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 57struct 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 75struct ip6mr_rule { 76 struct fib_rule common; 77}; 78 79struct 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 87static 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 */ 96static 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 106static struct kmem_cache *mrt_cachep __read_mostly; 107 108static struct mr6_table *ip6mr_new_table(struct net *net, u32 id); 109static void ip6mr_free_table(struct mr6_table *mrt); 110 111static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 112 struct sk_buff *skb, struct mfc6_cache *cache); 113static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 114 mifi_t mifi, int assert); 115static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 116 struct mfc6_cache *c, struct rtmsg *rtm); 117static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 118 int cmd); 119static int ip6mr_rtm_dumproute(struct sk_buff *skb, 120 struct netlink_callback *cb); 121static void mroute_clean_tables(struct mr6_table *mrt, bool all); 122static 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 128static 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 139static 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 157static 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 182static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags) 183{ 184 return 1; 185} 186 187static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = { 188 FRA_GENERIC_POLICY, 189}; 190 191static 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 197static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 198 struct nlattr **tb) 199{ 200 return 1; 201} 202 203static 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 212static 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 .default_pref = fib_default_rule_pref, 221 .fill = ip6mr_rule_fill, 222 .nlgroup = RTNLGRP_IPV6_RULE, 223 .policy = ip6mr_rule_policy, 224 .owner = THIS_MODULE, 225}; 226 227static int __net_init ip6mr_rules_init(struct net *net) 228{ 229 struct fib_rules_ops *ops; 230 struct mr6_table *mrt; 231 int err; 232 233 ops = fib_rules_register(&ip6mr_rules_ops_template, net); 234 if (IS_ERR(ops)) 235 return PTR_ERR(ops); 236 237 INIT_LIST_HEAD(&net->ipv6.mr6_tables); 238 239 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT); 240 if (!mrt) { 241 err = -ENOMEM; 242 goto err1; 243 } 244 245 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0); 246 if (err < 0) 247 goto err2; 248 249 net->ipv6.mr6_rules_ops = ops; 250 return 0; 251 252err2: 253 ip6mr_free_table(mrt); 254err1: 255 fib_rules_unregister(ops); 256 return err; 257} 258 259static void __net_exit ip6mr_rules_exit(struct net *net) 260{ 261 struct mr6_table *mrt, *next; 262 263 rtnl_lock(); 264 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) { 265 list_del(&mrt->list); 266 ip6mr_free_table(mrt); 267 } 268 fib_rules_unregister(net->ipv6.mr6_rules_ops); 269 rtnl_unlock(); 270} 271#else 272#define ip6mr_for_each_table(mrt, net) \ 273 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL) 274 275static struct mr6_table *ip6mr_get_table(struct net *net, u32 id) 276{ 277 return net->ipv6.mrt6; 278} 279 280static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6, 281 struct mr6_table **mrt) 282{ 283 *mrt = net->ipv6.mrt6; 284 return 0; 285} 286 287static int __net_init ip6mr_rules_init(struct net *net) 288{ 289 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT); 290 return net->ipv6.mrt6 ? 0 : -ENOMEM; 291} 292 293static void __net_exit ip6mr_rules_exit(struct net *net) 294{ 295 rtnl_lock(); 296 ip6mr_free_table(net->ipv6.mrt6); 297 net->ipv6.mrt6 = NULL; 298 rtnl_unlock(); 299} 300#endif 301 302static struct mr6_table *ip6mr_new_table(struct net *net, u32 id) 303{ 304 struct mr6_table *mrt; 305 unsigned int i; 306 307 mrt = ip6mr_get_table(net, id); 308 if (mrt) 309 return mrt; 310 311 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 312 if (!mrt) 313 return NULL; 314 mrt->id = id; 315 write_pnet(&mrt->net, net); 316 317 /* Forwarding cache */ 318 for (i = 0; i < MFC6_LINES; i++) 319 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]); 320 321 INIT_LIST_HEAD(&mrt->mfc6_unres_queue); 322 323 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 324 (unsigned long)mrt); 325 326#ifdef CONFIG_IPV6_PIMSM_V2 327 mrt->mroute_reg_vif_num = -1; 328#endif 329#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 330 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables); 331#endif 332 return mrt; 333} 334 335static void ip6mr_free_table(struct mr6_table *mrt) 336{ 337 del_timer_sync(&mrt->ipmr_expire_timer); 338 mroute_clean_tables(mrt, true); 339 kfree(mrt); 340} 341 342#ifdef CONFIG_PROC_FS 343 344struct ipmr_mfc_iter { 345 struct seq_net_private p; 346 struct mr6_table *mrt; 347 struct list_head *cache; 348 int ct; 349}; 350 351 352static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net, 353 struct ipmr_mfc_iter *it, loff_t pos) 354{ 355 struct mr6_table *mrt = it->mrt; 356 struct mfc6_cache *mfc; 357 358 read_lock(&mrt_lock); 359 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) { 360 it->cache = &mrt->mfc6_cache_array[it->ct]; 361 list_for_each_entry(mfc, it->cache, list) 362 if (pos-- == 0) 363 return mfc; 364 } 365 read_unlock(&mrt_lock); 366 367 spin_lock_bh(&mfc_unres_lock); 368 it->cache = &mrt->mfc6_unres_queue; 369 list_for_each_entry(mfc, it->cache, list) 370 if (pos-- == 0) 371 return mfc; 372 spin_unlock_bh(&mfc_unres_lock); 373 374 it->cache = NULL; 375 return NULL; 376} 377 378/* 379 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif 380 */ 381 382struct ipmr_vif_iter { 383 struct seq_net_private p; 384 struct mr6_table *mrt; 385 int ct; 386}; 387 388static struct mif_device *ip6mr_vif_seq_idx(struct net *net, 389 struct ipmr_vif_iter *iter, 390 loff_t pos) 391{ 392 struct mr6_table *mrt = iter->mrt; 393 394 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 395 if (!MIF_EXISTS(mrt, iter->ct)) 396 continue; 397 if (pos-- == 0) 398 return &mrt->vif6_table[iter->ct]; 399 } 400 return NULL; 401} 402 403static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) 404 __acquires(mrt_lock) 405{ 406 struct ipmr_vif_iter *iter = seq->private; 407 struct net *net = seq_file_net(seq); 408 struct mr6_table *mrt; 409 410 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 411 if (!mrt) 412 return ERR_PTR(-ENOENT); 413 414 iter->mrt = mrt; 415 416 read_lock(&mrt_lock); 417 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1) 418 : SEQ_START_TOKEN; 419} 420 421static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 422{ 423 struct ipmr_vif_iter *iter = seq->private; 424 struct net *net = seq_file_net(seq); 425 struct mr6_table *mrt = iter->mrt; 426 427 ++*pos; 428 if (v == SEQ_START_TOKEN) 429 return ip6mr_vif_seq_idx(net, iter, 0); 430 431 while (++iter->ct < mrt->maxvif) { 432 if (!MIF_EXISTS(mrt, iter->ct)) 433 continue; 434 return &mrt->vif6_table[iter->ct]; 435 } 436 return NULL; 437} 438 439static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) 440 __releases(mrt_lock) 441{ 442 read_unlock(&mrt_lock); 443} 444 445static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) 446{ 447 struct ipmr_vif_iter *iter = seq->private; 448 struct mr6_table *mrt = iter->mrt; 449 450 if (v == SEQ_START_TOKEN) { 451 seq_puts(seq, 452 