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 net_device *dev, struct sk_buff *skb, 105 struct mfc_cache *cache, 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->needs_free_netdev = true; 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), RTM_NEWNETCONF, 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 = skb_pull(skb, 673 sizeof(struct iphdr)); 674 nlh->nlmsg_type = NLMSG_ERROR; 675 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 676 skb_trim(skb, nlh->nlmsg_len); 677 e = nlmsg_data(nlh); 678 e->error = -ETIMEDOUT; 679 memset(&e->msg, 0, sizeof(e->msg)); 680 681 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 682 } else { 683 kfree_skb(skb); 684 } 685 } 686 687 ipmr_cache_free(c); 688 } 689 690 /* Timer process for the unresolved queue. */ 691 static void ipmr_expire_process(unsigned long arg) 692 { 693 struct mr_table *mrt = (struct mr_table *)arg; 694 unsigned long now; 695 unsigned long expires; 696 struct mfc_cache *c, *next; 697 698 if (!spin_trylock(&mfc_unres_lock)) { 699 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 700 return; 701 } 702 703 if (list_empty(&mrt->mfc_unres_queue)) 704 goto out; 705 706 now = jiffies; 707 expires = 10*HZ; 708 709 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 710 if (time_after(c->mfc_un.unres.expires, now)) { 711 unsigned long interval = c->mfc_un.unres.expires - now; 712 if (interval < expires) 713 expires = interval; 714 continue; 715 } 716 717 list_del(&c->list); 718 mroute_netlink_event(mrt, c, RTM_DELROUTE); 719 ipmr_destroy_unres(mrt, c); 720 } 721 722 if (!list_empty(&mrt->mfc_unres_queue)) 723 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 724 725 out: 726 spin_unlock(&mfc_unres_lock); 727 } 728 729 /* Fill oifs list. It is called under write locked mrt_lock. */ 730 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache, 731 unsigned char *ttls) 732 { 733 int vifi; 734 735 cache->mfc_un.res.minvif = MAXVIFS; 736 cache->mfc_un.res.maxvif = 0; 737 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 738 739 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 740 if (VIF_EXISTS(mrt, vifi) && 741 ttls[vifi] && ttls[vifi] < 255) { 742 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 743 if (cache->mfc_un.res.minvif > vifi) 744 cache->mfc_un.res.minvif = vifi; 745 if (cache->mfc_un.res.maxvif <= vifi) 746 cache->mfc_un.res.maxvif = vifi + 1; 747 } 748 } 749 cache->mfc_un.res.lastuse = jiffies; 750 } 751 752 static int vif_add(struct net *net, struct mr_table *mrt, 753 struct vifctl *vifc, int mrtsock) 754 { 755 int vifi = vifc->vifc_vifi; 756 struct vif_device *v = &mrt->vif_table[vifi]; 757 struct net_device *dev; 758 struct in_device *in_dev; 759 int err; 760 761 /* Is vif busy ? */ 762 if (VIF_EXISTS(mrt, vifi)) 763 return -EADDRINUSE; 764 765 switch (vifc->vifc_flags) { 766 case VIFF_REGISTER: 767 if (!ipmr_pimsm_enabled()) 768 return -EINVAL; 769 /* Special Purpose VIF in PIM 770 * All the packets will be sent to the daemon 771 */ 772 if (mrt->mroute_reg_vif_num >= 0) 773 return -EADDRINUSE; 774 dev = ipmr_reg_vif(net, mrt); 775 if (!dev) 776 return -ENOBUFS; 777 err = dev_set_allmulti(dev, 1); 778 if (err) { 779 unregister_netdevice(dev); 780 dev_put(dev); 781 return err; 782 } 783 break; 784 case VIFF_TUNNEL: 785 dev = ipmr_new_tunnel(net, vifc); 786 if (!dev) 787 return -ENOBUFS; 788 err = dev_set_allmulti(dev, 1); 789 if (err) { 790 ipmr_del_tunnel(dev, vifc); 791 dev_put(dev); 792 return err; 793 } 794 break; 795 case VIFF_USE_IFINDEX: 796 case 0: 797 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 798 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 799 if (dev && !__in_dev_get_rtnl(dev)) { 800 dev_put(dev); 801 return -EADDRNOTAVAIL; 802 } 803 } else { 804 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 805 } 806 if (!dev) 807 return -EADDRNOTAVAIL; 808 err = dev_set_allmulti(dev, 1); 809 if (err) { 810 dev_put(dev); 811 return err; 812 } 813 break; 814 default: 815 return -EINVAL; 816 } 817 818 in_dev = __in_dev_get_rtnl(dev); 819 if (!in_dev) { 820 dev_put(dev); 821 return -EADDRNOTAVAIL; 822 } 823 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 824 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING, 825 dev->ifindex, &in_dev->cnf); 826 ip_rt_multicast_event(in_dev); 827 828 /* Fill in the VIF structures */ 829 830 v->rate_limit = vifc->vifc_rate_limit; 831 v->local = vifc->vifc_lcl_addr.s_addr; 832 v->remote = vifc->vifc_rmt_addr.s_addr; 833 v->flags = vifc->vifc_flags; 834 if (!mrtsock) 835 v->flags |= VIFF_STATIC; 836 v->threshold = vifc->vifc_threshold; 837 v->bytes_in = 0; 838 v->bytes_out = 0; 839 v->pkt_in = 0; 840 v->pkt_out = 0; 841 v->link = dev->ifindex; 842 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER)) 843 v->link = dev_get_iflink(dev); 844 845 /* And finish update writing critical data */ 846 write_lock_bh(&mrt_lock); 847 v->dev = dev; 848 if (v->flags & VIFF_REGISTER) 849 mrt->mroute_reg_vif_num = vifi; 850 if (vifi+1 > mrt->maxvif) 851 mrt->maxvif = vifi+1; 852 write_unlock_bh(&mrt_lock); 853 return 0; 854 } 855 856 /* called with rcu_read_lock() */ 857 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 858 __be32 origin, 859 __be32 mcastgrp) 860 { 861 struct mfc_cache_cmp_arg arg = { 862 .mfc_mcastgrp = mcastgrp, 863 .mfc_origin = origin 864 }; 865 struct rhlist_head *tmp, *list; 866 struct mfc_cache *c; 867 868 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 869 rhl_for_each_entry_rcu(c, tmp, list, mnode) 870 return c; 871 872 return NULL; 873 } 874 875 /* Look for a (*,*,oif) entry */ 876 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt, 877 int vifi) 878 { 879 struct mfc_cache_cmp_arg arg = { 880 .mfc_mcastgrp = htonl(INADDR_ANY), 881 .mfc_origin = htonl(INADDR_ANY) 882 }; 883 struct rhlist_head *tmp, *list; 884 struct mfc_cache *c; 885 886 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 887 rhl_for_each_entry_rcu(c, tmp, list, mnode) 888 if (c->mfc_un.res.ttls[vifi] < 255) 889 return c; 890 891 return NULL; 892 } 893 894 /* Look for a (*,G) entry */ 895 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 896 __be32 mcastgrp, int vifi) 897 { 898 struct mfc_cache_cmp_arg arg = { 899 .mfc_mcastgrp = mcastgrp, 900 .mfc_origin = htonl(INADDR_ANY) 901 }; 902 struct rhlist_head *tmp, *list; 903 struct mfc_cache *c, *proxy; 904 905 if (mcastgrp == htonl(INADDR_ANY)) 906 goto skip; 907 908 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 909 rhl_for_each_entry_rcu(c, tmp, list, mnode) { 910 if (c->mfc_un.res.ttls[vifi] < 255) 911 return c; 912 913 /* It's ok if the vifi is part of the static tree */ 914 proxy = ipmr_cache_find_any_parent(mrt, c->mfc_parent); 915 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255) 916 return c; 917 } 918 919 skip: 920 return ipmr_cache_find_any_parent(mrt, vifi); 921 } 922 923 /* Look for a (S,G,iif) entry if parent != -1 */ 924 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt, 925 __be32 origin, __be32 mcastgrp, 926 int parent) 927 { 928 struct mfc_cache_cmp_arg arg = { 929 .mfc_mcastgrp = mcastgrp, 930 .mfc_origin = origin, 931 }; 932 struct rhlist_head *tmp, *list; 933 struct mfc_cache *c; 934 935 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params); 936 rhl_for_each_entry_rcu(c, tmp, list, mnode) 937 if (parent == -1 || parent == c->mfc_parent) 938 return c; 939 940 return NULL; 941 } 942 943 /* Allocate a multicast cache entry */ 944 static struct mfc_cache *ipmr_cache_alloc(void) 945 { 946 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 947 948 if (c) { 949 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1; 950 c->mfc_un.res.minvif = MAXVIFS; 951 } 952 return c; 953 } 954 955 static struct mfc_cache *ipmr_cache_alloc_unres(void) 956 { 957 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 958 959 if (c) { 960 skb_queue_head_init(&c->mfc_un.unres.unresolved); 961 c->mfc_un.unres.expires = jiffies + 10*HZ; 962 } 963 return c; 964 } 965 966 /* A cache entry has gone into a resolved state from queued */ 967 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 968 struct mfc_cache *uc, struct mfc_cache *c) 969 { 970 struct sk_buff *skb; 971 struct nlmsgerr *e; 972 973 /* Play the pending entries through our router */ 974 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 975 if (ip_hdr(skb)->version == 0) { 976 struct nlmsghdr *nlh = skb_pull(skb, 977 sizeof(struct iphdr)); 978 979 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 980 nlh->nlmsg_len = skb_tail_pointer(skb) - 981 (u8 *)nlh; 982 } else { 983 nlh->nlmsg_type = NLMSG_ERROR; 984 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 985 skb_trim(skb, nlh->nlmsg_len); 986 e = nlmsg_data(nlh); 987 e->error = -EMSGSIZE; 988 memset(&e->msg, 0, sizeof(e->msg)); 989 } 990 991 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 992 } else { 993 ip_mr_forward(net, mrt, skb->dev, skb, c, 0); 994 } 995 } 996 } 997 998 /* Bounce a cache query up to mrouted. We could use netlink for this but mrouted 999 * expects the following bizarre scheme. 