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_SPINLOCK(mrt_lock); 66 67 static struct net_device *vif_dev_read(const struct vif_device *vif) 68 { 69 return rcu_dereference(vif->dev); 70 } 71 72 /* Multicast router control variables */ 73 74 /* Special spinlock for queue of unresolved entries */ 75 static DEFINE_SPINLOCK(mfc_unres_lock); 76 77 /* We return to original Alan's scheme. Hash table of resolved 78 entries is changed only in process context and protected 79 with weak lock mrt_lock. Queue of unresolved entries is protected 80 with strong spinlock mfc_unres_lock. 81 82 In this case data path is free of exclusive locks at all. 83 */ 84 85 static struct kmem_cache *mrt_cachep __read_mostly; 86 87 static struct mr_table *ip6mr_new_table(struct net *net, u32 id); 88 static void ip6mr_free_table(struct mr_table *mrt); 89 90 static void ip6_mr_forward(struct net *net, struct mr_table *mrt, 91 struct net_device *dev, struct sk_buff *skb, 92 struct mfc6_cache *cache); 93 static int ip6mr_cache_report(const struct mr_table *mrt, struct sk_buff *pkt, 94 mifi_t mifi, int assert); 95 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc, 96 int cmd); 97 static void mrt6msg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt); 98 static int ip6mr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, 99 struct netlink_ext_ack *extack); 100 static int ip6mr_rtm_dumproute(struct sk_buff *skb, 101 struct netlink_callback *cb); 102 static void mroute_clean_tables(struct mr_table *mrt, int flags); 103 static void ipmr_expire_process(struct timer_list *t); 104 105 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 106 #define ip6mr_for_each_table(mrt, net) \ 107 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list, \ 108 lockdep_rtnl_is_held() || \ 109 list_empty(&net->ipv6.mr6_tables)) 110 111 static struct mr_table *ip6mr_mr_table_iter(struct net *net, 112 struct mr_table *mrt) 113 { 114 struct mr_table *ret; 115 116 if (!mrt) 117 ret = list_entry_rcu(net->ipv6.mr6_tables.next, 118 struct mr_table, list); 119 else 120 ret = list_entry_rcu(mrt->list.next, 121 struct mr_table, list); 122 123 if (&ret->list == &net->ipv6.mr6_tables) 124 return NULL; 125 return ret; 126 } 127 128 static struct mr_table *ip6mr_get_table(struct net *net, u32 id) 129 { 130 struct mr_table *mrt; 131 132 ip6mr_for_each_table(mrt, net) { 133 if (mrt->id == id) 134 return mrt; 135 } 136 return NULL; 137 } 138 139 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 140 struct mr_table **mrt) 141 { 142 int err; 143 struct ip6mr_result res; 144 struct fib_lookup_arg arg = { 145 .result = &res, 146 .flags = FIB_LOOKUP_NOREF, 147 }; 148 149 /* update flow if oif or iif point to device enslaved to l3mdev */ 150 l3mdev_update_flow(net, flowi6_to_flowi(flp6)); 151 152 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, 153 flowi6_to_flowi(flp6), 0, &arg); 154 if (err < 0) 155 return err; 156 *mrt = res.mrt; 157 return 0; 158 } 159 160 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp, 161 int flags, struct fib_lookup_arg *arg) 162 { 163 struct ip6mr_result *res = arg->result; 164 struct mr_table *mrt; 165 166 switch (rule->action) { 167 case FR_ACT_TO_TBL: 168 break; 169 case FR_ACT_UNREACHABLE: 170 return -ENETUNREACH; 171 case FR_ACT_PROHIBIT: 172 return -EACCES; 173 case FR_ACT_BLACKHOLE: 174 default: 175 return -EINVAL; 176 } 177 178 arg->table = fib_rule_get_table(rule, arg); 179 180 mrt = ip6mr_get_table(rule->fr_net, arg->table); 181 if (!mrt) 182 return -EAGAIN; 183 res->mrt = mrt; 184 return 0; 185 } 186 187 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags) 188 { 189 return 1; 190 } 191 192 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 193 struct fib_rule_hdr *frh, struct nlattr **tb, 194 struct netlink_ext_ack *extack) 195 { 196 return 0; 197 } 198 199 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 200 struct nlattr **tb) 201 { 202 return 1; 203 } 204 205 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 206 struct fib_rule_hdr *frh) 207 { 208 frh->dst_len = 0; 209 frh->src_len = 0; 210 frh->tos = 0; 211 return 0; 212 } 213 214 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = { 215 .family = RTNL_FAMILY_IP6MR, 216 .rule_size = sizeof(struct ip6mr_rule), 217 .addr_size = sizeof(struct in6_addr), 218 .action = ip6mr_rule_action, 219 .match = ip6mr_rule_match, 220 .configure = ip6mr_rule_configure, 221 .compare = ip6mr_rule_compare, 222 .fill = ip6mr_rule_fill, 223 .nlgroup = RTNLGRP_IPV6_RULE, 224 .owner = THIS_MODULE, 225 }; 226 227 static int __net_init ip6mr_rules_init(struct net *net) 228 { 229 struct fib_rules_ops *ops; 230 struct mr_table *mrt; 231 int err; 232 233 ops = fib_rules_register(&ip6mr_rules_ops_template, net); 234 if (IS_ERR(ops)) 235 return PTR_ERR(ops); 236 237 INIT_LIST_HEAD(&net->ipv6.mr6_tables); 238 239 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT); 240 if (IS_ERR(mrt)) { 241 err = PTR_ERR(mrt); 242 goto err1; 243 } 244 245 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT); 246 if (err < 0) 247 goto err2; 248 249 net->ipv6.mr6_rules_ops = ops; 250 return 0; 251 252 err2: 253 rtnl_lock(); 254 ip6mr_free_table(mrt); 255 rtnl_unlock(); 256 err1: 257 fib_rules_unregister(ops); 258 return err; 259 } 260 261 static void __net_exit ip6mr_rules_exit(struct net *net) 262 { 263 struct mr_table *mrt, *next; 264 265 ASSERT_RTNL(); 266 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) { 267 list_del(&mrt->list); 268 ip6mr_free_table(mrt); 269 } 270 fib_rules_unregister(net->ipv6.mr6_rules_ops); 271 } 272 273 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb, 274 struct netlink_ext_ack *extack) 275 { 276 return fib_rules_dump(net, nb, RTNL_FAMILY_IP6MR, extack); 277 } 278 279 static unsigned int ip6mr_rules_seq_read(const struct net *net) 280 { 281 return fib_rules_seq_read(net, RTNL_FAMILY_IP6MR); 282 } 283 284 bool ip6mr_rule_default(const struct fib_rule *rule) 285 { 286 return fib_rule_matchall(rule) && rule->action == FR_ACT_TO_TBL && 287 rule->table == RT6_TABLE_DFLT && !rule->l3mdev; 288 } 289 EXPORT_SYMBOL(ip6mr_rule_default); 290 #else 291 #define ip6mr_for_each_table(mrt, net) \ 292 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL) 293 294 static struct mr_table *ip6mr_mr_table_iter(struct net *net, 295 struct mr_table *mrt) 296 { 297 if (!mrt) 298 return net->ipv6.mrt6; 299 return NULL; 300 } 301 302 static struct mr_table *ip6mr_get_table(struct net *net, u32 id) 303 { 304 return net->ipv6.mrt6; 305 } 306 307 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 308 struct mr_table **mrt) 309 { 310 *mrt = net->ipv6.mrt6; 311 return 0; 312 } 313 314 static int __net_init ip6mr_rules_init(struct net *net) 315 { 316 struct mr_table *mrt; 317 318 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT); 319 if (IS_ERR(mrt)) 320 return PTR_ERR(mrt); 321 net->ipv6.mrt6 = mrt; 322 return 0; 323 } 324 325 static void __net_exit ip6mr_rules_exit(struct net *net) 326 { 327 ASSERT_RTNL(); 328 ip6mr_free_table(net->ipv6.mrt6); 329 net->ipv6.mrt6 = NULL; 330 } 331 332 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb, 333 struct netlink_ext_ack *extack) 334 { 335 return 0; 336 } 337 338 static unsigned int ip6mr_rules_seq_read(const struct net *net) 339 { 340 return 0; 341 } 342 #endif 343 344 static int ip6mr_hash_cmp(struct rhashtable_compare_arg *arg, 345 const void *ptr) 346 { 347 const struct mfc6_cache_cmp_arg *cmparg = arg->key; 348 struct mfc6_cache *c = (struct mfc6_cache *)ptr; 349 350 return !ipv6_addr_equal(&c->mf6c_mcastgrp, &cmparg->mf6c_mcastgrp) || 351 !ipv6_addr_equal(&c->mf6c_origin, &cmparg->mf6c_origin); 352 } 353 354 static const struct rhashtable_params ip6mr_rht_params = { 355 .head_offset = offsetof(struct mr_mfc, mnode), 356 .key_offset = offsetof(struct mfc6_cache, cmparg), 357 .key_len = sizeof(struct mfc6_cache_cmp_arg), 358 .nelem_hint = 3, 359 .obj_cmpfn = ip6mr_hash_cmp, 360 .automatic_shrinking = true, 361 }; 362 363 static void ip6mr_new_table_set(struct mr_table *mrt, 364 struct net *net) 365 { 366 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 367 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables); 368 #endif 369 } 370 371 static struct mfc6_cache_cmp_arg ip6mr_mr_table_ops_cmparg_any = { 372 .mf6c_origin = IN6ADDR_ANY_INIT, 373 .mf6c_mcastgrp = IN6ADDR_ANY_INIT, 374 }; 375 376 static struct mr_table_ops ip6mr_mr_table_ops = { 377 .rht_params = &ip6mr_rht_params, 378 .cmparg_any = &ip6mr_mr_table_ops_cmparg_any, 379 }; 380 381 static struct mr_table *ip6mr_new_table(struct net *net, u32 id) 382 { 383 struct mr_table *mrt; 384 385 mrt = ip6mr_get_table(net, id); 386 if (mrt) 387 return mrt; 388 389 return mr_table_alloc(net, id, &ip6mr_mr_table_ops, 