1 /* 2 * IP multicast routing support for mrouted 3.6/3.8 3 * 4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk> 5 * Linux Consultancy and Custom Driver Development 6 * 7 * This program is free software; you can redistribute it and/or 8 * modify it under the terms of the GNU General Public License 9 * as published by the Free Software Foundation; either version 10 * 2 of the License, or (at your option) any later version. 11 * 12 * Fixes: 13 * Michael Chastain : Incorrect size of copying. 14 * Alan Cox : Added the cache manager code 15 * Alan Cox : Fixed the clone/copy bug and device race. 16 * Mike McLagan : Routing by source 17 * Malcolm Beattie : Buffer handling fixes. 18 * Alexey Kuznetsov : Double buffer free and other fixes. 19 * SVR Anand : Fixed several multicast bugs and problems. 20 * Alexey Kuznetsov : Status, optimisations and more. 21 * Brad Parker : Better behaviour on mrouted upcall 22 * overflow. 23 * Carlos Picoto : PIMv1 Support 24 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header 25 * Relax this requirement to work with older peers. 26 * 27 */ 28 29 #include <linux/uaccess.h> 30 #include <linux/types.h> 31 #include <linux/capability.h> 32 #include <linux/errno.h> 33 #include <linux/timer.h> 34 #include <linux/mm.h> 35 #include <linux/kernel.h> 36 #include <linux/fcntl.h> 37 #include <linux/stat.h> 38 #include <linux/socket.h> 39 #include <linux/in.h> 40 #include <linux/inet.h> 41 #include <linux/netdevice.h> 42 #include <linux/inetdevice.h> 43 #include <linux/igmp.h> 44 #include <linux/proc_fs.h> 45 #include <linux/seq_file.h> 46 #include <linux/mroute.h> 47 #include <linux/init.h> 48 #include <linux/if_ether.h> 49 #include <linux/slab.h> 50 #include <net/net_namespace.h> 51 #include <net/ip.h> 52 #include <net/protocol.h> 53 #include <linux/skbuff.h> 54 #include <net/route.h> 55 #include <net/sock.h> 56 #include <net/icmp.h> 57 #include <net/udp.h> 58 #include <net/raw.h> 59 #include <linux/notifier.h> 60 #include <linux/if_arp.h> 61 #include <linux/netfilter_ipv4.h> 62 #include <linux/compat.h> 63 #include <linux/export.h> 64 #include <net/ip_tunnels.h> 65 #include <net/checksum.h> 66 #include <net/netlink.h> 67 #include <net/fib_rules.h> 68 #include <linux/netconf.h> 69 #include <net/nexthop.h> 70 71 struct ipmr_rule { 72 struct fib_rule common; 73 }; 74 75 struct ipmr_result { 76 struct mr_table *mrt; 77 }; 78 79 /* Big lock, protecting vif table, mrt cache and mroute socket state. 80 * Note that the changes are semaphored via rtnl_lock. 81 */ 82 83 static DEFINE_RWLOCK(mrt_lock); 84 85 /* Multicast router control variables */ 86 87 /* Special spinlock for queue of unresolved entries */ 88 static DEFINE_SPINLOCK(mfc_unres_lock); 89 90 /* We return to original Alan's scheme. Hash table of resolved 91 * entries is changed only in process context and protected 92 * with weak lock mrt_lock. Queue of unresolved entries is protected 93 * with strong spinlock mfc_unres_lock. 94 * 95 * In this case data path is free of exclusive locks at all. 96 */ 97 98 static struct kmem_cache *mrt_cachep __read_mostly; 99 100 static struct mr_table *ipmr_new_table(struct net *net, u32 id); 101 static void ipmr_free_table(struct mr_table *mrt); 102 103 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 104 struct sk_buff *skb, struct mfc_cache *cache, 105 int local); 106 static int ipmr_cache_report(struct mr_table *mrt, 107 struct sk_buff *pkt, vifi_t vifi, int assert); 108 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 109 struct mfc_cache *c, struct rtmsg *rtm); 110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 111 int cmd); 112 static void mroute_clean_tables(struct mr_table *mrt, bool all); 113 static void ipmr_expire_process(unsigned long arg); 114 115 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 116 #define ipmr_for_each_table(mrt, net) \ 117 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list) 118 119 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 120 { 121 struct mr_table *mrt; 122 123 ipmr_for_each_table(mrt, net) { 124 if (mrt->id == id) 125 return mrt; 126 } 127 return NULL; 128 } 129 130 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 131 struct mr_table **mrt) 132 { 133 int err; 134 struct ipmr_result res; 135 struct fib_lookup_arg arg = { 136 .result = &res, 137 .flags = FIB_LOOKUP_NOREF, 138 }; 139 140 /* update flow if oif or iif point to device enslaved to l3mdev */ 141 l3mdev_update_flow(net, flowi4_to_flowi(flp4)); 142 143 err = fib_rules_lookup(net->ipv4.mr_rules_ops, 144 flowi4_to_flowi(flp4), 0, &arg); 145 if (err < 0) 146 return err; 147 *mrt = res.mrt; 148 return 0; 149 } 150 151 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp, 152 int flags, struct fib_lookup_arg *arg) 153 { 154 struct ipmr_result *res = arg->result; 155 struct mr_table *mrt; 156 157 switch (rule->action) { 158 case FR_ACT_TO_TBL: 159 break; 160 case FR_ACT_UNREACHABLE: 161 return -ENETUNREACH; 162 case FR_ACT_PROHIBIT: 163 return -EACCES; 164 case FR_ACT_BLACKHOLE: 165 default: 166 return -EINVAL; 167 } 168 169 arg->table = fib_rule_get_table(rule, arg); 170 171 mrt = ipmr_get_table(rule->fr_net, arg->table); 172 if (!mrt) 173 return -EAGAIN; 174 res->mrt = mrt; 175 return 0; 176 } 177 178 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 179 { 180 return 1; 181 } 182 183 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = { 184 FRA_GENERIC_POLICY, 185 }; 186 187 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 188 struct fib_rule_hdr *frh, struct nlattr **tb) 189 { 190 return 0; 191 } 192 193 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 194 struct nlattr **tb) 195 { 196 return 1; 197 } 198 199 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 200 struct fib_rule_hdr *frh) 201 { 202 frh->dst_len = 0; 203 frh->src_len = 0; 204 frh->tos = 0; 205 return 0; 206 } 207 208 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = { 209 .family = RTNL_FAMILY_IPMR, 210 .rule_size = sizeof(struct ipmr_rule), 211 .addr_size = sizeof(u32), 212 .action = ipmr_rule_action, 213 .match = ipmr_rule_match, 214 .configure = ipmr_rule_configure, 215 .compare = ipmr_rule_compare, 216 .fill = ipmr_rule_fill, 217 .nlgroup = RTNLGRP_IPV4_RULE, 218 .policy = ipmr_rule_policy, 219 .owner = THIS_MODULE, 220 }; 221 222 static int __net_init ipmr_rules_init(struct net *net) 223 { 224 struct fib_rules_ops *ops; 225 struct mr_table *mrt; 226 int err; 227 228 ops = fib_rules_register(&ipmr_rules_ops_template, net); 229 if (IS_ERR(ops)) 230 return PTR_ERR(ops); 231 232 INIT_LIST_HEAD(&net->ipv4.mr_tables); 233 234 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 235 if (IS_ERR(mrt)) { 236 err = PTR_ERR(mrt); 237 goto err1; 238 } 239 240 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0); 241 if (err < 0) 242 goto err2; 243 244 net->ipv4.mr_rules_ops = ops; 245 return 0; 246 247 err2: 248 ipmr_free_table(mrt); 249 err1: 250 fib_rules_unregister(ops); 251 return err; 252 } 253 254 static void __net_exit ipmr_rules_exit(struct net *net) 255 { 256 struct mr_table *mrt, *next; 257 258 rtnl_lock(); 259 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) { 260 list_del(&mrt->list); 261 ipmr_free_table(mrt); 262 } 263 fib_rules_unregister(net->ipv4.mr_rules_ops); 264 rtnl_unlock(); 265 } 266 #else 267 #define ipmr_for_each_table(mrt, net) \ 268 for (mrt = net->ipv4.mrt; mrt; mrt = NULL) 269 270 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 271 { 272 return net->ipv4.mrt; 273 } 274 275 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 276 struct mr_table **mrt) 277 { 278 *mrt = net->ipv4.mrt; 279 return 0; 280 } 281 282 static int __net_init ipmr_rules_init(struct net *net) 283 { 284 struct mr_table *mrt; 285 286 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 287 if (IS_ERR(mrt)) 288 return PTR_ERR(mrt); 289 net->ipv4.mrt = mrt; 290 return 0; 291 } 292 293 static void __net_exit ipmr_rules_exit(struct net *net) 294 { 295 rtnl_lock(); 296 ipmr_free_table(net->ipv4.mrt); 297 net->ipv4.mrt = NULL; 298 rtnl_unlock(); 299 } 300 #endif 301 302 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg, 303 const void *ptr) 304 { 305 const struct mfc_cache_cmp_arg *cmparg = arg->key; 306 struct mfc_cache *c = (struct mfc_cache *)ptr; 307 308 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp || 309 cmparg->mfc_origin != c->mfc_origin; 310 } 311 312 static const struct rhashtable_params ipmr_rht_params = { 313 .head_offset = offsetof(struct mfc_cache, mnode), 314 .key_offset = offsetof(struct mfc_cache, cmparg), 315 .key_len = sizeof(struct mfc_cache_cmp_arg), 316 .nelem_hint = 3, 317 .locks_mul = 1, 318 .obj_cmpfn = ipmr_hash_cmp, 319 .automatic_shrinking = true, 320 }; 321 322 static struct mr_table *ipmr_new_table(struct net *net, u32 id) 323 { 324 struct mr_table *mrt; 325 326 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */ 327 if (id != RT_TABLE_DEFAULT && id >= 1000000000) 328 return ERR_PTR(-EINVAL); 329 330 mrt = ipmr_get_table(net, id); 331 if (mrt) 332 return mrt; 333 334 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 335 if (!mrt) 336 return ERR_PTR(-ENOMEM); 337 write_pnet(&mrt->net, net); 338 mrt->id = id; 339 340 rhltable_init(&mrt->mfc_hash, &ipmr_rht_params); 341 INIT_LIST_HEAD(&mrt->mfc_cache_list); 342 INIT_LIST_HEAD(&mrt->mfc_unres_queue); 343 344 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 345 (unsigned long)mrt); 346 347 mrt->mroute_reg_vif_num = -1; 348 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 349 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables); 350 #endif 351 return