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 #ifdef CONFIG_NET_NS 77 struct net *net; 78 #endif 79 u32 id; 80 struct sock __rcu *mroute_sk; 81 struct timer_list ipmr_expire_timer; 82 struct list_head mfc_unres_queue; 83 struct list_head mfc_cache_array[MFC_LINES]; 84 struct vif_device vif_table[MAXVIFS]; 85 int maxvif; 86 atomic_t cache_resolve_queue_len; 87 bool mroute_do_assert; 88 bool mroute_do_pim; 89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) 90 int mroute_reg_vif_num; 91 #endif 92 }; 93 94 struct ipmr_rule { 95 struct fib_rule common; 96 }; 97 98 struct ipmr_result { 99 struct mr_table *mrt; 100 }; 101 102 /* Big lock, protecting vif table, mrt cache and mroute socket state. 103 * Note that the changes are semaphored via rtnl_lock. 104 */ 105 106 static DEFINE_RWLOCK(mrt_lock); 107 108 /* 109 * Multicast router control variables 110 */ 111 112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL) 113 114 /* Special spinlock for queue of unresolved entries */ 115 static DEFINE_SPINLOCK(mfc_unres_lock); 116 117 /* We return to original Alan's scheme. Hash table of resolved 118 * entries is changed only in process context and protected 119 * with weak lock mrt_lock. Queue of unresolved entries is protected 120 * with strong spinlock mfc_unres_lock. 121 * 122 * In this case data path is free of exclusive locks at all. 123 */ 124 125 static struct kmem_cache *mrt_cachep __read_mostly; 126 127 static struct mr_table *ipmr_new_table(struct net *net, u32 id); 128 static void ipmr_free_table(struct mr_table *mrt); 129 130 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 131 struct sk_buff *skb, struct mfc_cache *cache, 132 int local); 133 static int ipmr_cache_report(struct mr_table *mrt, 134 struct sk_buff *pkt, vifi_t vifi, int assert); 135 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 136 struct mfc_cache *c, struct rtmsg *rtm); 137 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 138 int cmd); 139 static void mroute_clean_tables(struct mr_table *mrt); 140 static void ipmr_expire_process(unsigned long arg); 141 142 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 143 #define ipmr_for_each_table(mrt, net) \ 144 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list) 145 146 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 147 { 148 struct mr_table *mrt; 149 150 ipmr_for_each_table(mrt, net) { 151 if (mrt->id == id) 152 return mrt; 153 } 154 return NULL; 155 } 156 157 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 158 struct mr_table **mrt) 159 { 160 int err; 161 struct ipmr_result res; 162 struct fib_lookup_arg arg = { 163 .result = &res, 164 .flags = FIB_LOOKUP_NOREF, 165 }; 166 167 err = fib_rules_lookup(net->ipv4.mr_rules_ops, 168 flowi4_to_flowi(flp4), 0, &arg); 169 if (err < 0) 170 return err; 171 *mrt = res.mrt; 172 return 0; 173 } 174 175 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp, 176 int flags, struct fib_lookup_arg *arg) 177 { 178 struct ipmr_result *res = arg->result; 179 struct mr_table *mrt; 180 181 switch (rule->action) { 182 case FR_ACT_TO_TBL: 183 break; 184 case FR_ACT_UNREACHABLE: 185 return -ENETUNREACH; 186 case FR_ACT_PROHIBIT: 187 return -EACCES; 188 case FR_ACT_BLACKHOLE: 189 default: 190 return -EINVAL; 191 } 192 193 mrt = ipmr_get_table(rule->fr_net, rule->table); 194 if (mrt == NULL) 195 return -EAGAIN; 196 res->mrt = mrt; 197 return 0; 198 } 199 200 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 201 { 202 return 1; 203 } 204 205 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = { 206 FRA_GENERIC_POLICY, 207 }; 208 209 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 210 struct fib_rule_hdr *frh, struct nlattr **tb) 211 { 212 return 0; 213 } 214 215 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 216 struct nlattr **tb) 217 { 218 return 1; 219 } 220 221 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 222 struct fib_rule_hdr *frh) 223 { 224 frh->dst_len = 0; 225 frh->src_len = 0; 226 frh->tos = 0; 227 return 0; 228 } 229 230 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = { 231 .family = RTNL_FAMILY_IPMR, 232 .rule_size = sizeof(struct ipmr_rule), 233 .addr_size = sizeof(u32), 234 .action = ipmr_rule_action, 235 .match = ipmr_rule_match, 236 .configure = ipmr_rule_configure, 237 .compare = ipmr_rule_compare, 238 .default_pref = fib_default_rule_pref, 239 .fill = ipmr_rule_fill, 240 .nlgroup = RTNLGRP_IPV4_RULE, 241 .policy = ipmr_rule_policy, 242 .owner = THIS_MODULE, 243 }; 244 245 static int __net_init ipmr_rules_init(struct net *net) 246 { 247 struct fib_rules_ops *ops; 248 struct mr_table *mrt; 249 int err; 250 251 ops = fib_rules_register(&ipmr_rules_ops_template, net); 252 if (IS_ERR(ops)) 253 return PTR_ERR(ops); 254 255 INIT_LIST_HEAD(&net->ipv4.mr_tables); 256 257 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 258 if (mrt == NULL) { 259 err = -ENOMEM; 260 goto err1; 261 } 262 263 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0); 264 if (err < 0) 265 goto err2; 266 267 net->ipv4.mr_rules_ops = ops; 268 return 0; 269 270 err2: 271 ipmr_free_table(mrt); 272 err1: 273 fib_rules_unregister(ops); 274 return err; 275 } 276 277 static void __net_exit ipmr_rules_exit(struct net *net) 278 { 279 struct mr_table *mrt, *next; 280 281 rtnl_lock(); 282 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) { 283 list_del(&mrt->list); 284 ipmr_free_table(mrt); 285 } 286 fib_rules_unregister(net->ipv4.mr_rules_ops); 287 rtnl_unlock(); 288 } 289 #else 290 #define ipmr_for_each_table(mrt, net) \ 291 for (mrt = net->ipv4.mrt; mrt; mrt = NULL) 292 293 static struct mr_table *ipmr_get_table(struct net *net, u32 id) 294 { 295 return net->ipv4.mrt; 296 } 297 298 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4, 299 struct mr_table **mrt) 300 { 301 *mrt = net->ipv4.mrt; 302 return 0; 303 } 304 305 static int __net_init ipmr_rules_init(struct net *net) 306 { 307 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT); 308 return net->ipv4.mrt ? 0 : -ENOMEM; 309 } 310 311 static void __net_exit ipmr_rules_exit(struct net *net) 312 { 313 rtnl_lock(); 314 ipmr_free_table(net->ipv4.mrt); 315 net->ipv4.mrt = NULL; 316 rtnl_unlock(); 317 } 318 #endif 319 320 static struct mr_table *ipmr_new_table(struct net *net, u32 id) 321 { 322 struct mr_table *mrt; 323 unsigned int i; 324 325 mrt = ipmr_get_table(net, id); 326 if (mrt != NULL) 327 return mrt; 328 329 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL); 330 if (mrt == NULL) 331 return NULL; 332 write_pnet(&mrt->net, net); 333 mrt->id = id; 334 335 /* Forwarding cache */ 336 for (i = 0; i < MFC_LINES; i++) 337 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]); 338 339 INIT_LIST_HEAD(&mrt->mfc_unres_queue); 340 341 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process, 342 (unsigned long)mrt); 343 344 #ifdef CONFIG_IP_PIMSM 345 mrt->mroute_reg_vif_num = -1; 346 #endif 347 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 348 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables); 349 #endif 350 return mrt; 351 } 352 353 static void ipmr_free_table(struct mr_table *mrt) 354 { 355 del_timer_sync(&mrt->ipmr_expire_timer); 356 mroute_clean_tables(mrt); 357 kfree(mrt); 358 } 359 360 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ 361 362 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v) 363 { 364 struct net *net = dev_net(dev); 365 366 dev_close(dev); 367 368 dev = __dev_get_by_name(net, "tunl0"); 369 if (dev) { 370 const struct net_device_ops *ops = dev->netdev_ops; 371 struct ifreq ifr; 372 struct ip_tunnel_parm p; 373 374 memset(&p, 0, sizeof(p)); 375 p.iph.daddr = v->vifc_rmt_addr.s_addr; 376 p.iph.saddr = v->vifc_lcl_addr.s_addr; 377 p.iph.version = 4; 378 p.iph.ihl = 5; 379 p.iph.protocol = IPPROTO_IPIP; 380 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 381 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 382 383 if (ops->ndo_do_ioctl) { 384 mm_segment_t oldfs = get_fs(); 385 386 set_fs(KERNEL_DS); 387 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL); 388 set_fs(oldfs); 389 } 390 } 391 } 392 393 static 394 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v) 395 { 396 struct net_device *dev; 397 398 dev = __dev_get_by_name(net, "tunl0"); 399 400 if (dev) { 401 const struct net_device_ops *ops = dev->netdev_ops; 402 int err; 403 struct ifreq ifr; 404 struct ip_tunnel_parm p; 405 struct in_device *in_dev; 406 407 memset(&p, 0, sizeof(p)); 408 p.iph.daddr = v->vifc_rmt_addr.s_addr; 409 p.iph.saddr = v->vifc_lcl_addr.s_addr; 410 p.iph.version = 4; 411 p.iph.ihl = 5; 412 p.iph.protocol = IPPROTO_IPIP; 413 sprintf(p.name, "dvmrp%d", v->vifc_vifi); 414 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 415 416 if (ops->ndo_do_ioctl) { 417 mm_segment_t oldfs = get_fs(); 418 419 set_fs(KERNEL_DS); 420 