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/flow.h> 46 #include <net/net_namespace.h> 47 #include <net/ip.h> 48 #include <net/protocol.h> 49 #include <linux/skbuff.h> 50 #include <net/route.h> 51 #include <net/icmp.h> 52 #include <net/udp.h> 53 #include <net/raw.h> 54 #include <linux/notifier.h> 55 #include <linux/if_arp.h> 56 #include <linux/netfilter_ipv4.h> 57 #include <linux/compat.h> 58 #include <linux/export.h> 59 #include <linux/rhashtable.h> 60 #include <net/ip_tunnels.h> 61 #include <net/checksum.h> 62 #include <net/netlink.h> 63 #include <net/fib_rules.h> 64 #include <linux/netconf.h> 65 #include <net/rtnh.h> 66 #include <net/inet_dscp.h> 67 68 #include <linux/nospec.h> 69 70 struct ipmr_rule { 71 struct fib_rule common; 72 }; 73 74 struct ipmr_result { 75 struct mr_table *mrt; 76 }; 77 78 /* Big lock, protecting vif table, mrt cache and mroute socket state. 79 * Note that the changes are semaphored via rtnl_lock. 80 */ 81 82 static DEFINE_SPINLOCK(mrt_lock); 83 84 static struct net_device *vif_dev_read(const struct vif_device *vif) 85 { 86 return rcu_dereference(vif->dev); 87 } 88 89 /* Multicast router control variables */ 90 91 /* Special spinlock for queue of unresolved entries */ 92 static DEFINE_SPINLOCK(mfc_unres_lock); 93 94 /* We return to original Alan's scheme. Hash table of resolved 95 * entries is changed only in process context and protected 96 * with weak lock mrt_lock. Queue of unresolved entries is protected 97 * with strong spinlock mfc_unres_lock. 98 * 99 * In this case data path is free of exclusive locks at all. 100 */ 101 102 static struct kmem_cache *mrt_cachep __ro_after_init; 103 104 static struct mr_table *ipmr_new_table(struct net *net, u32 id); 105 static void ipmr_free_table(struct mr_table *mrt, 106 struct list_head *dev_kill_list); 107 108 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 109 struct net_device *dev, struct sk_buff *skb, 110 struct mfc_cache *cache, int local); 111 static int ipmr_cache_report(const struct mr_table *mrt, 112 struct sk_buff *pkt, vifi_t vifi, int assert); 113 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 114 int cmd); 115 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt); 116 static void mroute_clean_tables(struct mr_table *mrt, int flags, 117 struct list_head *dev_kill_list); 118 static void ipmr_expire_process(struct timer_list *t); 119 120 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 121 #define ipmr_for_each_table(mrt, net) \ 122 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \ 123 lockdep_rtnl_is_held() || \ 124 list_empty(&net->ipv4.mr_tables)) 125 126 static struct mr_table *ipmr_mr_table_iter(struct net *net, 127 struct mr_table *mrt) 128 { 129 struct mr_table *ret; 130 131 if (!mrt) 132 ret = list_entry_rcu(net->ipv4.mr_tables.next, 133 struct mr_table, list); 134 else 135 ret = list_entry_rcu(mrt->list.next, 136 struct mr_table, list); 137 138 if (&ret->list == &net->ipv4.mr_tables) 139 return NULL; 140 return ret; 141 } 142 143 static struct mr_table *__ipmr_get_table(struct net *net, u32 id) 144 { 145 struct mr_table *mrt; 146 147 ipmr_for_each_table(mrt, net) { 148 if (mrt->id == id) 149 return mrt; 150 } 151 return NULL; 152 } 153 154 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 155 struct mr_table **mrt) 156 { 157 int err; 158 struct ipmr_result res; 159 struct fib_lookup_arg arg = { 160 .result = &res, 161 .flags = FIB_LOOKUP_NOREF, 162 }; 163 164 /* update flow if oif or iif point to device enslaved to l3mdev */ 165 l3mdev_update_flow(net, flowi4_to_flowi(flp4)); 166 167 err = fib_rules_lookup(net->ipv4.mr_rules_ops, 168 flowi4_to_flowi(flp4), 0, &arg); 169 if (err < 0) 170 return err; 171 *mrt = res.mrt; 172 return 0; 173 } 174 175 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp, 176 int flags, struct fib_lookup_arg *arg) 177 { 178 struct ipmr_result *res = arg->result; 179 struct mr_table *mrt; 180 181 switch (rule->action) { 182 case FR_ACT_TO_TBL: 183 break; 184 case FR_ACT_UNREACHABLE: 185 return -ENETUNREACH; 186 case FR_ACT_PROHIBIT: 187 return -EACCES; 188 case FR_ACT_BLACKHOLE: 189 default: 190 return -EINVAL; 191 } 192 193 arg->table = fib_rule_get_table(rule, arg); 194 195 mrt = __ipmr_get_table(rule->fr_net, arg->table); 196 if (!mrt) 197 return -EAGAIN; 198 res->mrt = mrt; 199 return 0; 200 } 201 202 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 203 { 204 return 1; 205 } 206 207 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 208 struct fib_rule_hdr *frh, struct nlattr **tb, 209 struct netlink_ext_ack *extack) 210 { 211 return 0; 212 } 213 214 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 215 struct nlattr **tb) 216 { 217 return 1; 218 } 219 220 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 221 struct fib_rule_hdr *frh) 222 { 223 frh->dst_len = 0; 224 frh->src_len = 0; 225 frh->tos = 0; 226 return 0; 227 } 228 229 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = { 230 .family = RTNL_FAMILY_IPMR, 231 .rule_size = sizeof(struct ipmr_rule), 232 .addr_size = sizeof(u32), 233 .action = ipmr_rule_action, 234 .match = ipmr_rule_match, 235 .configure = ipmr_rule_configure, 236 .compare = ipmr_rule_compare, 237 .fill = ipmr_rule_fill, 238 .nlgroup = RTNLGRP_IPV4_RULE, 239 .owner = THIS_MODULE, 240 }; 241 242 static int __net_init ipmr_rules_init(struct net *net) 243 { 244 struct fib_rules_ops *ops; 245 LIST_HEAD(dev_kill_list); 246 struct mr_table *mrt; 247 int err; 248 249 ops = fib_rules_register(&ipmr_rules_ops_template, net); 250 if (IS_ERR(ops)) 251 return PTR_ERR(ops); 252 253 INIT_LIST_HEAD(&net->ipv4.mr_tables); 254 255 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 256 if (IS_ERR(mrt)) { 257 err = PTR_ERR(mrt); 258 goto err1; 259 } 260 261 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT); 262 if (err < 0) 263 goto err2; 264 265 net->ipv4.mr_rules_ops = ops; 266 return 0; 267 268 err2: 269 ipmr_free_table(mrt, &dev_kill_list); 270 err1: 271 fib_rules_unregister(ops); 272 return err; 273 } 274 275 static void __net_exit ipmr_rules_exit(struct net *net) 276 { 277 fib_rules_unregister(net->ipv4.mr_rules_ops); 278 } 279 280 static void __net_exit ipmr_rules_exit_rtnl(struct net *net, 281 struct list_head *dev_kill_list) 282 { 283 struct mr_table *mrt, *next; 284 285 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) { 286 list_del_rcu(&mrt->list); 287 ipmr_free_table(mrt, dev_kill_list); 288 } 289 } 290 291 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb, 292 struct netlink_ext_ack *extack) 293 { 294 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack); 295 } 296 297 static unsigned int ipmr_rules_seq_read(const struct net *net) 298 { 299 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR); 300 } 301 302 bool ipmr_rule_default(const struct fib_rule *rule) 303 { 304 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT; 305 } 306 EXPORT_SYMBOL(ipmr_rule_default); 307 #else 308 static struct mr_table *ipmr_mr_table_iter(struct net *net, 309 struct mr_table *mrt) 310 { 311 if (!mrt) 312 return rcu_dereference(net->ipv4.mrt); 313 return NULL; 314 } 315 316 static struct mr_table *__ipmr_get_table(struct net *net, u32 id) 317 { 318 return rcu_dereference_check(net->ipv4.mrt, 319 lockdep_rtnl_is_held() || 320 !rcu_access_pointer(net->ipv4.mrt)); 321 } 322 323 #define ipmr_for_each_table(mrt, net) \ 324 for (mrt = __ipmr_get_table(net, 0); mrt; mrt = NULL) 325 326 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 327 struct mr_table **mrt) 328 { 329 *mrt = rcu_dereference(net->ipv4.mrt); 330 if (!*mrt) 331 return -EAGAIN; 332 return 0; 333 } 334 335 static int __net_init ipmr_rules_init(struct net *net) 336 { 337 struct mr_table *mrt; 338 339 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 340 if (IS_ERR(mrt)) 341 return PTR_ERR(mrt); 342 343 rcu_assign_pointer(net->ipv4.mrt, mrt); 344 return 0; 345 } 346 347 static void __net_exit ipmr_rules_exit(struct net *net) 348 { 349 } 350 351 static void __net_exit ipmr_rules_exit_rtnl(struct net *net, 352 struct list_head *dev_kill_list) 353 { 354 struct mr_table *mrt = rcu_dereference_protected(net->ipv4.mrt, 1); 355 356 RCU_INIT_POINTER(net->ipv4.mrt, NULL); 357 ipmr_free_table(mrt, dev_kill_list); 358 } 359 360 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb, 361 struct netlink_ext_ack *extack) 362 { 363 return 0; 364 } 365 366 static unsigned int ipmr_rules_seq_read(const struct net *net) 367 { 368 return 0; 369 } 370 371 bool ipmr_rule_default(const struct fib_rule *rule) 372 { 373 return true; 374 } 375 EXPORT_SYMBOL(ipmr_rule_default); 376 #endif 377 378 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 379 { 380 struct mr_table *mrt; 381 382 rcu_read_lock(); 383 mrt = __ipmr_get_table(net, id); 384 rcu_read_unlock(); 385 386 return mrt; 387 } 388 389 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg, 390 const void *ptr) 391 { 392 const struct mfc_cache_cmp_arg *cmparg = arg->key; 393 const struct mfc_cache *c = ptr; 394 395 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp || 396 cmparg->mfc_origin != c->mfc_origin; 397 } 398 399 static const struct rhashtable_params ipmr_rht_params = { 400 .head_offset = offsetof(struct mr_mfc, mnode), 401 .key_offset = offsetof(struct mfc_cache, cmparg), 402 .key_len = sizeof(struct mfc_cache_cmp_arg), 403 .nelem_hint = 3, 404 .obj_cmpfn = ipmr_hash_cmp, 405 .automatic_shrinking = true, 406 }; 407 408 static void ipmr_new_table_set(struct mr_table *mrt, 409 struct net *net) 410 { 411 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 412 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables); 413 #endif 414 } 415 416 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = { 417 .mfc_mcastgrp = htonl(INADDR_ANY), 418 .mfc_origin = htonl(INADDR_ANY), 419 }; 420 421 static struct mr_table_ops ipmr_mr_table_ops = { 422 .rht_params = &ipmr_rht_params, 423 .cmparg_any = &ipmr_mr_table_ops_cmparg_any, 424 }; 425 426 static struct mr_table *ipmr_new_table(struct net *net, u32 id) 427 { 428 struct mr_table *mrt; 429 430 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */ 431 if (id != RT_TABLE_DEFAULT && id >= 1000000000) 432 return ERR_PTR(-EINVAL); 433 434 mrt = __ipmr_get_table(net, id); 435 if (mrt) 436 return mrt; 437 438 return mr_table_alloc(net, id, &ipmr_mr_table_ops, 439 ipmr_expire_process, ipmr_new_table_set); 440 } 441 442 static void ipmr_free_table(struct mr_table *mrt, struct list_head *dev_kill_list) 443 { 444 struct net *net = read_pnet(&mrt->net); 445 LIST_HEAD(ipmr_dev_kill_list); 446 447 WARN_ON_ONCE(!mr_can_free_table(net)); 448 449 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC | 450 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC, 451 &ipmr_dev_kill_list); 452 timer_shutdown_sync(&mrt->ipmr_expire_timer); 453 mr_table_free(mrt); 454 455 WARN_ON_ONCE(!net_initialized(net) && !list_empty(&ipmr_dev_kill_list)); 456 list_splice(&ipmr_dev_kill_list, dev_kill_list); 457 } 458 459 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ 460 461 /* Initialize ipmr pimreg/tunnel in_device */ 462 static bool ipmr_init_vif_indev(const struct net_device *dev) 463 { 464 struct in_device *in_dev; 465 466 ASSERT_RTNL(); 467 468 in_dev = __in_dev_get_rtnl(dev); 469 if (!in_dev) 470 return false; 471 ipv4_devconf_setall(in_dev); 472 neigh_parms_data_state_setall(in_dev->arp_parms); 473 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 474 475 return true; 476 } 477 478 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v) 479 { 480 