390 ipmr_expire_process, ip6mr_new_table_set); 391 } 392 393 static void ip6mr_free_table(struct mr_table *mrt) 394 { 395 timer_shutdown_sync(&mrt->ipmr_expire_timer); 396 mroute_clean_tables(mrt, MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC | 397 MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC); 398 rhltable_destroy(&mrt->mfc_hash); 399 kfree(mrt); 400 } 401 402 #ifdef CONFIG_PROC_FS 403 /* The /proc interfaces to multicast routing 404 * /proc/ip6_mr_cache /proc/ip6_mr_vif 405 */ 406 407 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) 408 __acquires(RCU) 409 { 410 struct mr_vif_iter *iter = seq->private; 411 struct net *net = seq_file_net(seq); 412 struct mr_table *mrt; 413 414 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 415 if (!mrt) 416 return ERR_PTR(-ENOENT); 417 418 iter->mrt = mrt; 419 420 rcu_read_lock(); 421 return mr_vif_seq_start(seq, pos); 422 } 423 424 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) 425 __releases(RCU) 426 { 427 rcu_read_unlock(); 428 } 429 430 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) 431 { 432 struct mr_vif_iter *iter = seq->private; 433 struct mr_table *mrt = iter->mrt; 434 435 if (v == SEQ_START_TOKEN) { 436 seq_puts(seq, 437 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); 438 } else { 439 const struct vif_device *vif = v; 440 const struct net_device *vif_dev; 441 const char *name; 442 443 vif_dev = vif_dev_read(vif); 444 name = vif_dev ? vif_dev->name : "none"; 445 446 seq_printf(seq, 447 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", 448 vif - mrt->vif_table, 449 name, vif->bytes_in, vif->pkt_in, 450 vif->bytes_out, vif->pkt_out, 451 vif->flags); 452 } 453 return 0; 454 } 455 456 static const struct seq_operations ip6mr_vif_seq_ops = { 457 .start = ip6mr_vif_seq_start, 458 .next = mr_vif_seq_next, 459 .stop = ip6mr_vif_seq_stop, 460 .show = ip6mr_vif_seq_show, 461 }; 462 463 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 464 { 465 struct net *net = seq_file_net(seq); 466 struct mr_table *mrt; 467 468 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 469 if (!mrt) 470 return ERR_PTR(-ENOENT); 471 472 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock); 473 } 474 475 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 476 { 477 int n; 478 479 if (v == SEQ_START_TOKEN) { 480 seq_puts(seq, 481 "Group " 482 "Origin " 483 "Iif Pkts Bytes Wrong Oifs\n"); 484 } else { 485 const struct mfc6_cache *mfc = v; 486 const struct mr_mfc_iter *it = seq->private; 487 struct mr_table *mrt = it->mrt; 488 489 seq_printf(seq, "%pI6 %pI6 %-3hd", 490 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, 491 mfc->_c.mfc_parent); 492 493 if (it->cache != &mrt->mfc_unres_queue) { 494 seq_printf(seq, " %8lu %8lu %8lu", 495 mfc->_c.mfc_un.res.pkt, 496 mfc->_c.mfc_un.res.bytes, 497 mfc->_c.mfc_un.res.wrong_if); 498 for (n = mfc->_c.mfc_un.res.minvif; 499 n < mfc->_c.mfc_un.res.maxvif; n++) { 500 if (VIF_EXISTS(mrt, n) && 501 mfc->_c.mfc_un.res.ttls[n] < 255) 502 seq_printf(seq, 503 " %2d:%-3d", n, 504 mfc->_c.mfc_un.res.ttls[n]); 505 } 506 } else { 507 /* unresolved mfc_caches don't contain 508 * pkt, bytes and wrong_if values 509 */ 510 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 511 } 512 seq_putc(seq, '\n'); 513 } 514 return 0; 515 } 516 517 static const struct seq_operations ipmr_mfc_seq_ops = { 518 .start = ipmr_mfc_seq_start, 519 .next = mr_mfc_seq_next, 520 .stop = mr_mfc_seq_stop, 521 .show = ipmr_mfc_seq_show, 522 }; 523 #endif 524 525 #ifdef CONFIG_IPV6_PIMSM_V2 526 527 static int pim6_rcv(struct sk_buff *skb) 528 { 529 struct pimreghdr *pim; 530 struct ipv6hdr *encap; 531 struct net_device *reg_dev = NULL; 532 struct net *net = dev_net(skb->dev); 533 struct mr_table *mrt; 534 struct flowi6 fl6 = { 535 .flowi6_iif = skb->dev->ifindex, 536 .flowi6_mark = skb->mark, 537 }; 538 int reg_vif_num; 539 540 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) 541 goto drop; 542 543 pim = (struct pimreghdr *)skb_transport_header(skb); 544 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) || 545 (pim->flags & PIM_NULL_REGISTER) || 546 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 547 sizeof(*pim), IPPROTO_PIM, 548 csum_partial((void *)pim, sizeof(*pim), 0)) && 549 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 550 goto drop; 551 552 /* check if the inner packet is destined to mcast group */ 553 encap = (struct ipv6hdr *)(skb_transport_header(skb) + 554 sizeof(*pim)); 555 556 if (!ipv6_addr_is_multicast(&encap->daddr) || 557 encap->payload_len == 0 || 558 ntohs(encap->payload_len) + sizeof(*pim) > skb->len) 559 goto drop; 560 561 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 562 goto drop; 563 564 /* Pairs with WRITE_ONCE() in mif6_add()/mif6_delete() */ 565 reg_vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 566 if (reg_vif_num >= 0) 567 reg_dev = vif_dev_read(&mrt->vif_table[reg_vif_num]); 568 569 if (!reg_dev) 570 goto drop; 571 572 skb->mac_header = skb->network_header; 573 skb_pull(skb, (u8 *)encap - skb->data); 574 skb_reset_network_header(skb); 575 skb->protocol = htons(ETH_P_IPV6); 576 skb->ip_summed = CHECKSUM_NONE; 577 578 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 579 580 netif_rx(skb); 581 582 return 0; 583 drop: 584 kfree_skb(skb); 585 return 0; 586 } 587 588 static const struct inet6_protocol pim6_protocol = { 589 .handler = pim6_rcv, 590 }; 591 592 /* Service routines creating virtual interfaces: PIMREG */ 593 594 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, 595 struct net_device *dev) 596 { 597 struct net *net = dev_net(dev); 598 struct mr_table *mrt; 599 struct flowi6 fl6 = { 600 .flowi6_oif = dev->ifindex, 601 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 602 .flowi6_mark = skb->mark, 603 }; 604 605 if (!pskb_inet_may_pull(skb)) 606 goto tx_err; 607 608 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 609 goto tx_err; 610 611 DEV_STATS_ADD(dev, tx_bytes, skb->len); 612 DEV_STATS_INC(dev, tx_packets); 613 rcu_read_lock(); 614 ip6mr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num), 615 MRT6MSG_WHOLEPKT); 616 rcu_read_unlock(); 617 kfree_skb(skb); 618 return NETDEV_TX_OK; 619 620 tx_err: 621 DEV_STATS_INC(dev, tx_errors); 622 kfree_skb(skb); 623 return NETDEV_TX_OK; 624 } 625 626 static int reg_vif_get_iflink(const struct net_device *dev) 627 { 628 return 0; 629 } 630 631 static const struct net_device_ops reg_vif_netdev_ops = { 632 .ndo_start_xmit = reg_vif_xmit, 633 .ndo_get_iflink = reg_vif_get_iflink, 634 }; 635 636 static void reg_vif_setup(struct net_device *dev) 637 { 638 dev->type = ARPHRD_PIMREG; 639 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; 640 dev->flags = IFF_NOARP; 641 dev->netdev_ops = ®_vif_netdev_ops; 642 dev->needs_free_netdev = true; 643 dev->netns_local = true; 644 } 645 646 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr_table *mrt) 647 { 648 struct net_device *dev; 649 char name[IFNAMSIZ]; 650 651 if (mrt->id == RT6_TABLE_DFLT) 652 sprintf(name, "pim6reg"); 653 else 654 sprintf(name, "pim6reg%u", mrt->id); 655 656 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 657 if (!dev) 658 return NULL; 659 660 dev_net_set(dev, net); 661 662 if (register_netdevice(dev)) { 663 free_netdev(dev); 664 return NULL; 665 } 666 667 if (dev_open(dev, NULL)) 668 goto failure; 669 670 dev_hold(dev); 671 return dev; 672 673 failure: 674 unregister_netdevice(dev); 675 return NULL; 676 } 677 #endif 678 679 static int call_ip6mr_vif_entry_notifiers(struct net *net, 680 enum fib_event_type event_type, 681 struct vif_device *vif, 682 struct net_device *vif_dev, 683 mifi_t vif_index, u32 tb_id) 684 { 685 return mr_call_vif_notifiers(net, RTNL_FAMILY_IP6MR, event_type, 686 vif, vif_dev, vif_index, tb_id, 687 &net->ipv6.ipmr_seq); 688 } 689 690 static int call_ip6mr_mfc_entry_notifiers(struct net *net, 691 enum fib_event_type event_type, 692 struct mfc6_cache *mfc, u32 tb_id) 693 { 694 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IP6MR, event_type, 695 &mfc->_c, tb_id, &net->ipv6.ipmr_seq); 696 } 697 698 /* Delete a VIF entry */ 699 static int mif6_delete(struct mr_table *mrt, int vifi, int notify, 700 struct list_head *head) 701 { 702 struct vif_device *v; 703 struct net_device *dev; 704 struct inet6_dev *in6_dev; 705 706 if (vifi < 0 || vifi >= mrt->maxvif) 707 return -EADDRNOTAVAIL; 708 709 v = &mrt->vif_table[vifi]; 710 711 dev = rtnl_dereference(v->dev); 712 if (!dev) 713 return -EADDRNOTAVAIL; 714 715 call_ip6mr_vif_entry_notifiers(read_pnet(&mrt->net), 716 FIB_EVENT_VIF_DEL, v, dev, 717 vifi, mrt->id); 718 spin_lock(&mrt_lock); 719 RCU_INIT_POINTER(v->dev, NULL); 720 721 #ifdef CONFIG_IPV6_PIMSM_V2 722 if (vifi == mrt->mroute_reg_vif_num) { 723 /* Pairs with READ_ONCE() in ip6mr_cache_report() and reg_vif_xmit() */ 724 WRITE_ONCE(mrt->mroute_reg_vif_num, -1); 725 } 726 #endif 727 728 if (vifi + 1 == mrt->maxvif) { 729 int tmp; 730 for (tmp = vifi - 1; tmp >= 0; tmp--) { 731 if (VIF_EXISTS(mrt, tmp)) 732 break; 733 } 734 WRITE_ONCE(mrt->maxvif, tmp + 1); 735 } 736 737 spin_unlock(&mrt_lock); 738 739 dev_set_allmulti(dev, -1); 740 741 in6_dev = __in6_dev_get(dev); 742 if (in6_dev) { 743 atomic_dec(&in6_dev->cnf.mc_forwarding); 744 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 745 NETCONFA_MC_FORWARDING, 746 dev->ifindex, &in6_dev->cnf); 747 } 748 749 if ((v->flags & MIFF_REGISTER) && !notify) 750 unregister_netdevice_queue(dev, head); 751 752 netdev_put(dev, &v->dev_tracker); 753 return 0; 754 } 755 756 static inline void ip6mr_cache_free_rcu(struct rcu_head *head) 757 { 758 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu); 759 760 kmem_cache_free(mrt_cachep, (struct mfc6_cache *)c); 761 } 762 763 static inline void ip6mr_cache_free(struct mfc6_cache *c) 764 { 765 call_rcu(&c->_c.rcu, ip6mr_cache_free_rcu); 766 } 767 768 /* Destroy an unresolved cache entry, killing queued skbs 769 and reporting error to netlink readers. 