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 <asm/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 70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 71 #define CONFIG_IP_PIMSM 1 72 #endif 73 74 struct mr_table { 75 struct list_head list; 76 possible_net_t net; 77 u32 id; 78 struct sock __rcu *mroute_sk; 79 struct timer_list ipmr_expire_timer; 80 struct list_head mfc_unres_queue; 81 struct list_head mfc_cache_array[MFC_LINES]; 82 struct vif_device vif_table[MAXVIFS]; 83 int maxvif; 84 atomic_t cache_resolve_queue_len; 85 bool mroute_do_assert; 86 bool mroute_do_pim; 87 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 88 int mroute_reg_vif_num; 89 #endif 90 }; 91 92 struct ipmr_rule { 93 struct fib_rule common; 94 }; 95 96 struct ipmr_result { 97 struct mr_table *mrt; 98 }; 99 100 /* Big lock, protecting vif table, mrt cache and mroute socket state. 101 * Note that the changes are semaphored via rtnl_lock. 102 */ 103 104 static DEFINE_RWLOCK(mrt_lock); 105 106 /* 107 * Multicast router control variables 108 */ 109 110 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL) 111 112 /* Special spinlock for queue of unresolved entries */ 113 static DEFINE_SPINLOCK(mfc_unres_lock); 114 115 /* We return to original Alan's scheme. Hash table of resolved 116 * entries is changed only in process context and protected 117 * with weak lock mrt_lock. Queue of unresolved entries is protected 118 * with strong spinlock mfc_unres_lock. 119 * 120 * In this case data path is free of exclusive locks at all. 121 */ 122 123 static struct kmem_cache *mrt_cachep __read_mostly; 124 125 static struct mr_table *ipmr_new_table(struct net *net, u32 id); 126 static void ipmr_free_table(struct mr_table *mrt); 127 128 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 129 struct sk_buff *skb, struct mfc_cache *cache, 130 int local); 131 static int ipmr_cache_report(struct mr_table *mrt, 132 struct sk_buff *pkt, vifi_t vifi, int assert); 133 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 134 struct mfc_cache *c, struct rtmsg *rtm); 135 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 136 int cmd); 137 static void mroute_clean_tables(struct mr_table *mrt); 138 static void ipmr_expire_process(unsigned long arg); 139 140 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 141 #define ipmr_for_each_table(mrt, net) \ 142 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list) 143 144 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 145 { 146 struct mr_table *mrt; 147 148 ipmr_for_each_table(mrt, net) { 149 if (mrt->id == id) 150 return mrt; 151 } 152 return NULL; 153 } 154 155 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 156 struct mr_table **mrt) 157 { 158 int err; 159 struct ipmr_result res; 160 struct fib_lookup_arg arg = { 161 .result = &res, 162 .flags = FIB_LOOKUP_NOREF, 163 }; 164 165 err = fib_rules_lookup(net->ipv4.mr_rules_ops, 166 flowi4_to_flowi(flp4), 0, &arg); 167 if (err < 0) 168 return err; 169 *mrt = res.mrt; 170 return 0; 171 } 172 173 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp, 174 int flags, struct fib_lookup_arg *arg) 175 { 176 struct ipmr_result *res = arg->result; 177 struct mr_table *mrt; 178 179 switch (rule->action) { 180 case FR_ACT_TO_TBL: 181 break; 182 case FR_ACT_UNREACHABLE: 183 return -ENETUNREACH; 184 case FR_ACT_PROHIBIT: 185 return -EACCES; 186 case FR_ACT_BLACKHOLE: 187 default: 188 return -EINVAL; 189 } 190 191 mrt = ipmr_get_table(rule->fr_net, rule->table); 192 if (mrt == NULL) 193 return -EAGAIN; 194 res->mrt = mrt; 195 return 0; 196 } 197 198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 199 { 200 return 1; 201 } 202 203 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = { 204 FRA_GENERIC_POLICY, 205 }; 206 207 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 208 struct fib_rule_hdr *frh, struct nlattr **tb) 209 { 210 return 0; 211 } 212 213 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 214 struct nlattr **tb) 215 { 216 return 1; 217 } 218 219 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 220 struct fib_rule_hdr *frh) 221 { 222 frh->dst_len = 0; 223 frh->src_len = 0; 224 frh->tos = 0; 225 return 0; 226 } 227 228 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = { 229 .family = RTNL_FAMILY_IPMR, 230 .rule_size = sizeof(struct ipmr_rule), 231 .addr_size = sizeof(u32), 232 .action = ipmr_rule_action, 233 .match = ipmr_rule_match, 234 .configure = ipmr_rule_configure, 235 .compare = ipmr_rule_compare, 236 .default_pref = fib_default_rule_pref, 237 .fill = ipmr_rule_fill, 238 .nlgroup = RTNLGRP_IPV4_RULE, 239 .policy = ipmr_rule_policy, 240 .owner = THIS_MODULE, 241 }; 242 243 static int __net_init ipmr_rules_init(struct net *net) 244 { 245 struct fib_rules_ops *ops; 246 struct mr_table *mrt; 247 int err; 248 249 ops = fib_rules_register(&ipmr_rules_ops_template, net); 250 if (IS_ERR(ops)) 251 return PTR_ERR(ops); 252 253 INIT_LIST_HEAD(&net->ipv4.mr_tables); 254 255 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 256 if (mrt == NULL) { 257 err = -ENOMEM; 258 goto err1; 259 } 260 261 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0); 262 if (err < 0) 263 goto err2; 264 265 net->ipv4.mr_rules_ops = ops; 266 return 0; 267 268 err2: 269 kfree(mrt); 270 err1: 271 fib_rules_unregister(ops); 272 return err; 273 } 274 275 static void __net_exit ipmr_rules_exit(struct net *net) 276 { 277 struct mr_table *mrt, *next; 278 279 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) { 280 list_del(&mrt->list); 281 ipmr_free_table(mrt); 282 } 283 fib_rules_unregister(net->ipv4.mr_rules_ops); 284 } 285 #else 286 #define ipmr_for_each_table(mrt, net) \ 287 for (mrt = net->ipv4.mrt; mrt; mrt = NULL) 288 289 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 290 { 291 return net->ipv4.mrt; 292 } 293 294 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 295 struct mr_table **mrt) 296 { 297 *mrt = net->ipv4.mrt; 298 return 0; 299 } 300 301 static int __net_init ipmr_rules_init(struct net *net) 302 { 303 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 304 return net->ipv4.mrt ? 0 : -ENOMEM; 305 } 306 307 static void __net_exit ipmr_rules_exit(struct net *net) 308 { 309 ipmr_free_table(net->ipv4.mrt); 310 } 311 #endif 312 313 static struct mr_table *ipmr_new_table(struct net *net, u32 id) 314 { 315 struct mr_table *mrt; 316 unsigned int i; 317 318 mrt = ipmr_get_table(net, id); 319 if (mrt != NULL) 320 return mrt; 321 322 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 323 if (mrt == NULL) 324 return NULL; 325 write_pnet(&mrt->net, net); 326 mrt->id = id; 327 328 /* Forwarding cache */ 329 for (i = 0; i < MFC_LINES; i++) 330 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]); 331 332 INIT_LIST_HEAD(&mrt->mfc_unres_queue); 333 334 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 335 (unsigned long)mrt); 336 337 #ifdef CONFIG_IP_PIMSM 338 mrt->mroute_reg_vif_num = -1; 339 #endif 340 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 341 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables); 342 #endif 343 return mrt; 344 } 345 346 static void ipmr_free_table(struct mr_table *mrt) 347 { 348 del_timer_sync(&mrt->ipmr_expire_timer); 349 mroute_clean_tables(mrt); 350 kfree(mrt); 351 } 352 353 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ 354 355 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v) 356 { 357 struct net *net = dev_net(dev); 358 359 dev_close(dev); 360 361 dev = __dev_get_by_name(net, "tunl0"); 362 if (dev) { 363 const struct net_device_ops *ops = dev->netdev_ops; 364 struct ifreq ifr; 365 struct ip_tunnel_parm p; 366 367 memset(&p, 0, sizeof(p)); 368 p.iph.daddr = v->vifc_rmt_addr.s_addr; 369 p.iph.saddr = v->vifc_lcl_addr.s_addr; 370 p.iph.version = 4; 371 p.iph.ihl = 5; 372 p.iph.protocol = IPPROTO_IPIP; 373 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 374 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 375 376 if (ops->ndo_do_ioctl) { 377 mm_segment_t oldfs = get_fs(); 378 379 set_fs(KERNEL_DS); 380 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL); 381 set_fs(oldfs); 382 } 383 } 384 } 385 386 static 387 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v) 388 { 389 struct net_device *dev; 390 391 dev = __dev_get_by_name(net, "tunl0"); 392 393 if (dev) { 394 const struct net_device_ops *ops = dev->netdev_ops; 395 int err; 396 struct ifreq ifr; 397 struct ip_tunnel_parm p; 398 struct in_device *in_dev; 399 400 memset(&p, 0, sizeof(p)); 401 p.iph.daddr = v->vifc_rmt_addr.s_addr; 402 p.iph.saddr = v->vifc_lcl_addr.s_addr; 403 p.iph.version = 4; 404 p.iph.ihl = 5; 405 p.iph.protocol = IPPROTO_IPIP; 406 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 