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->destructor = free_netdev; 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 = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); 850 nlh->nlmsg_type = NLMSG_ERROR; 851 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 852 skb_trim(skb, nlh->nlmsg_len); 853 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT; 854 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 855 } else 856 kfree_skb(skb); 857 } 858 859 ip6mr_cache_free(c); 860 } 861 862 863 /* Timer process for all the unresolved queue. */ 864 865 static void ipmr_do_expire_process(struct mr6_table *mrt) 866 { 867 unsigned long now = jiffies; 868 unsigned long expires = 10 * HZ; 869 struct mfc6_cache *c, *next; 870 871 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 872 if (time_after(c->mfc_un.unres.expires, now)) { 873 /* not yet... */ 874 unsigned long interval = c->mfc_un.unres.expires - now; 875 if (interval < expires) 876 expires = interval; 877 continue; 878 } 879 880 list_del(&c->list); 881 mr6_netlink_event(mrt, c, RTM_DELROUTE); 882 ip6mr_destroy_unres(mrt, c); 883 } 884 885 if (!list_empty(&mrt->mfc6_unres_queue)) 886 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 887 } 888 889 static void ipmr_expire_process(unsigned long arg) 890 { 891 struct mr6_table *mrt = (struct mr6_table *)arg; 892 893 if (!spin_trylock(&mfc_unres_lock)) { 894 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 895 return; 896 } 897 898 if (!list_empty(&mrt->mfc6_unres_queue)) 899 ipmr_do_expire_process(mrt); 900 901 spin_unlock(&mfc_unres_lock); 902 } 903 904 /* Fill oifs list. It is called under write locked mrt_lock. */ 905 906 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache, 907 unsigned char *ttls) 908 { 909 int vifi; 910 911 cache->mfc_un.res.minvif = MAXMIFS; 912 cache->mfc_un.res.maxvif = 0; 913 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 914 915 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 916 if (MIF_EXISTS(mrt, vifi) && 917 ttls[vifi] && ttls[vifi] < 255) { 918 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 919 if (cache->mfc_un.res.minvif > vifi) 920 cache->mfc_un.res.minvif = vifi; 921 if (cache->mfc_un.res.maxvif <= vifi) 922 cache->mfc_un.res.maxvif = vifi + 1; 923 } 924 } 925 cache->mfc_un.res.lastuse = jiffies; 926 } 927 928 static int mif6_add(struct net *net, struct mr6_table *mrt, 929 struct mif6ctl *vifc, int mrtsock) 930 { 931 int vifi = vifc->mif6c_mifi; 932 struct mif_device *v = &mrt->vif6_table[vifi]; 933 struct net_device *dev; 934 struct inet6_dev *in6_dev; 935 int err; 936 937 /* Is vif busy ? */ 938 if (MIF_EXISTS(mrt, vifi)) 939 return -EADDRINUSE; 940 941 switch (vifc->mif6c_flags) { 942 #ifdef CONFIG_IPV6_PIMSM_V2 943 case MIFF_REGISTER: 944 /* 945 * Special Purpose VIF in PIM 946 * All the packets will be sent to the daemon 947 */ 948 if (mrt->mroute_reg_vif_num >= 0) 949 return -EADDRINUSE; 950 dev = ip6mr_reg_vif(net, mrt); 951 if (!dev) 952 return -ENOBUFS; 953 err = dev_set_allmulti(dev, 1); 954 if (err) { 955 unregister_netdevice(dev); 956 dev_put(dev); 957 return err; 958 } 959 break; 960 #endif 961 case 0: 962 dev = dev_get_by_index(net, vifc->mif6c_pifi); 963 if (!dev) 964 return -EADDRNOTAVAIL; 965 err = dev_set_allmulti(dev, 1); 966 if (err) { 967 dev_put(dev); 968 return err; 969 } 970 break; 971 default: 972 return -EINVAL; 973 } 974 975 in6_dev = __in6_dev_get(dev); 976 if (in6_dev) { 977 in6_dev->cnf.mc_forwarding++; 978 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 979 NETCONFA_MC_FORWARDING, 980 dev->ifindex, &in6_dev->cnf); 981 } 982 983 /* 984 * Fill in the VIF structures 985 */ 986 v->rate_limit = vifc->vifc_rate_limit; 987 v->flags = vifc->mif6c_flags; 988 if (!mrtsock) 989 v->flags |= VIFF_STATIC; 990 v->threshold = vifc->vifc_threshold; 991 v->bytes_in = 0; 992 v->bytes_out = 0; 993 v->pkt_in = 0; 994 v->pkt_out = 0; 995 v->link = dev->ifindex; 996 if (v->flags & MIFF_REGISTER) 997 v->link = dev_get_iflink(dev); 998 999 /* And finish update writing critical data */ 1000 write_lock_bh(&mrt_lock); 1001 v->dev = dev; 1002 #ifdef CONFIG_IPV6_PIMSM_V2 1003 if (v->flags & MIFF_REGISTER) 1004 mrt->mroute_reg_vif_num = vifi; 1005 #endif 1006 if (vifi + 1 > mrt->maxvif) 1007 mrt->maxvif = vifi + 1; 1008 write_unlock_bh(&mrt_lock); 1009 return 0; 1010 } 1011 1012 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt, 1013 const struct in6_addr *origin, 1014 const struct in6_addr *mcastgrp) 1015 { 1016 int line = MFC6_HASH(mcastgrp, origin); 1017 struct mfc6_cache *c; 1018 1019 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1020 if (ipv6_addr_equal(&c->mf6c_origin, origin) && 1021 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) 1022 return c; 1023 } 1024 return NULL; 1025 } 1026 1027 /* Look for a (*,*,oif) entry */ 1028 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt, 1029 mifi_t mifi) 1030 { 1031 int line = MFC6_HASH(&in6addr_any, &in6addr_any); 1032 struct mfc6_cache *c; 1033 1034 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1035 if (ipv6_addr_any(&c->mf6c_origin) && 1036 ipv6_addr_any(&c->mf6c_mcastgrp) && 1037 (c->mfc_un.res.ttls[mifi] < 255)) 1038 return c; 1039 1040 return NULL; 1041 } 1042 1043 /* Look for a (*,G) entry */ 1044 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt, 1045 struct in6_addr *mcastgrp, 1046 mifi_t mifi) 1047 { 1048 int line = MFC6_HASH(mcastgrp, &in6addr_any); 1049 struct mfc6_cache *c, *proxy; 1050 1051 if (ipv6_addr_any(mcastgrp)) 1052 goto skip; 1053 1054 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1055 if (ipv6_addr_any(&c->mf6c_origin) && 1056 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) { 1057 if (c->mfc_un.res.ttls[mifi] < 255) 1058 return c; 1059 1060 /* It's ok if the mifi is part of the static tree */ 1061 proxy = ip6mr_cache_find_any_parent(mrt, 1062 c->mf6c_parent); 1063 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255) 1064 return c; 1065 } 1066 1067 skip: 1068 return ip6mr_cache_find_any_parent(mrt, mifi); 1069 } 1070 1071 /* 1072 * Allocate a multicast cache entry 1073 */ 1074 static struct mfc6_cache *ip6mr_cache_alloc(void) 1075 { 1076 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 1077 if (!c) 1078 return NULL; 1079 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 1080 c->mfc_un.res.minvif = MAXMIFS; 1081 return c; 1082 } 1083 1084 static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 1085 { 1086 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 1087 if (!c) 1088 return NULL; 1089 skb_queue_head_init(&c->mfc_un.unres.unresolved); 1090 c->mfc_un.unres.expires = jiffies + 10 * HZ; 1091 return c; 1092 } 1093 1094 /* 1095 * A cache entry has gone into a resolved state from queued 1096 */ 1097 1098 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt, 1099 struct mfc6_cache *uc, struct mfc6_cache *c) 1100 { 1101 struct sk_buff *skb; 1102 1103 /* 1104 * Play the pending entries through our router 1105 */ 1106 1107 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 1108 if (ipv6_hdr(skb)->version == 0) { 1109 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); 1110 1111 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 1112 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1113 } else { 1114 nlh->nlmsg_type = NLMSG_ERROR; 1115 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1116 skb_trim(skb, nlh->nlmsg_len); 1117 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE; 1118 } 1119 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1120 } else 1121 ip6_mr_forward(net, mrt, skb, c); 1122 } 1123 } 1124 1125 /* 1126 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd 1127 * expects the following bizarre scheme. 1128 * 1129 * Called under mrt_lock. 1130 */ 1131 1132 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 1133 mifi_t mifi, int assert) 1134 { 1135 struct sk_buff *skb; 1136 struct mrt6msg *msg; 1137 int ret; 1138 1139 #ifdef CONFIG_IPV6_PIMSM_V2 1140 if (assert == MRT6MSG_WHOLEPKT) 1141 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1142 +sizeof(*msg)); 1143 else 1144 #endif 1145 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1146 1147 if (!skb) 1148 return -ENOBUFS; 1149 1150 /* I suppose that internal messages 1151 * do not require checksums */ 1152 1153 skb->ip_summed = CHECKSUM_UNNECESSARY; 1154 1155 #ifdef CONFIG_IPV6_PIMSM_V2 1156 if (assert == MRT6MSG_WHOLEPKT) { 1157 /* Ugly, but we have no choice with this interface. 1158 Duplicate old header, fix length etc. 1159 And all this only to mangle msg->im6_msgtype and 1160 to set msg->im6_mbz to "mbz" :-) 1161 */ 1162 skb_push(skb, -skb_network_offset(pkt)); 1163 1164 skb_push(skb, sizeof(*msg)); 1165 skb_reset_transport_header(skb); 1166 msg = (struct mrt6msg *)skb_transport_header(skb); 1167 msg->im6_mbz = 0; 1168 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1169 msg->im6_mif = mrt->mroute_reg_vif_num; 1170 msg->im6_pad = 0; 1171 msg->im6_src = ipv6_hdr(pkt)->saddr; 1172 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1173 1174 skb->ip_summed = CHECKSUM_UNNECESSARY; 1175 } else 1176 #endif 1177 { 1178 /* 1179 * Copy the IP header 1180 */ 1181 1182 skb_put(skb, sizeof(struct ipv6hdr)); 1183 skb_reset_network_header(skb); 1184 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1185 1186 /* 1187 * Add our header 1188 */ 1189 skb_put(skb, sizeof(*msg)); 1190 skb_reset_transport_header(skb); 1191 msg = (struct mrt6msg *)skb_transport_header(skb); 1192 1193 msg->im6_mbz = 0; 1194 msg->im6_msgtype = assert; 1195 msg->im6_mif = mifi; 1196 msg->im6_pad = 0; 1197 msg->im6_src = ipv6_hdr(pkt)->saddr; 1198 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1199 1200 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1201 skb->ip_summed = CHECKSUM_UNNECESSARY; 1202 } 1203 1204 if (!mrt->mroute6_sk) { 1205 kfree_skb(skb); 1206 return -EINVAL; 1207 } 1208 1209 /* 1210 * Deliver to user space multicast routing algorithms 1211 */ 1212 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb); 1213 if (ret < 0) { 1214 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1215 kfree_skb(skb); 1216 } 1217 1218 return ret; 1219 } 1220 1221 /* 1222 * Queue a packet for resolution. It gets locked cache entry! 