1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IP multicast routing support for mrouted 3.6/3.8 4 * 5 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk> 6 * Linux Consultancy and Custom Driver Development 7 * 8 * Fixes: 9 * Michael Chastain : Incorrect size of copying. 10 * Alan Cox : Added the cache manager code 11 * Alan Cox : Fixed the clone/copy bug and device race. 12 * Mike McLagan : Routing by source 13 * Malcolm Beattie : Buffer handling fixes. 14 * Alexey Kuznetsov : Double buffer free and other fixes. 15 * SVR Anand : Fixed several multicast bugs and problems. 16 * Alexey Kuznetsov : Status, optimisations and more. 17 * Brad Parker : Better behaviour on mrouted upcall 18 * overflow. 19 * Carlos Picoto : PIMv1 Support 20 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header 21 * Relax this requirement to work with older peers. 22 */ 23 24 #include <linux/uaccess.h> 25 #include <linux/types.h> 26 #include <linux/cache.h> 27 #include <linux/capability.h> 28 #include <linux/errno.h> 29 #include <linux/mm.h> 30 #include <linux/kernel.h> 31 #include <linux/fcntl.h> 32 #include <linux/stat.h> 33 #include <linux/socket.h> 34 #include <linux/in.h> 35 #include <linux/inet.h> 36 #include <linux/netdevice.h> 37 #include <linux/inetdevice.h> 38 #include <linux/igmp.h> 39 #include <linux/proc_fs.h> 40 #include <linux/seq_file.h> 41 #include <linux/mroute.h> 42 #include <linux/init.h> 43 #include <linux/if_ether.h> 44 #include <linux/slab.h> 45 #include <net/net_namespace.h> 46 #include <net/ip.h> 47 #include <net/protocol.h> 48 #include <linux/skbuff.h> 49 #include <net/route.h> 50 #include <net/icmp.h> 51 #include <net/udp.h> 52 #include <net/raw.h> 53 #include <linux/notifier.h> 54 #include <linux/if_arp.h> 55 #include <linux/netfilter_ipv4.h> 56 #include <linux/compat.h> 57 #include <linux/export.h> 58 #include <linux/rhashtable.h> 59 #include <net/ip_tunnels.h> 60 #include <net/checksum.h> 61 #include <net/netlink.h> 62 #include <net/fib_rules.h> 63 #include <linux/netconf.h> 64 #include <net/rtnh.h> 65 66 #include <linux/nospec.h> 67 68 struct ipmr_rule { 69 struct fib_rule common; 70 }; 71 72 struct ipmr_result { 73 struct mr_table *mrt; 74 }; 75 76 /* Big lock, protecting vif table, mrt cache and mroute socket state. 77 * Note that the changes are semaphored via rtnl_lock. 78 */ 79 80 static DEFINE_SPINLOCK(mrt_lock); 81 82 static struct net_device *vif_dev_read(const struct vif_device *vif) 83 { 84 return rcu_dereference(vif->dev); 85 } 86 87 /* Multicast router control variables */ 88 89 /* Special spinlock for queue of unresolved entries */ 90 static DEFINE_SPINLOCK(mfc_unres_lock); 91 92 /* We return to original Alan's scheme. Hash table of resolved 93 * entries is changed only in process context and protected 94 * with weak lock mrt_lock. Queue of unresolved entries is protected 95 * with strong spinlock mfc_unres_lock. 96 * 97 * In this case data path is free of exclusive locks at all. 98 */ 99 100 static struct kmem_cache *mrt_cachep __ro_after_init; 101 102 static struct mr_table *ipmr_new_table(struct net *net, u32 id); 103 static void ipmr_free_table(struct mr_table *mrt); 104 105 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 106 struct net_device *dev, struct sk_buff *skb, 107 struct mfc_cache *cache, int local); 108 static int ipmr_cache_report(const struct mr_table *mrt, 109 struct sk_buff *pkt, vifi_t vifi, int assert); 110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 111 int cmd); 112 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt); 113 static void mroute_clean_tables(struct mr_table *mrt, int flags); 114 static void ipmr_expire_process(struct timer_list *t); 115 116 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 117 #define ipmr_for_each_table(mrt, net) \ 118 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \ 119 lockdep_rtnl_is_held() || \ 120 list_empty(&net->ipv4.mr_tables)) 121 122 static struct mr_table *ipmr_mr_table_iter(struct net *net, 123 struct mr_table *mrt) 124 { 125 struct mr_table *ret; 126 127 if (!mrt) 128 ret = list_entry_rcu(net->ipv4.mr_tables.next, 129 struct mr_table, list); 130 else 131 ret = list_entry_rcu(mrt->list.next, 132 struct mr_table, list); 133 134 if (&ret->list == &net->ipv4.mr_tables) 135 return NULL; 136 return ret; 137 } 138 139 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 140 { 141 struct mr_table *mrt; 142 143 ipmr_for_each_table(mrt, net) { 144 if (mrt->id == id) 145 return mrt; 146 } 147 return NULL; 148 } 149 150 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 151 struct mr_table **mrt) 152 { 153 int err; 154 struct ipmr_result res; 155 struct fib_lookup_arg arg = { 156 .result = &res, 157 .flags = FIB_LOOKUP_NOREF, 158 }; 159 160 /* update flow if oif or iif point to device enslaved to l3mdev */ 161 l3mdev_update_flow(net, flowi4_to_flowi(flp4)); 162 163 err = fib_rules_lookup(net->ipv4.mr_rules_ops, 164 flowi4_to_flowi(flp4), 0, &arg); 165 if (err < 0) 166 return err; 167 *mrt = res.mrt; 168 return 0; 169 } 170 171 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp, 172 int flags, struct fib_lookup_arg *arg) 173 { 174 struct ipmr_result *res = arg->result; 175 struct mr_table *mrt; 176 177 switch (rule->action) { 178 case FR_ACT_TO_TBL: 179 break; 180 case FR_ACT_UNREACHABLE: 181 return -ENETUNREACH; 182 case FR_ACT_PROHIBIT: 183 return -EACCES; 184 case FR_ACT_BLACKHOLE: 185 default: 186 return -EINVAL; 187 } 188 189 arg->table = fib_rule_get_table(rule, arg); 190 191 mrt = ipmr_get_table(rule->fr_net, arg->table); 192 if (!mrt) 193 return -EAGAIN; 194 res->mrt = mrt; 195 return 0; 196 } 197 198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 199 { 200 return 1; 201 } 202 203 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 204 struct fib_rule_hdr *frh, struct nlattr **tb, 205 struct netlink_ext_ack *extack) 206 { 207 return 0; 208 } 209 210 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 211 struct nlattr **tb) 212 { 213 return 1; 214 } 215 216 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 217 struct fib_rule_hdr *frh) 218 { 219 frh->dst_len = 0; 220 frh->src_len = 0; 221 frh->tos = 0; 222 return 0; 223 } 224 225 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = { 226 .family = RTNL_FAMILY_IPMR, 227 .rule_size = sizeof(struct ipmr_rule), 228 .addr_size = sizeof(u32), 229 .action = ipmr_rule_action, 230 .match = ipmr_rule_match, 231 .configure = ipmr_rule_configure, 232 .compare = ipmr_rule_compare, 233 .fill = ipmr_rule_fill, 234 .nlgroup = RTNLGRP_IPV4_RULE, 235 .owner = THIS_MODULE, 236 }; 237 238 static int __net_init ipmr_rules_init(struct net *net) 239 { 240 struct fib_rules_ops *ops; 241 struct mr_table *mrt; 242 int err; 243 244 ops = fib_rules_register(&ipmr_rules_ops_template, net); 245 if (IS_ERR(ops)) 246 return PTR_ERR(ops); 247 248 INIT_LIST_HEAD(&net->ipv4.mr_tables); 249 250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 251 if (IS_ERR(mrt)) { 252 err = PTR_ERR(mrt); 253 goto err1; 254 } 255 256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0); 257 if (err < 0) 258 goto err2; 259 260 net->ipv4.mr_rules_ops = ops; 261 return 0; 262 263 err2: 264 rtnl_lock(); 265 ipmr_free_table(mrt); 266 rtnl_unlock(); 267 err1: 268 fib_rules_unregister(ops); 269 return err; 270 } 271 272 static void __net_exit ipmr_rules_exit(struct net *net) 273 { 274 struct mr_table *mrt, *next; 275 276 ASSERT_RTNL(); 277 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) { 278 list_del(&mrt->list); 279 ipmr_free_table(mrt); 280 } 281 fib_rules_unregister(net->ipv4.mr_rules_ops); 282 } 283 284 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb, 285 struct netlink_ext_ack *extack) 286 { 287 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack); 288 } 289 290 static unsigned int ipmr_rules_seq_read(struct net *net) 291 { 292 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR); 293 } 294 295 bool ipmr_rule_default(const struct fib_rule *rule) 296 { 297 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT; 298 } 299 EXPORT_SYMBOL(ipmr_rule_default); 300 #else 301 #define ipmr_for_each_table(mrt, net) \ 302 for (mrt = net->ipv4.mrt; mrt; mrt = NULL) 303 304 static struct mr_table *ipmr_mr_table_iter(struct net *net, 305 struct mr_table *mrt) 306 { 307 if (!mrt) 308 return net->ipv4.mrt; 309 return NULL; 310 } 311 312 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 313 { 314 return net->ipv4.mrt; 315 } 316 317 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 318 struct mr_table **mrt) 319 { 320 *mrt = net->ipv4.mrt; 321 return 0; 322 } 323 324 static int __net_init ipmr_rules_init(struct net *net) 325 { 326 struct mr_table *mrt; 327 328 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 329 if (IS_ERR(mrt)) 330 return PTR_ERR(mrt); 331 net->ipv4.mrt = mrt; 332 return 0; 333 } 334 335 static void __net_exit ipmr_rules_exit(struct net *net) 336 { 337 ASSERT_RTNL(); 338 ipmr_free_table(net->ipv4.mrt); 339 net->ipv4.mrt = NULL; 340 } 341 342 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb, 343 struct netlink_ext_ack *extack) 344 { 345 return 0; 346 } 347 348 static unsigned int ipmr_rules_seq_read(struct net *net) 349 { 350 return 0; 351 } 352 353 bool ipmr_rule_default(const struct fib_rule *rule) 354 { 355 return true; 356 } 357 EXPORT_SYMBOL(ipmr_rule_default); 358 #endif 359 360 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg, 361 const void *ptr) 362 { 363 const struct mfc_cache_cmp_arg *cmparg = arg->key; 364 const struct mfc_cache *c = ptr; 365 366 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp || 367 cmparg->mfc_origin != c->mfc_origin; 368 } 369 370 static const struct rhashtable_params ipmr_rht_params = { 371 .head_offset = offsetof(struct mr_mfc, mnode), 372 .key_offset = offsetof(struct mfc_cache, cmparg), 373 .key_len = sizeof(struct mfc_cache_cmp_arg), 374 .nelem_hint = 3, 375 .obj_cmpfn = ipmr_hash_cmp, 376 .automatic_shrinking = true, 377 }; 378 379 static void ipmr_new_table_set(struct mr_table *mrt, 380 struct net *net) 381 { 382 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 383 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables); 384 #endif 385 } 386 387 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = { 388 .mfc_mcastgrp = htonl(INADDR_ANY), 389 .mfc_origin = htonl(INADDR_ANY), 390 }; 391 392 static struct mr_table_ops ipmr_mr_table_ops = { 393 .rht_params = &ipmr_rht_params, 394 .cmparg_any = &ipmr_mr_table_ops_cmparg_any, 395 }; 396 397 static struct mr_table *ipmr_new_table(struct net *net, u32 id) 398 { 399 struct mr_table *mrt; 400 401 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */ 402 if (id != RT_TABLE_DEFAULT && id >= 1000000000) 403 return ERR_PTR(-EINVAL); 404 405 mrt = ipmr_get_table(net, id); 406 if (mrt) 407 return mrt; 408 409 return mr_table_alloc(net, id, &ipmr_mr_table_ops, 410 