1000 * 1001 * Called under mrt_lock. 1002 */ 1003 static int ipmr_cache_report(struct mr_table *mrt, 1004 struct sk_buff *pkt, vifi_t vifi, int assert) 1005 { 1006 const int ihl = ip_hdrlen(pkt); 1007 struct sock *mroute_sk; 1008 struct igmphdr *igmp; 1009 struct igmpmsg *msg; 1010 struct sk_buff *skb; 1011 int ret; 1012 1013 if (assert == IGMPMSG_WHOLEPKT) 1014 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 1015 else 1016 skb = alloc_skb(128, GFP_ATOMIC); 1017 1018 if (!skb) 1019 return -ENOBUFS; 1020 1021 if (assert == IGMPMSG_WHOLEPKT) { 1022 /* Ugly, but we have no choice with this interface. 1023 * Duplicate old header, fix ihl, length etc. 1024 * And all this only to mangle msg->im_msgtype and 1025 * to set msg->im_mbz to "mbz" :-) 1026 */ 1027 skb_push(skb, sizeof(struct iphdr)); 1028 skb_reset_network_header(skb); 1029 skb_reset_transport_header(skb); 1030 msg = (struct igmpmsg *)skb_network_header(skb); 1031 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 1032 msg->im_msgtype = IGMPMSG_WHOLEPKT; 1033 msg->im_mbz = 0; 1034 msg->im_vif = mrt->mroute_reg_vif_num; 1035 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 1036 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 1037 sizeof(struct iphdr)); 1038 } else { 1039 /* Copy the IP header */ 1040 skb_set_network_header(skb, skb->len); 1041 skb_put(skb, ihl); 1042 skb_copy_to_linear_data(skb, pkt->data, ihl); 1043 /* Flag to the kernel this is a route add */ 1044 ip_hdr(skb)->protocol = 0; 1045 msg = (struct igmpmsg *)skb_network_header(skb); 1046 msg->im_vif = vifi; 1047 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1048 /* Add our header */ 1049 igmp = skb_put(skb, sizeof(struct igmphdr)); 1050 igmp->type = assert; 1051 msg->im_msgtype = assert; 1052 igmp->code = 0; 1053 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1054 skb->transport_header = skb->network_header; 1055 } 1056 1057 rcu_read_lock(); 1058 mroute_sk = rcu_dereference(mrt->mroute_sk); 1059 if (!mroute_sk) { 1060 rcu_read_unlock(); 1061 kfree_skb(skb); 1062 return -EINVAL; 1063 } 1064 1065 /* Deliver to mrouted */ 1066 ret = sock_queue_rcv_skb(mroute_sk, skb); 1067 rcu_read_unlock(); 1068 if (ret < 0) { 1069 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1070 kfree_skb(skb); 1071 } 1072 1073 return ret; 1074 } 1075 1076 /* Queue a packet for resolution. It gets locked cache entry! */ 1077 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, 1078 struct sk_buff *skb, struct net_device *dev) 1079 { 1080 const struct iphdr *iph = ip_hdr(skb); 1081 struct mfc_cache *c; 1082 bool found = false; 1083 int err; 1084 1085 spin_lock_bh(&mfc_unres_lock); 1086 list_for_each_entry(c, &mrt->mfc_unres_queue, list) { 1087 if (c->mfc_mcastgrp == iph->daddr && 1088 c->mfc_origin == iph->saddr) { 1089 found = true; 1090 break; 1091 } 1092 } 1093 1094 if (!found) { 1095 /* Create a new entry if allowable */ 1096 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1097 (c = ipmr_cache_alloc_unres()) == NULL) { 1098 spin_unlock_bh(&mfc_unres_lock); 1099 1100 kfree_skb(skb); 1101 return -ENOBUFS; 1102 } 1103 1104 /* Fill in the new cache entry */ 1105 c->mfc_parent = -1; 1106 c->mfc_origin = iph->saddr; 1107 c->mfc_mcastgrp = iph->daddr; 1108 1109 /* Reflect first query at mrouted. */ 1110 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1111 if (err < 0) { 1112 /* If the report failed throw the cache entry 1113 out - Brad Parker 1114 */ 1115 spin_unlock_bh(&mfc_unres_lock); 1116 1117 ipmr_cache_free(c); 1118 kfree_skb(skb); 1119 return err; 1120 } 1121 1122 atomic_inc(&mrt->cache_resolve_queue_len); 1123 list_add(&c->list, &mrt->mfc_unres_queue); 1124 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1125 1126 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1127 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires); 1128 } 1129 1130 /* See if we can append the packet */ 1131 if (c->mfc_un.unres.unresolved.qlen > 3) { 1132 kfree_skb(skb); 1133 err = -ENOBUFS; 1134 } else { 1135 if (dev) { 1136 skb->dev = dev; 1137 skb->skb_iif = dev->ifindex; 1138 } 1139 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1140 err = 0; 1141 } 1142 1143 spin_unlock_bh(&mfc_unres_lock); 1144 return err; 1145 } 1146 1147 /* MFC cache manipulation by user space mroute daemon */ 1148 1149 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1150 { 1151 struct mfc_cache *c; 1152 1153 /* The entries are added/deleted only under RTNL */ 1154 rcu_read_lock(); 1155 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1156 mfc->mfcc_mcastgrp.s_addr, parent); 1157 rcu_read_unlock(); 1158 if (!c) 1159 return -ENOENT; 1160 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1161 list_del_rcu(&c->list); 1162 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1163 ipmr_cache_free(c); 1164 1165 return 0; 1166 } 1167 1168 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1169 struct mfcctl *mfc, int mrtsock, int parent) 1170 { 1171 struct mfc_cache *uc, *c; 1172 bool found; 1173 int ret; 1174 1175 if (mfc->mfcc_parent >= MAXVIFS) 1176 return -ENFILE; 1177 1178 /* The entries are added/deleted only under RTNL */ 1179 rcu_read_lock(); 1180 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr, 1181 mfc->mfcc_mcastgrp.s_addr, parent); 1182 rcu_read_unlock(); 1183 if (c) { 1184 write_lock_bh(&mrt_lock); 1185 c->mfc_parent = mfc->mfcc_parent; 1186 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1187 if (!mrtsock) 1188 c->mfc_flags |= MFC_STATIC; 1189 write_unlock_bh(&mrt_lock); 1190 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1191 return 0; 1192 } 1193 1194 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1195 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1196 return -EINVAL; 1197 1198 c = ipmr_cache_alloc(); 1199 if (!c) 1200 return -ENOMEM; 1201 1202 c->mfc_origin = mfc->mfcc_origin.s_addr; 1203 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1204 c->mfc_parent = mfc->mfcc_parent; 1205 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1206 if (!mrtsock) 1207 c->mfc_flags |= MFC_STATIC; 1208 1209 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->mnode, 1210 ipmr_rht_params); 1211 if (ret) { 1212 pr_err("ipmr: rhtable insert error %d\n", ret); 1213 ipmr_cache_free(c); 1214 return ret; 1215 } 1216 list_add_tail_rcu(&c->list, &mrt->mfc_cache_list); 1217 /* Check to see if we resolved a queued list. If so we 1218 * need to send on the frames and tidy up. 1219 */ 1220 found = false; 1221 spin_lock_bh(&mfc_unres_lock); 1222 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) { 1223 if (uc->mfc_origin == c->mfc_origin && 1224 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1225 list_del(&uc->list); 1226 atomic_dec(&mrt->cache_resolve_queue_len); 1227 found = true; 1228 break; 1229 } 1230 } 1231 if (list_empty(&mrt->mfc_unres_queue)) 1232 