770 */ 771 772 static void ip6mr_destroy_unres(struct mr_table *mrt, struct mfc6_cache *c) 773 { 774 struct net *net = read_pnet(&mrt->net); 775 struct sk_buff *skb; 776 777 atomic_dec(&mrt->cache_resolve_queue_len); 778 779 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved)) != NULL) { 780 if (ipv6_hdr(skb)->version == 0) { 781 struct nlmsghdr *nlh = skb_pull(skb, 782 sizeof(struct ipv6hdr)); 783 nlh->nlmsg_type = NLMSG_ERROR; 784 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 785 skb_trim(skb, nlh->nlmsg_len); 786 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT; 787 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 788 } else 789 kfree_skb(skb); 790 } 791 792 ip6mr_cache_free(c); 793 } 794 795 796 /* Timer process for all the unresolved queue. */ 797 798 static void ipmr_do_expire_process(struct mr_table *mrt) 799 { 800 unsigned long now = jiffies; 801 unsigned long expires = 10 * HZ; 802 struct mr_mfc *c, *next; 803 804 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 805 if (time_after(c->mfc_un.unres.expires, now)) { 806 /* not yet... */ 807 unsigned long interval = c->mfc_un.unres.expires - now; 808 if (interval < expires) 809 expires = interval; 810 continue; 811 } 812 813 list_del(&c->list); 814 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE); 815 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c); 816 } 817 818 if (!list_empty(&mrt->mfc_unres_queue)) 819 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 820 } 821 822 static void ipmr_expire_process(struct timer_list *t) 823 { 824 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer); 825 826 if (!spin_trylock(&mfc_unres_lock)) { 827 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 828 return; 829 } 830 831 if (!list_empty(&mrt->mfc_unres_queue)) 832 ipmr_do_expire_process(mrt); 833 834 spin_unlock(&mfc_unres_lock); 835 } 836 837 /* Fill oifs list. It is called under locked mrt_lock. */ 838 839 static void ip6mr_update_thresholds(struct mr_table *mrt, 840 struct mr_mfc *cache, 841 unsigned char *ttls) 842 { 843 int vifi; 844 845 cache->mfc_un.res.minvif = MAXMIFS; 846 cache->mfc_un.res.maxvif = 0; 847 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 848 849 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 850 if (VIF_EXISTS(mrt, vifi) && 851 ttls[vifi] && ttls[vifi] < 255) { 852 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 853 if (cache->mfc_un.res.minvif > vifi) 854 cache->mfc_un.res.minvif = vifi; 855 if (cache->mfc_un.res.maxvif <= vifi) 856 cache->mfc_un.res.maxvif = vifi + 1; 857 } 858 } 859 cache->mfc_un.res.lastuse = jiffies; 860 } 861 862 static int mif6_add(struct net *net, struct mr_table *mrt, 863 struct mif6ctl *vifc, int mrtsock) 864 { 865 int vifi = vifc->mif6c_mifi; 866 struct vif_device *v = &mrt->vif_table[vifi]; 867 struct net_device *dev; 868 struct inet6_dev *in6_dev; 869 int err; 870 871 /* Is vif busy ? */ 872 if (VIF_EXISTS(mrt, vifi)) 873 return -EADDRINUSE; 874 875 switch (vifc->mif6c_flags) { 876 #ifdef CONFIG_IPV6_PIMSM_V2 877 case MIFF_REGISTER: 878 /* 879 * Special Purpose VIF in PIM 880 * All the packets will be sent to the daemon 881 */ 882 if (mrt->mroute_reg_vif_num >= 0) 883 return -EADDRINUSE; 884 dev = ip6mr_reg_vif(net, mrt); 885 if (!dev) 886 return -ENOBUFS; 887 err = dev_set_allmulti(dev, 1); 888 if (err) { 889 unregister_netdevice(dev); 890 dev_put(dev); 891 return err; 892 } 893 break; 894 #endif 895 case 0: 896 dev = dev_get_by_index(net, vifc->mif6c_pifi); 897 if (!dev) 898 return -EADDRNOTAVAIL; 899 err = dev_set_allmulti(dev, 1); 900 if (err) { 901 dev_put(dev); 902 return err; 903 } 904 break; 905 default: 906 return -EINVAL; 907 } 908 909 in6_dev = __in6_dev_get(dev); 910 if (in6_dev) { 911 atomic_inc(&in6_dev->cnf.mc_forwarding); 912 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 913 NETCONFA_MC_FORWARDING, 914 dev->ifindex, &in6_dev->cnf); 915 } 916 917 /* Fill in the VIF structures */ 918 vif_device_init(v, dev, vifc->vifc_rate_limit, vifc->vifc_threshold, 919 vifc->mif6c_flags | (!mrtsock ? VIFF_STATIC : 0), 920 MIFF_REGISTER); 921 922 /* And finish update writing critical data */ 923 spin_lock(&mrt_lock); 924 rcu_assign_pointer(v->dev, dev); 925 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC); 926 #ifdef CONFIG_IPV6_PIMSM_V2 927 if (v->flags & MIFF_REGISTER) 928 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi); 929 #endif 930 if (vifi + 1 > mrt->maxvif) 931 WRITE_ONCE(mrt->maxvif, vifi + 1); 932 spin_unlock(&mrt_lock); 933 call_ip6mr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, 934 v, dev, vifi, mrt->id); 935 return 0; 936 } 937 938 static struct mfc6_cache *ip6mr_cache_find(struct mr_table *mrt, 939 const struct in6_addr *origin, 940 const struct in6_addr *mcastgrp) 941 { 942 struct mfc6_cache_cmp_arg arg = { 943 .mf6c_origin = *origin, 944 .mf6c_mcastgrp = *mcastgrp, 945 }; 946 947 return mr_mfc_find(mrt, &arg); 948 } 949 950 /* Look for a (*,G) entry */ 951 static struct mfc6_cache *ip6mr_cache_find_any(struct mr_table *mrt, 952 struct in6_addr *mcastgrp, 953 mifi_t mifi) 954 { 955 struct mfc6_cache_cmp_arg arg = { 956 .mf6c_origin = in6addr_any, 957 .mf6c_mcastgrp = *mcastgrp, 958 }; 959 960 if (ipv6_addr_any(mcastgrp)) 961 return mr_mfc_find_any_parent(mrt, mifi); 962 return mr_mfc_find_any(mrt, mifi, &arg); 963 } 964 965 /* Look for a (S,G,iif) entry if parent != -1 */ 966 static struct mfc6_cache * 967 ip6mr_cache_find_parent(struct mr_table *mrt, 968 const struct in6_addr *origin, 969 const struct in6_addr *mcastgrp, 970 int parent) 971 { 972 struct mfc6_cache_cmp_arg arg = { 973 .mf6c_origin = *origin, 974 .mf6c_mcastgrp = *mcastgrp, 975 }; 976 977 return mr_mfc_find_parent(mrt, &arg, parent); 978 } 979 980 /* Allocate a multicast cache entry */ 981 static struct mfc6_cache *ip6mr_cache_alloc(void) 982 { 983 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 984 if (!c) 985 return NULL; 986 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 987 c->_c.mfc_un.res.minvif = MAXMIFS; 988 c->_c.free = ip6mr_cache_free_rcu; 989 refcount_set(&c->_c.mfc_un.res.refcount, 1); 990 return c; 991 } 992 993 static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 994 { 995 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 996 if (!c) 997 return NULL; 998 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved); 999 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ; 1000 return c; 1001 } 1002 1003 /* 1004 * A cache entry has gone into a resolved state from queued 1005 */ 1006 1007 static void ip6mr_cache_resolve(struct net *net, struct mr_table *mrt, 1008 struct mfc6_cache *uc, struct mfc6_cache *c) 1009 { 1010 struct sk_buff *skb; 1011 1012 /* 1013 * Play the pending entries through our router 1014 */ 1015 1016 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) { 1017 if (ipv6_hdr(skb)->version == 0) { 1018 struct nlmsghdr *nlh = skb_pull(skb, 1019 sizeof(struct ipv6hdr)); 1020 1021 if (mr_fill_mroute(mrt, skb, &c->_c, 1022 nlmsg_data(nlh)) > 0) { 1023 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1024 } else { 1025 nlh->nlmsg_type = NLMSG_ERROR; 1026 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1027 skb_trim(skb, nlh->nlmsg_len); 1028 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE; 1029 } 1030 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1031 } else { 1032 rcu_read_lock(); 1033 ip6_mr_forward(net, mrt, skb->dev, skb, c); 1034 rcu_read_unlock(); 1035 } 1036 } 1037 } 1038 1039 /* 1040 * Bounce a cache query up to pim6sd and netlink. 