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); 453 } else { 454 const struct mif_device *vif = v; 455 const char *name = vif->dev ? vif->dev->name : "none"; 456 457 seq_printf(seq, 458 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", 459 vif - mrt->vif6_table, 460 name, vif->bytes_in, vif->pkt_in, 461 vif->bytes_out, vif->pkt_out, 462 vif->flags); 463 } 464 return 0; 465} 466 467static const struct seq_operations ip6mr_vif_seq_ops = { 468 .start = ip6mr_vif_seq_start, 469 .next = ip6mr_vif_seq_next, 470 .stop = ip6mr_vif_seq_stop, 471 .show = ip6mr_vif_seq_show, 472}; 473 474static int ip6mr_vif_open(struct inode *inode, struct file *file) 475{ 476 return seq_open_net(inode, file, &ip6mr_vif_seq_ops, 477 sizeof(struct ipmr_vif_iter)); 478} 479 480static const struct file_operations ip6mr_vif_fops = { 481 .owner = THIS_MODULE, 482 .open = ip6mr_vif_open, 483 .read = seq_read, 484 .llseek = seq_lseek, 485 .release = seq_release_net, 486}; 487 488static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 489{ 490 struct ipmr_mfc_iter *it = seq->private; 491 struct net *net = seq_file_net(seq); 492 struct mr6_table *mrt; 493 494 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 495 if (!mrt) 496 return ERR_PTR(-ENOENT); 497 498 it->mrt = mrt; 499 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 500 : SEQ_START_TOKEN; 501} 502 503static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 504{ 505 struct mfc6_cache *mfc = v; 506 struct ipmr_mfc_iter *it = seq->private; 507 struct net *net = seq_file_net(seq); 508 struct mr6_table *mrt = it->mrt; 509 510 ++*pos; 511 512 if (v == SEQ_START_TOKEN) 513 return ipmr_mfc_seq_idx(net, seq->private, 0); 514 515 if (mfc->list.next != it->cache) 516 return list_entry(mfc->list.next, struct mfc6_cache, list); 517 518 if (it->cache == &mrt->mfc6_unres_queue) 519 goto end_of_list; 520 521 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]); 522 523 while (++it->ct < MFC6_LINES) { 524 it->cache = &mrt->mfc6_cache_array[it->ct]; 525 if (list_empty(it->cache)) 526 continue; 527 return list_first_entry(it->cache, struct mfc6_cache, list); 528 } 529 530 /* exhausted cache_array, show unresolved */ 531 read_unlock(&mrt_lock); 532 it->cache = &mrt->mfc6_unres_queue; 533 it->ct = 0; 534 535 spin_lock_bh(&mfc_unres_lock); 536 if (!list_empty(it->cache)) 537 return list_first_entry(it->cache, struct mfc6_cache, list); 538 539 end_of_list: 540 spin_unlock_bh(&mfc_unres_lock); 541 it->cache = NULL; 542 543 return NULL; 544} 545 546static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 547{ 548 struct ipmr_mfc_iter *it = seq->private; 549 struct mr6_table *mrt = it->mrt; 550 551 if (it->cache == &mrt->mfc6_unres_queue) 552 spin_unlock_bh(&mfc_unres_lock); 553 else if (it->cache == &mrt->mfc6_cache_array[it->ct]) 554 read_unlock(&mrt_lock); 555} 556 557static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 558{ 559 int n; 560 561 if (v == SEQ_START_TOKEN) { 562 seq_puts(seq, 563 "Group " 564 "Origin " 565 "Iif Pkts Bytes Wrong Oifs\n"); 566 } else { 567 const struct mfc6_cache *mfc = v; 568 const struct ipmr_mfc_iter *it = seq->private; 569 struct mr6_table *mrt = it->mrt; 570 571 seq_printf(seq, "%pI6 %pI6 %-3hd", 572 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, 573 mfc->mf6c_parent); 574 575 if (it->cache != &mrt->mfc6_unres_queue) { 576 seq_printf(seq, " %8lu %8lu %8lu", 577 mfc->mfc_un.res.pkt, 578 mfc->mfc_un.res.bytes, 579 mfc->mfc_un.res.wrong_if); 580 for (n = mfc->mfc_un.res.minvif; 581 n < mfc->mfc_un.res.maxvif; n++) { 582 if (MIF_EXISTS(mrt, n) && 583 mfc->mfc_un.res.ttls[n] < 255) 584 seq_printf(seq, 585 " %2d:%-3d", 586 n, mfc->mfc_un.res.ttls[n]); 587 } 588 } else { 589 /* unresolved mfc_caches don't contain 590 * pkt, bytes and wrong_if values 591 */ 592 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 593 } 594 seq_putc(seq, '\n'); 595 } 596 return 0; 597} 598 599static const struct seq_operations ipmr_mfc_seq_ops = { 600 .start = ipmr_mfc_seq_start, 601 .next = ipmr_mfc_seq_next, 602 .stop = ipmr_mfc_seq_stop, 603 .show = ipmr_mfc_seq_show, 604}; 605 606static int ipmr_mfc_open(struct inode *inode, struct file *file) 607{ 608 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 609 sizeof(struct ipmr_mfc_iter)); 610} 611 612static const struct file_operations ip6mr_mfc_fops = { 613 .owner = THIS_MODULE, 614 .open = ipmr_mfc_open, 615 .read = seq_read, 616 .llseek = seq_lseek, 617 .release = seq_release_net, 618}; 619#endif 620 621#ifdef CONFIG_IPV6_PIMSM_V2 622 623static int pim6_rcv(struct sk_buff *skb) 624{ 625 struct pimreghdr *pim; 626 struct ipv6hdr *encap; 627 struct net_device *reg_dev = NULL; 628 struct net *net = dev_net(skb->dev); 629 struct mr6_table *mrt; 630 struct flowi6 fl6 = { 631 .flowi6_iif = skb->dev->ifindex, 632 .flowi6_mark = skb->mark, 633 }; 634 int reg_vif_num; 635 636 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) 637 goto drop; 638 639 pim = (struct pimreghdr *)skb_transport_header(skb); 640 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || 641 (pim->flags & PIM_NULL_REGISTER) || 642 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 643 sizeof(*pim), IPPROTO_PIM, 644 csum_partial((void *)pim, sizeof(*pim), 0)) && 645 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 646 goto drop; 647 648 /* check if the inner packet is destined to mcast group */ 649 encap = (struct ipv6hdr *)(skb_transport_header(skb) + 650 sizeof(*pim)); 651 652 if (!ipv6_addr_is_multicast(&encap->daddr) || 653 encap->payload_len == 0 || 654 ntohs(encap->payload_len) + sizeof(*pim) > skb->len) 655 goto drop; 656 657 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 658 goto drop; 659 reg_vif_num = mrt->mroute_reg_vif_num; 660 661 read_lock(&mrt_lock); 662 if (reg_vif_num >= 0) 663 reg_dev = mrt->vif6_table[reg_vif_num].dev; 664 if (reg_dev) 665 dev_hold(reg_dev); 666 read_unlock(&mrt_lock); 667 668 if (!reg_dev) 669 goto drop; 670 671 skb->mac_header = skb->network_header; 672 skb_pull(skb, (u8 *)encap - skb->data); 673 skb_reset_network_header(skb); 674 skb->protocol = htons(ETH_P_IPV6); 675 skb->ip_summed = CHECKSUM_NONE; 676 677 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 678 679 netif_rx(skb); 680 681 dev_put(reg_dev); 682 return 0; 683 drop: 684 kfree_skb(skb); 685 return 0; 686} 687 688static const struct inet6_protocol pim6_protocol = { 689 .handler = pim6_rcv, 690}; 691 692/* Service routines creating virtual interfaces: PIMREG */ 693 694static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, 695 struct net_device *dev) 696{ 697 struct net *net = dev_net(dev); 698 struct mr6_table *mrt; 699 struct flowi6 fl6 = { 700 .flowi6_oif = dev->ifindex, 701 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 702 .flowi6_mark = skb->mark, 703 }; 704 int err; 705 706 err = ip6mr_fib_lookup(net, &fl6, &mrt); 707 if (err < 0) { 708 kfree_skb(skb); 709 return err; 710 } 711 712 read_lock(&mrt_lock); 713 dev->stats.tx_bytes += skb->len; 714 dev->stats.tx_packets++; 715 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT); 716 read_unlock(&mrt_lock); 717 kfree_skb(skb); 718 return NETDEV_TX_OK; 719} 720 721static int reg_vif_get_iflink(const struct net_device *dev) 722{ 723 return 0; 724} 725 726static const struct net_device_ops reg_vif_netdev_ops = { 727 .ndo_start_xmit = reg_vif_xmit, 728 .ndo_get_iflink = reg_vif_get_iflink, 729}; 730 731static void