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