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(struct timer_list *t); 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 timer_setup(&mrt->ipmr_expire_timer, ipmr_expire_process, 0); 324 325 #ifdef CONFIG_IPV6_PIMSM_V2 326 mrt->mroute_reg_vif_num = -1; 327 #endif 328 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 329 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables); 330 #endif 331 return mrt; 332 } 333 334 static void ip6mr_free_table(struct mr6_table *mrt) 335 { 336 del_timer_sync(&mrt->ipmr_expire_timer); 337 mroute_clean_tables(mrt, true); 338 kfree(mrt); 339 } 340 341 #ifdef CONFIG_PROC_FS 342 343 struct ipmr_mfc_iter { 344 struct seq_net_private p; 345 struct mr6_table *mrt; 346 struct list_head *cache; 347 int ct; 348 }; 349 350 351 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net, 352 struct ipmr_mfc_iter *it, loff_t pos) 353 { 354 struct mr6_table *mrt = it->mrt; 355 struct mfc6_cache *mfc; 356 357 read_lock(&mrt_lock); 358 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) { 359 it->cache = &mrt->mfc6_cache_array[it->ct]; 360 list_for_each_entry(mfc, it->cache, list) 361 if (pos-- == 0) 362 return mfc; 363 } 364 read_unlock(&mrt_lock); 365 366 spin_lock_bh(&mfc_unres_lock); 367 it->cache = &mrt->mfc6_unres_queue; 368 list_for_each_entry(mfc, it->cache, list) 369 if (pos-- == 0) 370 return mfc; 371 spin_unlock_bh(&mfc_unres_lock); 372 373 it->cache = NULL; 374 return NULL; 375 } 376 377 /* 378 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif 379 */ 380 381 struct ipmr_vif_iter { 382 struct seq_net_private p; 383 struct mr6_table *mrt; 384 int ct; 385 }; 386 387 static struct mif_device *ip6mr_vif_seq_idx(struct net *net, 388 struct ipmr_vif_iter *iter, 389 loff_t pos) 390 { 391 struct mr6_table *mrt = iter->mrt; 392 393 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 394 if (!MIF_EXISTS(mrt, iter->ct)) 395 continue; 396 if (pos-- == 0) 397 return &mrt->vif6_table[iter->ct]; 398 } 399 return NULL; 400 } 401 402 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) 403 __acquires(mrt_lock) 404 { 405 struct ipmr_vif_iter *iter = seq->private; 406 struct net *net = seq_file_net(seq); 407 struct mr6_table *mrt; 408 409 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 410 if (!mrt) 411 return ERR_PTR(-ENOENT); 412 413 iter->mrt = mrt; 414 415 read_lock(&mrt_lock); 416 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1) 417 : SEQ_START_TOKEN; 418 } 419 420 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 421 { 422 struct ipmr_vif_iter *iter = seq->private; 423 struct net *net = seq_file_net(seq); 424 struct mr6_table *mrt = iter->mrt; 425 426 ++*pos; 427 if (v == SEQ_START_TOKEN) 428 return ip6mr_vif_seq_idx(net, iter, 0); 429 430 while (++iter->ct < mrt->maxvif) { 431 if (!MIF_EXISTS(mrt, iter->ct)) 432 continue; 433 return &mrt->vif6_table[iter->ct]; 434 } 435 return NULL; 436 } 437 438 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) 439 __releases(mrt_lock) 440 { 441 read_unlock(&mrt_lock); 442 } 443 444 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) 445 { 446 struct ipmr_vif_iter *iter = seq->private; 447 struct mr6_table *mrt = iter->mrt; 448 449 if (v == SEQ_START_TOKEN) { 450 seq_puts(seq, 451 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); 452 } else { 453 const struct mif_device *vif = v; 454 const char *name = vif->dev ? vif->dev->name : "none"; 455 456 seq_printf(seq, 457 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", 458 vif - mrt->vif6_table, 459 name, vif->bytes_in, vif->pkt_in, 460 vif->bytes_out, vif->pkt_out, 461 vif->flags); 462 } 463 return 0; 464 } 465 466 static const struct seq_operations ip6mr_vif_seq_ops = { 467 .start = ip6mr_vif_seq_start, 468 .next = ip6mr_vif_seq_next, 469 .stop = ip6mr_vif_seq_stop, 470 .show = ip6mr_vif_seq_show, 471 }; 472 473 static int ip6mr_vif_open(struct inode *inode, struct file *file) 474 { 475 return seq_open_net(inode, file, &ip6mr_vif_seq_ops, 476 sizeof(struct ipmr_vif_iter)); 477 } 478 479 static const struct file_operations ip6mr_vif_fops = { 480 .owner = THIS_MODULE, 481 .open = ip6mr_vif_open, 482 .read = seq_read, 483 .llseek = seq_lseek, 484 .release = seq_release_net, 485 }; 486 487 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 488 { 489 struct ipmr_mfc_iter *it = seq->private; 490 struct net *net = seq_file_net(seq); 491 struct mr6_table *mrt; 492 493 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 494 if (!mrt) 495 return ERR_PTR(-ENOENT); 496 497 it->mrt = mrt; 498 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 499 : SEQ_START_TOKEN; 500 } 501 502 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 503 { 504 struct mfc6_cache *mfc = v; 505 struct ipmr_mfc_iter *it = seq->private; 506 struct net *net = seq_file_net(seq); 507 struct mr6_table *mrt = it->mrt; 508 509 ++*pos; 510 511 if (v == SEQ_START_TOKEN) 512 return ipmr_mfc_seq_idx(net, seq->private, 0); 513 514 if (mfc->list.next != it->cache) 515 return list_entry(mfc->list.next, struct mfc6_cache, list); 516 517 if (it->cache == &mrt->mfc6_unres_queue) 518 goto end_of_list; 519 520 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]); 521 522 while (++it->ct < MFC6_LINES) { 523 it->cache = &mrt->mfc6_cache_array[it->ct]; 524 if (list_empty(it->cache)) 525 continue; 526 return list_first_entry(it->cache, struct mfc6_cache, list); 527 } 528 529 /* exhausted cache_array, show unresolved */ 530 read_unlock(&mrt_lock); 531 it->cache = &mrt->mfc6_unres_queue; 532 it->ct = 0; 533 534 spin_lock_bh(&mfc_unres_lock); 535 if (!list_empty(it->cache)) 536 return list_first_entry(it->cache, struct mfc6_cache, list); 537 538 end_of_list: 539 spin_unlock_bh(&mfc_unres_lock); 540 it->cache = NULL; 541 542 return NULL; 543 } 544 545 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 546 { 547 struct ipmr_mfc_iter *it = seq->private; 548 struct mr6_table *mrt = it->mrt; 549 550 if (it->cache == &mrt->mfc6_unres_queue) 551 spin_unlock_bh(&mfc_unres_lock); 552 else if (it->cache == &mrt->mfc6_cache_array[it->ct]) 553 read_unlock(&mrt_lock); 554 } 555 556 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 557 { 558 int n; 559 560 if (v == SEQ_START_TOKEN) { 561 seq_puts(seq, 562 "Group " 563 "Origin " 564 "Iif Pkts Bytes Wrong Oifs\n"); 565 } else { 566 const struct mfc6_cache *mfc = v; 567 const struct ipmr_mfc_iter *it = seq->private; 568 struct mr6_table *mrt = it->mrt; 569 570 seq_printf(seq, "%pI6 %pI6 %-3hd", 571 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, 572 mfc->mf6c_parent); 573 574 if (it->cache != &mrt->mfc6_unres_queue) { 575 seq_printf(seq, " %8lu %8lu %8lu", 576 mfc->mfc_un.res.pkt, 577 mfc->mfc_un.res.bytes, 578 mfc->mfc_un.res.wrong_if); 579 for (n = mfc->mfc_un.res.minvif; 580 n < mfc->mfc_un.res.maxvif; n++) { 581 if (MIF_EXISTS(mrt, n) && 582 mfc->mfc_un.res.ttls[n] < 255) 583 seq_printf(seq, 584 " %2d:%-3d", 585 n, mfc->mfc_un.res.ttls[n]); 586 } 587 } else { 588 /* unresolved mfc_caches don't contain 589 * pkt, bytes and wrong_if values 590 */ 591 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 592 } 593 seq_putc(seq, '\n'); 594 } 595 return 0; 596 } 597 598 static const struct seq_operations ipmr_mfc_seq_ops = { 599 .start = ipmr_mfc_seq_start, 600 .next = ipmr_mfc_seq_next, 601 .stop = ipmr_mfc_seq_stop, 602 .show = ipmr_mfc_seq_show, 603 }; 604 605 static int ipmr_mfc_open(struct