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 err = rtnl_register_module(THIS_MODULE, RTNL_FAMILY_IP6MR, RTM_GETROUTE, 1429 NULL, ip6mr_rtm_dumproute, 0); 1430 if (err == 0) 1431 return 0; 1432 1433 #ifdef CONFIG_IPV6_PIMSM_V2 1434 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1435 add_proto_fail: 1436 unregister_netdevice_notifier(&ip6_mr_notifier); 1437 #endif 1438 reg_notif_fail: 1439 unregister_pernet_subsys(&ip6mr_net_ops); 1440 reg_pernet_fail: 1441 kmem_cache_destroy(mrt_cachep); 1442 return err; 1443 } 1444 1445 void ip6_mr_cleanup(void) 1446 { 1447 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1448 #ifdef CONFIG_IPV6_PIMSM_V2 1449 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1450 #endif 1451 unregister_netdevice_notifier(&ip6_mr_notifier); 1452 unregister_pernet_subsys(&ip6mr_net_ops); 1453 kmem_cache_destroy(mrt_cachep); 1454 } 1455 1456 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1457 struct mf6cctl *mfc, int mrtsock, int parent) 1458 { 1459 bool found = false; 1460 int line; 1461 struct mfc6_cache *uc, *c; 1462 unsigned char ttls[MAXMIFS]; 1463 int i; 1464 1465 if (mfc->mf6cc_parent >= MAXMIFS) 1466 return -ENFILE; 1467 1468 memset(ttls, 255, MAXMIFS); 1469 for (i = 0; i < MAXMIFS; i++) { 1470 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1471 ttls[i] = 1; 1472 1473 } 1474 1475 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1476 1477 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1478 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1479 ipv6_addr_equal(&c->mf6c_mcastgrp, 1480 &mfc->mf6cc_mcastgrp.sin6_addr) && 1481 (parent == -1 || parent == mfc->mf6cc_parent)) { 1482 found = true; 1483 break; 1484 } 1485 } 1486 1487 if (found) { 1488 write_lock_bh(&mrt_lock); 1489 c->mf6c_parent = mfc->mf6cc_parent; 1490 ip6mr_update_thresholds(mrt, c, ttls); 1491 if (!mrtsock) 1492 c->mfc_flags |= MFC_STATIC; 1493 write_unlock_bh(&mrt_lock); 1494 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1495 return 0; 1496 } 1497 1498 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1499 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1500 return -EINVAL; 1501 1502 c = ip6mr_cache_alloc(); 1503 if (!c) 1504 return -ENOMEM; 1505 1506 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1507 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1508 c->mf6c_parent = mfc->mf6cc_parent; 1509 ip6mr_update_thresholds(mrt, c, ttls); 1510 if (!mrtsock) 1511 c->mfc_flags |= MFC_STATIC; 1512 1513 write_lock_bh(&mrt_lock); 1514 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1515 write_unlock_bh(&mrt_lock); 1516 1517 /* 1518 * Check to see if we resolved a queued list. If so we 1519 * need to send on the frames and tidy up. 1520 */ 1521 found = false; 1522 spin_lock_bh(&mfc_unres_lock); 1523 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1524 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1525 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1526 list_del(&uc->list); 1527 atomic_dec(&mrt->cache_resolve_queue_len); 1528 found = true; 1529 break; 1530 } 1531 } 1532 if (list_empty(&mrt->mfc6_unres_queue)) 1533 del_timer(&mrt->ipmr_expire_timer); 1534 spin_unlock_bh(&mfc_unres_lock); 1535 1536 if (found) { 1537 ip6mr_cache_resolve(net, mrt, uc, c); 1538 ip6mr_cache_free(uc); 1539 } 1540 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1541 return 0; 1542 } 1543 1544 /* 1545 * Close the multicast socket, and clear the vif tables etc 1546 */ 1547 1548 static void mroute_clean_tables(struct mr6_table *mrt, bool all) 1549 { 1550 int i; 1551 LIST_HEAD(list); 1552 struct mfc6_cache *c, *next; 1553 1554 /* 1555 * Shut down all active vif entries 1556 */ 1557 for (i = 0; i < mrt->maxvif; i++) { 1558 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) 1559 continue; 1560 mif6_delete(mrt, i, 0, &list); 1561 } 1562 