1 /* 2 * Linux IPv6 multicast routing support for BSD pim6sd 3 * Based on net/ipv4/ipmr.c. 4 * 5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr> 6 * LSIIT Laboratory, Strasbourg, France 7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com> 8 * 6WIND, Paris, France 9 * Copyright (C)2007,2008 USAGI/WIDE Project 10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> 11 * 12 * This program is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU General Public License 14 * as published by the Free Software Foundation; either version 15 * 2 of the License, or (at your option) any later version. 16 * 17 */ 18 19 #include <linux/uaccess.h> 20 #include <linux/types.h> 21 #include <linux/sched.h> 22 #include <linux/errno.h> 23 #include <linux/timer.h> 24 #include <linux/mm.h> 25 #include <linux/kernel.h> 26 #include <linux/fcntl.h> 27 #include <linux/stat.h> 28 #include <linux/socket.h> 29 #include <linux/inet.h> 30 #include <linux/netdevice.h> 31 #include <linux/inetdevice.h> 32 #include <linux/proc_fs.h> 33 #include <linux/seq_file.h> 34 #include <linux/init.h> 35 #include <linux/slab.h> 36 #include <linux/compat.h> 37 #include <net/protocol.h> 38 #include <linux/skbuff.h> 39 #include <net/sock.h> 40 #include <net/raw.h> 41 #include <linux/notifier.h> 42 #include <linux/if_arp.h> 43 #include <net/checksum.h> 44 #include <net/netlink.h> 45 #include <net/fib_rules.h> 46 47 #include <net/ipv6.h> 48 #include <net/ip6_route.h> 49 #include <linux/mroute6.h> 50 #include <linux/pim.h> 51 #include <net/addrconf.h> 52 #include <linux/netfilter_ipv6.h> 53 #include <linux/export.h> 54 #include <net/ip6_checksum.h> 55 #include <linux/netconf.h> 56 57 struct mr6_table { 58 struct list_head list; 59 possible_net_t net; 60 u32 id; 61 struct sock *mroute6_sk; 62 struct timer_list ipmr_expire_timer; 63 struct list_head mfc6_unres_queue; 64 struct list_head mfc6_cache_array[MFC6_LINES]; 65 struct mif_device vif6_table[MAXMIFS]; 66 int maxvif; 67 atomic_t cache_resolve_queue_len; 68 bool mroute_do_assert; 69 bool mroute_do_pim; 70 #ifdef CONFIG_IPV6_PIMSM_V2 71 int mroute_reg_vif_num; 72 #endif 73 }; 74 75 struct ip6mr_rule { 76 struct fib_rule common; 77 }; 78 79 struct ip6mr_result { 80 struct mr6_table *mrt; 81 }; 82 83 /* Big lock, protecting vif table, mrt cache and mroute socket state. 84 Note that the changes are semaphored via rtnl_lock. 85 */ 86 87 static DEFINE_RWLOCK(mrt_lock); 88 89 /* 90 * Multicast router control variables 91 */ 92 93 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL) 94 95 /* Special spinlock for queue of unresolved entries */ 96 static DEFINE_SPINLOCK(mfc_unres_lock); 97 98 /* We return to original Alan's scheme. Hash table of resolved 99 entries is changed only in process context and protected 100 with weak lock mrt_lock. Queue of unresolved entries is protected 101 with strong spinlock mfc_unres_lock. 102 103 In this case data path is free of exclusive locks at all. 104 */ 105 106 static struct kmem_cache *mrt_cachep __read_mostly; 107 108 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id); 109 static void ip6mr_free_table(struct mr6_table *mrt); 110 111 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 112 struct sk_buff *skb, struct mfc6_cache *cache); 113 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 114 mifi_t mifi, int assert); 115 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 116 struct mfc6_cache *c, struct rtmsg *rtm); 117 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 118 int cmd); 119 static void mrt6msg_netlink_event(struct mr6_table *mrt, struct sk_buff *pkt); 120 static int ip6mr_rtm_dumproute(struct sk_buff *skb, 121 struct netlink_callback *cb); 122 static void mroute_clean_tables(struct mr6_table *mrt, bool all); 123 static void ipmr_expire_process(unsigned long arg); 124 125 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 126 #define ip6mr_for_each_table(mrt, net) \ 127 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list) 128 129 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 130 { 131 struct mr6_table *mrt; 132 133 ip6mr_for_each_table(mrt, net) { 134 if (mrt->id == id) 135 return mrt; 136 } 137 return NULL; 138 } 139 140 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 141 struct mr6_table **mrt) 142 { 143 int err; 144 struct ip6mr_result res; 145 struct fib_lookup_arg arg = { 146 .result = &res, 147 .flags = FIB_LOOKUP_NOREF, 148 }; 149 150 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, 151 flowi6_to_flowi(flp6), 0, &arg); 152 if (err < 0) 153 return err; 154 *mrt = res.mrt; 155 return 0; 156 } 157 158 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp, 159 int flags, struct fib_lookup_arg *arg) 160 { 161 struct ip6mr_result *res = arg->result; 162 struct mr6_table *mrt; 163 164 switch (rule->action) { 165 case FR_ACT_TO_TBL: 166 break; 167 case FR_ACT_UNREACHABLE: 168 return -ENETUNREACH; 169 case FR_ACT_PROHIBIT: 170 return -EACCES; 171 case FR_ACT_BLACKHOLE: 172 default: 173 return -EINVAL; 174 } 175 176 mrt = ip6mr_get_table(rule->fr_net, rule->table); 177 if (!mrt) 178 return -EAGAIN; 179 res->mrt = mrt; 180 return 0; 181 } 182 183 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags) 184 { 185 return 1; 186 } 187 188 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = { 189 FRA_GENERIC_POLICY, 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 { 195 return 0; 196 } 197 198 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 199 struct nlattr **tb) 200 { 201 return 1; 202 } 203 204 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 205 struct fib_rule_hdr *frh) 206 { 207 frh->dst_len = 0; 208 frh->src_len = 0; 209 frh->tos = 0; 210 return 0; 211 } 212 213 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = { 214 .family = RTNL_FAMILY_IP6MR, 215 .rule_size = sizeof(struct ip6mr_rule), 216 .addr_size = sizeof(struct in6_addr), 217 .action = ip6mr_rule_action, 218 .match = ip6mr_rule_match, 219 .configure = ip6mr_rule_configure, 220 .compare = ip6mr_rule_compare, 221 .fill = ip6mr_rule_fill, 222 .nlgroup = RTNLGRP_IPV6_RULE, 223 .policy = ip6mr_rule_policy, 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 mr6_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 (!mrt) { 241 err = -ENOMEM; 242 goto err1; 243 } 244 245 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0); 246 if (err < 0) 247 goto err2; 248 249 net->ipv6.mr6_rules_ops = ops; 250 return 0; 251 252 err2: 253 ip6mr_free_table(mrt); 254 err1: 255 fib_rules_unregister(ops); 256 return err; 257 } 258 259 static void __net_exit ip6mr_rules_exit(struct net *net) 260 { 261 struct mr6_table *mrt, *next; 262 263 rtnl_lock(); 264 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) { 265 list_del(&mrt->list); 266 ip6mr_free_table(mrt); 267 } 268 fib_rules_unregister(net->ipv6.mr6_rules_ops); 269 rtnl_unlock(); 270 } 271 #else 272 #define ip6mr_for_each_table(mrt, net) \ 273 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL) 274 275 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 276 { 277 return net->ipv6.mrt6; 278 } 279 280 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 281 struct mr6_table **mrt) 282 { 283 *mrt = net->ipv6.mrt6; 284 return 0; 285 } 286 287 static int __net_init ip6mr_rules_init(struct net *net) 288 { 289 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT); 290 return net->ipv6.mrt6 ? 