1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Linux IPv6 multicast routing support for BSD pim6sd 4 * Based on net/ipv4/ipmr.c. 5 * 6 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr> 7 * LSIIT Laboratory, Strasbourg, France 8 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com> 9 * 6WIND, Paris, France 10 * Copyright (C)2007,2008 USAGI/WIDE Project 11 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> 12 */ 13 14 #include <linux/uaccess.h> 15 #include <linux/types.h> 16 #include <linux/sched.h> 17 #include <linux/errno.h> 18 #include <linux/mm.h> 19 #include <linux/kernel.h> 20 #include <linux/fcntl.h> 21 #include <linux/stat.h> 22 #include <linux/socket.h> 23 #include <linux/inet.h> 24 #include <linux/netdevice.h> 25 #include <linux/inetdevice.h> 26 #include <linux/proc_fs.h> 27 #include <linux/seq_file.h> 28 #include <linux/init.h> 29 #include <linux/compat.h> 30 #include <linux/rhashtable.h> 31 #include <net/protocol.h> 32 #include <linux/skbuff.h> 33 #include <net/raw.h> 34 #include <linux/notifier.h> 35 #include <linux/if_arp.h> 36 #include <net/checksum.h> 37 #include <net/netlink.h> 38 #include <net/fib_rules.h> 39 40 #include <net/ipv6.h> 41 #include <net/ip6_route.h> 42 #include <linux/mroute6.h> 43 #include <linux/pim.h> 44 #include <net/addrconf.h> 45 #include <linux/netfilter_ipv6.h> 46 #include <linux/export.h> 47 #include <net/ip6_checksum.h> 48 #include <linux/netconf.h> 49 #include <net/ip_tunnels.h> 50 51 #include <linux/nospec.h> 52 53 struct ip6mr_rule { 54 struct fib_rule common; 55 }; 56 57 struct ip6mr_result { 58 struct mr_table *mrt; 59 }; 60 61 /* Big lock, protecting vif table, mrt cache and mroute socket state. 62 Note that the changes are semaphored via rtnl_lock. 63 */ 64 65 static DEFINE_RWLOCK(mrt_lock); 66 67 /* Multicast router control variables */ 68 69 /* Special spinlock for queue of unresolved entries */ 70 static DEFINE_SPINLOCK(mfc_unres_lock); 71 72 /* We return to original Alan's scheme. Hash table of resolved 73 entries is changed only in process context and protected 74 with weak lock mrt_lock. Queue of unresolved entries is protected 75 with strong spinlock mfc_unres_lock. 76 77 In this case data path is free of exclusive locks at all. 78 */ 79 80 static struct kmem_cache *mrt_cachep __read_mostly; 81 82 static struct mr_table *ip6mr_new_table(struct net *net, u32 id); 83 static void ip6mr_free_table(struct mr_table *mrt); 84 85 static void ip6_mr_forward(struct net *net, struct mr_table *mrt, 86 struct net_device *dev, struct sk_buff *skb, 87 struct mfc6_cache *cache); 88 static int ip6mr_cache_report(struct mr_table *mrt, struct sk_buff *pkt, 89 mifi_t mifi, int assert); 90 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc, 91 int cmd); 92 static void mrt6msg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt); 93 static int ip6mr_rtm_dumproute(struct sk_buff *skb, 94 struct netlink_callback *cb); 95 static void mroute_clean_tables(struct mr_table *mrt, int flags); 96 static void ipmr_expire_process(struct timer_list *t); 97 98 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 99 #define ip6mr_for_each_table(mrt, net) \ 100 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list) 101 102 static struct mr_table *ip6mr_mr_table_iter(struct net *net, 103 struct mr_table *mrt) 104 { 105 struct mr_table *ret; 106 107 if (!mrt) 108 ret = list_entry_rcu(net->ipv6.mr6_tables.next, 109 struct mr_table, list); 110 else 111 ret = list_entry_rcu(mrt->list.next, 112 struct mr_table, list); 113 114 if (&ret->list == &net->ipv6.mr6_tables) 115 return NULL; 116 return ret; 117 } 118 119 static struct mr_table *ip6mr_get_table(struct net *net, u32 id) 120 { 121 struct mr_table *mrt; 122 123 ip6mr_for_each_table(mrt, net) { 124 if (mrt->id == id) 125 return mrt; 126 } 127 return NULL; 128 } 129 130 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 131 struct mr_table **mrt) 132 { 133 int err; 134 struct ip6mr_result res; 135 struct fib_lookup_arg arg = { 136 .result = &res, 137 .flags = FIB_LOOKUP_NOREF, 138 }; 139 140 /* update flow if oif or iif point to device enslaved to l3mdev */ 141 l3mdev_update_flow(net, flowi6_to_flowi(flp6)); 142 143 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, 144 flowi6_to_flowi(flp6), 0, &arg); 145 if (err < 0) 146 return err; 147 *mrt = res.mrt; 148 return 0; 149 } 150 151 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp, 152 int flags, struct fib_lookup_arg *arg) 153 { 154 struct ip6mr_result *res = arg->result; 155 struct mr_table *mrt; 156 157 switch (rule->action) { 158 case FR_ACT_TO_TBL: 159 break; 160 case FR_ACT_UNREACHABLE: 161 return -ENETUNREACH; 162 case FR_ACT_PROHIBIT: 163 return -EACCES; 164 case FR_ACT_BLACKHOLE: 165 default: 166 return -EINVAL; 167 } 168 169 arg->table = fib_rule_get_table(rule, arg); 170 171 mrt = ip6mr_get_table(rule->fr_net, arg->table); 172 if (!mrt) 173 return -EAGAIN; 174 res->mrt = mrt; 175 return 0; 176 } 177 178 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags) 179 { 180 return 1; 181 } 182 183 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = { 184 FRA_GENERIC_POLICY, 185 }; 186 187 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 188 struct fib_rule_hdr *frh, struct nlattr **tb, 189 struct netlink_ext_ack *extack) 190 { 191 return 0; 192 } 193 194 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 195 struct nlattr **tb) 196 { 197 return 1; 198 } 199 200 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 201 struct fib_rule_hdr *frh) 202 { 203 frh->dst_len = 0; 204 frh->src_len = 0; 205 frh->tos = 0; 206 return 0; 207 } 208 209 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = { 210 .family = RTNL_FAMILY_IP6MR, 211 .rule_size = sizeof(struct ip6mr_rule), 212 .addr_size = sizeof(struct in6_addr), 213 .action = ip6mr_rule_action, 214 .match = ip6mr_rule_match, 215 .configure = ip6mr_rule_configure, 216 .compare = ip6mr_rule_compare, 217 .fill = ip6mr_rule_fill, 218 .nlgroup = RTNLGRP_IPV6_RULE, 219 .policy = ip6mr_rule_policy, 220 .owner = THIS_MODULE, 221 }; 222 223 static int __net_init ip6mr_rules_init(struct net *net) 224 { 225 struct fib_rules_ops *ops; 226 struct mr_table *mrt; 227 int err; 228 229 ops = fib_rules_register(&ip6mr_rules_ops_template, net); 230 if (IS_ERR(ops)) 231 return PTR_ERR(ops); 232 233 INIT_LIST_HEAD(&net->ipv6.mr6_tables); 234 235 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT); 236 if (IS_ERR(mrt)) { 237 err = PTR_ERR(mrt); 238 goto err1; 239 } 240 241 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0); 242 if (err < 0) 243 goto err2; 244 245 net->ipv6.mr6_rules_ops = ops; 246 return 0; 247 248 err2: 249 ip6mr_free_table(mrt); 250 err1: 251 fib_rules_unregister(ops); 252 return err; 253 } 254 255 static void __net_exit ip6mr_rules_exit(struct net *net) 256 { 257 struct mr_table *mrt, *next; 258 259 rtnl_lock(); 260 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) { 261 list_del(&mrt->list); 262 ip6mr_free_table(mrt); 263 } 264 fib_rules_unregister(net->ipv6.mr6_rules_ops); 265 rtnl_unlock(); 266 } 267 268 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb) 269 { 270 return fib_rules_dump(net, nb, RTNL_FAMILY_IP6MR); 271 } 272 273 static unsigned int ip6mr_rules_seq_read(struct net *net) 274 { 275 return fib_rules_seq_read(net, RTNL_FAMILY_IP6MR); 276 } 277 278 bool ip6mr_rule_default(const struct fib_rule *rule) 279 { 280 return fib_rule_matchall(rule) && rule->action == FR_ACT_TO_TBL && 281 rule->table == RT6_TABLE_DFLT && !rule->l3mdev; 282 } 283 EXPORT_SYMBOL(ip6mr_rule_default); 284 #else 285 #define ip6mr_for_each_table(mrt, net) \ 286 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL) 287 288 static struct mr_table *ip6mr_mr_table_iter(struct net *net, 289 struct mr_table *mrt) 290 { 291 if (!mrt) 292 return net->ipv6.mrt6; 293 return NULL; 294 } 295 296 static struct mr_table *ip6mr_get_table(struct net *net, u32 id) 297 { 298 return net->ipv6.mrt6; 299 } 300 301 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 302 struct mr_table **mrt) 303 { 304 *mrt = net->ipv6.mrt6; 305 return 0; 306 } 307 308 static int __net_init ip6mr_rules_init(struct net *net) 309 { 310 struct mr_table *mrt; 311 312 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT); 313 if (IS_ERR(mrt)) 314 return PTR_ERR(mrt); 315 net->ipv6.mrt6 = mrt; 316 return 0; 317 } 318 319 static void __net_exit