del_timer(&mrt->ipmr_expire_timer); 1233 spin_unlock_bh(&mfc_unres_lock); 1234 1235 if (found) { 1236 ipmr_cache_resolve(net, mrt, uc, c); 1237 ipmr_cache_free(uc); 1238 } 1239 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1240 return 0; 1241 } 1242 1243 /* Close the multicast socket, and clear the vif tables etc */ 1244 static void mroute_clean_tables(struct mr_table *mrt, bool all) 1245 { 1246 struct mfc_cache *c, *tmp; 1247 LIST_HEAD(list); 1248 int i; 1249 1250 /* Shut down all active vif entries */ 1251 for (i = 0; i < mrt->maxvif; i++) { 1252 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC)) 1253 continue; 1254 vif_delete(mrt, i, 0, &list); 1255 } 1256 unregister_netdevice_many(&list); 1257 1258 /* Wipe the cache */ 1259 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) { 1260 if (!all && (c->mfc_flags & MFC_STATIC)) 1261 continue; 1262 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params); 1263 list_del_rcu(&c->list); 1264 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1265 ipmr_cache_free(c); 1266 } 1267 1268 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1269 spin_lock_bh(&mfc_unres_lock); 1270 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) { 1271 list_del(&c->list); 1272 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1273 ipmr_destroy_unres(mrt, c); 1274 } 1275 spin_unlock_bh(&mfc_unres_lock); 1276 } 1277 } 1278 1279 /* called from ip_ra_control(), before an RCU grace period, 1280 * we dont need to call synchronize_rcu() here 1281 */ 1282 static void mrtsock_destruct(struct sock *sk) 1283 { 1284 struct net *net = sock_net(sk); 1285 struct mr_table *mrt; 1286 1287 ASSERT_RTNL(); 1288 ipmr_for_each_table(mrt, net) { 1289 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1290 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1291 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1292 NETCONFA_MC_FORWARDING, 1293 NETCONFA_IFINDEX_ALL, 1294 net->ipv4.devconf_all); 1295 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1296 mroute_clean_tables(mrt, false); 1297 } 1298 } 1299 } 1300 1301 /* Socket options and virtual interface manipulation. The whole 1302 * virtual interface system is a complete heap, but unfortunately 1303 * that's how BSD mrouted happens to think. Maybe one day with a proper 1304 * MOSPF/PIM router set up we can clean this up. 1305 */ 1306 1307 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, 1308 unsigned int optlen) 1309 { 1310 struct net *net = sock_net(sk); 1311 int val, ret = 0, parent = 0; 1312 struct mr_table *mrt; 1313 struct vifctl vif; 1314 struct mfcctl mfc; 1315 u32 uval; 1316 1317 /* There's one exception to the lock - MRT_DONE which needs to unlock */ 1318 rtnl_lock(); 1319 if (sk->sk_type != SOCK_RAW || 1320 inet_sk(sk)->inet_num != IPPROTO_IGMP) { 1321 ret = -EOPNOTSUPP; 1322 goto out_unlock; 1323 } 1324 1325 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1326 if (!mrt) { 1327 ret = -ENOENT; 1328 goto out_unlock; 1329 } 1330 if (optname != MRT_INIT) { 1331 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1332 !ns_capable(net->user_ns, CAP_NET_ADMIN)) { 1333 ret = -EACCES; 1334 goto out_unlock; 1335 } 1336 } 1337 1338 switch (optname) { 1339 case MRT_INIT: 1340 if (optlen != sizeof(int)) { 1341 ret = -EINVAL; 1342 break; 1343 } 1344 if (rtnl_dereference(mrt->mroute_sk)) { 1345 ret = -EADDRINUSE; 1346 break; 1347 } 1348 1349 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1350 if (ret == 0) { 1351 rcu_assign_pointer(mrt->mroute_sk, sk); 1352 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1353 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 1354 NETCONFA_MC_FORWARDING, 1355 NETCONFA_IFINDEX_ALL, 1356 net->ipv4.devconf_all); 1357 } 1358 break; 1359 case MRT_DONE: 1360 if (sk != rcu_access_pointer(mrt->mroute_sk)) { 1361 ret = -EACCES; 1362 } else { 1363 ret = ip_ra_control(sk, 0, NULL); 1364 goto out_unlock; 1365 } 1366 break; 1367 case MRT_ADD_VIF: 1368 case MRT_DEL_VIF: 1369 if (optlen != sizeof(vif)) { 1370 ret = -EINVAL; 1371 break; 1372 } 1373 if (copy_from_user(&vif, optval, sizeof(vif))) { 1374 ret = -EFAULT; 1375 break; 1376 } 1377 if (vif.vifc_vifi >= MAXVIFS) { 1378 ret = -ENFILE; 1379 break; 1380 } 1381 if (optname == MRT_ADD_VIF) { 1382 ret = vif_add(net, mrt, &vif, 1383 sk == rtnl_dereference(mrt->mroute_sk)); 1384 } else { 1385 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1386 } 1387 break; 1388 /* Manipulate the forwarding caches. These live 1389 * in a sort of kernel/user symbiosis. 1390 */ 1391 case MRT_ADD_MFC: 1392 case MRT_DEL_MFC: 1393 parent = -1; 1394 case MRT_ADD_MFC_PROXY: 1395 case MRT_DEL_MFC_PROXY: 1396 if (optlen != sizeof(mfc)) { 1397 ret = -EINVAL; 1398 break; 1399 } 1400 if (copy_from_user(&mfc, optval, sizeof(mfc))) { 1401 ret = -EFAULT; 1402 break; 1403 } 1404 if (parent == 0) 1405 parent = mfc.mfcc_parent; 1406 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1407 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1408 else 1409 ret = ipmr_mfc_add(net, mrt, &mfc, 1410 sk == rtnl_dereference(mrt->mroute_sk), 1411 parent); 1412 break; 1413 /* Control PIM assert. */ 1414 case MRT_ASSERT: 1415 if (optlen != sizeof(val)) { 1416 ret = -EINVAL; 1417 break; 1418 } 1419 if (get_user(val, (int __user *)optval)) { 1420 ret = -EFAULT; 1421 break; 1422 } 1423 mrt->mroute_do_assert = val; 1424 break; 1425 case MRT_PIM: 1426 if (!ipmr_pimsm_enabled()) { 1427 ret = -ENOPROTOOPT; 1428 break; 1429 } 1430 if (optlen != sizeof(val)) { 1431 ret = -EINVAL; 1432 break; 1433 } 1434 if (get_user(val, (int __user *)optval)) { 1435 ret = -EFAULT; 1436 break; 1437 } 1438 1439 val = !!val; 1440 if (val != mrt->mroute_do_pim) { 1441 mrt->mroute_do_pim = val; 1442 mrt->mroute_do_assert = val; 1443 } 1444 break; 1445 case MRT_TABLE: 1446 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) { 1447 ret = -ENOPROTOOPT; 1448 break; 1449 } 1450 if (optlen != sizeof(uval)) { 1451 ret = -EINVAL; 1452 break; 1453 } 1454 if (get_user(uval, (u32 __user *)optval)) { 1455 ret = -EFAULT; 1456 break; 1457 } 1458 1459 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1460 ret = -EBUSY; 1461 } else { 1462 mrt = ipmr_new_table(net, uval); 1463 if (IS_ERR(mrt)) 1464 ret = PTR_ERR(mrt); 1465 else 1466 raw_sk(sk)->ipmr_table = uval; 1467 } 1468 break; 1469 /* Spurious command, or MRT_VERSION which you cannot set. */ 1470 default: 1471 ret = -ENOPROTOOPT; 1472 } 1473 out_unlock: 1474 rtnl_unlock(); 1475 return ret; 1476 } 1477 1478 /* Getsock opt support for the multicast routing system. */ 1479 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen) 1480 { 1481 int olr; 1482 int val; 1483 struct net *net = sock_net(sk); 1484 struct mr_table *mrt; 1485 1486 if (sk->sk_type != SOCK_RAW || 1487 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1488 return -EOPNOTSUPP; 1489 1490 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1491 if (!mrt) 1492 return -ENOENT; 1493 1494 switch (optname) { 1495 case MRT_VERSION: 1496 val = 0x0305; 1497 break; 1498 case MRT_PIM: 1499 if (!ipmr_pimsm_enabled()) 1500 return -ENOPROTOOPT; 1501 val = mrt->mroute_do_pim; 1502 break; 1503 case MRT_ASSERT: 1504 val = mrt->mroute_do_assert; 1505 break; 1506 default: 1507 return -ENOPROTOOPT; 1508 } 1509 1510 if (get_user(olr, optlen)) 1511 return -EFAULT; 1512 olr = min_t(unsigned int, olr, sizeof(int)); 1513 if (olr < 0) 1514 return -EINVAL; 1515 if (put_user(olr, optlen)) 1516 return -EFAULT; 1517 if (copy_to_user(optval, &val, olr)) 1518 return -EFAULT; 1519 return 0; 1520 } 1521 1522 /* The IP multicast ioctl support routines. */ 1523 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) 1524 { 1525 struct sioc_sg_req sr; 1526 struct sioc_vif_req vr; 1527 struct vif_device *vif; 1528 struct mfc_cache *c; 1529 struct net *net = sock_net(sk); 1530 struct mr_table *mrt; 1531 1532 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1533 if (!mrt) 1534 return -ENOENT; 1535 1536 switch (cmd) { 1537 case SIOCGETVIFCNT: 1538 if (copy_from_user(&vr, arg, sizeof(vr))) 1539 return -EFAULT; 1540 if (vr.vifi >= mrt->maxvif) 1541 return -EINVAL; 1542 read_lock(&mrt_lock); 1543 vif = &mrt->vif_table[vr.vifi]; 1544 if (VIF_EXISTS(mrt, vr.vifi)) { 1545 vr.icount = vif->pkt_in; 1546 vr.ocount = vif->pkt_out; 1547 vr.ibytes = vif->bytes_in; 1548 vr.obytes = vif->bytes_out; 1549 read_unlock(&mrt_lock); 1550 1551 if (copy_to_user(arg, &vr, sizeof(vr))) 1552 return -EFAULT; 1553 return 0; 1554 } 1555 read_unlock(&mrt_lock); 1556 return -EADDRNOTAVAIL; 1557 case SIOCGETSGCNT: 1558 if (copy_from_user(&sr, arg, sizeof(sr))) 1559 return -EFAULT; 1560 1561 rcu_read_lock(); 1562 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1563 if (c) { 1564 sr.pktcnt = c->mfc_un.res.pkt; 1565 sr.bytecnt = c->mfc_un.res.bytes; 1566 sr.wrong_if = c->mfc_un.res.wrong_if; 1567 rcu_read_unlock(); 1568 1569 if (copy_to_user(arg, &sr, sizeof(sr))) 1570 return -EFAULT; 1571 return 0; 1572 } 1573 rcu_read_unlock(); 1574 return -EADDRNOTAVAIL; 1575 default: 1576 return -ENOIOCTLCMD; 1577 } 1578 } 1579 1580 #ifdef CONFIG_COMPAT 1581 struct compat_sioc_sg_req { 1582 struct in_addr src; 1583 struct in_addr grp; 1584 compat_ulong_t pktcnt; 1585 compat_ulong_t bytecnt; 1586 compat_ulong_t wrong_if; 1587 }; 1588 1589 struct compat_sioc_vif_req { 1590 vifi_t vifi; /* Which iface */ 1591 compat_ulong_t icount; 1592 compat_ulong_t ocount; 1593 compat_ulong_t ibytes; 1594 compat_ulong_t obytes; 1595 }; 1596 1597 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1598 { 1599 struct compat_sioc_sg_req sr; 1600 struct compat_sioc_vif_req vr; 1601 struct vif_device *vif; 1602 struct mfc_cache *c; 1603 struct net *net = sock_net(sk); 1604 struct mr_table *mrt; 1605 1606 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1607 if (!mrt) 1608 return -ENOENT; 1609 1610 switch (cmd) { 1611 case SIOCGETVIFCNT: 1612 if (copy_from_user(&vr, arg, sizeof(vr))) 1613 return -EFAULT; 1614 if (vr.vifi >= mrt->maxvif) 1615 return -EINVAL; 1616 read_lock(&mrt_lock); 1617 vif = &mrt->vif_table[vr.vifi]; 1618 if (VIF_EXISTS(mrt, vr.vifi)) { 1619 vr.icount = vif->pkt_in; 1620 vr.ocount = vif->pkt_out; 1621 vr.ibytes = vif->bytes_in; 1622 vr.obytes = vif->bytes_out; 1623 read_unlock(&mrt_lock); 1624 1625 if (copy_to_user(arg, &vr, sizeof(vr))) 1626 return -EFAULT; 1627 return 0; 1628 } 1629 read_unlock(&mrt_lock); 1630 return -EADDRNOTAVAIL; 1631 case SIOCGETSGCNT: 1632 if (copy_from_user(&sr, arg, sizeof(sr))) 1633 return -EFAULT; 1634 1635 rcu_read_lock(); 1636 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1637 if (c) { 1638 sr.pktcnt = c->mfc_un.res.pkt; 1639 sr.bytecnt = c->mfc_un.res.bytes; 1640 sr.wrong_if = c->mfc_un.res.wrong_if; 1641 rcu_read_unlock(); 1642 1643 if (copy_to_user(arg, &sr, sizeof(sr))) 1644 return -EFAULT; 1645 return 0; 1646 } 1647 rcu_read_unlock(); 1648 return -EADDRNOTAVAIL; 1649 default: 1650 return -ENOIOCTLCMD; 1651 } 1652 } 1653 #endif 1654 1655 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1656 { 1657 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1658 struct net *net = dev_net(dev); 1659 struct mr_table *mrt; 1660 struct vif_device *v; 1661 int ct; 1662 1663 if (event != NETDEV_UNREGISTER) 1664 return NOTIFY_DONE; 1665 1666 ipmr_for_each_table(mrt, net) { 1667 v = &mrt->vif_table[0]; 1668 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1669 if (v->dev == dev) 1670 vif_delete(mrt, ct, 1, NULL); 1671 } 1672 } 1673 return NOTIFY_DONE; 1674 } 1675 1676 static struct notifier_block ip_mr_notifier = { 1677 .notifier_call = ipmr_device_event, 1678 }; 1679 1680 /* Encapsulate a packet by attaching a valid IPIP header to it. 1681 * This avoids tunnel drivers and other mess and gives us the speed so 1682 * important for multicast video. 1683 */ 1684 static void ip_encap(struct net *net, struct sk_buff *skb, 1685 __be32 saddr, __be32 daddr) 1686 { 1687 struct iphdr *iph; 1688 const struct iphdr *old_iph = ip_hdr(skb); 1689 1690 skb_push(skb, sizeof(struct iphdr)); 1691 skb->transport_header = skb->network_header; 1692 skb_reset_network_header(skb); 1693 iph = ip_hdr(skb); 1694 1695 iph->version = 4; 1696 iph->tos = old_iph->tos; 1697 iph->ttl = old_iph->ttl; 1698 iph->frag_off = 0; 1699 iph->daddr = daddr; 1700 iph->saddr = saddr; 1701 iph->protocol = IPPROTO_IPIP; 1702 iph->ihl = 5; 1703 iph->tot_len = htons(skb->len); 1704 ip_select_ident(net, skb, NULL); 1705 ip_send_check(iph); 1706 1707 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1708 nf_reset(skb); 1709 } 1710 1711 static inline int ipmr_forward_finish(struct net *net, struct sock *sk, 1712 struct sk_buff *skb) 1713 { 1714 struct ip_options *opt = &(IPCB(skb)->opt); 1715 1716 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS); 1717 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len); 1718 1719 if (unlikely(opt->optlen)) 1720 ip_forward_options(skb); 1721 1722 return dst_output(net, sk, skb); 1723 } 1724 1725 /* Processing handlers for ipmr_forward */ 1726 1727 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt, 1728 struct sk_buff *skb, struct mfc_cache *c, int vifi) 1729 { 1730 const struct iphdr *iph = ip_hdr(skb); 1731 struct vif_device *vif = &mrt->vif_table[vifi]; 1732 struct net_device *dev; 1733 struct rtable *rt; 1734 struct flowi4 fl4; 1735 int encap = 0; 1736 1737 if (!vif->dev) 1738 goto out_free; 1739 1740 if (vif->flags & VIFF_REGISTER) { 1741 vif->pkt_out++; 1742 vif->bytes_out += skb->len; 1743 vif->dev->stats.tx_bytes += skb->len; 1744 vif->dev->stats.tx_packets++; 1745 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1746 goto out_free; 1747 } 1748 1749 if (vif->flags & VIFF_TUNNEL) { 1750 rt = ip_route_output_ports(net, &fl4, NULL, 1751 vif->remote, vif->local, 1752 0, 0, 1753 IPPROTO_IPIP, 1754 RT_TOS(iph->tos), vif->link); 1755 if (IS_ERR(rt)) 1756 goto out_free; 1757 encap = sizeof(struct iphdr); 1758 } else { 1759 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1760 0, 0, 1761 IPPROTO_IPIP, 1762 RT_TOS(iph->tos), vif->link); 1763 if (IS_ERR(rt)) 1764 goto out_free; 1765 } 1766 1767 dev = rt->dst.dev; 1768 1769 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1770 /* Do not fragment multicasts. Alas, IPv4 does not 1771 * allow to send ICMP, so that packets will disappear 1772 * to blackhole. 1773 */ 1774 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS); 1775 ip_rt_put(rt); 1776 goto out_free; 1777 } 1778 1779 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len; 1780 1781 if (skb_cow(skb, encap)) { 1782 ip_rt_put(rt); 1783 goto out_free; 1784 } 1785 1786 vif->pkt_out++; 1787 vif->bytes_out += skb->len; 1788 1789 skb_dst_drop(skb); 1790 skb_dst_set(skb, &rt->dst); 1791 ip_decrease_ttl(ip_hdr(skb)); 1792 1793 /* FIXME: forward and output firewalls used to be called here. 1794 * What do we do with netfilter? -- RR 1795 */ 1796 if (vif->flags & VIFF_TUNNEL) { 1797 ip_encap(net, skb, vif->local, vif->remote); 1798 /* FIXME: extra output firewall step used to be here. --RR */ 1799 vif->dev->stats.tx_packets++; 1800 vif->dev->stats.tx_bytes += skb->len; 1801 } 1802 1803 IPCB(skb)->flags |= IPSKB_FORWARDED; 1804 1805 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1806 * not only before forwarding, but after forwarding on all output 1807 * interfaces. It is clear, if mrouter runs a multicasting 1808 * program, it should receive packets not depending to what interface 1809 * program is joined. 1810 * If we will not make it, the program will have to join on all 1811 * interfaces. On the other hand, multihoming host (or router, but 1812 * not mrouter) cannot join to more than one interface - it will 1813 * result in receiving multiple packets. 