0 : -ENOMEM; 291 } 292 293 static void __net_exit ip6mr_rules_exit(struct net *net) 294 { 295 rtnl_lock(); 296 ip6mr_free_table(net->ipv6.mrt6); 297 net->ipv6.mrt6 = NULL; 298 rtnl_unlock(); 299 } 300 #endif 301 302 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id) 303 { 304 struct mr6_table *mrt; 305 unsigned int i; 306 307 mrt = ip6mr_get_table(net, id); 308 if (mrt) 309 return mrt; 310 311 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 312 if (!mrt) 313 return NULL; 314 mrt->id = id; 315 write_pnet(&mrt->net, net); 316 317 /* Forwarding cache */ 318 for (i = 0; i < MFC6_LINES; i++) 319 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]); 320 321 INIT_LIST_HEAD(&mrt->mfc6_unres_queue); 322 323 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 324 (unsigned long)mrt); 325 326 #ifdef CONFIG_IPV6_PIMSM_V2 327 mrt->mroute_reg_vif_num = -1; 328 #endif 329 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 330 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables); 331 #endif 332 return mrt; 333 } 334 335 static void ip6mr_free_table(struct mr6_table *mrt) 336 { 337 del_timer_sync(&mrt->ipmr_expire_timer); 338 mroute_clean_tables(mrt, true); 339 kfree(mrt); 340 } 341 342 #ifdef CONFIG_PROC_FS 343 344 struct ipmr_mfc_iter { 345 struct seq_net_private p; 346 struct mr6_table *mrt; 347 struct list_head *cache; 348 int ct; 349 }; 350 351 352 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net, 353 struct ipmr_mfc_iter *it, loff_t pos) 354 { 355 struct mr6_table *mrt = it->mrt; 356 struct mfc6_cache *mfc; 357 358 read_lock(&mrt_lock); 359 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) { 360 it->cache = &mrt->mfc6_cache_array[it->ct]; 361 list_for_each_entry(mfc, it->cache, list) 362 if (pos-- == 0) 363 return mfc; 364 } 365 read_unlock(&mrt_lock); 366 367 spin_lock_bh(&mfc_unres_lock); 368 it->cache = &mrt->mfc6_unres_queue; 369 list_for_each_entry(mfc, it->cache, list) 370 if (pos-- == 0) 371 return mfc; 372 spin_unlock_bh(&mfc_unres_lock); 373 374 it->cache = NULL; 375 return NULL; 376 } 377 378 /* 379 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif 380 */ 381 382 struct ipmr_vif_iter { 383 struct seq_net_private p; 384 struct mr6_table *mrt; 385 int ct; 386 }; 387 388 static struct mif_device *ip6mr_vif_seq_idx(struct net *net, 389 struct ipmr_vif_iter *iter, 390 loff_t pos) 391 { 392 struct mr6_table *mrt = iter->mrt; 393 394 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 395 if (!MIF_EXISTS(mrt, iter->ct)) 396 continue; 397 if (pos-- == 0) 398 return &mrt->vif6_table[iter->ct]; 399 } 400 return NULL; 401 } 402 403 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) 404 __acquires(mrt_lock) 405 { 406 struct ipmr_vif_iter *iter = seq->private; 407 struct net *net = seq_file_net(seq); 408 struct mr6_table *mrt; 409 410 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 411 if (!mrt) 412 return ERR_PTR(-ENOENT); 413 414 iter->mrt = mrt; 415 416 read_lock(&mrt_lock); 417 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1) 418 : SEQ_START_TOKEN; 419 } 420 421 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 422 { 423 struct ipmr_vif_iter *iter = seq->private; 424 struct net *net = seq_file_net(seq); 425 struct mr6_table *mrt = iter->mrt; 426 427 ++*pos; 428 if (v == SEQ_START_TOKEN) 429 return ip6mr_vif_seq_idx(net, iter, 0); 430 431 while (++iter->ct < mrt->maxvif) { 432 if (!MIF_EXISTS(mrt, iter->ct)) 433 continue; 434 return &mrt->vif6_table[iter->ct]; 435 } 436 return NULL; 437 } 438 439 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) 440 __releases(mrt_lock) 441 { 442 read_unlock(&mrt_lock); 443 } 444 445 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) 446 { 447 struct ipmr_vif_iter *iter = seq->private; 448 struct mr6_table *mrt = iter->mrt; 449 450 if (v == SEQ_START_TOKEN) { 451 seq_puts(seq, 452 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); 453 } else { 454 const struct mif_device *vif = v; 455 const char *name = vif->dev ? vif->dev->name : "none"; 456 457 seq_printf(seq, 458 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", 459 vif - mrt->vif6_table, 460 name, vif->bytes_in, vif->pkt_in, 461 vif->bytes_out, vif->pkt_out, 462 vif->flags); 463 } 464 return 0; 465 } 466 467 static const struct seq_operations ip6mr_vif_seq_ops = { 468 .start = ip6mr_vif_seq_start, 469 .next = ip6mr_vif_seq_next, 470 .stop = ip6mr_vif_seq_stop, 471 .show = ip6mr_vif_seq_show, 472 }; 473 474 static int ip6mr_vif_open(struct inode *inode, struct file *file) 475 { 476 return seq_open_net(inode, file, &ip6mr_vif_seq_ops, 477 sizeof(struct ipmr_vif_iter)); 478 } 479 480 static const struct file_operations ip6mr_vif_fops = { 481 .owner = THIS_MODULE, 482 .open = ip6mr_vif_open, 483 .read = seq_read, 484 .llseek = seq_lseek, 485 .release = seq_release_net, 486 }; 487 488 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 489 { 490 struct ipmr_mfc_iter *it = seq->private; 491 struct net *net = seq_file_net(seq); 492 struct mr6_table *mrt; 493 494 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 495 if (!mrt) 496 return ERR_PTR(-ENOENT); 497 498 it->mrt = mrt; 499 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 500 : SEQ_START_TOKEN; 501 } 502 503 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 504 { 505 struct mfc6_cache *mfc = v; 506 struct ipmr_mfc_iter *it = seq->private; 507 struct net *net = seq_file_net(seq); 508 struct mr6_table *mrt = it->mrt; 509 510 ++*pos; 511 512 if (v == SEQ_START_TOKEN) 513 return ipmr_mfc_seq_idx(net, seq->private, 0); 514 515 if (mfc->list.next != it->cache) 516 return list_entry(mfc->list.next, struct mfc6_cache, list); 517 518 if (it->cache == &mrt->mfc6_unres_queue) 519 goto end_of_list; 520 521 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]); 522 523 while (++it->ct < MFC6_LINES) { 524 it->cache = &mrt->mfc6_cache_array[it->ct]; 525 if (list_empty(it->cache)) 526 continue; 527 return list_first_entry(it->cache, struct mfc6_cache, list); 528 } 529 530 /* exhausted cache_array, show unresolved */ 531 read_unlock(&mrt_lock); 532 it->cache = &mrt->mfc6_unres_queue; 533 it->ct = 0; 534 535 spin_lock_bh(&mfc_unres_lock); 536 if (!list_empty(it->cache)) 537 return list_first_entry(it->cache, struct mfc6_cache, list); 538 539 end_of_list: 540 spin_unlock_bh(&mfc_unres_lock); 541 it->cache = NULL; 542 543 return NULL; 544 } 545 546 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 547 { 548 struct ipmr_mfc_iter *it = seq->private; 549 struct mr6_table *mrt = it->mrt; 550 551 if (it->cache == &mrt->mfc6_unres_queue) 552 spin_unlock_bh(&mfc_unres_lock); 553 else if (it->cache == &mrt->mfc6_cache_array[it->ct]) 554 read_unlock(&mrt_lock); 555 } 556 557 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 558 { 559 int n; 560 561 if (v == SEQ_START_TOKEN) { 562 seq_puts(seq, 563 "Group " 564 "Origin " 565 "Iif Pkts Bytes Wrong Oifs\n"); 566 } else { 567 const struct mfc6_cache *mfc = v; 568 const struct ipmr_mfc_iter *it = seq->private; 569 struct mr6_table *mrt = it->mrt; 570 571 seq_printf(seq, "%pI6 %pI6 %-3hd", 572 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, 573 mfc->mf6c_parent); 574 575 if (it->cache != &mrt->mfc6_unres_queue) { 576 seq_printf(seq, " %8lu %8lu %8lu", 577 mfc->mfc_un.res.pkt, 578 mfc->mfc_un.res.bytes, 579 mfc->mfc_un.res.wrong_if); 580 for (n = mfc->mfc_un.res.minvif; 581 n < mfc->mfc_un.res.maxvif; n++) { 582 if (MIF_EXISTS(mrt, n) && 583 mfc->mfc_un.res.ttls[n] < 255) 584 seq_printf(seq, 585 " %2d:%-3d", 586 n, mfc->mfc_un.res.ttls[n]); 587 } 588 } else { 589 /* unresolved mfc_caches don't contain 590 * pkt, bytes and wrong_if values 591 */ 592 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 593 } 594 seq_putc(seq, '\n'); 595 } 596 return 0; 597 } 598 599 static const struct seq_operations ipmr_mfc_seq_ops = { 600 .start = ipmr_mfc_seq_start, 601 .next = ipmr_mfc_seq_next, 602 .stop = ipmr_mfc_seq_stop, 603 .show = ipmr_mfc_seq_show, 604 }; 605 606 static int ipmr_mfc_open(struct inode *inode, struct file *file) 607 { 608 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 609 sizeof(struct ipmr_mfc_iter)); 610 } 611 612 static const struct file_operations ip6mr_mfc_fops = { 613 .owner = THIS_MODULE, 614 .open = ipmr_mfc_open, 615 .read = seq_read, 616 .llseek = seq_lseek, 617 .release = seq_release_net, 618 }; 619 #endif 620 621 #ifdef CONFIG_IPV6_PIMSM_V2 622 623 static int pim6_rcv(struct sk_buff *skb) 624 { 625 struct pimreghdr *pim; 626 struct ipv6hdr *encap; 627 struct net_device *reg_dev = NULL; 628 struct net *net = dev_net(skb->dev); 629 struct mr6_table *mrt; 630 struct flowi6 fl6 = { 631 .flowi6_iif = skb->dev->ifindex, 632 .flowi6_mark = skb->mark, 633 }; 634 int reg_vif_num; 635 636 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) 637 goto drop; 638 639 pim = (struct pimreghdr *)skb_transport_header(skb); 640 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) || 641 (pim->flags & PIM_NULL_REGISTER) || 642 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 643 sizeof(*pim), IPPROTO_PIM, 644 csum_partial((void *)pim, sizeof(*pim), 0)) && 645 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 646 goto drop; 647 648 /* check if the inner packet is destined to mcast group */ 649 encap = (struct ipv6hdr *)(skb_transport_header(skb) + 650 sizeof(*pim)); 651 652 if (!ipv6_addr_is_multicast(&encap->daddr) || 