inode *inode, struct file *file) 606 { 607 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 608 sizeof(struct ipmr_mfc_iter)); 609 } 610 611 static const struct file_operations ip6mr_mfc_fops = { 612 .owner = THIS_MODULE, 613 .open = ipmr_mfc_open, 614 .read = seq_read, 615 .llseek = seq_lseek, 616 .release = seq_release_net, 617 }; 618 #endif 619 620 #ifdef CONFIG_IPV6_PIMSM_V2 621 622 static int pim6_rcv(struct sk_buff *skb) 623 { 624 struct pimreghdr *pim; 625 struct ipv6hdr *encap; 626 struct net_device *reg_dev = NULL; 627 struct net *net = dev_net(skb->dev); 628 struct mr6_table *mrt; 629 struct flowi6 fl6 = { 630 .flowi6_iif = skb->dev->ifindex, 631 .flowi6_mark = skb->mark, 632 }; 633 int reg_vif_num; 634 635 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) 636 goto drop; 637 638 pim = (struct pimreghdr *)skb_transport_header(skb); 639 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) || 640 (pim->flags & PIM_NULL_REGISTER) || 641 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 642 sizeof(*pim), IPPROTO_PIM, 643 csum_partial((void *)pim, sizeof(*pim), 0)) && 644 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 645 goto drop; 646 647 /* check if the inner packet is destined to mcast group */ 648 encap = (struct ipv6hdr *)(skb_transport_header(skb) + 649 sizeof(*pim)); 650 651 if (!ipv6_addr_is_multicast(&encap->daddr) || 652 encap->payload_len == 0 || 653 ntohs(encap->payload_len) + sizeof(*pim) > skb->len) 654 goto drop; 655 656 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 657 goto drop; 658 reg_vif_num = mrt->mroute_reg_vif_num; 659 660 read_lock(&mrt_lock); 661 if (reg_vif_num >= 0) 662 reg_dev = mrt->vif6_table[reg_vif_num].dev; 663 if (reg_dev) 664 dev_hold(reg_dev); 665 read_unlock(&mrt_lock); 666 667 if (!reg_dev) 668 goto drop; 669 670 skb->mac_header = skb->network_header; 671 skb_pull(skb, (u8 *)encap - skb->data); 672 skb_reset_network_header(skb); 673 skb->protocol = htons(ETH_P_IPV6); 674 skb->ip_summed = CHECKSUM_NONE; 675 676 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 677 678 netif_rx(skb); 679 680 dev_put(reg_dev); 681 return 0; 682 drop: 683 kfree_skb(skb); 684 return 0; 685 } 686 687 static const struct inet6_protocol pim6_protocol = { 688 .handler = pim6_rcv, 689 }; 690 691 /* Service routines creating virtual interfaces: PIMREG */ 692 693 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, 694 struct net_device *dev) 695 { 696 struct net *net = dev_net(dev); 697 struct mr6_table *mrt; 698 struct flowi6 fl6 = { 699 .flowi6_oif = dev->ifindex, 700 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 701 .flowi6_mark = skb->mark, 702 }; 703 int err; 704 705 err = ip6mr_fib_lookup(net, &fl6, &mrt); 706 if (err < 0) { 707 kfree_skb(skb); 708 return err; 709 } 710 711 read_lock(&mrt_lock); 712 dev->stats.tx_bytes += skb->len; 713 dev->stats.tx_packets++; 714 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT); 715 read_unlock(&mrt_lock); 716 kfree_skb(skb); 717 return NETDEV_TX_OK; 718 } 719 720 static int reg_vif_get_iflink(const struct net_device *dev) 721 { 722 return 0; 723 } 724 725 static const struct net_device_ops reg_vif_netdev_ops = { 726 .ndo_start_xmit = reg_vif_xmit, 727 .ndo_get_iflink = reg_vif_get_iflink, 728 }; 729 730 static void reg_vif_setup(struct net_device *dev) 731 { 732 dev->type = ARPHRD_PIMREG; 733 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; 734 dev->flags = IFF_NOARP; 735 dev->netdev_ops = ®_vif_netdev_ops; 736 dev->needs_free_netdev = true; 737 dev->features |= NETIF_F_NETNS_LOCAL; 738 } 739 740 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt) 741 { 742 struct net_device *dev; 743 char name[IFNAMSIZ]; 744 745 if (mrt->id == RT6_TABLE_DFLT) 746 sprintf(name, "pim6reg"); 747 else 748 sprintf(name, "pim6reg%u", mrt->id); 749 750 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 751 if (!dev) 752 return NULL; 753 754 dev_net_set(dev, net); 755 756 if (register_netdevice(dev)) { 757 free_netdev(dev); 758 return NULL; 759 } 760 761 if (dev_open(dev)) 762 goto failure; 763 764 dev_hold(dev); 765 return dev; 766 767 failure: 768 unregister_netdevice(dev); 769 return NULL; 770 } 771 #endif 772 773 /* 774 * Delete a VIF entry 775 */ 776 777 static int mif6_delete(struct mr6_table *mrt, int vifi, int notify, 778 struct list_head *head) 779 { 780 struct mif_device *v; 781 struct net_device *dev; 782 struct inet6_dev *in6_dev; 783 784 if (vifi < 0 || vifi >= mrt->maxvif) 785 return -EADDRNOTAVAIL; 786 787 v = &mrt->vif6_table[vifi]; 788 789 write_lock_bh(&mrt_lock); 790 dev = v->dev; 791 v->dev = NULL; 792 793 if (!dev) { 794 write_unlock_bh(&mrt_lock); 795 return -EADDRNOTAVAIL; 796 } 797 798 #ifdef CONFIG_IPV6_PIMSM_V2 799 if (vifi == mrt->mroute_reg_vif_num) 800 mrt->mroute_reg_vif_num = -1; 801 #endif 802 803 if (vifi + 1 == mrt->maxvif) { 804 int tmp; 805 for (tmp = vifi - 1; tmp >= 0; tmp--) { 806 if (MIF_EXISTS(mrt, tmp)) 807 break; 808 } 809 mrt->maxvif = tmp + 1; 810 } 811 812 write_unlock_bh(&mrt_lock); 813 814 dev_set_allmulti(dev, -1); 815 816 in6_dev = __in6_dev_get(dev); 817 if (in6_dev) { 818 in6_dev->cnf.mc_forwarding--; 819 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 820 NETCONFA_MC_FORWARDING, 821 dev->ifindex, &in6_dev->cnf); 822 } 823 824 if ((v->flags & MIFF_REGISTER) && !notify) 825 unregister_netdevice_queue(dev, head); 826 827 dev_put(dev); 828 return 0; 829 } 830 831 static inline void ip6mr_cache_free(struct mfc6_cache *c) 832 { 833 kmem_cache_free(mrt_cachep, c); 834 } 835 836 /* Destroy an unresolved cache entry, killing queued skbs 837 and reporting error to netlink readers. 838 */ 839 840 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c) 841 { 842 struct net *net = read_pnet(&mrt->net); 843 struct sk_buff *skb; 844 845 atomic_dec(&mrt->cache_resolve_queue_len); 846 847 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { 848 if (ipv6_hdr(skb)->version == 0) { 849 struct nlmsghdr *nlh = skb_pull(skb, 850 sizeof(struct ipv6hdr)); 851 nlh->nlmsg_type = NLMSG_ERROR; 852 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 853 skb_trim(skb, nlh->nlmsg_len); 854 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT; 855 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 856 } else 857 kfree_skb(skb); 858 } 859 860 ip6mr_cache_free(c); 861 } 862 863 864 /* Timer process for all the unresolved queue. */ 865 866 static void ipmr_do_expire_process(struct mr6_table *mrt) 867 { 868 unsigned long now = jiffies; 869 unsigned long expires = 10 * HZ; 870 struct mfc6_cache *c, *next; 871 872 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 873 if (time_after(c->mfc_un.unres.expires, now)) { 874 /* not yet... */ 875 unsigned long interval = c->mfc_un.unres.expires - now; 876 if (interval < expires) 877 expires = interval; 878 continue; 879 } 880 881 list_del(&c->list); 882 mr6_netlink_event(mrt, c, RTM_DELROUTE); 883 ip6mr_destroy_unres(mrt, c); 884 } 885 886 if (!list_empty(&mrt->mfc6_unres_queue)) 887 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 888 } 889 890 static void ipmr_expire_process(struct timer_list *t) 891 { 892 struct mr6_table *mrt = from_timer(mrt, t, ipmr_expire_timer); 893 894 if (!spin_trylock(&mfc_unres_lock)) { 895 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 896 return; 897 } 898 899 if (!list_empty(&mrt->mfc6_unres_queue)) 900 ipmr_do_expire_process(mrt); 