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 <asm/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 #ifdef CONFIG_NET_NS 60 struct net *net; 61 #endif 62 u32 id; 63 struct sock *mroute6_sk; 64 struct timer_list ipmr_expire_timer; 65 struct list_head mfc6_unres_queue; 66 struct list_head mfc6_cache_array[MFC6_LINES]; 67 struct mif_device vif6_table[MAXMIFS]; 68 int maxvif; 69 atomic_t cache_resolve_queue_len; 70 bool mroute_do_assert; 71 bool mroute_do_pim; 72 #ifdef CONFIG_IPV6_PIMSM_V2 73 int mroute_reg_vif_num; 74 #endif 75 }; 76 77 struct ip6mr_rule { 78 struct fib_rule common; 79 }; 80 81 struct ip6mr_result { 82 struct mr6_table *mrt; 83 }; 84 85 /* Big lock, protecting vif table, mrt cache and mroute socket state. 86 Note that the changes are semaphored via rtnl_lock. 87 */ 88 89 static DEFINE_RWLOCK(mrt_lock); 90 91 /* 92 * Multicast router control variables 93 */ 94 95 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL) 96 97 /* Special spinlock for queue of unresolved entries */ 98 static DEFINE_SPINLOCK(mfc_unres_lock); 99 100 /* We return to original Alan's scheme. Hash table of resolved 101 entries is changed only in process context and protected 102 with weak lock mrt_lock. Queue of unresolved entries is protected 103 with strong spinlock mfc_unres_lock. 104 105 In this case data path is free of exclusive locks at all. 106 */ 107 108 static struct kmem_cache *mrt_cachep __read_mostly; 109 110 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id); 111 static void ip6mr_free_table(struct mr6_table *mrt); 112 113 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt, 114 struct sk_buff *skb, struct mfc6_cache *cache); 115 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 116 mifi_t mifi, int assert); 117 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 118 struct mfc6_cache *c, struct rtmsg *rtm); 119 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 120 int cmd); 121 static int ip6mr_rtm_dumproute(struct sk_buff *skb, 122 struct netlink_callback *cb); 123 static void mroute_clean_tables(struct mr6_table *mrt); 124 static void ipmr_expire_process(unsigned long arg); 125 126 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 127 #define ip6mr_for_each_table(mrt, net) \ 128 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list) 129 130 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 131 { 132 struct mr6_table *mrt; 133 134 ip6mr_for_each_table(mrt, net) { 135 if (mrt->id == id) 136 return mrt; 137 } 138 return NULL; 139 } 140 141 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 142 struct mr6_table **mrt) 143 { 144 struct ip6mr_result res; 145 struct fib_lookup_arg arg = { .result = &res, }; 146 int err; 147 148 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, 149 flowi6_to_flowi(flp6), 0, &arg); 150 if (err < 0) 151 return err; 152 *mrt = res.mrt; 153 return 0; 154 } 155 156 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp, 157 int flags, struct fib_lookup_arg *arg) 158 { 159 struct ip6mr_result *res = arg->result; 160 struct mr6_table *mrt; 161 162 switch (rule->action) { 163 case FR_ACT_TO_TBL: 164 break; 165 case FR_ACT_UNREACHABLE: 166 return -ENETUNREACH; 167 case FR_ACT_PROHIBIT: 168 return -EACCES; 169 case FR_ACT_BLACKHOLE: 170 default: 171 return -EINVAL; 172 } 173 174 mrt = ip6mr_get_table(rule->fr_net, rule->table); 175 if (mrt == NULL) 176 return -EAGAIN; 177 res->mrt = mrt; 178 return 0; 179 } 180 181 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags) 182 { 183 return 1; 184 } 185 186 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = { 187 FRA_GENERIC_POLICY, 188 }; 189 190 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 191 struct fib_rule_hdr *frh, struct nlattr **tb) 192 { 193 return 0; 194 } 195 196 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 197 struct nlattr **tb) 198 { 199 return 1; 200 } 201 202 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 203 struct fib_rule_hdr *frh) 204 { 205 frh->dst_len = 0; 206 frh->src_len = 0; 207 frh->tos = 0; 208 return 0; 209 } 210 211 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = { 212 .family = RTNL_FAMILY_IP6MR, 213 .rule_size = sizeof(struct ip6mr_rule), 214 .addr_size = sizeof(struct in6_addr), 215 .action = ip6mr_rule_action, 216 .match = ip6mr_rule_match, 217 .configure = ip6mr_rule_configure, 218 .compare = ip6mr_rule_compare, 219 .default_pref = fib_default_rule_pref, 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 == NULL) { 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 kfree(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 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) { 263 list_del(&mrt->list); 264 ip6mr_free_table(mrt); 265 } 266 fib_rules_unregister(net->ipv6.mr6_rules_ops); 267 } 268 #else 269 #define ip6mr_for_each_table(mrt, net) \ 270 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL) 271 272 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 273 { 274 return net->ipv6.mrt6; 275 } 276 277 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 278 struct mr6_table **mrt) 279 { 280 *mrt = net->ipv6.mrt6; 281 return 0; 282 } 283 284 static int __net_init ip6mr_rules_init(struct net *net) 285 { 286 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT); 287 return net->ipv6.mrt6 ? 0 : -ENOMEM; 288 } 289 290 static void __net_exit ip6mr_rules_exit(struct net *net) 291 { 292 ip6mr_free_table(net->ipv6.mrt6); 293 } 294 #endif 295 296 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id) 297 { 298 struct mr6_table *mrt; 299 unsigned int i; 300 301 mrt = ip6mr_get_table(net, id); 302 if (mrt != NULL) 303 return mrt; 304 305 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 306 if (mrt == NULL) 307 return NULL; 308 mrt->id = id; 309 write_pnet(&mrt->net, net); 310 311 /* Forwarding cache */ 312 for (i = 0; i < MFC6_LINES; i++) 313 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]); 314 315 INIT_LIST_HEAD(&mrt->mfc6_unres_queue); 316 317 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 318 (unsigned long)mrt); 319 320 #ifdef CONFIG_IPV6_PIMSM_V2 321 mrt->mroute_reg_vif_num = -1; 322 #endif 323 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 324 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables); 325 #endif 326 return mrt; 327 } 328 329 static void ip6mr_free_table(struct mr6_table *mrt) 330 { 331 del_timer(&mrt->ipmr_expire_timer); 332 mroute_clean_tables(mrt); 333 kfree(mrt); 334 } 335 336 #ifdef CONFIG_PROC_FS 337 338 struct ipmr_mfc_iter { 339 struct seq_net_private p; 340 struct mr6_table *mrt; 341 struct list_head *cache; 342 int ct; 343 }; 344 345 346 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net, 347 struct ipmr_mfc_iter *it, loff_t pos) 348 { 349 struct mr6_table *mrt = it->mrt; 350 struct mfc6_cache *mfc; 351 352 read_lock(&mrt_lock); 353 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) { 354 it->cache = &mrt->mfc6_cache_array[it->ct]; 355 list_for_each_entry(mfc, it->cache, list) 356 if (pos-- == 0) 357 return mfc; 358 } 359 read_unlock(&mrt_lock); 360 361 spin_lock_bh(&mfc_unres_lock); 362 it->cache = &mrt->mfc6_unres_queue; 363 list_for_each_entry(mfc, it->cache, list) 364 if (pos-- == 0) 365 return mfc; 366 spin_unlock_bh(&mfc_unres_lock); 367 368 it->cache = NULL; 369 return NULL; 370 } 371 372 /* 373 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif 374 */ 375 376 struct ipmr_vif_iter { 377 struct seq_net_private p; 378 struct mr6_table *mrt; 379 int ct; 380 }; 381 382 static struct mif_device *ip6mr_vif_seq_idx(struct net *net, 383 struct ipmr_vif_iter *iter, 384 loff_t pos) 385 { 386 struct mr6_table *mrt = iter->mrt; 387 388 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 389 if (!MIF_EXISTS(mrt, iter->ct)) 390 continue; 391 if (pos-- == 0) 392 return &mrt->vif6_table[iter->ct]; 393 } 394 return NULL; 395 } 396 397 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) 398 __acquires(mrt_lock) 399 { 400 struct ipmr_vif_iter *iter = seq->private; 401 struct net *net = seq_file_net(seq); 402 struct mr6_table *mrt; 403 404 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 405 if (mrt == NULL) 406 return ERR_PTR(-ENOENT); 407 408 iter->mrt = mrt; 409 410 read_lock(&mrt_lock); 411 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1) 412 : SEQ_START_TOKEN; 413 } 414 415 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 416 { 417 struct ipmr_vif_iter *iter = seq->private; 418 struct net *net = seq_file_net(seq); 419 struct mr6_table *mrt = iter->mrt; 420 421 ++*pos; 422 if (v == SEQ_START_TOKEN) 423 return ip6mr_vif_seq_idx(net, iter, 0); 424 425 while (++iter->ct < mrt->maxvif) { 426 if (!MIF_EXISTS(mrt, iter->ct)) 427 continue; 428 return &mrt->vif6_table[iter->ct]; 429 } 430 return NULL; 431 } 432 433 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) 434 __releases(mrt_lock) 435 { 436 read_unlock(&mrt_lock); 437 } 438 439 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) 440 { 441 struct ipmr_vif_iter *iter = seq->private; 442 struct mr6_table *mrt = iter->mrt; 443 444 if (v == SEQ_START_TOKEN) { 445 seq_puts(seq, 446 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); 447 } else { 448 const struct mif_device *vif = v; 449 const char *name = vif->dev ? vif->dev->name : "none"; 450 451 seq_printf(seq, 452 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", 453 vif - mrt->vif6_table, 454 name, vif->bytes_in, vif->pkt_in, 455 vif->bytes_out, vif->pkt_out, 456 vif->flags); 457 } 458 return 0; 459 } 460 461 static const struct seq_operations ip6mr_vif_seq_ops = { 462 .start = ip6mr_vif_seq_start, 463 .next = ip6mr_vif_seq_next, 464 .stop = ip6mr_vif_seq_stop, 465 .show = ip6mr_vif_seq_show, 466 }; 467 468 static int ip6mr_vif_open(struct inode *inode, struct file *file) 469 { 470 return seq_open_net(inode, file, &ip6mr_vif_seq_ops, 471 sizeof(struct ipmr_vif_iter)); 472 } 473 474 static const struct file_operations ip6mr_vif_fops = { 475 .owner = THIS_MODULE, 476 .open = ip6mr_vif_open, 477 .read = seq_read, 478 .llseek = seq_lseek, 479 .release = seq_release_net, 480 }; 481 482 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 483 { 484 struct ipmr_mfc_iter *it = seq->private; 485 struct net *net = seq_file_net(seq); 486 struct mr6_table *mrt; 487 488 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 489 if (mrt == NULL) 490 return ERR_PTR(-ENOENT); 491 492 it->mrt = mrt; 493 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 494 : SEQ_START_TOKEN; 495 } 496 497 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 498 { 499 struct mfc6_cache *mfc = v; 500 struct ipmr_mfc_iter *it = seq->private; 501 struct net *net = seq_file_net(seq); 502 struct mr6_table *mrt = it->mrt; 503 504 ++*pos; 505 506 if (v == SEQ_START_TOKEN) 507 return ipmr_mfc_seq_idx(net, seq->private, 0); 508 509 if (mfc->list.next != it->cache) 510 return list_entry(mfc->list.next, struct mfc6_cache, list); 511 512 if (it->cache == &mrt->mfc6_unres_queue) 513 goto end_of_list; 514 515 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]); 516 517 while (++it->ct < MFC6_LINES) { 518 it->cache = &mrt->mfc6_cache_array[it->ct]; 519 if (list_empty(it->cache)) 520 continue; 521 return list_first_entry(it->cache, struct mfc6_cache, list); 522 } 523 524 /* exhausted cache_array, show unresolved */ 525 read_unlock(&mrt_lock); 526 it->cache = &mrt->mfc6_unres_queue; 527 it->ct = 0; 528 529 spin_lock_bh(&mfc_unres_lock); 530 if (!list_empty(it->cache)) 531 return list_first_entry(it->cache, struct mfc6_cache, list); 532 533 end_of_list: 534 spin_unlock_bh(&mfc_unres_lock); 535 it->cache = NULL; 536 537 return NULL; 538 } 539 540 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 541 { 542 struct ipmr_mfc_iter *it = seq->private; 543 struct mr6_table *mrt = it->mrt; 544 545 if (it->cache == &mrt->mfc6_unres_queue) 546 spin_unlock_bh(&mfc_unres_lock); 547 else if (it->cache == mrt->mfc6_cache_array) 548 read_unlock(&mrt_lock); 549 } 550 551 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 552 { 553 int n; 554 555 if (v == SEQ_START_TOKEN) { 556 seq_puts(seq, 557 "Group " 558 "Origin " 559 "Iif Pkts Bytes Wrong Oifs\n"); 560 } else { 561 const struct mfc6_cache *mfc = v; 562 const struct ipmr_mfc_iter *it = seq->private; 563 struct mr6_table *mrt = it->mrt; 564 565 seq_printf(seq, "%pI6 %pI6 %-3hd", 566 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, 567 mfc->mf6c_parent); 568 569 if (it->cache != &mrt->mfc6_unres_queue) { 570 seq_printf(seq, " %8lu %8lu %8lu", 571 mfc->mfc_un.res.pkt, 572 mfc->mfc_un.res.bytes, 573 mfc->mfc_un.res.wrong_if); 574 for (n = mfc->mfc_un.res.minvif; 575 n < mfc->mfc_un.res.maxvif; n++) { 576 if (MIF_EXISTS(mrt, n) && 577 mfc->mfc_un.res.ttls[n] < 255) 578 seq_printf(seq, 579 " %2d:%-3d", 580 n, mfc->mfc_un.res.ttls[n]); 581 } 582 } else { 583 /* unresolved mfc_caches don't contain 584 * pkt, bytes and wrong_if values 585 */ 586 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 587 } 588 seq_putc(seq, '\n'); 589 } 590 return 0; 591 } 592 593 static const struct seq_operations ipmr_mfc_seq_ops = { 594 .start = ipmr_mfc_seq_start, 595 .next = ipmr_mfc_seq_next, 596 .stop = ipmr_mfc_seq_stop, 597 .show = ipmr_mfc_seq_show, 598 }; 599 600 static int ipmr_mfc_open(struct inode *inode, struct file *file) 601 { 602 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 603 sizeof(struct ipmr_mfc_iter)); 604 } 605 606 static const struct file_operations ip6mr_mfc_fops = { 607 .owner = THIS_MODULE, 608 .open = ipmr_mfc_open, 609 .read = seq_read, 610 .llseek = seq_lseek, 611 .release = seq_release_net, 612 }; 613 #endif 614 615 #ifdef CONFIG_IPV6_PIMSM_V2 616 617 static int pim6_rcv(struct sk_buff *skb) 618 { 619 struct pimreghdr *pim; 620 struct ipv6hdr *encap; 621 struct net_device *reg_dev = NULL; 622 struct net *net = dev_net(skb->dev); 623 struct mr6_table *mrt; 624 struct flowi6 fl6 = { 625 .flowi6_iif = skb->dev->ifindex, 626 .flowi6_mark = skb->mark, 627 }; 628 int reg_vif_num; 629 630 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) 631 goto drop; 632 633 pim = (struct pimreghdr *)skb_transport_header(skb); 634 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || 635 (pim->flags & PIM_NULL_REGISTER) || 636 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 637 sizeof(*pim), IPPROTO_PIM, 638 csum_partial((void *)pim, sizeof(*pim), 0)) && 639 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 640 goto drop; 641 642 /* check if the inner packet is destined to mcast group */ 643 encap = (struct ipv6hdr *)(skb_transport_header(skb) + 644 sizeof(*pim)); 645 646 if (!ipv6_addr_is_multicast(&encap->daddr) || 647 encap->payload_len == 0 || 648 