unregister_netdevice_many(&list); 1563 1564 /* 1565 * Wipe the cache 1566 */ 1567 for (i = 0; i < MFC6_LINES; i++) { 1568 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1569 if (!all && (c->mfc_flags & MFC_STATIC)) 1570 continue; 1571 write_lock_bh(&mrt_lock); 1572 list_del(&c->list); 1573 write_unlock_bh(&mrt_lock); 1574 1575 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1576 ip6mr_cache_free(c); 1577 } 1578 } 1579 1580 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1581 spin_lock_bh(&mfc_unres_lock); 1582 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1583 list_del(&c->list); 1584 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1585 ip6mr_destroy_unres(mrt, c); 1586 } 1587 spin_unlock_bh(&mfc_unres_lock); 1588 } 1589 } 1590 1591 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1592 { 1593 int err = 0; 1594 struct net *net = sock_net(sk); 1595 1596 rtnl_lock(); 1597 write_lock_bh(&mrt_lock); 1598 if (likely(mrt->mroute6_sk == NULL)) { 1599 mrt->mroute6_sk = sk; 1600 net->ipv6.devconf_all->mc_forwarding++; 1601 } else { 1602 err = -EADDRINUSE; 1603 } 1604 write_unlock_bh(&mrt_lock); 1605 1606 if (!err) 1607 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1608 NETCONFA_MC_FORWARDING, 1609 NETCONFA_IFINDEX_ALL, 1610 net->ipv6.devconf_all); 1611 rtnl_unlock(); 1612 1613 return err; 1614 } 1615 1616 int ip6mr_sk_done(struct sock *sk) 1617 { 1618 int err = -EACCES; 1619 struct net *net = sock_net(sk); 1620 struct mr6_table *mrt; 1621 1622 if (sk->sk_type != SOCK_RAW || 1623 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1624 return err; 1625 1626 rtnl_lock(); 1627 ip6mr_for_each_table(mrt, net) { 1628 if (sk == mrt->mroute6_sk) { 1629 write_lock_bh(&mrt_lock); 1630 mrt->mroute6_sk = NULL; 1631 net->ipv6.devconf_all->mc_forwarding--; 1632 write_unlock_bh(&mrt_lock); 1633 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1634 NETCONFA_MC_FORWARDING, 1635 NETCONFA_IFINDEX_ALL, 1636 net->ipv6.devconf_all); 1637 1638 mroute_clean_tables(mrt, false); 1639 err = 0; 1640 break; 1641 } 1642 } 1643 rtnl_unlock(); 1644 1645 return err; 1646 } 1647 1648 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1649 { 1650 struct mr6_table *mrt; 1651 struct flowi6 fl6 = { 1652 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1653 .flowi6_oif = skb->dev->ifindex, 1654 .flowi6_mark = skb->mark, 1655 }; 1656 1657 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1658 return NULL; 1659 1660 return mrt->mroute6_sk; 1661 } 1662 1663 /* 1664 * Socket options and virtual interface manipulation. The whole 1665 * virtual interface system is a complete heap, but unfortunately 1666 * that's how BSD mrouted happens to think. Maybe one day with a proper 1667 * MOSPF/PIM router set up we can clean this up. 1668 */ 1669 1670 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1671 { 1672 int ret, parent = 0; 1673 struct mif6ctl vif; 1674 struct mf6cctl mfc; 1675 mifi_t mifi; 1676 struct net *net = sock_net(sk); 1677 struct mr6_table *mrt; 1678 1679 if (sk->sk_type != SOCK_RAW || 1680 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1681 return -EOPNOTSUPP; 1682 1683 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1684 if (!mrt) 1685 return -ENOENT; 1686 1687 if (optname != MRT6_INIT) { 1688 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1689 return -EACCES; 1690 } 1691 1692 switch (optname) { 1693 case MRT6_INIT: 1694 if (optlen < sizeof(int)) 1695 return -EINVAL; 1696 1697 return ip6mr_sk_init(mrt, sk); 1698 1699 case MRT6_DONE: 1700 return ip6mr_sk_done(sk); 1701 1702 case MRT6_ADD_MIF: 1703 if (optlen < sizeof(vif)) 1704 return -EINVAL; 1705 if (copy_from_user(&vif, optval, sizeof(vif))) 1706 return -EFAULT; 1707 if (vif.mif6c_mifi >= MAXMIFS) 1708 return -ENFILE; 1709 rtnl_lock(); 1710 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1711 rtnl_unlock(); 1712 return ret; 1713 1714 case MRT6_DEL_MIF: 1715 if (optlen < sizeof(mifi_t)) 1716 return -EINVAL; 1717 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1718 return -EFAULT; 1719 rtnl_lock(); 1720 ret = mif6_delete(mrt, mifi, 0, NULL); 1721 rtnl_unlock(); 1722 return ret; 1723 1724 /* 1725 * Manipulate the forwarding caches. These live 1726 * in a sort of kernel/user symbiosis. 