1223 */ 1224 1225 static int 1226 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb) 1227 { 1228 bool found = false; 1229 int err; 1230 struct mfc6_cache *c; 1231 1232 spin_lock_bh(&mfc_unres_lock); 1233 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) { 1234 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1235 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1236 found = true; 1237 break; 1238 } 1239 } 1240 1241 if (!found) { 1242 /* 1243 * Create a new entry if allowable 1244 */ 1245 1246 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1247 (c = ip6mr_cache_alloc_unres()) == NULL) { 1248 spin_unlock_bh(&mfc_unres_lock); 1249 1250 kfree_skb(skb); 1251 return -ENOBUFS; 1252 } 1253 1254 /* 1255 * Fill in the new cache entry 1256 */ 1257 c->mf6c_parent = -1; 1258 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1259 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1260 1261 /* 1262 * Reflect first query at pim6sd 1263 */ 1264 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1265 if (err < 0) { 1266 /* If the report failed throw the cache entry 1267 out - Brad Parker 1268 */ 1269 spin_unlock_bh(&mfc_unres_lock); 1270 1271 ip6mr_cache_free(c); 1272 kfree_skb(skb); 1273 return err; 1274 } 1275 1276 atomic_inc(&mrt->cache_resolve_queue_len); 1277 list_add(&c->list, &mrt->mfc6_unres_queue); 1278 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1279 1280 ipmr_do_expire_process(mrt); 1281 } 1282 1283 /* 1284 * See if we can append the packet 1285 */ 1286 if (c->mfc_un.unres.unresolved.qlen > 3) { 1287 kfree_skb(skb); 1288 err = -ENOBUFS; 1289 } else { 1290 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1291 err = 0; 1292 } 1293 1294 spin_unlock_bh(&mfc_unres_lock); 1295 return err; 1296 } 1297 1298 /* 1299 * MFC6 cache manipulation by user space 1300 */ 1301 1302 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc, 1303 int parent) 1304 { 1305 int line; 1306 struct mfc6_cache *c, *next; 1307 1308 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1309 1310 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) { 1311 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1312 ipv6_addr_equal(&c->mf6c_mcastgrp, 1313 &mfc->mf6cc_mcastgrp.sin6_addr) && 1314 (parent == -1 || parent == c->mf6c_parent)) { 1315 write_lock_bh(&mrt_lock); 1316 list_del(&c->list); 1317 write_unlock_bh(&mrt_lock); 1318 1319 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1320 ip6mr_cache_free(c); 1321 return 0; 1322 } 1323 } 1324 return -ENOENT; 1325 } 1326 1327 static int ip6mr_device_event(struct notifier_block *this, 1328 unsigned long event, void *ptr) 1329 { 1330 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1331 struct net *net = dev_net(dev); 1332 struct mr6_table *mrt; 1333 struct mif_device *v; 1334 int ct; 1335 1336 if (event != NETDEV_UNREGISTER) 1337 return NOTIFY_DONE; 1338 1339 ip6mr_for_each_table(mrt, net) { 1340 v = &mrt->vif6_table[0]; 1341 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1342 if (v->dev == dev) 1343 mif6_delete(mrt, ct, 1, NULL); 1344 } 1345 } 1346 1347 return NOTIFY_DONE; 1348 } 1349 1350 static struct notifier_block ip6_mr_notifier = { 1351 .notifier_call = ip6mr_device_event 1352 }; 1353 1354 /* 1355 * Setup for IP multicast routing 1356 */ 1357 1358 static int __net_init ip6mr_net_init(struct net *net) 1359 { 1360 int err; 1361 1362 err = ip6mr_rules_init(net); 1363 if (err < 0) 1364 goto fail; 1365 1366 #ifdef CONFIG_PROC_FS 1367 err = -ENOMEM; 1368 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops)) 1369 goto proc_vif_fail; 1370 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops)) 1371 goto proc_cache_fail; 1372 #endif 1373 1374 return 0; 1375 1376 #ifdef CONFIG_PROC_FS 1377 proc_cache_fail: 1378 remove_proc_entry("ip6_mr_vif", net->proc_net); 1379 proc_vif_fail: 1380 ip6mr_rules_exit(net); 1381 #endif 1382 fail: 1383 return err; 1384 } 1385 1386 static void __net_exit ip6mr_net_exit(struct net *net) 1387 { 1388 #ifdef CONFIG_PROC_FS 1389 remove_proc_entry("ip6_mr_cache", net->proc_net); 1390 remove_proc_entry("ip6_mr_vif", net->proc_net); 1391 #endif 1392 ip6mr_rules_exit(net); 1393 } 1394 1395 static struct pernet_operations ip6mr_net_ops = { 1396 .init = ip6mr_net_init, 1397 .exit = ip6mr_net_exit, 1398 }; 1399 1400 int __init ip6_mr_init(void) 1401 { 1402 int err; 1403 1404 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1405 sizeof(struct mfc6_cache), 1406 0, SLAB_HWCACHE_ALIGN, 1407 NULL); 1408 if (!mrt_cachep) 1409 return -ENOMEM; 1410 1411 err = register_pernet_subsys(&ip6mr_net_ops); 1412 if (err) 1413 goto reg_pernet_fail; 1414 1415 err = register_netdevice_notifier(&ip6_mr_notifier); 1416 if (err) 1417 goto reg_notif_fail; 1418 #ifdef CONFIG_IPV6_PIMSM_V2 1419 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1420 pr_err("%s: can't add PIM protocol\n", __func__); 1421 err = -EAGAIN; 1422 goto add_proto_fail; 1423 } 1424 #endif 1425 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, 1426 ip6mr_rtm_dumproute, NULL); 1427 return 0; 1428 #ifdef CONFIG_IPV6_PIMSM_V2 1429 add_proto_fail: 1430 unregister_netdevice_notifier(&ip6_mr_notifier); 1431 #endif 1432 reg_notif_fail: 1433 unregister_pernet_subsys(&ip6mr_net_ops); 1434 reg_pernet_fail: 1435 kmem_cache_destroy(mrt_cachep); 1436 return err; 1437 } 1438 1439 void ip6_mr_cleanup(void) 1440 { 1441 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1442 #ifdef CONFIG_IPV6_PIMSM_V2 1443 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1444 #endif 1445 unregister_netdevice_notifier(&ip6_mr_notifier); 1446 unregister_pernet_subsys(&ip6mr_net_ops); 1447 kmem_cache_destroy(mrt_cachep); 1448 } 1449 1450 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1451 struct mf6cctl *mfc, int mrtsock, int parent) 1452 { 1453 bool found = false; 1454 int line; 1455 struct mfc6_cache *uc, *c; 1456 unsigned char ttls[MAXMIFS]; 1457 int i; 1458 1459 if (mfc->mf6cc_parent >= MAXMIFS) 1460 return -ENFILE; 1461 1462 memset(ttls, 255, MAXMIFS); 1463 for (i = 0; i < MAXMIFS; i++) { 1464 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1465 ttls[i] = 1; 1466 1467 } 1468 1469 