1041 * 1042 * Called under rcu_read_lock() 1043 */ 1044 1045 static int ip6mr_cache_report(const struct mr_table *mrt, struct sk_buff *pkt, 1046 mifi_t mifi, int assert) 1047 { 1048 struct sock *mroute6_sk; 1049 struct sk_buff *skb; 1050 struct mrt6msg *msg; 1051 int ret; 1052 1053 #ifdef CONFIG_IPV6_PIMSM_V2 1054 if (assert == MRT6MSG_WHOLEPKT || assert == MRT6MSG_WRMIFWHOLE) 1055 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1056 +sizeof(*msg)); 1057 else 1058 #endif 1059 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1060 1061 if (!skb) 1062 return -ENOBUFS; 1063 1064 /* I suppose that internal messages 1065 * do not require checksums */ 1066 1067 skb->ip_summed = CHECKSUM_UNNECESSARY; 1068 1069 #ifdef CONFIG_IPV6_PIMSM_V2 1070 if (assert == MRT6MSG_WHOLEPKT || assert == MRT6MSG_WRMIFWHOLE) { 1071 /* Ugly, but we have no choice with this interface. 1072 Duplicate old header, fix length etc. 1073 And all this only to mangle msg->im6_msgtype and 1074 to set msg->im6_mbz to "mbz" :-) 1075 */ 1076 __skb_pull(skb, skb_network_offset(pkt)); 1077 1078 skb_push(skb, sizeof(*msg)); 1079 skb_reset_transport_header(skb); 1080 msg = (struct mrt6msg *)skb_transport_header(skb); 1081 msg->im6_mbz = 0; 1082 msg->im6_msgtype = assert; 1083 if (assert == MRT6MSG_WRMIFWHOLE) 1084 msg->im6_mif = mifi; 1085 else 1086 msg->im6_mif = READ_ONCE(mrt->mroute_reg_vif_num); 1087 msg->im6_pad = 0; 1088 msg->im6_src = ipv6_hdr(pkt)->saddr; 1089 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1090 1091 skb->ip_summed = CHECKSUM_UNNECESSARY; 1092 } else 1093 #endif 1094 { 1095 /* 1096 * Copy the IP header 1097 */ 1098 1099 skb_put(skb, sizeof(struct ipv6hdr)); 1100 skb_reset_network_header(skb); 1101 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1102 1103 /* 1104 * Add our header 1105 */ 1106 skb_put(skb, sizeof(*msg)); 1107 skb_reset_transport_header(skb); 1108 msg = (struct mrt6msg *)skb_transport_header(skb); 1109 1110 msg->im6_mbz = 0; 1111 msg->im6_msgtype = assert; 1112 msg->im6_mif = mifi; 1113 msg->im6_pad = 0; 1114 msg->im6_src = ipv6_hdr(pkt)->saddr; 1115 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1116 1117 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1118 skb->ip_summed = CHECKSUM_UNNECESSARY; 1119 } 1120 1121 mroute6_sk = rcu_dereference(mrt->mroute_sk); 1122 if (!mroute6_sk) { 1123 kfree_skb(skb); 1124 return -EINVAL; 1125 } 1126 1127 mrt6msg_netlink_event(mrt, skb); 1128 1129 /* Deliver to user space multicast routing algorithms */ 1130 ret = sock_queue_rcv_skb(mroute6_sk, skb); 1131 1132 if (ret < 0) { 1133 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1134 kfree_skb(skb); 1135 } 1136 1137 return ret; 1138 } 1139 1140 /* Queue a packet for resolution. It gets locked cache entry! */ 1141 static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi, 1142 struct sk_buff *skb, struct net_device *dev) 1143 { 1144 struct mfc6_cache *c; 1145 bool found = false; 1146 int err; 1147 1148 spin_lock_bh(&mfc_unres_lock); 1149 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) { 1150 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1151 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1152 found = true; 1153 break; 1154 } 1155 } 1156 1157 if (!found) { 1158 /* 1159 * Create a new entry if allowable 1160 */ 1161 1162 c = ip6mr_cache_alloc_unres(); 1163 if (!c) { 1164 spin_unlock_bh(&mfc_unres_lock); 1165 1166 kfree_skb(skb); 1167 return -ENOBUFS; 1168 } 1169 1170 /* Fill in the new cache entry */ 1171 c->_c.mfc_parent = -1; 1172 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1173 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1174 1175 /* 1176 * Reflect first query at pim6sd 1177 */ 1178 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1179 if (err < 0) { 1180 /* If the report failed throw the cache entry 1181 out - Brad Parker 1182 */ 1183 spin_unlock_bh(&mfc_unres_lock); 1184 1185 ip6mr_cache_free(c); 1186 kfree_skb(skb); 1187 return err; 1188 } 1189 1190 atomic_inc(&mrt->cache_resolve_queue_len); 1191 list_add(&c->_c.list, &mrt->mfc_unres_queue); 1192 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1193 1194 ipmr_do_expire_process(mrt); 1195 } 1196 1197 /* See if we can append the packet */ 1198 if (c->_c.mfc_un.unres.unresolved.qlen > 3) { 1199 kfree_skb(skb); 1200 err = -ENOBUFS; 1201 } else { 1202 if (dev) { 1203 skb->dev = dev; 1204 skb->skb_iif = dev->ifindex; 1205 } 1206 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb); 1207 err = 0; 1208 } 1209 1210 spin_unlock_bh(&mfc_unres_lock); 1211 return err; 1212 } 1213 1214 /* 1215 * MFC6 cache manipulation by user space 1216 */ 1217 1218 static int ip6mr_mfc_delete(struct mr_table *mrt, struct mf6cctl *mfc, 1219 int parent) 1220 { 1221 struct mfc6_cache *c; 1222 1223 /* The entries are added/deleted only under RTNL */ 1224 rcu_read_lock(); 1225 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr, 1226 &mfc->mf6cc_mcastgrp.sin6_addr, parent); 1227 rcu_read_unlock(); 1228 if (!c) 1229 return -ENOENT; 1230 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ip6mr_rht_params); 1231 list_del_rcu(&c->_c.list); 1232 1233 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net), 1234 FIB_EVENT_ENTRY_DEL, c, mrt->id); 1235 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1236 mr_cache_put(&c->_c); 1237 return 0; 1238 } 1239 1240 static int ip6mr_device_event(struct notifier_block *this, 1241 unsigned long event, void *ptr) 1242 { 1243 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1244 struct net *net = dev_net(dev); 1245 struct mr_table *mrt; 1246 struct vif_device *v; 1247 int ct; 1248 1249 if (event != NETDEV_UNREGISTER) 1250 return NOTIFY_DONE; 1251 1252 ip6mr_for_each_table(mrt, net) { 1253 v = &mrt->vif_table[0]; 1254 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1255 if (rcu_access_pointer(v->dev) == dev) 1256 mif6_delete(mrt, ct, 1, NULL); 1257 } 1258 } 1259 1260 return NOTIFY_DONE; 1261 } 1262 1263 static unsigned int ip6mr_seq_read(const struct net *net) 1264 { 1265 return READ_ONCE(net->ipv6.ipmr_seq) + ip6mr_rules_seq_read(net); 1266 } 1267 1268 static int ip6mr_dump(struct net *net, struct notifier_block *nb, 1269 struct netlink_ext_ack *extack) 1270 { 1271 return mr_dump(net, nb, RTNL_FAMILY_IP6MR, ip6mr_rules_dump, 1272 ip6mr_mr_table_iter, extack); 1273 } 1274 1275 static struct notifier_block ip6_mr_notifier = { 1276 .notifier_call = ip6mr_device_event 1277 }; 1278 1279 static const struct fib_notifier_ops ip6mr_notifier_ops_template = { 1280 .family = RTNL_FAMILY_IP6MR, 1281 .fib_seq_read = ip6mr_seq_read, 1282 .fib_dump = ip6mr_dump, 1283 .owner = THIS_MODULE, 1284 }; 1285 1286 static int __net_init ip6mr_notifier_init(struct net *net) 1287 { 1288 struct fib_notifier_ops *ops; 1289 1290 net->ipv6.ipmr_seq = 0; 1291 1292 ops = fib_notifier_ops_register(&ip6mr_notifier_ops_template, net); 1293 if (IS_ERR(ops)) 1294 return PTR_ERR(ops); 1295 1296 net->ipv6.ip6mr_notifier_ops = ops; 1297 1298 return 0; 1299 } 1300 1301 static void __net_exit ip6mr_notifier_exit(struct net *net) 1302 { 1303 fib_notifier_ops_unregister(net->ipv6.ip6mr_notifier_ops); 1304 net->ipv6.ip6mr_notifier_ops = NULL; 1305 } 1306 1307 /* Setup for IP multicast routing */ 1308 static int __net_init ip6mr_net_init(struct net *net) 1309 { 1310 int err; 1311 1312 err = ip6mr_notifier_init(net); 1313 if (err) 1314 return err; 1315 1316 err = ip6mr_rules_init(net); 1317 if (err < 0) 1318 goto ip6mr_rules_fail; 1319 1320 #ifdef CONFIG_PROC_FS 1321 err = -ENOMEM; 1322 if (!proc_create_net("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_seq_ops, 1323 sizeof(struct mr_vif_iter))) 1324 goto proc_vif_fail; 1325 if (!proc_create_net("ip6_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops, 1326 sizeof(struct mr_mfc_iter))) 1327 goto proc_cache_fail; 1328 #endif 1329 1330 return 0; 1331 1332 #ifdef CONFIG_PROC_FS 1333 proc_cache_fail: 1334 remove_proc_entry("ip6_mr_vif", net->proc_net); 1335 proc_vif_fail: 1336 rtnl_lock(); 1337 ip6mr_rules_exit(net); 1338 rtnl_unlock(); 1339 #endif 1340 ip6mr_rules_fail: 1341 ip6mr_notifier_exit(net); 1342 return err; 1343 } 1344 1345 static void __net_exit ip6mr_net_exit(struct net *net) 1346 { 1347 #ifdef CONFIG_PROC_FS 1348 remove_proc_entry("ip6_mr_cache", net->proc_net); 1349 remove_proc_entry("ip6_mr_vif", net->proc_net); 1350 #endif 1351 ip6mr_notifier_exit(net); 1352 } 1353 1354 static void __net_exit ip6mr_net_exit_batch(struct list_head *net_list) 1355 { 1356 struct net *net; 1357 1358 rtnl_lock(); 1359 list_for_each_entry(net, net_list, exit_list) 1360 ip6mr_rules_exit(net); 1361 rtnl_unlock(); 1362 } 1363 1364 static struct