1727 */ 1728 case MRT6_ADD_MFC: 1729 case MRT6_DEL_MFC: 1730 parent = -1; 1731 /* fall through */ 1732 case MRT6_ADD_MFC_PROXY: 1733 case MRT6_DEL_MFC_PROXY: 1734 if (optlen < sizeof(mfc)) 1735 return -EINVAL; 1736 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1737 return -EFAULT; 1738 if (parent == 0) 1739 parent = mfc.mf6cc_parent; 1740 rtnl_lock(); 1741 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1742 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1743 else 1744 ret = ip6mr_mfc_add(net, mrt, &mfc, 1745 sk == mrt->mroute6_sk, parent); 1746 rtnl_unlock(); 1747 return ret; 1748 1749 /* 1750 * Control PIM assert (to activate pim will activate assert) 1751 */ 1752 case MRT6_ASSERT: 1753 { 1754 int v; 1755 1756 if (optlen != sizeof(v)) 1757 return -EINVAL; 1758 if (get_user(v, (int __user *)optval)) 1759 return -EFAULT; 1760 mrt->mroute_do_assert = v; 1761 return 0; 1762 } 1763 1764 #ifdef CONFIG_IPV6_PIMSM_V2 1765 case MRT6_PIM: 1766 { 1767 int v; 1768 1769 if (optlen != sizeof(v)) 1770 return -EINVAL; 1771 if (get_user(v, (int __user *)optval)) 1772 return -EFAULT; 1773 v = !!v; 1774 rtnl_lock(); 1775 ret = 0; 1776 if (v != mrt->mroute_do_pim) { 1777 mrt->mroute_do_pim = v; 1778 mrt->mroute_do_assert = v; 1779 } 1780 rtnl_unlock(); 1781 return ret; 1782 } 1783 1784 #endif 1785 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1786 case MRT6_TABLE: 1787 { 1788 u32 v; 1789 1790 if (optlen != sizeof(u32)) 1791 return -EINVAL; 1792 if (get_user(v, (u32 __user *)optval)) 1793 return -EFAULT; 1794 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1795 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1796 return -EINVAL; 1797 if (sk == mrt->mroute6_sk) 1798 return -EBUSY; 1799 1800 rtnl_lock(); 1801 ret = 0; 1802 if (!ip6mr_new_table(net, v)) 1803 ret = -ENOMEM; 1804 raw6_sk(sk)->ip6mr_table = v; 1805 rtnl_unlock(); 1806 return ret; 1807 } 1808 #endif 1809 /* 1810 * Spurious command, or MRT6_VERSION which you cannot 1811 * set. 1812 */ 1813 default: 1814 return -ENOPROTOOPT; 1815 } 1816 } 1817 1818 /* 1819 * Getsock opt support for the multicast routing system. 1820 */ 1821 1822 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1823 int __user *optlen) 1824 { 1825 int olr; 1826 int val; 1827 struct net *net = sock_net(sk); 1828 struct mr6_table *mrt; 1829 1830 if (sk->sk_type != SOCK_RAW || 1831 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1832 return -EOPNOTSUPP; 1833 1834 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1835 if (!mrt) 1836 return -ENOENT; 1837 1838 switch (optname) { 1839 case MRT6_VERSION: 1840 val = 0x0305; 1841 break; 1842 #ifdef CONFIG_IPV6_PIMSM_V2 1843 case MRT6_PIM: 1844 val = mrt->mroute_do_pim; 1845 break; 1846 #endif 1847 case MRT6_ASSERT: 1848 val = mrt->mroute_do_assert; 1849 break; 1850 default: 1851 return -ENOPROTOOPT; 1852 } 1853 1854 if (get_user(olr, optlen)) 1855 return -EFAULT; 1856 1857 olr = min_t(int, olr, sizeof(int)); 1858 if (olr < 0) 1859 return -EINVAL; 1860 1861 if (put_user(olr, optlen)) 1862 return -EFAULT; 1863 if (copy_to_user(optval, &val, olr)) 1864 return -EFAULT; 1865 return 0; 1866 } 1867 1868 /* 1869 * The IP multicast ioctl support routines. 