2 * IP multicast routing support for mrouted 3.6/3.8
4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5 * Linux Consultancy and Custom Driver Development
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.
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
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.
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>
35 #include <linux/kernel.h>
36 #include <linux/fcntl.h>
37 #include <linux/stat.h>
38 #include <linux/socket.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>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.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>
71 struct list_head list;
74 struct sock __rcu *mroute_sk;
75 struct timer_list ipmr_expire_timer;
76 struct list_head mfc_unres_queue;
77 struct list_head mfc_cache_array[MFC_LINES];
78 struct vif_device vif_table[MAXVIFS];
80 atomic_t cache_resolve_queue_len;
81 bool mroute_do_assert;
83 int mroute_reg_vif_num;
87 struct fib_rule common;
94 static inline bool pimsm_enabled(void)
96 return IS_BUILTIN(CONFIG_IP_PIMSM_V1) || IS_BUILTIN(CONFIG_IP_PIMSM_V2);
99 /* Big lock, protecting vif table, mrt cache and mroute socket state.
100 * Note that the changes are semaphored via rtnl_lock.
103 static DEFINE_RWLOCK(mrt_lock);
106 * Multicast router control variables
109 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
111 /* Special spinlock for queue of unresolved entries */
112 static DEFINE_SPINLOCK(mfc_unres_lock);
114 /* We return to original Alan's scheme. Hash table of resolved
115 * entries is changed only in process context and protected
116 * with weak lock mrt_lock. Queue of unresolved entries is protected
117 * with strong spinlock mfc_unres_lock.
119 * In this case data path is free of exclusive locks at all.
122 static struct kmem_cache *mrt_cachep __read_mostly;
124 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
125 static void ipmr_free_table(struct mr_table *mrt);
127 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
128 struct sk_buff *skb, struct mfc_cache *cache,
130 static int ipmr_cache_report(struct mr_table *mrt,
131 struct sk_buff *pkt, vifi_t vifi, int assert);
132 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
133 struct mfc_cache *c, struct rtmsg *rtm);
134 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
136 static void mroute_clean_tables(struct mr_table *mrt);
137 static void ipmr_expire_process(unsigned long arg);
139 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
140 #define ipmr_for_each_table(mrt, net) \
141 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
143 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
145 struct mr_table *mrt;
147 ipmr_for_each_table(mrt, net) {
154 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
155 struct mr_table **mrt)
158 struct ipmr_result res;
159 struct fib_lookup_arg arg = {
161 .flags = FIB_LOOKUP_NOREF,
164 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
165 flowi4_to_flowi(flp4), 0, &arg);
172 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
173 int flags, struct fib_lookup_arg *arg)
175 struct ipmr_result *res = arg->result;
176 struct mr_table *mrt;
178 switch (rule->action) {
181 case FR_ACT_UNREACHABLE:
183 case FR_ACT_PROHIBIT:
185 case FR_ACT_BLACKHOLE:
190 mrt = ipmr_get_table(rule->fr_net, rule->table);
197 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
202 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
206 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
207 struct fib_rule_hdr *frh, struct nlattr **tb)
212 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
218 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
219 struct fib_rule_hdr *frh)
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 .fill = ipmr_rule_fill,
236 .nlgroup = RTNLGRP_IPV4_RULE,
237 .policy = ipmr_rule_policy,
238 .owner = THIS_MODULE,
241 static int __net_init ipmr_rules_init(struct net *net)
243 struct fib_rules_ops *ops;
244 struct mr_table *mrt;
247 ops = fib_rules_register(&ipmr_rules_ops_template, net);
251 INIT_LIST_HEAD(&net->ipv4.mr_tables);
253 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
259 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
263 net->ipv4.mr_rules_ops = ops;
267 ipmr_free_table(mrt);
269 fib_rules_unregister(ops);
273 static void __net_exit ipmr_rules_exit(struct net *net)
275 struct mr_table *mrt, *next;
278 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
279 list_del(&mrt->list);
280 ipmr_free_table(mrt);
282 fib_rules_unregister(net->ipv4.mr_rules_ops);
286 #define ipmr_for_each_table(mrt, net) \
287 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
289 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
291 return net->ipv4.mrt;
294 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
295 struct mr_table **mrt)
297 *mrt = net->ipv4.mrt;
301 static int __net_init ipmr_rules_init(struct net *net)
303 struct mr_table *mrt;
305 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
312 static void __net_exit ipmr_rules_exit(struct net *net)
315 ipmr_free_table(net->ipv4.mrt);
316 net->ipv4.mrt = NULL;
321 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
323 struct mr_table *mrt;
326 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
327 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
328 return ERR_PTR(-EINVAL);
330 mrt = ipmr_get_table(net, id);
334 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
336 return ERR_PTR(-ENOMEM);
337 write_pnet(&mrt->net, net);
340 /* Forwarding cache */
341 for (i = 0; i < MFC_LINES; i++)
342 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
344 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
346 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
349 mrt->mroute_reg_vif_num = -1;
350 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
351 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
356 static void ipmr_free_table(struct mr_table *mrt)
358 del_timer_sync(&mrt->ipmr_expire_timer);
359 mroute_clean_tables(mrt);
363 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
365 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
367 struct net *net = dev_net(dev);
371 dev = __dev_get_by_name(net, "tunl0");
373 const struct net_device_ops *ops = dev->netdev_ops;
375 struct ip_tunnel_parm p;
377 memset(&p, 0, sizeof(p));
378 p.iph.daddr = v->vifc_rmt_addr.s_addr;
379 p.iph.saddr = v->vifc_lcl_addr.s_addr;
382 p.iph.protocol = IPPROTO_IPIP;
383 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
384 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
386 if (ops->ndo_do_ioctl) {
387 mm_segment_t oldfs = get_fs();