1814 */ 1815 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, 1816 net, NULL, skb, skb->dev, dev, 1817 ipmr_forward_finish); 1818 return; 1819 1820 out_free: 1821 kfree_skb(skb); 1822 } 1823 1824 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev) 1825 { 1826 int ct; 1827 1828 for (ct = mrt->maxvif-1; ct >= 0; ct--) { 1829 if (mrt->vif_table[ct].dev == dev) 1830 break; 1831 } 1832 return ct; 1833 } 1834 1835 /* "local" means that we should preserve one skb (for local delivery) */ 1836 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 1837 struct net_device *dev, struct sk_buff *skb, 1838 struct mfc_cache *cache, int local) 1839 { 1840 int true_vifi = ipmr_find_vif(mrt, dev); 1841 int psend = -1; 1842 int vif, ct; 1843 1844 vif = cache->mfc_parent; 1845 cache->mfc_un.res.pkt++; 1846 cache->mfc_un.res.bytes += skb->len; 1847 cache->mfc_un.res.lastuse = jiffies; 1848 1849 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 1850 struct mfc_cache *cache_proxy; 1851 1852 /* For an (*,G) entry, we only check that the incomming 1853 * interface is part of the static tree. 1854 */ 1855 cache_proxy = ipmr_cache_find_any_parent(mrt, vif); 1856 if (cache_proxy && 1857 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 1858 goto forward; 1859 } 1860 1861 /* Wrong interface: drop packet and (maybe) send PIM assert. */ 1862 if (mrt->vif_table[vif].dev != dev) { 1863 if (rt_is_output_route(skb_rtable(skb))) { 1864 /* It is our own packet, looped back. 1865 * Very complicated situation... 1866 * 1867 * The best workaround until routing daemons will be 1868 * fixed is not to redistribute packet, if it was 1869 * send through wrong interface. It means, that 1870 * multicast applications WILL NOT work for 1871 * (S,G), which have default multicast route pointing 1872 * to wrong oif. In any case, it is not a good 1873 * idea to use multicasting applications on router. 1874 */ 1875 goto dont_forward; 1876 } 1877 1878 cache->mfc_un.res.wrong_if++; 1879 1880 if (true_vifi >= 0 && mrt->mroute_do_assert && 1881 /* pimsm uses asserts, when switching from RPT to SPT, 1882 * so that we cannot check that packet arrived on an oif. 1883 * It is bad, but otherwise we would need to move pretty 1884 * large chunk of pimd to kernel. Ough... --ANK 1885 */ 1886 (mrt->mroute_do_pim || 1887 cache->mfc_un.res.ttls[true_vifi] < 255) && 1888 time_after(jiffies, 1889 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 1890 cache->mfc_un.res.last_assert = jiffies; 1891 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 1892 } 1893 goto dont_forward; 1894 } 1895 1896 forward: 1897 mrt->vif_table[vif].pkt_in++; 1898 mrt->vif_table[vif].bytes_in += skb->len; 1899 1900 /* Forward the frame */ 1901 if (cache->mfc_origin == htonl(INADDR_ANY) && 1902 cache->mfc_mcastgrp == htonl(INADDR_ANY)) { 1903 if (true_vifi >= 0 && 1904 true_vifi != cache->mfc_parent && 1905 ip_hdr(skb)->ttl > 1906 cache->mfc_un.res.ttls[cache->mfc_parent]) { 1907 /* It's an (*,*) entry and the packet is not coming from 1908 * the upstream: forward the packet to the upstream 1909 * only. 1910 */ 1911 psend = cache->mfc_parent; 1912 goto last_forward; 1913 } 1914 goto dont_forward; 1915 } 1916 for (ct = cache->mfc_un.res.maxvif - 1; 1917 ct >= cache->mfc_un.res.minvif; ct--) { 1918 /* For (*,G) entry, don't forward to the incoming interface */ 1919 if ((cache->mfc_origin != htonl(INADDR_ANY) || 1920 ct != true_vifi) && 1921 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) { 1922 if (psend != -1) { 1923 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1924 1925 if (skb2) 1926 ipmr_queue_xmit(net, mrt, skb2, cache, 1927 psend); 1928 } 1929 psend = ct; 1930 } 1931 } 1932 last_forward: 1933 if (psend != -1) { 1934 if (local) { 1935 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1936 1937 if (skb2) 1938 ipmr_queue_xmit(net, mrt, skb2, cache, psend); 1939 } else { 1940 ipmr_queue_xmit(net, mrt, skb, cache, psend); 1941 return; 1942 } 1943 } 1944 1945 dont_forward: 1946 if (!local) 1947 kfree_skb(skb); 1948 } 1949 1950 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 1951 { 1952 struct rtable *rt = skb_rtable(skb); 1953 struct iphdr *iph = ip_hdr(skb); 1954 struct flowi4 fl4 = { 1955 .daddr = iph->daddr, 1956 .saddr = iph->saddr, 1957 .flowi4_tos = RT_TOS(iph->tos), 1958 .flowi4_oif = (rt_is_output_route(rt) ? 1959 skb->dev->ifindex : 0), 1960 .flowi4_iif = (rt_is_output_route(rt) ? 1961 LOOPBACK_IFINDEX : 1962 skb->dev->ifindex), 1963 .flowi4_mark = skb->mark, 1964 }; 1965 struct mr_table *mrt; 1966 int err; 1967 1968 err = ipmr_fib_lookup(net, &fl4, &mrt); 1969 if (err) 1970 return ERR_PTR(err); 1971 return mrt; 1972 } 1973 1974 /* Multicast packets for forwarding arrive here 1975 * Called with rcu_read_lock(); 1976 */ 1977 int ip_mr_input(struct sk_buff *skb) 1978 { 1979 struct mfc_cache *cache; 1980 struct net *net = dev_net(skb->dev); 1981 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 1982 struct mr_table *mrt; 1983 struct net_device *dev; 1984 1985 /* skb->dev passed in is the loX master dev for vrfs. 1986 * As there are no vifs associated with loopback devices, 1987 * get the proper interface that does have a vif associated with it. 1988 */ 1989 dev = skb->dev; 1990 if (netif_is_l3_master(skb->dev)) { 1991 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 1992 if (!dev) { 1993 kfree_skb(skb); 1994 return -ENODEV; 1995 } 1996 } 1997 1998 /* Packet is looped back after forward, it should not be 1999 * forwarded second time, but still can be delivered locally. 2000 */ 2001 if (IPCB(skb)->flags & IPSKB_FORWARDED) 2002 goto dont_forward; 2003 2004 mrt = ipmr_rt_fib_lookup(net, skb); 2005 if (IS_ERR(mrt)) { 2006 kfree_skb(skb); 2007 return PTR_ERR(mrt); 2008 } 2009 if (!local) { 2010 if (IPCB(skb)->opt.router_alert) { 2011 if (ip_call_ra_chain(skb)) 2012 return 0; 2013 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 2014 /* IGMPv1 (and broken IGMPv2 implementations sort of 2015 * Cisco IOS <= 11.2(8)) do not put router alert 2016 * option to IGMP packets destined to routable 2017 * groups. It is very bad, because it means 2018 * that we can forward NO IGMP messages. 2019 */ 2020 struct sock *mroute_sk; 2021 2022 mroute_sk = rcu_dereference(mrt->mroute_sk); 2023 if (mroute_sk) { 2024 nf_reset(skb); 2025 raw_rcv(mroute_sk, skb); 2026 return 0; 2027 } 2028 } 2029 } 2030 2031 /* already under rcu_read_lock() */ 2032 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 2033 if (!cache) { 2034 int vif = ipmr_find_vif(mrt, dev); 2035 2036 if (vif >= 0) 2037 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2038 vif); 2039 } 2040 2041 /* No usable cache entry */ 2042 if (!cache) { 2043 int vif; 2044 2045 if (local) { 2046 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2047 ip_local_deliver(skb); 2048 if (!skb2) 2049 return -ENOBUFS; 2050 skb = skb2; 2051 } 2052 2053 read_lock(&mrt_lock); 2054 vif = ipmr_find_vif(mrt, dev); 2055 if (vif >= 0) { 2056 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev); 2057 read_unlock(&mrt_lock); 2058 2059 return err2; 2060 } 2061 read_unlock(&mrt_lock); 2062 kfree_skb(skb); 2063 return -ENODEV; 2064 } 2065 2066 read_lock(&mrt_lock); 2067 ip_mr_forward(net, mrt, dev, skb, cache, local); 2068 read_unlock(&mrt_lock); 2069 2070 if (local) 2071 return ip_local_deliver(skb); 2072 2073 return 0; 2074 2075 dont_forward: 2076 if (local) 2077 return ip_local_deliver(skb); 2078 kfree_skb(skb); 2079 return 0; 2080 } 2081 2082 #ifdef CONFIG_IP_PIMSM_V1 2083 /* Handle IGMP messages of PIMv1 */ 2084 int pim_rcv_v1(struct sk_buff *skb) 2085 { 2086 struct igmphdr *pim; 2087 struct net *net = dev_net(skb->dev); 2088 struct mr_table *mrt; 2089 2090 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2091 goto drop; 2092 2093 pim = igmp_hdr(skb); 2094 2095 mrt = ipmr_rt_fib_lookup(net, skb); 2096 if (IS_ERR(mrt)) 2097 goto drop; 2098 if (!mrt->mroute_do_pim || 2099 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2100 goto drop; 2101 2102 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2103 drop: 2104 kfree_skb(skb); 2105 } 2106 return 0; 2107 } 2108 #endif 2109 2110 #ifdef CONFIG_IP_PIMSM_V2 2111 static int pim_rcv(struct