reg_vif_setup(struct net_device *dev) 732{ 733 dev->type = ARPHRD_PIMREG; 734 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; 735 dev->flags = IFF_NOARP; 736 dev->netdev_ops = ®_vif_netdev_ops; 737 dev->destructor = free_netdev; 738 dev->features |= NETIF_F_NETNS_LOCAL; 739} 740 741static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt) 742{ 743 struct net_device *dev; 744 char name[IFNAMSIZ]; 745 746 if (mrt->id == RT6_TABLE_DFLT) 747 sprintf(name, "pim6reg"); 748 else 749 sprintf(name, "pim6reg%u", mrt->id); 750 751 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 752 if (!dev) 753 return NULL; 754 755 dev_net_set(dev, net); 756 757 if (register_netdevice(dev)) { 758 free_netdev(dev); 759 return NULL; 760 } 761 762 if (dev_open(dev)) 763 goto failure; 764 765 dev_hold(dev); 766 return dev; 767 768failure: 769 /* allow the register to be completed before unregistering. */ 770 rtnl_unlock(); 771 rtnl_lock(); 772 773 unregister_netdevice(dev); 774 return NULL; 775} 776#endif 777 778/* 779 * Delete a VIF entry 780 */ 781 782static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head) 783{ 784 struct mif_device *v; 785 struct net_device *dev; 786 struct inet6_dev *in6_dev; 787 788 if (vifi < 0 || vifi >= mrt->maxvif) 789 return -EADDRNOTAVAIL; 790 791 v = &mrt->vif6_table[vifi]; 792 793 write_lock_bh(&mrt_lock); 794 dev = v->dev; 795 v->dev = NULL; 796 797 if (!dev) { 798 write_unlock_bh(&mrt_lock); 799 return -EADDRNOTAVAIL; 800 } 801 802#ifdef CONFIG_IPV6_PIMSM_V2 803 if (vifi == mrt->mroute_reg_vif_num) 804 mrt->mroute_reg_vif_num = -1; 805#endif 806 807 if (vifi + 1 == mrt->maxvif) { 808 int tmp; 809 for (tmp = vifi - 1; tmp >= 0; tmp--) { 810 if (MIF_EXISTS(mrt, tmp)) 811 break; 812 } 813 mrt->maxvif = tmp + 1; 814 } 815 816 write_unlock_bh(&mrt_lock); 817 818 dev_set_allmulti(dev, -1); 819 820 in6_dev = __in6_dev_get(dev); 821 if (in6_dev) { 822 in6_dev->cnf.mc_forwarding--; 823 inet6_netconf_notify_devconf(dev_net(dev), 824 NETCONFA_MC_FORWARDING, 825 dev->ifindex, &in6_dev->cnf); 826 } 827 828 if (v->flags & MIFF_REGISTER) 829 unregister_netdevice_queue(dev, head); 830 831 dev_put(dev); 832 return 0; 833} 834 835static inline void ip6mr_cache_free(struct mfc6_cache *c) 836{ 837 kmem_cache_free(mrt_cachep, c); 838} 839 840/* Destroy an unresolved cache entry, killing queued skbs 841 and reporting error to netlink readers. 842 */ 843 844static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c) 845{ 846 struct net *net = read_pnet(&mrt->net); 847 struct sk_buff *skb; 848 849 atomic_dec(&mrt->cache_resolve_queue_len); 850 851 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { 852 if (ipv6_hdr(skb)->version == 0) { 853 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); 854 nlh->nlmsg_type = NLMSG_ERROR; 855 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 856 skb_trim(skb, nlh->nlmsg_len); 857 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT; 858 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 859 } else 860 kfree_skb(skb); 861 } 862 863 ip6mr_cache_free(c); 864} 865 866 867/* Timer process for all the unresolved queue. */ 868 869static void ipmr_do_expire_process(struct mr6_table *mrt) 870{ 871 unsigned long now = jiffies; 872 unsigned long expires = 10 * HZ; 873 struct mfc6_cache *c, *next; 874 875 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 876 if (time_after(c->mfc_un.unres.expires, now)) { 877 /* not yet... */ 878 unsigned long interval = c->mfc_un.unres.expires - now; 879 if (interval < expires) 880 expires = interval; 881 continue; 882 } 883 884 list_del(&c->list); 885 mr6_netlink_event(mrt, c, RTM_DELROUTE); 886 ip6mr_destroy_unres(mrt, c); 887 } 888 889 if (!list_empty(&mrt->mfc6_unres_queue)) 890 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 891} 892 893static void ipmr_expire_process(unsigned long arg) 894{ 895 struct mr6_table *mrt = (struct mr6_table *)arg; 896 897 if (!spin_trylock(&mfc_unres_lock)) { 898 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1); 899 return; 900 } 901 902 if (!list_empty(&mrt->mfc6_unres_queue)) 903 ipmr_do_expire_process(mrt); 904 905 spin_unlock(&mfc_unres_lock); 906} 907 908/* Fill oifs list. It is called under write locked mrt_lock. */ 909 910static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache, 911 unsigned char *ttls) 912{ 913 int vifi; 914 915 cache->mfc_un.res.minvif = MAXMIFS; 916 cache->mfc_un.res.maxvif = 0; 917 memset(cache->mfc_un.res.ttls, 255, MAXMIFS); 918 919 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 920 if (MIF_EXISTS(mrt, vifi) && 921 ttls[vifi] && ttls[vifi] < 255) { 922 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 923 if (cache->mfc_un.res.minvif > vifi) 924 cache->mfc_un.res.minvif = vifi; 925 if (cache->mfc_un.res.maxvif <= vifi) 926 cache->mfc_un.res.maxvif = vifi + 1; 927 } 928 } 929} 930 931static int mif6_add(struct net *net, struct mr6_table *mrt, 932 struct mif6ctl *vifc, int mrtsock) 933{ 934 int vifi = vifc->mif6c_mifi; 935 struct mif_device *v = &mrt->vif6_table[vifi]; 936 struct net_device *dev; 937 struct inet6_dev *in6_dev; 938 int err; 939 940 /* Is vif busy ? */ 941 if (MIF_EXISTS(mrt, vifi)) 942 return -EADDRINUSE; 943 944 switch (vifc->mif6c_flags) { 945#ifdef CONFIG_IPV6_PIMSM_V2 946 case MIFF_REGISTER: 947 /* 948 * Special Purpose VIF in PIM 949 * All the packets will be sent to the daemon 950 */ 951 if (mrt->mroute_reg_vif_num >= 0) 952 return -EADDRINUSE; 953 dev = ip6mr_reg_vif(net, mrt); 954 if (!dev) 955 return -ENOBUFS; 956 err = dev_set_allmulti(dev, 1); 957 if (err) { 958 unregister_netdevice(dev); 959 dev_put(dev); 960 return err; 961 } 962 break; 963#endif 964 case 0: 965 dev = dev_get_by_index(net, vifc->mif6c_pifi); 966 if (!dev) 967 return -EADDRNOTAVAIL; 968 err = dev_set_allmulti(dev, 1); 969 if (err) { 970 dev_put(dev); 971 return err; 972 } 973 break; 974 default: 975 return -EINVAL; 976 } 977 978 in6_dev = __in6_dev_get(dev); 979 if (in6_dev) { 980 in6_dev->cnf.mc_forwarding++; 981 inet6_netconf_notify_devconf(dev_net(dev), 982 NETCONFA_MC_FORWARDING, 983 dev->ifindex, &in6_dev->cnf); 984 } 985 986 /* 987 * Fill in the VIF structures 988 */ 989 v->rate_limit = vifc->vifc_rate_limit; 990 v->flags = vifc->mif6c_flags; 991 if (!mrtsock) 992 v->flags |= VIFF_STATIC; 993 v->threshold = vifc->vifc_threshold; 994 v->bytes_in = 0; 995 v->bytes_out = 0; 996 v->pkt_in = 0; 997 v->pkt_out = 0; 998 v->link = dev->ifindex; 999 if (v->flags & MIFF_REGISTER) 1000 v->link = dev_get_iflink(dev); 1001 1002 /* And finish update writing critical data */ 1003 write_lock_bh(&mrt_lock); 1004 v->dev = dev; 1005#ifdef CONFIG_IPV6_PIMSM_V2 1006 if (v->flags & MIFF_REGISTER) 1007 mrt->mroute_reg_vif_num = vifi; 1008#endif 1009 if (vifi + 1 > mrt->maxvif) 1010 mrt->maxvif = vifi + 1; 1011 write_unlock_bh(&mrt_lock); 1012 return 0; 1013} 1014 1015static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt, 1016 const struct in6_addr *origin, 1017 const struct in6_addr *mcastgrp) 1018{ 1019 int line = MFC6_HASH(mcastgrp, origin); 1020 struct mfc6_cache *c; 1021 1022 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1023 if (ipv6_addr_equal(&c->mf6c_origin, origin) && 1024 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) 1025 return c; 1026 } 1027 return NULL; 1028} 1029 1030/* Look for a (*,*,oif) entry */ 1031static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt, 1032 mifi_t