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>
69 #include <linux/nospec.h>
71 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
72 #define CONFIG_IP_PIMSM 1
76 struct list_head list;
79 struct sock __rcu *mroute_sk;
80 struct timer_list ipmr_expire_timer;
81 struct list_head mfc_unres_queue;
82 struct list_head mfc_cache_array[MFC_LINES];
83 struct vif_device vif_table[MAXVIFS];
85 atomic_t cache_resolve_queue_len;
86 bool mroute_do_assert;
88 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
89 int mroute_reg_vif_num;
94 struct fib_rule common;
101 /* Big lock, protecting vif table, mrt cache and mroute socket state.
102 * Note that the changes are semaphored via rtnl_lock.
105 static DEFINE_RWLOCK(mrt_lock);
108 * Multicast router control variables
111 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
113 /* Special spinlock for queue of unresolved entries */
114 static DEFINE_SPINLOCK(mfc_unres_lock);
116 /* We return to original Alan's scheme. Hash table of resolved
117 * entries is changed only in process context and protected
118 * with weak lock mrt_lock. Queue of unresolved entries is protected
119 * with strong spinlock mfc_unres_lock.
121 * In this case data path is free of exclusive locks at all.
124 static struct kmem_cache *mrt_cachep __read_mostly;
126 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
127 static void ipmr_free_table(struct mr_table *mrt);
129 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
130 struct sk_buff *skb, struct mfc_cache *cache,
132 static int ipmr_cache_report(struct mr_table *mrt,
133 struct sk_buff *pkt, vifi_t vifi, int assert);
134 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
135 struct mfc_cache *c, struct rtmsg *rtm);
136 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
138 static void mroute_clean_tables(struct mr_table *mrt, bool all);
139 static void ipmr_expire_process(unsigned long arg);
141 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
142 #define ipmr_for_each_table(mrt, net) \
143 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
145 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
147 struct mr_table *mrt;
149 ipmr_for_each_table(mrt, net) {
156 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
157 struct mr_table **mrt)
160 struct ipmr_result res;
161 struct fib_lookup_arg arg = {
163 .flags = FIB_LOOKUP_NOREF,
166 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
167 flowi4_to_flowi(flp4), 0, &arg);
174 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
175 int flags, struct fib_lookup_arg *arg)
177 struct ipmr_result *res = arg->result;
178 struct mr_table *mrt;
180 switch (rule->action) {
183 case FR_ACT_UNREACHABLE:
185 case FR_ACT_PROHIBIT:
187 case FR_ACT_BLACKHOLE:
192 mrt = ipmr_get_table(rule->fr_net, rule->table);
199 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
204 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
208 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
209 struct fib_rule_hdr *frh, struct nlattr **tb)
214 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
220 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
221 struct fib_rule_hdr *frh)
229 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
230 .family = RTNL_FAMILY_IPMR,
231 .rule_size = sizeof(struct ipmr_rule),
232 .addr_size = sizeof(u32),
233 .action = ipmr_rule_action,
234 .match = ipmr_rule_match,
235 .configure = ipmr_rule_configure,
236 .compare = ipmr_rule_compare,
237 .fill = ipmr_rule_fill,
238 .nlgroup = RTNLGRP_IPV4_RULE,
239 .policy = ipmr_rule_policy,
240 .owner = THIS_MODULE,
243 static int __net_init ipmr_rules_init(struct net *net)
245 struct fib_rules_ops *ops;
246 struct mr_table *mrt;
249 ops = fib_rules_register(&ipmr_rules_ops_template, net);
253 INIT_LIST_HEAD(&net->ipv4.mr_tables);
255 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
261 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
265 net->ipv4.mr_rules_ops = ops;
269 ipmr_free_table(mrt);
271 fib_rules_unregister(ops);
275 static void __net_exit ipmr_rules_exit(struct net *net)
277 struct mr_table *mrt, *next;
280 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
281 list_del(&mrt->list);
282 ipmr_free_table(mrt);
284 fib_rules_unregister(net->ipv4.mr_rules_ops);
288 #define ipmr_for_each_table(mrt, net) \
289 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
291 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
293 return net->ipv4.mrt;
296 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
297 struct mr_table **mrt)
299 *mrt = net->ipv4.mrt;
303 static int __net_init ipmr_rules_init(struct net *net)
305 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
306 return net->ipv4.mrt ? 0 : -ENOMEM;
309 static void __net_exit ipmr_rules_exit(struct net *net)
312 ipmr_free_table(net->ipv4.mrt);
313 net->ipv4.mrt = NULL;
318 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