struct net_device *tunnel_dev, *new_dev; 481 struct ip_tunnel_parm_kern p = { }; 482 int err; 483 484 tunnel_dev = __dev_get_by_name(net, "tunl0"); 485 if (!tunnel_dev) 486 goto out; 487 488 p.iph.daddr = v->vifc_rmt_addr.s_addr; 489 p.iph.saddr = v->vifc_lcl_addr.s_addr; 490 p.iph.version = 4; 491 p.iph.ihl = 5; 492 p.iph.protocol = IPPROTO_IPIP; 493 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 494 495 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl) 496 goto out; 497 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p, 498 SIOCADDTUNNEL); 499 if (err) 500 goto out; 501 502 new_dev = __dev_get_by_name(net, p.name); 503 if (!new_dev) 504 goto out; 505 506 new_dev->flags |= IFF_MULTICAST; 507 if (!ipmr_init_vif_indev(new_dev)) 508 goto out_unregister; 509 if (dev_open(new_dev, NULL)) 510 goto out_unregister; 511 dev_hold(new_dev); 512 err = dev_set_allmulti(new_dev, 1); 513 if (err) { 514 dev_close(new_dev); 515 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p, 516 SIOCDELTUNNEL); 517 dev_put(new_dev); 518 new_dev = ERR_PTR(err); 519 } 520 return new_dev; 521 522 out_unregister: 523 unregister_netdevice(new_dev); 524 out: 525 return ERR_PTR(-ENOBUFS); 526 } 527 528 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 529 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) 530 { 531 struct net *net = dev_net(dev); 532 struct mr_table *mrt; 533 struct flowi4 fl4 = { 534 .flowi4_oif = dev->ifindex, 535 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 536 .flowi4_mark = skb->mark, 537 }; 538 int err; 539 540 err = ipmr_fib_lookup(net, &fl4, &mrt); 541 if (err < 0) { 542 kfree_skb(skb); 543 return err; 544 } 545 546 DEV_STATS_ADD(dev, tx_bytes, skb->len); 547 DEV_STATS_INC(dev, tx_packets); 548 rcu_read_lock(); 549 550 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */ 551 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num), 552 IGMPMSG_WHOLEPKT); 553 554 rcu_read_unlock(); 555 kfree_skb(skb); 556 return NETDEV_TX_OK; 557 } 558 559 static int reg_vif_get_iflink(const struct net_device *dev) 560 { 561 return 0; 562 } 563 564 static const struct net_device_ops reg_vif_netdev_ops = { 565 .ndo_start_xmit = reg_vif_xmit, 566 .ndo_get_iflink = reg_vif_get_iflink, 567 }; 568 569 static void reg_vif_setup(struct net_device *dev) 570 { 571 dev->type = ARPHRD_PIMREG; 572 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; 573 dev->flags = IFF_NOARP; 574 dev->netdev_ops = ®_vif_netdev_ops; 575 dev->needs_free_netdev = true; 576 dev->netns_immutable = true; 577 } 578 579 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 580 { 581 struct net_device *dev; 582 char name[IFNAMSIZ]; 583 584 if (mrt->id == RT_TABLE_DEFAULT) 585 sprintf(name, "pimreg"); 586 else 587 sprintf(name, "pimreg%u", mrt->id); 588 589 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 590 591 if (!dev) 592 return NULL; 593 594 dev_net_set(dev, net); 595 596 if (register_netdevice(dev)) { 597 free_netdev(dev); 598 return NULL; 599 } 600 601 if (!ipmr_init_vif_indev(dev)) 602 goto failure; 603 if (dev_open(dev, NULL)) 604 goto failure; 605 606 dev_hold(dev); 607 608 return dev; 609 610 failure: 611 unregister_netdevice(dev); 612 return NULL; 613 } 614 615 /* called with rcu_read_lock() */ 616 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb, 617 unsigned int pimlen) 618 { 619 struct net_device *reg_dev = NULL; 620 struct iphdr *encap; 621 int vif_num; 622 623 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); 624 /* Check that: 625 * a. packet is really sent to a multicast group 626 * b. packet is not a NULL-REGISTER 627 * c. packet is not truncated 628 */ 629 if (!ipv4_is_multicast(encap->daddr) || 630 encap->tot_len == 0 || 631 ntohs(encap->tot_len) + pimlen > skb->len) 632 return 1; 633 634 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */ 635 vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 636 if (vif_num >= 0) 637 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]); 638 if (!reg_dev) 639 return 1; 640 641 skb->mac_header = skb->network_header; 642 skb_pull(skb, (u8 *)encap - skb->data); 643 skb_reset_network_header(skb); 644 skb->protocol = htons(ETH_P_IP); 645 skb->ip_summed = CHECKSUM_NONE; 646 647 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 648 649 netif_rx(skb); 650 651 return NET_RX_SUCCESS; 652 } 653 #else 654 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 655 { 656 return NULL; 657 } 658 #endif 659 660 static int call_ipmr_vif_entry_notifiers(struct net *net, 661 enum fib_event_type event_type, 662 struct vif_device *vif, 663 struct net_device *vif_dev, 664 vifi_t vif_index, u32 tb_id) 665 { 666 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type, 667 vif, vif_dev, vif_index, tb_id, 668 &net->ipv4.ipmr_seq); 669 } 670 671 static int call_ipmr_mfc_entry_notifiers(struct net *net, 672 enum fib_event_type event_type, 673 struct mfc_cache *mfc, u32 tb_id) 674 { 675 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type, 676 &mfc->_c, tb_id, &net->ipv4.ipmr_seq); 677 } 678 679 /** 680 * vif_delete - Delete a VIF entry 681 * @mrt: Table to delete from 682 * @vifi: VIF identifier to delete 683 * @notify: Set to 1, if the caller is a notifier_call 684 * @head: if unregistering the VIF, place it on this queue 685 */ 686 static int vif_delete(struct mr_table *mrt, int vifi, int notify, 687 struct list_head *head) 688 { 689 struct net *net = read_pnet(&mrt->net); 690 struct vif_device *v; 691 struct net_device *dev; 692 struct in_device *in_dev; 693 694 if (vifi < 0 || vifi >= mrt->maxvif) 695 return -EADDRNOTAVAIL; 696 697 v = &mrt->vif_table[vifi]; 698 699 dev = rtnl_dereference(v->dev); 700 if (!dev) 701 return -EADDRNOTAVAIL; 702 703 spin_lock(&mrt_lock); 704 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev, 705 vifi, mrt->id); 706 RCU_INIT_POINTER(v->dev, NULL); 707 708 if (vifi == mrt->mroute_reg_vif_num) { 709 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */ 710 WRITE_ONCE(mrt->mroute_reg_vif_num, -1); 711 } 712 if (vifi + 1 == mrt->maxvif) { 713 int tmp; 714 715 for (tmp = vifi - 1; tmp >= 0; tmp--) { 716 if (VIF_EXISTS(mrt, tmp)) 717 break; 718 } 719 WRITE_ONCE(mrt->maxvif, tmp + 1); 720 } 721 722 spin_unlock(&mrt_lock); 723 724 dev_set_allmulti(dev, -1); 725 726 in_dev = __in_dev_get_rtnl(dev); 727 if (in_dev) { 728 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; 729 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 730 NETCONFA_MC_FORWARDING, 731 dev->ifindex, &in_dev->cnf); 732 ip_rt_multicast_event(in_dev); 733 } 734 735 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify) 736 unregister_netdevice_queue(dev, head); 737 738 netdev_put(dev, &v->dev_tracker); 739 return 0; 740 } 741 742 static void ipmr_cache_free_rcu(struct rcu_head *head) 743 { 744 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu); 745 746 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c); 747 } 748 749 static void ipmr_cache_free(struct mfc_cache *c) 750 { 751 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu); 752 } 753 754 /* Destroy an unresolved cache entry, killing queued skbs 755 * and reporting error to netlink readers. 756 */ 757 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c) 758 { 759 struct net *net = read_pnet(&mrt->net); 760 struct sk_buff *skb; 761 struct nlmsgerr *e; 762 763 atomic_dec(&mrt->cache_resolve_queue_len); 764 765 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) { 766 if (ip_hdr(skb)->version == 0) { 767 struct nlmsghdr *nlh = skb_pull(skb, 768 sizeof(struct iphdr)); 769 nlh->nlmsg_type = NLMSG_ERROR; 770 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 771 skb_trim(skb, nlh->nlmsg_len); 772 e = nlmsg_data(nlh); 773 e->error = -ETIMEDOUT; 774 memset(&e->msg, 0, sizeof(e->msg)); 775 776 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 777 } else { 778 kfree_skb(skb); 779 } 780 } 781 782 ipmr_cache_free(c); 783 } 784 785 /* Timer process for the unresolved queue. */ 786 static void ipmr_expire_process(struct timer_list *t) 787 { 788 struct mr_table *mrt = timer_container_of(mrt, t, ipmr_expire_timer); 789 struct mr_mfc *c, *next; 790 unsigned long expires; 791 unsigned long now; 792 793 if (!spin_trylock(&mfc_unres_lock)) { 794 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 795 return; 796 } 797 798 if (list_empty(&mrt->mfc_unres_queue)) 799 goto out; 800 801 now = jiffies; 802 expires = 10*HZ; 803 804 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 805 if (time_after(c->mfc_un.unres.expires, now)) { 806 unsigned long interval = c->mfc_un.unres.expires - now; 807 if (interval < expires) 808 expires = interval; 809 continue; 810 } 811 812 list_del(&c->list); 813 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE); 814 ipmr_destroy_unres(mrt, (struct mfc_cache *)c); 815 } 816 817 if (!list_empty(&mrt->mfc_unres_queue)) 818 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 819 820 out: 821 spin_unlock(&mfc_unres_lock); 822 } 823 824 /* Fill oifs list. It is called under locked mrt_lock. */ 825 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache, 826 unsigned char *ttls) 827 { 828 int vifi; 829 830 cache->mfc_un.res.minvif = MAXVIFS; 831 cache->mfc_un.res.maxvif = 0; 832 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 833 834 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 835 if (VIF_EXISTS(mrt, vifi) && 836 ttls[vifi] && ttls[vifi] < 255) { 837 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 838 if (cache->mfc_un.res.minvif > vifi) 839 cache->mfc_un.res.minvif = vifi; 840 if (cache->mfc_un.res.maxvif <= vifi) 841 cache->mfc_un.res.maxvif = vifi + 1; 842 } 843 } 844 WRITE_ONCE(cache->mfc_un.res.lastuse, jiffies); 845 } 846 847 static int vif_add(struct net *net, struct mr_table *mrt, 848 struct vifctl *vifc, int mrtsock) 849 { 850 struct netdev_phys_item_id ppid = { }; 851 int vifi = vifc->vifc_vifi; 852 struct vif_device *v = &mrt->vif_table[vifi]; 853 struct net_device *dev; 854 struct in_device *in_dev; 855 int err; 856 857 /* Is vif busy ? */ 858 if (VIF_EXISTS(mrt, vifi)) 859 return -EADDRINUSE; 860 861 switch (vifc->vifc_flags) { 862 case VIFF_REGISTER: 863 if (!ipmr_pimsm_enabled()) 864 return -EINVAL; 865 /* Special Purpose VIF in PIM 866 * All the packets will be sent to the daemon 867 */ 868 if (mrt->mroute_reg_vif_num >= 0) 869 return -EADDRINUSE; 870 dev = ipmr_reg_vif(net, mrt); 871 if (!dev) 872 return -ENOBUFS; 873 err = dev_set_allmulti(dev, 1); 874 if (err) { 875 unregister_netdevice(dev); 876 dev_put(dev); 877 return err; 878 } 879 break; 880 case VIFF_TUNNEL: 881 dev = ipmr_new_tunnel(net, vifc); 882 if (IS_ERR(dev)) 883 return PTR_ERR(dev); 884 break; 885 case VIFF_USE_IFINDEX: 886 case 0: 887 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 888 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 889 if (dev && !