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1470 1471 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1472 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1473 ipv6_addr_equal(&c->mf6c_mcastgrp, 1474 &mfc->mf6cc_mcastgrp.sin6_addr) && 1475 (parent == -1 || parent == mfc->mf6cc_parent)) { 1476 found = true; 1477 break; 1478 } 1479 } 1480 1481 if (found) { 1482 write_lock_bh(&mrt_lock); 1483 c->mf6c_parent = mfc->mf6cc_parent; 1484 ip6mr_update_thresholds(mrt, c, ttls); 1485 if (!mrtsock) 1486 c->mfc_flags |= MFC_STATIC; 1487 write_unlock_bh(&mrt_lock); 1488 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1489 return 0; 1490 } 1491 1492 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1493 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1494 return -EINVAL; 1495 1496 c = ip6mr_cache_alloc(); 1497 if (!c) 1498 return -ENOMEM; 1499 1500 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1501 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1502 c->mf6c_parent = mfc->mf6cc_parent; 1503 ip6mr_update_thresholds(mrt, c, ttls); 1504 if (!mrtsock) 1505 c->mfc_flags |= MFC_STATIC; 1506 1507 write_lock_bh(&mrt_lock); 1508 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1509 write_unlock_bh(&mrt_lock); 1510 1511 /* 1512 * Check to see if we resolved a queued list. If so we 1513 * need to send on the frames and tidy up. 1514 */ 1515 found = false; 1516 spin_lock_bh(&mfc_unres_lock); 1517 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1518 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1519 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1520 list_del(&uc->list); 1521 atomic_dec(&mrt->cache_resolve_queue_len); 1522 found = true; 1523 break; 1524 } 1525 } 1526 if (list_empty(&mrt->mfc6_unres_queue)) 1527 del_timer(&mrt->ipmr_expire_timer); 1528 spin_unlock_bh(&mfc_unres_lock); 1529 1530 if (found) { 1531 ip6mr_cache_resolve(net, mrt, uc, c); 1532 ip6mr_cache_free(uc); 1533 } 1534 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1535 return 0; 1536 } 1537 1538 /* 1539 * Close the multicast socket, and clear the vif tables etc 1540 */ 1541 1542 static void mroute_clean_tables(struct mr6_table *mrt, bool all) 1543 { 1544 int i; 1545 LIST_HEAD(list); 1546 struct mfc6_cache *c, *next; 1547 1548 /* 1549 * Shut down all active vif entries 1550 */ 1551 for (i = 0; i < mrt->maxvif; i++) { 1552 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) 1553 continue; 1554 mif6_delete(mrt, i, 0, &list); 1555 } 1556 unregister_netdevice_many(&list); 1557 1558 /* 1559 * Wipe the cache 1560 */ 1561 for (i = 0; i < MFC6_LINES; i++) { 1562 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1563 if (!all && (c->mfc_flags & MFC_STATIC)) 1564 continue; 1565 write_lock_bh(&mrt_lock); 1566 list_del(&c->list); 1567 write_unlock_bh(&mrt_lock); 1568 1569 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1570 ip6mr_cache_free(c); 1571 } 1572 } 1573 1574 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1575 spin_lock_bh(&mfc_unres_lock); 1576 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1577 list_del(&c->list); 1578 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1579 ip6mr_destroy_unres(mrt, c); 1580 } 1581 spin_unlock_bh(&mfc_unres_lock); 1582 } 1583 } 1584 1585 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1586 { 1587 int err = 0; 1588 struct net *net = sock_net(sk); 1589 1590 rtnl_lock(); 1591 write_lock_bh(&mrt_lock); 1592 if (likely(mrt->mroute6_sk == NULL)) { 1593 mrt->mroute6_sk = sk; 1594 net->ipv6.devconf_all->mc_forwarding++; 1595 } else { 1596 err = -EADDRINUSE; 1597 } 1598 write_unlock_bh(&mrt_lock); 1599 1600 if (!err) 1601 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1602 NETCONFA_MC_FORWARDING, 1603 NETCONFA_IFINDEX_ALL, 1604 net->ipv6.devconf_all); 1605 rtnl_unlock(); 1606 1607 return err; 1608 } 1609 1610 int ip6mr_sk_done(struct sock *sk) 1611 { 1612 int err = -EACCES; 1613 struct net *net = sock_net(sk); 1614 struct mr6_table *mrt; 1615 1616 rtnl_lock(); 1617 ip6mr_for_each_table(mrt, net) { 1618 if (sk == mrt->mroute6_sk) { 1619 write_lock_bh(&mrt_lock); 1620 mrt->mroute6_sk = NULL; 1621 net->ipv6.devconf_all->mc_forwarding--; 1622 write_unlock_bh(&mrt_lock); 1623 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 1624 NETCONFA_MC_FORWARDING, 1625 NETCONFA_IFINDEX_ALL, 1626 net->ipv6.devconf_all); 1627 1628 mroute_clean_tables(mrt, false); 1629 err = 0; 1630 break; 1631 } 1632 } 1633 rtnl_unlock(); 1634 1635 return err; 1636 } 1637 1638 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1639 { 1640 struct mr6_table *mrt; 1641 struct flowi6 fl6 = { 1642 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1643 .flowi6_oif = skb->dev->ifindex, 1644 .flowi6_mark = skb->mark, 1645 }; 1646 1647 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1648 return NULL; 1649 1650 return mrt->mroute6_sk; 1651 } 1652 1653 /* 1654 * Socket options and virtual interface manipulation. The whole 1655 * virtual interface system is a complete heap, but unfortunately 1656 * that's how BSD mrouted happens to think. Maybe one day with a proper 1657 * MOSPF/PIM router set up we can clean this up. 1658 */ 1659 1660 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1661 { 1662 int ret, parent = 0; 1663 struct mif6ctl vif; 1664 struct mf6cctl mfc; 1665 mifi_t mifi; 1666 struct net *net = sock_net(sk); 1667 struct mr6_table *mrt; 1668 1669 if (sk->sk_type != SOCK_RAW || 1670 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1671 