sk_buff *skb) 2112 { 2113 struct pimreghdr *pim; 2114 struct net *net = dev_net(skb->dev); 2115 struct mr_table *mrt; 2116 2117 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2118 goto drop; 2119 2120 pim = (struct pimreghdr *)skb_transport_header(skb); 2121 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) || 2122 (pim->flags & PIM_NULL_REGISTER) || 2123 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2124 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2125 goto drop; 2126 2127 mrt = ipmr_rt_fib_lookup(net, skb); 2128 if (IS_ERR(mrt)) 2129 goto drop; 2130 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2131 drop: 2132 kfree_skb(skb); 2133 } 2134 return 0; 2135 } 2136 #endif 2137 2138 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2139 struct mfc_cache *c, struct rtmsg *rtm) 2140 { 2141 struct rta_mfc_stats mfcs; 2142 struct nlattr *mp_attr; 2143 struct rtnexthop *nhp; 2144 unsigned long lastuse; 2145 int ct; 2146 2147 /* If cache is unresolved, don't try to parse IIF and OIF */ 2148 if (c->mfc_parent >= MAXVIFS) { 2149 rtm->rtm_flags |= RTNH_F_UNRESOLVED; 2150 return -ENOENT; 2151 } 2152 2153 if (VIF_EXISTS(mrt, c->mfc_parent) && 2154 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0) 2155 return -EMSGSIZE; 2156 2157 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH))) 2158 return -EMSGSIZE; 2159 2160 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2161 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2162 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) { 2163 nla_nest_cancel(skb, mp_attr); 2164 return -EMSGSIZE; 2165 } 2166 2167 nhp->rtnh_flags = 0; 2168 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2169 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex; 2170 nhp->rtnh_len = sizeof(*nhp); 2171 } 2172 } 2173 2174 nla_nest_end(skb, mp_attr); 2175 2176 lastuse = READ_ONCE(c->mfc_un.res.lastuse); 2177 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0; 2178 2179 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2180 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2181 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2182 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) || 2183 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse), 2184 RTA_PAD)) 2185 return -EMSGSIZE; 2186 2187 rtm->rtm_type = RTN_MULTICAST; 2188 return 1; 2189 } 2190 2191 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2192 __be32 saddr, __be32 daddr, 2193 struct rtmsg *rtm, u32 portid) 2194 { 2195 struct mfc_cache *cache; 2196 struct mr_table *mrt; 2197 int err; 2198 2199 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2200 if (!mrt) 2201 return -ENOENT; 2202 2203 rcu_read_lock(); 2204 cache = ipmr_cache_find(mrt, saddr, daddr); 2205 if (!cache && skb->dev) { 2206 int vif = ipmr_find_vif(mrt, skb->dev); 2207 2208 if (vif >= 0) 2209 cache = ipmr_cache_find_any(mrt, daddr, vif); 2210 } 2211 if (!cache) { 2212 struct sk_buff *skb2; 2213 struct iphdr *iph; 2214 struct net_device *dev; 2215 int vif = -1; 2216 2217 dev = skb->dev; 2218 read_lock(&mrt_lock); 2219 if (dev) 2220 vif = ipmr_find_vif(mrt, dev); 2221 if (vif < 0) { 2222 read_unlock(&mrt_lock); 2223 rcu_read_unlock(); 2224 return -ENODEV; 2225 } 2226 skb2 = skb_clone(skb, GFP_ATOMIC); 2227 if (!skb2) { 2228 read_unlock(&mrt_lock); 2229 rcu_read_unlock(); 2230 return -ENOMEM; 2231 } 2232 2233 NETLINK_CB(skb2).portid = portid; 2234 skb_push(skb2, sizeof(struct iphdr)); 2235 skb_reset_network_header(skb2); 2236 iph = ip_hdr(skb2); 2237 iph->ihl = sizeof(struct iphdr) >> 2; 2238 iph->saddr = saddr; 2239 iph->daddr = daddr; 2240 iph->version = 0; 2241 err = ipmr_cache_unresolved(mrt, vif, skb2, dev); 2242 read_unlock(&mrt_lock); 2243 rcu_read_unlock(); 2244 return err; 2245 } 2246 2247 read_lock(&mrt_lock); 2248 err = __ipmr_fill_mroute(mrt, skb, cache, rtm); 2249 read_unlock(&mrt_lock); 2250 rcu_read_unlock(); 2251 return err; 2252 } 2253 2254 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2255 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2256 int flags) 2257 { 2258 struct nlmsghdr *nlh; 2259 struct rtmsg *rtm; 2260 int err; 2261 2262 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2263 if (!nlh) 2264 return -EMSGSIZE; 2265 2266 rtm = nlmsg_data(nlh); 2267 rtm->rtm_family = RTNL_FAMILY_IPMR; 2268 rtm->rtm_dst_len = 32; 2269 rtm->rtm_src_len = 32; 2270 rtm->rtm_tos = 0; 2271 rtm->rtm_table = mrt->id; 2272 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2273 goto nla_put_failure; 2274 rtm->rtm_type = RTN_MULTICAST; 2275 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2276 if (c->mfc_flags & MFC_STATIC) 2277 rtm->rtm_protocol = RTPROT_STATIC; 2278 else 2279 rtm->rtm_protocol = RTPROT_MROUTED; 2280 rtm->rtm_flags = 0; 2281 2282 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) || 2283 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp)) 2284 goto nla_put_failure; 2285 err = __ipmr_fill_mroute(mrt, skb, c, rtm); 2286 /* do not break the dump if cache is unresolved */ 2287 if (err < 0 && err != -ENOENT) 2288 goto nla_put_failure; 2289 2290 nlmsg_end(skb, nlh); 2291 return 0; 2292 2293 nla_put_failure: 2294 nlmsg_cancel(skb, nlh); 2295 return -EMSGSIZE; 2296 } 2297 2298 static size_t mroute_msgsize(bool unresolved, int maxvif) 2299 { 2300 size_t len = 2301 NLMSG_ALIGN(sizeof(struct rtmsg)) 2302 + nla_total_size(4) /* RTA_TABLE */ 2303 + nla_total_size(4) /* RTA_SRC */ 2304 + nla_total_size(4) /* RTA_DST */ 2305 ; 2306 2307 if (!unresolved) 2308 len = len 2309 + nla_total_size(4) /* RTA_IIF */ 2310 + nla_total_size(0) /* RTA_MULTIPATH */ 2311 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2312 /* RTA_MFC_STATS */ 2313 + nla_total_size_64bit(sizeof(struct rta_mfc_stats)) 2314 ; 2315 2316 return len; 2317 } 2318 2319 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2320 int cmd) 2321 { 2322 struct net *net = read_pnet(&mrt->net); 2323 struct sk_buff *skb; 2324 int err = -ENOBUFS; 2325 2326 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif), 2327 GFP_ATOMIC); 2328 if (!skb) 2329 goto errout; 2330 2331 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2332 if (err < 0) 2333 goto errout; 2334 2335 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2336 return; 2337 2338 errout: 2339 kfree_skb(skb); 2340 if (err < 0) 2341 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2342 } 2343 2344 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2345 { 2346 struct net *net = sock_net(skb->sk); 2347 struct mr_table *mrt; 2348 struct mfc_cache *mfc; 2349 unsigned int t = 0, s_t; 2350 unsigned int e = 0, s_e; 2351 2352 s_t = cb->args[0]; 2353 s_e = cb->args[1]; 2354 2355 rcu_read_lock(); 2356 ipmr_for_each_table(mrt, net) { 2357 if (t < s_t) 2358 goto next_table; 2359 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) { 2360 if (e < s_e) 2361 goto next_entry; 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 goto done; 2368 next_entry: 2369 e++; 2370 } 2371 e = 0; 2372 s_e = 0; 2373 2374 spin_lock_bh(&mfc_unres_lock); 2375 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) { 2376 if (e < s_e) 2377 goto next_entry2; 2378 if (ipmr_fill_mroute(mrt, skb, 2379 NETLINK_CB(cb->skb).portid, 2380 cb->nlh->nlmsg_seq, 2381 mfc, RTM_NEWROUTE, 2382 NLM_F_MULTI) < 0) { 2383 spin_unlock_bh(&mfc_unres_lock); 2384 goto done; 2385 } 2386 next_entry2: 2387 e++; 2388 } 2389 spin_unlock_bh(&mfc_unres_lock); 2390 e = 0; 2391 s_e = 0; 2392 next_table: 2393 t++; 2394 } 2395 done: 2396 rcu_read_unlock(); 2397 2398 cb->args[1] = e; 2399 cb->args[0] = t; 2400 2401 return skb->len; 2402 } 2403 2404 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = { 2405 [RTA_SRC] = { .type = NLA_U32 }, 2406 [RTA_DST] = { .type = NLA_U32 }, 2407 [RTA_IIF] = { .type = NLA_U32 }, 2408 [RTA_TABLE] = { .type = NLA_U32 }, 2409 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, 2410 }; 2411 2412 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol) 2413 { 2414 switch (rtm_protocol) { 2415 case RTPROT_STATIC: 2416 case RTPROT_MROUTED: 2417 return true; 2418 } 2419 return false; 2420 } 2421 2422 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc) 2423 { 2424 struct rtnexthop *rtnh = nla_data(nla); 2425 int remaining = nla_len(nla), vifi = 0; 2426 2427 while (rtnh_ok(rtnh, remaining)) { 2428 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops; 2429 if (++vifi == MAXVIFS) 2430 break; 2431 rtnh = rtnh_next(rtnh, &remaining); 2432 } 2433 2434 return remaining > 0 ? -EINVAL : vifi; 2435 } 2436 2437 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */ 2438 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh, 2439 struct mfcctl *mfcc, int *mrtsock, 2440 struct mr_table **mrtret, 2441 struct netlink_ext_ack *extack) 2442 { 2443 struct net_device *dev = NULL; 2444 u32 tblid = RT_TABLE_DEFAULT; 2445 struct mr_table *mrt; 2446 struct nlattr *attr; 2447 struct rtmsg *rtm; 2448 int ret, rem; 2449 2450 ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy, 2451 extack); 2452 if (ret < 0) 2453 goto out; 2454 rtm = nlmsg_data(nlh); 2455 2456 ret = -EINVAL; 2457 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 || 2458 rtm->rtm_type != RTN_MULTICAST || 2459 rtm->rtm_scope != RT_SCOPE_UNIVERSE || 2460 !ipmr_rtm_validate_proto(rtm->rtm_protocol)) 2461 goto out; 2462 2463 memset(mfcc, 0, sizeof(*mfcc)); 2464 mfcc->mfcc_parent = -1; 2465 ret = 0; 2466 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) { 2467 switch (nla_type(attr)) { 2468 case RTA_SRC: 2469 mfcc->mfcc_origin.s_addr = nla_get_be32(attr); 2470 break; 2471 case RTA_DST: 2472 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr); 2473 break; 2474 case RTA_IIF: 2475 dev = __dev_get_by_index(net, nla_get_u32(attr)); 2476 if (!dev) { 2477 ret = -ENODEV; 2478 goto out; 2479 } 2480 break; 2481 case RTA_MULTIPATH: 2482 if (ipmr_nla_get_ttls(attr, mfcc) < 0) { 2483 ret = -EINVAL; 2484 goto out; 2485 } 2486 break; 2487 case RTA_PREFSRC: 2488 ret = 1; 2489 break; 2490 case RTA_TABLE: 2491 tblid = nla_get_u32(attr); 2492 break; 2493 } 2494 } 2495 mrt = ipmr_get_table(net, tblid); 2496 if (!mrt) { 2497 ret = -ENOENT; 2498 goto out; 2499 } 2500 *mrtret = mrt; 2501 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0; 2502 if (dev) 2503 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev); 2504 2505 out: 2506 return ret; 2507 } 2508 2509 /* takes care of both newroute and delroute */ 2510 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh, 2511 struct netlink_ext_ack *extack) 2512 { 2513 struct net *net = sock_net(skb->sk); 2514 int ret, mrtsock, parent; 2515 struct mr_table *tbl; 2516 struct mfcctl mfcc; 2517 2518 mrtsock = 0; 2519 tbl = NULL; 2520 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack); 2521 if (ret < 0) 2522 return ret; 2523 2524 parent = ret ? mfcc.mfcc_parent : -1; 2525 if (nlh->nlmsg_type == RTM_NEWROUTE) 2526 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent); 2527 else 2528 return ipmr_mfc_delete(tbl, &mfcc, parent); 2529 } 2530 2531 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb) 2532 { 2533 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len); 2534 2535 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) || 2536 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) || 2537 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM, 2538 mrt->mroute_reg_vif_num) || 2539 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT, 2540 mrt->mroute_do_assert) || 2541 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim)) 2542 return false; 2543 2544 return true; 2545 } 2546 2547 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb) 2548 { 2549 struct nlattr *vif_nest; 2550 struct vif_device *vif; 2551 2552 /* if the VIF doesn't exist just continue */ 2553 if (!VIF_EXISTS(mrt, vifid)) 2554 return true; 2555 2556 vif = &mrt->vif_table[vifid]; 2557 vif_nest = nla_nest_start(skb, IPMRA_VIF); 2558 if (!vif_nest) 2559 return false; 2560 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) || 2561 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) || 2562 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) || 2563 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in, 2564 IPMRA_VIFA_PAD) || 2565 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out, 2566 IPMRA_VIFA_PAD) || 2567 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in, 2568 IPMRA_VIFA_PAD) || 2569 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out, 2570 IPMRA_VIFA_PAD) || 2571 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) || 2572 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) { 2573 nla_nest_cancel(skb, vif_nest); 2574 return false; 2575 } 2576 nla_nest_end(skb, vif_nest); 2577 2578 return true; 2579 } 2580 2581 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb) 2582 { 2583 struct net *net = sock_net(skb->sk); 2584 struct nlmsghdr *nlh = NULL; 2585 unsigned int t = 0, s_t; 2586 unsigned int e = 0, s_e; 2587 struct mr_table *mrt; 2588 2589 s_t = cb->args[0]; 2590 s_e = cb->args[1]; 2591 2592 ipmr_for_each_table(mrt, net) { 2593 struct nlattr *vifs, *af; 2594 struct ifinfomsg *hdr; 2595 u32 i; 2596 2597 if (t < s_t) 2598 goto skip_table; 2599 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, 2600 cb->nlh->nlmsg_seq, RTM_NEWLINK, 2601 sizeof(*hdr), NLM_F_MULTI); 2602 if (!nlh) 2603 break; 2604 2605 hdr = nlmsg_data(nlh); 2606 memset(hdr, 0, sizeof(*hdr)); 2607 hdr->ifi_family = RTNL_FAMILY_IPMR; 2608 2609 af = nla_nest_start(skb, IFLA_AF_SPEC); 2610 if (!af) { 2611 nlmsg_cancel(skb, nlh); 2612 goto out; 2613 } 2614 2615 if (!ipmr_fill_table(mrt, skb)) { 2616 nlmsg_cancel(skb, nlh); 2617 goto out; 2618 } 2619 2620 vifs = nla_nest_start(skb, IPMRA_TABLE_VIFS); 2621 if (!vifs) { 2622 nla_nest_end(skb, af); 2623 nlmsg_end(skb, nlh); 2624 goto out; 2625 } 2626 for (i = 0; i < mrt->maxvif; i++) { 2627 if (e < s_e) 2628 goto skip_entry; 2629 if (!ipmr_fill_vif(mrt, i, skb)) { 2630 nla_nest_end(skb, vifs); 2631 nla_nest_end(skb, af); 2632 nlmsg_end(skb, nlh); 2633 goto out; 2634 } 2635 skip_entry: 2636 e++; 2637 } 2638 s_e = 0; 2639 e = 0; 2640 nla_nest_end(skb, vifs); 2641 nla_nest_end(skb, af); 2642 nlmsg_end(skb, nlh); 2643 skip_table: 2644 t++; 2645 } 2646 2647 out: 2648 cb->args[1] = e; 2649 cb->args[0] = t; 2650 2651 return skb->len; 2652 } 2653 2654 #ifdef CONFIG_PROC_FS 2655 /* The /proc interfaces to multicast routing : 2656 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 2657 */ 2658 struct ipmr_vif_iter { 2659 struct seq_net_private p; 2660 struct mr_table *mrt; 2661 int ct; 2662 }; 2663 2664 static struct vif_device *ipmr_vif_seq_idx(struct net *net, 2665 struct ipmr_vif_iter *iter, 2666 loff_t pos) 2667 { 2668 struct mr_table *mrt = iter->mrt; 2669 2670 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 2671 if (!VIF_EXISTS(mrt, iter->ct)) 2672 continue; 2673 if (pos-- == 0) 2674 return &mrt->vif_table[iter->ct]; 2675 } 2676 return NULL; 2677 } 2678 2679 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 2680 __acquires(mrt_lock) 2681 { 2682 struct ipmr_vif_iter *iter = seq->private; 2683 struct net *net = seq_file_net(seq); 2684 struct mr_table *mrt; 2685 2686 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2687 if (!mrt) 2688 return ERR_PTR(-ENOENT); 2689 2690 iter->mrt = mrt; 2691 2692 read_lock(&mrt_lock); 2693 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1) 2694 : SEQ_START_TOKEN; 2695 } 2696 2697 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2698 { 2699 struct ipmr_vif_iter *iter = seq->private; 2700 struct net *net = seq_file_net(seq); 2701 struct mr_table *mrt = iter->mrt; 2702 2703 ++*pos; 2704 if (v == SEQ_START_TOKEN) 2705 return ipmr_vif_seq_idx(net, iter, 0); 2706 2707 while (++iter->ct < mrt->maxvif) { 2708 if (!VIF_EXISTS(mrt, iter->ct)) 2709 continue; 2710 return &mrt->vif_table[iter->ct]; 2711 } 2712 return NULL; 2713 } 2714 2715 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 