addr *psfsrc, 87 int delta); 88 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca, 89 int sfmode, int sfcount, const struct in6_addr *psfsrc, 90 int delta); 91 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml, 92 struct inet6_dev *idev); 93 static int __ipv6_dev_mc_inc(struct net_device *dev, 94 const struct in6_addr *addr, unsigned int mode); 95 96 #define MLD_QRV_DEFAULT 2 97 /* RFC3810, 9.2. Query Interval */ 98 #define MLD_QI_DEFAULT (125 * HZ) 99 /* RFC3810, 9.3. Query Response Interval */ 100 #define MLD_QRI_DEFAULT (10 * HZ) 101 102 /* RFC3810, 8.1 Query Version Distinctions */ 103 #define MLD_V1_QUERY_LEN 24 104 #define MLD_V2_QUERY_LEN_MIN 28 105 106 #define IPV6_MLD_MAX_MSF 64 107 108 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF; 109 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT; 110 111 #define mc_assert_locked(idev) \ 112 lockdep_assert_held(&(idev)->mc_lock) 113 114 #define mc_dereference(e, idev) \ 115 rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock)) 116 117 #define sock_dereference(e, sk) \ 118 rcu_dereference_protected(e, lockdep_sock_is_held(sk)) 119 120 #define for_each_pmc_socklock(np, sk, pmc) \ 121 for (pmc = sock_dereference((np)->ipv6_mc_list, sk); \ 122 pmc; \ 123 pmc = sock_dereference(pmc->next, sk)) 124 125 #define for_each_pmc_rcu(np, pmc) \ 126 for (pmc = rcu_dereference((np)->ipv6_mc_list); \ 127 pmc; \ 128 pmc = rcu_dereference(pmc->next)) 129 130 #define for_each_psf_mclock(mc, psf) \ 131 for (psf = mc_dereference((mc)->mca_sources, mc->idev); \ 132 psf; \ 133 psf = mc_dereference(psf->sf_next, mc->idev)) 134 135 #define for_each_psf_rcu(mc, psf) \ 136 for (psf = rcu_dereference((mc)->mca_sources); \ 137 psf; \ 138 psf = rcu_dereference(psf->sf_next)) 139 140 #define for_each_psf_tomb(mc, psf) \ 141 for (psf = mc_dereference((mc)->mca_tomb, mc->idev); \ 142 psf; \ 143 psf = mc_dereference(psf->sf_next, mc->idev)) 144 145 #define for_each_mc_mclock(idev, mc) \ 146 for (mc = mc_dereference((idev)->mc_list, idev); \ 147 mc; \ 148 mc = mc_dereference(mc->next, idev)) 149 150 #define for_each_mc_rcu(idev, mc) \ 151 for (mc = rcu_dereference((idev)->mc_list); \ 152 mc; \ 153 mc = rcu_dereference(mc->next)) 154 155 #define for_each_mc_tomb(idev, mc) \ 156 for (mc = mc_dereference((idev)->mc_tomb, idev); \ 157 mc; \ 158 mc = mc_dereference(mc->next, idev)) 159 160 static int unsolicited_report_interval(struct inet6_dev *idev) 161 { 162 int iv; 163 164 if (mld_in_v1_mode(idev)) 165 iv = READ_ONCE(idev->cnf.mldv1_unsolicited_report_interval); 166 else 167 iv = READ_ONCE(idev->cnf.mldv2_unsolicited_report_interval); 168 169 return iv > 0 ? iv : 1; 170 } 171 172 static struct net_device *ip6_mc_find_dev(struct net *net, 173 const struct in6_addr *group, 174 int ifindex) 175 { 176 struct net_device *dev = NULL; 177 struct rt6_info *rt; 178 179 if (ifindex == 0) { 180 rcu_read_lock(); 181 rt = rt6_lookup(net, group, NULL, 0, NULL, 0); 182 if (rt) { 183 dev = dst_dev_rcu(&rt->dst); 184 dev_hold(dev); 185 ip6_rt_put(rt); 186 } 187 rcu_read_unlock(); 188 } else { 189 dev = dev_get_by_index(net, ifindex); 190 } 191 192 return dev; 193 } 194 195 /* 196 * socket join on multicast group 197 */ 198 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex, 199 const struct in6_addr *addr, unsigned int mode) 200 { 201 struct ipv6_pinfo *np = inet6_sk(sk); 202 struct ipv6_mc_socklist *mc_lst; 203 struct net *net = sock_net(sk); 204 struct net_device *dev = NULL; 205 int err; 206 207 if (!ipv6_addr_is_multicast(addr)) 208 return -EINVAL; 209 210 for_each_pmc_socklock(np, sk, mc_lst) { 211 if ((ifindex == 0 || mc_lst->ifindex == ifindex) && 212 ipv6_addr_equal(&mc_lst->addr, addr)) 213 return -EADDRINUSE; 214 } 215 216 mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL); 217 if (!mc_lst) 218 return -ENOMEM; 219 220 mc_lst->next = NULL; 221 mc_lst->addr = *addr; 222 223 dev = ip6_mc_find_dev(net, addr, ifindex); 224 if (!dev) { 225 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst)); 226 return -ENODEV; 227 } 228 229 mc_lst->ifindex = dev->ifindex; 230 mc_lst->sfmode = mode; 231 RCU_INIT_POINTER(mc_lst->sflist, NULL); 232 233 /* now add/increase the group membership on the device */ 234 err = __ipv6_dev_mc_inc(dev, addr, mode); 235 236 dev_put(dev); 237 238 if (err) { 239 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst)); 240 return err; 241 } 242 243 mc_lst->next = np->ipv6_mc_list; 244 rcu_assign_pointer(np->ipv6_mc_list, mc_lst); 245 246 return 0; 247 } 248 249 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr) 250 { 251 return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE); 252 } 253 EXPORT_SYMBOL(ipv6_sock_mc_join); 254 255 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex, 256 const struct in6_addr *addr, unsigned int mode) 257 { 258 return __ipv6_sock_mc_join(sk, ifindex, addr, mode); 259 } 260 261 /* 262 * socket leave on multicast group 263 */ 264 static void __ipv6_sock_mc_drop(struct sock *sk, struct ipv6_mc_socklist *mc_lst) 265 { 266 struct net *net = sock_net(sk); 267 struct net_device *dev; 268 269 dev = dev_get_by_index(net, mc_lst->ifindex); 270 if (dev) { 271 struct inet6_dev *idev = in6_dev_get(dev); 272 273 ip6_mc_leave_src(sk, mc_lst, idev); 274 275 if (idev) { 276 __ipv6_dev_mc_dec(idev, &mc_lst->addr); 277 in6_dev_put(idev); 278 } 279 280 dev_put(dev); 281 } else { 282 ip6_mc_leave_src(sk, mc_lst, NULL); 283 } 284 285 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc); 286 kfree_rcu(mc_lst, rcu); 287 } 288 289 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr) 290 { 291 struct ipv6_pinfo *np = inet6_sk(sk); 292 struct ipv6_mc_socklist __rcu **lnk; 293 struct ipv6_mc_socklist *mc_lst; 294 295 if (!ipv6_addr_is_multicast(addr)) 296 return -EINVAL; 297 298 for (lnk = &np->ipv6_mc_list; 299 (mc_lst = sock_dereference(*lnk, sk)) != NULL; 300 lnk = &mc_lst->next) { 301 if ((ifindex == 0 || mc_lst->ifindex == ifindex) && 302 ipv6_addr_equal(&mc_lst->addr, addr)) { 303 *lnk = mc_lst->next; 304 __ipv6_sock_mc_drop(sk, mc_lst); 305 return 0; 306 } 307 } 308 309 return -EADDRNOTAVAIL; 310 } 311 EXPORT_SYMBOL(ipv6_sock_mc_drop); 312 313 static struct inet6_dev *ip6_mc_find_idev(struct net *net, 314 const struct in6_addr *group, 315 int ifindex) 316 { 317 struct net_device *dev; 318 struct inet6_dev *idev; 319 320 dev = ip6_mc_find_dev(net, group, ifindex); 321 if (!dev) 322 return NULL; 323 324 idev = in6_dev_get(dev); 325 dev_put(dev); 326 327 return idev; 328 } 329 330 void __ipv6_sock_mc_close(struct sock *sk) 331 { 332 struct ipv6_pinfo *np = inet6_sk(sk); 333 struct ipv6_mc_socklist *mc_lst; 334 335 while ((mc_lst = sock_dereference(np->ipv6_mc_list, sk)) != NULL) { 336 np->ipv6_mc_list = mc_lst->next; 337 __ipv6_sock_mc_drop(sk, mc_lst); 338 } 339 } 340 341 void ipv6_sock_mc_close(struct sock *sk) 342 { 343 struct ipv6_pinfo *np = inet6_sk(sk); 344 345 if (!rcu_access_pointer(np->ipv6_mc_list)) 346 return; 347 348 lock_sock(sk); 349 __ipv6_sock_mc_close(sk); 350 release_sock(sk); 351 } 352 353 int ip6_mc_source(int add, int omode, struct sock *sk, 354 struct group_source_req *pgsr) 355 { 356 struct ipv6_pinfo *inet6 = inet6_sk(sk); 357 struct in6_addr *source, *group; 358 struct net *net = sock_net(sk); 359 struct ipv6_mc_socklist *pmc; 360 struct ip6_sf_socklist *psl; 361 struct inet6_dev *idev; 362 int leavegroup = 0; 363 int i, j, rv; 364 int err; 365 366 source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr; 367 group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr; 368 369 if (!ipv6_addr_is_multicast(group)) 370 return -EINVAL; 371 372 idev = ip6_mc_find_idev(net, group, pgsr->gsr_interface); 373 if (!idev) 374 return -ENODEV; 375 376 mutex_lock(&idev->mc_lock); 377 378 if (idev->dead) { 379 err = -ENODEV; 380 goto done; 381 } 382 383 err = -EADDRNOTAVAIL; 384 385 for_each_pmc_socklock(inet6, sk, pmc) { 386 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface) 387 continue; 388 if (ipv6_addr_equal(&pmc->addr, group)) 389 break; 390 } 391 if (!pmc) { /* must have a prior join */ 392 err = -EINVAL; 393 goto done; 394 } 395 /* if a source filter was set, must be the same mode as before */ 396 if (rcu_access_pointer(pmc->sflist)) { 397 if (pmc->sfmode != omode) { 398 err = -EINVAL; 399 goto done; 400 } 401 } else if (pmc->sfmode != omode) { 402 /* allow mode switches for empty-set filters */ 403 ip6_mc_add_src(idev, group, omode, 0, NULL, 0); 404 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0); 405 pmc->sfmode = omode; 406 } 407 408 psl = sock_dereference(pmc->sflist, sk); 409 if (!add) { 410 if (!psl) 411 goto done; /* err = -EADDRNOTAVAIL */ 412 rv = !0; 413 for (i = 0; i < psl->sl_count; i++) { 414 rv = !ipv6_addr_equal(&psl->sl_addr[i], source); 415 if (rv == 0) 416 break; 417 } 418 if (rv) /* source not found */ 419 goto done; /* err = -EADDRNOTAVAIL */ 420 421 /* special case - (INCLUDE, empty) == LEAVE_GROUP */ 422 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) { 423 leavegroup = 1; 424 goto done; 425 } 426 427 /* update the interface filter */ 428 ip6_mc_del_src(idev, group, omode, 1, source, 1); 429 430 for (j = i+1; j < psl->sl_count; j++) 431 psl->sl_addr[j-1] = psl->sl_addr[j]; 432 psl->sl_count--; 433 err = 0; 434 goto done; 435 } 436 /* else, add a new source to the filter */ 437 438 if (psl && psl->sl_count >= sysctl_mld_max_msf) { 439 err = -ENOBUFS; 440 goto done; 441 } 442 if (!psl || psl->sl_count == psl->sl_max) { 443 struct ip6_sf_socklist *newpsl; 444 int count = IP6_SFBLOCK; 445 446 if (psl) 447 count += psl->sl_max; 448 newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, count), 449 GFP_KERNEL); 450 if (!newpsl) { 451 err = -ENOBUFS; 452 goto done; 453 } 454 newpsl->sl_max = count; 455 newpsl->sl_count = count - IP6_SFBLOCK; 456 if (psl) { 457 for (i = 0; i < psl->sl_count; i++) 458 newpsl->sl_addr[i] = psl->sl_addr[i]; 459 atomic_sub(struct_size(psl, sl_addr, psl->sl_max), 460 &sk->sk_omem_alloc); 461 } 462 rcu_assign_pointer(pmc->sflist, newpsl); 463 kfree_rcu(psl, rcu); 464 psl = newpsl; 465 } 466 rv = 1; /* > 0 for insert logic below if sl_count is 0 */ 467 for (i = 0; i < psl->sl_count; i++) { 468 rv = !ipv6_addr_equal(&psl->sl_addr[i], source); 469 if (rv == 0) /* There is an error in the address. */ 470 goto done; 471 } 472 for (j = psl->sl_count-1; j >= i; j--) 473 psl->sl_addr[j+1] = psl->sl_addr[j]; 474 psl->sl_addr[i] = *source; 475 psl->sl_count++; 476 err = 0; 477 /* update the interface list */ 478 ip6_mc_add_src(idev, group, omode, 1, source, 1); 479 done: 480 mutex_unlock(&idev->mc_lock); 481 