return -EOPNOTSUPP; 1672 1673 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1674 if (!mrt) 1675 return -ENOENT; 1676 1677 if (optname != MRT6_INIT) { 1678 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1679 return -EACCES; 1680 } 1681 1682 switch (optname) { 1683 case MRT6_INIT: 1684 if (optlen < sizeof(int)) 1685 return -EINVAL; 1686 1687 return ip6mr_sk_init(mrt, sk); 1688 1689 case MRT6_DONE: 1690 return ip6mr_sk_done(sk); 1691 1692 case MRT6_ADD_MIF: 1693 if (optlen < sizeof(vif)) 1694 return -EINVAL; 1695 if (copy_from_user(&vif, optval, sizeof(vif))) 1696 return -EFAULT; 1697 if (vif.mif6c_mifi >= MAXMIFS) 1698 return -ENFILE; 1699 rtnl_lock(); 1700 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1701 rtnl_unlock(); 1702 return ret; 1703 1704 case MRT6_DEL_MIF: 1705 if (optlen < sizeof(mifi_t)) 1706 return -EINVAL; 1707 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1708 return -EFAULT; 1709 rtnl_lock(); 1710 ret = mif6_delete(mrt, mifi, 0, NULL); 1711 rtnl_unlock(); 1712 return ret; 1713 1714 /* 1715 * Manipulate the forwarding caches. These live 1716 * in a sort of kernel/user symbiosis. 1717 */ 1718 case MRT6_ADD_MFC: 1719 case MRT6_DEL_MFC: 1720 parent = -1; 1721 case MRT6_ADD_MFC_PROXY: 1722 case MRT6_DEL_MFC_PROXY: 1723 if (optlen < sizeof(mfc)) 1724 return -EINVAL; 1725 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1726 return -EFAULT; 1727 if (parent == 0) 1728 parent = mfc.mf6cc_parent; 1729 rtnl_lock(); 1730 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1731 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1732 else 1733 ret = ip6mr_mfc_add(net, mrt, &mfc, 1734 sk == mrt->mroute6_sk, parent); 1735 rtnl_unlock(); 1736 return ret; 1737 1738 /* 1739 * Control PIM assert (to activate pim will activate assert) 1740 */ 1741 case MRT6_ASSERT: 1742 { 1743 int v; 1744 1745 if (optlen != sizeof(v)) 1746 return -EINVAL; 1747 if (get_user(v, (int __user *)optval)) 1748 return -EFAULT; 1749 mrt->mroute_do_assert = v; 1750 return 0; 1751 } 1752 1753 #ifdef CONFIG_IPV6_PIMSM_V2 1754 case MRT6_PIM: 1755 { 1756 int v; 1757 1758 if (optlen != sizeof(v)) 1759 return -EINVAL; 1760 if (get_user(v, (int __user *)optval)) 1761 return -EFAULT; 1762 v = !!v; 1763 rtnl_lock(); 1764 ret = 0; 1765 if (v != mrt->mroute_do_pim) { 1766 mrt->mroute_do_pim = v; 1767 mrt->mroute_do_assert = v; 1768 } 1769 rtnl_unlock(); 1770 return ret; 1771 } 1772 1773 #endif 1774 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1775 case MRT6_TABLE: 1776 { 1777 u32 v; 1778 1779 if (optlen != sizeof(u32)) 1780 return -EINVAL; 1781 if (get_user(v, (u32 __user *)optval)) 1782 return -EFAULT; 1783 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1784 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1785 return -EINVAL; 1786 if (sk == mrt->mroute6_sk) 1787 return -EBUSY; 1788 1789 rtnl_lock(); 1790 ret = 0; 1791 if (!ip6mr_new_table(net, v)) 1792 ret = -ENOMEM; 1793 raw6_sk(sk)->ip6mr_table = v; 1794 rtnl_unlock(); 1795 return ret; 1796 } 1797 #endif 1798 /* 1799 * Spurious command, or MRT6_VERSION which you cannot 1800 * set. 1801 */ 1802 default: 1803 return -ENOPROTOOPT; 1804 } 1805 } 1806 1807 /* 1808 * Getsock opt support for the multicast routing system. 1809 */ 1810 1811 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1812 int __user *optlen) 1813 { 1814 int olr; 1815 int val; 1816 struct net *net = sock_net(sk); 1817 struct mr6_table *mrt; 1818 1819 if (sk->sk_type != SOCK_RAW || 1820 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1821 return -EOPNOTSUPP; 1822 1823 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1824 if (!mrt) 1825 return -ENOENT; 1826 1827 switch (optname) { 1828 case MRT6_VERSION: 1829 val = 0x0305; 1830 break; 1831 #ifdef CONFIG_IPV6_PIMSM_V2 1832 case MRT6_PIM: 1833 val = mrt->mroute_do_pim; 1834 break; 1835 #endif 1836 case MRT6_ASSERT: 1837 val = mrt->mroute_do_assert; 1838 break; 1839 default: 1840 return -ENOPROTOOPT; 1841 } 1842 1843 if (get_user(olr, optlen)) 1844 return -EFAULT; 1845 1846 olr = min_t(int, olr, sizeof(int)); 1847 if (olr < 0) 1848 return -EINVAL; 1849 1850 if (put_user(olr, optlen)) 1851 return -EFAULT; 1852 if (copy_to_user(optval, &val, olr)) 1853 return -EFAULT; 1854 return 0; 1855 } 1856 1857 /* 1858 * The IP multicast ioctl support routines. 1859 */ 1860 1861 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1862 { 1863 struct sioc_sg_req6 sr; 1864 struct sioc_mif_req6 vr; 1865 struct mif_device *vif; 1866 struct mfc6_cache *c; 1867 struct net *net = sock_net(sk); 1868 struct mr6_table *mrt; 1869 1870 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1871 if (!mrt) 1872 return -ENOENT; 1873 1874 switch (cmd) { 1875 case SIOCGETMIFCNT_IN6: 1876 if (copy_from_user(&vr, arg, sizeof(vr))) 1877 return -EFAULT; 1878 if (vr.mifi >= mrt->maxvif) 1879 return -EINVAL; 1880 read_lock(&mrt_lock); 1881 vif = &mrt->vif6_table[vr.mifi]; 1882 if (MIF_EXISTS(mrt, vr.mifi)) { 1883 vr.icount = vif->pkt_in; 1884 vr.ocount = vif->pkt_out; 1885 vr.ibytes = vif->bytes_in; 1886 vr.obytes = vif->bytes_out; 1887 read_unlock(&mrt_lock); 1888 1889 if (copy_to_user(arg, &vr, sizeof(vr))) 1890 return -EFAULT; 1891 return 0; 1892 } 1893 read_unlock(&mrt_lock); 1894 return -EADDRNOTAVAIL; 1895 case SIOCGETSGCNT_IN6: 1896 if (copy_from_user(&sr, arg, sizeof(sr))) 1897 return -EFAULT; 1898 1899 read_lock(&mrt_lock); 1900 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1901 if (c) { 1902 sr.pktcnt = c->mfc_un.res.pkt; 1903 sr.bytecnt = c->mfc_un.res.bytes; 1904 sr.wrong_if = c->mfc_un.res.wrong_if; 1905 read_unlock(&mrt_lock); 1906 1907 if (copy_to_user(arg, &sr, sizeof(sr))) 1908 return -EFAULT; 1909 return 0; 1910 } 1911 read_unlock(&mrt_lock); 1912 return -EADDRNOTAVAIL; 1913 default: 1914 return -ENOIOCTLCMD; 1915 } 1916 } 1917 1918 #ifdef CONFIG_COMPAT 1919 struct compat_sioc_sg_req6 { 1920 struct sockaddr_in6 src; 1921 struct sockaddr_in6 grp; 1922 compat_ulong_t pktcnt; 1923 compat_ulong_t bytecnt; 1924 compat_ulong_t wrong_if; 1925 }; 1926 1927 struct compat_sioc_mif_req6 { 1928 mifi_t mifi; 1929 compat_ulong_t icount; 1930 compat_ulong_t ocount; 1931 compat_ulong_t ibytes; 1932 compat_ulong_t obytes; 1933 }; 1934 1935 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1936 { 1937 struct compat_sioc_sg_req6 sr; 1938 struct compat_sioc_mif_req6 vr; 1939 struct mif_device *vif; 1940 struct mfc6_cache *c; 1941 struct net *net = sock_net(sk); 1942 struct mr6_table *mrt; 1943 1944 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1945 if (!mrt) 1946 return -ENOENT; 1947 1948 switch (cmd) { 1949 case SIOCGETMIFCNT_IN6: 1950 if (copy_from_user(&vr, arg, sizeof(vr))) 1951 return -EFAULT; 1952 if (vr.mifi >= mrt->maxvif) 1953 return -EINVAL; 1954 read_lock(&mrt_lock); 1955 vif = &mrt->vif6_table[vr.mifi]; 1956 if (MIF_EXISTS(mrt, vr.mifi)) { 1957 vr.icount = vif->pkt_in; 1958 vr.ocount = vif->pkt_out; 1959 vr.ibytes = vif->bytes_in; 1960 vr.obytes = vif->bytes_out; 1961 read_unlock(&mrt_lock); 1962 1963 if (copy_to_user(arg, &vr, sizeof(vr))) 1964 return -EFAULT; 1965 return 0; 1966 } 1967 read_unlock(&mrt_lock); 1968 return -EADDRNOTAVAIL; 1969 case SIOCGETSGCNT_IN6: 1970 if (copy_from_user(&sr, arg, sizeof(sr))) 1971 return -EFAULT; 1972 1973 read_lock(&mrt_lock); 1974 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1975 if (c) { 1976 sr.pktcnt = c->mfc_un.res.pkt; 1977 sr.bytecnt = c->mfc_un.res.bytes; 1978 sr.wrong_if = c->mfc_un.res.wrong_if; 1979 read_unlock(&mrt_lock); 1980 1981 if (copy_to_user(arg, &sr, sizeof(sr))) 1982 return -EFAULT; 1983 return 0; 1984 } 1985 read_unlock(&mrt_lock); 1986 return -EADDRNOTAVAIL; 1987 default: 1988 return -ENOIOCTLCMD; 1989 } 1990 } 1991 #endif 1992 1993 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 1994 { 1995 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 1996 IPSTATS_MIB_OUTFORWDATAGRAMS); 1997 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)), 1998 IPSTATS_MIB_OUTOCTETS, skb->len); 1999 return dst_output(net, sk, skb); 2000 } 2001 2002 /* 2003 * Processing handlers for ip6mr_forward 2004 */ 2005 2006 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 2007 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 2008 { 2009 struct ipv6hdr *ipv6h; 2010 struct mif_device *vif = &mrt->vif6_table[vifi]; 2011 struct net_device *dev; 2012 struct dst_entry *dst; 2013 struct flowi6 fl6; 2014 2015 if (!vif->dev) 2016 goto out_free; 2017 2018 #ifdef CONFIG_IPV6_PIMSM_V2 2019 if (vif->flags & MIFF_REGISTER) { 2020 vif->pkt_out++; 2021 vif->bytes_out += skb->len; 2022 vif->dev->stats.tx_bytes += skb->len; 2023 vif->dev->stats.tx_packets++; 2024 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2025 goto out_free; 2026 } 2027 #endif 2028 2029 ipv6h = ipv6_hdr(skb); 2030 2031 fl6 = (struct flowi6) { 2032 .flowi6_oif = vif->link, 2033 .daddr = ipv6h->daddr, 2034 }; 2035 2036 dst = ip6_route_output(net, NULL, &fl6); 2037 if (dst->error) { 2038 dst_release(dst); 2039 goto out_free; 2040 } 2041 2042 skb_dst_drop(skb); 2043 skb_dst_set(skb, dst); 2044 2045 /* 2046 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2047 * not only before forwarding, but after forwarding on all output 2048 * interfaces. It is clear, if mrouter runs a multicasting 2049 * program, it should receive packets not depending to what interface 2050 * program is joined. 2051 * If we will not make it, the program will have to join on all 2052 * interfaces. On the other hand, multihoming host (or router, but 2053 * not mrouter) cannot join to more than one interface - it will 2054 * result in receiving multiple packets. 2055 */ 2056 dev = vif->dev; 2057 skb->dev = dev; 2058 vif->pkt_out++; 2059 vif->bytes_out += skb->len; 2060 2061 /* We are about to write */ 2062 /* XXX: extension headers? */ 2063 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2064 goto out_free; 2065 2066 ipv6h = ipv6_hdr(skb); 2067 ipv6h->hop_limit--; 2068 2069 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2070 2071 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, 2072 net, NULL, skb, skb->dev, dev, 2073 ip6mr_forward2_finish); 2074 2075 out_free: 2076 kfree_skb(skb); 2077 return 0; 2078 } 2079 2080 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2081 { 2082 int ct; 2083 2084 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2085 if (mrt->vif6_table[ct].dev == dev) 2086 break; 2087 } 2088 return ct; 2089 } 2090 2091 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2092 struct sk_buff *skb, struct mfc6_cache *cache) 2093 { 2094 int psend = -1; 2095 int vif, ct; 2096 int true_vifi = ip6mr_find_vif(mrt, skb->dev); 2097 2098 vif = cache->mf6c_parent; 2099 cache->mfc_un.res.pkt++; 2100 cache->mfc_un.res.bytes += skb->len; 2101 cache->mfc_un.res.lastuse = jiffies; 2102 2103 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) { 2104 struct mfc6_cache *cache_proxy; 2105 2106 /* For an (*,G) entry, we only check that the incoming 2107 * interface is part of the static tree. 2108 */ 2109 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif); 2110 if (cache_proxy && 2111 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 2112 goto forward; 2113 } 2114 2115 /* 2116 * Wrong interface: drop packet and (maybe) send PIM assert. 