901 902 spin_unlock(&mfc_unres_lock); 903 } 904 905 /* Fill oifs list. It is called under write locked mrt_lock. */ 906 907 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache, 908 unsigned char *ttls) 909 { 910 int vifi; 911 912 cache->mfc_un.res.minvif = MAXMIFS; 913 cache->mfc_un.res.maxvif = 0; 914 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 915 916 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 917 if (MIF_EXISTS(mrt, vifi) && 918 ttls[vifi] && ttls[vifi] < 255) { 919 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 920 if (cache->mfc_un.res.minvif > vifi) 921 cache->mfc_un.res.minvif = vifi; 922 if (cache->mfc_un.res.maxvif <= vifi) 923 cache->mfc_un.res.maxvif = vifi + 1; 924 } 925 } 926 cache->mfc_un.res.lastuse = jiffies; 927 } 928 929 static int mif6_add(struct net *net, struct mr6_table *mrt, 930 struct mif6ctl *vifc, int mrtsock) 931 { 932 int vifi = vifc->mif6c_mifi; 933 struct mif_device *v = &mrt->vif6_table[vifi]; 934 struct net_device *dev; 935 struct inet6_dev *in6_dev; 936 int err; 937 938 /* Is vif busy ? */ 939 if (MIF_EXISTS(mrt, vifi)) 940 return -EADDRINUSE; 941 942 switch (vifc->mif6c_flags) { 943 #ifdef CONFIG_IPV6_PIMSM_V2 944 case MIFF_REGISTER: 945 /* 946 * Special Purpose VIF in PIM 947 * All the packets will be sent to the daemon 948 */ 949 if (mrt->mroute_reg_vif_num >= 0) 950 return -EADDRINUSE; 951 dev = ip6mr_reg_vif(net, mrt); 952 if (!dev) 953 return -ENOBUFS; 954 err = dev_set_allmulti(dev, 1); 955 if (err) { 956 unregister_netdevice(dev); 957 dev_put(dev); 958 return err; 959 } 960 break; 961 #endif 962 case 0: 963 dev = dev_get_by_index(net, vifc->mif6c_pifi); 964 if (!dev) 965 return -EADDRNOTAVAIL; 966 err = dev_set_allmulti(dev, 1); 967 if (err) { 968 dev_put(dev); 969 return err; 970 } 971 break; 972 default: 973 return -EINVAL; 974 } 975 976 in6_dev = __in6_dev_get(dev); 977 if (in6_dev) { 978 in6_dev->cnf.mc_forwarding++; 979 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 980 NETCONFA_MC_FORWARDING, 981 dev->ifindex, &in6_dev->cnf); 982 } 983 984 /* 985 * Fill in the VIF structures 986 */ 987 v->rate_limit = vifc->vifc_rate_limit; 988 v->flags = vifc->mif6c_flags; 989 if (!mrtsock) 990 v->flags |= VIFF_STATIC; 991 v->threshold = vifc->vifc_threshold; 992 v->bytes_in = 0; 993 v->bytes_out = 0; 994 v->pkt_in = 0; 995 v->pkt_out = 0; 996 v->link = dev->ifindex; 997 if (v->flags & MIFF_REGISTER) 998 v->link = dev_get_iflink(dev); 999 1000 /* And finish update writing critical data */ 1001 write_lock_bh(&mrt_lock); 1002 v->dev = dev; 1003 #ifdef CONFIG_IPV6_PIMSM_V2 1004 if (v->flags & MIFF_REGISTER) 1005 mrt->mroute_reg_vif_num = vifi; 1006 #endif 1007 if (vifi + 1 > mrt->maxvif) 1008 mrt->maxvif = vifi + 1; 1009 write_unlock_bh(&mrt_lock); 1010 return 0; 1011 } 1012 1013 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt, 1014 const struct in6_addr *origin, 1015 const struct in6_addr *mcastgrp) 1016 { 1017 int line = MFC6_HASH(mcastgrp, origin); 1018 struct mfc6_cache *c; 1019 1020 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1021 if (ipv6_addr_equal(&c->mf6c_origin, origin) && 1022 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) 1023 return c; 1024 } 1025 return NULL; 1026 } 1027 1028 /* Look for a (*,*,oif) entry */ 1029 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt, 1030 mifi_t mifi) 1031 { 1032 int line = MFC6_HASH(&in6addr_any, &in6addr_any); 1033 struct mfc6_cache *c; 1034 1035 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1036 if (ipv6_addr_any(&c->mf6c_origin) && 1037 ipv6_addr_any(&c->mf6c_mcastgrp) && 1038 (c->mfc_un.res.ttls[mifi] < 255)) 1039 return c; 1040 1041 return NULL; 1042 } 1043 1044 /* Look for a (*,G) entry */ 1045 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt, 1046 struct in6_addr *mcastgrp, 1047 mifi_t mifi) 1048 { 1049 int line = MFC6_HASH(mcastgrp, &in6addr_any); 1050 struct mfc6_cache *c, *proxy; 1051 1052 if (ipv6_addr_any(mcastgrp)) 1053 goto skip; 1054 1055 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1056 if (ipv6_addr_any(&c->mf6c_origin) && 1057 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) { 1058 if (c->mfc_un.res.ttls[mifi] < 255) 1059 return c; 1060 1061 /* It's ok if the mifi is part of the static tree */ 1062 proxy = ip6mr_cache_find_any_parent(mrt, 1063 c->mf6c_parent); 1064 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255) 1065 return c; 1066 } 1067 1068 skip: 1069 return ip6mr_cache_find_any_parent(mrt, mifi); 1070 } 1071 1072 /* 1073 * Allocate a multicast cache entry 1074 */ 1075 static struct mfc6_cache *ip6mr_cache_alloc(void) 1076 { 1077 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 1078 if (!c) 1079 return NULL; 1080 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 1081 c->mfc_un.res.minvif = MAXMIFS; 1082 return c; 1083 } 1084 1085 static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 1086 { 1087 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 1088 if (!c) 1089 return NULL; 1090 skb_queue_head_init(&c->mfc_un.unres.unresolved); 1091 c->mfc_un.unres.expires = jiffies + 10 * HZ; 1092 return c; 1093 } 1094 1095 /* 1096 * A cache entry has gone into a resolved state from queued 1097 */ 1098 1099 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt, 1100 struct mfc6_cache *uc, struct mfc6_cache *c) 1101 { 1102 struct sk_buff *skb; 1103 1104 /* 1105 * Play the pending entries through our router 1106 */ 1107 1108 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 1109 if (ipv6_hdr(skb)->version == 0) { 1110 struct nlmsghdr *nlh = skb_pull(skb, 1111 sizeof(struct ipv6hdr)); 1112 1113 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 1114 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1115 } else { 1116 nlh->nlmsg_type = NLMSG_ERROR; 1117 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1118 skb_trim(skb, nlh->nlmsg_len); 1119 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE; 1120 } 1121 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1122 } else 1123 ip6_mr_forward(net, mrt, skb, c); 1124 } 1125 } 1126 1127 /* 1128 * Bounce a cache query up to pim6sd and netlink. 1129 * 1130 * Called under mrt_lock. 1131 */ 1132 1133 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 1134 mifi_t mifi, int assert) 1135 { 1136 struct sk_buff *skb; 1137 struct mrt6msg *msg; 1138 int ret; 1139 1140 #ifdef CONFIG_IPV6_PIMSM_V2 1141 if (assert == MRT6MSG_WHOLEPKT) 1142 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1143 +sizeof(*msg)); 1144 else 1145 #endif 1146 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1147 1148 if (!skb) 1149 return -ENOBUFS; 1150 1151 /* I suppose that internal messages 1152 * do not require checksums */ 1153 1154 skb->ip_summed = CHECKSUM_UNNECESSARY; 1155 1156 #ifdef CONFIG_IPV6_PIMSM_V2 1157 if (assert == MRT6MSG_WHOLEPKT) { 1158 /* Ugly, but we have no choice with this interface. 