1870 */ 1871 1872 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1873 { 1874 struct sioc_sg_req6 sr; 1875 struct sioc_mif_req6 vr; 1876 struct mif_device *vif; 1877 struct mfc6_cache *c; 1878 struct net *net = sock_net(sk); 1879 struct mr6_table *mrt; 1880 1881 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1882 if (!mrt) 1883 return -ENOENT; 1884 1885 switch (cmd) { 1886 case SIOCGETMIFCNT_IN6: 1887 if (copy_from_user(&vr, arg, sizeof(vr))) 1888 return -EFAULT; 1889 if (vr.mifi >= mrt->maxvif) 1890 return -EINVAL; 1891 read_lock(&mrt_lock); 1892 vif = &mrt->vif6_table[vr.mifi]; 1893 if (MIF_EXISTS(mrt, vr.mifi)) { 1894 vr.icount = vif->pkt_in; 1895 vr.ocount = vif->pkt_out; 1896 vr.ibytes = vif->bytes_in; 1897 vr.obytes = vif->bytes_out; 1898 read_unlock(&mrt_lock); 1899 1900 if (copy_to_user(arg, &vr, sizeof(vr))) 1901 return -EFAULT; 1902 return 0; 1903 } 1904 read_unlock(&mrt_lock); 1905 return -EADDRNOTAVAIL; 1906 case SIOCGETSGCNT_IN6: 1907 if (copy_from_user(&sr, arg, sizeof(sr))) 1908 return -EFAULT; 1909 1910 read_lock(&mrt_lock); 1911 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1912 if (c) { 1913 sr.pktcnt = c->mfc_un.res.pkt; 1914 sr.bytecnt = c->mfc_un.res.bytes; 1915 sr.wrong_if = c->mfc_un.res.wrong_if; 1916 read_unlock(&mrt_lock); 1917 1918 if (copy_to_user(arg, &sr, sizeof(sr))) 1919 return -EFAULT; 1920 return 0; 1921 } 1922 read_unlock(&mrt_lock); 1923 return -EADDRNOTAVAIL; 1924 default: 1925 return -ENOIOCTLCMD; 1926 } 1927 } 1928 1929 #ifdef CONFIG_COMPAT 1930 struct compat_sioc_sg_req6 { 1931 struct sockaddr_in6 src; 1932 struct sockaddr_in6 grp; 1933 compat_ulong_t pktcnt; 1934 compat_ulong_t bytecnt; 1935 compat_ulong_t wrong_if; 1936 }; 1937 1938 struct compat_sioc_mif_req6 { 1939 mifi_t mifi; 1940 compat_ulong_t icount; 1941 compat_ulong_t ocount; 1942 compat_ulong_t ibytes; 1943 compat_ulong_t obytes; 1944 }; 1945 1946 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1947 { 1948 struct compat_sioc_sg_req6 sr; 1949 struct compat_sioc_mif_req6 vr; 1950 struct mif_device *vif; 1951 struct mfc6_cache *c; 1952 struct net *net = sock_net(sk); 1953 struct mr6_table *mrt; 1954 1955 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1956 if (!mrt) 1957 return -ENOENT; 1958 1959 switch (cmd) { 1960 case SIOCGETMIFCNT_IN6: 1961 if (copy_from_user(&vr, arg, sizeof(vr))) 1962 return -EFAULT; 1963 if (vr.mifi >= mrt->maxvif) 1964 return -EINVAL; 1965 read_lock(&mrt_lock); 1966 vif = &mrt->vif6_table[vr.mifi]; 1967 if (MIF_EXISTS(mrt, vr.mifi)) { 1968 vr.icount = vif->pkt_in; 1969 vr.ocount = vif->pkt_out; 1970 vr.ibytes = vif->bytes_in; 1971 vr.obytes = vif->bytes_out; 1972 read_unlock(&mrt_lock); 1973 1974 if (copy_to_user(arg, &vr, sizeof(vr))) 1975 return -EFAULT; 1976 return 0; 1977 } 1978 read_unlock(&mrt_lock); 1979 return -EADDRNOTAVAIL; 1980 case SIOCGETSGCNT_IN6: 1981 if (copy_from_user(&sr, arg, sizeof(sr))) 1982 return -EFAULT; 1983 1984 read_lock(&mrt_lock); 1985 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1986 if (c) { 1987 sr.pktcnt = c->mfc_un.res.pkt; 1988 sr.bytecnt = c->mfc_un.res.bytes; 1989 sr.wrong_if = c->mfc_un.res.wrong_if; 1990 read_unlock(&mrt_lock); 1991 1992 if (copy_to_user(arg, &sr, sizeof(sr))) 1993 return -EFAULT; 1994 return 0; 1995 } 1996 read_unlock(&mrt_lock); 1997 return -EADDRNOTAVAIL; 1998 default: 1999 return -ENOIOCTLCMD; 2000 } 2001 } 2002 #endif 2003 2004 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 2005 { 2006 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 2007 IPSTATS_MIB_OUTFORWDATAGRAMS); 2008 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)), 2009 IPSTATS_MIB_OUTOCTETS, skb->len); 2010 return dst_output(net, sk, skb); 2011 } 2012 2013 /* 2014 * Processing handlers for ip6mr_forward 2015 */ 2016 2017 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 2018 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 2019 { 2020 struct ipv6hdr *ipv6h; 2021 struct mif_device *vif = &mrt->vif6_table[vifi]; 2022 struct net_device *dev; 2023 struct dst_entry *dst; 2024 struct flowi6 fl6; 2025 2026 if (!vif->dev) 2027 goto out_free; 2028 2029 #ifdef