mrt; 352 } 353 354 static void ipmr_free_table(struct mr_table *mrt) 355 { 356 del_timer_sync(&mrt->ipmr_expire_timer); 357 mroute_clean_tables(mrt, true); 358 rhltable_destroy(&mrt->mfc_hash); 359 kfree(mrt); 360 } 361 362 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ 363 364 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v) 365 { 366 struct net *net = dev_net(dev); 367 368 dev_close(dev); 369 370 dev = __dev_get_by_name(net, "tunl0"); 371 if (dev) { 372 const struct net_device_ops *ops = dev->netdev_ops; 373 struct ifreq ifr; 374 struct ip_tunnel_parm p; 375 376 memset(&p, 0, sizeof(p)); 377 p.iph.daddr = v->vifc_rmt_addr.s_addr; 378 p.iph.saddr = v->vifc_lcl_addr.s_addr; 379 p.iph.version = 4; 380 p.iph.ihl = 5; 381 p.iph.protocol = IPPROTO_IPIP; 382 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 383 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 384 385 if (ops->ndo_do_ioctl) { 386 mm_segment_t oldfs = get_fs(); 387 388 set_fs(KERNEL_DS); 389 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL); 390 set_fs(oldfs); 391 } 392 } 393 } 394 395 /* Initialize ipmr pimreg/tunnel in_device */ 396 static bool ipmr_init_vif_indev(const struct net_device *dev) 397 { 398 struct in_device *in_dev; 399 400 ASSERT_RTNL(); 401 402 in_dev = __in_dev_get_rtnl(dev); 403 if (!in_dev) 404 return false; 405 ipv4_devconf_setall(in_dev); 406 neigh_parms_data_state_setall(in_dev->arp_parms); 407 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 408 409 return true; 410 } 411 412 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v) 413 { 414 struct net_device *dev; 415 416 dev = __dev_get_by_name(net, "tunl0"); 417 418 if (dev) { 419 const struct net_device_ops *ops = dev->netdev_ops; 420 int err; 421 struct ifreq ifr; 422 struct ip_tunnel_parm p; 423 424 memset(&p, 0, sizeof(p)); 425 p.iph.daddr = v->vifc_rmt_addr.s_addr; 426 p.iph.saddr = v->vifc_lcl_addr.s_addr; 427 p.iph.version = 4; 428 p.iph.ihl = 5; 429 p.iph.protocol = IPPROTO_IPIP; 430 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 431 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 432 433 if (ops->ndo_do_ioctl) { 434 mm_segment_t oldfs = get_fs(); 435 436 set_fs(KERNEL_DS); 437 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL); 438 set_fs(oldfs); 439 } else { 440 err = -EOPNOTSUPP; 441 } 442 dev = NULL; 443 444 if (err == 0 && 445 (dev = __dev_get_by_name(net, p.name)) != NULL) { 446 dev->flags |= IFF_MULTICAST; 447 if (!ipmr_init_vif_indev(dev)) 448 goto failure; 449 if (dev_open(dev)) 450 goto failure; 451 dev_hold(dev); 452 } 453 } 454 return dev; 455 456 failure: 457 unregister_netdevice(dev); 458 return NULL; 459 } 460 461 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 462 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) 463 { 464 struct net *net = dev_net(dev); 465 struct mr_table *mrt; 466 struct flowi4 fl4 = { 467 .flowi4_oif = dev->ifindex, 468 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 469 .flowi4_mark = skb->mark, 470 }; 471 int err; 472 473 err = ipmr_fib_lookup(net, &fl4, &mrt); 474 if (err < 0) { 475 kfree_skb(skb); 476 return err; 477 } 478 479 read_lock(&mrt_lock); 480 dev->stats.tx_bytes += skb->len; 481 dev->stats.tx_packets++; 482 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT); 483 read_unlock(&mrt_lock); 484 kfree_skb(skb); 485 return NETDEV_TX_OK; 486 } 487 488 static int reg_vif_get_iflink(const struct net_device *dev) 489 { 490 return 0; 491 } 492 493 static const struct net_device_ops reg_vif_netdev_ops = { 494 .ndo_start_xmit = reg_vif_xmit, 495 .ndo_get_iflink = reg_vif_get_iflink, 496 }; 497 498 static void reg_vif_setup(struct net_device *dev) 499 { 500 dev->type = ARPHRD_PIMREG; 501 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; 502 dev->flags = IFF_NOARP; 503 dev->netdev_ops = ®_vif_netdev_ops; 504 dev->destructor = free_netdev; 505 dev->features |= NETIF_F_NETNS_LOCAL; 506 } 507 508 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 509 { 510 struct net_device *dev; 511 char name[IFNAMSIZ]; 512 513 if (mrt->id == RT_TABLE_DEFAULT) 514 sprintf(name, "pimreg"); 515 else 516 sprintf(name, "pimreg%u", mrt->id); 517 518 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 519 520 if (!dev) 521 return NULL; 522 523 dev_net_set(dev, net); 524 525 if (register_netdevice(dev)) { 526 free_netdev(dev); 527 return NULL; 528 } 529 530 if (!ipmr_init_vif_indev(dev)) 531 goto failure; 532 if (dev_open(dev)) 533 goto failure; 534 535 dev_hold(dev); 536 537 return dev; 538 539 failure: 540 unregister_netdevice(dev); 541 return NULL; 542 } 543 544 /* called with rcu_read_lock() */ 545 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb, 546 unsigned int pimlen) 547 { 548 struct net_device *reg_dev = NULL; 549 struct iphdr *encap; 550 551 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); 552 /* Check that: 553 * a. packet is really sent to a multicast group 554 * b. packet is not a NULL-REGISTER 555 * c. packet is not truncated 556 */ 557 if (!ipv4_is_multicast(encap->daddr) || 558 encap->tot_len == 0 || 559 ntohs(encap->tot_len) + pimlen > skb->len) 560 return 1; 561 562 read_lock(&mrt_lock); 563 if (mrt->mroute_reg_vif_num >= 0) 564 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev; 565 read_unlock(&mrt_lock); 566 567 if (!reg_dev) 568 return 1; 569 570 skb->mac_header = skb->network_header; 571 skb_pull(skb, (u8 *)encap - skb->data); 572 skb_reset_network_header(skb); 573 skb->protocol = htons(ETH_P_IP); 574 skb->ip_summed = CHECKSUM_NONE; 575 576 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 577 578 netif_rx(skb); 579 580 return NET_RX_SUCCESS; 581 } 582 #else 583 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 584 { 585 return NULL; 586 } 587 #endif 588 589 /** 590 * vif_delete - Delete a VIF entry 591 * @notify: Set to 1, if the caller is a notifier_call 592 */ 593 static int vif_delete(struct mr_table *mrt, int vifi, int notify, 594 struct list_head *head) 595 { 596 struct vif_device *v; 597 struct net_device *dev; 598 struct in_device *in_dev; 599 600 if (vifi < 0 || vifi >= mrt->maxvif) 601 return -EADDRNOTAVAIL; 602 603 v = &mrt->vif_table[vifi]; 604 605 write_lock_bh(&mrt_lock); 606 dev = v->dev; 607 v->dev = NULL; 608 609 if (!dev) { 610 write_unlock_bh(&mrt_lock); 611 return -EADDRNOTAVAIL; 612 } 613 614 if (vifi == mrt->mroute_reg_vif_num) 615 mrt->mroute_reg_vif_num = -1; 616 617 if (vifi + 1 == mrt->maxvif) { 618 int tmp; 619 620 for (tmp = vifi - 1; tmp >= 0; tmp--) { 621 if (VIF_EXISTS(mrt, tmp)) 622 break; 623 } 624 mrt->maxvif = tmp+1; 625 } 626 627 write_unlock_bh(&mrt_lock); 628 629 dev_set_allmulti(dev, -1); 630 631 in_dev = __in_dev_get_rtnl(dev); 632 if (in_dev) { 633 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; 634 inet_netconf_notify_devconf(dev_net(dev), 635 NETCONFA_MC_FORWARDING, 636 dev->ifindex, &in_dev->cnf); 637 ip_rt_multicast_event(in_dev); 638 } 639 640 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify) 641 unregister_netdevice_queue(dev, head); 642 643 dev_put(dev); 644 return 0; 645 } 646 647 static void ipmr_cache_free_rcu(struct rcu_head *head) 648 { 649 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu); 650 651 kmem_cache_free(mrt_cachep, c); 652 } 653 654 static inline void ipmr_cache_free(struct mfc_cache *c) 655 { 656 call_rcu(&c->rcu, ipmr_cache_free_rcu); 657 } 658 659 /* Destroy an unresolved cache entry, killing queued skbs 660 * and reporting error to netlink readers. 661 */ 662 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c) 663 { 664 struct net *net = read_pnet(&mrt->net); 665 struct sk_buff *skb; 666 struct nlmsgerr *e; 667 668 atomic_dec(&mrt->cache_resolve_queue_len); 669 670 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) { 671 if (ip_hdr(skb)->version == 0) { 672 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); 673 nlh->nlmsg_type = NLMSG_ERROR; 674 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 675 skb_trim(skb, nlh->nlmsg_len); 676 e = nlmsg_data(nlh); 677 e->error = -ETIMEDOUT; 678 memset(&e->msg, 0, sizeof(e->msg)); 679 680 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 681 } else { 682 kfree_skb(skb); 683 } 684 } 685 686 ipmr_cache_free(c); 687 } 688 689 /* Timer process for the unresolved queue. */ 690 static void ipmr_expire_process(unsigned long arg) 691 { 692 struct mr_table *mrt = (struct mr_table *)arg; 693 unsigned long now; 694 unsigned long expires; 695 struct mfc_cache *c, *next; 696 697 if (!spin_trylock(&mfc_unres_lock)) { 698 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 699 return; 700 } 701 702 if (list_empty(&mrt->mfc_unres_queue)) 703 goto out; 704 705 now = jiffies; 706 expires = 10*HZ; 707 708 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 709 if (time_after(c->mfc_un.unres.expires, now)) { 710 unsigned long interval = c->mfc_un.unres.expires - now; 711 if (interval < expires) 712 expires = interval; 713 continue; 714 } 715 716 list_del(&c->list); 717 mroute_netlink_event(mrt, c, RTM_DELROUTE); 718 ipmr_destroy_unres(mrt, c); 719 } 720 721 if (!list_empty(&mrt->mfc_unres_queue)) 722 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 723 724 out: 725 spin_unlock(&mfc_unres_lock); 726 } 727 728 /* Fill oifs list. It is called under write locked mrt_lock. */ 729 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache, 730 unsigned char *ttls) 731 { 732 int vifi; 733 734 cache->mfc_un.res.minvif = MAXVIFS; 735 cache->mfc_un.res.maxvif = 0; 736 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 737 738 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 739 if (VIF_EXISTS(mrt, vifi) && 740 ttls[vifi] && ttls[vifi] < 255) { 741 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 742 if (cache->mfc_un.res.minvif > vifi) 743 cache->mfc_un.res.minvif = vifi; 744 if (cache->mfc_un.res.maxvif <= vifi) 745 cache->mfc_un.res.maxvif = vifi + 1; 746 } 747 } 748 cache->mfc_un.res.lastuse = jiffies; 749 } 750 751 static int vif_add(struct net *net, struct mr_table *mrt, 752 struct vifctl *vifc, int mrtsock) 753 { 754 int vifi = vifc->vifc_vifi; 755 struct vif_device *v = &mrt->vif_table[vifi]; 756 struct net_device *dev; 757 struct in_device *in_dev; 758 int err; 759 760 /* Is vif busy ? */ 761 if (VIF_EXISTS(mrt, vifi)) 762 return -EADDRINUSE; 763 764 switch (vifc->vifc_flags) { 765 case VIFF_REGISTER: 766 if (!ipmr_pimsm_enabled()) 767 return -EINVAL; 768 /* Special Purpose VIF in PIM 769 * All the packets will be sent to the daemon 770 */ 771 if (mrt->mroute_reg_vif_num >= 0) 772 return -EADDRINUSE; 773 dev = ipmr_reg_vif(net, mrt); 774 if (!dev) 775 return -ENOBUFS; 776 err = dev_set_allmulti(dev, 1); 777 if (err) { 778 unregister_netdevice(dev); 779 dev_put(dev); 780 return err; 781 } 782 break; 783 case VIFF_TUNNEL: 784 dev = ipmr_new_tunnel(net, vifc); 785 if (!dev) 786 return -ENOBUFS; 787 err = dev_set_allmulti(dev, 1); 788 if (err) { 789 ipmr_del_tunnel(dev, vifc); 790 dev_put(dev); 791 return err; 792 } 793 break; 794 case VIFF_USE_IFINDEX: 795 case 0: 796 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 797 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 798 if (dev && !__in_dev_get_rtnl(dev)) { 799 dev_put(dev); 800 return -EADDRNOTAVAIL; 801 } 802 } else { 803 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 804 } 805 if (!dev) 806 return -EADDRNOTAVAIL; 807 err = dev_set_allmulti(dev, 1); 808 if (err) { 809 dev_put(dev); 810 return err; 811 } 812 break; 813 default: 814 return -EINVAL; 815 } 816 817 in_dev = __in_dev_get_rtnl(dev); 818 if (!in_dev) { 819 dev_put(dev); 820 return -EADDRNOTAVAIL; 821 } 822 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 823 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex, 824 &in_dev->cnf); 825 ip_rt_multicast_event(in_dev); 826 827 /* Fill in the VIF structures */ 828 829 v->rate_limit = vifc->vifc_rate_limit; 830 v->local = vifc->vifc_lcl_addr.s_addr; 831 v->remote = vifc->vifc_rmt_addr.s_addr; 832 v->flags = vifc->vifc_flags; 833 if (!mrtsock) 834 v->flags |= VIFF_STATIC; 835 v->threshold = vifc->vifc_threshold; 836 v->bytes_in = 0; 837 v->bytes_out = 0; 838 v->pkt_in = 0; 839 v->pkt_out = 0; 840 v->link = dev->ifindex; 841 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER)) 842 v->link = dev_get_iflink(dev); 843 844 /* And finish update writing critical data */ 845 write_lock_bh(&mrt_lock); 846 v->dev = dev; 847 if (v->flags & VIFF_REGISTER) 848 mrt->mroute_reg_vif_num = vifi; 849 if (vifi+1 > mrt->maxvif) 850 mrt->maxvif = vifi+1; 851 write_unlock_bh(&mrt_lock); 852 return 0; 853 } 854 855 /* called with rcu_read_lock() */ 856 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 857 __be32 origin, 858 __be32 mcastgrp) 859 { 860 struct mfc_cache_cmp_arg arg = { 861 .mfc_mcastgrp = mcastgrp, 862 .mfc_origin = origin 863 }; 864 struct rhlist_head *tmp, *list; 865 struct mfc_cache *c; 866 867 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 868 rhl_for_each_entry_rcu(c, tmp, list, mnode) 869 return c; 870 871 return NULL; 872 } 873 874 /* Look for a (*,*,oif) entry */ 875 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt, 876 int vifi) 877 { 878 struct mfc_cache_cmp_arg arg = { 879 .mfc_mcastgrp = htonl(INADDR_ANY), 880 .mfc_origin = htonl(INADDR_ANY) 881 }; 882 struct rhlist_head *tmp, *list; 883 struct mfc_cache *c; 884 885 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 886 rhl_for_each_entry_rcu(c, tmp, list, mnode) 887 if (c->mfc_un.res.ttls[vifi] < 255) 888 return c; 889 890 return NULL; 891 } 892 893 /* Look for a (*,G) entry */ 894 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 895 __be32 mcastgrp, int vifi) 896 { 897 struct mfc_cache_cmp_arg arg = { 898 .mfc_mcastgrp = mcastgrp, 899 .mfc_origin = htonl(INADDR_ANY) 900 }; 901 struct rhlist_head *tmp, *list; 902 struct mfc_cache *c, *proxy; 903 904 if (mcastgrp == htonl(INADDR_ANY)) 905 goto skip; 906 907 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 908 rhl_for_each_entry_rcu(c, tmp, list, mnode) { 909 if (c->mfc_un.res.ttls[vifi] < 255) 910 return c; 911 912 /* It's ok if the vifi is part of the static tree */ 913 proxy = ipmr_cache_find_any_parent(mrt, c->mfc_parent); 914 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255) 915 return c; 916 } 917 918 skip: 919 return ipmr_cache_find_any_parent(mrt, vifi); 920 } 921 922 /* Look for a (S,G,iif) entry if parent != -1 */ 923 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt, 924 __be32 origin, __be32 mcastgrp, 925 int parent) 926 { 927 struct mfc_cache_cmp_arg arg = { 928 .mfc_mcastgrp = mcastgrp, 929 .mfc_origin = origin, 930 }; 931 struct rhlist_head *tmp, *list; 932 struct mfc_cache *c; 933 934 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 935 rhl_for_each_entry_rcu(c, tmp, list, mnode) 936 if (parent == -1 || parent == c->mfc_parent) 937 return c; 938 939 return NULL; 940 } 941 942 /* Allocate a multicast cache entry */ 943 static struct mfc_cache *ipmr_cache_alloc(void) 944 { 945 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 946 947 if (c) { 948 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 949 c->mfc_un.res.minvif = MAXVIFS; 950 } 951 return c; 952 } 953 954 static struct mfc_cache *ipmr_cache_alloc_unres(void) 955 { 956 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 957 958 if (c) { 959 skb_queue_head_init(&c->mfc_un.unres.unresolved); 960 c->mfc_un.unres.expires = jiffies + 10*HZ; 961 } 962 return c; 963 } 964 965 /* A cache entry has gone into a resolved state from queued */ 966 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 967 struct mfc_cache *uc, struct mfc_cache *c) 968 { 969 struct sk_buff *skb; 970 struct nlmsgerr *e; 971 972 /* Play the pending entries through our router */ 973 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 974 if (ip_hdr(skb)->version == 0) { 975 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); 976 977 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 978 nlh->nlmsg_len = skb_tail_pointer(skb) - 979 (u8 *)nlh; 980 } else { 981 nlh->nlmsg_type = NLMSG_ERROR; 982 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 983 skb_trim(skb, nlh->nlmsg_len); 984 e = nlmsg_data(nlh); 985 e->error = -EMSGSIZE; 986 memset(&e->msg, 0, sizeof(e->msg)); 987 } 988 989 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 990 } else { 991 ip_mr_forward(net, mrt, skb, c, 0); 992 } 993 } 994 } 995 996 /* Bounce a cache query up to mrouted. We could use netlink for this but mrouted 997 * expects the following bizarre scheme. 998 * 999 * Called under mrt_lock. 1000 */ 1001 static int ipmr_cache_report(struct mr_table *mrt, 1002 struct sk_buff *pkt, vifi_t vifi, int assert) 1003 { 1004 const int ihl = ip_hdrlen(pkt); 1005 struct sock *mroute_sk; 1006 struct igmphdr *igmp; 1007 struct igmpmsg *msg; 1008 struct sk_buff *skb; 1009 int ret; 1010 1011 if (assert == IGMPMSG_WHOLEPKT) 1012 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 1013 else 1014 skb = alloc_skb(128, GFP_ATOMIC); 1015 1016 if (!skb) 1017 return -ENOBUFS; 1018 1019 if (assert == IGMPMSG_WHOLEPKT) { 1020 /* Ugly, but we have no choice with this interface. 1021 * Duplicate old header, fix ihl, length etc. 1022 * And all this only to mangle msg->im_msgtype and 1023 * to set msg->im_mbz to "mbz" :-) 1024 */ 1025 skb_push(skb, sizeof(struct iphdr)); 1026 skb_reset_network_header(skb); 1027 skb_reset_transport_header(skb); 1028 msg = (struct igmpmsg *)skb_network_header(skb); 1029 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 1030 msg->im_msgtype = IGMPMSG_WHOLEPKT; 1031 msg->im_mbz = 0; 1032 msg->im_vif = mrt->mroute_reg_vif_num; 1033 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 1034 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 1035 sizeof(struct iphdr)); 1036 } else { 1037 /* Copy the IP header */ 1038 skb_set_network_header(skb, skb->len); 1039 skb_put(skb, ihl); 1040 skb_copy_to_linear_data(skb, pkt->data, ihl); 1041 /* Flag to the kernel this is a route add */ 1042 ip_hdr(skb)->protocol = 0; 1043 msg = (struct igmpmsg *)skb_network_header(skb); 1044 msg->im_vif = vifi; 1045 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1046 /* Add our header */ 1047 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr)); 1048 igmp->type = assert; 1049 msg->im_msgtype = assert; 1050 igmp->code = 0; 1051 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1052 skb->transport_header = skb->network_header; 1053 } 1054 1055 rcu_read_lock(); 1056 mroute_sk = rcu_dereference(mrt->mroute_sk); 1057 if (!mroute_sk) { 1058 rcu_read_unlock(); 1059 kfree_skb(skb); 1060 return -EINVAL; 1061 } 1062 1063 /* Deliver to mrouted */ 1064 ret = sock_queue_rcv_skb(mroute_sk, skb); 1065 rcu_read_unlock(); 1066 if (ret < 0) { 1067 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1068 kfree_skb(skb); 1069 } 1070 1071 return ret; 1072 } 1073 1074 /* Queue a packet for resolution. It gets locked cache entry! */ 1075 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, 1076 struct sk_buff *skb) 1077 { 1078 const struct iphdr *iph = ip_hdr(skb); 1079 struct mfc_cache *c; 1080 bool found = false; 1081 int err; 1082 1083 spin_lock_bh(&mfc_unres_lock); 1084 list_for_each_entry(c, &mrt->mfc_unres_queue, list) { 1085 