ip6mr_rules_exit(struct net *net) 320 { 321 rtnl_lock(); 322 ip6mr_free_table(net->ipv6.mrt6); 323 net->ipv6.mrt6 = NULL; 324 rtnl_unlock(); 325 } 326 327 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb) 328 { 329 return 0; 330 } 331 332 static unsigned int ip6mr_rules_seq_read(struct net *net) 333 { 334 return 0; 335 } 336 #endif 337 338 static int ip6mr_hash_cmp(struct rhashtable_compare_arg *arg, 339 const void *ptr) 340 { 341 const struct mfc6_cache_cmp_arg *cmparg = arg->key; 342 struct mfc6_cache *c = (struct mfc6_cache *)ptr; 343 344 return !ipv6_addr_equal(&c->mf6c_mcastgrp, &cmparg->mf6c_mcastgrp) || 345 !ipv6_addr_equal(&c->mf6c_origin, &cmparg->mf6c_origin); 346 } 347 348 static const struct rhashtable_params ip6mr_rht_params = { 349 .head_offset = offsetof(struct mr_mfc, mnode), 350 .key_offset = offsetof(struct mfc6_cache, cmparg), 351 .key_len = sizeof(struct mfc6_cache_cmp_arg), 352 .nelem_hint = 3, 353 .obj_cmpfn = ip6mr_hash_cmp, 354 .automatic_shrinking = true, 355 }; 356 357 static void ip6mr_new_table_set(struct mr_table *mrt, 358 struct net *net) 359 { 360 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 361 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables); 362 #endif 363 } 364 365 static struct mfc6_cache_cmp_arg ip6mr_mr_table_ops_cmparg_any = { 366 .mf6c_origin = IN6ADDR_ANY_INIT, 367 .mf6c_mcastgrp = IN6ADDR_ANY_INIT, 368 }; 369 370 static struct mr_table_ops ip6mr_mr_table_ops = { 371 .rht_params = &ip6mr_rht_params, 372 .cmparg_any = &ip6mr_mr_table_ops_cmparg_any, 373 }; 374 375 static struct mr_table *ip6mr_new_table(struct net *net, u32 id) 376 { 377 struct mr_table *mrt; 378 379 mrt = ip6mr_get_table(net, id); 380 if (mrt) 381 return mrt; 382 383 return mr_table_alloc(net, id, &ip6mr_mr_table_ops, 384 ipmr_expire_process, ip6mr_new_table_set); 385 } 386 387 static void ip6mr_free_table(struct mr_table *mrt) 388 { 389 del_timer_sync(&mrt->ipmr_expire_timer); 390 mroute_clean_tables(mrt, MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC | 391 MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC); 392 rhltable_destroy(&mrt->mfc_hash); 393 kfree(mrt); 394 } 395 396 #ifdef CONFIG_PROC_FS 397 /* The /proc interfaces to multicast routing 398 * /proc/ip6_mr_cache /proc/ip6_mr_vif 399 */ 400 401 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) 402 __acquires(mrt_lock) 403 { 404 struct mr_vif_iter *iter = seq->private; 405 struct net *net = seq_file_net(seq); 406 struct mr_table *mrt; 407 408 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 409 if (!mrt) 410 return ERR_PTR(-ENOENT); 411 412 iter->mrt = mrt; 413 414 read_lock(&mrt_lock); 415 return mr_vif_seq_start(seq, pos); 416 } 417 418 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) 419 __releases(mrt_lock) 420 { 421 read_unlock(&mrt_lock); 422 } 423 424 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) 425 { 426 struct mr_vif_iter *iter = seq->private; 427 struct mr_table *mrt = iter->mrt; 428 429 if (v == SEQ_START_TOKEN) { 430 seq_puts(seq, 431 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); 432 } else { 433 const struct vif_device *vif = v; 434 const char *name = vif->dev ? vif->dev->name : "none"; 435 436 seq_printf(seq, 437 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", 438 vif - mrt->vif_table, 439 name, vif->bytes_in, vif->pkt_in, 440 vif->bytes_out, vif->pkt_out, 441 vif->flags); 442 } 443 return 0; 444 } 445 446 static const struct seq_operations ip6mr_vif_seq_ops = { 447 .start = ip6mr_vif_seq_start, 448 .next = mr_vif_seq_next, 449 .stop = ip6mr_vif_seq_stop, 450 .show = ip6mr_vif_seq_show, 451 }; 452 453 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 454 { 455 struct net *net = seq_file_net(seq); 456 struct mr_table *mrt; 457 458 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 459 if (!mrt) 460 return ERR_PTR(-ENOENT); 461 462 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock); 463 } 464 465 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 466 { 467 int n; 468 469 if (v == SEQ_START_TOKEN) { 470 seq_puts(seq, 471 "Group " 472 "Origin " 473 "Iif Pkts Bytes Wrong Oifs\n"); 474 } else { 475 const struct mfc6_cache *mfc = v; 476 const struct mr_mfc_iter *it = seq->private; 477 struct mr_table *mrt = it->mrt; 478 479 seq_printf(seq, "%pI6 %pI6 %-3hd", 480 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, 481 mfc->_c.mfc_parent); 482 483 if (it->cache != &mrt->mfc_unres_queue) { 484 seq_printf(seq, " %8lu %8lu %8lu", 485 mfc->_c.mfc_un.res.pkt, 486 mfc->_c.mfc_un.res.bytes, 487 mfc->_c.mfc_un.res.wrong_if); 488 for (n = mfc->_c.mfc_un.res.minvif; 489 n < mfc->_c.mfc_un.res.maxvif; n++) { 490 if (VIF_EXISTS(mrt, n) && 491 mfc->_c.mfc_un.res.ttls[n] < 255) 492 seq_printf(seq, 493 " %2d:%-3d", n, 494 mfc->_c.mfc_un.res.ttls[n]); 495 } 496 } else { 497 /* unresolved mfc_caches don't contain 498 * pkt, bytes and wrong_if values 499 */ 500 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 501 } 502 seq_putc(seq, '\n'); 503 } 504 return 0; 505 } 506 507 static const struct seq_operations ipmr_mfc_seq_ops = { 508 .start = ipmr_mfc_seq_start, 509 .next = mr_mfc_seq_next, 510 .stop = mr_mfc_seq_stop, 511 .show = ipmr_mfc_seq_show, 512 }; 513 #endif 514 515 #ifdef CONFIG_IPV6_PIMSM_V2 516 517 static int pim6_rcv(struct sk_buff *skb) 518 { 519 struct pimreghdr *pim; 520 struct ipv6hdr *encap; 521 struct net_device *reg_dev = NULL; 522 struct net *net = dev_net(skb->dev); 523 struct mr_table *mrt; 524 struct flowi6 fl6 = { 525 .flowi6_iif = skb->dev->ifindex, 526 .flowi6_mark = skb->mark, 527 }; 528 int reg_vif_num; 529 530 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) 531 goto drop; 532 533 pim = (struct pimreghdr *)skb_transport_header(skb); 534 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) || 535 (pim->flags & PIM_NULL_REGISTER) || 536 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 537 sizeof(*pim), IPPROTO_PIM, 538 csum_partial((void *)pim, sizeof(*pim), 0)) && 539 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 540 goto drop; 541 542 /* check if the inner packet is destined to mcast group */ 543 encap = (struct ipv6hdr *)(skb_transport_header(skb) + 544 sizeof(*pim)); 545 546 if (!ipv6_addr_is_multicast(&encap->daddr) || 547 encap->payload_len == 0 || 548 ntohs(encap->payload_len) + sizeof(*pim) > skb->len) 549 goto drop; 550 551 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 552 goto drop; 553 reg_vif_num = mrt->mroute_reg_vif_num; 554 555 read_lock(&mrt_lock); 556 if (reg_vif_num >= 0) 557 reg_dev = mrt->vif_table[reg_vif_num].dev; 558 if (reg_dev) 559 dev_hold(reg_dev); 560 read_unlock(&mrt_lock); 561 562 if (!reg_dev) 563 goto drop; 564 565 skb->mac_header = skb->network_header; 566 skb_pull(skb, (u8 *)encap - skb->data); 567 skb_reset_network_header(skb); 568 skb->protocol = htons(ETH_P_IPV6); 569 skb->ip_summed = CHECKSUM_NONE; 570 571 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 572 573 netif_rx(skb); 574 575 dev_put(reg_dev); 576 return 0; 577 drop: 578 kfree_skb(skb); 579 return 0; 580 } 581 582 static const struct inet6_protocol pim6_protocol = { 583 .handler = pim6_rcv, 584 }; 585 586 /* Service routines creating virtual interfaces: PIMREG */ 587 588 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, 589 struct net_device *dev) 590 { 591 struct net *net = dev_net(dev); 592 struct mr_table *mrt; 593 struct flowi6 fl6 = { 594 .flowi6_oif = dev->ifindex, 595 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 596 .flowi6_mark = skb->mark, 597 }; 598 599 if (!pskb_inet_may_pull(skb)) 600 goto tx_err; 601 602 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 603 goto tx_err; 604 605 read_lock(&mrt_lock); 606 dev->stats.tx_bytes += skb->len; 607 dev->stats.tx_packets++; 608 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT); 609 read_unlock(&mrt_lock); 610 kfree_skb(skb); 611 return NETDEV_TX_OK; 612 613 tx_err: 614 dev->stats.tx_errors++; 615 kfree_skb(skb); 616 return NETDEV_TX_OK; 617 } 618 619 static int reg_vif_get_iflink(const struct net_device *dev) 620 { 621 return 0; 622 } 623 624 static const struct net_device_ops reg_vif_netdev_ops = { 625 .ndo_start_xmit = reg_vif_xmit, 626 .ndo_get_iflink = reg_vif_get_iflink, 627 }; 628 629 static void reg_vif_setup(struct net_device *dev) 630 { 631 dev->type = ARPHRD_PIMREG; 632 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; 633 dev->flags = IFF_NOARP; 634 dev->netdev_ops = ®_vif_netdev_ops; 635 dev->needs_free_netdev = true; 636 dev->features |= NETIF_F_NETNS_LOCAL; 637 } 638 639 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr_table *mrt) 640 { 641 struct net_device *dev; 642 char name[IFNAMSIZ]; 643 644 if (mrt->id == RT6_TABLE_DFLT) 645 sprintf(name, "pim6reg"); 646 else 647 sprintf(name, "pim6reg%u", mrt->id); 648 649 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 650 if (!dev) 651 return NULL; 652 653 dev_net_set(dev, net); 654 655 if (register_netdevice(dev)) { 656 free_netdev(dev); 657 return NULL; 658 } 659 660 if (dev_open(dev, NULL)) 661 goto failure; 662 663 dev_hold(dev); 664 return dev; 665 666 failure: 667 unregister_netdevice(dev); 668 return NULL; 669 } 670 #endif 671 672 static int call_ip6mr_vif_entry_notifiers(struct net *net, 673 enum fib_event_type event_type, 674 struct vif_device *vif, 675 mifi_t vif_index, u32 tb_id) 676 { 677 return mr_call_vif_notifiers(net, RTNL_FAMILY_IP6MR, event_type, 678 vif, vif_index, tb_id, 679 &net->ipv6.ipmr_seq); 680 } 681 682 static int call_ip6mr_mfc_entry_notifiers(struct net *net, 683 enum fib_event_type event_type, 684 struct mfc6_cache *mfc, u32 tb_id) 685 { 686 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IP6MR, event_type, 687 &mfc->_c, tb_id, &net->ipv6.ipmr_seq); 688 } 689 690 /* Delete a VIF entry */ 691 static int mif6_delete(struct mr_table *mrt, int vifi, int notify, 692 struct list_head *head) 693 { 694 struct vif_device *v; 695 struct net_device *dev; 696 struct inet6_dev *in6_dev; 697 698 if (vifi < 0 || vifi >= mrt->maxvif) 699 return -EADDRNOTAVAIL; 700 701 v = &mrt->vif_table[vifi]; 702 703 if (VIF_EXISTS(mrt, vifi)) 704 call_ip6mr_vif_entry_notifiers(read_pnet(&mrt->net), 705 FIB_EVENT_VIF_DEL, v, vifi, 706 mrt->id); 707 708 write_lock_bh(&mrt_lock); 709 dev = v->dev; 710 v->dev = NULL; 711 712 if (!dev) { 713 write_unlock_bh(&mrt_lock); 714 return -EADDRNOTAVAIL; 715 } 716 717 #ifdef CONFIG_IPV6_PIMSM_V2 718 if (vifi == mrt->mroute_reg_vif_num) 719 mrt->mroute_reg_vif_num = -1; 720 #endif 721 722 if (vifi + 1 == mrt->maxvif) { 723 int tmp; 724 for (tmp = vifi - 1; tmp >= 0; tmp--) { 725 if (VIF_EXISTS(mrt, tmp)) 726 break; 727 } 728 mrt->maxvif = tmp + 1; 729 } 730 731 write_unlock_bh(&mrt_lock); 732 733 dev_set_allmulti(dev, -1); 734 735 in6_dev = __in6_dev_get(dev); 736 if (in6_dev) { 737 in6_dev->cnf.mc_forwarding--; 738 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 739 NETCONFA_MC_FORWARDING, 740 dev->ifindex, &in6_dev->cnf); 741 } 742 743 if ((v->flags & MIFF_REGISTER) && !notify) 744 unregister_netdevice_queue(dev, head); 745 746 dev_put(dev); 747 return 0; 748 } 749 750 static inline void ip6mr_cache_free_rcu(struct rcu_head *head) 751 { 752 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu); 753 754 kmem_cache_free(mrt_cachep, (struct mfc6_cache *)c); 755 } 756 757 static inline void ip6mr_cache_free(struct mfc6_cache *c) 758 { 759 call_rcu(&c->_c.rcu, ip6mr_cache_free_rcu); 760 } 761 762 /* Destroy an unresolved cache entry, killing queued skbs 763 and reporting error to netlink readers. 764 */ 765 766 static void ip6mr_destroy_unres(struct mr_table *mrt, struct mfc6_cache *c) 767 { 768 struct net *net = read_pnet(&mrt->net); 769 struct sk_buff *skb; 770 771 atomic_dec(&mrt->cache_resolve_queue_len); 772 773 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved)) != NULL) { 774 if (ipv6_hdr(skb)->version == 0) { 775 struct nlmsghdr *nlh = skb_pull(skb, 776 sizeof(struct ipv6hdr)); 777 nlh->nlmsg_type = NLMSG_ERROR; 778 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 779 skb_trim(skb, nlh->nlmsg_len); 780 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT; 781 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 782 } else 783 kfree_skb(skb); 784 } 785 786 ip6mr_cache_free(c); 787 } 788 789 790 /* Timer process for all the unresolved queue. */ 791 792 static void ipmr_do_expire_process(struct mr_table *mrt) 793 { 794 unsigned long now = jiffies; 795 unsigned long expires = 10 * HZ; 796 struct mr_mfc *c, *next; 797 798 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 799 if (time_after(c->mfc_un.unres.expires, now)) { 800 /* not yet... */ 801 unsigned long interval = c->mfc_un.unres.expires - now; 802 if (interval < expires) 803 expires = interval; 804 continue; 805 } 806 807 list_del(&c->list); 808 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE); 809 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c); 810 } 811 812 if (!list_empty(&mrt->mfc_unres_queue)) 813 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 814 } 815 816 static void ipmr_expire_process(struct timer_list *t) 817 { 818 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer); 819 820 if (!spin_trylock(&mfc_unres_lock)) { 821 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 822 return; 823 } 824 825 if (!list_empty(&mrt->mfc_unres_queue)) 826 ipmr_do_expire_process(mrt); 827 828 spin_unlock(&mfc_unres_lock); 829 } 830 831 /* Fill oifs list. It is called under write locked mrt_lock. */ 832 833 static void ip6mr_update_thresholds(struct mr_table *mrt, 834 struct mr_mfc *cache, 835 unsigned char *ttls) 836 { 837 int vifi; 838 839 cache->mfc_un.res.minvif = MAXMIFS; 840 cache->mfc_un.res.maxvif = 0; 841 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 842 843 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 844 if (VIF_EXISTS(mrt, vifi) && 845 ttls[vifi] && ttls[vifi] < 255) { 846 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 847 if (cache->mfc_un.res.minvif > vifi) 848 cache->mfc_un.res.minvif = vifi; 849 if (cache->mfc_un.res.maxvif <= vifi) 850 cache->mfc_un.res.maxvif = vifi + 1; 851 } 852 } 853 cache->mfc_un.res.lastuse = jiffies; 854 } 855 856 static int mif6_add(struct net *net, struct mr_table *mrt, 857 struct mif6ctl *vifc, int mrtsock) 858 { 859 int vifi = vifc->mif6c_mifi; 860 struct vif_device *v = &mrt->vif_table[vifi]; 861 struct net_device *dev; 862 struct inet6_dev *in6_dev; 863 int err; 864 865 /* Is vif busy ? */ 866 if (VIF_EXISTS(mrt, vifi)) 867 return -EADDRINUSE; 868 869 switch (vifc->mif6c_flags) { 870 #ifdef CONFIG_IPV6_PIMSM_V2 871 case MIFF_REGISTER: 872 /* 873 * Special Purpose VIF in PIM 874 * All the packets will be sent to the daemon 875 */ 876 if (mrt->mroute_reg_vif_num >= 0) 877 return -EADDRINUSE; 878 dev = ip6mr_reg_vif(net, mrt); 879 if (!dev) 880 return -ENOBUFS; 881 err = dev_set_allmulti(dev, 1); 882 if (err) { 883 unregister_netdevice(dev); 884 dev_put(dev); 885 return err; 886 } 887 break; 888 #endif 889 case 0: 890 dev = dev_get_by_index(net, vifc->mif6c_pifi); 891 if (!dev) 892 return -EADDRNOTAVAIL; 893 err = dev_set_allmulti(dev, 1); 894 if (err) { 895 dev_put(dev); 896 return err; 897 } 898 break; 899 default: 900 return -EINVAL; 901 } 902 903 in6_dev = __in6_dev_get(dev); 904 if (in6_dev) { 905 in6_dev->cnf.mc_forwarding++; 906 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 907 NETCONFA_MC_FORWARDING, 908 dev->ifindex, &in6_dev->cnf); 909 } 910 911 /* Fill in the VIF structures */ 912 vif_device_init(v, dev, vifc->vifc_rate_limit, vifc->vifc_threshold, 913 vifc->mif6c_flags | (!mrtsock ? VIFF_STATIC : 0), 914 MIFF_REGISTER); 915 916 /* And finish update writing critical data */ 917 write_lock_bh(&mrt_lock); 918 v->dev = dev; 919 #ifdef CONFIG_IPV6_PIMSM_V2 920 if (v->flags & MIFF_REGISTER) 921 mrt->mroute_reg_vif_num = vifi; 922 #endif 923 if (vifi + 1 > mrt->maxvif) 924 mrt->maxvif = vifi + 1; 925 write_unlock_bh(&mrt_lock); 926 call_ip6mr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, 927 v, vifi, mrt->id); 928 return 0; 929 } 930 931 static struct mfc6_cache *ip6mr_cache_find(struct mr_table *mrt, 932 const struct in6_addr *origin, 933 const struct in6_addr *mcastgrp) 934 { 935 struct mfc6_cache_cmp_arg arg = { 936 .mf6c_origin = *origin, 937 .mf6c_mcastgrp = *mcastgrp, 938 }; 939 940 return mr_mfc_find(mrt, &arg); 941 } 942 943 /* Look for a (*,G) entry */ 944 static struct mfc6_cache *ip6mr_cache_find_any(struct mr_table *mrt, 945 struct in6_addr *mcastgrp, 946 mifi_t mifi) 947 { 948 struct mfc6_cache_cmp_arg arg = { 949 .mf6c_origin = in6addr_any, 950 .mf6c_mcastgrp = *mcastgrp, 951 }; 952 953 if (ipv6_addr_any(mcastgrp)) 954 return mr_mfc_find_any_parent(mrt, mifi); 955 return mr_mfc_find_any(mrt, mifi, &arg); 956 } 957 958 /* Look for a (S,G,iif) entry if parent != -1 */ 