390 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
397 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
399 struct net_device *dev;
401 dev = __dev_get_by_name(net, "tunl0");
404 const struct net_device_ops *ops = dev->netdev_ops;
407 struct ip_tunnel_parm p;
408 struct in_device *in_dev;
410 memset(&p, 0, sizeof(p));
411 p.iph.daddr = v->vifc_rmt_addr.s_addr;
412 p.iph.saddr = v->vifc_lcl_addr.s_addr;
415 p.iph.protocol = IPPROTO_IPIP;
416 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
417 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
419 if (ops->ndo_do_ioctl) {
420 mm_segment_t oldfs = get_fs();
423 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
431 (dev = __dev_get_by_name(net, p.name)) != NULL) {
432 dev->flags |= IFF_MULTICAST;
434 in_dev = __in_dev_get_rtnl(dev);
438 ipv4_devconf_setall(in_dev);
439 neigh_parms_data_state_setall(in_dev->arp_parms);
440 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
450 /* allow the register to be completed before unregistering. */
454 unregister_netdevice(dev);
458 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
459 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
461 struct net *net = dev_net(dev);
462 struct mr_table *mrt;
463 struct flowi4 fl4 = {
464 .flowi4_oif = dev->ifindex,
465 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
466 .flowi4_mark = skb->mark,
470 err = ipmr_fib_lookup(net, &fl4, &mrt);
476 read_lock(&mrt_lock);
477 dev->stats.tx_bytes += skb->len;
478 dev->stats.tx_packets++;
479 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
480 read_unlock(&mrt_lock);
485 static int reg_vif_get_iflink(const struct net_device *dev)
490 static const struct net_device_ops reg_vif_netdev_ops = {
491 .ndo_start_xmit = reg_vif_xmit,
492 .ndo_get_iflink = reg_vif_get_iflink,
495 static void reg_vif_setup(struct net_device *dev)
497 dev->type = ARPHRD_PIMREG;
498 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
499 dev->flags = IFF_NOARP;
500 dev->netdev_ops = ®_vif_netdev_ops;
501 dev->destructor = free_netdev;
502 dev->features |= NETIF_F_NETNS_LOCAL;
505 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
507 struct net_device *dev;
508 struct in_device *in_dev;
511 if (mrt->id == RT_TABLE_DEFAULT)
512 sprintf(name, "pimreg");
514 sprintf(name, "pimreg%u", mrt->id);
516 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
521 dev_net_set(dev, net);
523 if (register_netdevice(dev)) {
529 in_dev = __in_dev_get_rcu(dev);
535 ipv4_devconf_setall(in_dev);
536 neigh_parms_data_state_setall(in_dev->arp_parms);
537 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
548 /* allow the register to be completed before unregistering. */
552 unregister_netdevice(dev);
556 /* called with rcu_read_lock() */
557 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
560 struct net_device *reg_dev = NULL;
563 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
566 * a. packet is really sent to a multicast group
567 * b. packet is not a NULL-REGISTER
568 * c. packet is not truncated
570 if (!ipv4_is_multicast(encap->daddr) ||
571 encap->tot_len == 0 ||
572 ntohs(encap->tot_len) + pimlen > skb->len)
575 read_lock(&mrt_lock);
576 if (mrt->mroute_reg_vif_num >= 0)
577 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
578 read_unlock(&mrt_lock);
583 skb->mac_header = skb->network_header;
584 skb_pull(skb, (u8 *)encap - skb->data);
585 skb_reset_network_header(skb);
586 skb->protocol = htons(ETH_P_IP);
587 skb->ip_summed = CHECKSUM_NONE;
589 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
593 return NET_RX_SUCCESS;
596 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
603 * vif_delete - Delete a VIF entry
604 * @notify: Set to 1, if the caller is a notifier_call
607 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
608 struct list_head *head)
610 struct vif_device *v;
611 struct net_device *dev;
612 struct in_device *in_dev;
614 if (vifi < 0 || vifi >= mrt->maxvif)
615 return -EADDRNOTAVAIL;
617 v = &mrt->vif_table[vifi];
619 write_lock_bh(&mrt_lock);
624 write_unlock_bh(&mrt_lock);
625 return -EADDRNOTAVAIL;
628 if (vifi == mrt->mroute_reg_vif_num)
629 mrt->mroute_reg_vif_num = -1;
631 if (vifi + 1 == mrt->maxvif) {
634 for (tmp = vifi - 1; tmp >= 0; tmp--) {
635 if (VIF_EXISTS(mrt, tmp))
641 write_unlock_bh(&mrt_lock);
643 dev_set_allmulti(dev, -1);
645 in_dev = __in_dev_get_rtnl(dev);
647 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
648 inet_netconf_notify_devconf(dev_net(dev),
649 NETCONFA_MC_FORWARDING,
650 dev->ifindex, &in_dev->cnf);
651 ip_rt_multicast_event(in_dev);
654 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
655 unregister_netdevice_queue(dev, head);
661 static void ipmr_cache_free_rcu(struct rcu_head *head)
663 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
665 kmem_cache_free(mrt_cachep, c);
668 static inline void ipmr_cache_free(struct mfc_cache *c)
670 call_rcu(&c->rcu, ipmr_cache_free_rcu);
673 /* Destroy an unresolved cache entry, killing queued skbs
674 * and reporting error to netlink readers.
677 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
679 struct net *net = read_pnet(&mrt->net);
683 atomic_dec(&mrt->cache_resolve_queue_len);
685 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
686 if (ip_hdr(skb)->version == 0) {
687 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
688 nlh->nlmsg_type = NLMSG_ERROR;
689 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
690 skb_trim(skb, nlh->nlmsg_len);
692 e->error = -ETIMEDOUT;
693 memset(&e->msg, 0, sizeof(e->msg));
695 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
705 /* Timer process for the unresolved queue. */
707 static void ipmr_expire_process(unsigned long arg)
709 struct mr_table *mrt = (struct mr_table *)arg;
711 unsigned long expires;
712 struct mfc_cache *c, *next;
714 if (!spin_trylock(&mfc_unres_lock)) {
715 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
719 if (list_empty(&mrt->mfc_unres_queue))
725 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
726 if (time_after(c->mfc_un.unres.expires, now)) {
727 unsigned long interval = c->mfc_un.unres.expires - now;
728 if (interval < expires)
734 mroute_netlink_event(mrt, c, RTM_DELROUTE);
735 ipmr_destroy_unres(mrt, c);
738 if (!list_empty(&mrt->mfc_unres_queue))
739 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
742 spin_unlock(&mfc_unres_lock);
745 /* Fill oifs list. It is called under write locked mrt_lock. */
747 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
752 cache->mfc_un.res.minvif = MAXVIFS;
753 cache->mfc_un.res.maxvif = 0;
754 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