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL); 421 set_fs(oldfs); 422 } else { 423 err = -EOPNOTSUPP; 424 } 425 dev = NULL; 426 427 if (err == 0 && 428 (dev = __dev_get_by_name(net, p.name)) != NULL) { 429 dev->flags |= IFF_MULTICAST; 430 431 in_dev = __in_dev_get_rtnl(dev); 432 if (in_dev == NULL) 433 goto failure; 434 435 ipv4_devconf_setall(in_dev); 436 neigh_parms_data_state_setall(in_dev->arp_parms); 437 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 438 439 if (dev_open(dev)) 440 goto failure; 441 dev_hold(dev); 442 } 443 } 444 return dev; 445 446 failure: 447 /* allow the register to be completed before unregistering. */ 448 rtnl_unlock(); 449 rtnl_lock(); 450 451 unregister_netdevice(dev); 452 return NULL; 453 } 454 455 #ifdef CONFIG_IP_PIMSM 456 457 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) 458 { 459 struct net *net = dev_net(dev); 460 struct mr_table *mrt; 461 struct flowi4 fl4 = { 462 .flowi4_oif = dev->ifindex, 463 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX, 464 .flowi4_mark = skb->mark, 465 }; 466 int err; 467 468 err = ipmr_fib_lookup(net, &fl4, &mrt); 469 if (err < 0) { 470 kfree_skb(skb); 471 return err; 472 } 473 474 read_lock(&mrt_lock); 475 dev->stats.tx_bytes += skb->len; 476 dev->stats.tx_packets++; 477 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT); 478 read_unlock(&mrt_lock); 479 kfree_skb(skb); 480 return NETDEV_TX_OK; 481 } 482 483 static const struct net_device_ops reg_vif_netdev_ops = { 484 .ndo_start_xmit = reg_vif_xmit, 485 }; 486 487 static void reg_vif_setup(struct net_device *dev) 488 { 489 dev->type = ARPHRD_PIMREG; 490 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; 491 dev->flags = IFF_NOARP; 492 dev->netdev_ops = ®_vif_netdev_ops; 493 dev->destructor = free_netdev; 494 dev->features |= NETIF_F_NETNS_LOCAL; 495 } 496 497 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt) 498 { 499 struct net_device *dev; 500 struct in_device *in_dev; 501 char name[IFNAMSIZ]; 502 503 if (mrt->id == RT_TABLE_DEFAULT) 504 sprintf(name, "pimreg"); 505 else 506 sprintf(name, "pimreg%u", mrt->id); 507 508 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup); 509 510 if (dev == NULL) 511 return NULL; 512 513 dev_net_set(dev, net); 514 515 if (register_netdevice(dev)) { 516 free_netdev(dev); 517 return NULL; 518 } 519 dev->iflink = 0; 520 521 rcu_read_lock(); 522 in_dev = __in_dev_get_rcu(dev); 523 if (!in_dev) { 524 rcu_read_unlock(); 525 goto failure; 526 } 527 528 ipv4_devconf_setall(in_dev); 529 neigh_parms_data_state_setall(in_dev->arp_parms); 530 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; 531 rcu_read_unlock(); 532 533 if (dev_open(dev)) 534 goto failure; 535 536 dev_hold(dev); 537 538 return dev; 539 540 failure: 541 /* allow the register to be completed before unregistering. */ 542 rtnl_unlock(); 543 rtnl_lock(); 544 545 unregister_netdevice(dev); 546 return NULL; 547 } 548 #endif 549 550 /** 551 * vif_delete - Delete a VIF entry 552 * @notify: Set to 1, if the caller is a notifier_call 553 */ 554 555 static int vif_delete(struct mr_table *mrt, int vifi, int notify, 556 struct list_head *head) 557 { 558 struct vif_device *v; 559 struct net_device *dev; 560 struct in_device *in_dev; 561 562 if (vifi < 0 || vifi >= mrt->maxvif) 563 return -EADDRNOTAVAIL; 564 565 v = &mrt->vif_table[vifi]; 566 567 write_lock_bh(&mrt_lock); 568 dev = v->dev; 569 v->dev = NULL; 570 571 if (!dev) { 572 write_unlock_bh(&mrt_lock); 573 return -EADDRNOTAVAIL; 574 } 575 576 #ifdef CONFIG_IP_PIMSM 577 if (vifi == mrt->mroute_reg_vif_num) 578 mrt->mroute_reg_vif_num = -1; 579 #endif 580 581 if (vifi + 1 == mrt->maxvif) { 582 int tmp; 583 584 for (tmp = vifi - 1; tmp >= 0; tmp--) { 585 if (VIF_EXISTS(mrt, tmp)) 586 break; 587 } 588 mrt->maxvif = tmp+1; 589 } 590 591 write_unlock_bh(&mrt_lock); 592 593 dev_set_allmulti(dev, -1); 594 595 in_dev = __in_dev_get_rtnl(dev); 596 if (in_dev) { 597 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; 598 inet_netconf_notify_devconf(dev_net(dev), 599 NETCONFA_MC_FORWARDING, 600 dev->ifindex, &in_dev->cnf); 601 ip_rt_multicast_event(in_dev); 602 } 603 604 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify) 605 unregister_netdevice_queue(dev, head); 606 607 dev_put(dev); 608 return 0; 609 } 610 611 static void ipmr_cache_free_rcu(struct rcu_head *head) 612 { 613 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu); 614 615 kmem_cache_free(mrt_cachep, c); 616 } 617 618 static inline void ipmr_cache_free(struct mfc_cache *c) 619 { 620 call_rcu(&c->rcu, ipmr_cache_free_rcu); 621 } 622 623 /* Destroy an unresolved cache entry, killing queued skbs 624 * and reporting error to netlink readers. 625 */ 626 627 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c) 628 { 629 struct net *net = read_pnet(&mrt->net); 630 struct sk_buff *skb; 631 struct nlmsgerr *e; 632 633 atomic_dec(&mrt->cache_resolve_queue_len); 634 635 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) { 636 if (ip_hdr(skb)->version == 0) { 637 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); 638 nlh->nlmsg_type = NLMSG_ERROR; 639 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 640 skb_trim(skb, nlh->nlmsg_len); 641 e = nlmsg_data(nlh); 642 e->error = -ETIMEDOUT; 643 memset(&e->msg, 0, sizeof(e->msg)); 644 645 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 646 } else { 647 kfree_skb(skb); 648 } 649 } 650 651 ipmr_cache_free(c); 652 } 653 654 655 /* Timer process for the unresolved queue. */ 656 657 static void ipmr_expire_process(unsigned long arg) 658 { 659 struct mr_table *mrt = (struct mr_table *)arg; 660 unsigned long now; 661 unsigned long expires; 662 struct mfc_cache *c, *next; 663 664 if (!spin_trylock(&mfc_unres_lock)) { 665 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10); 666 return; 667 } 668 669 if (list_empty(&mrt->mfc_unres_queue)) 670 goto out; 671 672 now = jiffies; 673 expires = 10*HZ; 674 675 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 676 if (time_after(c->mfc_un.unres.expires, now)) { 677 unsigned long interval = c->mfc_un.unres.expires - now; 678 if (interval < expires) 679 expires = interval; 680 continue; 681 } 682 683 list_del(&c->list); 684 mroute_netlink_event(mrt, c, RTM_DELROUTE); 685 ipmr_destroy_unres(mrt, c); 686 } 687 688 if (!list_empty(&mrt->mfc_unres_queue)) 689 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires); 690 691 out: 692 spin_unlock(&mfc_unres_lock); 693 } 694 695 /* Fill oifs list. It is called under write locked mrt_lock. */ 696 697 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache, 698 unsigned char *ttls) 699 { 700 int vifi; 701 702 cache->mfc_un.res.minvif = MAXVIFS; 703 cache->mfc_un.res.maxvif = 0; 704 memset(cache->mfc_un.res.ttls, 255, MAXVIFS); 705 706 for (vifi = 0; vifi < mrt->maxvif; vifi++) { 707 if (VIF_EXISTS(mrt, vifi) && 708 ttls[vifi] && ttls[vifi] < 255) { 709 cache->mfc_un.res.ttls[vifi] = ttls[vifi]; 710 if (cache->mfc_un.res.minvif > vifi) 711 cache->mfc_un.res.minvif = vifi; 712 if (cache->mfc_un.res.maxvif <= vifi) 713 cache->mfc_un.res.maxvif = vifi + 1; 714 } 715 } 716 } 717 718 static int vif_add(struct net *net, struct mr_table *mrt, 719 struct vifctl *vifc, int mrtsock) 720 { 721 int vifi = vifc->vifc_vifi; 722 struct vif_device *v = &mrt->vif_table[vifi]; 723 struct net_device *dev; 724 struct in_device *in_dev; 725 int err; 726 727 /* Is vif busy ? */ 728 if (VIF_EXISTS(mrt, vifi)) 729 return -EADDRINUSE; 730 731 switch (vifc->vifc_flags) { 732 #ifdef CONFIG_IP_PIMSM 733 case VIFF_REGISTER: 734 /* 735 * Special Purpose VIF in PIM 736 * All the packets will be sent to the daemon 737 */ 738 if (mrt->mroute_reg_vif_num >= 0) 739 return -EADDRINUSE; 740 dev = ipmr_reg_vif(net, mrt); 741 if (!dev) 742 return -ENOBUFS; 743 err = dev_set_allmulti(dev, 1); 744 if (err) { 745 unregister_netdevice(dev); 746 dev_put(dev); 747 return err; 748 } 749 break; 750 #endif 751 case VIFF_TUNNEL: 752 dev = ipmr_new_tunnel(net, vifc); 753 if (!dev) 754 return -ENOBUFS; 755 err = dev_set_allmulti(dev, 1); 756 if (err) { 757 ipmr_del_tunnel(dev, vifc); 758 dev_put(dev); 759 return err; 760 } 761 break; 762 763 case VIFF_USE_IFINDEX: 764 case 0: 765 if (vifc->vifc_flags == VIFF_USE_IFINDEX) { 766 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex); 767 if (dev && __in_dev_get_rtnl(dev) == NULL) { 768 dev_put(dev); 769 return -EADDRNOTAVAIL; 770 } 771 } else { 772 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr); 773 } 774 if (!dev) 775 return -EADDRNOTAVAIL; 776 err = dev_set_allmulti(dev, 1); 777 if (err) { 778 dev_put(dev); 779 return err; 780 } 781 break; 782 default: 783 return -EINVAL; 784 } 785 786 in_dev = __in_dev_get_rtnl(dev); 787 if (!in_dev) { 788 dev_put(dev); 789 return -EADDRNOTAVAIL; 790 } 791 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; 