mifi) 1033{ 1034 int line = MFC6_HASH(&in6addr_any, &in6addr_any); 1035 struct mfc6_cache *c; 1036 1037 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1038 if (ipv6_addr_any(&c->mf6c_origin) && 1039 ipv6_addr_any(&c->mf6c_mcastgrp) && 1040 (c->mfc_un.res.ttls[mifi] < 255)) 1041 return c; 1042 1043 return NULL; 1044} 1045 1046/* Look for a (*,G) entry */ 1047static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt, 1048 struct in6_addr *mcastgrp, 1049 mifi_t mifi) 1050{ 1051 int line = MFC6_HASH(mcastgrp, &in6addr_any); 1052 struct mfc6_cache *c, *proxy; 1053 1054 if (ipv6_addr_any(mcastgrp)) 1055 goto skip; 1056 1057 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) 1058 if (ipv6_addr_any(&c->mf6c_origin) && 1059 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) { 1060 if (c->mfc_un.res.ttls[mifi] < 255) 1061 return c; 1062 1063 /* It's ok if the mifi is part of the static tree */ 1064 proxy = ip6mr_cache_find_any_parent(mrt, 1065 c->mf6c_parent); 1066 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255) 1067 return c; 1068 } 1069 1070skip: 1071 return ip6mr_cache_find_any_parent(mrt, mifi); 1072} 1073 1074/* 1075 * Allocate a multicast cache entry 1076 */ 1077static struct mfc6_cache *ip6mr_cache_alloc(void) 1078{ 1079 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 1080 if (!c) 1081 return NULL; 1082 c->mfc_un.res.minvif = MAXMIFS; 1083 return c; 1084} 1085 1086static struct mfc6_cache *ip6mr_cache_alloc_unres(void) 1087{ 1088 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 1089 if (!c) 1090 return NULL; 1091 skb_queue_head_init(&c->mfc_un.unres.unresolved); 1092 c->mfc_un.unres.expires = jiffies + 10 * HZ; 1093 return c; 1094} 1095 1096/* 1097 * A cache entry has gone into a resolved state from queued 1098 */ 1099 1100static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt, 1101 struct mfc6_cache *uc, struct mfc6_cache *c) 1102{ 1103 struct sk_buff *skb; 1104 1105 /* 1106 * Play the pending entries through our router 1107 */ 1108 1109 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 1110 if (ipv6_hdr(skb)->version == 0) { 1111 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); 1112 1113 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 1114 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; 1115 } else { 1116 nlh->nlmsg_type = NLMSG_ERROR; 1117 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1118 skb_trim(skb, nlh->nlmsg_len); 1119 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE; 1120 } 1121 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1122 } else 1123 ip6_mr_forward(net, mrt, skb, c); 1124 } 1125} 1126 1127/* 1128 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd 1129 * expects the following bizarre scheme. 1130 * 1131 * Called under mrt_lock. 1132 */ 1133 1134static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt, 1135 mifi_t mifi, int assert) 1136{ 1137 struct sk_buff *skb; 1138 struct mrt6msg *msg; 1139 int ret; 1140 1141#ifdef CONFIG_IPV6_PIMSM_V2 1142 if (assert == MRT6MSG_WHOLEPKT) 1143 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) 1144 +sizeof(*msg)); 1145 else 1146#endif 1147 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); 1148 1149 if (!skb) 1150 return -ENOBUFS; 1151 1152 /* I suppose that internal messages 1153 * do not require checksums */ 1154 1155 skb->ip_summed = CHECKSUM_UNNECESSARY; 1156 1157#ifdef CONFIG_IPV6_PIMSM_V2 1158 if (assert == MRT6MSG_WHOLEPKT) { 1159 /* Ugly, but we have no choice with this interface. 1160 Duplicate old header, fix length etc. 1161 And all this only to mangle msg->im6_msgtype and 1162 to set msg->im6_mbz to "mbz" :-) 1163 */ 1164 skb_push(skb, -skb_network_offset(pkt)); 1165 1166 skb_push(skb, sizeof(*msg)); 1167 skb_reset_transport_header(skb); 1168 msg = (struct mrt6msg *)skb_transport_header(skb); 1169 msg->im6_mbz = 0; 1170 msg->im6_msgtype = MRT6MSG_WHOLEPKT; 1171 msg->im6_mif = mrt->mroute_reg_vif_num; 1172 msg->im6_pad = 0; 1173 msg->im6_src = ipv6_hdr(pkt)->saddr; 1174 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1175 1176 skb->ip_summed = CHECKSUM_UNNECESSARY; 1177 } else 1178#endif 1179 { 1180 /* 1181 * Copy the IP header 1182 */ 1183 1184 skb_put(skb, sizeof(struct ipv6hdr)); 1185 skb_reset_network_header(skb); 1186 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); 1187 1188 /* 1189 * Add our header 1190 */ 1191 skb_put(skb, sizeof(*msg)); 1192 skb_reset_transport_header(skb); 1193 msg = (struct mrt6msg *)skb_transport_header(skb); 1194 1195 msg->im6_mbz = 0; 1196 msg->im6_msgtype = assert; 1197 msg->im6_mif = mifi; 1198 msg->im6_pad = 0; 1199 msg->im6_src = ipv6_hdr(pkt)->saddr; 1200 msg->im6_dst = ipv6_hdr(pkt)->daddr; 1201 1202 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1203 skb->ip_summed = CHECKSUM_UNNECESSARY; 1204 } 1205 1206 if (!mrt->mroute6_sk) { 1207 kfree_skb(skb); 1208 return -EINVAL; 1209 } 1210 1211 /* 1212 * Deliver to user space multicast routing algorithms 1213 */ 1214 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb); 1215 if (ret < 0) { 1216 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n"); 1217 kfree_skb(skb); 1218 } 1219 1220 return ret; 1221} 1222 1223/* 1224 * Queue a packet for resolution. It gets locked cache entry! 1225 */ 1226 1227static int 1228ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb) 1229{ 1230 bool found = false; 1231 int err; 1232 struct mfc6_cache *c; 1233 1234 spin_lock_bh(&mfc_unres_lock); 1235 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) { 1236 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && 1237 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) { 1238 found = true; 1239 break; 1240 } 1241 } 1242 1243 if (!found) { 1244 /* 1245 * Create a new entry if allowable 1246 */ 1247 1248 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1249 (c = ip6mr_cache_alloc_unres()) == NULL) { 1250 spin_unlock_bh(&mfc_unres_lock); 1251 1252 kfree_skb(skb); 1253 return -ENOBUFS; 1254 } 1255 1256 /* 1257 * Fill in the new cache entry 1258 */ 1259 c->mf6c_parent = -1; 1260 c->mf6c_origin = ipv6_hdr(skb)->saddr; 1261 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; 1262 1263 /* 1264 * Reflect first query at pim6sd 1265 */ 1266 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE); 1267 if (err < 0) { 1268 /* If the report failed throw the cache entry 1269 out - Brad Parker 1270 */ 1271 spin_unlock_bh(&mfc_unres_lock); 1272 1273 ip6mr_cache_free(c); 1274 kfree_skb(skb); 1275 return err; 1276 } 1277 1278 atomic_inc(&mrt->cache_resolve_queue_len); 1279 list_add(&c->list, &mrt->mfc6_unres_queue); 1280 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1281 1282 ipmr_do_expire_process(mrt); 1283 } 1284 1285 /* 1286 * See if we can append the packet 1287 */ 1288 if (c->mfc_un.unres.unresolved.qlen > 3) { 1289 kfree_skb(skb); 1290 err = -ENOBUFS; 1291 } else { 1292 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1293 err = 0; 1294 } 1295 1296 spin_unlock_bh(&mfc_unres_lock); 1297 return err; 1298} 1299 1300/* 1301 * MFC6 cache manipulation by user space 1302 */ 1303 1304static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc, 1305 int parent) 1306{ 1307 int line; 1308 struct mfc6_cache *c, *next; 1309 1310 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1311 1312 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) { 1313 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1314 