756 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
757 if (VIF_EXISTS(mrt, vifi) &&
758 ttls[vifi] && ttls[vifi] < 255) {
759 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
760 if (cache->mfc_un.res.minvif > vifi)
761 cache->mfc_un.res.minvif = vifi;
762 if (cache->mfc_un.res.maxvif <= vifi)
763 cache->mfc_un.res.maxvif = vifi + 1;
768 static int vif_add(struct net *net, struct mr_table *mrt,
769 struct vifctl *vifc, int mrtsock)
771 int vifi = vifc->vifc_vifi;
772 struct vif_device *v = &mrt->vif_table[vifi];
773 struct net_device *dev;
774 struct in_device *in_dev;
778 if (VIF_EXISTS(mrt, vifi))
781 switch (vifc->vifc_flags) {
783 if (!pimsm_enabled())
785 /* Special Purpose VIF in PIM
786 * All the packets will be sent to the daemon
788 if (mrt->mroute_reg_vif_num >= 0)
790 dev = ipmr_reg_vif(net, mrt);
793 err = dev_set_allmulti(dev, 1);
795 unregister_netdevice(dev);
801 dev = ipmr_new_tunnel(net, vifc);
804 err = dev_set_allmulti(dev, 1);
806 ipmr_del_tunnel(dev, vifc);
812 case VIFF_USE_IFINDEX:
814 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
815 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
816 if (dev && !__in_dev_get_rtnl(dev)) {
818 return -EADDRNOTAVAIL;
821 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
824 return -EADDRNOTAVAIL;
825 err = dev_set_allmulti(dev, 1);
835 in_dev = __in_dev_get_rtnl(dev);
838 return -EADDRNOTAVAIL;
840 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
841 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
843 ip_rt_multicast_event(in_dev);
845 /* Fill in the VIF structures */
847 v->rate_limit = vifc->vifc_rate_limit;
848 v->local = vifc->vifc_lcl_addr.s_addr;
849 v->remote = vifc->vifc_rmt_addr.s_addr;
850 v->flags = vifc->vifc_flags;
852 v->flags |= VIFF_STATIC;
853 v->threshold = vifc->vifc_threshold;
858 v->link = dev->ifindex;
859 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
860 v->link = dev_get_iflink(dev);
862 /* And finish update writing critical data */
863 write_lock_bh(&mrt_lock);
865 if (v->flags & VIFF_REGISTER)
866 mrt->mroute_reg_vif_num = vifi;
867 if (vifi+1 > mrt->maxvif)
868 mrt->maxvif = vifi+1;
869 write_unlock_bh(&mrt_lock);
873 /* called with rcu_read_lock() */
874 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
878 int line = MFC_HASH(mcastgrp, origin);
881 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
882 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
888 /* Look for a (*,*,oif) entry */
889 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
892 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
895 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
896 if (c->mfc_origin == htonl(INADDR_ANY) &&
897 c->mfc_mcastgrp == htonl(INADDR_ANY) &&
898 c->mfc_un.res.ttls[vifi] < 255)
904 /* Look for a (*,G) entry */
905 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
906 __be32 mcastgrp, int vifi)
908 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
909 struct mfc_cache *c, *proxy;
911 if (mcastgrp == htonl(INADDR_ANY))
914 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
915 if (c->mfc_origin == htonl(INADDR_ANY) &&
916 c->mfc_mcastgrp == mcastgrp) {
917 if (c->mfc_un.res.ttls[vifi] < 255)
920 /* It's ok if the vifi is part of the static tree */
921 proxy = ipmr_cache_find_any_parent(mrt,
923 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
928 return ipmr_cache_find_any_parent(mrt, vifi);
932 * Allocate a multicast cache entry
934 static struct mfc_cache *ipmr_cache_alloc(void)
936 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
939 c->mfc_un.res.minvif = MAXVIFS;
943 static struct mfc_cache *ipmr_cache_alloc_unres(void)
945 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
948 skb_queue_head_init(&c->mfc_un.unres.unresolved);
949 c->mfc_un.unres.expires = jiffies + 10*HZ;
955 * A cache entry has gone into a resolved state from queued
958 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
959 struct mfc_cache *uc, struct mfc_cache *c)
964 /* Play the pending entries through our router */
966 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
967 if (ip_hdr(skb)->version == 0) {
968 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
970 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
971 nlh->nlmsg_len = skb_tail_pointer(skb) -
974 nlh->nlmsg_type = NLMSG_ERROR;
975 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
976 skb_trim(skb, nlh->nlmsg_len);
978 e->error = -EMSGSIZE;
979 memset(&e->msg, 0, sizeof(e->msg));
982 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
984 ip_mr_forward(net, mrt, skb, c, 0);
989 /* Bounce a cache query up to mrouted. We could use netlink for this but mrouted
990 * expects the following bizarre scheme.
992 * Called under mrt_lock.
994 static int ipmr_cache_report(struct mr_table *mrt,
995 struct sk_buff *pkt, vifi_t vifi, int assert)
997 const int ihl = ip_hdrlen(pkt);
998 struct sock *mroute_sk;
999 struct igmphdr *igmp;
1000 struct igmpmsg *msg;
1001 struct sk_buff *skb;
1004 if (assert == IGMPMSG_WHOLEPKT)
1005 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1007 skb = alloc_skb(128, GFP_ATOMIC);
1012 if (assert == IGMPMSG_WHOLEPKT) {
1013 /* Ugly, but we have no choice with this interface.
1014 * Duplicate old header, fix ihl, length etc.
1015 * And all this only to mangle msg->im_msgtype and
1016 * to set msg->im_mbz to "mbz" :-)
1018 skb_push(skb, sizeof(struct iphdr));
1019 skb_reset_network_header(skb);
1020 skb_reset_transport_header(skb);
1021 msg = (struct igmpmsg *)skb_network_header(skb);
1022 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1023 msg->im_msgtype = IGMPMSG_WHOLEPKT;
1025 msg->im_vif = mrt->mroute_reg_vif_num;
1026 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1027 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1028 sizeof(struct iphdr));
1030 /* Copy the IP header */
1031 skb_set_network_header(skb, skb->len);
1033 skb_copy_to_linear_data(skb, pkt->data, ihl);
1034 /* Flag to the kernel this is a route add */
1035 ip_hdr(skb)->protocol = 0;
1036 msg = (struct igmpmsg *)skb_network_header(skb);
1038 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1039 /* Add our header */
1040 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
1041 igmp->type = assert;
1042 msg->im_msgtype = assert;
1044 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1045 skb->transport_header = skb->network_header;
1049 mroute_sk = rcu_dereference(mrt->mroute_sk);
1056 /* Deliver to mrouted */
1057 ret = sock_queue_rcv_skb(mroute_sk, skb);
1060 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1068 * Queue a packet for resolution. It gets locked cache entry!