959 static struct mfc6_cache * 960 ip6mr_cache_find_parent(struct mr_table *mrt, 961 const struct in6_addr *origin, 962 const struct in6_addr *mcastgrp, 963 int parent) 964 { 965 struct mfc6_cache_cmp_arg arg = { 966 .mf6c_origin = *origin, 967 .mf6c_mcastgrp = *mcastgrp, 968 }; 969 970 return mr_mfc_find_parent(mrt, &arg, parent); 971 } 972 973 /* Allocate a multicast cache entry */ 974 static struct mfc6_cache *ip6mr_cache_alloc(void) 975 { 976 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 977 if (!c) 978 return NULL; 979 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 980 c->_c.mfc_un.res.minvif = MAXMIFS; 981 c->_c.free = ip6mr_cache_free_rcu; 982 refcount_set(&c->_c.mfc_un.res.refcount, 1); 983 return c; 984 } 985 986 static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 987 { 988 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 989 if (!c) 990 return NULL; 991 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved); 992 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ; 993 return c; 994 } 995 996 /* 997 * A cache entry has gone into a resolved state from queued 998 */ 999 1000 static void ip6mr_cache_resolve(struct net *net, struct mr_table *mrt, 1001 struct mfc6_cache *uc, struct mfc6_cache *c) 1002 { 1003 struct sk_buff *skb; 1004 1005 /* 1006 * Play the pending entries through our router 1007 */ 1008 1009 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) { 1010 if (ipv6_hdr(skb)->version == 0) { 1011 struct nlmsghdr *nlh = skb_pull(skb, 1012 sizeof(struct ipv6hdr)); 1013 1014 if (mr_fill_mroute(mrt, skb, &c->_c, 1015 nlmsg_data(nlh)) > 0) { 1016 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1017 } else { 1018 nlh->nlmsg_type = NLMSG_ERROR; 1019 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1020 skb_trim(skb, nlh->nlmsg_len); 1021 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE; 1022 } 1023 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1024 } else 1025 ip6_mr_forward(net, mrt, skb->dev, skb, c); 1026 } 1027 } 1028 1029 /* 1030 * Bounce a cache query up to pim6sd and netlink. 1031 * 1032 * Called under mrt_lock. 1033 */ 1034 1035 static int ip6mr_cache_report(struct mr_table *mrt, struct sk_buff *pkt, 1036 mifi_t mifi, int assert) 1037 { 1038 struct sock *mroute6_sk; 1039 struct sk_buff *skb; 1040 struct mrt6msg *msg; 1041 int ret; 1042 1043 #ifdef CONFIG_IPV6_PIMSM_V2 1044 if (assert == MRT6MSG_WHOLEPKT) 1045 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1046 +sizeof(*msg)); 1047 else 1048 #endif 1049 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1050 1051 if (!skb) 1052 return -ENOBUFS; 1053 1054 /* I suppose that internal messages 1055 * do not require checksums */ 1056 1057 skb->ip_summed = CHECKSUM_UNNECESSARY; 1058 1059 #ifdef CONFIG_IPV6_PIMSM_V2 1060 if (assert == MRT6MSG_WHOLEPKT) { 1061 /* Ugly, but we have no choice with this interface. 1062 Duplicate old header, fix length etc. 1063 And all this only to mangle msg->im6_msgtype and 1064 to set msg->im6_mbz to "mbz" :-) 1065 */ 1066 skb_push(skb, -skb_network_offset(pkt)); 1067 1068 skb_push(skb, sizeof(*msg)); 1069 skb_reset_transport_header(skb); 1070 msg = (struct mrt6msg *)skb_transport_header(skb); 1071 msg->im6_mbz = 0; 1072 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1073 msg->im6_mif = mrt->mroute_reg_vif_num; 1074 msg->im6_pad = 0; 1075 msg->im6_src = ipv6_hdr(pkt)->saddr; 1076 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1077 1078 skb->ip_summed = CHECKSUM_UNNECESSARY; 1079 } else 1080 #endif 1081 { 1082 /* 1083 * Copy the IP header 1084 */ 1085 1086 skb_put(skb, sizeof(struct ipv6hdr)); 1087 skb_reset_network_header(skb); 1088 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1089 1090 /* 1091 * Add our header 1092 */ 1093 skb_put(skb, sizeof(*msg)); 1094 skb_reset_transport_header(skb); 1095 msg = (struct mrt6msg *)skb_transport_header(skb); 1096 1097 msg->im6_mbz = 0; 1098 msg->im6_msgtype = assert; 1099 msg->im6_mif = mifi; 1100 msg->im6_pad = 0; 1101 msg->im6_src = ipv6_hdr(pkt)->saddr; 1102 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1103 1104 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1105 skb->ip_summed = CHECKSUM_UNNECESSARY; 1106 } 1107 1108 rcu_read_lock(); 1109 mroute6_sk = rcu_dereference(mrt->mroute_sk); 1110 if (!mroute6_sk) { 1111 rcu_read_unlock(); 1112 kfree_skb(skb); 1113 return -EINVAL; 1114 } 1115 1116 mrt6msg_netlink_event(mrt, skb); 1117 1118 /* Deliver to user space multicast routing algorithms */ 1119 ret = sock_queue_rcv_skb(mroute6_sk, skb); 1120 rcu_read_unlock(); 1121 if (ret < 0) { 1122 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1123 kfree_skb(skb); 1124 } 1125 1126 return ret; 1127 } 1128 1129 /* Queue a packet for resolution. It gets locked cache entry! */ 1130 static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi, 1131 struct sk_buff *skb, struct net_device *dev) 1132 { 1133 struct mfc6_cache *c; 1134 bool found = false; 1135 int err; 1136 1137 spin_lock_bh(&mfc_unres_lock); 1138 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) { 1139 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1140 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1141 found = true; 1142 break; 1143 } 1144 } 1145 1146 if (!found) { 1147 /* 1148 * Create a new entry if allowable 1149 */ 1150 1151 c = ip6mr_cache_alloc_unres(); 1152 if (!c) { 1153 spin_unlock_bh(&mfc_unres_lock); 1154 1155 kfree_skb(skb); 1156 return -ENOBUFS; 1157 } 1158 1159 /* Fill in the new cache entry */ 1160 c->_c.mfc_parent = -1; 1161 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1162 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1163 1164 /* 1165 * Reflect first query at pim6sd 1166 */ 1167 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1168 if (err < 0) { 1169 /* If the report failed throw the cache entry 1170 out - Brad Parker 1171 */ 1172 spin_unlock_bh(&mfc_unres_lock); 1173 1174 ip6mr_cache_free(c); 1175 kfree_skb(skb); 1176 return err; 1177 } 1178 1179 atomic_inc(&mrt->cache_resolve_queue_len); 1180 list_add(&c->_c.list, &mrt->mfc_unres_queue); 1181 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1182 1183 ipmr_do_expire_process(mrt); 1184 } 1185 1186 /* See if we can append the packet */ 1187 if (c->_c.mfc_un.unres.unresolved.qlen > 3) { 1188 kfree_skb(skb); 1189 err = -ENOBUFS; 1190 } else { 1191 if (dev) { 1192 skb->dev = dev; 1193 skb->skb_iif = dev->ifindex; 1194 } 1195 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb); 1196 err = 0; 1197 } 1198 1199 spin_unlock_bh(&mfc_unres_lock); 1200 return err; 1201 } 1202 1203 /* 1204 * MFC6 cache manipulation by user space 1205 */ 1206 1207 static int ip6mr_mfc_delete(struct mr_table *mrt, struct mf6cctl *mfc, 1208 int parent) 1209 { 1210 struct mfc6_cache *c; 1211 1212 /* The entries are added/deleted only under RTNL */ 1213 rcu_read_lock(); 1214 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr, 1215 &mfc->mf6cc_mcastgrp.sin6_addr, parent); 1216 rcu_read_unlock(); 1217 if (!c) 1218 return -ENOENT; 1219 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ip6mr_rht_params); 1220 list_del_rcu(&c->_c.list); 1221 1222 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net), 1223 FIB_EVENT_ENTRY_DEL, c, mrt->id); 1224 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1225 mr_cache_put(&c->_c); 1226 return 0; 1227 } 1228 1229 static int ip6mr_device_event(struct notifier_block *this, 1230 unsigned long event, void *ptr) 1231 { 1232 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1233 struct net *net = dev_net(dev); 1234 struct mr_table *mrt; 1235 struct vif_device *v; 1236 int ct; 1237 1238 if (event != NETDEV_UNREGISTER) 1239 return NOTIFY_DONE; 1240 1241 ip6mr_for_each_table(mrt, net) { 1242 v = &mrt->vif_table[0]; 1243 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1244 if (v->dev == dev) 1245 mif6_delete(mrt, ct, 1, NULL); 1246 } 1247 } 1248 1249 return NOTIFY_DONE; 1250 } 1251 1252 static unsigned int ip6mr_seq_read(struct net *net) 1253 { 1254 ASSERT_RTNL(); 1255 1256 return net->ipv6.ipmr_seq + ip6mr_rules_seq_read(net); 1257 } 1258 1259 static int ip6mr_dump(struct net *net, struct notifier_block *nb) 1260 { 1261 return mr_dump(net, nb, RTNL_FAMILY_IP6MR, ip6mr_rules_dump, 1262 ip6mr_mr_table_iter, &mrt_lock); 1263 } 1264 1265 static struct notifier_block ip6_mr_notifier = { 1266 .notifier_call = ip6mr_device_event 1267 }; 1268 