pernet_operations ip6mr_net_ops = { 1365 .init = ip6mr_net_init, 1366 .exit = ip6mr_net_exit, 1367 .exit_batch = ip6mr_net_exit_batch, 1368 }; 1369 1370 int __init ip6_mr_init(void) 1371 { 1372 int err; 1373 1374 mrt_cachep = KMEM_CACHE(mfc6_cache, SLAB_HWCACHE_ALIGN); 1375 if (!mrt_cachep) 1376 return -ENOMEM; 1377 1378 err = register_pernet_subsys(&ip6mr_net_ops); 1379 if (err) 1380 goto reg_pernet_fail; 1381 1382 err = register_netdevice_notifier(&ip6_mr_notifier); 1383 if (err) 1384 goto reg_notif_fail; 1385 #ifdef CONFIG_IPV6_PIMSM_V2 1386 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1387 pr_err("%s: can't add PIM protocol\n", __func__); 1388 err = -EAGAIN; 1389 goto add_proto_fail; 1390 } 1391 #endif 1392 err = rtnl_register_module(THIS_MODULE, RTNL_FAMILY_IP6MR, RTM_GETROUTE, 1393 ip6mr_rtm_getroute, ip6mr_rtm_dumproute, 0); 1394 if (err == 0) 1395 return 0; 1396 1397 #ifdef CONFIG_IPV6_PIMSM_V2 1398 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1399 add_proto_fail: 1400 unregister_netdevice_notifier(&ip6_mr_notifier); 1401 #endif 1402 reg_notif_fail: 1403 unregister_pernet_subsys(&ip6mr_net_ops); 1404 reg_pernet_fail: 1405 kmem_cache_destroy(mrt_cachep); 1406 return err; 1407 } 1408 1409 void ip6_mr_cleanup(void) 1410 { 1411 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1412 #ifdef CONFIG_IPV6_PIMSM_V2 1413 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1414 #endif 1415 unregister_netdevice_notifier(&ip6_mr_notifier); 1416 unregister_pernet_subsys(&ip6mr_net_ops); 1417 kmem_cache_destroy(mrt_cachep); 1418 } 1419 1420 static int ip6mr_mfc_add(struct net *net, struct mr_table *mrt, 1421 struct mf6cctl *mfc, int mrtsock, int parent) 1422 { 1423 unsigned char ttls[MAXMIFS]; 1424 struct mfc6_cache *uc, *c; 1425 struct mr_mfc *_uc; 1426 bool found; 1427 int i, err; 1428 1429 if (mfc->mf6cc_parent >= MAXMIFS) 1430 return -ENFILE; 1431 1432 memset(ttls, 255, MAXMIFS); 1433 for (i = 0; i < MAXMIFS; i++) { 1434 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1435 ttls[i] = 1; 1436 } 1437 1438 /* The entries are added/deleted only under RTNL */ 1439 rcu_read_lock(); 1440 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr, 1441 &mfc->mf6cc_mcastgrp.sin6_addr, parent); 1442 rcu_read_unlock(); 1443 if (c) { 1444 spin_lock(&mrt_lock); 1445 c->_c.mfc_parent = mfc->mf6cc_parent; 1446 ip6mr_update_thresholds(mrt, &c->_c, ttls); 1447 if (!mrtsock) 1448 c->_c.mfc_flags |= MFC_STATIC; 1449 spin_unlock(&mrt_lock); 1450 call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, 1451 c, mrt->id); 1452 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1453 return 0; 1454 } 1455 1456 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1457 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1458 return -EINVAL; 1459 1460 c = ip6mr_cache_alloc(); 1461 if (!c) 1462 return -ENOMEM; 1463 1464 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1465 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1466 c->_c.mfc_parent = mfc->mf6cc_parent; 1467 ip6mr_update_thresholds(mrt, &c->_c, ttls); 1468 if (!mrtsock) 1469 c->_c.mfc_flags |= MFC_STATIC; 1470 1471 err = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode, 1472 ip6mr_rht_params); 1473 if (err) { 1474 pr_err("ip6mr: rhtable insert error %d\n", err); 1475 ip6mr_cache_free(c); 1476 return err; 1477 } 1478 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list); 1479 1480 /* Check to see if we resolved a queued list. If so we 1481 * need to send on the frames and tidy up. 1482 */ 1483 found = false; 1484 spin_lock_bh(&mfc_unres_lock); 1485 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) { 1486 uc = (struct mfc6_cache *)_uc; 1487 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1488 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1489 list_del(&_uc->list); 1490 atomic_dec(&mrt->cache_resolve_queue_len); 1491 found = true; 1492 break; 1493 } 1494 } 1495 if (list_empty(&mrt->mfc_unres_queue)) 1496 del_timer(&mrt->ipmr_expire_timer); 1497 spin_unlock_bh(&mfc_unres_lock); 1498 1499 if (found) { 1500 ip6mr_cache_resolve(net, mrt, uc, c); 1501 ip6mr_cache_free(uc); 1502 } 1503 call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, 1504 c, mrt->id); 1505 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1506 return 0; 1507 } 1508 1509 /* 1510 * Close the multicast socket, and clear the vif tables etc 1511 */ 1512 1513 static void mroute_clean_tables(struct mr_table *mrt, int flags) 1514 { 1515 struct mr_mfc *c, *tmp; 1516 LIST_HEAD(list); 1517 int i; 1518 1519 /* Shut down all active vif entries */ 1520 if (flags & (MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC)) { 1521 for (i = 0; i < mrt->maxvif; i++) { 1522 if (((mrt->vif_table[i].flags & VIFF_STATIC) && 1523 !(flags & MRT6_FLUSH_MIFS_STATIC)) || 1524 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT6_FLUSH_MIFS))) 1525 continue; 1526 mif6_delete(mrt, i, 0, &list); 1527 } 1528 unregister_netdevice_many(&list); 1529 } 1530 1531 /* Wipe the cache */ 1532 if (flags & (MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC)) { 1533 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1534 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC_STATIC)) || 1535 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC))) 1536 continue; 1537 rhltable_remove(&mrt->mfc_hash, &c->mnode, ip6mr_rht_params); 1538 list_del_rcu(&c->list); 1539 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net), 1540 FIB_EVENT_ENTRY_DEL, 1541 (struct mfc6_cache *)c, mrt->id); 1542 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE); 1543 mr_cache_put(c); 1544 } 1545 } 1546 1547 if (flags & MRT6_FLUSH_MFC) { 1548 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1549 spin_lock_bh(&mfc_unres_lock); 1550 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1551 list_del(&c->list); 1552 mr6_netlink_event(mrt, (struct mfc6_cache *)c, 1553 RTM_DELROUTE); 1554 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c); 1555 } 1556 spin_unlock_bh(&mfc_unres_lock); 1557 } 1558 } 1559 } 1560 1561 static int ip6mr_sk_init(struct mr_table *mrt, struct sock *sk) 1562 { 1563 int err = 0; 1564 struct net *net = sock_net(sk); 1565 1566 rtnl_lock(); 1567 spin_lock(&mrt_lock); 1568 if (rtnl_dereference(mrt->mroute_sk)) { 1569 err = -EADDRINUSE; 1570 } else { 1571 rcu_assign_pointer(mrt->mroute_sk, sk); 1572 sock_set_flag(sk, SOCK_RCU_FREE); 1573 atomic_inc(&net->ipv6.devconf_all->mc_forwarding); 1574 } 1575 spin_unlock(&mrt_lock); 1576 1577 if (!err) 1578 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1579 NETCONFA_MC_FORWARDING, 1580 NETCONFA_IFINDEX_ALL, 1581 net->ipv6.devconf_all); 1582 rtnl_unlock(); 1583 1584 return err; 1585 } 1586 1587 int ip6mr_sk_done(struct sock *sk) 1588 { 1589 struct net *net = sock_net(sk); 1590 struct ipv6_devconf *devconf; 1591 struct mr_table *mrt; 1592 int err = -EACCES; 1593 1594 if (sk->sk_type != SOCK_RAW || 1595 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1596 return err; 1597 1598 devconf = net->ipv6.devconf_all; 1599 if (!devconf || !atomic_read(&devconf->mc_forwarding)) 1600 return err; 1601 1602 rtnl_lock(); 1603 ip6mr_for_each_table(mrt, net) { 1604 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1605 spin_lock(&mrt_lock); 1606 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1607 /* Note that mroute_sk had SOCK_RCU_FREE set, 1608 * so the RCU grace period before sk freeing 1609 * is guaranteed by sk_destruct() 1610 */ 1611 atomic_dec(&devconf->mc_forwarding); 1612 spin_unlock(&mrt_lock); 1613 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1614 NETCONFA_MC_FORWARDING, 1615 NETCONFA_IFINDEX_ALL, 1616 net->ipv6.devconf_all); 1617 1618 mroute_clean_tables(mrt, MRT6_FLUSH_MIFS | MRT6_FLUSH_MFC); 1619 err = 0; 1620 break; 1621 } 1622 } 1623 rtnl_unlock(); 1624 1625 return err; 1626 } 1627 1628 bool mroute6_is_socket(struct net *net, struct sk_buff *skb) 1629 { 1630 struct mr_table *mrt; 1631 struct flowi6 fl6 = { 1632 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1633 .flowi6_oif = skb->dev->ifindex, 1634 .flowi6_mark = skb->mark, 1635 }; 1636 1637 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1638 return NULL; 1639 1640 return rcu_access_pointer(mrt->mroute_sk); 1641 } 1642 EXPORT_SYMBOL(mroute6_is_socket); 1643 1644 /* 1645 * Socket options and virtual interface manipulation. The whole 1646 * virtual interface system is a complete heap, but unfortunately 1647 * that's how BSD mrouted happens to think. Maybe one day with a proper 1648 * MOSPF/PIM router set up we can clean this up. 1649 */ 1650 1651 int ip6_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval, 1652 unsigned int optlen) 1653 { 1654 int ret, parent = 0; 1655 struct mif6ctl vif; 1656 struct mf6cctl mfc; 1657 mifi_t mifi; 1658 struct net *net = sock_net(sk); 1659 struct mr_table *mrt; 1660 1661 if (sk->sk_type != SOCK_RAW || 1662 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1663 return -EOPNOTSUPP; 1664 1665 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1666 if (!mrt) 1667 return -ENOENT; 1668 1669 if (optname != MRT6_INIT) { 1670 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1671 !