in6_dev_put(idev); 482 if (leavegroup) 483 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group); 484 return err; 485 } 486 487 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf, 488 struct sockaddr_storage *list) 489 { 490 struct ipv6_pinfo *inet6 = inet6_sk(sk); 491 struct ip6_sf_socklist *newpsl, *psl; 492 struct net *net = sock_net(sk); 493 const struct in6_addr *group; 494 struct ipv6_mc_socklist *pmc; 495 struct inet6_dev *idev; 496 int leavegroup = 0; 497 int i, err; 498 499 group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr; 500 501 if (!ipv6_addr_is_multicast(group)) 502 return -EINVAL; 503 if (gsf->gf_fmode != MCAST_INCLUDE && 504 gsf->gf_fmode != MCAST_EXCLUDE) 505 return -EINVAL; 506 507 idev = ip6_mc_find_idev(net, group, gsf->gf_interface); 508 if (!idev) 509 return -ENODEV; 510 511 mutex_lock(&idev->mc_lock); 512 513 if (idev->dead) { 514 err = -ENODEV; 515 goto done; 516 } 517 518 err = 0; 519 520 if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) { 521 leavegroup = 1; 522 goto done; 523 } 524 525 for_each_pmc_socklock(inet6, sk, pmc) { 526 if (pmc->ifindex != gsf->gf_interface) 527 continue; 528 if (ipv6_addr_equal(&pmc->addr, group)) 529 break; 530 } 531 if (!pmc) { /* must have a prior join */ 532 err = -EINVAL; 533 goto done; 534 } 535 if (gsf->gf_numsrc) { 536 newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, 537 gsf->gf_numsrc), 538 GFP_KERNEL); 539 if (!newpsl) { 540 err = -ENOBUFS; 541 goto done; 542 } 543 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc; 544 for (i = 0; i < newpsl->sl_count; ++i, ++list) { 545 struct sockaddr_in6 *psin6; 546 547 psin6 = (struct sockaddr_in6 *)list; 548 newpsl->sl_addr[i] = psin6->sin6_addr; 549 } 550 551 err = ip6_mc_add_src(idev, group, gsf->gf_fmode, 552 newpsl->sl_count, newpsl->sl_addr, 0); 553 if (err) { 554 sock_kfree_s(sk, newpsl, struct_size(newpsl, sl_addr, 555 newpsl->sl_max)); 556 goto done; 557 } 558 } else { 559 newpsl = NULL; 560 ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0); 561 } 562 563 psl = sock_dereference(pmc->sflist, sk); 564 if (psl) { 565 ip6_mc_del_src(idev, group, pmc->sfmode, 566 psl->sl_count, psl->sl_addr, 0); 567 atomic_sub(struct_size(psl, sl_addr, psl->sl_max), 568 &sk->sk_omem_alloc); 569 } else { 570 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0); 571 } 572 573 rcu_assign_pointer(pmc->sflist, newpsl); 574 kfree_rcu(psl, rcu); 575 pmc->sfmode = gsf->gf_fmode; 576 err = 0; 577 done: 578 mutex_unlock(&idev->mc_lock); 579 in6_dev_put(idev); 580 if (leavegroup) 581 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group); 582 return err; 583 } 584 585 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf, 586 sockptr_t optval, size_t ss_offset) 587 { 588 struct ipv6_pinfo *inet6 = inet6_sk(sk); 589 const struct in6_addr *group; 590 struct ipv6_mc_socklist *pmc; 591 struct ip6_sf_socklist *psl; 592 unsigned int count; 593 int i, copycount; 594 595 group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr; 596 597 if (!ipv6_addr_is_multicast(group)) 598 return -EINVAL; 599 600 for_each_pmc_socklock(inet6, sk, pmc) { 601 if (pmc->ifindex != gsf->gf_interface) 602 continue; 603 if (ipv6_addr_equal(group, &pmc->addr)) 604 break; 605 } 606 if (!pmc) /* must have a prior join */ 607 return -EADDRNOTAVAIL; 608 609 gsf->gf_fmode = pmc->sfmode; 610 psl = sock_dereference(pmc->sflist, sk); 611 count = psl ? psl->sl_count : 0; 612 613 copycount = min(count, gsf->gf_numsrc); 614 gsf->gf_numsrc = count; 615 for (i = 0; i < copycount; i++) { 616 struct sockaddr_in6 *psin6; 617 struct sockaddr_storage ss; 618 619 psin6 = (struct sockaddr_in6 *)&ss; 620 memset(&ss, 0, sizeof(ss)); 621 psin6->sin6_family = AF_INET6; 622 psin6->sin6_addr = psl->sl_addr[i]; 623 if (copy_to_sockptr_offset(optval, ss_offset, &ss, sizeof(ss))) 624 return -EFAULT; 625 ss_offset += sizeof(ss); 626 } 627 return 0; 628 } 629 630 bool inet6_mc_check(const struct sock *sk, const struct in6_addr *mc_addr, 631 const struct in6_addr *src_addr) 632 { 633 const struct ipv6_pinfo *np = inet6_sk(sk); 634 const struct ipv6_mc_socklist *mc; 635 const struct ip6_sf_socklist *psl; 636 bool rv = true; 637 638 rcu_read_lock(); 639 for_each_pmc_rcu(np, mc) { 640 if (ipv6_addr_equal(&mc->addr, mc_addr)) 641 break; 642 } 643 if (!mc) { 644 rcu_read_unlock(); 645 return inet6_test_bit(MC6_ALL, sk); 646 } 647 psl = rcu_dereference(mc->sflist); 648 if (!psl) { 649 rv = mc->sfmode == MCAST_EXCLUDE; 650 } else { 651 int i; 652 653 for (i = 0; i < psl->sl_count; i++) { 654 if (ipv6_addr_equal(&psl->sl_addr[i], src_addr)) 655 break; 656 } 657 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count) 658 rv = false; 659 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count) 660 rv = false; 661 } 662 rcu_read_unlock(); 663 664 return rv; 665 } 666 667 static void igmp6_group_added(struct ifmcaddr6 *mc) 668 { 669 struct net_device *dev = mc->idev->dev; 670 char buf[MAX_ADDR_LEN]; 671 672 mc_assert_locked(mc->idev); 673 674 if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < 675 IPV6_ADDR_SCOPE_LINKLOCAL) 676 return; 677 678 if (!(mc->mca_flags&MAF_LOADED)) { 679 mc->mca_flags |= MAF_LOADED; 680 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0) 681 dev_mc_add(dev, buf); 682 } 683 684 if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT)) 685 return; 686 687 if (mld_in_v1_mode(mc->idev)) { 688 igmp6_join_group(mc); 689 return; 690 } 691 /* else v2 */ 692 693 /* Based on RFC3810 6.1, for newly added INCLUDE SSM, we 694 * should not send filter-mode change record as the mode 695 * should be from IN() to IN(A). 696 */ 697 if (mc->mca_sfmode == MCAST_EXCLUDE) 698 mc->mca_crcount = mc->idev->mc_qrv; 699 700 mld_ifc_event(mc->idev); 701 } 702 703 static void igmp6_group_dropped(struct ifmcaddr6 *mc) 704 { 705 struct net_device *dev = mc->idev->dev; 706 char buf[MAX_ADDR_LEN]; 707 708 mc_assert_locked(mc->idev); 709 710 if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < 711 IPV6_ADDR_SCOPE_LINKLOCAL) 712 return; 713 714 if (mc->mca_flags&MAF_LOADED) { 715 mc->mca_flags &= ~MAF_LOADED; 716 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0) 717 dev_mc_del(dev, buf); 718 } 719 720 if (mc->mca_flags & MAF_NOREPORT) 721 return; 722 723 if (!mc->idev->dead) 724 igmp6_leave_group(mc); 725 726 if (cancel_delayed_work(&mc->mca_work)) 727 refcount_dec(&mc->mca_refcnt); 728 } 729 730 /* deleted ifmcaddr6 manipulation */ 731 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im) 732 { 733 struct ifmcaddr6 *pmc; 734 735 mc_assert_locked(idev); 736 737 /* this is an "ifmcaddr6" for convenience; only the fields below 738 * are actually used. In particular, the refcnt and users are not 739 * used for management of the delete list. Using the same structure 740 * for deleted items allows change reports to use common code with 741 * non-deleted or query-response MCA's. 742 */ 743 pmc = kzalloc_obj(*pmc); 744 if (!pmc) 745 return; 746 747 pmc->idev = im->idev; 748 in6_dev_hold(idev); 749 pmc->mca_addr = im->mca_addr; 750 pmc->mca_crcount = idev->mc_qrv; 751 pmc->mca_sfmode = im->mca_sfmode; 752 if (pmc->mca_sfmode == MCAST_INCLUDE) { 753 struct ip6_sf_list *psf; 754 755 rcu_assign_pointer(pmc->mca_tomb, 756 mc_dereference(im->mca_tomb, idev)); 757 rcu_assign_pointer(pmc->mca_sources, 758 mc_dereference(im->mca_sources, idev)); 759 RCU_INIT_POINTER(im->mca_tomb, NULL); 760 RCU_INIT_POINTER(im->mca_sources, NULL); 761 762 for_each_psf_mclock(pmc, psf) 763 psf->sf_crcount = pmc->mca_crcount; 764 } 765 766 rcu_assign_pointer(pmc->next, idev->mc_tomb); 767 rcu_assign_pointer(idev->mc_tomb, pmc); 768 } 769 770 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im) 771 { 772 struct ip6_sf_list *psf, *sources, *tomb; 773 struct in6_addr *pmca = &im->mca_addr; 774 struct ifmcaddr6 *pmc, *pmc_prev; 775 776 mc_assert_locked(idev); 777 778 pmc_prev = NULL; 779 for_each_mc_tomb(idev, pmc) { 780 if (ipv6_addr_equal(&pmc->mca_addr, pmca)) 781 break; 782 pmc_prev = pmc; 783 } 784 if (!pmc) 785 return; 786 if (pmc_prev) 787 rcu_assign_pointer(pmc_prev->next, pmc->next); 788 else 789 rcu_assign_pointer(idev->mc_tomb, pmc->next); 790 791 im->idev = pmc->idev; 792 if (im->mca_sfmode == MCAST_INCLUDE) { 793 tomb = rcu_replace_pointer(im->mca_tomb, 794 mc_dereference(pmc->mca_tomb, pmc->idev), 795 lockdep_is_held(&im->idev->mc_lock)); 796 rcu_assign_pointer(pmc->mca_tomb, tomb); 797 798 sources = rcu_replace_pointer(im->mca_sources, 799 mc_dereference(pmc->mca_sources, pmc->idev), 800 lockdep_is_held(&im->idev->mc_lock)); 801 rcu_assign_pointer(pmc->mca_sources, sources); 802 for_each_psf_mclock(im, psf) 803 psf->sf_crcount = idev->mc_qrv; 804 } else { 805 im->mca_crcount = idev->mc_qrv; 806 } 807 ip6_mc_clear_src(pmc); 808 in6_dev_put(pmc->idev); 809 kfree_rcu(pmc, rcu); 810 } 811 812 static void mld_clear_delrec(struct inet6_dev *idev) 813 { 814 struct ifmcaddr6 *pmc, *nextpmc; 815 816 mc_assert_locked(idev); 817 818 pmc = mc_dereference(idev->mc_tomb, idev); 819 RCU_INIT_POINTER(idev->mc_tomb, NULL); 820 821 for (; pmc; pmc = nextpmc) { 822 nextpmc = mc_dereference(pmc->next, idev); 823 ip6_mc_clear_src(pmc); 824 in6_dev_put(pmc->idev); 825 kfree_rcu(pmc, rcu); 826 } 827 828 /* clear dead sources, too */ 829 for_each_mc_mclock(idev, pmc) { 830 struct ip6_sf_list *psf, *psf_next; 831 832 psf = mc_dereference(pmc->mca_tomb, idev); 833 RCU_INIT_POINTER(pmc->mca_tomb, NULL); 834 for (; psf; psf = psf_next) { 835 psf_next = mc_dereference(psf->sf_next, idev); 836 kfree_rcu(psf, rcu); 837 } 838 } 839 } 840 841 static void mld_clear_query(struct inet6_dev *idev) 842 { 843 spin_lock_bh(&idev->mc_query_lock); 844 __skb_queue_purge(&idev->mc_query_queue); 845 spin_unlock_bh(&idev->mc_query_lock); 846 } 847 848 static void mld_clear_report(struct inet6_dev *idev) 849 { 850 spin_lock_bh(&idev->mc_report_lock); 851 __skb_queue_purge(&idev->mc_report_queue); 852 spin_unlock_bh(&idev->mc_report_lock); 853 } 854 855 static void ma_put(struct ifmcaddr6 *mc) 856 { 857 if (refcount_dec_and_test(&mc->mca_refcnt)) { 858 in6_dev_put(mc->idev); 859 kfree_rcu(mc, rcu); 860 } 861 } 862 863 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev, 864 const struct in6_addr *addr, 865 unsigned int mode) 866 { 867 struct ifmcaddr6 *mc; 868 869 mc_assert_locked(idev); 870 871 mc = kzalloc_obj(*mc); 872 if (!mc) 873 return NULL; 874 875 INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work); 876 877 mc->mca_addr = *addr; 878 mc->idev = idev; /* reference taken by caller */ 879 mc->mca_users = 1; 880 /* mca_stamp should be updated upon changes */ 