1072 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1076 struct mfc_cache *c;
1077 const struct iphdr *iph = ip_hdr(skb);
1079 spin_lock_bh(&mfc_unres_lock);
1080 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1081 if (c->mfc_mcastgrp == iph->daddr &&
1082 c->mfc_origin == iph->saddr) {
1089 /* Create a new entry if allowable */
1091 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1092 (c = ipmr_cache_alloc_unres()) == NULL) {
1093 spin_unlock_bh(&mfc_unres_lock);
1099 /* Fill in the new cache entry */
1102 c->mfc_origin = iph->saddr;
1103 c->mfc_mcastgrp = iph->daddr;
1105 /* Reflect first query at mrouted. */
1107 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1109 /* If the report failed throw the cache entry
1112 spin_unlock_bh(&mfc_unres_lock);
1119 atomic_inc(&mrt->cache_resolve_queue_len);
1120 list_add(&c->list, &mrt->mfc_unres_queue);
1121 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1123 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1124 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1127 /* See if we can append the packet */
1129 if (c->mfc_un.unres.unresolved.qlen > 3) {
1133 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1137 spin_unlock_bh(&mfc_unres_lock);
1142 * MFC cache manipulation by user space mroute daemon
1145 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1148 struct mfc_cache *c, *next;
1150 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1152 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1153 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1154 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1155 (parent == -1 || parent == c->mfc_parent)) {
1156 list_del_rcu(&c->list);
1157 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1165 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1166 struct mfcctl *mfc, int mrtsock, int parent)
1170 struct mfc_cache *uc, *c;
1172 if (mfc->mfcc_parent >= MAXVIFS)
1175 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1177 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1178 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1179 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1180 (parent == -1 || parent == c->mfc_parent)) {
1187 write_lock_bh(&mrt_lock);
1188 c->mfc_parent = mfc->mfcc_parent;
1189 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1191 c->mfc_flags |= MFC_STATIC;
1192 write_unlock_bh(&mrt_lock);
1193 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1197 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1198 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1201 c = ipmr_cache_alloc();
1205 c->mfc_origin = mfc->mfcc_origin.s_addr;
1206 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1207 c->mfc_parent = mfc->mfcc_parent;
1208 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1210 c->mfc_flags |= MFC_STATIC;
1212 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1215 * Check to see if we resolved a queued list. If so we
1216 * need to send on the frames and tidy up.
1219 spin_lock_bh(&mfc_unres_lock);
1220 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1221 if (uc->mfc_origin == c->mfc_origin &&
1222 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1223 list_del(&uc->list);
1224 atomic_dec(&mrt->cache_resolve_queue_len);
1229 if (list_empty(&mrt->mfc_unres_queue))
1230 del_timer(&mrt->ipmr_expire_timer);
1231 spin_unlock_bh(&mfc_unres_lock);
1234 ipmr_cache_resolve(net, mrt, uc, c);
1235 ipmr_cache_free(uc);
1237 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1242 * Close the multicast socket, and clear the vif tables etc
1245 static void mroute_clean_tables(struct mr_table *mrt)
1249 struct mfc_cache *c, *next;
1251 /* Shut down all active vif entries */
1253 for (i = 0; i < mrt->maxvif; i++) {
1254 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1255 vif_delete(mrt, i, 0, &list);
1257 unregister_netdevice_many(&list);
1259 /* Wipe the cache */
1261 for (i = 0; i < MFC_LINES; i++) {
1262 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1263 if (c->mfc_flags & MFC_STATIC)
1265 list_del_rcu(&c->list);
1266 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1271 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1272 spin_lock_bh(&mfc_unres_lock);
1273 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1275 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1276 ipmr_destroy_unres(mrt, c);
1278 spin_unlock_bh(&mfc_unres_lock);
1282 /* called from ip_ra_control(), before an RCU grace period,
1283 * we dont need to call synchronize_rcu() here
1285 static void mrtsock_destruct(struct sock *sk)
1287 struct net *net = sock_net(sk);
1288 struct mr_table *mrt;
1291 ipmr_for_each_table(mrt, net) {
1292 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1293 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1294 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1295 NETCONFA_IFINDEX_ALL,
1296 net->ipv4.devconf_all);
1297 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1298 mroute_clean_tables(mrt);
1305 * Socket options and virtual interface manipulation. The whole
1306 * virtual interface system is a complete heap, but unfortunately
1307 * that's how BSD mrouted happens to think. Maybe one day with a proper
1308 * MOSPF/PIM router set up we can clean this up.
1311 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1313 int ret, parent = 0;
1316 struct net *net = sock_net(sk);
1317 struct mr_table *mrt;
1319 if (sk->sk_type != SOCK_RAW ||
1320 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1323 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1327 if (optname != MRT_INIT) {
1328 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1329 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1335 if (optlen != sizeof(int))
1339 if (rtnl_dereference(mrt->mroute_sk)) {
1344 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1346 rcu_assign_pointer(mrt->mroute_sk, sk);
1347 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1348 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1349 NETCONFA_IFINDEX_ALL,
1350 net->ipv4.devconf_all);
1355 if (sk != rcu_access_pointer(mrt->mroute_sk))
1357 return ip_ra_control(sk, 0, NULL);
1360 if (optlen != sizeof(vif))
1362 if (copy_from_user(&vif, optval, sizeof(vif)))
1364 if (vif.vifc_vifi >= MAXVIFS)
1367 if (optname == MRT_ADD_VIF) {
1368 ret = vif_add(net, mrt, &vif,
1369 sk == rtnl_dereference(mrt->mroute_sk));
1371 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1377 * Manipulate the forwarding caches. These live
1378 * in a sort of kernel/user symbiosis.
1383 case MRT_ADD_MFC_PROXY:
1384 case MRT_DEL_MFC_PROXY:
1385 if (optlen != sizeof(mfc))
1387 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1390 parent = mfc.mfcc_parent;
1392 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1393 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1395 ret = ipmr_mfc_add(net, mrt, &mfc,
1396 sk == rtnl_dereference(mrt->mroute_sk),
1401 * Control PIM assert.
1406 if (optlen != sizeof(v))
1408 if (get_user(v, (int __user *)optval))
1410 mrt->mroute_do_assert = v;
1417 if (!pimsm_enabled())
1418 return -ENOPROTOOPT;
1419 if (optlen != sizeof(v))
1421 if (get_user(v, (int __user *)optval))
1427 if (v != mrt->mroute_do_pim) {
1428 mrt->mroute_do_pim = v;
1429 mrt->mroute_do_assert = v;
1434 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1439 if (optlen != sizeof(u32))
1441 if (get_user(v, (u32 __user *)optval))
1446 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1449 mrt = ipmr_new_table(net, v);
1453 raw_sk(sk)->ipmr_table = v;
1460 * Spurious command, or MRT_VERSION which you cannot
1464 return -ENOPROTOOPT;
1469 * Getsock opt support for the multicast routing system.
