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 #include <net/inet_dscp.h> 66 67 #include <linux/nospec.h> 68 69 struct ipmr_rule { 70 struct fib_rule common; 71 }; 72 73 struct ipmr_result { 74 struct mr_table *mrt; 75 }; 76 77 /* Big lock, protecting vif table, mrt cache and mroute socket state. 78 * Note that the changes are semaphored via rtnl_lock. 79 */ 80 81 static DEFINE_SPINLOCK(mrt_lock); 82 83 static struct net_device *vif_dev_read(const struct vif_device *vif) 84 { 85 return rcu_dereference(vif->dev); 86 } 87 88 /* Multicast router control variables */ 89 90 /* Special spinlock for queue of unresolved entries */ 91 static DEFINE_SPINLOCK(mfc_unres_lock); 92 93 /* We return to original Alan's scheme. Hash table of resolved 94 * entries is changed only in process context and protected 95 * with weak lock mrt_lock. Queue of unresolved entries is protected 96 * with strong spinlock mfc_unres_lock. 97 * 98 * In this case data path is free of exclusive locks at all. 99 */ 100 101 static struct kmem_cache *mrt_cachep __ro_after_init; 102 103 static struct mr_table *ipmr_new_table(struct net *net, u32 id); 104 static void ipmr_free_table(struct mr_table *mrt); 105 106 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 107 struct net_device *dev, struct sk_buff *skb, 108 struct mfc_cache *cache, int local); 109 static int ipmr_cache_report(const struct mr_table *mrt, 110 struct sk_buff *pkt, vifi_t vifi, int assert); 111 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 112 int cmd); 113 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt); 114 static void mroute_clean_tables(struct mr_table *mrt, int flags); 115 static void ipmr_expire_process(struct timer_list *t); 116 117 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 118 #define ipmr_for_each_table(mrt, net) \ 119 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \ 120 lockdep_rtnl_is_held() || \ 121 list_empty(&net->ipv4.mr_tables)) 122 123 static struct mr_table *ipmr_mr_table_iter(struct net *net, 124 struct mr_table *mrt) 125 { 126 struct mr_table *ret; 127 128 if (!mrt) 129 ret = list_entry_rcu(net->ipv4.mr_tables.next, 130 struct mr_table, list); 131 else 132 ret = list_entry_rcu(mrt->list.next, 133 struct mr_table, list); 134 135 if (&ret->list == &net->ipv4.mr_tables) 136 return NULL; 137 return ret; 138 } 139 140 static struct mr_table *__ipmr_get_table(struct net *net, u32 id) 141 { 142 struct mr_table *mrt; 143 144 ipmr_for_each_table(mrt, net) { 145 if (mrt->id == id) 146 return mrt; 147 } 148 return NULL; 149 } 150 151 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 152 { 153 struct mr_table *mrt; 154 155 rcu_read_lock(); 156 mrt = __ipmr_get_table(net, id); 157 rcu_read_unlock(); 158 return mrt; 159 } 160 161 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 162 struct mr_table **mrt) 163 { 164 int err; 165 struct ipmr_result res; 166 struct fib_lookup_arg arg = { 167 .result = &res, 168 .flags = FIB_LOOKUP_NOREF, 169 }; 170 171 /* update flow if oif or iif point to device enslaved to l3mdev */ 172 l3mdev_update_flow(net, flowi4_to_flowi(flp4)); 173 174 err = fib_rules_lookup(net->ipv4.mr_rules_ops, 175 flowi4_to_flowi(flp4), 0, &arg); 176 if (err < 0) 177 return err; 178 *mrt = res.mrt; 179 return 0; 180 } 181 182 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp, 183 int flags, struct fib_lookup_arg *arg) 184 { 185 struct ipmr_result *res = arg->result; 186 struct mr_table *mrt; 187 188 switch (rule->action) { 189 case FR_ACT_TO_TBL: 190 break; 191 case FR_ACT_UNREACHABLE: 192 return -ENETUNREACH; 193 case FR_ACT_PROHIBIT: 194 return -EACCES; 195 case FR_ACT_BLACKHOLE: 196 default: 197 return -EINVAL; 198 } 199 200 arg->table = fib_rule_get_table(rule, arg); 201 202 mrt = __ipmr_get_table(rule->fr_net, arg->table); 203 if (!mrt) 204 return -EAGAIN; 205 res->mrt = mrt; 206 return 0; 207 } 208 209 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 210 { 211 return 1; 212 } 213 214 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 215 struct fib_rule_hdr *frh, struct nlattr **tb, 216 struct netlink_ext_ack *extack) 217 { 218 return 0; 219 } 220 221 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 222 struct nlattr **tb) 223 { 224 return 1; 225 } 226 227 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 228 struct fib_rule_hdr *frh) 229 { 230 frh->dst_len = 0; 231 frh->src_len = 0; 232 frh->tos = 0; 233 return 0; 234 } 235 236 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = { 237 .family = RTNL_FAMILY_IPMR, 238 .rule_size = sizeof(struct ipmr_rule), 239 .addr_size = sizeof(u32), 240 .action = ipmr_rule_action, 241 .match = ipmr_rule_match, 242 .configure = ipmr_rule_configure, 243 .compare = ipmr_rule_compare, 244 .fill = ipmr_rule_fill, 245 .nlgroup = RTNLGRP_IPV4_RULE, 246 .owner = THIS_MODULE, 247 }; 248 249 static int __net_init ipmr_rules_init(struct net *net) 250 { 251 struct fib_rules_ops *ops; 252 struct mr_table *mrt; 253 int err; 254 255 ops = fib_rules_register(&ipmr_rules_ops_template, net); 256 if (IS_ERR(ops)) 257 return PTR_ERR(ops); 258 259 INIT_LIST_HEAD(&net->ipv4.mr_tables); 260 261 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 262 if (IS_ERR(mrt)) { 263 err = PTR_ERR(mrt); 264 goto err1; 265 } 266 267 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT); 268 if (err < 0) 269 goto err2; 270 271 net->ipv4.mr_rules_ops = ops; 272 return 0; 273 274 err2: 275 rtnl_lock(); 276 ipmr_free_table(mrt); 277 rtnl_unlock(); 278 err1: 279 fib_rules_unregister(ops); 280 return err; 281 } 282 283 static void __net_exit ipmr_rules_exit(struct net *net) 284 { 285 struct mr_table *mrt, *next; 286 287 ASSERT_RTNL(); 288 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) { 289 list_del(&mrt->list); 290 ipmr_free_table(mrt); 291 } 292 fib_rules_unregister(net->ipv4.mr_rules_ops); 293 } 294 295 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb, 296 struct netlink_ext_ack *extack) 297 { 298 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack); 299 } 300 301 static unsigned int ipmr_rules_seq_read(const struct net *net) 302 { 303 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR); 304 } 305 306 bool ipmr_rule_default(const struct fib_rule *rule) 307 { 308 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT; 309 } 310 EXPORT_SYMBOL(ipmr_rule_default); 311 #else 312 #define ipmr_for_each_table(mrt, net) \ 313 for (mrt = net->ipv4.mrt; mrt; mrt = NULL) 314 315 static struct mr_table *ipmr_mr_table_iter(struct net *net, 316 struct mr_table *mrt) 317 { 318 if (!mrt) 319 return net->ipv4.mrt; 320 return NULL; 321 } 322 323 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 324 { 325 return net->ipv4.mrt; 326 } 327 328 #define __ipmr_get_table ipmr_get_table 329 330 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 331 struct mr_table **mrt) 332 { 333 *mrt = net->ipv4.mrt; 334 return 0; 335 } 336 337 static int __net_init ipmr_rules_init(struct net *net) 338 { 339 struct mr_table *mrt; 340 341 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 342 if (IS_ERR(mrt)) 343 return PTR_ERR(mrt); 344 net->ipv4.mrt = mrt; 345 return 0; 346 } 347 348 static void __net_exit ipmr_rules_exit(struct net *net) 349 { 350 ASSERT_RTNL(); 351 ipmr_free_table(net->ipv4.mrt); 352 net->ipv4.mrt = NULL; 353 } 354 355 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb, 356 struct netlink_ext_ack *extack) 357 { 358 return 0; 359 } 360 361 static unsigned int ipmr_rules_seq_read(const struct net *net) 362 { 363 return 0; 364 } 365 366 bool ipmr_rule_default(const struct fib_rule *rule) 367 { 368 return true; 369 } 370 EXPORT_SYMBOL(ipmr_rule_default); 371 #endif 372 373 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg, 374 const void *ptr) 375 { 376 const struct mfc_cache_cmp_arg *cmparg = arg->key; 377 const struct mfc_cache *c = ptr; 378 379 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp || 380 cmparg->mfc_origin != c->mfc_origin; 381 } 382 383 static const struct rhashtable_params ipmr_rht_params = { 384 .head_offset = offsetof(struct mr_mfc, mnode), 385 .key_offset = offsetof(struct mfc_cache, cmparg), 386 .key_len = sizeof(struct mfc_cache_cmp_arg), 387 .nelem_hint = 3, 388 .obj_cmpfn = ipmr_hash_cmp, 389 .automatic_shrinking = true, 390 }; 391 392 static void ipmr_new_table_set(struct mr_table *mrt, 393 struct net *net) 394 { 395 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 396 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables); 397 #endif 398 } 399 400 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = { 401 .mfc_mcastgrp = htonl(INADDR_ANY), 402 .mfc_origin = htonl(INADDR_ANY), 403 }; 404 405 static struct mr_table_ops ipmr_mr_table_ops = { 406 .rht_params = &ipmr_rht_params, 407 .cmparg_any = &ipmr_mr_table_ops_cmparg_any, 408 }; 409 410 static struct mr_table *ipmr_new_table(struct net *net, u32 id) 411 { 412 struct mr_table *mrt; 413 414 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */ 415 if (id != RT_TABLE_DEFAULT && id >= 1000000000) 416 return ERR_PTR(-EINVAL); 417 418 mrt = __ipmr_get_table(net, id); 419 if (mrt) 420 return mrt; 421 422 return mr_table_alloc(net, id, &ipmr_mr_table_ops, 423 ipmr_expire_process, ipmr_new_table_set); 424 } 425 426 static void ipmr_free_table(struct mr_table *mrt) 427 { 428 struct net *net = read_pnet(&mrt->net); 429 430 WARN_ON_ONCE(!mr_can_free_table(net)); 431 432 timer_shutdown_sync(&mrt->ipmr_expire_timer); 433 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC | 434 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC); 435 rhltable_destroy(&mrt->mfc_hash); 436 kfree(mrt); 437 } 438 439 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ 440 441 /* Initialize ipmr pimreg/tunnel in_device */ 442 static bool ipmr_init_vif_indev(const struct net_device *dev) 443 { 444 struct in_device *in_dev; 445 446 ASSERT_RTNL(); 447 448 in_dev = __in_dev_get_rtnl(dev); 449 if (!in_dev) 450 return false; 451 ipv4_devconf_setall(in_dev); 452 neigh_parms_data_state_setall(in_dev->arp_parms); 453 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 454 455 return true; 456 } 457 458 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v) 459 { 460 struct net_device *tunnel_dev, *new_dev; 461 struct ip_tunnel_parm_kern p = { }; 462 int err; 463 464 tunnel_dev = __dev_get_by_name(net, "tunl0"); 465 if (!tunnel_dev) 466 goto out; 467 468 p.iph.daddr = v->vifc_rmt_addr.s_addr; 469 p.iph.saddr = v->vifc_lcl_addr.s_addr; 470 p.iph.version = 4; 471 p.iph.ihl = 5; 472 p.iph.protocol = IPPROTO_IPIP; 473 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 474 475 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl) 476 goto out; 477 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p, 478 SIOCADDTUNNEL); 479 if (err) 480 goto out; 481 482 new_dev = __dev_get_by_name(net, p.name); 