__in_dev_get_rtnl(dev)) { 890 dev_put(dev); 891 return -EADDRNOTAVAIL; 892 } 893 } else { 894 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 895 } 896 if (!dev) 897 return -EADDRNOTAVAIL; 898 err = dev_set_allmulti(dev, 1); 899 if (err) { 900 dev_put(dev); 901 return err; 902 } 903 break; 904 default: 905 return -EINVAL; 906 } 907 908 in_dev = __in_dev_get_rtnl(dev); 909 if (!in_dev) { 910 dev_put(dev); 911 return -EADDRNOTAVAIL; 912 } 913 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 914 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING, 915 dev->ifindex, &in_dev->cnf); 916 ip_rt_multicast_event(in_dev); 917 918 /* Fill in the VIF structures */ 919 vif_device_init(v, dev, vifc->vifc_rate_limit, 920 vifc->vifc_threshold, 921 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0), 922 (VIFF_TUNNEL | VIFF_REGISTER)); 923 924 err = netif_get_port_parent_id(dev, &ppid, true); 925 if (err == 0) { 926 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len); 927 v->dev_parent_id.id_len = ppid.id_len; 928 } else { 929 v->dev_parent_id.id_len = 0; 930 } 931 932 v->local = vifc->vifc_lcl_addr.s_addr; 933 v->remote = vifc->vifc_rmt_addr.s_addr; 934 935 /* And finish update writing critical data */ 936 spin_lock(&mrt_lock); 937 rcu_assign_pointer(v->dev, dev); 938 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC); 939 if (v->flags & VIFF_REGISTER) { 940 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */ 941 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi); 942 } 943 if (vifi+1 > mrt->maxvif) 944 WRITE_ONCE(mrt->maxvif, vifi + 1); 945 spin_unlock(&mrt_lock); 946 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev, 947 vifi, mrt->id); 948 return 0; 949 } 950 951 /* called with rcu_read_lock() */ 952 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 953 __be32 origin, 954 __be32 mcastgrp) 955 { 956 struct mfc_cache_cmp_arg arg = { 957 .mfc_mcastgrp = mcastgrp, 958 .mfc_origin = origin 959 }; 960 961 return mr_mfc_find(mrt, &arg); 962 } 963 964 /* Look for a (*,G) entry */ 965 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 966 __be32 mcastgrp, int vifi) 967 { 968 struct mfc_cache_cmp_arg arg = { 969 .mfc_mcastgrp = mcastgrp, 970 .mfc_origin = htonl(INADDR_ANY) 971 }; 972 973 if (mcastgrp == htonl(INADDR_ANY)) 974 return mr_mfc_find_any_parent(mrt, vifi); 975 return mr_mfc_find_any(mrt, vifi, &arg); 976 } 977 978 /* Look for a (S,G,iif) entry if parent != -1 */ 979 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt, 980 __be32 origin, __be32 mcastgrp, 981 int parent) 982 { 983 struct mfc_cache_cmp_arg arg = { 984 .mfc_mcastgrp = mcastgrp, 985 .mfc_origin = origin, 986 }; 987 988 return mr_mfc_find_parent(mrt, &arg, parent); 989 } 990 991 /* Allocate a multicast cache entry */ 992 static struct mfc_cache *ipmr_cache_alloc(void) 993 { 994 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 995 996 if (c) { 997 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 998 c->_c.mfc_un.res.minvif = MAXVIFS; 999 c->_c.free = ipmr_cache_free_rcu; 1000 refcount_set(&c->_c.mfc_un.res.refcount, 1); 1001 } 1002 return c; 1003 } 1004 1005 static struct mfc_cache *ipmr_cache_alloc_unres(void) 1006 { 1007 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 1008 1009 if (c) { 1010 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved); 1011 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ; 1012 } 1013 return c; 1014 } 1015 1016 /* A cache entry has gone into a resolved state from queued */ 1017 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 1018 struct mfc_cache *uc, struct mfc_cache *c) 1019 { 1020 struct sk_buff *skb; 1021 struct nlmsgerr *e; 1022 1023 /* Play the pending entries through our router */ 1024 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) { 1025 if (ip_hdr(skb)->version == 0) { 1026 struct nlmsghdr *nlh = skb_pull(skb, 1027 sizeof(struct iphdr)); 1028 1029 if (mr_fill_mroute(mrt, skb, &c->_c, 1030 nlmsg_data(nlh)) > 0) { 1031 nlh->nlmsg_len = skb_tail_pointer(skb) - 1032 (u8 *)nlh; 1033 } else { 1034 nlh->nlmsg_type = NLMSG_ERROR; 1035 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 1036 skb_trim(skb, nlh->nlmsg_len); 1037 e = nlmsg_data(nlh); 1038 e->error = -EMSGSIZE; 1039 memset(&e->msg, 0, sizeof(e->msg)); 1040 } 1041 1042 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 1043 } else { 1044 rcu_read_lock(); 1045 ip_mr_forward(net, mrt, skb->dev, skb, c, 0); 1046 rcu_read_unlock(); 1047 } 1048 } 1049 } 1050 1051 /* Bounce a cache query up to mrouted and netlink. 1052 * 1053 * Called under rcu_read_lock(). 1054 */ 1055 static int ipmr_cache_report(const struct mr_table *mrt, 1056 struct sk_buff *pkt, vifi_t vifi, int assert) 1057 { 1058 const int ihl = ip_hdrlen(pkt); 1059 struct sock *mroute_sk; 1060 struct igmphdr *igmp; 1061 struct igmpmsg *msg; 1062 struct sk_buff *skb; 1063 int ret; 1064 1065 mroute_sk = rcu_dereference(mrt->mroute_sk); 1066 if (!mroute_sk) 1067 return -EINVAL; 1068 1069 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) 1070 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 1071 else 1072 skb = alloc_skb(128, GFP_ATOMIC); 1073 1074 if (!skb) 1075 return -ENOBUFS; 1076 1077 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) { 1078 /* Ugly, but we have no choice with this interface. 1079 * Duplicate old header, fix ihl, length etc. 1080 * And all this only to mangle msg->im_msgtype and 1081 * to set msg->im_mbz to "mbz" :-) 1082 */ 1083 skb_push(skb, sizeof(struct iphdr)); 1084 skb_reset_network_header(skb); 1085 skb_reset_transport_header(skb); 1086 msg = (struct igmpmsg *)skb_network_header(skb); 1087 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 1088 msg->im_msgtype = assert; 1089 msg->im_mbz = 0; 1090 if (assert == IGMPMSG_WRVIFWHOLE) { 1091 msg->im_vif = vifi; 1092 msg->im_vif_hi = vifi >> 8; 1093 } else { 1094 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */ 1095 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num); 1096 1097 msg->im_vif = vif_num; 1098 msg->im_vif_hi = vif_num >> 8; 1099 } 1100 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 1101 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 1102 sizeof(struct iphdr)); 1103 } else { 1104 /* Copy the IP header */ 1105 skb_set_network_header(skb, skb->len); 1106 skb_put(skb, ihl); 1107 skb_copy_to_linear_data(skb, pkt->data, ihl); 1108 /* Flag to the kernel this is a route add */ 1109 ip_hdr(skb)->protocol = 0; 1110 msg = (struct igmpmsg *)skb_network_header(skb); 1111 msg->im_vif = vifi; 1112 msg->im_vif_hi = vifi >> 8; 1113 ipv4_pktinfo_prepare(mroute_sk, pkt, false); 1114 memcpy(skb->cb, pkt->cb, sizeof(skb->cb)); 1115 /* Add our header */ 1116 igmp = skb_put(skb, sizeof(struct igmphdr)); 1117 igmp->type = assert; 1118 msg->im_msgtype = assert; 1119 igmp->code = 0; 1120 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1121 skb->transport_header = skb->network_header; 1122 } 1123 1124 igmpmsg_netlink_event(mrt, skb); 1125 1126 /* Deliver to mrouted */ 1127 ret = sock_queue_rcv_skb(mroute_sk, skb); 1128 1129 if (ret < 0) { 1130 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1131 kfree_skb(skb); 1132 } 1133 1134 return ret; 1135 } 1136 1137 /* Queue a packet for resolution. It gets locked cache entry! */ 1138 /* Called under rcu_read_lock() */ 1139 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, 1140 struct sk_buff *skb, struct net_device *dev) 1141 { 1142 struct net *net = read_pnet(&mrt->net); 1143 const struct iphdr *iph = ip_hdr(skb); 1144 struct mfc_cache *c = NULL; 1145 bool found = false; 1146 int err; 1147 1148 spin_lock_bh(&mfc_unres_lock); 1149 1150 if (!check_net(net)) { 1151 err = -EINVAL; 1152 goto err; 1153 } 1154 1155 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) { 1156 if (c->mfc_mcastgrp == iph->daddr && 1157 c->mfc_origin == iph->saddr) { 1158 found = true; 1159 break; 1160 } 1161 } 1162 1163 if (!found) { 1164 /* Create a new entry if allowable */ 1165 c = ipmr_cache_alloc_unres(); 1166 if (!c) { 1167 err = -ENOBUFS; 1168 goto err; 1169 } 1170 1171 /* Fill in the new cache entry */ 1172 c->_c.mfc_parent = -1; 1173 c->mfc_origin = iph->saddr; 1174 c->mfc_mcastgrp = iph->daddr; 1175 1176 /* Reflect first query at mrouted. */ 1177 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1178 if (err < 0) 1179 goto err; 1180 1181 atomic_inc(&mrt->cache_resolve_queue_len); 1182 list_add(&c->_c.list, &mrt->mfc_unres_queue); 1183 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1184 1185 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1186 mod_timer(&mrt->ipmr_expire_timer, 1187 c->_c.mfc_un.unres.expires); 1188 } 1189 1190 /* See if we can append the packet */ 1191 if (c->_c.mfc_un.unres.unresolved.qlen > 3) { 1192 c = NULL; 1193 err = -ENOBUFS; 1194 goto err; 1195 } 1196 1197 if (dev) { 1198 skb->dev = dev; 1199 skb->skb_iif = dev->ifindex; 1200 } 1201 1202 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb); 1203 1204 spin_unlock_bh(&mfc_unres_lock); 1205 return 0; 1206 1207 err: 1208 spin_unlock_bh(&mfc_unres_lock); 1209 if (c) 1210 ipmr_cache_free(c); 1211 kfree_skb(skb); 1212 return err; 1213 } 1214 1215 /* MFC cache manipulation by user space mroute daemon */ 1216 1217 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1218 { 1219 struct net *net = read_pnet(&mrt->net); 1220 struct mfc_cache *c; 1221 1222 rcu_read_lock(); 1223 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1224 mfc->mfcc_mcastgrp.s_addr, parent); 1225 rcu_read_unlock(); 1226 if (!c) 1227 return -ENOENT; 1228 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params); 1229 list_del_rcu(&c->_c.list); 1230 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id); 1231 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1232 mr_cache_put(&c->_c); 1233 1234 return 0; 1235 } 1236 1237 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1238 struct mfcctl *mfc, int mrtsock, int parent) 1239 { 1240 struct mfc_cache *uc, *c; 1241 struct mr_mfc *_uc; 1242 bool found; 1243 int ret; 1244 1245 if (mfc->mfcc_parent >= MAXVIFS) 1246 return -ENFILE; 1247 1248 rcu_read_lock(); 1249 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1250 mfc->mfcc_mcastgrp.s_addr, parent); 1251 rcu_read_unlock(); 1252 if (c) { 1253 spin_lock(&mrt_lock); 1254 c->_c.mfc_parent = mfc->mfcc_parent; 1255 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls); 1256 if (!mrtsock) 1257 c->_c.mfc_flags |= MFC_STATIC; 1258 spin_unlock(&mrt_lock); 1259 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c, 1260 mrt->id); 1261 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1262 return 0; 1263 } 1264 1265 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1266 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1267 return -EINVAL; 1268 1269 c = ipmr_cache_alloc(); 1270 if (!c) 1271 return -ENOMEM; 1272 1273 c->mfc_origin = mfc->mfcc_origin.s_addr; 1274 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1275 c->_c.mfc_parent = mfc->mfcc_parent; 1276 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls); 1277 if (!mrtsock) 1278 c->_c.mfc_flags |= MFC_STATIC; 1279 1280 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode, 1281 ipmr_rht_params); 1282 if (ret) { 1283 pr_err("ipmr: rhtable insert error %d\n", ret); 1284 ipmr_cache_free(c); 1285 return ret; 1286 } 1287 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list); 1288 /* Check to see if we resolved a queued list. If so we 1289 * need to send on the frames and tidy up. 1290 */ 1291 found = false; 1292 spin_lock_bh(&mfc_unres_lock); 1293 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) { 1294 uc = (struct mfc_cache *)_uc; 1295 if (uc->mfc_origin == c->mfc_origin && 1296 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1297 list_del(&_uc->list); 1298 atomic_dec(&mrt->cache_resolve_queue_len); 1299 found = true; 1300 break; 1301 } 1302 } 1303 if (list_empty(&mrt->mfc_unres_queue)) 1304 timer_delete(&mrt->ipmr_expire_timer); 1305 spin_unlock_bh(&mfc_unres_lock); 1306 1307 if (found) { 1308 ipmr_cache_resolve(net, mrt, uc, c); 1309 ipmr_cache_free(uc); 1310 } 1311 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id); 1312 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1313 return 0; 1314 } 1315 1316 /* Close the multicast socket, and clear the vif tables etc */ 1317 static void mroute_clean_tables(struct mr_table *mrt, int flags, 1318 struct list_head *dev_kill_list) 1319 { 1320 struct net *net = read_pnet(&mrt->net); 1321 struct mfc_cache *cache; 1322 struct mr_mfc *c, *tmp; 1323 int i; 1324 1325 /* Shut down all active vif entries */ 1326 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) { 1327 for (i = 0; i < mrt->maxvif; i++) { 1328 if (((mrt->vif_table[i].flags & VIFF_STATIC) && 1329 !