ipmr_expire_process, ipmr_new_table_set); 411 } 412 413 static void ipmr_free_table(struct mr_table *mrt) 414 { 415 del_timer_sync(&mrt->ipmr_expire_timer); 416 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC | 417 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC); 418 rhltable_destroy(&mrt->mfc_hash); 419 kfree(mrt); 420 } 421 422 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ 423 424 /* Initialize ipmr pimreg/tunnel in_device */ 425 static bool ipmr_init_vif_indev(const struct net_device *dev) 426 { 427 struct in_device *in_dev; 428 429 ASSERT_RTNL(); 430 431 in_dev = __in_dev_get_rtnl(dev); 432 if (!in_dev) 433 return false; 434 ipv4_devconf_setall(in_dev); 435 neigh_parms_data_state_setall(in_dev->arp_parms); 436 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 437 438 return true; 439 } 440 441 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v) 442 { 443 struct net_device *tunnel_dev, *new_dev; 444 struct ip_tunnel_parm p = { }; 445 int err; 446 447 tunnel_dev = __dev_get_by_name(net, "tunl0"); 448 if (!tunnel_dev) 449 goto out; 450 451 p.iph.daddr = v->vifc_rmt_addr.s_addr; 452 p.iph.saddr = v->vifc_lcl_addr.s_addr; 453 p.iph.version = 4; 454 p.iph.ihl = 5; 455 p.iph.protocol = IPPROTO_IPIP; 456 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 457 458 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl) 459 goto out; 460 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p, 461 SIOCADDTUNNEL); 462 if (err) 463 goto out; 464 465 new_dev = __dev_get_by_name(net, p.name); 466 if (!new_dev) 467 goto out; 468 469 new_dev->flags |= IFF_MULTICAST; 470 if (!ipmr_init_vif_indev(new_dev)) 471 goto out_unregister; 472 if (dev_open(new_dev, NULL)) 473 goto out_unregister; 474 dev_hold(new_dev); 475 err = dev_set_allmulti(new_dev, 1); 476 if (err) { 477 dev_close(new_dev); 478 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p, 479 SIOCDELTUNNEL); 480 dev_put(new_dev); 481 new_dev = ERR_PTR(err); 482 } 483 return new_dev; 484 485 out_unregister: 486 unregister_netdevice(new_dev); 487 out: 488 return ERR_PTR(-ENOBUFS); 489 } 490 491 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 492 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) 493 { 494 struct net *net = dev_net(dev); 495 struct mr_table *mrt; 496 struct flowi4 fl4 = { 497 .flowi4_oif = dev->ifindex, 498 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 499 .flowi4_mark = skb->mark, 500 }; 501 int err; 502 503 err = ipmr_fib_lookup(net, &fl4, &mrt); 504 if (err < 0) { 505 kfree_skb(skb); 506 return err; 507 } 508 509 dev->stats.tx_bytes += skb->len; 510 dev->stats.tx_packets++; 511 rcu_read_lock(); 512 513 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */ 514 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num), 515 IGMPMSG_WHOLEPKT); 516 517 rcu_read_unlock(); 518 kfree_skb(skb); 519 return NETDEV_TX_OK; 520 } 521 522 static int reg_vif_get_iflink(const struct net_device *dev) 523 { 524 return 0; 525 } 526 527 static const struct net_device_ops reg_vif_netdev_ops = { 528 .ndo_start_xmit = reg_vif_xmit, 529 .ndo_get_iflink = reg_vif_get_iflink, 530 }; 531 532 static void reg_vif_setup(struct net_device *dev) 533 { 534 dev->type = ARPHRD_PIMREG; 535 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; 536 dev->flags = IFF_NOARP; 537 dev->netdev_ops = ®_vif_netdev_ops; 538 dev->needs_free_netdev = true; 539 dev->features |= NETIF_F_NETNS_LOCAL; 540 } 541 542 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 543 { 544 struct net_device *dev; 545 char name[IFNAMSIZ]; 546 547 if (mrt->id == RT_TABLE_DEFAULT) 548 sprintf(name, "pimreg"); 549 else 550 sprintf(name, "pimreg%u", mrt->id); 551 552 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 553 554 if (!dev) 555 return NULL; 556 557 dev_net_set(dev, net); 558 559 if (register_netdevice(dev)) { 560 free_netdev(dev); 561 return NULL; 562 } 563 564 if (!ipmr_init_vif_indev(dev)) 565 goto failure; 566 if (dev_open(dev, NULL)) 567 goto failure; 568 569 dev_hold(dev); 570 571 return dev; 572 573 failure: 574 unregister_netdevice(dev); 575 return NULL; 576 } 577 578 /* called with rcu_read_lock() */ 579 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb, 580 unsigned int pimlen) 581 { 582 struct net_device *reg_dev = NULL; 583 struct iphdr *encap; 584 int vif_num; 585 586 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); 587 /* Check that: 588 * a. packet is really sent to a multicast group 589 * b. packet is not a NULL-REGISTER 590 * c. packet is not truncated 591 */ 592 if (!ipv4_is_multicast(encap->daddr) || 593 encap->tot_len == 0 || 594 ntohs(encap->tot_len) + pimlen > skb->len) 595 return 1; 596 597 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */ 598 vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 599 if (vif_num >= 0) 600 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]); 601 if (!reg_dev) 602 return 1; 603 604 skb->mac_header = skb->network_header; 605 skb_pull(skb, (u8 *)encap - skb->data); 606 skb_reset_network_header(skb); 607 skb->protocol = htons(ETH_P_IP); 608 skb->ip_summed = CHECKSUM_NONE; 609 610 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 611 612 netif_rx(skb); 613 614 return NET_RX_SUCCESS; 615 } 616 #else 617 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 618 { 619 return NULL; 620 } 621 #endif 622 623 static int call_ipmr_vif_entry_notifiers(struct net *net, 624 enum fib_event_type event_type, 625 struct vif_device *vif, 626 struct net_device *vif_dev, 627 vifi_t vif_index, u32 tb_id) 628 { 629 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type, 630 vif, vif_dev, vif_index, tb_id, 631 &net->ipv4.ipmr_seq); 632 } 633 634 static int call_ipmr_mfc_entry_notifiers(struct net *net, 635 enum fib_event_type event_type, 636 struct mfc_cache *mfc, u32 tb_id) 637 { 638 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type, 639 &mfc->_c, tb_id, &net->ipv4.ipmr_seq); 640 } 641 642 /** 643 * vif_delete - Delete a VIF entry 644 * @mrt: Table to delete from 645 * @vifi: VIF identifier to delete 646 * @notify: Set to 1, if the caller is a notifier_call 647 * @head: if unregistering the VIF, place it on this queue 648 */ 649 static int vif_delete(struct mr_table *mrt, int vifi, int notify, 650 struct list_head *head) 651 { 652 struct net *net = read_pnet(&mrt->net); 653 struct vif_device *v; 654 struct net_device *dev; 655 struct in_device *in_dev; 656 657 if (vifi < 0 || vifi >= mrt->maxvif) 658 return -EADDRNOTAVAIL; 659 660 v = &mrt->vif_table[vifi]; 661 662 dev = rtnl_dereference(v->dev); 663 if (!dev) 664 return -EADDRNOTAVAIL; 665 666 spin_lock(&mrt_lock); 667 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev, 668 vifi, mrt->id); 669 RCU_INIT_POINTER(v->dev, NULL); 670 671 if (vifi == mrt->mroute_reg_vif_num) { 672 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */ 673 WRITE_ONCE(mrt->mroute_reg_vif_num, -1); 674 } 675 if (vifi + 1 == mrt->maxvif) { 676 int tmp; 677 678 for (tmp = vifi - 1; tmp >= 0; tmp--) { 679 if (VIF_EXISTS(mrt, tmp)) 680 break; 681 } 682 WRITE_ONCE(mrt->maxvif, tmp + 1); 683 } 684 685 spin_unlock(&mrt_lock); 686 687 dev_set_allmulti(dev, -1); 688 689 in_dev = __in_dev_get_rtnl(dev); 690 if (in_dev) { 691 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; 692 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 693 NETCONFA_MC_FORWARDING, 694 dev->ifindex, &in_dev->cnf); 695 ip_rt_multicast_event(in_dev); 696 } 697 698 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify) 699 unregister_netdevice_queue(dev, head); 700 701 netdev_put(dev, &v->dev_tracker); 702 return 0; 703 } 704 705 static void ipmr_cache_free_rcu(struct rcu_head *head) 706 { 707 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu); 708 709 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c); 710 } 711 712 static void ipmr_cache_free(struct mfc_cache *c) 713 { 714 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu); 715 } 716 717 /* Destroy an unresolved cache entry, killing queued skbs 718 * and reporting error to netlink readers. 719 */ 720 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c) 721 { 722 struct net *net = read_pnet(&mrt->net); 723 struct sk_buff *skb; 724 struct nlmsgerr *e; 725 726 atomic_dec(&mrt->cache_resolve_queue_len); 727 728 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) { 729 if (ip_hdr(skb)->version == 0) { 730 struct nlmsghdr *nlh = skb_pull(skb, 731 sizeof(struct iphdr)); 732 nlh->nlmsg_type = NLMSG_ERROR; 733 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 734 skb_trim(skb, nlh->nlmsg_len); 735 e = nlmsg_data(nlh); 736 e->error = -ETIMEDOUT; 737 memset(&e->msg, 0, sizeof(e->msg)); 738 739 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 740 } else { 741 kfree_skb(skb); 742 } 743 } 744 745 ipmr_cache_free(c); 746 } 747 748 /* Timer process for the unresolved queue. */ 749 static void ipmr_expire_process(struct timer_list *t) 750 { 751 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer); 752 struct mr_mfc *c, *next; 753 unsigned long expires; 754 unsigned long now; 755 756 if (!spin_trylock(&mfc_unres_lock)) { 757 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 758 return; 759 } 760 761 if (list_empty(&mrt->mfc_unres_queue)) 762 goto out; 763 764 now = jiffies; 765 expires = 10*HZ; 766 767 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 768 if (time_after(c->mfc_un.unres.expires, now)) { 769 unsigned long interval = c->mfc_un.unres.expires - now; 770 if (interval < expires) 771 expires = interval; 772 continue; 773 } 774 775 list_del(&c->list); 776 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE); 777 ipmr_destroy_unres(mrt, (struct mfc_cache *)c); 778 } 779 780 if (!list_empty(&mrt->mfc_unres_queue)) 781 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 782 783 out: 784 spin_unlock(&mfc_unres_lock); 785 } 786 787 /* Fill oifs list. It is called under locked mrt_lock. */ 788 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache, 789 unsigned char *ttls) 790 { 791 int vifi; 792 793 cache->mfc_un.res.minvif = MAXVIFS; 794 cache->mfc_un.res.maxvif = 0; 795 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 796 797 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 798 if (VIF_EXISTS(mrt, vifi) && 799 ttls[vifi] && ttls[vifi] < 255) { 800 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 801 if (cache->mfc_un.res.minvif > vifi) 802 cache->mfc_un.res.minvif = vifi; 803 if (cache->mfc_un.res.maxvif <= vifi) 804 cache->mfc_un.res.maxvif = vifi + 1; 805 } 806 } 807 cache->mfc_un.res.lastuse = jiffies; 808 } 809 810 static int vif_add(struct net *net, struct mr_table *mrt, 811 struct vifctl *vifc, int mrtsock) 812 { 813 struct netdev_phys_item_id ppid = { }; 814 int vifi = vifc->vifc_vifi; 815 struct vif_device *v = &mrt->vif_table[vifi]; 816 struct net_device *dev; 817 struct in_device *in_dev; 818 int err; 819 820 /* Is vif busy ? */ 821 if (VIF_EXISTS(mrt, vifi)) 822 return -EADDRINUSE; 823 824 switch (vifc->vifc_flags) { 825 case VIFF_REGISTER: 826 if (!ipmr_pimsm_enabled()) 827 return -EINVAL; 828 /* Special Purpose VIF in PIM 829 * All the packets will be sent to the daemon 830 */ 831 if (mrt->mroute_reg_vif_num >= 0) 832 return -EADDRINUSE; 833 dev = ipmr_reg_vif(net, mrt); 834 if (!dev) 835 return -ENOBUFS; 836 err = dev_set_allmulti(dev, 1); 837 if (err) { 838 unregister_netdevice(dev); 839 dev_put(dev); 840 return err; 841 } 842 break; 843 case VIFF_TUNNEL: 844 dev = ipmr_new_tunnel(net, vifc); 845 if (IS_ERR(dev)) 846 return PTR_ERR(dev); 847 break; 848 case VIFF_USE_IFINDEX: 849 case 0: 850 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 851 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 852 if (dev && !