483 if (!new_dev) 484 goto out; 485 486 new_dev->flags |= IFF_MULTICAST; 487 if (!ipmr_init_vif_indev(new_dev)) 488 goto out_unregister; 489 if (dev_open(new_dev, NULL)) 490 goto out_unregister; 491 dev_hold(new_dev); 492 err = dev_set_allmulti(new_dev, 1); 493 if (err) { 494 dev_close(new_dev); 495 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p, 496 SIOCDELTUNNEL); 497 dev_put(new_dev); 498 new_dev = ERR_PTR(err); 499 } 500 return new_dev; 501 502 out_unregister: 503 unregister_netdevice(new_dev); 504 out: 505 return ERR_PTR(-ENOBUFS); 506 } 507 508 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 509 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) 510 { 511 struct net *net = dev_net(dev); 512 struct mr_table *mrt; 513 struct flowi4 fl4 = { 514 .flowi4_oif = dev->ifindex, 515 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 516 .flowi4_mark = skb->mark, 517 }; 518 int err; 519 520 err = ipmr_fib_lookup(net, &fl4, &mrt); 521 if (err < 0) { 522 kfree_skb(skb); 523 return err; 524 } 525 526 DEV_STATS_ADD(dev, tx_bytes, skb->len); 527 DEV_STATS_INC(dev, tx_packets); 528 rcu_read_lock(); 529 530 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */ 531 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num), 532 IGMPMSG_WHOLEPKT); 533 534 rcu_read_unlock(); 535 kfree_skb(skb); 536 return NETDEV_TX_OK; 537 } 538 539 static int reg_vif_get_iflink(const struct net_device *dev) 540 { 541 return 0; 542 } 543 544 static const struct net_device_ops reg_vif_netdev_ops = { 545 .ndo_start_xmit = reg_vif_xmit, 546 .ndo_get_iflink = reg_vif_get_iflink, 547 }; 548 549 static void reg_vif_setup(struct net_device *dev) 550 { 551 dev->type = ARPHRD_PIMREG; 552 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; 553 dev->flags = IFF_NOARP; 554 dev->netdev_ops = ®_vif_netdev_ops; 555 dev->needs_free_netdev = true; 556 dev->netns_immutable = true; 557 } 558 559 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 560 { 561 struct net_device *dev; 562 char name[IFNAMSIZ]; 563 564 if (mrt->id == RT_TABLE_DEFAULT) 565 sprintf(name, "pimreg"); 566 else 567 sprintf(name, "pimreg%u", mrt->id); 568 569 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 570 571 if (!dev) 572 return NULL; 573 574 dev_net_set(dev, net); 575 576 if (register_netdevice(dev)) { 577 free_netdev(dev); 578 return NULL; 579 } 580 581 if (!ipmr_init_vif_indev(dev)) 582 goto failure; 583 if (dev_open(dev, NULL)) 584 goto failure; 585 586 dev_hold(dev); 587 588 return dev; 589 590 failure: 591 unregister_netdevice(dev); 592 return NULL; 593 } 594 595 /* called with rcu_read_lock() */ 596 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb, 597 unsigned int pimlen) 598 { 599 struct net_device *reg_dev = NULL; 600 struct iphdr *encap; 601 int vif_num; 602 603 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); 604 /* Check that: 605 * a. packet is really sent to a multicast group 606 * b. packet is not a NULL-REGISTER 607 * c. packet is not truncated 608 */ 609 if (!ipv4_is_multicast(encap->daddr) || 610 encap->tot_len == 0 || 611 ntohs(encap->tot_len) + pimlen > skb->len) 612 return 1; 613 614 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */ 615 vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 616 if (vif_num >= 0) 617 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]); 618 if (!reg_dev) 619 return 1; 620 621 skb->mac_header = skb->network_header; 622 skb_pull(skb, (u8 *)encap - skb->data); 623 skb_reset_network_header(skb); 624 skb->protocol = htons(ETH_P_IP); 625 skb->ip_summed = CHECKSUM_NONE; 626 627 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 628 629 netif_rx(skb); 630 631 return NET_RX_SUCCESS; 632 } 633 #else 634 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 635 { 636 return NULL; 637 } 638 #endif 639 640 static int call_ipmr_vif_entry_notifiers(struct net *net, 641 enum fib_event_type event_type, 642 struct vif_device *vif, 643 struct net_device *vif_dev, 644 vifi_t vif_index, u32 tb_id) 645 { 646 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type, 647 vif, vif_dev, vif_index, tb_id, 648 &net->ipv4.ipmr_seq); 649 } 650 651 static int call_ipmr_mfc_entry_notifiers(struct net *net, 652 enum fib_event_type event_type, 653 struct mfc_cache *mfc, u32 tb_id) 654 { 655 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type, 656 &mfc->_c, tb_id, &net->ipv4.ipmr_seq); 657 } 658 659 /** 660 * vif_delete - Delete a VIF entry 661 * @mrt: Table to delete from 662 * @vifi: VIF identifier to delete 663 * @notify: Set to 1, if the caller is a notifier_call 664 * @head: if unregistering the VIF, place it on this queue 665 */ 666 static int vif_delete(struct mr_table *mrt, int vifi, int notify, 667 struct list_head *head) 668 { 669 struct net *net = read_pnet(&mrt->net); 670 struct vif_device *v; 671 struct net_device *dev; 672 struct in_device *in_dev; 673 674 if (vifi < 0 || vifi >= mrt->maxvif) 675 return -EADDRNOTAVAIL; 676 677 v = &mrt->vif_table[vifi]; 678 679 dev = rtnl_dereference(v->dev); 680 if (!dev) 681 return -EADDRNOTAVAIL; 682 683 spin_lock(&mrt_lock); 684 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev, 685 vifi, mrt->id); 686 RCU_INIT_POINTER(v->dev, NULL); 687 688 if (vifi == mrt->mroute_reg_vif_num) { 689 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */ 690 WRITE_ONCE(mrt->mroute_reg_vif_num, -1); 691 } 692 if (vifi + 1 == mrt->maxvif) { 693 int tmp; 694 695 for (tmp = vifi - 1; tmp >= 0; tmp--) { 696 if (VIF_EXISTS(mrt, tmp)) 697 break; 698 } 699 WRITE_ONCE(mrt->maxvif, tmp + 1); 700 } 701 702 spin_unlock(&mrt_lock); 703 704 dev_set_allmulti(dev, -1); 705 706 in_dev = __in_dev_get_rtnl(dev); 707 if (in_dev) { 708 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; 709 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 710 NETCONFA_MC_FORWARDING, 711 dev->ifindex, &in_dev->cnf); 712 ip_rt_multicast_event(in_dev); 713 } 714 715 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify) 716 unregister_netdevice_queue(dev, head); 717 718 netdev_put(dev, &v->dev_tracker); 719 return 0; 720 } 721 722 static void ipmr_cache_free_rcu(struct rcu_head *head) 723 { 724 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu); 725 726 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c); 727 } 728 729 static void ipmr_cache_free(struct mfc_cache *c) 730 { 731 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu); 732 } 733 734 /* Destroy an unresolved cache entry, killing queued skbs 735 * and reporting error to netlink readers. 736 */ 737 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c) 738 { 739 struct net *net = read_pnet(&mrt->net); 740 struct sk_buff *skb; 741 struct nlmsgerr *e; 742 743 atomic_dec(&mrt->cache_resolve_queue_len); 744 745 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) { 746 if (ip_hdr(skb)->version == 0) { 747 struct nlmsghdr *nlh = skb_pull(skb, 748 sizeof(struct iphdr)); 749 nlh->nlmsg_type = NLMSG_ERROR; 750 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 751 skb_trim(skb, nlh->nlmsg_len); 752 e = nlmsg_data(nlh); 753 e->error = -ETIMEDOUT; 754 memset(&e->msg, 0, sizeof(e->msg)); 755 756 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 757 } else { 758 kfree_skb(skb); 759 } 760 } 761 762 ipmr_cache_free(c); 763 } 764 765 /* Timer process for the unresolved queue. */ 766 static void ipmr_expire_process(struct timer_list *t) 767 { 768 struct mr_table *mrt = timer_container_of(mrt, t, ipmr_expire_timer); 769 struct mr_mfc *c, *next; 770 unsigned long expires; 771 unsigned long now; 772 773 if (!spin_trylock(&mfc_unres_lock)) { 774 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 775 return; 776 } 777 778 if (list_empty(&mrt->mfc_unres_queue)) 779 goto out; 780 781 now = jiffies; 782 expires = 10*HZ; 783 784 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 785 if (time_after(c->mfc_un.unres.expires, now)) { 786 unsigned long interval = c->mfc_un.unres.expires - now; 787 if (interval < expires) 788 expires = interval; 789 continue; 790 } 791 792 list_del(&c->list); 793 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE); 794 ipmr_destroy_unres(mrt, (struct mfc_cache *)c); 795 } 796 797 if (!list_empty(&mrt->mfc_unres_queue)) 798 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 799 800 out: 801 spin_unlock(&mfc_unres_lock); 802 } 803 804 /* Fill oifs list. It is called under locked mrt_lock. */ 805 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache, 806 unsigned char *ttls) 807 { 808 int vifi; 809 810 cache->mfc_un.res.minvif = MAXVIFS; 811 cache->mfc_un.res.maxvif = 0; 812 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 813 814 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 815 if (VIF_EXISTS(mrt, vifi) && 816 ttls[vifi] && ttls[vifi] < 255) { 817 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 818 if (cache->mfc_un.res.minvif > vifi) 819 cache->mfc_un.res.minvif = vifi; 820 if (cache->mfc_un.res.maxvif <= vifi) 821 cache->mfc_un.res.maxvif = vifi + 1; 822 } 823 } 824 WRITE_ONCE(cache->mfc_un.res.lastuse, jiffies); 825 } 826 827 static int vif_add(struct net *net, struct mr_table *mrt, 828 struct vifctl *vifc, int mrtsock) 829 { 830 struct netdev_phys_item_id ppid = { }; 831 int vifi = vifc->vifc_vifi; 832 struct vif_device *v = &mrt->vif_table[vifi]; 833 struct net_device *dev; 834 struct in_device *in_dev; 835 int err; 836 837 /* Is vif busy ? */ 838 if (VIF_EXISTS(mrt, vifi)) 839 return -EADDRINUSE; 840 841 switch (vifc->vifc_flags) { 842 case VIFF_REGISTER: 843 if (!ipmr_pimsm_enabled()) 844 return -EINVAL; 845 /* Special Purpose VIF in PIM 846 * All the packets will be sent to the daemon 847 */ 848 if (mrt->mroute_reg_vif_num >= 0) 849 return -EADDRINUSE; 850 dev = ipmr_reg_vif(net, mrt); 851 if (!dev) 852 return -ENOBUFS; 853 err = dev_set_allmulti(dev, 1); 854 if (err) { 855 unregister_netdevice(dev); 856 dev_put(dev); 857 return err; 858 } 859 break; 860 case VIFF_TUNNEL: 861 dev = ipmr_new_tunnel(net, vifc); 862 if (IS_ERR(dev)) 863 return PTR_ERR(dev); 864 break; 865 case VIFF_USE_IFINDEX: 866 case 0: 867 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 868 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 869 if (dev && !