1269 static const struct fib_notifier_ops ip6mr_notifier_ops_template = { 1270 .family = RTNL_FAMILY_IP6MR, 1271 .fib_seq_read = ip6mr_seq_read, 1272 .fib_dump = ip6mr_dump, 1273 .owner = THIS_MODULE, 1274 }; 1275 1276 static int __net_init ip6mr_notifier_init(struct net *net) 1277 { 1278 struct fib_notifier_ops *ops; 1279 1280 net->ipv6.ipmr_seq = 0; 1281 1282 ops = fib_notifier_ops_register(&ip6mr_notifier_ops_template, net); 1283 if (IS_ERR(ops)) 1284 return PTR_ERR(ops); 1285 1286 net->ipv6.ip6mr_notifier_ops = ops; 1287 1288 return 0; 1289 } 1290 1291 static void __net_exit ip6mr_notifier_exit(struct net *net) 1292 { 1293 fib_notifier_ops_unregister(net->ipv6.ip6mr_notifier_ops); 1294 net->ipv6.ip6mr_notifier_ops = NULL; 1295 } 1296 1297 /* Setup for IP multicast routing */ 1298 static int __net_init ip6mr_net_init(struct net *net) 1299 { 1300 int err; 1301 1302 err = ip6mr_notifier_init(net); 1303 if (err) 1304 return err; 1305 1306 err = ip6mr_rules_init(net); 1307 if (err < 0) 1308 goto ip6mr_rules_fail; 1309 1310 #ifdef CONFIG_PROC_FS 1311 err = -ENOMEM; 1312 if (!proc_create_net("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_seq_ops, 1313 sizeof(struct mr_vif_iter))) 1314 goto proc_vif_fail; 1315 if (!proc_create_net("ip6_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops, 1316 sizeof(struct mr_mfc_iter))) 1317 goto proc_cache_fail; 1318 #endif 1319 1320 return 0; 1321 1322 #ifdef CONFIG_PROC_FS 1323 proc_cache_fail: 1324 remove_proc_entry("ip6_mr_vif", net->proc_net); 1325 proc_vif_fail: 1326 ip6mr_rules_exit(net); 1327 #endif 1328 ip6mr_rules_fail: 1329 ip6mr_notifier_exit(net); 1330 return err; 1331 } 1332 1333 static void __net_exit ip6mr_net_exit(struct net *net) 1334 { 1335 #ifdef CONFIG_PROC_FS 1336 remove_proc_entry("ip6_mr_cache", net->proc_net); 1337 remove_proc_entry("ip6_mr_vif", net->proc_net); 1338 #endif 1339 ip6mr_rules_exit(net); 1340 ip6mr_notifier_exit(net); 1341 } 1342 1343 static struct pernet_operations ip6mr_net_ops = { 1344 .init = ip6mr_net_init, 1345 .exit = ip6mr_net_exit, 1346 }; 1347 1348 int __init ip6_mr_init(void) 1349 { 1350 int err; 1351 1352 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1353 sizeof(struct mfc6_cache), 1354 0, SLAB_HWCACHE_ALIGN, 1355 NULL); 1356 if (!mrt_cachep) 1357 return -ENOMEM; 1358 1359 err = register_pernet_subsys(&ip6mr_net_ops); 1360 if (err) 1361 goto reg_pernet_fail; 1362 1363 err = register_netdevice_notifier(&ip6_mr_notifier); 1364 if (err) 1365 goto reg_notif_fail; 1366 #ifdef CONFIG_IPV6_PIMSM_V2 1367 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1368 pr_err("%s: can't add PIM protocol\n", __func__); 1369 err = -EAGAIN; 1370 goto add_proto_fail; 1371 } 1372 #endif 1373 err = rtnl_register_module(THIS_MODULE, RTNL_FAMILY_IP6MR, RTM_GETROUTE, 1374 NULL, ip6mr_rtm_dumproute, 0); 1375 if (err == 0) 1376 return 0; 1377 1378 #ifdef CONFIG_IPV6_PIMSM_V2 1379 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1380 add_proto_fail: 1381 unregister_netdevice_notifier(&ip6_mr_notifier); 1382 #endif 1383 reg_notif_fail: 1384 unregister_pernet_subsys(&ip6mr_net_ops); 1385 reg_pernet_fail: 1386 kmem_cache_destroy(mrt_cachep); 1387 return err; 1388 } 1389 1390 void ip6_mr_cleanup(void) 1391 { 1392 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1393 #ifdef CONFIG_IPV6_PIMSM_V2 1394 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1395 #endif 1396 unregister_netdevice_notifier(&ip6_mr_notifier); 1397 unregister_pernet_subsys(&ip6mr_net_ops); 1398 kmem_cache_destroy(mrt_cachep); 1399 } 1400 1401 static int ip6mr_mfc_add(struct net *net, struct mr_table *mrt, 1402 struct mf6cctl *mfc, int mrtsock, int parent) 1403 { 1404 unsigned char ttls[MAXMIFS]; 1405 struct mfc6_cache *uc, *c; 1406 struct mr_mfc *_uc; 1407 bool found; 1408 int i, err; 1409 1410 if (mfc->mf6cc_parent >= MAXMIFS) 1411 return -ENFILE; 1412 1413 memset(ttls, 255, MAXMIFS); 1414 for (i = 0; i < MAXMIFS; i++) { 1415 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1416 ttls[i] = 1; 1417 } 1418 1419 /* The entries are added/deleted only under RTNL */ 1420 rcu_read_lock(); 1421 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr, 1422 &mfc->mf6cc_mcastgrp.sin6_addr, parent); 1423 rcu_read_unlock(); 1424 if (c) { 1425 write_lock_bh(&mrt_lock); 1426 c->_c.mfc_parent = mfc->mf6cc_parent; 1427 ip6mr_update_thresholds(mrt, &c->_c, ttls); 1428 if (!mrtsock) 1429 c->_c.mfc_flags |= MFC_STATIC; 1430 write_unlock_bh(&mrt_lock); 1431 call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, 1432 c, mrt->id); 1433 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1434 return 0; 1435 } 1436 1437 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1438 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1439 return -EINVAL; 1440 1441 c = ip6mr_cache_alloc(); 1442 if (!c) 1443 return -ENOMEM; 1444 1445 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1446 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1447 c->_c.mfc_parent = mfc->mf6cc_parent; 1448 ip6mr_update_thresholds(mrt, &c->_c, ttls); 1449 if (!mrtsock) 1450 c->_c.mfc_flags |= MFC_STATIC; 1451 1452 err = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode, 1453 ip6mr_rht_params); 1454 if (err) { 1455 pr_err("ip6mr: rhtable insert error %d\n", err); 1456 ip6mr_cache_free(c); 1457 return err; 1458 } 1459 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list); 1460 1461 /* Check to see if we resolved a queued list. If so we 1462 * need to send on the frames and tidy up. 1463 */ 1464 found = false; 1465 spin_lock_bh(&mfc_unres_lock); 1466 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) { 1467 uc = (struct mfc6_cache *)_uc; 1468 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1469 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1470 list_del(&_uc->list); 1471 atomic_dec(&mrt->cache_resolve_queue_len); 1472 found = true; 1473 break; 1474 } 1475 } 1476 if (list_empty(&mrt->mfc_unres_queue)) 1477 del_timer(&mrt->ipmr_expire_timer); 1478 spin_unlock_bh(&mfc_unres_lock); 1479 1480 if (found) { 1481 ip6mr_cache_resolve(net, mrt, uc, c); 1482 ip6mr_cache_free(uc); 1483 } 1484 call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, 1485 c, mrt->id); 1486 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1487 return 0; 1488 } 1489 1490 /* 1491 * Close the multicast socket, and clear the vif tables etc 1492 */ 1493 1494 static void mroute_clean_tables(struct mr_table *mrt, int flags) 1495 { 1496 struct mr_mfc *c, *tmp; 1497 LIST_HEAD(list); 1498 int i; 1499 1500 /* Shut down all active vif entries */ 1501 if (flags & (MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC)) { 1502 for (i = 0; i < mrt->maxvif; i++) { 1503 if (((mrt->vif_table[i].flags & VIFF_STATIC) && 1504 !(flags & MRT6_FLUSH_MIFS_STATIC)) || 1505 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT6_FLUSH_MIFS))) 1506 continue; 1507 mif6_delete(mrt, i, 0, &list); 1508 } 1509 unregister_netdevice_many(&list); 1510 } 1511 1512 /* Wipe the cache */ 1513 if (flags & (MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC)) { 1514 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1515 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC_STATIC)) || 1516 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC))) 1517 continue; 1518 rhltable_remove(&mrt->mfc_hash, &c->mnode, ip6mr_rht_params); 1519 list_del_rcu(&c->list); 1520 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net), 1521 FIB_EVENT_ENTRY_DEL, 1522 (struct mfc6_cache *)c, mrt->id); 1523 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE); 1524 mr_cache_put(c); 1525 } 1526 } 1527 1528 if (flags & MRT6_FLUSH_MFC) { 1529 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1530 spin_lock_bh(&mfc_unres_lock); 1531 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1532 list_del(&c->list); 1533 mr6_netlink_event(mrt, (struct mfc6_cache *)c, 1534 RTM_DELROUTE); 1535 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c); 1536 } 1537 spin_unlock_bh(&mfc_unres_lock); 1538 } 1539 } 1540 } 1541 1542 static int ip6mr_sk_init(struct mr_table *mrt, struct sock *sk) 1543 { 1544 int err = 0; 1545 struct net *net = sock_net(sk); 1546 1547 rtnl_lock(); 1548 write_lock_bh(&mrt_lock); 1549 if (rtnl_dereference(mrt->mroute_sk)) { 1550 err = -EADDRINUSE; 1551 } else { 1552 rcu_assign_pointer(mrt->mroute_sk, sk); 1553 sock_set_flag(sk, SOCK_RCU_FREE); 1554 