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 int ip6mr_rtm_dumproute(struct sk_buff *skb, 120 struct netlink_callback *cb); 121 static void mroute_clean_tables(struct mr6_table *mrt, bool all); 122 static void ipmr_expire_process(unsigned long arg); 123 124 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 125 #define ip6mr_for_each_table(mrt, net) \ 126 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list) 127 128 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 129 { 130 struct mr6_table *mrt; 131 132 ip6mr_for_each_table(mrt, net) { 133 if (mrt->id == id) 134 return mrt; 135 } 136 return NULL; 137 } 138 139 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 140 struct mr6_table **mrt) 141 { 142 int err; 143 struct ip6mr_result res; 144 struct fib_lookup_arg arg = { 145 .result = &res, 146 .flags = FIB_LOOKUP_NOREF, 147 }; 148 149 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, 150 flowi6_to_flowi(flp6), 0, &arg); 151 if (err < 0) 152 return err; 153 *mrt = res.mrt; 154 return 0; 155 } 156 157 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp, 158 int flags, struct fib_lookup_arg *arg) 159 { 160 struct ip6mr_result *res = arg->result; 161 struct mr6_table *mrt; 162 163 switch (rule->action) { 164 case FR_ACT_TO_TBL: 165 break; 166 case FR_ACT_UNREACHABLE: 167 return -ENETUNREACH; 168 case FR_ACT_PROHIBIT: 169 return -EACCES; 170 case FR_ACT_BLACKHOLE: 171 default: 172 return -EINVAL; 173 } 174 175 mrt = ip6mr_get_table(rule->fr_net, rule->table); 176 if (!mrt) 177 return -EAGAIN; 178 res->mrt = mrt; 179 return 0; 180 } 181 182 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags) 183 { 184 return 1; 185 } 186 187 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = { 188 FRA_GENERIC_POLICY, 189 }; 190 191 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 192 struct fib_rule_hdr *frh, struct nlattr **tb) 193 { 194 return 0; 195 } 196 197 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 198 struct nlattr **tb) 199 { 200 return 1; 201 } 202 203 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 204 struct fib_rule_hdr *frh) 205 { 206 frh->dst_len = 0; 207 frh->src_len = 0; 208 frh->tos = 0; 209 return 0; 210 } 211 212 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = { 213 .family = RTNL_FAMILY_IP6MR, 214 .rule_size = sizeof(struct ip6mr_rule), 215 .addr_size = sizeof(struct in6_addr), 216 .action = ip6mr_rule_action, 217 .match = ip6mr_rule_match, 218 .configure = ip6mr_rule_configure, 219 .compare = ip6mr_rule_compare, 220 .fill = ip6mr_rule_fill, 221 .nlgroup = RTNLGRP_IPV6_RULE, 222 .policy = ip6mr_rule_policy, 223 .owner = THIS_MODULE, 224 }; 225 226 static int __net_init ip6mr_rules_init(struct net *net) 227 { 228 struct fib_rules_ops *ops; 229 struct mr6_table *mrt; 230 int err; 231 232 ops = fib_rules_register(&ip6mr_rules_ops_template, net); 233 if (IS_ERR(ops)) 234 return PTR_ERR(ops); 235 236 INIT_LIST_HEAD(&net->ipv6.mr6_tables); 237 238 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT); 239 if (!mrt) { 240 err = -ENOMEM; 241 goto err1; 242 } 243 244 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0); 245 if (err < 0) 246 goto err2; 247 248 net->ipv6.mr6_rules_ops = ops; 249 return 0; 250 251 err2: 252 ip6mr_free_table(mrt); 253 err1: 254 fib_rules_unregister(ops); 255 return err; 256 } 257 258 static void __net_exit ip6mr_rules_exit(struct net *net) 259 { 260 struct mr6_table *mrt, *next; 261 262 rtnl_lock(); 263 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) { 264 list_del(&mrt->list); 265 ip6mr_free_table(mrt); 266 } 267 fib_rules_unregister(net->ipv6.mr6_rules_ops); 268 rtnl_unlock(); 269 } 270 #else 271 #define ip6mr_for_each_table(mrt, net) \ 272 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL) 273 274 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 275 { 276 return net->ipv6.mrt6; 277 } 278 279 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 280 struct mr6_table **mrt) 281 { 282 *mrt = net->ipv6.mrt6; 283 return 0; 284 } 285 286 static int __net_init ip6mr_rules_init(struct net *net) 287 { 288 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT); 289 return net->ipv6.mrt6 ? 0 : -ENOMEM; 290 } 291 292 static void __net_exit ip6mr_rules_exit(struct net *net) 293 { 294 rtnl_lock(); 295 ip6mr_free_table(net->ipv6.mrt6); 296 net->ipv6.mrt6 = NULL; 297 rtnl_unlock(); 298 } 299 #endif 300 301 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id) 302 { 303 struct mr6_table *mrt; 304 unsigned int i; 305 306 mrt = ip6mr_get_table(net, id); 307 if (mrt) 308 return mrt; 309 310 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 311 if (!mrt) 312 return NULL; 313 mrt->id = id; 314 write_pnet(&mrt->net, net); 315 316 /* Forwarding cache */ 317 for (i = 0; i < MFC6_LINES; i++) 318 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]); 319 320 INIT_LIST_HEAD(&mrt->mfc6_unres_queue); 321 322 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 323 (unsigned long)mrt); 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(unsigned long arg) 891 { 892 struct mr6_table *mrt = (struct mr6_table *)arg; 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. We could use netlink for this but pim6sd 1129 * expects the following bizarre scheme. 