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1672 return -EACCES; 1673 } 1674 1675 switch (optname) { 1676 case MRT6_INIT: 1677 if (optlen < sizeof(int)) 1678 return -EINVAL; 1679 1680 return ip6mr_sk_init(mrt, sk); 1681 1682 case MRT6_DONE: 1683 return ip6mr_sk_done(sk); 1684 1685 case MRT6_ADD_MIF: 1686 if (optlen < sizeof(vif)) 1687 return -EINVAL; 1688 if (copy_from_sockptr(&vif, optval, sizeof(vif))) 1689 return -EFAULT; 1690 if (vif.mif6c_mifi >= MAXMIFS) 1691 return -ENFILE; 1692 rtnl_lock(); 1693 ret = mif6_add(net, mrt, &vif, 1694 sk == rtnl_dereference(mrt->mroute_sk)); 1695 rtnl_unlock(); 1696 return ret; 1697 1698 case MRT6_DEL_MIF: 1699 if (optlen < sizeof(mifi_t)) 1700 return -EINVAL; 1701 if (copy_from_sockptr(&mifi, optval, sizeof(mifi_t))) 1702 return -EFAULT; 1703 rtnl_lock(); 1704 ret = mif6_delete(mrt, mifi, 0, NULL); 1705 rtnl_unlock(); 1706 return ret; 1707 1708 /* 1709 * Manipulate the forwarding caches. These live 1710 * in a sort of kernel/user symbiosis. 1711 */ 1712 case MRT6_ADD_MFC: 1713 case MRT6_DEL_MFC: 1714 parent = -1; 1715 fallthrough; 1716 case MRT6_ADD_MFC_PROXY: 1717 case MRT6_DEL_MFC_PROXY: 1718 if (optlen < sizeof(mfc)) 1719 return -EINVAL; 1720 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) 1721 return -EFAULT; 1722 if (parent == 0) 1723 parent = mfc.mf6cc_parent; 1724 rtnl_lock(); 1725 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1726 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1727 else 1728 ret = ip6mr_mfc_add(net, mrt, &mfc, 1729 sk == 1730 rtnl_dereference(mrt->mroute_sk), 1731 parent); 1732 rtnl_unlock(); 1733 return ret; 1734 1735 case MRT6_FLUSH: 1736 { 1737 int flags; 1738 1739 if (optlen != sizeof(flags)) 1740 return -EINVAL; 1741 if (copy_from_sockptr(&flags, optval, sizeof(flags))) 1742 return -EFAULT; 1743 rtnl_lock(); 1744 mroute_clean_tables(mrt, flags); 1745 rtnl_unlock(); 1746 return 0; 1747 } 1748 1749 /* 1750 * Control PIM assert (to activate pim will activate assert) 1751 */ 1752 case MRT6_ASSERT: 1753 { 1754 int v; 1755 1756 if (optlen != sizeof(v)) 1757 return -EINVAL; 1758 if (copy_from_sockptr(&v, optval, sizeof(v))) 1759 return -EFAULT; 1760 mrt->mroute_do_assert = v; 1761 return 0; 1762 } 1763 1764 #ifdef CONFIG_IPV6_PIMSM_V2 1765 case MRT6_PIM: 1766 { 1767 bool do_wrmifwhole; 1768 int v; 1769 1770 if (optlen != sizeof(v)) 1771 return -EINVAL; 1772 if (copy_from_sockptr(&v, optval, sizeof(v))) 1773 return -EFAULT; 1774 1775 do_wrmifwhole = (v == MRT6MSG_WRMIFWHOLE); 1776 v = !!v; 1777 rtnl_lock(); 1778 ret = 0; 1779 if (v != mrt->mroute_do_pim) { 1780 mrt->mroute_do_pim = v; 1781 mrt->mroute_do_assert = v; 1782 mrt->mroute_do_wrvifwhole = do_wrmifwhole; 1783 } 1784 rtnl_unlock(); 1785 return ret; 1786 } 1787 1788 #endif 1789 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1790 case MRT6_TABLE: 1791 { 1792 u32 v; 1793 1794 if (optlen != sizeof(u32)) 1795 return -EINVAL; 1796 if (copy_from_sockptr(&v, optval, sizeof(v))) 1797 return -EFAULT; 1798 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1799 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1800 return -EINVAL; 1801 if (sk == rcu_access_pointer(mrt->mroute_sk)) 1802 return -EBUSY; 1803 1804 rtnl_lock(); 1805 ret = 0; 1806 mrt = ip6mr_new_table(net, v); 1807 if (IS_ERR(mrt)) 1808 ret = PTR_ERR(mrt); 1809 else 1810 raw6_sk(sk)->ip6mr_table = v; 1811 rtnl_unlock(); 1812 return ret; 1813 } 1814 #endif 1815 /* 1816 * Spurious command, or MRT6_VERSION which you cannot 1817 * set. 1818 */ 1819 default: 1820 return -ENOPROTOOPT; 1821 } 1822 } 1823 1824 /* 1825 * Getsock opt support for the multicast routing system. 1826 */ 1827 1828 int ip6_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval, 1829 sockptr_t optlen) 1830 { 1831 int olr; 1832 int val; 1833 struct net *net = sock_net(sk); 1834 struct mr_table *mrt; 1835 1836 if (sk->sk_type != SOCK_RAW || 1837 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1838 return -EOPNOTSUPP; 1839 1840 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1841 if (!mrt) 1842 return -ENOENT; 1843 1844 switch (optname) { 1845 case MRT6_VERSION: 1846 val = 0x0305; 1847 break; 1848 #ifdef CONFIG_IPV6_PIMSM_V2 1849 case MRT6_PIM: 1850 val = mrt->mroute_do_pim; 1851 break; 1852 #endif 1853 case MRT6_ASSERT: 1854 val = mrt->mroute_do_assert; 1855 break; 1856 default: 1857 return -ENOPROTOOPT; 1858 } 1859 1860 if (copy_from_sockptr(&olr, optlen, sizeof(int))) 1861 return -EFAULT; 1862 1863 olr = min_t(int, olr, sizeof(int)); 1864 if (olr < 0) 1865 return -EINVAL; 1866 1867 if (copy_to_sockptr(optlen, &olr, sizeof(int))) 1868 return -EFAULT; 1869 if (copy_to_sockptr(optval, &val, olr)) 1870 return -EFAULT; 1871 return 0; 1872 } 1873 1874 /* 1875 * The IP multicast ioctl support routines. 1876 */ 1877 int ip6mr_ioctl(struct sock *sk, int cmd, void *arg) 1878 { 1879 struct sioc_sg_req6 *sr; 1880 struct sioc_mif_req6 *vr; 1881 struct vif_device *vif; 1882 struct mfc6_cache *c; 1883 struct net *net = sock_net(sk); 1884 struct mr_table *mrt; 1885 1886 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1887 if (!mrt) 1888 return -ENOENT; 1889 1890 switch (cmd) { 1891 case SIOCGETMIFCNT_IN6: 1892 vr = (struct sioc_mif_req6 *)arg; 1893 if (vr->mifi >= mrt->maxvif) 1894 return -EINVAL; 1895 vr->mifi = array_index_nospec(vr->mifi, mrt->maxvif); 1896 rcu_read_lock(); 1897 vif = &mrt->vif_table[vr->mifi]; 1898 if (VIF_EXISTS(mrt, vr->mifi)) { 1899 vr->icount = READ_ONCE(vif->pkt_in); 1900 vr->ocount = READ_ONCE(vif->pkt_out); 1901 vr->ibytes = READ_ONCE(vif->bytes_in); 1902 vr->obytes = READ_ONCE(vif->bytes_out); 1903 rcu_read_unlock(); 1904 return 0; 1905 } 1906 rcu_read_unlock(); 1907 return -EADDRNOTAVAIL; 1908 case SIOCGETSGCNT_IN6: 1909 sr = (struct sioc_sg_req6 *)arg; 1910 1911 rcu_read_lock(); 1912 c = ip6mr_cache_find(mrt, &sr->src.sin6_addr, 1913 &sr->grp.sin6_addr); 1914 if (c) { 1915 sr->pktcnt = c->_c.mfc_un.res.pkt; 1916 sr->bytecnt = c->_c.mfc_un.res.bytes; 1917 sr->wrong_if = c->_c.mfc_un.res.wrong_if; 1918 rcu_read_unlock(); 1919 return 0; 1920 } 1921 rcu_read_unlock(); 1922 return -EADDRNOTAVAIL; 1923 default: 1924 return -ENOIOCTLCMD; 1925 } 1926 } 1927 1928 #ifdef CONFIG_COMPAT 1929 struct compat_sioc_sg_req6 { 1930 struct sockaddr_in6 src; 1931 struct sockaddr_in6 grp; 1932 compat_ulong_t pktcnt; 1933 compat_ulong_t bytecnt; 1934 compat_ulong_t wrong_if; 1935 }; 1936 1937 struct compat_sioc_mif_req6 { 1938 mifi_t mifi; 1939 compat_ulong_t icount; 1940 compat_ulong_t ocount; 1941 compat_ulong_t ibytes; 1942 compat_ulong_t obytes; 1943 }; 1944 1945 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1946 { 1947 struct compat_sioc_sg_req6 sr; 1948 struct compat_sioc_mif_req6 vr; 1949 struct vif_device *vif; 1950 struct mfc6_cache *c; 1951 struct net *net = sock_net(sk); 1952 struct mr_table *mrt; 1953 1954 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1955 if (!mrt) 1956 return -ENOENT; 1957 1958 switch (cmd) { 1959 case SIOCGETMIFCNT_IN6: 1960 if (copy_from_user(&vr, arg, sizeof(vr))) 1961 return -EFAULT; 1962 if (vr.mifi >= mrt->maxvif) 1963 return -EINVAL; 1964 vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif); 1965 rcu_read_lock(); 1966 vif = &mrt->vif_table[vr.mifi]; 1967 if (VIF_EXISTS(mrt, vr.mifi)) { 1968 vr.icount = READ_ONCE(vif->pkt_in); 1969 vr.ocount = READ_ONCE(vif->pkt_out); 1970 vr.ibytes = READ_ONCE(vif->bytes_in); 1971 vr.obytes = READ_ONCE(vif->bytes_out); 1972 rcu_read_unlock(); 1973 1974 if (copy_to_user(arg, &vr, sizeof(vr))) 1975 return -EFAULT; 1976 return 0; 1977 } 1978 rcu_read_unlock(); 1979 return -EADDRNOTAVAIL; 1980 case SIOCGETSGCNT_IN6: 1981 if (copy_from_user(&sr, arg, sizeof(sr))) 1982 return -EFAULT; 1983 1984 rcu_read_lock(); 1985 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1986 if (c) { 1987 sr.pktcnt = c->_c.mfc_un.res.pkt; 1988 sr.bytecnt = c->_c.mfc_un.res.bytes; 1989 sr.wrong_if = c->_c.mfc_un.res.wrong_if; 1990 rcu_read_unlock(); 1991 1992 if (copy_to_user(arg, &sr, sizeof(sr))) 1993 return -EFAULT; 1994 return 0; 1995 } 1996 rcu_read_unlock(); 1997 return -EADDRNOTAVAIL; 1998 default: 1999 return -ENOIOCTLCMD; 2000 } 2001 } 2002 #endif 2003 2004 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 2005 { 2006 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 