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