1159 Duplicate old header, fix length etc. 1160 And all this only to mangle msg->im6_msgtype and 1161 to set msg->im6_mbz to "mbz" :-) 1162 */ 1163 skb_push(skb, -skb_network_offset(pkt)); 1164 1165 skb_push(skb, sizeof(*msg)); 1166 skb_reset_transport_header(skb); 1167 msg = (struct mrt6msg *)skb_transport_header(skb); 1168 msg->im6_mbz = 0; 1169 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1170 msg->im6_mif = mrt->mroute_reg_vif_num; 1171 msg->im6_pad = 0; 1172 msg->im6_src = ipv6_hdr(pkt)->saddr; 1173 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1174 1175 skb->ip_summed = CHECKSUM_UNNECESSARY; 1176 } else 1177 #endif 1178 { 1179 /* 1180 * Copy the IP header 1181 */ 1182 1183 skb_put(skb, sizeof(struct ipv6hdr)); 1184 skb_reset_network_header(skb); 1185 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1186 1187 /* 1188 * Add our header 1189 */ 1190 skb_put(skb, sizeof(*msg)); 1191 skb_reset_transport_header(skb); 1192 msg = (struct mrt6msg *)skb_transport_header(skb); 1193 1194 msg->im6_mbz = 0; 1195 msg->im6_msgtype = assert; 1196 msg->im6_mif = mifi; 1197 msg->im6_pad = 0; 1198 msg->im6_src = ipv6_hdr(pkt)->saddr; 1199 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1200 1201 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1202 skb->ip_summed = CHECKSUM_UNNECESSARY; 1203 } 1204 1205 if (!mrt->mroute6_sk) { 1206 kfree_skb(skb); 1207 return -EINVAL; 1208 } 1209 1210 mrt6msg_netlink_event(mrt, skb); 1211 1212 /* 1213 * Deliver to user space multicast routing algorithms 1214 */ 1215 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb); 1216 if (ret < 0) { 1217 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1218 kfree_skb(skb); 1219 } 1220 1221 return ret; 1222 } 1223 1224 /* 1225 * Queue a packet for resolution. It gets locked cache entry! 1226 */ 1227 1228 static int 1229 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb) 1230 { 1231 bool found = false; 1232 int err; 1233 struct mfc6_cache *c; 1234 1235 spin_lock_bh(&mfc_unres_lock); 1236 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) { 1237 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1238 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1239 found = true; 1240 break; 1241 } 1242 } 1243 1244 if (!found) { 1245 /* 1246 * Create a new entry if allowable 1247 */ 1248 1249 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1250 (c = ip6mr_cache_alloc_unres()) == NULL) { 1251 spin_unlock_bh(&mfc_unres_lock); 1252 1253 kfree_skb(skb); 1254 return -ENOBUFS; 1255 } 1256 1257 /* 1258 * Fill in the new cache entry 1259 */ 1260 c->mf6c_parent = -1; 1261 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1262 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1263 1264 /* 1265 * Reflect first query at pim6sd 1266 */ 1267 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1268 if (err < 0) { 1269 /* If the report failed throw the cache entry 1270 out - Brad Parker 1271 */ 1272 spin_unlock_bh(&mfc_unres_lock); 1273 1274 ip6mr_cache_free(c); 1275 kfree_skb(skb); 1276 return err; 1277 } 1278 1279 atomic_inc(&mrt->cache_resolve_queue_len); 1280 list_add(&c->list, &mrt->mfc6_unres_queue); 1281 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1282 1283 ipmr_do_expire_process(mrt); 1284 } 1285 1286 /* 1287 * See if we can append the packet 1288 */ 1289 if (c->mfc_un.unres.unresolved.qlen > 3) { 1290 kfree_skb(skb); 1291 err = -ENOBUFS; 1292 } else { 1293 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1294 err = 0; 1295 } 1296 1297 spin_unlock_bh(&mfc_unres_lock); 1298 return err; 1299 } 1300 1301 /* 1302 * MFC6 cache manipulation by user space 1303 */ 1304 1305 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc, 1306 int parent) 1307 { 1308 int line; 1309 struct mfc6_cache *c, *next; 1310 1311 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1312 1313 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) { 1314 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1315 ipv6_addr_equal(&c->mf6c_mcastgrp, 1316 &mfc->mf6cc_mcastgrp.sin6_addr) && 1317 (parent == -1 || parent == c->mf6c_parent)) { 1318 write_lock_bh(&mrt_lock); 1319 list_del(&c->list); 1320 write_unlock_bh(&mrt_lock); 1321 1322 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1323 ip6mr_cache_free(c); 1324 return 0; 1325 } 1326 } 1327 return -ENOENT; 1328 } 1329 1330 static int ip6mr_device_event(struct notifier_block *this, 1331 unsigned long event, void *ptr) 1332 { 1333 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1334 struct net *net = dev_net(dev); 1335 struct mr6_table *mrt; 1336 struct mif_device *v; 1337 int ct; 1338 1339 if (event != NETDEV_UNREGISTER) 1340 return NOTIFY_DONE; 1341 1342 ip6mr_for_each_table(mrt, net) { 1343 v = &mrt->vif6_table[0]; 1344 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1345 if (v->dev == dev) 1346 mif6_delete(mrt, ct, 1, NULL); 1347 } 1348 } 1349 1350 return NOTIFY_DONE; 1351 } 1352 1353 static struct notifier_block ip6_mr_notifier = { 1354 .notifier_call = ip6mr_device_event 1355 }; 1356 1357 /* 1358 * Setup for IP multicast routing 1359 */ 1360 1361 static int __net_init ip6mr_net_init(struct net *net) 1362 { 1363 int err; 1364 1365 err = ip6mr_rules_init(net); 1366 if (err < 0) 1367 goto fail; 1368 1369 #ifdef CONFIG_PROC_FS 1370 err = -ENOMEM; 1371 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops)) 1372 goto proc_vif_fail; 1373 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops)) 1374 goto proc_cache_fail; 1375 #endif 1376 1377 return 0; 1378 1379 #ifdef CONFIG_PROC_FS 1380 proc_cache_fail: 1381 remove_proc_entry("ip6_mr_vif", net->proc_net); 1382 proc_vif_fail: 1383 ip6mr_rules_exit(net); 1384 #endif 1385 fail: 1386 return err; 1387 } 1388 1389 static void __net_exit ip6mr_net_exit(struct net *net) 1390 { 1391 #ifdef CONFIG_PROC_FS 1392 remove_proc_entry("ip6_mr_cache", net->proc_net); 1393 remove_proc_entry("ip6_mr_vif", net->proc_net); 1394 #endif 1395 ip6mr_rules_exit(net); 1396 } 1397 1398 static struct pernet_operations ip6mr_net_ops = { 1399 .init = ip6mr_net_init, 1400 .exit = ip6mr_net_exit, 1401 }; 1402 1403 int __init ip6_mr_init(void) 1404 { 1405 int err; 1406 1407 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1408 sizeof(struct mfc6_cache), 1409 0, SLAB_HWCACHE_ALIGN, 1410 NULL); 1411 if (!mrt_cachep) 1412 return -ENOMEM; 1413 1414 err = register_pernet_subsys(&ip6mr_net_ops); 1415 if (err) 1416 goto reg_pernet_fail; 1417 1418 err = register_netdevice_notifier(&ip6_mr_notifier); 1419 if (err) 1420 goto reg_notif_fail; 1421 #ifdef CONFIG_IPV6_PIMSM_V2 1422 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1423 pr_err("%s: can't add PIM protocol\n", __func__); 1424 err = -EAGAIN; 1425 goto add_proto_fail; 1426 } 1427 #endif 1428 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, 1429 ip6mr_rtm_dumproute, 0); 1430 return 0; 1431 #ifdef CONFIG_IPV6_PIMSM_V2 1432 add_proto_fail: 1433 unregister_netdevice_notifier(&ip6_mr_notifier); 1434 #endif 1435 reg_notif_fail: 1436 unregister_pernet_subsys(&ip6mr_net_ops); 1437 reg_pernet_fail: 1438 kmem_cache_destroy(mrt_cachep); 1439 return err; 1440 } 1441 1442 void ip6_mr_cleanup(void) 1443 { 1444 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1445 #ifdef CONFIG_IPV6_PIMSM_V2 1446 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1447 #endif 1448 unregister_netdevice_notifier(&ip6_mr_notifier); 1449 unregister_pernet_subsys(&ip6mr_net_ops); 1450 kmem_cache_destroy(mrt_cachep); 1451 } 1452 1453 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1454 struct mf6cctl *mfc, int mrtsock, int parent) 1455 { 1456 bool found = false; 1457 int line; 1458 struct mfc6_cache *uc, *c; 1459 unsigned char ttls[MAXMIFS]; 1460 int i; 1461 1462 if (mfc->mf6cc_parent >= MAXMIFS) 1463 return -ENFILE; 1464 1465 memset(ttls, 255, MAXMIFS); 1466 for (i = 0; i < MAXMIFS; i++) { 1467 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1468 