881 mc->mca_cstamp = mc->mca_tstamp = jiffies; 882 refcount_set(&mc->mca_refcnt, 1); 883 884 mc->mca_sfmode = mode; 885 mc->mca_sfcount[mode] = 1; 886 887 if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) || 888 IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL) 889 mc->mca_flags |= MAF_NOREPORT; 890 891 return mc; 892 } 893 894 static void inet6_ifmcaddr_notify(struct net_device *dev, 895 const struct ifmcaddr6 *ifmca, int event) 896 { 897 struct inet6_fill_args fillargs = { 898 .portid = 0, 899 .seq = 0, 900 .event = event, 901 .flags = 0, 902 .netnsid = -1, 903 .force_rt_scope_universe = true, 904 }; 905 struct net *net = dev_net(dev); 906 struct sk_buff *skb; 907 int err = -ENOMEM; 908 909 skb = nlmsg_new(NLMSG_ALIGN(sizeof(struct ifaddrmsg)) + 910 nla_total_size(sizeof(struct in6_addr)) + 911 nla_total_size(sizeof(struct ifa_cacheinfo)), 912 GFP_KERNEL); 913 if (!skb) 914 goto error; 915 916 err = inet6_fill_ifmcaddr(skb, ifmca, &fillargs); 917 if (err < 0) { 918 WARN_ON_ONCE(err == -EMSGSIZE); 919 nlmsg_free(skb); 920 goto error; 921 } 922 923 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MCADDR, NULL, GFP_KERNEL); 924 return; 925 error: 926 rtnl_set_sk_err(net, RTNLGRP_IPV6_MCADDR, err); 927 } 928 929 /* 930 * device multicast group inc (add if not found) 931 */ 932 static int __ipv6_dev_mc_inc(struct net_device *dev, 933 const struct in6_addr *addr, unsigned int mode) 934 { 935 struct inet6_dev *idev; 936 struct ifmcaddr6 *mc; 937 938 /* we need to take a reference on idev */ 939 idev = in6_dev_get(dev); 940 if (!idev) 941 return -EINVAL; 942 943 mutex_lock(&idev->mc_lock); 944 945 if (READ_ONCE(idev->dead)) { 946 mutex_unlock(&idev->mc_lock); 947 in6_dev_put(idev); 948 return -ENODEV; 949 } 950 951 for_each_mc_mclock(idev, mc) { 952 if (ipv6_addr_equal(&mc->mca_addr, addr)) { 953 mc->mca_users++; 954 ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0); 955 mutex_unlock(&idev->mc_lock); 956 in6_dev_put(idev); 957 return 0; 958 } 959 } 960 961 mc = mca_alloc(idev, addr, mode); 962 if (!mc) { 963 mutex_unlock(&idev->mc_lock); 964 in6_dev_put(idev); 965 return -ENOMEM; 966 } 967 968 rcu_assign_pointer(mc->next, idev->mc_list); 969 rcu_assign_pointer(idev->mc_list, mc); 970 971 mld_del_delrec(idev, mc); 972 igmp6_group_added(mc); 973 inet6_ifmcaddr_notify(dev, mc, RTM_NEWMULTICAST); 974 mutex_unlock(&idev->mc_lock); 975 976 return 0; 977 } 978 979 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr) 980 { 981 return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE); 982 } 983 EXPORT_SYMBOL(ipv6_dev_mc_inc); 984 985 /* 986 * device multicast group del 987 */ 988 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr) 989 { 990 struct ifmcaddr6 *ma, __rcu **map; 991 992 mutex_lock(&idev->mc_lock); 993 994 for (map = &idev->mc_list; 995 (ma = mc_dereference(*map, idev)); 996 map = &ma->next) { 997 if (ipv6_addr_equal(&ma->mca_addr, addr)) { 998 if (--ma->mca_users == 0) { 999 *map = ma->next; 1000 1001 igmp6_group_dropped(ma); 1002 inet6_ifmcaddr_notify(idev->dev, ma, 1003 RTM_DELMULTICAST); 1004 ip6_mc_clear_src(ma); 1005 mutex_unlock(&idev->mc_lock); 1006 1007 ma_put(ma); 1008 return 0; 1009 } 1010 mutex_unlock(&idev->mc_lock); 1011 return 0; 1012 } 1013 } 1014 1015 mutex_unlock(&idev->mc_lock); 1016 return -ENOENT; 1017 } 1018 1019 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr) 1020 { 1021 struct inet6_dev *idev; 1022 int err; 1023 1024 idev = in6_dev_get(dev); 1025 if (!idev) 1026 return -ENODEV; 1027 1028 err = __ipv6_dev_mc_dec(idev, addr); 1029 in6_dev_put(idev); 1030 1031 return err; 1032 } 1033 EXPORT_SYMBOL(ipv6_dev_mc_dec); 1034 1035 /* 1036 * check if the interface/address pair is valid 1037 */ 1038 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group, 1039 const struct in6_addr *src_addr) 1040 { 1041 struct inet6_dev *idev; 1042 struct ifmcaddr6 *mc; 1043 bool rv = false; 1044 1045 rcu_read_lock(); 1046 idev = __in6_dev_get(dev); 1047 if (!idev) 1048 goto unlock; 1049 for_each_mc_rcu(idev, mc) { 1050 if (ipv6_addr_equal(&mc->mca_addr, group)) 1051 break; 1052 } 1053 if (!mc) 1054 goto unlock; 1055 if (src_addr && !ipv6_addr_any(src_addr)) { 1056 struct ip6_sf_list *psf; 1057 1058 for_each_psf_rcu(mc, psf) { 1059 if (ipv6_addr_equal(&psf->sf_addr, src_addr)) 1060 break; 1061 } 1062 if (psf) 1063 rv = READ_ONCE(psf->sf_count[MCAST_INCLUDE]) || 1064 READ_ONCE(psf->sf_count[MCAST_EXCLUDE]) != 1065 READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]); 1066 else 1067 rv = READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]) != 0; 1068 } else { 1069 rv = true; /* don't filter unspecified source */ 1070 } 1071 unlock: 1072 rcu_read_unlock(); 1073 return rv; 1074 } 1075 1076 static void mld_gq_start_work(struct inet6_dev *idev) 1077 { 1078 unsigned long tv = get_random_u32_below(idev->mc_maxdelay); 1079 1080 mc_assert_locked(idev); 1081 1082 idev->mc_gq_running = 1; 1083 if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2)) 1084 in6_dev_hold(idev); 1085 } 1086 1087 static void mld_gq_stop_work(struct inet6_dev *idev) 1088 { 1089 mc_assert_locked(idev); 1090 1091 idev->mc_gq_running = 0; 1092 if (cancel_delayed_work(&idev->mc_gq_work)) 1093 __in6_dev_put(idev); 1094 } 1095 1096 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay) 1097 { 1098 unsigned long tv = get_random_u32_below(delay); 1099 1100 mc_assert_locked(idev); 1101 1102 if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2)) 1103 in6_dev_hold(idev); 1104 } 1105 1106 static void mld_ifc_stop_work(struct inet6_dev *idev) 1107 { 1108 mc_assert_locked(idev); 1109 1110 idev->mc_ifc_count = 0; 1111 if (cancel_delayed_work(&idev->mc_ifc_work)) 1112 __in6_dev_put(idev); 1113 } 1114 1115 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay) 1116 { 1117 unsigned long tv = get_random_u32_below(delay); 1118 1119 mc_assert_locked(idev); 1120 1121 if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2)) 1122 in6_dev_hold(idev); 1123 } 1124 1125 static void mld_dad_stop_work(struct inet6_dev *idev) 1126 { 1127 if (cancel_delayed_work(&idev->mc_dad_work)) 1128 __in6_dev_put(idev); 1129 } 1130 1131 static void mld_query_stop_work(struct inet6_dev *idev) 1132 { 1133 spin_lock_bh(&idev->mc_query_lock); 1134 if (cancel_delayed_work(&idev->mc_query_work)) 1135 __in6_dev_put(idev); 1136 spin_unlock_bh(&idev->mc_query_lock); 1137 } 1138 1139 static void mld_report_stop_work(struct inet6_dev *idev) 1140 { 1141 if (cancel_delayed_work_sync(&idev->mc_report_work)) 1142 __in6_dev_put(idev); 1143 } 1144 1145 /* IGMP handling (alias multicast ICMPv6 messages) */ 1146 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime) 1147 { 1148 unsigned long delay = resptime; 1149 1150 mc_assert_locked(ma->idev); 1151 1152 /* Do not start work for these addresses */ 1153 if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) || 1154 IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL) 1155 return; 1156 1157 if (cancel_delayed_work(&ma->mca_work)) { 1158 refcount_dec(&ma->mca_refcnt); 1159 delay = ma->mca_work.timer.expires - jiffies; 1160 } 1161 1162 if (delay >= resptime) 1163 delay = get_random_u32_below(resptime); 1164 1165 if (!mod_delayed_work(mld_wq, &ma->mca_work, delay)) 1166 refcount_inc(&ma->mca_refcnt); 1167 ma->mca_flags |= MAF_TIMER_RUNNING; 1168 } 1169 1170 /* mark EXCLUDE-mode sources */ 1171 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs, 1172 const struct in6_addr *srcs) 1173 { 1174 struct ip6_sf_list *psf; 1175 int i, scount; 1176 1177 mc_assert_locked(pmc->idev); 1178 1179 scount = 0; 1180 for_each_psf_mclock(pmc, psf) { 1181 if (scount == nsrcs) 1182 break; 1183 for (i = 0; i < nsrcs; i++) { 1184 /* skip inactive filters */ 1185 if (psf->sf_count[MCAST_INCLUDE] || 1186 pmc->mca_sfcount[MCAST_EXCLUDE] != 1187 psf->sf_count[MCAST_EXCLUDE]) 1188 break; 1189 if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) { 1190 scount++; 1191 break; 1192 } 1193 } 1194 } 1195 pmc->mca_flags &= ~MAF_GSQUERY; 1196 if (scount == nsrcs) /* all sources excluded */ 1197 return false; 1198 return true; 1199 } 1200 1201 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs, 1202 const struct in6_addr *srcs) 1203 { 1204 struct ip6_sf_list *psf; 1205 int i, scount; 1206 1207 mc_assert_locked(pmc->idev); 1208 1209 if (pmc->mca_sfmode == MCAST_EXCLUDE) 1210 return mld_xmarksources(pmc, nsrcs, srcs); 1211 1212 /* mark INCLUDE-mode sources */ 1213 1214 scount = 0; 1215 for_each_psf_mclock(pmc, psf) { 1216 if (scount == nsrcs) 1217 break; 1218 for (i = 0; i < nsrcs; i++) { 1219 if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) { 1220 psf->sf_gsresp = 1; 1221 scount++; 1222 break; 1223 } 1224 } 1225 } 1226 if (!scount) { 1227 pmc->mca_flags &= ~MAF_GSQUERY; 1228 return false; 1229 } 1230 pmc->mca_flags |= MAF_GSQUERY; 1231 return true; 1232 } 1233 1234 static int mld_force_mld_version(const struct inet6_dev *idev) 1235 { 1236 const struct net *net = dev_net(idev->dev); 1237 int all_force; 1238 1239 all_force = READ_ONCE(net->ipv6.devconf_all->force_mld_version); 1240 /* Normally, both are 0 here. If enforcement to a particular is 1241 * being used, individual device enforcement will have a lower 1242 * precedence over 'all' device (.../conf/all/force_mld_version). 1243 */ 1244 return all_force ?: READ_ONCE(idev->cnf.force_mld_version); 1245 } 1246 1247 static bool mld_in_v2_mode_only(const struct inet6_dev *idev) 1248 { 1249 return mld_force_mld_version(idev) == 2; 1250 } 1251 1252 static bool mld_in_v1_mode_only(const struct inet6_dev *idev) 1253 { 1254 return mld_force_mld_version(idev) == 1; 1255 } 1256 1257 static bool mld_in_v1_mode(const struct inet6_dev *idev) 1258 { 1259 if (mld_in_v2_mode_only(idev)) 1260 return false; 1261 if (mld_in_v1_mode_only(idev)) 1262 return true; 1263 if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen)) 1264 return true; 1265 1266 return false; 1267 } 1268 1269 static void mld_set_v1_mode(struct inet6_dev *idev) 1270 { 1271 /* RFC3810, relevant sections: 1272 * - 9.1. Robustness Variable 1273 * - 9.2. Query Interval 1274 * - 9.3. Query Response Interval 1275 * - 9.12. Older Version Querier Present Timeout 1276 */ 1277 unsigned long switchback; 1278 1279 switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri; 1280 1281 idev->mc_v1_seen = jiffies + switchback; 1282 } 1283 1284 static void mld_update_qrv(struct inet6_dev *idev, 1285 const struct mld2_query *mlh2) 1286 { 1287 /* RFC3810, relevant sections: 1288 * - 5.1.8. QRV (Querier's Robustness Variable) 1289 * - 9.1. Robustness Variable 1290 */ 1291 1292 /* The value of the Robustness Variable MUST NOT be zero, 1293 * and SHOULD NOT be one. Catch this here if we ever run 1294 * into such a case in future. 