792 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex, 793 &in_dev->cnf); 794 ip_rt_multicast_event(in_dev); 795 796 /* Fill in the VIF structures */ 797 798 v->rate_limit = vifc->vifc_rate_limit; 799 v->local = vifc->vifc_lcl_addr.s_addr; 800 v->remote = vifc->vifc_rmt_addr.s_addr; 801 v->flags = vifc->vifc_flags; 802 if (!mrtsock) 803 v->flags |= VIFF_STATIC; 804 v->threshold = vifc->vifc_threshold; 805 v->bytes_in = 0; 806 v->bytes_out = 0; 807 v->pkt_in = 0; 808 v->pkt_out = 0; 809 v->link = dev->ifindex; 810 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER)) 811 v->link = dev->iflink; 812 813 /* And finish update writing critical data */ 814 write_lock_bh(&mrt_lock); 815 v->dev = dev; 816 #ifdef CONFIG_IP_PIMSM 817 if (v->flags & VIFF_REGISTER) 818 mrt->mroute_reg_vif_num = vifi; 819 #endif 820 if (vifi+1 > mrt->maxvif) 821 mrt->maxvif = vifi+1; 822 write_unlock_bh(&mrt_lock); 823 return 0; 824 } 825 826 /* called with rcu_read_lock() */ 827 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt, 828 __be32 origin, 829 __be32 mcastgrp) 830 { 831 int line = MFC_HASH(mcastgrp, origin); 832 struct mfc_cache *c; 833 834 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) { 835 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp) 836 return c; 837 } 838 return NULL; 839 } 840 841 /* Look for a (*,*,oif) entry */ 842 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt, 843 int vifi) 844 { 845 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY)); 846 struct mfc_cache *c; 847 848 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) 849 if (c->mfc_origin == htonl(INADDR_ANY) && 850 c->mfc_mcastgrp == htonl(INADDR_ANY) && 851 c->mfc_un.res.ttls[vifi] < 255) 852 return c; 853 854 return NULL; 855 } 856 857 /* Look for a (*,G) entry */ 858 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt, 859 __be32 mcastgrp, int vifi) 860 { 861 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY)); 862 struct mfc_cache *c, *proxy; 863 864 if (mcastgrp == htonl(INADDR_ANY)) 865 goto skip; 866 867 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) 868 if (c->mfc_origin == htonl(INADDR_ANY) && 869 c->mfc_mcastgrp == mcastgrp) { 870 if (c->mfc_un.res.ttls[vifi] < 255) 871 return c; 872 873 /* It's ok if the vifi is part of the static tree */ 874 proxy = ipmr_cache_find_any_parent(mrt, 875 c->mfc_parent); 876 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255) 877 return c; 878 } 879 880 skip: 881 return ipmr_cache_find_any_parent(mrt, vifi); 882 } 883 884 /* 885 * Allocate a multicast cache entry 886 */ 887 static struct mfc_cache *ipmr_cache_alloc(void) 888 { 889 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); 890 891 if (c) 892 c->mfc_un.res.minvif = MAXVIFS; 893 return c; 894 } 895 896 static struct mfc_cache *ipmr_cache_alloc_unres(void) 897 { 898 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); 899 900 if (c) { 901 skb_queue_head_init(&c->mfc_un.unres.unresolved); 902 c->mfc_un.unres.expires = jiffies + 10*HZ; 903 } 904 return c; 905 } 906 907 /* 908 * A cache entry has gone into a resolved state from queued 909 */ 910 911 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt, 912 struct mfc_cache *uc, struct mfc_cache *c) 913 { 914 struct sk_buff *skb; 915 struct nlmsgerr *e; 916 917 /* Play the pending entries through our router */ 918 919 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { 920 if (ip_hdr(skb)->version == 0) { 921 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); 922 923 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) { 924 nlh->nlmsg_len = skb_tail_pointer(skb) - 925 (u8 *)nlh; 926 } else { 927 nlh->nlmsg_type = NLMSG_ERROR; 928 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr)); 929 skb_trim(skb, nlh->nlmsg_len); 930 e = nlmsg_data(nlh); 931 e->error = -EMSGSIZE; 932 memset(&e->msg, 0, sizeof(e->msg)); 933 } 934 935 rtnl_unicast(skb, net, NETLINK_CB(skb).portid); 936 } else { 937 ip_mr_forward(net, mrt, skb, c, 0); 938 } 939 } 940 } 941 942 /* 943 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted 944 * expects the following bizarre scheme. 945 * 946 * Called under mrt_lock. 947 */ 948 949 static int ipmr_cache_report(struct mr_table *mrt, 950 struct sk_buff *pkt, vifi_t vifi, int assert) 951 { 952 struct sk_buff *skb; 953 const int ihl = ip_hdrlen(pkt); 954 struct igmphdr *igmp; 955 struct igmpmsg *msg; 956 struct sock *mroute_sk; 957 int ret; 958 959 #ifdef CONFIG_IP_PIMSM 960 if (assert == IGMPMSG_WHOLEPKT) 961 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); 962 else 963 #endif 964 skb = alloc_skb(128, GFP_ATOMIC); 965 966 if (!skb) 967 return -ENOBUFS; 968 969 #ifdef CONFIG_IP_PIMSM 970 if (assert == IGMPMSG_WHOLEPKT) { 971 /* Ugly, but we have no choice with this interface. 972 * Duplicate old header, fix ihl, length etc. 973 * And all this only to mangle msg->im_msgtype and 974 * to set msg->im_mbz to "mbz" :-) 975 */ 976 skb_push(skb, sizeof(struct iphdr)); 977 skb_reset_network_header(skb); 978 skb_reset_transport_header(skb); 979 msg = (struct igmpmsg *)skb_network_header(skb); 980 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); 981 msg->im_msgtype = IGMPMSG_WHOLEPKT; 982 msg->im_mbz = 0; 983 msg->im_vif = mrt->mroute_reg_vif_num; 984 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; 985 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + 986 sizeof(struct iphdr)); 987 } else 988 #endif 989 { 990 991 /* Copy the IP header */ 992 993 skb_set_network_header(skb, skb->len); 994 skb_put(skb, ihl); 995 skb_copy_to_linear_data(skb, pkt->data, ihl); 996 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */ 997 msg = (struct igmpmsg *)skb_network_header(skb); 998 msg->im_vif = vifi; 999 skb_dst_set(skb, dst_clone(skb_dst(pkt))); 1000 1001 /* Add our header */ 1002 1003 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr)); 1004 igmp->type = 1005 msg->im_msgtype = assert; 1006 igmp->code = 0; 1007 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ 1008 skb->transport_header = skb->network_header; 1009 } 1010 1011 rcu_read_lock(); 1012 mroute_sk = rcu_dereference(mrt->mroute_sk); 1013 if (mroute_sk == NULL) { 1014 rcu_read_unlock(); 1015 kfree_skb(skb); 1016 return -EINVAL; 1017 } 1018 1019 /* Deliver to mrouted */ 1020 1021 ret = sock_queue_rcv_skb(mroute_sk, skb); 1022 rcu_read_unlock(); 1023 if (ret < 0) { 1024 net_warn_ratelimited("mroute: pending queue full, dropping entries\n"); 1025 kfree_skb(skb); 1026 } 1027 1028 return ret; 1029 } 1030 1031 /* 1032 * Queue a packet for resolution. It gets locked cache entry! 1033 */ 1034 1035 static int 1036 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb) 1037 { 1038 bool found = false; 1039 int err; 1040 struct mfc_cache *c; 1041 const struct iphdr *iph = ip_hdr(skb); 1042 1043 spin_lock_bh(&mfc_unres_lock); 1044 list_for_each_entry(c, &mrt->mfc_unres_queue, list) { 1045 if (c->mfc_mcastgrp == iph->daddr && 1046 c->mfc_origin == iph->saddr) { 1047 found = true; 1048 break; 1049 } 1050 } 1051 1052 if (!found) { 1053 /* Create a new entry if allowable */ 1054 1055 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 || 1056 (c = ipmr_cache_alloc_unres()) == NULL) { 1057 spin_unlock_bh(&mfc_unres_lock); 1058 1059 kfree_skb(skb); 1060 return -ENOBUFS; 1061 } 1062 1063 /* Fill in the new cache entry */ 1064 1065 c->mfc_parent = -1; 1066 c->mfc_origin = iph->saddr; 1067 c->mfc_mcastgrp = iph->daddr; 1068 1069 /* Reflect first query at mrouted. */ 1070 1071 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE); 1072 if (err < 0) { 1073 /* If the report failed throw the cache entry 1074 out - Brad Parker 1075 */ 1076 spin_unlock_bh(&mfc_unres_lock); 1077 1078 ipmr_cache_free(c); 1079 kfree_skb(skb); 1080 return err; 1081 } 1082 1083 atomic_inc(&mrt->cache_resolve_queue_len); 1084 list_add(&c->list, &mrt->mfc_unres_queue); 1085 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1086 1087 if (atomic_read(&mrt->cache_resolve_queue_len) == 1) 1088 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires); 1089 } 1090 1091 /* See if we can append the packet */ 1092 1093 if (c->mfc_un.unres.unresolved.qlen > 3) { 1094 kfree_skb(skb); 1095 err = -ENOBUFS; 1096 } else { 1097 skb_queue_tail(&c->mfc_un.unres.unresolved, skb); 1098 err = 0; 1099 } 1100 1101 spin_unlock_bh(&mfc_unres_lock); 1102 return err; 1103 } 1104 1105 /* 1106 * MFC cache manipulation by user space mroute daemon 1107 */ 1108 1109 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent) 1110 { 1111 int line; 1112 struct mfc_cache *c, *next; 1113 1114 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); 1115 1116 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) { 1117 if (c->mfc_origin == mfc->mfcc_origin.s_addr && 1118 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr && 1119 (parent == -1 || parent == c->mfc_parent)) { 1120 list_del_rcu(&c->list); 1121 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1122 ipmr_cache_free(c); 1123 return 0; 1124 } 1125 } 1126 return -ENOENT; 1127 } 1128 1129 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, 1130 struct mfcctl *mfc, int mrtsock, int parent) 1131 { 1132 bool found = false; 1133 int line; 1134 struct mfc_cache *uc, *c; 1135 1136 if (mfc->mfcc_parent >= MAXVIFS) 1137 return -ENFILE; 1138 1139 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); 1140 1141 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) { 1142 if (c->mfc_origin == mfc->mfcc_origin.s_addr && 1143 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr && 1144 (parent == -1 || parent == c->mfc_parent)) { 1145 found = true; 1146 break; 1147 } 1148 } 1149 1150 if (found) { 1151 write_lock_bh(&mrt_lock); 1152 c->mfc_parent = mfc->mfcc_parent; 1153 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1154 if (!mrtsock) 1155 c->mfc_flags |= MFC_STATIC; 1156 write_unlock_bh(&mrt_lock); 1157 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1158 return 0; 1159 } 1160 1161 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) && 1162 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) 1163 return -EINVAL; 1164 1165 c = ipmr_cache_alloc(); 1166 if (c == NULL) 1167 return -ENOMEM; 1168 1169 c->mfc_origin = mfc->mfcc_origin.s_addr; 1170 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; 1171 c->mfc_parent = mfc->mfcc_parent; 1172 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls); 1173 if (!mrtsock) 1174 c->mfc_flags |= MFC_STATIC; 1175 1176 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]); 1177 1178 /* 1179 * Check to see if we resolved a queued list. If so we 1180 * need to send on the frames and tidy up. 1181 */ 1182 found = false; 1183 spin_lock_bh(&mfc_unres_lock); 1184 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) { 1185 if (uc->mfc_origin == c->mfc_origin && 1186 uc->mfc_mcastgrp == c->mfc_mcastgrp) { 1187 list_del(&uc->list); 1188 atomic_dec(&mrt->cache_resolve_queue_len); 1189 found = true; 1190 break; 1191 } 1192 } 1193 if (list_empty(&mrt->mfc_unres_queue)) 1194 del_timer(&mrt->ipmr_expire_timer); 1195 spin_unlock_bh(&mfc_unres_lock); 1196 1197 if (found) { 1198 ipmr_cache_resolve(net, mrt, uc, c); 1199 ipmr_cache_free(uc); 1200 } 1201 mroute_netlink_event(mrt, c, RTM_NEWROUTE); 1202 return 0; 1203 } 1204 1205 /* 1206 * Close the multicast socket, and clear the vif tables etc 1207 */ 1208 1209 static void mroute_clean_tables(struct mr_table *mrt) 1210 { 1211 int i; 1212 LIST_HEAD(list); 1213 struct mfc_cache *c, *next; 1214 1215 /* Shut down all active vif entries */ 1216 1217 for (i = 0; i < mrt->maxvif; i++) { 1218 if (!(mrt->vif_table[i].flags & VIFF_STATIC)) 1219 vif_delete(mrt, i, 0, &list); 1220 } 1221 unregister_netdevice_many(&list); 1222 1223 /* Wipe the cache */ 1224 1225 for (i = 0; i < MFC_LINES; i++) { 1226 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) { 1227 if (c->mfc_flags & MFC_STATIC) 1228 continue; 1229 list_del_rcu(&c->list); 1230 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1231 ipmr_cache_free(c); 1232 } 1233 } 1234 1235 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) { 1236 spin_lock_bh(&mfc_unres_lock); 1237 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) { 1238 list_del(&c->list); 1239 mroute_netlink_event(mrt, c, RTM_DELROUTE); 1240 ipmr_destroy_unres(mrt, c); 1241 } 1242 spin_unlock_bh(&mfc_unres_lock); 1243 } 1244 } 1245 1246 /* called from ip_ra_control(), before an RCU grace period, 1247 * we dont need to call synchronize_rcu() here 1248 */ 1249 static void mrtsock_destruct(struct sock *sk) 1250 { 1251 struct net *net = sock_net(sk); 1252 struct mr_table *mrt; 1253 1254 rtnl_lock(); 1255 ipmr_for_each_table(mrt, net) { 1256 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1257 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--; 1258 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1259 NETCONFA_IFINDEX_ALL, 1260 net->ipv4.devconf_all); 1261 RCU_INIT_POINTER(mrt->mroute_sk, NULL); 1262 mroute_clean_tables(mrt); 1263 } 1264 } 1265 rtnl_unlock(); 1266 } 1267 1268 /* 1269 * Socket options and virtual interface manipulation. The whole 1270 * virtual interface system is a complete heap, but unfortunately 1271 * that's how BSD mrouted happens to think. Maybe one day with a proper 1272 * MOSPF/PIM router set up we can clean this up. 1273 */ 1274 1275 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen) 1276 { 1277 int ret, parent = 0; 1278 struct vifctl vif; 1279 struct mfcctl mfc; 1280 struct net *net = sock_net(sk); 1281 struct mr_table *mrt; 1282 1283 if (sk->sk_type != SOCK_RAW || 1284 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1285 return -EOPNOTSUPP; 1286 1287 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1288 if (mrt == NULL) 1289 return -ENOENT; 1290 1291 if (optname != MRT_INIT) { 1292 if (sk != rcu_access_pointer(mrt->mroute_sk) && 1293 !ns_capable(net->user_ns, CAP_NET_ADMIN)) 1294 return -EACCES; 1295 } 1296 1297 switch (optname) { 1298 case MRT_INIT: 1299 if (optlen != sizeof(int)) 1300 return -EINVAL; 1301 1302 rtnl_lock(); 1303 if (rtnl_dereference(mrt->mroute_sk)) { 1304 rtnl_unlock(); 1305 return -EADDRINUSE; 1306 } 1307 1308 ret = ip_ra_control(sk, 1, mrtsock_destruct); 1309 if (ret == 0) { 1310 rcu_assign_pointer(mrt->mroute_sk, sk); 1311 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++; 1312 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, 1313 NETCONFA_IFINDEX_ALL, 1314 net->ipv4.devconf_all); 1315 } 1316 rtnl_unlock(); 1317 return ret; 1318 case MRT_DONE: 1319 if (sk != rcu_access_pointer(mrt->mroute_sk)) 1320 return -EACCES; 1321 return ip_ra_control(sk, 0, NULL); 1322 case MRT_ADD_VIF: 1323 case MRT_DEL_VIF: 1324 if (optlen != sizeof(vif)) 1325 return -EINVAL; 1326 if (copy_from_user(&vif, optval, sizeof(vif))) 1327 return -EFAULT; 1328 if (vif.vifc_vifi >= MAXVIFS) 1329 return -ENFILE; 1330 rtnl_lock(); 1331 if (optname == MRT_ADD_VIF) { 1332 ret = vif_add(net, mrt, &vif, 1333 sk == rtnl_dereference(mrt->mroute_sk)); 1334 } else { 1335 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL); 1336 } 1337 rtnl_unlock(); 1338 return ret; 1339 1340 /* 1341 * Manipulate the forwarding caches. These live 1342 * in a sort of kernel/user symbiosis. 1343 */ 1344 case MRT_ADD_MFC: 1345 case MRT_DEL_MFC: 1346 parent = -1; 1347 case MRT_ADD_MFC_PROXY: 1348 case MRT_DEL_MFC_PROXY: 1349 if (optlen != sizeof(mfc)) 1350 return -EINVAL; 1351 if (copy_from_user(&mfc, optval, sizeof(mfc))) 1352 return -EFAULT; 1353 if (parent == 0) 1354 parent = mfc.mfcc_parent; 1355 rtnl_lock(); 1356 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY) 1357 ret = ipmr_mfc_delete(mrt, &mfc, parent); 1358 else 1359 ret = ipmr_mfc_add(net, mrt, &mfc, 1360 sk == rtnl_dereference(mrt->mroute_sk), 1361 parent); 1362 rtnl_unlock(); 1363 return ret; 1364 /* 1365 * Control PIM assert. 1366 */ 1367 case MRT_ASSERT: 1368 { 1369 int v; 1370 if (optlen != sizeof(v)) 1371 return -EINVAL; 1372 if (get_user(v, (int __user *)optval)) 1373 return -EFAULT; 1374 mrt->mroute_do_assert = v; 1375 return 0; 1376 } 1377 #ifdef CONFIG_IP_PIMSM 1378 case MRT_PIM: 1379 { 1380 int v; 1381 1382 if (optlen != sizeof(v)) 1383 return -EINVAL; 1384 if (get_user(v, (int __user *)optval)) 1385 return -EFAULT; 1386 v = !!v; 1387 1388 rtnl_lock(); 1389 ret = 0; 1390 if (v != mrt->mroute_do_pim) { 1391 mrt->mroute_do_pim = v; 1392 mrt->mroute_do_assert = v; 1393 } 1394 rtnl_unlock(); 1395 return ret; 1396 } 1397 #endif 1398 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES 1399 case MRT_TABLE: 1400 { 1401 u32 v; 1402 1403 if (optlen != sizeof(u32)) 1404 return -EINVAL; 1405 if (get_user(v, (u32 __user *)optval)) 1406 return -EFAULT; 1407 1408 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */ 1409 if (v != RT_TABLE_DEFAULT && v >= 1000000000) 1410 return -EINVAL; 1411 1412 rtnl_lock(); 1413 ret = 0; 1414 if (sk == rtnl_dereference(mrt->mroute_sk)) { 1415 ret = -EBUSY; 1416 } else { 1417 if (!ipmr_new_table(net, v)) 1418 ret = -ENOMEM; 1419 else 1420 raw_sk(sk)->ipmr_table = v; 1421 } 1422 rtnl_unlock(); 1423 return ret; 1424 } 1425 #endif 1426 /* 1427 * Spurious command, or MRT_VERSION which you cannot 1428 * set. 1429 */ 1430 default: 1431 return -ENOPROTOOPT; 1432 } 1433 } 1434 1435 /* 1436 * Getsock opt support for the multicast routing system. 