1472 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1476 struct net *net = sock_net(sk);
1477 struct mr_table *mrt;
1479 if (sk->sk_type != SOCK_RAW ||
1480 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1483 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1487 if (optname != MRT_VERSION &&
1488 optname != MRT_PIM &&
1489 optname != MRT_ASSERT)
1490 return -ENOPROTOOPT;
1492 if (get_user(olr, optlen))
1495 olr = min_t(unsigned int, olr, sizeof(int));
1499 if (put_user(olr, optlen))
1501 if (optname == MRT_VERSION) {
1503 } else if (optname == MRT_PIM) {
1504 if (!pimsm_enabled())
1505 return -ENOPROTOOPT;
1506 val = mrt->mroute_do_pim;
1508 val = mrt->mroute_do_assert;
1510 if (copy_to_user(optval, &val, olr))
1516 * The IP multicast ioctl support routines.
1519 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1521 struct sioc_sg_req sr;
1522 struct sioc_vif_req vr;
1523 struct vif_device *vif;
1524 struct mfc_cache *c;
1525 struct net *net = sock_net(sk);
1526 struct mr_table *mrt;
1528 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1534 if (copy_from_user(&vr, arg, sizeof(vr)))
1536 if (vr.vifi >= mrt->maxvif)
1538 read_lock(&mrt_lock);
1539 vif = &mrt->vif_table[vr.vifi];
1540 if (VIF_EXISTS(mrt, vr.vifi)) {
1541 vr.icount = vif->pkt_in;
1542 vr.ocount = vif->pkt_out;
1543 vr.ibytes = vif->bytes_in;
1544 vr.obytes = vif->bytes_out;
1545 read_unlock(&mrt_lock);
1547 if (copy_to_user(arg, &vr, sizeof(vr)))
1551 read_unlock(&mrt_lock);
1552 return -EADDRNOTAVAIL;
1554 if (copy_from_user(&sr, arg, sizeof(sr)))
1558 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1560 sr.pktcnt = c->mfc_un.res.pkt;
1561 sr.bytecnt = c->mfc_un.res.bytes;
1562 sr.wrong_if = c->mfc_un.res.wrong_if;
1565 if (copy_to_user(arg, &sr, sizeof(sr)))
1570 return -EADDRNOTAVAIL;
1572 return -ENOIOCTLCMD;
1576 #ifdef CONFIG_COMPAT
1577 struct compat_sioc_sg_req {
1580 compat_ulong_t pktcnt;
1581 compat_ulong_t bytecnt;
1582 compat_ulong_t wrong_if;
1585 struct compat_sioc_vif_req {
1586 vifi_t vifi; /* Which iface */
1587 compat_ulong_t icount;
1588 compat_ulong_t ocount;
1589 compat_ulong_t ibytes;
1590 compat_ulong_t obytes;
1593 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1595 struct compat_sioc_sg_req sr;
1596 struct compat_sioc_vif_req vr;
1597 struct vif_device *vif;
1598 struct mfc_cache *c;
1599 struct net *net = sock_net(sk);
1600 struct mr_table *mrt;
1602 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1608 if (copy_from_user(&vr, arg, sizeof(vr)))
1610 if (vr.vifi >= mrt->maxvif)
1612 read_lock(&mrt_lock);
1613 vif = &mrt->vif_table[vr.vifi];
1614 if (VIF_EXISTS(mrt, vr.vifi)) {
1615 vr.icount = vif->pkt_in;
1616 vr.ocount = vif->pkt_out;
1617 vr.ibytes = vif->bytes_in;
1618 vr.obytes = vif->bytes_out;
1619 read_unlock(&mrt_lock);
1621 if (copy_to_user(arg, &vr, sizeof(vr)))
1625 read_unlock(&mrt_lock);
1626 return -EADDRNOTAVAIL;
1628 if (copy_from_user(&sr, arg, sizeof(sr)))
1632 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1634 sr.pktcnt = c->mfc_un.res.pkt;
1635 sr.bytecnt = c->mfc_un.res.bytes;
1636 sr.wrong_if = c->mfc_un.res.wrong_if;
1639 if (copy_to_user(arg, &sr, sizeof(sr)))
1644 return -EADDRNOTAVAIL;
1646 return -ENOIOCTLCMD;
1652 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1654 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1655 struct net *net = dev_net(dev);
1656 struct mr_table *mrt;
1657 struct vif_device *v;
1660 if (event != NETDEV_UNREGISTER)
1663 ipmr_for_each_table(mrt, net) {
1664 v = &mrt->vif_table[0];
1665 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1667 vif_delete(mrt, ct, 1, NULL);
1674 static struct notifier_block ip_mr_notifier = {
1675 .notifier_call = ipmr_device_event,
1679 * Encapsulate a packet by attaching a valid IPIP header to it.
1680 * This avoids tunnel drivers and other mess and gives us the speed so
1681 * important for multicast video.
1684 static void ip_encap(struct net *net, struct sk_buff *skb,
1685 __be32 saddr, __be32 daddr)
1688 const struct iphdr *old_iph = ip_hdr(skb);
1690 skb_push(skb, sizeof(struct iphdr));
1691 skb->transport_header = skb->network_header;
1692 skb_reset_network_header(skb);
1696 iph->tos = old_iph->tos;
1697 iph->ttl = old_iph->ttl;
1701 iph->protocol = IPPROTO_IPIP;
1703 iph->tot_len = htons(skb->len);
1704 ip_select_ident(net, skb, NULL);
1707 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1711 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1712 struct sk_buff *skb)
1714 struct ip_options *opt = &(IPCB(skb)->opt);
1716 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1717 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1719 if (unlikely(opt->optlen))
1720 ip_forward_options(skb);
1722 return dst_output(net, sk, skb);
1726 * Processing handlers for ipmr_forward
1729 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1730 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1732 const struct iphdr *iph = ip_hdr(skb);
1733 struct vif_device *vif = &mrt->vif_table[vifi];
1734 struct net_device *dev;
1742 if (vif->flags & VIFF_REGISTER) {
1744 vif->bytes_out += skb->len;
1745 vif->dev->stats.tx_bytes += skb->len;
1746 vif->dev->stats.tx_packets++;
1747 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1751 if (vif->flags & VIFF_TUNNEL) {
1752 rt = ip_route_output_ports(net, &fl4, NULL,
1753 vif->remote, vif->local,
1756 RT_TOS(iph->tos), vif->link);
1759 encap = sizeof(struct iphdr);
1761 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1764 RT_TOS(iph->tos), vif->link);
1771 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1772 /* Do not fragment multicasts. Alas, IPv4 does not
1773 * allow to send ICMP, so that packets will disappear
1777 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1782 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1784 if (skb_cow(skb, encap)) {
1790 vif->bytes_out += skb->len;
1793 skb_dst_set(skb, &rt->dst);
1794 ip_decrease_ttl(ip_hdr(skb));
1796 /* FIXME: forward and output firewalls used to be called here.