(flags & MRT_FLUSH_VIFS_STATIC)) || 1330 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS))) 1331 continue; 1332 vif_delete(mrt, i, 0, dev_kill_list); 1333 } 1334 } 1335 1336 /* Wipe the cache */ 1337 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) { 1338 mutex_lock(&net->ipv4.mfc_mutex); 1339 1340 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1341 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) || 1342 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC))) 1343 continue; 1344 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1345 list_del_rcu(&c->list); 1346 cache = (struct mfc_cache *)c; 1347 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache, 1348 mrt->id); 1349 mroute_netlink_event(mrt, cache, RTM_DELROUTE); 1350 mr_cache_put(c); 1351 } 1352 1353 mutex_unlock(&net->ipv4.mfc_mutex); 1354 } 1355 1356 if (flags & MRT_FLUSH_MFC) { 1357 if (atomic_read(&mrt->cache_resolve_queue_len) != 0 || !check_net(net)) { 1358 spin_lock_bh(&mfc_unres_lock); 1359 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1360 list_del(&c->list); 1361 cache = (struct mfc_cache *)c; 1362 mroute_netlink_event(mrt, cache, RTM_DELROUTE); 1363 ipmr_destroy_unres(mrt, cache); 1364 } 1365 spin_unlock_bh(&mfc_unres_lock); 1366 } 1367 } 1368 } 1369 1370 /* called from ip_ra_control(), before an RCU grace period, 1371 * we don't need to call synchronize_rcu() here 1372 */ 1373 static void mrtsock_destruct(struct sock *sk) 1374 { 1375 struct net *net = sock_net(sk); 1376 LIST_HEAD(dev_kill_list); 1377 struct mr_table *mrt; 1378 1379 rtnl_lock(); 1380 1381 ipmr_for_each_table(mrt, net) { 1382 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1383 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1384 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1385 NETCONFA_MC_FORWARDING, 1386 NETCONFA_IFINDEX_ALL, 1387 net->ipv4.devconf_all); 1388 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1389 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC, 1390 &dev_kill_list); 1391 } 1392 } 1393 1394 unregister_netdevice_many(&dev_kill_list); 1395 1396 rtnl_unlock(); 1397 } 1398 1399 /* Socket options and virtual interface manipulation. The whole 1400 * virtual interface system is a complete heap, but unfortunately 1401 * that's how BSD mrouted happens to think. Maybe one day with a proper 1402 * MOSPF/PIM router set up we can clean this up. 1403 */ 1404 1405 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval, 1406 unsigned int optlen) 1407 { 1408 struct net *net = sock_net(sk); 1409 int val, ret = 0, parent = 0; 1410 struct mr_table *mrt; 1411 struct vifctl vif; 1412 struct mfcctl mfc; 1413 bool do_wrvifwhole; 1414 u32 uval; 1415 1416 /* There's one exception to the lock - MRT_DONE which needs to unlock */ 1417 rtnl_lock(); 1418 if (sk->sk_type != SOCK_RAW || 1419 inet_sk(sk)->inet_num != IPPROTO_IGMP) { 1420 ret = -EOPNOTSUPP; 1421 goto out_unlock; 1422 } 1423 1424 mrt = __ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1425 if (!mrt) { 1426 ret = -ENOENT; 1427 goto out_unlock; 1428 } 1429 if (optname != MRT_INIT) { 1430 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1431 !ns_capable(net->user_ns, CAP_NET_ADMIN)) { 1432 ret = -EACCES; 1433 goto out_unlock; 1434 } 1435 } 1436 1437 switch (optname) { 1438 case MRT_INIT: 1439 if (optlen != sizeof(int)) { 1440 ret = -EINVAL; 1441 break; 1442 } 1443 if (rtnl_dereference(mrt->mroute_sk)) { 1444 ret = -EADDRINUSE; 1445 break; 1446 } 1447 1448 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1449 if (ret == 0) { 1450 rcu_assign_pointer(mrt->mroute_sk, sk); 1451 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1452 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1453 NETCONFA_MC_FORWARDING, 1454 NETCONFA_IFINDEX_ALL, 1455 net->ipv4.devconf_all); 1456 } 1457 break; 1458 case MRT_DONE: 1459 if (sk != rcu_access_pointer(mrt->mroute_sk)) { 1460 ret = -EACCES; 1461 } else { 1462 /* We need to unlock here because mrtsock_destruct takes 1463 * care of rtnl itself and we can't change that due to 1464 * the IP_ROUTER_ALERT setsockopt which runs without it. 1465 */ 1466 rtnl_unlock(); 1467 ret = ip_ra_control(sk, 0, NULL); 1468 goto out; 1469 } 1470 break; 1471 case MRT_ADD_VIF: 1472 case MRT_DEL_VIF: 1473 if (optlen != sizeof(vif)) { 1474 ret = -EINVAL; 1475 break; 1476 } 1477 if (copy_from_sockptr(&vif, optval, sizeof(vif))) { 1478 ret = -EFAULT; 1479 break; 1480 } 1481 if (vif.vifc_vifi >= MAXVIFS) { 1482 ret = -ENFILE; 1483 break; 1484 } 1485 if (optname == MRT_ADD_VIF) { 1486 ret = vif_add(net, mrt, &vif, 1487 sk == rtnl_dereference(mrt->mroute_sk)); 1488 } else { 1489 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1490 } 1491 break; 1492 /* Manipulate the forwarding caches. These live 1493 * in a sort of kernel/user symbiosis. 1494 */ 1495 case MRT_ADD_MFC: 1496 case MRT_DEL_MFC: 1497 parent = -1; 1498 fallthrough; 1499 case MRT_ADD_MFC_PROXY: 1500 case MRT_DEL_MFC_PROXY: 1501 if (optlen != sizeof(mfc)) { 1502 ret = -EINVAL; 1503 break; 1504 } 1505 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) { 1506 ret = -EFAULT; 1507 break; 1508 } 1509 if (parent == 0) 1510 parent = mfc.mfcc_parent; 1511 1512 mutex_lock(&net->ipv4.mfc_mutex); 1513 1514 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1515 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1516 else 1517 ret = ipmr_mfc_add(net, mrt, &mfc, 1518 sk == rtnl_dereference(mrt->mroute_sk), 1519 parent); 1520 1521 mutex_unlock(&net->ipv4.mfc_mutex); 1522 break; 1523 case MRT_FLUSH: { 1524 LIST_HEAD(dev_kill_list); 1525 1526 if (optlen != sizeof(val)) { 1527 ret = -EINVAL; 1528 break; 1529 } 1530 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1531 ret = -EFAULT; 1532 break; 1533 } 1534 1535 mroute_clean_tables(mrt, val, &dev_kill_list); 1536 unregister_netdevice_many(&dev_kill_list); 1537 break; 1538 } 1539 /* Control PIM assert. */ 1540 case MRT_ASSERT: 1541 if (optlen != sizeof(val)) { 1542 ret = -EINVAL; 1543 break; 1544 } 1545 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1546 ret = -EFAULT; 1547 break; 1548 } 1549 WRITE_ONCE(mrt->mroute_do_assert, val); 1550 break; 1551 case MRT_PIM: 1552 if (!ipmr_pimsm_enabled()) { 1553 ret = -ENOPROTOOPT; 1554 break; 1555 } 1556 if (optlen != sizeof(val)) { 1557 ret = -EINVAL; 1558 break; 1559 } 1560 if (copy_from_sockptr(&val, optval, sizeof(val))) { 1561 ret = -EFAULT; 1562 break; 1563 } 1564 1565 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE); 1566 val = !!val; 1567 if (val != mrt->mroute_do_pim) { 1568 WRITE_ONCE(mrt->mroute_do_pim, val); 1569 WRITE_ONCE(mrt->mroute_do_assert, val); 1570 WRITE_ONCE(mrt->mroute_do_wrvifwhole, do_wrvifwhole); 1571 } 1572 break; 1573 case MRT_TABLE: 1574 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) { 1575 ret = -ENOPROTOOPT; 1576 break; 1577 } 1578 if (optlen != sizeof(uval)) { 1579 ret = -EINVAL; 1580 break; 1581 } 1582 if (copy_from_sockptr(&uval, optval, sizeof(uval))) { 1583 ret = -EFAULT; 1584 break; 1585 } 1586 1587 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1588 ret = -EBUSY; 1589 } else { 1590 mrt = ipmr_new_table(net, uval); 1591 if (IS_ERR(mrt)) 1592 ret = PTR_ERR(mrt); 1593 else 1594 raw_sk(sk)->ipmr_table = uval; 1595 } 1596 break; 1597 /* Spurious command, or MRT_VERSION which you cannot set. */ 1598 default: 1599 ret = -ENOPROTOOPT; 1600 } 1601 out_unlock: 1602 rtnl_unlock(); 1603 out: 1604 return ret; 1605 } 1606 1607 /* Execute if this ioctl is a special mroute ioctl */ 1608 int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1609 { 1610 switch (cmd) { 1611 /* These userspace buffers will be consumed by ipmr_ioctl() */ 1612 case SIOCGETVIFCNT: { 1613 struct sioc_vif_req buffer; 1614 1615 return sock_ioctl_inout(sk, cmd, arg, &buffer, 1616 sizeof(buffer)); 1617 } 1618 case SIOCGETSGCNT: { 1619 struct sioc_sg_req buffer; 1620 1621 return sock_ioctl_inout(sk, cmd, arg, &buffer, 1622 sizeof(buffer)); 1623 } 1624 } 1625 /* return code > 0 means that the ioctl was not executed */ 1626 return 1; 1627 } 1628 1629 /* Getsock opt support for the multicast routing system. */ 1630 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval, 1631 sockptr_t optlen) 1632 { 1633 int olr; 1634 int val; 1635 struct net *net = sock_net(sk); 1636 struct mr_table *mrt; 1637 1638 if (sk->sk_type != SOCK_RAW || 1639 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1640 return -EOPNOTSUPP; 1641 1642 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1643 if (!mrt) 1644 return -ENOENT; 1645 1646 switch (optname) { 1647 case MRT_VERSION: 1648 val = 0x0305; 1649 break; 1650 case MRT_PIM: 1651 if (!ipmr_pimsm_enabled()) 1652 return -ENOPROTOOPT; 1653 val = READ_ONCE(mrt->mroute_do_pim); 1654 break; 1655 case MRT_ASSERT: 1656 val = READ_ONCE(mrt->mroute_do_assert); 1657 break; 1658 default: 1659 return -ENOPROTOOPT; 1660 } 1661 1662 if (copy_from_sockptr(&olr, optlen, sizeof(int))) 1663 return -EFAULT; 1664 if (olr < 0) 1665 return -EINVAL; 1666 1667 olr = min_t(unsigned int, olr, sizeof(int)); 1668 1669 if (copy_to_sockptr(optlen, &olr, sizeof(int))) 1670 return -EFAULT; 1671 if (copy_to_sockptr(optval, &val, olr)) 1672 return -EFAULT; 1673 return 0; 1674 } 1675 1676 /* The IP multicast ioctl support routines. */ 1677 int ipmr_ioctl(struct sock *sk, int cmd, void *arg) 1678 { 1679 struct vif_device *vif; 1680 struct mfc_cache *c; 1681 struct net *net = sock_net(sk); 1682 struct sioc_vif_req *vr; 1683 struct sioc_sg_req *sr; 1684 struct mr_table *mrt; 1685 1686 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1687 if (!mrt) 1688 return -ENOENT; 1689 1690 switch (cmd) { 1691 case SIOCGETVIFCNT: 1692 vr = (struct sioc_vif_req *)arg; 1693 if (vr->vifi >= mrt->maxvif) 1694 return -EINVAL; 1695 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif); 1696 rcu_read_lock(); 1697 vif = &mrt->vif_table[vr->vifi]; 1698 if (VIF_EXISTS(mrt, vr->vifi)) { 1699 vr->icount = READ_ONCE(vif->pkt_in); 1700 vr->ocount = READ_ONCE(vif->pkt_out); 1701 vr->ibytes = READ_ONCE(vif->bytes_in); 1702 vr->obytes = READ_ONCE(vif->bytes_out); 1703 rcu_read_unlock(); 1704 1705 return 0; 1706 } 1707 rcu_read_unlock(); 1708 return -EADDRNOTAVAIL; 1709 case SIOCGETSGCNT: 1710 sr = (struct sioc_sg_req *)arg; 1711 1712 rcu_read_lock(); 1713 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr); 1714 if (c) { 1715 sr->pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt); 1716 sr->bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes); 1717 sr->wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if); 1718 rcu_read_unlock(); 1719 return 0; 1720 } 1721 rcu_read_unlock(); 1722 return -EADDRNOTAVAIL; 1723 default: 1724 return -ENOIOCTLCMD; 1725 } 1726 } 1727 1728 #ifdef CONFIG_COMPAT 1729 struct compat_sioc_sg_req { 1730 struct in_addr src; 1731 struct in_addr grp; 1732 compat_ulong_t pktcnt; 1733 compat_ulong_t bytecnt; 1734 compat_ulong_t wrong_if; 1735 }; 1736 1737 struct compat_sioc_vif_req { 1738 vifi_t vifi; /* Which iface */ 1739 compat_ulong_t icount; 1740 compat_ulong_t