CONFIG_IPV6_PIMSM_V2 2030 if (vif->flags & MIFF_REGISTER) { 2031 vif->pkt_out++; 2032 vif->bytes_out += skb->len; 2033 vif->dev->stats.tx_bytes += skb->len; 2034 vif->dev->stats.tx_packets++; 2035 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2036 goto out_free; 2037 } 2038 #endif 2039 2040 ipv6h = ipv6_hdr(skb); 2041 2042 fl6 = (struct flowi6) { 2043 .flowi6_oif = vif->link, 2044 .daddr = ipv6h->daddr, 2045 }; 2046 2047 dst = ip6_route_output(net, NULL, &fl6); 2048 if (dst->error) { 2049 dst_release(dst); 2050 goto out_free; 2051 } 2052 2053 skb_dst_drop(skb); 2054 skb_dst_set(skb, dst); 2055 2056 /* 2057 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2058 * not only before forwarding, but after forwarding on all output 2059 * interfaces. It is clear, if mrouter runs a multicasting 2060 * program, it should receive packets not depending to what interface 2061 * program is joined. 2062 * If we will not make it, the program will have to join on all 2063 * interfaces. On the other hand, multihoming host (or router, but 2064 * not mrouter) cannot join to more than one interface - it will 2065 * result in receiving multiple packets. 2066 */ 2067 dev = vif->dev; 2068 skb->dev = dev; 2069 vif->pkt_out++; 2070 vif->bytes_out += skb->len; 2071 2072 /* We are about to write */ 2073 /* XXX: extension headers? */ 2074 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2075 goto out_free; 2076 2077 ipv6h = ipv6_hdr(skb); 2078 ipv6h->hop_limit--; 2079 2080 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2081 2082 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2083 net, NULL, skb, skb->dev, dev, 2084 ip6mr_forward2_finish); 2085 2086 out_free: 2087 kfree_skb(skb); 2088 return 0; 2089 } 2090 2091 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2092 { 2093 int ct; 2094 2095 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2096 if (mrt->vif6_table[ct].dev == dev) 2097 break; 2098 } 2099 return ct; 2100 } 2101 2102 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2103 struct sk_buff *skb, struct mfc6_cache *cache) 2104 { 2105 int psend = -1; 2106 int vif, ct; 2107 int true_vifi = ip6mr_find_vif(mrt, skb->dev); 2108 2109 vif = cache->mf6c_parent; 2110 cache->mfc_un.res.pkt++; 2111 cache->mfc_un.res.bytes += skb->len; 2112 cache->mfc_un.res.lastuse = jiffies; 2113 2114 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) { 2115 struct mfc6_cache *cache_proxy; 2116 2117 /* For an (*,G) entry, we only check that the incoming 2118 * interface is part of the static tree. 2119 */ 2120 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif); 2121 if (cache_proxy && 2122 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 2123 goto forward; 2124 } 2125 2126 /* 2127 * Wrong interface: drop packet and (maybe) send PIM assert. 2128 */ 2129 if (mrt->vif6_table[vif].dev != skb->dev) { 2130 cache->mfc_un.res.wrong_if++; 2131 2132 if (true_vifi >= 0 && mrt->mroute_do_assert && 2133 /* pimsm uses asserts, when switching from RPT to SPT, 2134 so that we cannot check that packet arrived on an oif. 2135 It is bad, but otherwise we would need to move pretty 2136 large chunk of pimd to kernel. Ough... --ANK 2137 */ 2138 (mrt->mroute_do_pim || 2139 cache->mfc_un.res.ttls[true_vifi] < 255) && 2140 time_after(jiffies, 2141 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2142 cache->mfc_un.res.last_assert = jiffies; 2143 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2144 } 2145 goto dont_forward; 2146 } 2147 2148 forward: 2149 mrt->vif6_table[vif].pkt_in++; 2150 mrt->vif6_table[vif].bytes_in += skb->len; 2151 2152 /* 2153 * Forward the frame 2154 */ 2155 if (ipv6_addr_any(&cache->mf6c_origin) && 2156 ipv6_addr_any(&cache->mf6c_mcastgrp)) { 2157 if (true_vifi >= 0 && 2158 true_vifi != cache->mf6c_parent && 2159 ipv6_hdr(skb)->hop_limit > 2160 cache->mfc_un.res.ttls[cache->mf6c_parent]) { 