320 struct mr_table *mrt;
323 mrt = ipmr_get_table(net, id);
327 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
330 write_pnet(&mrt->net, net);
333 /* Forwarding cache */
334 for (i = 0; i < MFC_LINES; i++)
335 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
337 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
339 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
342 #ifdef CONFIG_IP_PIMSM
343 mrt->mroute_reg_vif_num = -1;
345 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
346 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
351 static void ipmr_free_table(struct mr_table *mrt)
353 del_timer_sync(&mrt->ipmr_expire_timer);
354 mroute_clean_tables(mrt, true);
358 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
360 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
362 struct net *net = dev_net(dev);
366 dev = __dev_get_by_name(net, "tunl0");
368 const struct net_device_ops *ops = dev->netdev_ops;
370 struct ip_tunnel_parm p;
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;
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;
381 if (ops->ndo_do_ioctl) {
382 mm_segment_t oldfs = get_fs();
385 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
392 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
394 struct net_device *dev;
396 dev = __dev_get_by_name(net, "tunl0");
399 const struct net_device_ops *ops = dev->netdev_ops;
402 struct ip_tunnel_parm p;
403 struct in_device *in_dev;
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;
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;
414 if (ops->ndo_do_ioctl) {
415 mm_segment_t oldfs = get_fs();
418 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
426 (dev = __dev_get_by_name(net, p.name)) != NULL) {
427 dev->flags |= IFF_MULTICAST;
429 in_dev = __in_dev_get_rtnl(dev);
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;
445 unregister_netdevice(dev);
449 #ifdef CONFIG_IP_PIMSM
451 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
453 struct net *net = dev_net(dev);
454 struct mr_table *mrt;
455 struct flowi4 fl4 = {
456 .flowi4_oif = dev->ifindex,
457 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
458 .flowi4_mark = skb->mark,
462 err = ipmr_fib_lookup(net, &fl4, &mrt);
468 read_lock(&mrt_lock);
469 dev->stats.tx_bytes += skb->len;
470 dev->stats.tx_packets++;
471 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
472 read_unlock(&mrt_lock);
477 static int reg_vif_get_iflink(const struct net_device *dev)
482 static const struct net_device_ops reg_vif_netdev_ops = {
483 .ndo_start_xmit = reg_vif_xmit,
484 .ndo_get_iflink = reg_vif_get_iflink,
487 static void reg_vif_setup(struct net_device *dev)
489 dev->type = ARPHRD_PIMREG;
490 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
491 dev->flags = IFF_NOARP;
492 dev->netdev_ops = ®_vif_netdev_ops;
493 dev->destructor = free_netdev;
494 dev->features |= NETIF_F_NETNS_LOCAL;
497 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
499 struct net_device *dev;
500 struct in_device *in_dev;
503 if (mrt->id == RT_TABLE_DEFAULT)
504 sprintf(name, "pimreg");
506 sprintf(name, "pimreg%u", mrt->id);
508 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
513 dev_net_set(dev, net);
515 if (register_netdevice(dev)) {
521 in_dev = __in_dev_get_rcu(dev);
527 ipv4_devconf_setall(in_dev);
528 neigh_parms_data_state_setall(in_dev->arp_parms);
529 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
540 unregister_netdevice(dev);
546 * vif_delete - Delete a VIF entry
547 * @notify: Set to 1, if the caller is a notifier_call
550 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
551 struct list_head *head)
553 struct vif_device *v;
554 struct net_device *dev;
555 struct in_device *in_dev;
557 if (vifi < 0 || vifi >= mrt->maxvif)
558 return -EADDRNOTAVAIL;
560 v = &mrt->vif_table[vifi];
562 write_lock_bh(&mrt_lock);
567 write_unlock_bh(&mrt_lock);
568 return -EADDRNOTAVAIL;
571 #ifdef CONFIG_IP_PIMSM
572 if (vifi == mrt->mroute_reg_vif_num)
573 mrt->mroute_reg_vif_num = -1;
576 if (vifi + 1 == mrt->maxvif) {
579 for (tmp = vifi - 1; tmp >= 0; tmp--) {
580 if (VIF_EXISTS(mrt, tmp))
586 write_unlock_bh(&mrt_lock);
588 dev_set_allmulti(dev, -1);
590 in_dev = __in_dev_get_rtnl(dev);
592 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
593 inet_netconf_notify_devconf(dev_net(dev),
594 NETCONFA_MC_FORWARDING,
595 dev->ifindex, &in_dev->cnf);
596 ip_rt_multicast_event(in_dev);
599 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
600 unregister_netdevice_queue(dev, head);
606 static void ipmr_cache_free_rcu(struct rcu_head *head)
608 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
610 kmem_cache_free(mrt_cachep, c);
613 static inline void ipmr_cache_free(struct mfc_cache *c)
615 call_rcu(&c->rcu, ipmr_cache_free_rcu);
618 /* Destroy an unresolved cache entry, killing queued skbs
619 * and reporting error to netlink readers.