ocount; 1741 compat_ulong_t ibytes; 1742 compat_ulong_t obytes; 1743 }; 1744 1745 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1746 { 1747 struct compat_sioc_sg_req sr; 1748 struct compat_sioc_vif_req vr; 1749 struct vif_device *vif; 1750 struct mfc_cache *c; 1751 struct net *net = sock_net(sk); 1752 struct mr_table *mrt; 1753 1754 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1755 if (!mrt) 1756 return -ENOENT; 1757 1758 switch (cmd) { 1759 case SIOCGETVIFCNT: 1760 if (copy_from_user(&vr, arg, sizeof(vr))) 1761 return -EFAULT; 1762 if (vr.vifi >= mrt->maxvif) 1763 return -EINVAL; 1764 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif); 1765 rcu_read_lock(); 1766 vif = &mrt->vif_table[vr.vifi]; 1767 if (VIF_EXISTS(mrt, vr.vifi)) { 1768 vr.icount = READ_ONCE(vif->pkt_in); 1769 vr.ocount = READ_ONCE(vif->pkt_out); 1770 vr.ibytes = READ_ONCE(vif->bytes_in); 1771 vr.obytes = READ_ONCE(vif->bytes_out); 1772 rcu_read_unlock(); 1773 1774 if (copy_to_user(arg, &vr, sizeof(vr))) 1775 return -EFAULT; 1776 return 0; 1777 } 1778 rcu_read_unlock(); 1779 return -EADDRNOTAVAIL; 1780 case SIOCGETSGCNT: 1781 if (copy_from_user(&sr, arg, sizeof(sr))) 1782 return -EFAULT; 1783 1784 rcu_read_lock(); 1785 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1786 if (c) { 1787 sr.pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt); 1788 sr.bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes); 1789 sr.wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if); 1790 rcu_read_unlock(); 1791 1792 if (copy_to_user(arg, &sr, sizeof(sr))) 1793 return -EFAULT; 1794 return 0; 1795 } 1796 rcu_read_unlock(); 1797 return -EADDRNOTAVAIL; 1798 default: 1799 return -ENOIOCTLCMD; 1800 } 1801 } 1802 #endif 1803 1804 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1805 { 1806 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1807 struct net *net = dev_net(dev); 1808 struct mr_table *mrt; 1809 struct vif_device *v; 1810 int ct; 1811 1812 if (event != NETDEV_UNREGISTER) 1813 return NOTIFY_DONE; 1814 1815 ipmr_for_each_table(mrt, net) { 1816 v = &mrt->vif_table[0]; 1817 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1818 if (rcu_access_pointer(v->dev) == dev) 1819 vif_delete(mrt, ct, 1, NULL); 1820 } 1821 } 1822 return NOTIFY_DONE; 1823 } 1824 1825 static struct notifier_block ip_mr_notifier = { 1826 .notifier_call = ipmr_device_event, 1827 }; 1828 1829 /* Encapsulate a packet by attaching a valid IPIP header to it. 1830 * This avoids tunnel drivers and other mess and gives us the speed so 1831 * important for multicast video. 1832 */ 1833 static void ip_encap(struct net *net, struct sk_buff *skb, 1834 __be32 saddr, __be32 daddr) 1835 { 1836 struct iphdr *iph; 1837 const struct iphdr *old_iph = ip_hdr(skb); 1838 1839 skb_push(skb, sizeof(struct iphdr)); 1840 skb->transport_header = skb->network_header; 1841 skb_reset_network_header(skb); 1842 iph = ip_hdr(skb); 1843 1844 iph->version = 4; 1845 iph->tos = old_iph->tos; 1846 iph->ttl = old_iph->ttl; 1847 iph->frag_off = 0; 1848 iph->daddr = daddr; 1849 iph->saddr = saddr; 1850 iph->protocol = IPPROTO_IPIP; 1851 iph->ihl = 5; 1852 iph->tot_len = htons(skb->len); 1853 ip_select_ident(net, skb, NULL); 1854 ip_send_check(iph); 1855 1856 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1857 nf_reset_ct(skb); 1858 } 1859 1860 static inline int ipmr_forward_finish(struct net *net, struct sock *sk, 1861 struct sk_buff *skb) 1862 { 1863 struct ip_options *opt = &(IPCB(skb)->opt); 1864 1865 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS); 1866 1867 if (unlikely(opt->optlen)) 1868 ip_forward_options(skb); 1869 1870 return dst_output(net, sk, skb); 1871 } 1872 1873 #ifdef CONFIG_NET_SWITCHDEV 1874 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt, 1875 int in_vifi, int out_vifi) 1876 { 1877 struct vif_device *out_vif = &mrt->vif_table[out_vifi]; 1878 struct vif_device *in_vif = &mrt->vif_table[in_vifi]; 1879 1880 if (!skb->offload_l3_fwd_mark) 1881 return false; 1882 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len) 1883 return false; 1884 return netdev_phys_item_id_same(&out_vif->dev_parent_id, 1885 &in_vif->dev_parent_id); 1886 } 1887 #else 1888 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt, 1889 int in_vifi, int out_vifi) 1890 { 1891 return false; 1892 } 1893 #endif 1894 1895 /* Processing handlers for ipmr_forward, under rcu_read_lock() */ 1896 1897 static int ipmr_prepare_xmit(struct net *net, struct mr_table *mrt, 1898 struct sk_buff *skb, int vifi) 1899 { 1900 const struct iphdr *iph = ip_hdr(skb); 1901 struct vif_device *vif = &mrt->vif_table[vifi]; 1902 struct net_device *vif_dev; 1903 struct rtable *rt; 1904 struct flowi4 fl4; 1905 int encap = 0; 1906 1907 vif_dev = vif_dev_read(vif); 1908 if (!vif_dev) 1909 return -1; 1910 1911 if (vif->flags & VIFF_REGISTER) { 1912 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 1913 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 1914 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len); 1915 DEV_STATS_INC(vif_dev, tx_packets); 1916 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1917 return -1; 1918 } 1919 1920 if (vif->flags & VIFF_TUNNEL) { 1921 rt = ip_route_output_ports(net, &fl4, NULL, 1922 vif->remote, vif->local, 1923 0, 0, 1924 IPPROTO_IPIP, 1925 iph->tos & INET_DSCP_MASK, vif->link); 1926 if (IS_ERR(rt)) 1927 return -1; 1928 encap = sizeof(struct iphdr); 1929 } else { 1930 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1931 0, 0, 1932 IPPROTO_IPIP, 1933 iph->tos & INET_DSCP_MASK, vif->link); 1934 if (IS_ERR(rt)) 1935 return -1; 1936 } 1937 1938 if (skb->len+encap > dst4_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1939 /* Do not fragment multicasts. Alas, IPv4 does not 1940 * allow to send ICMP, so that packets will disappear 1941 * to blackhole. 1942 */ 1943 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS); 1944 ip_rt_put(rt); 1945 return -1; 1946 } 1947 1948 encap += LL_RESERVED_SPACE(dst_dev_rcu(&rt->dst)) + rt->dst.header_len; 1949 1950 if (skb_cow(skb, encap)) { 1951 ip_rt_put(rt); 1952 return -1; 1953 } 1954 1955 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1); 1956 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len); 1957 1958 skb_dst_drop(skb); 1959 skb_dst_set(skb, &rt->dst); 1960 ip_decrease_ttl(ip_hdr(skb)); 1961 1962 /* FIXME: forward and output firewalls used to be called here. 1963 * What do we do with netfilter? -- RR 1964 */ 1965 if (vif->flags & VIFF_TUNNEL) { 1966 ip_encap(net, skb, vif->local, vif->remote); 1967 /* FIXME: extra output firewall step used to be here. --RR */ 1968 DEV_STATS_INC(vif_dev, tx_packets); 1969 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len); 1970 } 1971 1972 return 0; 1973 } 1974 1975 static void ipmr_queue_fwd_xmit(struct net *net, struct mr_table *mrt, 1976 int in_vifi, struct sk_buff *skb, int vifi) 1977 { 1978 struct rtable *rt; 1979 1980 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi)) 1981 goto out_free; 1982 1983 if (ipmr_prepare_xmit(net, mrt, skb, vifi)) 1984 goto out_free; 1985 1986 rt = skb_rtable(skb); 1987 1988 IPCB(skb)->flags |= IPSKB_FORWARDED; 1989 1990 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1991 * not only before forwarding, but after forwarding on all output 1992 * interfaces. It is clear, if mrouter runs a multicasting 1993 * program, it should receive packets not depending to what interface 1994 * program is joined. 1995 * If we will not make it, the program will have to join on all 1996 * interfaces. On the other hand, multihoming host (or router, but 1997 * not mrouter) cannot join to more than one interface - it will 1998 * result in receiving multiple packets. 1999 */ 2000 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, 2001 net, NULL, skb, skb->dev, dst_dev_rcu(&rt->dst), 2002 ipmr_forward_finish); 2003 return; 2004 2005 out_free: 2006 kfree_skb(skb); 2007 } 2008 2009 static void ipmr_queue_output_xmit(struct net *net, struct mr_table *mrt, 2010 struct sk_buff *skb, int vifi) 2011 { 2012 if (ipmr_prepare_xmit(net, mrt, skb, vifi)) 2013 goto out_free; 2014 2015 ip_mc_output(net, NULL, skb); 2016 return; 2017 2018 out_free: 2019 kfree_skb(skb); 2020 } 2021 2022 /* Called with mrt_lock or rcu_read_lock() */ 2023 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev) 2024 { 2025 int ct; 2026 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */ 2027 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) { 2028 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev) 2029 break; 2030 } 2031 return ct; 2032 } 2033 2034 /* "local" means that we should preserve one skb (for local delivery) */ 2035 /* Called uner rcu_read_lock() */ 2036 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 2037 struct net_device *dev, struct sk_buff *skb, 2038 struct mfc_cache *c, int local) 2039 { 2040 int true_vifi = ipmr_find_vif(mrt, dev); 2041 int psend = -1; 2042 int vif, ct; 2043 2044 vif = c->_c.mfc_parent; 2045 atomic_long_inc(&c->_c.mfc_un.res.pkt); 2046 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes); 2047 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies); 2048 2049 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 2050 struct mfc_cache *cache_proxy; 2051 2052 /* For an (*,G) entry, we only check that the incoming 2053 * interface is part of the static tree. 2054 */ 2055 cache_proxy = mr_mfc_find_any_parent(mrt, vif); 2056 if (cache_proxy && 2057 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) 2058 goto forward; 2059 } 2060 2061 /* Wrong interface: drop packet and (maybe) send PIM assert. */ 2062 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) { 2063 if (rt_is_output_route(skb_rtable(skb))) { 2064 /* It is our own packet, looped back. 2065 * Very complicated situation... 2066 * 2067 * The best workaround until routing daemons will be 2068 * fixed is not to redistribute packet, if it was 2069 * send through wrong interface. It means, that 2070 * multicast applications WILL NOT work for 2071 * (S,G), which have default multicast route pointing 2072 * to wrong oif. In any case, it is not a good 2073 * idea to use multicasting applications on router. 2074 */ 2075 goto dont_forward; 2076 } 2077 2078 atomic_long_inc(&c->_c.mfc_un.res.wrong_if); 2079 2080 if (true_vifi >= 0 && READ_ONCE(mrt->mroute_do_assert) && 2081 /* pimsm uses asserts, when switching from RPT to SPT, 2082 * so that we cannot check that packet arrived on an oif. 2083 * It is bad, but otherwise we would need to move pretty 2084 * large chunk of pimd to kernel. Ough... --ANK 2085 */ 2086 (READ_ONCE(mrt->mroute_do_pim) || 2087 c->_c.mfc_un.res.ttls[true_vifi] < 255) && 2088 time_after(jiffies, 2089 c->_c.mfc_un.res.last_assert + 2090 MFC_ASSERT_THRESH)) { 2091 c->_c.mfc_un.res.last_assert = jiffies; 2092 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 2093 if (READ_ONCE(mrt->mroute_do_wrvifwhole)) 2094 ipmr_cache_report(mrt, skb, true_vifi, 2095 IGMPMSG_WRVIFWHOLE); 2096 } 2097 goto dont_forward; 2098 } 2099 2100 forward: 2101 WRITE_ONCE(mrt->vif_table[vif].pkt_in, 2102 mrt->vif_table[vif].pkt_in + 1); 2103 WRITE_ONCE(mrt->vif_table[vif].bytes_in, 2104 mrt->vif_table[vif].bytes_in + skb->len); 2105 2106 /* Forward the frame */ 2107 if (c->mfc_origin == htonl(INADDR_ANY) && 2108 c->mfc_mcastgrp == htonl(INADDR_ANY)) { 2109 if (true_vifi >= 0 && 2110 true_vifi != c->_c.mfc_parent && 2111 ip_hdr(skb)->ttl > 2112 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) { 2113 /* It's an (*,*) entry and the packet is not coming from 2114 * the upstream: forward the packet to the upstream 2115 * only. 2116 */ 2117 psend = c->_c.mfc_parent; 2118 goto last_forward; 2119 } 2120 goto dont_forward; 2121 } 2122 for (ct = c->_c.mfc_un.res.maxvif - 1; 2123 ct >= c->_c.mfc_un.res.minvif; ct--) { 2124 /* For (*,G) entry, don't forward to the incoming interface */ 2125 if ((c->mfc_origin != htonl(INADDR_ANY) || 2126 ct != true_vifi) && 2127 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) { 2128 if (psend != -1) { 2129 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2130 2131 if (skb2) 2132 ipmr_queue_fwd_xmit(net, mrt, true_vifi, 2133 skb2, psend); 2134 } 2135 psend = ct; 2136 } 2137 } 2138 last_forward: 2139 if (psend != -1) { 2140 if (local) { 2141 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2142 2143 if (skb2) 2144 ipmr_queue_fwd_xmit(net, mrt, true_vifi, skb2, 2145 psend); 2146 } else { 2147 ipmr_queue_fwd_xmit(net, mrt, true_vifi, skb, psend); 2148 return; 2149 } 2150 } 2151 2152 dont_forward: 2153 if (!local) 2154 kfree_skb(skb); 2155 } 2156 2157 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 2158 { 2159 struct rtable *rt = skb_rtable(skb); 2160 struct iphdr *iph = ip_hdr(skb); 2161 struct flowi4 fl4 = { 2162 .daddr = iph->daddr, 2163 .saddr = iph->saddr, 2164 .flowi4_dscp = ip4h_dscp(iph), 2165 .flowi4_oif = (rt_is_output_route(rt) ? 