2161 /* It's an (*,*) entry and the packet is not coming from 2162 * the upstream: forward the packet to the upstream 2163 * only. 2164 */ 2165 psend = cache->mf6c_parent; 2166 goto last_forward; 2167 } 2168 goto dont_forward; 2169 } 2170 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2171 /* For (*,G) entry, don't forward to the incoming interface */ 2172 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) && 2173 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2174 if (psend != -1) { 2175 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2176 if (skb2) 2177 ip6mr_forward2(net, mrt, skb2, cache, psend); 2178 } 2179 psend = ct; 2180 } 2181 } 2182 last_forward: 2183 if (psend != -1) { 2184 ip6mr_forward2(net, mrt, skb, cache, psend); 2185 return; 2186 } 2187 2188 dont_forward: 2189 kfree_skb(skb); 2190 } 2191 2192 2193 /* 2194 * Multicast packets for forwarding arrive here 2195 */ 2196 2197 int ip6_mr_input(struct sk_buff *skb) 2198 { 2199 struct mfc6_cache *cache; 2200 struct net *net = dev_net(skb->dev); 2201 struct mr6_table *mrt; 2202 struct flowi6 fl6 = { 2203 .flowi6_iif = skb->dev->ifindex, 2204 .flowi6_mark = skb->mark, 2205 }; 2206 int err; 2207 2208 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2209 if (err < 0) { 2210 kfree_skb(skb); 2211 return err; 2212 } 2213 2214 read_lock(&mrt_lock); 2215 cache = ip6mr_cache_find(mrt, 2216 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2217 if (!cache) { 2218 int vif = ip6mr_find_vif(mrt, skb->dev); 2219 2220 if (vif >= 0) 2221 cache = ip6mr_cache_find_any(mrt, 2222 &ipv6_hdr(skb)->daddr, 2223 vif); 2224 } 2225 2226 /* 2227 * No usable cache entry 2228 */ 2229 if (!cache) { 2230 int vif; 2231 2232 vif = ip6mr_find_vif(mrt, skb->dev); 2233 if (vif >= 0) { 2234 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2235 read_unlock(&mrt_lock); 2236 2237 return err; 2238 } 2239 read_unlock(&mrt_lock); 2240 kfree_skb(skb); 2241 return -ENODEV; 2242 } 2243 2244 ip6_mr_forward(net, mrt, skb, cache); 2245 2246 read_unlock(&mrt_lock); 2247 2248 return 0; 2249 } 2250 2251 2252 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2253 struct mfc6_cache *c, struct rtmsg *rtm) 2254 { 2255 struct rta_mfc_stats mfcs; 2256 struct nlattr *mp_attr; 2257 struct rtnexthop *nhp; 2258 unsigned long lastuse; 2259 int ct; 2260 2261 /* If cache is unresolved, don't try to parse IIF and OIF */ 2262 if (c->mf6c_parent >= MAXMIFS) { 2263 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2264 return -ENOENT; 2265 } 2266 2267 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2268 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2269 return -EMSGSIZE; 2270 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2271 if (!mp_attr) 2272 return -EMSGSIZE; 2273 2274 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2275 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2276 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2277 if (!nhp) { 2278 nla_nest_cancel(skb, mp_attr); 2279 return -EMSGSIZE; 2280 } 2281 2282 nhp->rtnh_flags = 0; 2283 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2284 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2285 nhp->rtnh_len = sizeof(*nhp); 2286 } 2287 } 2288 2289 nla_nest_end(skb, mp_attr); 2290 2291 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2292 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2293 2294 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2295 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2296 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2297 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2298 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2299 RTA_PAD)) 2300 return -EMSGSIZE; 2301 2302 rtm->rtm_type = RTN_MULTICAST; 2303 return 1; 2304 } 2305 2306 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2307 u32 portid) 2308 { 