622 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
624 struct net *net = read_pnet(&mrt->net);
628 atomic_dec(&mrt->cache_resolve_queue_len);
630 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
631 if (ip_hdr(skb)->version == 0) {
632 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
633 nlh->nlmsg_type = NLMSG_ERROR;
634 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
635 skb_trim(skb, nlh->nlmsg_len);
637 e->error = -ETIMEDOUT;
638 memset(&e->msg, 0, sizeof(e->msg));
640 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
650 /* Timer process for the unresolved queue. */
652 static void ipmr_expire_process(unsigned long arg)
654 struct mr_table *mrt = (struct mr_table *)arg;
656 unsigned long expires;
657 struct mfc_cache *c, *next;
659 if (!spin_trylock(&mfc_unres_lock)) {
660 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
664 if (list_empty(&mrt->mfc_unres_queue))
670 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
671 if (time_after(c->mfc_un.unres.expires, now)) {
672 unsigned long interval = c->mfc_un.unres.expires - now;
673 if (interval < expires)
679 mroute_netlink_event(mrt, c, RTM_DELROUTE);
680 ipmr_destroy_unres(mrt, c);
683 if (!list_empty(&mrt->mfc_unres_queue))
684 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
687 spin_unlock(&mfc_unres_lock);
690 /* Fill oifs list. It is called under write locked mrt_lock. */
692 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
697 cache->mfc_un.res.minvif = MAXVIFS;
698 cache->mfc_un.res.maxvif = 0;
699 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
701 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
702 if (VIF_EXISTS(mrt, vifi) &&
703 ttls[vifi] && ttls[vifi] < 255) {
704 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
705 if (cache->mfc_un.res.minvif > vifi)
706 cache->mfc_un.res.minvif = vifi;
707 if (cache->mfc_un.res.maxvif <= vifi)
708 cache->mfc_un.res.maxvif = vifi + 1;
713 static int vif_add(struct net *net, struct mr_table *mrt,
714 struct vifctl *vifc, int mrtsock)
716 int vifi = vifc->vifc_vifi;
717 struct vif_device *v = &mrt->vif_table[vifi];
718 struct net_device *dev;
719 struct in_device *in_dev;
723 if (VIF_EXISTS(mrt, vifi))
726 switch (vifc->vifc_flags) {
727 #ifdef CONFIG_IP_PIMSM
730 * Special Purpose VIF in PIM
731 * All the packets will be sent to the daemon
733 if (mrt->mroute_reg_vif_num >= 0)
735 dev = ipmr_reg_vif(net, mrt);
738 err = dev_set_allmulti(dev, 1);
740 unregister_netdevice(dev);
747 dev = ipmr_new_tunnel(net, vifc);
750 err = dev_set_allmulti(dev, 1);
752 ipmr_del_tunnel(dev, vifc);
758 case VIFF_USE_IFINDEX:
760 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
761 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
762 if (dev && !__in_dev_get_rtnl(dev)) {
764 return -EADDRNOTAVAIL;
767 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
770 return -EADDRNOTAVAIL;
771 err = dev_set_allmulti(dev, 1);
781 in_dev = __in_dev_get_rtnl(dev);
784 return -EADDRNOTAVAIL;
786 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
787 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
789 ip_rt_multicast_event(in_dev);
791 /* Fill in the VIF structures */
793 v->rate_limit = vifc->vifc_rate_limit;
794 v->local = vifc->vifc_lcl_addr.s_addr;
795 v->remote = vifc->vifc_rmt_addr.s_addr;
796 v->flags = vifc->vifc_flags;
798 v->flags |= VIFF_STATIC;
799 v->threshold = vifc->vifc_threshold;
804 v->link = dev->ifindex;
805 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
806 v->link = dev_get_iflink(dev);
808 /* And finish update writing critical data */
809 write_lock_bh(&mrt_lock);
811 #ifdef CONFIG_IP_PIMSM
812 if (v->flags & VIFF_REGISTER)
813 mrt->mroute_reg_vif_num = vifi;
815 if (vifi+1 > mrt->maxvif)
816 mrt->maxvif = vifi+1;
817 write_unlock_bh(&mrt_lock);
821 /* called with rcu_read_lock() */
822 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
826 int line = MFC_HASH(mcastgrp, origin);
829 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
830 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
836 /* Look for a (*,*,oif) entry */
837 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
840 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
843 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
844 if (c->mfc_origin == htonl(INADDR_ANY) &&
845 c->mfc_mcastgrp == htonl(INADDR_ANY) &&
846 c->mfc_un.res.ttls[vifi] < 255)
852 /* Look for a (*,G) entry */
853 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
854 __be32 mcastgrp, int vifi)
856 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
857 struct mfc_cache *c, *proxy;
859 if (mcastgrp == htonl(INADDR_ANY))
862 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
863 if (c->mfc_origin == htonl(INADDR_ANY) &&
864 c->mfc_mcastgrp == mcastgrp) {
865 if (c->mfc_un.res.ttls[vifi] < 255)
868 /* It's ok if the vifi is part of the static tree */
869 proxy = ipmr_cache_find_any_parent(mrt,
871 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
876 return ipmr_cache_find_any_parent(mrt, vifi);
880 * Allocate a multicast cache entry
882 static struct mfc_cache *ipmr_cache_alloc(void)
884 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
887 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
888 c->mfc_un.res.minvif = MAXVIFS;
893 static struct mfc_cache *ipmr_cache_alloc_unres(void)
895 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
898 skb_queue_head_init(&c->mfc_un.unres.unresolved);
899 c->mfc_un.unres.expires = jiffies + 10*HZ;
905 * A cache entry has gone into a resolved state from queued
908 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
909 struct mfc_cache *uc, struct mfc_cache *c)
914 /* Play the pending entries through our router */
916 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
917 if (ip_hdr(skb)->version == 0) {
918 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
920 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
921 nlh->nlmsg_len = skb_tail_pointer(skb) -
924 nlh->nlmsg_type = NLMSG_ERROR;
925 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
926 skb_trim(skb, nlh->nlmsg_len);
928 e->error = -EMSGSIZE;
929 memset(&e->msg, 0, sizeof(e->msg));
932 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
934 ip_mr_forward(net, mrt, skb, c, 0);
940 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
941 * expects the following bizarre scheme.