ttls[i] = 1; 1469 1470 } 1471 1472 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1473 1474 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1475 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1476 ipv6_addr_equal(&c->mf6c_mcastgrp, 1477 &mfc->mf6cc_mcastgrp.sin6_addr) && 1478 (parent == -1 || parent == mfc->mf6cc_parent)) { 1479 found = true; 1480 break; 1481 } 1482 } 1483 1484 if (found) { 1485 write_lock_bh(&mrt_lock); 1486 c->mf6c_parent = mfc->mf6cc_parent; 1487 ip6mr_update_thresholds(mrt, c, ttls); 1488 if (!mrtsock) 1489 c->mfc_flags |= MFC_STATIC; 1490 write_unlock_bh(&mrt_lock); 1491 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1492 return 0; 1493 } 1494 1495 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1496 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1497 return -EINVAL; 1498 1499 c = ip6mr_cache_alloc(); 1500 if (!c) 1501 return -ENOMEM; 1502 1503 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1504 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1505 c->mf6c_parent = mfc->mf6cc_parent; 1506 ip6mr_update_thresholds(mrt, c, ttls); 1507 if (!mrtsock) 1508 c->mfc_flags |= MFC_STATIC; 1509 1510 write_lock_bh(&mrt_lock); 1511 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1512 write_unlock_bh(&mrt_lock); 1513 1514 /* 1515 * Check to see if we resolved a queued list. If so we 1516 * need to send on the frames and tidy up. 1517 */ 1518 found = false; 1519 spin_lock_bh(&mfc_unres_lock); 1520 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1521 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1522 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1523 list_del(&uc->list); 1524 atomic_dec(&mrt->cache_resolve_queue_len); 1525 found = true; 1526 break; 1527 } 1528 } 1529 if (list_empty(&mrt->mfc6_unres_queue)) 1530 del_timer(&mrt->ipmr_expire_timer); 1531 spin_unlock_bh(&mfc_unres_lock); 1532 1533 if (found) { 1534 ip6mr_cache_resolve(net, mrt, uc, c); 1535 ip6mr_cache_free(uc); 1536 } 1537 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1538 return 0; 1539 } 1540 1541 /* 1542 * Close the multicast socket, and clear the vif tables etc 1543 */ 1544 1545 static void mroute_clean_tables(struct mr6_table *mrt, bool all) 1546 { 1547 int i; 1548 LIST_HEAD(list); 1549 struct mfc6_cache *c, *next; 1550 1551 /* 1552 * Shut down all active vif entries 1553 */ 1554 for (i = 0; i < mrt->maxvif; i++) { 1555 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) 1556 continue; 1557 mif6_delete(mrt, i, 0, &list); 1558 } 1559 unregister_netdevice_many(&list); 1560 1561 /* 1562 * Wipe the cache 1563 */ 1564 for (i = 0; i < MFC6_LINES; i++) { 1565 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1566 if (!all && (c->mfc_flags & MFC_STATIC)) 1567 continue; 1568 write_lock_bh(&mrt_lock); 1569 list_del(&c->list); 1570 write_unlock_bh(&mrt_lock); 1571 1572 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1573 ip6mr_cache_free(c); 1574 } 1575 } 1576 1577 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1578 spin_lock_bh(&mfc_unres_lock); 1579 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1580 list_del(&c->list); 1581 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1582 ip6mr_destroy_unres(mrt, c); 1583 } 1584 spin_unlock_bh(&mfc_unres_lock); 1585 } 1586 } 1587 1588 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1589 { 1590 int err = 0; 1591 struct net *net = sock_net(sk); 1592 1593 rtnl_lock(); 1594 write_lock_bh(&mrt_lock); 1595 if (likely(mrt->mroute6_sk == NULL)) { 1596 mrt->mroute6_sk = sk; 1597 net->ipv6.devconf_all->mc_forwarding++; 1598 } else { 1599 err = -EADDRINUSE; 1600 } 1601 write_unlock_bh(&mrt_lock); 1602 1603 if (!err) 1604 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1605 NETCONFA_MC_FORWARDING, 1606 NETCONFA_IFINDEX_ALL, 1607 net->ipv6.devconf_all); 1608 rtnl_unlock(); 1609 1610 return err; 1611 } 1612 1613 int ip6mr_sk_done(struct sock *sk) 1614 { 1615 int err = -EACCES; 1616 struct net *net = sock_net(sk); 1617 struct mr6_table *mrt; 1618 1619 if (sk->sk_type != SOCK_RAW || 1620 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1621 return err; 1622 1623 rtnl_lock(); 1624 ip6mr_for_each_table(mrt, net) { 1625 if (sk == mrt->mroute6_sk) { 1626 write_lock_bh(&mrt_lock); 1627 mrt->mroute6_sk = NULL; 1628 net->ipv6.devconf_all->mc_forwarding--; 1629 write_unlock_bh(&mrt_lock); 1630 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1631 NETCONFA_MC_FORWARDING, 1632 NETCONFA_IFINDEX_ALL, 1633 net->ipv6.devconf_all); 1634 1635 mroute_clean_tables(mrt, false); 1636 err = 0; 1637 break; 1638 } 1639 } 1640 rtnl_unlock(); 1641 1642 return err; 1643 } 1644 1645 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1646 { 1647 struct mr6_table *mrt; 1648 struct flowi6 fl6 = { 1649 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1650 .flowi6_oif = skb->dev->ifindex, 1651 .flowi6_mark = skb->mark, 1652 }; 1653 1654 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1655 return NULL; 1656 1657 return mrt->mroute6_sk; 1658 } 1659 1660 /* 1661 * Socket options and virtual interface manipulation. The whole 1662 * virtual interface system is a complete heap, but unfortunately 1663 * that's how BSD mrouted happens to think. Maybe one day with a proper 1664 * MOSPF/PIM router set up we can clean this up. 1665 */ 1666 1667 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1668 { 1669 int ret, parent = 0; 1670 struct mif6ctl vif; 1671 struct mf6cctl mfc; 1672 mifi_t mifi; 1673 struct net *net = sock_net(sk); 1674 struct mr6_table *mrt; 1675 1676 if (sk->sk_type != SOCK_RAW || 1677 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1678 return -EOPNOTSUPP; 1679 1680 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1681 if (!mrt) 1682 return -ENOENT; 1683 1684 if (optname != MRT6_INIT) { 1685 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1686 return -EACCES; 1687 } 1688 1689 switch (optname) { 1690 case MRT6_INIT: 1691 if (optlen < sizeof(int)) 1692 return -EINVAL; 1693 1694 return ip6mr_sk_init(mrt, sk); 1695 1696 case MRT6_DONE: 1697 return ip6mr_sk_done(sk); 1698 1699 case MRT6_ADD_MIF: 1700 if (optlen < sizeof(vif)) 1701 return -EINVAL; 1702 if (copy_from_user(&vif, optval, sizeof(vif))) 1703 return -EFAULT; 1704 if (vif.mif6c_mifi >= MAXMIFS) 1705 return -ENFILE; 1706 rtnl_lock(); 1707 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1708 rtnl_unlock(); 1709 return ret; 1710 1711 case MRT6_DEL_MIF: 1712 if (optlen < sizeof(mifi_t)) 1713 return -EINVAL; 1714 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1715 return -EFAULT; 1716 rtnl_lock(); 1717 ret = mif6_delete(mrt, mifi, 0, NULL); 1718 rtnl_unlock(); 1719 return ret; 1720 1721 /* 1722 * Manipulate the forwarding caches. These live 1723 * in a sort of kernel/user symbiosis. 