1797 * What do we do with netfilter? -- RR
1799 if (vif->flags & VIFF_TUNNEL) {
1800 ip_encap(net, skb, vif->local, vif->remote);
1801 /* FIXME: extra output firewall step used to be here. --RR */
1802 vif->dev->stats.tx_packets++;
1803 vif->dev->stats.tx_bytes += skb->len;
1806 IPCB(skb)->flags |= IPSKB_FORWARDED;
1809 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1810 * not only before forwarding, but after forwarding on all output
1811 * interfaces. It is clear, if mrouter runs a multicasting
1812 * program, it should receive packets not depending to what interface
1813 * program is joined.
1814 * If we will not make it, the program will have to join on all
1815 * interfaces. On the other hand, multihoming host (or router, but
1816 * not mrouter) cannot join to more than one interface - it will
1817 * result in receiving multiple packets.
1819 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1820 net, NULL, skb, skb->dev, dev,
1821 ipmr_forward_finish);
1828 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1832 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1833 if (mrt->vif_table[ct].dev == dev)
1839 /* "local" means that we should preserve one skb (for local delivery) */
1841 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1842 struct sk_buff *skb, struct mfc_cache *cache,
1847 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1849 vif = cache->mfc_parent;
1850 cache->mfc_un.res.pkt++;
1851 cache->mfc_un.res.bytes += skb->len;
1853 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1854 struct mfc_cache *cache_proxy;
1856 /* For an (*,G) entry, we only check that the incomming
1857 * interface is part of the static tree.
1859 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1861 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1866 * Wrong interface: drop packet and (maybe) send PIM assert.
1868 if (mrt->vif_table[vif].dev != skb->dev) {
1869 if (rt_is_output_route(skb_rtable(skb))) {
1870 /* It is our own packet, looped back.
1871 * Very complicated situation...
1873 * The best workaround until routing daemons will be
1874 * fixed is not to redistribute packet, if it was
1875 * send through wrong interface. It means, that
1876 * multicast applications WILL NOT work for
1877 * (S,G), which have default multicast route pointing
1878 * to wrong oif. In any case, it is not a good
1879 * idea to use multicasting applications on router.
1884 cache->mfc_un.res.wrong_if++;
1886 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1887 /* pimsm uses asserts, when switching from RPT to SPT,
1888 * so that we cannot check that packet arrived on an oif.
1889 * It is bad, but otherwise we would need to move pretty
1890 * large chunk of pimd to kernel. Ough... --ANK
1892 (mrt->mroute_do_pim ||
1893 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1895 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1896 cache->mfc_un.res.last_assert = jiffies;
1897 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1903 mrt->vif_table[vif].pkt_in++;
1904 mrt->vif_table[vif].bytes_in += skb->len;
1909 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1910 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1911 if (true_vifi >= 0 &&
1912 true_vifi != cache->mfc_parent &&
1914 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1915 /* It's an (*,*) entry and the packet is not coming from
1916 * the upstream: forward the packet to the upstream
1919 psend = cache->mfc_parent;
1924 for (ct = cache->mfc_un.res.maxvif - 1;
1925 ct >= cache->mfc_un.res.minvif; ct--) {
1926 /* For (*,G) entry, don't forward to the incoming interface */
1927 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1929 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1931 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1934 ipmr_queue_xmit(net, mrt, skb2, cache,
1943 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1946 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1948 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1958 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1960 struct rtable *rt = skb_rtable(skb);
1961 struct iphdr *iph = ip_hdr(skb);
1962 struct flowi4 fl4 = {
1963 .daddr = iph->daddr,
1964 .saddr = iph->saddr,
1965 .flowi4_tos = RT_TOS(iph->tos),
1966 .flowi4_oif = (rt_is_output_route(rt) ?
1967 skb->dev->ifindex : 0),
1968 .flowi4_iif = (rt_is_output_route(rt) ?
1971 .flowi4_mark = skb->mark,
1973 struct mr_table *mrt;
1976 err = ipmr_fib_lookup(net, &fl4, &mrt);
1978 return ERR_PTR(err);
1983 * Multicast packets for forwarding arrive here
1984 * Called with rcu_read_lock();
1987 int ip_mr_input(struct sk_buff *skb)
1989 struct mfc_cache *cache;
1990 struct net *net = dev_net(skb->dev);
1991 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1992 struct mr_table *mrt;
1994 /* Packet is looped back after forward, it should not be
1995 * forwarded second time, but still can be delivered locally.
1997 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2000 mrt = ipmr_rt_fib_lookup(net, skb);
2003 return PTR_ERR(mrt);
2006 if (IPCB(skb)->opt.router_alert) {
2007 if (ip_call_ra_chain(skb))
2009 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2010 /* IGMPv1 (and broken IGMPv2 implementations sort of
2011 * Cisco IOS <= 11.2(8)) do not put router alert
2012 * option to IGMP packets destined to routable
2013 * groups. It is very bad, because it means
2014 * that we can forward NO IGMP messages.