2007 IPSTATS_MIB_OUTFORWDATAGRAMS); 2008 return dst_output(net, sk, skb); 2009 } 2010 2011 /* 2012 * Processing handlers for ip6mr_forward 2013 */ 2014 2015 static int ip6mr_forward2(struct net *net, struct mr_table *mrt, 2016 struct sk_buff *skb, int vifi) 2017 { 2018 struct vif_device *vif = &mrt->vif_table[vifi]; 2019 struct net_device *vif_dev; 2020 struct ipv6hdr *ipv6h; 2021 struct dst_entry *dst; 2022 struct flowi6 fl6; 2023 2024 vif_dev = vif_dev_read(vif); 2025 if (!vif_dev) 2026 goto out_free; 2027 2028 #ifdef CONFIG_IPV6_PIMSM_V2 2029 if (vif->flags & MIFF_REGISTER) { 2030 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 2031 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 2032 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len); 2033 DEV_STATS_INC(vif_dev, tx_packets); 2034 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2035 goto out_free; 2036 } 2037 #endif 2038 2039 ipv6h = ipv6_hdr(skb); 2040 2041 fl6 = (struct flowi6) { 2042 .flowi6_oif = vif->link, 2043 .daddr = ipv6h->daddr, 2044 }; 2045 2046 dst = ip6_route_output(net, NULL, &fl6); 2047 if (dst->error) { 2048 dst_release(dst); 2049 goto out_free; 2050 } 2051 2052 skb_dst_drop(skb); 2053 skb_dst_set(skb, dst); 2054 2055 /* 2056 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2057 * not only before forwarding, but after forwarding on all output 2058 * interfaces. It is clear, if mrouter runs a multicasting 2059 * program, it should receive packets not depending to what interface 2060 * program is joined. 2061 * If we will not make it, the program will have to join on all 2062 * interfaces. On the other hand, multihoming host (or router, but 2063 * not mrouter) cannot join to more than one interface - it will 2064 * result in receiving multiple packets. 2065 */ 2066 skb->dev = vif_dev; 2067 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 2068 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 2069 2070 /* We are about to write */ 2071 /* XXX: extension headers? */ 2072 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(vif_dev))) 2073 goto out_free; 2074 2075 ipv6h = ipv6_hdr(skb); 2076 ipv6h->hop_limit--; 2077 2078 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2079 2080 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2081 net, NULL, skb, skb->dev, vif_dev, 2082 ip6mr_forward2_finish); 2083 2084 out_free: 2085 kfree_skb(skb); 2086 return 0; 2087 } 2088 2089 /* Called with rcu_read_lock() */ 2090 static int ip6mr_find_vif(struct mr_table *mrt, struct net_device *dev) 2091 { 2092 int ct; 2093 2094 /* Pairs with WRITE_ONCE() in mif6_delete()/mif6_add() */ 2095 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) { 2096 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev) 2097 break; 2098 } 2099 return ct; 2100 } 2101 2102 /* Called under rcu_read_lock() */ 2103 static void ip6_mr_forward(struct net *net, struct mr_table *mrt, 2104 struct net_device *dev, struct sk_buff *skb, 2105 struct mfc6_cache *c) 2106 { 2107 int psend = -1; 2108 int vif, ct; 2109 int true_vifi = ip6mr_find_vif(mrt, dev); 2110 2111 vif = c->_c.mfc_parent; 2112 c->_c.mfc_un.res.pkt++; 2113 c->_c.mfc_un.res.bytes += skb->len; 2114 c->_c.mfc_un.res.lastuse = jiffies; 2115 2116 if (ipv6_addr_any(&c->mf6c_origin) && true_vifi >= 0) { 2117 struct mfc6_cache *cache_proxy; 2118 2119 /* For an (*,G) entry, we only check that the incoming 2120 * interface is part of the static tree. 2121 */ 2122 cache_proxy = mr_mfc_find_any_parent(mrt, vif); 2123 if (cache_proxy && 2124 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) 2125 goto forward; 2126 } 2127 2128 /* 2129 * Wrong interface: drop packet and (maybe) send PIM assert. 2130 */ 2131 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) { 2132 c->_c.mfc_un.res.wrong_if++; 2133 2134 if (true_vifi >= 0 && mrt->mroute_do_assert && 2135 /* pimsm uses asserts, when switching from RPT to SPT, 2136 so that we cannot check that packet arrived on an oif. 2137 It is bad, but otherwise we would need to move pretty 2138 large chunk of pimd to kernel. Ough... --ANK 2139 */ 2140 (mrt->mroute_do_pim || 2141 c->_c.mfc_un.res.ttls[true_vifi] < 255) && 2142 time_after(jiffies, 2143 c->_c.mfc_un.res.last_assert + 2144 MFC_ASSERT_THRESH)) { 2145 c->_c.mfc_un.res.last_assert = jiffies; 2146 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2147 if (mrt->mroute_do_wrvifwhole) 2148 ip6mr_cache_report(mrt, skb, true_vifi, 2149 MRT6MSG_WRMIFWHOLE); 2150 } 2151 goto dont_forward; 2152 } 2153 2154 forward: 2155 WRITE_ONCE(mrt->vif_table[vif].pkt_in, 2156 mrt->vif_table[vif].pkt_in + 1); 2157 WRITE_ONCE(mrt->vif_table[vif].bytes_in, 2158 mrt->vif_table[vif].bytes_in + skb->len); 2159 2160 /* 2161 * Forward the frame 2162 */ 2163 if (ipv6_addr_any(&c->mf6c_origin) && 2164 ipv6_addr_any(&c->mf6c_mcastgrp)) { 2165 if (true_vifi >= 0 && 2166 true_vifi != c->_c.mfc_parent && 2167 ipv6_hdr(skb)->hop_limit > 2168 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) { 2169 /* It's an (*,*) entry and the packet is not coming from 2170 * the upstream: forward the packet to the upstream 2171 * only. 2172 */ 2173 psend = c->_c.mfc_parent; 2174 goto last_forward; 2175 } 2176 goto dont_forward; 2177 } 2178 for (ct = c->_c.mfc_un.res.maxvif - 1; 2179 ct >= c->_c.mfc_un.res.minvif; ct--) { 2180 /* For (*,G) entry, don't forward to the incoming interface */ 2181 if ((!ipv6_addr_any(&c->mf6c_origin) || ct != true_vifi) && 2182 ipv6_hdr(skb)->hop_limit > c->_c.mfc_un.res.ttls[ct]) { 2183 if (psend != -1) { 2184 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2185 if (skb2) 2186 ip6mr_forward2(net, mrt, skb2, psend); 2187 } 2188 psend = ct; 2189 } 2190 } 2191 last_forward: 2192 if (psend != -1) { 2193 ip6mr_forward2(net, mrt, skb, psend); 2194 return; 2195 } 2196 2197 dont_forward: 2198 kfree_skb(skb); 2199 } 2200 2201 2202 /* 2203 * Multicast packets for forwarding arrive here 2204 */ 2205 2206 int ip6_mr_input(struct sk_buff *skb) 2207 { 2208 struct mfc6_cache *cache; 2209 struct net *net = dev_net(skb->dev); 2210 struct mr_table *mrt; 2211 struct flowi6 fl6 = { 2212 .flowi6_iif = skb->dev->ifindex, 2213 .flowi6_mark = skb->mark, 2214 }; 2215 int err; 2216 struct net_device *dev; 2217 2218 /* skb->dev passed in is the master dev for vrfs. 2219 * Get the proper interface that does have a vif associated with it. 2220 */ 2221 dev = skb->dev; 2222 if (netif_is_l3_master(skb->dev)) { 2223 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 2224 if (!dev) { 2225 kfree_skb(skb); 2226 return -ENODEV; 2227 } 2228 } 2229 2230 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2231 if (err < 0) { 2232 kfree_skb(skb); 2233 return err; 2234 } 2235 2236 cache = ip6mr_cache_find(mrt, 2237 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2238 if (!cache) { 2239 int vif = ip6mr_find_vif(mrt, dev); 2240 2241 if (vif >= 0) 2242 cache = ip6mr_cache_find_any(mrt, 2243 &ipv6_hdr(skb)->daddr, 2244 vif); 2245 } 2246 2247 /* 2248 * No usable cache entry 2249 */ 2250 if (!cache) { 2251 int vif; 2252 2253 vif = ip6mr_find_vif(mrt, dev); 2254 if (vif >= 0) { 2255 int err = ip6mr_cache_unresolved(mrt, vif, skb, dev); 2256 2257 return err; 2258 } 2259 kfree_skb(skb); 2260 return -ENODEV; 2261 } 2262 2263 ip6_mr_forward(net, mrt, dev, skb, cache); 2264 2265 return 0; 2266 } 2267 2268 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2269 u32 portid) 2270 { 2271 int err; 2272 struct mr_table *mrt; 2273 struct mfc6_cache *cache; 2274 struct rt6_info *rt = dst_rt6_info(skb_dst(skb)); 2275 2276 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2277 if (!mrt) 2278 return -ENOENT; 2279 2280 rcu_read_lock(); 2281 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2282 if (!cache && skb->dev) { 2283 int vif = ip6mr_find_vif(mrt, skb->dev); 2284 2285 if (vif >= 0) 2286 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2287 vif); 2288 } 2289 2290 if (!cache) { 2291 struct sk_buff *skb2; 2292 struct ipv6hdr *iph; 2293 struct net_device *dev; 2294 int vif; 2295 2296 dev = skb->dev; 2297 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2298 rcu_read_unlock(); 2299 return -ENODEV; 2300 } 2301 2302 /* really correct? */ 2303 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2304 if (!skb2) { 2305 rcu_read_unlock(); 2306 return -ENOMEM; 2307 } 2308 2309 NETLINK_CB(skb2).portid = portid; 2310 skb_reset_transport_header(skb2); 2311 2312 skb_put(skb2, sizeof(struct ipv6hdr)); 2313 skb_reset_network_header(skb2); 2314 2315 iph = ipv6_hdr(skb2); 2316 iph->version = 0; 2317 iph->priority = 0; 2318 iph->flow_lbl[0] = 0; 2319 iph->flow_lbl[1] = 0; 2320 iph->flow_lbl[2] = 0; 2321 iph->payload_len = 0; 2322 iph->nexthdr = IPPROTO_NONE; 2323 iph->hop_limit = 0; 2324 iph->saddr = rt->rt6i_src.addr; 2325 iph->daddr = rt->rt6i_dst.addr; 2326 2327 err = ip6mr_cache_unresolved(mrt, vif, skb2, dev); 2328 rcu_read_unlock(); 2329 2330 return err; 2331 } 2332 2333 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm); 2334 rcu_read_unlock(); 2335 return err; 2336 } 2337 2338 static int ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2339 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2340 int flags) 2341 { 2342 struct nlmsghdr *nlh; 2343 struct rtmsg *rtm; 2344 int err; 2345 2346 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2347 if (!nlh) 2348 return -EMSGSIZE; 2349 2350 rtm = nlmsg_data(nlh); 2351 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2352 rtm->rtm_dst_len = 128; 2353 rtm->rtm_src_len = 128; 2354 rtm->rtm_tos = 0; 2355 rtm->rtm_table = mrt->id; 2356 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2357 goto nla_put_failure; 2358 rtm->rtm_type = RTN_MULTICAST; 2359 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2360 if (c->_c.mfc_flags & MFC_STATIC) 2361 rtm->rtm_protocol = RTPROT_STATIC; 2362 else 2363 rtm->rtm_protocol = RTPROT_MROUTED; 2364 rtm->rtm_flags = 0; 2365 2366 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2367 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2368 goto nla_put_failure; 2369 err = mr_fill_mroute(mrt, skb, &c->_c, rtm); 2370 /* do not break the dump if cache is unresolved */ 2371 if (err < 0 && err != -ENOENT) 2372 goto nla_put_failure; 2373 2374 nlmsg_end(skb, nlh); 2375 return 0; 2376 2377 nla_put_failure: 2378 nlmsg_cancel(skb, nlh); 2379 return -EMSGSIZE; 2380 } 2381 2382 static int _ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2383 u32 portid, u32 seq, struct mr_mfc *c, 2384 int cmd, int flags) 2385 { 2386 return ip6mr_fill_mroute(mrt, skb, portid, seq, (struct mfc6_cache *)c, 2387 cmd, flags); 2388 } 2389 2390 static int mr6_msgsize(bool unresolved, int maxvif) 2391 { 2392 size_t len = 2393 NLMSG_ALIGN(sizeof(struct rtmsg)) 2394 + nla_total_size(4) /* RTA_TABLE */ 2395 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2396 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2397 ; 2398 2399 if (!unresolved) 2400 len = len 2401 + nla_total_size(4) /* RTA_IIF */ 2402 + nla_total_size(0) /* RTA_MULTIPATH */ 2403 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2404 /* RTA_MFC_STATS */ 2405 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2406 ; 2407 2408 return len; 2409 } 2410 2411 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc, 2412 int cmd) 2413 { 2414 struct net *net = read_pnet(&mrt->net); 2415 struct sk_buff *skb; 2416 int err = -ENOBUFS; 2417 2418 skb = nlmsg_new(mr6_msgsize(mfc->_c.mfc_parent >= MAXMIFS, mrt->maxvif), 2419 GFP_ATOMIC); 2420 if (!skb) 2421 goto errout; 2422 2423 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2424 if (err < 0) 2425 goto errout; 2426 2427 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2428 return; 2429 2430 errout: 2431 kfree_skb(skb); 2432 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2433 } 2434 2435 static size_t mrt6msg_netlink_msgsize(size_t payloadlen) 2436 { 2437 size_t len = 2438 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2439 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */ 2440 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */ 2441 /* IP6MRA_CREPORT_SRC_ADDR */ 2442 + nla_total_size(sizeof(struct in6_addr)) 2443 /* IP6MRA_CREPORT_DST_ADDR */ 2444 + nla_total_size(sizeof(struct in6_addr)) 2445 /* IP6MRA_CREPORT_PKT */ 2446 + nla_total_size(payloadlen) 2447 ; 2448 2449 return len; 2450 } 2451 2452 static void mrt6msg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt) 2453 { 2454 struct net *net = read_pnet(&mrt->net); 2455 struct nlmsghdr *nlh; 2456 struct rtgenmsg *rtgenm; 2457 struct mrt6msg *msg; 2458 struct sk_buff *skb; 2459 struct nlattr *nla; 2460 int payloadlen; 2461 2462 payloadlen = pkt->len - sizeof(struct mrt6msg); 2463 msg = (struct mrt6msg *)skb_transport_header(pkt); 2464 2465 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2466 if (!skb) 2467 goto errout; 2468 2469 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2470 sizeof(struct rtgenmsg), 0); 2471 if (!nlh) 2472 goto errout; 2473 rtgenm = nlmsg_data(nlh); 2474 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR; 2475 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) || 2476 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) || 2477 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR, 2478 &msg->im6_src) || 2479 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR, 2480 &msg->im6_dst)) 2481 goto nla_put_failure; 2482 2483 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen); 2484 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg), 2485 nla_data(nla), payloadlen)) 2486 goto nla_put_failure; 2487 2488 nlmsg_end(skb, nlh); 2489 2490 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC); 2491 return; 2492 2493 nla_put_failure: 2494 nlmsg_cancel(skb, nlh); 2495 errout: 2496 kfree_skb(skb); 2497 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS); 2498 } 2499 2500 static const struct nla_policy ip6mr_getroute_policy[RTA_MAX + 1] = { 2501 [RTA_SRC] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)), 2502 [RTA_DST] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)), 2503 [RTA_TABLE] = { .type = NLA_U32 }, 2504 }; 2505 2506 static int ip6mr_rtm_valid_getroute_req(struct sk_buff *skb, 2507 const struct nlmsghdr *nlh, 2508 struct nlattr **tb, 2509 struct netlink_ext_ack *extack) 2510 { 2511 struct rtmsg *rtm; 2512 int err; 2513 2514 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, ip6mr_getroute_policy, 2515 extack); 2516 if (err) 2517 return err; 2518 2519 rtm = nlmsg_data(nlh); 2520 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) || 2521 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) || 2522 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol || 2523 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) { 2524 NL_SET_ERR_MSG_MOD(extack, 2525 "Invalid values in header for multicast route get request"); 2526 return -EINVAL; 2527 } 2528 2529 if ((tb[RTA_SRC] && !rtm->rtm_src_len) || 2530 (tb[RTA_DST] && !rtm->rtm_dst_len)) { 2531 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6"); 2532 return -EINVAL; 2533 } 2534 2535 return 0; 2536 } 2537 2538 static int ip6mr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, 2539 struct netlink_ext_ack *extack) 2540 { 2541 struct net *net = sock_net(in_skb->sk); 2542 struct in6_addr src = {}, grp = {}; 2543 struct nlattr *tb[RTA_MAX + 1]; 2544 struct mfc6_cache *cache; 2545 struct mr_table *mrt; 2546 struct sk_buff *skb; 2547 u32 tableid; 2548 int err; 2549 2550 err = ip6mr_rtm_valid_getroute_req(in_skb, nlh, tb, extack); 2551 if (err < 0) 2552 return err; 2553 2554 if (tb[RTA_SRC]) 2555 src = nla_get_in6_addr(tb[RTA_SRC]); 2556 if (tb[RTA_DST]) 2557 grp = nla_get_in6_addr(tb[RTA_DST]); 2558 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0; 2559 2560 mrt = ip6mr_get_table(net, tableid ?: RT_TABLE_DEFAULT); 2561 if (!mrt) { 2562 NL_SET_ERR_MSG_MOD(extack, "MR table does not exist"); 2563 return -ENOENT; 2564 } 2565 2566 /* entries are added/deleted only under RTNL */ 2567 rcu_read_lock(); 2568 cache = ip6mr_cache_find(mrt, &src, &grp); 2569 rcu_read_unlock(); 2570 if (!cache) { 2571 NL_SET_ERR_MSG_MOD(extack, "MR cache entry not found"); 2572 return -ENOENT; 2573 } 2574 2575 skb = nlmsg_new(mr6_msgsize(false, mrt->maxvif), GFP_KERNEL); 2576 if (!skb) 2577 return -ENOBUFS; 2578 2579 err = ip6mr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid, 2580 nlh->nlmsg_seq, cache, RTM_NEWROUTE, 0); 2581 if (err < 0) { 2582 kfree_skb(skb); 2583 return err; 2584 } 2585 2586 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 2587 } 2588 2589 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2590 { 2591 const struct nlmsghdr *nlh = cb->nlh; 2592 struct fib_dump_filter filter = { 2593 .rtnl_held = true, 2594 }; 2595 int err; 2596 2597 if (cb->strict_check) { 2598 err = ip_valid_fib_dump_req(sock_net(skb->sk), nlh, 2599 &filter, cb); 2600 if (err < 0) 2601 return err; 2602 } 2603 2604 if (filter.table_id) { 2605 struct mr_table *mrt; 2606 2607 mrt = ip6mr_get_table(sock_net(skb->sk), filter.table_id); 2608 if (!mrt) { 2609 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IP6MR) 2610 return skb->len; 2611 2612 NL_SET_ERR_MSG_MOD(cb->extack, "MR table does not exist"); 2613 return -ENOENT; 2614 } 2615 err = mr_table_dump(mrt, skb, cb, _ip6mr_fill_mroute, 2616 &mfc_unres_lock, &filter); 2617 return skb->len ? : err; 2618 } 2619 2620 return mr_rtm_dumproute(skb, cb, ip6mr_mr_table_iter, 2621 _ip6mr_fill_mroute, &mfc_unres_lock, &filter); 2622 } 2623