1295 */ 1296 const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv); 1297 WARN_ON(idev->mc_qrv == 0); 1298 1299 if (mlh2->mld2q_qrv > 0) 1300 idev->mc_qrv = mlh2->mld2q_qrv; 1301 1302 if (unlikely(idev->mc_qrv < min_qrv)) { 1303 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n", 1304 idev->mc_qrv, min_qrv); 1305 idev->mc_qrv = min_qrv; 1306 } 1307 } 1308 1309 static void mld_update_qi(struct inet6_dev *idev, 1310 const struct mld2_query *mlh2) 1311 { 1312 /* RFC3810, relevant sections: 1313 * - 5.1.9. QQIC (Querier's Query Interval Code) 1314 * - 9.2. Query Interval 1315 * - 9.12. Older Version Querier Present Timeout 1316 * (the [Query Interval] in the last Query received) 1317 */ 1318 idev->mc_qi = mldv2_qqi(mlh2) * HZ; 1319 } 1320 1321 static void mld_update_qri(struct inet6_dev *idev, 1322 const struct mld2_query *mlh2) 1323 { 1324 /* RFC3810, relevant sections: 1325 * - 5.1.3. Maximum Response Code 1326 * - 9.3. Query Response Interval 1327 */ 1328 idev->mc_qri = msecs_to_jiffies(mldv2_mrd(mlh2)); 1329 } 1330 1331 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld, 1332 unsigned long *max_delay, bool v1_query) 1333 { 1334 unsigned long mldv1_md; 1335 1336 /* Ignore v1 queries */ 1337 if (mld_in_v2_mode_only(idev)) 1338 return -EINVAL; 1339 1340 mldv1_md = ntohs(mld->mld_maxdelay); 1341 1342 /* When in MLDv1 fallback and a MLDv2 router start-up being 1343 * unaware of current MLDv1 operation, the MRC == MRD mapping 1344 * only works when the exponential algorithm is not being 1345 * used (as MLDv1 is unaware of such things). 1346 * 1347 * According to the RFC author, the MLDv2 implementations 1348 * he's aware of all use a MRC < 32768 on start up queries. 1349 * 1350 * Thus, should we *ever* encounter something else larger 1351 * than that, just assume the maximum possible within our 1352 * reach. 1353 */ 1354 if (!v1_query) 1355 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT); 1356 1357 *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL); 1358 1359 /* MLDv1 router present: we need to go into v1 mode *only* 1360 * when an MLDv1 query is received as per section 9.12. of 1361 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1 1362 * queries MUST be of exactly 24 octets. 1363 */ 1364 if (v1_query) 1365 mld_set_v1_mode(idev); 1366 1367 /* cancel MLDv2 report work */ 1368 mld_gq_stop_work(idev); 1369 /* cancel the interface change work */ 1370 mld_ifc_stop_work(idev); 1371 /* clear deleted report items */ 1372 mld_clear_delrec(idev); 1373 1374 return 0; 1375 } 1376 1377 static void mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld, 1378 unsigned long *max_delay) 1379 { 1380 *max_delay = max(msecs_to_jiffies(mldv2_mrd(mld)), 1UL); 1381 1382 mld_update_qrv(idev, mld); 1383 mld_update_qi(idev, mld); 1384 mld_update_qri(idev, mld); 1385 1386 idev->mc_maxdelay = *max_delay; 1387 1388 return; 1389 } 1390 1391 /* called with rcu_read_lock() */ 1392 void igmp6_event_query(struct sk_buff *skb) 1393 { 1394 struct inet6_dev *idev = __in6_dev_get(skb->dev); 1395 1396 if (!idev || idev->dead) 1397 goto out; 1398 1399 spin_lock_bh(&idev->mc_query_lock); 1400 if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) { 1401 __skb_queue_tail(&idev->mc_query_queue, skb); 1402 if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0)) 1403 in6_dev_hold(idev); 1404 skb = NULL; 1405 } 1406 spin_unlock_bh(&idev->mc_query_lock); 1407 out: 1408 kfree_skb(skb); 1409 } 1410 1411 static void __mld_query_work(struct sk_buff *skb) 1412 { 1413 struct mld2_query *mlh2 = NULL; 1414 const struct in6_addr *group; 1415 unsigned long max_delay; 1416 struct inet6_dev *idev; 1417 struct ifmcaddr6 *ma; 1418 struct mld_msg *mld; 1419 int group_type; 1420 int mark = 0; 1421 int len, err; 1422 1423 if (!pskb_may_pull(skb, sizeof(struct in6_addr))) 1424 goto kfree_skb; 1425 1426 /* compute payload length excluding extension headers */ 1427 len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr); 1428 len -= skb_network_header_len(skb); 1429 1430 /* RFC3810 6.2 1431 * Upon reception of an MLD message that contains a Query, the node 1432 * checks if the source address of the message is a valid link-local 1433 * address, if the Hop Limit is set to 1, and if the Router Alert 1434 * option is present in the Hop-By-Hop Options header of the IPv6 1435 * packet. If any of these checks fails, the packet is dropped. 1436 */ 1437 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) || 1438 ipv6_hdr(skb)->hop_limit != 1 || 1439 !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) || 1440 IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD)) 1441 goto kfree_skb; 1442 1443 idev = in6_dev_get(skb->dev); 1444 if (!idev) 1445 goto kfree_skb; 1446 1447 mld = (struct mld_msg *)icmp6_hdr(skb); 1448 group = &mld->mld_mca; 1449 group_type = ipv6_addr_type(group); 1450 1451 if (group_type != IPV6_ADDR_ANY && 1452 !(group_type&IPV6_ADDR_MULTICAST)) 1453 goto out; 1454 1455 if (len < MLD_V1_QUERY_LEN) { 1456 goto out; 1457 } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) { 1458 err = mld_process_v1(idev, mld, &max_delay, 1459 len == MLD_V1_QUERY_LEN); 1460 if (err < 0) 1461 goto out; 1462 } else if (len >= MLD_V2_QUERY_LEN_MIN) { 1463 int srcs_offset = sizeof(struct mld2_query) - 1464 sizeof(struct icmp6hdr); 1465 1466 if (!pskb_may_pull(skb, srcs_offset)) 1467 goto out; 1468 1469 mlh2 = (struct mld2_query *)skb_transport_header(skb); 1470 1471 mld_process_v2(idev, mlh2, &max_delay); 1472 1473 if (group_type == IPV6_ADDR_ANY) { /* general query */ 1474 if (mlh2->mld2q_nsrcs) 1475 goto out; /* no sources allowed */ 1476 1477 mld_gq_start_work(idev); 1478 goto out; 1479 } 1480 /* mark sources to include, if group & source-specific */ 1481 if (mlh2->mld2q_nsrcs != 0) { 1482 if (!pskb_may_pull(skb, srcs_offset + 1483 ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr))) 1484 goto out; 1485 1486 mlh2 = (struct mld2_query *)skb_transport_header(skb); 1487 mark = 1; 1488 } 1489 } else { 1490 goto out; 1491 } 1492 1493 if (group_type == IPV6_ADDR_ANY) { 1494 for_each_mc_mclock(idev, ma) { 1495 igmp6_group_queried(ma, max_delay); 1496 } 1497 } else { 1498 for_each_mc_mclock(idev, ma) { 1499 if (!ipv6_addr_equal(group, &ma->mca_addr)) 1500 continue; 1501 if (ma->mca_flags & MAF_TIMER_RUNNING) { 1502 /* gsquery <- gsquery && mark */ 1503 if (!mark) 1504 ma->mca_flags &= ~MAF_GSQUERY; 1505 } else { 1506 /* gsquery <- mark */ 1507 if (mark) 1508 ma->mca_flags |= MAF_GSQUERY; 1509 else 1510 ma->mca_flags &= ~MAF_GSQUERY; 1511 } 1512 if (!(ma->mca_flags & MAF_GSQUERY) || 1513 mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs)) 1514 igmp6_group_queried(ma, max_delay); 1515 break; 1516 } 1517 } 1518 1519 out: 1520 in6_dev_put(idev); 1521 kfree_skb: 1522 consume_skb(skb); 1523 } 1524 1525 static void mld_query_work(struct work_struct *work) 1526 { 1527 struct inet6_dev *idev = container_of(to_delayed_work(work), 1528 struct inet6_dev, 1529 mc_query_work); 1530 struct sk_buff_head q; 1531 struct sk_buff *skb; 1532 bool rework = false; 1533 int cnt = 0; 1534 1535 skb_queue_head_init(&q); 1536 1537 spin_lock_bh(&idev->mc_query_lock); 1538 while ((skb = __skb_dequeue(&idev->mc_query_queue))) { 1539 __skb_queue_tail(&q, skb); 1540 1541 if (++cnt >= MLD_MAX_QUEUE) { 1542 rework = true; 1543 break; 1544 } 1545 } 1546 spin_unlock_bh(&idev->mc_query_lock); 1547 1548 mutex_lock(&idev->mc_lock); 1549 while ((skb = __skb_dequeue(&q))) 1550 __mld_query_work(skb); 1551 mutex_unlock(&idev->mc_lock); 1552 1553 if (rework && queue_delayed_work(mld_wq, &idev->mc_query_work, 0)) 1554 return; 1555 1556 in6_dev_put(idev); 1557 } 1558 1559 /* called with rcu_read_lock() */ 1560 void igmp6_event_report(struct sk_buff *skb) 1561 { 1562 struct inet6_dev *idev = __in6_dev_get(skb->dev); 1563 1564 if (!idev || idev->dead) 1565 goto out; 1566 1567 spin_lock_bh(&idev->mc_report_lock); 1568 if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) { 1569 __skb_queue_tail(&idev->mc_report_queue, skb); 1570 if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0)) 1571 in6_dev_hold(idev); 1572 skb = NULL; 1573 } 1574 spin_unlock_bh(&idev->mc_report_lock); 1575 out: 1576 kfree_skb(skb); 1577 } 1578 1579 static void __mld_report_work(struct sk_buff *skb) 1580 { 1581 struct inet6_dev *idev; 1582 struct ifmcaddr6 *ma; 1583 struct mld_msg *mld; 1584 int addr_type; 1585 1586 /* Our own report looped back. Ignore it. */ 1587 if (skb->pkt_type == PACKET_LOOPBACK) 1588 goto kfree_skb; 1589 1590 /* send our report if the MC router may not have heard this report */ 1591 if (skb->pkt_type != PACKET_MULTICAST && 1592 skb->pkt_type != PACKET_BROADCAST) 1593 goto kfree_skb; 1594 1595 if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr))) 1596 goto kfree_skb; 1597 1598 mld = (struct mld_msg *)icmp6_hdr(skb); 1599 1600 /* Drop reports with not link local source */ 1601 addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr); 1602 if (addr_type != IPV6_ADDR_ANY && 1603 !