943 * Called under mrt_lock.
946 static int ipmr_cache_report(struct mr_table *mrt,
947 struct sk_buff *pkt, vifi_t vifi, int assert)
950 const int ihl = ip_hdrlen(pkt);
951 struct igmphdr *igmp;
953 struct sock *mroute_sk;
956 #ifdef CONFIG_IP_PIMSM
957 if (assert == IGMPMSG_WHOLEPKT)
958 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
961 skb = alloc_skb(128, GFP_ATOMIC);
966 #ifdef CONFIG_IP_PIMSM
967 if (assert == IGMPMSG_WHOLEPKT) {
968 /* Ugly, but we have no choice with this interface.
969 * Duplicate old header, fix ihl, length etc.
970 * And all this only to mangle msg->im_msgtype and
971 * to set msg->im_mbz to "mbz" :-)
973 skb_push(skb, sizeof(struct iphdr));
974 skb_reset_network_header(skb);
975 skb_reset_transport_header(skb);
976 msg = (struct igmpmsg *)skb_network_header(skb);
977 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
978 msg->im_msgtype = IGMPMSG_WHOLEPKT;
980 msg->im_vif = mrt->mroute_reg_vif_num;
981 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
982 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
983 sizeof(struct iphdr));
988 /* Copy the IP header */
990 skb_set_network_header(skb, skb->len);
992 skb_copy_to_linear_data(skb, pkt->data, ihl);
993 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
994 msg = (struct igmpmsg *)skb_network_header(skb);
996 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1000 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
1002 msg->im_msgtype = assert;
1004 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1005 skb->transport_header = skb->network_header;
1009 mroute_sk = rcu_dereference(mrt->mroute_sk);
1016 /* Deliver to mrouted */
1018 ret = sock_queue_rcv_skb(mroute_sk, skb);
1021 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1029 * Queue a packet for resolution. It gets locked cache entry!
1033 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1037 struct mfc_cache *c;
1038 const struct iphdr *iph = ip_hdr(skb);
1040 spin_lock_bh(&mfc_unres_lock);
1041 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1042 if (c->mfc_mcastgrp == iph->daddr &&
1043 c->mfc_origin == iph->saddr) {
1050 /* Create a new entry if allowable */
1052 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1053 (c = ipmr_cache_alloc_unres()) == NULL) {
1054 spin_unlock_bh(&mfc_unres_lock);
1060 /* Fill in the new cache entry */
1063 c->mfc_origin = iph->saddr;
1064 c->mfc_mcastgrp = iph->daddr;
1066 /* Reflect first query at mrouted. */
1068 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1070 /* If the report failed throw the cache entry
1073 spin_unlock_bh(&mfc_unres_lock);
1080 atomic_inc(&mrt->cache_resolve_queue_len);
1081 list_add(&c->list, &mrt->mfc_unres_queue);
1082 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1084 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1085 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1088 /* See if we can append the packet */
1090 if (c->mfc_un.unres.unresolved.qlen > 3) {
1094 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1098 spin_unlock_bh(&mfc_unres_lock);
1103 * MFC cache manipulation by user space mroute daemon
1106 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1109 struct mfc_cache *c, *next;
1111 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1113 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1114 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1115 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1116 (parent == -1 || parent == c->mfc_parent)) {
1117 list_del_rcu(&c->list);
1118 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1126 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1127 struct mfcctl *mfc, int mrtsock, int parent)
1131 struct mfc_cache *uc, *c;
1133 if (mfc->mfcc_parent >= MAXVIFS)
1136 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1138 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1139 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1140 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1141 (parent == -1 || parent == c->mfc_parent)) {
1148 write_lock_bh(&mrt_lock);
1149 c->mfc_parent = mfc->mfcc_parent;
1150 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1152 c->mfc_flags |= MFC_STATIC;
1153 write_unlock_bh(&mrt_lock);
1154 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1158 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1159 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1162 c = ipmr_cache_alloc();
1166 c->mfc_origin = mfc->mfcc_origin.s_addr;
1167 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1168 c->mfc_parent = mfc->mfcc_parent;
1169 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1171 c->mfc_flags |= MFC_STATIC;
1173 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1176 * Check to see if we resolved a queued list. If so we
1177 * need to send on the frames and tidy up.