1437 */ 1438 1439 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen) 1440 { 1441 int olr; 1442 int val; 1443 struct net *net = sock_net(sk); 1444 struct mr_table *mrt; 1445 1446 if (sk->sk_type != SOCK_RAW || 1447 inet_sk(sk)->inet_num != IPPROTO_IGMP) 1448 return -EOPNOTSUPP; 1449 1450 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1451 if (mrt == NULL) 1452 return -ENOENT; 1453 1454 if (optname != MRT_VERSION && 1455 #ifdef CONFIG_IP_PIMSM 1456 optname != MRT_PIM && 1457 #endif 1458 optname != MRT_ASSERT) 1459 return -ENOPROTOOPT; 1460 1461 if (get_user(olr, optlen)) 1462 return -EFAULT; 1463 1464 olr = min_t(unsigned int, olr, sizeof(int)); 1465 if (olr < 0) 1466 return -EINVAL; 1467 1468 if (put_user(olr, optlen)) 1469 return -EFAULT; 1470 if (optname == MRT_VERSION) 1471 val = 0x0305; 1472 #ifdef CONFIG_IP_PIMSM 1473 else if (optname == MRT_PIM) 1474 val = mrt->mroute_do_pim; 1475 #endif 1476 else 1477 val = mrt->mroute_do_assert; 1478 if (copy_to_user(optval, &val, olr)) 1479 return -EFAULT; 1480 return 0; 1481 } 1482 1483 /* 1484 * The IP multicast ioctl support routines. 1485 */ 1486 1487 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) 1488 { 1489 struct sioc_sg_req sr; 1490 struct sioc_vif_req vr; 1491 struct vif_device *vif; 1492 struct mfc_cache *c; 1493 struct net *net = sock_net(sk); 1494 struct mr_table *mrt; 1495 1496 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1497 if (mrt == NULL) 1498 return -ENOENT; 1499 1500 switch (cmd) { 1501 case SIOCGETVIFCNT: 1502 if (copy_from_user(&vr, arg, sizeof(vr))) 1503 return -EFAULT; 1504 if (vr.vifi >= mrt->maxvif) 1505 return -EINVAL; 1506 read_lock(&mrt_lock); 1507 vif = &mrt->vif_table[vr.vifi]; 1508 if (VIF_EXISTS(mrt, vr.vifi)) { 1509 vr.icount = vif->pkt_in; 1510 vr.ocount = vif->pkt_out; 1511 vr.ibytes = vif->bytes_in; 1512 vr.obytes = vif->bytes_out; 1513 read_unlock(&mrt_lock); 1514 1515 if (copy_to_user(arg, &vr, sizeof(vr))) 1516 return -EFAULT; 1517 return 0; 1518 } 1519 read_unlock(&mrt_lock); 1520 return -EADDRNOTAVAIL; 1521 case SIOCGETSGCNT: 1522 if (copy_from_user(&sr, arg, sizeof(sr))) 1523 return -EFAULT; 1524 1525 rcu_read_lock(); 1526 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1527 if (c) { 1528 sr.pktcnt = c->mfc_un.res.pkt; 1529 sr.bytecnt = c->mfc_un.res.bytes; 1530 sr.wrong_if = c->mfc_un.res.wrong_if; 1531 rcu_read_unlock(); 1532 1533 if (copy_to_user(arg, &sr, sizeof(sr))) 1534 return -EFAULT; 1535 return 0; 1536 } 1537 rcu_read_unlock(); 1538 return -EADDRNOTAVAIL; 1539 default: 1540 return -ENOIOCTLCMD; 1541 } 1542 } 1543 1544 #ifdef CONFIG_COMPAT 1545 struct compat_sioc_sg_req { 1546 struct in_addr src; 1547 struct in_addr grp; 1548 compat_ulong_t pktcnt; 1549 compat_ulong_t bytecnt; 1550 compat_ulong_t wrong_if; 1551 }; 1552 1553 struct compat_sioc_vif_req { 1554 vifi_t vifi; /* Which iface */ 1555 compat_ulong_t icount; 1556 compat_ulong_t ocount; 1557 compat_ulong_t ibytes; 1558 compat_ulong_t obytes; 1559 }; 1560 1561 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg) 1562 { 1563 struct compat_sioc_sg_req sr; 1564 struct compat_sioc_vif_req vr; 1565 struct vif_device *vif; 1566 struct mfc_cache *c; 1567 struct net *net = sock_net(sk); 1568 struct mr_table *mrt; 1569 1570 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT); 1571 if (mrt == NULL) 1572 return -ENOENT; 1573 1574 switch (cmd) { 1575 case SIOCGETVIFCNT: 1576 if (copy_from_user(&vr, arg, sizeof(vr))) 1577 return -EFAULT; 1578 if (vr.vifi >= mrt->maxvif) 1579 return -EINVAL; 1580 read_lock(&mrt_lock); 1581 vif = &mrt->vif_table[vr.vifi]; 1582 if (VIF_EXISTS(mrt, vr.vifi)) { 1583 vr.icount = vif->pkt_in; 1584 vr.ocount = vif->pkt_out; 1585 vr.ibytes = vif->bytes_in; 1586 vr.obytes = vif->bytes_out; 1587 read_unlock(&mrt_lock); 1588 1589 if (copy_to_user(arg, &vr, sizeof(vr))) 1590 return -EFAULT; 1591 return 0; 1592 } 1593 read_unlock(&mrt_lock); 1594 return -EADDRNOTAVAIL; 1595 case SIOCGETSGCNT: 1596 if (copy_from_user(&sr, arg, sizeof(sr))) 1597 return -EFAULT; 1598 1599 rcu_read_lock(); 1600 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr); 1601 if (c) { 1602 sr.pktcnt = c->mfc_un.res.pkt; 1603 sr.bytecnt = c->mfc_un.res.bytes; 1604 sr.wrong_if = c->mfc_un.res.wrong_if; 1605 rcu_read_unlock(); 1606 1607 if (copy_to_user(arg, &sr, sizeof(sr))) 1608 return -EFAULT; 1609 return 0; 1610 } 1611 rcu_read_unlock(); 1612 return -EADDRNOTAVAIL; 1613 default: 1614 return -ENOIOCTLCMD; 1615 } 1616 } 1617 #endif 1618 1619 1620 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) 1621 { 1622 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1623 struct net *net = dev_net(dev); 1624 struct mr_table *mrt; 1625 struct vif_device *v; 1626 int ct; 1627 1628 if (event != NETDEV_UNREGISTER) 1629 return NOTIFY_DONE; 1630 1631 ipmr_for_each_table(mrt, net) { 1632 v = &mrt->vif_table[0]; 1633 for (ct = 0; ct < mrt->maxvif; ct++, v++) { 1634 if (v->dev == dev) 1635 vif_delete(mrt, ct, 1, NULL); 1636 } 1637 } 1638 return NOTIFY_DONE; 1639 } 1640 1641 1642 static struct notifier_block ip_mr_notifier = { 1643 .notifier_call = ipmr_device_event, 1644 }; 1645 1646 /* 1647 * Encapsulate a packet by attaching a valid IPIP header to it. 1648 * This avoids tunnel drivers and other mess and gives us the speed so 1649 * important for multicast video. 1650 */ 1651 1652 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr) 1653 { 1654 struct iphdr *iph; 1655 const struct iphdr *old_iph = ip_hdr(skb); 1656 1657 skb_push(skb, sizeof(struct iphdr)); 1658 skb->transport_header = skb->network_header; 1659 skb_reset_network_header(skb); 1660 iph = ip_hdr(skb); 1661 1662 iph->version = 4; 1663 iph->tos = old_iph->tos; 1664 iph->ttl = old_iph->ttl; 1665 iph->frag_off = 0; 1666 iph->daddr = daddr; 1667 iph->saddr = saddr; 1668 iph->protocol = IPPROTO_IPIP; 1669 iph->ihl = 5; 1670 iph->tot_len = htons(skb->len); 1671 ip_select_ident(skb, NULL); 1672 ip_send_check(iph); 1673 1674 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1675 nf_reset(skb); 1676 } 1677 1678 static inline int ipmr_forward_finish(struct sk_buff *skb) 1679 { 1680 struct ip_options *opt = &(IPCB(skb)->opt); 1681 1682 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS); 1683 IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len); 1684 1685 if (unlikely(opt->optlen)) 1686 ip_forward_options(skb); 1687 1688 return dst_output(skb); 1689 } 1690 1691 /* 1692 * Processing handlers for ipmr_forward 1693 */ 1694 1695 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt, 1696 struct sk_buff *skb, struct mfc_cache *c, int vifi) 1697 { 1698 const struct iphdr *iph = ip_hdr(skb); 1699 struct vif_device *vif = &mrt->vif_table[vifi]; 1700 struct net_device *dev; 1701 struct rtable *rt; 1702 struct flowi4 fl4; 1703 int encap = 0; 1704 1705 if (vif->dev == NULL) 1706 goto out_free; 1707 1708 #ifdef CONFIG_IP_PIMSM 1709 if (vif->flags & VIFF_REGISTER) { 1710 vif->pkt_out++; 1711 vif->bytes_out += skb->len; 1712 vif->dev->stats.tx_bytes += skb->len; 1713 vif->dev->stats.tx_packets++; 1714 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT); 1715 goto out_free; 1716 } 1717 #endif 1718 1719 if (vif->flags & VIFF_TUNNEL) { 1720 rt = ip_route_output_ports(net, &fl4, NULL, 1721 vif->remote, vif->local, 1722 0, 0, 1723 IPPROTO_IPIP, 1724 RT_TOS(iph->tos), vif->link); 1725 if (IS_ERR(rt)) 1726 goto out_free; 1727 encap = sizeof(struct iphdr); 1728 } else { 1729 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0, 1730 0, 0, 1731 IPPROTO_IPIP, 1732 RT_TOS(iph->tos), vif->link); 1733 if (IS_ERR(rt)) 1734 goto out_free; 1735 } 1736 1737 dev = rt->dst.dev; 1738 1739 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) { 1740 /* Do not fragment multicasts. Alas, IPv4 does not 1741 * allow to send ICMP, so that packets will disappear 1742 * to blackhole. 1743 */ 1744 1745 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS); 1746 ip_rt_put(rt); 1747 goto out_free; 1748 } 1749 1750 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len; 1751 1752 if (skb_cow(skb, encap)) { 1753 ip_rt_put(rt); 1754 goto out_free; 1755 } 1756 1757 vif->pkt_out++; 1758 vif->bytes_out += skb->len; 1759 1760 skb_dst_drop(skb); 1761 skb_dst_set(skb, &rt->dst); 1762 ip_decrease_ttl(ip_hdr(skb)); 1763 1764 /* FIXME: forward and output firewalls used to be called here. 1765 * What do we do with netfilter? -- RR 1766 */ 1767 if (vif->flags & VIFF_TUNNEL) { 1768 ip_encap(skb, vif->local, vif->remote); 1769 /* FIXME: extra output firewall step used to be here. --RR */ 1770 vif->dev->stats.tx_packets++; 1771 vif->dev->stats.tx_bytes += skb->len; 1772 } 1773 1774 IPCB(skb)->flags |= IPSKB_FORWARDED; 1775 1776 /* 1777 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally 1778 * not only before forwarding, but after forwarding on all output 1779 * interfaces. It is clear, if mrouter runs a multicasting 1780 * program, it should receive packets not depending to what interface 1781 * program is joined. 1782 * If we will not make it, the program will have to join on all 1783 * interfaces. On the other hand, multihoming host (or router, but 1784 * not mrouter) cannot join to more than one interface - it will 1785 * result in receiving multiple packets. 