2117 */ 2118 if (mrt->vif6_table[vif].dev != skb->dev) { 2119 cache->mfc_un.res.wrong_if++; 2120 2121 if (true_vifi >= 0 && mrt->mroute_do_assert && 2122 /* pimsm uses asserts, when switching from RPT to SPT, 2123 so that we cannot check that packet arrived on an oif. 2124 It is bad, but otherwise we would need to move pretty 2125 large chunk of pimd to kernel. Ough... --ANK 2126 */ 2127 (mrt->mroute_do_pim || 2128 cache->mfc_un.res.ttls[true_vifi] < 255) && 2129 time_after(jiffies, 2130 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2131 cache->mfc_un.res.last_assert = jiffies; 2132 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2133 } 2134 goto dont_forward; 2135 } 2136 2137 forward: 2138 mrt->vif6_table[vif].pkt_in++; 2139 mrt->vif6_table[vif].bytes_in += skb->len; 2140 2141 /* 2142 * Forward the frame 2143 */ 2144 if (ipv6_addr_any(&cache->mf6c_origin) && 2145 ipv6_addr_any(&cache->mf6c_mcastgrp)) { 2146 if (true_vifi >= 0 && 2147 true_vifi != cache->mf6c_parent && 2148 ipv6_hdr(skb)->hop_limit > 2149 cache->mfc_un.res.ttls[cache->mf6c_parent]) { 2150 /* It's an (*,*) entry and the packet is not coming from 2151 * the upstream: forward the packet to the upstream 2152 * only. 2153 */ 2154 psend = cache->mf6c_parent; 2155 goto last_forward; 2156 } 2157 goto dont_forward; 2158 } 2159 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2160 /* For (*,G) entry, don't forward to the incoming interface */ 2161 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) && 2162 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2163 if (psend != -1) { 2164 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2165 if (skb2) 2166 ip6mr_forward2(net, mrt, skb2, cache, psend); 2167 } 2168 psend = ct; 2169 } 2170 } 2171 last_forward: 2172 if (psend != -1) { 2173 ip6mr_forward2(net, mrt, skb, cache, psend); 2174 return; 2175 } 2176 2177 dont_forward: 2178 kfree_skb(skb); 2179 } 2180 2181 2182 /* 2183 * Multicast packets for forwarding arrive here 2184 */ 2185 2186 int ip6_mr_input(struct sk_buff *skb) 2187 { 2188 struct mfc6_cache *cache; 2189 struct net *net = dev_net(skb->dev); 2190 struct mr6_table *mrt; 2191 struct flowi6 fl6 = { 2192 .flowi6_iif = skb->dev->ifindex, 2193 .flowi6_mark = skb->mark, 2194 }; 2195 int err; 2196 2197 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2198 if (err < 0) { 2199 kfree_skb(skb); 2200 return err; 2201 } 2202 2203 read_lock(&mrt_lock); 2204 cache = ip6mr_cache_find(mrt, 2205 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2206 if (!cache) { 2207 int vif = ip6mr_find_vif(mrt, skb->dev); 2208 2209 if (vif >= 0) 2210 cache = ip6mr_cache_find_any(mrt, 2211 &ipv6_hdr(skb)->daddr, 2212 vif); 2213 } 2214 2215 /* 2216 * No usable cache entry 2217 */ 2218 if (!cache) { 2219 int vif; 2220 2221 vif = ip6mr_find_vif(mrt, skb->dev); 2222 if (vif >= 0) { 2223 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2224 read_unlock(&mrt_lock); 2225 2226 return err; 2227 } 2228 read_unlock(&mrt_lock); 2229 kfree_skb(skb); 2230 return -ENODEV; 2231 } 2232 2233 ip6_mr_forward(net, mrt, skb, cache); 2234 2235 read_unlock(&mrt_lock); 2236 2237 return 0; 2238 } 2239 2240 2241 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2242 struct mfc6_cache *c, struct rtmsg *rtm) 2243 { 2244 struct rta_mfc_stats mfcs; 2245 struct nlattr *mp_attr; 2246 struct rtnexthop *nhp; 2247 unsigned long lastuse; 2248 int ct; 2249 2250 /* If cache is unresolved, don't try to parse IIF and OIF */ 2251 if (c->mf6c_parent >= MAXMIFS) { 2252 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2253 return -ENOENT; 2254 } 2255 2256 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2257 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2258 return -EMSGSIZE; 2259 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2260 if (!mp_attr) 2261 return -EMSGSIZE; 2262 2263 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2264 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2265 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2266 if (!nhp) { 2267 nla_nest_cancel(skb, mp_attr); 2268 return -EMSGSIZE; 2269 } 2270 2271 nhp->rtnh_flags = 0; 2272 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2273 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2274 nhp->rtnh_len = sizeof(*nhp); 2275 } 2276 } 2277 2278 nla_nest_end(skb, mp_attr); 2279 2280 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2281 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2282 2283 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2284 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2285 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2286 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2287 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2288 RTA_PAD)) 2289 return -EMSGSIZE; 2290 2291 rtm->rtm_type = RTN_MULTICAST; 2292 return 1; 2293 } 2294 2295 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm, 2296 u32 portid) 2297 { 2298 int err; 2299 struct mr6_table *mrt; 2300 struct mfc6_cache *cache; 2301 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2302 2303 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2304 if (!mrt) 2305 