407 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 408 409 if (ops->ndo_do_ioctl) { 410 mm_segment_t oldfs = get_fs(); 411 412 set_fs(KERNEL_DS); 413 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL); 414 set_fs(oldfs); 415 } else { 416 err = -EOPNOTSUPP; 417 } 418 dev = NULL; 419 420 if (err == 0 && 421 (dev = __dev_get_by_name(net, p.name)) != NULL) { 422 dev->flags |= IFF_MULTICAST; 423 424 in_dev = __in_dev_get_rtnl(dev); 425 if (in_dev == NULL) 426 goto failure; 427 428 ipv4_devconf_setall(in_dev); 429 neigh_parms_data_state_setall(in_dev->arp_parms); 430 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 431 432 if (dev_open(dev)) 433 goto failure; 434 dev_hold(dev); 435 } 436 } 437 return dev; 438 439 failure: 440 /* allow the register to be completed before unregistering. */ 441 rtnl_unlock(); 442 rtnl_lock(); 443 444 unregister_netdevice(dev); 445 return NULL; 446 } 447 448 #ifdef CONFIG_IP_PIMSM 449 450 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) 451 { 452 struct net *net = dev_net(dev); 453 struct mr_table *mrt; 454 struct flowi4 fl4 = { 455 .flowi4_oif = dev->ifindex, 456 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 457 .flowi4_mark = skb->mark, 458 }; 459 int err; 460 461 err = ipmr_fib_lookup(net, &fl4, &mrt); 462 if (err < 0) { 463 kfree_skb(skb); 464 return err; 465 } 466 467 read_lock(&mrt_lock); 468 dev->stats.tx_bytes += skb->len; 469 dev->stats.tx_packets++; 470 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT); 471 read_unlock(&mrt_lock); 472 kfree_skb(skb); 473 return NETDEV_TX_OK; 474 } 475 476 static const struct net_device_ops reg_vif_netdev_ops = { 477 .ndo_start_xmit = reg_vif_xmit, 478 }; 479 480 static void reg_vif_setup(struct net_device *dev) 481 { 482 dev->type = ARPHRD_PIMREG; 483 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; 484 dev->flags = IFF_NOARP; 485 dev->netdev_ops = ®_vif_netdev_ops; 486 dev->destructor = free_netdev; 487 dev->features |= NETIF_F_NETNS_LOCAL; 488 } 489 490 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 491 { 492 struct net_device *dev; 493 struct in_device *in_dev; 494 char name[IFNAMSIZ]; 495 496 if (mrt->id == RT_TABLE_DEFAULT) 497 sprintf(name, "pimreg"); 498 else 499 sprintf(name, "pimreg%u", mrt->id); 500 501 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 502 503 if (dev == NULL) 504 return NULL; 505 506 dev_net_set(dev, net); 507 508 if (register_netdevice(dev)) { 509 free_netdev(dev); 510 return NULL; 511 } 512 dev->iflink = 0; 513 514 rcu_read_lock(); 515 in_dev = __in_dev_get_rcu(dev); 516 if (!in_dev) { 517 rcu_read_unlock(); 518 goto failure; 519 } 520 521 ipv4_devconf_setall(in_dev); 522 neigh_parms_data_state_setall(in_dev->arp_parms); 523 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 524 rcu_read_unlock(); 525 526 if (dev_open(dev)) 527 goto failure; 528 529 dev_hold(dev); 530 531 return dev; 532 533 failure: 534 /* allow the register to be completed before unregistering. */ 535 rtnl_unlock(); 536 rtnl_lock(); 537 538 unregister_netdevice(dev); 539 return NULL; 540 } 541 #endif 542 543 /** 544 * vif_delete - Delete a VIF entry 545 * @notify: Set to 1, if the caller is a notifier_call 546 */ 547 548 static int vif_delete(struct mr_table *mrt, int vifi, int notify, 549 struct list_head *head) 550 { 551 struct vif_device *v; 552 struct net_device *dev; 553 struct in_device *in_dev; 554 555 if (vifi < 0 || vifi >= mrt->maxvif) 556 return -EADDRNOTAVAIL; 557 558 v = &mrt->vif_table[vifi]; 559 560 write_lock_bh(&mrt_lock); 561 dev = v->dev; 562 v->dev = NULL; 563 564 if (!dev) { 565 write_unlock_bh(&mrt_lock); 566 return -EADDRNOTAVAIL; 567 } 568 569 #ifdef CONFIG_IP_PIMSM 570 if (vifi == mrt->mroute_reg_vif_num) 571 mrt->mroute_reg_vif_num = -1; 572 #endif 573 574 if (vifi + 1 == mrt->maxvif) { 575 int tmp; 576 577 for (tmp = vifi - 1; tmp >= 0; tmp--) { 578 if (VIF_EXISTS(mrt, tmp)) 579 break; 580 } 581 mrt->maxvif = tmp+1; 582 } 583 584 write_unlock_bh(&mrt_lock); 585 586 dev_set_allmulti(dev, -1); 587 588 in_dev = __in_dev_get_rtnl(dev); 589 if (in_dev) { 590 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; 591 inet_netconf_notify_devconf(dev_net(dev), 592 NETCONFA_MC_FORWARDING, 593 dev->ifindex, &in_dev->cnf); 594 ip_rt_multicast_event(in_dev); 595 } 596 597 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify) 598 unregister_netdevice_queue(dev, head); 599 600 dev_put(dev); 601 return 0; 602 } 603 604 static void ipmr_cache_free_rcu(struct rcu_head *head) 605 { 606 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu); 607 608 kmem_cache_free(mrt_cachep, c); 609 } 610 611 static inline void ipmr_cache_free(struct mfc_cache *c) 612 { 613 call_rcu(&c->rcu, ipmr_cache_free_rcu); 614 } 615 616 /* Destroy an unresolved cache entry, killing queued skbs 617 * and reporting error to netlink readers. 618 */ 619 620 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c) 621 { 622 struct net *net = read_pnet(&mrt->net); 623 struct sk_buff *skb; 624 struct nlmsgerr *e; 625 626 atomic_dec(&mrt->cache_resolve_queue_len); 627 628 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) { 629 if (ip_hdr(skb)->version == 0) { 630 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); 631 nlh->nlmsg_type = NLMSG_ERROR; 632 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 633 skb_trim(skb, nlh->nlmsg_len); 634 e = nlmsg_data(nlh); 635 e->error = -ETIMEDOUT; 636 memset(&e->msg, 0, sizeof(e->msg)); 637 638 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 639 } else { 640 kfree_skb(skb); 641 } 642 } 643 644 ipmr_cache_free(c); 645 } 646 647 648 /* Timer process for the unresolved queue. */ 649 650 static void ipmr_expire_process(unsigned long arg) 651 { 652 struct mr_table *mrt = (struct mr_table *)arg; 653 unsigned long now; 654 unsigned long expires; 655 struct mfc_cache *c, *next; 656 657 if (!spin_trylock(&mfc_unres_lock)) { 658 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 659 return; 660 } 661 662 if (list_empty(&mrt->mfc_unres_queue)) 663 goto out; 664 665 now = jiffies; 666 expires = 10*HZ; 667 668 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 669 if (time_after(c->mfc_un.unres.expires, now)) { 670 unsigned long interval = c->mfc_un.unres.expires - now; 671 if (interval < expires) 672 expires = interval; 673 continue; 674 } 675 676 list_del(&c->list); 677 mroute_netlink_event(mrt, c, RTM_DELROUTE); 678 ipmr_destroy_unres(mrt, c); 679 } 680 681 if (!list_empty(&mrt->mfc_unres_queue)) 682 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 683 684 out: 685 spin_unlock(&mfc_unres_lock); 686 } 687 688 /* Fill oifs list. It is called under write locked mrt_lock. */ 689 690 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache, 691 unsigned char *ttls) 692 { 693 int vifi; 694 695 cache->mfc_un.res.minvif = MAXVIFS; 696 cache->mfc_un.res.maxvif = 0; 697 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 698 699 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 700 if (VIF_EXISTS(mrt, vifi) && 701 ttls[vifi] && ttls[vifi] < 255) { 702 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 703 if (cache->mfc_un.res.minvif > vifi) 704 cache->mfc_un.res.minvif = vifi; 705 if (cache->mfc_un.res.maxvif <= vifi) 706 cache->mfc_un.res.maxvif = vifi + 1; 707 } 708 } 709 } 710 711 static int vif_add(struct net *net, struct mr_table *mrt, 712 struct vifctl *vifc, int mrtsock) 713 { 714 int vifi = vifc->vifc_vifi; 715 struct vif_device *v = &mrt->vif_table[vifi]; 716 struct net_device *dev; 717 struct in_device *in_dev; 718 int err; 719 720 /* Is vif busy ? */ 721 if (VIF_EXISTS(mrt, vifi)) 722 return -EADDRINUSE; 723 724 switch (vifc->vifc_flags) { 725 #ifdef CONFIG_IP_PIMSM 726 case VIFF_REGISTER: 727 /* 728 * Special Purpose VIF in PIM 729 * All the packets will be sent to the daemon 730 */ 731 if (mrt->mroute_reg_vif_num >= 0) 732 return -EADDRINUSE; 733 dev = ipmr_reg_vif(net, mrt); 734 if (!dev) 735 return -ENOBUFS; 736 err = dev_set_allmulti(dev, 1); 737 if (err) { 738 unregister_netdevice(dev); 739 dev_put(dev); 740 return err; 741 } 742 break; 743 #endif 744 case VIFF_TUNNEL: 745 dev = ipmr_new_tunnel(net, vifc); 746 if (!dev) 747 return -ENOBUFS; 748 err = dev_set_allmulti(dev, 1); 749 if (err) { 750 ipmr_del_tunnel(dev, vifc); 751 dev_put(dev); 752 return err; 753 } 754 break; 755 756 case VIFF_USE_IFINDEX: 757 case 0: 758 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 759 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 760 if (dev && __in_dev_get_rtnl(dev) == NULL) { 761 dev_put(dev); 762 return -EADDRNOTAVAIL; 763 } 764 } else { 765 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 766 } 767 if (!dev) 768 return -EADDRNOTAVAIL; 769 err = dev_set_allmulti(dev, 1); 770 if (err) { 771 dev_put(dev); 772 return err; 773 } 774 break; 775 default: 776 return -EINVAL; 777 } 778 779 in_dev = __in_dev_get_rtnl(dev); 780 if (!in_dev) { 781 dev_put(dev); 782 return -EADDRNOTAVAIL; 783 } 784 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 785 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex, 786 &in_dev->cnf); 787 ip_rt_multicast_event(in_dev); 788 789 /* Fill in the VIF structures */ 790 791 v->rate_limit = vifc->vifc_rate_limit; 792 v->local = vifc->vifc_lcl_addr.s_addr; 793 v->remote = vifc->vifc_rmt_addr.s_addr; 794 v->flags = vifc->vifc_flags; 795 if (!mrtsock) 796 v->flags |= VIFF_STATIC; 797 v->threshold = vifc->vifc_threshold; 798 v->bytes_in = 0; 799 v->bytes_out = 0; 800 v->pkt_in = 0; 801 v->pkt_out = 0; 802 v->link = dev->ifindex; 803 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER)) 804 v->link = dev->iflink; 805 806 /* And finish update writing critical data */ 807 write_lock_bh(&mrt_lock); 808 v->dev = dev; 809 #ifdef CONFIG_IP_PIMSM 810 if (v->flags & VIFF_REGISTER) 811 mrt->mroute_reg_vif_num = vifi; 812 #endif 813 if (vifi+1 > mrt->maxvif) 814 mrt->maxvif = vifi+1; 815 write_unlock_bh(&mrt_lock); 816 return 0; 817 } 818 819 /* called with rcu_read_lock() */ 820 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 821 __be32 origin, 822 __be32 mcastgrp) 823 { 824 int line = MFC_HASH(mcastgrp, origin); 825 struct mfc_cache *c; 826 827 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) { 828 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp) 829 return c; 830 } 831 return NULL; 832 } 833 834 /* Look for a (*,*,oif) entry */ 835 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt, 836 int vifi) 837 { 838 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY)); 839 struct mfc_cache *c; 840 841 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) 842 if (c->mfc_origin == htonl(INADDR_ANY) && 843 c->mfc_mcastgrp == htonl(INADDR_ANY) && 844 c->mfc_un.res.ttls[vifi] < 255) 845 return c; 846 847 return NULL; 848 } 849 850 /* Look for a (*,G) entry */ 851 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 852 __be32 mcastgrp, int vifi) 853 { 854 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY)); 855 struct mfc_cache *c, *proxy; 856 857 if (mcastgrp == htonl(INADDR_ANY)) 858 goto skip; 859 860 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) 861 if (c->mfc_origin == htonl(INADDR_ANY) && 862 c->mfc_mcastgrp == mcastgrp) { 863 if (c->mfc_un.res.ttls[vifi] < 255) 864 return c; 865 866 /* It's ok if the vifi is part of the static tree */ 867 proxy = ipmr_cache_find_any_parent(mrt, 868 c->mfc_parent); 869 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255) 870 return c; 871 } 872 873 skip: 874 return ipmr_cache_find_any_parent(mrt, vifi); 875 } 876 877 /* 878 * Allocate a multicast cache entry 879 */ 880 static struct mfc_cache *ipmr_cache_alloc(void) 881 { 882 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 883 884 if (c) 885 c->mfc_un.res.minvif = MAXVIFS; 886 return c; 887 } 888 889 static struct mfc_cache *ipmr_cache_alloc_unres(void) 890 { 891 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 892 893 if (c) { 894 skb_queue_head_init(&c->mfc_un.unres.unresolved); 895 c->mfc_un.unres.expires = jiffies + 10*HZ; 896 } 897 return c; 898 } 899 900 /* 901 * A cache entry has gone into a resolved state from queued 902 */ 903 904 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 905 struct mfc_cache *uc, struct mfc_cache *c) 906 { 907 struct sk_buff *skb; 908 struct nlmsgerr *e; 909 910 /* Play the pending entries through our router */ 911 912 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 913 if (ip_hdr(skb)->version == 0) { 914 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); 915 916 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 917 nlh->nlmsg_len = skb_tail_pointer(skb) - 918 (u8 *)nlh; 919 } else { 920 nlh->nlmsg_type = NLMSG_ERROR; 921 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 922 skb_trim(skb, nlh->nlmsg_len); 923 e = nlmsg_data(nlh); 924 e->error = -EMSGSIZE; 925 memset(&e->msg, 0, sizeof(e->msg)); 926 } 927 928 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 929 } else { 930 ip_mr_forward(net, mrt, skb, c, 0); 931 } 932 } 933 } 934 935 /* 936 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted 937 * expects the following bizarre scheme. 938 * 939 * Called under mrt_lock. 940 */ 941 942 static int ipmr_cache_report(struct mr_table *mrt, 943 struct sk_buff *pkt, vifi_t vifi, int assert) 944 { 945 struct sk_buff *skb; 946 const int ihl = ip_hdrlen(pkt); 947 struct igmphdr *igmp; 948 struct igmpmsg *msg; 949 struct sock *mroute_sk; 950 int ret; 951 952 #ifdef CONFIG_IP_PIMSM 953 if (assert == IGMPMSG_WHOLEPKT) 954 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 955 else 956 #endif 957 skb = alloc_skb(128, GFP_ATOMIC); 958 959 if (!skb) 960 return -ENOBUFS; 961 962 #ifdef CONFIG_IP_PIMSM 963 if (assert == IGMPMSG_WHOLEPKT) { 964 /* Ugly, but we have no choice with this interface. 965 * Duplicate old header, fix ihl, length etc. 966 * And all this only to mangle msg->im_msgtype and 967 * to set msg->im_mbz to "mbz" :-) 968 */ 969 skb_push(skb, sizeof(struct iphdr)); 970 skb_reset_network_header(skb); 971 skb_reset_transport_header(skb); 972 msg = (struct igmpmsg *)skb_network_header(skb); 973 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 974 msg->im_msgtype = IGMPMSG_WHOLEPKT; 975 msg->im_mbz = 0; 976 msg->im_vif = mrt->mroute_reg_vif_num; 977 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 978 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 979 sizeof(struct iphdr)); 980 } else 981 #endif 982 { 983 984 /* Copy the IP header */ 985 986 skb_set_network_header(skb, skb->len); 987 skb_put(skb, ihl); 988 skb_copy_to_linear_data(skb, pkt->data, ihl); 989 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */ 990 msg = (struct igmpmsg *)skb_network_header(skb); 991 msg->im_vif = vifi; 992 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 993 994 /* Add our header */ 995 996 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr)); 997 igmp->type = 998 msg->im_msgtype = assert; 999 igmp->code = 0; 1000 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1001 skb->transport_header = skb->network_header; 1002 } 1003 1004 rcu_read_lock(); 1005 mroute_sk = rcu_dereference(mrt->mroute_sk); 1006 if (mroute_sk == NULL) { 1007 rcu_read_unlock(); 1008 kfree_skb(skb); 1009 return -EINVAL; 1010 } 1011 1012 /* Deliver to mrouted */ 1013 1014 ret = sock_queue_rcv_skb(mroute_sk, skb); 1015 rcu_read_unlock(); 1016 if (ret < 0) { 1017 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1018 kfree_skb(skb); 1019 } 1020 1021 return ret; 1022 } 1023 1024 /* 1025 * Queue a packet for resolution. It gets locked cache entry! 1026 */ 1027 1028 static int 1029 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb) 1030 { 1031 bool found = false; 1032 int err; 1033 struct mfc_cache *c; 1034 const struct iphdr *iph = ip_hdr(skb); 1035 1036 spin_lock_bh(&mfc_unres_lock); 1037 list_for_each_entry(c, &mrt->mfc_unres_queue, list) { 1038 if (c->mfc_mcastgrp == iph->daddr && 1039 c->mfc_origin == iph->saddr) { 1040 found = true; 1041 break; 1042 } 1043 } 1044 1045 if (!found) { 1046 /* Create a new entry if allowable */ 1047 1048 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1049 (c = ipmr_cache_alloc_unres()) == NULL) { 1050 spin_unlock_bh(&mfc_unres_lock); 1051 1052 kfree_skb(skb); 1053 return -ENOBUFS; 1054 } 1055 1056 /* Fill in the new cache entry */ 1057 1058 c->mfc_parent = -1; 1059 c->mfc_origin = iph->saddr; 1060 c->mfc_mcastgrp = iph->daddr; 1061 1062 /* Reflect first query at mrouted. */ 1063 1064 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1065 if (err < 0) { 1066 /* If the report failed throw the cache entry 1067 out - Brad Parker 1068 */ 1069 spin_unlock_bh(&mfc_unres_lock); 1070 1071 ipmr_cache_free(c); 1072 kfree_skb(skb); 1073 return err; 1074 } 1075 1076 atomic_inc(&mrt->cache_resolve_queue_len); 1077 list_add(&c->list, &mrt->mfc_unres_queue); 1078 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1079 1080 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1081 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires); 1082 } 1083 1084 /* See if we can append the packet */ 1085 1086 if (c->mfc_un.unres.unresolved.qlen > 3) { 1087 kfree_skb(skb); 1088 err = -ENOBUFS; 1089 } else { 1090 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1091 err = 0; 1092 } 1093 1094 spin_unlock_bh(&mfc_unres_lock); 1095 return err; 1096 } 1097 1098 /* 1099 * MFC cache manipulation by user space mroute daemon 1100 */ 1101 1102 