2166 skb->dev->ifindex : 0), 2167 .flowi4_iif = (rt_is_output_route(rt) ? 2168 LOOPBACK_IFINDEX : 2169 skb->dev->ifindex), 2170 .flowi4_mark = skb->mark, 2171 }; 2172 struct mr_table *mrt; 2173 int err; 2174 2175 err = ipmr_fib_lookup(net, &fl4, &mrt); 2176 if (err) 2177 return ERR_PTR(err); 2178 return mrt; 2179 } 2180 2181 /* Multicast packets for forwarding arrive here 2182 * Called with rcu_read_lock(); 2183 */ 2184 int ip_mr_input(struct sk_buff *skb) 2185 { 2186 struct mfc_cache *cache; 2187 struct net *net = dev_net(skb->dev); 2188 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 2189 struct mr_table *mrt; 2190 struct net_device *dev; 2191 2192 /* skb->dev passed in is the loX master dev for vrfs. 2193 * As there are no vifs associated with loopback devices, 2194 * get the proper interface that does have a vif associated with it. 2195 */ 2196 dev = skb->dev; 2197 if (netif_is_l3_master(skb->dev)) { 2198 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 2199 if (!dev) { 2200 kfree_skb(skb); 2201 return -ENODEV; 2202 } 2203 } 2204 2205 /* Packet is looped back after forward, it should not be 2206 * forwarded second time, but still can be delivered locally. 2207 */ 2208 if (IPCB(skb)->flags & IPSKB_FORWARDED) 2209 goto dont_forward; 2210 2211 mrt = ipmr_rt_fib_lookup(net, skb); 2212 if (IS_ERR(mrt)) { 2213 kfree_skb(skb); 2214 return PTR_ERR(mrt); 2215 } 2216 if (!local) { 2217 if (IPCB(skb)->opt.router_alert) { 2218 if (ip_call_ra_chain(skb)) 2219 return 0; 2220 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 2221 /* IGMPv1 (and broken IGMPv2 implementations sort of 2222 * Cisco IOS <= 11.2(8)) do not put router alert 2223 * option to IGMP packets destined to routable 2224 * groups. It is very bad, because it means 2225 * that we can forward NO IGMP messages. 2226 */ 2227 struct sock *mroute_sk; 2228 2229 mroute_sk = rcu_dereference(mrt->mroute_sk); 2230 if (mroute_sk) { 2231 nf_reset_ct(skb); 2232 raw_rcv(mroute_sk, skb); 2233 return 0; 2234 } 2235 } 2236 } 2237 2238 /* already under rcu_read_lock() */ 2239 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 2240 if (!cache) { 2241 int vif = ipmr_find_vif(mrt, dev); 2242 2243 if (vif >= 0) 2244 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2245 vif); 2246 } 2247 2248 /* No usable cache entry */ 2249 if (!cache) { 2250 int vif; 2251 2252 if (local) { 2253 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2254 ip_local_deliver(skb); 2255 if (!skb2) 2256 return -ENOBUFS; 2257 skb = skb2; 2258 } 2259 2260 vif = ipmr_find_vif(mrt, dev); 2261 if (vif >= 0) 2262 return ipmr_cache_unresolved(mrt, vif, skb, dev); 2263 kfree_skb(skb); 2264 return -ENODEV; 2265 } 2266 2267 ip_mr_forward(net, mrt, dev, skb, cache, local); 2268 2269 if (local) 2270 return ip_local_deliver(skb); 2271 2272 return 0; 2273 2274 dont_forward: 2275 if (local) 2276 return ip_local_deliver(skb); 2277 kfree_skb(skb); 2278 return 0; 2279 } 2280 2281 static void ip_mr_output_finish(struct net *net, struct mr_table *mrt, 2282 struct net_device *dev, struct sk_buff *skb, 2283 struct mfc_cache *c) 2284 { 2285 int psend = -1; 2286 int ct; 2287 2288 atomic_long_inc(&c->_c.mfc_un.res.pkt); 2289 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes); 2290 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies); 2291 2292 /* Forward the frame */ 2293 if (c->mfc_origin == htonl(INADDR_ANY) && 2294 c->mfc_mcastgrp == htonl(INADDR_ANY)) { 2295 if (ip_hdr(skb)->ttl > 2296 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) { 2297 /* It's an (*,*) entry and the packet is not coming from 2298 * the upstream: forward the packet to the upstream 2299 * only. 2300 */ 2301 psend = c->_c.mfc_parent; 2302 goto last_xmit; 2303 } 2304 goto dont_xmit; 2305 } 2306 2307 for (ct = c->_c.mfc_un.res.maxvif - 1; 2308 ct >= c->_c.mfc_un.res.minvif; ct--) { 2309 if (ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) { 2310 if (psend != -1) { 2311 struct sk_buff *skb2; 2312 2313 skb2 = skb_clone(skb, GFP_ATOMIC); 2314 if (skb2) 2315 ipmr_queue_output_xmit(net, mrt, 2316 skb2, psend); 2317 } 2318 psend = ct; 2319 } 2320 } 2321 2322 last_xmit: 2323 if (psend != -1) { 2324 ipmr_queue_output_xmit(net, mrt, skb, psend); 2325 return; 2326 } 2327 2328 dont_xmit: 2329 kfree_skb(skb); 2330 } 2331 2332 /* Multicast packets for forwarding arrive here 2333 * Called with rcu_read_lock(); 2334 */ 2335 int ip_mr_output(struct net *net, struct sock *sk, struct sk_buff *skb) 2336 { 2337 struct rtable *rt = skb_rtable(skb); 2338 struct mfc_cache *cache; 2339 struct net_device *dev; 2340 struct mr_table *mrt; 2341 int vif; 2342 2343 guard(rcu)(); 2344 2345 dev = dst_dev_rcu(&rt->dst); 2346 2347 if (IPCB(skb)->flags & IPSKB_FORWARDED) 2348 goto mc_output; 2349 if (!(IPCB(skb)->flags & IPSKB_MCROUTE)) 2350 goto mc_output; 2351 2352 skb->dev = dev; 2353 2354 mrt = ipmr_rt_fib_lookup(net, skb); 2355 if (IS_ERR(mrt)) 2356 goto mc_output; 2357 2358 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 2359 if (!cache) { 2360 vif = ipmr_find_vif(mrt, dev); 2361 if (vif >= 0) 2362 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2363 vif); 2364 } 2365 2366 /* No usable cache entry */ 2367 if (!cache) { 2368 vif = ipmr_find_vif(mrt, dev); 2369 if (vif >= 0) 2370 return ipmr_cache_unresolved(mrt, vif, skb, dev); 2371 goto mc_output; 2372 } 2373 2374 vif = cache->_c.mfc_parent; 2375 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) 2376 goto mc_output; 2377 2378 ip_mr_output_finish(net, mrt, dev, skb, cache); 2379 return 0; 2380 2381 mc_output: 2382 return ip_mc_output(net, sk, skb); 2383 } 2384 2385 #ifdef CONFIG_IP_PIMSM_V1 2386 /* Handle IGMP messages of PIMv1 */ 2387 int pim_rcv_v1(struct sk_buff *skb) 2388 { 2389 struct igmphdr *pim; 2390 struct net *net = dev_net(skb->dev); 2391 struct mr_table *mrt; 2392 2393 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2394 goto drop; 2395 2396 pim = igmp_hdr(skb); 2397 2398 mrt = ipmr_rt_fib_lookup(net, skb); 2399 if (IS_ERR(mrt)) 2400 goto drop; 2401 if (!READ_ONCE(mrt->mroute_do_pim) || 2402 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2403 goto drop; 2404 2405 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2406 drop: 2407 kfree_skb(skb); 2408 } 2409 return 0; 2410 } 2411 #endif 2412 2413 #ifdef CONFIG_IP_PIMSM_V2 2414 static int pim_rcv(struct sk_buff *skb) 2415 { 2416 struct pimreghdr *pim; 2417 struct net *net = dev_net(skb->dev); 2418 struct mr_table *mrt; 2419 2420 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2421 goto drop; 2422 2423 pim = (struct pimreghdr *)skb_transport_header(skb); 2424 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) || 2425 (pim->flags & PIM_NULL_REGISTER) || 2426 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2427 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2428 goto drop; 2429 2430 mrt = ipmr_rt_fib_lookup(net, skb); 2431 if (IS_ERR(mrt)) 2432 goto drop; 2433 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2434 drop: 2435 kfree_skb(skb); 2436 } 2437 return 0; 2438 } 2439 #endif 2440 2441 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2442 __be32 saddr, __be32 daddr, 2443 struct rtmsg *rtm, u32 portid) 2444 { 2445 struct mfc_cache *cache; 2446 struct mr_table *mrt; 2447 int err; 2448 2449 rcu_read_lock(); 2450 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT); 2451 if (!mrt) { 2452 rcu_read_unlock(); 2453 return -ENOENT; 2454 } 2455 2456 cache = ipmr_cache_find(mrt, saddr, daddr); 2457 if (!cache && skb->dev) { 2458 int vif = ipmr_find_vif(mrt, skb->dev); 2459 2460 if (vif >= 0) 2461 cache = ipmr_cache_find_any(mrt, daddr, vif); 2462 } 2463 if (!cache) { 2464 struct sk_buff *skb2; 2465 struct iphdr *iph; 2466 struct net_device *dev; 2467 int vif = -1; 2468 2469 dev = skb->dev; 2470 if (dev) 2471 vif = ipmr_find_vif(mrt, dev); 2472 if (vif < 0) { 2473 rcu_read_unlock(); 2474 return -ENODEV; 2475 } 2476 2477 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr)); 2478 if (!skb2) { 2479 rcu_read_unlock(); 2480 return -ENOMEM; 2481 } 2482 2483 NETLINK_CB(skb2).portid = portid; 2484 skb_push(skb2, sizeof(struct iphdr)); 2485 skb_reset_network_header(skb2); 2486 iph = ip_hdr(skb2); 2487 iph->ihl = sizeof(struct iphdr) >> 2; 2488 iph->saddr = saddr; 2489 iph->daddr = daddr; 2490 iph->version = 0; 2491 err = ipmr_cache_unresolved(mrt, vif, skb2, dev); 2492 rcu_read_unlock(); 2493 return err; 2494 } 2495 2496 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm); 2497 rcu_read_unlock(); 2498 return err; 2499 } 2500 2501 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2502 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2503 int flags) 2504 { 2505 struct nlmsghdr *nlh; 2506 struct rtmsg *rtm; 2507 int err; 2508 2509 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2510 if (!nlh) 2511 return -EMSGSIZE; 2512 2513 rtm = nlmsg_data(nlh); 2514 rtm->rtm_family = RTNL_FAMILY_IPMR; 2515 rtm->rtm_dst_len = 32; 2516 rtm->rtm_src_len = 32; 2517 rtm->rtm_tos = 0; 2518 rtm->rtm_table = mrt->id; 2519 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2520 goto nla_put_failure; 2521 rtm->rtm_type = RTN_MULTICAST; 2522 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2523 if (c->_c.mfc_flags & MFC_STATIC) 2524 rtm->rtm_protocol = RTPROT_STATIC; 2525 else 2526 rtm->rtm_protocol = RTPROT_MROUTED; 2527 rtm->rtm_flags = 0; 2528 2529 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) || 2530 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp)) 2531 goto nla_put_failure; 2532 err = mr_fill_mroute(mrt, skb, &c->_c, rtm); 2533 /* do not break the dump if cache is unresolved */ 2534 if (err < 0 && err != -ENOENT) 2535 goto nla_put_failure; 2536 2537 nlmsg_end(skb, nlh); 2538 return 0; 2539 2540 nla_put_failure: 2541 nlmsg_cancel(skb, nlh); 2542 return -EMSGSIZE; 2543 } 2544 2545 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2546 u32 portid, u32 seq, struct mr_mfc *c, int cmd, 2547 int flags) 2548 { 2549 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c, 2550 cmd, flags); 2551 } 2552 2553 static size_t mroute_msgsize(bool unresolved) 2554 { 2555 size_t len = 2556 NLMSG_ALIGN(sizeof(struct rtmsg)) 2557 + nla_total_size(4) /* RTA_TABLE */ 2558 + nla_total_size(4) /* RTA_SRC */ 2559 + nla_total_size(4) /* RTA_DST */ 2560 ; 2561 2562 if (!unresolved) 2563 len = len 2564 + nla_total_size(4) /* RTA_IIF */ 2565 + nla_total_size(0) /* RTA_MULTIPATH */ 2566 + MAXVIFS * NLA_ALIGN(sizeof(struct rtnexthop)) 2567 /* RTA_MFC_STATS */ 2568 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2569 ; 2570 2571 return len; 2572 } 2573 2574 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2575 int cmd) 2576 { 2577 struct net *net = read_pnet(&mrt->net); 2578 struct sk_buff *skb; 2579 int err = -ENOBUFS; 2580 2581 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS), 2582 GFP_ATOMIC); 2583 if (!skb) 2584 goto errout; 2585 2586 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2587 if (err < 0) 2588 goto errout; 2589 2590 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2591 return; 2592 2593 errout: 2594 kfree_skb(skb); 2595 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2596 } 2597 2598 static size_t igmpmsg_netlink_msgsize(size_t payloadlen) 2599 { 2600 size_t len = 2601 NLMSG_ALIGN(sizeof(struct rtgenmsg)) 2602 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */ 2603 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */ 2604 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */ 2605 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */ 2606 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */ 2607 /* IPMRA_CREPORT_PKT */ 2608 + nla_total_size(payloadlen) 2609 ; 2610 2611 return len; 2612 } 2613 2614 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt) 2615 { 2616 struct net *net = read_pnet(&mrt->net); 2617 struct nlmsghdr *nlh; 2618 struct rtgenmsg *rtgenm; 2619 struct igmpmsg *msg; 2620 struct sk_buff *skb; 2621 struct nlattr *nla; 2622 int payloadlen; 2623 2624 payloadlen = pkt->len - sizeof(struct igmpmsg); 2625 msg = (struct igmpmsg *)skb_network_header(pkt); 2626 2627 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC); 2628 if (!skb) 2629 goto errout; 2630 2631 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT, 2632 sizeof(struct rtgenmsg), 0); 2633 if (!nlh) 2634 goto errout; 2635 rtgenm = nlmsg_data(nlh); 2636 rtgenm->rtgen_family = RTNL_FAMILY_IPMR; 2637 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) || 2638 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) || 2639 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR, 2640 msg->im_src.s_addr) || 2641 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR, 2642 msg->im_dst.s_addr) || 2643 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id)) 2644 goto nla_put_failure; 2645 2646 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen); 2647 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg), 2648 nla_data(nla), payloadlen)) 2649 goto nla_put_failure; 2650 2651 nlmsg_end(skb, nlh); 2652 2653 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC); 2654 return; 2655 2656 nla_put_failure: 2657 nlmsg_cancel(skb, nlh); 2658 errout: 2659 kfree_skb(skb); 2660 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS); 2661 } 2662 2663 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb, 2664 const struct nlmsghdr *nlh, 2665 struct nlattr **tb, 2666 struct netlink_ext_ack *extack) 2667 { 2668 struct rtmsg *rtm; 2669 int i, err; 2670 2671 rtm = nlmsg_payload(nlh, sizeof(*rtm)); 2672 if (!rtm) { 2673 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request"); 2674 return -EINVAL; 2675 } 2676 2677 if (!netlink_strict_get_check(skb)) 2678 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, 2679 rtm_ipv4_policy, extack); 2680 2681 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) || 2682 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) || 2683 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol || 2684 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) { 2685 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request"); 2686 return -EINVAL; 2687 } 2688 2689 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, 2690 rtm_ipv4_policy, extack); 2691 if (err) 2692 return err; 2693 2694 if ((tb[RTA_SRC] && !rtm->rtm_src_len) || 2695 (tb[RTA_DST] && !rtm->rtm_dst_len)) { 2696 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4"); 2697 return -EINVAL; 2698 } 2699 2700 for (i = 0; i <= RTA_MAX; i++) { 2701 if (!tb[i]) 2702 continue; 2703 2704 switch (i) { 2705 case RTA_SRC: 2706 case RTA_DST: 2707 case RTA_TABLE: 2708 break; 2709 default: 2710 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request"); 2711 return -EINVAL; 2712 } 2713 } 2714 2715 return 0; 2716 } 2717 2718 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, 2719 struct netlink_ext_ack *extack) 2720 { 2721 struct net *net = sock_net(in_skb->sk); 2722 struct nlattr *tb[RTA_MAX + 1]; 2723 struct mfc_cache *cache; 2724 struct mr_table *mrt; 2725 struct sk_buff *skb; 2726 __be32 src, grp; 2727 u32 tableid; 2728 int err; 2729 2730 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack); 2731 if (err < 0) 2732 goto errout; 2733 2734 src = nla_get_in_addr_default(tb[RTA_SRC], 0); 2735 grp = nla_get_in_addr_default(tb[RTA_DST], 0); 2736 tableid = nla_get_u32_default(tb[RTA_TABLE], 0); 2737 2738 skb = nlmsg_new(mroute_msgsize(false), GFP_KERNEL); 2739 if (!skb) { 2740 err = -ENOBUFS; 2741 goto errout; 2742 } 2743 2744 rcu_read_lock(); 2745 2746 mrt = __ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT); 2747 if (!mrt) { 2748 err = -ENOENT; 2749 goto errout_unlock; 2750 } 2751 2752 cache = ipmr_cache_find(mrt, src, grp); 2753 if (!cache) { 2754 err = -ENOENT; 2755 goto errout_unlock; 2756 } 2757 2758 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid, 2759 nlh->nlmsg_seq, cache, 2760 RTM_NEWROUTE, 0); 2761 if (err < 0) 2762 goto errout_unlock; 2763 2764 rcu_read_unlock(); 2765 2766 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 2767 errout: 2768 return err; 2769 2770 errout_unlock: 2771 rcu_read_unlock(); 2772 kfree_skb(skb); 2773 goto errout; 2774 } 2775 2776 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2777 { 2778 struct fib_dump_filter filter = { 2779 .rtnl_held = false, 2780 }; 2781 int err; 2782 2783 rcu_read_lock(); 2784 2785 if (cb->strict_check) { 2786 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh, 2787 &filter, cb); 2788 if (err < 0) 2789 goto out; 2790 } 2791 2792 if (filter.table_id) { 2793 struct mr_table *mrt; 2794 2795 mrt = __ipmr_get_table(sock_net(skb->sk), filter.table_id); 2796 if (!mrt) { 2797 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR) { 2798 err = skb->len; 2799 goto out; 2800 } 2801 2802 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist"); 2803 err = -ENOENT; 2804 goto out; 2805 } 2806 2807 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute, 2808 &mfc_unres_lock, &filter); 2809 err = skb->len ? : err; 2810 goto out; 2811 } 2812 2813 err = mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter, 2814 _ipmr_fill_mroute, &mfc_unres_lock, &filter); 2815 out: 2816 rcu_read_unlock(); 2817 2818 return err; 2819 } 2820 2821 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = { 2822 [RTA_SRC] = { .type = NLA_U32 }, 2823 [RTA_DST] = { .type = NLA_U32 }, 2824 [RTA_IIF] = { .type = NLA_U32 }, 2825 [RTA_TABLE] = { .type = NLA_U32 }, 2826 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 2827 }; 2828 2829 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol) 2830 { 2831 switch (rtm_protocol) { 2832 case RTPROT_STATIC: 2833 case RTPROT_MROUTED: 2834 return true; 2835 } 2836 return false; 2837 } 2838 2839 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc) 2840 { 2841 struct rtnexthop *rtnh = nla_data(nla); 2842 int remaining = nla_len(nla), vifi = 0; 2843 2844 while (rtnh_ok(rtnh, remaining)) { 2845 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops; 2846 if (++vifi == MAXVIFS) 2847 break; 2848 rtnh = rtnh_next(rtnh, &remaining); 2849 } 2850 2851 return remaining > 0 ? -EINVAL : vifi; 2852 } 2853 2854 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */ 2855 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh, 2856 struct mfcctl *mfcc, int *mrtsock, 2857 struct mr_table **mrtret, 2858 struct netlink_ext_ack *extack) 2859 { 2860 struct net_device *dev = NULL; 2861 u32 tblid = RT_TABLE_DEFAULT; 2862 int ret, rem, iif = 0; 2863 struct mr_table *mrt; 2864 struct nlattr *attr; 2865 struct rtmsg *rtm; 2866 2867 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX, 2868 rtm_ipmr_policy, extack); 2869 if (ret < 0) 2870 goto out; 2871 rtm = nlmsg_data(nlh); 2872 2873 ret = -EINVAL; 2874 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 || 2875 rtm->rtm_type != RTN_MULTICAST || 2876 rtm->rtm_scope != RT_SCOPE_UNIVERSE || 2877 !ipmr_rtm_validate_proto(rtm->rtm_protocol)) 2878 goto out; 2879 2880 memset(mfcc, 0, sizeof(*mfcc)); 2881 mfcc->mfcc_parent = -1; 2882 ret = 0; 2883 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) { 2884 switch (nla_type(attr)) { 2885 case RTA_SRC: 2886 mfcc->mfcc_origin.s_addr = nla_get_be32(attr); 2887 break; 2888 case RTA_DST: 2889 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr); 2890 break; 2891 case RTA_IIF: 2892 iif = nla_get_u32(attr); 2893 break; 2894 case RTA_MULTIPATH: 2895 if (ipmr_nla_get_ttls(attr, mfcc) < 0) { 2896 ret = -EINVAL; 2897 goto out; 2898 } 2899 break; 2900 case RTA_PREFSRC: 2901 ret = 1; 2902 break; 2903 case RTA_TABLE: 2904 tblid = nla_get_u32(attr); 2905 break; 2906 } 2907 } 2908 2909 rcu_read_lock(); 2910 2911 mrt = __ipmr_get_table(net, tblid); 2912 if (!mrt) { 2913 ret = -ENOENT; 2914 goto unlock; 2915 } 2916 2917 if (iif) { 2918 dev = dev_get_by_index_rcu(net, iif); 2919 if (!dev) { 2920 ret = -ENODEV; 2921 goto unlock; 2922 } 2923 2924 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev); 2925 } 2926 2927 *mrtret = mrt; 2928 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0; 2929 2930 unlock: 2931 rcu_read_unlock(); 2932 out: 2933 return ret; 2934 } 2935 2936 /* takes care of both newroute and delroute */ 2937 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh, 2938 struct netlink_ext_ack *extack) 2939 { 2940 struct net *net = sock_net(skb->sk); 2941 int ret, mrtsock = 0, parent; 2942 struct mr_table *tbl = NULL; 2943 struct mfcctl mfcc; 2944 2945 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack); 2946 if (ret < 0) 2947 return ret; 2948 2949 parent = ret ? mfcc.mfcc_parent : -1; 2950 2951 mutex_lock(&net->ipv4.mfc_mutex); 2952 2953 if (nlh->nlmsg_type == RTM_NEWROUTE) 2954 ret = ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent); 2955 else 2956 ret = ipmr_mfc_delete(tbl, &mfcc, parent); 2957 2958 mutex_unlock(&net->ipv4.mfc_mutex); 2959 2960 return ret; 2961 } 2962 2963 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb) 2964 { 2965 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len); 2966 2967 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) || 2968 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) || 2969 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM, 2970 READ_ONCE(mrt->mroute_reg_vif_num)) || 2971 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT, 2972 READ_ONCE(mrt->mroute_do_assert)) || 2973 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, 2974 READ_ONCE(mrt->mroute_do_pim)) || 2975 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE, 2976 READ_ONCE(mrt->mroute_do_wrvifwhole))) 2977 return false; 2978 2979 return true; 2980 } 2981 2982 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb) 2983 { 2984 struct net_device *vif_dev; 2985 struct nlattr *vif_nest; 2986 struct vif_device *vif; 