net->ipv6.devconf_all->mc_forwarding++; 1555 } 1556 write_unlock_bh(&mrt_lock); 1557 1558 if (!err) 1559 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1560 NETCONFA_MC_FORWARDING, 1561 NETCONFA_IFINDEX_ALL, 1562 net->ipv6.devconf_all); 1563 rtnl_unlock(); 1564 1565 return err; 1566 } 1567 1568 int ip6mr_sk_done(struct sock *sk) 1569 { 1570 int err = -EACCES; 1571 struct net *net = sock_net(sk); 1572 struct mr_table *mrt; 1573 1574 if (sk->sk_type != SOCK_RAW || 1575 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1576 return err; 1577 1578 rtnl_lock(); 1579 ip6mr_for_each_table(mrt, net) { 1580 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1581 write_lock_bh(&mrt_lock); 1582 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1583 /* Note that mroute_sk had SOCK_RCU_FREE set, 1584 * so the RCU grace period before sk freeing 1585 * is guaranteed by sk_destruct() 1586 */ 1587 net->ipv6.devconf_all->mc_forwarding--; 1588 write_unlock_bh(&mrt_lock); 1589 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1590 NETCONFA_MC_FORWARDING, 1591 NETCONFA_IFINDEX_ALL, 1592 net->ipv6.devconf_all); 1593 1594 mroute_clean_tables(mrt, MRT6_FLUSH_MIFS | MRT6_FLUSH_MFC); 1595 err = 0; 1596 break; 1597 } 1598 } 1599 rtnl_unlock(); 1600 1601 return err; 1602 } 1603 1604 bool mroute6_is_socket(struct net *net, struct sk_buff *skb) 1605 { 1606 struct mr_table *mrt; 1607 struct flowi6 fl6 = { 1608 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1609 .flowi6_oif = skb->dev->ifindex, 1610 .flowi6_mark = skb->mark, 1611 }; 1612 1613 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1614 return NULL; 1615 1616 return rcu_access_pointer(mrt->mroute_sk); 1617 } 1618 EXPORT_SYMBOL(mroute6_is_socket); 1619 1620 /* 1621 * Socket options and virtual interface manipulation. The whole 1622 * virtual interface system is a complete heap, but unfortunately 1623 * that's how BSD mrouted happens to think. Maybe one day with a proper 1624 * MOSPF/PIM router set up we can clean this up. 1625 */ 1626 1627 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1628 { 1629 int ret, parent = 0; 1630 struct mif6ctl vif; 1631 struct mf6cctl mfc; 1632 mifi_t mifi; 1633 struct net *net = sock_net(sk); 1634 struct mr_table *mrt; 1635 1636 if (sk->sk_type != SOCK_RAW || 1637 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1638 return -EOPNOTSUPP; 1639 1640 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1641 if (!mrt) 1642 return -ENOENT; 1643 1644 if (optname != MRT6_INIT) { 1645 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1646 !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1647 return -EACCES; 1648 } 1649 1650 switch (optname) { 1651 case MRT6_INIT: 1652 if (optlen < sizeof(int)) 1653 return -EINVAL; 1654 1655 return ip6mr_sk_init(mrt, sk); 1656 1657 case MRT6_DONE: 1658 return ip6mr_sk_done(sk); 1659 1660 case MRT6_ADD_MIF: 1661 if (optlen < sizeof(vif)) 1662 return -EINVAL; 1663 if (copy_from_user(&vif, optval, sizeof(vif))) 1664 return -EFAULT; 1665 if (vif.mif6c_mifi >= MAXMIFS) 1666 return -ENFILE; 1667 rtnl_lock(); 1668 ret = mif6_add(net, mrt, &vif, 1669 sk == rtnl_dereference(mrt->mroute_sk)); 1670 rtnl_unlock(); 1671 return ret; 1672 1673 case MRT6_DEL_MIF: 1674 if (optlen < sizeof(mifi_t)) 1675 return -EINVAL; 1676 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1677 return -EFAULT; 1678 rtnl_lock(); 1679 ret = mif6_delete(mrt, mifi, 0, NULL); 1680 rtnl_unlock(); 1681 return ret; 1682 1683 /* 1684 * Manipulate the forwarding caches. These live 1685 * in a sort of kernel/user symbiosis. 1686 */ 1687 case MRT6_ADD_MFC: 1688 case MRT6_DEL_MFC: 1689 parent = -1; 1690 /* fall through */ 1691 case MRT6_ADD_MFC_PROXY: 1692 case MRT6_DEL_MFC_PROXY: 1693 if (optlen < sizeof(mfc)) 1694 return -EINVAL; 1695 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1696 return -EFAULT; 1697 if (parent == 0) 1698 parent = mfc.mf6cc_parent; 1699 rtnl_lock(); 1700 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1701 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1702 else 1703 ret = ip6mr_mfc_add(net, mrt, &mfc, 1704 sk == 1705 rtnl_dereference(mrt->mroute_sk), 1706 parent); 1707 rtnl_unlock(); 1708 return ret; 1709 1710 case MRT6_FLUSH: 1711 { 1712 int flags; 1713 1714 if (optlen != sizeof(flags)) 1715 return -EINVAL; 1716 if (get_user(flags, (int __user *)optval)) 1717 return -EFAULT; 1718 rtnl_lock(); 1719 mroute_clean_tables(mrt, flags); 1720 rtnl_unlock(); 1721 return 0; 1722 } 1723 1724 /* 1725 * Control PIM assert (to activate pim will activate assert) 1726 */ 1727 case MRT6_ASSERT: 1728 { 1729 int v; 1730 1731 if (optlen != sizeof(v)) 1732 return -EINVAL; 1733 if (get_user(v, (int __user *)optval)) 1734 return -EFAULT; 1735 mrt->mroute_do_assert = v; 1736 return 0; 1737 } 1738 1739 #ifdef CONFIG_IPV6_PIMSM_V2 1740 case MRT6_PIM: 1741 { 1742 int v; 1743 1744 if (optlen != sizeof(v)) 1745 return -EINVAL; 1746 if (get_user(v, (int __user *)optval)) 1747 return -EFAULT; 1748 v = !!v; 1749 rtnl_lock(); 1750 ret = 0; 1751 if (v != mrt->mroute_do_pim) { 1752 mrt->mroute_do_pim = v; 1753 mrt->mroute_do_assert = v; 1754 } 1755 rtnl_unlock(); 1756 return ret; 1757 } 1758 1759 #endif 1760 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1761 case MRT6_TABLE: 1762 { 1763 u32 v; 1764 1765 if (optlen != sizeof(u32)) 1766 return -EINVAL; 1767 if (get_user(v, (u32 __user *)optval)) 1768 return -EFAULT; 1769 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1770 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1771 return -EINVAL; 1772 if (sk == rcu_access_pointer(mrt->mroute_sk)) 1773 return -EBUSY; 1774 1775 rtnl_lock(); 1776 ret = 0; 1777 mrt = ip6mr_new_table(net, v); 1778 if (IS_ERR(mrt)) 1779 ret = PTR_ERR(mrt); 1780 else 1781 raw6_sk(sk)->ip6mr_table = v; 1782 rtnl_unlock(); 1783 return ret; 1784 } 1785 #endif 1786 /* 1787 * Spurious command, or MRT6_VERSION which you cannot 1788 * set. 1789 */ 1790 default: 1791 return -ENOPROTOOPT; 1792 } 1793 } 1794 1795 /* 1796 * Getsock opt support for the multicast routing system. 1797 */ 1798 1799 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1800 int __user *optlen) 1801 { 1802 int olr; 1803 int val; 1804 struct net *net = sock_net(sk); 1805 struct mr_table *mrt; 1806 1807 if (sk->sk_type != SOCK_RAW || 1808 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1809 return -EOPNOTSUPP; 1810 1811 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1812 if (!mrt) 1813 return -ENOENT; 1814 1815 switch (optname) { 1816 case MRT6_VERSION: 1817 val = 0x0305; 1818 break; 1819 #ifdef CONFIG_IPV6_PIMSM_V2 1820 case MRT6_PIM: 1821 val = mrt->mroute_do_pim; 1822 break; 1823 #endif 1824 case MRT6_ASSERT: 1825 val = mrt->mroute_do_assert; 1826 break; 1827 default: 1828 return -ENOPROTOOPT; 1829 } 1830 1831 if (get_user(olr, optlen)) 1832 return -EFAULT; 1833 1834 olr = min_t(int, olr, sizeof(int)); 1835 if (olr < 0) 1836 return -EINVAL; 1837 1838 if (put_user(olr, optlen)) 1839 return -EFAULT; 1840 if (copy_to_user(optval, &val, olr)) 1841 return -EFAULT; 1842 return 0; 1843 } 1844 1845 /* 1846 * The IP multicast ioctl support routines. 1847 */ 1848 1849 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1850 { 1851 struct sioc_sg_req6 sr; 1852 struct sioc_mif_req6 vr; 1853 struct vif_device *vif; 1854 struct mfc6_cache *c; 1855 struct net *net = sock_net(sk); 1856 struct mr_table *mrt; 1857 1858 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1859 if (!mrt) 1860 return -ENOENT; 1861 1862 switch (cmd) { 1863 case SIOCGETMIFCNT_IN6: 1864 if (copy_from_user(&vr, arg, sizeof(vr))) 1865 return -EFAULT; 1866 if (vr.mifi >= mrt->maxvif) 1867 return -EINVAL; 1868 vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif); 1869 read_lock(&mrt_lock); 1870 vif = &mrt->vif_table[vr.mifi]; 1871 if (VIF_EXISTS(mrt, vr.mifi)) { 1872 vr.icount = vif->pkt_in; 1873 vr.ocount = vif->pkt_out; 1874 vr.ibytes = vif->bytes_in; 1875 vr.obytes = vif->bytes_out; 1876 read_unlock(&mrt_lock); 1877 1878 if (copy_to_user(arg, &vr, sizeof(vr))) 1879 return -EFAULT; 1880 return 0; 1881 } 1882 read_unlock(&mrt_lock); 1883 return -EADDRNOTAVAIL; 1884 case SIOCGETSGCNT_IN6: 1885 if (copy_from_user(&sr, arg, sizeof(sr))) 1886 return -EFAULT; 1887 1888 rcu_read_lock(); 1889 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1890 if (c) { 1891 sr.pktcnt = c->_c.mfc_un.res.pkt; 1892 sr.bytecnt = c->_c.mfc_un.res.bytes; 1893 