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