1724 */ 1725 case MRT6_ADD_MFC: 1726 case MRT6_DEL_MFC: 1727 parent = -1; 1728 /* fall through */ 1729 case MRT6_ADD_MFC_PROXY: 1730 case MRT6_DEL_MFC_PROXY: 1731 if (optlen < sizeof(mfc)) 1732 return -EINVAL; 1733 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1734 return -EFAULT; 1735 if (parent == 0) 1736 parent = mfc.mf6cc_parent; 1737 rtnl_lock(); 1738 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1739 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1740 else 1741 ret = ip6mr_mfc_add(net, mrt, &mfc, 1742 sk == mrt->mroute6_sk, parent); 1743 rtnl_unlock(); 1744 return ret; 1745 1746 /* 1747 * Control PIM assert (to activate pim will activate assert) 1748 */ 1749 case MRT6_ASSERT: 1750 { 1751 int v; 1752 1753 if (optlen != sizeof(v)) 1754 return -EINVAL; 1755 if (get_user(v, (int __user *)optval)) 1756 return -EFAULT; 1757 mrt->mroute_do_assert = v; 1758 return 0; 1759 } 1760 1761 #ifdef CONFIG_IPV6_PIMSM_V2 1762 case MRT6_PIM: 1763 { 1764 int v; 1765 1766 if (optlen != sizeof(v)) 1767 return -EINVAL; 1768 if (get_user(v, (int __user *)optval)) 1769 return -EFAULT; 1770 v = !!v; 1771 rtnl_lock(); 1772 ret = 0; 1773 if (v != mrt->mroute_do_pim) { 1774 mrt->mroute_do_pim = v; 1775 mrt->mroute_do_assert = v; 1776 } 1777 rtnl_unlock(); 1778 return ret; 1779 } 1780 1781 #endif 1782 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1783 case MRT6_TABLE: 1784 { 1785 u32 v; 1786 1787 if (optlen != sizeof(u32)) 1788 return -EINVAL; 1789 if (get_user(v, (u32 __user *)optval)) 1790 return -EFAULT; 1791 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1792 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1793 return -EINVAL; 1794 if (sk == mrt->mroute6_sk) 1795 return -EBUSY; 1796 1797 rtnl_lock(); 1798 ret = 0; 1799 if (!ip6mr_new_table(net, v)) 1800 ret = -ENOMEM; 1801 raw6_sk(sk)->ip6mr_table = v; 1802 rtnl_unlock(); 1803 return ret; 1804 } 1805 #endif 1806 /* 1807 * Spurious command, or MRT6_VERSION which you cannot 1808 * set. 1809 */ 1810 default: 1811 return -ENOPROTOOPT; 1812 } 1813 } 1814 1815 /* 1816 * Getsock opt support for the multicast routing system. 1817 */ 1818 1819 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1820 int __user *optlen) 1821 { 1822 int olr; 1823 int val; 1824 struct net *net = sock_net(sk); 1825 struct mr6_table *mrt; 1826 1827 if (sk->sk_type != SOCK_RAW || 1828 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1829 return -EOPNOTSUPP; 1830 1831 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1832 if (!mrt) 1833 return -ENOENT; 1834 1835 switch (optname) { 1836 case MRT6_VERSION: 1837 val = 0x0305; 1838 break; 1839 #ifdef CONFIG_IPV6_PIMSM_V2 1840 case MRT6_PIM: 1841 val = mrt->mroute_do_pim; 1842 break; 1843 #endif 1844 case MRT6_ASSERT: 1845 val = mrt->mroute_do_assert; 1846 break; 1847 default: 1848 return -ENOPROTOOPT; 1849 } 1850 1851 if (get_user(olr, optlen)) 1852 return -EFAULT; 1853 1854 olr = min_t(int, olr, sizeof(int)); 1855 if (olr < 0) 1856 return -EINVAL; 1857 1858 if (put_user(olr, optlen)) 1859 return -EFAULT; 1860 if (copy_to_user(optval, &val, olr)) 1861 return -EFAULT; 1862 return 0; 1863 } 1864 1865 /* 1866 * The IP multicast ioctl support routines. 1867 */ 1868 1869 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1870 { 1871 struct sioc_sg_req6 sr; 1872 struct sioc_mif_req6 vr; 1873 struct mif_device *vif; 1874 struct mfc6_cache *c; 1875 struct net *net = sock_net(sk); 1876 struct mr6_table *mrt; 1877 1878 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1879 if (!mrt) 1880 return -ENOENT; 1881 1882 switch (cmd) { 1883 case SIOCGETMIFCNT_IN6: 1884 if (copy_from_user(&vr, arg, sizeof(vr))) 1885 return -EFAULT; 1886 if (vr.mifi >= mrt->maxvif) 1887 return -EINVAL; 1888 read_lock(&mrt_lock); 1889 vif = &mrt->vif6_table[vr.mifi]; 1890 if (MIF_EXISTS(mrt, vr.mifi)) { 1891 vr.icount = vif->pkt_in; 1892 vr.ocount = vif->pkt_out; 1893 vr.ibytes = vif->bytes_in; 1894 vr.obytes = vif->bytes_out; 1895 read_unlock(&mrt_lock); 1896 1897 if (copy_to_user(arg, &vr, sizeof(vr))) 1898 return -EFAULT; 1899 return 0; 1900 } 1901 read_unlock(&mrt_lock); 1902 return -EADDRNOTAVAIL; 1903 case SIOCGETSGCNT_IN6: 1904 if (copy_from_user(&sr, arg, sizeof(sr))) 1905 return -EFAULT; 1906 1907 read_lock(&mrt_lock); 1908 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1909 if (c) { 1910 sr.pktcnt = c->mfc_un.res.pkt; 1911 sr.bytecnt = c->mfc_un.res.bytes; 1912 sr.wrong_if = c->mfc_un.res.wrong_if; 1913 read_unlock(&mrt_lock); 1914 1915 if (copy_to_user(arg, &sr, sizeof(sr))) 1916 return -EFAULT; 1917 return 0; 1918 } 1919 read_unlock(&mrt_lock); 1920 return -EADDRNOTAVAIL; 1921 default: 1922 return -ENOIOCTLCMD; 1923 } 1924 } 1925 1926 #ifdef CONFIG_COMPAT 1927 struct compat_sioc_sg_req6 { 1928 struct sockaddr_in6 src; 1929 struct sockaddr_in6 grp; 1930 compat_ulong_t pktcnt; 1931 compat_ulong_t bytecnt; 1932 compat_ulong_t wrong_if; 1933 }; 1934 1935 struct compat_sioc_mif_req6 { 1936 mifi_t mifi; 1937 compat_ulong_t icount; 1938 compat_ulong_t ocount; 1939 compat_ulong_t ibytes; 1940 compat_ulong_t obytes; 1941 }; 1942 1943 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1944 { 1945 struct compat_sioc_sg_req6 sr; 1946 struct compat_sioc_mif_req6 vr; 1947 struct mif_device *vif; 1948 struct mfc6_cache *c; 1949 struct net *net = sock_net(sk); 1950 struct mr6_table *mrt; 1951 1952 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1953 if (!mrt) 1954 return -ENOENT; 1955 1956 switch (cmd) { 1957 case SIOCGETMIFCNT_IN6: 1958 if (copy_from_user(&vr, arg, sizeof(vr))) 1959 return -EFAULT; 1960 if (vr.mifi >= mrt->maxvif) 1961 return -EINVAL; 1962 read_lock(&mrt_lock); 1963 vif = &mrt->vif6_table[vr.mifi]; 1964 if (MIF_EXISTS(mrt, vr.mifi)) { 1965 vr.icount = vif->pkt_in; 1966 vr.ocount = vif->pkt_out; 1967 vr.ibytes = vif->bytes_in; 1968 vr.obytes = vif->bytes_out; 1969 read_unlock(&mrt_lock); 1970 1971 if (copy_to_user(arg, &vr, sizeof(vr))) 1972 return -EFAULT; 1973 return 0; 1974 } 1975 read_unlock(&mrt_lock); 1976 return -EADDRNOTAVAIL; 1977 case SIOCGETSGCNT_IN6: 1978 if (copy_from_user(&sr, arg, sizeof(sr))) 1979 return -EFAULT; 1980 1981 read_lock(&mrt_lock); 1982 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1983 if (c) { 1984 sr.pktcnt = c->mfc_un.res.pkt; 1985 sr.bytecnt = c->mfc_un.res.bytes; 1986 sr.wrong_if = c->mfc_un.res.wrong_if; 1987 read_unlock(&mrt_lock); 1988 1989 if (copy_to_user(arg, &sr, sizeof(sr))) 1990 return -EFAULT; 1991 return 0; 1992 } 1993 read_unlock(&mrt_lock); 1994 return -EADDRNOTAVAIL; 1995 default: 1996 return -ENOIOCTLCMD; 1997 } 1998 } 1999 #endif 2000 2001 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 2002 { 2003 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 2004 IPSTATS_MIB_OUTFORWDATAGRAMS); 2005 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)), 2006 IPSTATS_MIB_OUTOCTETS, skb->len); 2007 return dst_output(net, sk, skb); 2008 } 2009 2010 /* 2011 * Processing handlers for ip6mr_forward 2012 */ 2013 2014 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 2015 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 2016 { 2017 struct ipv6hdr *ipv6h; 2018 struct mif_device *vif = &mrt->vif6_table[vifi]; 2019 struct net_device *dev; 2020 struct dst_entry *dst; 2021 struct flowi6 fl6; 2022 2023 if (!vif->dev) 2024 goto out_free; 2025 2026 #ifdef CONFIG_IPV6_PIMSM_V2 2027 if (vif->flags & MIFF_REGISTER) { 2028 vif->pkt_out++; 2029 vif->bytes_out += skb->len; 2030 vif->dev->stats.tx_bytes += skb->len; 2031 vif->dev->stats.tx_packets++; 2032 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2033 goto out_free; 2034 } 2035 #endif 2036 2037 ipv6h = ipv6_hdr(skb); 2038 2039 fl6 = (struct flowi6) { 2040 .flowi6_oif = vif->link, 2041 .daddr = ipv6h->daddr, 2042 }; 2043 2044 dst = ip6_route_output(net, NULL, &fl6); 2045 if (dst->error) { 2046 dst_release(dst); 2047 goto out_free; 2048 } 2049 2050 skb_dst_drop(skb); 2051 skb_dst_set(skb, dst); 2052 2053 /* 2054 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2055 * not only before forwarding, but after forwarding on all output 2056 * interfaces. It is clear, if mrouter runs a multicasting 2057 * program, it should receive packets not depending to what interface 2058 * program is joined. 