653 encap->payload_len == 0 || 654 ntohs(encap->payload_len) + sizeof(*pim) > skb->len) 655 goto drop; 656 657 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 658 goto drop; 659 reg_vif_num = mrt->mroute_reg_vif_num; 660 661 read_lock(&mrt_lock); 662 if (reg_vif_num >= 0) 663 reg_dev = mrt->vif6_table[reg_vif_num].dev; 664 if (reg_dev) 665 dev_hold(reg_dev); 666 read_unlock(&mrt_lock); 667 668 if (!reg_dev) 669 goto drop; 670 671 skb->mac_header = skb->network_header; 672 skb_pull(skb, (u8 *)encap - skb->data); 673 skb_reset_network_header(skb); 674 skb->protocol = htons(ETH_P_IPV6); 675 skb->ip_summed = CHECKSUM_NONE; 676 677 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 678 679 netif_rx(skb); 680 681 dev_put(reg_dev); 682 return 0; 683 drop: 684 kfree_skb(skb); 685 return 0; 686 } 687 688 static const struct inet6_protocol pim6_protocol = { 689 .handler = pim6_rcv, 690 }; 691 692 /* Service routines creating virtual interfaces: PIMREG */ 693 694 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, 695 struct net_device *dev) 696 { 697 struct net *net = dev_net(dev); 698 struct mr6_table *mrt; 699 struct flowi6 fl6 = { 700 .flowi6_oif = dev->ifindex, 701 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 702 .flowi6_mark = skb->mark, 703 }; 704 int err; 705 706 err = ip6mr_fib_lookup(net, &fl6, &mrt); 707 if (err < 0) { 708 kfree_skb(skb); 709 return err; 710 } 711 712 read_lock(&mrt_lock); 713 dev->stats.tx_bytes += skb->len; 714 dev->stats.tx_packets++; 715 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT); 716 read_unlock(&mrt_lock); 717 kfree_skb(skb); 718 return NETDEV_TX_OK; 719 } 720 721 static int reg_vif_get_iflink(const struct net_device *dev) 722 { 723 return 0; 724 } 725 726 static const struct net_device_ops reg_vif_netdev_ops = { 727 .ndo_start_xmit = reg_vif_xmit, 728 .ndo_get_iflink = reg_vif_get_iflink, 729 }; 730 731 static void reg_vif_setup(struct net_device *dev) 732 { 733 dev->type = ARPHRD_PIMREG; 734 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; 735 dev->flags = IFF_NOARP; 736 dev->netdev_ops = ®_vif_netdev_ops; 737 dev->needs_free_netdev = true; 738 dev->features |= NETIF_F_NETNS_LOCAL; 739 } 740 741 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt) 742 { 743 struct net_device *dev; 744 char name[IFNAMSIZ]; 745 746 if (mrt->id == RT6_TABLE_DFLT) 747 sprintf(name, "pim6reg"); 748 else 749 sprintf(name, "pim6reg%u", mrt->id); 750 751 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 752 if (!dev) 753 return NULL; 754 755 dev_net_set(dev, net); 756 757 if (register_netdevice(dev)) { 758 free_netdev(dev); 759 return NULL; 760 } 761 762 if (dev_open(dev)) 763 goto failure; 764 765 dev_hold(dev); 766 return dev; 767 768 failure: 769 unregister_netdevice(dev); 770 return NULL; 771 } 772 #endif 773 774 /* 775 * Delete a VIF entry 776 */ 777 778 static int mif6_delete(struct mr6_table *mrt, int vifi, int notify, 779 struct list_head *head) 780 { 781 struct mif_device *v; 782 struct net_device *dev; 783 struct inet6_dev *in6_dev; 784 785 if (vifi < 0 || vifi >= mrt->maxvif) 786 return -EADDRNOTAVAIL; 787 788 v = &mrt->vif6_table[vifi]; 789 790 write_lock_bh(&mrt_lock); 791 dev = v->dev; 792 v->dev = NULL; 793 794 if (!dev) { 795 write_unlock_bh(&mrt_lock); 796 return -EADDRNOTAVAIL; 797 } 798 799 #ifdef CONFIG_IPV6_PIMSM_V2 800 if (vifi == mrt->mroute_reg_vif_num) 801 mrt->mroute_reg_vif_num = -1; 802 #endif 803 804 if (vifi + 1 == mrt->maxvif) { 805 int tmp; 806 for (tmp = vifi - 1; tmp >= 0; tmp--) { 807 if (MIF_EXISTS(mrt, tmp)) 808 break; 809 } 810 mrt->maxvif = tmp + 1; 811 } 812 813 write_unlock_bh(&mrt_lock); 814 815 dev_set_allmulti(dev, -1); 816 817 in6_dev = __in6_dev_get(dev); 818 if (in6_dev) { 819 in6_dev->cnf.mc_forwarding--; 820 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 821 NETCONFA_MC_FORWARDING, 822 dev->ifindex, &in6_dev->cnf); 823 } 824 825 if ((v->flags & MIFF_REGISTER) && !notify) 826 unregister_netdevice_queue(dev, head); 827 828 dev_put(dev); 829 return 0; 830 } 831 832 static inline void ip6mr_cache_free(struct mfc6_cache *c) 833 { 834 kmem_cache_free(mrt_cachep, c); 835 } 836 837 /* Destroy an unresolved cache entry, killing queued skbs 838 and reporting error to netlink readers. 839 */ 840 841 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c) 842 { 843 struct net *net = read_pnet(&mrt->net); 844 struct sk_buff *skb; 845 846 atomic_dec(&mrt->cache_resolve_queue_len); 847 848 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { 849 if (ipv6_hdr(skb)->version == 0) { 850 struct nlmsghdr *nlh = skb_pull(skb, 851 sizeof(struct ipv6hdr)); 852 nlh->nlmsg_type = NLMSG_ERROR; 853 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 854 skb_trim(skb, nlh->nlmsg_len); 855 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT; 856 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 857 } else 858 kfree_skb(skb); 859 } 860 861 ip6mr_cache_free(c); 862 } 863 864 865 /* Timer process for all the unresolved queue. */ 866 867 static void ipmr_do_expire_process(struct mr6_table *mrt) 868 { 869 unsigned long now = jiffies; 870 unsigned long expires = 10 * HZ; 871 struct mfc6_cache *c, *next; 872 873 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 874 if (time_after(c->mfc_un.unres.expires, now)) { 875 /* not yet... */ 876 unsigned long interval = c->mfc_un.unres.expires - now; 877 if (interval < expires) 878 expires = interval; 879 continue; 880 } 881 882 list_del(&c->list); 883 mr6_netlink_event(mrt, c, RTM_DELROUTE); 884 ip6mr_destroy_unres(mrt, c); 885 } 886 887 if (!list_empty(&mrt->mfc6_unres_queue)) 888 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 889 } 890 891 static void ipmr_expire_process(unsigned long arg) 892 { 893 struct mr6_table *mrt = (struct mr6_table *)arg; 894 895 if (!spin_trylock(&mfc_unres_lock)) { 896 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 897 return; 898 } 899 900 if (!list_empty(&mrt->mfc6_unres_queue)) 901 ipmr_do_expire_process(mrt); 902 903 spin_unlock(&mfc_unres_lock); 904 } 905 906 /* Fill oifs list. It is called under write locked mrt_lock. */ 907 908 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache, 909 unsigned char *ttls) 910 { 911 int vifi; 912 913 cache->mfc_un.res.minvif = MAXMIFS; 914 cache->mfc_un.res.maxvif = 0; 915 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 916 917 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 918 if (MIF_EXISTS(mrt, vifi) && 919 ttls[vifi] && ttls[vifi] < 255) { 920 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 921 if (cache->mfc_un.res.minvif > vifi) 922 cache->mfc_un.res.minvif = vifi; 923 if (cache->mfc_un.res.maxvif <= vifi) 924 cache->mfc_un.res.maxvif = vifi + 1; 925 } 926 } 927 cache->mfc_un.res.lastuse = jiffies; 928 } 929 930 static int mif6_add(struct net *net, struct mr6_table *mrt, 931 struct mif6ctl *vifc, int mrtsock) 932 { 933 int vifi = vifc->mif6c_mifi; 934 struct mif_device *v = &mrt->vif6_table[vifi]; 935 struct net_device *dev; 936 struct inet6_dev *in6_dev; 937 int err; 938 939 /* Is vif busy ? */ 940 if (MIF_EXISTS(mrt, vifi)) 941 return -EADDRINUSE; 942 943 switch (vifc->mif6c_flags) { 944 #ifdef CONFIG_IPV6_PIMSM_V2 945 case MIFF_REGISTER: 946 /* 947 * Special Purpose VIF in PIM 948 * All the packets will be sent to the daemon 949 */ 950 if (mrt->mroute_reg_vif_num >= 0) 951 return -EADDRINUSE; 952 dev = ip6mr_reg_vif(net, mrt); 953 if (!dev) 954 return -ENOBUFS; 955 err = dev_set_allmulti(dev, 1); 956 if (err) { 957 unregister_netdevice(dev); 958 dev_put(dev); 959 return err; 960 } 961 break; 962 #endif 963 case 0: 964 dev = dev_get_by_index(net, vifc->mif6c_pifi); 965 if (!dev) 966 return -EADDRNOTAVAIL; 967 err = dev_set_allmulti(dev, 1); 968 if (err) { 969 dev_put(dev); 970 return err; 971 } 972 break; 973 default: 974 return -EINVAL; 975 } 976 977 in6_dev = __in6_dev_get(dev); 978 if (in6_dev) { 979 in6_dev->cnf.mc_forwarding++; 980 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 981 NETCONFA_MC_FORWARDING, 982 dev->ifindex, &in6_dev->cnf); 983 } 984 985 /* 986 * Fill in the VIF structures 987 */ 988 v->rate_limit = vifc->vifc_rate_limit; 989 v->flags = vifc->mif6c_flags; 990 if (!mrtsock) 991 v->flags |= VIFF_STATIC; 992 v->threshold = vifc->vifc_threshold; 993 v->bytes_in = 0; 994 v->bytes_out = 0; 995 v->pkt_in = 0; 996 v->pkt_out = 0; 997 