1130 * 1131 * Called under mrt_lock. 1132 */ 1133 1134 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 1135 mifi_t mifi, int assert) 1136 { 1137 struct sk_buff *skb; 1138 struct mrt6msg *msg; 1139 int ret; 1140 1141 #ifdef CONFIG_IPV6_PIMSM_V2 1142 if (assert == MRT6MSG_WHOLEPKT) 1143 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1144 +sizeof(*msg)); 1145 else 1146 #endif 1147 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1148 1149 if (!skb) 1150 return -ENOBUFS; 1151 1152 /* I suppose that internal messages 1153 * do not require checksums */ 1154 1155 skb->ip_summed = CHECKSUM_UNNECESSARY; 1156 1157 #ifdef CONFIG_IPV6_PIMSM_V2 1158 if (assert == MRT6MSG_WHOLEPKT) { 1159 /* Ugly, but we have no choice with this interface. 1160 Duplicate old header, fix length etc. 1161 And all this only to mangle msg->im6_msgtype and 1162 to set msg->im6_mbz to "mbz" :-) 1163 */ 1164 skb_push(skb, -skb_network_offset(pkt)); 1165 1166 skb_push(skb, sizeof(*msg)); 1167 skb_reset_transport_header(skb); 1168 msg = (struct mrt6msg *)skb_transport_header(skb); 1169 msg->im6_mbz = 0; 1170 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1171 msg->im6_mif = mrt->mroute_reg_vif_num; 1172 msg->im6_pad = 0; 1173 msg->im6_src = ipv6_hdr(pkt)->saddr; 1174 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1175 1176 skb->ip_summed = CHECKSUM_UNNECESSARY; 1177 } else 1178 #endif 1179 { 1180 /* 1181 * Copy the IP header 1182 */ 1183 1184 skb_put(skb, sizeof(struct ipv6hdr)); 1185 skb_reset_network_header(skb); 1186 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1187 1188 /* 1189 * Add our header 1190 */ 1191 skb_put(skb, sizeof(*msg)); 1192 skb_reset_transport_header(skb); 1193 msg = (struct mrt6msg *)skb_transport_header(skb); 1194 1195 msg->im6_mbz = 0; 1196 msg->im6_msgtype = assert; 1197 msg->im6_mif = mifi; 1198 msg->im6_pad = 0; 1199 msg->im6_src = ipv6_hdr(pkt)->saddr; 1200 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1201 1202 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1203 skb->ip_summed = CHECKSUM_UNNECESSARY; 1204 } 1205 1206 if (!mrt->mroute6_sk) { 1207 kfree_skb(skb); 1208 return -EINVAL; 1209 } 1210 1211 /* 1212 * Deliver to user space multicast routing algorithms 1213 */ 1214 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb); 1215 if (ret < 0) { 1216 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1217 kfree_skb(skb); 1218 } 1219 1220 return ret; 1221 } 1222 1223 /* 1224 * Queue a packet for resolution. It gets locked cache entry! 1225 */ 1226 1227 static int 1228 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb) 1229 { 1230 bool found = false; 1231 int err; 1232 struct mfc6_cache *c; 1233 1234 spin_lock_bh(&mfc_unres_lock); 1235 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) { 1236 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1237 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1238 found = true; 1239 break; 1240 } 1241 } 1242 1243 if (!found) { 1244 /* 1245 * Create a new entry if allowable 1246 */ 1247 1248 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1249 (c = ip6mr_cache_alloc_unres()) == NULL) { 1250 spin_unlock_bh(&mfc_unres_lock); 1251 1252 kfree_skb(skb); 1253 return -ENOBUFS; 1254 } 1255 1256 /* 1257 * Fill in the new cache entry 1258 */ 1259 c->mf6c_parent = -1; 1260 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1261 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1262 1263 /* 1264 * Reflect first query at pim6sd 1265 */ 1266 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1267 if (err < 0) { 1268 /* If the report failed throw the cache entry 1269 out - Brad Parker 1270 */ 1271 spin_unlock_bh(&mfc_unres_lock); 1272 1273 ip6mr_cache_free(c); 1274 kfree_skb(skb); 1275 return err; 1276 } 1277 1278 atomic_inc(&mrt->cache_resolve_queue_len); 1279 list_add(&c->list, &mrt->mfc6_unres_queue); 1280 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1281 1282 ipmr_do_expire_process(mrt); 1283 } 1284 1285 /* 1286 * See if we can append the packet 1287 */ 1288 if (c->mfc_un.unres.unresolved.qlen > 3) { 1289 kfree_skb(skb); 1290 err = -ENOBUFS; 1291 } else { 1292 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1293 err = 0; 1294 } 1295 1296 spin_unlock_bh(&mfc_unres_lock); 1297 return err; 1298 } 1299 1300 /* 1301 * MFC6 cache manipulation by user space 1302 */ 1303 1304 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc, 1305 int parent) 1306 { 1307 int line; 1308 struct mfc6_cache *c, *next; 1309 1310 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1311 1312 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) { 1313 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1314 ipv6_addr_equal(&c->mf6c_mcastgrp, 1315 &mfc->mf6cc_mcastgrp.sin6_addr) && 1316 (parent == -1 || parent == c->mf6c_parent)) { 1317 write_lock_bh(&mrt_lock); 1318 list_del(&c->list); 1319 write_unlock_bh(&mrt_lock); 1320 1321 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1322 ip6mr_cache_free(c); 1323 return 0; 1324 } 1325 } 1326 return -ENOENT; 1327 } 1328 1329 static int