__in_dev_get_rtnl(dev)) { 870 dev_put(dev); 871 return -EADDRNOTAVAIL; 872 } 873 } else { 874 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 875 } 876 if (!dev) 877 return -EADDRNOTAVAIL; 878 err = dev_set_allmulti(dev, 1); 879 if (err) { 880 dev_put(dev); 881 return err; 882 } 883 break; 884 default: 885 return -EINVAL; 886 } 887 888 in_dev = __in_dev_get_rtnl(dev); 889 if (!in_dev) { 890 dev_put(dev); 891 return -EADDRNOTAVAIL; 892 } 893 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 894 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING, 895 dev->ifindex, &in_dev->cnf); 896 ip_rt_multicast_event(in_dev); 897 898 /* Fill in the VIF structures */ 899 vif_device_init(v, dev, vifc->vifc_rate_limit, 900 vifc->vifc_threshold, 901 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0), 902 (VIFF_TUNNEL | VIFF_REGISTER)); 903 904 err = dev_get_port_parent_id(dev, &ppid, true); 905 if (err == 0) { 906 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len); 907 v->dev_parent_id.id_len = ppid.id_len; 908 } else { 909 v->dev_parent_id.id_len = 0; 910 } 911 912 v->local = vifc->vifc_lcl_addr.s_addr; 913 v->remote = vifc->vifc_rmt_addr.s_addr; 914 915 /* And finish update writing critical data */ 916 spin_lock(&mrt_lock); 917 rcu_assign_pointer(v->dev, dev); 918 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC); 919 if (v->flags & VIFF_REGISTER) { 920 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */ 921 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi); 922 } 923 if (vifi+1 > mrt->maxvif) 924 WRITE_ONCE(mrt->maxvif, vifi + 1); 925 spin_unlock(&mrt_lock); 926 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev, 927 vifi, mrt->id); 928 return 0; 929 } 930 931 /* called with rcu_read_lock() */ 932 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 933 __be32 origin, 934 __be32 mcastgrp) 935 { 936 struct mfc_cache_cmp_arg arg = { 937 .mfc_mcastgrp = mcastgrp, 938 .mfc_origin = origin 939 }; 940 941 return mr_mfc_find(mrt, &arg); 942 } 943 944 /* Look for a (*,G) entry */ 945 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 946 __be32 mcastgrp, int vifi) 947 { 948 struct mfc_cache_cmp_arg arg = { 949 .mfc_mcastgrp = mcastgrp, 950 .mfc_origin = htonl(INADDR_ANY) 951 }; 952 953 if (mcastgrp == htonl(INADDR_ANY)) 954 return mr_mfc_find_any_parent(mrt, vifi); 955 return mr_mfc_find_any(mrt, vifi, &arg); 956 } 957 958 /* Look for a (S,G,iif) entry if parent != -1 */ 959 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt, 960 __be32 origin, __be32 mcastgrp, 961 int parent) 962 { 963 struct mfc_cache_cmp_arg arg = { 964 .mfc_mcastgrp = mcastgrp, 965 .mfc_origin = origin, 966 }; 967 968 return mr_mfc_find_parent(mrt, &arg, parent); 969 } 970 971 /* Allocate a multicast cache entry */ 972 static struct mfc_cache *ipmr_cache_alloc(void) 973 { 974 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 975 976 if (c) { 977 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 978 c->_c.mfc_un.res.minvif = MAXVIFS; 979 c->_c.free = ipmr_cache_free_rcu; 980 refcount_set(&c->_c.mfc_un.res.refcount, 1); 981 } 982 return c; 983 } 984 985 static struct mfc_cache *ipmr_cache_alloc_unres(void) 986 { 987 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 988 989 if (c) { 990 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved); 991 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ; 992 } 993 return c; 994 } 995 996 /* A cache entry has gone into a resolved state from queued */ 997 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 998 struct mfc_cache *uc, struct mfc_cache *c) 999 { 1000 struct sk_buff *skb; 1001 struct nlmsgerr *e; 1002 1003 /* Play the pending entries through our router */ 1004 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) { 1005 if (ip_hdr(skb)->version == 0) { 1006 struct nlmsghdr *nlh = skb_pull(skb, 1007 sizeof(struct iphdr)); 1008 1009 if (mr_fill_mroute(mrt, skb, &c->_c, 1010 nlmsg_data(nlh)) > 0) { 1011 nlh->nlmsg_len = skb_tail_pointer(skb) - 1012 (u8 *)nlh; 1013 } else { 1014 nlh->nlmsg_type = NLMSG_ERROR; 1015 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1016 skb_trim(skb, nlh->nlmsg_len); 1017 e = nlmsg_data(nlh); 1018 e->error = -EMSGSIZE; 1019 memset(&e->msg, 0, sizeof(e->msg)); 1020 } 1021 1022 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1023 } else { 1024 rcu_read_lock(); 1025 ip_mr_forward(net, mrt, skb->dev, skb, c, 0); 1026 rcu_read_unlock(); 1027 } 1028 } 1029 } 1030 1031 /* Bounce a cache query up to mrouted and netlink. 1032 * 1033 * Called under rcu_read_lock(). 1034 */ 1035 static int ipmr_cache_report(const struct mr_table *mrt, 1036 struct sk_buff *pkt, vifi_t vifi, int assert) 1037 { 1038 const int ihl = ip_hdrlen(pkt); 1039 struct sock *mroute_sk; 1040 struct igmphdr *igmp; 1041 struct igmpmsg *msg; 1042 struct sk_buff *skb; 1043 int ret; 1044 1045 mroute_sk = rcu_dereference(mrt->mroute_sk); 1046 if (!mroute_sk) 1047 return -EINVAL; 1048 1049 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) 1050 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 1051 else 1052 skb = alloc_skb(128, GFP_ATOMIC); 1053 1054 if (!skb) 1055 return -ENOBUFS; 1056 1057 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) { 1058 /* Ugly, but we have no choice with this interface. 1059 * Duplicate old header, fix ihl, length etc. 1060 * And all this only to mangle msg->im_msgtype and 1061 * to set msg->im_mbz to "mbz" :-) 1062 */ 1063 skb_push(skb, sizeof(struct iphdr)); 1064 skb_reset_network_header(skb); 1065 skb_reset_transport_header(skb); 1066 msg = (struct igmpmsg *)skb_network_header(skb); 1067 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 1068 msg->im_msgtype = assert; 1069 msg->im_mbz = 0; 1070 if (assert == IGMPMSG_WRVIFWHOLE) { 1071 msg->im_vif = vifi; 1072 msg->im_vif_hi = vifi >> 8; 1073 } else { 1074 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */ 1075 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 1076 1077 msg->im_vif = vif_num; 1078 msg->im_vif_hi = vif_num >> 8; 1079 } 1080 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 1081 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 1082 sizeof(struct iphdr)); 1083 } else { 1084 /* Copy the IP header */ 1085 skb_set_network_header(skb, skb->len); 1086 skb_put(skb, ihl); 1087 skb_copy_to_linear_data(skb, pkt->data, ihl); 1088 /* Flag to the kernel this is a route add */ 1089 ip_hdr(skb)->protocol = 0; 1090 msg = (struct igmpmsg *)skb_network_header(skb); 1091 msg->im_vif = vifi; 1092 msg->im_vif_hi = vifi >> 8; 1093 ipv4_pktinfo_prepare(mroute_sk, pkt, false); 1094 memcpy(skb->cb, pkt->cb, sizeof(skb->cb)); 1095 /* Add our header */ 1096 igmp = skb_put(skb, sizeof(struct igmphdr)); 1097 igmp->type = assert; 1098 msg->im_msgtype = assert; 1099 igmp->code = 0; 1100 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1101 skb->transport_header = skb->network_header; 1102 } 1103 1104 igmpmsg_netlink_event(mrt, skb); 1105 1106 /* Deliver to mrouted */ 1107 ret = sock_queue_rcv_skb(mroute_sk, skb); 1108 1109 if (ret < 0) { 1110 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1111 kfree_skb(skb); 1112 } 1113 1114 return ret; 1115 } 1116 1117 /* Queue a packet for resolution. It gets locked cache entry! */ 1118 /* Called under rcu_read_lock() */ 1119 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, 1120 struct sk_buff *skb, struct net_device *dev) 1121 { 1122 const struct iphdr *iph = ip_hdr(skb); 1123 struct mfc_cache *c; 1124 bool found = false; 1125 int err; 1126 1127 spin_lock_bh(&mfc_unres_lock); 1128 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) { 1129 if (c->mfc_mcastgrp == iph->daddr && 1130 c->mfc_origin == iph->saddr) { 1131 found = true; 1132 break; 1133 } 1134 } 1135 1136 if (!found) { 1137 /* Create a new entry if allowable */ 1138 c = ipmr_cache_alloc_unres(); 1139 if (!c) { 1140 spin_unlock_bh(&mfc_unres_lock); 1141 1142 kfree_skb(skb); 1143 return -ENOBUFS; 1144 } 1145 1146 /* Fill in the new cache entry */ 1147 c->_c.mfc_parent = -1; 1148 c->mfc_origin = iph->saddr; 1149 c->mfc_mcastgrp = iph->daddr; 1150 1151 /* Reflect first query at mrouted. */ 1152 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1153 1154 if (err < 0) { 1155 /* If the report failed throw the cache entry 1156 out - Brad Parker 1157 */ 1158 spin_unlock_bh(&mfc_unres_lock); 1159 1160 ipmr_cache_free(c); 1161 kfree_skb(skb); 1162 return err; 1163 } 1164 1165 atomic_inc(&mrt->cache_resolve_queue_len); 1166 list_add(&c->_c.list, &mrt->mfc_unres_queue); 1167 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1168 1169 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1170 mod_timer(&mrt->ipmr_expire_timer, 1171 c->_c.mfc_un.unres.expires); 1172 } 1173 1174 /* See if we can append the packet */ 1175 if (c->_c.mfc_un.unres.unresolved.qlen > 3) { 1176 kfree_skb(skb); 1177 err = -ENOBUFS; 1178 } else { 1179 if (dev) { 1180 skb->dev = dev; 1181 skb->skb_iif = dev->ifindex; 1182 } 1183 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb); 1184 err = 0; 1185 } 1186 1187 spin_unlock_bh(&mfc_unres_lock); 1188 return err; 1189 } 1190 1191 /* MFC cache manipulation by user space mroute daemon */ 1192 1193 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1194 { 1195 struct net *net = read_pnet(&mrt->net); 1196 struct mfc_cache *c; 1197 1198 /* The entries are added/deleted only under RTNL */ 1199 rcu_read_lock(); 1200 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1201 mfc->mfcc_mcastgrp.s_addr, parent); 1202 rcu_read_unlock(); 1203 if (!c) 1204 return -ENOENT; 1205 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params); 1206 list_del_rcu(&c->_c.list); 1207 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id); 1208 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1209 mr_cache_put(&c->_c); 1210 1211 return 0; 1212 } 1213 1214 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1215 struct mfcctl *mfc, int mrtsock, int parent) 1216 { 1217 struct mfc_cache *uc, *c; 1218 struct mr_mfc *_uc; 1219 bool found; 1220 int ret; 1221 1222 if (mfc->mfcc_parent >= MAXVIFS) 1223 return -ENFILE; 1224 1225 /* The entries are added/deleted only under RTNL */ 1226 rcu_read_lock(); 1227 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1228 mfc->mfcc_mcastgrp.s_addr, parent); 1229 rcu_read_unlock(); 1230 if (c) { 1231 spin_lock(&mrt_lock); 1232 c->_c.mfc_parent = mfc->mfcc_parent; 1233 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls); 1234 if (!mrtsock) 1235 c->_c.mfc_flags |= MFC_STATIC; 1236 spin_unlock(&mrt_lock); 1237 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c, 1238 mrt->id); 1239 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1240 return 0; 1241 } 1242 1243 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1244 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1245 return -EINVAL; 1246 1247 c = ipmr_cache_alloc(); 1248 if (!c) 1249 return -ENOMEM; 1250 1251 c->mfc_origin = mfc->mfcc_origin.s_addr; 1252 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1253 c->_c.mfc_parent = mfc->mfcc_parent; 1254 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls); 1255 if (!mrtsock) 1256 c->_c.mfc_flags |= MFC_STATIC; 1257 1258 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode, 1259 ipmr_rht_params); 1260 if (ret) { 1261 pr_err("ipmr: rhtable insert error %d\n", ret); 1262 ipmr_cache_free(c); 1263 return ret; 1264 } 1265 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list); 1266 /* Check to see if we resolved a queued list. If so we 1267 * need to send on the frames and tidy up. 1268 */ 1269 found = false; 1270 spin_lock_bh(&mfc_unres_lock); 1271 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) { 1272 uc = (struct mfc_cache *)_uc; 1273 if (uc->mfc_origin == c->mfc_origin && 1274 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1275 list_del(&_uc->list); 1276 atomic_dec(&mrt->cache_resolve_queue_len); 1277 found = true; 1278 break; 1279 } 1280 } 1281 if (list_empty(&mrt->mfc_unres_queue)) 1282 timer_delete(&mrt->ipmr_expire_timer); 1283 spin_unlock_bh(&mfc_unres_lock); 1284 1285 if (found) { 1286 ipmr_cache_resolve(net, mrt, uc, c); 1287 ipmr_cache_free(uc); 1288 } 1289 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id); 1290 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1291 return 0; 1292 } 1293 1294 /* Close the multicast socket, and clear the vif tables etc */ 1295 static void mroute_clean_tables(struct mr_table *mrt, int flags) 1296 { 1297 struct net *net = read_pnet(&mrt->net); 1298 struct mr_mfc *c, *tmp; 1299 struct mfc_cache *cache; 1300 LIST_HEAD(list); 1301 int i; 1302 1303 /* Shut down all active vif entries */ 1304 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) { 1305 for (i = 0; i < mrt->maxvif; i++) { 1306 if (((mrt->vif_table[i].flags & VIFF_STATIC) && 1307 !