if (c->mfc_mcastgrp == iph->daddr && 1086 c->mfc_origin == iph->saddr) { 1087 found = true; 1088 break; 1089 } 1090 } 1091 1092 if (!found) { 1093 /* Create a new entry if allowable */ 1094 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1095 (c = ipmr_cache_alloc_unres()) == NULL) { 1096 spin_unlock_bh(&mfc_unres_lock); 1097 1098 kfree_skb(skb); 1099 return -ENOBUFS; 1100 } 1101 1102 /* Fill in the new cache entry */ 1103 c->mfc_parent = -1; 1104 c->mfc_origin = iph->saddr; 1105 c->mfc_mcastgrp = iph->daddr; 1106 1107 /* Reflect first query at mrouted. */ 1108 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1109 if (err < 0) { 1110 /* If the report failed throw the cache entry 1111 out - Brad Parker 1112 */ 1113 spin_unlock_bh(&mfc_unres_lock); 1114 1115 ipmr_cache_free(c); 1116 kfree_skb(skb); 1117 return err; 1118 } 1119 1120 atomic_inc(&mrt->cache_resolve_queue_len); 1121 list_add(&c->list, &mrt->mfc_unres_queue); 1122 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1123 1124 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1125 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires); 1126 } 1127 1128 /* See if we can append the packet */ 1129 if (c->mfc_un.unres.unresolved.qlen > 3) { 1130 kfree_skb(skb); 1131 err = -ENOBUFS; 1132 } else { 1133 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1134 err = 0; 1135 } 1136 1137 spin_unlock_bh(&mfc_unres_lock); 1138 return err; 1139 } 1140 1141 /* MFC cache manipulation by user space mroute daemon */ 1142 1143 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1144 { 1145 struct mfc_cache *c; 1146 1147 /* The entries are added/deleted only under RTNL */ 1148 rcu_read_lock(); 1149 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1150 mfc->mfcc_mcastgrp.s_addr, parent); 1151 rcu_read_unlock(); 1152 if (!c) 1153 return -ENOENT; 1154 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1155 list_del_rcu(&c->list); 1156 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1157 ipmr_cache_free(c); 1158 1159 return 0; 1160 } 1161 1162 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1163 struct mfcctl *mfc, int mrtsock, int parent) 1164 { 1165 struct mfc_cache *uc, *c; 1166 bool found; 1167 int ret; 1168 1169 if (mfc->mfcc_parent >= MAXVIFS) 1170 return -ENFILE; 1171 1172 /* The entries are added/deleted only under RTNL */ 1173 rcu_read_lock(); 1174 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1175 mfc->mfcc_mcastgrp.s_addr, parent); 1176 rcu_read_unlock(); 1177 if (c) { 1178 write_lock_bh(&mrt_lock); 1179 c->mfc_parent = mfc->mfcc_parent; 1180 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1181 if (!mrtsock) 1182 c->mfc_flags |= MFC_STATIC; 1183 write_unlock_bh(&mrt_lock); 1184 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1185 return 0; 1186 } 1187 1188 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1189 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1190 return -EINVAL; 1191 1192 c = ipmr_cache_alloc(); 1193 if (!c) 1194 return -ENOMEM; 1195 1196 c->mfc_origin = mfc->mfcc_origin.s_addr; 1197 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1198 c->mfc_parent = mfc->mfcc_parent; 1199 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1200 if (!mrtsock) 1201 c->mfc_flags |= MFC_STATIC; 1202 1203 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->mnode, 1204 ipmr_rht_params); 1205 if (ret) { 1206 pr_err("ipmr: rhtable insert error %d\n", ret); 1207 ipmr_cache_free(c); 1208 return ret; 1209 } 1210 list_add_tail_rcu(&c->list, &mrt->mfc_cache_list); 1211 /* Check to see if we resolved a queued list. If so we 1212 * need to send on the frames and tidy up. 1213 */ 1214 found = false; 1215 spin_lock_bh(&mfc_unres_lock); 1216 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) { 1217 if (uc->mfc_origin == c->mfc_origin && 1218 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1219 list_del(&uc->list); 1220 atomic_dec(&mrt->cache_resolve_queue_len); 1221 found = true; 1222 break; 1223 } 1224 } 1225 if (list_empty(&mrt->mfc_unres_queue)) 1226 del_timer(&mrt->ipmr_expire_timer); 1227 spin_unlock_bh(&mfc_unres_lock); 1228 1229 if (found) { 1230 ipmr_cache_resolve(net, mrt, uc, c); 1231 ipmr_cache_free(uc); 1232 } 1233 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1234 return 0; 1235 } 1236 1237 /* Close the multicast socket, and clear the vif tables etc */ 1238 static void mroute_clean_tables(struct mr_table *mrt, bool all) 1239 { 1240 struct mfc_cache *c, *tmp; 1241 LIST_HEAD(list); 1242 int i; 1243 1244 /* Shut down all active vif entries */ 1245 for (i = 0; i < mrt->maxvif; i++) { 1246 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC)) 1247 continue; 1248 vif_delete(mrt, i, 0, &list); 1249 } 1250 unregister_netdevice_many(&list); 1251 1252 /* Wipe the cache */ 1253 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1254 if (!all && (c->mfc_flags & MFC_STATIC)) 1255 continue; 1256 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1257 list_del_rcu(&c->list); 1258 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1259 ipmr_cache_free(c); 1260 } 1261 1262 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1263 spin_lock_bh(&mfc_unres_lock); 1264 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1265 list_del(&c->list); 1266 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1267 ipmr_destroy_unres(mrt, c); 1268 } 1269 spin_unlock_bh(&mfc_unres_lock); 1270 } 1271 } 1272 1273 /* called from ip_ra_control(), before an RCU grace period, 1274 * we dont need to call synchronize_rcu() here 1275 */ 1276 static void mrtsock_destruct(struct sock *sk) 1277 { 1278 struct net *net = sock_net(sk); 1279 struct mr_table *mrt; 1280 1281 ASSERT_RTNL(); 1282 ipmr_for_each_table(mrt, net) { 1283 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1284 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1285 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1286 NETCONFA_IFINDEX_ALL, 1287 net->ipv4.devconf_all); 1288 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1289 mroute_clean_tables(mrt, false); 1290 } 1291 } 1292 } 1293 1294 /* Socket options and virtual interface manipulation. The whole 1295 * virtual interface system is a complete heap, but unfortunately 1296 * that's how BSD mrouted happens to think. Maybe one day with a proper 1297 * MOSPF/PIM router set up we can clean this up. 1298 */ 1299 1300 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, 1301 unsigned int optlen) 1302 { 1303 struct net *net = sock_net(sk); 1304 int val, ret = 0, parent = 0; 1305 struct mr_table *mrt; 1306 struct vifctl vif; 1307 struct mfcctl mfc; 1308 u32 uval; 1309 1310 /* There's one exception to the lock - MRT_DONE which needs to unlock */ 1311 rtnl_lock(); 1312 if (sk->sk_type != SOCK_RAW || 1313 inet_sk(sk)->inet_num != IPPROTO_IGMP) { 1314 ret = -EOPNOTSUPP; 1315 goto out_unlock; 1316 } 1317 1318 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1319 if (!mrt) { 1320 ret = -ENOENT; 1321 goto out_unlock; 1322 } 1323 if (optname != MRT_INIT) { 1324 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1325 !ns_capable(net->user_ns, CAP_NET_ADMIN)) { 1326 ret = -EACCES; 1327 goto out_unlock; 1328 } 1329 } 1330 1331 switch (optname) { 1332 case MRT_INIT: 1333 if (optlen != sizeof(int)) { 1334 ret = -EINVAL; 1335 break; 1336 } 1337 if (rtnl_dereference(mrt->mroute_sk)) { 1338 ret = -EADDRINUSE; 1339 break; 1340 } 1341 1342 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1343 if (ret == 0) { 1344 rcu_assign_pointer(mrt->mroute_sk, sk); 1345 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1346 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1347 NETCONFA_IFINDEX_ALL, 1348 net->ipv4.devconf_all); 1349 } 1350 break; 1351 case MRT_DONE: 1352 if (sk != rcu_access_pointer(mrt->mroute_sk)) { 1353 ret = -EACCES; 1354 } else { 1355 ret = ip_ra_control(sk, 0, NULL); 1356 goto out_unlock; 1357 } 1358 break; 1359 case MRT_ADD_VIF: 1360 case MRT_DEL_VIF: 1361 if (optlen != sizeof(vif)) { 1362 ret = -EINVAL; 1363 break; 1364 } 1365 if (copy_from_user(&vif, optval, sizeof(vif))) { 1366 ret = -EFAULT; 1367 break; 1368 } 1369 if (vif.vifc_vifi >= MAXVIFS) { 1370 ret = -ENFILE; 1371 break; 1372 } 1373 if (optname == MRT_ADD_VIF) { 1374 ret = vif_add(net, mrt, &vif, 1375 sk == rtnl_dereference(mrt->mroute_sk)); 1376 } else { 1377 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1378 } 1379 break; 1380 /* Manipulate the forwarding caches. These live 1381 * in a sort of kernel/user symbiosis. 