ipv6_addr_equal(&c->mf6c_mcastgrp, 1315 &mfc->mf6cc_mcastgrp.sin6_addr) && 1316 (parent == -1 || parent == c->mf6c_parent)) { 1317 write_lock_bh(&mrt_lock); 1318 list_del(&c->list); 1319 write_unlock_bh(&mrt_lock); 1320 1321 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1322 ip6mr_cache_free(c); 1323 return 0; 1324 } 1325 } 1326 return -ENOENT; 1327} 1328 1329static int ip6mr_device_event(struct notifier_block *this, 1330 unsigned long event, void *ptr) 1331{ 1332 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1333 struct net *net = dev_net(dev); 1334 struct mr6_table *mrt; 1335 struct mif_device *v; 1336 int ct; 1337 LIST_HEAD(list); 1338 1339 if (event != NETDEV_UNREGISTER) 1340 return NOTIFY_DONE; 1341 1342 ip6mr_for_each_table(mrt, net) { 1343 v = &mrt->vif6_table[0]; 1344 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1345 if (v->dev == dev) 1346 mif6_delete(mrt, ct, &list); 1347 } 1348 } 1349 unregister_netdevice_many(&list); 1350 1351 return NOTIFY_DONE; 1352} 1353 1354static struct notifier_block ip6_mr_notifier = { 1355 .notifier_call = ip6mr_device_event 1356}; 1357 1358/* 1359 * Setup for IP multicast routing 1360 */ 1361 1362static int __net_init ip6mr_net_init(struct net *net) 1363{ 1364 int err; 1365 1366 err = ip6mr_rules_init(net); 1367 if (err < 0) 1368 goto fail; 1369 1370#ifdef CONFIG_PROC_FS 1371 err = -ENOMEM; 1372 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops)) 1373 goto proc_vif_fail; 1374 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops)) 1375 goto proc_cache_fail; 1376#endif 1377 1378 return 0; 1379 1380#ifdef CONFIG_PROC_FS 1381proc_cache_fail: 1382 remove_proc_entry("ip6_mr_vif", net->proc_net); 1383proc_vif_fail: 1384 ip6mr_rules_exit(net); 1385#endif 1386fail: 1387 return err; 1388} 1389 1390static void __net_exit ip6mr_net_exit(struct net *net) 1391{ 1392#ifdef CONFIG_PROC_FS 1393 remove_proc_entry("ip6_mr_cache", net->proc_net); 1394 remove_proc_entry("ip6_mr_vif", net->proc_net); 1395#endif 1396 ip6mr_rules_exit(net); 1397} 1398 1399static struct pernet_operations ip6mr_net_ops = { 1400 .init = ip6mr_net_init, 1401 .exit = ip6mr_net_exit, 1402}; 1403 1404int __init ip6_mr_init(void) 1405{ 1406 int err; 1407 1408 mrt_cachep = kmem_cache_create("ip6_mrt_cache", 1409 sizeof(struct mfc6_cache), 1410 0, SLAB_HWCACHE_ALIGN, 1411 NULL); 1412 if (!mrt_cachep) 1413 return -ENOMEM; 1414 1415 err = register_pernet_subsys(&ip6mr_net_ops); 1416 if (err) 1417 goto reg_pernet_fail; 1418 1419 err = register_netdevice_notifier(&ip6_mr_notifier); 1420 if (err) 1421 goto reg_notif_fail; 1422#ifdef CONFIG_IPV6_PIMSM_V2 1423 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) { 1424 pr_err("%s: can't add PIM protocol\n", __func__); 1425 err = -EAGAIN; 1426 goto add_proto_fail; 1427 } 1428#endif 1429 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, 1430 ip6mr_rtm_dumproute, NULL); 1431 return 0; 1432#ifdef CONFIG_IPV6_PIMSM_V2 1433add_proto_fail: 1434 unregister_netdevice_notifier(&ip6_mr_notifier); 1435#endif 1436reg_notif_fail: 1437 unregister_pernet_subsys(&ip6mr_net_ops); 1438reg_pernet_fail: 1439 kmem_cache_destroy(mrt_cachep); 1440 return err; 1441} 1442 1443void ip6_mr_cleanup(void) 1444{ 1445 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE); 1446#ifdef CONFIG_IPV6_PIMSM_V2 1447 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM); 1448#endif 1449 unregister_netdevice_notifier(&ip6_mr_notifier); 1450 unregister_pernet_subsys(&ip6mr_net_ops); 1451 kmem_cache_destroy(mrt_cachep); 1452} 1453 1454static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, 1455 struct mf6cctl *mfc, int mrtsock, int parent) 1456{ 1457 bool found = false; 1458 int line; 1459 struct mfc6_cache *uc, *c; 1460 unsigned char ttls[MAXMIFS]; 1461 int i; 1462 1463 if (mfc->mf6cc_parent >= MAXMIFS) 1464 return -ENFILE; 1465 1466 memset(ttls, 255, MAXMIFS); 1467 for (i = 0; i < MAXMIFS; i++) { 1468 if (IF_ISSET(i, &mfc->mf6cc_ifset)) 1469 ttls[i] = 1; 1470 1471 } 1472 1473 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); 1474 1475 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) { 1476 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && 1477 ipv6_addr_equal(&c->mf6c_mcastgrp, 1478 &mfc->mf6cc_mcastgrp.sin6_addr) && 1479 (parent == -1 || parent == mfc->mf6cc_parent)) { 1480 found = true; 1481 break; 1482 } 1483 } 1484 1485 if (found) { 1486 write_lock_bh(&mrt_lock); 1487 c->mf6c_parent = mfc->mf6cc_parent; 1488 ip6mr_update_thresholds(mrt, c, ttls); 1489 if (!mrtsock) 1490 c->mfc_flags |= MFC_STATIC; 1491 write_unlock_bh(&mrt_lock); 1492 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1493 return 0; 1494 } 1495 1496 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) && 1497 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) 1498 return -EINVAL; 1499 1500 c = ip6mr_cache_alloc(); 1501 if (!c) 1502 return -ENOMEM; 1503 1504 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; 1505 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; 1506 c->mf6c_parent = mfc->mf6cc_parent; 1507 ip6mr_update_thresholds(mrt, c, ttls); 1508 if (!mrtsock) 1509 c->mfc_flags |= MFC_STATIC; 1510 1511 write_lock_bh(&mrt_lock); 1512 list_add(&c->list, &mrt->mfc6_cache_array[line]); 1513 write_unlock_bh(&mrt_lock); 1514 1515 /* 1516 * Check to see if we resolved a queued list. If so we 1517 * need to send on the frames and tidy up. 1518 */ 1519 found = false; 1520 spin_lock_bh(&mfc_unres_lock); 1521 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) { 1522 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && 1523 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { 1524 list_del(&uc->list); 1525 atomic_dec(&mrt->cache_resolve_queue_len); 1526 found = true; 1527 break; 1528 } 1529 } 1530 if (list_empty(&mrt->mfc6_unres_queue)) 1531 del_timer(&mrt->ipmr_expire_timer); 1532 spin_unlock_bh(&mfc_unres_lock); 1533 1534 if (found) { 1535 ip6mr_cache_resolve(net, mrt, uc, c); 1536 ip6mr_cache_free(uc); 1537 } 1538 mr6_netlink_event(mrt, c, RTM_NEWROUTE); 1539 return 0; 1540} 1541 1542/* 1543 * Close the multicast socket, and clear the vif tables etc 1544 */ 1545 1546static void mroute_clean_tables(struct mr6_table *mrt, bool all) 1547{ 1548 int i; 1549 LIST_HEAD(list); 1550 struct mfc6_cache *c, *next; 1551 1552 /* 1553 * Shut down all active vif entries 1554 */ 1555 for (i = 0; i < mrt->maxvif; i++) { 1556 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) 1557 continue; 1558 mif6_delete(mrt, i, &list); 1559 } 1560 unregister_netdevice_many(&list); 1561 1562 /* 1563 * Wipe the cache 1564 */ 1565 for (i = 0; i < MFC6_LINES; i++) { 1566 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { 1567 if (!all && (c->mfc_flags & MFC_STATIC)) 1568 continue; 1569 write_lock_bh(&mrt_lock); 1570 list_del(&c->list); 1571 write_unlock_bh(&mrt_lock); 1572 1573 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1574 ip6mr_cache_free(c); 1575 } 1576 } 1577 1578 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1579 spin_lock_bh(&mfc_unres_lock); 1580 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) { 1581 list_del(&c->list); 1582 mr6_netlink_event(mrt, c, RTM_DELROUTE); 1583 ip6mr_destroy_unres(mrt, c); 1584 } 1585 spin_unlock_bh(&mfc_unres_lock); 1586 } 1587} 1588 1589static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk) 1590{ 1591 int err = 0; 1592 struct net *net = sock_net(sk); 1593 1594 rtnl_lock(); 1595 write_lock_bh(&mrt_lock); 