2309 int err; 2310 struct mr6_table *mrt; 2311 struct mfc6_cache *cache; 2312 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2313 2314 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2315 if (!mrt) 2316 return -ENOENT; 2317 2318 read_lock(&mrt_lock); 2319 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2320 if (!cache && skb->dev) { 2321 int vif = ip6mr_find_vif(mrt, skb->dev); 2322 2323 if (vif >= 0) 2324 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2325 vif); 2326 } 2327 2328 if (!cache) { 2329 struct sk_buff *skb2; 2330 struct ipv6hdr *iph; 2331 struct net_device *dev; 2332 int vif; 2333 2334 dev = skb->dev; 2335 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2336 read_unlock(&mrt_lock); 2337 return -ENODEV; 2338 } 2339 2340 /* really correct? */ 2341 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2342 if (!skb2) { 2343 read_unlock(&mrt_lock); 2344 return -ENOMEM; 2345 } 2346 2347 NETLINK_CB(skb2).portid = portid; 2348 skb_reset_transport_header(skb2); 2349 2350 skb_put(skb2, sizeof(struct ipv6hdr)); 2351 skb_reset_network_header(skb2); 2352 2353 iph = ipv6_hdr(skb2); 2354 iph->version = 0; 2355 iph->priority = 0; 2356 iph->flow_lbl[0] = 0; 2357 iph->flow_lbl[1] = 0; 2358 iph->flow_lbl[2] = 0; 2359 iph->payload_len = 0; 2360 iph->nexthdr = IPPROTO_NONE; 2361 iph->hop_limit = 0; 2362 iph->saddr = rt->rt6i_src.addr; 2363 iph->daddr = rt->rt6i_dst.addr; 2364 2365 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2366 read_unlock(&mrt_lock); 2367 2368 return err; 2369 } 2370 2371 if (rtm->rtm_flags & RTM_F_NOTIFY) 2372 cache->mfc_flags |= MFC_NOTIFY; 2373 2374 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2375 read_unlock(&mrt_lock); 2376 return err; 2377 } 2378 2379 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2380 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2381 int flags) 2382 { 2383 struct nlmsghdr *nlh; 2384 struct rtmsg *rtm; 2385 int err; 2386 2387 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2388 if (!nlh) 2389 return -EMSGSIZE; 2390 2391 rtm = nlmsg_data(nlh); 2392 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2393 rtm->rtm_dst_len = 128; 2394 rtm->rtm_src_len = 128; 2395 rtm->rtm_tos = 0; 2396 rtm->rtm_table = mrt->id; 2397 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2398 goto nla_put_failure; 2399 rtm->rtm_type = RTN_MULTICAST; 2400 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2401 if (c->mfc_flags & MFC_STATIC) 2402 rtm->rtm_protocol = RTPROT_STATIC; 2403 else 2404 rtm->rtm_protocol = RTPROT_MROUTED; 2405 rtm->rtm_flags = 0; 2406 2407 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2408 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2409 goto nla_put_failure; 2410 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2411 /* do not break the dump if cache is unresolved */ 2412 if (err < 0 && err != -ENOENT) 2413 goto nla_put_failure; 2414 2415 nlmsg_end(skb, nlh); 2416 return 0; 2417 2418 nla_put_failure: 2419 nlmsg_cancel(skb, nlh); 2420 return -EMSGSIZE; 2421 } 2422 2423 static int mr6_msgsize(bool unresolved, int maxvif) 2424 { 2425 size_t len = 2426 NLMSG_ALIGN(sizeof(struct rtmsg)) 2427 + nla_total_size(4) /* RTA_TABLE */ 2428 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2429 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2430 ; 2431 2432 if (!unresolved) 2433 len = len 2434 + nla_total_size(4) /* RTA_IIF */ 2435 + nla_total_size(0) /* RTA_MULTIPATH */ 2436 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2437 /* RTA_MFC_STATS */ 2438 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2439 ; 2440 2441 return len; 2442 } 2443 2444 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2445 int cmd) 2446 { 2447 struct net *net = read_pnet(&mrt->net); 2448 struct sk_buff *skb; 2449 int err = -ENOBUFS; 2450 2451 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2452 