1180 spin_lock_bh(&mfc_unres_lock);
1181 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1182 if (uc->mfc_origin == c->mfc_origin &&
1183 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1184 list_del(&uc->list);
1185 atomic_dec(&mrt->cache_resolve_queue_len);
1190 if (list_empty(&mrt->mfc_unres_queue))
1191 del_timer(&mrt->ipmr_expire_timer);
1192 spin_unlock_bh(&mfc_unres_lock);
1195 ipmr_cache_resolve(net, mrt, uc, c);
1196 ipmr_cache_free(uc);
1198 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1203 * Close the multicast socket, and clear the vif tables etc
1206 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1210 struct mfc_cache *c, *next;
1212 /* Shut down all active vif entries */
1214 for (i = 0; i < mrt->maxvif; i++) {
1215 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1217 vif_delete(mrt, i, 0, &list);
1219 unregister_netdevice_many(&list);
1221 /* Wipe the cache */
1223 for (i = 0; i < MFC_LINES; i++) {
1224 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1225 if (!all && (c->mfc_flags & MFC_STATIC))
1227 list_del_rcu(&c->list);
1228 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1233 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1234 spin_lock_bh(&mfc_unres_lock);
1235 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1237 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1238 ipmr_destroy_unres(mrt, c);
1240 spin_unlock_bh(&mfc_unres_lock);
1244 /* called from ip_ra_control(), before an RCU grace period,
1245 * we dont need to call synchronize_rcu() here
1247 static void mrtsock_destruct(struct sock *sk)
1249 struct net *net = sock_net(sk);
1250 struct mr_table *mrt;
1253 ipmr_for_each_table(mrt, net) {
1254 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1255 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1256 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1257 NETCONFA_IFINDEX_ALL,
1258 net->ipv4.devconf_all);
1259 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1260 mroute_clean_tables(mrt, false);
1267 * Socket options and virtual interface manipulation. The whole
1268 * virtual interface system is a complete heap, but unfortunately
1269 * that's how BSD mrouted happens to think. Maybe one day with a proper
1270 * MOSPF/PIM router set up we can clean this up.
1273 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1275 int ret, parent = 0;
1278 struct net *net = sock_net(sk);
1279 struct mr_table *mrt;
1281 if (sk->sk_type != SOCK_RAW ||
1282 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1285 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1289 if (optname != MRT_INIT) {
1290 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1291 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1297 if (optlen != sizeof(int))
1301 if (rtnl_dereference(mrt->mroute_sk)) {
1306 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1308 rcu_assign_pointer(mrt->mroute_sk, sk);
1309 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1310 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1311 NETCONFA_IFINDEX_ALL,
1312 net->ipv4.devconf_all);
1317 if (sk != rcu_access_pointer(mrt->mroute_sk))
1319 return ip_ra_control(sk, 0, NULL);
1322 if (optlen != sizeof(vif))
1324 if (copy_from_user(&vif, optval, sizeof(vif)))
1326 if (vif.vifc_vifi >= MAXVIFS)
1329 if (optname == MRT_ADD_VIF) {
1330 ret = vif_add(net, mrt, &vif,
1331 sk == rtnl_dereference(mrt->mroute_sk));
1333 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1339 * Manipulate the forwarding caches. These live
1340 * in a sort of kernel/user symbiosis.
1345 case MRT_ADD_MFC_PROXY:
1346 case MRT_DEL_MFC_PROXY:
1347 if (optlen != sizeof(mfc))
1349 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1352 parent = mfc.mfcc_parent;
1354 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1355 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1357 ret = ipmr_mfc_add(net, mrt, &mfc,
1358 sk == rtnl_dereference(mrt->mroute_sk),
1363 * Control PIM assert.
1368 if (optlen != sizeof(v))
1370 if (get_user(v, (int __user *)optval))
1372 mrt->mroute_do_assert = v;
1375 #ifdef CONFIG_IP_PIMSM
1380 if (optlen != sizeof(v))
1382 if (get_user(v, (int __user *)optval))
1388 if (v != mrt->mroute_do_pim) {
1389 mrt->mroute_do_pim = v;
1390 mrt->mroute_do_assert = v;
1396 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1401 if (optlen != sizeof(u32))
1403 if (get_user(v, (u32 __user *)optval))
1406 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1407 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1412 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1415 if (!ipmr_new_table(net, v))
1418 raw_sk(sk)->ipmr_table = v;
1425 * Spurious command, or MRT_VERSION which you cannot
1429 return -ENOPROTOOPT;
1434 * Getsock opt support for the multicast routing system.
1437 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1441 struct net *net = sock_net(sk);
1442 struct mr_table *mrt;
1444 if (sk->sk_type != SOCK_RAW ||
1445 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1448 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1452 if (optname != MRT_VERSION &&
1453 #ifdef CONFIG_IP_PIMSM
1454 optname != MRT_PIM &&
1456 optname != MRT_ASSERT)
1457 return -ENOPROTOOPT;
1459 if (get_user(olr, optlen))
1462 olr = min_t(unsigned int, olr, sizeof(int));
1466 if (put_user(olr, optlen))
1468 if (optname == MRT_VERSION)
1470 #ifdef CONFIG_IP_PIMSM
1471 else if (optname == MRT_PIM)
1472 val = mrt->mroute_do_pim;
1475 val = mrt->mroute_do_assert;
1476 if (copy_to_user(optval, &val, olr))
1482 * The IP multicast ioctl support routines.