1786 */ 1787 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev, 1788 ipmr_forward_finish); 1789 return; 1790 1791 out_free: 1792 kfree_skb(skb); 1793 } 1794 1795 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev) 1796 { 1797 int ct; 1798 1799 for (ct = mrt->maxvif-1; ct >= 0; ct--) { 1800 if (mrt->vif_table[ct].dev == dev) 1801 break; 1802 } 1803 return ct; 1804 } 1805 1806 /* "local" means that we should preserve one skb (for local delivery) */ 1807 1808 static void ip_mr_forward(struct net *net, struct mr_table *mrt, 1809 struct sk_buff *skb, struct mfc_cache *cache, 1810 int local) 1811 { 1812 int psend = -1; 1813 int vif, ct; 1814 int true_vifi = ipmr_find_vif(mrt, skb->dev); 1815 1816 vif = cache->mfc_parent; 1817 cache->mfc_un.res.pkt++; 1818 cache->mfc_un.res.bytes += skb->len; 1819 1820 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) { 1821 struct mfc_cache *cache_proxy; 1822 1823 /* For an (*,G) entry, we only check that the incomming 1824 * interface is part of the static tree. 1825 */ 1826 cache_proxy = ipmr_cache_find_any_parent(mrt, vif); 1827 if (cache_proxy && 1828 cache_proxy->mfc_un.res.ttls[true_vifi] < 255) 1829 goto forward; 1830 } 1831 1832 /* 1833 * Wrong interface: drop packet and (maybe) send PIM assert. 1834 */ 1835 if (mrt->vif_table[vif].dev != skb->dev) { 1836 if (rt_is_output_route(skb_rtable(skb))) { 1837 /* It is our own packet, looped back. 1838 * Very complicated situation... 1839 * 1840 * The best workaround until routing daemons will be 1841 * fixed is not to redistribute packet, if it was 1842 * send through wrong interface. It means, that 1843 * multicast applications WILL NOT work for 1844 * (S,G), which have default multicast route pointing 1845 * to wrong oif. In any case, it is not a good 1846 * idea to use multicasting applications on router. 1847 */ 1848 goto dont_forward; 1849 } 1850 1851 cache->mfc_un.res.wrong_if++; 1852 1853 if (true_vifi >= 0 && mrt->mroute_do_assert && 1854 /* pimsm uses asserts, when switching from RPT to SPT, 1855 * so that we cannot check that packet arrived on an oif. 1856 * It is bad, but otherwise we would need to move pretty 1857 * large chunk of pimd to kernel. Ough... --ANK 1858 */ 1859 (mrt->mroute_do_pim || 1860 cache->mfc_un.res.ttls[true_vifi] < 255) && 1861 time_after(jiffies, 1862 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { 1863 cache->mfc_un.res.last_assert = jiffies; 1864 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF); 1865 } 1866 goto dont_forward; 1867 } 1868 1869 forward: 1870 mrt->vif_table[vif].pkt_in++; 1871 mrt->vif_table[vif].bytes_in += skb->len; 1872 1873 /* 1874 * Forward the frame 1875 */ 1876 if (cache->mfc_origin == htonl(INADDR_ANY) && 1877 cache->mfc_mcastgrp == htonl(INADDR_ANY)) { 1878 if (true_vifi >= 0 && 1879 true_vifi != cache->mfc_parent && 1880 ip_hdr(skb)->ttl > 1881 cache->mfc_un.res.ttls[cache->mfc_parent]) { 1882 /* It's an (*,*) entry and the packet is not coming from 1883 * the upstream: forward the packet to the upstream 1884 * only. 1885 */ 1886 psend = cache->mfc_parent; 1887 goto last_forward; 1888 } 1889 goto dont_forward; 1890 } 1891 for (ct = cache->mfc_un.res.maxvif - 1; 1892 ct >= cache->mfc_un.res.minvif; ct--) { 1893 /* For (*,G) entry, don't forward to the incoming interface */ 1894 if ((cache->mfc_origin != htonl(INADDR_ANY) || 1895 ct != true_vifi) && 1896 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) { 1897 if (psend != -1) { 1898 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1899 1900 if (skb2) 1901 ipmr_queue_xmit(net, mrt, skb2, cache, 1902 psend); 1903 } 1904 psend = ct; 1905 } 1906 } 1907 last_forward: 1908 if (psend != -1) { 1909 if (local) { 1910 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1911 1912 if (skb2) 1913 ipmr_queue_xmit(net, mrt, skb2, cache, psend); 1914 } else { 1915 ipmr_queue_xmit(net, mrt, skb, cache, psend); 1916 return; 1917 } 1918 } 1919 1920 dont_forward: 1921 if (!local) 1922 kfree_skb(skb); 1923 } 1924 1925 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb) 1926 { 1927 struct rtable *rt = skb_rtable(skb); 1928 struct iphdr *iph = ip_hdr(skb); 1929 struct flowi4 fl4 = { 1930 .daddr = iph->daddr, 1931 .saddr = iph->saddr, 1932 .flowi4_tos = RT_TOS(iph->tos), 1933 .flowi4_oif = (rt_is_output_route(rt) ? 1934 skb->dev->ifindex : 0), 1935 .flowi4_iif = (rt_is_output_route(rt) ? 1936 LOOPBACK_IFINDEX : 1937 skb->dev->ifindex), 1938 .flowi4_mark = skb->mark, 1939 }; 1940 struct mr_table *mrt; 1941 int err; 1942 1943 err = ipmr_fib_lookup(net, &fl4, &mrt); 1944 if (err) 1945 return ERR_PTR(err); 1946 return mrt; 1947 } 1948 1949 /* 1950 * Multicast packets for forwarding arrive here 1951 * Called with rcu_read_lock(); 1952 */ 1953 1954 int ip_mr_input(struct sk_buff *skb) 1955 { 1956 struct mfc_cache *cache; 1957 struct net *net = dev_net(skb->dev); 1958 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; 1959 struct mr_table *mrt; 1960 1961 /* Packet is looped back after forward, it should not be 1962 * forwarded second time, but still can be delivered locally. 1963 */ 1964 if (IPCB(skb)->flags & IPSKB_FORWARDED) 1965 goto dont_forward; 1966 1967 mrt = ipmr_rt_fib_lookup(net, skb); 1968 if (IS_ERR(mrt)) { 1969 kfree_skb(skb); 1970 return PTR_ERR(mrt); 1971 } 1972 if (!local) { 1973 if (IPCB(skb)->opt.router_alert) { 1974 if (ip_call_ra_chain(skb)) 1975 return 0; 1976 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) { 1977 /* IGMPv1 (and broken IGMPv2 implementations sort of 1978 * Cisco IOS <= 11.2(8)) do not put router alert 1979 * option to IGMP packets destined to routable 1980 * groups. It is very bad, because it means 1981 * that we can forward NO IGMP messages. 1982 */ 1983 struct sock *mroute_sk; 1984 1985 mroute_sk = rcu_dereference(mrt->mroute_sk); 1986 if (mroute_sk) { 1987 nf_reset(skb); 1988 raw_rcv(mroute_sk, skb); 1989 return 0; 1990 } 1991 } 1992 } 1993 1994 /* already under rcu_read_lock() */ 1995 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); 1996 if (cache == NULL) { 1997 int vif = ipmr_find_vif(mrt, skb->dev); 1998 1999 if (vif >= 0) 2000 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr, 2001 vif); 2002 } 2003 2004 /* 2005 * No usable cache entry 2006 */ 2007 if (cache == NULL) { 2008 int vif; 2009 2010 if (local) { 2011 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 2012 ip_local_deliver(skb); 2013 if (skb2 == NULL) 2014 return -ENOBUFS; 2015 skb = skb2; 2016 } 2017 2018 read_lock(&mrt_lock); 2019 vif = ipmr_find_vif(mrt, skb->dev); 2020 if (vif >= 0) { 2021 int err2 = ipmr_cache_unresolved(mrt, vif, skb); 2022 read_unlock(&mrt_lock); 2023 2024 return err2; 2025 } 2026 read_unlock(&mrt_lock); 2027 kfree_skb(skb); 2028 return -ENODEV; 2029 } 2030 2031 read_lock(&mrt_lock); 2032 ip_mr_forward(net, mrt, skb, cache, local); 2033 read_unlock(&mrt_lock); 2034 2035 if (local) 2036 return ip_local_deliver(skb); 2037 2038 return 0; 2039 2040 dont_forward: 2041 if (local) 2042 return ip_local_deliver(skb); 2043 kfree_skb(skb); 2044 return 0; 2045 } 2046 2047 #ifdef CONFIG_IP_PIMSM 2048 /* called with rcu_read_lock() */ 2049 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb, 2050 unsigned int pimlen) 2051 { 2052 struct net_device *reg_dev = NULL; 2053 struct iphdr *encap; 2054 2055 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); 2056 /* 2057 * Check that: 2058 * a. packet is really sent to a multicast group 2059 * b. packet is not a NULL-REGISTER 2060 * c. packet is not truncated 2061 */ 2062 if (!ipv4_is_multicast(encap->daddr) || 2063 encap->tot_len == 0 || 2064 ntohs(encap->tot_len) + pimlen > skb->len) 2065 return 1; 2066 2067 read_lock(&mrt_lock); 2068 if (mrt->mroute_reg_vif_num >= 0) 2069 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev; 2070 read_unlock(&mrt_lock); 2071 2072 if (reg_dev == NULL) 2073 return 1; 2074 2075 skb->mac_header = skb->network_header; 2076 skb_pull(skb, (u8 *)encap - skb->data); 2077 skb_reset_network_header(skb); 2078 skb->protocol = htons(ETH_P_IP); 2079 skb->ip_summed = CHECKSUM_NONE; 2080 2081 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev)); 2082 2083 netif_rx(skb); 2084 2085 return NET_RX_SUCCESS; 2086 } 2087 #endif 2088 2089 #ifdef CONFIG_IP_PIMSM_V1 2090 /* 2091 * Handle IGMP messages of PIMv1 2092 */ 2093 2094 int pim_rcv_v1(struct sk_buff *skb) 2095 { 2096 struct igmphdr *pim; 2097 struct net *net = dev_net(skb->dev); 2098 struct mr_table *mrt; 2099 2100 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2101 goto drop; 2102 2103 pim = igmp_hdr(skb); 2104 2105 mrt = ipmr_rt_fib_lookup(net, skb); 2106 if (IS_ERR(mrt)) 2107 goto drop; 2108 if (!mrt->mroute_do_pim || 2109 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) 2110 goto drop; 2111 2112 