sr.wrong_if = c->_c.mfc_un.res.wrong_if; 1894 rcu_read_unlock(); 1895 1896 if (copy_to_user(arg, &sr, sizeof(sr))) 1897 return -EFAULT; 1898 return 0; 1899 } 1900 rcu_read_unlock(); 1901 return -EADDRNOTAVAIL; 1902 default: 1903 return -ENOIOCTLCMD; 1904 } 1905 } 1906 1907 #ifdef CONFIG_COMPAT 1908 struct compat_sioc_sg_req6 { 1909 struct sockaddr_in6 src; 1910 struct sockaddr_in6 grp; 1911 compat_ulong_t pktcnt; 1912 compat_ulong_t bytecnt; 1913 compat_ulong_t wrong_if; 1914 }; 1915 1916 struct compat_sioc_mif_req6 { 1917 mifi_t mifi; 1918 compat_ulong_t icount; 1919 compat_ulong_t ocount; 1920 compat_ulong_t ibytes; 1921 compat_ulong_t obytes; 1922 }; 1923 1924 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1925 { 1926 struct compat_sioc_sg_req6 sr; 1927 struct compat_sioc_mif_req6 vr; 1928 struct vif_device *vif; 1929 struct mfc6_cache *c; 1930 struct net *net = sock_net(sk); 1931 struct mr_table *mrt; 1932 1933 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1934 if (!mrt) 1935 return -ENOENT; 1936 1937 switch (cmd) { 1938 case SIOCGETMIFCNT_IN6: 1939 if (copy_from_user(&vr, arg, sizeof(vr))) 1940 return -EFAULT; 1941 if (vr.mifi >= mrt->maxvif) 1942 return -EINVAL; 1943 vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif); 1944 read_lock(&mrt_lock); 1945 vif = &mrt->vif_table[vr.mifi]; 1946 if (VIF_EXISTS(mrt, vr.mifi)) { 1947 vr.icount = vif->pkt_in; 1948 vr.ocount = vif->pkt_out; 1949 vr.ibytes = vif->bytes_in; 1950 vr.obytes = vif->bytes_out; 1951 read_unlock(&mrt_lock); 1952 1953 if (copy_to_user(arg, &vr, sizeof(vr))) 1954 return -EFAULT; 1955 return 0; 1956 } 1957 read_unlock(&mrt_lock); 1958 return -EADDRNOTAVAIL; 1959 case SIOCGETSGCNT_IN6: 1960 if (copy_from_user(&sr, arg, sizeof(sr))) 1961 return -EFAULT; 1962 1963 rcu_read_lock(); 1964 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1965 if (c) { 1966 sr.pktcnt = c->_c.mfc_un.res.pkt; 1967 sr.bytecnt = c->_c.mfc_un.res.bytes; 1968 sr.wrong_if = c->_c.mfc_un.res.wrong_if; 1969 rcu_read_unlock(); 1970 1971 if (copy_to_user(arg, &sr, sizeof(sr))) 1972 return -EFAULT; 1973 return 0; 1974 } 1975 rcu_read_unlock(); 1976 return -EADDRNOTAVAIL; 1977 default: 1978 return -ENOIOCTLCMD; 1979 } 1980 } 1981 #endif 1982 1983 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 1984 { 1985 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 1986 IPSTATS_MIB_OUTFORWDATAGRAMS); 1987 IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)), 1988 IPSTATS_MIB_OUTOCTETS, skb->len); 1989 return dst_output(net, sk, skb); 1990 } 1991 1992 /* 1993 * Processing handlers for ip6mr_forward 1994 */ 1995 1996 static int ip6mr_forward2(struct net *net, struct mr_table *mrt, 1997 struct sk_buff *skb, int vifi) 1998 { 1999 struct ipv6hdr *ipv6h; 2000 struct vif_device *vif = &mrt->vif_table[vifi]; 2001 struct net_device *dev; 2002 struct dst_entry *dst; 2003 struct flowi6 fl6; 2004 2005 if (!vif->dev) 2006 goto out_free; 2007 2008 #ifdef CONFIG_IPV6_PIMSM_V2 2009 if (vif->flags & MIFF_REGISTER) { 2010 vif->pkt_out++; 2011 vif->bytes_out += skb->len; 2012 vif->dev->stats.tx_bytes += skb->len; 2013 vif->dev->stats.tx_packets++; 2014 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2015 goto out_free; 2016 } 2017 #endif 2018 2019 ipv6h = ipv6_hdr(skb); 2020 2021 fl6 = (struct flowi6) { 2022 .flowi6_oif = vif->link, 2023 .daddr = ipv6h->daddr, 2024 }; 2025 2026 dst = ip6_route_output(net, NULL, &fl6); 2027 if (dst->error) { 2028 dst_release(dst); 2029 goto out_free; 2030 } 2031 2032 skb_dst_drop(skb); 2033 skb_dst_set(skb, dst); 2034 2035 /* 2036 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2037 * not only before forwarding, but after forwarding on all output 2038 * interfaces. It is clear, if mrouter runs a multicasting 2039 * program, it should receive packets not depending to what interface 2040 * program is joined. 2041 * If we will not make it, the program will have to join on all 2042 * interfaces. On the other hand, multihoming host (or router, but 2043 * not mrouter) cannot join to more than one interface - it will 2044 * result in receiving multiple packets. 2045 */ 2046 dev = vif->dev; 2047 skb->dev = dev; 2048 vif->pkt_out++; 2049 vif->bytes_out += skb->len; 2050 2051 /* We are about to write */ 2052 /* XXX: extension headers? */ 2053 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2054 goto out_free; 2055 2056 ipv6h = ipv6_hdr(skb); 2057 ipv6h->hop_limit--; 2058 2059 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2060 2061 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2062 net, NULL, skb, skb->dev, dev, 2063 ip6mr_forward2_finish); 2064 2065 out_free: 2066 kfree_skb(skb); 2067 return 0; 2068 } 2069 2070 static int ip6mr_find_vif(struct mr_table *mrt, struct net_device *dev) 2071 { 2072 int ct; 2073 2074 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2075 if (mrt->vif_table[ct].dev == dev) 2076 break; 2077 } 2078 return ct; 2079 } 2080 2081 static void ip6_mr_forward(struct net *net, struct mr_table *mrt, 2082 struct net_device *dev, struct sk_buff *skb, 2083 struct mfc6_cache *c) 2084 { 2085 int psend = -1; 2086 int vif, ct; 2087 int true_vifi = ip6mr_find_vif(mrt, dev); 2088 2089 vif = c->_c.mfc_parent; 2090 c->_c.mfc_un.res.pkt++; 2091 c->_c.mfc_un.res.bytes += skb->len; 2092 c->_c.mfc_un.res.lastuse = jiffies; 2093 2094 if (ipv6_addr_any(&c->mf6c_origin) && true_vifi >= 0) { 2095 struct mfc6_cache *cache_proxy; 2096 2097 /* For an (*,G) entry, we only check that the incoming 2098 * interface is part of the static tree. 2099 */ 2100 rcu_read_lock(); 2101 cache_proxy = mr_mfc_find_any_parent(mrt, vif); 2102 if (cache_proxy && 2103 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) { 2104 rcu_read_unlock(); 2105 goto forward; 2106 } 2107 rcu_read_unlock(); 2108 } 2109 2110 /* 2111 * Wrong interface: drop packet and (maybe) send PIM assert. 2112 */ 2113 if (mrt->vif_table[vif].dev != dev) { 2114 c->_c.mfc_un.res.wrong_if++; 2115 2116 if (true_vifi >= 0 && mrt->mroute_do_assert && 2117 /* pimsm uses asserts, when switching from RPT to SPT, 2118 so that we cannot check that packet arrived on an oif. 2119 It is bad, but otherwise we would need to move pretty 2120 large chunk of pimd to kernel. Ough... --ANK 2121 */ 2122 (mrt->mroute_do_pim || 2123 c->_c.mfc_un.res.ttls[true_vifi] < 255) && 2124 time_after(jiffies, 2125 c->_c.mfc_un.res.last_assert + 2126 MFC_ASSERT_THRESH)) { 2127 c->_c.mfc_un.res.last_assert = jiffies; 2128 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2129 } 2130 goto dont_forward; 2131 } 2132 2133 forward: 2134 mrt->vif_table[vif].pkt_in++; 2135 mrt->vif_table[vif].bytes_in += skb->len; 2136 2137 /* 2138 * Forward the frame 2139 */ 2140 if (ipv6_addr_any(&c->mf6c_origin) && 2141 ipv6_addr_any(&c->mf6c_mcastgrp)) { 2142 if (true_vifi >= 0 && 2143 true_vifi != c->_c.mfc_parent && 2144 ipv6_hdr(skb)->hop_limit > 2145 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) { 2146 /* It's an (*,*) entry and the packet is not coming from 2147 * the upstream: forward the packet to the upstream 2148 * only. 2149 */ 2150 psend = c->_c.mfc_parent; 2151 goto last_forward; 2152 } 2153 goto dont_forward; 2154 } 2155 for (ct = c->_c.mfc_un.res.maxvif - 1; 2156 ct >= c->_c.mfc_un.res.minvif; ct--) { 2157 /* For (*,G) entry, don't forward to the incoming interface */ 2158 if ((!ipv6_addr_any(&c->mf6c_origin) || ct != true_vifi) && 2159 ipv6_hdr(skb)->hop_limit > c->_c.mfc_un.res.ttls[ct]) { 2160 if (psend != -1) { 2161 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2162 if (skb2) 2163 ip6mr_forward2(net, mrt, skb2, psend); 2164 } 2165 psend = ct; 2166 } 2167 } 2168 last_forward: 2169 if (psend != -1) { 2170 ip6mr_forward2(net, mrt, skb, psend); 2171 return; 2172 } 2173 2174 dont_forward: 2175 kfree_skb(skb); 2176 } 2177 2178 2179 /* 2180 * Multicast packets for forwarding arrive here 2181 */ 2182 2183 int ip6_mr_input(struct sk_buff *skb) 2184 { 2185 struct mfc6_cache *cache; 2186 struct net *net = dev_net(skb->dev); 2187 struct mr_table *mrt; 2188 struct flowi6 fl6 = { 2189 .flowi6_iif = skb->dev->ifindex, 2190 .flowi6_mark = skb->mark, 2191 }; 2192 int err; 2193 struct net_device *dev; 2194 2195 /* skb->dev passed in is the master dev for vrfs. 2196 * Get the proper interface that does have a vif associated with it. 2197 */ 2198 dev = skb->dev; 2199 if (netif_is_l3_master(skb->dev)) { 2200 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 