__in_dev_get_rtnl(dev)) { 853 dev_put(dev); 854 return -EADDRNOTAVAIL; 855 } 856 } else { 857 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 858 } 859 if (!dev) 860 return -EADDRNOTAVAIL; 861 err = dev_set_allmulti(dev, 1); 862 if (err) { 863 dev_put(dev); 864 return err; 865 } 866 break; 867 default: 868 return -EINVAL; 869 } 870 871 in_dev = __in_dev_get_rtnl(dev); 872 if (!in_dev) { 873 dev_put(dev); 874 return -EADDRNOTAVAIL; 875 } 876 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 877 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING, 878 dev->ifindex, &in_dev->cnf); 879 ip_rt_multicast_event(in_dev); 880 881 /* Fill in the VIF structures */ 882 vif_device_init(v, dev, vifc->vifc_rate_limit, 883 vifc->vifc_threshold, 884 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0), 885 (VIFF_TUNNEL | VIFF_REGISTER)); 886 887 err = dev_get_port_parent_id(dev, &ppid, true); 888 if (err == 0) { 889 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len); 890 v->dev_parent_id.id_len = ppid.id_len; 891 } else { 892 v->dev_parent_id.id_len = 0; 893 } 894 895 v->local = vifc->vifc_lcl_addr.s_addr; 896 v->remote = vifc->vifc_rmt_addr.s_addr; 897 898 /* And finish update writing critical data */ 899 spin_lock(&mrt_lock); 900 rcu_assign_pointer(v->dev, dev); 901 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC); 902 if (v->flags & VIFF_REGISTER) { 903 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */ 904 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi); 905 } 906 if (vifi+1 > mrt->maxvif) 907 WRITE_ONCE(mrt->maxvif, vifi + 1); 908 spin_unlock(&mrt_lock); 909 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev, 910 vifi, mrt->id); 911 return 0; 912 } 913 914 /* called with rcu_read_lock() */ 915 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 916 __be32 origin, 917 __be32 mcastgrp) 918 { 919 struct mfc_cache_cmp_arg arg = { 920 .mfc_mcastgrp = mcastgrp, 921 .mfc_origin = origin 922 }; 923 924 return mr_mfc_find(mrt, &arg); 925 } 926 927 /* Look for a (*,G) entry */ 928 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 929 __be32 mcastgrp, int vifi) 930 { 931 struct mfc_cache_cmp_arg arg = { 932 .mfc_mcastgrp = mcastgrp, 933 .mfc_origin = htonl(INADDR_ANY) 934 }; 935 936 if (mcastgrp == htonl(INADDR_ANY)) 937 return mr_mfc_find_any_parent(mrt, vifi); 938 return mr_mfc_find_any(mrt, vifi, &arg); 939 } 940 941 /* Look for a (S,G,iif) entry if parent != -1 */ 942 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt, 943 __be32 origin, __be32 mcastgrp, 944 int parent) 945 { 946 struct mfc_cache_cmp_arg arg = { 947 .mfc_mcastgrp = mcastgrp, 948 .mfc_origin = origin, 949 }; 950 951 return mr_mfc_find_parent(mrt, &arg, parent); 952 } 953 954 /* Allocate a multicast cache entry */ 955 static struct mfc_cache *ipmr_cache_alloc(void) 956 { 957 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 958 959 if (c) { 960 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 961 c->_c.mfc_un.res.minvif = MAXVIFS; 962 c->_c.free = ipmr_cache_free_rcu; 963 refcount_set(&c->_c.mfc_un.res.refcount, 1); 964 } 965 return c; 966 } 967 968 static struct mfc_cache *ipmr_cache_alloc_unres(void) 969 { 970 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 971 972 if (c) { 973 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved); 974 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ; 975 } 976 return c; 977 } 978 979 /* A cache entry has gone into a resolved state from queued */ 980 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 981 struct mfc_cache *uc, struct mfc_cache *c) 982 { 983 struct sk_buff *skb; 984 struct nlmsgerr *e; 985 986 /* Play the pending entries through our router */ 987 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) { 988 if (ip_hdr(skb)->version == 0) { 989 struct nlmsghdr *nlh = skb_pull(skb, 990 sizeof(struct iphdr)); 991 992 if (mr_fill_mroute(mrt, skb, &c->_c, 993 nlmsg_data(nlh)) > 0) { 994 nlh->nlmsg_len = skb_tail_pointer(skb) - 995 (u8 *)nlh; 996 } else { 997 nlh->nlmsg_type = NLMSG_ERROR; 998 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 999 skb_trim(skb, nlh->nlmsg_len); 1000 e = nlmsg_data(nlh); 1001 e->error = -EMSGSIZE; 1002 memset(&e->msg, 0, sizeof(e->msg)); 1003 } 1004 1005 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1006 } else { 1007 ip_mr_forward(net, mrt, skb->dev, skb, c, 0); 1008 } 1009 } 1010 } 1011 1012 /* Bounce a cache query up to mrouted and netlink. 1013 * 1014 * Called under rcu_read_lock(). 1015 */ 1016 static int ipmr_cache_report(const struct mr_table *mrt, 1017 struct sk_buff *pkt, vifi_t vifi, int assert) 1018 { 1019 const int ihl = ip_hdrlen(pkt); 1020 struct sock *mroute_sk; 1021 struct igmphdr *igmp; 1022 struct igmpmsg *msg; 1023 struct sk_buff *skb; 1024 int ret; 1025 1026 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) 1027 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 1028 else 1029 skb = alloc_skb(128, GFP_ATOMIC); 1030 1031 if (!skb) 1032 return -ENOBUFS; 1033 1034 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) { 1035 /* Ugly, but we have no choice with this interface. 1036 * Duplicate old header, fix ihl, length etc. 1037 * And all this only to mangle msg->im_msgtype and 1038 * to set msg->im_mbz to "mbz" :-) 1039 */ 1040 skb_push(skb, sizeof(struct iphdr)); 1041 skb_reset_network_header(skb); 1042 skb_reset_transport_header(skb); 1043 msg = (struct igmpmsg *)skb_network_header(skb); 1044 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 1045 msg->im_msgtype = assert; 1046 msg->im_mbz = 0; 1047 if (assert == IGMPMSG_WRVIFWHOLE) { 1048 msg->im_vif = vifi; 1049 msg->im_vif_hi = vifi >> 8; 1050 } else { 1051 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */ 1052 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 1053 1054 msg->im_vif = vif_num; 1055 msg->im_vif_hi = vif_num >> 8; 1056 } 1057 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 1058 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 1059 sizeof(struct iphdr)); 1060 } else { 1061 /* Copy the IP header */ 1062 skb_set_network_header(skb, skb->len); 1063 skb_put(skb, ihl); 1064 skb_copy_to_linear_data(skb, pkt->data, ihl); 1065 /* Flag to the kernel this is a route add */ 1066 ip_hdr(skb)->protocol = 0; 1067 msg = (struct igmpmsg *)skb_network_header(skb); 1068 msg->im_vif = vifi; 1069 msg->im_vif_hi = vifi >> 8; 1070 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1071 /* Add our header */ 1072 igmp = skb_put(skb, sizeof(struct igmphdr)); 1073 igmp->type = assert; 1074 msg->im_msgtype = assert; 1075 igmp->code = 0; 1076 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1077 skb->transport_header = skb->network_header; 1078 } 1079 1080 mroute_sk = rcu_dereference(mrt->mroute_sk); 1081 if (!mroute_sk) { 1082 kfree_skb(skb); 1083 return -EINVAL; 1084 } 1085 1086 igmpmsg_netlink_event(mrt, skb); 1087 1088 /* Deliver to mrouted */ 1089 ret = sock_queue_rcv_skb(mroute_sk, skb); 1090 1091 if (ret < 0) { 1092 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1093 kfree_skb(skb); 1094 } 1095 1096 return ret; 1097 } 1098 1099 /* Queue a packet for resolution. It gets locked cache entry! */ 1100 /* Called under rcu_read_lock() */ 1101 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, 1102 struct sk_buff *skb, struct net_device *dev) 1103 { 1104 const struct iphdr *iph = ip_hdr(skb); 1105 struct mfc_cache *c; 1106 bool found = false; 1107 int err; 1108 1109 spin_lock_bh(&mfc_unres_lock); 1110 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) { 1111 if (c->mfc_mcastgrp == iph->daddr && 1112 c->mfc_origin == iph->saddr) { 1113 found = true; 1114 break; 1115 } 1116 } 1117 1118 if (!found) { 1119 /* Create a new entry if allowable */ 1120 c = ipmr_cache_alloc_unres(); 1121 if (!c) { 1122 spin_unlock_bh(&mfc_unres_lock); 1123 1124 kfree_skb(skb); 1125 return -ENOBUFS; 1126 } 1127 1128 /* Fill in the new cache entry */ 1129 c->_c.mfc_parent = -1; 1130 c->mfc_origin = iph->saddr; 1131 c->mfc_mcastgrp = iph->daddr; 1132 1133 /* Reflect first query at mrouted. */ 1134 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1135 1136 if (err < 0) { 1137 /* If the report failed throw the cache entry 1138 out - Brad Parker 1139 */ 1140 spin_unlock_bh(&mfc_unres_lock); 1141 1142 ipmr_cache_free(c); 1143 kfree_skb(skb); 1144 return err; 1145 } 1146 1147 atomic_inc(&mrt->cache_resolve_queue_len); 1148 list_add(&c->_c.list, &mrt->mfc_unres_queue); 1149 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1150 1151 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1152 mod_timer(&mrt->ipmr_expire_timer, 1153 c->_c.mfc_un.unres.expires); 1154 } 1155 1156 /* See if we can append the packet */ 1157 if (c->_c.mfc_un.unres.unresolved.qlen > 3) { 1158 kfree_skb(skb); 1159 err = -ENOBUFS; 1160 } else { 1161 if (dev) { 1162 skb->dev = dev; 1163 skb->skb_iif = dev->ifindex; 1164 } 1165 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb); 1166 err = 0; 1167 } 1168 1169 spin_unlock_bh(&mfc_unres_lock); 1170 return err; 1171 } 1172 1173 /* MFC cache manipulation by user space mroute daemon */ 1174 1175 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1176 { 1177 struct net *net = read_pnet(&mrt->net); 1178 struct mfc_cache *c; 1179 1180 /* The entries are added/deleted only under RTNL */ 1181 rcu_read_lock(); 1182 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1183 mfc->mfcc_mcastgrp.s_addr, parent); 1184 rcu_read_unlock(); 1185 if (!c) 1186 return -ENOENT; 1187 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params); 1188 list_del_rcu(&c->_c.list); 1189 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id); 1190 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1191 mr_cache_put(&c->_c); 1192 1193 return 0; 1194 } 1195 1196 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1197 struct mfcctl *mfc, int mrtsock, int parent) 1198 { 1199 struct mfc_cache *uc, *c; 1200 struct mr_mfc *_uc; 1201 bool found; 1202 int ret; 1203 1204 if (mfc->mfcc_parent >= MAXVIFS) 1205 return -ENFILE; 1206 1207 /* The entries are added/deleted only under RTNL */ 1208 