ip6mr_device_event(struct notifier_block *this, 1330 unsigned long event, void *ptr) 1331 { 1332 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1333 struct net *net = dev_net(dev); 1334 struct mr6_table *mrt; 1335 struct mif_device *v; 1336 int ct; 1337 1338 if (event != NETDEV_UNREGISTER) 1339 return NOTIFY_DONE; 1340 1341 ip6mr_for_each_table(mrt, net) { 1342 v = &mrt->vif6_table[0]; 1343 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1344 if (v->dev == dev) 1345 mif6_delete(mrt, ct, 1, NULL); 1346 } 1347 } 1348 1349 return NOTIFY_DONE; 1350 } 1351 1352 static struct notifier_block ip6_mr_notifier = { 1353 .notifier_call = ip6mr_device_event 1354 }; 1355 1356 /* 1357 * Setup for IP multicast routing 1358 */ 1359 1360 static int __net_init ip6mr_net_init(struct net *net) 1361 { 1362 int err; 1363 1364 err = ip6mr_rules_init(net); 1365 if (err < 0) 1366 goto fail; 1367 1368 #ifdef CONFIG_PROC_FS 1369 err = -ENOMEM; 1370 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops)) 1371 goto proc_vif_fail; 1372 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops)) 1373 goto proc_cache_fail; 1374 #endif 1375 1376 return 0; 1377 1378 #ifdef CONFIG_PROC_FS 1379 proc_cache_fail: 1380 remove_proc_entry("ip6_mr_vif", net->proc_net); 1381 proc_vif_fail: 1382 ip6mr_rules_exit(net); 1383 #endif 1384 fail: 1385 return err; 1386 } 1387 1388 static void __net_exit ip6mr_net_exit(struct net *net) 1389 { 1390 #ifdef CONFIG_PROC_FS 1391 remove_proc_entry("ip6_mr_cache", net->proc_net); 1392 remove_proc_entry("ip6_mr_vif", net->proc_net); 1393 #endif 1394 ip6mr_rules_exit(net); 1395 } 1396 1397 static struct pernet_operations ip6mr_net_ops = { 1398 .init = ip6mr_net_init, 1399 .exit = ip6mr_net_exit, 1400 }; 1401 1402 int __init ip6_mr_init(void) 1403 { 1404 int err; 1405 1406 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1407 sizeof(struct mfc6_cache), 1408 0, SLAB_HWCACHE_ALIGN, 1409 NULL); 1410 if (!mrt_cachep) 1411 return -ENOMEM; 1412 1413 err = register_pernet_subsys(&ip6mr_net_ops); 1414 if (err) 1415 goto reg_pernet_fail; 1416 1417 err = register_netdevice_notifier(&ip6_mr_notifier); 1418 if (err) 1419 goto reg_notif_fail; 1420 #ifdef CONFIG_IPV6_PIMSM_V2 1421 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1422 pr_err("%s: can't add PIM protocol\n", __func__); 1423 err = -EAGAIN; 1424 goto add_proto_fail; 1425 } 1426 #endif 1427 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, 1428 ip6mr_rtm_dumproute, NULL); 1429 return 0; 1430 #ifdef CONFIG_IPV6_PIMSM_V2 1431 add_proto_fail: 1432 unregister_netdevice_notifier(&ip6_mr_notifier); 1433 #endif 1434 reg_notif_fail: 1435 unregister_pernet_subsys(&ip6mr_net_ops); 1436 reg_pernet_fail: 1437 kmem_cache_destroy(mrt_cachep); 1438 return err; 1439 } 1440 1441 void ip6_mr_cleanup(void) 1442 { 1443 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1444 #ifdef CONFIG_IPV6_PIMSM_V2 1445 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1446 #endif 1447 unregister_netdevice_notifier(&ip6_mr_notifier); 1448 unregister_pernet_subsys(&ip6mr_net_ops); 1449 kmem_cache_destroy(mrt_cachep); 1450 } 1451 1452 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1453 struct mf6cctl *mfc, int mrtsock, int parent) 1454 { 1455 bool found = false; 1456 int line; 1457 struct mfc6_cache *uc, *c; 1458 unsigned char ttls[MAXMIFS]; 1459 int i; 1460 1461 if (mfc->mf6cc_parent >= MAXMIFS) 1462 return -ENFILE; 1463 1464 memset(ttls, 255, MAXMIFS); 1465 for (i = 0; i < MAXMIFS; i++) { 1466 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1467 ttls[i] = 1; 1468 1469 } 1470 1471 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1472 1473 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1474 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1475 ipv6_addr_equal(&c->mf6c_mcastgrp, 1476 &mfc->mf6cc_mcastgrp.sin6_addr) && 1477 (parent == -1 || parent == mfc->mf6cc_parent)) { 1478 found = true; 1479 break; 1480 } 1481 } 1482 1483 if (found) { 1484 write_lock_bh(&mrt_lock); 1485 c->mf6c_parent = mfc->mf6cc_parent; 1486 ip6mr_update_thresholds(mrt, c, ttls); 1487 if (!mrtsock) 1488 c->mfc_flags |= MFC_STATIC; 1489 write_unlock_bh(&mrt_lock); 1490 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1491 return 0; 1492 } 1493 1494 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1495 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1496 return -EINVAL; 1497 1498 c = ip6mr_cache_alloc(); 1499 if (!c) 1500 return -ENOMEM; 1501 1502 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1503 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1504 c->mf6c_parent = mfc->mf6cc_parent; 1505 ip6mr_update_thresholds(mrt, c, ttls); 1506 if (!mrtsock) 1507 c->mfc_flags |= MFC_STATIC; 1508 1509 write_lock_bh(&mrt_lock); 1510 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1511 write_unlock_bh(&mrt_lock); 1512 1513 /* 1514 * Check to see if we resolved a queued list. If so we 1515 * need to send on the frames and tidy up. 1516 */ 1517 found = false; 1518 spin_lock_bh(&mfc_unres_lock); 1519 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1520 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1521 