v->link = dev->ifindex; 998 if (v->flags & MIFF_REGISTER) 999 v->link = dev_get_iflink(dev); 1000 1001 /* And finish update writing critical data */ 1002 write_lock_bh(&mrt_lock); 1003 v->dev = dev; 1004 #ifdef CONFIG_IPV6_PIMSM_V2 1005 if (v->flags & MIFF_REGISTER) 1006 mrt->mroute_reg_vif_num = vifi; 1007 #endif 1008 if (vifi + 1 > mrt->maxvif) 1009 mrt->maxvif = vifi + 1; 1010 write_unlock_bh(&mrt_lock); 1011 return 0; 1012 } 1013 1014 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt, 1015 const struct in6_addr *origin, 1016 const struct in6_addr *mcastgrp) 1017 { 1018 int line = MFC6_HASH(mcastgrp, origin); 1019 struct mfc6_cache *c; 1020 1021 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1022 if (ipv6_addr_equal(&c->mf6c_origin, origin) && 1023 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) 1024 return c; 1025 } 1026 return NULL; 1027 } 1028 1029 /* Look for a (*,*,oif) entry */ 1030 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt, 1031 mifi_t mifi) 1032 { 1033 int line = MFC6_HASH(&in6addr_any, &in6addr_any); 1034 struct mfc6_cache *c; 1035 1036 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1037 if (ipv6_addr_any(&c->mf6c_origin) && 1038 ipv6_addr_any(&c->mf6c_mcastgrp) && 1039 (c->mfc_un.res.ttls[mifi] < 255)) 1040 return c; 1041 1042 return NULL; 1043 } 1044 1045 /* Look for a (*,G) entry */ 1046 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt, 1047 struct in6_addr *mcastgrp, 1048 mifi_t mifi) 1049 { 1050 int line = MFC6_HASH(mcastgrp, &in6addr_any); 1051 struct mfc6_cache *c, *proxy; 1052 1053 if (ipv6_addr_any(mcastgrp)) 1054 goto skip; 1055 1056 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1057 if (ipv6_addr_any(&c->mf6c_origin) && 1058 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) { 1059 if (c->mfc_un.res.ttls[mifi] < 255) 1060 return c; 1061 1062 /* It's ok if the mifi is part of the static tree */ 1063 proxy = ip6mr_cache_find_any_parent(mrt, 1064 c->mf6c_parent); 1065 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255) 1066 return c; 1067 } 1068 1069 skip: 1070 return ip6mr_cache_find_any_parent(mrt, mifi); 1071 } 1072 1073 /* 1074 * Allocate a multicast cache entry 1075 */ 1076 static struct mfc6_cache *ip6mr_cache_alloc(void) 1077 { 1078 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 1079 if (!c) 1080 return NULL; 1081 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 1082 c->mfc_un.res.minvif = MAXMIFS; 1083 return c; 1084 } 1085 1086 static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 1087 { 1088 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 1089 if (!c) 1090 return NULL; 1091 skb_queue_head_init(&c->mfc_un.unres.unresolved); 1092 c->mfc_un.unres.expires = jiffies + 10 * HZ; 1093 return c; 1094 } 1095 1096 /* 1097 * A cache entry has gone into a resolved state from queued 1098 */ 1099 1100 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt, 1101 struct mfc6_cache *uc, struct mfc6_cache *c) 1102 { 1103 struct sk_buff *skb; 1104 1105 /* 1106 * Play the pending entries through our router 1107 */ 1108 1109 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 1110 if (ipv6_hdr(skb)->version == 0) { 1111 struct nlmsghdr *nlh = skb_pull(skb, 1112 sizeof(struct ipv6hdr)); 1113 1114 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 1115 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1116 } else { 1117 nlh->nlmsg_type = NLMSG_ERROR; 1118 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1119 skb_trim(skb, nlh->nlmsg_len); 1120 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE; 1121 } 1122 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1123 } else 1124 ip6_mr_forward(net, mrt, skb, c); 1125 } 1126 } 1127 1128 /* 1129 * Bounce a cache query up to pim6sd and netlink. 1130 * 1131 * Called under mrt_lock. 1132 */ 1133 1134 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 1135 mifi_t mifi, int assert) 1136 { 1137 struct sk_buff *skb; 1138 struct mrt6msg *msg; 1139 int ret; 1140 1141 #ifdef CONFIG_IPV6_PIMSM_V2 1142 if (assert == MRT6MSG_WHOLEPKT) 1143 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1144 +sizeof(*msg)); 1145 else 1146 #endif 1147 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1148 1149 if (!skb) 1150 return -ENOBUFS; 1151 1152 /* I suppose that internal messages 1153 * do not require checksums */ 1154 1155 skb->ip_summed = CHECKSUM_UNNECESSARY; 1156 1157 #ifdef CONFIG_IPV6_PIMSM_V2 1158 if (assert == MRT6MSG_WHOLEPKT) { 1159 /* Ugly, but we have no choice with this interface. 1160 Duplicate old header, fix length etc. 1161 And all this only to mangle msg->im6_msgtype and 1162 to set msg->im6_mbz to "mbz" :-) 1163 */ 1164 skb_push(skb, -skb_network_offset(pkt)); 1165 1166 skb_push(skb, sizeof(*msg)); 1167 skb_reset_transport_header(skb); 1168 msg = (struct mrt6msg *)skb_transport_header(skb); 1169 msg->im6_mbz = 0; 1170 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1171 msg->im6_mif = mrt->mroute_reg_vif_num; 1172 msg->im6_pad = 0; 1173 msg->im6_src = ipv6_hdr(pkt)->saddr; 1174 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1175 1176 skb->ip_summed = CHECKSUM_UNNECESSARY; 1177 } else 1178 #endif 1179 { 1180 /* 1181 * Copy the IP header 1182 */ 1183 1184 skb_put(skb, sizeof(struct ipv6hdr)); 1185 skb_reset_network_header(skb); 1186 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1187 1188 /* 1189 * Add our header 1190 */ 1191 skb_put(skb, sizeof(*msg)); 1192 skb_reset_transport_header(skb); 1193 msg = (struct mrt6msg *)skb_transport_header(skb); 1194 1195 msg->im6_mbz = 0; 1196 msg->im6_msgtype = assert; 1197 msg->im6_mif = mifi; 1198 msg->im6_pad = 0; 1199 msg->im6_src = ipv6_hdr(pkt)->saddr; 1200 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1201 1202 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1203 skb->ip_summed = CHECKSUM_UNNECESSARY; 1204 } 1205 1206 if (!mrt->mroute6_sk) { 1207 kfree_skb(skb); 1208 return -EINVAL; 1209 } 1210 1211 mrt6msg_netlink_event(mrt, skb); 1212 1213 /* 1214 * Deliver to user space multicast routing algorithms 1215 */ 1216 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb); 1217 if (ret < 0) { 1218 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1219 kfree_skb(skb); 1220 } 1221 1222 return ret; 1223 } 1224 1225 /* 1226 * Queue a packet for resolution. It gets locked cache entry! 1227 */ 1228 1229 static int 1230 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb) 1231 { 1232 bool found = false; 1233 int err; 1234 struct mfc6_cache *c; 1235 1236 spin_lock_bh(&mfc_unres_lock); 1237 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) { 1238 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1239 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1240 found = true; 1241 break; 1242 } 1243 } 1244 1245 if (!found) { 1246 /* 1247 * Create a new entry if allowable 1248 */ 1249 1250 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1251 (c = ip6mr_cache_alloc_unres()) == NULL) { 1252 spin_unlock_bh(&mfc_unres_lock); 1253 1254 kfree_skb(skb); 1255 return -ENOBUFS; 1256 } 1257 1258 /* 1259 * Fill in the new cache entry 1260 */ 1261 c->mf6c_parent = -1; 1262 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1263 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1264 1265 /* 1266 * Reflect first query at pim6sd 1267 */ 1268 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1269 if (err < 0) { 1270 /* If the report failed throw the cache entry 1271 out - Brad Parker 1272 */ 1273 spin_unlock_bh(&mfc_unres_lock); 1274 1275 ip6mr_cache_free(c); 1276 kfree_skb(skb); 1277 return err; 1278 } 1279 1280 atomic_inc(&mrt->cache_resolve_queue_len); 1281 list_add(&c->list, &mrt->mfc6_unres_queue); 1282 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1283 1284 ipmr_do_expire_process(mrt); 1285 } 1286 1287 /* 1288 * See if we can append the packet 1289 */ 1290 if (c->mfc_un.unres.unresolved.qlen > 3) { 1291 kfree_skb(skb); 1292 err = -ENOBUFS; 1293 } else { 1294 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1295 err = 0; 1296 } 1297 1298 spin_unlock_bh(&mfc_unres_lock); 1299 return err; 1300 } 1301 1302 /* 1303 * MFC6 cache manipulation by user space 1304 */ 1305 1306 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc, 1307 int parent) 1308 { 1309 int line; 1310 struct mfc6_cache *c, *next; 1311 1312 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1313 1314 