rcu_read_lock(); 1209 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1210 mfc->mfcc_mcastgrp.s_addr, parent); 1211 rcu_read_unlock(); 1212 if (c) { 1213 spin_lock(&mrt_lock); 1214 c->_c.mfc_parent = mfc->mfcc_parent; 1215 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls); 1216 if (!mrtsock) 1217 c->_c.mfc_flags |= MFC_STATIC; 1218 spin_unlock(&mrt_lock); 1219 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c, 1220 mrt->id); 1221 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1222 return 0; 1223 } 1224 1225 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1226 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1227 return -EINVAL; 1228 1229 c = ipmr_cache_alloc(); 1230 if (!c) 1231 return -ENOMEM; 1232 1233 c->mfc_origin = mfc->mfcc_origin.s_addr; 1234 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1235 c->_c.mfc_parent = mfc->mfcc_parent; 1236 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls); 1237 if (!mrtsock) 1238 c->_c.mfc_flags |= MFC_STATIC; 1239 1240 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode, 1241 ipmr_rht_params); 1242 if (ret) { 1243 pr_err("ipmr: rhtable insert error %d\n", ret); 1244 ipmr_cache_free(c); 1245 return ret; 1246 } 1247 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list); 1248 /* Check to see if we resolved a queued list. If so we 1249 * need to send on the frames and tidy up. 1250 */ 1251 found = false; 1252 spin_lock_bh(&mfc_unres_lock); 1253 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) { 1254 uc = (struct mfc_cache *)_uc; 1255 if (uc->mfc_origin == c->mfc_origin && 1256 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1257 list_del(&_uc->list); 1258 atomic_dec(&mrt->cache_resolve_queue_len); 1259 found = true; 1260 break; 1261 } 1262 } 1263 if (list_empty(&mrt->mfc_unres_queue)) 1264 del_timer(&mrt->ipmr_expire_timer); 1265 spin_unlock_bh(&mfc_unres_lock); 1266 1267 if (found) { 1268 ipmr_cache_resolve(net, mrt, uc, c); 1269 ipmr_cache_free(uc); 1270 } 1271 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id); 1272 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1273 return 0; 1274 } 1275 1276 /* Close the multicast socket, and clear the vif tables etc */ 1277 static void mroute_clean_tables(struct mr_table *mrt, int flags) 1278 { 1279 struct net *net = read_pnet(&mrt->net); 1280 struct mr_mfc *c, *tmp; 1281 struct mfc_cache *cache; 1282 LIST_HEAD(list); 1283 int i; 1284 1285 /* Shut down all active vif entries */ 1286 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) { 1287 for (i = 0; i < mrt->maxvif; i++) { 1288 if (((mrt->vif_table[i].flags & VIFF_STATIC) && 1289 !(flags & MRT_FLUSH_VIFS_STATIC)) || 1290 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS))) 1291 continue; 1292 vif_delete(mrt, i, 0, &list); 1293 } 1294 unregister_netdevice_many(&list); 1295 } 1296 1297 /* Wipe the cache */ 1298 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) { 1299 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1300 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) || 1301 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC))) 1302 continue; 1303 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1304 list_del_rcu(&c->list); 1305 cache = (struct mfc_cache *)c; 1306 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache, 1307 mrt->id); 1308 mroute_netlink_event(mrt, cache, RTM_DELROUTE); 1309 mr_cache_put(c); 1310 } 1311 } 1312 1313 if (flags & MRT_FLUSH_MFC) { 1314 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1315 spin_lock_bh(&mfc_unres_lock); 1316 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1317 list_del(&c->list); 1318 cache = (struct mfc_cache *)c; 1319 mroute_netlink_event(mrt, cache, RTM_DELROUTE); 1320 ipmr_destroy_unres(mrt, cache); 1321 } 1322 spin_unlock_bh(&mfc_unres_lock); 1323 } 1324 } 1325 } 1326 1327 /* called from ip_ra_control(), before an RCU grace period, 1328 * we don't need to call synchronize_rcu() here 1329 */ 1330 static void mrtsock_destruct(struct sock *sk) 1331 { 1332 struct net *net = sock_net(sk); 1333 struct mr_table *mrt; 1334 1335 rtnl_lock(); 1336 ipmr_for_each_table(mrt, net) { 1337 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1338 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1339 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1340 NETCONFA_MC_FORWARDING, 1341 NETCONFA_IFINDEX_ALL, 1342 net->ipv4.devconf_all); 1343 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1344 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC); 1345 } 1346 } 1347 rtnl_unlock(); 1348 } 1349 1350 /* Socket options and virtual interface manipulation. The whole 1351 * virtual interface system is a complete heap, but unfortunately 1352 * that's how BSD mrouted happens to think. Maybe one day with a proper 1353 * MOSPF/PIM router set up we can clean this up. 1354 */ 1355 1356 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval, 1357 unsigned int optlen) 1358 { 1359 struct net *net = sock_net(sk); 1360 int val, ret = 0, parent = 0; 1361 struct mr_table *mrt; 1362 struct vifctl vif; 1363 struct mfcctl mfc; 1364 bool do_wrvifwhole; 1365 u32 uval; 1366 1367 /* There's one exception to the lock - MRT_DONE which needs to unlock */ 1368 rtnl_lock(); 1369 if (sk->sk_type != SOCK_RAW || 1370 inet_sk(sk)->inet_num != IPPROTO_IGMP) { 1371 ret = -EOPNOTSUPP; 1372 goto out_unlock; 1373 } 1374 1375 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1376 if (!mrt) { 1377 ret = -ENOENT; 1378 goto out_unlock; 1379 } 1380 if (optname != MRT_INIT) { 1381 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1382 !ns_capable(net->user_ns, CAP_NET_ADMIN)) { 1383 ret = -EACCES; 1384 goto out_unlock; 1385 } 1386 } 1387 1388 switch (optname) { 1389 case MRT_INIT: 1390 if (optlen != sizeof(int)) { 1391 ret = -EINVAL; 1392 break; 1393 } 1394 if (rtnl_dereference(mrt->mroute_sk)) { 1395 ret = -EADDRINUSE; 1396 break; 1397 } 1398 1399 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1400 if (ret == 0) { 1401 rcu_assign_pointer(mrt->mroute_sk, sk); 1402 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1403 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1404 NETCONFA_MC_FORWARDING, 1405 NETCONFA_IFINDEX_ALL, 1406 net->ipv4.devconf_all); 1407 } 1408 break; 1409 case MRT_DONE: 1410 if (sk != rcu_access_pointer(mrt->mroute_sk)) { 1411 ret = -EACCES; 1412 } else { 1413 /* We need to unlock here because mrtsock_destruct takes 1414 * care of rtnl itself and we can't change that due to 1415 * the IP_ROUTER_ALERT setsockopt which runs without it. 1416 */ 1417 rtnl_unlock(); 1418 ret = ip_ra_control(sk, 0, NULL); 1419 goto out; 1420 } 1421 break; 1422 case MRT_ADD_VIF: 1423 case MRT_DEL_VIF: 1424 if (optlen != sizeof(vif)) { 1425 ret = -EINVAL; 1426 break; 1427 } 1428 if (copy_from_sockptr(&vif, optval, sizeof(vif))) { 1429 ret = -EFAULT; 1430 break; 1431 } 1432 if (vif.vifc_vifi >= MAXVIFS) { 1433 ret = -ENFILE; 1434 break; 1435 } 1436 if (optname == MRT_ADD_VIF) { 1437 ret = vif_add(net, mrt, &vif, 1438 sk == rtnl_dereference(mrt->mroute_sk)); 1439 } else { 1440 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1441 } 1442 break; 1443 /* Manipulate the forwarding caches. These live 1444 * in a sort of kernel/user symbiosis. 1445 */ 1446 case MRT_ADD_MFC: 1447 case MRT_DEL_MFC: 1448 parent = -1; 1449 fallthrough; 1450 case MRT_ADD_MFC_PROXY: 1451 case MRT_DEL_MFC_PROXY: 1452 if (optlen != sizeof(mfc)) { 1453 ret = -EINVAL; 1454 break; 1455 } 1456 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) { 1457 ret = -EFAULT; 1458 break; 1459 } 1460 if (parent == 0) 1461 parent = mfc.mfcc_parent; 1462 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1463 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1464 else 1465 ret = ipmr_mfc_add(net, mrt, &mfc, 1466 sk == rtnl_dereference(mrt->mroute_sk), 1467 parent); 1468 break; 1469 case MRT_FLUSH: 1470 if (optlen != sizeof(val)) { 1471 ret = -EINVAL; 1472 break; 1473 } 1474 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1475 ret = -EFAULT; 1476 break; 1477 } 1478 mroute_clean_tables(mrt, val); 1479 break; 1480 /* Control PIM assert. */ 1481 case MRT_ASSERT: 1482 if (optlen != sizeof(val)) { 1483 ret = -EINVAL; 1484 break; 1485 } 1486 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1487 ret = -EFAULT; 1488 break; 1489 } 1490 mrt->mroute_do_assert = val; 1491 break; 1492 case MRT_PIM: 1493 if (!ipmr_pimsm_enabled()) { 1494 ret = -ENOPROTOOPT; 1495 break; 1496 } 1497 if (optlen != sizeof(val)) { 1498 ret = -EINVAL; 1499 break; 1500 } 1501 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1502 ret = -EFAULT; 1503 break; 1504 } 1505 1506 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE); 1507 val = !!val; 1508 if (val != mrt->mroute_do_pim) { 1509 mrt->mroute_do_pim = val; 1510 mrt->mroute_do_assert = val; 1511 mrt->mroute_do_wrvifwhole = do_wrvifwhole; 1512 } 1513 break; 1514 case MRT_TABLE: 1515 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) { 1516 ret = -ENOPROTOOPT; 1517 break; 1518 } 1519 if (optlen != sizeof(uval)) { 1520 ret = -EINVAL; 1521 break; 1522 } 1523 if (copy_from_sockptr(&uval, optval, sizeof(uval))) { 1524 ret = -EFAULT; 1525 break; 1526 } 1527 1528 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1529 ret = -EBUSY; 1530 } else { 1531 mrt = ipmr_new_table(net, uval); 1532 if (IS_ERR(mrt)) 1533 ret = PTR_ERR(mrt); 1534 else 1535 raw_sk(sk)->ipmr_table = uval; 1536 } 1537 break; 1538 /* Spurious command, or MRT_VERSION which you cannot set. */ 1539 default: 1540 ret = -ENOPROTOOPT; 1541 } 1542 out_unlock: 1543 rtnl_unlock(); 1544 out: 1545 return ret; 1546 } 1547 1548 /* Getsock opt support for the multicast routing system. */ 1549 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval, 1550 sockptr_t optlen) 1551 { 1552 int olr; 1553 int val; 1554 struct net *net = sock_net(sk); 1555 struct mr_table *mrt; 1556 1557 if (sk->sk_type != SOCK_RAW || 1558 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1559 return -EOPNOTSUPP; 1560 1561 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1562 if (!mrt) 1563 return -ENOENT; 1564 1565 switch (optname) { 1566 case MRT_VERSION: 1567 val = 0x0305; 1568 break; 1569 case MRT_PIM: 1570 if (!ipmr_pimsm_enabled()) 1571 return -ENOPROTOOPT; 1572 val = mrt->mroute_do_pim; 1573 break; 1574 case MRT_ASSERT: 1575 val = mrt->mroute_do_assert; 1576 break; 1577 default: 1578 return -ENOPROTOOPT; 1579 } 1580 1581 if (copy_from_sockptr(&olr, optlen, sizeof(int))) 1582 return -EFAULT; 1583 olr = min_t(unsigned int, olr, sizeof(int)); 1584 if (olr < 0) 1585 return -EINVAL; 1586 if (copy_to_sockptr(optlen, &olr, sizeof(int))) 1587 return -EFAULT; 1588 if (copy_to_sockptr(optval, &val, olr)) 1589 return -EFAULT; 1590 return 0; 1591 } 1592 1593 /* The IP multicast ioctl support routines. */ 1594 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) 1595 { 1596 struct sioc_sg_req sr; 