(flags & MRT_FLUSH_VIFS_STATIC)) || 1308 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS))) 1309 continue; 1310 vif_delete(mrt, i, 0, &list); 1311 } 1312 unregister_netdevice_many(&list); 1313 } 1314 1315 /* Wipe the cache */ 1316 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) { 1317 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1318 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) || 1319 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC))) 1320 continue; 1321 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1322 list_del_rcu(&c->list); 1323 cache = (struct mfc_cache *)c; 1324 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache, 1325 mrt->id); 1326 mroute_netlink_event(mrt, cache, RTM_DELROUTE); 1327 mr_cache_put(c); 1328 } 1329 } 1330 1331 if (flags & MRT_FLUSH_MFC) { 1332 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1333 spin_lock_bh(&mfc_unres_lock); 1334 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1335 list_del(&c->list); 1336 cache = (struct mfc_cache *)c; 1337 mroute_netlink_event(mrt, cache, RTM_DELROUTE); 1338 ipmr_destroy_unres(mrt, cache); 1339 } 1340 spin_unlock_bh(&mfc_unres_lock); 1341 } 1342 } 1343 } 1344 1345 /* called from ip_ra_control(), before an RCU grace period, 1346 * we don't need to call synchronize_rcu() here 1347 */ 1348 static void mrtsock_destruct(struct sock *sk) 1349 { 1350 struct net *net = sock_net(sk); 1351 struct mr_table *mrt; 1352 1353 rtnl_lock(); 1354 ipmr_for_each_table(mrt, net) { 1355 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1356 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1357 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1358 NETCONFA_MC_FORWARDING, 1359 NETCONFA_IFINDEX_ALL, 1360 net->ipv4.devconf_all); 1361 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1362 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC); 1363 } 1364 } 1365 rtnl_unlock(); 1366 } 1367 1368 /* Socket options and virtual interface manipulation. The whole 1369 * virtual interface system is a complete heap, but unfortunately 1370 * that's how BSD mrouted happens to think. Maybe one day with a proper 1371 * MOSPF/PIM router set up we can clean this up. 1372 */ 1373 1374 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval, 1375 unsigned int optlen) 1376 { 1377 struct net *net = sock_net(sk); 1378 int val, ret = 0, parent = 0; 1379 struct mr_table *mrt; 1380 struct vifctl vif; 1381 struct mfcctl mfc; 1382 bool do_wrvifwhole; 1383 u32 uval; 1384 1385 /* There's one exception to the lock - MRT_DONE which needs to unlock */ 1386 rtnl_lock(); 1387 if (sk->sk_type != SOCK_RAW || 1388 inet_sk(sk)->inet_num != IPPROTO_IGMP) { 1389 ret = -EOPNOTSUPP; 1390 goto out_unlock; 1391 } 1392 1393 mrt = __ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1394 if (!mrt) { 1395 ret = -ENOENT; 1396 goto out_unlock; 1397 } 1398 if (optname != MRT_INIT) { 1399 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1400 !ns_capable(net->user_ns, CAP_NET_ADMIN)) { 1401 ret = -EACCES; 1402 goto out_unlock; 1403 } 1404 } 1405 1406 switch (optname) { 1407 case MRT_INIT: 1408 if (optlen != sizeof(int)) { 1409 ret = -EINVAL; 1410 break; 1411 } 1412 if (rtnl_dereference(mrt->mroute_sk)) { 1413 ret = -EADDRINUSE; 1414 break; 1415 } 1416 1417 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1418 if (ret == 0) { 1419 rcu_assign_pointer(mrt->mroute_sk, sk); 1420 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1421 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1422 NETCONFA_MC_FORWARDING, 1423 NETCONFA_IFINDEX_ALL, 1424 net->ipv4.devconf_all); 1425 } 1426 break; 1427 case MRT_DONE: 1428 if (sk != rcu_access_pointer(mrt->mroute_sk)) { 1429 ret = -EACCES; 1430 } else { 1431 /* We need to unlock here because mrtsock_destruct takes 1432 * care of rtnl itself and we can't change that due to 1433 * the IP_ROUTER_ALERT setsockopt which runs without it. 1434 */ 1435 rtnl_unlock(); 1436 ret = ip_ra_control(sk, 0, NULL); 1437 goto out; 1438 } 1439 break; 1440 case MRT_ADD_VIF: 1441 case MRT_DEL_VIF: 1442 if (optlen != sizeof(vif)) { 1443 ret = -EINVAL; 1444 break; 1445 } 1446 if (copy_from_sockptr(&vif, optval, sizeof(vif))) { 1447 ret = -EFAULT; 1448 break; 1449 } 1450 if (vif.vifc_vifi >= MAXVIFS) { 1451 ret = -ENFILE; 1452 break; 1453 } 1454 if (optname == MRT_ADD_VIF) { 1455 ret = vif_add(net, mrt, &vif, 1456 sk == rtnl_dereference(mrt->mroute_sk)); 1457 } else { 1458 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1459 } 1460 break; 1461 /* Manipulate the forwarding caches. These live 1462 * in a sort of kernel/user symbiosis. 1463 */ 1464 case MRT_ADD_MFC: 1465 case MRT_DEL_MFC: 1466 parent = -1; 1467 fallthrough; 1468 case MRT_ADD_MFC_PROXY: 1469 case MRT_DEL_MFC_PROXY: 1470 if (optlen != sizeof(mfc)) { 1471 ret = -EINVAL; 1472 break; 1473 } 1474 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) { 1475 ret = -EFAULT; 1476 break; 1477 } 1478 if (parent == 0) 1479 parent = mfc.mfcc_parent; 1480 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1481 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1482 else 1483 ret = ipmr_mfc_add(net, mrt, &mfc, 1484 sk == rtnl_dereference(mrt->mroute_sk), 1485 parent); 1486 break; 1487 case MRT_FLUSH: 1488 if (optlen != sizeof(val)) { 1489 ret = -EINVAL; 1490 break; 1491 } 1492 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1493 ret = -EFAULT; 1494 break; 1495 } 1496 mroute_clean_tables(mrt, val); 1497 break; 1498 /* Control PIM assert. */ 1499 case MRT_ASSERT: 1500 if (optlen != sizeof(val)) { 1501 ret = -EINVAL; 1502 break; 1503 } 1504 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1505 ret = -EFAULT; 1506 break; 1507 } 1508 mrt->mroute_do_assert = val; 1509 break; 1510 case MRT_PIM: 1511 if (!ipmr_pimsm_enabled()) { 1512 ret = -ENOPROTOOPT; 1513 break; 1514 } 1515 if (optlen != sizeof(val)) { 1516 ret = -EINVAL; 1517 break; 1518 } 1519 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1520 ret = -EFAULT; 1521 break; 1522 } 1523 1524 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE); 1525 val = !!val; 1526 if (val != mrt->mroute_do_pim) { 1527 mrt->mroute_do_pim = val; 1528 mrt->mroute_do_assert = val; 1529 mrt->mroute_do_wrvifwhole = do_wrvifwhole; 1530 } 1531 break; 1532 case MRT_TABLE: 1533 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) { 1534 ret = -ENOPROTOOPT; 1535 break; 1536 } 1537 if (optlen != sizeof(uval)) { 1538 ret = -EINVAL; 1539 break; 1540 } 1541 if (copy_from_sockptr(&uval, optval, sizeof(uval))) { 1542 ret = -EFAULT; 1543 break; 1544 } 1545 1546 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1547 ret = -EBUSY; 1548 } else { 1549 mrt = ipmr_new_table(net, uval); 1550 if (IS_ERR(mrt)) 1551 ret = PTR_ERR(mrt); 1552 else 1553 raw_sk(sk)->ipmr_table = uval; 1554 } 1555 break; 1556 /* Spurious command, or MRT_VERSION which you cannot set. */ 1557 default: 1558 ret = -ENOPROTOOPT; 1559 } 1560 out_unlock: 1561 rtnl_unlock(); 1562 out: 1563 return ret; 1564 } 1565 1566 /* Execute if this ioctl is a special mroute ioctl */ 1567 int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1568 { 1569 switch (cmd) { 1570 /* These userspace buffers will be consumed by ipmr_ioctl() */ 1571 case SIOCGETVIFCNT: { 1572 struct sioc_vif_req buffer; 1573 1574 return sock_ioctl_inout(sk, cmd, arg, &buffer, 1575 sizeof(buffer)); 1576 } 1577 case SIOCGETSGCNT: { 1578 struct sioc_sg_req buffer; 1579 1580 return sock_ioctl_inout(sk, cmd, arg, &buffer, 1581 sizeof(buffer)); 1582 } 1583 } 1584 /* return code > 0 means that the ioctl was not executed */ 1585 return 1; 1586 } 1587 1588 /* Getsock opt support for the multicast routing system. */ 1589 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval, 1590 sockptr_t optlen) 1591 { 1592 int olr; 1593 int val; 1594 struct net *net = sock_net(sk); 1595 struct mr_table *mrt; 1596 1597 if (sk->sk_type != SOCK_RAW || 1598 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1599 return -EOPNOTSUPP; 1600 1601 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1602 if (!mrt) 1603 return -ENOENT; 1604 1605 switch (optname) { 1606 case MRT_VERSION: 1607 val = 0x0305; 1608 break; 1609 case MRT_PIM: 1610 if (!ipmr_pimsm_enabled()) 1611 return -ENOPROTOOPT; 1612 val = mrt->mroute_do_pim; 1613 break; 1614 case MRT_ASSERT: 1615 val = mrt->mroute_do_assert; 1616 break; 1617 default: 1618 return -ENOPROTOOPT; 1619 } 1620 1621 if (copy_from_sockptr(&olr, optlen, sizeof(int))) 1622 return -EFAULT; 1623 if (olr < 0) 1624 return -EINVAL; 1625 1626 olr = min_t(unsigned int, olr, sizeof(int)); 1627 1628 if (copy_to_sockptr(optlen, &olr, sizeof(int))) 1629 return -EFAULT; 1630 if (copy_to_sockptr(optval, &val, olr)) 1631 return -EFAULT; 1632 return 0; 1633 } 1634 1635 /* The IP multicast ioctl support routines. */ 1636 int ipmr_ioctl(struct sock *sk, int cmd, void *arg) 1637 { 1638 struct vif_device *vif; 1639 struct mfc_cache *c; 1640 struct net *net = sock_net(sk); 1641 struct sioc_vif_req *vr; 1642 struct sioc_sg_req *sr; 1643 struct mr_table *mrt; 1644 1645 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1646 if (!mrt) 1647 return -ENOENT; 1648 1649 switch (cmd) { 1650 case SIOCGETVIFCNT: 1651 vr = (struct sioc_vif_req *)arg; 1652 if (vr->vifi >= mrt->maxvif) 1653 return -EINVAL; 1654 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif); 1655 rcu_read_lock(); 1656 vif = &mrt->vif_table[vr->vifi]; 1657 if (VIF_EXISTS(mrt, vr->vifi)) { 1658 