1596 if (likely(mrt->mroute6_sk == NULL)) { 1597 mrt->mroute6_sk = sk; 1598 net->ipv6.devconf_all->mc_forwarding++; 1599 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1600 NETCONFA_IFINDEX_ALL, 1601 net->ipv6.devconf_all); 1602 } 1603 else 1604 err = -EADDRINUSE; 1605 write_unlock_bh(&mrt_lock); 1606 1607 rtnl_unlock(); 1608 1609 return err; 1610} 1611 1612int ip6mr_sk_done(struct sock *sk) 1613{ 1614 int err = -EACCES; 1615 struct net *net = sock_net(sk); 1616 struct mr6_table *mrt; 1617 1618 rtnl_lock(); 1619 ip6mr_for_each_table(mrt, net) { 1620 if (sk == mrt->mroute6_sk) { 1621 write_lock_bh(&mrt_lock); 1622 mrt->mroute6_sk = NULL; 1623 net->ipv6.devconf_all->mc_forwarding--; 1624 inet6_netconf_notify_devconf(net, 1625 NETCONFA_MC_FORWARDING, 1626 NETCONFA_IFINDEX_ALL, 1627 net->ipv6.devconf_all); 1628 write_unlock_bh(&mrt_lock); 1629 1630 mroute_clean_tables(mrt, false); 1631 err = 0; 1632 break; 1633 } 1634 } 1635 rtnl_unlock(); 1636 1637 return err; 1638} 1639 1640struct sock *mroute6_socket(struct net *net, struct sk_buff *skb) 1641{ 1642 struct mr6_table *mrt; 1643 struct flowi6 fl6 = { 1644 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 1645 .flowi6_oif = skb->dev->ifindex, 1646 .flowi6_mark = skb->mark, 1647 }; 1648 1649 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0) 1650 return NULL; 1651 1652 return mrt->mroute6_sk; 1653} 1654 1655/* 1656 * Socket options and virtual interface manipulation. The whole 1657 * virtual interface system is a complete heap, but unfortunately 1658 * that's how BSD mrouted happens to think. Maybe one day with a proper 1659 * MOSPF/PIM router set up we can clean this up. 1660 */ 1661 1662int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1663{ 1664 int ret, parent = 0; 1665 struct mif6ctl vif; 1666 struct mf6cctl mfc; 1667 mifi_t mifi; 1668 struct net *net = sock_net(sk); 1669 struct mr6_table *mrt; 1670 1671 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1672 if (!mrt) 1673 return -ENOENT; 1674 1675 if (optname != MRT6_INIT) { 1676 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1677 return -EACCES; 1678 } 1679 1680 switch (optname) { 1681 case MRT6_INIT: 1682 if (sk->sk_type != SOCK_RAW || 1683 inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1684 return -EOPNOTSUPP; 1685 if (optlen < sizeof(int)) 1686 return -EINVAL; 1687 1688 return ip6mr_sk_init(mrt, sk); 1689 1690 case MRT6_DONE: 1691 return ip6mr_sk_done(sk); 1692 1693 case MRT6_ADD_MIF: 1694 if (optlen < sizeof(vif)) 1695 return -EINVAL; 1696 if (copy_from_user(&vif, optval, sizeof(vif))) 1697 return -EFAULT; 1698 if (vif.mif6c_mifi >= MAXMIFS) 1699 return -ENFILE; 1700 rtnl_lock(); 1701 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk); 1702 rtnl_unlock(); 1703 return ret; 1704 1705 case MRT6_DEL_MIF: 1706 if (optlen < sizeof(mifi_t)) 1707 return -EINVAL; 1708 if (copy_from_user(&mifi, optval, sizeof(mifi_t))) 1709 return -EFAULT; 1710 rtnl_lock(); 1711 ret = mif6_delete(mrt, mifi, NULL); 1712 rtnl_unlock(); 1713 return ret; 1714 1715 /* 1716 * Manipulate the forwarding caches. These live 1717 * in a sort of kernel/user symbiosis. 1718 */ 1719 case MRT6_ADD_MFC: 1720 case MRT6_DEL_MFC: 1721 parent = -1; 1722 case MRT6_ADD_MFC_PROXY: 1723 case MRT6_DEL_MFC_PROXY: 1724 if (optlen < sizeof(mfc)) 1725 return -EINVAL; 1726 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1727 return -EFAULT; 1728 if (parent == 0) 1729 parent = mfc.mf6cc_parent; 1730 rtnl_lock(); 1731 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY) 1732 ret = ip6mr_mfc_delete(mrt, &mfc, parent); 1733 else 1734 ret = ip6mr_mfc_add(net, mrt, &mfc, 1735 sk == mrt->mroute6_sk, parent); 1736 rtnl_unlock(); 1737 return ret; 1738 1739 /* 1740 * Control PIM assert (to activate pim will activate assert) 1741 */ 1742 case MRT6_ASSERT: 1743 { 1744 int v; 1745 1746 if (optlen != sizeof(v)) 1747 return -EINVAL; 1748 if (get_user(v, (int __user *)optval)) 1749 return -EFAULT; 1750 mrt->mroute_do_assert = v; 1751 return 0; 1752 } 1753 1754#ifdef CONFIG_IPV6_PIMSM_V2 1755 case MRT6_PIM: 1756 { 1757 int v; 1758 1759 if (optlen != sizeof(v)) 1760 return -EINVAL; 1761 if (get_user(v, (int __user *)optval)) 1762 return -EFAULT; 1763 v = !!v; 1764 rtnl_lock(); 1765 ret = 0; 1766 if (v != mrt->mroute_do_pim) { 1767 mrt->mroute_do_pim = v; 1768 mrt->mroute_do_assert = v; 1769 } 1770 rtnl_unlock(); 1771 return ret; 1772 } 1773 1774#endif 1775#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES 1776 case MRT6_TABLE: 1777 { 1778 u32 v; 1779 1780 if (optlen != sizeof(u32)) 1781 return -EINVAL; 1782 if (get_user(v, (u32 __user *)optval)) 1783 return -EFAULT; 1784 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1785 if (v != RT_TABLE_DEFAULT && v >= 100000000) 1786 return -EINVAL; 1787 if (sk == mrt->mroute6_sk) 1788 return -EBUSY; 1789 1790 rtnl_lock(); 1791 ret = 0; 1792 if (!ip6mr_new_table(net, v)) 1793 ret = -ENOMEM; 1794 raw6_sk(sk)->ip6mr_table = v; 1795 rtnl_unlock(); 1796 return ret; 1797 } 1798#endif 1799 /* 1800 * Spurious command, or MRT6_VERSION which you cannot 1801 * set. 1802 */ 1803 default: 1804 return -ENOPROTOOPT; 1805 } 1806} 1807 1808/* 1809 * Getsock opt support for the multicast routing system. 1810 */ 1811 1812int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, 1813 int __user *optlen) 1814{ 1815 int olr; 1816 int val; 1817 struct net *net = sock_net(sk); 1818 struct mr6_table *mrt; 1819 1820 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1821 if (!mrt) 1822 return -ENOENT; 1823 1824 switch (optname) { 1825 case MRT6_VERSION: 1826 val = 0x0305; 1827 break; 1828#ifdef CONFIG_IPV6_PIMSM_V2 1829 case MRT6_PIM: 1830 val = mrt->mroute_do_pim; 1831 break; 1832#endif 1833 case MRT6_ASSERT: 1834 val = mrt->mroute_do_assert; 1835 break; 1836 default: 1837 return -ENOPROTOOPT; 1838 } 1839 1840 if (get_user(olr, optlen)) 1841 return -EFAULT; 1842 1843 olr = min_t(int, olr, sizeof(int)); 1844 if (olr < 0) 1845 return -EINVAL; 1846 1847 if (put_user(olr, optlen)) 1848 return -EFAULT; 1849 if (copy_to_user(optval, &val, olr)) 1850 return -EFAULT; 1851 return 0; 1852} 1853 1854/* 1855 * The IP multicast ioctl support routines. 1856 */ 1857 1858int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) 1859{ 1860 struct sioc_sg_req6 sr; 1861 struct 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) 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 1915#ifdef CONFIG_COMPAT 1916struct compat_sioc_sg_req6 { 1917 struct sockaddr_in6 src; 1918 struct sockaddr_in6 grp; 1919 compat_ulong_t pktcnt; 1920 compat_ulong_t bytecnt; 1921 compat_ulong_t wrong_if; 1922}; 1923 1924struct compat_sioc_mif_req6 { 1925 mifi_t mifi; 1926 compat_ulong_t icount; 1927 compat_ulong_t ocount; 1928 compat_ulong_t ibytes; 1929 compat_ulong_t obytes; 1930}; 1931 1932int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1933{ 1934 struct compat_sioc_sg_req6 sr; 1935 struct compat_sioc_mif_req6 vr; 1936 struct mif_device *vif; 1937 struct mfc6_cache *c; 1938 struct net *net = sock_net(sk); 1939 struct mr6_table *mrt; 1940 1941 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT); 1942 if (!mrt) 1943 return -ENOENT; 1944 1945 switch (cmd) { 1946 case SIOCGETMIFCNT_IN6: 1947 if (copy_from_user(&vr, arg, sizeof(vr))) 1948 return -EFAULT; 1949 if (vr.mifi >= mrt->maxvif) 1950 return -EINVAL; 1951 read_lock(&mrt_lock); 1952 vif = &mrt->vif6_table[vr.mifi]; 1953 if (MIF_EXISTS(mrt, vr.mifi)) { 1954 vr.icount = vif->pkt_in; 1955 vr.ocount = vif->pkt_out; 1956 vr.ibytes = vif->bytes_in; 1957 vr.obytes = vif->bytes_out; 1958 read_unlock(&mrt_lock); 1959 1960 if (copy_to_user(arg, &vr, sizeof(vr))) 1961 return -EFAULT; 1962 return 0; 1963 } 1964 read_unlock(&mrt_lock); 1965 return -EADDRNOTAVAIL; 1966 case SIOCGETSGCNT_IN6: 1967 if (copy_from_user(&sr, arg, sizeof(sr))) 1968 return -EFAULT; 1969 1970 read_lock(&mrt_lock); 1971 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr); 1972 if (c) { 1973 sr.pktcnt = c->mfc_un.res.pkt; 1974 sr.bytecnt = c->mfc_un.res.bytes; 1975 sr.wrong_if = c->mfc_un.res.wrong_if; 1976 read_unlock(&mrt_lock); 1977 1978 if (copy_to_user(arg, &sr, sizeof(sr))) 1979 return -EFAULT; 1980 return 0; 1981 } 1982 read_unlock(&mrt_lock); 1983 return -EADDRNOTAVAIL; 1984 default: 1985 return -ENOIOCTLCMD; 1986 } 1987} 1988#endif 1989 1990static inline int ip6mr_forward2_finish(struct sock *sk, struct sk_buff *skb) 1991{ 1992 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)), 1993 IPSTATS_MIB_OUTFORWDATAGRAMS); 1994 IP6_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)), 1995 IPSTATS_MIB_OUTOCTETS, skb->len); 1996 return dst_output_sk(sk, skb); 1997} 1998 1999/* 2000 * Processing handlers for ip6mr_forward 2001 */ 2002 2003static int ip6mr_forward2(struct net *net, struct mr6_table *mrt, 2004 struct sk_buff *skb, struct mfc6_cache *c, int vifi) 2005{ 2006 struct ipv6hdr *ipv6h; 2007 struct mif_device *vif = &mrt->vif6_table[vifi]; 2008 struct net_device *dev; 2009 struct dst_entry *dst; 2010 struct flowi6 fl6; 2011 2012 if (!vif->dev) 2013 goto out_free; 2014 2015#ifdef CONFIG_IPV6_PIMSM_V2 2016 if (vif->flags & MIFF_REGISTER) { 2017 vif->pkt_out++; 2018 vif->bytes_out += skb->len; 2019 vif->dev->stats.tx_bytes += skb->len; 2020 vif->dev->stats.tx_packets++; 2021 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT); 2022 goto out_free; 2023 } 2024#endif 2025 2026 ipv6h = ipv6_hdr(skb); 2027 2028 fl6 = (struct flowi6) { 2029 .flowi6_oif = vif->link, 2030 .daddr = ipv6h->daddr, 2031 }; 2032 2033 dst = ip6_route_output(net, NULL, &fl6); 2034 if (dst->error) { 2035 dst_release(dst); 2036 goto out_free; 2037 } 2038 2039 skb_dst_drop(skb); 2040 skb_dst_set(skb, dst); 2041 2042 /* 2043 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 2044 * not only before forwarding, but after forwarding on all output 2045 * interfaces. It is clear, if mrouter runs a multicasting 2046 * program, it should receive packets not depending to what interface 2047 * program is joined. 2048 * If we will not make it, the program will have to join on all 2049 * interfaces. On the other hand, multihoming host (or router, but 2050 * not mrouter) cannot join to more than one interface - it will 2051 * result in receiving multiple packets. 2052 */ 2053 dev = vif->dev; 2054 skb->dev = dev; 2055 vif->pkt_out++; 2056 vif->bytes_out += skb->len; 2057 2058 /* We are about to write */ 2059 /* XXX: extension headers? */ 2060 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) 2061 goto out_free; 2062 2063 ipv6h = ipv6_hdr(skb); 2064 ipv6h->hop_limit--; 2065 2066 IP6CB(skb)->flags |= IP6SKB_FORWARDED; 2067 2068 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, NULL, skb, 2069 skb->dev, dev, 2070 ip6mr_forward2_finish); 2071 2072out_free: 2073 kfree_skb(skb); 2074 return 0; 2075} 2076 2077static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev) 2078{ 2079 int ct; 2080 2081 for (ct = mrt->maxvif - 1; ct >= 0; ct--) { 2082 if (mrt->vif6_table[ct].dev == dev) 2083 break; 2084 } 2085 return ct; 2086} 2087 2088static void ip6_mr_forward(struct net *net, struct mr6_table *mrt, 2089 struct sk_buff *skb, struct mfc6_cache *cache) 2090{ 2091 int psend = -1; 2092 int vif, ct; 2093 int true_vifi = ip6mr_find_vif(mrt, skb->dev); 2094 2095 vif = cache->mf6c_parent; 2096 cache->mfc_un.res.pkt++; 2097 cache->mfc_un.res.bytes += skb->len; 2098 2099 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) { 2100 struct mfc6_cache *cache_proxy; 2101 2102 /* For an (*,G) entry, we only check that the incoming 2103 * interface is part of the static tree. 2104 */ 2105 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif); 2106 if (cache_proxy && 2107 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 2108 goto forward; 2109 } 2110 2111 /* 2112 * Wrong interface: drop packet and (maybe) send PIM assert. 2113 */ 2114 if (mrt->vif6_table[vif].dev != skb->dev) { 2115 cache->mfc_un.res.wrong_if++; 2116 2117 if (true_vifi >= 0 && mrt->mroute_do_assert && 2118 /* pimsm uses asserts, when switching from RPT to SPT, 2119 so that we cannot check that packet arrived on an oif. 2120 It is bad, but otherwise we would need to move pretty 2121 large chunk of pimd to kernel. Ough... --ANK 2122 */ 2123 (mrt->mroute_do_pim || 2124 cache->mfc_un.res.ttls[true_vifi] < 255) && 2125 time_after(jiffies, 2126 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 2127 cache->mfc_un.res.last_assert = jiffies; 2128 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF); 2129 } 2130 goto dont_forward; 2131 } 2132 2133forward: 2134 mrt->vif6_table[vif].pkt_in++; 2135 mrt->vif6_table[vif].bytes_in += skb->len; 2136 2137 /* 2138 * Forward the frame 2139 */ 2140 if (ipv6_addr_any(&cache->mf6c_origin) && 2141 ipv6_addr_any(&cache->mf6c_mcastgrp)) { 2142 if (true_vifi >= 0 && 2143 true_vifi != cache->mf6c_parent && 2144 ipv6_hdr(skb)->hop_limit > 2145 cache->mfc_un.res.ttls[cache->mf6c_parent]) { 2146 /* It's an (*,*) entry and the packet is not coming from 2147 * the upstream: forward the packet to the upstream 2148 * only. 2149 */ 2150 psend = cache->mf6c_parent; 2151 goto last_forward; 2152 } 2153 goto dont_forward; 2154 } 2155 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { 2156 /* For (*,G) entry, don't forward to the incoming interface */ 2157 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) && 2158 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { 2159 if (psend != -1) { 2160 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2161 if (skb2) 2162 ip6mr_forward2(net, mrt, skb2, cache, psend); 2163 } 2164 psend = ct; 2165 } 2166 } 2167last_forward: 2168 if (psend != -1) { 2169 ip6mr_forward2(net, mrt, skb, cache, psend); 2170 return; 2171 } 2172 2173dont_forward: 2174 kfree_skb(skb); 2175} 2176 2177 2178/* 2179 * Multicast packets for forwarding arrive here 2180 */ 2181 2182int ip6_mr_input(struct sk_buff *skb) 2183{ 2184 struct mfc6_cache *cache; 2185 struct net *net = dev_net(skb->dev); 2186 struct mr6_table *mrt; 2187 struct flowi6 fl6 = { 2188 .flowi6_iif = skb->dev->ifindex, 2189 .flowi6_mark = skb->mark, 2190 }; 2191 int err; 2192 2193 err = ip6mr_fib_lookup(net, &fl6, &mrt); 2194 if (err < 0) { 2195 kfree_skb(skb); 2196 return err; 2197 } 2198 2199 read_lock(&mrt_lock); 2200 cache = ip6mr_cache_find(mrt, 2201 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); 2202 if (!cache) { 2203 int vif = ip6mr_find_vif(mrt, skb->dev); 2204 2205 if (vif >= 0) 2206 cache = ip6mr_cache_find_any(mrt, 2207 &ipv6_hdr(skb)->daddr, 2208 vif); 2209 } 2210 2211 /* 2212 * No usable cache entry 2213 */ 2214 if (!cache) { 2215 int vif; 2216 2217 vif = ip6mr_find_vif(mrt, skb->dev); 2218 if (vif >= 0) { 2219 int err = ip6mr_cache_unresolved(mrt, vif, skb); 2220 read_unlock(&mrt_lock); 2221 2222 return err; 2223 } 2224 read_unlock(&mrt_lock); 2225 kfree_skb(skb); 