1597 struct sioc_vif_req vr; 1598 struct vif_device *vif; 1599 struct mfc_cache *c; 1600 struct net *net = sock_net(sk); 1601 struct mr_table *mrt; 1602 1603 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1604 if (!mrt) 1605 return -ENOENT; 1606 1607 switch (cmd) { 1608 case SIOCGETVIFCNT: 1609 if (copy_from_user(&vr, arg, sizeof(vr))) 1610 return -EFAULT; 1611 if (vr.vifi >= mrt->maxvif) 1612 return -EINVAL; 1613 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif); 1614 rcu_read_lock(); 1615 vif = &mrt->vif_table[vr.vifi]; 1616 if (VIF_EXISTS(mrt, vr.vifi)) { 1617 vr.icount = READ_ONCE(vif->pkt_in); 1618 vr.ocount = READ_ONCE(vif->pkt_out); 1619 vr.ibytes = READ_ONCE(vif->bytes_in); 1620 vr.obytes = READ_ONCE(vif->bytes_out); 1621 rcu_read_unlock(); 1622 1623 if (copy_to_user(arg, &vr, sizeof(vr))) 1624 return -EFAULT; 1625 return 0; 1626 } 1627 rcu_read_unlock(); 1628 return -EADDRNOTAVAIL; 1629 case SIOCGETSGCNT: 1630 if (copy_from_user(&sr, arg, sizeof(sr))) 1631 return -EFAULT; 1632 1633 rcu_read_lock(); 1634 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1635 if (c) { 1636 sr.pktcnt = c->_c.mfc_un.res.pkt; 1637 sr.bytecnt = c->_c.mfc_un.res.bytes; 1638 sr.wrong_if = c->_c.mfc_un.res.wrong_if; 1639 rcu_read_unlock(); 1640 1641 if (copy_to_user(arg, &sr, sizeof(sr))) 1642 return -EFAULT; 1643 return 0; 1644 } 1645 rcu_read_unlock(); 1646 return -EADDRNOTAVAIL; 1647 default: 1648 return -ENOIOCTLCMD; 1649 } 1650 } 1651 1652 #ifdef CONFIG_COMPAT 1653 struct compat_sioc_sg_req { 1654 struct in_addr src; 1655 struct in_addr grp; 1656 compat_ulong_t pktcnt; 1657 compat_ulong_t bytecnt; 1658 compat_ulong_t wrong_if; 1659 }; 1660 1661 struct compat_sioc_vif_req { 1662 vifi_t vifi; /* Which iface */ 1663 compat_ulong_t icount; 1664 compat_ulong_t ocount; 1665 compat_ulong_t ibytes; 1666 compat_ulong_t obytes; 1667 }; 1668 1669 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1670 { 1671 struct compat_sioc_sg_req sr; 1672 struct compat_sioc_vif_req vr; 1673 struct vif_device *vif; 1674 struct mfc_cache *c; 1675 struct net *net = sock_net(sk); 1676 struct mr_table *mrt; 1677 1678 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1679 if (!mrt) 1680 return -ENOENT; 1681 1682 switch (cmd) { 1683 case SIOCGETVIFCNT: 1684 if (copy_from_user(&vr, arg, sizeof(vr))) 1685 return -EFAULT; 1686 if (vr.vifi >= mrt->maxvif) 1687 return -EINVAL; 1688 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif); 1689 rcu_read_lock(); 1690 vif = &mrt->vif_table[vr.vifi]; 1691 if (VIF_EXISTS(mrt, vr.vifi)) { 1692 vr.icount = READ_ONCE(vif->pkt_in); 1693 vr.ocount = READ_ONCE(vif->pkt_out); 1694 vr.ibytes = READ_ONCE(vif->bytes_in); 1695 vr.obytes = READ_ONCE(vif->bytes_out); 1696 rcu_read_unlock(); 1697 1698 if (copy_to_user(arg, &vr, sizeof(vr))) 1699 return -EFAULT; 1700 return 0; 1701 } 1702 rcu_read_unlock(); 1703 return -EADDRNOTAVAIL; 1704 case SIOCGETSGCNT: 1705 if (copy_from_user(&sr, arg, sizeof(sr))) 1706 return -EFAULT; 1707 1708 rcu_read_lock(); 1709 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1710 if (c) { 1711 sr.pktcnt = c->_c.mfc_un.res.pkt; 1712 sr.bytecnt = c->_c.mfc_un.res.bytes; 1713 sr.wrong_if = c->_c.mfc_un.res.wrong_if; 1714 rcu_read_unlock(); 1715 1716 if (copy_to_user(arg, &sr, sizeof(sr))) 1717 return -EFAULT; 1718 return 0; 1719 } 1720 rcu_read_unlock(); 1721 return -EADDRNOTAVAIL; 1722 default: 1723 return -ENOIOCTLCMD; 1724 } 1725 } 1726 #endif 1727 1728 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1729 { 1730 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1731 struct net *net = dev_net(dev); 1732 struct mr_table *mrt; 1733 struct vif_device *v; 1734 int ct; 1735 1736 if (event != NETDEV_UNREGISTER) 1737 return NOTIFY_DONE; 1738 1739 ipmr_for_each_table(mrt, net) { 1740 v = &mrt->vif_table[0]; 1741 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1742 if (rcu_access_pointer(v->dev) == dev) 1743 vif_delete(mrt, ct, 1, NULL); 1744 } 1745 } 1746 return NOTIFY_DONE; 1747 } 1748 1749 static struct notifier_block ip_mr_notifier = { 1750 .notifier_call = ipmr_device_event, 1751 }; 1752 1753 /* Encapsulate a packet by attaching a valid IPIP header to it. 1754 * This avoids tunnel drivers and other mess and gives us the speed so 1755 * important for multicast video. 1756 */ 1757 static void ip_encap(struct net *net, struct sk_buff *skb, 1758 __be32 saddr, __be32 daddr) 1759 { 1760 struct iphdr *iph; 1761 const struct iphdr *old_iph = ip_hdr(skb); 1762 1763 skb_push(skb, sizeof(struct iphdr)); 1764 skb->transport_header = skb->network_header; 1765 skb_reset_network_header(skb); 1766 iph = ip_hdr(skb); 1767 1768 iph->version = 4; 1769 iph->tos = old_iph->tos; 1770 iph->ttl = old_iph->ttl; 1771 iph->frag_off = 0; 1772 iph->daddr = daddr; 1773 iph->saddr = saddr; 1774 iph->protocol = IPPROTO_IPIP; 1775 iph->ihl = 5; 1776 iph->tot_len = htons(skb->len); 1777 ip_select_ident(net, skb, NULL); 1778 ip_send_check(iph); 1779 1780 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1781 nf_reset_ct(skb); 1782 } 1783 1784 static inline int ipmr_forward_finish(struct net *net, struct sock *sk, 1785 struct sk_buff *skb) 1786 { 1787 struct ip_options *opt = &(IPCB(skb)->opt); 1788 1789 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS); 1790 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len); 1791 1792 if (unlikely(opt->optlen)) 1793 ip_forward_options(skb); 1794 1795 return dst_output(net, sk, skb); 1796 } 1797 1798 #ifdef CONFIG_NET_SWITCHDEV 1799 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt, 1800 int in_vifi, int out_vifi) 1801 { 1802 struct vif_device *out_vif = &mrt->vif_table[out_vifi]; 1803 struct vif_device *in_vif = &mrt->vif_table[in_vifi]; 1804 1805 if (!skb->offload_l3_fwd_mark) 1806 return false; 1807 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len) 1808 return false; 1809 return netdev_phys_item_id_same(&out_vif->dev_parent_id, 1810 &in_vif->dev_parent_id); 1811 } 1812 #else 1813 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt, 1814 int in_vifi, int out_vifi) 1815 { 1816 return false; 1817 } 1818 #endif 1819 1820 /* Processing handlers for ipmr_forward, under rcu_read_lock() */ 1821 1822 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt, 1823 int in_vifi, struct sk_buff *skb, int vifi) 1824 { 1825 const struct iphdr *iph = ip_hdr(skb); 1826 struct vif_device *vif = &mrt->vif_table[vifi]; 1827 struct net_device *vif_dev; 1828 struct net_device *dev; 1829 struct rtable *rt; 1830 struct flowi4 fl4; 1831 int encap = 0; 1832 1833 vif_dev = vif_dev_read(vif); 1834 if (!vif_dev) 1835 goto out_free; 1836 1837 if (vif->flags & VIFF_REGISTER) { 1838 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 1839 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 1840 vif_dev->stats.tx_bytes += skb->len; 1841 vif_dev->stats.tx_packets++; 1842 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1843 goto out_free; 1844 } 1845 1846 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi)) 1847 goto out_free; 1848 1849 if (vif->flags & VIFF_TUNNEL) { 1850 rt = ip_route_output_ports(net, &fl4, NULL, 1851 vif->remote, vif->local, 1852 0, 0, 1853 IPPROTO_IPIP, 1854 RT_TOS(iph->tos), vif->link); 1855 if (IS_ERR(rt)) 1856 goto out_free; 1857 encap = sizeof(struct iphdr); 1858 } else { 1859 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1860 0, 0, 1861 IPPROTO_IPIP, 1862 RT_TOS(iph->tos), vif->link); 1863 if (IS_ERR(rt)) 1864 goto out_free; 1865 } 1866 1867 dev = rt->dst.dev; 1868 1869 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1870 /* Do not fragment multicasts. Alas, IPv4 does not 1871 * allow to send ICMP, so that packets will disappear 1872 * to blackhole. 1873 */ 1874 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS); 1875 ip_rt_put(rt); 1876 goto out_free; 1877 } 1878 1879 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len; 1880 1881 if (skb_cow(skb, encap)) { 1882 ip_rt_put(rt); 1883 goto out_free; 1884 } 1885 1886 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 1887 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 1888 1889 skb_dst_drop(skb); 1890 skb_dst_set(skb, &rt->dst); 1891 ip_decrease_ttl(ip_hdr(skb)); 1892 1893 /* FIXME: forward and output firewalls used to be called here. 1894 * What do we do with netfilter? -- RR 1895 */ 1896 if (vif->flags & VIFF_TUNNEL) { 1897 ip_encap(net, skb, vif->local, vif->remote); 1898 /* FIXME: extra output firewall step used to be here. --RR */ 1899 vif_dev->stats.tx_packets++; 1900 vif_dev->stats.tx_bytes += skb->len; 1901 } 1902 1903 IPCB(skb)->flags |= IPSKB_FORWARDED; 1904 1905 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1906 * not only before forwarding, but after forwarding on all output 1907 * interfaces. It is clear, if mrouter runs a multicasting 1908 * program, it should receive packets not depending to what interface 1909 * program is joined. 1910 * If we will not make it, the program will have to join on all 1911 * interfaces. On the other hand, multihoming host (or router, but 1912 * not mrouter) cannot join to more than one interface - it will 1913 * result in receiving multiple packets. 1914 */ 1915 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, 1916 net, NULL, skb, skb->dev, dev, 1917 ipmr_forward_finish); 1918 return; 1919 1920 out_free: 1921 kfree_skb(skb); 1922 } 1923 1924 /* Called with mrt_lock or rcu_read_lock() */ 1925 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev) 1926 { 1927 int ct; 1928 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */ 1929 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) { 1930 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev) 1931 break; 1932 } 1933 return ct; 1934 } 1935 1936 /* "local" means that we should preserve one skb (for local delivery) */ 1937 /* Called uner rcu_read_lock() */ 1938 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 1939 struct net_device *dev, struct sk_buff *skb, 1940 struct mfc_cache *c, int local) 1941 { 1942 int true_vifi = ipmr_find_vif(mrt, dev); 1943 int psend = -1; 1944 int vif, ct; 1945 1946 vif = c->_c.mfc_parent; 1947 c->_c.mfc_un.res.pkt++; 1948 c->_c.mfc_un.res.bytes += skb->len; 1949 c->_c.mfc_un.res.lastuse = jiffies; 1950 1951 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 1952 struct mfc_cache *cache_proxy; 1953 1954 /* For an (*,G) entry, we only check that the incoming 1955 * interface is part of the static tree. 1956 */ 1957 cache_proxy = mr_mfc_find_any_parent(mrt, vif); 1958 if (cache_proxy && 1959 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) 1960 goto forward; 1961 } 1962 1963 /* Wrong interface: drop packet and (maybe) send PIM assert. */ 1964 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) { 1965 if (rt_is_output_route(skb_rtable(skb))) { 1966 /* It is our own packet, looped back. 1967 * Very complicated situation... 