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1103 { 1104 int line; 1105 struct mfc_cache *c, *next; 1106 1107 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); 1108 1109 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) { 1110 if (c->mfc_origin == mfc->mfcc_origin.s_addr && 1111 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr && 1112 (parent == -1 || parent == c->mfc_parent)) { 1113 list_del_rcu(&c->list); 1114 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1115 ipmr_cache_free(c); 1116 return 0; 1117 } 1118 } 1119 return -ENOENT; 1120 } 1121 1122 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1123 struct mfcctl *mfc, int mrtsock, int parent) 1124 { 1125 bool found = false; 1126 int line; 1127 struct mfc_cache *uc, *c; 1128 1129 if (mfc->mfcc_parent >= MAXVIFS) 1130 return -ENFILE; 1131 1132 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); 1133 1134 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) { 1135 if (c->mfc_origin == mfc->mfcc_origin.s_addr && 1136 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr && 1137 (parent == -1 || parent == c->mfc_parent)) { 1138 found = true; 1139 break; 1140 } 1141 } 1142 1143 if (found) { 1144 write_lock_bh(&mrt_lock); 1145 c->mfc_parent = mfc->mfcc_parent; 1146 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1147 if (!mrtsock) 1148 c->mfc_flags |= MFC_STATIC; 1149 write_unlock_bh(&mrt_lock); 1150 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1151 return 0; 1152 } 1153 1154 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1155 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1156 return -EINVAL; 1157 1158 c = ipmr_cache_alloc(); 1159 if (c == NULL) 1160 return -ENOMEM; 1161 1162 c->mfc_origin = mfc->mfcc_origin.s_addr; 1163 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1164 c->mfc_parent = mfc->mfcc_parent; 1165 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1166 if (!mrtsock) 1167 c->mfc_flags |= MFC_STATIC; 1168 1169 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]); 1170 1171 /* 1172 * Check to see if we resolved a queued list. If so we 1173 * need to send on the frames and tidy up. 1174 */ 1175 found = false; 1176 spin_lock_bh(&mfc_unres_lock); 1177 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) { 1178 if (uc->mfc_origin == c->mfc_origin && 1179 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1180 list_del(&uc->list); 1181 atomic_dec(&mrt->cache_resolve_queue_len); 1182 found = true; 1183 break; 1184 } 1185 } 1186 if (list_empty(&mrt->mfc_unres_queue)) 1187 del_timer(&mrt->ipmr_expire_timer); 1188 spin_unlock_bh(&mfc_unres_lock); 1189 1190 if (found) { 1191 ipmr_cache_resolve(net, mrt, uc, c); 1192 ipmr_cache_free(uc); 1193 } 1194 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1195 return 0; 1196 } 1197 1198 /* 1199 * Close the multicast socket, and clear the vif tables etc 1200 */ 1201 1202 static void mroute_clean_tables(struct mr_table *mrt) 1203 { 1204 int i; 1205 LIST_HEAD(list); 1206 struct mfc_cache *c, *next; 1207 1208 /* Shut down all active vif entries */ 1209 1210 for (i = 0; i < mrt->maxvif; i++) { 1211 if (!(mrt->vif_table[i].flags & VIFF_STATIC)) 1212 vif_delete(mrt, i, 0, &list); 1213 } 1214 unregister_netdevice_many(&list); 1215 1216 /* Wipe the cache */ 1217 1218 for (i = 0; i < MFC_LINES; i++) { 1219 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) { 1220 if (c->mfc_flags & MFC_STATIC) 1221 continue; 1222 list_del_rcu(&c->list); 1223 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1224 ipmr_cache_free(c); 1225 } 1226 } 1227 1228 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1229 spin_lock_bh(&mfc_unres_lock); 1230 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 1231 list_del(&c->list); 1232 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1233 ipmr_destroy_unres(mrt, c); 1234 } 1235 spin_unlock_bh(&mfc_unres_lock); 1236 } 1237 } 1238 1239 /* called from ip_ra_control(), before an RCU grace period, 1240 * we dont need to call synchronize_rcu() here 1241 */ 1242 static void mrtsock_destruct(struct sock *sk) 1243 { 1244 struct net *net = sock_net(sk); 1245 struct mr_table *mrt; 1246 1247 rtnl_lock(); 1248 ipmr_for_each_table(mrt, net) { 1249 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1250 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1251 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1252 NETCONFA_IFINDEX_ALL, 1253 net->ipv4.devconf_all); 1254 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1255 mroute_clean_tables(mrt); 1256 } 1257 } 1258 rtnl_unlock(); 1259 } 1260 1261 /* 1262 * Socket options and virtual interface manipulation. The whole 1263 * virtual interface system is a complete heap, but unfortunately 1264 * that's how BSD mrouted happens to think. Maybe one day with a proper 1265 * MOSPF/PIM router set up we can clean this up. 1266 */ 1267 1268 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1269 { 1270 int ret, parent = 0; 1271 struct vifctl vif; 1272 struct mfcctl mfc; 1273 struct net *net = sock_net(sk); 1274 struct mr_table *mrt; 1275 1276 if (sk->sk_type != SOCK_RAW || 1277 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1278 return -EOPNOTSUPP; 1279 1280 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1281 if (mrt == NULL) 1282 return -ENOENT; 1283 1284 if (optname != MRT_INIT) { 1285 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1286 !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1287 return -EACCES; 1288 } 1289 1290 switch (optname) { 1291 case MRT_INIT: 1292 if (optlen != sizeof(int)) 1293 return -EINVAL; 1294 1295 rtnl_lock(); 1296 if (rtnl_dereference(mrt->mroute_sk)) { 1297 rtnl_unlock(); 1298 return -EADDRINUSE; 1299 } 1300 1301 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1302 if (ret == 0) { 1303 rcu_assign_pointer(mrt->mroute_sk, sk); 1304 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1305 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1306 NETCONFA_IFINDEX_ALL, 1307 net->ipv4.devconf_all); 1308 } 1309 rtnl_unlock(); 1310 return ret; 1311 case MRT_DONE: 1312 if (sk != rcu_access_pointer(mrt->mroute_sk)) 1313 return -EACCES; 1314 return ip_ra_control(sk, 0, NULL); 1315 case MRT_ADD_VIF: 1316 case MRT_DEL_VIF: 1317 if (optlen != sizeof(vif)) 1318 return -EINVAL; 1319 if (copy_from_user(&vif, optval, sizeof(vif))) 1320 return -EFAULT; 1321 if (vif.vifc_vifi >= MAXVIFS) 1322 return -ENFILE; 1323 rtnl_lock(); 1324 if (optname == MRT_ADD_VIF) { 1325 ret = vif_add(net, mrt, &vif, 1326 sk == rtnl_dereference(mrt->mroute_sk)); 1327 } else { 1328 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1329 } 1330 rtnl_unlock(); 1331 return ret; 1332 1333 /* 1334 * Manipulate the forwarding caches. These live 1335 * in a sort of kernel/user symbiosis. 1336 */ 1337 case MRT_ADD_MFC: 1338 case MRT_DEL_MFC: 1339 parent = -1; 1340 case MRT_ADD_MFC_PROXY: 1341 case MRT_DEL_MFC_PROXY: 1342 if (optlen != sizeof(mfc)) 1343 return -EINVAL; 1344 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1345 return -EFAULT; 1346 if (parent == 0) 1347 parent = mfc.mfcc_parent; 1348 rtnl_lock(); 1349 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1350 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1351 else 1352 ret = ipmr_mfc_add(net, mrt, &mfc, 1353 sk == rtnl_dereference(mrt->mroute_sk), 1354 parent); 1355 rtnl_unlock(); 1356 return ret; 1357 /* 1358 * Control PIM assert. 1359 */ 1360 case MRT_ASSERT: 1361 { 1362 int v; 1363 if (optlen != sizeof(v)) 1364 return -EINVAL; 1365 if (get_user(v, (int __user *)optval)) 1366 return -EFAULT; 1367 mrt->mroute_do_assert = v; 1368 return 0; 1369 } 1370 #ifdef CONFIG_IP_PIMSM 1371 case MRT_PIM: 1372 { 1373 int v; 1374 1375 if (optlen != sizeof(v)) 1376 return -EINVAL; 1377 if (get_user(v, (int __user *)optval)) 1378 return -EFAULT; 1379 v = !!v; 1380 1381 rtnl_lock(); 1382 ret = 0; 1383 if (v != mrt->mroute_do_pim) { 1384 mrt->mroute_do_pim = v; 1385 mrt->mroute_do_assert = v; 1386 } 1387 rtnl_unlock(); 1388 return ret; 1389 } 1390 #endif 1391 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 1392 case MRT_TABLE: 1393 { 1394 u32 v; 1395 1396 if (optlen != sizeof(u32)) 1397 return -EINVAL; 1398 if (get_user(v, (u32 __user *)optval)) 1399 return -EFAULT; 1400 1401 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1402 if (v != RT_TABLE_DEFAULT && v >= 1000000000) 1403 return -EINVAL; 1404 1405 rtnl_lock(); 1406 ret = 0; 1407 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1408 ret = -EBUSY; 1409 } else { 1410 if (!ipmr_new_table(net, v)) 1411 ret = -ENOMEM; 1412 else 1413 raw_sk(sk)->ipmr_table = v; 1414 } 1415 rtnl_unlock(); 1416 return ret; 1417 } 1418 #endif 1419 /* 1420 * Spurious command, or MRT_VERSION which you cannot 1421 * set. 1422 */ 1423 default: 1424 return -ENOPROTOOPT; 1425 } 1426 } 1427 1428 /* 1429 * Getsock opt support for the multicast routing system. 