(addr_type&IPV6_ADDR_LINKLOCAL)) 1604 goto kfree_skb; 1605 1606 idev = in6_dev_get(skb->dev); 1607 if (!idev) 1608 goto kfree_skb; 1609 1610 /* 1611 * Cancel the work for this group 1612 */ 1613 1614 for_each_mc_mclock(idev, ma) { 1615 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) { 1616 if (cancel_delayed_work(&ma->mca_work)) 1617 refcount_dec(&ma->mca_refcnt); 1618 ma->mca_flags &= ~(MAF_LAST_REPORTER | 1619 MAF_TIMER_RUNNING); 1620 break; 1621 } 1622 } 1623 1624 in6_dev_put(idev); 1625 kfree_skb: 1626 consume_skb(skb); 1627 } 1628 1629 static void mld_report_work(struct work_struct *work) 1630 { 1631 struct inet6_dev *idev = container_of(to_delayed_work(work), 1632 struct inet6_dev, 1633 mc_report_work); 1634 struct sk_buff_head q; 1635 struct sk_buff *skb; 1636 bool rework = false; 1637 int cnt = 0; 1638 1639 skb_queue_head_init(&q); 1640 spin_lock_bh(&idev->mc_report_lock); 1641 while ((skb = __skb_dequeue(&idev->mc_report_queue))) { 1642 __skb_queue_tail(&q, skb); 1643 1644 if (++cnt >= MLD_MAX_QUEUE) { 1645 rework = true; 1646 break; 1647 } 1648 } 1649 spin_unlock_bh(&idev->mc_report_lock); 1650 1651 mutex_lock(&idev->mc_lock); 1652 while ((skb = __skb_dequeue(&q))) 1653 __mld_report_work(skb); 1654 mutex_unlock(&idev->mc_lock); 1655 1656 if (rework && queue_delayed_work(mld_wq, &idev->mc_report_work, 0)) 1657 return; 1658 1659 in6_dev_put(idev); 1660 } 1661 1662 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type, 1663 int gdeleted, int sdeleted) 1664 { 1665 switch (type) { 1666 case MLD2_MODE_IS_INCLUDE: 1667 case MLD2_MODE_IS_EXCLUDE: 1668 if (gdeleted || sdeleted) 1669 return false; 1670 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) { 1671 if (pmc->mca_sfmode == MCAST_INCLUDE) 1672 return true; 1673 /* don't include if this source is excluded 1674 * in all filters 1675 */ 1676 if (psf->sf_count[MCAST_INCLUDE]) 1677 return type == MLD2_MODE_IS_INCLUDE; 1678 return pmc->mca_sfcount[MCAST_EXCLUDE] == 1679 psf->sf_count[MCAST_EXCLUDE]; 1680 } 1681 return false; 1682 case MLD2_CHANGE_TO_INCLUDE: 1683 if (gdeleted || sdeleted) 1684 return false; 1685 return psf->sf_count[MCAST_INCLUDE] != 0; 1686 case MLD2_CHANGE_TO_EXCLUDE: 1687 if (gdeleted || sdeleted) 1688 return false; 1689 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 || 1690 psf->sf_count[MCAST_INCLUDE]) 1691 return false; 1692 return pmc->mca_sfcount[MCAST_EXCLUDE] == 1693 psf->sf_count[MCAST_EXCLUDE]; 1694 case MLD2_ALLOW_NEW_SOURCES: 1695 if (gdeleted || !psf->sf_crcount) 1696 return false; 1697 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted; 1698 case MLD2_BLOCK_OLD_SOURCES: 1699 if (pmc->mca_sfmode == MCAST_INCLUDE) 1700 return gdeleted || (psf->sf_crcount && sdeleted); 1701 return psf->sf_crcount && !gdeleted && !sdeleted; 1702 } 1703 return false; 1704 } 1705 1706 static int 1707 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted) 1708 { 1709 struct ip6_sf_list *psf; 1710 int scount = 0; 1711 1712 for_each_psf_mclock(pmc, psf) { 1713 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) 1714 continue; 1715 scount++; 1716 } 1717 return scount; 1718 } 1719 1720 static void ip6_mc_hdr(const struct sock *sk, struct sk_buff *skb, 1721 struct net_device *dev, const struct in6_addr *saddr, 1722 const struct in6_addr *daddr, int proto, int len) 1723 { 1724 struct ipv6hdr *hdr; 1725 1726 skb->protocol = htons(ETH_P_IPV6); 1727 skb->dev = dev; 1728 1729 skb_reset_network_header(skb); 1730 skb_put(skb, sizeof(struct ipv6hdr)); 1731 hdr = ipv6_hdr(skb); 1732 1733 ip6_flow_hdr(hdr, 0, 0); 1734 1735 hdr->payload_len = htons(len); 1736 hdr->nexthdr = proto; 1737 hdr->hop_limit = READ_ONCE(inet6_sk(sk)->hop_limit); 1738 1739 hdr->saddr = *saddr; 1740 hdr->daddr = *daddr; 1741 } 1742 1743 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu) 1744 { 1745 u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT, 1746 2, 0, 0, IPV6_TLV_PADN, 0 }; 1747 struct net_device *dev = idev->dev; 1748 int hlen = LL_RESERVED_SPACE(dev); 1749 int tlen = dev->needed_tailroom; 1750 const struct in6_addr *saddr; 1751 struct in6_addr addr_buf; 1752 struct mld2_report *pmr; 1753 struct sk_buff *skb; 1754 unsigned int size; 1755 struct sock *sk; 1756 struct net *net; 1757 1758 /* we assume size > sizeof(ra) here 1759 * Also try to not allocate high-order pages for big MTU 1760 */ 1761 size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen; 1762 skb = alloc_skb(size, GFP_KERNEL); 1763 if (!skb) 1764 return NULL; 1765 1766 skb->priority = TC_PRIO_CONTROL; 1767 skb_reserve(skb, hlen); 1768 skb_tailroom_reserve(skb, mtu, tlen); 1769 1770 rcu_read_lock(); 1771 1772 net = dev_net_rcu(dev); 1773 sk = net->ipv6.igmp_sk; 1774 skb_set_owner_w(skb, sk); 1775 1776 if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) { 1777 /* <draft-ietf-magma-mld-source-05.txt>: 1778 * use unspecified address as the source address 1779 * when a valid link-local address is not available. 1780 */ 1781 saddr = &in6addr_any; 1782 } else 1783 saddr = &addr_buf; 1784 1785 ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0); 1786 1787 rcu_read_unlock(); 1788 1789 skb_put_data(skb, ra, sizeof(ra)); 1790 1791 skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data); 1792 skb_put(skb, sizeof(*pmr)); 1793 pmr = (struct mld2_report *)skb_transport_header(skb); 1794 pmr->mld2r_type = ICMPV6_MLD2_REPORT; 1795 pmr->mld2r_resv1 = 0; 1796 pmr->mld2r_cksum = 0; 1797 pmr->mld2r_resv2 = 0; 1798 pmr->mld2r_ngrec = 0; 1799 return skb; 1800 } 1801 1802 static void mld_sendpack(struct sk_buff *skb) 1803 { 1804 struct ipv6hdr *pip6 = ipv6_hdr(skb); 1805 struct mld2_report *pmr = 1806 (struct mld2_report *)skb_transport_header(skb); 1807 int payload_len, mldlen; 1808 struct inet6_dev *idev; 1809 struct net *net = dev_net(skb->dev); 1810 int err; 1811 struct flowi6 fl6; 1812 struct dst_entry *dst; 1813 1814 rcu_read_lock(); 1815 idev = __in6_dev_get(skb->dev); 1816 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS); 1817 1818 payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) - 1819 sizeof(*pip6); 1820 mldlen = skb_tail_pointer(skb) - skb_transport_header(skb); 1821 pip6->payload_len = htons(payload_len); 1822 1823 pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen, 1824 IPPROTO_ICMPV6, 1825 csum_partial(skb_transport_header(skb), 1826 mldlen, 0)); 1827 1828 icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT, 1829 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 1830 skb->dev->ifindex); 1831 dst = icmp6_dst_alloc(skb->dev, &fl6); 1832 1833 err = 0; 1834 if (IS_ERR(dst)) { 1835 err = PTR_ERR(dst); 1836 dst = NULL; 1837 } 1838 skb_dst_set(skb, dst); 1839 if (err) 1840 goto err_out; 1841 1842 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, 1843 net, net->ipv6.igmp_sk, skb, NULL, skb->dev, 1844 dst_output); 1845 out: 1846 if (!err) { 1847 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT); 1848 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); 1849 } else { 1850 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); 1851 } 1852 1853 rcu_read_unlock(); 1854 return; 1855 1856 err_out: 1857 kfree_skb(skb); 1858 goto out; 1859 } 1860 1861 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel) 1862 { 1863 return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel); 1864 } 1865 1866 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc, 1867 int type, struct mld2_grec **ppgr, unsigned int mtu) 1868 { 1869 struct mld2_report *pmr; 1870 struct mld2_grec *pgr; 1871 1872 if (!skb) { 1873 skb = mld_newpack(pmc->idev, mtu); 1874 if (!skb) 1875 return NULL; 1876 } 1877 pgr = skb_put(skb, sizeof(struct mld2_grec)); 1878 pgr->grec_type = type; 1879 pgr->grec_auxwords = 0; 1880 pgr->grec_nsrcs = 0; 1881 pgr->grec_mca = pmc->mca_addr; /* structure copy */ 1882 pmr = (struct mld2_report *)skb_transport_header(skb); 1883 pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1); 1884 *ppgr = pgr; 1885 return skb; 1886 } 1887 1888 #define AVAILABLE(skb) ((skb) ? skb_availroom(skb) : 0) 1889 1890 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc, 1891 int type, int gdeleted, int sdeleted, 1892 int crsend) 1893 { 1894 struct ip6_sf_list *psf, *psf_prev, *psf_next; 1895 int scount, stotal, first, isquery, truncate; 1896 struct ip6_sf_list __rcu **psf_list; 1897 struct inet6_dev *idev = pmc->idev; 1898 struct net_device *dev = idev->dev; 1899 struct mld2_grec *pgr = NULL; 1900 struct mld2_report *pmr; 1901 unsigned int mtu; 1902 1903 mc_assert_locked(idev); 1904 1905 if (pmc->mca_flags & MAF_NOREPORT) 1906 return skb; 1907 1908 mtu = READ_ONCE(dev->mtu); 1909 if (mtu < IPV6_MIN_MTU) 1910 return skb; 1911 1912 isquery = type == MLD2_MODE_IS_INCLUDE || 1913 type == MLD2_MODE_IS_EXCLUDE; 1914 truncate = type == MLD2_MODE_IS_EXCLUDE || 1915 type == MLD2_CHANGE_TO_EXCLUDE; 1916 1917 stotal = scount = 0; 1918 1919 psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources; 1920 1921 if (!rcu_access_pointer(*psf_list)) 1922 goto empty_source; 1923 1924 pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL; 1925 1926 /* EX and TO_EX get a fresh packet, if needed */ 1927 if (truncate) { 1928 if (pmr && pmr->mld2r_ngrec && 1929 AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) { 1930 if (skb) 1931 mld_sendpack(skb); 1932 skb = mld_newpack(idev, mtu); 1933 } 1934 } 1935 first = 1; 1936 psf_prev = NULL; 1937 for (psf = mc_dereference(*psf_list, idev); 1938 psf; 1939 psf = psf_next) { 1940 struct in6_addr *psrc; 1941 1942 psf_next = mc_dereference(psf->sf_next, idev); 1943 1944 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) { 1945 psf_prev = psf; 1946 continue; 1947 } 1948 1949 /* Based on RFC3810 6.1. Should not send source-list change 1950 * records when there is a filter mode change. 1951 */ 1952 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) || 1953 (!gdeleted && pmc->mca_crcount)) && 1954 (type == MLD2_ALLOW_NEW_SOURCES || 1955 type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) 1956 goto decrease_sf_crcount; 1957 1958 /* clear marks on query responses */ 1959 if (isquery) 1960 psf->sf_gsresp = 0; 1961 1962 if (AVAILABLE(skb) < sizeof(*psrc) + 1963 first*sizeof(struct mld2_grec)) { 1964 if (truncate && !first) 1965 break; /* truncate these */ 1966 if (pgr) 1967 pgr->grec_nsrcs = htons(scount); 1968 if (skb) 1969 mld_sendpack(skb); 1970 skb = mld_newpack(idev, mtu); 1971 first = 1; 1972 scount = 0; 1973 } 1974 if (first) { 1975 skb = add_grhead(skb, pmc, type, &pgr, mtu); 1976 first = 0; 1977 } 1978 if (!skb) 1979 return NULL; 1980 psrc = skb_put(skb, sizeof(*psrc)); 1981 *psrc = psf->sf_addr; 1982 scount++; stotal++; 1983 if ((type == MLD2_ALLOW_NEW_SOURCES || 1984 type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) { 1985 decrease_sf_crcount: 1986 psf->sf_crcount--; 1987 if ((sdeleted || gdeleted) && psf->sf_crcount == 0) { 1988 if (psf_prev) 1989 rcu_assign_pointer(psf_prev->sf_next, 1990 mc_dereference(psf->sf_next, idev)); 1991 else 1992 rcu_assign_pointer(*psf_list, 1993 mc_dereference(psf->sf_next, idev)); 1994 kfree_rcu(psf, rcu); 1995 continue; 1996 } 1997 } 1998 psf_prev = psf; 1999 } 2000 2001 empty_source: 2002 if (!stotal) { 2003 if (type == MLD2_ALLOW_NEW_SOURCES || 2004 type == MLD2_BLOCK_OLD_SOURCES) 2005 return skb; 2006 if (pmc->mca_crcount || isquery || crsend) { 2007 /* make sure we have room for group header */ 2008 if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) { 2009 mld_sendpack(skb); 2010 skb = NULL; /* add_grhead will get a new one */ 2011 } 2012 skb = add_grhead(skb, pmc, type, &pgr, mtu); 2013 } 2014 } 2015 if (pgr) 2016 pgr->grec_nsrcs = htons(scount); 2017 2018 if (isquery) 2019 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */ 2020 return skb; 2021 } 2022 2023 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc) 2024 { 2025 struct sk_buff *skb = NULL; 2026 int type; 2027 2028 mc_assert_locked(idev); 2029 2030 if (!pmc) { 2031 for_each_mc_mclock(idev, pmc) { 2032 if (pmc->mca_flags & MAF_NOREPORT) 2033 continue; 2034 if (pmc->mca_sfcount[MCAST_EXCLUDE]) 2035 type = MLD2_MODE_IS_EXCLUDE; 2036 else 2037 type = MLD2_MODE_IS_INCLUDE; 2038 skb = add_grec(skb, pmc, type, 0, 0, 0); 2039 } 2040 } else { 2041 if (pmc->mca_sfcount[MCAST_EXCLUDE]) 2042 type = MLD2_MODE_IS_EXCLUDE; 2043 else 2044 type = MLD2_MODE_IS_INCLUDE; 2045 skb = add_grec(skb, pmc, type, 0, 0, 0); 2046 } 2047 if (skb) 2048 mld_sendpack(skb); 2049 } 2050 2051 /* remove zero-count source records from a source filter list */ 2052 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev) 2053 { 2054 struct ip6_sf_list *psf_prev, *psf_next, *psf; 2055 2056 psf_prev = NULL; 2057 for (psf = mc_dereference(*ppsf, idev); 2058 psf; 2059 psf = psf_next) { 2060 psf_next = mc_dereference(psf->sf_next, idev); 2061 if (psf->sf_crcount == 0) { 2062 if (psf_prev) 2063 rcu_assign_pointer(psf_prev->sf_next, 2064 mc_dereference(psf->sf_next, idev)); 2065 else 2066 rcu_assign_pointer(*ppsf, 2067 mc_dereference(psf->sf_next, idev)); 2068 kfree_rcu(psf, rcu); 2069 } else { 2070 psf_prev = psf; 2071 } 2072 } 2073 } 2074 2075 static void mld_send_cr(struct inet6_dev *idev) 2076 { 2077 struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next; 2078 struct sk_buff *skb = NULL; 2079 int type, dtype; 2080 2081 /* deleted MCA's */ 2082 pmc_prev = NULL; 2083 for (pmc = mc_dereference(idev->mc_tomb, idev); 2084 pmc; 2085 pmc = pmc_next) { 2086 pmc_next = mc_dereference(pmc->next, idev); 2087 if (pmc->mca_sfmode == MCAST_INCLUDE) { 2088 type = MLD2_BLOCK_OLD_SOURCES; 2089 dtype = MLD2_BLOCK_OLD_SOURCES; 2090 skb = add_grec(skb, pmc, type, 1, 0, 0); 2091 skb = add_grec(skb, pmc, dtype, 1, 1, 0); 2092 } 2093 if (pmc->mca_crcount) { 2094 if (pmc->mca_sfmode == MCAST_EXCLUDE) { 2095 type = MLD2_CHANGE_TO_INCLUDE; 2096 skb = add_grec(skb, pmc, type, 1, 0, 0); 2097 } 2098 pmc->mca_crcount--; 2099 if (pmc->mca_crcount == 0) { 2100 mld_clear_zeros(&pmc->mca_tomb, idev); 2101 mld_clear_zeros(&pmc->mca_sources, idev); 2102 } 2103 } 2104 if (pmc->mca_crcount == 0 && 2105 !rcu_access_pointer(pmc->mca_tomb) && 2106 !rcu_access_pointer(pmc->mca_sources)) { 2107 if (pmc_prev) 2108 rcu_assign_pointer(pmc_prev->next, pmc_next); 2109 else 2110 rcu_assign_pointer(idev->mc_tomb, pmc_next); 2111 in6_dev_put(pmc->idev); 2112 kfree_rcu(pmc, rcu); 2113 } else 2114 pmc_prev = pmc; 2115 } 2116 2117 /* change recs */ 2118 for_each_mc_mclock(idev, pmc) { 2119 if (pmc->mca_sfcount[MCAST_EXCLUDE]) { 2120 type = MLD2_BLOCK_OLD_SOURCES; 2121 dtype = MLD2_ALLOW_NEW_SOURCES; 2122 } else { 2123 type = MLD2_ALLOW_NEW_SOURCES; 2124 dtype = MLD2_BLOCK_OLD_SOURCES; 2125 } 2126 skb = add_grec(skb, pmc, type, 0, 0, 0); 2127 skb = add_grec(skb, pmc, dtype, 0, 1, 0); /* deleted sources */ 2128 2129 /* filter mode changes */ 2130 if (pmc->mca_crcount) { 2131 if (pmc->mca_sfmode == MCAST_EXCLUDE) 2132 type = MLD2_CHANGE_TO_EXCLUDE; 2133 else 2134 type = MLD2_CHANGE_TO_INCLUDE; 2135 skb = add_grec(skb, pmc, type, 0, 0, 0); 2136 pmc->mca_crcount--; 2137 } 2138 } 2139 if (!skb) 2140 return; 2141 (void) mld_sendpack(skb); 2142 } 2143 2144 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type) 2145 { 2146 const struct in6_addr *snd_addr, *saddr; 2147 int err, len, payload_len, full_len; 2148 struct in6_addr addr_buf; 2149 struct inet6_dev *idev; 2150 struct sk_buff *skb; 2151 struct mld_msg *hdr; 2152 int hlen = LL_RESERVED_SPACE(dev); 2153 int tlen = dev->needed_tailroom; 2154 u8 ra[8] = { IPPROTO_ICMPV6, 0, 2155 IPV6_TLV_ROUTERALERT, 2, 0, 0, 2156 IPV6_TLV_PADN, 0 }; 2157 struct dst_entry *dst; 2158 struct flowi6 fl6; 2159 struct net *net; 2160 struct sock *sk; 2161 2162 if (type == ICMPV6_MGM_REDUCTION) 2163 snd_addr = &in6addr_linklocal_allrouters; 2164 else 2165 snd_addr = addr; 2166 2167 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr); 2168 payload_len = len + sizeof(ra); 2169 full_len = sizeof(struct ipv6hdr) + payload_len; 2170 2171 skb = alloc_skb(hlen + tlen + full_len, GFP_KERNEL); 2172 2173 rcu_read_lock(); 2174 2175 net = dev_net_rcu(dev); 2176 idev = __in6_dev_get(dev); 2177 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS); 2178 if (!skb) { 2179 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); 2180 rcu_read_unlock(); 2181 return; 2182 } 2183 sk = net->ipv6.igmp_sk; 2184 skb_set_owner_w(skb, sk); 2185 2186 skb->priority = TC_PRIO_CONTROL; 2187 skb_reserve(skb, hlen); 2188 2189 if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) { 2190 /* <draft-ietf-magma-mld-source-05.txt>: 2191 * use unspecified address as the source address 2192 * when a valid link-local address is not available. 2193 */ 2194 saddr = &in6addr_any; 2195 } else 2196 saddr = &addr_buf; 2197 2198 ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len); 2199 2200 skb_put_data(skb, ra, sizeof(ra)); 2201 2202 hdr = skb_put_zero(skb, sizeof(struct mld_msg)); 2203 hdr->mld_type = type; 2204 hdr->mld_mca = *addr; 2205 2206 hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len, 2207 IPPROTO_ICMPV6, 2208 csum_partial(hdr, len, 0)); 2209 2210 icmpv6_flow_init(sk, &fl6, type, 2211 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 2212 skb->dev->ifindex); 2213 dst = icmp6_dst_alloc(skb->dev, &fl6); 2214 if (IS_ERR(dst)) { 2215 err = PTR_ERR(dst); 2216 goto err_out; 2217 } 2218 2219 skb_dst_set(skb, dst); 2220 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, 2221 net, sk, skb, NULL, skb->dev, 2222 dst_output); 2223 out: 2224 if (!err) { 2225 ICMP6MSGOUT_INC_STATS(net, idev, type); 2226 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); 2227 } else 2228 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); 2229 2230 rcu_read_unlock(); 2231 return; 2232 2233 err_out: 2234 kfree_skb(skb); 2235 goto out; 2236 } 2237 2238 static void mld_send_initial_cr(struct inet6_dev *idev) 2239 { 2240 struct ifmcaddr6 *pmc; 2241 struct sk_buff *skb; 2242 int type; 2243 2244 mc_assert_locked(idev); 2245 2246 if (mld_in_v1_mode(idev)) 2247 return; 2248 2249 skb = NULL; 2250 for_each_mc_mclock(idev, pmc) { 2251 if (pmc->mca_sfcount[MCAST_EXCLUDE]) 2252 type = MLD2_CHANGE_TO_EXCLUDE; 2253 else 2254 type = MLD2_ALLOW_NEW_SOURCES; 2255 skb = add_grec(skb, pmc, type, 0, 0, 1); 2256 } 2257 if (skb) 2258 mld_sendpack(skb); 2259 } 2260 2261 void ipv6_mc_dad_complete(struct inet6_dev *idev) 2262 { 2263 mutex_lock(&idev->mc_lock); 2264 idev->mc_dad_count = idev->mc_qrv; 2265 if (idev->mc_dad_count) { 2266 mld_send_initial_cr(idev); 2267 idev->mc_dad_count--; 2268 if (idev->mc_dad_count) 2269 mld_dad_start_work(idev, 2270 unsolicited_report_interval(idev)); 2271 } 2272 mutex_unlock(&idev->mc_lock); 2273 } 2274 2275 static void mld_dad_work(struct work_struct *work) 2276 { 2277 struct inet6_dev *idev = container_of(to_delayed_work(work), 2278 struct inet6_dev, 2279 mc_dad_work); 2280 mutex_lock(&idev->mc_lock); 2281 mld_send_initial_cr(idev); 2282 if (idev->mc_dad_count) { 2283 idev->mc_dad_count--; 2284 if (idev->mc_dad_count) 2285 mld_dad_start_work(idev, 2286 unsolicited_report_interval(idev)); 2287 } 2288 mutex_unlock(&idev->mc_lock); 2289 in6_dev_put(idev); 2290 } 2291 2292 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode, 2293 const struct in6_addr *psfsrc) 2294 { 2295 struct ip6_sf_list *psf, *psf_prev; 2296 int rv = 0; 2297 2298 mc_assert_locked(pmc->idev); 2299 2300 psf_prev = NULL; 2301 for_each_psf_mclock(pmc, psf) { 2302 if (ipv6_addr_equal(&psf->sf_addr, psfsrc)) 2303 break; 2304 psf_prev = psf; 2305 } 2306 if (!psf || psf->sf_count[sfmode] == 0) { 2307 /* source filter not found, or count wrong => bug */ 2308 return -ESRCH; 2309 } 2310 WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] - 1); 2311 if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) { 2312 struct inet6_dev *idev = pmc->idev; 2313 2314 /* no more filters for this source */ 2315 if (psf_prev) 2316 rcu_assign_pointer(psf_prev->sf_next, 2317 mc_dereference(psf->sf_next, idev)); 2318 else 2319 rcu_assign_pointer(pmc->mca_sources, 2320 mc_dereference(psf->sf_next, idev)); 2321 2322 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) && 2323 !mld_in_v1_mode(idev)) { 2324 psf->sf_crcount = idev->mc_qrv; 2325 rcu_assign_pointer(psf->sf_next, 2326 mc_dereference(pmc->mca_tomb, idev)); 2327 rcu_assign_pointer(pmc->mca_tomb, psf); 2328 rv = 1; 2329 } else { 2330 kfree_rcu(psf, rcu); 2331 } 2332 } 2333 return rv; 2334 } 2335 2336 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca, 2337 int sfmode, int sfcount, const struct in6_addr *psfsrc, 2338 int delta) 2339 { 2340 struct ifmcaddr6 *pmc; 2341 int changerec = 0; 2342 int i, err; 2343 2344 if (!idev) 2345 return -ENODEV; 2346 2347 mc_assert_locked(idev); 2348 2349 for_each_mc_mclock(idev, pmc) { 2350 if (ipv6_addr_equal(pmca, &pmc->mca_addr)) 2351 break; 2352 } 2353 if (!pmc) 2354 return -ESRCH; 2355 2356 sf_markstate(pmc); 2357 if (!delta) { 2358 if (!pmc->mca_sfcount[sfmode]) 2359 return -EINVAL; 2360 2361 pmc->mca_sfcount[sfmode]--; 2362 } 2363 err = 0; 2364 for (i = 0; i < sfcount; i++) { 2365 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]); 2366 2367 changerec |= rv > 0; 2368 if (!err && rv < 0) 2369 err = rv; 2370 } 2371 if (pmc->mca_sfmode == MCAST_EXCLUDE && 2372 pmc->mca_sfcount[MCAST_EXCLUDE] == 0 && 2373 pmc->mca_sfcount[MCAST_INCLUDE]) { 2374 struct ip6_sf_list *psf; 2375 2376 /* filter mode change */ 2377 pmc->mca_sfmode = MCAST_INCLUDE; 2378 pmc->mca_crcount = idev->mc_qrv; 2379 idev->mc_ifc_count = pmc->mca_crcount; 2380 for_each_psf_mclock(pmc, psf) 2381 psf->sf_crcount = 0; 2382 mld_ifc_event(pmc->idev); 2383 } else if (sf_setstate(pmc) || changerec) { 2384 mld_ifc_event(pmc->idev); 2385 } 2386 2387 return err; 2388 } 2389 2390 /* Add multicast single-source filter to the interface list */ 2391 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode, 2392 const struct in6_addr *psfsrc) 2393 { 2394 struct ip6_sf_list *psf, *psf_prev; 2395 2396 mc_assert_locked(pmc->idev); 2397 2398 psf_prev = NULL; 2399 for_each_psf_mclock(pmc, psf) { 2400 if (ipv6_addr_equal(&psf->sf_addr, psfsrc)) 2401 break; 2402 psf_prev = psf; 2403 } 2404 if (!psf) { 2405 psf = kzalloc_obj(*psf); 2406 if (!psf) 2407 return -ENOBUFS; 2408 2409 psf->sf_addr = *psfsrc; 2410 if (psf_prev) { 2411 rcu_assign_pointer(psf_prev->sf_next, psf); 2412 } else { 2413 rcu_assign_pointer(pmc->mca_sources, psf); 2414 } 2415 } 2416 WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] + 1); 2417 return 0; 2418 } 2419 2420 static void sf_markstate(struct ifmcaddr6 *pmc) 2421 { 2422 int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE]; 2423 struct ip6_sf_list *psf; 2424 2425 mc_assert_locked(pmc->idev); 2426 2427 for_each_psf_mclock(pmc, psf) { 2428 if (pmc->mca_sfcount[MCAST_EXCLUDE]) { 2429 psf->sf_oldin = mca_xcount == 2430 psf->sf_count[MCAST_EXCLUDE] && 2431 !psf->sf_count[MCAST_INCLUDE]; 2432 } else { 2433 psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0; 2434 } 2435 } 2436 } 2437 2438 static int sf_setstate(struct ifmcaddr6 *pmc) 2439 { 2440 int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE]; 2441 struct ip6_sf_list *psf, *dpsf; 2442 int qrv = pmc->idev->mc_qrv; 2443 int new_in, rv; 2444 2445 mc_assert_locked(pmc->idev); 2446 2447 rv = 0; 2448 for_each_psf_mclock(pmc, psf) { 2449 if (pmc->mca_sfcount[MCAST_EXCLUDE]) { 2450 new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] && 2451 !psf->sf_count[MCAST_INCLUDE]; 2452 } else 2453 new_in = psf->sf_count[MCAST_INCLUDE] != 0; 2454 if (new_in) { 2455 if (!psf->sf_oldin) { 2456 struct ip6_sf_list *prev = NULL; 2457 2458 for_each_psf_tomb(pmc, dpsf) { 2459 if (ipv6_addr_equal(&dpsf->sf_addr, 2460 &psf->sf_addr)) 2461 break; 2462 prev = dpsf; 2463 } 2464 if (dpsf) { 2465 if (prev) 2466 rcu_assign_pointer(prev->sf_next, 2467 mc_dereference(dpsf->sf_next, 2468 pmc->idev)); 2469 else 2470 rcu_assign_pointer(pmc->mca_tomb, 2471 mc_dereference(dpsf->sf_next, 2472 pmc->idev)); 2473 kfree_rcu(dpsf, rcu); 2474 } 2475 psf->sf_crcount = qrv; 2476 rv++; 2477 } 2478 } else if (psf->sf_oldin) { 2479 psf->sf_crcount = 0; 2480 /* 2481 * add or update "delete" records if an active filter 2482 * is now inactive 2483 */ 2484 2485 for_each_psf_tomb(pmc, dpsf) 2486 if (ipv6_addr_equal(&dpsf->sf_addr, 2487 &psf->sf_addr)) 2488 break; 2489 if (!dpsf) { 2490 dpsf = kmalloc_obj(*dpsf); 2491 if (!dpsf) 2492 continue; 2493 *dpsf = *psf; 2494 rcu_assign_pointer(dpsf->sf_next, 2495 mc_dereference(pmc->mca_tomb, pmc->idev)); 2496 rcu_assign_pointer(pmc->mca_tomb, dpsf); 2497 } 2498 dpsf->sf_crcount = qrv; 2499 rv++; 2500 } 2501 } 2502 return rv; 2503 } 2504 2505 /* Add multicast source filter list to the interface list */ 2506 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca, 2507 int sfmode, int sfcount, const struct in6_addr *psfsrc, 2508 int delta) 2509 { 2510 struct ifmcaddr6 *pmc; 2511 int isexclude; 2512 int i, err; 2513 2514 if (!idev) 2515 return -ENODEV; 2516 2517 mc_assert_locked(idev); 2518 2519 for_each_mc_mclock(idev, pmc) { 2520 if (ipv6_addr_equal(pmca, &pmc->mca_addr)) 2521 break; 2522 } 2523 if (!pmc) 2524 return -ESRCH; 2525 2526 sf_markstate(pmc); 2527 isexclude = pmc->mca_sfmode == MCAST_EXCLUDE; 2528 if (!delta) 2529 WRITE_ONCE(pmc->mca_sfcount[sfmode], 2530 pmc->mca_sfcount[sfmode] + 1); 2531 err = 0; 2532 for (i = 0; i < sfcount; i++) { 2533 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]); 2534 if (err) 2535 break; 2536 } 2537 if (err) { 2538 int j; 2539 2540 if (!delta) 2541 WRITE_ONCE(pmc->mca_sfcount[sfmode], 2542 pmc->mca_sfcount[sfmode] - 1); 2543 for (j = 0; j < i; j++) 2544 ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]); 2545 } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) { 2546 struct ip6_sf_list *psf; 2547 2548 /* filter mode change */ 2549 if (pmc->mca_sfcount[MCAST_EXCLUDE]) 2550 pmc->mca_sfmode = MCAST_EXCLUDE; 2551 else if (pmc->mca_sfcount[MCAST_INCLUDE]) 2552 pmc->mca_sfmode = MCAST_INCLUDE; 2553 /* else no filters; keep old mode for reports */ 2554 2555 pmc->mca_crcount = idev->mc_qrv; 2556 idev->mc_ifc_count = pmc->mca_crcount; 2557 for_each_psf_mclock(pmc, psf) 2558 psf->sf_crcount = 0; 2559 mld_ifc_event(idev); 2560 } else if (sf_setstate(pmc)) { 2561 mld_ifc_event(idev); 2562 } 2563 return err; 2564 } 2565 2566 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc) 2567 { 2568 struct ip6_sf_list *psf, *nextpsf; 2569 2570 mc_assert_locked(pmc->idev); 2571 2572 for (psf = mc_dereference(pmc->mca_tomb, pmc->idev); 2573 psf; 2574 psf = nextpsf) { 2575 nextpsf = mc_dereference(psf->sf_next, pmc->idev); 2576 kfree_rcu(psf, rcu); 2577 } 2578 RCU_INIT_POINTER(pmc->mca_tomb, NULL); 2579 for (psf = mc_dereference(pmc->mca_sources, pmc->idev); 2580 psf; 2581 psf = nextpsf) { 2582 nextpsf = mc_dereference(psf->sf_next, pmc->idev); 2583 kfree_rcu(psf, rcu); 2584 } 2585 RCU_INIT_POINTER(pmc->mca_sources, NULL); 2586 pmc->mca_sfmode = MCAST_EXCLUDE; 2587 pmc->mca_sfcount[MCAST_INCLUDE] = 0; 2588 /* Paired with the READ_ONCE() from ipv6_chk_mcast_addr() */ 2589 WRITE_ONCE(pmc->mca_sfcount[MCAST_EXCLUDE], 1); 2590 } 2591 2592 static void igmp6_join_group(struct ifmcaddr6 *ma) 2593 { 2594 unsigned long delay; 2595 2596 mc_assert_locked(ma->idev); 2597 2598 if (ma->mca_flags & MAF_NOREPORT) 2599 return; 2600 2601 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT); 2602 2603 delay = get_random_u32_below(unsolicited_report_interval(ma->idev)); 2604 2605 if (cancel_delayed_work(&ma->mca_work)) { 2606 refcount_dec(&ma->mca_refcnt); 2607 delay = ma->mca_work.timer.expires - jiffies; 2608 } 2609 2610 if (!mod_delayed_work(mld_wq, &ma->mca_work, delay)) 2611 refcount_inc(&ma->mca_refcnt); 2612 ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER; 2613 } 2614 2615 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml, 2616 struct inet6_dev *idev) 2617 { 2618 struct ip6_sf_socklist *psl; 2619 int err; 2620 2621 psl = sock_dereference(iml->sflist, sk); 2622 2623 if (idev) 2624 mutex_lock(&idev->mc_lock); 2625 2626 if (!psl) { 2627 /* any-source empty exclude case */ 2628 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0); 2629 } else { 2630 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 2631 psl->sl_count, psl->sl_addr, 0); 2632 RCU_INIT_POINTER(iml->sflist, NULL); 2633 atomic_sub(struct_size(psl, sl_addr, psl->sl_max), 2634 &sk->sk_omem_alloc); 2635 kfree_rcu(psl, rcu); 2636 } 2637 2638 if (idev) 2639 mutex_unlock(&idev->mc_lock); 2640 2641 return err; 2642 } 2643 2644 static void igmp6_leave_group(struct ifmcaddr6 *ma) 2645 { 2646 mc_assert_locked(ma->idev); 2647 2648 if (mld_in_v1_mode(ma->idev)) { 2649 if (ma->mca_flags & MAF_LAST_REPORTER) { 2650 igmp6_send(&ma->mca_addr, ma->idev->dev, 2651 ICMPV6_MGM_REDUCTION); 2652 } 2653 } else { 2654 mld_add_delrec(ma->idev, ma); 2655 mld_ifc_event(ma->idev); 2656 } 2657 } 2658 2659 static void mld_gq_work(struct work_struct *work) 2660 { 2661 struct inet6_dev *idev = container_of(to_delayed_work(work), 2662 struct inet6_dev, 2663 mc_gq_work); 2664 2665 mutex_lock(&idev->mc_lock); 2666 mld_send_report(idev, NULL); 2667 idev->mc_gq_running = 0; 2668 mutex_unlock(&idev->mc_lock); 2669 2670 in6_dev_put(idev); 2671 } 2672 2673 static void mld_ifc_work(struct work_struct *work) 2674 { 2675 struct inet6_dev *idev = container_of(to_delayed_work(work), 2676 struct inet6_dev, 2677 mc_ifc_work); 2678 2679 mutex_lock(&idev->mc_lock); 2680 mld_send_cr(idev); 2681 2682 if (idev->mc_ifc_count) { 2683 idev->mc_ifc_count--; 2684 if (idev->mc_ifc_count) 2685 mld_ifc_start_work(idev, 2686 unsolicited_report_interval(idev)); 2687 } 2688 mutex_unlock(&idev->mc_lock); 2689 in6_dev_put(idev); 2690 } 2691 2692 static void mld_ifc_event(struct inet6_dev *idev) 2693 { 2694 mc_assert_locked(idev); 2695 2696 if (mld_in_v1_mode(idev)) 2697 return; 2698 2699 idev->mc_ifc_count = idev->mc_qrv; 2700 mld_ifc_start_work(idev, 1); 2701 } 2702 2703 static void mld_mca_work(struct work_struct *work) 2704 { 2705 struct ifmcaddr6 *ma = container_of(to_delayed_work(work), 2706 struct ifmcaddr6, mca_work); 2707 2708 mutex_lock(&ma->idev->mc_lock); 2709 if (mld_in_v1_mode(ma->idev)) 2710 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT); 2711 else 2712 mld_send_report(ma->idev, ma); 2713 ma->mca_flags |= MAF_LAST_REPORTER; 2714 ma->mca_flags &= ~MAF_TIMER_RUNNING; 2715 mutex_unlock(&ma->idev->mc_lock); 2716 2717 ma_put(ma); 2718 } 2719 2720 /* Device changing type */ 2721 2722 void ipv6_mc_unmap(struct inet6_dev *idev) 2723 { 2724 struct ifmcaddr6 *i; 2725 2726 /* Install multicast list, except for all-nodes (already installed) */ 2727 2728 mutex_lock(&idev->mc_lock); 2729 for_each_mc_mclock(idev, i) 2730 igmp6_group_dropped(i); 2731 mutex_unlock(&idev->mc_lock); 2732 } 2733 2734 void ipv6_mc_remap(struct inet6_dev *idev) 2735 { 2736 ipv6_mc_up(idev); 2737 } 2738 2739 /* Device going down */ 2740 void ipv6_mc_down(struct inet6_dev *idev) 2741 { 2742 struct ifmcaddr6 *i; 2743 2744 mutex_lock(&idev->mc_lock); 2745 /* Withdraw multicast list */ 2746 for_each_mc_mclock(idev, i) 2747 igmp6_group_dropped(i); 2748 mutex_unlock(&idev->mc_lock); 2749 2750 /* Should stop work after group drop. or we will 2751 * start work again in mld_ifc_event() 2752 */ 2753 mld_query_stop_work(idev); 2754 mld_report_stop_work(idev); 2755 2756 mutex_lock(&idev->mc_lock); 2757 mld_ifc_stop_work(idev); 2758 mld_gq_stop_work(idev); 2759 mutex_unlock(&idev->mc_lock); 2760 2761 mld_dad_stop_work(idev); 2762 } 2763 2764 static void ipv6_mc_reset(struct inet6_dev *idev) 2765 { 2766 idev->mc_qrv = sysctl_mld_qrv; 2767 idev->mc_qi = MLD_QI_DEFAULT; 2768 idev->mc_qri = MLD_QRI_DEFAULT; 2769 idev->mc_v1_seen = 0; 2770 idev->mc_maxdelay = unsolicited_report_interval(idev); 2771 } 2772 2773 /* Device going up */ 2774 2775 void ipv6_mc_up(struct inet6_dev *idev) 2776 { 2777 struct ifmcaddr6 *i; 2778 2779 /* Install multicast list, except for all-nodes (already installed) */ 2780 2781 ipv6_mc_reset(idev); 2782 mutex_lock(&idev->mc_lock); 2783 for_each_mc_mclock(idev, i) { 2784 mld_del_delrec(idev, i); 2785 igmp6_group_added(i); 2786 } 2787 mutex_unlock(&idev->mc_lock); 2788 } 2789 2790 /* IPv6 device initialization. */ 2791 2792 void ipv6_mc_init_dev(struct inet6_dev *idev) 2793 { 2794 idev->mc_gq_running = 0; 2795 INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work); 2796 RCU_INIT_POINTER(idev->mc_tomb, NULL); 2797 idev->mc_ifc_count = 0; 2798 INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work); 2799 INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work); 2800 INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work); 2801 INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work); 2802 skb_queue_head_init(&idev->mc_query_queue); 2803 skb_queue_head_init(&idev->mc_report_queue); 2804 spin_lock_init(&idev->mc_query_lock); 2805 spin_lock_init(&idev->mc_report_lock); 2806 mutex_init(&idev->mc_lock); 2807 ipv6_mc_reset(idev); 2808 } 2809 2810 /* 2811 * Device is about to be destroyed: clean up. 2812 */ 2813 2814 void ipv6_mc_destroy_dev(struct inet6_dev *idev) 2815 { 2816 struct ifmcaddr6 *i; 2817 2818 /* Deactivate works */ 2819 ipv6_mc_down(idev); 2820 mutex_lock(&idev->mc_lock); 2821 mld_clear_delrec(idev); 2822 mutex_unlock(&idev->mc_lock); 2823 mld_clear_query(idev); 2824 mld_clear_report(idev); 2825 2826 /* Delete all-nodes