ntohs(encap->payload_len) + sizeof(*pim) > skb->len) 649 goto drop; 650 651 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 652 goto drop; 653 reg_vif_num = mrt->mroute_reg_vif_num; 654 655 read_lock(&mrt_lock); 656 if (reg_vif_num >= 0) 657 reg_dev = mrt->vif6_table[reg_vif_num].dev; 658 if (reg_dev) 659 dev_hold(reg_dev); 660 read_unlock(&mrt_lock); 661 662 if (reg_dev == NULL) 663 goto drop; 664 665 skb->mac_header = skb->network_header; 666 skb_pull(skb, (u8 *)encap - skb->data); 667 skb_reset_network_header(skb); 668 skb->protocol = htons(ETH_P_IPV6); 669 skb->ip_summed = CHECKSUM_NONE; 670 skb->pkt_type = PACKET_HOST; 671 672 skb_tunnel_rx(skb, reg_dev); 673 674 netif_rx(skb); 675 676 dev_put(reg_dev); 677 return 0; 678 drop: 679 kfree_skb(skb); 680 return 0; 681 } 682 683 static const struct inet6_protocol pim6_protocol = { 684 .handler = pim6_rcv, 685 }; 686 687 /* Service routines creating virtual interfaces: PIMREG */ 688 689 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, 690 struct net_device *dev) 691 { 692 struct net *net = dev_net(dev); 693 struct mr6_table *mrt; 694 struct flowi6 fl6 = { 695 .flowi6_oif = dev->ifindex, 696 .flowi6_iif = skb->skb_iif, 697 .flowi6_mark = skb->mark, 698 }; 699 int err; 700 701 err = ip6mr_fib_lookup(net, &fl6, &mrt); 702 if (err < 0) { 703 kfree_skb(skb); 704 return err; 705 } 706 707 read_lock(&mrt_lock); 708 dev->stats.tx_bytes += skb->len; 709 dev->stats.tx_packets++; 710 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT); 711 read_unlock(&mrt_lock); 712 kfree_skb(skb); 713 return NETDEV_TX_OK; 714 } 715 716 static const struct net_device_ops reg_vif_netdev_ops = { 717 .ndo_start_xmit = reg_vif_xmit, 718 }; 719 720 static void reg_vif_setup(struct net_device *dev) 721 { 722 dev->type = ARPHRD_PIMREG; 723 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; 724 dev->flags = IFF_NOARP; 725 dev->netdev_ops = ®_vif_netdev_ops; 726 dev->destructor = free_netdev; 727 dev->features |= NETIF_F_NETNS_LOCAL; 728 } 729 730 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt) 731 { 732 struct net_device *dev; 733 char name[IFNAMSIZ]; 734 735 if (mrt->id == RT6_TABLE_DFLT) 736 sprintf(name, "pim6reg"); 737 else 738 sprintf(name, "pim6reg%u", mrt->id); 739 740 dev = alloc_netdev(0, name, reg_vif_setup); 741 if (dev == NULL) 742 return NULL; 743 744 dev_net_set(dev, net); 745 746 if (register_netdevice(dev)) { 747 free_netdev(dev); 748 return NULL; 749 } 750 dev->iflink = 0; 751 752 if (dev_open(dev)) 753 goto failure; 754 755 dev_hold(dev); 756 return dev; 757 758 failure: 759 /* allow the register to be completed before unregistering. */ 760 rtnl_unlock(); 761 rtnl_lock(); 762 763 unregister_netdevice(dev); 764 return NULL; 765 } 766 #endif 767 768 /* 769 * Delete a VIF entry 770 */ 771 772 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head) 773 { 774 struct mif_device *v; 775 struct net_device *dev; 776 struct inet6_dev *in6_dev; 777 778 if (vifi < 0 || vifi >= mrt->maxvif) 779 return -EADDRNOTAVAIL; 780 781 v = &mrt->vif6_table[vifi]; 782 783 write_lock_bh(&mrt_lock); 784 dev = v->dev; 785 v->dev = NULL; 786 787 if (!dev) { 788 write_unlock_bh(&mrt_lock); 789 return -EADDRNOTAVAIL; 790 } 791 792 #ifdef CONFIG_IPV6_PIMSM_V2 793 if (vifi == mrt->mroute_reg_vif_num) 794 mrt->mroute_reg_vif_num = -1; 795 #endif 796 797 if (vifi + 1 == mrt->maxvif) { 798 int tmp; 799 for (tmp = vifi - 1; tmp >= 0; tmp--) { 800 if (MIF_EXISTS(mrt, tmp)) 801 break; 802 } 803 mrt->maxvif = tmp + 1; 804 } 805 806 write_unlock_bh(&mrt_lock); 807 808 dev_set_allmulti(dev, -1); 809 810 in6_dev = __in6_dev_get(dev); 811 if (in6_dev) { 812 in6_dev->cnf.mc_forwarding--; 813 inet6_netconf_notify_devconf(dev_net(dev), 814 NETCONFA_MC_FORWARDING, 815 dev->ifindex, &in6_dev->cnf); 816 } 817 818 if (v->flags & MIFF_REGISTER) 819 unregister_netdevice_queue(dev, head); 820 821 dev_put(dev); 822 return 0; 823 } 824 825 static inline void ip6mr_cache_free(struct mfc6_cache *c) 826 { 827 kmem_cache_free(mrt_cachep, c); 828 } 829 830 /* Destroy an unresolved cache entry, killing queued skbs 831 and reporting error to netlink readers. 832 */ 833 834 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c) 835 { 836 struct net *net = read_pnet(&mrt->net); 837 struct sk_buff *skb; 838 839 atomic_dec(&mrt->cache_resolve_queue_len); 840 841 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { 842 if (ipv6_hdr(skb)->version == 0) { 843 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); 844 nlh->nlmsg_type = NLMSG_ERROR; 845 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); 846 skb_trim(skb, nlh->nlmsg_len); 847 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; 848 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 849 } else 850 kfree_skb(skb); 851 } 852 853 ip6mr_cache_free(c); 854 } 855 856 857 /* Timer process for all the unresolved queue. */ 858 859 static void ipmr_do_expire_process(struct mr6_table *mrt) 860 { 861 unsigned long now = jiffies; 862 unsigned long expires = 10 * HZ; 863 struct mfc6_cache *c, *next; 864 865 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 866 if (time_after(c->mfc_un.unres.expires, now)) { 867 /* not yet... */ 868 unsigned long interval = c->mfc_un.unres.expires - now; 869 if (interval < expires) 870 expires = interval; 871 continue; 872 } 873 874 list_del(&c->list); 875 mr6_netlink_event(mrt, c, RTM_DELROUTE); 876 ip6mr_destroy_unres(mrt, c); 877 } 878 879 if (!list_empty(&mrt->mfc6_unres_queue)) 880 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 881 } 882 883 static void ipmr_expire_process(unsigned long arg) 884 { 885 struct mr6_table *mrt = (struct mr6_table *)arg; 886 887 if (!spin_trylock(&mfc_unres_lock)) { 888 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 889 return; 890 } 891 892 if (!list_empty(&mrt->mfc6_unres_queue)) 893 ipmr_do_expire_process(mrt); 894 895 spin_unlock(&mfc_unres_lock); 896 } 897 898 /* Fill oifs list. It is called under write locked mrt_lock. */ 899 900 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache, 901 unsigned char *ttls) 902 { 903 int vifi; 904 905 cache->mfc_un.res.minvif = MAXMIFS; 906 cache->mfc_un.res.maxvif = 0; 907 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 908 909 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 910 if (MIF_EXISTS(mrt, vifi) && 911 ttls[vifi] && ttls[vifi] < 255) { 912 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 913 if (cache->mfc_un.res.minvif > vifi) 914 cache->mfc_un.res.minvif = vifi; 915 if (cache->mfc_un.res.maxvif <= vifi) 916 cache->mfc_un.res.maxvif = vifi + 1; 917 } 918 } 919 } 920 921 static int mif6_add(struct net *net, struct mr6_table *mrt, 922 struct mif6ctl *vifc, int mrtsock) 923 { 924 int vifi = vifc->mif6c_mifi; 925 struct mif_device *v = &mrt->vif6_table[vifi]; 926 struct net_device *dev; 927 struct inet6_dev *in6_dev; 928 int err; 929 930 /* Is vif busy ? */ 931 if (MIF_EXISTS(mrt, vifi)) 932 return -EADDRINUSE; 933 934 switch (vifc->mif6c_flags) { 935 #ifdef CONFIG_IPV6_PIMSM_V2 936 case MIFF_REGISTER: 937 /* 938 * Special Purpose VIF in PIM 939 * All the packets will be sent to the daemon 940 */ 941 if (mrt->mroute_reg_vif_num >= 0) 942 return -EADDRINUSE; 943 dev = ip6mr_reg_vif(net, mrt); 944 if (!dev) 945 return -ENOBUFS; 946 err = dev_set_allmulti(dev, 1); 947 if (err) { 948 unregister_netdevice(dev); 949 