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) { 1315 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1316 ipv6_addr_equal(&c->mf6c_mcastgrp, 1317 &mfc->mf6cc_mcastgrp.sin6_addr) && 1318 (parent == -1 || parent == c->mf6c_parent)) { 1319 write_lock_bh(&mrt_lock); 1320 list_del(&c->list); 1321 write_unlock_bh(&mrt_lock); 1322 1323 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1324 ip6mr_cache_free(c); 1325 return 0; 1326 } 1327 } 1328 return -ENOENT; 1329 } 1330 1331 static int ip6mr_device_event(struct notifier_block *this, 1332 unsigned long event, void *ptr) 1333 { 1334 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1335 struct net *net = dev_net(dev); 1336 struct mr6_table *mrt; 1337 struct mif_device *v; 1338 int ct; 1339 1340 if (event != NETDEV_UNREGISTER) 1341 return NOTIFY_DONE; 1342 1343 ip6mr_for_each_table(mrt, net) { 1344 v = &mrt->vif6_table[0]; 1345 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1346 if (v->dev == dev) 1347 mif6_delete(mrt, ct, 1, NULL); 1348 } 1349 } 1350 1351 return NOTIFY_DONE; 1352 } 1353 1354 static struct notifier_block ip6_mr_notifier = { 1355 .notifier_call = ip6mr_device_event 1356 }; 1357 1358 /* 1359 * Setup for IP multicast routing 1360 */ 1361 1362 static int __net_init ip6mr_net_init(struct net *net) 1363 { 1364 int err; 1365 1366 err = ip6mr_rules_init(net); 1367 if (err < 0) 1368 goto fail; 1369 1370 #ifdef CONFIG_PROC_FS 1371 err = -ENOMEM; 1372 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops)) 1373 goto proc_vif_fail; 1374 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops)) 1375 goto proc_cache_fail; 1376 #endif 1377 1378 return 0; 1379 1380 #ifdef CONFIG_PROC_FS 1381 proc_cache_fail: 1382 remove_proc_entry("ip6_mr_vif", net->proc_net); 1383 proc_vif_fail: 1384 ip6mr_rules_exit(net); 1385 #endif 1386 fail: 1387 return err; 1388 } 1389 1390 static void __net_exit ip6mr_net_exit(struct net *net) 1391 { 1392 #ifdef CONFIG_PROC_FS 1393 remove_proc_entry("ip6_mr_cache", net->proc_net); 1394 remove_proc_entry("ip6_mr_vif", net->proc_net); 1395 #endif 1396 ip6mr_rules_exit(net); 1397 } 1398 1399 static struct pernet_operations ip6mr_net_ops = { 1400 .init = ip6mr_net_init, 1401 .exit = ip6mr_net_exit, 1402 }; 1403 1404 int __init ip6_mr_init(void) 1405 { 1406 int err; 1407 1408 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1409 sizeof(struct mfc6_cache), 1410 0, SLAB_HWCACHE_ALIGN, 1411 NULL); 1412 if (!mrt_cachep) 1413 return -ENOMEM; 1414 1415 err = register_pernet_subsys(&ip6mr_net_ops); 1416 if (err) 1417 goto reg_pernet_fail; 1418 1419 err = register_netdevice_notifier(&ip6_mr_notifier); 1420 if (err) 1421 goto reg_notif_fail; 1422 #ifdef CONFIG_IPV6_PIMSM_V2 1423 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1424 pr_err("%s: can't add PIM protocol\n", __func__); 1425 err = -EAGAIN; 1426 goto add_proto_fail; 1427 } 1428 #endif 1429 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, 1430 ip6mr_rtm_dumproute, 0); 1431 return 0; 1432 #ifdef CONFIG_IPV6_PIMSM_V2 1433 add_proto_fail: 1434 unregister_netdevice_notifier(&ip6_mr_notifier); 1435 #endif 1436 reg_notif_fail: 1437 unregister_pernet_subsys(&ip6mr_net_ops); 1438 reg_pernet_fail: 1439 kmem_cache_destroy(mrt_cachep); 1440 return err; 1441 } 1442 1443 void ip6_mr_cleanup(void) 1444 { 1445 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1446 #ifdef CONFIG_IPV6_PIMSM_V2 1447 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1448 #endif 1449 unregister_netdevice_notifier(&ip6_mr_notifier); 1450 unregister_pernet_subsys(&ip6mr_net_ops); 1451 kmem_cache_destroy(mrt_cachep); 1452 } 1453 1454 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1455 struct mf6cctl *mfc, int mrtsock, int parent) 1456 { 1457 bool found = false; 1458 int line; 1459 struct mfc6_cache *uc, *c; 1460 unsigned char ttls[MAXMIFS]; 1461 int i; 1462 1463 if (mfc->mf6cc_parent >= MAXMIFS) 1464 return -ENFILE; 1465 1466 memset(ttls, 255, MAXMIFS); 1467 for (i = 0; i < MAXMIFS; i++) { 1468 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1469 ttls[i] = 1; 1470 1471 } 1472 1473 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1474 1475 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1476 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1477 ipv6_addr_equal(&c->mf6c_mcastgrp, 1478 &mfc->mf6cc_mcastgrp.sin6_addr) && 1479 (parent == -1 || parent == mfc->mf6cc_parent)) { 1480 found = true; 1481 break; 1482 } 1483 } 1484 1485 if (found) { 1486 write_lock_bh(&mrt_lock); 1487 c->mf6c_parent = mfc->mf6cc_parent; 1488 ip6mr_update_thresholds(mrt, c, ttls); 1489 if (!mrtsock) 1490 c->mfc_flags |= MFC_STATIC; 1491 write_unlock_bh(&mrt_lock); 1492 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1493 return 0; 1494 } 1495 1496 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1497 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1498 return -EINVAL; 1499 1500 c = ip6mr_cache_alloc(); 1501 if (!c) 1502 return -ENOMEM; 1503 1504 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1505 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1506 c->mf6c_parent = mfc->mf6cc_parent; 1507 ip6mr_update_thresholds(mrt, c, ttls); 1508 if (!mrtsock) 1509 c->mfc_flags |= MFC_STATIC; 1510 1511 write_lock_bh(&mrt_lock); 1512 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1513 write_unlock_bh(&mrt_lock); 1514 1515 /* 1516 * Check to see if we resolved a queued list. If so we 1517 * need to send on the frames and tidy up. 1518 */ 1519 found = false; 1520 spin_lock_bh(&mfc_unres_lock); 1521 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1522 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1523 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1524 list_del(&uc->list); 1525 atomic_dec(&mrt->cache_resolve_queue_len); 1526 found = true; 1527 break; 1528 } 1529 } 1530 if (list_empty(&mrt->mfc6_unres_queue)) 1531 del_timer(&mrt->ipmr_expire_timer); 1532 spin_unlock_bh(&mfc_unres_lock); 1533 1534 if (found) { 1535 ip6mr_cache_resolve(net, mrt, uc, c); 1536 ip6mr_cache_free(uc); 1537 } 1538 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1539 return 0; 1540 } 1541 1542 /* 1543 * Close the multicast socket, and clear the vif tables etc 1544 */ 1545 1546 static void mroute_clean_tables(struct mr6_table *mrt, bool all) 1547 { 1548 int i; 1549 LIST_HEAD(list); 1550 struct mfc6_cache *c, *next; 1551 1552 /* 1553 * Shut down all active vif entries 1554 */ 1555 for (i = 0; i < mrt->maxvif; i++) { 1556 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) 1557 continue; 1558 mif6_delete(mrt, i, 0, &list); 1559 } 1560 unregister_netdevice_many(&list); 1561 1562 /* 1563 * Wipe the cache 1564 */ 1565 for (i = 0; i < MFC6_LINES; i++) { 1566 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1567 if (!all && (c->mfc_flags & MFC_STATIC)) 1568 continue; 1569 write_lock_bh(&mrt_lock); 1570 list_del(&c->list); 1571 write_unlock_bh(&mrt_lock); 1572 1573 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1574 ip6mr_cache_free(c); 1575 } 1576 } 1577 1578 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1579 spin_lock_bh(&mfc_unres_lock); 1580 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1581 list_del(&c->list); 1582 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1583 ip6mr_destroy_unres(mrt, c); 1584 } 1585 spin_unlock_bh(&mfc_unres_lock); 1586 } 1587 } 1588 1589 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1590 { 1591 int err = 0; 1592 struct net *net = sock_net(sk); 1593 1594 rtnl_lock(); 1595 write_lock_bh(&mrt_lock); 1596 if (likely(mrt->mroute6_sk == NULL)) { 1597 mrt->mroute6_sk = sk; 1598 net->ipv6.devconf_all->mc_forwarding++; 1599 } else { 1600 err = -EADDRINUSE; 1601 } 1602 write_unlock_bh(&mrt_lock); 1603 1604 if (!err) 1605 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1606 NETCONFA_MC_FORWARDING, 1607 NETCONFA_IFINDEX_ALL, 1608 net->ipv6.devconf_all); 1609 rtnl_unlock(); 1610 1611 return err; 1612 } 1613 1614 int ip6mr_sk_done(struct sock *sk) 1615 { 1616 int err = -EACCES; 1617 struct net *net = sock_net(sk); 1618 struct mr6_table *mrt; 1619 1620 rtnl_lock(); 1621 ip6mr_for_each_table(mrt, net) { 1622 if (sk == mrt->mroute6_sk) { 1623 write_lock_bh(&mrt_lock); 1624 mrt->mroute6_sk = NULL; 1625 net->ipv6.devconf_all->mc_forwarding--; 1626 write_unlock_bh(&mrt_lock); 1627 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1628 