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1522 list_del(&uc->list); 1523 atomic_dec(&mrt->cache_resolve_queue_len); 1524 found = true; 1525 break; 1526 } 1527 } 1528 if (list_empty(&mrt->mfc6_unres_queue)) 1529 del_timer(&mrt->ipmr_expire_timer); 1530 spin_unlock_bh(&mfc_unres_lock); 1531 1532 if (found) { 1533 ip6mr_cache_resolve(net, mrt, uc, c); 1534 ip6mr_cache_free(uc); 1535 } 1536 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1537 return 0; 1538 } 1539 1540 /* 1541 * Close the multicast socket, and clear the vif tables etc 1542 */ 1543 1544 static void mroute_clean_tables(struct mr6_table *mrt, bool all) 1545 { 1546 int i; 1547 LIST_HEAD(list); 1548 struct mfc6_cache *c, *next; 1549 1550 /* 1551 * Shut down all active vif entries 1552 */ 1553 for (i = 0; i < mrt->maxvif; i++) { 1554 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) 1555 continue; 1556 mif6_delete(mrt, i, 0, &list); 1557 } 1558 unregister_netdevice_many(&list); 1559 1560 /* 1561 * Wipe the cache 1562 */ 1563 for (i = 0; i < MFC6_LINES; i++) { 1564 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1565 if (!all && (c->mfc_flags & MFC_STATIC)) 1566 continue; 1567 write_lock_bh(&mrt_lock); 1568 list_del(&c->list); 1569 write_unlock_bh(&mrt_lock); 1570 1571 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1572 ip6mr_cache_free(c); 1573 } 1574 } 1575 1576 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1577 spin_lock_bh(&mfc_unres_lock); 1578 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1579 list_del(&c->list); 1580 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1581 ip6mr_destroy_unres(mrt, c); 1582 } 1583 spin_unlock_bh(&mfc_unres_lock); 1584 } 1585 } 1586 1587 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1588 { 1589 int err = 0; 1590 struct net *net = sock_net(sk); 1591 1592 rtnl_lock(); 1593 write_lock_bh(&mrt_lock); 1594 if (likely(mrt->mroute6_sk == NULL)) { 1595 mrt->mroute6_sk = sk; 1596 net->ipv6.devconf_all->mc_forwarding++; 1597 } else { 1598 err = -EADDRINUSE; 1599 } 1600 write_unlock_bh(&mrt_lock); 1601 1602 if (!err) 1603 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1604 NETCONFA_MC_FORWARDING, 1605 NETCONFA_IFINDEX_ALL, 1606 net->ipv6.devconf_all); 1607 rtnl_unlock(); 1608 1609 return err; 1610 } 1611 1612 int ip6mr_sk_done(struct sock *sk) 1613 { 1614 int err = -EACCES; 1615 struct net *net = sock_net(sk); 1616 struct mr6_table *mrt; 1617 1618 rtnl_lock(); 1619 ip6mr_for_each_table(mrt, net) { 1620 if (sk == mrt->mroute6_sk) { 1621 write_lock_bh(&mrt_lock); 1622 mrt->mroute6_sk = NULL; 1623 net->ipv6.devconf_all->mc_forwarding--; 1624 write_unlock_bh(&mrt_lock); 1625 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1626 NETCONFA_MC_FORWARDING, 1627 NETCONFA_IFINDEX_ALL, 1628 net->ipv6.devconf_all); 1629 1630 mroute_clean_tables(mrt, false); 1631 err = 0; 1632 break; 1633 } 1634 } 1635 rtnl_unlock(); 1636 1637 return err; 1638 } 1639 1640 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1641 { 1642 struct mr6_table *mrt; 1643 struct flowi6 fl6 = { 1644 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1645 .flowi6_oif = skb->dev->ifindex, 1646 .flowi6_mark = skb->mark, 1647 }; 1648 1649 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1650 return NULL; 1651 1652 return mrt->mroute6_sk; 1653 } 1654 1655 /* 1656 * Socket options and virtual interface manipulation. The whole 1657 * virtual interface system is a complete heap, but unfortunately 1658 * that's how BSD mrouted happens to think. Maybe one day with a proper 1659 * MOSPF/PIM router set up we can clean this up. 1660 */ 1661 1662 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1663 { 1664 int ret, parent = 0; 1665 struct mif6ctl vif; 1666 struct mf6cctl mfc; 1667 mifi_t mifi; 1668 struct net *net = sock_net(sk); 1669 struct mr6_table *mrt; 1670 1671 if (sk->sk_type != SOCK_RAW || 1672 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1673 return -EOPNOTSUPP; 1674 1675 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1676 if (!mrt) 1677 return -ENOENT; 1678 1679 if (optname != MRT6_INIT) { 1680 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1681 return -EACCES; 1682 } 1683 1684 switch (optname) { 1685 case MRT6_INIT: 1686 if (optlen < sizeof(int)) 1687 return -EINVAL; 1688 1689 return ip6mr_sk_init(mrt, sk); 1690 1691 case MRT6_DONE: 1692 return ip6mr_sk_done(sk); 1693 1694 case MRT6_ADD_MIF: 1695 if (optlen < sizeof(vif)) 1696 return -EINVAL; 1697 if (copy_from_user(&vif, optval, sizeof(vif))) 1698 return -EFAULT; 1699 if (vif.mif6c_mifi >= MAXMIFS) 1700 return -ENFILE; 1701 rtnl_lock(); 1702 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1703 rtnl_unlock(); 1704 return ret; 1705 1706 case MRT6_DEL_MIF: 1707 if (optlen < sizeof(mifi_t)) 1708 return -EINVAL; 1709 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1710 return -EFAULT; 1711 rtnl_lock(); 1712 ret = mif6_delete(mrt, mifi, 0, NULL); 1713 rtnl_unlock(); 1714 return ret; 1715 1716 /* 1717 * Manipulate the forwarding caches. These live 1718 * in a sort of kernel/user symbiosis. 