2987 2988 vif = &mrt->vif_table[vifid]; 2989 vif_dev = vif_dev_read(vif); 2990 /* if the VIF doesn't exist just continue */ 2991 if (!vif_dev) 2992 return true; 2993 2994 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF); 2995 if (!vif_nest) 2996 return false; 2997 2998 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, READ_ONCE(vif_dev->ifindex)) || 2999 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) || 3000 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) || 3001 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, READ_ONCE(vif->bytes_in), 3002 IPMRA_VIFA_PAD) || 3003 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, READ_ONCE(vif->bytes_out), 3004 IPMRA_VIFA_PAD) || 3005 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, READ_ONCE(vif->pkt_in), 3006 IPMRA_VIFA_PAD) || 3007 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, READ_ONCE(vif->pkt_out), 3008 IPMRA_VIFA_PAD) || 3009 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) || 3010 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) { 3011 nla_nest_cancel(skb, vif_nest); 3012 return false; 3013 } 3014 nla_nest_end(skb, vif_nest); 3015 3016 return true; 3017 } 3018 3019 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh, 3020 struct netlink_ext_ack *extack) 3021 { 3022 struct ifinfomsg *ifm; 3023 3024 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 3025 if (!ifm) { 3026 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump"); 3027 return -EINVAL; 3028 } 3029 3030 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 3031 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump"); 3032 return -EINVAL; 3033 } 3034 3035 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 3036 ifm->ifi_change || ifm->ifi_index) { 3037 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request"); 3038 return -EINVAL; 3039 } 3040 3041 return 0; 3042 } 3043 3044 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb) 3045 { 3046 struct net *net = sock_net(skb->sk); 3047 struct nlmsghdr *nlh = NULL; 3048 unsigned int t = 0, s_t; 3049 unsigned int e = 0, s_e; 3050 struct mr_table *mrt; 3051 3052 if (cb->strict_check) { 3053 int err = ipmr_valid_dumplink(cb->nlh, cb->extack); 3054 3055 if (err < 0) 3056 return err; 3057 } 3058 3059 s_t = cb->args[0]; 3060 s_e = cb->args[1]; 3061 3062 rcu_read_lock(); 3063 3064 ipmr_for_each_table(mrt, net) { 3065 struct nlattr *vifs, *af; 3066 struct ifinfomsg *hdr; 3067 u32 i; 3068 3069 if (t < s_t) 3070 goto skip_table; 3071 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, 3072 cb->nlh->nlmsg_seq, RTM_NEWLINK, 3073 sizeof(*hdr), NLM_F_MULTI); 3074 if (!nlh) 3075 break; 3076 3077 hdr = nlmsg_data(nlh); 3078 memset(hdr, 0, sizeof(*hdr)); 3079 hdr->ifi_family = RTNL_FAMILY_IPMR; 3080 3081 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC); 3082 if (!af) { 3083 nlmsg_cancel(skb, nlh); 3084 goto out; 3085 } 3086 3087 if (!ipmr_fill_table(mrt, skb)) { 3088 nlmsg_cancel(skb, nlh); 3089 goto out; 3090 } 3091 3092 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS); 3093 if (!vifs) { 3094 nla_nest_end(skb, af); 3095 nlmsg_end(skb, nlh); 3096 goto out; 3097 } 3098 for (i = 0; i < READ_ONCE(mrt->maxvif); i++) { 3099 if (e < s_e) 3100 goto skip_entry; 3101 if (!ipmr_fill_vif(mrt, i, skb)) { 3102 nla_nest_end(skb, vifs); 3103 nla_nest_end(skb, af); 3104 nlmsg_end(skb, nlh); 3105 goto out; 3106 } 3107 skip_entry: 3108 e++; 3109 } 3110 s_e = 0; 3111 e = 0; 3112 nla_nest_end(skb, vifs); 3113 nla_nest_end(skb, af); 3114 nlmsg_end(skb, nlh); 3115 skip_table: 3116 t++; 3117 } 3118 3119 out: 3120 rcu_read_unlock(); 3121 3122 cb->args[1] = e; 3123 cb->args[0] = t; 3124 3125 return skb->len; 3126 } 3127 3128 #ifdef CONFIG_PROC_FS 3129 /* The /proc interfaces to multicast routing : 3130 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 3131 */ 3132 3133 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 3134 __acquires(RCU) 3135 { 3136 struct mr_vif_iter *iter = seq->private; 3137 struct net *net = seq_file_net(seq); 3138 struct mr_table *mrt; 3139 3140 rcu_read_lock(); 3141 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT); 3142 if (!mrt) { 3143 rcu_read_unlock(); 3144 return ERR_PTR(-ENOENT); 3145 } 3146 3147 iter->mrt = mrt; 3148 3149 return mr_vif_seq_start(seq, pos); 3150 } 3151 3152 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 3153 __releases(RCU) 3154 { 3155 rcu_read_unlock(); 3156 } 3157 3158 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 3159 { 3160 struct mr_vif_iter *iter = seq->private; 3161 struct mr_table *mrt = iter->mrt; 3162 3163 if (v == SEQ_START_TOKEN) { 3164 seq_puts(seq, 3165 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 3166 } else { 3167 const struct vif_device *vif = v; 3168 const struct net_device *vif_dev; 3169 const char *name; 3170 3171 vif_dev = vif_dev_read(vif); 3172 name = vif_dev ? vif_dev->name : "none"; 3173 seq_printf(seq, 3174 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 3175 vif - mrt->vif_table, 3176 name, vif->bytes_in, vif->pkt_in, 3177 vif->bytes_out, vif->pkt_out, 3178 vif->flags, vif->local, vif->remote); 3179 } 3180 return 0; 3181 } 3182 3183 static const struct seq_operations ipmr_vif_seq_ops = { 3184 .start = ipmr_vif_seq_start, 3185 .next = mr_vif_seq_next, 3186 .stop = ipmr_vif_seq_stop, 3187 .show = ipmr_vif_seq_show, 3188 }; 3189 3190 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 3191 { 3192 struct net *net = seq_file_net(seq); 3193 struct mr_table *mrt; 3194 3195 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 3196 if (!mrt) 3197 return ERR_PTR(-ENOENT); 3198 3199 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock); 3200 } 3201 3202 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 3203 { 3204 int n; 3205 3206 if (v == SEQ_START_TOKEN) { 3207 seq_puts(seq, 3208 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 3209 } else { 3210 const struct mfc_cache *mfc = v; 3211 const struct mr_mfc_iter *it = seq->private; 3212 const struct mr_table *mrt = it->mrt; 3213 3214 seq_printf(seq, "%08X %08X %-3hd", 3215 (__force u32) mfc->mfc_mcastgrp, 3216 (__force u32) mfc->mfc_origin, 3217 mfc->_c.mfc_parent); 3218 3219 if (it->cache != &mrt->mfc_unres_queue) { 3220 seq_printf(seq, " %8lu %8lu %8lu", 3221 atomic_long_read(&mfc->_c.mfc_un.res.pkt), 3222 atomic_long_read(&mfc->_c.mfc_un.res.bytes), 3223 atomic_long_read(&mfc->_c.mfc_un.res.wrong_if)); 3224 for (n = mfc->_c.mfc_un.res.minvif; 3225 n < mfc->_c.mfc_un.res.maxvif; n++) { 3226 if (VIF_EXISTS(mrt, n) && 3227 mfc->_c.mfc_un.res.ttls[n] < 255) 3228 seq_printf(seq, 3229 " %2d:%-3d", 3230 n, mfc->_c.mfc_un.res.ttls[n]); 3231 } 3232 } else { 3233 /* unresolved mfc_caches don't contain 3234 * pkt, bytes and wrong_if values 3235 */ 3236 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 3237 } 3238 seq_putc(seq, '\n'); 3239 } 3240 return 0; 3241 } 3242 3243 static const struct seq_operations ipmr_mfc_seq_ops = { 3244 .start = ipmr_mfc_seq_start, 3245 .next = mr_mfc_seq_next, 3246 .stop = mr_mfc_seq_stop, 3247 .show = ipmr_mfc_seq_show, 3248 }; 3249 #endif 3250 3251 #ifdef CONFIG_IP_PIMSM_V2 3252 static const struct net_protocol pim_protocol = { 3253 .handler = pim_rcv, 3254 }; 3255 #endif 3256 3257 static unsigned int ipmr_seq_read(const struct net *net) 3258 { 3259 return atomic_read(&net->ipv4.ipmr_seq) + ipmr_rules_seq_read(net); 3260 } 3261 3262 static int ipmr_dump(struct net *net, struct notifier_block *nb, 3263 struct netlink_ext_ack *extack) 3264 { 3265 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump, 3266 ipmr_mr_table_iter, extack); 3267 } 3268 3269 static const struct fib_notifier_ops ipmr_notifier_ops_template = { 3270 .family = RTNL_FAMILY_IPMR, 3271 .fib_seq_read = ipmr_seq_read, 3272 .fib_dump = ipmr_dump, 3273 .owner = THIS_MODULE, 3274 }; 3275 3276 static int __net_init ipmr_notifier_init(struct net *net) 3277 { 3278 struct fib_notifier_ops *ops; 3279 3280 atomic_set(&net->ipv4.ipmr_seq, 0); 3281 3282 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net); 3283 if (IS_ERR(ops)) 3284 return PTR_ERR(ops); 3285 net->ipv4.ipmr_notifier_ops = ops; 3286 3287 return 0; 3288 } 3289 3290 static void __net_exit ipmr_notifier_exit(struct net *net) 3291 { 3292 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops); 3293 net->ipv4.ipmr_notifier_ops = NULL; 3294 } 3295 3296 /* Setup for IP multicast routing */ 3297 static int __net_init ipmr_net_init(struct net *net) 3298 { 3299 LIST_HEAD(dev_kill_list); 3300 int err; 3301 3302 mutex_init(&net->ipv4.mfc_mutex); 3303 3304 err = ipmr_notifier_init(net); 3305 if (err) 3306 goto ipmr_notifier_fail; 3307 3308 err = ipmr_rules_init(net); 3309 if (err < 0) 3310 goto ipmr_rules_fail; 3311 3312 #ifdef CONFIG_PROC_FS 3313 err = -ENOMEM; 3314 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops, 3315 sizeof(struct mr_vif_iter))) 3316 goto proc_vif_fail; 3317 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops, 3318 sizeof(struct mr_mfc_iter))) 3319 goto proc_cache_fail; 3320 #endif 3321 return 0; 3322 3323 #ifdef CONFIG_PROC_FS 3324 proc_cache_fail: 3325 remove_proc_entry("ip_mr_vif", net->proc_net); 3326 proc_vif_fail: 3327 ipmr_rules_exit_rtnl(net, &dev_kill_list); 3328 ipmr_rules_exit(net); 3329 #endif 3330 ipmr_rules_fail: 3331 ipmr_notifier_exit(net); 3332 ipmr_notifier_fail: 3333 return err; 3334 } 3335 3336 static void __net_exit ipmr_net_exit(struct net *net) 3337 { 3338 #ifdef CONFIG_PROC_FS 3339 remove_proc_entry("ip_mr_cache", net->proc_net); 3340 remove_proc_entry("ip_mr_vif", net->proc_net); 3341 #endif 3342 ipmr_rules_exit(net); 3343 ipmr_notifier_exit(net); 3344 } 3345 3346 static void __net_exit ipmr_net_exit_rtnl(struct net *net, 3347 struct list_head *dev_kill_list) 3348 { 3349 ipmr_rules_exit_rtnl(net, dev_kill_list); 3350 } 3351 3352 static struct pernet_operations ipmr_net_ops = { 3353 .init = ipmr_net_init, 3354 .exit = ipmr_net_exit, 3355 .exit_rtnl = ipmr_net_exit_rtnl, 3356 }; 3357 3358 static const struct rtnl_msg_handler ipmr_rtnl_msg_handlers[] __initconst = { 3359 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETLINK, 3360 .dumpit = ipmr_rtm_dumplink, .flags = RTNL_FLAG_DUMP_UNLOCKED}, 3361 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_NEWROUTE, 3362 .doit = ipmr_rtm_route, .flags = RTNL_FLAG_DOIT_UNLOCKED}, 3363 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_DELROUTE, 3364 .doit = ipmr_rtm_route, .flags = RTNL_FLAG_DOIT_UNLOCKED}, 3365 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETROUTE, 3366 .doit = ipmr_rtm_getroute, .dumpit = ipmr_rtm_dumproute, 3367 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 3368 }; 3369 3370 int __init ip_mr_init(void) 3371 { 3372 int err; 3373 3374 mrt_cachep = KMEM_CACHE(mfc_cache, SLAB_HWCACHE_ALIGN | SLAB_PANIC); 3375 3376 err = register_pernet_subsys(&ipmr_net_ops); 3377 if (err) 3378 goto reg_pernet_fail; 3379 3380 err = register_netdevice_notifier(&ip_mr_notifier); 3381 if (err) 3382 goto reg_notif_fail; 3383 #ifdef CONFIG_IP_PIMSM_V2 3384 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 3385 pr_err("%s: can't add PIM protocol\n", __func__); 3386 err = -EAGAIN; 3387 goto add_proto_fail; 3388 } 3389 #endif 3390 rtnl_register_many(ipmr_rtnl_msg_handlers); 3391 3392 return 0; 3393 3394 #ifdef CONFIG_IP_PIMSM_V2 3395 add_proto_fail: 3396 unregister_netdevice_notifier(&ip_mr_notifier); 3397 #endif 3398 reg_notif_fail: 3399 unregister_pernet_subsys(&ipmr_net_ops); 3400 reg_pernet_fail: 3401 kmem_cache_destroy(mrt_cachep); 3402 return err; 3403 } 3404