1382 */ 1383 case MRT_ADD_MFC: 1384 case MRT_DEL_MFC: 1385 parent = -1; 1386 case MRT_ADD_MFC_PROXY: 1387 case MRT_DEL_MFC_PROXY: 1388 if (optlen != sizeof(mfc)) { 1389 ret = -EINVAL; 1390 break; 1391 } 1392 if (copy_from_user(&mfc, optval, sizeof(mfc))) { 1393 ret = -EFAULT; 1394 break; 1395 } 1396 if (parent == 0) 1397 parent = mfc.mfcc_parent; 1398 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1399 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1400 else 1401 ret = ipmr_mfc_add(net, mrt, &mfc, 1402 sk == rtnl_dereference(mrt->mroute_sk), 1403 parent); 1404 break; 1405 /* Control PIM assert. */ 1406 case MRT_ASSERT: 1407 if (optlen != sizeof(val)) { 1408 ret = -EINVAL; 1409 break; 1410 } 1411 if (get_user(val, (int __user *)optval)) { 1412 ret = -EFAULT; 1413 break; 1414 } 1415 mrt->mroute_do_assert = val; 1416 break; 1417 case MRT_PIM: 1418 if (!ipmr_pimsm_enabled()) { 1419 ret = -ENOPROTOOPT; 1420 break; 1421 } 1422 if (optlen != sizeof(val)) { 1423 ret = -EINVAL; 1424 break; 1425 } 1426 if (get_user(val, (int __user *)optval)) { 1427 ret = -EFAULT; 1428 break; 1429 } 1430 1431 val = !!val; 1432 if (val != mrt->mroute_do_pim) { 1433 mrt->mroute_do_pim = val; 1434 mrt->mroute_do_assert = val; 1435 } 1436 break; 1437 case MRT_TABLE: 1438 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) { 1439 ret = -ENOPROTOOPT; 1440 break; 1441 } 1442 if (optlen != sizeof(uval)) { 1443 ret = -EINVAL; 1444 break; 1445 } 1446 if (get_user(uval, (u32 __user *)optval)) { 1447 ret = -EFAULT; 1448 break; 1449 } 1450 1451 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1452 ret = -EBUSY; 1453 } else { 1454 mrt = ipmr_new_table(net, uval); 1455 if (IS_ERR(mrt)) 1456 ret = PTR_ERR(mrt); 1457 else 1458 raw_sk(sk)->ipmr_table = uval; 1459 } 1460 break; 1461 /* Spurious command, or MRT_VERSION which you cannot set. */ 1462 default: 1463 ret = -ENOPROTOOPT; 1464 } 1465 out_unlock: 1466 rtnl_unlock(); 1467 return ret; 1468 } 1469 1470 /* Getsock opt support for the multicast routing system. */ 1471 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen) 1472 { 1473 int olr; 1474 int val; 1475 struct net *net = sock_net(sk); 1476 struct mr_table *mrt; 1477 1478 if (sk->sk_type != SOCK_RAW || 1479 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1480 return -EOPNOTSUPP; 1481 1482 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1483 if (!mrt) 1484 return -ENOENT; 1485 1486 switch (optname) { 1487 case MRT_VERSION: 1488 val = 0x0305; 1489 break; 1490 case MRT_PIM: 1491 if (!ipmr_pimsm_enabled()) 1492 return -ENOPROTOOPT; 1493 val = mrt->mroute_do_pim; 1494 break; 1495 case MRT_ASSERT: 1496 val = mrt->mroute_do_assert; 1497 break; 1498 default: 1499 return -ENOPROTOOPT; 1500 } 1501 1502 if (get_user(olr, optlen)) 1503 return -EFAULT; 1504 olr = min_t(unsigned int, olr, sizeof(int)); 1505 if (olr < 0) 1506 return -EINVAL; 1507 if (put_user(olr, optlen)) 1508 return -EFAULT; 1509 if (copy_to_user(optval, &val, olr)) 1510 return -EFAULT; 1511 return 0; 1512 } 1513 1514 /* The IP multicast ioctl support routines. */ 1515 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) 1516 { 1517 struct sioc_sg_req sr; 1518 struct sioc_vif_req vr; 1519 struct vif_device *vif; 1520 struct mfc_cache *c; 1521 struct net *net = sock_net(sk); 1522 struct mr_table *mrt; 1523 1524 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1525 if (!mrt) 1526 return -ENOENT; 1527 1528 switch (cmd) { 1529 case SIOCGETVIFCNT: 1530 if (copy_from_user(&vr, arg, sizeof(vr))) 1531 return -EFAULT; 1532 if (vr.vifi >= mrt->maxvif) 1533 return -EINVAL; 1534 read_lock(&mrt_lock); 1535 vif = &mrt->vif_table[vr.vifi]; 1536 if (VIF_EXISTS(mrt, vr.vifi)) { 1537 vr.icount = vif->pkt_in; 1538 vr.ocount = vif->pkt_out; 1539 vr.ibytes = vif->bytes_in; 1540 vr.obytes = vif->bytes_out; 1541 read_unlock(&mrt_lock); 1542 1543 if (copy_to_user(arg, &vr, sizeof(vr))) 1544 return -EFAULT; 1545 return 0; 1546 } 1547 read_unlock(&mrt_lock); 1548 return -EADDRNOTAVAIL; 1549 case SIOCGETSGCNT: 1550 if (copy_from_user(&sr, arg, sizeof(sr))) 1551 return -EFAULT; 1552 1553 rcu_read_lock(); 1554 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1555 if (c) { 1556 sr.pktcnt = c->mfc_un.res.pkt; 1557 sr.bytecnt = c->mfc_un.res.bytes; 1558 sr.wrong_if = c->mfc_un.res.wrong_if; 1559 rcu_read_unlock(); 1560 1561 if (copy_to_user(arg, &sr, sizeof(sr))) 1562 return -EFAULT; 1563 return 0; 1564 } 1565 rcu_read_unlock(); 1566 return -EADDRNOTAVAIL; 1567 default: 1568 return -ENOIOCTLCMD; 1569 } 1570 } 1571 1572 #ifdef CONFIG_COMPAT 1573 struct compat_sioc_sg_req { 1574 struct in_addr src; 1575 struct in_addr grp; 1576 compat_ulong_t pktcnt; 1577 compat_ulong_t bytecnt; 1578 compat_ulong_t wrong_if; 1579 }; 1580 1581 struct compat_sioc_vif_req { 1582 vifi_t vifi; /* Which iface */ 1583 compat_ulong_t icount; 1584 compat_ulong_t ocount; 1585 compat_ulong_t ibytes; 1586 compat_ulong_t obytes; 1587 }; 1588 1589 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1590 { 1591 struct compat_sioc_sg_req sr; 1592 struct compat_sioc_vif_req vr; 1593 struct vif_device *vif; 1594 struct mfc_cache *c; 1595 struct net *net = sock_net(sk); 1596 struct mr_table *mrt; 1597 1598 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1599 if (!mrt) 1600 return -ENOENT; 1601 1602 switch (cmd) { 1603 case SIOCGETVIFCNT: 1604 if (copy_from_user(&vr, arg, sizeof(vr))) 1605 return -EFAULT; 1606 if (vr.vifi >= mrt->maxvif) 1607 return -EINVAL; 1608 read_lock(&mrt_lock); 1609 vif = &mrt->vif_table[vr.vifi]; 1610 if (VIF_EXISTS(mrt, vr.vifi)) { 1611 vr.icount = vif->pkt_in; 1612 vr.ocount = vif->pkt_out; 1613 vr.ibytes = vif->bytes_in; 1614 vr.obytes = vif->bytes_out; 1615 read_unlock(&mrt_lock); 1616 1617 if (copy_to_user(arg, &vr, sizeof(vr))) 1618 return -EFAULT; 1619 return 0; 1620 } 1621 read_unlock(&mrt_lock); 1622 return -EADDRNOTAVAIL; 1623 case SIOCGETSGCNT: 1624 if (copy_from_user(&sr, arg, sizeof(sr))) 1625 return -EFAULT; 1626 1627 rcu_read_lock(); 1628 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1629 if (c) { 1630 sr.pktcnt = c->mfc_un.res.pkt; 1631 sr.bytecnt = c->mfc_un.res.bytes; 1632 sr.wrong_if = c->mfc_un.res.wrong_if; 1633 rcu_read_unlock(); 1634 1635 if (copy_to_user(arg, &sr, sizeof(sr))) 1636 return -EFAULT; 1637 return 0; 1638 } 1639 rcu_read_unlock(); 1640 return -EADDRNOTAVAIL; 1641 default: 1642 return -ENOIOCTLCMD; 1643 } 1644 } 1645 #endif 1646 1647 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1648 { 1649 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1650 struct net *net = dev_net(dev); 1651 struct mr_table *mrt; 1652 struct vif_device *v; 1653 int ct; 1654 1655 if (event != NETDEV_UNREGISTER) 1656 return NOTIFY_DONE; 1657 1658 ipmr_for_each_table(mrt, net) { 1659 v = &mrt->vif_table[0]; 1660 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1661 if (v->dev == dev) 1662 vif_delete(mrt, ct, 1, NULL); 1663 } 1664 } 1665 return NOTIFY_DONE; 1666 } 1667 1668 static struct notifier_block ip_mr_notifier = { 1669 .notifier_call = ipmr_device_event, 1670 }; 1671 1672 /* Encapsulate a packet by attaching a valid IPIP header to it. 1673 * This avoids tunnel drivers and other mess and gives us the speed so 1674 * important for multicast video. 1675 */ 1676 static void ip_encap(struct net *net, struct sk_buff *skb, 1677 __be32 saddr, __be32 daddr) 1678 { 1679 struct iphdr *iph; 1680 const struct iphdr *old_iph = ip_hdr(skb); 1681 1682 skb_push(skb, sizeof(struct iphdr)); 1683 skb->transport_header = skb->network_header; 1684 skb_reset_network_header(skb); 1685 iph = ip_hdr(skb); 1686 1687 iph->version = 4; 1688 iph->tos = old_iph->tos; 1689 iph->ttl = old_iph->ttl; 1690 iph->frag_off = 0; 1691 iph->daddr = daddr; 1692 iph->saddr = saddr; 1693 iph->protocol = IPPROTO_IPIP; 1694 iph->ihl = 5; 1695 iph->tot_len = htons(skb->len); 1696 ip_select_ident(net, skb, NULL); 1697 ip_send_check(iph); 1698 1699 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1700 nf_reset(skb); 1701 } 1702 1703 static inline int ipmr_forward_finish(struct net *net, struct sock *sk, 1704 struct sk_buff *skb) 1705 { 1706 struct ip_options *opt = &(IPCB(skb)->opt); 1707 1708 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS); 1709 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len); 1710 1711 if (unlikely(opt->optlen)) 1712 ip_forward_options(skb); 1713 1714 return dst_output(net, sk, skb); 1715 } 1716 1717 /* Processing handlers for ipmr_forward */ 1718 1719 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt, 1720 struct sk_buff *skb, struct mfc_cache *c, int vifi) 1721 { 1722 const struct iphdr *iph = ip_hdr(skb); 1723 struct vif_device *vif = &mrt->vif_table[vifi]; 1724 struct net_device *dev; 1725 struct rtable *rt; 1726 struct flowi4 fl4; 1727 int encap = 0; 1728 1729 if (!vif->dev) 1730 goto out_free; 1731 1732 if (vif->flags & VIFF_REGISTER) { 1733 vif->pkt_out++; 1734 vif->bytes_out += skb->len; 1735 vif->dev->stats.tx_bytes += skb->len; 1736 vif->dev->stats.tx_packets++; 1737 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1738 goto out_free; 1739 } 1740 1741 if (vif->flags & VIFF_TUNNEL) { 1742 rt = ip_route_output_ports(net, &fl4, NULL, 1743 vif->remote, vif->local, 1744 0, 0, 1745 IPPROTO_IPIP, 1746 RT_TOS(iph->tos), vif->link); 1747 if (IS_ERR(rt)) 1748 goto out_free; 1749 encap = sizeof(struct iphdr); 1750 } else { 1751 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1752 0, 0, 1753 IPPROTO_IPIP, 1754 RT_TOS(iph->tos), vif->link); 1755 if (IS_ERR(rt)) 1756 goto out_free; 1757 } 1758 1759 dev = rt->dst.dev; 1760 1761 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1762 /* Do not fragment multicasts. Alas, IPv4 does not 1763 * allow to send ICMP, so that packets will disappear 1764 * to blackhole. 1765 */ 1766 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS); 1767 ip_rt_put(rt); 1768 goto out_free; 1769 } 1770 1771 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len; 1772 1773 if (skb_cow(skb, encap)) { 1774 ip_rt_put(rt); 1775 goto out_free; 1776 } 1777 1778 vif->pkt_out++; 1779 vif->bytes_out += skb->len; 1780 1781 skb_dst_drop(skb); 1782 skb_dst_set(skb, &rt->dst); 1783 ip_decrease_ttl(ip_hdr(skb)); 1784 1785 /* FIXME: forward and output firewalls used to be called here. 1786 * What do we do with netfilter? -- RR 1787 */ 1788 if (vif->flags & VIFF_TUNNEL) { 1789 ip_encap(net, skb, vif->local, vif->remote); 1790 /* FIXME: extra output firewall step used to be here. --RR */ 1791 vif->dev->stats.tx_packets++; 1792 vif->dev->stats.tx_bytes += skb->len; 1793 } 1794 1795 IPCB(skb)->flags |= IPSKB_FORWARDED; 1796 1797 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1798 * not only before forwarding, but after forwarding on all output 1799 * interfaces. It is clear, if mrouter runs a multicasting 1800 * program, it should receive packets not depending to what interface 1801 * program is joined. 