NETCONFA_MC_FORWARDING, 1629 NETCONFA_IFINDEX_ALL, 1630 net->ipv6.devconf_all); 1631 1632 mroute_clean_tables(mrt, false); 1633 err = 0; 1634 break; 1635 } 1636 } 1637 rtnl_unlock(); 1638 1639 return err; 1640 } 1641 1642 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1643 { 1644 struct mr6_table *mrt; 1645 struct flowi6 fl6 = { 1646 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1647 .flowi6_oif = skb->dev->ifindex, 1648 .flowi6_mark = skb->mark, 1649 }; 1650 1651 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1652 return NULL; 1653 1654 return mrt->mroute6_sk; 1655 } 1656 1657 /* 1658 * Socket options and virtual interface manipulation. The whole 1659 * virtual interface system is a complete heap, but unfortunately 1660 * that's how BSD mrouted happens to think. Maybe one day with a proper 1661 * MOSPF/PIM router set up we can clean this up. 1662 */ 1663 1664 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1665 { 1666 int ret, parent = 0; 1667 struct mif6ctl vif; 1668 struct mf6cctl mfc; 1669 mifi_t mifi; 1670 struct net *net = sock_net(sk); 1671 struct mr6_table *mrt; 1672 1673 if (sk->sk_type != SOCK_RAW || 1674 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1675 return -EOPNOTSUPP; 1676 1677 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1678 if (!mrt) 1679 return -ENOENT; 1680 1681 if (optname != MRT6_INIT) { 1682 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1683 return -EACCES; 1684 } 1685 1686 switch (optname) { 1687 case MRT6_INIT: 1688 if (optlen < sizeof(int)) 1689 return -EINVAL; 1690 1691 return ip6mr_sk_init(mrt, sk); 1692 1693 case MRT6_DONE: 1694 return ip6mr_sk_done(sk); 1695 1696 case MRT6_ADD_MIF: 1697 if (optlen < sizeof(vif)) 1698 return -EINVAL; 1699 if (copy_from_user(&vif, optval, sizeof(vif))) 1700 return -EFAULT; 1701 if (vif.mif6c_mifi >= MAXMIFS) 1702 return -ENFILE; 1703 rtnl_lock(); 1704 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1705 rtnl_unlock(); 1706 return ret; 1707 1708 case MRT6_DEL_MIF: 1709 if (optlen < sizeof(mifi_t)) 1710 return -EINVAL; 1711 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1712 return -EFAULT; 1713 rtnl_lock(); 1714 ret = mif6_delete(mrt, mifi, 0, NULL); 1715 rtnl_unlock(); 1716 return ret; 1717 1718 /* 1719 * Manipulate the forwarding caches. These live 1720 * in a sort of kernel/user symbiosis. 1721 */ 1722 case MRT6_ADD_MFC: 1723 case MRT6_DEL_MFC: 1724 parent = -1; 1725 case MRT6_ADD_MFC_PROXY: 1726 case MRT6_DEL_MFC_PROXY: 1727 if (optlen < sizeof(mfc)) 1728 return -EINVAL; 1729 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1730 return -EFAULT; 1731 if (parent == 0) 1732 parent = mfc.mf6cc_parent; 1733 rtnl_lock(); 1734 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1735 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1736 else 1737 ret = ip6mr_mfc_add(net, mrt, &mfc, 1738 sk == mrt->mroute6_sk, parent); 1739 rtnl_unlock(); 1740 return ret; 1741 1742 /* 1743 * Control PIM assert (to activate pim will activate assert) 1744 */ 1745 case MRT6_ASSERT: 1746 { 1747 int v; 1748 1749 if (optlen != sizeof(v)) 1750 return -EINVAL; 1751 if (get_user(v, (int __user *)optval)) 1752 return -EFAULT; 1753 mrt->mroute_do_assert = v; 1754 return 0; 1755 } 1756 1757 #ifdef CONFIG_IPV6_PIMSM_V2 1758 case MRT6_PIM: 1759 { 1760 int v; 1761 1762 if (optlen != sizeof(v)) 1763 return -EINVAL; 1764 if (get_user(v, (int __user *)optval)) 1765 return -EFAULT; 1766 v = !!v; 1767 rtnl_lock(); 1768 ret = 0; 1769 if (v != mrt->mroute_do_pim) { 1770 mrt->mroute_do_pim = v; 1771 mrt->mroute_do_assert = v; 1772 } 1773 rtnl_unlock(); 1774 return ret; 1775 } 1776 1777 #endif 1778 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1779 case MRT6_TABLE: 1780 { 1781 u32 v; 1782 1783 if (optlen != sizeof(u32)) 1784 return -EINVAL; 1785 if (get_user(v, (u32 __user *)optval)) 1786 return -EFAULT; 1787 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1788 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1789 return -EINVAL; 1790 if (sk == mrt->mroute6_sk) 1791 return -EBUSY; 1792 1793 rtnl_lock(); 1794 ret = 0; 1795 if (!ip6mr_new_table(net, v)) 1796 ret = -ENOMEM; 1797 raw6_sk(sk)->ip6mr_table = v; 1798 rtnl_unlock(); 1799 return ret; 1800 } 1801 #endif 1802 /* 1803 * Spurious command, or MRT6_VERSION which you cannot 1804 * set. 1805 */ 1806 default: 1807 return -ENOPROTOOPT; 1808 } 1809 } 1810 1811 /* 1812 * Getsock opt support for the multicast routing system. 1813 */ 1814 1815 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1816 int __user *optlen) 1817 { 1818 int olr; 1819 int val; 1820 struct net *net = sock_net(sk); 1821 struct mr6_table *mrt; 1822 1823 if (sk->sk_type != SOCK_RAW || 1824 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1825 return -EOPNOTSUPP; 1826 1827 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1828 if (!mrt) 1829 return -ENOENT; 1830 1831 switch (optname) { 1832 case MRT6_VERSION: 1833 val = 0x0305; 1834 break; 1835 #ifdef CONFIG_IPV6_PIMSM_V2 1836 case MRT6_PIM: 1837 val = mrt->mroute_do_pim; 1838 break; 1839 #endif 1840 case MRT6_ASSERT: 1841 val = mrt->mroute_do_assert; 1842 break; 1843 default: 1844 return -ENOPROTOOPT; 1845 } 1846 1847 if (get_user(olr, optlen)) 1848 return -EFAULT; 1849 1850 olr = min_t(int, olr, sizeof(int)); 1851 if (olr < 0) 1852 return -EINVAL; 1853 1854 if (put_user(olr, optlen)) 1855 return -EFAULT; 1856 if (copy_to_user(optval, &val, olr)) 1857 return -EFAULT; 1858 return 0; 1859 } 1860 1861 /* 1862 * The IP multicast ioctl support routines. 1863 */ 1864 1865 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1866 { 1867 struct sioc_sg_req6 sr; 1868 struct sioc_mif_req6 vr; 1869 struct mif_device *vif; 1870 struct mfc6_cache *c; 1871 struct net *net = sock_net(sk); 1872 struct mr6_table *mrt; 1873 1874 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1875 if (!mrt) 1876 return -ENOENT; 1877 1878 switch (cmd) { 1879 case SIOCGETMIFCNT_IN6: 1880 if (copy_from_user(&vr, arg, sizeof(vr))) 1881 return -EFAULT; 1882 if (vr.mifi >= mrt->maxvif) 1883 return -EINVAL; 1884 read_lock(&mrt_lock); 1885 vif = &mrt->vif6_table[vr.mifi]; 1886 if (MIF_EXISTS(mrt, vr.mifi)) { 1887 vr.icount = vif->pkt_in; 1888 vr.ocount = vif->pkt_out; 1889 vr.ibytes = vif->bytes_in; 1890 vr.obytes = vif->bytes_out; 1891 read_unlock(&mrt_lock); 1892 1893 if (copy_to_user(arg, &vr, sizeof(vr))) 1894 return -EFAULT; 1895 return 0; 1896 } 1897 read_unlock(&mrt_lock); 1898 return -EADDRNOTAVAIL; 1899 case SIOCGETSGCNT_IN6: 1900 if (copy_from_user(&sr, arg, sizeof(sr))) 1901 return -EFAULT; 1902 1903 read_lock(&mrt_lock); 1904 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1905 if (c) { 1906 sr.pktcnt = c->mfc_un.res.pkt; 1907 sr.bytecnt = c->mfc_un.res.bytes; 1908 sr.wrong_if = c->mfc_un.res.wrong_if; 1909 read_unlock(&mrt_lock); 1910 1911 if (copy_to_user(arg, &sr, sizeof(sr))) 1912 return -EFAULT; 1913 return 0; 1914 } 1915 read_unlock(&mrt_lock); 1916 return -EADDRNOTAVAIL; 1917 default: 1918 return -ENOIOCTLCMD; 1919 } 1920 } 1921 1922 #ifdef CONFIG_COMPAT 1923 struct compat_sioc_sg_req6 { 1924 struct sockaddr_in6 src; 1925 struct sockaddr_in6 grp; 1926 compat_ulong_t pktcnt; 1927 compat_ulong_t bytecnt; 1928 compat_ulong_t wrong_if; 1929 }; 1930 1931 struct compat_sioc_mif_req6 { 1932 mifi_t mifi; 1933 compat_ulong_t icount; 1934 compat_ulong_t ocount; 1935 compat_ulong_t ibytes; 1936 compat_ulong_t obytes; 1937 }; 1938 1939 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1940 { 1941 struct compat_sioc_sg_req6 sr; 1942 struct compat_sioc_mif_req6 vr; 1943 struct mif_device *vif; 1944 struct mfc6_cache *c; 1945 struct net *net = sock_net(sk); 1946 struct mr6_table *mrt; 1947 1948 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1949 if (!mrt) 1950 return -ENOENT; 1951 1952 switch (cmd) { 1953 case SIOCGETMIFCNT_IN6: 1954 if (copy_from_user(&vr, arg, sizeof(vr))) 1955 return -EFAULT; 1956 if (vr.mifi >= mrt->maxvif) 1957 return -EINVAL; 1958 read_lock(&mrt_lock); 1959 vif = &mrt->vif6_table[vr.mifi]; 1960 if (MIF_EXISTS(mrt, vr.mifi)) { 1961 vr.icount = vif->pkt_in; 1962 vr.ocount = vif->pkt_out; 1963 vr.ibytes = vif->bytes_in; 1964 vr.obytes = vif->bytes_out; 1965 read_unlock(&mrt_lock); 1966 1967 if (copy_to_user(arg, &vr, sizeof(vr))) 1968 return -EFAULT; 1969 return 0; 1970 } 1971 read_unlock(&mrt_lock); 1972 return -EADDRNOTAVAIL; 1973 case