1968 * 1969 * The best workaround until routing daemons will be 1970 * fixed is not to redistribute packet, if it was 1971 * send through wrong interface. It means, that 1972 * multicast applications WILL NOT work for 1973 * (S,G), which have default multicast route pointing 1974 * to wrong oif. In any case, it is not a good 1975 * idea to use multicasting applications on router. 1976 */ 1977 goto dont_forward; 1978 } 1979 1980 c->_c.mfc_un.res.wrong_if++; 1981 1982 if (true_vifi >= 0 && mrt->mroute_do_assert && 1983 /* pimsm uses asserts, when switching from RPT to SPT, 1984 * so that we cannot check that packet arrived on an oif. 1985 * It is bad, but otherwise we would need to move pretty 1986 * large chunk of pimd to kernel. Ough... --ANK 1987 */ 1988 (mrt->mroute_do_pim || 1989 c->_c.mfc_un.res.ttls[true_vifi] < 255) && 1990 time_after(jiffies, 1991 c->_c.mfc_un.res.last_assert + 1992 MFC_ASSERT_THRESH)) { 1993 c->_c.mfc_un.res.last_assert = jiffies; 1994 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 1995 if (mrt->mroute_do_wrvifwhole) 1996 ipmr_cache_report(mrt, skb, true_vifi, 1997 IGMPMSG_WRVIFWHOLE); 1998 } 1999 goto dont_forward; 2000 } 2001 2002 forward: 2003 WRITE_ONCE(mrt->vif_table[vif].pkt_in, 2004 mrt->vif_table[vif].pkt_in + 1); 2005 WRITE_ONCE(mrt->vif_table[vif].bytes_in, 2006 mrt->vif_table[vif].bytes_in + skb->len); 2007 2008 /* Forward the frame */ 2009 if (c->mfc_origin == htonl(INADDR_ANY) && 2010 c->mfc_mcastgrp == htonl(INADDR_ANY)) { 2011 if (true_vifi >= 0 && 2012 true_vifi != c->_c.mfc_parent && 2013 ip_hdr(skb)->ttl > 2014 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) { 2015 /* It's an (*,*) entry and the packet is not coming from 2016 * the upstream: forward the packet to the upstream 2017 * only. 2018 */ 2019 psend = c->_c.mfc_parent; 2020 goto last_forward; 2021 } 2022 goto dont_forward; 2023 } 2024 for (ct = c->_c.mfc_un.res.maxvif - 1; 2025 ct >= c->_c.mfc_un.res.minvif; ct--) { 2026 /* For (*,G) entry, don't forward to the incoming interface */ 2027 if ((c->mfc_origin != htonl(INADDR_ANY) || 2028 ct != true_vifi) && 2029 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) { 2030 if (psend != -1) { 2031 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2032 2033 if (skb2) 2034 ipmr_queue_xmit(net, mrt, true_vifi, 2035 skb2, psend); 2036 } 2037 psend = ct; 2038 } 2039 } 2040 last_forward: 2041 if (psend != -1) { 2042 if (local) { 2043 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2044 2045 if (skb2) 2046 ipmr_queue_xmit(net, mrt, true_vifi, skb2, 2047 psend); 2048 } else { 2049 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend); 2050 return; 2051 } 2052 } 2053 2054 dont_forward: 2055 if (!local) 2056 kfree_skb(skb); 2057 } 2058 2059 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 2060 { 2061 struct rtable *rt = skb_rtable(skb); 2062 struct iphdr *iph = ip_hdr(skb); 2063 struct flowi4 fl4 = { 2064 .daddr = iph->daddr, 2065 .saddr = iph->saddr, 2066 .flowi4_tos = RT_TOS(iph->tos), 2067 .flowi4_oif = (rt_is_output_route(rt) ? 2068 skb->dev->ifindex : 0), 2069 .flowi4_iif = (rt_is_output_route(rt) ? 2070 LOOPBACK_IFINDEX : 2071 skb->dev->ifindex), 2072 .flowi4_mark = skb->mark, 2073 }; 2074 struct mr_table *mrt; 2075 int err; 2076 2077 err = ipmr_fib_lookup(net, &fl4, &mrt); 2078 if (err) 2079 return ERR_PTR(err); 2080 return mrt; 2081 } 2082 2083 /* Multicast packets for forwarding arrive here 2084 * Called with rcu_read_lock(); 2085 */ 2086 int ip_mr_input(struct sk_buff *skb) 2087 { 2088 struct mfc_cache *cache; 2089 struct net *net = dev_net(skb->dev); 2090 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 2091 struct mr_table *mrt; 2092 struct net_device *dev; 2093 2094 /* skb->dev passed in is the loX master dev for vrfs. 2095 * As there are no vifs associated with loopback devices, 2096 * get the proper interface that does have a vif associated with it. 2097 */ 2098 dev = skb->dev; 2099 if (netif_is_l3_master(skb->dev)) { 2100 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 2101 if (!dev) { 2102 kfree_skb(skb); 2103 return -ENODEV; 2104 } 2105 } 2106 2107 /* Packet is looped back after forward, it should not be 2108 * forwarded second time, but still can be delivered locally. 2109 */ 2110 if (IPCB(skb)->flags & IPSKB_FORWARDED) 2111 goto dont_forward; 2112 2113 mrt = ipmr_rt_fib_lookup(net, skb); 2114 if (IS_ERR(mrt)) { 2115 kfree_skb(skb); 2116 return PTR_ERR(mrt); 2117 } 2118 if (!local) { 2119 if (IPCB(skb)->opt.router_alert) { 2120 if (ip_call_ra_chain(skb)) 2121 return 0; 2122 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 2123 /* IGMPv1 (and broken IGMPv2 implementations sort of 2124 * Cisco IOS <= 11.2(8)) do not put router alert 2125 * option to IGMP packets destined to routable 2126 * groups. It is very bad, because it means 2127 * that we can forward NO IGMP messages. 2128 */ 2129 struct sock *mroute_sk; 2130 2131 mroute_sk = rcu_dereference(mrt->mroute_sk); 2132 if (mroute_sk) { 2133 nf_reset_ct(skb); 2134 raw_rcv(mroute_sk, skb); 2135 return 0; 2136 } 2137 } 2138 } 2139 2140 /* already under rcu_read_lock() */ 2141 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 2142 if (!cache) { 2143 int vif = ipmr_find_vif(mrt, dev); 2144 2145 if (vif >= 0) 2146 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2147 vif); 2148 } 2149 2150 /* No usable cache entry */ 2151 if (!cache) { 2152 int vif; 2153 2154 if (local) { 2155 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2156 ip_local_deliver(skb); 2157 if (!skb2) 2158 return -ENOBUFS; 2159 skb = skb2; 2160 } 2161 2162 vif = ipmr_find_vif(mrt, dev); 2163 if (vif >= 0) 2164 return ipmr_cache_unresolved(mrt, vif, skb, dev); 2165 kfree_skb(skb); 2166 return -ENODEV; 2167 } 2168 2169 ip_mr_forward(net, mrt, dev, skb, cache, local); 2170 2171 if (local) 2172 return ip_local_deliver(skb); 2173 2174 return 0; 2175 2176 dont_forward: 2177 if (local) 2178 return ip_local_deliver(skb); 2179 kfree_skb(skb); 2180 return 0; 2181 } 2182 2183 #ifdef CONFIG_IP_PIMSM_V1 2184 /* Handle IGMP messages of PIMv1 */ 2185 int pim_rcv_v1(struct sk_buff *skb) 2186 { 2187 struct igmphdr *pim; 2188 struct net *net = dev_net(skb->dev); 2189 struct mr_table *mrt; 2190 2191 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2192 goto drop; 2193 2194 pim = igmp_hdr(skb); 2195 2196 mrt = ipmr_rt_fib_lookup(net, skb); 2197 if (IS_ERR(mrt)) 2198 goto drop; 2199 if (!mrt->mroute_do_pim || 2200 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2201 goto drop; 2202 2203 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2204 drop: 2205 kfree_skb(skb); 2206 } 2207 return 0; 2208 } 2209 #endif 2210 2211 #ifdef CONFIG_IP_PIMSM_V2 2212 static int pim_rcv(struct sk_buff *skb) 2213 { 2214 struct pimreghdr *pim; 2215 struct net *net = dev_net(skb->dev); 2216 struct mr_table *mrt; 2217 2218 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2219 goto drop; 2220 2221 pim = (struct pimreghdr *)skb_transport_header(skb); 2222 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) || 2223 (pim->flags & PIM_NULL_REGISTER) || 2224 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2225 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2226 goto drop; 2227 2228 mrt = ipmr_rt_fib_lookup(net, skb); 2229 if (IS_ERR(mrt)) 2230 goto drop; 2231 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2232 drop: 2233 kfree_skb(skb); 2234 } 2235 return 0; 2236 } 2237 #endif 2238 2239 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2240 __be32 saddr, __be32 daddr, 2241 struct rtmsg *rtm, u32 portid) 2242 { 2243 struct mfc_cache *cache; 2244 struct mr_table *mrt; 2245 int err; 2246 2247 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2248 if (!mrt) 2249 return -ENOENT; 2250 2251 rcu_read_lock(); 2252 cache = ipmr_cache_find(mrt, saddr, daddr); 2253 if (!cache && skb->dev) { 2254 int vif = ipmr_find_vif(mrt, skb->dev); 2255 2256 if (vif >= 0) 2257 cache = ipmr_cache_find_any(mrt, daddr, vif); 2258 } 2259 if (!cache) { 2260 struct sk_buff *skb2; 2261 struct iphdr *iph; 2262 struct net_device *dev; 2263 int vif = -1; 2264 2265 dev = skb->dev; 2266 if (dev) 2267 vif = ipmr_find_vif(mrt, dev); 2268 if (vif < 0) { 2269 rcu_read_unlock(); 2270 return -ENODEV; 2271 } 2272 2273 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr)); 2274 if (!skb2) { 2275 rcu_read_unlock(); 2276 return -ENOMEM; 2277 } 2278 2279 NETLINK_CB(skb2).portid = portid; 2280 skb_push(skb2, sizeof(struct iphdr)); 2281 skb_reset_network_header(skb2); 2282 iph = ip_hdr(skb2); 2283 iph->ihl = sizeof(struct iphdr) >> 2; 2284 iph->saddr = saddr; 2285 iph->daddr = daddr; 2286 iph->version = 0; 2287 err = ipmr_cache_unresolved(mrt, vif, skb2, dev); 2288 rcu_read_unlock(); 2289 return err; 2290 } 2291 2292 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm); 2293 rcu_read_unlock(); 2294 return err; 2295 } 2296 2297 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2298 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2299 int flags) 2300 { 2301 struct nlmsghdr *nlh; 2302 struct rtmsg *rtm; 2303 int err; 2304 2305 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2306 if (!nlh) 2307 return -EMSGSIZE; 2308 2309 rtm = nlmsg_data(nlh); 2310 rtm->rtm_family = RTNL_FAMILY_IPMR; 2311 rtm->rtm_dst_len = 32; 2312 rtm->rtm_src_len = 32; 2313 rtm->rtm_tos = 0; 2314 rtm->rtm_table = mrt->id; 2315 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2316 goto nla_put_failure; 2317 rtm->rtm_type = RTN_MULTICAST; 2318 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2319 if (c->_c.mfc_flags & MFC_STATIC) 2320 rtm->rtm_protocol = RTPROT_STATIC; 2321 else 2322 rtm->rtm_protocol = RTPROT_MROUTED; 2323 rtm->rtm_flags = 0; 2324 2325 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) || 2326 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp)) 2327 goto nla_put_failure; 2328 err = mr_fill_mroute(mrt, skb, &c->_c, rtm); 2329 /* do not break the dump if cache is unresolved */ 2330 if (err < 0 && err != -ENOENT) 2331 goto nla_put_failure; 2332 2333 nlmsg_end(skb, nlh); 2334 return 0; 2335 2336 nla_put_failure: 2337 nlmsg_cancel(skb, nlh); 2338 return -EMSGSIZE; 2339 } 2340 2341 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2342 u32 portid, u32 seq, struct mr_mfc *c, int cmd, 2343 int flags) 2344 { 2345 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c, 2346 cmd, flags); 2347 } 2348 2349 static size_t mroute_msgsize(bool unresolved, int maxvif) 2350 { 2351 size_t len = 2352 NLMSG_ALIGN(sizeof(struct rtmsg)) 2353 + nla_total_size(4) /* RTA_TABLE */ 2354 + nla_total_size(4) /* RTA_SRC */ 2355 + nla_total_size(4) /* RTA_DST */ 2356 ; 2357 2358 if (!unresolved) 2359 len = len 2360 + nla_total_size(4) /* RTA_IIF */ 2361 + nla_total_size(0) /* RTA_MULTIPATH */ 2362 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2363 /* RTA_MFC_STATS */ 2364 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2365 ; 2366 