1802 * If we will not make it, the program will have to join on all 1803 * interfaces. On the other hand, multihoming host (or router, but 1804 * not mrouter) cannot join to more than one interface - it will 1805 * result in receiving multiple packets. 1806 */ 1807 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, 1808 net, NULL, skb, skb->dev, dev, 1809 ipmr_forward_finish); 1810 return; 1811 1812 out_free: 1813 kfree_skb(skb); 1814 } 1815 1816 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev) 1817 { 1818 int ct; 1819 1820 for (ct = mrt->maxvif-1; ct >= 0; ct--) { 1821 if (mrt->vif_table[ct].dev == dev) 1822 break; 1823 } 1824 return ct; 1825 } 1826 1827 /* "local" means that we should preserve one skb (for local delivery) */ 1828 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 1829 struct sk_buff *skb, struct mfc_cache *cache, 1830 int local) 1831 { 1832 int true_vifi = ipmr_find_vif(mrt, skb->dev); 1833 int psend = -1; 1834 int vif, ct; 1835 1836 vif = cache->mfc_parent; 1837 cache->mfc_un.res.pkt++; 1838 cache->mfc_un.res.bytes += skb->len; 1839 cache->mfc_un.res.lastuse = jiffies; 1840 1841 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 1842 struct mfc_cache *cache_proxy; 1843 1844 /* For an (*,G) entry, we only check that the incomming 1845 * interface is part of the static tree. 1846 */ 1847 cache_proxy = ipmr_cache_find_any_parent(mrt, vif); 1848 if (cache_proxy && 1849 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 1850 goto forward; 1851 } 1852 1853 /* Wrong interface: drop packet and (maybe) send PIM assert. */ 1854 if (mrt->vif_table[vif].dev != skb->dev) { 1855 struct net_device *mdev; 1856 1857 mdev = l3mdev_master_dev_rcu(mrt->vif_table[vif].dev); 1858 if (mdev == skb->dev) 1859 goto forward; 1860 1861 if (rt_is_output_route(skb_rtable(skb))) { 1862 /* It is our own packet, looped back. 1863 * Very complicated situation... 1864 * 1865 * The best workaround until routing daemons will be 1866 * fixed is not to redistribute packet, if it was 1867 * send through wrong interface. It means, that 1868 * multicast applications WILL NOT work for 1869 * (S,G), which have default multicast route pointing 1870 * to wrong oif. In any case, it is not a good 1871 * idea to use multicasting applications on router. 1872 */ 1873 goto dont_forward; 1874 } 1875 1876 cache->mfc_un.res.wrong_if++; 1877 1878 if (true_vifi >= 0 && mrt->mroute_do_assert && 1879 /* pimsm uses asserts, when switching from RPT to SPT, 1880 * so that we cannot check that packet arrived on an oif. 1881 * It is bad, but otherwise we would need to move pretty 1882 * large chunk of pimd to kernel. Ough... --ANK 1883 */ 1884 (mrt->mroute_do_pim || 1885 cache->mfc_un.res.ttls[true_vifi] < 255) && 1886 time_after(jiffies, 1887 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 1888 cache->mfc_un.res.last_assert = jiffies; 1889 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 1890 } 1891 goto dont_forward; 1892 } 1893 1894 forward: 1895 mrt->vif_table[vif].pkt_in++; 1896 mrt->vif_table[vif].bytes_in += skb->len; 1897 1898 /* Forward the frame */ 1899 if (cache->mfc_origin == htonl(INADDR_ANY) && 1900 cache->mfc_mcastgrp == htonl(INADDR_ANY)) { 1901 if (true_vifi >= 0 && 1902 true_vifi != cache->mfc_parent && 1903 ip_hdr(skb)->ttl > 1904 cache->mfc_un.res.ttls[cache->mfc_parent]) { 1905 /* It's an (*,*) entry and the packet is not coming from 1906 * the upstream: forward the packet to the upstream 1907 * only. 1908 */ 1909 psend = cache->mfc_parent; 1910 goto last_forward; 1911 } 1912 goto dont_forward; 1913 } 1914 for (ct = cache->mfc_un.res.maxvif - 1; 1915 ct >= cache->mfc_un.res.minvif; ct--) { 1916 /* For (*,G) entry, don't forward to the incoming interface */ 1917 if ((cache->mfc_origin != htonl(INADDR_ANY) || 1918 ct != true_vifi) && 1919 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) { 1920 if (psend != -1) { 1921 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1922 1923 if (skb2) 1924 ipmr_queue_xmit(net, mrt, skb2, cache, 1925 psend); 1926 } 1927 psend = ct; 1928 } 1929 } 1930 last_forward: 1931 if (psend != -1) { 1932 if (local) { 1933 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1934 1935 if (skb2) 1936 ipmr_queue_xmit(net, mrt, skb2, cache, psend); 1937 } else { 1938 ipmr_queue_xmit(net, mrt, skb, cache, psend); 1939 return; 1940 } 1941 } 1942 1943 dont_forward: 1944 if (!local) 1945 kfree_skb(skb); 1946 } 1947 1948 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 1949 { 1950 struct rtable *rt = skb_rtable(skb); 1951 struct iphdr *iph = ip_hdr(skb); 1952 struct flowi4 fl4 = { 1953 .daddr = iph->daddr, 1954 .saddr = iph->saddr, 1955 .flowi4_tos = RT_TOS(iph->tos), 1956 .flowi4_oif = (rt_is_output_route(rt) ? 1957 skb->dev->ifindex : 0), 1958 .flowi4_iif = (rt_is_output_route(rt) ? 1959 LOOPBACK_IFINDEX : 1960 skb->dev->ifindex), 1961 .flowi4_mark = skb->mark, 1962 }; 1963 struct mr_table *mrt; 1964 int err; 1965 1966 err = ipmr_fib_lookup(net, &fl4, &mrt); 1967 if (err) 1968 return ERR_PTR(err); 1969 return mrt; 1970 } 1971 1972 /* Multicast packets for forwarding arrive here 1973 * Called with rcu_read_lock(); 1974 */ 1975 int ip_mr_input(struct sk_buff *skb) 1976 { 1977 struct mfc_cache *cache; 1978 struct net *net = dev_net(skb->dev); 1979 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 1980 struct mr_table *mrt; 1981 1982 /* Packet is looped back after forward, it should not be 1983 * forwarded second time, but still can be delivered locally. 1984 */ 1985 if (IPCB(skb)->flags & IPSKB_FORWARDED) 1986 goto dont_forward; 1987 1988 mrt = ipmr_rt_fib_lookup(net, skb); 1989 if (IS_ERR(mrt)) { 1990 kfree_skb(skb); 1991 return PTR_ERR(mrt); 1992 } 1993 if (!local) { 1994 if (IPCB(skb)->opt.router_alert) { 1995 if (ip_call_ra_chain(skb)) 1996 return 0; 1997 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 1998 /* IGMPv1 (and broken IGMPv2 implementations sort of 1999 * Cisco IOS <= 11.2(8)) do not put router alert 2000 * option to IGMP packets destined to routable 2001 * groups. It is very bad, because it means 2002 * that we can forward NO IGMP messages. 2003 */ 2004 struct sock *mroute_sk; 2005 2006 mroute_sk = rcu_dereference(mrt->mroute_sk); 2007 if (mroute_sk) { 2008 nf_reset(skb); 2009 raw_rcv(mroute_sk, skb); 2010 return 0; 2011 } 2012 } 2013 } 2014 2015 /* already under rcu_read_lock() */ 2016 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 2017 if (!cache) { 2018 int vif = ipmr_find_vif(mrt, skb->dev); 2019 2020 if (vif >= 0) 2021 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2022 vif); 2023 } 2024 2025 /* No usable cache entry */ 2026 if (!cache) { 2027 int vif; 2028 2029 if (local) { 2030 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2031 ip_local_deliver(skb); 2032 if (!skb2) 2033 return -ENOBUFS; 2034 skb = skb2; 2035 } 2036 2037 read_lock(&mrt_lock); 2038 vif = ipmr_find_vif(mrt, skb->dev); 2039 if (vif >= 0) { 2040 int err2 = ipmr_cache_unresolved(mrt, vif, skb); 2041 read_unlock(&mrt_lock); 2042 2043 return err2; 2044 } 2045 read_unlock(&mrt_lock); 2046 kfree_skb(skb); 2047 return -ENODEV; 2048 } 2049 2050 read_lock(&mrt_lock); 2051 ip_mr_forward(net, mrt, skb, cache, local); 2052 read_unlock(&mrt_lock); 2053 2054 if (local) 2055 return ip_local_deliver(skb); 2056 2057 return 0; 2058 2059 dont_forward: 2060 if (local) 2061 return ip_local_deliver(skb); 2062 kfree_skb(skb); 2063 return 0; 2064 } 2065 2066 #ifdef CONFIG_IP_PIMSM_V1 2067 /* Handle IGMP messages of PIMv1 */ 2068 int pim_rcv_v1(struct sk_buff *skb) 2069 { 2070 struct igmphdr *pim; 2071 struct net *net = dev_net(skb->dev); 2072 struct mr_table *mrt; 2073 2074 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2075 goto drop; 2076 2077 pim = igmp_hdr(skb); 2078 2079 mrt = ipmr_rt_fib_lookup(net, skb); 2080 if (IS_ERR(mrt)) 2081 goto drop; 2082 if (!mrt->mroute_do_pim || 2083 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2084 goto drop; 2085 2086 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2087 drop: 2088 kfree_skb(skb); 2089 } 2090 return 0; 2091 } 2092 #endif 2093 2094 #ifdef CONFIG_IP_PIMSM_V2 2095 static int pim_rcv(struct sk_buff *skb) 2096 { 2097 struct pimreghdr *pim; 2098 struct net *net = dev_net(skb->dev); 2099 struct mr_table *mrt; 2100 2101 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2102 goto drop; 2103 2104 pim = (struct pimreghdr *)skb_transport_header(skb); 2105 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) || 2106 (pim->flags & PIM_NULL_REGISTER) || 2107 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2108 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2109 goto drop; 2110 2111 mrt = ipmr_rt_fib_lookup(net, skb); 2112 if (IS_ERR(mrt)) 2113 goto drop; 2114 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2115 drop: 2116 kfree_skb(skb); 2117 } 2118 return 0; 2119 } 2120 #endif 2121 2122 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2123 struct mfc_cache *c, struct rtmsg *rtm) 2124 { 2125 struct rta_mfc_stats mfcs; 2126 struct nlattr *mp_attr; 2127 struct rtnexthop *nhp; 2128 unsigned long lastuse; 2129 int ct; 2130 2131 /* If cache is unresolved, don't try to parse IIF and OIF */ 2132 if (c->mfc_parent >= MAXVIFS) { 2133 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2134 return -ENOENT; 2135 } 2136 2137 if (VIF_EXISTS(mrt, c->mfc_parent) && 2138 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0) 2139 