dev_put(dev); 950 return err; 951 } 952 break; 953 #endif 954 case 0: 955 dev = dev_get_by_index(net, vifc->mif6c_pifi); 956 if (!dev) 957 return -EADDRNOTAVAIL; 958 err = dev_set_allmulti(dev, 1); 959 if (err) { 960 dev_put(dev); 961 return err; 962 } 963 break; 964 default: 965 return -EINVAL; 966 } 967 968 in6_dev = __in6_dev_get(dev); 969 if (in6_dev) { 970 in6_dev->cnf.mc_forwarding++; 971 inet6_netconf_notify_devconf(dev_net(dev), 972 NETCONFA_MC_FORWARDING, 973 dev->ifindex, &in6_dev->cnf); 974 } 975 976 /* 977 * Fill in the VIF structures 978 */ 979 v->rate_limit = vifc->vifc_rate_limit; 980 v->flags = vifc->mif6c_flags; 981 if (!mrtsock) 982 v->flags |= VIFF_STATIC; 983 v->threshold = vifc->vifc_threshold; 984 v->bytes_in = 0; 985 v->bytes_out = 0; 986 v->pkt_in = 0; 987 v->pkt_out = 0; 988 v->link = dev->ifindex; 989 if (v->flags & MIFF_REGISTER) 990 v->link = dev->iflink; 991 992 /* And finish update writing critical data */ 993 write_lock_bh(&mrt_lock); 994 v->dev = dev; 995 #ifdef CONFIG_IPV6_PIMSM_V2 996 if (v->flags & MIFF_REGISTER) 997 mrt->mroute_reg_vif_num = vifi; 998 #endif 999 if (vifi + 1 > mrt->maxvif) 1000 mrt->maxvif = vifi + 1; 1001 write_unlock_bh(&mrt_lock); 1002 return 0; 1003 } 1004 1005 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt, 1006 const struct in6_addr *origin, 1007 const struct in6_addr *mcastgrp) 1008 { 1009 int line = MFC6_HASH(mcastgrp, origin); 1010 struct mfc6_cache *c; 1011 1012 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1013 if (ipv6_addr_equal(&c->mf6c_origin, origin) && 1014 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) 1015 return c; 1016 } 1017 return NULL; 1018 } 1019 1020 /* 1021 * Allocate a multicast cache entry 1022 */ 1023 static struct mfc6_cache *ip6mr_cache_alloc(void) 1024 { 1025 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 1026 if (c == NULL) 1027 return NULL; 1028 c->mfc_un.res.minvif = MAXMIFS; 1029 return c; 1030 } 1031 1032 static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 1033 { 1034 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 1035 if (c == NULL) 1036 return NULL; 1037 skb_queue_head_init(&c->mfc_un.unres.unresolved); 1038 c->mfc_un.unres.expires = jiffies + 10 * HZ; 1039 return c; 1040 } 1041 1042 /* 1043 * A cache entry has gone into a resolved state from queued 1044 */ 1045 1046 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt, 1047 struct mfc6_cache *uc, struct mfc6_cache *c) 1048 { 1049 struct sk_buff *skb; 1050 1051 /* 1052 * Play the pending entries through our router 1053 */ 1054 1055 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 1056 if (ipv6_hdr(skb)->version == 0) { 1057 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); 1058 1059 if (__ip6mr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) { 1060 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1061 } else { 1062 nlh->nlmsg_type = NLMSG_ERROR; 1063 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); 1064 skb_trim(skb, nlh->nlmsg_len); 1065 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; 1066 } 1067 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1068 } else 1069 ip6_mr_forward(net, mrt, skb, c); 1070 } 1071 } 1072 1073 /* 1074 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd 1075 * expects the following bizarre scheme. 1076 * 1077 * Called under mrt_lock. 1078 */ 1079 1080 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 1081 mifi_t mifi, int assert) 1082 { 1083 struct sk_buff *skb; 1084 struct mrt6msg *msg; 1085 int ret; 1086 1087 #ifdef CONFIG_IPV6_PIMSM_V2 1088 if (assert == MRT6MSG_WHOLEPKT) 1089 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1090 +sizeof(*msg)); 1091 else 1092 #endif 1093 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1094 1095 if (!skb) 1096 return -ENOBUFS; 1097 1098 /* I suppose that internal messages 1099 * do not require checksums */ 1100 1101 skb->ip_summed = CHECKSUM_UNNECESSARY; 1102 1103 #ifdef CONFIG_IPV6_PIMSM_V2 1104 if (assert == MRT6MSG_WHOLEPKT) { 1105 /* Ugly, but we have no choice with this interface. 1106 Duplicate old header, fix length etc. 1107 And all this only to mangle msg->im6_msgtype and 1108 to set msg->im6_mbz to "mbz" :-) 1109 */ 1110 skb_push(skb, -skb_network_offset(pkt)); 1111 1112 skb_push(skb, sizeof(*msg)); 1113 skb_reset_transport_header(skb); 1114 msg = (struct mrt6msg *)skb_transport_header(skb); 1115 msg->im6_mbz = 0; 1116 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1117 msg->im6_mif = mrt->mroute_reg_vif_num; 1118 msg->im6_pad = 0; 1119 msg->im6_src = ipv6_hdr(pkt)->saddr; 1120 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1121 1122 skb->ip_summed = CHECKSUM_UNNECESSARY; 1123 } else 1124 #endif 1125 { 1126 /* 1127 * Copy the IP header 1128 */ 1129 1130 skb_put(skb, sizeof(struct ipv6hdr)); 1131 skb_reset_network_header(skb); 1132 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1133 1134 /* 1135 * Add our header 1136 */ 1137 skb_put(skb, sizeof(*msg)); 1138 skb_reset_transport_header(skb); 1139 msg = (struct mrt6msg *)skb_transport_header(skb); 1140 1141 msg->im6_mbz = 0; 1142 msg->im6_msgtype = assert; 1143 msg->im6_mif = mifi; 1144 msg->im6_pad = 0; 1145 msg->im6_src = ipv6_hdr(pkt)->saddr; 1146 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1147 1148 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1149 skb->ip_summed = CHECKSUM_UNNECESSARY; 1150 } 1151 1152 if (mrt->mroute6_sk == NULL) { 1153 kfree_skb(skb); 1154 return -EINVAL; 1155 } 1156 1157 /* 1158 * Deliver to user space multicast routing algorithms 1159 */ 1160 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb); 1161 if (ret < 0) { 1162 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1163 kfree_skb(skb); 1164 } 1165 1166 return ret; 1167 } 1168 1169 /* 1170 * Queue a packet for resolution. It gets locked cache entry! 1171 */ 1172 1173 static int 1174 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb) 1175 { 1176 bool found = false; 1177 int err; 1178 struct mfc6_cache *c; 1179 1180 spin_lock_bh(&mfc_unres_lock); 1181 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) { 1182 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1183 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1184 found = true; 1185 break; 1186 } 1187 } 1188 1189 if (!found) { 1190 /* 1191 * Create a new entry if allowable 1192 */ 1193 1194 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1195 (c = ip6mr_cache_alloc_unres()) == NULL) { 1196 spin_unlock_bh(&mfc_unres_lock); 1197 1198 kfree_skb(skb); 1199 return -ENOBUFS; 1200 } 1201 1202 /* 1203 * Fill in the new cache entry 1204 */ 1205 c->mf6c_parent = -1; 1206 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1207 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1208 1209 /* 1210 * Reflect first query at pim6sd 1211 */ 1212 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1213 if (err < 0) { 1214 /* If the report failed throw the cache entry 1215 out - Brad Parker 1216 */ 1217 spin_unlock_bh(&mfc_unres_lock); 1218 1219 ip6mr_cache_free(c); 1220 kfree_skb(skb); 1221 return err; 1222 } 1223 1224 atomic_inc(&mrt->cache_resolve_queue_len); 1225 list_add(&c->list, &mrt->mfc6_unres_queue); 1226 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1227 1228 ipmr_do_expire_process(mrt); 1229 } 1230 1231 /* 1232 * See if we can append the packet 1233 */ 