1719 */ 1720 case MRT6_ADD_MFC: 1721 case MRT6_DEL_MFC: 1722 parent = -1; 1723 case MRT6_ADD_MFC_PROXY: 1724 case MRT6_DEL_MFC_PROXY: 1725 if (optlen < sizeof(mfc)) 1726 return -EINVAL; 1727 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1728 return -EFAULT; 1729 if (parent == 0) 1730 parent = mfc.mf6cc_parent; 1731 rtnl_lock(); 1732 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1733 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1734 else 1735 ret = ip6mr_mfc_add(net, mrt, &mfc, 1736 sk == mrt->mroute6_sk, parent); 1737 rtnl_unlock(); 1738 return ret; 1739 1740 /* 1741 * Control PIM assert (to activate pim will activate assert) 1742 */ 1743 case MRT6_ASSERT: 1744 { 1745 int v; 1746 1747 if (optlen != sizeof(v)) 1748 return -EINVAL; 1749 if (get_user(v, (int __user *)optval)) 1750 return -EFAULT; 1751 mrt->mroute_do_assert = v; 1752 return 0; 1753 } 1754 1755 #ifdef CONFIG_IPV6_PIMSM_V2 1756 case MRT6_PIM: 1757 { 1758 int v; 1759 1760 if (optlen != sizeof(v)) 1761 return -EINVAL; 1762 if (get_user(v, (int __user *)optval)) 1763 return -EFAULT; 1764 v = !!v; 1765 rtnl_lock(); 1766 ret = 0; 1767 if (v != mrt->mroute_do_pim) { 1768 mrt->mroute_do_pim = v; 1769 mrt->mroute_do_assert = v; 1770 } 1771 rtnl_unlock(); 1772 return ret; 1773 } 1774 1775 #endif 1776 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1777 case MRT6_TABLE: 1778 { 1779 u32 v; 1780 1781 if (optlen != sizeof(u32)) 1782 return -EINVAL; 1783 if (get_user(v, (u32 __user *)optval)) 1784 return -EFAULT; 1785 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1786 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1787 return -EINVAL; 1788 if (sk == mrt->mroute6_sk) 1789 return -EBUSY; 1790 1791 rtnl_lock(); 1792 ret = 0; 1793 if (!ip6mr_new_table(net, v)) 1794 ret = -ENOMEM; 1795 raw6_sk(sk)->ip6mr_table = v; 1796 rtnl_unlock(); 1797 return ret; 1798 } 1799 #endif 1800 /* 1801 * Spurious command, or MRT6_VERSION which you cannot 1802 * set. 1803 */ 1804 default: 1805 return -ENOPROTOOPT; 1806 } 1807 } 1808 1809 /* 1810 * Getsock opt support for the multicast routing system. 1811 */ 1812 1813 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1814 int __user *optlen) 1815 { 1816 int olr; 1817 int val; 1818 struct net *net = sock_net(sk); 1819 struct mr6_table *mrt; 1820 1821 if (sk->sk_type != SOCK_RAW || 1822 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1823 return -EOPNOTSUPP; 1824 1825 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1826 if (!mrt) 1827 return -ENOENT; 1828 1829 switch (optname) { 1830 case MRT6_VERSION: 1831 val = 0x0305; 1832 break; 1833 #ifdef CONFIG_IPV6_PIMSM_V2 1834 case MRT6_PIM: 1835 val = mrt->mroute_do_pim; 1836 break; 1837 #endif 1838 case MRT6_ASSERT: 1839 val = mrt->mroute_do_assert; 1840 break; 1841 default: 1842 return -ENOPROTOOPT; 1843 } 1844 1845 if (get_user(olr, optlen)) 1846 return -EFAULT; 1847 1848 olr = min_t(int, olr, sizeof(int)); 1849 if (olr < 0) 1850 return -EINVAL; 1851 1852 if (put_user(olr, optlen)) 1853 return -EFAULT; 1854 if (copy_to_user(optval, &val, olr)) 1855 return -EFAULT; 1856 return 0; 1857 } 1858 1859 /* 1860 * The IP multicast ioctl support routines. 1861 */ 1862 1863 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1864 { 1865 struct sioc_sg_req6 sr; 1866 struct sioc_mif_req6 vr; 1867 struct mif_device *vif; 1868 struct mfc6_cache *c; 1869 struct net *net = sock_net(sk); 1870 struct mr6_table *mrt; 1871 1872 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1873 if (!mrt) 1874 return -ENOENT; 1875 1876 switch (cmd) { 1877 case SIOCGETMIFCNT_IN6: 1878 if (copy_from_user(&vr, arg, sizeof(vr))) 1879 return -EFAULT; 1880 if (vr.mifi >= mrt->maxvif) 1881 return -EINVAL; 1882 read_lock(&mrt_lock); 1883 vif = &mrt->vif6_table[vr.mifi]; 1884 if (MIF_EXISTS(mrt, vr.mifi)) { 1885 vr.icount = vif->pkt_in; 1886 vr.ocount = vif->pkt_out; 1887 vr.ibytes = vif->bytes_in; 1888 vr.obytes = vif->bytes_out; 1889 read_unlock(&mrt_lock); 1890 1891 if (copy_to_user(arg, &vr, sizeof(vr))) 1892 return -EFAULT; 1893 return 0; 1894 } 1895 read_unlock(&mrt_lock); 1896 return -EADDRNOTAVAIL; 1897 case SIOCGETSGCNT_IN6: 1898 if (copy_from_user(&sr, arg, sizeof(sr))) 1899 return -EFAULT; 1900 1901 read_lock(&mrt_lock); 1902 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1903 if (c) { 1904 sr.pktcnt = c->mfc_un.res.pkt; 1905 sr.bytecnt = c->mfc_un.res.bytes; 1906 sr.wrong_if = c->mfc_un.res.wrong_if; 1907 read_unlock(&mrt_lock); 1908 1909 if (copy_to_user(arg, &sr, sizeof(sr))) 1910 return -EFAULT; 1911 return 0; 1912 } 1913 read_unlock(&mrt_lock); 1914 return -EADDRNOTAVAIL; 1915 default: 1916 return -ENOIOCTLCMD; 1917 } 1918 } 1919 1920 #ifdef CONFIG_COMPAT 1921 struct compat_sioc_sg_req6 { 1922 struct sockaddr_in6 src; 1923 struct sockaddr_in6 grp; 1924 compat_ulong_t pktcnt; 1925 compat_ulong_t bytecnt; 1926 compat_ulong_t wrong_if; 1927 }; 1928 1929 struct compat_sioc_mif_req6 { 1930 mifi_t mifi; 1931 compat_ulong_t icount; 1932 compat_ulong_t ocount; 1933 compat_ulong_t ibytes; 1934 compat_ulong_t obytes; 1935 }; 1936 1937 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1938 { 1939 struct compat_sioc_sg_req6 sr; 1940 struct compat_sioc_mif_req6 vr; 1941 struct mif_device *vif; 1942 struct