return -EMSGSIZE; 2140 2141 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH))) 2142 return -EMSGSIZE; 2143 2144 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2145 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2146 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) { 2147 nla_nest_cancel(skb, mp_attr); 2148 return -EMSGSIZE; 2149 } 2150 2151 nhp->rtnh_flags = 0; 2152 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2153 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex; 2154 nhp->rtnh_len = sizeof(*nhp); 2155 } 2156 } 2157 2158 nla_nest_end(skb, mp_attr); 2159 2160 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2161 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2162 2163 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2164 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2165 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2166 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2167 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2168 RTA_PAD)) 2169 return -EMSGSIZE; 2170 2171 rtm->rtm_type = RTN_MULTICAST; 2172 return 1; 2173 } 2174 2175 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2176 __be32 saddr, __be32 daddr, 2177 struct rtmsg *rtm, u32 portid) 2178 { 2179 struct mfc_cache *cache; 2180 struct mr_table *mrt; 2181 int err; 2182 2183 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2184 if (!mrt) 2185 return -ENOENT; 2186 2187 rcu_read_lock(); 2188 cache = ipmr_cache_find(mrt, saddr, daddr); 2189 if (!cache && skb->dev) { 2190 int vif = ipmr_find_vif(mrt, skb->dev); 2191 2192 if (vif >= 0) 2193 cache = ipmr_cache_find_any(mrt, daddr, vif); 2194 } 2195 if (!cache) { 2196 struct sk_buff *skb2; 2197 struct iphdr *iph; 2198 struct net_device *dev; 2199 int vif = -1; 2200 2201 dev = skb->dev; 2202 read_lock(&mrt_lock); 2203 if (dev) 2204 vif = ipmr_find_vif(mrt, dev); 2205 if (vif < 0) { 2206 read_unlock(&mrt_lock); 2207 rcu_read_unlock(); 2208 return -ENODEV; 2209 } 2210 skb2 = skb_clone(skb, GFP_ATOMIC); 2211 if (!skb2) { 2212 read_unlock(&mrt_lock); 2213 rcu_read_unlock(); 2214 return -ENOMEM; 2215 } 2216 2217 NETLINK_CB(skb2).portid = portid; 2218 skb_push(skb2, sizeof(struct iphdr)); 2219 skb_reset_network_header(skb2); 2220 iph = ip_hdr(skb2); 2221 iph->ihl = sizeof(struct iphdr) >> 2; 2222 iph->saddr = saddr; 2223 iph->daddr = daddr; 2224 iph->version = 0; 2225 err = ipmr_cache_unresolved(mrt, vif, skb2); 2226 read_unlock(&mrt_lock); 2227 rcu_read_unlock(); 2228 return err; 2229 } 2230 2231 read_lock(&mrt_lock); 2232 err = __ipmr_fill_mroute(mrt, skb, cache, rtm); 2233 read_unlock(&mrt_lock); 2234 rcu_read_unlock(); 2235 return err; 2236 } 2237 2238 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2239 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2240 int flags) 2241 { 2242 struct nlmsghdr *nlh; 2243 struct rtmsg *rtm; 2244 int err; 2245 2246 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2247 if (!nlh) 2248 return -EMSGSIZE; 2249 2250 rtm = nlmsg_data(nlh); 2251 rtm->rtm_family = RTNL_FAMILY_IPMR; 2252 rtm->rtm_dst_len = 32; 2253 rtm->rtm_src_len = 32; 2254 rtm->rtm_tos = 0; 2255 rtm->rtm_table = mrt->id; 2256 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2257 goto nla_put_failure; 2258 rtm->rtm_type = RTN_MULTICAST; 2259 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2260 if (c->mfc_flags & MFC_STATIC) 2261 rtm->rtm_protocol = RTPROT_STATIC; 2262 else 2263 rtm->rtm_protocol = RTPROT_MROUTED; 2264 rtm->rtm_flags = 0; 2265 2266 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) || 2267 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp)) 2268 goto nla_put_failure; 2269 err = __ipmr_fill_mroute(mrt, skb, c, rtm); 2270 /* do not break the dump if cache is unresolved */ 2271 if (err < 0 && err != -ENOENT) 2272 goto nla_put_failure; 2273 2274 nlmsg_end(skb, nlh); 2275 return 0; 2276 2277 nla_put_failure: 2278 nlmsg_cancel(skb, nlh); 2279 return -EMSGSIZE; 2280 } 2281 2282 static size_t mroute_msgsize(bool unresolved, int maxvif) 2283 { 2284 size_t len = 2285 NLMSG_ALIGN(sizeof(struct rtmsg)) 2286 + nla_total_size(4) /* RTA_TABLE */ 2287 + nla_total_size(4) /* RTA_SRC */ 2288 + nla_total_size(4) /* RTA_DST */ 2289 ; 2290 2291 if (!unresolved) 2292 len = len 2293 + nla_total_size(4) /* RTA_IIF */ 2294 + nla_total_size(0) /* RTA_MULTIPATH */ 2295 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2296 /* RTA_MFC_STATS */ 2297 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2298 ; 2299 2300 return len; 2301 } 2302 2303 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2304 int cmd) 2305 { 2306 struct net *net = read_pnet(&mrt->net); 2307 struct sk_buff *skb; 2308 int err = -ENOBUFS; 2309 2310 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif), 2311 GFP_ATOMIC); 2312 if (!skb) 2313 goto errout; 2314 2315 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2316 if (err < 0) 2317 goto errout; 2318 2319 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2320 return; 2321 2322 errout: 2323 kfree_skb(skb); 2324 if (err < 0) 2325 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2326 } 2327 2328 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2329 { 2330 struct net *net = sock_net(skb->sk); 2331 struct mr_table *mrt; 2332 struct mfc_cache *mfc; 2333 unsigned int t = 0, s_t; 2334 unsigned int e = 0, s_e; 2335 2336 s_t = cb->args[0]; 2337 s_e = cb->args[1]; 2338 2339 rcu_read_lock(); 2340 ipmr_for_each_table(mrt, net) { 2341 if (t < s_t) 2342 goto next_table; 2343 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) { 2344 if (e < s_e) 2345 goto next_entry; 2346 if (ipmr_fill_mroute(mrt, skb, 2347 NETLINK_CB(cb->skb).portid, 2348 cb->nlh->nlmsg_seq, 2349 mfc, RTM_NEWROUTE, 2350 NLM_F_MULTI) < 0) 2351 goto done; 2352 next_entry: 2353 e++; 2354 } 2355 e = 0; 2356 s_e = 0; 2357 2358 spin_lock_bh(&mfc_unres_lock); 2359 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) { 2360 if (e < s_e) 2361 goto next_entry2; 2362 if (ipmr_fill_mroute(mrt, skb, 2363 NETLINK_CB(cb->skb).portid, 2364 cb->nlh->nlmsg_seq, 2365 mfc, RTM_NEWROUTE, 2366 NLM_F_MULTI) < 0) { 2367 spin_unlock_bh(&mfc_unres_lock); 2368 goto done; 2369 } 2370 next_entry2: 2371 e++; 2372 } 2373 spin_unlock_bh(&mfc_unres_lock); 2374 e = 0; 2375 s_e = 0; 2376 next_table: 2377 t++; 2378 } 2379 done: 2380 rcu_read_unlock(); 2381 2382 cb->args[1] = e; 2383 cb->args[0] = t; 2384 2385 return skb->len; 2386 } 2387 2388 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = { 2389 [RTA_SRC] = { .type = NLA_U32 }, 2390 [RTA_DST] = { .type = NLA_U32 }, 2391 [RTA_IIF] = { .type = NLA_U32 }, 2392 [RTA_TABLE] = { .type = NLA_U32 }, 2393 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 2394 }; 2395 2396 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol) 2397 { 2398 switch (rtm_protocol) { 2399 case RTPROT_STATIC: 2400 case RTPROT_MROUTED: 2401 return true; 2402 } 2403 return false; 2404 } 2405 2406 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc) 2407 { 2408 struct rtnexthop *rtnh = nla_data(nla); 2409 int remaining = nla_len(nla), vifi = 0; 2410 2411 while (rtnh_ok(rtnh, remaining)) { 2412 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops; 2413 if (++vifi == MAXVIFS) 2414 break; 2415 rtnh = rtnh_next(rtnh, &remaining); 2416 } 2417 2418 return remaining > 0 ? -EINVAL : vifi; 2419 } 2420 2421 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */ 2422 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh, 2423 struct mfcctl *mfcc, int *mrtsock, 2424 struct mr_table **mrtret) 2425 { 2426 struct net_device *dev = NULL; 2427 u32 tblid = RT_TABLE_DEFAULT; 2428 struct mr_table *mrt; 2429 struct nlattr *attr; 2430 struct rtmsg *rtm; 2431 int ret, rem; 2432 2433 ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy); 2434 if (ret < 0) 2435 goto out; 2436 rtm = nlmsg_data(nlh); 2437 2438 ret = -EINVAL; 2439 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 || 2440 rtm->rtm_type != RTN_MULTICAST || 2441 rtm->rtm_scope != RT_SCOPE_UNIVERSE || 2442 !ipmr_rtm_validate_proto(rtm->rtm_protocol)) 2443 goto out; 2444 2445 memset(mfcc, 0, sizeof(*mfcc)); 2446 mfcc->mfcc_parent = -1; 2447 ret = 0; 2448 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) { 2449 switch (nla_type(attr)) { 2450 case RTA_SRC: 2451 mfcc->mfcc_origin.s_addr = nla_get_be32(attr); 2452 break; 2453 case RTA_DST: 2454 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr); 2455 break; 2456 case RTA_IIF: 2457 dev = __dev_get_by_index(net, nla_get_u32(attr)); 2458 if (!dev) { 2459 ret = -ENODEV; 2460 goto out; 2461 } 2462 break; 2463 case RTA_MULTIPATH: 2464 if (ipmr_nla_get_ttls(attr, mfcc) < 0) { 2465 ret = -EINVAL; 2466 goto out; 2467 } 2468 break; 2469 case RTA_PREFSRC: 2470 ret = 1; 2471 break; 2472 case RTA_TABLE: 2473 tblid = nla_get_u32(attr); 2474 break; 2475 } 2476 } 2477 mrt = ipmr_get_table(net, tblid); 2478 if (!mrt) { 2479 ret = -ENOENT; 2480 goto out; 2481 } 2482 *mrtret = mrt; 2483 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0; 2484 if (dev) 2485 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev); 2486 2487 out: 2488 return ret; 2489 } 2490 2491 /* takes care of both newroute and delroute */ 2492 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh) 2493 { 2494 struct net *net = sock_net(skb->sk); 2495 int ret, mrtsock, parent; 2496 struct mr_table *tbl; 2497 struct mfcctl mfcc; 2498 2499 mrtsock = 0; 2500 tbl = NULL; 2501 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl); 2502 if (ret < 0) 2503 return ret; 2504 2505 parent = ret ? mfcc.mfcc_parent : -1; 2506 if (nlh->nlmsg_type == RTM_NEWROUTE) 2507 