1430 */ 1431 1432 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen) 1433 { 1434 int olr; 1435 int val; 1436 struct net *net = sock_net(sk); 1437 struct mr_table *mrt; 1438 1439 if (sk->sk_type != SOCK_RAW || 1440 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1441 return -EOPNOTSUPP; 1442 1443 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1444 if (mrt == NULL) 1445 return -ENOENT; 1446 1447 if (optname != MRT_VERSION && 1448 #ifdef CONFIG_IP_PIMSM 1449 optname != MRT_PIM && 1450 #endif 1451 optname != MRT_ASSERT) 1452 return -ENOPROTOOPT; 1453 1454 if (get_user(olr, optlen)) 1455 return -EFAULT; 1456 1457 olr = min_t(unsigned int, olr, sizeof(int)); 1458 if (olr < 0) 1459 return -EINVAL; 1460 1461 if (put_user(olr, optlen)) 1462 return -EFAULT; 1463 if (optname == MRT_VERSION) 1464 val = 0x0305; 1465 #ifdef CONFIG_IP_PIMSM 1466 else if (optname == MRT_PIM) 1467 val = mrt->mroute_do_pim; 1468 #endif 1469 else 1470 val = mrt->mroute_do_assert; 1471 if (copy_to_user(optval, &val, olr)) 1472 return -EFAULT; 1473 return 0; 1474 } 1475 1476 /* 1477 * The IP multicast ioctl support routines. 1478 */ 1479 1480 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) 1481 { 1482 struct sioc_sg_req sr; 1483 struct sioc_vif_req vr; 1484 struct vif_device *vif; 1485 struct mfc_cache *c; 1486 struct net *net = sock_net(sk); 1487 struct mr_table *mrt; 1488 1489 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1490 if (mrt == NULL) 1491 return -ENOENT; 1492 1493 switch (cmd) { 1494 case SIOCGETVIFCNT: 1495 if (copy_from_user(&vr, arg, sizeof(vr))) 1496 return -EFAULT; 1497 if (vr.vifi >= mrt->maxvif) 1498 return -EINVAL; 1499 read_lock(&mrt_lock); 1500 vif = &mrt->vif_table[vr.vifi]; 1501 if (VIF_EXISTS(mrt, vr.vifi)) { 1502 vr.icount = vif->pkt_in; 1503 vr.ocount = vif->pkt_out; 1504 vr.ibytes = vif->bytes_in; 1505 vr.obytes = vif->bytes_out; 1506 read_unlock(&mrt_lock); 1507 1508 if (copy_to_user(arg, &vr, sizeof(vr))) 1509 return -EFAULT; 1510 return 0; 1511 } 1512 read_unlock(&mrt_lock); 1513 return -EADDRNOTAVAIL; 1514 case SIOCGETSGCNT: 1515 if (copy_from_user(&sr, arg, sizeof(sr))) 1516 return -EFAULT; 1517 1518 rcu_read_lock(); 1519 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1520 if (c) { 1521 sr.pktcnt = c->mfc_un.res.pkt; 1522 sr.bytecnt = c->mfc_un.res.bytes; 1523 sr.wrong_if = c->mfc_un.res.wrong_if; 1524 rcu_read_unlock(); 1525 1526 if (copy_to_user(arg, &sr, sizeof(sr))) 1527 return -EFAULT; 1528 return 0; 1529 } 1530 rcu_read_unlock(); 1531 return -EADDRNOTAVAIL; 1532 default: 1533 return -ENOIOCTLCMD; 1534 } 1535 } 1536 1537 #ifdef CONFIG_COMPAT 1538 struct compat_sioc_sg_req { 1539 struct in_addr src; 1540 struct in_addr grp; 1541 compat_ulong_t pktcnt; 1542 compat_ulong_t bytecnt; 1543 compat_ulong_t wrong_if; 1544 }; 1545 1546 struct compat_sioc_vif_req { 1547 vifi_t vifi; /* Which iface */ 1548 compat_ulong_t icount; 1549 compat_ulong_t ocount; 1550 compat_ulong_t ibytes; 1551 compat_ulong_t obytes; 1552 }; 1553 1554 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1555 { 1556 struct compat_sioc_sg_req sr; 1557 struct compat_sioc_vif_req vr; 1558 struct vif_device *vif; 1559 struct mfc_cache *c; 1560 struct net *net = sock_net(sk); 1561 struct mr_table *mrt; 1562 1563 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1564 if (mrt == NULL) 1565 return -ENOENT; 1566 1567 switch (cmd) { 1568 case SIOCGETVIFCNT: 1569 if (copy_from_user(&vr, arg, sizeof(vr))) 1570 return -EFAULT; 1571 if (vr.vifi >= mrt->maxvif) 1572 return -EINVAL; 1573 read_lock(&mrt_lock); 1574 vif = &mrt->vif_table[vr.vifi]; 1575 if (VIF_EXISTS(mrt, vr.vifi)) { 1576 vr.icount = vif->pkt_in; 1577 vr.ocount = vif->pkt_out; 1578 vr.ibytes = vif->bytes_in; 1579 vr.obytes = vif->bytes_out; 1580 read_unlock(&mrt_lock); 1581 1582 if (copy_to_user(arg, &vr, sizeof(vr))) 1583 return -EFAULT; 1584 return 0; 1585 } 1586 read_unlock(&mrt_lock); 1587 return -EADDRNOTAVAIL; 1588 case SIOCGETSGCNT: 1589 if (copy_from_user(&sr, arg, sizeof(sr))) 1590 return -EFAULT; 1591 1592 rcu_read_lock(); 1593 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1594 if (c) { 1595 sr.pktcnt = c->mfc_un.res.pkt; 1596 sr.bytecnt = c->mfc_un.res.bytes; 1597 sr.wrong_if = c->mfc_un.res.wrong_if; 1598 rcu_read_unlock(); 1599 1600 if (copy_to_user(arg, &sr, sizeof(sr))) 1601 return -EFAULT; 1602 return 0; 1603 } 1604 rcu_read_unlock(); 1605 return -EADDRNOTAVAIL; 1606 default: 1607 return -ENOIOCTLCMD; 1608 } 1609 } 1610 #endif 1611 1612 1613 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1614 { 1615 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1616 struct net *net = dev_net(dev); 1617 struct mr_table *mrt; 1618 struct vif_device *v; 1619 int ct; 1620 1621 if (event != NETDEV_UNREGISTER) 1622 return NOTIFY_DONE; 1623 1624 ipmr_for_each_table(mrt, net) { 1625 v = &mrt->vif_table[0]; 1626 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1627 if (v->dev == dev) 1628 vif_delete(mrt, ct, 1, NULL); 1629 } 1630 } 1631 return NOTIFY_DONE; 1632 } 1633 1634 1635 static struct notifier_block ip_mr_notifier = { 1636 .notifier_call = ipmr_device_event, 1637 }; 1638 1639 /* 1640 * Encapsulate a packet by attaching a valid IPIP header to it. 1641 * This avoids tunnel drivers and other mess and gives us the speed so 1642 * important for multicast video. 1643 */ 1644 1645 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr) 1646 { 1647 struct iphdr *iph; 1648 const struct iphdr *old_iph = ip_hdr(skb); 1649 1650 skb_push(skb, sizeof(struct iphdr)); 1651 skb->transport_header = skb->network_header; 1652 skb_reset_network_header(skb); 1653 iph = ip_hdr(skb); 1654 1655 iph->version = 4; 1656 iph->tos = old_iph->tos; 1657 iph->ttl = old_iph->ttl; 1658 iph->frag_off = 0; 1659 iph->daddr = daddr; 1660 iph->saddr = saddr; 1661 iph->protocol = IPPROTO_IPIP; 1662 iph->ihl = 5; 1663 iph->tot_len = htons(skb->len); 1664 ip_select_ident(skb, NULL); 1665 ip_send_check(iph); 1666 1667 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1668 nf_reset(skb); 1669 } 1670 1671 static inline int ipmr_forward_finish(struct sk_buff *skb) 1672 { 1673 struct ip_options *opt = &(IPCB(skb)->opt); 1674 1675 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS); 1676 IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len); 1677 1678 if (unlikely(opt->optlen)) 1679 ip_forward_options(skb); 1680 1681 return dst_output(skb); 1682 } 1683 1684 /* 1685 * Processing handlers for ipmr_forward 1686 */ 1687 1688 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt, 1689 struct sk_buff *skb, struct mfc_cache *c, int vifi) 1690 { 1691 const struct iphdr *iph = ip_hdr(skb); 1692 struct vif_device *vif = &mrt->vif_table[vifi]; 1693 struct net_device *dev; 1694 struct rtable *rt; 1695 struct flowi4 fl4; 1696 int encap = 0; 1697 1698 if (vif->dev == NULL) 1699 goto out_free; 1700 1701 #ifdef CONFIG_IP_PIMSM 1702 if (vif->flags & VIFF_REGISTER) { 1703 vif->pkt_out++; 1704 vif->bytes_out += skb->len; 1705 vif->dev->stats.tx_bytes += skb->len; 1706 vif->dev->stats.tx_packets++; 1707 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1708 goto out_free; 1709 } 1710 #endif 1711 1712 if (vif->flags & VIFF_TUNNEL) { 1713 rt = ip_route_output_ports(net, &fl4, NULL, 1714 vif->remote, vif->local, 1715 0, 0, 1716 IPPROTO_IPIP, 1717 RT_TOS(iph->tos), vif->link); 1718 if (IS_ERR(rt)) 1719 goto out_free; 1720 encap = sizeof(struct iphdr); 1721 } else { 1722 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1723 0, 0, 1724 IPPROTO_IPIP, 1725 RT_TOS(iph->tos), vif->link); 1726 if (IS_ERR(rt)) 1727 goto out_free; 1728 } 1729 1730 dev = rt->dst.dev; 1731 1732 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1733 /* Do not fragment multicasts. Alas, IPv4 does not 1734 * allow to send ICMP, so that packets will disappear 1735 * to blackhole. 1736 */ 1737 1738 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS); 1739 ip_rt_put(rt); 1740 goto out_free; 1741 } 1742 1743 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len; 1744 1745 if (skb_cow(skb, encap)) { 1746 ip_rt_put(rt); 1747 goto out_free; 1748 } 1749 1750 vif->pkt_out++; 1751 vif->bytes_out += skb->len; 1752 1753 skb_dst_drop(skb); 1754 skb_dst_set(skb, &rt->dst); 1755 ip_decrease_ttl(ip_hdr(skb)); 1756 1757 /* FIXME: forward and output firewalls used to be called here. 1758 * What do we do with netfilter? -- RR 1759 */ 1760 if (vif->flags & VIFF_TUNNEL) { 1761 ip_encap(skb, vif->local, vif->remote); 1762 /* FIXME: extra output firewall step used to be here. --RR */ 1763 vif->dev->stats.tx_packets++; 1764 vif->dev->stats.tx_bytes += skb->len; 1765 } 1766 1767 IPCB(skb)->flags |= IPSKB_FORWARDED; 1768 1769 /* 1770 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1771 * not only before forwarding, but after forwarding on all output 1772 * interfaces. It is clear, if mrouter runs a multicasting 1773 * program, it should receive packets not depending to what interface 1774 * program is joined. 