1485 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1487 struct sioc_sg_req sr;
1488 struct sioc_vif_req vr;
1489 struct vif_device *vif;
1490 struct mfc_cache *c;
1491 struct net *net = sock_net(sk);
1492 struct mr_table *mrt;
1494 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1500 if (copy_from_user(&vr, arg, sizeof(vr)))
1502 if (vr.vifi >= mrt->maxvif)
1504 read_lock(&mrt_lock);
1505 vif = &mrt->vif_table[vr.vifi];
1506 if (VIF_EXISTS(mrt, vr.vifi)) {
1507 vr.icount = vif->pkt_in;
1508 vr.ocount = vif->pkt_out;
1509 vr.ibytes = vif->bytes_in;
1510 vr.obytes = vif->bytes_out;
1511 read_unlock(&mrt_lock);
1513 if (copy_to_user(arg, &vr, sizeof(vr)))
1517 read_unlock(&mrt_lock);
1518 return -EADDRNOTAVAIL;
1520 if (copy_from_user(&sr, arg, sizeof(sr)))
1524 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1526 sr.pktcnt = c->mfc_un.res.pkt;
1527 sr.bytecnt = c->mfc_un.res.bytes;
1528 sr.wrong_if = c->mfc_un.res.wrong_if;
1531 if (copy_to_user(arg, &sr, sizeof(sr)))
1536 return -EADDRNOTAVAIL;
1538 return -ENOIOCTLCMD;
1542 #ifdef CONFIG_COMPAT
1543 struct compat_sioc_sg_req {
1546 compat_ulong_t pktcnt;
1547 compat_ulong_t bytecnt;
1548 compat_ulong_t wrong_if;
1551 struct compat_sioc_vif_req {
1552 vifi_t vifi; /* Which iface */
1553 compat_ulong_t icount;
1554 compat_ulong_t ocount;
1555 compat_ulong_t ibytes;
1556 compat_ulong_t obytes;
1559 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1561 struct compat_sioc_sg_req sr;
1562 struct compat_sioc_vif_req vr;
1563 struct vif_device *vif;
1564 struct mfc_cache *c;
1565 struct net *net = sock_net(sk);
1566 struct mr_table *mrt;
1568 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1574 if (copy_from_user(&vr, arg, sizeof(vr)))
1576 if (vr.vifi >= mrt->maxvif)
1578 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1579 read_lock(&mrt_lock);
1580 vif = &mrt->vif_table[vr.vifi];
1581 if (VIF_EXISTS(mrt, vr.vifi)) {
1582 vr.icount = vif->pkt_in;
1583 vr.ocount = vif->pkt_out;
1584 vr.ibytes = vif->bytes_in;
1585 vr.obytes = vif->bytes_out;
1586 read_unlock(&mrt_lock);
1588 if (copy_to_user(arg, &vr, sizeof(vr)))
1592 read_unlock(&mrt_lock);
1593 return -EADDRNOTAVAIL;
1595 if (copy_from_user(&sr, arg, sizeof(sr)))
1599 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1601 sr.pktcnt = c->mfc_un.res.pkt;
1602 sr.bytecnt = c->mfc_un.res.bytes;
1603 sr.wrong_if = c->mfc_un.res.wrong_if;
1606 if (copy_to_user(arg, &sr, sizeof(sr)))
1611 return -EADDRNOTAVAIL;
1613 return -ENOIOCTLCMD;
1619 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1621 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1622 struct net *net = dev_net(dev);
1623 struct mr_table *mrt;
1624 struct vif_device *v;
1627 if (event != NETDEV_UNREGISTER)
1630 ipmr_for_each_table(mrt, net) {
1631 v = &mrt->vif_table[0];
1632 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1634 vif_delete(mrt, ct, 1, NULL);
1641 static struct notifier_block ip_mr_notifier = {
1642 .notifier_call = ipmr_device_event,
1646 * Encapsulate a packet by attaching a valid IPIP header to it.
1647 * This avoids tunnel drivers and other mess and gives us the speed so
1648 * important for multicast video.