GFP_ATOMIC); 2453 if (!skb) 2454 goto errout; 2455 2456 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2457 if (err < 0) 2458 goto errout; 2459 2460 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2461 return; 2462 2463 errout: 2464 kfree_skb(skb); 2465 if (err < 0) 2466 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2467 } 2468 2469 static size_t mrt6msg_netlink_msgsize(size_t payloadlen) 2470 { 2471 size_t len = 2472 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2473 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */ 2474 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */ 2475 /* IP6MRA_CREPORT_SRC_ADDR */ 2476 + nla_total_size(sizeof(struct in6_addr)) 2477 /* IP6MRA_CREPORT_DST_ADDR */ 2478 + nla_total_size(sizeof(struct in6_addr)) 2479 /* IP6MRA_CREPORT_PKT */ 2480 + nla_total_size(payloadlen) 2481 ; 2482 2483 return len; 2484 } 2485 2486 static void mrt6msg_netlink_event(struct mr6_table *mrt, struct sk_buff *pkt) 2487 { 2488 struct net *net = read_pnet(&mrt->net); 2489 struct nlmsghdr *nlh; 2490 struct rtgenmsg *rtgenm; 2491 struct mrt6msg *msg; 2492 struct sk_buff *skb; 2493 struct nlattr *nla; 2494 int payloadlen; 2495 2496 payloadlen = pkt->len - sizeof(struct mrt6msg); 2497 msg = (struct mrt6msg *)skb_transport_header(pkt); 2498 2499 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2500 if (!skb) 2501 goto errout; 2502 2503 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2504 sizeof(struct rtgenmsg), 0); 2505 if (!nlh) 2506 goto errout; 2507 rtgenm = nlmsg_data(nlh); 2508 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR; 2509 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) || 2510 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) || 2511 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR, 2512 &msg->im6_src) || 2513 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR, 2514 &msg->im6_dst)) 2515 goto nla_put_failure; 2516 2517 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen); 2518 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg), 2519 nla_data(nla), payloadlen)) 2520 goto nla_put_failure; 2521 2522 nlmsg_end(skb, nlh); 2523 2524 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC); 2525 return; 2526 2527 nla_put_failure: 2528 nlmsg_cancel(skb, nlh); 2529 errout: 2530 kfree_skb(skb); 2531 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS); 2532 } 2533 2534 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2535 { 2536 struct net *net = sock_net(skb->sk); 2537 struct mr6_table *mrt; 2538 struct mfc6_cache *mfc; 2539 unsigned int t = 0, s_t; 2540 unsigned int h = 0, s_h; 2541 unsigned int e = 0, s_e; 2542 2543 s_t = cb->args[0]; 2544 s_h = cb->args[1]; 2545 s_e = cb->args[2]; 2546 2547 read_lock(&mrt_lock); 2548 ip6mr_for_each_table(mrt, net) { 2549 if (t < s_t) 2550 goto next_table; 2551 if (t > s_t) 2552 s_h = 0; 2553 for (h = s_h; h < MFC6_LINES; h++) { 2554 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2555 if (e < s_e) 2556 goto next_entry; 2557 if (ip6mr_fill_mroute(mrt, skb, 2558 NETLINK_CB(cb->skb).portid, 2559 cb->nlh->nlmsg_seq, 2560 mfc, RTM_NEWROUTE, 2561 NLM_F_MULTI) < 0) 2562 goto done; 2563 next_entry: 2564 e++; 2565 } 2566 e = s_e = 0; 2567 } 2568 spin_lock_bh(&mfc_unres_lock); 2569 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2570 if (e < s_e) 2571 goto next_entry2; 2572 if (ip6mr_fill_mroute(mrt, skb, 2573 NETLINK_CB(cb->skb).portid, 2574 cb->nlh->nlmsg_seq, 2575 mfc, RTM_NEWROUTE, 2576 NLM_F_MULTI) < 0) { 2577 spin_unlock_bh(&mfc_unres_lock); 2578 goto done; 2579 } 2580 next_entry2: 2581 e++; 2582 } 2583 spin_unlock_bh(&mfc_unres_lock); 2584 e = s_e = 0; 2585 s_h = 0; 2586 next_table: 2587 t++; 2588 } 2589 done: 2590 read_unlock(&mrt_lock); 2591 2592 cb->args[2] = e; 2593 cb->args[1] = h; 2594 cb->args[0] = t; 2595 2596 return skb->len; 2597 } 2598