1775 * If we will not make it, the program will have to join on all 1776 * interfaces. On the other hand, multihoming host (or router, but 1777 * not mrouter) cannot join to more than one interface - it will 1778 * result in receiving multiple packets. 1779 */ 1780 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev, 1781 ipmr_forward_finish); 1782 return; 1783 1784 out_free: 1785 kfree_skb(skb); 1786 } 1787 1788 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev) 1789 { 1790 int ct; 1791 1792 for (ct = mrt->maxvif-1; ct >= 0; ct--) { 1793 if (mrt->vif_table[ct].dev == dev) 1794 break; 1795 } 1796 return ct; 1797 } 1798 1799 /* "local" means that we should preserve one skb (for local delivery) */ 1800 1801 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 1802 struct sk_buff *skb, struct mfc_cache *cache, 1803 int local) 1804 { 1805 int psend = -1; 1806 int vif, ct; 1807 int true_vifi = ipmr_find_vif(mrt, skb->dev); 1808 1809 vif = cache->mfc_parent; 1810 cache->mfc_un.res.pkt++; 1811 cache->mfc_un.res.bytes += skb->len; 1812 1813 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 1814 struct mfc_cache *cache_proxy; 1815 1816 /* For an (*,G) entry, we only check that the incomming 1817 * interface is part of the static tree. 1818 */ 1819 cache_proxy = ipmr_cache_find_any_parent(mrt, vif); 1820 if (cache_proxy && 1821 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 1822 goto forward; 1823 } 1824 1825 /* 1826 * Wrong interface: drop packet and (maybe) send PIM assert. 1827 */ 1828 if (mrt->vif_table[vif].dev != skb->dev) { 1829 if (rt_is_output_route(skb_rtable(skb))) { 1830 /* It is our own packet, looped back. 1831 * Very complicated situation... 1832 * 1833 * The best workaround until routing daemons will be 1834 * fixed is not to redistribute packet, if it was 1835 * send through wrong interface. It means, that 1836 * multicast applications WILL NOT work for 1837 * (S,G), which have default multicast route pointing 1838 * to wrong oif. In any case, it is not a good 1839 * idea to use multicasting applications on router. 1840 */ 1841 goto dont_forward; 1842 } 1843 1844 cache->mfc_un.res.wrong_if++; 1845 1846 if (true_vifi >= 0 && mrt->mroute_do_assert && 1847 /* pimsm uses asserts, when switching from RPT to SPT, 1848 * so that we cannot check that packet arrived on an oif. 1849 * It is bad, but otherwise we would need to move pretty 1850 * large chunk of pimd to kernel. Ough... --ANK 1851 */ 1852 (mrt->mroute_do_pim || 1853 cache->mfc_un.res.ttls[true_vifi] < 255) && 1854 time_after(jiffies, 1855 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 1856 cache->mfc_un.res.last_assert = jiffies; 1857 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 1858 } 1859 goto dont_forward; 1860 } 1861 1862 forward: 1863 mrt->vif_table[vif].pkt_in++; 1864 mrt->vif_table[vif].bytes_in += skb->len; 1865 1866 /* 1867 * Forward the frame 1868 */ 1869 if (cache->mfc_origin == htonl(INADDR_ANY) && 1870 cache->mfc_mcastgrp == htonl(INADDR_ANY)) { 1871 if (true_vifi >= 0 && 1872 true_vifi != cache->mfc_parent && 1873 ip_hdr(skb)->ttl > 1874 cache->mfc_un.res.ttls[cache->mfc_parent]) { 1875 /* It's an (*,*) entry and the packet is not coming from 1876 * the upstream: forward the packet to the upstream 1877 * only. 1878 */ 1879 psend = cache->mfc_parent; 1880 goto last_forward; 1881 } 1882 goto dont_forward; 1883 } 1884 for (ct = cache->mfc_un.res.maxvif - 1; 1885 ct >= cache->mfc_un.res.minvif; ct--) { 1886 /* For (*,G) entry, don't forward to the incoming interface */ 1887 if ((cache->mfc_origin != htonl(INADDR_ANY) || 1888 ct != true_vifi) && 1889 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) { 1890 if (psend != -1) { 1891 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1892 1893 if (skb2) 1894 ipmr_queue_xmit(net, mrt, skb2, cache, 1895 psend); 1896 } 1897 psend = ct; 1898 } 1899 } 1900 last_forward: 1901 if (psend != -1) { 1902 if (local) { 1903 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1904 1905 if (skb2) 1906 ipmr_queue_xmit(net, mrt, skb2, cache, psend); 1907 } else { 1908 ipmr_queue_xmit(net, mrt, skb, cache, psend); 1909 return; 1910 } 1911 } 1912 1913 dont_forward: 1914 if (!local) 1915 kfree_skb(skb); 1916 } 1917 1918 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 1919 { 1920 struct rtable *rt = skb_rtable(skb); 1921 struct iphdr *iph = ip_hdr(skb); 1922 struct flowi4 fl4 = { 1923 .daddr = iph->daddr, 1924 .saddr = iph->saddr, 1925 .flowi4_tos = RT_TOS(iph->tos), 1926 .flowi4_oif = (rt_is_output_route(rt) ? 1927 skb->dev->ifindex : 0), 1928 .flowi4_iif = (rt_is_output_route(rt) ? 1929 LOOPBACK_IFINDEX : 1930 skb->dev->ifindex), 1931 .flowi4_mark = skb->mark, 1932 }; 1933 struct mr_table *mrt; 1934 int err; 1935 1936 err = ipmr_fib_lookup(net, &fl4, &mrt); 1937 if (err) 1938 return ERR_PTR(err); 1939 return mrt; 1940 } 1941 1942 /* 1943 * Multicast packets for forwarding arrive here 1944 * Called with rcu_read_lock(); 1945 */ 1946 1947 int ip_mr_input(struct sk_buff *skb) 1948 { 1949 struct mfc_cache *cache; 1950 struct net *net = dev_net(skb->dev); 1951 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 1952 struct mr_table *mrt; 1953 1954 /* Packet is looped back after forward, it should not be 1955 * forwarded second time, but still can be delivered locally. 1956 */ 1957 if (IPCB(skb)->flags & IPSKB_FORWARDED) 1958 goto dont_forward; 1959 1960 mrt = ipmr_rt_fib_lookup(net, skb); 1961 if (IS_ERR(mrt)) { 1962 kfree_skb(skb); 1963 return PTR_ERR(mrt); 1964 } 1965 if (!local) { 1966 if (IPCB(skb)->opt.router_alert) { 1967 if (ip_call_ra_chain(skb)) 1968 return 0; 1969 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 1970 /* IGMPv1 (and broken IGMPv2 implementations sort of 1971 * Cisco IOS <= 11.2(8)) do not put router alert 1972 * option to IGMP packets destined to routable 1973 * groups. It is very bad, because it means 1974 * that we can forward NO IGMP messages. 1975 */ 1976 struct sock *mroute_sk; 1977 1978 mroute_sk = rcu_dereference(mrt->mroute_sk); 1979 if (mroute_sk) { 1980 nf_reset(skb); 1981 raw_rcv(mroute_sk, skb); 1982 return 0; 1983 } 1984 } 1985 } 1986 1987 /* already under rcu_read_lock() */ 1988 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 1989 if (cache == NULL) { 1990 int vif = ipmr_find_vif(mrt, skb->dev); 1991 1992 if (vif >= 0) 1993 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 1994 vif); 1995 } 1996 1997 /* 1998 * No usable cache entry 1999 */ 2000 if (cache == NULL) { 2001 int vif; 2002 2003 if (local) { 2004 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2005 ip_local_deliver(skb); 2006 if (skb2 == NULL) 2007 return -ENOBUFS; 2008 skb = skb2; 2009 } 2010 2011 read_lock(&mrt_lock); 2012 vif = ipmr_find_vif(mrt, skb->dev); 2013 if (vif >= 0) { 2014 int err2 = ipmr_cache_unresolved(mrt, vif, skb); 2015 read_unlock(&mrt_lock); 2016 2017 return err2; 2018 } 2019 read_unlock(&mrt_lock); 2020 kfree_skb(skb); 2021 return -ENODEV; 2022 } 2023 2024 read_lock(&mrt_lock); 2025 ip_mr_forward(net, mrt, skb, cache, local); 2026 read_unlock(&mrt_lock); 2027 2028 if (local) 2029 return ip_local_deliver(skb); 2030 2031 return 0; 2032 2033 dont_forward: 2034 if (local) 2035 return ip_local_deliver(skb); 2036 kfree_skb(skb); 2037 return 0; 2038 } 2039 2040 #ifdef CONFIG_IP_PIMSM 2041 /* called with rcu_read_lock() */ 2042 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb, 2043 unsigned int pimlen) 2044 { 2045 struct net_device *reg_dev = NULL; 2046 struct iphdr *encap; 2047 2048 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); 2049 /* 2050 * Check that: 2051 * a. packet is really sent to a multicast group 2052 * b. packet is not a NULL-REGISTER 2053 * c. packet is not truncated 2054 */ 2055 if (!ipv4_is_multicast(encap->daddr) || 2056 encap->tot_len == 0 || 2057 ntohs(encap->tot_len) + pimlen > skb->len) 2058 return 1; 2059 2060 read_lock(&mrt_lock); 2061 if (mrt->mroute_reg_vif_num >= 0) 2062 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev; 2063 read_unlock(&mrt_lock); 2064 2065 if (reg_dev == NULL) 2066 return 1; 2067 2068 skb->mac_header = skb->network_header; 2069 skb_pull(skb, (u8 *)encap - skb->data); 2070 skb_reset_network_header(skb); 2071 skb->protocol = htons(ETH_P_IP); 2072 skb->ip_summed = CHECKSUM_NONE; 2073 2074 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 2075 2076 netif_rx(skb); 2077 2078 return NET_RX_SUCCESS; 2079 } 2080 #endif 2081 2082 #ifdef CONFIG_IP_PIMSM_V1 2083 /* 2084 * Handle IGMP messages of PIMv1 2085 */ 2086 2087 int pim_rcv_v1(struct sk_buff *skb) 2088 { 2089 struct igmphdr *pim; 2090 struct net *net = dev_net(skb->dev); 2091 struct mr_table *mrt; 2092 2093 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2094 goto drop; 2095 2096 pim = igmp_hdr(skb); 2097 2098 mrt = ipmr_rt_fib_lookup(net, skb); 2099 if (IS_ERR(mrt)) 2100 goto drop; 2101 if (!mrt->mroute_do_pim || 2102 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2103 goto drop; 2104 2105 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2106 drop: 2107 kfree_skb(skb); 2108 } 2109 return 0; 2110 } 2111 #endif 2112 2113 #ifdef CONFIG_IP_PIMSM_V2 2114 static int pim_rcv(struct sk_buff *skb) 2115 { 2116 struct pimreghdr *pim; 2117 struct net *net = dev_net(skb->dev); 2118 struct mr_table *mrt; 2119 2120 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2121 goto drop; 2122 2123 pim = (struct pimreghdr *)skb_transport_header(skb); 2124 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) || 2125 (pim->flags & PIM_NULL_REGISTER) || 2126 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2127 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2128 goto drop; 2129 2130 mrt = ipmr_rt_fib_lookup(net, skb); 2131 if (IS_ERR(mrt)) 2132 goto drop; 2133 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2134 drop: 2135 kfree_skb(skb); 2136 } 2137 return 0; 2138 } 2139 #endif 2140 2141 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2142 struct mfc_cache *c, struct rtmsg *rtm) 2143 { 2144 int ct; 2145 struct rtnexthop *nhp; 2146 struct nlattr *mp_attr; 2147 struct rta_mfc_stats mfcs; 2148 2149 /* If cache is unresolved, don't try to parse IIF and OIF */ 2150 if (c->mfc_parent >= MAXVIFS) 2151 return -ENOENT; 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 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2177 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2178 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2179 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0) 2180 return -EMSGSIZE; 2181 2182 rtm->rtm_type = RTN_MULTICAST; 2183 return 1; 2184 } 2185 2186 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2187 __be32 saddr, __be32 daddr, 2188 struct rtmsg *rtm, int nowait) 2189 { 2190 struct mfc_cache *cache; 2191 struct mr_table *mrt; 2192 int err; 2193 2194 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2195 if (mrt == NULL) 2196 return -ENOENT; 2197 2198 rcu_read_lock(); 2199 cache = ipmr_cache_find(mrt, saddr, daddr); 2200 if (cache == NULL && skb->dev) { 2201 int vif = ipmr_find_vif(mrt, skb->dev); 2202 2203 if (vif >= 0) 2204 cache = ipmr_cache_find_any(mrt, daddr, vif); 2205 } 2206 if (cache == NULL) { 2207 struct sk_buff *skb2; 2208 struct iphdr *iph; 2209 struct net_device *dev; 2210 int vif = -1; 2211 2212 if (nowait) { 2213 rcu_read_unlock(); 2214 return -EAGAIN; 2215 } 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 skb_push(skb2, sizeof(struct iphdr)); 2234 skb_reset_network_header(skb2); 2235 iph = ip_hdr(skb2); 2236 iph->ihl = sizeof(struct iphdr) >> 2; 2237 iph->saddr = saddr; 2238 iph->daddr = daddr; 2239 iph->version = 0; 2240 err = ipmr_cache_unresolved(mrt, vif, skb2); 2241 read_unlock(&mrt_lock); 2242 rcu_read_unlock(); 2243 return err; 2244 } 2245 2246 read_lock(&mrt_lock); 2247 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY)) 2248 cache->mfc_flags |= MFC_NOTIFY; 2249 err = __ipmr_fill_mroute(mrt, skb, cache, rtm); 2250 read_unlock(&mrt_lock); 2251 rcu_read_unlock(); 2252 return err; 2253 } 2254 2255 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2256 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2257 int flags) 2258 { 2259 struct nlmsghdr *nlh; 2260 struct rtmsg *rtm; 2261 int err; 2262 2263 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2264 if (nlh == NULL) 2265 return -EMSGSIZE; 2266 2267 rtm = nlmsg_data(nlh); 2268 rtm->rtm_family = RTNL_FAMILY_IPMR; 2269 rtm->rtm_dst_len = 32; 2270 rtm->rtm_src_len = 32; 2271 rtm->rtm_tos = 0; 2272 rtm->rtm_table = mrt->id; 2273 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2274 goto nla_put_failure; 2275 rtm->rtm_type = RTN_MULTICAST; 2276 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2277 if (c->mfc_flags & MFC_STATIC) 2278 rtm->rtm_protocol = RTPROT_STATIC; 2279 else 2280 rtm->rtm_protocol = RTPROT_MROUTED; 2281 rtm->rtm_flags = 0; 2282 2283 if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) || 2284 nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp)) 2285 goto nla_put_failure; 2286 err = __ipmr_fill_mroute(mrt, skb, c, rtm); 2287 /* do not break the dump if cache is unresolved */ 2288 if (err < 0 && err != -ENOENT) 2289 goto nla_put_failure; 2290 2291 nlmsg_end(skb, nlh); 2292 return 0; 2293 2294 nla_put_failure: 2295 nlmsg_cancel(skb, nlh); 2296 return -EMSGSIZE; 2297 } 2298 2299 static size_t mroute_msgsize(bool unresolved, int maxvif) 2300 { 2301 size_t len = 2302 NLMSG_ALIGN(sizeof(struct rtmsg)) 2303 + nla_total_size(4) /* RTA_TABLE */ 2304 + nla_total_size(4) /* RTA_SRC */ 2305 + nla_total_size(4) /* RTA_DST */ 2306 ; 2307 2308 if (!unresolved) 2309 len = len 2310 + nla_total_size(4) /* RTA_IIF */ 2311 + nla_total_size(0) /* RTA_MULTIPATH */ 2312 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2313 /* RTA_MFC_STATS */ 2314 + nla_total_size(sizeof(struct rta_mfc_stats)) 2315 ; 2316 2317 return len; 2318 } 2319 2320 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2321 int cmd) 2322 { 2323 struct net *net = read_pnet(&mrt->net); 2324 struct sk_buff *skb; 2325 int err = -ENOBUFS; 2326 2327 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif), 2328 GFP_ATOMIC); 2329 if (skb == NULL) 2330 goto errout; 2331 2332 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2333 if (err < 0) 2334 goto errout; 2335 2336 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2337 return; 2338 2339 errout: 2340 kfree_skb(skb); 2341 if (err < 0) 2342 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2343 } 2344 2345 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2346 { 2347 struct net *net = sock_net(skb->sk); 2348 struct mr_table *mrt; 2349 struct mfc_cache *mfc; 2350 unsigned int t = 0, s_t; 2351 unsigned int h = 0, s_h; 2352 unsigned int e = 0, s_e; 2353 2354 s_t = cb->args[0]; 2355 s_h = cb->args[1]; 2356 s_e = cb->args[2]; 2357 2358 rcu_read_lock(); 2359 ipmr_for_each_table(mrt, net) { 2360 if (t < s_t) 2361 goto next_table; 2362 if (t > s_t) 2363 s_h = 0; 2364 for (h = s_h; h < MFC_LINES; h++) { 2365 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) { 2366 if (e < s_e) 2367 goto next_entry; 2368 if (ipmr_fill_mroute(mrt, skb, 2369 NETLINK_CB(cb->skb).portid, 2370 cb->nlh->nlmsg_seq, 2371 mfc, RTM_NEWROUTE, 2372 NLM_F_MULTI) < 0) 2373 goto done; 2374 next_entry: 2375 e++; 2376 } 2377 e = s_e = 0; 2378 } 2379 spin_lock_bh(&mfc_unres_lock); 2380 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) { 2381 if (e < s_e) 2382 goto next_entry2; 2383 if (ipmr_fill_mroute(mrt, skb, 2384 NETLINK_CB(cb->skb).portid, 2385 cb->nlh->nlmsg_seq, 2386 mfc, RTM_NEWROUTE, 2387 NLM_F_MULTI) < 0) { 2388 spin_unlock_bh(&mfc_unres_lock); 2389 goto done; 2390 } 2391 next_entry2: 2392 e++; 2393 } 2394 spin_unlock_bh(&mfc_unres_lock); 2395 e = s_e = 0; 2396 s_h = 0; 2397 next_table: 2398 t++; 2399 } 2400 done: 2401 rcu_read_unlock(); 2402 2403 cb->args[2] = e; 2404 cb->args[1] = h; 2405 cb->args[0] = t; 2406 2407 return skb->len; 2408 } 2409 2410 #ifdef CONFIG_PROC_FS 2411 /* 2412 * The /proc interfaces to multicast routing : 2413 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 2414 */ 2415 struct ipmr_vif_iter { 2416 struct seq_net_private p; 2417 struct mr_table *mrt; 2418 int ct; 2419 }; 2420 2421 static struct vif_device *ipmr_vif_seq_idx(struct net *net, 2422 struct ipmr_vif_iter *iter, 2423 loff_t pos) 2424 { 2425 struct mr_table *mrt = iter->mrt; 2426 2427 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 2428 if (!VIF_EXISTS(mrt, iter->ct)) 2429 continue; 2430 if (pos-- == 0) 2431 return &mrt->vif_table[iter->ct]; 2432 } 2433 return NULL; 2434 } 2435 2436 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 2437 __acquires(mrt_lock) 2438 { 2439 struct ipmr_vif_iter *iter = seq->private; 2440 struct net *net = seq_file_net(seq); 2441 struct mr_table *mrt; 2442 2443 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2444 if (mrt == NULL) 2445 return ERR_PTR(-ENOENT); 2446 2447 iter->mrt = mrt; 2448 2449 read_lock(&mrt_lock); 2450 