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