1651 static void ip_encap(struct net *net, struct sk_buff *skb,
1652 __be32 saddr, __be32 daddr)
1655 const struct iphdr *old_iph = ip_hdr(skb);
1657 skb_push(skb, sizeof(struct iphdr));
1658 skb->transport_header = skb->network_header;
1659 skb_reset_network_header(skb);
1663 iph->tos = old_iph->tos;
1664 iph->ttl = old_iph->ttl;
1668 iph->protocol = IPPROTO_IPIP;
1670 iph->tot_len = htons(skb->len);
1671 ip_select_ident(net, skb, NULL);
1674 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1678 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1679 struct sk_buff *skb)
1681 struct ip_options *opt = &(IPCB(skb)->opt);
1683 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1684 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1686 if (unlikely(opt->optlen))
1687 ip_forward_options(skb);
1689 return dst_output(net, sk, skb);
1693 * Processing handlers for ipmr_forward
1696 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1697 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1699 const struct iphdr *iph = ip_hdr(skb);
1700 struct vif_device *vif = &mrt->vif_table[vifi];
1701 struct net_device *dev;
1709 #ifdef CONFIG_IP_PIMSM
1710 if (vif->flags & VIFF_REGISTER) {
1712 vif->bytes_out += skb->len;
1713 vif->dev->stats.tx_bytes += skb->len;
1714 vif->dev->stats.tx_packets++;
1715 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1720 if (vif->flags & VIFF_TUNNEL) {
1721 rt = ip_route_output_ports(net, &fl4, NULL,
1722 vif->remote, vif->local,
1725 RT_TOS(iph->tos), vif->link);
1728 encap = sizeof(struct iphdr);
1730 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1733 RT_TOS(iph->tos), vif->link);
1740 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1741 /* Do not fragment multicasts. Alas, IPv4 does not
1742 * allow to send ICMP, so that packets will disappear
1746 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1751 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1753 if (skb_cow(skb, encap)) {
1759 vif->bytes_out += skb->len;
1762 skb_dst_set(skb, &rt->dst);
1763 ip_decrease_ttl(ip_hdr(skb));
1765 /* FIXME: forward and output firewalls used to be called here.
1766 * What do we do with netfilter? -- RR
1768 if (vif->flags & VIFF_TUNNEL) {
1769 ip_encap(net, skb, vif->local, vif->remote);
1770 /* FIXME: extra output firewall step used to be here. --RR */
1771 vif->dev->stats.tx_packets++;
1772 vif->dev->stats.tx_bytes += skb->len;
1775 IPCB(skb)->flags |= IPSKB_FORWARDED;
1778 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1779 * not only before forwarding, but after forwarding on all output
1780 * interfaces. It is clear, if mrouter runs a multicasting
1781 * program, it should receive packets not depending to what interface
1782 * program is joined.
1783 * If we will not make it, the program will have to join on all
1784 * interfaces. On the other hand, multihoming host (or router, but
1785 * not mrouter) cannot join to more than one interface - it will
1786 * result in receiving multiple packets.
1788 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1789 net, NULL, skb, skb->dev, dev,
1790 ipmr_forward_finish);
1797 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1801 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1802 if (mrt->vif_table[ct].dev == dev)
1808 /* "local" means that we should preserve one skb (for local delivery) */
1810 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1811 struct sk_buff *skb, struct mfc_cache *cache,
1816 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1818 vif = cache->mfc_parent;
1819 cache->mfc_un.res.pkt++;
1820 cache->mfc_un.res.bytes += skb->len;
1822 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1823 struct mfc_cache *cache_proxy;
1825 /* For an (*,G) entry, we only check that the incomming
1826 * interface is part of the static tree.
1828 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1830 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1835 * Wrong interface: drop packet and (maybe) send PIM assert.
1837 if (mrt->vif_table[vif].dev != skb->dev) {
1838 if (rt_is_output_route(skb_rtable(skb))) {
1839 /* It is our own packet, looped back.
1840 * Very complicated situation...
1842 * The best workaround until routing daemons will be
1843 * fixed is not to redistribute packet, if it was
1844 * send through wrong interface. It means, that
1845 * multicast applications WILL NOT work for
1846 * (S,G), which have default multicast route pointing
1847 * to wrong oif. In any case, it is not a good
1848 * idea to use multicasting applications on router.
1853 cache->mfc_un.res.wrong_if++;
1855 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1856 /* pimsm uses asserts, when switching from RPT to SPT,
1857 * so that we cannot check that packet arrived on an oif.
1858 * It is bad, but otherwise we would need to move pretty
1859 * large chunk of pimd to kernel. Ough... --ANK
1861 (mrt->mroute_do_pim ||
1862 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1864 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1865 cache->mfc_un.res.last_assert = jiffies;
1866 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1872 mrt->vif_table[vif].pkt_in++;
1873 mrt->vif_table[vif].bytes_in += skb->len;
1878 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1879 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1880 if (true_vifi >= 0 &&
1881 true_vifi != cache->mfc_parent &&
1883 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1884 /* It's an (*,*) entry and the packet is not coming from
1885 * the upstream: forward the packet to the upstream
1888 psend = cache->mfc_parent;
1893 for (ct = cache->mfc_un.res.maxvif - 1;
1894 ct >= cache->mfc_un.res.minvif; ct--) {
1895 /* For (*,G) entry, don't forward to the incoming interface */
1896 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1898 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1900 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1903 ipmr_queue_xmit(net, mrt, skb2, cache,
1912 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1915 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1917 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1927 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1929 struct rtable *rt = skb_rtable(skb);
1930 struct iphdr *iph = ip_hdr(skb);
1931 struct flowi4 fl4 = {
1932 .daddr = iph->daddr,
1933 .saddr = iph->saddr,
1934 .flowi4_tos = RT_TOS(iph->tos),
1935 .flowi4_oif = (rt_is_output_route(rt) ?
1936 skb->dev->ifindex : 0),
1937 .flowi4_iif = (rt_is_output_route(rt) ?