2016 struct sock *mroute_sk;
2018 mroute_sk = rcu_dereference(mrt->mroute_sk);
2021 raw_rcv(mroute_sk, skb);
2027 /* already under rcu_read_lock() */
2028 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2030 int vif = ipmr_find_vif(mrt, skb->dev);
2033 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2038 * No usable cache entry
2044 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2045 ip_local_deliver(skb);
2051 read_lock(&mrt_lock);
2052 vif = ipmr_find_vif(mrt, skb->dev);
2054 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2055 read_unlock(&mrt_lock);
2059 read_unlock(&mrt_lock);
2064 read_lock(&mrt_lock);
2065 ip_mr_forward(net, mrt, skb, cache, local);
2066 read_unlock(&mrt_lock);
2069 return ip_local_deliver(skb);
2075 return ip_local_deliver(skb);
2080 #ifdef CONFIG_IP_PIMSM_V1
2082 * Handle IGMP messages of PIMv1
2085 int pim_rcv_v1(struct sk_buff *skb)
2087 struct igmphdr *pim;
2088 struct net *net = dev_net(skb->dev);
2089 struct mr_table *mrt;
2091 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2094 pim = igmp_hdr(skb);
2096 mrt = ipmr_rt_fib_lookup(net, skb);
2099 if (!mrt->mroute_do_pim ||
2100 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2103 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2111 #ifdef CONFIG_IP_PIMSM_V2
2112 static int pim_rcv(struct sk_buff *skb)
2114 struct pimreghdr *pim;
2115 struct net *net = dev_net(skb->dev);
2116 struct mr_table *mrt;
2118 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2121 pim = (struct pimreghdr *)skb_transport_header(skb);
2122 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2123 (pim->flags & PIM_NULL_REGISTER) ||
2124 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2125 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2128 mrt = ipmr_rt_fib_lookup(net, skb);
2131 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2139 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2140 struct mfc_cache *c, struct rtmsg *rtm)
2143 struct rtnexthop *nhp;
2144 struct nlattr *mp_attr;
2145 struct rta_mfc_stats mfcs;
2147 /* If cache is unresolved, don't try to parse IIF and OIF */
2148 if (c->mfc_parent >= MAXVIFS)
2151 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2152 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2155 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2158 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2159 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2160 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2161 nla_nest_cancel(skb, mp_attr);
2165 nhp->rtnh_flags = 0;
2166 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2167 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2168 nhp->rtnh_len = sizeof(*nhp);
2172 nla_nest_end(skb, mp_attr);
2174 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2175 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2176 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2177 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2180 rtm->rtm_type = RTN_MULTICAST;
2184 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2185 __be32 saddr, __be32 daddr,
2186 struct rtmsg *rtm, int nowait)
2188 struct mfc_cache *cache;
2189 struct mr_table *mrt;
2192 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2197 cache = ipmr_cache_find(mrt, saddr, daddr);
2198 if (!cache && skb->dev) {
2199 int vif = ipmr_find_vif(mrt, skb->dev);
2202 cache = ipmr_cache_find_any(mrt, daddr, vif);
2205 struct sk_buff *skb2;
2207 struct net_device *dev;
2216 read_lock(&mrt_lock);
2218 vif = ipmr_find_vif(mrt, dev);
2220 read_unlock(&mrt_lock);
2224 skb2 = skb_clone(skb, GFP_ATOMIC);
2226 read_unlock(&mrt_lock);
2231 skb_push(skb2, sizeof(struct iphdr));
2232 skb_reset_network_header(skb2);
2234 iph->ihl = sizeof(struct iphdr) >> 2;
2238 err = ipmr_cache_unresolved(mrt, vif, skb2);
2239 read_unlock(&mrt_lock);
2244 read_lock(&mrt_lock);
2245 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2246 cache->mfc_flags |= MFC_NOTIFY;
2247 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2248 read_unlock(&mrt_lock);
2253 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2254 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2257 struct nlmsghdr *nlh;
2261 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2265 rtm = nlmsg_data(nlh);
2266 rtm->rtm_family = RTNL_FAMILY_IPMR;
2267 rtm->rtm_dst_len = 32;
2268 rtm->rtm_src_len = 32;
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;
2278 rtm->rtm_protocol = RTPROT_MROUTED;
2281 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2282 nla_put_in_addr(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;
2289 nlmsg_end(skb, nlh);
2293 nlmsg_cancel(skb, nlh);
2297 static size_t mroute_msgsize(bool unresolved, int maxvif)
2300 NLMSG_ALIGN(sizeof(struct rtmsg))
2301 + nla_total_size(4) /* RTA_TABLE */
2302 + nla_total_size(4) /* RTA_SRC */
2303 + nla_total_size(4) /* RTA_DST */
2308 + nla_total_size(4) /* RTA_IIF */
2309 + nla_total_size(0) /* RTA_MULTIPATH */
2310 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2312 + nla_total_size(sizeof(struct rta_mfc_stats))
2318 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2321 struct net *net = read_pnet(&mrt->net);
2322 struct sk_buff *skb;
2325 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2330 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2334 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2340 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2343 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2345 struct net *net = sock_net(skb->sk);
2346 struct mr_table *mrt;
2347 struct mfc_cache *mfc;
2348 unsigned int t = 0, s_t;
2349 unsigned int h = 0, s_h;
2350 unsigned int e = 0, s_e;
2357 ipmr_for_each_table(mrt, net) {
2362 for (h = s_h; h < MFC_LINES; h++) {
2363 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2366 if (ipmr_fill_mroute(mrt, skb,
2367 NETLINK_CB(cb->skb).portid,
2377 spin_lock_bh(&mfc_unres_lock);
2378 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2381 if (ipmr_fill_mroute(mrt, skb,
2382 NETLINK_CB(cb->skb).portid,
2386 spin_unlock_bh(&mfc_unres_lock);
2392 spin_unlock_bh(&mfc_unres_lock);
2408 #ifdef CONFIG_PROC_FS
2410 * The /proc interfaces to multicast routing :
2411 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2413 struct ipmr_vif_iter {
2414 struct seq_net_private p;
2415 struct mr_table *mrt;
2419 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2420 struct ipmr_vif_iter *iter,
2423 struct mr_table *mrt = iter->mrt;
2425 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2426 if (!VIF_EXISTS(mrt, iter->ct))
2429 return &mrt->vif_table[iter->ct];
2434 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2435 __acquires(mrt_lock)
2437 struct ipmr_vif_iter *iter = seq->private;
2438 struct net *net = seq_file_net(seq);
2439 struct mr_table *mrt;