mfc6_cache *c; 1943 struct net *net = sock_net(sk); 1944 struct mr6_table *mrt; 1945 1946 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1947 if (!mrt) 1948 return -ENOENT; 1949 1950 switch (cmd) { 1951 case SIOCGETMIFCNT_IN6: 1952 if (copy_from_user(&vr, arg, sizeof(vr))) 1953 return -EFAULT; 1954 if (vr.mifi >= mrt->maxvif) 1955 return -EINVAL; 1956 read_lock(&mrt_lock); 1957 vif = &mrt->vif6_table[vr.mifi]; 1958 if (MIF_EXISTS(mrt, vr.mifi)) { 1959 vr.icount = vif->pkt_in; 1960 vr.ocount = vif->pkt_out; 1961 vr.ibytes = vif->bytes_in; 1962 vr.obytes = vif->bytes_out; 1963 read_unlock(&mrt_lock); 1964 1965 if (copy_to_user(arg, &vr, sizeof(vr))) 1966 return -EFAULT; 1967 return 0; 1968 } 1969 read_unlock(&mrt_lock); 1970 return -EADDRNOTAVAIL; 1971 case SIOCGETSGCNT_IN6: 1972 if (copy_from_user(&sr, arg, sizeof(sr))) 1973 return -EFAULT; 1974 1975 read_lock(&mrt_lock); 1976 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1977 if (c) { 1978 sr.pktcnt = c->mfc_un.res.pkt; 1979 sr.bytecnt = c->mfc_un.res.bytes; 1980 sr.wrong_if = c->mfc_un.res.wrong_if; 1981 read_unlock(&mrt_lock); 1982 1983 if (copy_to_user(arg, &sr, sizeof(sr))) 1984 return -EFAULT; 1985 return 0; 1986 } 1987 read_unlock(&mrt_lock); 1988 return -EADDRNOTAVAIL; 1989 default: 1990 return -ENOIOCTLCMD; 1991 } 1992 } 1993 #endif 1994 1995 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 1996 { 1997 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 1998 IPSTATS_MIB_OUTFORWDATAGRAMS); 1999 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)), 2000 IPSTATS_MIB_OUTOCTETS, skb->len); 2001 return dst_output(net, sk, skb); 2002 } 2003 2004 /* 2005 * Processing handlers for ip6mr_forward 2006 */ 2007 2008 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 2009 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 2010 { 2011 struct ipv6hdr *ipv6h; 2012 struct mif_device *vif = &mrt->vif6_table[vifi]; 2013 struct net_device *dev; 2014 struct dst_entry *dst; 2015 struct flowi6 fl6; 2016 2017 if (!vif->dev) 2018 goto out_free; 2019 2020 #ifdef CONFIG_IPV6_PIMSM_V2 2021 if (vif->flags & MIFF_REGISTER) { 2022 vif->pkt_out++; 2023 vif->bytes_out += skb->len; 2024 vif->dev->stats.tx_bytes += skb->len; 2025 vif->dev->stats.tx_packets++; 2026 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2027 goto out_free; 2028 } 2029 #endif 2030 2031 ipv6h = ipv6_hdr(skb); 2032 2033 fl6 = (struct flowi6) { 2034 .flowi6_oif = vif->link, 2035 .daddr = ipv6h->daddr, 2036 }; 2037 2038 dst = ip6_route_output(net, NULL, &fl6); 2039 if (dst->error) { 2040 dst_release(dst); 2041 goto out_free; 2042 } 2043 2044 skb_dst_drop(skb); 2045 skb_dst_set(skb, dst); 2046 2047 /* 2048 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2049 * not only before forwarding, but after forwarding on all output 2050 * interfaces. It is clear, if mrouter runs a multicasting 2051 * program, it should receive packets not depending to what interface 2052 * program is joined. 2053 * If we will not make it, the program will have to join on all 2054 * interfaces. On the other hand, multihoming host (or router, but 2055 * not mrouter) cannot join to more than one interface - it will 2056 * result in receiving multiple packets. 2057 */ 2058 dev = vif->dev; 2059 skb->dev = dev; 2060 vif->pkt_out++; 2061 vif->bytes_out += skb->len; 2062 2063 /* We are about to write */ 2064 /* XXX: extension headers? */ 2065 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2066 goto out_free; 2067 2068 ipv6h = ipv6_hdr(skb); 2069 ipv6h->hop_limit--; 2070 2071 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2072 2073 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2074 net, NULL, skb, skb->dev, dev, 2075 ip6mr_forward2_finish); 2076 2077 out_free: 2078 kfree_skb(skb); 2079 return 0; 2080 } 2081 2082 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2083 { 2084 int ct; 2085 2086 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2087 if (mrt->vif6_table[ct].dev == dev) 2088 break; 2089 } 2090 return ct; 2091 } 2092 2093 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2094 struct sk_buff *skb, struct mfc6_cache *cache) 2095 { 2096 int psend = -1; 2097 int vif, ct; 2098 int true_vifi = ip6mr_find_vif(mrt, skb->dev); 2099 2100 vif = cache->mf6c_parent; 2101 cache->mfc_un.res.pkt++; 2102 cache->mfc_un.res.bytes += skb->len; 2103 cache->mfc_un.res.lastuse = jiffies; 2104 2105 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) { 2106 struct mfc6_cache *cache_proxy; 2107 2108 /* For an (*,G) entry, we only check that the incoming 2109 * interface is part of the static tree. 2110 */ 2111 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif); 2112 if (cache_proxy && 2113 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 2114 goto forward; 2115 } 2116 2117 /* 2118 * Wrong interface: drop packet and (maybe) send PIM assert. 