return -ENOENT; 2306 2307 read_lock(&mrt_lock); 2308 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2309 if (!cache && skb->dev) { 2310 int vif = ip6mr_find_vif(mrt, skb->dev); 2311 2312 if (vif >= 0) 2313 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2314 vif); 2315 } 2316 2317 if (!cache) { 2318 struct sk_buff *skb2; 2319 struct ipv6hdr *iph; 2320 struct net_device *dev; 2321 int vif; 2322 2323 dev = skb->dev; 2324 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2325 read_unlock(&mrt_lock); 2326 return -ENODEV; 2327 } 2328 2329 /* really correct? */ 2330 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2331 if (!skb2) { 2332 read_unlock(&mrt_lock); 2333 return -ENOMEM; 2334 } 2335 2336 NETLINK_CB(skb2).portid = portid; 2337 skb_reset_transport_header(skb2); 2338 2339 skb_put(skb2, sizeof(struct ipv6hdr)); 2340 skb_reset_network_header(skb2); 2341 2342 iph = ipv6_hdr(skb2); 2343 iph->version = 0; 2344 iph->priority = 0; 2345 iph->flow_lbl[0] = 0; 2346 iph->flow_lbl[1] = 0; 2347 iph->flow_lbl[2] = 0; 2348 iph->payload_len = 0; 2349 iph->nexthdr = IPPROTO_NONE; 2350 iph->hop_limit = 0; 2351 iph->saddr = rt->rt6i_src.addr; 2352 iph->daddr = rt->rt6i_dst.addr; 2353 2354 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2355 read_unlock(&mrt_lock); 2356 2357 return err; 2358 } 2359 2360 if (rtm->rtm_flags & RTM_F_NOTIFY) 2361 cache->mfc_flags |= MFC_NOTIFY; 2362 2363 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2364 read_unlock(&mrt_lock); 2365 return err; 2366 } 2367 2368 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2369 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2370 int flags) 2371 { 2372 struct nlmsghdr *nlh; 2373 struct rtmsg *rtm; 2374 int err; 2375 2376 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2377 if (!nlh) 2378 return -EMSGSIZE; 2379 2380 rtm = nlmsg_data(nlh); 2381 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2382 rtm->rtm_dst_len = 128; 2383 rtm->rtm_src_len = 128; 2384 rtm->rtm_tos = 0; 2385 rtm->rtm_table = mrt->id; 2386 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2387 goto nla_put_failure; 2388 rtm->rtm_type = RTN_MULTICAST; 2389 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2390 if (c->mfc_flags & MFC_STATIC) 2391 rtm->rtm_protocol = RTPROT_STATIC; 2392 else 2393 rtm->rtm_protocol = RTPROT_MROUTED; 2394 rtm->rtm_flags = 0; 2395 2396 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2397 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2398 goto nla_put_failure; 2399 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2400 /* do not break the dump if cache is unresolved */ 2401 if (err < 0 && err != -ENOENT) 2402 goto nla_put_failure; 2403 2404 nlmsg_end(skb, nlh); 2405 return 0; 2406 2407 nla_put_failure: 2408 nlmsg_cancel(skb, nlh); 2409 return -EMSGSIZE; 2410 } 2411 2412 static int mr6_msgsize(bool unresolved, int maxvif) 2413 { 2414 size_t len = 2415 NLMSG_ALIGN(sizeof(struct rtmsg)) 2416 + nla_total_size(4) /* RTA_TABLE */ 2417 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2418 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2419 ; 2420 2421 if (!unresolved) 2422 len = len 2423 + nla_total_size(4) /* RTA_IIF */ 2424 + nla_total_size(0) /* RTA_MULTIPATH */ 2425 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2426 /* RTA_MFC_STATS */ 2427 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2428 ; 2429 2430 return len; 2431 } 2432 2433 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2434 int cmd) 2435 { 2436 struct net *net = read_pnet(&mrt->net); 2437 struct sk_buff *skb; 2438 int err = -ENOBUFS; 2439 2440 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2441 GFP_ATOMIC); 2442 if (!skb) 2443 goto errout; 2444 2445 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2446 if (err < 0) 2447 goto errout; 2448 2449 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2450 return; 2451 2452 errout: 2453 kfree_skb(skb); 2454 if (err < 0) 2455 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2456 } 2457 2458 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2459 { 2460 struct net *net = sock_net(skb->sk); 2461 struct mr6_table *mrt; 2462 struct mfc6_cache *mfc; 2463 unsigned int t = 0, s_t; 2464 unsigned int h = 0, s_h; 2465 unsigned int e = 0, s_e; 2466 2467 s_t = cb->args[0]; 2468 s_h = cb->args[1]; 2469 s_e = cb->args[2]; 2470 2471 read_lock(&mrt_lock); 2472 ip6mr_for_each_table(mrt, net) { 2473 if (t < s_t) 2474 goto next_table; 2475 if (t > s_t) 2476 s_h = 0; 2477 for (h = s_h; h < MFC6_LINES; h++) { 2478 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2479 if (e < s_e) 2480 goto next_entry; 2481 if (ip6mr_fill_mroute(mrt, skb, 2482 NETLINK_CB(cb->skb).portid, 2483 cb->nlh->nlmsg_seq, 2484 mfc, RTM_NEWROUTE, 2485 NLM_F_MULTI) < 0) 2486 goto done; 2487 next_entry: 2488 e++; 2489 } 2490 e = s_e = 0; 2491 } 2492 spin_lock_bh(&mfc_unres_lock); 2493 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2494 if (e < s_e) 2495 goto next_entry2; 2496 if (ip6mr_fill_mroute(mrt, skb, 2497 NETLINK_CB(cb->skb).portid, 2498 cb->nlh->nlmsg_seq, 2499 mfc, RTM_NEWROUTE, 2500 NLM_F_MULTI) < 0) { 2501 spin_unlock_bh(&mfc_unres_lock); 2502 goto done; 2503 } 2504 next_entry2: 2505 e++; 2506 } 2507 spin_unlock_bh(&mfc_unres_lock); 2508 e = s_e = 0; 2509 s_h = 0; 2510 next_table: 2511 t++; 2512 } 2513 done: 2514 read_unlock(&mrt_lock); 2515 2516 cb->args[2] = e; 2517 cb->args[1] = h; 2518 cb->args[0] = t; 2519 2520 return skb->len; 2521 } 2522