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2113 drop: 2114 kfree_skb(skb); 2115 } 2116 return 0; 2117 } 2118 #endif 2119 2120 #ifdef CONFIG_IP_PIMSM_V2 2121 static int pim_rcv(struct sk_buff *skb) 2122 { 2123 struct pimreghdr *pim; 2124 struct net *net = dev_net(skb->dev); 2125 struct mr_table *mrt; 2126 2127 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) 2128 goto drop; 2129 2130 pim = (struct pimreghdr *)skb_transport_header(skb); 2131 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) || 2132 (pim->flags & PIM_NULL_REGISTER) || 2133 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && 2134 csum_fold(skb_checksum(skb, 0, skb->len, 0)))) 2135 goto drop; 2136 2137 mrt = ipmr_rt_fib_lookup(net, skb); 2138 if (IS_ERR(mrt)) 2139 goto drop; 2140 if (__pim_rcv(mrt, skb, sizeof(*pim))) { 2141 drop: 2142 kfree_skb(skb); 2143 } 2144 return 0; 2145 } 2146 #endif 2147 2148 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2149 struct mfc_cache *c, struct rtmsg *rtm) 2150 { 2151 int ct; 2152 struct rtnexthop *nhp; 2153 struct nlattr *mp_attr; 2154 struct rta_mfc_stats mfcs; 2155 2156 /* If cache is unresolved, don't try to parse IIF and OIF */ 2157 if (c->mfc_parent >= MAXVIFS) 2158 return -ENOENT; 2159 2160 if (VIF_EXISTS(mrt, c->mfc_parent) && 2161 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0) 2162 return -EMSGSIZE; 2163 2164 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH))) 2165 return -EMSGSIZE; 2166 2167 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { 2168 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) { 2169 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) { 2170 nla_nest_cancel(skb, mp_attr); 2171 return -EMSGSIZE; 2172 } 2173 2174 nhp->rtnh_flags = 0; 2175 nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; 2176 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex; 2177 nhp->rtnh_len = sizeof(*nhp); 2178 } 2179 } 2180 2181 nla_nest_end(skb, mp_attr); 2182 2183 mfcs.mfcs_packets = c->mfc_un.res.pkt; 2184 mfcs.mfcs_bytes = c->mfc_un.res.bytes; 2185 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if; 2186 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0) 2187 return -EMSGSIZE; 2188 2189 rtm->rtm_type = RTN_MULTICAST; 2190 return 1; 2191 } 2192 2193 int ipmr_get_route(struct net *net, struct sk_buff *skb, 2194 __be32 saddr, __be32 daddr, 2195 struct rtmsg *rtm, int nowait) 2196 { 2197 struct mfc_cache *cache; 2198 struct mr_table *mrt; 2199 int err; 2200 2201 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2202 if (mrt == NULL) 2203 return -ENOENT; 2204 2205 rcu_read_lock(); 2206 cache = ipmr_cache_find(mrt, saddr, daddr); 2207 if (cache == NULL && skb->dev) { 2208 int vif = ipmr_find_vif(mrt, skb->dev); 2209 2210 if (vif >= 0) 2211 cache = ipmr_cache_find_any(mrt, daddr, vif); 2212 } 2213 if (cache == NULL) { 2214 struct sk_buff *skb2; 2215 struct iphdr *iph; 2216 struct net_device *dev; 2217 int vif = -1; 2218 2219 if (nowait) { 2220 rcu_read_unlock(); 2221 return -EAGAIN; 2222 } 2223 2224 dev = skb->dev; 2225 read_lock(&mrt_lock); 2226 if (dev) 2227 vif = ipmr_find_vif(mrt, dev); 2228 if (vif < 0) { 2229 read_unlock(&mrt_lock); 2230 rcu_read_unlock(); 2231 return -ENODEV; 2232 } 2233 skb2 = skb_clone(skb, GFP_ATOMIC); 2234 if (!skb2) { 2235 read_unlock(&mrt_lock); 2236 rcu_read_unlock(); 2237 return -ENOMEM; 2238 } 2239 2240 skb_push(skb2, sizeof(struct iphdr)); 2241 skb_reset_network_header(skb2); 2242 iph = ip_hdr(skb2); 2243 iph->ihl = sizeof(struct iphdr) >> 2; 2244 iph->saddr = saddr; 2245 iph->daddr = daddr; 2246 iph->version = 0; 2247 err = ipmr_cache_unresolved(mrt, vif, skb2); 2248 read_unlock(&mrt_lock); 2249 rcu_read_unlock(); 2250 return err; 2251 } 2252 2253 read_lock(&mrt_lock); 2254 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY)) 2255 cache->mfc_flags |= MFC_NOTIFY; 2256 err = __ipmr_fill_mroute(mrt, skb, cache, rtm); 2257 read_unlock(&mrt_lock); 2258 rcu_read_unlock(); 2259 return err; 2260 } 2261 2262 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, 2263 u32 portid, u32 seq, struct mfc_cache *c, int cmd, 2264 int flags) 2265 { 2266 struct nlmsghdr *nlh; 2267 struct rtmsg *rtm; 2268 int err; 2269 2270 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags); 2271 if (nlh == NULL) 2272 return -EMSGSIZE; 2273 2274 rtm = nlmsg_data(nlh); 2275 rtm->rtm_family = RTNL_FAMILY_IPMR; 2276 rtm->rtm_dst_len = 32; 2277 rtm->rtm_src_len = 32; 2278 rtm->rtm_tos = 0; 2279 rtm->rtm_table = mrt->id; 2280 if (nla_put_u32(skb, RTA_TABLE, mrt->id)) 2281 goto nla_put_failure; 2282 rtm->rtm_type = RTN_MULTICAST; 2283 rtm->rtm_scope = RT_SCOPE_UNIVERSE; 2284 if (c->mfc_flags & MFC_STATIC) 2285 rtm->rtm_protocol = RTPROT_STATIC; 2286 else 2287 rtm->rtm_protocol = RTPROT_MROUTED; 2288 rtm->rtm_flags = 0; 2289 2290 if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) || 2291 nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp)) 2292 goto nla_put_failure; 2293 err = __ipmr_fill_mroute(mrt, skb, c, rtm); 2294 /* do not break the dump if cache is unresolved */ 2295 if (err < 0 && err != -ENOENT) 2296 goto nla_put_failure; 2297 2298 nlmsg_end(skb, nlh); 2299 return 0; 2300 2301 nla_put_failure: 2302 nlmsg_cancel(skb, nlh); 2303 return -EMSGSIZE; 2304 } 2305 2306 static size_t mroute_msgsize(bool unresolved, int maxvif) 2307 { 2308 size_t len = 2309 NLMSG_ALIGN(sizeof(struct rtmsg)) 2310 + nla_total_size(4) /* RTA_TABLE */ 2311 + nla_total_size(4) /* RTA_SRC */ 2312 + nla_total_size(4) /* RTA_DST */ 2313 ; 2314 2315 if (!unresolved) 2316 len = len 2317 + nla_total_size(4) /* RTA_IIF */ 2318 + nla_total_size(0) /* RTA_MULTIPATH */ 2319 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop)) 2320 /* RTA_MFC_STATS */ 2321 + nla_total_size(sizeof(struct rta_mfc_stats)) 2322 ; 2323 2324 return len; 2325 } 2326 2327 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, 2328 int cmd) 2329 { 2330 struct net *net = read_pnet(&mrt->net); 2331 struct sk_buff *skb; 2332 int err = -ENOBUFS; 2333 2334 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif), 2335 GFP_ATOMIC); 2336 if (skb == NULL) 2337 goto errout; 2338 2339 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0); 2340 if (err < 0) 2341 goto errout; 2342 2343 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC); 2344 return; 2345 2346 errout: 2347 kfree_skb(skb); 2348 if (err < 0) 2349 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err); 2350 } 2351 2352 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb) 2353 { 2354 struct net *net = sock_net(skb->sk); 2355 struct mr_table *mrt; 2356 struct mfc_cache *mfc; 2357 unsigned int t = 0, s_t; 2358 unsigned int h = 0, s_h; 2359 unsigned int e = 0, s_e; 2360 2361 s_t = cb->args[0]; 2362 s_h = cb->args[1]; 2363 s_e = cb->args[2]; 2364 2365 rcu_read_lock(); 2366 ipmr_for_each_table(mrt, net) { 2367 if (t < s_t) 2368 goto next_table; 2369 if (t > s_t) 2370 s_h = 0; 2371 for (h = s_h; h < MFC_LINES; h++) { 2372 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) { 2373 if (e < s_e) 2374 goto next_entry; 2375 if (ipmr_fill_mroute(mrt, skb, 2376 NETLINK_CB(cb->skb).portid, 2377 cb->nlh->nlmsg_seq, 2378 mfc, RTM_NEWROUTE, 2379 NLM_F_MULTI) < 0) 2380 goto done; 2381 next_entry: 2382 e++; 2383 } 2384 e = s_e = 0; 2385 } 2386 spin_lock_bh(&mfc_unres_lock); 2387 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) { 2388 if (e < s_e) 2389 goto next_entry2; 2390 if (ipmr_fill_mroute(mrt, skb, 2391 NETLINK_CB(cb->skb).portid, 2392 cb->nlh->nlmsg_seq, 2393 mfc, RTM_NEWROUTE, 2394 NLM_F_MULTI) < 0) { 2395 spin_unlock_bh(&mfc_unres_lock); 2396 goto done; 2397 } 2398 next_entry2: 2399 e++; 2400 } 2401 spin_unlock_bh(&mfc_unres_lock); 2402 e = s_e = 0; 2403 s_h = 0; 2404 next_table: 2405 t++; 2406 } 2407 done: 2408 rcu_read_unlock(); 2409 2410 cb->args[2] = e; 2411 cb->args[1] = h; 2412 cb->args[0] = t; 2413 2414 return skb->len; 2415 } 2416 2417 #ifdef CONFIG_PROC_FS 2418 /* 2419 * The /proc interfaces to multicast routing : 2420 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif 2421 */ 2422 struct ipmr_vif_iter { 2423 struct seq_net_private p; 2424 struct mr_table *mrt; 2425 int ct; 2426 }; 2427 2428 static struct vif_device *ipmr_vif_seq_idx(struct net *net, 2429 struct ipmr_vif_iter *iter, 2430 loff_t pos) 2431 { 2432 struct mr_table *mrt = iter->mrt; 2433 2434 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) { 2435 if (!VIF_EXISTS(mrt, iter->ct)) 2436 continue; 2437 if (pos-- == 0) 2438 return &mrt->vif_table[iter->ct]; 2439 } 2440 return NULL; 2441 } 2442 2443 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) 2444 __acquires(mrt_lock) 2445 { 2446 struct ipmr_vif_iter *iter = seq->private; 2447 struct net *net = seq_file_net(seq); 2448 struct mr_table *mrt; 2449 2450 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2451 if (mrt == NULL) 2452 return