vr->icount = READ_ONCE(vif->pkt_in); 1659 vr->ocount = READ_ONCE(vif->pkt_out); 1660 vr->ibytes = READ_ONCE(vif->bytes_in); 1661 vr->obytes = READ_ONCE(vif->bytes_out); 1662 rcu_read_unlock(); 1663 1664 return 0; 1665 } 1666 rcu_read_unlock(); 1667 return -EADDRNOTAVAIL; 1668 case SIOCGETSGCNT: 1669 sr = (struct sioc_sg_req *)arg; 1670 1671 rcu_read_lock(); 1672 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr); 1673 if (c) { 1674 sr->pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt); 1675 sr->bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes); 1676 sr->wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if); 1677 rcu_read_unlock(); 1678 return 0; 1679 } 1680 rcu_read_unlock(); 1681 return -EADDRNOTAVAIL; 1682 default: 1683 return -ENOIOCTLCMD; 1684 } 1685 } 1686 1687 #ifdef CONFIG_COMPAT 1688 struct compat_sioc_sg_req { 1689 struct in_addr src; 1690 struct in_addr grp; 1691 compat_ulong_t pktcnt; 1692 compat_ulong_t bytecnt; 1693 compat_ulong_t wrong_if; 1694 }; 1695 1696 struct compat_sioc_vif_req { 1697 vifi_t vifi; /* Which iface */ 1698 compat_ulong_t icount; 1699 compat_ulong_t ocount; 1700 compat_ulong_t ibytes; 1701 compat_ulong_t obytes; 1702 }; 1703 1704 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1705 { 1706 struct compat_sioc_sg_req sr; 1707 struct compat_sioc_vif_req vr; 1708 struct vif_device *vif; 1709 struct mfc_cache *c; 1710 struct net *net = sock_net(sk); 1711 struct mr_table *mrt; 1712 1713 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1714 if (!mrt) 1715 return -ENOENT; 1716 1717 switch (cmd) { 1718 case SIOCGETVIFCNT: 1719 if (copy_from_user(&vr, arg, sizeof(vr))) 1720 return -EFAULT; 1721 if (vr.vifi >= mrt->maxvif) 1722 return -EINVAL; 1723 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif); 1724 rcu_read_lock(); 1725 vif = &mrt->vif_table[vr.vifi]; 1726 if (VIF_EXISTS(mrt, vr.vifi)) { 1727 vr.icount = READ_ONCE(vif->pkt_in); 1728 vr.ocount = READ_ONCE(vif->pkt_out); 1729 vr.ibytes = READ_ONCE(vif->bytes_in); 1730 vr.obytes = READ_ONCE(vif->bytes_out); 1731 rcu_read_unlock(); 1732 1733 if (copy_to_user(arg, &vr, sizeof(vr))) 1734 return -EFAULT; 1735 return 0; 1736 } 1737 rcu_read_unlock(); 1738 return -EADDRNOTAVAIL; 1739 case SIOCGETSGCNT: 1740 if (copy_from_user(&sr, arg, sizeof(sr))) 1741 return -EFAULT; 1742 1743 rcu_read_lock(); 1744 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1745 if (c) { 1746 sr.pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt); 1747 sr.bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes); 1748 sr.wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if); 1749 rcu_read_unlock(); 1750 1751 if (copy_to_user(arg, &sr, sizeof(sr))) 1752 return -EFAULT; 1753 return 0; 1754 } 1755 rcu_read_unlock(); 1756 return -EADDRNOTAVAIL; 1757 default: 1758 return -ENOIOCTLCMD; 1759 } 1760 } 1761 #endif 1762 1763 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1764 { 1765 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1766 struct net *net = dev_net(dev); 1767 struct mr_table *mrt; 1768 struct vif_device *v; 1769 int ct; 1770 1771 if (event != NETDEV_UNREGISTER) 1772 return NOTIFY_DONE; 1773 1774 ipmr_for_each_table(mrt, net) { 1775 v = &mrt->vif_table[0]; 1776 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1777 if (rcu_access_pointer(v->dev) == dev) 1778 vif_delete(mrt, ct, 1, NULL); 1779 } 1780 } 1781 return NOTIFY_DONE; 1782 } 1783 1784 static struct notifier_block ip_mr_notifier = { 1785 .notifier_call = ipmr_device_event, 1786 }; 1787 1788 /* Encapsulate a packet by attaching a valid IPIP header to it. 1789 * This avoids tunnel drivers and other mess and gives us the speed so 1790 * important for multicast video. 1791 */ 1792 static void ip_encap(struct net *net, struct sk_buff *skb, 1793 __be32 saddr, __be32 daddr) 1794 { 1795 struct iphdr *iph; 1796 const struct iphdr *old_iph = ip_hdr(skb); 1797 1798 skb_push(skb, sizeof(struct iphdr)); 1799 skb->transport_header = skb->network_header; 1800 skb_reset_network_header(skb); 1801 iph = ip_hdr(skb); 1802 1803 iph->version = 4; 1804 iph->tos = old_iph->tos; 1805 iph->ttl = old_iph->ttl; 1806 iph->frag_off = 0; 1807 iph->daddr = daddr; 1808 iph->saddr = saddr; 1809 iph->protocol = IPPROTO_IPIP; 1810 iph->ihl = 5; 1811 iph->tot_len = htons(skb->len); 1812 ip_select_ident(net, skb, NULL); 1813 ip_send_check(iph); 1814 1815 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1816 nf_reset_ct(skb); 1817 } 1818 1819 static inline int ipmr_forward_finish(struct net *net, struct sock *sk, 1820 struct sk_buff *skb) 1821 { 1822 struct ip_options *opt = &(IPCB(skb)->opt); 1823 1824 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS); 1825 1826 if (unlikely(opt->optlen)) 1827 ip_forward_options(skb); 1828 1829 return dst_output(net, sk, skb); 1830 } 1831 1832 #ifdef CONFIG_NET_SWITCHDEV 1833 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt, 1834 int in_vifi, int out_vifi) 1835 { 1836 struct vif_device *out_vif = &mrt->vif_table[out_vifi]; 1837 struct vif_device *in_vif = &mrt->vif_table[in_vifi]; 1838 1839 if (!skb->offload_l3_fwd_mark) 1840 return false; 1841 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len) 1842 return false; 1843 return netdev_phys_item_id_same(&out_vif->dev_parent_id, 1844 &in_vif->dev_parent_id); 1845 } 1846 #else 1847 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt, 1848 int in_vifi, int out_vifi) 1849 { 1850 return false; 1851 } 1852 #endif 1853 1854 /* Processing handlers for ipmr_forward, under rcu_read_lock() */ 1855 1856 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt, 1857 int in_vifi, struct sk_buff *skb, int vifi) 1858 { 1859 const struct iphdr *iph = ip_hdr(skb); 1860 struct vif_device *vif = &mrt->vif_table[vifi]; 1861 struct net_device *vif_dev; 1862 struct net_device *dev; 1863 struct rtable *rt; 1864 struct flowi4 fl4; 1865 int encap = 0; 1866 1867 vif_dev = vif_dev_read(vif); 1868 if (!vif_dev) 1869 goto out_free; 1870 1871 if (vif->flags & VIFF_REGISTER) { 1872 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 1873 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 1874 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len); 1875 DEV_STATS_INC(vif_dev, tx_packets); 1876 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1877 goto out_free; 1878 } 1879 1880 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi)) 1881 goto out_free; 1882 1883 if (vif->flags & VIFF_TUNNEL) { 1884 rt = ip_route_output_ports(net, &fl4, NULL, 1885 vif->remote, vif->local, 1886 0, 0, 1887 IPPROTO_IPIP, 1888 iph->tos & INET_DSCP_MASK, vif->link); 1889 if (IS_ERR(rt)) 1890 goto out_free; 1891 encap = sizeof(struct iphdr); 1892 } else { 1893 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1894 0, 0, 1895 IPPROTO_IPIP, 1896 iph->tos & INET_DSCP_MASK, vif->link); 1897 if (IS_ERR(rt)) 1898 goto out_free; 1899 } 1900 1901 dev = rt->dst.dev; 1902 1903 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1904 /* Do not fragment multicasts. Alas, IPv4 does not 1905 * allow to send ICMP, so that packets will disappear 1906 * to blackhole. 1907 */ 1908 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS); 1909 ip_rt_put(rt); 1910 goto out_free; 1911 } 1912 1913 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len; 1914 1915 if (skb_cow(skb, encap)) { 1916 ip_rt_put(rt); 1917 goto out_free; 1918 } 1919 1920 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 1921 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 1922 1923 skb_dst_drop(skb); 1924 skb_dst_set(skb, &rt->dst); 1925 ip_decrease_ttl(ip_hdr(skb)); 1926 1927 /* FIXME: forward and output firewalls used to be called here. 1928 * What do we do with netfilter? -- RR 1929 */ 1930 if (vif->flags & VIFF_TUNNEL) { 1931 ip_encap(net, skb, vif->local, vif->remote); 1932 /* FIXME: extra output firewall step used to be here. --RR */ 1933 DEV_STATS_INC(vif_dev, tx_packets); 1934 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len); 1935 } 1936 1937 IPCB(skb)->flags |= IPSKB_FORWARDED; 1938 1939 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1940 * not only before forwarding, but after forwarding on all output 1941 * interfaces. It is clear, if mrouter runs a multicasting 1942 * program, it should receive packets not depending to what interface 1943 * program is joined. 1944 * If we will not make it, the program will have to join on all 1945 * interfaces. On the other hand, multihoming host (or router, but 1946 * not mrouter) cannot join to more than one interface - it will 1947 * result in receiving multiple packets. 1948 */ 1949 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, 1950 net, NULL, skb, skb->dev, dev, 1951 ipmr_forward_finish); 1952 return; 1953 1954 out_free: 1955 kfree_skb(skb); 1956 } 1957 1958 /* Called with mrt_lock or rcu_read_lock() */ 1959 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev) 1960 { 1961 int ct; 1962 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */ 1963 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) { 1964 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev) 1965 break; 1966 } 1967 return ct; 1968 } 1969 1970 /* "local" means that we should preserve one skb (for local delivery) */ 1971 /* Called uner rcu_read_lock() */ 1972 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 1973 struct net_device *dev, struct sk_buff *skb, 1974 struct mfc_cache *c, int local) 1975 { 1976 int true_vifi = ipmr_find_vif(mrt, dev); 1977 int psend = -1; 1978 int vif, ct; 1979 1980 vif = c->_c.mfc_parent; 1981 atomic_long_inc(&c->_c.mfc_un.res.pkt); 1982 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes); 1983 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies); 1984 1985 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 1986 struct mfc_cache *cache_proxy; 1987 1988 /* For an (*,G) entry, we only check that the incoming 1989 * interface is part of the static tree. 1990 */ 1991 cache_proxy = mr_mfc_find_any_parent(mrt, vif); 1992 if (cache_proxy && 1993 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) 1994 goto forward; 1995 } 1996 1997 /* Wrong interface: drop packet and (maybe) send PIM assert. */ 1998 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) { 1999 if (rt_is_output_route(skb_rtable(skb))) { 2000 /* It is our own packet, looped back. 2001 * Very complicated situation... 2002 * 2003 * The best workaround until routing daemons will be 2004 * fixed is not to redistribute packet, if it was 2005 * send through wrong interface. It means, that 2006 * multicast applications WILL NOT work for 2007 * (S,G), which have default multicast route pointing 2008 * to wrong oif. In any case, it is not a good 2009 * idea to use multicasting applications on router. 