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1) 2451 : SEQ_START_TOKEN; 2452 } 2453 2454 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2455 { 2456 struct ipmr_vif_iter *iter = seq->private; 2457 struct net *net = seq_file_net(seq); 2458 struct mr_table *mrt = iter->mrt; 2459 2460 ++*pos; 2461 if (v == SEQ_START_TOKEN) 2462 return ipmr_vif_seq_idx(net, iter, 0); 2463 2464 while (++iter->ct < mrt->maxvif) { 2465 if (!VIF_EXISTS(mrt, iter->ct)) 2466 continue; 2467 return &mrt->vif_table[iter->ct]; 2468 } 2469 return NULL; 2470 } 2471 2472 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 2473 __releases(mrt_lock) 2474 { 2475 read_unlock(&mrt_lock); 2476 } 2477 2478 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 2479 { 2480 struct ipmr_vif_iter *iter = seq->private; 2481 struct mr_table *mrt = iter->mrt; 2482 2483 if (v == SEQ_START_TOKEN) { 2484 seq_puts(seq, 2485 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 2486 } else { 2487 const struct vif_device *vif = v; 2488 const char *name = vif->dev ? vif->dev->name : "none"; 2489 2490 seq_printf(seq, 2491 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 2492 vif - mrt->vif_table, 2493 name, vif->bytes_in, vif->pkt_in, 2494 vif->bytes_out, vif->pkt_out, 2495 vif->flags, vif->local, vif->remote); 2496 } 2497 return 0; 2498 } 2499 2500 static const struct seq_operations ipmr_vif_seq_ops = { 2501 .start = ipmr_vif_seq_start, 2502 .next = ipmr_vif_seq_next, 2503 .stop = ipmr_vif_seq_stop, 2504 .show = ipmr_vif_seq_show, 2505 }; 2506 2507 static int ipmr_vif_open(struct inode *inode, struct file *file) 2508 { 2509 return seq_open_net(inode, file, &ipmr_vif_seq_ops, 2510 sizeof(struct ipmr_vif_iter)); 2511 } 2512 2513 static const struct file_operations ipmr_vif_fops = { 2514 .owner = THIS_MODULE, 2515 .open = ipmr_vif_open, 2516 .read = seq_read, 2517 .llseek = seq_lseek, 2518 .release = seq_release_net, 2519 }; 2520 2521 struct ipmr_mfc_iter { 2522 struct seq_net_private p; 2523 struct mr_table *mrt; 2524 struct list_head *cache; 2525 int ct; 2526 }; 2527 2528 2529 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net, 2530 struct ipmr_mfc_iter *it, loff_t pos) 2531 { 2532 struct mr_table *mrt = it->mrt; 2533 struct mfc_cache *mfc; 2534 2535 rcu_read_lock(); 2536 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) { 2537 it->cache = &mrt->mfc_cache_array[it->ct]; 2538 list_for_each_entry_rcu(mfc, it->cache, list) 2539 if (pos-- == 0) 2540 return mfc; 2541 } 2542 rcu_read_unlock(); 2543 2544 spin_lock_bh(&mfc_unres_lock); 2545 it->cache = &mrt->mfc_unres_queue; 2546 list_for_each_entry(mfc, it->cache, list) 2547 if (pos-- == 0) 2548 return mfc; 2549 spin_unlock_bh(&mfc_unres_lock); 2550 2551 it->cache = NULL; 2552 return NULL; 2553 } 2554 2555 2556 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 2557 { 2558 struct ipmr_mfc_iter *it = seq->private; 2559 struct net *net = seq_file_net(seq); 2560 struct mr_table *mrt; 2561 2562 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2563 if (mrt == NULL) 2564 return ERR_PTR(-ENOENT); 2565 2566 it->mrt = mrt; 2567 it->cache = NULL; 2568 it->ct = 0; 2569 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 2570 : SEQ_START_TOKEN; 2571 } 2572 2573 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2574 { 2575 struct mfc_cache *mfc = v; 2576 struct ipmr_mfc_iter *it = seq->private; 2577 struct net *net = seq_file_net(seq); 2578 struct mr_table *mrt = it->mrt; 2579 2580 ++*pos; 2581 2582 if (v == SEQ_START_TOKEN) 2583 return ipmr_mfc_seq_idx(net, seq->private, 0); 2584 2585 if (mfc->list.next != it->cache) 2586 return list_entry(mfc->list.next, struct mfc_cache, list); 2587 2588 if (it->cache == &mrt->mfc_unres_queue) 2589 goto end_of_list; 2590 2591 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]); 2592 2593 while (++it->ct < MFC_LINES) { 2594 it->cache = &mrt->mfc_cache_array[it->ct]; 2595 if (list_empty(it->cache)) 2596 continue; 2597 return list_first_entry(it->cache, struct mfc_cache, list); 2598 } 2599 2600 /* exhausted cache_array, show unresolved */ 2601 rcu_read_unlock(); 2602 it->cache = &mrt->mfc_unres_queue; 2603 it->ct = 0; 2604 2605 spin_lock_bh(&mfc_unres_lock); 2606 if (!list_empty(it->cache)) 2607 return list_first_entry(it->cache, struct mfc_cache, list); 2608 2609 end_of_list: 2610 spin_unlock_bh(&mfc_unres_lock); 2611 it->cache = NULL; 2612 2613 return NULL; 2614 } 2615 2616 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 2617 { 2618 struct ipmr_mfc_iter *it = seq->private; 2619 struct mr_table *mrt = it->mrt; 2620 2621 if (it->cache == &mrt->mfc_unres_queue) 2622 spin_unlock_bh(&mfc_unres_lock); 2623 else if (it->cache == &mrt->mfc_cache_array[it->ct]) 2624 rcu_read_unlock(); 2625 } 2626 2627 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 2628 { 2629 int n; 2630 2631 if (v == SEQ_START_TOKEN) { 2632 seq_puts(seq, 2633 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 2634 } else { 2635 const struct mfc_cache *mfc = v; 2636 const struct ipmr_mfc_iter *it = seq->private; 2637 const struct mr_table *mrt = it->mrt; 2638 2639 seq_printf(seq, "%08X %08X %-3hd", 2640 (__force u32) mfc->mfc_mcastgrp, 2641 (__force u32) mfc->mfc_origin, 2642 mfc->mfc_parent); 2643 2644 if (it->cache != &mrt->mfc_unres_queue) { 2645 seq_printf(seq, " %8lu %8lu %8lu", 2646 mfc->mfc_un.res.pkt, 2647 mfc->mfc_un.res.bytes, 2648 mfc->mfc_un.res.wrong_if); 2649 for (n = mfc->mfc_un.res.minvif; 2650 n < mfc->mfc_un.res.maxvif; n++) { 2651 if (VIF_EXISTS(mrt, n) && 2652 mfc->mfc_un.res.ttls[n] < 255) 2653 seq_printf(seq, 2654 " %2d:%-3d", 2655 n, mfc->mfc_un.res.ttls[n]); 2656 } 2657 } else { 2658 /* unresolved mfc_caches don't contain 2659 * pkt, bytes and wrong_if values 2660 */ 2661 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 2662 } 2663 seq_putc(seq, '\n'); 2664 } 2665 return 0; 2666 } 2667 2668 static const struct seq_operations ipmr_mfc_seq_ops = { 2669 .start = ipmr_mfc_seq_start, 2670 .next = ipmr_mfc_seq_next, 2671 .stop = ipmr_mfc_seq_stop, 2672 .show = ipmr_mfc_seq_show, 2673 }; 2674 2675 static int ipmr_mfc_open(struct inode *inode, struct file *file) 2676 { 2677 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 2678 sizeof(struct ipmr_mfc_iter)); 2679 } 2680 2681 static const struct file_operations ipmr_mfc_fops = { 2682 .owner = THIS_MODULE, 2683 .open = ipmr_mfc_open, 2684 .read = seq_read, 2685 .llseek = seq_lseek, 2686 .release = seq_release_net, 2687 }; 2688 #endif 2689 2690 #ifdef CONFIG_IP_PIMSM_V2 2691 static const struct net_protocol pim_protocol = { 2692 .handler = pim_rcv, 2693 .netns_ok = 1, 2694 }; 2695 #endif 2696 2697 2698 /* 2699 * Setup for IP multicast routing 2700 */ 2701 static int __net_init ipmr_net_init(struct net *net) 2702 { 2703 int err; 2704 2705 err = ipmr_rules_init(net); 2706 if (err < 0) 2707 goto fail; 2708 2709 #ifdef CONFIG_PROC_FS 2710 err = -ENOMEM; 2711 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops)) 2712 goto proc_vif_fail; 2713 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops)) 2714 goto proc_cache_fail; 2715 #endif 2716 return 0; 2717 2718 #ifdef CONFIG_PROC_FS 2719 proc_cache_fail: 2720 remove_proc_entry("ip_mr_vif", net->proc_net); 2721 proc_vif_fail: 2722 ipmr_rules_exit(net); 2723 #endif 2724 fail: 2725 return err; 2726 } 2727 2728 static void __net_exit ipmr_net_exit(struct net *net) 2729 { 2730 #ifdef CONFIG_PROC_FS 2731 remove_proc_entry("ip_mr_cache", net->proc_net); 2732 remove_proc_entry("ip_mr_vif", net->proc_net); 2733 #endif 2734 ipmr_rules_exit(net); 2735 } 2736 2737 static struct pernet_operations ipmr_net_ops = { 2738 .init = ipmr_net_init, 2739 .exit = ipmr_net_exit, 2740 }; 2741 2742 int __init ip_mr_init(void) 2743 { 2744 int err; 2745 2746 mrt_cachep = kmem_cache_create("ip_mrt_cache", 2747 sizeof(struct mfc_cache), 2748 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, 2749 NULL); 2750 if (!mrt_cachep) 2751 return -ENOMEM; 2752 2753 err = register_pernet_subsys(&ipmr_net_ops); 2754 if (err) 2755 goto reg_pernet_fail; 2756 2757 err = register_netdevice_notifier(&ip_mr_notifier); 2758 if (err) 2759 goto reg_notif_fail; 2760 #ifdef CONFIG_IP_PIMSM_V2 2761 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 2762 pr_err("%s: can't add PIM protocol\n", __func__); 2763 err = -EAGAIN; 2764 goto add_proto_fail; 2765 } 2766 #endif 2767 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE, 2768 NULL, ipmr_rtm_dumproute, NULL); 2769 return 0; 2770 2771 #ifdef CONFIG_IP_PIMSM_V2 2772 add_proto_fail: 2773 unregister_netdevice_notifier(&ip_mr_notifier); 2774 #endif 2775 reg_notif_fail: 2776 unregister_pernet_subsys(&ipmr_net_ops); 2777 reg_pernet_fail: 2778 kmem_cache_destroy(mrt_cachep); 2779 return err; 2780 } 2781