address. */ 2827 /* We cannot call ipv6_dev_mc_dec() directly, our caller in 2828 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will 2829 * fail. 2830 */ 2831 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes); 2832 2833 if (idev->cnf.forwarding) 2834 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters); 2835 2836 mutex_lock(&idev->mc_lock); 2837 while ((i = mc_dereference(idev->mc_list, idev))) { 2838 rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev)); 2839 2840 ip6_mc_clear_src(i); 2841 ma_put(i); 2842 } 2843 mutex_unlock(&idev->mc_lock); 2844 } 2845 2846 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev) 2847 { 2848 struct ifmcaddr6 *pmc; 2849 2850 mutex_lock(&idev->mc_lock); 2851 if (mld_in_v1_mode(idev)) { 2852 for_each_mc_mclock(idev, pmc) 2853 igmp6_join_group(pmc); 2854 } else { 2855 mld_send_report(idev, NULL); 2856 } 2857 mutex_unlock(&idev->mc_lock); 2858 } 2859 2860 static int ipv6_mc_netdev_event(struct notifier_block *this, 2861 unsigned long event, 2862 void *ptr) 2863 { 2864 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 2865 struct inet6_dev *idev = __in6_dev_get(dev); 2866 2867 switch (event) { 2868 case NETDEV_RESEND_IGMP: 2869 if (idev) 2870 ipv6_mc_rejoin_groups(idev); 2871 break; 2872 default: 2873 break; 2874 } 2875 2876 return NOTIFY_DONE; 2877 } 2878 2879 static struct notifier_block igmp6_netdev_notifier = { 2880 .notifier_call = ipv6_mc_netdev_event, 2881 }; 2882 2883 #ifdef CONFIG_PROC_FS 2884 struct igmp6_mc_iter_state { 2885 struct seq_net_private p; 2886 struct net_device *dev; 2887 struct inet6_dev *idev; 2888 }; 2889 2890 #define igmp6_mc_seq_private(seq) ((struct igmp6_mc_iter_state *)(seq)->private) 2891 2892 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq) 2893 { 2894 struct ifmcaddr6 *im = NULL; 2895 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq); 2896 struct net *net = seq_file_net(seq); 2897 2898 state->idev = NULL; 2899 for_each_netdev_rcu(net, state->dev) { 2900 struct inet6_dev *idev; 2901 idev = __in6_dev_get(state->dev); 2902 if (!idev) 2903 continue; 2904 2905 im = rcu_dereference(idev->mc_list); 2906 if (im) { 2907 state->idev = idev; 2908 break; 2909 } 2910 } 2911 return im; 2912 } 2913 2914 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im) 2915 { 2916 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq); 2917 2918 im = rcu_dereference(im->next); 2919 while (!im) { 2920 state->dev = next_net_device_rcu(state->dev); 2921 if (!state->dev) { 2922 state->idev = NULL; 2923 break; 2924 } 2925 state->idev = __in6_dev_get(state->dev); 2926 if (!state->idev) 2927 continue; 2928 im = rcu_dereference(state->idev->mc_list); 2929 } 2930 return im; 2931 } 2932 2933 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos) 2934 { 2935 struct ifmcaddr6 *im = igmp6_mc_get_first(seq); 2936 if (im) 2937 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL) 2938 --pos; 2939 return pos ? NULL : im; 2940 } 2941 2942 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos) 2943 __acquires(RCU) 2944 { 2945 rcu_read_lock(); 2946 return igmp6_mc_get_idx(seq, *pos); 2947 } 2948 2949 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2950 { 2951 struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v); 2952 2953 ++*pos; 2954 return im; 2955 } 2956 2957 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v) 2958 __releases(RCU) 2959 { 2960 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq); 2961 2962 if (likely(state->idev)) 2963 state->idev = NULL; 2964 state->dev = NULL; 2965 rcu_read_unlock(); 2966 } 2967 2968 static int igmp6_mc_seq_show(struct seq_file *seq, void *v) 2969 { 2970 struct ifmcaddr6 *im = (struct ifmcaddr6 *)v; 2971 struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq); 2972 2973 seq_printf(seq, 2974 "%-4d %-15s %pi6 %5d %08X %ld\n", 2975 state->dev->ifindex, state->dev->name, 2976 &im->mca_addr, 2977 im->mca_users, im->mca_flags, 2978 (im->mca_flags & MAF_TIMER_RUNNING) ? 2979 jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0); 2980 return 0; 2981 } 2982 2983 static const struct seq_operations igmp6_mc_seq_ops = { 2984 .start = igmp6_mc_seq_start, 2985 .next = igmp6_mc_seq_next, 2986 .stop = igmp6_mc_seq_stop, 2987 .show = igmp6_mc_seq_show, 2988 }; 2989 2990 struct igmp6_mcf_iter_state { 2991 struct seq_net_private p; 2992 struct net_device *dev; 2993 struct inet6_dev *idev; 2994 struct ifmcaddr6 *im; 2995 }; 2996 2997 #define igmp6_mcf_seq_private(seq) ((struct igmp6_mcf_iter_state *)(seq)->private) 2998 2999 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq) 3000 { 3001 struct ip6_sf_list *psf = NULL; 3002 struct ifmcaddr6 *im = NULL; 3003 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq); 3004 struct net *net = seq_file_net(seq); 3005 3006 state->idev = NULL; 3007 state->im = NULL; 3008 for_each_netdev_rcu(net, state->dev) { 3009 struct inet6_dev *idev; 3010 idev = __in6_dev_get(state->dev); 3011 if (unlikely(idev == NULL)) 3012 continue; 3013 3014 im = rcu_dereference(idev->mc_list); 3015 if (likely(im)) { 3016 psf = rcu_dereference(im->mca_sources); 3017 if (likely(psf)) { 3018 state->im = im; 3019 state->idev = idev; 3020 break; 3021 } 3022 } 3023 } 3024 return psf; 3025 } 3026 3027 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf) 3028 { 3029 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq); 3030 3031 psf = rcu_dereference(psf->sf_next); 3032 while (!psf) { 3033 state->im = rcu_dereference(state->im->next); 3034 while (!state->im) { 3035 state->dev = next_net_device_rcu(state->dev); 3036 if (!state->dev) { 3037 state->idev = NULL; 3038 goto out; 3039 } 3040 state->idev = __in6_dev_get(state->dev); 3041 if (!state->idev) 3042 continue; 3043 state->im = rcu_dereference(state->idev->mc_list); 3044 } 3045 psf = rcu_dereference(state->im->mca_sources); 3046 } 3047 out: 3048 return psf; 3049 } 3050 3051 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos) 3052 { 3053 struct ip6_sf_list *psf = igmp6_mcf_get_first(seq); 3054 if (psf) 3055 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL) 3056 --pos; 3057 return pos ? NULL : psf; 3058 } 3059 3060 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos) 3061 __acquires(RCU) 3062 { 3063 rcu_read_lock(); 3064 return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 3065 } 3066 3067 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos) 3068 { 3069 struct ip6_sf_list *psf; 3070 if (v == SEQ_START_TOKEN) 3071 psf = igmp6_mcf_get_first(seq); 3072 else 3073 psf = igmp6_mcf_get_next(seq, v); 3074 ++*pos; 3075 return psf; 3076 } 3077 3078 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v) 3079 __releases(RCU) 3080 { 3081 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq); 3082 3083 if (likely(state->im)) 3084 state->im = NULL; 3085 if (likely(state->idev)) 3086 state->idev = NULL; 3087 3088 state->dev = NULL; 3089 rcu_read_unlock(); 3090 } 3091 3092 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v) 3093 { 3094 struct ip6_sf_list *psf = (struct ip6_sf_list *)v; 3095 struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq); 3096 3097 if (v == SEQ_START_TOKEN) { 3098 seq_puts(seq, "Idx Device Multicast Address Source Address INC EXC\n"); 3099 } else { 3100 seq_printf(seq, 3101 "%3d %6.6s %pi6 %pi6 %6lu %6lu\n", 3102 state->dev->ifindex, state->dev->name, 3103 &state->im->mca_addr, 3104 &psf->sf_addr, 3105 READ_ONCE(psf->sf_count[MCAST_INCLUDE]), 3106 READ_ONCE(psf->sf_count[MCAST_EXCLUDE])); 3107 } 3108 return 0; 3109 } 3110 3111 static const struct seq_operations igmp6_mcf_seq_ops = { 3112 .start = igmp6_mcf_seq_start, 3113 .next = igmp6_mcf_seq_next, 3114 .stop = igmp6_mcf_seq_stop, 3115 .show = igmp6_mcf_seq_show, 3116 }; 3117 3118 static int __net_init igmp6_proc_init(struct net *net) 3119 { 3120 int err; 3121 3122 err = -ENOMEM; 3123 if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops, 3124 sizeof(struct igmp6_mc_iter_state))) 3125 goto out; 3126 if (!proc_create_net("mcfilter6", 0444, net->proc_net, 3127 &igmp6_mcf_seq_ops, 3128 sizeof(struct igmp6_mcf_iter_state))) 3129 goto out_proc_net_igmp6; 3130 3131 err = 0; 3132 out: 3133 return err; 3134 3135 out_proc_net_igmp6: 3136 remove_proc_entry("igmp6", net->proc_net); 3137 goto out; 3138 } 3139 3140 static void __net_exit igmp6_proc_exit(struct net *net) 3141 { 3142 remove_proc_entry("mcfilter6", net->proc_net); 3143 remove_proc_entry("igmp6", net->proc_net); 3144 } 3145 #else 3146 static inline int igmp6_proc_init(struct net *net) 3147 { 3148 return 0; 3149 } 3150 static inline void igmp6_proc_exit(struct net *net) 3151 { 3152 } 3153 #endif 3154 3155 static int __net_init igmp6_net_init(struct net *net) 3156 { 3157 int err; 3158 3159 err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6, 3160 SOCK_RAW, IPPROTO_ICMPV6, net); 3161 if (err < 0) { 3162 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n", 3163 err); 3164 goto out; 3165 } 3166 3167 inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1; 3168 net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL; 3169 3170 err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6, 3171 SOCK_RAW, IPPROTO_ICMPV6, net); 3172 if (err < 0) { 3173 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n", 3174 err); 3175 goto out_sock_create; 3176 } 3177 3178 err = igmp6_proc_init(net); 3179 if (err) 3180 goto out_sock_create_autojoin; 3181 3182 return 0; 3183 3184 out_sock_create_autojoin: 3185 inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk); 3186 out_sock_create: 3187 inet_ctl_sock_destroy(net->ipv6.igmp_sk); 3188 out: 3189 return err; 3190 } 3191 3192 static void __net_exit igmp6_net_exit(struct net *net) 3193 { 3194 inet_ctl_sock_destroy(net->ipv6.igmp_sk); 3195 inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk); 3196 igmp6_proc_exit(net); 3197 } 3198 3199 static struct pernet_operations igmp6_net_ops = { 3200 .init = igmp6_net_init, 3201 .exit = igmp6_net_exit, 3202 }; 3203 3204 int __init igmp6_init(void) 3205 { 3206 int err; 3207 3208 err = register_pernet_subsys(&igmp6_net_ops); 3209 if (err) 3210 return err; 3211 3212 mld_wq = create_workqueue("mld"); 3213 if (!mld_wq) { 3214 unregister_pernet_subsys(&igmp6_net_ops); 3215 return -ENOMEM; 3216 } 3217 3218 return err; 3219 } 3220 3221 int __init igmp6_late_init(void) 3222 { 3223 return register_netdevice_notifier(&igmp6_netdev_notifier); 3224 } 3225 3226 void igmp6_cleanup(void) 3227 { 3228 unregister_pernet_subsys(&igmp6_net_ops); 3229 destroy_workqueue(mld_wq); 3230 } 3231 3232 void igmp6_late_cleanup(void) 3233 { 3234 unregister_netdevice_notifier(&igmp6_netdev_notifier); 3235 } 3236