2059 * If we will not make it, the program will have to join on all 2060 * interfaces. On the other hand, multihoming host (or router, but 2061 * not mrouter) cannot join to more than one interface - it will 2062 * result in receiving multiple packets. 2063 */ 2064 dev = vif->dev; 2065 skb->dev = dev; 2066 vif->pkt_out++; 2067 vif->bytes_out += skb->len; 2068 2069 /* We are about to write */ 2070 /* XXX: extension headers? */ 2071 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2072 goto out_free; 2073 2074 ipv6h = ipv6_hdr(skb); 2075 ipv6h->hop_limit--; 2076 2077 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2078 2079 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2080 net, NULL, skb, skb->dev, dev, 2081 ip6mr_forward2_finish); 2082 2083 out_free: 2084 kfree_skb(skb); 2085 return 0; 2086 } 2087 2088 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2089 { 2090 int ct; 2091 2092 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2093 if (mrt->vif6_table[ct].dev == dev) 2094 break; 2095 } 2096 return ct; 2097 } 2098 2099 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2100 struct sk_buff *skb, struct mfc6_cache *cache) 2101 { 2102 int psend = -1; 2103 int vif, ct; 2104 int true_vifi = ip6mr_find_vif(mrt, skb->dev); 2105 2106 vif = cache->mf6c_parent; 2107 cache->mfc_un.res.pkt++; 2108 cache->mfc_un.res.bytes += skb->len; 2109 cache->mfc_un.res.lastuse = jiffies; 2110 2111 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) { 2112 struct mfc6_cache *cache_proxy; 2113 2114 /* For an (*,G) entry, we only check that the incoming 2115 * interface is part of the static tree. 2116 */ 2117 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif); 2118 if (cache_proxy && 2119 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 2120 goto forward; 2121 } 2122 2123 /* 2124 * Wrong interface: drop packet and (maybe) send PIM assert. 2125 */ 2126 if (mrt->vif6_table[vif].dev != skb->dev) { 2127 cache->mfc_un.res.wrong_if++; 2128 2129 if (true_vifi >= 0 && mrt->mroute_do_assert && 2130 /* pimsm uses asserts, when switching from RPT to SPT, 2131 so that we cannot check that packet arrived on an oif. 2132 It is bad, but otherwise we would need to move pretty 2133 large chunk of pimd to kernel. Ough... --ANK 2134 */ 2135 (mrt->mroute_do_pim || 2136 cache->mfc_un.res.ttls[true_vifi] < 255) && 2137 time_after(jiffies, 2138 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2139 cache->mfc_un.res.last_assert = jiffies; 2140 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2141 } 2142 goto dont_forward; 2143 } 2144 2145 forward: 2146 mrt->vif6_table[vif].pkt_in++; 2147 mrt->vif6_table[vif].bytes_in += skb->len; 2148 2149 /* 2150 * Forward the frame 2151 */ 2152 if (ipv6_addr_any(&cache->mf6c_origin) && 2153 ipv6_addr_any(&cache->mf6c_mcastgrp)) { 2154 if (true_vifi >= 0 && 2155 true_vifi != cache->mf6c_parent && 2156 ipv6_hdr(skb)->hop_limit > 2157 cache->mfc_un.res.ttls[cache->mf6c_parent]) { 2158 /* It's an (*,*) entry and the packet is not coming from 2159 * the upstream: forward the packet to the upstream 2160 * only. 2161 */ 2162 psend = cache->mf6c_parent; 2163 goto last_forward; 2164 } 2165 goto dont_forward; 2166 } 2167 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2168 /* For (*,G) entry, don't forward to the incoming interface */ 2169 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) && 2170 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2171 if (psend != -1) { 2172 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2173 if (skb2) 2174 ip6mr_forward2(net, mrt, skb2, cache, psend); 2175 } 2176 psend = ct; 2177 } 2178 } 2179 last_forward: 2180 if (psend != -1) { 2181 ip6mr_forward2(net, mrt, skb, cache, psend); 2182 return; 2183 } 2184 2185 dont_forward: 2186 kfree_skb(skb); 2187 } 2188 2189 2190 /* 2191 * Multicast packets for forwarding arrive here 2192 */ 2193 2194 int ip6_mr_input(struct sk_buff *skb) 2195 { 2196 struct mfc6_cache *cache; 2197 struct net *net = dev_net(skb->dev); 2198 struct mr6_table *mrt; 2199 struct flowi6 fl6 = { 2200 .flowi6_iif = skb->dev->ifindex, 2201 .flowi6_mark = skb->mark, 2202 }; 2203 int err; 2204 2205 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2206 if (err < 0) { 2207 kfree_skb(skb); 2208 return err; 2209 } 2210 2211 read_lock(&mrt_lock); 2212 cache = ip6mr_cache_find(mrt, 2213 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2214 if (!cache) { 2215 int vif = ip6mr_find_vif(mrt, skb->dev); 2216 2217 if (vif >= 0) 2218 cache = ip6mr_cache_find_any(mrt, 2219 &ipv6_hdr(skb)->daddr, 2220 vif); 2221 } 2222 2223 /* 2224 * No usable cache entry 2225 */ 2226 if (!cache) { 2227 int vif; 2228 2229 vif = ip6mr_find_vif(mrt, skb->dev); 2230 if (vif >= 0) { 2231 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2232 read_unlock(&mrt_lock); 2233 2234 return err; 2235 } 2236 read_unlock(&mrt_lock); 2237 kfree_skb(skb); 2238 return -ENODEV; 2239 } 2240 2241 ip6_mr_forward(net, mrt, skb, cache); 2242 2243 read_unlock(&mrt_lock); 2244 2245 return 0; 2246 } 2247 2248 2249 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2250 struct mfc6_cache *c, struct rtmsg *rtm) 2251 { 2252 struct rta_mfc_stats mfcs; 2253 struct nlattr *mp_attr; 2254 struct rtnexthop *nhp; 2255 unsigned long lastuse; 2256 int ct; 2257 2258 /* If cache is unresolved, don't try to parse IIF and OIF */ 2259 if (c->mf6c_parent >= MAXMIFS) { 2260 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2261 return -ENOENT; 2262 } 2263 2264 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2265 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2266 return -EMSGSIZE; 2267 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2268 if (!mp_attr) 2269 return -EMSGSIZE; 2270 2271 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2272 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2273 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2274 if (!nhp) { 2275 nla_nest_cancel(skb, mp_attr); 2276 return -EMSGSIZE; 2277 } 2278 2279 nhp->rtnh_flags = 0; 2280 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2281 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2282 nhp->rtnh_len = sizeof(*nhp); 2283 } 2284 } 2285 2286 nla_nest_end(skb, mp_attr); 2287 2288 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2289 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2290 2291 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2292 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2293 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2294 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2295 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2296 RTA_PAD)) 2297 return -EMSGSIZE; 2298 2299 rtm->rtm_type = RTN_MULTICAST; 2300 return 1; 2301 } 2302 2303 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2304 u32 portid) 2305 { 2306 int err; 2307 struct mr6_table *mrt; 2308 struct mfc6_cache *cache; 2309 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2310 2311 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2312 if (!mrt) 2313 return -ENOENT; 2314 2315 read_lock(&mrt_lock); 2316 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2317 if (!cache && skb->dev) { 2318 int vif = ip6mr_find_vif(mrt, skb->dev); 2319 2320 if (vif >= 0) 2321 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2322 vif); 2323 } 2324 2325 if (!cache) { 2326 struct sk_buff *skb2; 2327 struct ipv6hdr *iph; 2328 struct net_device *dev; 2329 int vif; 2330 2331 dev = skb->dev; 2332 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2333 read_unlock(&mrt_lock); 2334 return -ENODEV; 2335 } 2336 2337 /* really correct? */ 2338 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2339 if (!skb2) { 2340 read_unlock(&mrt_lock); 2341 return -ENOMEM; 2342 } 2343 2344 NETLINK_CB(skb2).portid = portid; 2345 skb_reset_transport_header(skb2); 2346 2347 skb_put(skb2, sizeof(struct ipv6hdr)); 2348 skb_reset_network_header(skb2); 2349 2350 iph = ipv6_hdr(skb2); 2351 iph->version = 0; 2352 iph->priority = 0; 2353 iph->flow_lbl[0] = 0; 2354 iph->flow_lbl[1] = 0; 2355 iph->flow_lbl[2] = 0; 2356 iph->payload_len = 0; 2357 iph->nexthdr = IPPROTO_NONE; 2358 iph->hop_limit = 0; 2359 iph->saddr = rt->rt6i_src.addr; 2360 iph->daddr = rt->rt6i_dst.addr; 2361 2362 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2363 read_unlock(&mrt_lock); 2364 2365 return err; 2366 } 2367 2368 if (rtm->rtm_flags & RTM_F_NOTIFY) 2369 cache->mfc_flags |= MFC_NOTIFY; 2370 2371 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2372 read_unlock(&mrt_lock); 2373 return err; 2374 } 2375 2376 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2377 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2378 int flags) 2379 { 2380 struct nlmsghdr *nlh; 2381 struct rtmsg *rtm; 2382 int err; 2383 2384 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2385 if (!nlh) 2386 return -EMSGSIZE; 2387 2388 rtm = nlmsg_data(nlh); 2389 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2390 rtm->rtm_dst_len = 128; 2391 rtm->rtm_src_len = 128; 2392 rtm->rtm_tos = 0; 2393 rtm->rtm_table = mrt->id; 2394 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2395 goto nla_put_failure; 2396 rtm->rtm_type = RTN_MULTICAST; 2397 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2398 if (c->mfc_flags & MFC_STATIC) 2399 rtm->rtm_protocol = RTPROT_STATIC; 2400 else 2401 rtm->rtm_protocol = RTPROT_MROUTED; 2402 rtm->rtm_flags = 0; 2403 2404 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2405 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2406 goto nla_put_failure; 2407 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2408 /* do not break the dump if cache is unresolved */ 2409 if (err < 0 && err != -ENOENT) 2410 goto nla_put_failure; 2411 2412 nlmsg_end(skb, nlh); 2413 return 0; 2414 2415 nla_put_failure: 2416 nlmsg_cancel(skb, nlh); 2417 return -EMSGSIZE; 2418 } 2419 2420 static int mr6_msgsize(bool unresolved, int maxvif) 2421 { 2422 size_t len = 2423 NLMSG_ALIGN(sizeof(struct rtmsg)) 2424 + nla_total_size(4) /* RTA_TABLE */ 2425 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2426 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2427 ; 2428 2429 if (!unresolved) 2430 len = len 2431 + nla_total_size(4) /* RTA_IIF */ 2432 + nla_total_size(0) /* RTA_MULTIPATH */ 2433 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2434 /* RTA_MFC_STATS */ 2435 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2436 ; 2437 2438 return len; 2439 } 2440 2441 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2442 int cmd) 2443 { 2444 struct net *net = read_pnet(&mrt->net); 2445 struct sk_buff *skb; 2446 int err = -ENOBUFS; 2447 2448 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2449 GFP_ATOMIC); 2450 if (!skb) 2451 goto errout; 2452 2453 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2454 if (err < 0) 2455 goto errout; 2456 2457 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2458 return; 2459 2460 errout: 2461 kfree_skb(skb); 2462 if (err < 0) 2463 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2464 } 2465 2466 static size_t mrt6msg_netlink_msgsize(size_t payloadlen) 2467 { 2468 size_t len = 2469 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2470 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */ 2471 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */ 2472 /* IP6MRA_CREPORT_SRC_ADDR */ 2473 + nla_total_size(sizeof(struct in6_addr)) 2474 /* IP6MRA_CREPORT_DST_ADDR */ 2475 + nla_total_size(sizeof(struct in6_addr)) 2476 /* IP6MRA_CREPORT_PKT */ 2477 + nla_total_size(payloadlen) 2478 ; 2479 2480 return len; 2481 } 2482 2483 static void mrt6msg_netlink_event(struct mr6_table *mrt, struct sk_buff *pkt) 2484 { 2485 struct net *net = read_pnet(&mrt->net); 2486 struct nlmsghdr *nlh; 2487 struct rtgenmsg *rtgenm; 2488 struct mrt6msg *msg; 2489 struct sk_buff *skb; 2490 struct nlattr *nla; 2491 int payloadlen; 2492 2493 payloadlen = pkt->len - sizeof(struct mrt6msg); 2494 msg = (struct mrt6msg *)skb_transport_header(pkt); 2495 2496 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2497 if (!skb) 2498 goto errout; 2499 2500 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2501 sizeof(struct rtgenmsg), 0); 2502 if (!nlh) 2503 goto errout; 2504 rtgenm = nlmsg_data(nlh); 2505 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR; 2506 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) || 2507 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) || 2508 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR, 2509 &msg->im6_src) || 2510 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR, 2511 &msg->im6_dst)) 2512 goto nla_put_failure; 2513 2514 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen); 2515 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg), 2516 nla_data(nla), payloadlen)) 2517 goto nla_put_failure; 2518 2519 nlmsg_end(skb, nlh); 2520 2521 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC); 2522 return; 2523 2524 nla_put_failure: 2525 nlmsg_cancel(skb, nlh); 2526 errout: 2527 kfree_skb(skb); 2528 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS); 2529 } 2530 2531 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2532 { 2533 struct net *net = sock_net(skb->sk); 2534 struct mr6_table *mrt; 2535 struct mfc6_cache *mfc; 2536 unsigned int t = 0, s_t; 2537 unsigned int h = 0, s_h; 2538 unsigned int e = 0, s_e; 2539 2540 s_t = cb->args[0]; 2541 s_h = cb->args[1]; 2542 s_e = cb->args[2]; 2543 2544 read_lock(&mrt_lock); 2545 ip6mr_for_each_table(mrt, net) { 2546 if (t < s_t) 2547 goto next_table; 2548 if (t > s_t) 2549 s_h = 0; 2550 for (h = s_h; h < MFC6_LINES; h++) { 2551 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2552 if (e < s_e) 2553 goto next_entry; 2554 if (ip6mr_fill_mroute(mrt, skb, 2555 NETLINK_CB(cb->skb).portid, 2556 cb->nlh->nlmsg_seq, 2557 mfc, RTM_NEWROUTE, 2558 NLM_F_MULTI) < 0) 2559 goto done; 2560 next_entry: 2561 e++; 2562 } 2563 e = s_e = 0; 2564 } 2565 spin_lock_bh(&mfc_unres_lock); 2566 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2567 if (e < s_e) 2568 goto next_entry2; 2569 if (ip6mr_fill_mroute(mrt, skb, 2570 NETLINK_CB(cb->skb).portid, 2571 cb->nlh->nlmsg_seq, 2572 mfc, RTM_NEWROUTE, 2573 NLM_F_MULTI) < 0) { 2574 spin_unlock_bh(&mfc_unres_lock); 2575 goto done; 2576 } 2577 next_entry2: 2578 e++; 2579 } 2580 spin_unlock_bh(&mfc_unres_lock); 2581 e = s_e = 0; 2582 s_h = 0; 2583 next_table: 2584 t++; 2585 } 2586 done: 2587 read_unlock(&mrt_lock); 2588 2589 cb->args[2] = e; 2590 cb->args[1] = h; 2591 cb->args[0] = t; 2592 2593 return skb->len; 2594 } 2595