2119 */ 2120 if (mrt->vif6_table[vif].dev != skb->dev) { 2121 cache->mfc_un.res.wrong_if++; 2122 2123 if (true_vifi >= 0 && mrt->mroute_do_assert && 2124 /* pimsm uses asserts, when switching from RPT to SPT, 2125 so that we cannot check that packet arrived on an oif. 2126 It is bad, but otherwise we would need to move pretty 2127 large chunk of pimd to kernel. Ough... --ANK 2128 */ 2129 (mrt->mroute_do_pim || 2130 cache->mfc_un.res.ttls[true_vifi] < 255) && 2131 time_after(jiffies, 2132 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2133 cache->mfc_un.res.last_assert = jiffies; 2134 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2135 } 2136 goto dont_forward; 2137 } 2138 2139 forward: 2140 mrt->vif6_table[vif].pkt_in++; 2141 mrt->vif6_table[vif].bytes_in += skb->len; 2142 2143 /* 2144 * Forward the frame 2145 */ 2146 if (ipv6_addr_any(&cache->mf6c_origin) && 2147 ipv6_addr_any(&cache->mf6c_mcastgrp)) { 2148 if (true_vifi >= 0 && 2149 true_vifi != cache->mf6c_parent && 2150 ipv6_hdr(skb)->hop_limit > 2151 cache->mfc_un.res.ttls[cache->mf6c_parent]) { 2152 /* It's an (*,*) entry and the packet is not coming from 2153 * the upstream: forward the packet to the upstream 2154 * only. 2155 */ 2156 psend = cache->mf6c_parent; 2157 goto last_forward; 2158 } 2159 goto dont_forward; 2160 } 2161 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2162 /* For (*,G) entry, don't forward to the incoming interface */ 2163 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) && 2164 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2165 if (psend != -1) { 2166 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2167 if (skb2) 2168 ip6mr_forward2(net, mrt, skb2, cache, psend); 2169 } 2170 psend = ct; 2171 } 2172 } 2173 last_forward: 2174 if (psend != -1) { 2175 ip6mr_forward2(net, mrt, skb, cache, psend); 2176 return; 2177 } 2178 2179 dont_forward: 2180 kfree_skb(skb); 2181 } 2182 2183 2184 /* 2185 * Multicast packets for forwarding arrive here 2186 */ 2187 2188 int ip6_mr_input(struct sk_buff *skb) 2189 { 2190 struct mfc6_cache *cache; 2191 struct net *net = dev_net(skb->dev); 2192 struct mr6_table *mrt; 2193 struct flowi6 fl6 = { 2194 .flowi6_iif = skb->dev->ifindex, 2195 .flowi6_mark = skb->mark, 2196 }; 2197 int err; 2198 2199 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2200 if (err < 0) { 2201 kfree_skb(skb); 2202 return err; 2203 } 2204 2205 read_lock(&mrt_lock); 2206 cache = ip6mr_cache_find(mrt, 2207 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2208 if (!cache) { 2209 int vif = ip6mr_find_vif(mrt, skb->dev); 2210 2211 if (vif >= 0) 2212 cache = ip6mr_cache_find_any(mrt, 2213 &ipv6_hdr(skb)->daddr, 2214 vif); 2215 } 2216 2217 /* 2218 * No usable cache entry 2219 */ 2220 if (!cache) { 2221 int vif; 2222 2223 vif = ip6mr_find_vif(mrt, skb->dev); 2224 if (vif >= 0) { 2225 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2226 read_unlock(&mrt_lock); 2227 2228 return err; 2229 } 2230 read_unlock(&mrt_lock); 2231 kfree_skb(skb); 2232 return -ENODEV; 2233 } 2234 2235 ip6_mr_forward(net, mrt, skb, cache); 2236 2237 read_unlock(&mrt_lock); 2238 2239 return 0; 2240 } 2241 2242 2243 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2244 struct mfc6_cache *c, struct rtmsg *rtm) 2245 { 2246 struct rta_mfc_stats mfcs; 2247 struct nlattr *mp_attr; 2248 struct rtnexthop *nhp; 2249 unsigned long lastuse; 2250 int ct; 2251 2252 /* If cache is unresolved, don't try to parse IIF and OIF */ 2253 if (c->mf6c_parent >= MAXMIFS) { 2254 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2255 return -ENOENT; 2256 } 2257 2258 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2259 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2260 return -EMSGSIZE; 2261 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2262 if (!mp_attr) 2263 return -EMSGSIZE; 2264 2265 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2266 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2267 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2268 if (!nhp) { 2269 nla_nest_cancel(skb, mp_attr); 2270 return -EMSGSIZE; 2271 } 2272 2273 nhp->rtnh_flags = 0; 2274 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2275 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2276 nhp->rtnh_len = sizeof(*nhp); 2277 } 2278 } 2279 2280 nla_nest_end(skb, mp_attr); 2281 2282 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2283 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2284 2285 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2286 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2287 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2288 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2289 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2290 RTA_PAD)) 2291 return -EMSGSIZE; 2292 2293 rtm->rtm_type = RTN_MULTICAST; 2294 return 1; 2295 } 2296 2297 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2298 u32 portid) 2299 { 2300 int err; 2301 struct mr6_table *mrt; 2302 struct mfc6_cache *cache; 2303 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2304 2305 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2306 if (!mrt) 2307 return -ENOENT; 2308 2309 read_lock(&mrt_lock); 2310 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2311 if (!cache && skb->dev) { 2312 int vif = ip6mr_find_vif(mrt, skb->dev); 2313 2314 if (vif >= 0) 2315 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2316 vif); 2317 } 2318 2319 if (!cache) { 2320 