SIOCGETSGCNT_IN6: 1974 if (copy_from_user(&sr, arg, sizeof(sr))) 1975 return -EFAULT; 1976 1977 read_lock(&mrt_lock); 1978 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1979 if (c) { 1980 sr.pktcnt = c->mfc_un.res.pkt; 1981 sr.bytecnt = c->mfc_un.res.bytes; 1982 sr.wrong_if = c->mfc_un.res.wrong_if; 1983 read_unlock(&mrt_lock); 1984 1985 if (copy_to_user(arg, &sr, sizeof(sr))) 1986 return -EFAULT; 1987 return 0; 1988 } 1989 read_unlock(&mrt_lock); 1990 return -EADDRNOTAVAIL; 1991 default: 1992 return -ENOIOCTLCMD; 1993 } 1994 } 1995 #endif 1996 1997 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 1998 { 1999 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 2000 IPSTATS_MIB_OUTFORWDATAGRAMS); 2001 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)), 2002 IPSTATS_MIB_OUTOCTETS, skb->len); 2003 return dst_output(net, sk, skb); 2004 } 2005 2006 /* 2007 * Processing handlers for ip6mr_forward 2008 */ 2009 2010 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 2011 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 2012 { 2013 struct ipv6hdr *ipv6h; 2014 struct mif_device *vif = &mrt->vif6_table[vifi]; 2015 struct net_device *dev; 2016 struct dst_entry *dst; 2017 struct flowi6 fl6; 2018 2019 if (!vif->dev) 2020 goto out_free; 2021 2022 #ifdef CONFIG_IPV6_PIMSM_V2 2023 if (vif->flags & MIFF_REGISTER) { 2024 vif->pkt_out++; 2025 vif->bytes_out += skb->len; 2026 vif->dev->stats.tx_bytes += skb->len; 2027 vif->dev->stats.tx_packets++; 2028 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2029 goto out_free; 2030 } 2031 #endif 2032 2033 ipv6h = ipv6_hdr(skb); 2034 2035 fl6 = (struct flowi6) { 2036 .flowi6_oif = vif->link, 2037 .daddr = ipv6h->daddr, 2038 }; 2039 2040 dst = ip6_route_output(net, NULL, &fl6); 2041 if (dst->error) { 2042 dst_release(dst); 2043 goto out_free; 2044 } 2045 2046 skb_dst_drop(skb); 2047 skb_dst_set(skb, dst); 2048 2049 /* 2050 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2051 * not only before forwarding, but after forwarding on all output 2052 * interfaces. It is clear, if mrouter runs a multicasting 2053 * program, it should receive packets not depending to what interface 2054 * program is joined. 2055 * If we will not make it, the program will have to join on all 2056 * interfaces. On the other hand, multihoming host (or router, but 2057 * not mrouter) cannot join to more than one interface - it will 2058 * result in receiving multiple packets. 2059 */ 2060 dev = vif->dev; 2061 skb->dev = dev; 2062 vif->pkt_out++; 2063 vif->bytes_out += skb->len; 2064 2065 /* We are about to write */ 2066 /* XXX: extension headers? */ 2067 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2068 goto out_free; 2069 2070 ipv6h = ipv6_hdr(skb); 2071 ipv6h->hop_limit--; 2072 2073 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2074 2075 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2076 net, NULL, skb, skb->dev, dev, 2077 ip6mr_forward2_finish); 2078 2079 out_free: 2080 kfree_skb(skb); 2081 return 0; 2082 } 2083 2084 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2085 { 2086 int ct; 2087 2088 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2089 if (mrt->vif6_table[ct].dev == dev) 2090 break; 2091 } 2092 return ct; 2093 } 2094 2095 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2096 struct sk_buff *skb, struct mfc6_cache *cache) 2097 { 2098 int psend = -1; 2099 int vif, ct; 2100 int true_vifi = ip6mr_find_vif(mrt, skb->dev); 2101 2102 vif = cache->mf6c_parent; 2103 cache->mfc_un.res.pkt++; 2104 cache->mfc_un.res.bytes += skb->len; 2105 cache->mfc_un.res.lastuse = jiffies; 2106 2107 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) { 2108 struct mfc6_cache *cache_proxy; 2109 2110 /* For an (*,G) entry, we only check that the incoming 2111 * interface is part of the static tree. 2112 */ 2113 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif); 2114 if (cache_proxy && 2115 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 2116 goto forward; 2117 } 2118 2119 /* 2120 * Wrong interface: drop packet and (maybe) send PIM assert. 2121 */ 2122 if (mrt->vif6_table[vif].dev != skb->dev) { 2123 cache->mfc_un.res.wrong_if++; 2124 2125 if (true_vifi >= 0 && mrt->mroute_do_assert && 2126 /* pimsm uses asserts, when switching from RPT to SPT, 2127 so that we cannot check that packet arrived on an oif. 2128 It is bad, but otherwise we would need to move pretty 2129 large chunk of pimd to kernel. Ough... --ANK 2130 */ 2131 (mrt->mroute_do_pim || 2132 cache->mfc_un.res.ttls[true_vifi] < 255) && 2133 time_after(jiffies, 2134 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2135 cache->mfc_un.res.last_assert = jiffies; 2136 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2137 } 2138 goto dont_forward; 2139 } 2140 2141 forward: 2142 mrt->vif6_table[vif].pkt_in++; 2143 mrt->vif6_table[vif].bytes_in += skb->len; 2144 2145 /* 2146 * Forward the frame 2147 */ 2148 if (ipv6_addr_any(&cache->mf6c_origin) && 2149 ipv6_addr_any(&cache->mf6c_mcastgrp)) { 2150 if (true_vifi >= 0 && 2151 true_vifi != cache->mf6c_parent && 2152 ipv6_hdr(skb)->hop_limit > 2153 cache->mfc_un.res.ttls[cache->mf6c_parent]) { 2154 /* It's an (*,*) entry and the packet is not coming from 2155 * the upstream: forward the packet to the upstream 2156 * only. 2157 */ 2158 psend = cache->mf6c_parent; 2159 goto last_forward; 2160 } 2161 goto dont_forward; 2162 } 2163 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2164 /* For (*,G) entry, don't forward to the incoming interface */ 2165 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) && 2166 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2167 if (psend != -1) { 2168 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2169 if (skb2) 2170 ip6mr_forward2(net, mrt, skb2, cache, psend); 2171 } 2172 psend = ct; 2173 } 2174 } 2175 last_forward: 2176 if (psend != -1) { 2177 ip6mr_forward2(net, mrt, skb, cache, psend); 2178 return; 2179 } 2180 2181 dont_forward: 2182 kfree_skb(skb); 2183 } 2184 2185 2186 /* 2187 * Multicast packets for forwarding arrive here 2188 */ 2189 2190 int ip6_mr_input(struct sk_buff *skb) 2191 { 2192 struct mfc6_cache *cache; 2193 struct net *net = dev_net(skb->dev); 2194 struct mr6_table *mrt; 2195 struct flowi6 fl6 = { 2196 .flowi6_iif = skb->dev->ifindex, 2197 .flowi6_mark = skb->mark, 2198 }; 2199 int err; 2200 2201 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2202 if (err < 0) { 2203 kfree_skb(skb); 2204 return err; 2205 } 2206 2207 read_lock(&mrt_lock); 2208 cache = ip6mr_cache_find(mrt, 2209 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2210 if (!cache) { 2211 int vif = ip6mr_find_vif(mrt, skb->dev); 2212 2213 if (vif >= 0) 2214 cache = ip6mr_cache_find_any(mrt, 2215 &ipv6_hdr(skb)->daddr, 2216 vif); 2217 } 2218 2219 /* 2220 * No usable cache entry 2221 */ 2222 if (!cache) { 2223 int vif; 2224 2225 vif = ip6mr_find_vif(mrt, skb->dev); 2226 if (vif >= 0) { 2227 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2228 read_unlock(&mrt_lock); 2229 2230 return err; 2231 } 2232 read_unlock(&mrt_lock); 2233 kfree_skb(skb); 2234 return -ENODEV; 2235 } 2236 2237 ip6_mr_forward(net, mrt, skb, cache); 2238 2239 read_unlock(&mrt_lock); 2240 2241 return 0; 2242 } 2243 2244 2245 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2246 struct mfc6_cache *c, struct rtmsg *rtm) 2247 { 2248 struct rta_mfc_stats mfcs; 2249 struct nlattr *mp_attr; 2250 struct rtnexthop *nhp; 2251 unsigned long lastuse; 2252 int ct; 2253 2254 /* If cache is unresolved, don't try to parse IIF and OIF */ 2255 if (c->mf6c_parent >= MAXMIFS) { 2256 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2257 return -ENOENT; 2258 } 2259 2260 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2261 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2262 return -EMSGSIZE; 2263 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2264 if (!mp_attr) 2265 return -EMSGSIZE; 2266 2267 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2268 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2269 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2270 if (!nhp) { 2271 nla_nest_cancel(skb, mp_attr); 2272 return -EMSGSIZE; 2273 } 2274 2275 nhp->rtnh_flags = 0; 2276 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2277 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2278 nhp->rtnh_len = sizeof(*nhp); 2279 } 2280 } 