2010 */ 2011 goto dont_forward; 2012 } 2013 2014 atomic_long_inc(&c->_c.mfc_un.res.wrong_if); 2015 2016 if (true_vifi >= 0 && mrt->mroute_do_assert && 2017 /* pimsm uses asserts, when switching from RPT to SPT, 2018 * so that we cannot check that packet arrived on an oif. 2019 * It is bad, but otherwise we would need to move pretty 2020 * large chunk of pimd to kernel. Ough... --ANK 2021 */ 2022 (mrt->mroute_do_pim || 2023 c->_c.mfc_un.res.ttls[true_vifi] < 255) && 2024 time_after(jiffies, 2025 c->_c.mfc_un.res.last_assert + 2026 MFC_ASSERT_THRESH)) { 2027 c->_c.mfc_un.res.last_assert = jiffies; 2028 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 2029 if (mrt->mroute_do_wrvifwhole) 2030 ipmr_cache_report(mrt, skb, true_vifi, 2031 IGMPMSG_WRVIFWHOLE); 2032 } 2033 goto dont_forward; 2034 } 2035 2036 forward: 2037 WRITE_ONCE(mrt->vif_table[vif].pkt_in, 2038 mrt->vif_table[vif].pkt_in + 1); 2039 WRITE_ONCE(mrt->vif_table[vif].bytes_in, 2040 mrt->vif_table[vif].bytes_in + skb->len); 2041 2042 /* Forward the frame */ 2043 if (c->mfc_origin == htonl(INADDR_ANY) && 2044 c->mfc_mcastgrp == htonl(INADDR_ANY)) { 2045 if (true_vifi >= 0 && 2046 true_vifi != c->_c.mfc_parent && 2047 ip_hdr(skb)->ttl > 2048 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) { 2049 /* It's an (*,*) entry and the packet is not coming from 2050 * the upstream: forward the packet to the upstream 2051 * only. 2052 */ 2053 psend = c->_c.mfc_parent; 2054 goto last_forward; 2055 } 2056 goto dont_forward; 2057 } 2058 for (ct = c->_c.mfc_un.res.maxvif - 1; 2059 ct >= c->_c.mfc_un.res.minvif; ct--) { 2060 /* For (*,G) entry, don't forward to the incoming interface */ 2061 if ((c->mfc_origin != htonl(INADDR_ANY) || 2062 ct != true_vifi) && 2063 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) { 2064 if (psend != -1) { 2065 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2066 2067 if (skb2) 2068 ipmr_queue_xmit(net, mrt, true_vifi, 2069 skb2, psend); 2070 } 2071 psend = ct; 2072 } 2073 } 2074 last_forward: 2075 if (psend != -1) { 2076 if (local) { 2077 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2078 2079 if (skb2) 2080 ipmr_queue_xmit(net, mrt, true_vifi, skb2, 2081 psend); 2082 } else { 2083 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend); 2084 return; 2085 } 2086 } 2087 2088 dont_forward: 2089 if (!local) 2090 kfree_skb(skb); 2091 } 2092 2093 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 2094 { 2095 struct rtable *rt = skb_rtable(skb); 2096 struct iphdr *iph = ip_hdr(skb); 2097 struct flowi4 fl4 = { 2098 .daddr = iph->daddr, 2099 .saddr = iph->saddr, 2100 .flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)), 2101 .flowi4_oif = (rt_is_output_route(rt) ? 2102 skb->dev->ifindex : 0), 2103 .flowi4_iif = (rt_is_output_route(rt) ? 2104 LOOPBACK_IFINDEX : 2105 skb->dev->ifindex), 2106 .flowi4_mark = skb->mark, 2107 }; 2108 struct mr_table *mrt; 2109 int err; 2110 2111 err = ipmr_fib_lookup(net, &fl4, &mrt); 2112 if (err) 2113 return ERR_PTR(err); 2114 return mrt; 2115 } 2116 2117 /* Multicast packets for forwarding arrive here 2118 * Called with rcu_read_lock(); 2119 */ 2120 int ip_mr_input(struct sk_buff *skb) 2121 { 2122 struct mfc_cache *cache; 2123 struct net *net = dev_net(skb->dev); 2124 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 2125 struct mr_table *mrt; 2126 struct net_device *dev; 2127 2128 /* skb->dev passed in is the loX master dev for vrfs. 2129 * As there are no vifs associated with loopback devices, 2130 * get the proper interface that does have a vif associated with it. 2131 */ 2132 dev = skb->dev; 2133 if (netif_is_l3_master(skb->dev)) { 2134 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 2135 if (!dev) { 2136 kfree_skb(skb); 2137 return -ENODEV; 2138 } 2139 } 2140 2141 /* Packet is looped back after forward, it should not be 2142 * forwarded second time, but still can be delivered locally. 2143 */ 2144 if (IPCB(skb)->flags & IPSKB_FORWARDED) 2145 goto dont_forward; 2146 2147 mrt = ipmr_rt_fib_lookup(net, skb); 2148 if (IS_ERR(mrt)) { 2149 kfree_skb(skb); 2150 return PTR_ERR(mrt); 2151 } 2152 if (!local) { 2153 if (IPCB(skb)->opt.router_alert) { 2154 if (ip_call_ra_chain(skb)) 2155 return 0; 2156 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 2157 /* IGMPv1 (and broken IGMPv2 implementations sort of 2158 * Cisco IOS <= 11.2(8)) do not put router alert 2159 * option to IGMP packets destined to routable 2160 * groups. It is very bad, because it means 2161 * that we can forward NO IGMP messages. 2162 */ 2163 struct sock *mroute_sk; 2164 2165 mroute_sk = rcu_dereference(mrt->mroute_sk); 2166 if (mroute_sk) { 2167 nf_reset_ct(skb); 2168 raw_rcv(mroute_sk, skb); 2169 return 0; 2170 } 2171 } 2172 } 2173 2174 /* already under rcu_read_lock() */ 2175 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 2176 if (!cache) { 2177 int vif = ipmr_find_vif(mrt, dev); 2178 2179 if (vif >= 0) 2180 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2181 vif); 2182 } 2183 2184 /* No usable cache entry */ 2185 if (!cache) { 2186 int vif; 2187 2188 if (local) { 2189 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2190 ip_local_deliver(skb); 2191 if (!skb2) 2192 return -ENOBUFS; 2193 skb = skb2; 2194 } 2195 2196 vif = ipmr_find_vif(mrt, dev); 2197 if (vif >= 0) 2198 return ipmr_cache_unresolved(mrt, vif, skb, dev); 2199 kfree_skb(skb); 2200 return -ENODEV; 2201 } 2202 2203 ip_mr_forward(net, mrt, dev, skb, cache, local); 2204 2205 if (local) 2206 return ip_local_deliver(skb); 2207 2208 return 0; 2209 2210 dont_forward: 2211 if (local) 2212 return ip_local_deliver(skb); 2213 kfree_skb(skb); 2214 return 0; 2215 } 2216 2217 #ifdef CONFIG_IP_PIMSM_V1 2218 /* Handle IGMP messages of PIMv1 */ 2219 int pim_rcv_v1(struct sk_buff *skb) 2220 { 2221 struct igmphdr *pim; 2222 struct net *net = dev_net(skb->dev); 2223 struct mr_table *mrt; 2224 2225 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2226 goto drop; 2227 2228 pim = igmp_hdr(skb); 2229 2230 mrt = ipmr_rt_fib_lookup(net, skb); 2231 if (IS_ERR(mrt)) 2232 goto drop; 2233 if (!mrt->mroute_do_pim || 2234 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2235 goto drop; 2236 2237 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2238 drop: 2239 kfree_skb(skb); 2240 } 2241 return 0; 2242 } 2243 #endif 2244 2245 #ifdef CONFIG_IP_PIMSM_V2 2246 static int pim_rcv(struct sk_buff *skb) 2247 { 2248 struct pimreghdr *pim; 2249 struct net *net = dev_net(skb->dev); 2250 struct mr_table *mrt; 2251 2252 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2253 goto drop; 2254 2255 pim = (struct pimreghdr *)skb_transport_header(skb); 2256 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) || 2257 (pim->flags & PIM_NULL_REGISTER) || 2258 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2259 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2260 goto drop; 2261 2262 mrt = ipmr_rt_fib_lookup(net, skb); 2263 if (IS_ERR(mrt)) 2264 goto drop; 2265 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2266 drop: 2267 kfree_skb(skb); 2268 } 2269 return 0; 2270 } 2271 #endif 2272 2273 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2274 __be32 saddr, __be32 daddr, 2275 struct rtmsg *rtm, u32 portid) 2276 { 2277 struct mfc_cache *cache; 2278 struct mr_table *mrt; 2279 int err; 2280 2281 rcu_read_lock(); 2282 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT); 2283 if (!mrt) { 2284 rcu_read_unlock(); 2285 return -ENOENT; 2286 } 2287 2288 cache = ipmr_cache_find(mrt, saddr, daddr); 2289 if (!cache && skb->dev) { 2290 int vif = ipmr_find_vif(mrt, skb->dev); 2291 2292 if (vif >= 0) 2293 cache = ipmr_cache_find_any(mrt, daddr, vif); 2294 } 2295 if (!cache) { 2296 struct sk_buff *skb2; 2297 struct iphdr *iph; 2298 struct net_device *dev; 2299 int vif = -1; 2300 2301 dev = skb->dev; 2302 if (dev) 2303 vif = ipmr_find_vif(mrt, dev); 2304 if (vif < 0) { 2305 rcu_read_unlock(); 2306 return -ENODEV; 2307 } 2308 2309 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr)); 2310 if (!skb2) { 2311 rcu_read_unlock(); 2312 return -ENOMEM; 2313 } 2314 2315 NETLINK_CB(skb2).portid = portid; 2316 skb_push(skb2, sizeof(struct iphdr)); 2317 skb_reset_network_header(skb2); 2318 iph = ip_hdr(skb2); 2319 iph->ihl = sizeof(struct iphdr) >> 2; 2320 iph->saddr = saddr; 2321 iph->daddr = daddr; 2322 iph->version = 0; 2323 err = ipmr_cache_unresolved(mrt, vif, skb2, dev); 2324 rcu_read_unlock(); 2325 return err; 2326 } 2327 2328 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm); 2329 rcu_read_unlock(); 2330 return err; 2331 } 2332 2333 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2334 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2335 int flags) 2336 { 2337 struct nlmsghdr *nlh; 2338 struct rtmsg *rtm; 2339 int err; 2340 2341 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2342 if (!nlh) 2343 return -EMSGSIZE; 2344 2345 rtm = nlmsg_data(nlh); 2346 rtm->rtm_family = RTNL_FAMILY_IPMR; 2347 rtm->rtm_dst_len = 32; 2348 rtm->rtm_src_len = 32; 2349 rtm->rtm_tos = 0; 2350 rtm->rtm_table = mrt->id; 2351 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2352 goto nla_put_failure; 2353 rtm->rtm_type = RTN_MULTICAST; 2354 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2355 if (c->_c.mfc_flags & MFC_STATIC) 2356 rtm->rtm_protocol = RTPROT_STATIC; 2357 else 2358 rtm->rtm_protocol = RTPROT_MROUTED; 2359 rtm->rtm_flags = 0; 2360 2361 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) || 2362 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp)) 2363 goto nla_put_failure; 2364 err = mr_fill_mroute(mrt, skb, &c->_c, rtm); 2365 /* do not break the dump if cache is unresolved */ 2366 if (err < 0 && err != -ENOENT) 2367 goto nla_put_failure; 2368 2369 nlmsg_end(skb, nlh); 2370 return 0; 2371 2372 nla_put_failure: 2373 nlmsg_cancel(skb, nlh); 2374 return -EMSGSIZE; 2375 } 2376 2377 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2378 u32 portid, u32 seq, struct mr_mfc *c, int cmd, 2379 int flags) 2380 { 2381 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c, 2382 cmd, flags); 2383 } 2384 2385 static size_t mroute_msgsize(bool unresolved, int maxvif) 2386 { 2387 size_t len = 2388 NLMSG_ALIGN(sizeof(struct rtmsg)) 2389 + nla_total_size(4) /* RTA_TABLE */ 2390 + nla_total_size(4) /* RTA_SRC */ 2391 + nla_total_size(4) /* RTA_DST */ 2392 ; 2393 2394 if (!unresolved) 2395 len = len 2396 + nla_total_size(4) /* RTA_IIF */ 2397 + nla_total_size(0) /* RTA_MULTIPATH */ 2398 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2399 /* RTA_MFC_STATS */ 2400 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2401 ; 2402 2403 return len; 2404 } 2405 2406 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2407 int cmd) 2408 { 2409 struct net *net = read_pnet(&mrt->net); 2410 struct sk_buff *skb; 2411 int err = -ENOBUFS; 2412 2413 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS, 2414 mrt->maxvif), 