1234 if (c->mfc_un.unres.unresolved.qlen > 3) { 1235 kfree_skb(skb); 1236 err = -ENOBUFS; 1237 } else { 1238 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1239 err = 0; 1240 } 1241 1242 spin_unlock_bh(&mfc_unres_lock); 1243 return err; 1244 } 1245 1246 /* 1247 * MFC6 cache manipulation by user space 1248 */ 1249 1250 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc) 1251 { 1252 int line; 1253 struct mfc6_cache *c, *next; 1254 1255 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1256 1257 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) { 1258 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1259 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) { 1260 write_lock_bh(&mrt_lock); 1261 list_del(&c->list); 1262 write_unlock_bh(&mrt_lock); 1263 1264 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1265 ip6mr_cache_free(c); 1266 return 0; 1267 } 1268 } 1269 return -ENOENT; 1270 } 1271 1272 static int ip6mr_device_event(struct notifier_block *this, 1273 unsigned long event, void *ptr) 1274 { 1275 struct net_device *dev = ptr; 1276 struct net *net = dev_net(dev); 1277 struct mr6_table *mrt; 1278 struct mif_device *v; 1279 int ct; 1280 LIST_HEAD(list); 1281 1282 if (event != NETDEV_UNREGISTER) 1283 return NOTIFY_DONE; 1284 1285 ip6mr_for_each_table(mrt, net) { 1286 v = &mrt->vif6_table[0]; 1287 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1288 if (v->dev == dev) 1289 mif6_delete(mrt, ct, &list); 1290 } 1291 } 1292 unregister_netdevice_many(&list); 1293 1294 return NOTIFY_DONE; 1295 } 1296 1297 static struct notifier_block ip6_mr_notifier = { 1298 .notifier_call = ip6mr_device_event 1299 }; 1300 1301 /* 1302 * Setup for IP multicast routing 1303 */ 1304 1305 static int __net_init ip6mr_net_init(struct net *net) 1306 { 1307 int err; 1308 1309 err = ip6mr_rules_init(net); 1310 if (err < 0) 1311 goto fail; 1312 1313 #ifdef CONFIG_PROC_FS 1314 err = -ENOMEM; 1315 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops)) 1316 goto proc_vif_fail; 1317 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops)) 1318 goto proc_cache_fail; 1319 #endif 1320 1321 return 0; 1322 1323 #ifdef CONFIG_PROC_FS 1324 proc_cache_fail: 1325 proc_net_remove(net, "ip6_mr_vif"); 1326 proc_vif_fail: 1327 ip6mr_rules_exit(net); 1328 #endif 1329 fail: 1330 return err; 1331 } 1332 1333 static void __net_exit ip6mr_net_exit(struct net *net) 1334 { 1335 #ifdef CONFIG_PROC_FS 1336 proc_net_remove(net, "ip6_mr_cache"); 1337 proc_net_remove(net, "ip6_mr_vif"); 1338 #endif 1339 ip6mr_rules_exit(net); 1340 } 1341 1342 static struct pernet_operations ip6mr_net_ops = { 1343 .init = ip6mr_net_init, 1344 .exit = ip6mr_net_exit, 1345 }; 1346 1347 int __init ip6_mr_init(void) 1348 { 1349 int err; 1350 1351 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1352 sizeof(struct mfc6_cache), 1353 0, SLAB_HWCACHE_ALIGN, 1354 NULL); 1355 if (!mrt_cachep) 1356 return -ENOMEM; 1357 1358 err = register_pernet_subsys(&ip6mr_net_ops); 1359 if (err) 1360 goto reg_pernet_fail; 1361 1362 err = register_netdevice_notifier(&ip6_mr_notifier); 1363 if (err) 1364 goto reg_notif_fail; 1365 #ifdef CONFIG_IPV6_PIMSM_V2 1366 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1367 pr_err("%s: can't add PIM protocol\n", __func__); 1368 err = -EAGAIN; 1369 goto add_proto_fail; 1370 } 1371 #endif 1372 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, 1373 ip6mr_rtm_dumproute, NULL); 1374 return 0; 1375 #ifdef CONFIG_IPV6_PIMSM_V2 1376 add_proto_fail: 1377 unregister_netdevice_notifier(&ip6_mr_notifier); 1378 #endif 1379 reg_notif_fail: 1380 unregister_pernet_subsys(&ip6mr_net_ops); 1381 reg_pernet_fail: 1382 kmem_cache_destroy(mrt_cachep); 1383 return err; 1384 } 1385 1386 void ip6_mr_cleanup(void) 1387 { 1388 unregister_netdevice_notifier(&ip6_mr_notifier); 1389 unregister_pernet_subsys(&ip6mr_net_ops); 1390 kmem_cache_destroy(mrt_cachep); 1391 } 1392 1393 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1394 struct mf6cctl *mfc, int mrtsock) 1395 { 1396 bool found = false; 1397 int line; 1398 struct mfc6_cache *uc, *c; 1399 unsigned char ttls[MAXMIFS]; 1400 int i; 1401 1402 if (mfc->mf6cc_parent >= MAXMIFS) 1403 return -ENFILE; 1404 1405 memset(ttls, 255, MAXMIFS); 1406 for (i = 0; i < MAXMIFS; i++) { 1407 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1408 ttls[i] = 1; 1409 1410 } 1411 1412 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1413 1414 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1415 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1416 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) { 1417 found = true; 1418 break; 1419 } 1420 } 1421 1422 if (found) { 1423 write_lock_bh(&mrt_lock); 1424 c->mf6c_parent = mfc->mf6cc_parent; 1425 ip6mr_update_thresholds(mrt, c, ttls); 1426 if (!mrtsock) 1427 c->mfc_flags |= MFC_STATIC; 1428 write_unlock_bh(&mrt_lock); 1429 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1430 return 0; 1431 } 1432 1433 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1434 return -EINVAL; 1435 1436 c = ip6mr_cache_alloc(); 1437 if (c == NULL) 1438 return -ENOMEM; 1439 1440 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1441 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1442 c->mf6c_parent = mfc->mf6cc_parent; 1443 ip6mr_update_thresholds(mrt, c, ttls); 1444 if (!mrtsock) 1445 c->mfc_flags |= MFC_STATIC; 1446 1447 write_lock_bh(&mrt_lock); 1448 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1449 write_unlock_bh(&mrt_lock); 1450 1451 /* 1452 * Check to see if we resolved a queued list. If so we 1453 * need to send on the frames and tidy up. 1454 */ 1455 found = false; 1456 spin_lock_bh(&mfc_unres_lock); 1457 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1458 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1459 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1460 list_del(&uc->list); 1461 atomic_dec(&mrt->cache_resolve_queue_len); 1462 found = true; 1463 break; 1464 } 1465 } 1466 if (list_empty(&mrt->mfc6_unres_queue)) 1467 del_timer(&mrt->ipmr_expire_timer); 1468 spin_unlock_bh(&mfc_unres_lock); 1469 1470 if (found) { 1471 ip6mr_cache_resolve(net, mrt, uc, c); 1472 ip6mr_cache_free(uc); 1473 } 1474 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1475 return 0; 1476 } 1477 1478 /* 1479 * Close the multicast socket, and clear the vif tables etc 1480 */ 1481 1482 static void mroute_clean_tables(struct mr6_table *mrt) 1483 { 1484 int i; 1485 LIST_HEAD(list); 1486 struct mfc6_cache *c, *next; 1487 1488 /* 1489 * Shut down all active vif entries 1490 */ 1491 for (i = 0; i < mrt->maxvif; i++) { 1492 if (!