2226 return -ENODEV; 2227 } 2228 2229 ip6_mr_forward(net, mrt, skb, cache); 2230 2231 read_unlock(&mrt_lock); 2232 2233 return 0; 2234} 2235 2236 2237static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2238 struct mfc6_cache *c, struct rtmsg *rtm) 2239{ 2240 int ct; 2241 struct rtnexthop *nhp; 2242 struct nlattr *mp_attr; 2243 struct rta_mfc_stats mfcs; 2244 2245 /* If cache is unresolved, don't try to parse IIF and OIF */ 2246 if (c->mf6c_parent >= MAXMIFS) 2247 return -ENOENT; 2248 2249 if (MIF_EXISTS(mrt, c->mf6c_parent) && 2250 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0) 2251 return -EMSGSIZE; 2252 mp_attr = nla_nest_start(skb, RTA_MULTIPATH); 2253 if (!mp_attr) 2254 return -EMSGSIZE; 2255 2256 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2257 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2258 nhp = nla_reserve_nohdr(skb, sizeof(*nhp)); 2259 if (!nhp) { 2260 nla_nest_cancel(skb, mp_attr); 2261 return -EMSGSIZE; 2262 } 2263 2264 nhp->rtnh_flags = 0; 2265 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2266 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex; 2267 nhp->rtnh_len = sizeof(*nhp); 2268 } 2269 } 2270 2271 nla_nest_end(skb, mp_attr); 2272 2273 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2274 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2275 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2276 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0) 2277 return -EMSGSIZE; 2278 2279 rtm->rtm_type = RTN_MULTICAST; 2280 return 1; 2281} 2282 2283int ip6mr_get_route(struct net *net, 2284 struct sk_buff *skb, struct rtmsg *rtm, int nowait) 2285{ 2286 int err; 2287 struct mr6_table *mrt; 2288 struct mfc6_cache *cache; 2289 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); 2290 2291 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT); 2292 if (!mrt) 2293 return -ENOENT; 2294 2295 read_lock(&mrt_lock); 2296 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr); 2297 if (!cache && skb->dev) { 2298 int vif = ip6mr_find_vif(mrt, skb->dev); 2299 2300 if (vif >= 0) 2301 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr, 2302 vif); 2303 } 2304 2305 if (!cache) { 2306 struct sk_buff *skb2; 2307 struct ipv6hdr *iph; 2308 struct net_device *dev; 2309 int vif; 2310 2311 if (nowait) { 2312 read_unlock(&mrt_lock); 2313 return -EAGAIN; 2314 } 2315 2316 dev = skb->dev; 2317 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) { 2318 read_unlock(&mrt_lock); 2319 return -ENODEV; 2320 } 2321 2322 /* really correct? */ 2323 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 2324 if (!skb2) { 2325 read_unlock(&mrt_lock); 2326 return -ENOMEM; 2327 } 2328 2329 skb_reset_transport_header(skb2); 2330 2331 skb_put(skb2, sizeof(struct ipv6hdr)); 2332 skb_reset_network_header(skb2); 2333 2334 iph = ipv6_hdr(skb2); 2335 iph->version = 0; 2336 iph->priority = 0; 2337 iph->flow_lbl[0] = 0; 2338 iph->flow_lbl[1] = 0; 2339 iph->flow_lbl[2] = 0; 2340 iph->payload_len = 0; 2341 iph->nexthdr = IPPROTO_NONE; 2342 iph->hop_limit = 0; 2343 iph->saddr = rt->rt6i_src.addr; 2344 iph->daddr = rt->rt6i_dst.addr; 2345 2346 err = ip6mr_cache_unresolved(mrt, vif, skb2); 2347 read_unlock(&mrt_lock); 2348 2349 return err; 2350 } 2351 2352 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) 2353 cache->mfc_flags |= MFC_NOTIFY; 2354 2355 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm); 2356 read_unlock(&mrt_lock); 2357 return err; 2358} 2359 2360static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb, 2361 u32 portid, u32 seq, struct mfc6_cache *c, int cmd, 2362 int flags) 2363{ 2364 struct nlmsghdr *nlh; 2365 struct rtmsg *rtm; 2366 int err; 2367 2368 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2369 if (!nlh) 2370 return -EMSGSIZE; 2371 2372 rtm = nlmsg_data(nlh); 2373 rtm->rtm_family = RTNL_FAMILY_IP6MR; 2374 rtm->rtm_dst_len = 128; 2375 rtm->rtm_src_len = 128; 2376 rtm->rtm_tos = 0; 2377 rtm->rtm_table = mrt->id; 2378 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2379 goto nla_put_failure; 2380 rtm->rtm_type = RTN_MULTICAST; 2381 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2382 if (c->mfc_flags & MFC_STATIC) 2383 rtm->rtm_protocol = RTPROT_STATIC; 2384 else 2385 rtm->rtm_protocol = RTPROT_MROUTED; 2386 rtm->rtm_flags = 0; 2387 2388 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) || 2389 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp)) 2390 goto nla_put_failure; 2391 err = __ip6mr_fill_mroute(mrt, skb, c, rtm); 2392 /* do not break the dump if cache is unresolved */ 2393 if (err < 0 && err != -ENOENT) 2394 goto nla_put_failure; 2395 2396 nlmsg_end(skb, nlh); 2397 return 0; 2398 2399nla_put_failure: 2400 nlmsg_cancel(skb, nlh); 2401 return -EMSGSIZE; 2402} 2403 2404static int mr6_msgsize(bool unresolved, int maxvif) 2405{ 2406 size_t len = 2407 NLMSG_ALIGN(sizeof(struct rtmsg)) 2408 + nla_total_size(4) /* RTA_TABLE */ 2409 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */ 2410 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */ 2411 ; 2412 2413 if (!unresolved) 2414 len = len 2415 + nla_total_size(4) /* RTA_IIF */ 2416 + nla_total_size(0) /* RTA_MULTIPATH */ 2417 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2418 /* RTA_MFC_STATS */ 2419 + nla_total_size(sizeof(struct rta_mfc_stats)) 2420 ; 2421 2422 return len; 2423} 2424 2425static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, 2426 int cmd) 2427{ 2428 struct net *net = read_pnet(&mrt->net); 2429 struct sk_buff *skb; 2430 int err = -ENOBUFS; 2431 2432 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif), 2433 GFP_ATOMIC); 2434 if (!skb) 2435 goto errout; 2436 2437 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2438 if (err < 0) 2439 goto errout; 2440 2441 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC); 2442 return; 2443 2444errout: 2445 kfree_skb(skb); 2446 if (err < 0) 2447 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err); 2448} 2449 2450static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2451{ 2452 struct net *net = sock_net(skb->sk); 2453 struct mr6_table *mrt; 2454 struct mfc6_cache *mfc; 2455 unsigned int t = 0, s_t; 2456 unsigned int h = 0, s_h; 2457 unsigned int e = 0, s_e; 2458 2459 s_t = cb->args[0]; 2460 s_h = cb->args[1]; 2461 s_e = cb->args[2]; 2462 2463 read_lock(&mrt_lock); 2464 ip6mr_for_each_table(mrt, net) { 2465 if (t < s_t) 2466 goto next_table; 2467 if (t > s_t) 2468 s_h = 0; 2469 for (h = s_h; h < MFC6_LINES; h++) { 2470 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) { 2471 if (e < s_e) 2472 goto next_entry; 2473 if (ip6mr_fill_mroute(mrt, skb, 2474 NETLINK_CB(cb->skb).portid, 2475 cb->nlh->nlmsg_seq, 2476 mfc, RTM_NEWROUTE, 2477 NLM_F_MULTI) < 0) 2478 goto done; 2479next_entry: 2480 e++; 2481 } 2482 e = s_e = 0; 2483 } 2484 spin_lock_bh(&mfc_unres_lock); 2485 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) { 2486 if (e < s_e) 2487 goto next_entry2; 2488 if (ip6mr_fill_mroute(mrt, skb, 2489 NETLINK_CB(cb->skb).portid, 2490 cb->nlh->nlmsg_seq, 2491 mfc, RTM_NEWROUTE, 2492 NLM_F_MULTI) < 0) { 2493 spin_unlock_bh(&mfc_unres_lock); 2494 goto done; 2495 } 2496next_entry2: 2497 e++; 2498 } 2499 spin_unlock_bh(&mfc_unres_lock); 2500 e = s_e = 0; 2501 s_h = 0; 2502next_table: 2503 t++; 2504 } 2505done: 2506 read_unlock(&mrt_lock); 2507 2508 cb->args[2] = e; 2509 cb->args[1] = h; 2510 cb->args[0] = t; 2511 2512 return skb->len; 2513} 2514