2415 GFP_ATOMIC); 2416 if (!skb) 2417 goto errout; 2418 2419 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2420 if (err < 0) 2421 goto errout; 2422 2423 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2424 return; 2425 2426 errout: 2427 kfree_skb(skb); 2428 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2429 } 2430 2431 static size_t igmpmsg_netlink_msgsize(size_t payloadlen) 2432 { 2433 size_t len = 2434 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2435 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */ 2436 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */ 2437 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */ 2438 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */ 2439 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */ 2440 /* IPMRA_CREPORT_PKT */ 2441 + nla_total_size(payloadlen) 2442 ; 2443 2444 return len; 2445 } 2446 2447 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt) 2448 { 2449 struct net *net = read_pnet(&mrt->net); 2450 struct nlmsghdr *nlh; 2451 struct rtgenmsg *rtgenm; 2452 struct igmpmsg *msg; 2453 struct sk_buff *skb; 2454 struct nlattr *nla; 2455 int payloadlen; 2456 2457 payloadlen = pkt->len - sizeof(struct igmpmsg); 2458 msg = (struct igmpmsg *)skb_network_header(pkt); 2459 2460 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2461 if (!skb) 2462 goto errout; 2463 2464 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2465 sizeof(struct rtgenmsg), 0); 2466 if (!nlh) 2467 goto errout; 2468 rtgenm = nlmsg_data(nlh); 2469 rtgenm->rtgen_family = RTNL_FAMILY_IPMR; 2470 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) || 2471 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) || 2472 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR, 2473 msg->im_src.s_addr) || 2474 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR, 2475 msg->im_dst.s_addr) || 2476 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id)) 2477 goto nla_put_failure; 2478 2479 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen); 2480 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg), 2481 nla_data(nla), payloadlen)) 2482 goto nla_put_failure; 2483 2484 nlmsg_end(skb, nlh); 2485 2486 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC); 2487 return; 2488 2489 nla_put_failure: 2490 nlmsg_cancel(skb, nlh); 2491 errout: 2492 kfree_skb(skb); 2493 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS); 2494 } 2495 2496 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb, 2497 const struct nlmsghdr *nlh, 2498 struct nlattr **tb, 2499 struct netlink_ext_ack *extack) 2500 { 2501 struct rtmsg *rtm; 2502 int i, err; 2503 2504 rtm = nlmsg_payload(nlh, sizeof(*rtm)); 2505 if (!rtm) { 2506 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request"); 2507 return -EINVAL; 2508 } 2509 2510 if (!netlink_strict_get_check(skb)) 2511 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, 2512 rtm_ipv4_policy, extack); 2513 2514 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) || 2515 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) || 2516 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol || 2517 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) { 2518 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request"); 2519 return -EINVAL; 2520 } 2521 2522 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, 2523 rtm_ipv4_policy, extack); 2524 if (err) 2525 return err; 2526 2527 if ((tb[RTA_SRC] && !rtm->rtm_src_len) || 2528 (tb[RTA_DST] && !rtm->rtm_dst_len)) { 2529 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4"); 2530 return -EINVAL; 2531 } 2532 2533 for (i = 0; i <= RTA_MAX; i++) { 2534 if (!tb[i]) 2535 continue; 2536 2537 switch (i) { 2538 case RTA_SRC: 2539 case RTA_DST: 2540 case RTA_TABLE: 2541 break; 2542 default: 2543 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request"); 2544 return -EINVAL; 2545 } 2546 } 2547 2548 return 0; 2549 } 2550 2551 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, 2552 struct netlink_ext_ack *extack) 2553 { 2554 struct net *net = sock_net(in_skb->sk); 2555 struct nlattr *tb[RTA_MAX + 1]; 2556 struct sk_buff *skb = NULL; 2557 struct mfc_cache *cache; 2558 struct mr_table *mrt; 2559 __be32 src, grp; 2560 u32 tableid; 2561 int err; 2562 2563 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack); 2564 if (err < 0) 2565 goto errout; 2566 2567 src = nla_get_in_addr_default(tb[RTA_SRC], 0); 2568 grp = nla_get_in_addr_default(tb[RTA_DST], 0); 2569 tableid = nla_get_u32_default(tb[RTA_TABLE], 0); 2570 2571 mrt = __ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT); 2572 if (!mrt) { 2573 err = -ENOENT; 2574 goto errout_free; 2575 } 2576 2577 /* entries are added/deleted only under RTNL */ 2578 rcu_read_lock(); 2579 cache = ipmr_cache_find(mrt, src, grp); 2580 rcu_read_unlock(); 2581 if (!cache) { 2582 err = -ENOENT; 2583 goto errout_free; 2584 } 2585 2586 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL); 2587 if (!skb) { 2588 err = -ENOBUFS; 2589 goto errout_free; 2590 } 2591 2592 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid, 2593 nlh->nlmsg_seq, cache, 2594 RTM_NEWROUTE, 0); 2595 if (err < 0) 2596 goto errout_free; 2597 2598 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 2599 2600 errout: 2601 return err; 2602 2603 errout_free: 2604 kfree_skb(skb); 2605 goto errout; 2606 } 2607 2608 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2609 { 2610 struct fib_dump_filter filter = { 2611 .rtnl_held = true, 2612 }; 2613 int err; 2614 2615 if (cb->strict_check) { 2616 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh, 2617 &filter, cb); 2618 if (err < 0) 2619 return err; 2620 } 2621 2622 if (filter.table_id) { 2623 struct mr_table *mrt; 2624 2625 mrt = __ipmr_get_table(sock_net(skb->sk), filter.table_id); 2626 if (!mrt) { 2627 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR) 2628 return skb->len; 2629 2630 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist"); 2631 return -ENOENT; 2632 } 2633 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute, 2634 &mfc_unres_lock, &filter); 2635 return skb->len ? : err; 2636 } 2637 2638 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter, 2639 _ipmr_fill_mroute, &mfc_unres_lock, &filter); 2640 } 2641 2642 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = { 2643 [RTA_SRC] = { .type = NLA_U32 }, 2644 [RTA_DST] = { .type = NLA_U32 }, 2645 [RTA_IIF] = { .type = NLA_U32 }, 2646 [RTA_TABLE] = { .type = NLA_U32 }, 2647 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 2648 }; 2649 2650 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol) 2651 { 2652 switch (rtm_protocol) { 2653 case RTPROT_STATIC: 2654 case RTPROT_MROUTED: 2655 return true; 2656 } 2657 return false; 2658 } 2659 2660 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc) 2661 { 2662 struct rtnexthop *rtnh = nla_data(nla); 2663 int remaining = nla_len(nla), vifi = 0; 2664 2665 while (rtnh_ok(rtnh, remaining)) { 2666 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops; 2667 if (++vifi == MAXVIFS) 2668 break; 2669 rtnh = rtnh_next(rtnh, &remaining); 2670 } 2671 2672 return remaining > 0 ? -EINVAL : vifi; 2673 } 2674 2675 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */ 2676 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh, 2677 struct mfcctl *mfcc, int *mrtsock, 2678 struct mr_table **mrtret, 2679 struct netlink_ext_ack *extack) 2680 { 2681 struct net_device *dev = NULL; 2682 u32 tblid = RT_TABLE_DEFAULT; 2683 struct mr_table *mrt; 2684 struct nlattr *attr; 2685 struct rtmsg *rtm; 2686 int ret, rem; 2687 2688 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX, 2689 rtm_ipmr_policy, extack); 2690 if (ret < 0) 2691 goto out; 2692 rtm = nlmsg_data(nlh); 2693 2694 ret = -EINVAL; 2695 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 || 2696 rtm->rtm_type != RTN_MULTICAST || 2697 rtm->rtm_scope != RT_SCOPE_UNIVERSE || 2698 !ipmr_rtm_validate_proto(rtm->rtm_protocol)) 2699 goto out; 2700 2701 memset(mfcc, 0, sizeof(*mfcc)); 2702 mfcc->mfcc_parent = -1; 2703 ret = 0; 2704 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) { 2705 switch (nla_type(attr)) { 2706 case RTA_SRC: 2707 mfcc->mfcc_origin.s_addr = nla_get_be32(attr); 2708 break; 2709 case RTA_DST: 2710 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr); 2711 break; 2712 case RTA_IIF: 2713 dev = __dev_get_by_index(net, nla_get_u32(attr)); 2714 if (!dev) { 2715 ret = -ENODEV; 2716 goto out; 2717 } 2718 break; 2719 case RTA_MULTIPATH: 2720 if (ipmr_nla_get_ttls(attr, mfcc) < 0) { 2721 ret = -EINVAL; 2722 goto out; 2723 } 2724 break; 2725 case RTA_PREFSRC: 2726 ret = 1; 2727 break; 2728 case RTA_TABLE: 2729 tblid = nla_get_u32(attr); 2730 break; 2731 } 2732 } 2733 mrt = __ipmr_get_table(net, tblid); 2734 if (!mrt) { 2735 ret = -ENOENT; 2736 goto out; 2737 } 2738 *mrtret = mrt; 2739 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0; 2740 if (dev) 2741 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev); 2742 2743 out: 2744 return ret; 2745 } 2746 2747 /* takes care of both newroute and delroute */ 2748 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh, 2749 struct netlink_ext_ack *extack) 2750 { 2751 struct net *net = sock_net(skb->sk); 2752 int ret, mrtsock, parent; 2753 struct mr_table *tbl; 2754 struct mfcctl mfcc; 2755 2756 mrtsock = 0; 2757 tbl = NULL; 2758 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack); 2759 if (ret < 0) 2760 return ret; 2761 2762 parent = ret ? mfcc.mfcc_parent : -1; 2763 if (nlh->nlmsg_type == RTM_NEWROUTE) 2764 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent); 2765 else 2766 return ipmr_mfc_delete(tbl, &mfcc, parent); 2767 } 2768 2769 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb) 2770 { 2771 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len); 2772 2773 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) || 2774 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) || 2775 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM, 2776 mrt->mroute_reg_vif_num) || 2777 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT, 2778 mrt->mroute_do_assert) || 2779 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) || 2780 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE, 2781 mrt->mroute_do_wrvifwhole)) 2782 return false; 2783 2784 return true; 2785 } 2786 2787 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb) 2788 { 2789 struct net_device *vif_dev; 2790 struct nlattr *vif_nest; 2791 struct vif_device *vif; 2792 2793 vif = &mrt->vif_table[vifid]; 2794 vif_dev = rtnl_dereference(vif->dev); 2795 /* if the VIF doesn't exist just continue */ 2796 if (!vif_dev) 2797 return true; 2798 2799 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF); 2800 if (!vif_nest) 2801 return false; 2802 2803 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) || 