2201 if (!dev) { 2202 kfree_skb(skb); 2203 return -ENODEV; 2204 } 2205 } 2206 2207 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2208 if (err < 0) { 2209 kfree_skb(skb); 2210 return err; 2211 } 2212 2213 read_lock(&mrt_lock); 2214 cache = ip6mr_cache_find(mrt, 2215 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2216 if (!cache) { 2217 int vif = ip6mr_find_vif(mrt, dev); 2218 2219 if (vif >= 0) 2220 cache = ip6mr_cache_find_any(mrt, 2221 &ipv6_hdr(skb)->daddr, 2222 vif); 2223 } 2224 2225 /* 2226 * No usable cache entry 2227 */ 2228 if (!cache) { 2229 int vif; 2230 2231 vif = ip6mr_find_vif(mrt, dev); 2232 if (vif >= 0) { 2233 int err = ip6mr_cache_unresolved(mrt, vif, skb, dev); 2234 read_unlock(&mrt_lock); 2235 2236 return err; 2237 } 2238 read_unlock(&mrt_lock); 2239 kfree_skb(skb); 2240 return -ENODEV; 2241 } 2242 2243 ip6_mr_forward(net, mrt, dev, skb, cache); 2244 2245 read_unlock(&mrt_lock); 2246 2247 return 0; 2248 } 2249 2250 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2251 u32 portid) 2252 { 2253 int err; 2254 struct mr_table *mrt; 2255 struct mfc6_cache *cache; 2256 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2257 2258 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2259 if (!mrt) 2260 return -ENOENT; 2261 2262 read_lock(&mrt_lock); 2263 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2264 if (!cache && skb->dev) { 2265 int vif = ip6mr_find_vif(mrt, skb->dev); 2266 2267 if (vif >= 0) 2268 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2269 vif); 2270 } 2271 2272 if (!cache) { 2273 struct sk_buff *skb2; 2274 struct ipv6hdr *iph; 2275 struct net_device *dev; 2276 int vif; 2277 2278 dev = skb->dev; 2279 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2280 read_unlock(&mrt_lock); 2281 return -ENODEV; 2282 } 2283 2284 /* really correct? */ 2285 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2286 if (!skb2) { 2287 read_unlock(&mrt_lock); 2288 return -ENOMEM; 2289 } 2290 2291 NETLINK_CB(skb2).portid = portid; 2292 skb_reset_transport_header(skb2); 2293 2294 skb_put(skb2, sizeof(struct ipv6hdr)); 2295 skb_reset_network_header(skb2); 2296 2297 iph = ipv6_hdr(skb2); 2298 iph->version = 0; 2299 iph->priority = 0; 2300 iph->flow_lbl[0] = 0; 2301 iph->flow_lbl[1] = 0; 2302 iph->flow_lbl[2] = 0; 2303 iph->payload_len = 0; 2304 iph->nexthdr = IPPROTO_NONE; 2305 iph->hop_limit = 0; 2306 iph->saddr = rt->rt6i_src.addr; 2307 iph->daddr = rt->rt6i_dst.addr; 2308 2309 err = ip6mr_cache_unresolved(mrt, vif, skb2, dev); 2310 read_unlock(&mrt_lock); 2311 2312 return err; 2313 } 2314 2315 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm); 2316 read_unlock(&mrt_lock); 2317 return err; 2318 } 2319 2320 static int ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2321 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2322 int flags) 2323 { 2324 struct nlmsghdr *nlh; 2325 struct rtmsg *rtm; 2326 int err; 2327 2328 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2329 if (!nlh) 2330 return -EMSGSIZE; 2331 2332 rtm = nlmsg_data(nlh); 2333 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2334 rtm->rtm_dst_len = 128; 2335 rtm->rtm_src_len = 128; 2336 rtm->rtm_tos = 0; 2337 rtm->rtm_table = mrt->id; 2338 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2339 goto nla_put_failure; 2340 rtm->rtm_type = RTN_MULTICAST; 2341 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2342 if (c->_c.mfc_flags & MFC_STATIC) 2343 rtm->rtm_protocol = RTPROT_STATIC; 2344 else 2345 rtm->rtm_protocol = RTPROT_MROUTED; 2346 rtm->rtm_flags = 0; 2347 2348 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2349 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2350 goto nla_put_failure; 2351 err = mr_fill_mroute(mrt, skb, &c->_c, rtm); 2352 /* do not break the dump if cache is unresolved */ 2353 if (err < 0 && err != -ENOENT) 2354 goto nla_put_failure; 2355 2356 nlmsg_end(skb, nlh); 2357 return 0; 2358 2359 nla_put_failure: 2360 nlmsg_cancel(skb, nlh); 2361 return -EMSGSIZE; 2362 } 2363 2364 static int _ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2365 u32 portid, u32 seq, struct mr_mfc *c, 2366 int cmd, int flags) 2367 { 2368 return ip6mr_fill_mroute(mrt, skb, portid, seq, (struct mfc6_cache *)c, 2369 cmd, flags); 2370 } 2371 2372 static int mr6_msgsize(bool unresolved, int maxvif) 2373 { 2374 size_t len = 2375 NLMSG_ALIGN(sizeof(struct rtmsg)) 2376 + nla_total_size(4) /* RTA_TABLE */ 2377 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2378 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2379 ; 2380 2381 if (!unresolved) 2382 len = len 2383 + nla_total_size(4) /* RTA_IIF */ 2384 + nla_total_size(0) /* RTA_MULTIPATH */ 2385 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2386 /* RTA_MFC_STATS */ 2387 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2388 ; 2389 2390 return len; 2391 } 2392 2393 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc, 2394 int cmd) 2395 { 2396 struct net *net = read_pnet(&mrt->net); 2397 struct sk_buff *skb; 2398 int err = -ENOBUFS; 2399 2400 skb = nlmsg_new(mr6_msgsize(mfc->_c.mfc_parent >= MAXMIFS, mrt->maxvif), 2401 GFP_ATOMIC); 2402 if (!skb) 2403 goto errout; 2404 2405 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2406 if (err < 0) 2407 goto errout; 2408 2409 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2410 return; 2411 2412 errout: 2413 kfree_skb(skb); 2414 if (err < 0) 2415 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2416 } 2417 2418 static size_t mrt6msg_netlink_msgsize(size_t payloadlen) 2419 { 2420 size_t len = 2421 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2422 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */ 2423 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */ 2424 /* IP6MRA_CREPORT_SRC_ADDR */ 2425 + nla_total_size(sizeof(struct in6_addr)) 2426 /* IP6MRA_CREPORT_DST_ADDR */ 2427 + nla_total_size(sizeof(struct in6_addr)) 2428 /* IP6MRA_CREPORT_PKT */ 2429 + nla_total_size(payloadlen) 2430 ; 2431 2432 return len; 2433 } 2434 2435 static void mrt6msg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt) 2436 { 2437 struct net *net = read_pnet(&mrt->net); 2438 struct nlmsghdr *nlh; 2439 struct rtgenmsg *rtgenm; 2440 struct mrt6msg *msg; 2441 struct sk_buff *skb; 2442 struct nlattr *nla; 2443 int payloadlen; 2444 2445 payloadlen = pkt->len - sizeof(struct mrt6msg); 2446 msg = (struct mrt6msg *)skb_transport_header(pkt); 2447 2448 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2449 if (!skb) 2450 goto errout; 2451 2452 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2453 sizeof(struct rtgenmsg), 0); 2454 if (!nlh) 2455 goto errout; 2456 rtgenm = nlmsg_data(nlh); 2457 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR; 2458 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) || 2459 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) || 2460 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR, 2461 &msg->im6_src) || 2462 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR, 2463 &msg->im6_dst)) 2464 goto nla_put_failure; 2465 2466 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen); 2467 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg), 2468 nla_data(nla), payloadlen)) 2469 goto nla_put_failure; 2470 2471 nlmsg_end(skb, nlh); 2472 2473 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC); 2474 return; 2475 2476 nla_put_failure: 2477 nlmsg_cancel(skb, nlh); 2478 errout: 2479 kfree_skb(skb); 2480 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS); 2481 } 2482 2483 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2484 { 2485 const struct nlmsghdr *nlh = cb->nlh; 2486 struct fib_dump_filter filter = {}; 2487 int err; 2488 2489 if (cb->strict_check) { 2490 err = ip_valid_fib_dump_req(sock_net(skb->sk), nlh, 2491 &filter, cb); 2492 if (err < 0) 2493 return err; 2494 } 2495 2496 if (filter.table_id) { 2497 struct mr_table *mrt; 2498 2499 mrt = ip6mr_get_table(sock_net(skb->sk), filter.table_id); 2500 if (!mrt) { 2501 if (filter.dump_all_families) 2502 return skb->len; 2503 2504 NL_SET_ERR_MSG_MOD(cb->extack, "MR table does not exist"); 2505 return -ENOENT; 2506 } 2507 err = mr_table_dump(mrt, skb, cb, _ip6mr_fill_mroute, 2508 &mfc_unres_lock, &filter); 2509 return skb->len ? : err; 2510 } 2511 2512 return mr_rtm_dumproute(skb, cb, ip6mr_mr_table_iter, 2513 _ip6mr_fill_mroute, &mfc_unres_lock, &filter); 2514 } 2515