2441 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2443 return ERR_PTR(-ENOENT);
2447 read_lock(&mrt_lock);
2448 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2452 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2454 struct ipmr_vif_iter *iter = seq->private;
2455 struct net *net = seq_file_net(seq);
2456 struct mr_table *mrt = iter->mrt;
2459 if (v == SEQ_START_TOKEN)
2460 return ipmr_vif_seq_idx(net, iter, 0);
2462 while (++iter->ct < mrt->maxvif) {
2463 if (!VIF_EXISTS(mrt, iter->ct))
2465 return &mrt->vif_table[iter->ct];
2470 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2471 __releases(mrt_lock)
2473 read_unlock(&mrt_lock);
2476 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2478 struct ipmr_vif_iter *iter = seq->private;
2479 struct mr_table *mrt = iter->mrt;
2481 if (v == SEQ_START_TOKEN) {
2483 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2485 const struct vif_device *vif = v;
2486 const char *name = vif->dev ? vif->dev->name : "none";
2489 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2490 vif - mrt->vif_table,
2491 name, vif->bytes_in, vif->pkt_in,
2492 vif->bytes_out, vif->pkt_out,
2493 vif->flags, vif->local, vif->remote);
2498 static const struct seq_operations ipmr_vif_seq_ops = {
2499 .start = ipmr_vif_seq_start,
2500 .next = ipmr_vif_seq_next,
2501 .stop = ipmr_vif_seq_stop,
2502 .show = ipmr_vif_seq_show,
2505 static int ipmr_vif_open(struct inode *inode, struct file *file)
2507 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2508 sizeof(struct ipmr_vif_iter));
2511 static const struct file_operations ipmr_vif_fops = {
2512 .owner = THIS_MODULE,
2513 .open = ipmr_vif_open,
2515 .llseek = seq_lseek,
2516 .release = seq_release_net,
2519 struct ipmr_mfc_iter {
2520 struct seq_net_private p;
2521 struct mr_table *mrt;
2522 struct list_head *cache;
2527 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2528 struct ipmr_mfc_iter *it, loff_t pos)
2530 struct mr_table *mrt = it->mrt;
2531 struct mfc_cache *mfc;
2534 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2535 it->cache = &mrt->mfc_cache_array[it->ct];
2536 list_for_each_entry_rcu(mfc, it->cache, list)
2542 spin_lock_bh(&mfc_unres_lock);
2543 it->cache = &mrt->mfc_unres_queue;
2544 list_for_each_entry(mfc, it->cache, list)
2547 spin_unlock_bh(&mfc_unres_lock);
2554 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2556 struct ipmr_mfc_iter *it = seq->private;
2557 struct net *net = seq_file_net(seq);
2558 struct mr_table *mrt;
2560 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2562 return ERR_PTR(-ENOENT);
2567 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2571 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2573 struct mfc_cache *mfc = v;
2574 struct ipmr_mfc_iter *it = seq->private;
2575 struct net *net = seq_file_net(seq);
2576 struct mr_table *mrt = it->mrt;
2580 if (v == SEQ_START_TOKEN)
2581 return ipmr_mfc_seq_idx(net, seq->private, 0);
2583 if (mfc->list.next != it->cache)
2584 return list_entry(mfc->list.next, struct mfc_cache, list);
2586 if (it->cache == &mrt->mfc_unres_queue)
2589 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2591 while (++it->ct < MFC_LINES) {
2592 it->cache = &mrt->mfc_cache_array[it->ct];
2593 if (list_empty(it->cache))
2595 return list_first_entry(it->cache, struct mfc_cache, list);
2598 /* exhausted cache_array, show unresolved */
2600 it->cache = &mrt->mfc_unres_queue;
2603 spin_lock_bh(&mfc_unres_lock);
2604 if (!list_empty(it->cache))
2605 return list_first_entry(it->cache, struct mfc_cache, list);
2608 spin_unlock_bh(&mfc_unres_lock);
2614 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2616 struct ipmr_mfc_iter *it = seq->private;
2617 struct mr_table *mrt = it->mrt;
2619 if (it->cache == &mrt->mfc_unres_queue)
2620 spin_unlock_bh(&mfc_unres_lock);
2621 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2625 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2629 if (v == SEQ_START_TOKEN) {
2631 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2633 const struct mfc_cache *mfc = v;
2634 const struct ipmr_mfc_iter *it = seq->private;
2635 const struct mr_table *mrt = it->mrt;
2637 seq_printf(seq, "%08X %08X %-3hd",
2638 (__force u32) mfc->mfc_mcastgrp,
2639 (__force u32) mfc->mfc_origin,
2642 if (it->cache != &mrt->mfc_unres_queue) {
2643 seq_printf(seq, " %8lu %8lu %8lu",
2644 mfc->mfc_un.res.pkt,
2645 mfc->mfc_un.res.bytes,
2646 mfc->mfc_un.res.wrong_if);
2647 for (n = mfc->mfc_un.res.minvif;
2648 n < mfc->mfc_un.res.maxvif; n++) {
2649 if (VIF_EXISTS(mrt, n) &&
2650 mfc->mfc_un.res.ttls[n] < 255)
2653 n, mfc->mfc_un.res.ttls[n]);
2656 /* unresolved mfc_caches don't contain
2657 * pkt, bytes and wrong_if values
2659 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2661 seq_putc(seq, '\n');
2666 static const struct seq_operations ipmr_mfc_seq_ops = {
2667 .start = ipmr_mfc_seq_start,
2668 .next = ipmr_mfc_seq_next,
2669 .stop = ipmr_mfc_seq_stop,
2670 .show = ipmr_mfc_seq_show,
2673 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2675 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2676 sizeof(struct ipmr_mfc_iter));
2679 static const struct file_operations ipmr_mfc_fops = {
2680 .owner = THIS_MODULE,
2681 .open = ipmr_mfc_open,
2683 .llseek = seq_lseek,
2684 .release = seq_release_net,
2688 #ifdef CONFIG_IP_PIMSM_V2
2689 static const struct net_protocol pim_protocol = {
2697 * Setup for IP multicast routing
2699 static int __net_init ipmr_net_init(struct net *net)
2703 err = ipmr_rules_init(net);
2707 #ifdef CONFIG_PROC_FS
2709 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2711 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2712 goto proc_cache_fail;
2716 #ifdef CONFIG_PROC_FS
2718 remove_proc_entry("ip_mr_vif", net->proc_net);
2720 ipmr_rules_exit(net);
2726 static void __net_exit ipmr_net_exit(struct net *net)
2728 #ifdef CONFIG_PROC_FS
2729 remove_proc_entry("ip_mr_cache", net->proc_net);
2730 remove_proc_entry("ip_mr_vif", net->proc_net);
2732 ipmr_rules_exit(net);
2735 static struct pernet_operations ipmr_net_ops = {
2736 .init = ipmr_net_init,
2737 .exit = ipmr_net_exit,
2740 int __init ip_mr_init(void)
2744 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2745 sizeof(struct mfc_cache),
2746 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2751 err = register_pernet_subsys(&ipmr_net_ops);
2753 goto reg_pernet_fail;
2755 err = register_netdevice_notifier(&ip_mr_notifier);
2757 goto reg_notif_fail;
2758 #ifdef CONFIG_IP_PIMSM_V2
2759 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2760 pr_err("%s: can't add PIM protocol\n", __func__);
2762 goto add_proto_fail;
2765 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2766 NULL, ipmr_rtm_dumproute, NULL);
2769 #ifdef CONFIG_IP_PIMSM_V2
2771 unregister_netdevice_notifier(&ip_mr_notifier);
2774 unregister_pernet_subsys(&ipmr_net_ops);
2776 kmem_cache_destroy(mrt_cachep);