struct sk_buff *skb2; 2321 struct ipv6hdr *iph; 2322 struct net_device *dev; 2323 int vif; 2324 2325 dev = skb->dev; 2326 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2327 read_unlock(&mrt_lock); 2328 return -ENODEV; 2329 } 2330 2331 /* really correct? */ 2332 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2333 if (!skb2) { 2334 read_unlock(&mrt_lock); 2335 return -ENOMEM; 2336 } 2337 2338 NETLINK_CB(skb2).portid = portid; 2339 skb_reset_transport_header(skb2); 2340 2341 skb_put(skb2, sizeof(struct ipv6hdr)); 2342 skb_reset_network_header(skb2); 2343 2344 iph = ipv6_hdr(skb2); 2345 iph->version = 0; 2346 iph->priority = 0; 2347 iph->flow_lbl[0] = 0; 2348 iph->flow_lbl[1] = 0; 2349 iph->flow_lbl[2] = 0; 2350 iph->payload_len = 0; 2351 iph->nexthdr = IPPROTO_NONE; 2352 iph->hop_limit = 0; 2353 iph->saddr = rt->rt6i_src.addr; 2354 iph->daddr = rt->rt6i_dst.addr; 2355 2356 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2357 read_unlock(&mrt_lock); 2358 2359 return err; 2360 } 2361 2362 if (rtm->rtm_flags & RTM_F_NOTIFY) 2363 cache->mfc_flags |= MFC_NOTIFY; 2364 2365 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2366 read_unlock(&mrt_lock); 2367 return err; 2368 } 2369 2370 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2371 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2372 int flags) 2373 { 2374 struct nlmsghdr *nlh; 2375 struct rtmsg *rtm; 2376 int err; 2377 2378 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2379 if (!nlh) 2380 return -EMSGSIZE; 2381 2382 rtm = nlmsg_data(nlh); 2383 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2384 rtm->rtm_dst_len = 128; 2385 rtm->rtm_src_len = 128; 2386 rtm->rtm_tos = 0; 2387 rtm->rtm_table = mrt->id; 2388 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2389 goto nla_put_failure; 2390 rtm->rtm_type = RTN_MULTICAST; 2391 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2392 if (c->mfc_flags & MFC_STATIC) 2393 rtm->rtm_protocol = RTPROT_STATIC; 2394 else 2395 rtm->rtm_protocol = RTPROT_MROUTED; 2396 rtm->rtm_flags = 0; 2397 2398 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2399 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2400 goto nla_put_failure; 2401 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2402 /* do not break the dump if cache is unresolved */ 2403 if (err < 0 && err != -ENOENT) 2404 goto nla_put_failure; 2405 2406 nlmsg_end(skb, nlh); 2407 return 0; 2408 2409 nla_put_failure: 2410 nlmsg_cancel(skb, nlh); 2411 return -EMSGSIZE; 2412 } 2413 2414 static int mr6_msgsize(bool unresolved, int maxvif) 2415 { 2416 size_t len = 2417 NLMSG_ALIGN(sizeof(struct rtmsg)) 2418 + nla_total_size(4) /* RTA_TABLE */ 2419 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2420 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2421 ; 2422 2423 if (!unresolved) 2424 len = len 2425 + nla_total_size(4) /* RTA_IIF */ 2426 + nla_total_size(0) /* RTA_MULTIPATH */ 2427 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2428 /* RTA_MFC_STATS */ 2429 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2430 ; 2431 2432 return len; 2433 } 2434 2435 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2436 int cmd) 2437 { 2438 struct net *net = read_pnet(&mrt->net); 2439 struct sk_buff *skb; 2440 int err = -ENOBUFS; 2441 2442 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2443 GFP_ATOMIC); 2444 if (!skb) 2445 goto errout; 2446 2447 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2448 if (err < 0) 2449 goto errout; 2450 2451 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2452 return; 2453 2454 errout: 2455 kfree_skb(skb); 2456 if (err < 0) 2457 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2458 } 2459 2460 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2461 { 2462 struct net *net = sock_net(skb->sk); 2463 struct mr6_table *mrt; 2464 struct mfc6_cache *mfc; 2465 unsigned int t = 0, s_t; 2466 unsigned int h = 0, s_h; 2467 unsigned int e = 0, s_e; 2468 2469 s_t = cb->args[0]; 2470 s_h = cb->args[1]; 2471 s_e = cb->args[2]; 2472 2473 read_lock(&mrt_lock); 2474 ip6mr_for_each_table(mrt, net) { 2475 if (t < s_t) 2476 goto next_table; 2477 if (t > s_t) 2478 s_h = 0; 2479 for (h = s_h; h < MFC6_LINES; h++) { 2480 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2481 if (e < s_e) 2482 goto next_entry; 2483 if (ip6mr_fill_mroute(mrt, skb, 2484 NETLINK_CB(cb->skb).portid, 2485 cb->nlh->nlmsg_seq, 2486 mfc, RTM_NEWROUTE, 2487 NLM_F_MULTI) < 0) 2488 goto done; 2489 next_entry: 2490 e++; 2491 } 2492 e = s_e = 0; 2493 } 2494 spin_lock_bh(&mfc_unres_lock); 2495 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2496 if (e < s_e) 2497 goto next_entry2; 2498 if (ip6mr_fill_mroute(mrt, skb, 2499 NETLINK_CB(cb->skb).portid, 2500 cb->nlh->nlmsg_seq, 2501 mfc, RTM_NEWROUTE, 2502 NLM_F_MULTI) < 0) { 2503 spin_unlock_bh(&mfc_unres_lock); 2504 goto done; 2505 } 2506 next_entry2: 2507 e++; 2508 } 2509 spin_unlock_bh(&mfc_unres_lock); 2510 e = s_e = 0; 2511 s_h = 0; 2512 next_table: 2513 t++; 2514 } 2515 done: 2516 read_unlock(&mrt_lock); 2517 2518 cb->args[2] = e; 2519 cb->args[1] = h; 2520 cb->args[0] = t; 2521 2522 return skb->len; 2523 } 2524