2716 __releases(mrt_lock) 2717 { 2718 read_unlock(&mrt_lock); 2719 } 2720 2721 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 2722 { 2723 struct ipmr_vif_iter *iter = seq->private; 2724 struct mr_table *mrt = iter->mrt; 2725 2726 if (v == SEQ_START_TOKEN) { 2727 seq_puts(seq, 2728 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 2729 } else { 2730 const struct vif_device *vif = v; 2731 const char *name = vif->dev ? vif->dev->name : "none"; 2732 2733 seq_printf(seq, 2734 "%2zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 2735 vif - mrt->vif_table, 2736 name, vif->bytes_in, vif->pkt_in, 2737 vif->bytes_out, vif->pkt_out, 2738 vif->flags, vif->local, vif->remote); 2739 } 2740 return 0; 2741 } 2742 2743 static const struct seq_operations ipmr_vif_seq_ops = { 2744 .start = ipmr_vif_seq_start, 2745 .next = ipmr_vif_seq_next, 2746 .stop = ipmr_vif_seq_stop, 2747 .show = ipmr_vif_seq_show, 2748 }; 2749 2750 static int ipmr_vif_open(struct inode *inode, struct file *file) 2751 { 2752 return seq_open_net(inode, file, &ipmr_vif_seq_ops, 2753 sizeof(struct ipmr_vif_iter)); 2754 } 2755 2756 static const struct file_operations ipmr_vif_fops = { 2757 .owner = THIS_MODULE, 2758 .open = ipmr_vif_open, 2759 .read = seq_read, 2760 .llseek = seq_lseek, 2761 .release = seq_release_net, 2762 }; 2763 2764 struct ipmr_mfc_iter { 2765 struct seq_net_private p; 2766 struct mr_table *mrt; 2767 struct list_head *cache; 2768 }; 2769 2770 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net, 2771 struct ipmr_mfc_iter *it, loff_t pos) 2772 { 2773 struct mr_table *mrt = it->mrt; 2774 struct mfc_cache *mfc; 2775 2776 rcu_read_lock(); 2777 it->cache = &mrt->mfc_cache_list; 2778 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) 2779 if (pos-- == 0) 2780 return mfc; 2781 rcu_read_unlock(); 2782 2783 spin_lock_bh(&mfc_unres_lock); 2784 it->cache = &mrt->mfc_unres_queue; 2785 list_for_each_entry(mfc, it->cache, list) 2786 if (pos-- == 0) 2787 return mfc; 2788 spin_unlock_bh(&mfc_unres_lock); 2789 2790 it->cache = NULL; 2791 return NULL; 2792 } 2793 2794 2795 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 2796 { 2797 struct ipmr_mfc_iter *it = seq->private; 2798 struct net *net = seq_file_net(seq); 2799 struct mr_table *mrt; 2800 2801 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2802 if (!mrt) 2803 return ERR_PTR(-ENOENT); 2804 2805 it->mrt = mrt; 2806 it->cache = NULL; 2807 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 2808 : SEQ_START_TOKEN; 2809 } 2810 2811 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2812 { 2813 struct ipmr_mfc_iter *it = seq->private; 2814 struct net *net = seq_file_net(seq); 2815 struct mr_table *mrt = it->mrt; 2816 struct mfc_cache *mfc = v; 2817 2818 ++*pos; 2819 2820 if (v == SEQ_START_TOKEN) 2821 return ipmr_mfc_seq_idx(net, seq->private, 0); 2822 2823 if (mfc->list.next != it->cache) 2824 return list_entry(mfc->list.next, struct mfc_cache, list); 2825 2826 if (it->cache == &mrt->mfc_unres_queue) 2827 goto end_of_list; 2828 2829 /* exhausted cache_array, show unresolved */ 2830 rcu_read_unlock(); 2831 it->cache = &mrt->mfc_unres_queue; 2832 2833 spin_lock_bh(&mfc_unres_lock); 2834 if (!list_empty(it->cache)) 2835 return list_first_entry(it->cache, struct mfc_cache, list); 2836 2837 end_of_list: 2838 spin_unlock_bh(&mfc_unres_lock); 2839 it->cache = NULL; 2840 2841 return NULL; 2842 } 2843 2844 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 2845 { 2846 struct ipmr_mfc_iter *it = seq->private; 2847 struct mr_table *mrt = it->mrt; 2848 2849 if (it->cache == &mrt->mfc_unres_queue) 2850 spin_unlock_bh(&mfc_unres_lock); 2851 else if (it->cache == &mrt->mfc_cache_list) 2852 rcu_read_unlock(); 2853 } 2854 2855 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 2856 { 2857 int n; 2858 2859 if (v == SEQ_START_TOKEN) { 2860 seq_puts(seq, 2861 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 2862 } else { 2863 const struct mfc_cache *mfc = v; 2864 const struct ipmr_mfc_iter *it = seq->private; 2865 const struct mr_table *mrt = it->mrt; 2866 2867 seq_printf(seq, "%08X %08X %-3hd", 2868 (__force u32) mfc->mfc_mcastgrp, 2869 (__force u32) mfc->mfc_origin, 2870 mfc->mfc_parent); 2871 2872 if (it->cache != &mrt->mfc_unres_queue) { 2873 seq_printf(seq, " %8lu %8lu %8lu", 2874 mfc->mfc_un.res.pkt, 2875 mfc->mfc_un.res.bytes, 2876 mfc->mfc_un.res.wrong_if); 2877 for (n = mfc->mfc_un.res.minvif; 2878 n < mfc->mfc_un.res.maxvif; n++) { 2879 if (VIF_EXISTS(mrt, n) && 2880 mfc->mfc_un.res.ttls[n] < 255) 2881 seq_printf(seq, 2882 " %2d:%-3d", 2883 n, mfc->mfc_un.res.ttls[n]); 2884 } 2885 } else { 2886 /* unresolved mfc_caches don't contain 2887 * pkt, bytes and wrong_if values 2888 */ 2889 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 2890 } 2891 seq_putc(seq, '\n'); 2892 } 2893 return 0; 2894 } 2895 2896 static const struct seq_operations ipmr_mfc_seq_ops = { 2897 .start = ipmr_mfc_seq_start, 2898 .next = ipmr_mfc_seq_next, 2899 .stop = ipmr_mfc_seq_stop, 2900 .show = ipmr_mfc_seq_show, 2901 }; 2902 2903 static int ipmr_mfc_open(struct inode *inode, struct file *file) 2904 { 2905 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 2906 sizeof(struct ipmr_mfc_iter)); 2907 } 2908 2909 static const struct file_operations ipmr_mfc_fops = { 2910 .owner = THIS_MODULE, 2911 .open = ipmr_mfc_open, 2912 .read = seq_read, 2913 .llseek = seq_lseek, 2914 .release = seq_release_net, 2915 }; 2916 #endif 2917 2918 #ifdef CONFIG_IP_PIMSM_V2 2919 static const struct net_protocol pim_protocol = { 2920 .handler = pim_rcv, 2921 .netns_ok = 1, 2922 }; 2923 #endif 2924 2925 /* Setup for IP multicast routing */ 2926 static int __net_init ipmr_net_init(struct net *net) 2927 { 2928 int err; 2929 2930 err = ipmr_rules_init(net); 2931 if (err < 0) 2932 goto fail; 2933 2934 #ifdef CONFIG_PROC_FS 2935 err = -ENOMEM; 2936 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops)) 2937 goto proc_vif_fail; 2938 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops)) 2939 goto proc_cache_fail; 2940 #endif 2941 return 0; 2942 2943 #ifdef CONFIG_PROC_FS 2944 proc_cache_fail: 2945 remove_proc_entry("ip_mr_vif", net->proc_net); 2946 proc_vif_fail: 2947 ipmr_rules_exit(net); 2948 #endif 2949 fail: 2950 return err; 2951 } 2952 2953 static void __net_exit ipmr_net_exit(struct net *net) 2954 { 2955 #ifdef CONFIG_PROC_FS 2956 remove_proc_entry("ip_mr_cache", net->proc_net); 2957 remove_proc_entry("ip_mr_vif", net->proc_net); 2958 #endif 2959 ipmr_rules_exit(net); 2960 } 2961 2962 static struct pernet_operations ipmr_net_ops = { 2963 .init = ipmr_net_init, 2964 .exit = ipmr_net_exit, 2965 }; 2966 2967 int __init ip_mr_init(void) 2968 { 2969 int err; 2970 2971 mrt_cachep = kmem_cache_create("ip_mrt_cache", 2972 sizeof(struct mfc_cache), 2973 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, 2974 NULL); 2975 2976 err = register_pernet_subsys(&ipmr_net_ops); 2977 if (err) 2978 goto reg_pernet_fail; 2979 2980 err = register_netdevice_notifier(&ip_mr_notifier); 2981 if (err) 2982 goto reg_notif_fail; 2983 #ifdef CONFIG_IP_PIMSM_V2 2984 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 2985 pr_err("%s: can't add PIM protocol\n", __func__); 2986 err = -EAGAIN; 2987 goto add_proto_fail; 2988 } 2989 #endif 2990 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE, 2991 NULL, ipmr_rtm_dumproute, NULL); 2992 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE, 2993 ipmr_rtm_route, NULL, NULL); 2994 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE, 2995 ipmr_rtm_route, NULL, NULL); 2996 2997 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK, 2998 NULL, ipmr_rtm_dumplink, NULL); 2999 return 0; 3000 3001 #ifdef CONFIG_IP_PIMSM_V2 3002 add_proto_fail: 3003 unregister_netdevice_notifier(&ip_mr_notifier); 3004 #endif 3005 reg_notif_fail: 3006 unregister_pernet_subsys(&ipmr_net_ops); 3007 reg_pernet_fail: 3008 kmem_cache_destroy(mrt_cachep); 3009 return err; 3010 } 3011