2804 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) || 2805 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) || 2806 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in, 2807 IPMRA_VIFA_PAD) || 2808 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out, 2809 IPMRA_VIFA_PAD) || 2810 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in, 2811 IPMRA_VIFA_PAD) || 2812 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out, 2813 IPMRA_VIFA_PAD) || 2814 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) || 2815 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) { 2816 nla_nest_cancel(skb, vif_nest); 2817 return false; 2818 } 2819 nla_nest_end(skb, vif_nest); 2820 2821 return true; 2822 } 2823 2824 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh, 2825 struct netlink_ext_ack *extack) 2826 { 2827 struct ifinfomsg *ifm; 2828 2829 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 2830 if (!ifm) { 2831 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump"); 2832 return -EINVAL; 2833 } 2834 2835 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 2836 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump"); 2837 return -EINVAL; 2838 } 2839 2840 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 2841 ifm->ifi_change || ifm->ifi_index) { 2842 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request"); 2843 return -EINVAL; 2844 } 2845 2846 return 0; 2847 } 2848 2849 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb) 2850 { 2851 struct net *net = sock_net(skb->sk); 2852 struct nlmsghdr *nlh = NULL; 2853 unsigned int t = 0, s_t; 2854 unsigned int e = 0, s_e; 2855 struct mr_table *mrt; 2856 2857 if (cb->strict_check) { 2858 int err = ipmr_valid_dumplink(cb->nlh, cb->extack); 2859 2860 if (err < 0) 2861 return err; 2862 } 2863 2864 s_t = cb->args[0]; 2865 s_e = cb->args[1]; 2866 2867 ipmr_for_each_table(mrt, net) { 2868 struct nlattr *vifs, *af; 2869 struct ifinfomsg *hdr; 2870 u32 i; 2871 2872 if (t < s_t) 2873 goto skip_table; 2874 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, 2875 cb->nlh->nlmsg_seq, RTM_NEWLINK, 2876 sizeof(*hdr), NLM_F_MULTI); 2877 if (!nlh) 2878 break; 2879 2880 hdr = nlmsg_data(nlh); 2881 memset(hdr, 0, sizeof(*hdr)); 2882 hdr->ifi_family = RTNL_FAMILY_IPMR; 2883 2884 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC); 2885 if (!af) { 2886 nlmsg_cancel(skb, nlh); 2887 goto out; 2888 } 2889 2890 if (!ipmr_fill_table(mrt, skb)) { 2891 nlmsg_cancel(skb, nlh); 2892 goto out; 2893 } 2894 2895 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS); 2896 if (!vifs) { 2897 nla_nest_end(skb, af); 2898 nlmsg_end(skb, nlh); 2899 goto out; 2900 } 2901 for (i = 0; i < mrt->maxvif; i++) { 2902 if (e < s_e) 2903 goto skip_entry; 2904 if (!ipmr_fill_vif(mrt, i, skb)) { 2905 nla_nest_end(skb, vifs); 2906 nla_nest_end(skb, af); 2907 nlmsg_end(skb, nlh); 2908 goto out; 2909 } 2910 skip_entry: 2911 e++; 2912 } 2913 s_e = 0; 2914 e = 0; 2915 nla_nest_end(skb, vifs); 2916 nla_nest_end(skb, af); 2917 nlmsg_end(skb, nlh); 2918 skip_table: 2919 t++; 2920 } 2921 2922 out: 2923 cb->args[1] = e; 2924 cb->args[0] = t; 2925 2926 return skb->len; 2927 } 2928 2929 #ifdef CONFIG_PROC_FS 2930 /* The /proc interfaces to multicast routing : 2931 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 2932 */ 2933 2934 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 2935 __acquires(RCU) 2936 { 2937 struct mr_vif_iter *iter = seq->private; 2938 struct net *net = seq_file_net(seq); 2939 struct mr_table *mrt; 2940 2941 rcu_read_lock(); 2942 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT); 2943 if (!mrt) { 2944 rcu_read_unlock(); 2945 return ERR_PTR(-ENOENT); 2946 } 2947 2948 iter->mrt = mrt; 2949 2950 return mr_vif_seq_start(seq, pos); 2951 } 2952 2953 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 2954 __releases(RCU) 2955 { 2956 rcu_read_unlock(); 2957 } 2958 2959 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 2960 { 2961 struct mr_vif_iter *iter = seq->private; 2962 struct mr_table *mrt = iter->mrt; 2963 2964 if (v == SEQ_START_TOKEN) { 2965 seq_puts(seq, 2966 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 2967 } else { 2968 const struct vif_device *vif = v; 2969 const struct net_device *vif_dev; 2970 const char *name; 2971 2972 vif_dev = vif_dev_read(vif); 2973 name = vif_dev ? vif_dev->name : "none"; 2974 seq_printf(seq, 2975 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 2976 vif - mrt->vif_table, 2977 name, vif->bytes_in, vif->pkt_in, 2978 vif->bytes_out, vif->pkt_out, 2979 vif->flags, vif->local, vif->remote); 2980 } 2981 return 0; 2982 } 2983 2984 static const struct seq_operations ipmr_vif_seq_ops = { 2985 .start = ipmr_vif_seq_start, 2986 .next = mr_vif_seq_next, 2987 .stop = ipmr_vif_seq_stop, 2988 .show = ipmr_vif_seq_show, 2989 }; 2990 2991 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 2992 { 2993 struct net *net = seq_file_net(seq); 2994 struct mr_table *mrt; 2995 2996 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2997 if (!mrt) 2998 return ERR_PTR(-ENOENT); 2999 3000 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock); 3001 } 3002 3003 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 3004 { 3005 int n; 3006 3007 if (v == SEQ_START_TOKEN) { 3008 seq_puts(seq, 3009 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 3010 } else { 3011 const struct mfc_cache *mfc = v; 3012 const struct mr_mfc_iter *it = seq->private; 3013 const struct mr_table *mrt = it->mrt; 3014 3015 seq_printf(seq, "%08X %08X %-3hd", 3016 (__force u32) mfc->mfc_mcastgrp, 3017 (__force u32) mfc->mfc_origin, 3018 mfc->_c.mfc_parent); 3019 3020 if (it->cache != &mrt->mfc_unres_queue) { 3021 seq_printf(seq, " %8lu %8lu %8lu", 3022 atomic_long_read(&mfc->_c.mfc_un.res.pkt), 3023 atomic_long_read(&mfc->_c.mfc_un.res.bytes), 3024 atomic_long_read(&mfc->_c.mfc_un.res.wrong_if)); 3025 for (n = mfc->_c.mfc_un.res.minvif; 3026 n < mfc->_c.mfc_un.res.maxvif; n++) { 3027 if (VIF_EXISTS(mrt, n) && 3028 mfc->_c.mfc_un.res.ttls[n] < 255) 3029 seq_printf(seq, 3030 " %2d:%-3d", 3031 n, mfc->_c.mfc_un.res.ttls[n]); 3032 } 3033 } else { 3034 /* unresolved mfc_caches don't contain 3035 * pkt, bytes and wrong_if values 3036 */ 3037 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 3038 } 3039 seq_putc(seq, '\n'); 3040 } 3041 return 0; 3042 } 3043 3044 static const struct seq_operations ipmr_mfc_seq_ops = { 3045 .start = ipmr_mfc_seq_start, 3046 .next = mr_mfc_seq_next, 3047 .stop = mr_mfc_seq_stop, 3048 .show = ipmr_mfc_seq_show, 3049 }; 3050 #endif 3051 3052 #ifdef CONFIG_IP_PIMSM_V2 3053 static const struct net_protocol pim_protocol = { 3054 .handler = pim_rcv, 3055 }; 3056 #endif 3057 3058 static unsigned int ipmr_seq_read(const struct net *net) 3059 { 3060 return READ_ONCE(net->ipv4.ipmr_seq) + ipmr_rules_seq_read(net); 3061 } 3062 3063 static int ipmr_dump(struct net *net, struct notifier_block *nb, 3064 struct netlink_ext_ack *extack) 3065 { 3066 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump, 3067 ipmr_mr_table_iter, extack); 3068 } 3069 3070 static const struct fib_notifier_ops ipmr_notifier_ops_template = { 3071 .family = RTNL_FAMILY_IPMR, 3072 .fib_seq_read = ipmr_seq_read, 3073 .fib_dump = ipmr_dump, 3074 .owner = THIS_MODULE, 3075 }; 3076 3077 static int __net_init ipmr_notifier_init(struct net *net) 3078 { 3079 struct fib_notifier_ops *ops; 3080 3081 net->ipv4.ipmr_seq = 0; 3082 3083 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net); 3084 if (IS_ERR(ops)) 3085 return PTR_ERR(ops); 3086 net->ipv4.ipmr_notifier_ops = ops; 3087 3088 return 0; 3089 } 3090 3091 static void __net_exit ipmr_notifier_exit(struct net *net) 3092 { 3093 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops); 3094 net->ipv4.ipmr_notifier_ops = NULL; 3095 } 3096 3097 /* Setup for IP multicast routing */ 3098 static int __net_init ipmr_net_init(struct net *net) 3099 { 3100 int err; 3101 3102 err = ipmr_notifier_init(net); 3103 if (err) 3104 goto ipmr_notifier_fail; 3105 3106 err = ipmr_rules_init(net); 3107 if (err < 0) 3108 goto ipmr_rules_fail; 3109 3110 #ifdef CONFIG_PROC_FS 3111 err = -ENOMEM; 3112 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops, 3113 sizeof(struct mr_vif_iter))) 3114 goto proc_vif_fail; 3115 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops, 3116 sizeof(struct mr_mfc_iter))) 3117 goto proc_cache_fail; 3118 #endif 3119 return 0; 3120 3121 #ifdef CONFIG_PROC_FS 3122 proc_cache_fail: 3123 remove_proc_entry("ip_mr_vif", net->proc_net); 3124 proc_vif_fail: 3125 rtnl_lock(); 3126 ipmr_rules_exit(net); 3127 rtnl_unlock(); 3128 #endif 3129 ipmr_rules_fail: 3130 ipmr_notifier_exit(net); 3131 ipmr_notifier_fail: 3132 return err; 3133 } 3134 3135 static void __net_exit ipmr_net_exit(struct net *net) 3136 { 3137 #ifdef CONFIG_PROC_FS 3138 remove_proc_entry("ip_mr_cache", net->proc_net); 3139 remove_proc_entry("ip_mr_vif", net->proc_net); 3140 #endif 3141 ipmr_notifier_exit(net); 3142 } 3143 3144 static void __net_exit ipmr_net_exit_batch(struct list_head *net_list) 3145 { 3146 struct net *net; 3147 3148 rtnl_lock(); 3149 list_for_each_entry(net, net_list, exit_list) 3150 ipmr_rules_exit(net); 3151 rtnl_unlock(); 3152 } 3153 3154 static struct pernet_operations ipmr_net_ops = { 3155 .init = ipmr_net_init, 3156 .exit = ipmr_net_exit, 3157 .exit_batch = ipmr_net_exit_batch, 3158 }; 3159 3160 static const struct rtnl_msg_handler ipmr_rtnl_msg_handlers[] __initconst = { 3161 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETLINK, 3162 .dumpit = ipmr_rtm_dumplink}, 3163 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_NEWROUTE, 3164 .doit = ipmr_rtm_route}, 3165 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_DELROUTE, 3166 .doit = ipmr_rtm_route}, 3167 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETROUTE, 3168 .doit = ipmr_rtm_getroute, .dumpit = ipmr_rtm_dumproute}, 3169 }; 3170 3171 int __init ip_mr_init(void) 3172 { 3173 int err; 3174 3175 mrt_cachep = KMEM_CACHE(mfc_cache, SLAB_HWCACHE_ALIGN | SLAB_PANIC); 3176 3177 err = register_pernet_subsys(&ipmr_net_ops); 3178 if (err) 3179 goto reg_pernet_fail; 3180 3181 err = register_netdevice_notifier(&ip_mr_notifier); 3182 if (err) 3183 goto reg_notif_fail; 3184 #ifdef CONFIG_IP_PIMSM_V2 3185 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 3186 pr_err("%s: can't add PIM protocol\n", __func__); 3187 err = -EAGAIN; 3188 goto add_proto_fail; 3189 } 3190 #endif 3191 rtnl_register_many(ipmr_rtnl_msg_handlers); 3192 3193 return 0; 3194 3195 #ifdef CONFIG_IP_PIMSM_V2 3196 add_proto_fail: 3197 unregister_netdevice_notifier(&ip_mr_notifier); 3198 #endif 3199 reg_notif_fail: 3200 unregister_pernet_subsys(&ipmr_net_ops); 3201 reg_pernet_fail: 3202 kmem_cache_destroy(mrt_cachep); 3203 return err; 3204 } 3205