ERR_PTR(-ENOENT); 2453 2454 iter->mrt = mrt; 2455 2456 read_lock(&mrt_lock); 2457 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1) 2458 : SEQ_START_TOKEN; 2459 } 2460 2461 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2462 { 2463 struct ipmr_vif_iter *iter = seq->private; 2464 struct net *net = seq_file_net(seq); 2465 struct mr_table *mrt = iter->mrt; 2466 2467 ++*pos; 2468 if (v == SEQ_START_TOKEN) 2469 return ipmr_vif_seq_idx(net, iter, 0); 2470 2471 while (++iter->ct < mrt->maxvif) { 2472 if (!VIF_EXISTS(mrt, iter->ct)) 2473 continue; 2474 return &mrt->vif_table[iter->ct]; 2475 } 2476 return NULL; 2477 } 2478 2479 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) 2480 __releases(mrt_lock) 2481 { 2482 read_unlock(&mrt_lock); 2483 } 2484 2485 static int ipmr_vif_seq_show(struct seq_file *seq, void *v) 2486 { 2487 struct ipmr_vif_iter *iter = seq->private; 2488 struct mr_table *mrt = iter->mrt; 2489 2490 if (v == SEQ_START_TOKEN) { 2491 seq_puts(seq, 2492 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); 2493 } else { 2494 const struct vif_device *vif = v; 2495 const char *name = vif->dev ? vif->dev->name : "none"; 2496 2497 seq_printf(seq, 2498 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", 2499 vif - mrt->vif_table, 2500 name, vif->bytes_in, vif->pkt_in, 2501 vif->bytes_out, vif->pkt_out, 2502 vif->flags, vif->local, vif->remote); 2503 } 2504 return 0; 2505 } 2506 2507 static const struct seq_operations ipmr_vif_seq_ops = { 2508 .start = ipmr_vif_seq_start, 2509 .next = ipmr_vif_seq_next, 2510 .stop = ipmr_vif_seq_stop, 2511 .show = ipmr_vif_seq_show, 2512 }; 2513 2514 static int ipmr_vif_open(struct inode *inode, struct file *file) 2515 { 2516 return seq_open_net(inode, file, &ipmr_vif_seq_ops, 2517 sizeof(struct ipmr_vif_iter)); 2518 } 2519 2520 static const struct file_operations ipmr_vif_fops = { 2521 .owner = THIS_MODULE, 2522 .open = ipmr_vif_open, 2523 .read = seq_read, 2524 .llseek = seq_lseek, 2525 .release = seq_release_net, 2526 }; 2527 2528 struct ipmr_mfc_iter { 2529 struct seq_net_private p; 2530 struct mr_table *mrt; 2531 struct list_head *cache; 2532 int ct; 2533 }; 2534 2535 2536 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net, 2537 struct ipmr_mfc_iter *it, loff_t pos) 2538 { 2539 struct mr_table *mrt = it->mrt; 2540 struct mfc_cache *mfc; 2541 2542 rcu_read_lock(); 2543 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) { 2544 it->cache = &mrt->mfc_cache_array[it->ct]; 2545 list_for_each_entry_rcu(mfc, it->cache, list) 2546 if (pos-- == 0) 2547 return mfc; 2548 } 2549 rcu_read_unlock(); 2550 2551 spin_lock_bh(&mfc_unres_lock); 2552 it->cache = &mrt->mfc_unres_queue; 2553 list_for_each_entry(mfc, it->cache, list) 2554 if (pos-- == 0) 2555 return mfc; 2556 spin_unlock_bh(&mfc_unres_lock); 2557 2558 it->cache = NULL; 2559 return NULL; 2560 } 2561 2562 2563 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) 2564 { 2565 struct ipmr_mfc_iter *it = seq->private; 2566 struct net *net = seq_file_net(seq); 2567 struct mr_table *mrt; 2568 2569 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT); 2570 if (mrt == NULL) 2571 return ERR_PTR(-ENOENT); 2572 2573 it->mrt = mrt; 2574 it->cache = NULL; 2575 it->ct = 0; 2576 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) 2577 : SEQ_START_TOKEN; 2578 } 2579 2580 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2581 { 2582 struct mfc_cache *mfc = v; 2583 struct ipmr_mfc_iter *it = seq->private; 2584 struct net *net = seq_file_net(seq); 2585 struct mr_table *mrt = it->mrt; 2586 2587 ++*pos; 2588 2589 if (v == SEQ_START_TOKEN) 2590 return ipmr_mfc_seq_idx(net, seq->private, 0); 2591 2592 if (mfc->list.next != it->cache) 2593 return list_entry(mfc->list.next, struct mfc_cache, list); 2594 2595 if (it->cache == &mrt->mfc_unres_queue) 2596 goto end_of_list; 2597 2598 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]); 2599 2600 while (++it->ct < MFC_LINES) { 2601 it->cache = &mrt->mfc_cache_array[it->ct]; 2602 if (list_empty(it->cache)) 2603 continue; 2604 return list_first_entry(it->cache, struct mfc_cache, list); 2605 } 2606 2607 /* exhausted cache_array, show unresolved */ 2608 rcu_read_unlock(); 2609 it->cache = &mrt->mfc_unres_queue; 2610 it->ct = 0; 2611 2612 spin_lock_bh(&mfc_unres_lock); 2613 if (!list_empty(it->cache)) 2614 return list_first_entry(it->cache, struct mfc_cache, list); 2615 2616 end_of_list: 2617 spin_unlock_bh(&mfc_unres_lock); 2618 it->cache = NULL; 2619 2620 return NULL; 2621 } 2622 2623 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) 2624 { 2625 struct ipmr_mfc_iter *it = seq->private; 2626 struct mr_table *mrt = it->mrt; 2627 2628 if (it->cache == &mrt->mfc_unres_queue) 2629 spin_unlock_bh(&mfc_unres_lock); 2630 else if (it->cache == &mrt->mfc_cache_array[it->ct]) 2631 rcu_read_unlock(); 2632 } 2633 2634 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) 2635 { 2636 int n; 2637 2638 if (v == SEQ_START_TOKEN) { 2639 seq_puts(seq, 2640 "Group Origin Iif Pkts Bytes Wrong Oifs\n"); 2641 } else { 2642 const struct mfc_cache *mfc = v; 2643 const struct ipmr_mfc_iter *it = seq->private; 2644 const struct mr_table *mrt = it->mrt; 2645 2646 seq_printf(seq, "%08X %08X %-3hd", 2647 (__force u32) mfc->mfc_mcastgrp, 2648 (__force u32) mfc->mfc_origin, 2649 mfc->mfc_parent); 2650 2651 if (it->cache != &mrt->mfc_unres_queue) { 2652 seq_printf(seq, " %8lu %8lu %8lu", 2653 mfc->mfc_un.res.pkt, 2654 mfc->mfc_un.res.bytes, 2655 mfc->mfc_un.res.wrong_if); 2656 for (n = mfc->mfc_un.res.minvif; 2657 n < mfc->mfc_un.res.maxvif; n++) { 2658 if (VIF_EXISTS(mrt, n) && 2659 mfc->mfc_un.res.ttls[n] < 255) 2660 seq_printf(seq, 2661 " %2d:%-3d", 2662 n, mfc->mfc_un.res.ttls[n]); 2663 } 2664 } else { 2665 /* unresolved mfc_caches don't contain 2666 * pkt, bytes and wrong_if values 2667 */ 2668 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); 2669 } 2670 seq_putc(seq, '\n'); 2671 } 2672 return 0; 2673 } 2674 2675 static const struct seq_operations ipmr_mfc_seq_ops = { 2676 .start = ipmr_mfc_seq_start, 2677 .next = ipmr_mfc_seq_next, 2678 .stop = ipmr_mfc_seq_stop, 2679 .show = ipmr_mfc_seq_show, 2680 }; 2681 2682 static int ipmr_mfc_open(struct inode *inode, struct file *file) 2683 { 2684 return seq_open_net(inode, file, &ipmr_mfc_seq_ops, 2685 sizeof(struct ipmr_mfc_iter)); 2686 } 2687 2688 static const struct file_operations ipmr_mfc_fops = { 2689 .owner = THIS_MODULE, 2690 .open = ipmr_mfc_open, 2691 .read = seq_read, 2692 .llseek = seq_lseek, 2693 .release = seq_release_net, 2694 }; 2695 #endif 2696 2697 #ifdef CONFIG_IP_PIMSM_V2 2698 static const struct net_protocol pim_protocol = { 2699 .handler = pim_rcv, 2700 .netns_ok = 1, 2701 }; 2702 #endif 2703 2704 2705 /* 2706 * Setup for IP multicast routing 2707 */ 2708 static int __net_init ipmr_net_init(struct net *net) 2709 { 2710 int err; 2711 2712 err = ipmr_rules_init(net); 2713 if (err < 0) 2714 goto fail; 2715 2716 #ifdef CONFIG_PROC_FS 2717 err = -ENOMEM; 2718 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops)) 2719 goto proc_vif_fail; 2720 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops)) 2721 goto proc_cache_fail; 2722 #endif 2723 return 0; 2724 2725 #ifdef CONFIG_PROC_FS 2726 proc_cache_fail: 2727 remove_proc_entry("ip_mr_vif", net->proc_net); 2728 proc_vif_fail: 2729 ipmr_rules_exit(net); 2730 #endif 2731 fail: 2732 return err; 2733 } 2734 2735 static void __net_exit ipmr_net_exit(struct net *net) 2736 { 2737 #ifdef CONFIG_PROC_FS 2738 remove_proc_entry("ip_mr_cache", net->proc_net); 2739 remove_proc_entry("ip_mr_vif", net->proc_net); 2740 #endif 2741 ipmr_rules_exit(net); 2742 } 2743 2744 static struct pernet_operations ipmr_net_ops = { 2745 .init = ipmr_net_init, 2746 .exit = ipmr_net_exit, 2747 }; 2748 2749 int __init ip_mr_init(void) 2750 { 2751 int err; 2752 2753 mrt_cachep = kmem_cache_create("ip_mrt_cache", 2754 sizeof(struct mfc_cache), 2755 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, 2756 NULL); 2757 if (!mrt_cachep) 2758 return -ENOMEM; 2759 2760 err = register_pernet_subsys(&ipmr_net_ops); 2761 if (err) 2762 goto reg_pernet_fail; 2763 2764 err = register_netdevice_notifier(&ip_mr_notifier); 2765 if (err) 2766 goto reg_notif_fail; 2767 #ifdef CONFIG_IP_PIMSM_V2 2768 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) { 2769 pr_err("%s: can't add PIM protocol\n", __func__); 2770 err = -EAGAIN; 2771 goto add_proto_fail; 2772 } 2773 #endif 2774 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE, 2775 NULL, ipmr_rtm_dumproute, NULL); 2776 return 0; 2777 2778 #ifdef CONFIG_IP_PIMSM_V2 2779 add_proto_fail: 2780 unregister_netdevice_notifier(&ip_mr_notifier); 2781 #endif 2782 reg_notif_fail: 2783 unregister_pernet_subsys(&ipmr_net_ops); 2784 reg_pernet_fail: 2785 kmem_cache_destroy(mrt_cachep); 2786 return err; 2787 } 2788