(mrt->vif6_table[i].flags & VIFF_STATIC)) 1493 mif6_delete(mrt, i, &list); 1494 } 1495 unregister_netdevice_many(&list); 1496 1497 /* 1498 * Wipe the cache 1499 */ 1500 for (i = 0; i < MFC6_LINES; i++) { 1501 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1502 if (c->mfc_flags & MFC_STATIC) 1503 continue; 1504 write_lock_bh(&mrt_lock); 1505 list_del(&c->list); 1506 write_unlock_bh(&mrt_lock); 1507 1508 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1509 ip6mr_cache_free(c); 1510 } 1511 } 1512 1513 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1514 spin_lock_bh(&mfc_unres_lock); 1515 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1516 list_del(&c->list); 1517 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1518 ip6mr_destroy_unres(mrt, c); 1519 } 1520 spin_unlock_bh(&mfc_unres_lock); 1521 } 1522 } 1523 1524 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1525 { 1526 int err = 0; 1527 struct net *net = sock_net(sk); 1528 1529 rtnl_lock(); 1530 write_lock_bh(&mrt_lock); 1531 if (likely(mrt->mroute6_sk == NULL)) { 1532 mrt->mroute6_sk = sk; 1533 net->ipv6.devconf_all->mc_forwarding++; 1534 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1535 NETCONFA_IFINDEX_ALL, 1536 net->ipv6.devconf_all); 1537 } 1538 else 1539 err = -EADDRINUSE; 1540 write_unlock_bh(&mrt_lock); 1541 1542 rtnl_unlock(); 1543 1544 return err; 1545 } 1546 1547 int ip6mr_sk_done(struct sock *sk) 1548 { 1549 int err = -EACCES; 1550 struct net *net = sock_net(sk); 1551 struct mr6_table *mrt; 1552 1553 rtnl_lock(); 1554 ip6mr_for_each_table(mrt, net) { 1555 if (sk == mrt->mroute6_sk) { 1556 write_lock_bh(&mrt_lock); 1557 mrt->mroute6_sk = NULL; 1558 net->ipv6.devconf_all->mc_forwarding--; 1559 inet6_netconf_notify_devconf(net, 1560 NETCONFA_MC_FORWARDING, 1561 NETCONFA_IFINDEX_ALL, 1562 net->ipv6.devconf_all); 1563 write_unlock_bh(&mrt_lock); 1564 1565 mroute_clean_tables(mrt); 1566 err = 0; 1567 break; 1568 } 1569 } 1570 rtnl_unlock(); 1571 1572 return err; 1573 } 1574 1575 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1576 { 1577 struct mr6_table *mrt; 1578 struct flowi6 fl6 = { 1579 .flowi6_iif = skb->skb_iif, 1580 .flowi6_oif = skb->dev->ifindex, 1581 .flowi6_mark = skb->mark, 1582 }; 1583 1584 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1585 return NULL; 1586 1587 return mrt->mroute6_sk; 1588 } 1589 1590 /* 1591 * Socket options and virtual interface manipulation. The whole 1592 * virtual interface system is a complete heap, but unfortunately 1593 * that's how BSD mrouted happens to think. Maybe one day with a proper 1594 * MOSPF/PIM router set up we can clean this up. 1595 */ 1596 1597 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1598 { 1599 int ret; 1600 struct mif6ctl vif; 1601 struct mf6cctl mfc; 1602 mifi_t mifi; 1603 struct net *net = sock_net(sk); 1604 struct mr6_table *mrt; 1605 1606 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1607 if (mrt == NULL) 1608 return -ENOENT; 1609 1610 if (optname != MRT6_INIT) { 1611 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1612 return -EACCES; 1613 } 1614 1615 switch (optname) { 1616 case MRT6_INIT: 1617 if (sk->sk_type != SOCK_RAW || 1618 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1619 return -EOPNOTSUPP; 1620 if (optlen < sizeof(int)) 1621 return -EINVAL; 1622 1623 return ip6mr_sk_init(mrt, sk); 1624 1625 case MRT6_DONE: 1626 return ip6mr_sk_done(sk); 1627 1628 case MRT6_ADD_MIF: 1629 if (optlen < sizeof(vif)) 1630 return -EINVAL; 1631 if (copy_from_user(&vif, optval, sizeof(vif))) 1632 return -EFAULT; 1633 if (vif.mif6c_mifi >= MAXMIFS) 1634 return -ENFILE; 1635 rtnl_lock(); 1636 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1637 rtnl_unlock(); 1638 return ret; 1639 1640 case MRT6_DEL_MIF: 1641 if (optlen < sizeof(mifi_t)) 1642 return -EINVAL; 1643 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1644 return -EFAULT; 1645 rtnl_lock(); 1646 ret = mif6_delete(mrt, mifi, NULL); 1647 rtnl_unlock(); 1648 return ret; 1649 1650 /* 1651 * Manipulate the forwarding caches. These live 1652 * in a sort of kernel/user symbiosis. 1653 */ 1654 case MRT6_ADD_MFC: 1655 case MRT6_DEL_MFC: 1656 if (optlen < sizeof(mfc)) 1657 return -EINVAL; 1658 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1659 return -EFAULT; 1660 rtnl_lock(); 1661 if (optname == MRT6_DEL_MFC) 1662 ret = ip6mr_mfc_delete(mrt, &mfc); 1663 else 1664 ret = ip6mr_mfc_add(net, mrt, &mfc, sk == mrt->mroute6_sk); 1665 rtnl_unlock(); 1666 return ret; 1667 1668 /* 1669 * Control PIM assert (to activate pim will activate assert) 1670 */ 1671 case MRT6_ASSERT: 1672 { 1673 int v; 1674 1675 if (optlen != sizeof(v)) 1676 return -EINVAL; 1677 if (get_user(v, (int __user *)optval)) 1678 return -EFAULT; 1679 mrt->mroute_do_assert = v; 1680 return 0; 1681 } 1682 1683 #ifdef CONFIG_IPV6_PIMSM_V2 1684 case MRT6_PIM: 1685 { 1686 int v; 1687 1688 if (optlen != sizeof(v)) 1689 return -EINVAL; 1690 if (get_user(v, (int __user *)optval)) 1691 return -EFAULT; 1692 v = !!v; 1693 rtnl_lock(); 1694 ret = 0; 1695 if (v != mrt->mroute_do_pim) { 1696 mrt->mroute_do_pim = v; 1697 mrt->mroute_do_assert = v; 1698 } 1699 rtnl_unlock(); 1700 return ret; 1701 } 1702 1703 #endif 1704 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1705 case MRT6_TABLE: 1706 { 1707 u32 v; 1708 1709 if (optlen != sizeof(u32)) 1710 return -EINVAL; 1711 if (get_user(v, (u32 __user *)optval)) 1712 return -EFAULT; 1713 if (sk == mrt->mroute6_sk) 1714 return -EBUSY; 1715 1716 rtnl_lock(); 1717 ret = 0; 1718 if (!ip6mr_new_table(net, v)) 1719 ret = -ENOMEM; 1720 raw6_sk(sk)->ip6mr_table = v; 1721 rtnl_unlock(); 1722 return ret; 1723 } 1724 #endif 1725 /* 1726 * Spurious command, or MRT6_VERSION which you cannot 1727 * set. 1728 */ 1729 default: 1730 return -ENOPROTOOPT; 1731 } 1732 } 1733 1734 /* 1735 * Getsock opt support for the multicast routing system. 1736 */ 1737 1738 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1739 int __user *optlen) 1740 { 1741 int olr; 1742 int val; 1743 struct net *net = sock_net(sk); 1744 struct mr6_table *mrt; 1745 1746 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1747 if (mrt == NULL) 1748 return -ENOENT; 1749 1750 switch (optname) { 1751 case MRT6_VERSION: 1752 val = 0x0305; 1753 break; 1754 #ifdef CONFIG_IPV6_PIMSM_V2 1755 case MRT6_PIM: 1756 val = mrt->mroute_do_pim; 1757 break; 1758 #endif 1759 case MRT6_ASSERT: 1760 val = mrt->mroute_do_assert; 1761 break; 1762 default: 1763 return -ENOPROTOOPT; 1764 } 1765 1766 if (get_user(olr, optlen)) 1767 return -EFAULT; 1768 1769 olr = min_t(int, olr, sizeof(int)); 1770 if (olr < 0) 1771 return -EINVAL; 1772 1773 if (put_user(olr, optlen)) 1774 return -EFAULT; 1775 if (copy_to_user(optval, &val, olr)) 1776 return -EFAULT; 1777 return 0; 1778 } 1779 1780 /* 1781 * The IP multicast ioctl support routines. 1782 */ 1783 1784 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1785 { 1786 struct sioc_sg_req6 sr; 1787 struct sioc_mif_req6 vr; 1788 struct mif_device *vif; 1789 struct mfc6_cache *c; 1790 struct net *net = sock_net(sk); 1791 struct mr6_table *mrt; 1792 1793 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1794 if (mrt == NULL) 1795 return -ENOENT; 1796 1797 switch (cmd) { 1798 case SIOCGETMIFCNT_IN6: 1799 if (copy_from_user(&vr, arg, sizeof(vr))) 1800 return -EFAULT; 1801 if (vr.mifi >= mrt->maxvif) 1802 return -EINVAL; 1803 read_lock(&mrt_lock); 1804 vif = &mrt->vif6_table[vr.mifi]; 1805 if (MIF_EXISTS(mrt, vr.mifi)) { 1806 vr.icount = vif->pkt_in; 1807 vr.ocount = vif->pkt_out; 1808 vr.ibytes = vif->bytes_in; 1809 vr.obytes = vif->bytes_out; 1810 read_unlock(&mrt_lock); 1811 1812 if (copy_to_user(arg, &vr, sizeof(vr))) 1813 return -EFAULT; 1814 return 0; 1815 } 1816 read_unlock(&mrt_lock); 1817 return -EADDRNOTAVAIL; 1818 case SIOCGETSGCNT_IN6: 1819 if (copy_from_user(&sr, arg, sizeof(sr))) 1820 return -EFAULT; 1821 1822 read_lock(&mrt_lock); 1823 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1824 if (c) { 1825 sr.pktcnt = c->mfc_un.res.pkt; 1826 sr.bytecnt = c->mfc_un.res.bytes; 1827 sr.wrong_if = c->mfc_un.res.wrong_if; 1828 read_unlock(&mrt_lock); 1829 1830 if (copy_to_user(arg, &sr, sizeof(sr))) 1831 return -EFAULT; 1832 return 0; 1833 } 1834 read_unlock(&mrt_lock); 1835 return -EADDRNOTAVAIL; 1836 default: 1837 return -ENOIOCTLCMD; 1838 } 1839 } 1840 1841 #ifdef CONFIG_COMPAT 1842 struct compat_sioc_sg_req6 { 1843 struct sockaddr_in6 src; 1844 struct sockaddr_in6 grp; 1845 compat_ulong_t pktcnt; 1846 compat_ulong_t bytecnt; 1847 compat_ulong_t wrong_if; 1848 }; 1849 1850 struct compat_sioc_mif_req6 { 1851 mifi_t mifi; 1852 compat_ulong_t icount; 1853 compat_ulong_t ocount; 1854 compat_ulong_t ibytes; 1855 compat_ulong_t obytes; 1856 }; 1857 1858 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1859 { 1860 struct compat_sioc_sg_req6 sr; 1861 struct compat_sioc_mif_req6 vr; 1862 struct mif_device *vif; 1863 struct mfc6_cache *c; 1864 struct net *net = sock_net(sk); 1865 struct mr6_table *mrt; 1866 1867 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1868 if (mrt == NULL) 1869 return -ENOENT; 1870 1871 switch (cmd) { 1872 case SIOCGETMIFCNT_IN6: 1873 if (copy_from_user(&vr, arg, sizeof(vr))) 1874 return -EFAULT; 1875 if (vr.mifi >= mrt->maxvif) 1876 return -EINVAL; 1877 read_lock(&mrt_lock); 1878 vif = &mrt->vif6_table[vr.mifi]; 1879 if (MIF_EXISTS(mrt, vr.mifi)) { 1880 vr.icount = vif->pkt_in; 1881 vr.ocount = vif->pkt_out; 1882 vr.ibytes = vif->bytes_in; 1883 vr.obytes = vif->bytes_out; 1884 read_unlock(&mrt_lock); 1885 1886 if (copy_to_user(arg, &vr, sizeof(vr))) 1887 return -EFAULT; 1888 return 0; 1889 } 1890 read_unlock(&mrt_lock); 1891 return -EADDRNOTAVAIL; 1892 case SIOCGETSGCNT_IN6: 1893 if (copy_from_user(&sr, arg, sizeof(sr))) 1894 return -EFAULT; 1895 1896 read_lock(&mrt_lock); 1897 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1898 if (c) { 1899 sr.pktcnt = c->mfc_un.res.pkt; 1900 sr.bytecnt = c->mfc_un.res.bytes; 1901 sr.wrong_if = c->mfc_un.res.wrong_if; 1902 read_unlock(&mrt_lock); 1903 1904 if (copy_to_user(arg, &sr, sizeof(sr))) 1905 return -EFAULT; 1906 return 0; 1907 } 1908 read_unlock(&mrt_lock); 1909 return -EADDRNOTAVAIL; 1910 default: 1911 return -ENOIOCTLCMD; 1912 } 1913 } 1914 #endif 1915 1916 static inline int ip6mr_forward2_finish(struct sk_buff *skb) 1917 { 1918 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)), 1919 IPSTATS_MIB_OUTFORWDATAGRAMS); 1920 IP6_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)), 1921 IPSTATS_MIB_OUTOCTETS, skb->len); 1922 return dst_output(skb); 1923 } 1924 1925 /* 1926 * Processing handlers for ip6mr_forward 1927 */ 1928 1929 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 1930 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 1931 { 1932 struct ipv6hdr *ipv6h; 1933 struct mif_device *vif = &mrt->vif6_table[vifi]; 1934 struct net_device *dev; 1935 struct dst_entry *dst; 1936 struct flowi6 fl6; 1937 1938 if (vif->dev == NULL) 1939 goto out_free; 1940 1941 #ifdef CONFIG_IPV6_PIMSM_V2 1942 if (vif->flags & MIFF_REGISTER) { 1943 vif->pkt_out++; 1944 vif->bytes_out += skb->len; 1945 vif->dev->stats.tx_bytes += skb->len; 1946 vif->dev->stats.tx_packets++; 1947 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 1948 goto out_free; 1949 } 1950 #endif 1951 1952 ipv6h = ipv6_hdr(skb); 1953 1954 fl6 = (struct flowi6) { 1955 .flowi6_oif = vif->link, 1956 .daddr = ipv6h->daddr, 1957 }; 1958 1959 dst = ip6_route_output(net, NULL, &fl6); 1960 if (dst->error) { 1961 dst_release(dst); 1962 goto out_free; 1963 } 1964 1965 skb_dst_drop(skb); 1966 skb_dst_set(skb, dst); 1967 1968 /* 1969 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1970 * not only before forwarding, but after forwarding on all output 1971 * interfaces. It is clear, if mrouter runs a multicasting 1972 * program, it should receive packets not depending to what interface 1973 * program is joined. 1974 * If we will not make it, the program will have to join on all 1975 * interfaces. On the other hand, multihoming host (or router, but 1976 * not mrouter) cannot join to more than one interface - it will 1977 * result in receiving multiple packets. 1978 */ 1979 dev = vif->dev; 1980 skb->dev = dev; 1981 vif->pkt_out++; 1982 vif->bytes_out += skb->len; 1983 1984 /* We are about to write */ 1985 /* XXX: extension headers? */ 1986 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 1987 goto out_free; 1988 1989 ipv6h = ipv6_hdr(skb); 1990 ipv6h->hop_limit--; 1991 1992 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 1993 1994 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev, 1995 ip6mr_forward2_finish); 1996 1997 out_free: 1998 kfree_skb(skb); 1999 return 0; 2000 } 2001 2002 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2003 { 2004 int ct; 2005 2006 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2007 if (mrt->vif6_table[ct].dev == dev) 2008 break; 2009 } 2010 return ct; 2011 } 2012 2013 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2014 struct sk_buff *skb, struct mfc6_cache *cache) 2015 { 2016 int psend = -1; 2017 int vif, ct; 2018 2019 vif = cache->mf6c_parent; 2020 cache->mfc_un.res.pkt++; 2021 cache->mfc_un.res.bytes += skb->len; 2022 2023 /* 2024 * Wrong interface: drop packet and (maybe) send PIM assert. 2025 */ 2026 if (mrt->vif6_table[vif].dev != skb->dev) { 2027 int true_vifi; 2028 2029 cache->mfc_un.res.wrong_if++; 2030 true_vifi = ip6mr_find_vif(mrt, skb->dev); 2031 2032 if (true_vifi >= 0 && mrt->mroute_do_assert && 2033 /* pimsm uses asserts, when switching from RPT to SPT, 2034 so that we cannot check that packet arrived on an oif. 2035 It is bad, but otherwise we would need to move pretty 2036 large chunk of pimd to kernel. Ough... --ANK 2037 */ 2038 (mrt->mroute_do_pim || 2039 cache->mfc_un.res.ttls[true_vifi] < 255) && 2040 time_after(jiffies, 2041 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2042 cache->mfc_un.res.last_assert = jiffies; 2043 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2044 } 2045 goto dont_forward; 2046 } 2047 2048 mrt->vif6_table[vif].pkt_in++; 2049 mrt->vif6_table[vif].bytes_in += skb->len; 2050 2051 /* 2052 * Forward the frame 2053 */ 2054 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2055 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2056 if (psend != -1) { 2057 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2058 if (skb2) 2059 ip6mr_forward2(net, mrt, skb2, cache, psend); 2060 } 2061 psend = ct; 2062 } 2063 } 2064 if (psend != -1) { 2065 ip6mr_forward2(net, mrt, skb, cache, psend); 2066 return 0; 2067 } 2068 2069 dont_forward: 2070 kfree_skb(skb); 2071 return 0; 2072 } 2073 2074 2075 /* 2076 * Multicast packets for forwarding arrive here 2077 */ 2078 2079 int ip6_mr_input(struct sk_buff *skb) 2080 { 2081 struct mfc6_cache *cache; 2082 struct net *net = dev_net(skb->dev); 2083 struct mr6_table *mrt; 2084 struct flowi6 fl6 = { 2085 .flowi6_iif = skb->dev->ifindex, 2086 .flowi6_mark = skb->mark, 2087 }; 2088 int err; 2089 2090 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2091 if (err < 0) { 2092 kfree_skb(skb); 2093 return err; 2094 } 2095 2096 read_lock(&mrt_lock); 2097 cache = ip6mr_cache_find(mrt, 2098 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2099 2100 /* 2101 * No usable cache entry 2102 */ 2103 if (cache == NULL) { 2104 int vif; 2105 2106 vif = ip6mr_find_vif(mrt, skb->dev); 2107 if (vif >= 0) { 2108 