2281 2282 nla_nest_end(skb, mp_attr); 2283 2284 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2285 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2286 2287 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2288 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2289 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2290 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2291 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2292 RTA_PAD)) 2293 return -EMSGSIZE; 2294 2295 rtm->rtm_type = RTN_MULTICAST; 2296 return 1; 2297 } 2298 2299 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2300 u32 portid) 2301 { 2302 int err; 2303 struct mr6_table *mrt; 2304 struct mfc6_cache *cache; 2305 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2306 2307 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2308 if (!mrt) 2309 return -ENOENT; 2310 2311 read_lock(&mrt_lock); 2312 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2313 if (!cache && skb->dev) { 2314 int vif = ip6mr_find_vif(mrt, skb->dev); 2315 2316 if (vif >= 0) 2317 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2318 vif); 2319 } 2320 2321 if (!cache) { 2322 struct sk_buff *skb2; 2323 struct ipv6hdr *iph; 2324 struct net_device *dev; 2325 int vif; 2326 2327 dev = skb->dev; 2328 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2329 read_unlock(&mrt_lock); 2330 return -ENODEV; 2331 } 2332 2333 /* really correct? */ 2334 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2335 if (!skb2) { 2336 read_unlock(&mrt_lock); 2337 return -ENOMEM; 2338 } 2339 2340 NETLINK_CB(skb2).portid = portid; 2341 skb_reset_transport_header(skb2); 2342 2343 skb_put(skb2, sizeof(struct ipv6hdr)); 2344 skb_reset_network_header(skb2); 2345 2346 iph = ipv6_hdr(skb2); 2347 iph->version = 0; 2348 iph->priority = 0; 2349 iph->flow_lbl[0] = 0; 2350 iph->flow_lbl[1] = 0; 2351 iph->flow_lbl[2] = 0; 2352 iph->payload_len = 0; 2353 iph->nexthdr = IPPROTO_NONE; 2354 iph->hop_limit = 0; 2355 iph->saddr = rt->rt6i_src.addr; 2356 iph->daddr = rt->rt6i_dst.addr; 2357 2358 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2359 read_unlock(&mrt_lock); 2360 2361 return err; 2362 } 2363 2364 if (rtm->rtm_flags & RTM_F_NOTIFY) 2365 cache->mfc_flags |= MFC_NOTIFY; 2366 2367 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2368 read_unlock(&mrt_lock); 2369 return err; 2370 } 2371 2372 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2373 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2374 int flags) 2375 { 2376 struct nlmsghdr *nlh; 2377 struct rtmsg *rtm; 2378 int err; 2379 2380 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2381 if (!nlh) 2382 return -EMSGSIZE; 2383 2384 rtm = nlmsg_data(nlh); 2385 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2386 rtm->rtm_dst_len = 128; 2387 rtm->rtm_src_len = 128; 2388 rtm->rtm_tos = 0; 2389 rtm->rtm_table = mrt->id; 2390 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2391 goto nla_put_failure; 2392 rtm->rtm_type = RTN_MULTICAST; 2393 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2394 if (c->mfc_flags & MFC_STATIC) 2395 rtm->rtm_protocol = RTPROT_STATIC; 2396 else 2397 rtm->rtm_protocol = RTPROT_MROUTED; 2398 rtm->rtm_flags = 0; 2399 2400 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2401 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2402 goto nla_put_failure; 2403 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2404 /* do not break the dump if cache is unresolved */ 2405 if (err < 0 && err != -ENOENT) 2406 goto nla_put_failure; 2407 2408 nlmsg_end(skb, nlh); 2409 return 0; 2410 2411 nla_put_failure: 2412 nlmsg_cancel(skb, nlh); 2413 return -EMSGSIZE; 2414 } 2415 2416 static int mr6_msgsize(bool unresolved, int maxvif) 2417 { 2418 size_t len = 2419 NLMSG_ALIGN(sizeof(struct rtmsg)) 2420 + nla_total_size(4) /* RTA_TABLE */ 2421 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2422 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2423 ; 2424 2425 if (!unresolved) 2426 len = len 2427 + nla_total_size(4) /* RTA_IIF */ 2428 + nla_total_size(0) /* RTA_MULTIPATH */ 2429 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2430 /* RTA_MFC_STATS */ 2431 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2432 ; 2433 2434 return len; 2435 } 2436 2437 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2438 int cmd) 2439 { 2440 struct net *net = read_pnet(&mrt->net); 2441 struct sk_buff *skb; 2442 int err = -ENOBUFS; 2443 2444 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2445 GFP_ATOMIC); 2446 if (!skb) 2447 goto errout; 2448 2449 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2450 if (err < 0) 2451 goto errout; 2452 2453 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2454 return; 2455 2456 errout: 2457 kfree_skb(skb); 2458 if (err < 0) 2459 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2460 } 2461 2462 static size_t mrt6msg_netlink_msgsize(size_t payloadlen) 2463 { 2464 size_t len = 2465 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2466 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */ 2467 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */ 2468 /* IP6MRA_CREPORT_SRC_ADDR */ 2469 + nla_total_size(sizeof(struct in6_addr)) 2470 /* IP6MRA_CREPORT_DST_ADDR */ 2471 + nla_total_size(sizeof(struct in6_addr)) 2472 /* IP6MRA_CREPORT_PKT */ 2473 + nla_total_size(payloadlen) 2474 ; 2475 2476 return len; 2477 } 2478 2479 static void mrt6msg_netlink_event(struct mr6_table *mrt, struct sk_buff *pkt) 2480 { 2481 struct net *net = read_pnet(&mrt->net); 2482 struct nlmsghdr *nlh; 2483 struct rtgenmsg *rtgenm; 2484 struct mrt6msg *msg; 2485 struct sk_buff *skb; 2486 struct nlattr *nla; 2487 int payloadlen; 2488 2489 payloadlen = pkt->len - sizeof(struct mrt6msg); 2490 msg = (struct mrt6msg *)skb_transport_header(pkt); 2491 2492 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2493 if (!skb) 2494 goto errout; 2495 2496 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2497 sizeof(struct rtgenmsg), 0); 2498 if (!nlh) 2499 goto errout; 2500 rtgenm = nlmsg_data(nlh); 2501 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR; 2502 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) || 2503 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) || 2504 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR, 2505 &msg->im6_src) || 2506 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR, 2507 &msg->im6_dst)) 2508 goto nla_put_failure; 2509 2510 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen); 2511 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg), 2512 nla_data(nla), payloadlen)) 2513 goto nla_put_failure; 2514 2515 nlmsg_end(skb, nlh); 2516 2517 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC); 2518 return; 2519 2520 nla_put_failure: 2521 nlmsg_cancel(skb, nlh); 2522 errout: 2523 kfree_skb(skb); 2524 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS); 2525 } 2526 2527 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2528 { 2529 struct net *net = sock_net(skb->sk); 2530 struct mr6_table *mrt; 2531 struct mfc6_cache *mfc; 2532 unsigned int t = 0, s_t; 2533 unsigned int h = 0, s_h; 2534 unsigned int e = 0, s_e; 2535 2536 s_t = cb->args[0]; 2537 s_h = cb->args[1]; 2538 s_e = cb->args[2]; 2539 2540 read_lock(&mrt_lock); 2541 ip6mr_for_each_table(mrt, net) { 2542 if (t < s_t) 2543 goto next_table; 2544 if (t > s_t) 2545 s_h = 0; 2546 for (h = s_h; h < MFC6_LINES; h++) { 2547 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2548 if (e < s_e) 2549 goto next_entry; 2550 if (ip6mr_fill_mroute(mrt, skb, 2551 NETLINK_CB(cb->skb).portid, 2552 cb->nlh->nlmsg_seq, 2553 mfc, RTM_NEWROUTE, 2554 NLM_F_MULTI) < 0) 2555 goto done; 2556 next_entry: 2557 e++; 2558 } 2559 e = s_e = 0; 2560 } 2561 spin_lock_bh(&mfc_unres_lock); 2562 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2563 if (e < s_e) 2564 goto next_entry2; 2565 if (ip6mr_fill_mroute(mrt, skb, 2566 NETLINK_CB(cb->skb).portid, 2567 cb->nlh->nlmsg_seq, 2568 mfc, RTM_NEWROUTE, 2569 NLM_F_MULTI) < 0) { 2570 spin_unlock_bh(&mfc_unres_lock); 2571 goto done; 2572 } 2573 next_entry2: 2574 e++; 2575 } 2576 spin_unlock_bh(&mfc_unres_lock); 2577 e = s_e = 0; 2578 s_h = 0; 2579 next_table: 2580 t++; 2581 } 2582 done: 2583 read_unlock(&mrt_lock); 2584 2585 cb->args[2] = e; 2586 cb->args[1] = h; 2587 cb->args[0] = t; 2588 2589 return skb->len; 2590 } 2591