2367 return len; 2368 } 2369 2370 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2371 int cmd) 2372 { 2373 struct net *net = read_pnet(&mrt->net); 2374 struct sk_buff *skb; 2375 int err = -ENOBUFS; 2376 2377 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS, 2378 mrt->maxvif), 2379 GFP_ATOMIC); 2380 if (!skb) 2381 goto errout; 2382 2383 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2384 if (err < 0) 2385 goto errout; 2386 2387 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2388 return; 2389 2390 errout: 2391 kfree_skb(skb); 2392 if (err < 0) 2393 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2394 } 2395 2396 static size_t igmpmsg_netlink_msgsize(size_t payloadlen) 2397 { 2398 size_t len = 2399 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2400 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */ 2401 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */ 2402 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */ 2403 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */ 2404 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */ 2405 /* IPMRA_CREPORT_PKT */ 2406 + nla_total_size(payloadlen) 2407 ; 2408 2409 return len; 2410 } 2411 2412 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt) 2413 { 2414 struct net *net = read_pnet(&mrt->net); 2415 struct nlmsghdr *nlh; 2416 struct rtgenmsg *rtgenm; 2417 struct igmpmsg *msg; 2418 struct sk_buff *skb; 2419 struct nlattr *nla; 2420 int payloadlen; 2421 2422 payloadlen = pkt->len - sizeof(struct igmpmsg); 2423 msg = (struct igmpmsg *)skb_network_header(pkt); 2424 2425 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2426 if (!skb) 2427 goto errout; 2428 2429 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2430 sizeof(struct rtgenmsg), 0); 2431 if (!nlh) 2432 goto errout; 2433 rtgenm = nlmsg_data(nlh); 2434 rtgenm->rtgen_family = RTNL_FAMILY_IPMR; 2435 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) || 2436 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) || 2437 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR, 2438 msg->im_src.s_addr) || 2439 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR, 2440 msg->im_dst.s_addr) || 2441 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id)) 2442 goto nla_put_failure; 2443 2444 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen); 2445 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg), 2446 nla_data(nla), payloadlen)) 2447 goto nla_put_failure; 2448 2449 nlmsg_end(skb, nlh); 2450 2451 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC); 2452 return; 2453 2454 nla_put_failure: 2455 nlmsg_cancel(skb, nlh); 2456 errout: 2457 kfree_skb(skb); 2458 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS); 2459 } 2460 2461 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb, 2462 const struct nlmsghdr *nlh, 2463 struct nlattr **tb, 2464 struct netlink_ext_ack *extack) 2465 { 2466 struct rtmsg *rtm; 2467 int i, err; 2468 2469 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) { 2470 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request"); 2471 return -EINVAL; 2472 } 2473 2474 if (!netlink_strict_get_check(skb)) 2475 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, 2476 rtm_ipv4_policy, extack); 2477 2478 rtm = nlmsg_data(nlh); 2479 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) || 2480 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) || 2481 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol || 2482 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) { 2483 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request"); 2484 return -EINVAL; 2485 } 2486 2487 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, 2488 rtm_ipv4_policy, extack); 2489 if (err) 2490 return err; 2491 2492 if ((tb[RTA_SRC] && !rtm->rtm_src_len) || 2493 (tb[RTA_DST] && !rtm->rtm_dst_len)) { 2494 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4"); 2495 return -EINVAL; 2496 } 2497 2498 for (i = 0; i <= RTA_MAX; i++) { 2499 if (!tb[i]) 2500 continue; 2501 2502 switch (i) { 2503 case RTA_SRC: 2504 case RTA_DST: 2505 case RTA_TABLE: 2506 break; 2507 default: 2508 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request"); 2509 return -EINVAL; 2510 } 2511 } 2512 2513 return 0; 2514 } 2515 2516 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, 2517 struct netlink_ext_ack *extack) 2518 { 2519 struct net *net = sock_net(in_skb->sk); 2520 struct nlattr *tb[RTA_MAX + 1]; 2521 struct sk_buff *skb = NULL; 2522 struct mfc_cache *cache; 2523 struct mr_table *mrt; 2524 __be32 src, grp; 2525 u32 tableid; 2526 int err; 2527 2528 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack); 2529 if (err < 0) 2530 goto errout; 2531 2532 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0; 2533 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0; 2534 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0; 2535 2536 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT); 2537 if (!mrt) { 2538 err = -ENOENT; 2539 goto errout_free; 2540 } 2541 2542 /* entries are added/deleted only under RTNL */ 2543 rcu_read_lock(); 2544 cache = ipmr_cache_find(mrt, src, grp); 2545 rcu_read_unlock(); 2546 if (!cache) { 2547 err = -ENOENT; 2548 goto errout_free; 2549 } 2550 2551 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL); 2552 if (!skb) { 2553 err = -ENOBUFS; 2554 goto errout_free; 2555 } 2556 2557 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid, 2558 nlh->nlmsg_seq, cache, 2559 RTM_NEWROUTE, 0); 2560 if (err < 0) 2561 goto errout_free; 2562 2563 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 2564 2565 errout: 2566 return err; 2567 2568 errout_free: 2569 kfree_skb(skb); 2570 goto errout; 2571 } 2572 2573 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2574 { 2575 struct fib_dump_filter filter = {}; 2576 int err; 2577 2578 if (cb->strict_check) { 2579 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh, 2580 &filter, cb); 2581 if (err < 0) 2582 return err; 2583 } 2584 2585 if (filter.table_id) { 2586 struct mr_table *mrt; 2587 2588 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id); 2589 if (!mrt) { 2590 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR) 2591 return skb->len; 2592 2593 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist"); 2594 return -ENOENT; 2595 } 2596 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute, 2597 &mfc_unres_lock, &filter); 2598 return skb->len ? : err; 2599 } 2600 2601 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter, 2602 _ipmr_fill_mroute, &mfc_unres_lock, &filter); 2603 } 2604 2605 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = { 2606 [RTA_SRC] = { .type = NLA_U32 }, 2607 [RTA_DST] = { .type = NLA_U32 }, 2608 [RTA_IIF] = { .type = NLA_U32 }, 2609 [RTA_TABLE] = { .type = NLA_U32 }, 2610 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 2611 }; 2612 2613 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol) 2614 { 2615 switch (rtm_protocol) { 2616 case RTPROT_STATIC: 2617 case RTPROT_MROUTED: 2618 return true; 2619 } 2620 return false; 2621 } 2622 2623 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc) 2624 { 2625 struct rtnexthop *rtnh = nla_data(nla); 2626 int remaining = nla_len(nla), vifi = 0; 2627 2628 while (rtnh_ok(rtnh, remaining)) { 2629 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops; 2630 if (++vifi == MAXVIFS) 2631 break; 2632 rtnh = rtnh_next(rtnh, &remaining); 2633 } 2634 2635 return remaining > 0 ? -EINVAL : vifi; 2636 } 2637 2638 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */ 2639 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh, 2640 struct mfcctl *mfcc, int *mrtsock, 2641 struct mr_table **mrtret, 2642 struct netlink_ext_ack *extack) 2643 { 2644 struct net_device *dev = NULL; 2645 u32 tblid = RT_TABLE_DEFAULT; 2646 struct mr_table *mrt; 2647 struct nlattr *attr; 2648 struct rtmsg *rtm; 2649 int ret, rem; 2650 2651 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX, 2652 rtm_ipmr_policy, extack); 2653 if (ret < 0) 2654 goto out; 2655 rtm = nlmsg_data(nlh); 2656 2657 ret = -EINVAL; 2658 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 || 2659 rtm->rtm_type != RTN_MULTICAST || 2660 rtm->rtm_scope != RT_SCOPE_UNIVERSE || 2661 !ipmr_rtm_validate_proto(rtm->rtm_protocol)) 2662 goto out; 2663 2664 memset(mfcc, 0, sizeof(*mfcc)); 2665 mfcc->mfcc_parent = -1; 2666 ret = 0; 2667 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) { 2668 switch (nla_type(attr)) { 2669 case RTA_SRC: 2670 mfcc->mfcc_origin.s_addr = nla_get_be32(attr); 2671 break; 2672 case RTA_DST: 2673 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr); 2674 break; 2675 case RTA_IIF: 2676 dev = __dev_get_by_index(net, nla_get_u32(attr)); 2677 if (!dev) { 2678 ret = -ENODEV; 2679 goto out; 2680 } 2681 break; 2682 case RTA_MULTIPATH: 2683 if (ipmr_nla_get_ttls(attr, mfcc) < 0) { 2684 ret = -EINVAL; 2685 goto out; 2686 } 2687 break; 2688 case RTA_PREFSRC: 2689 ret = 1; 2690 break; 2691 case RTA_TABLE: 2692 tblid = nla_get_u32(attr); 2693 break; 2694 } 2695 } 2696 mrt = ipmr_get_table(net, tblid); 2697 if (!mrt) { 2698 ret = -ENOENT; 2699 goto out; 2700 } 2701 *mrtret = mrt; 2702 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0; 2703 if (dev) 2704 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev); 2705 2706 out: 2707 return ret; 2708 } 2709 2710 /* takes care of both newroute and delroute */ 2711 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh, 2712 struct netlink_ext_ack *extack) 2713 { 2714 struct net *net = sock_net(skb->sk); 2715 int ret, mrtsock, parent; 2716 struct mr_table *tbl; 2717 struct mfcctl mfcc; 2718 2719 mrtsock = 0; 2720 tbl = NULL; 2721 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack); 2722 if (ret < 0) 2723 return ret; 2724 2725 parent = ret ? mfcc.mfcc_parent : -1; 2726 if (nlh->nlmsg_type == RTM_NEWROUTE) 2727 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent); 2728 else 2729 return ipmr_mfc_delete(tbl, &mfcc, parent); 2730 } 2731 2732 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb) 2733 { 2734 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len); 2735 2736 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) || 2737 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) || 2738 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM, 2739 mrt->mroute_reg_vif_num) || 2740 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT, 2741 mrt->mroute_do_assert) || 2742 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) || 2743 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE, 2744 mrt->mroute_do_wrvifwhole)) 2745 return false; 2746 2747 return true; 2748 } 2749 2750 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb) 2751 { 2752 struct net_device *vif_dev; 2753 struct nlattr *vif_nest; 2754 struct vif_device *vif; 2755 2756 vif = &mrt->vif_table[vifid]; 2757 vif_dev = rtnl_dereference(vif->dev); 2758 /* if the VIF doesn't exist just continue */ 2759 if (!vif_dev) 2760 return true; 2761 2762 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF); 2763 if (!vif_nest) 2764 return false; 2765 2766 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) || 2767 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) || 2768 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) || 2769 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in, 2770 IPMRA_VIFA_PAD) || 2771 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out, 2772 IPMRA_VIFA_PAD) || 2773 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in, 2774 IPMRA_VIFA_PAD) || 2775 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out, 2776 IPMRA_VIFA_PAD) || 2777 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) || 2778 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) { 2779 nla_nest_cancel(skb, vif_nest); 2780 return false; 2781 } 2782 nla_nest_end(skb, vif_nest); 2783 2784 return true; 2785 } 2786 2787 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh, 2788 struct netlink_ext_ack *extack) 2789 { 2790 struct ifinfomsg *ifm; 2791 2792 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 2793 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump"); 2794 return -EINVAL; 2795 } 2796 2797 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 2798 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump"); 2799 return -EINVAL; 2800 } 2801 2802 ifm = nlmsg_data(nlh); 2803 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 2804 ifm->ifi_change || ifm->ifi_index) { 2805 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request"); 2806 return -EINVAL; 2807 } 2808 2809 return 0; 2810 } 2811 2812 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb) 2813 { 2814 struct net *net = sock_net(skb->sk); 2815 struct nlmsghdr *nlh = NULL; 2816 unsigned int t = 0, s_t; 2817 unsigned int e = 0, s_e; 2818 struct mr_table *mrt; 2819 2820 if (cb->strict_check) { 2821 int err = ipmr_valid_dumplink(cb->nlh, cb->extack); 2822 2823 if (err < 0) 2824 return err; 2825 } 2826 2827 s_t = cb->args[0]; 2828 s_e = cb->args[1]; 2829 2830 ipmr_for_each_table(mrt, net) { 2831 struct nlattr *vifs, *af; 2832 struct ifinfomsg *hdr; 2833 u32 i; 2834 2835 if (t < s_t) 2836 goto skip_table; 2837 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, 2838 cb->nlh->nlmsg_seq, RTM_NEWLINK, 2839 sizeof(*hdr), NLM_F_MULTI); 2840 if (!nlh) 2841 break; 2842 2843 hdr = nlmsg_data(nlh); 2844 memset(hdr, 0, sizeof(*hdr)); 2845 hdr->ifi_family = RTNL_FAMILY_IPMR; 2846 2847 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC); 2848 if (!af) { 2849 nlmsg_cancel(skb, nlh); 2850 goto out; 2851 } 2852 2853 if (!ipmr_fill_table(mrt, skb)) { 2854 nlmsg_cancel(skb, nlh); 2855 goto out; 2856 } 2857 2858 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS); 2859 if (!vifs) { 2860 nla_nest_end(skb, af); 2861 nlmsg_end(skb, nlh); 2862 goto out; 2863 } 2864 for (i = 0; i < mrt->maxvif; i++) { 2865 if (e < s_e) 2866 goto skip_entry; 2867 if (!ipmr_fill_vif(mrt, i, skb)) { 2868 nla_nest_end(skb, vifs); 2869 nla_nest_end(skb, af); 2870 nlmsg_end(skb, nlh); 2871 goto out; 2872 } 2873 skip_entry: 2874 e++; 2875 } 2876 s_e = 0; 2877 e = 0; 2878 nla_nest_end(skb, vifs); 2879 nla_nest_end(skb, af); 2880 nlmsg_end(skb, nlh); 2881 skip_table: 2882 t++; 2883 } 2884 2885 out: 2886 cb->args[1] = e; 2887 cb->args[0] = t; 2888 2889 return skb->len; 2890 } 2891 2892 #ifdef CONFIG_PROC_FS 2893 /* The /proc interfaces to multicast routing : 2894 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 2895 */ 2896 2897 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 2898 __acquires(RCU) 2899 { 2900 struct mr_vif_iter *iter = seq->private; 2901 struct net *net = seq_file_net(seq); 2902 struct mr_table *mrt; 2903 2904 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2905 if (!mrt) 2906 return ERR_PTR(-ENOENT); 2907 2908 iter->mrt = mrt; 2909 2910 rcu_read_lock(); 2911 return mr_vif_seq_start(seq, pos); 2912 } 2913 2914 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 2915 __releases(RCU) 2916 { 2917 rcu_read_unlock(); 2918 } 2919 2920 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 2921 { 2922 struct mr_vif_iter *iter = seq->private; 2923 struct mr_table *mrt = iter->mrt; 2924 2925 if (v == SEQ_START_TOKEN) { 2926 seq_puts(seq, 2927 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 2928 } else { 2929 const struct vif_device *vif = v; 2930 const struct net_device *vif_dev; 2931 const char *name; 2932 2933 vif_dev = vif_dev_read(vif); 2934 name = vif_dev ? vif_dev->name : "none"; 2935 seq_printf(seq, 2936 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 2937 vif - mrt->vif_table, 2938 name, vif->bytes_in, vif->pkt_in, 2939 vif->bytes_out, vif->pkt_out, 2940 vif->flags, vif->local, vif->remote); 2941 } 2942 return 0; 2943 } 2944 2945 static const struct seq_operations ipmr_vif_seq_ops = { 2946 .start = ipmr_vif_seq_start, 2947 .next = mr_vif_seq_next, 2948 .stop = ipmr_vif_seq_stop, 2949 .show = ipmr_vif_seq_show, 2950 }; 2951 2952 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 2953 { 2954 struct net *net = seq_file_net(seq); 2955 struct mr_table *mrt; 2956 2957 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2958 if (!mrt) 2959 return ERR_PTR(-ENOENT); 2960 2961 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock); 2962 } 2963 2964 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 2965 { 2966 int n; 2967 2968 if (v == SEQ_START_TOKEN) { 2969 seq_puts(seq, 2970 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 2971 } else { 2972 const struct mfc_cache *mfc = v; 2973 const struct mr_mfc_iter *it = seq->private; 2974 const struct mr_table *mrt = it->mrt; 2975 2976 seq_printf(seq, "%08X %08X %-3hd", 2977 (__force u32) mfc->mfc_mcastgrp, 2978 (__force u32) mfc->mfc_origin, 2979 mfc->_c.mfc_parent); 2980 2981 if (it->cache != &mrt->mfc_unres_queue) { 2982 seq_printf(seq, " %8lu %8lu %8lu", 2983 mfc->_c.mfc_un.res.pkt, 2984 mfc->_c.mfc_un.res.bytes, 2985 mfc->_c.mfc_un.res.wrong_if); 2986 for (n = mfc->_c.mfc_un.res.minvif; 2987 n < mfc->_c.mfc_un.res.maxvif; n++) { 2988 if (VIF_EXISTS(mrt, n) && 2989 mfc->_c.mfc_un.res.ttls[n] < 255) 2990 seq_printf(seq, 2991 " %2d:%-3d", 2992 n, mfc->_c.mfc_un.res.ttls[n]); 2993 } 2994 } else { 2995 /* unresolved mfc_caches don't contain 2996 * pkt, bytes and wrong_if values 2997 */ 2998 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 2999 } 3000 seq_putc(seq, '\n'); 3001 } 3002 return 0; 3003 } 3004 3005 static const struct seq_operations ipmr_mfc_seq_ops = { 3006 .start = ipmr_mfc_seq_start, 3007 .next = mr_mfc_seq_next, 3008 .stop = mr_mfc_seq_stop, 3009 .show = ipmr_mfc_seq_show, 3010 }; 3011 #endif 3012 3013 #ifdef CONFIG_IP_PIMSM_V2 3014 static const struct net_protocol pim_protocol = { 3015 .handler = pim_rcv, 3016 }; 3017 #endif 3018 3019 static unsigned int ipmr_seq_read(struct net *net) 3020 { 3021 ASSERT_RTNL(); 3022 3023 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net); 3024 } 3025 3026 static int ipmr_dump(struct net *net, struct notifier_block *nb, 3027 struct netlink_ext_ack *extack) 3028 { 3029 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump, 3030 ipmr_mr_table_iter, extack); 3031 } 3032 3033 static const struct fib_notifier_ops ipmr_notifier_ops_template = { 3034 .family = RTNL_FAMILY_IPMR, 3035 .fib_seq_read = ipmr_seq_read, 3036 .fib_dump = ipmr_dump, 3037 .owner = THIS_MODULE, 3038 }; 3039 3040 static int __net_init ipmr_notifier_init(struct net *net) 3041 { 3042 struct fib_notifier_ops *ops; 3043 3044 net->ipv4.ipmr_seq = 0; 3045 3046 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net); 3047 if (IS_ERR(ops)) 3048 return PTR_ERR(ops); 3049 net->ipv4.ipmr_notifier_ops = ops; 3050 3051 return 0; 3052 } 3053 3054 static void __net_exit ipmr_notifier_exit(struct net *net) 3055 { 3056 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops); 3057 net->ipv4.ipmr_notifier_ops = NULL; 3058 } 3059 3060 /* Setup for IP multicast routing */ 3061 static int __net_init ipmr_net_init(struct net *net) 3062 { 3063 int err; 3064 3065 err = ipmr_notifier_init(net); 3066 if (err) 3067 goto ipmr_notifier_fail; 3068 3069 err = ipmr_rules_init(net); 3070 if (err < 0) 3071 goto ipmr_rules_fail; 3072 3073 #ifdef CONFIG_PROC_FS 3074 err = -ENOMEM; 3075 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops, 3076 sizeof(struct mr_vif_iter))) 3077 goto proc_vif_fail; 3078 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops, 3079 sizeof(struct mr_mfc_iter))) 3080 goto proc_cache_fail; 3081 #endif 3082 return 0; 3083 3084 #ifdef CONFIG_PROC_FS 3085 proc_cache_fail: 3086 remove_proc_entry("ip_mr_vif", net->proc_net); 3087 proc_vif_fail: 3088 rtnl_lock(); 3089 ipmr_rules_exit(net); 3090 rtnl_unlock(); 3091 #endif 3092 ipmr_rules_fail: 3093 ipmr_notifier_exit(net); 3094 ipmr_notifier_fail: 3095 return err; 3096 } 3097 3098 static void __net_exit ipmr_net_exit(struct net *net) 3099 { 3100 #ifdef CONFIG_PROC_FS 3101 remove_proc_entry("ip_mr_cache", net->proc_net); 3102 remove_proc_entry("ip_mr_vif", net->proc_net); 3103 #endif 3104 ipmr_notifier_exit(net); 3105 } 3106 3107 static void __net_exit ipmr_net_exit_batch(struct list_head *net_list) 3108 { 3109 struct net *net; 3110 3111 rtnl_lock(); 3112 list_for_each_entry(net, net_list, exit_list) 3113 ipmr_rules_exit(net); 3114 rtnl_unlock(); 3115 } 3116 3117 static struct pernet_operations ipmr_net_ops = { 3118 .init = ipmr_net_init, 3119 .exit = ipmr_net_exit, 3120 .exit_batch = ipmr_net_exit_batch, 3121 }; 3122 3123 int __init ip_mr_init(void) 3124 { 3125 int err; 3126 3127 mrt_cachep = kmem_cache_create("ip_mrt_cache", 3128 sizeof(struct mfc_cache), 3129 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, 3130 NULL); 3131 3132 err = register_pernet_subsys(&ipmr_net_ops); 3133 if (err) 3134 goto reg_pernet_fail; 3135 3136 err = register_netdevice_notifier(&ip_mr_notifier); 3137 if (err) 3138 goto reg_notif_fail; 3139 #ifdef CONFIG_IP_PIMSM_V2 3140 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 3141 pr_err("%s: can't add PIM protocol\n", __func__); 3142 err = -EAGAIN; 3143 goto add_proto_fail; 3144 } 3145 #endif 3146 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE, 3147 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0); 3148 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE, 3149 ipmr_rtm_route, NULL, 0); 3150 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE, 3151 ipmr_rtm_route, NULL, 0); 3152 3153 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK, 3154 NULL, ipmr_rtm_dumplink, 0); 3155 return 0; 3156 3157 #ifdef CONFIG_IP_PIMSM_V2 3158 add_proto_fail: 3159 unregister_netdevice_notifier(&ip_mr_notifier); 3160 #endif 3161 reg_notif_fail: 3162 unregister_pernet_subsys(&ipmr_net_ops); 3163 reg_pernet_fail: 3164 kmem_cache_destroy(mrt_cachep); 3165 return err; 3166 } 3167