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2109 read_unlock(&mrt_lock); 2110 2111 return err; 2112 } 2113 read_unlock(&mrt_lock); 2114 kfree_skb(skb); 2115 return -ENODEV; 2116 } 2117 2118 ip6_mr_forward(net, mrt, skb, cache); 2119 2120 read_unlock(&mrt_lock); 2121 2122 return 0; 2123 } 2124 2125 2126 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2127 struct mfc6_cache *c, struct rtmsg *rtm) 2128 { 2129 int ct; 2130 struct rtnexthop *nhp; 2131 struct nlattr *mp_attr; 2132 struct rta_mfc_stats mfcs; 2133 2134 /* If cache is unresolved, don't try to parse IIF and OIF */ 2135 if (c->mf6c_parent >= MAXMIFS) 2136 return -ENOENT; 2137 2138 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2139 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2140 return -EMSGSIZE; 2141 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2142 if (mp_attr == NULL) 2143 return -EMSGSIZE; 2144 2145 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2146 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2147 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2148 if (nhp == NULL) { 2149 nla_nest_cancel(skb, mp_attr); 2150 return -EMSGSIZE; 2151 } 2152 2153 nhp->rtnh_flags = 0; 2154 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2155 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2156 nhp->rtnh_len = sizeof(*nhp); 2157 } 2158 } 2159 2160 nla_nest_end(skb, mp_attr); 2161 2162 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2163 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2164 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2165 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0) 2166 return -EMSGSIZE; 2167 2168 rtm->rtm_type = RTN_MULTICAST; 2169 return 1; 2170 } 2171 2172 int ip6mr_get_route(struct net *net, 2173 struct sk_buff *skb, struct rtmsg *rtm, int nowait) 2174 { 2175 int err; 2176 struct mr6_table *mrt; 2177 struct mfc6_cache *cache; 2178 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2179 2180 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2181 if (mrt == NULL) 2182 return -ENOENT; 2183 2184 read_lock(&mrt_lock); 2185 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2186 2187 if (!cache) { 2188 struct sk_buff *skb2; 2189 struct ipv6hdr *iph; 2190 struct net_device *dev; 2191 int vif; 2192 2193 if (nowait) { 2194 read_unlock(&mrt_lock); 2195 return -EAGAIN; 2196 } 2197 2198 dev = skb->dev; 2199 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2200 read_unlock(&mrt_lock); 2201 return -ENODEV; 2202 } 2203 2204 /* really correct? */ 2205 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2206 if (!skb2) { 2207 read_unlock(&mrt_lock); 2208 return -ENOMEM; 2209 } 2210 2211 skb_reset_transport_header(skb2); 2212 2213 skb_put(skb2, sizeof(struct ipv6hdr)); 2214 skb_reset_network_header(skb2); 2215 2216 iph = ipv6_hdr(skb2); 2217 iph->version = 0; 2218 iph->priority = 0; 2219 iph->flow_lbl[0] = 0; 2220 iph->flow_lbl[1] = 0; 2221 iph->flow_lbl[2] = 0; 2222 iph->payload_len = 0; 2223 iph->nexthdr = IPPROTO_NONE; 2224 iph->hop_limit = 0; 2225 iph->saddr = rt->rt6i_src.addr; 2226 iph->daddr = rt->rt6i_dst.addr; 2227 2228 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2229 read_unlock(&mrt_lock); 2230 2231 return err; 2232 } 2233 2234 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) 2235 cache->mfc_flags |= MFC_NOTIFY; 2236 2237 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2238 read_unlock(&mrt_lock); 2239 return err; 2240 } 2241 2242 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2243 u32 portid, u32 seq, struct mfc6_cache *c, int cmd) 2244 { 2245 struct nlmsghdr *nlh; 2246 struct rtmsg *rtm; 2247 int err; 2248 2249 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI); 2250 if (nlh == NULL) 2251 return -EMSGSIZE; 2252 2253 rtm = nlmsg_data(nlh); 2254 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2255 rtm->rtm_dst_len = 128; 2256 rtm->rtm_src_len = 128; 2257 rtm->rtm_tos = 0; 2258 rtm->rtm_table = mrt->id; 2259 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2260 goto nla_put_failure; 2261 rtm->rtm_type = RTN_MULTICAST; 2262 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2263 if (c->mfc_flags & MFC_STATIC) 2264 rtm->rtm_protocol = RTPROT_STATIC; 2265 else 2266 rtm->rtm_protocol = RTPROT_MROUTED; 2267 rtm->rtm_flags = 0; 2268 2269 if (nla_put(skb, RTA_SRC, 16, &c->mf6c_origin) || 2270 nla_put(skb, RTA_DST, 16, &c->mf6c_mcastgrp)) 2271 goto nla_put_failure; 2272 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2273 /* do not break the dump if cache is unresolved */ 2274 if (err < 0 && err != -ENOENT) 2275 goto nla_put_failure; 2276 2277 return nlmsg_end(skb, nlh); 2278 2279 nla_put_failure: 2280 nlmsg_cancel(skb, nlh); 2281 return -EMSGSIZE; 2282 } 2283 2284 static int mr6_msgsize(bool unresolved, int maxvif) 2285 { 2286 size_t len = 2287 NLMSG_ALIGN(sizeof(struct rtmsg)) 2288 + nla_total_size(4) /* RTA_TABLE */ 2289 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2290 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2291 ; 2292 2293 if (!unresolved) 2294 len = len 2295 + nla_total_size(4) /* RTA_IIF */ 2296 + nla_total_size(0) /* RTA_MULTIPATH */ 2297 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2298 /* RTA_MFC_STATS */ 2299 + nla_total_size(sizeof(struct rta_mfc_stats)) 2300 ; 2301 2302 return len; 2303 } 2304 2305 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2306 int cmd) 2307 { 2308 struct net *net = read_pnet(&mrt->net); 2309 struct sk_buff *skb; 2310 int err = -ENOBUFS; 2311 2312 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2313 GFP_ATOMIC); 2314 if (skb == NULL) 2315 goto errout; 2316 2317 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd); 2318 if (err < 0) 2319 goto errout; 2320 2321 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2322 return; 2323 2324 errout: 2325 kfree_skb(skb); 2326 if (err < 0) 2327 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2328 } 2329 2330 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2331 { 2332 struct net *net = sock_net(skb->sk); 2333 struct mr6_table *mrt; 2334 struct mfc6_cache *mfc; 2335 unsigned int t = 0, s_t; 2336 unsigned int h = 0, s_h; 2337 unsigned int e = 0, s_e; 2338 2339 s_t = cb->args[0]; 2340 s_h = cb->args[1]; 2341 s_e = cb->args[2]; 2342 2343 read_lock(&mrt_lock); 2344 ip6mr_for_each_table(mrt, net) { 2345 if (t < s_t) 2346 goto next_table; 2347 if (t > s_t) 2348 s_h = 0; 2349 for (h = s_h; h < MFC6_LINES; h++) { 2350 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2351 if (e < s_e) 2352 goto next_entry; 2353 if (ip6mr_fill_mroute(mrt, skb, 2354 NETLINK_CB(cb->skb).portid, 2355 cb->nlh->nlmsg_seq, 2356 mfc, RTM_NEWROUTE) < 0) 2357 goto done; 2358 next_entry: 2359 e++; 2360 } 2361 e = s_e = 0; 2362 } 2363 spin_lock_bh(&mfc_unres_lock); 2364 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2365 if (e < s_e) 2366 goto next_entry2; 2367 if (ip6mr_fill_mroute(mrt, skb, 2368 NETLINK_CB(cb->skb).portid, 2369 cb->nlh->nlmsg_seq, 2370 mfc, RTM_NEWROUTE) < 0) { 2371 spin_unlock_bh(&mfc_unres_lock); 2372 goto done; 2373 } 2374 next_entry2: 2375 e++; 2376 } 2377 spin_unlock_bh(&mfc_unres_lock); 2378 e = s_e = 0; 2379 s_h = 0; 2380 next_table: 2381 t++; 2382 } 2383 done: 2384 read_unlock(&mrt_lock); 2385 2386 cb->args[2] = e; 2387 cb->args[1] = h; 2388 cb->args[0] = t; 2389 2390 return skb->len; 2391 } 2392