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent); 2508 else 2509 return ipmr_mfc_delete(tbl, &mfcc, parent); 2510 } 2511 2512 #ifdef CONFIG_PROC_FS 2513 /* The /proc interfaces to multicast routing : 2514 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 2515 */ 2516 struct ipmr_vif_iter { 2517 struct seq_net_private p; 2518 struct mr_table *mrt; 2519 int ct; 2520 }; 2521 2522 static struct vif_device *ipmr_vif_seq_idx(struct net *net, 2523 struct ipmr_vif_iter *iter, 2524 loff_t pos) 2525 { 2526 struct mr_table *mrt = iter->mrt; 2527 2528 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 2529 if (!VIF_EXISTS(mrt, iter->ct)) 2530 continue; 2531 if (pos-- == 0) 2532 return &mrt->vif_table[iter->ct]; 2533 } 2534 return NULL; 2535 } 2536 2537 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 2538 __acquires(mrt_lock) 2539 { 2540 struct ipmr_vif_iter *iter = seq->private; 2541 struct net *net = seq_file_net(seq); 2542 struct mr_table *mrt; 2543 2544 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2545 if (!mrt) 2546 return ERR_PTR(-ENOENT); 2547 2548 iter->mrt = mrt; 2549 2550 read_lock(&mrt_lock); 2551 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1) 2552 : SEQ_START_TOKEN; 2553 } 2554 2555 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2556 { 2557 struct ipmr_vif_iter *iter = seq->private; 2558 struct net *net = seq_file_net(seq); 2559 struct mr_table *mrt = iter->mrt; 2560 2561 ++*pos; 2562 if (v == SEQ_START_TOKEN) 2563 return ipmr_vif_seq_idx(net, iter, 0); 2564 2565 while (++iter->ct < mrt->maxvif) { 2566 if (!VIF_EXISTS(mrt, iter->ct)) 2567 continue; 2568 return &mrt->vif_table[iter->ct]; 2569 } 2570 return NULL; 2571 } 2572 2573 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 2574 __releases(mrt_lock) 2575 { 2576 read_unlock(&mrt_lock); 2577 } 2578 2579 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 2580 { 2581 struct ipmr_vif_iter *iter = seq->private; 2582 struct mr_table *mrt = iter->mrt; 2583 2584 if (v == SEQ_START_TOKEN) { 2585 seq_puts(seq, 2586 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 2587 } else { 2588 const struct vif_device *vif = v; 2589 const char *name = vif->dev ? vif->dev->name : "none"; 2590 2591 seq_printf(seq, 2592 "%2zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 2593 vif - mrt->vif_table, 2594 name, vif->bytes_in, vif->pkt_in, 2595 vif->bytes_out, vif->pkt_out, 2596 vif->flags, vif->local, vif->remote); 2597 } 2598 return 0; 2599 } 2600 2601 static const struct seq_operations ipmr_vif_seq_ops = { 2602 .start = ipmr_vif_seq_start, 2603 .next = ipmr_vif_seq_next, 2604 .stop = ipmr_vif_seq_stop, 2605 .show = ipmr_vif_seq_show, 2606 }; 2607 2608 static int ipmr_vif_open(struct inode *inode, struct file *file) 2609 { 2610 return seq_open_net(inode, file, &ipmr_vif_seq_ops, 2611 sizeof(struct ipmr_vif_iter)); 2612 } 2613 2614 static const struct file_operations ipmr_vif_fops = { 2615 .owner = THIS_MODULE, 2616 .open = ipmr_vif_open, 2617 .read = seq_read, 2618 .llseek = seq_lseek, 2619 .release = seq_release_net, 2620 }; 2621 2622 struct ipmr_mfc_iter { 2623 struct seq_net_private p; 2624 struct mr_table *mrt; 2625 struct list_head *cache; 2626 }; 2627 2628 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net, 2629 struct ipmr_mfc_iter *it, loff_t pos) 2630 { 2631 struct mr_table *mrt = it->mrt; 2632 struct mfc_cache *mfc; 2633 2634 rcu_read_lock(); 2635 it->cache = &mrt->mfc_cache_list; 2636 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) 2637 if (pos-- == 0) 2638 return mfc; 2639 rcu_read_unlock(); 2640 2641 spin_lock_bh(&mfc_unres_lock); 2642 it->cache = &mrt->mfc_unres_queue; 2643 list_for_each_entry(mfc, it->cache, list) 2644 if (pos-- == 0) 2645 return mfc; 2646 spin_unlock_bh(&mfc_unres_lock); 2647 2648 it->cache = NULL; 2649 return NULL; 2650 } 2651 2652 2653 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 2654 { 2655 struct ipmr_mfc_iter *it = seq->private; 2656 struct net *net = seq_file_net(seq); 2657 struct mr_table *mrt; 2658 2659 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2660 if (!mrt) 2661 return ERR_PTR(-ENOENT); 2662 2663 it->mrt = mrt; 2664 it->cache = NULL; 2665 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 2666 : SEQ_START_TOKEN; 2667 } 2668 2669 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2670 { 2671 struct ipmr_mfc_iter *it = seq->private; 2672 struct net *net = seq_file_net(seq); 2673 struct mr_table *mrt = it->mrt; 2674 struct mfc_cache *mfc = v; 2675 2676 ++*pos; 2677 2678 if (v == SEQ_START_TOKEN) 2679 return ipmr_mfc_seq_idx(net, seq->private, 0); 2680 2681 if (mfc->list.next != it->cache) 2682 return list_entry(mfc->list.next, struct mfc_cache, list); 2683 2684 if (it->cache == &mrt->mfc_unres_queue) 2685 goto end_of_list; 2686 2687 /* exhausted cache_array, show unresolved */ 2688 rcu_read_unlock(); 2689 it->cache = &mrt->mfc_unres_queue; 2690 2691 spin_lock_bh(&mfc_unres_lock); 2692 if (!list_empty(it->cache)) 2693 return list_first_entry(it->cache, struct mfc_cache, list); 2694 2695 end_of_list: 2696 spin_unlock_bh(&mfc_unres_lock); 2697 it->cache = NULL; 2698 2699 return NULL; 2700 } 2701 2702 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 2703 { 2704 struct ipmr_mfc_iter *it = seq->private; 2705 struct mr_table *mrt = it->mrt; 2706 2707 if (it->cache == &mrt->mfc_unres_queue) 2708 spin_unlock_bh(&mfc_unres_lock); 2709 else if (it->cache == &mrt->mfc_cache_list) 2710 rcu_read_unlock(); 2711 } 2712 2713 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 2714 { 2715 int n; 2716 2717 if (v == SEQ_START_TOKEN) { 2718 seq_puts(seq, 2719 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 2720 } else { 2721 const struct mfc_cache *mfc = v; 2722 const struct ipmr_mfc_iter *it = seq->private; 2723 const struct mr_table *mrt = it->mrt; 2724 2725 seq_printf(seq, "%08X %08X %-3hd", 2726 (__force u32) mfc->mfc_mcastgrp, 2727 (__force u32) mfc->mfc_origin, 2728 mfc->mfc_parent); 2729 2730 if (it->cache != &mrt->mfc_unres_queue) { 2731 seq_printf(seq, " %8lu %8lu %8lu", 2732 mfc->mfc_un.res.pkt, 2733 mfc->mfc_un.res.bytes, 2734 mfc->mfc_un.res.wrong_if); 2735 for (n = mfc->mfc_un.res.minvif; 2736 n < mfc->mfc_un.res.maxvif; n++) { 2737 if (VIF_EXISTS(mrt, n) && 2738 mfc->mfc_un.res.ttls[n] < 255) 2739 seq_printf(seq, 2740 " %2d:%-3d", 2741 n, mfc->mfc_un.res.ttls[n]); 2742 } 2743 } else { 2744 /* unresolved mfc_caches don't contain 2745 * pkt, bytes and wrong_if values 2746 */ 2747 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 2748 } 2749 seq_putc(seq, '\n'); 2750 } 2751 return 0; 2752 } 2753 2754 static const struct seq_operations ipmr_mfc_seq_ops = { 2755 .start = ipmr_mfc_seq_start, 2756 .next = ipmr_mfc_seq_next, 2757 .stop = ipmr_mfc_seq_stop, 2758 .show = ipmr_mfc_seq_show, 2759 }; 2760 2761 static int ipmr_mfc_open(struct inode *inode, struct file *file) 2762 { 2763 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 2764 sizeof(struct ipmr_mfc_iter)); 2765 } 2766 2767 static const struct file_operations ipmr_mfc_fops = { 2768 .owner = THIS_MODULE, 2769 .open = ipmr_mfc_open, 2770 .read = seq_read, 2771 .llseek = seq_lseek, 2772 .release = seq_release_net, 2773 }; 2774 #endif 2775 2776 #ifdef CONFIG_IP_PIMSM_V2 2777 static const struct net_protocol pim_protocol = { 2778 .handler = pim_rcv, 2779 .netns_ok = 1, 2780 }; 2781 #endif 2782 2783 /* Setup for IP multicast routing */ 2784 static int __net_init ipmr_net_init(struct net *net) 2785 { 2786 int err; 2787 2788 err = ipmr_rules_init(net); 2789 if (err < 0) 2790 goto fail; 2791 2792 #ifdef CONFIG_PROC_FS 2793 err = -ENOMEM; 2794 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops)) 2795 goto proc_vif_fail; 2796 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops)) 2797 goto proc_cache_fail; 2798 #endif 2799 return 0; 2800 2801 #ifdef CONFIG_PROC_FS 2802 proc_cache_fail: 2803 remove_proc_entry("ip_mr_vif", net->proc_net); 2804 proc_vif_fail: 2805 ipmr_rules_exit(net); 2806 #endif 2807 fail: 2808 return err; 2809 } 2810 2811 static void __net_exit ipmr_net_exit(struct net *net) 2812 { 2813 #ifdef CONFIG_PROC_FS 2814 remove_proc_entry("ip_mr_cache", net->proc_net); 2815 remove_proc_entry("ip_mr_vif", net->proc_net); 2816 #endif 2817 ipmr_rules_exit(net); 2818 } 2819 2820 static struct pernet_operations ipmr_net_ops = { 2821 .init = ipmr_net_init, 2822 .exit = ipmr_net_exit, 2823 }; 2824 2825 int __init ip_mr_init(void) 2826 { 2827 int err; 2828 2829 mrt_cachep = kmem_cache_create("ip_mrt_cache", 2830 sizeof(struct mfc_cache), 2831 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, 2832 NULL); 2833 2834 err = register_pernet_subsys(&ipmr_net_ops); 2835 if (err) 2836 goto reg_pernet_fail; 2837 2838 err = register_netdevice_notifier(&ip_mr_notifier); 2839 if (err) 2840 goto reg_notif_fail; 2841 #ifdef CONFIG_IP_PIMSM_V2 2842 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 2843 pr_err("%s: can't add PIM protocol\n", __func__); 2844 err = -EAGAIN; 2845 goto add_proto_fail; 2846 } 2847 #endif 2848 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE, 2849 NULL, ipmr_rtm_dumproute, NULL); 2850 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE, 2851 ipmr_rtm_route, NULL, NULL); 2852 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE, 2853 ipmr_rtm_route, NULL, NULL); 2854 return 0; 2855 2856 #ifdef CONFIG_IP_PIMSM_V2 2857 add_proto_fail: 2858 unregister_netdevice_notifier(&ip_mr_notifier); 2859 #endif 2860 reg_notif_fail: 2861 unregister_pernet_subsys(&ipmr_net_ops); 2862 reg_pernet_fail: 2863 kmem_cache_destroy(mrt_cachep); 2864 return err; 2865 } 2866