1940 .flowi4_mark = skb->mark,
1942 struct mr_table *mrt;
1945 err = ipmr_fib_lookup(net, &fl4, &mrt);
1947 return ERR_PTR(err);
1952 * Multicast packets for forwarding arrive here
1953 * Called with rcu_read_lock();
1956 int ip_mr_input(struct sk_buff *skb)
1958 struct mfc_cache *cache;
1959 struct net *net = dev_net(skb->dev);
1960 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1961 struct mr_table *mrt;
1963 /* Packet is looped back after forward, it should not be
1964 * forwarded second time, but still can be delivered locally.
1966 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1969 mrt = ipmr_rt_fib_lookup(net, skb);
1972 return PTR_ERR(mrt);
1975 if (IPCB(skb)->opt.router_alert) {
1976 if (ip_call_ra_chain(skb))
1978 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1979 /* IGMPv1 (and broken IGMPv2 implementations sort of
1980 * Cisco IOS <= 11.2(8)) do not put router alert
1981 * option to IGMP packets destined to routable
1982 * groups. It is very bad, because it means
1983 * that we can forward NO IGMP messages.
1985 struct sock *mroute_sk;
1987 mroute_sk = rcu_dereference(mrt->mroute_sk);
1990 raw_rcv(mroute_sk, skb);
1996 /* already under rcu_read_lock() */
1997 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1999 int vif = ipmr_find_vif(mrt, skb->dev);
2002 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2007 * No usable cache entry
2013 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2014 ip_local_deliver(skb);
2020 read_lock(&mrt_lock);
2021 vif = ipmr_find_vif(mrt, skb->dev);
2023 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2024 read_unlock(&mrt_lock);
2028 read_unlock(&mrt_lock);
2033 read_lock(&mrt_lock);
2034 ip_mr_forward(net, mrt, skb, cache, local);
2035 read_unlock(&mrt_lock);
2038 return ip_local_deliver(skb);
2044 return ip_local_deliver(skb);
2049 #ifdef CONFIG_IP_PIMSM
2050 /* called with rcu_read_lock() */
2051 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2052 unsigned int pimlen)
2054 struct net_device *reg_dev = NULL;
2055 struct iphdr *encap;
2057 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2060 * a. packet is really sent to a multicast group
2061 * b. packet is not a NULL-REGISTER
2062 * c. packet is not truncated
2064 if (!ipv4_is_multicast(encap->daddr) ||
2065 encap->tot_len == 0 ||
2066 ntohs(encap->tot_len) + pimlen > skb->len)
2069 read_lock(&mrt_lock);
2070 if (mrt->mroute_reg_vif_num >= 0)
2071 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2072 read_unlock(&mrt_lock);
2077 skb->mac_header = skb->network_header;
2078 skb_pull(skb, (u8 *)encap - skb->data);
2079 skb_reset_network_header(skb);
2080 skb->protocol = htons(ETH_P_IP);
2081 skb->ip_summed = CHECKSUM_NONE;
2083 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2087 return NET_RX_SUCCESS;
2091 #ifdef CONFIG_IP_PIMSM_V1
2093 * Handle IGMP messages of PIMv1
2096 int pim_rcv_v1(struct sk_buff *skb)
2098 struct igmphdr *pim;
2099 struct net *net = dev_net(skb->dev);
2100 struct mr_table *mrt;
2102 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2105 pim = igmp_hdr(skb);
2107 mrt = ipmr_rt_fib_lookup(net, skb);
2110 if (!mrt->mroute_do_pim ||
2111 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2114 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2122 #ifdef CONFIG_IP_PIMSM_V2
2123 static int pim_rcv(struct sk_buff *skb)
2125 struct pimreghdr *pim;
2126 struct net *net = dev_net(skb->dev);
2127 struct mr_table *mrt;
2129 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2132 pim = (struct pimreghdr *)skb_transport_header(skb);
2133 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2134 (pim->flags & PIM_NULL_REGISTER) ||
2135 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2136 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2139 mrt = ipmr_rt_fib_lookup(net, skb);
2142 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2150 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2151 struct mfc_cache *c, struct rtmsg *rtm)
2154 struct rtnexthop *nhp;
2155 struct nlattr *mp_attr;
2156 struct rta_mfc_stats mfcs;
2158 /* If cache is unresolved, don't try to parse IIF and OIF */
2159 if (c->mfc_parent >= MAXVIFS)
2162 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2163 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2166 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2169 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2170 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2171 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2172 nla_nest_cancel(skb, mp_attr);
2176 nhp->rtnh_flags = 0;
2177 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2178 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2179 nhp->rtnh_len = sizeof(*nhp);
2183 nla_nest_end(skb, mp_attr);
2185 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2186 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2187 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2188 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2191 rtm->rtm_type = RTN_MULTICAST;
2195 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2196 __be32 saddr, __be32 daddr,
2197 struct rtmsg *rtm, int nowait, u32 portid)
2199 struct mfc_cache *cache;
2200 struct mr_table *mrt;
2203 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2208 cache = ipmr_cache_find(mrt, saddr, daddr);
2209 if (!cache && skb->dev) {
2210 int vif = ipmr_find_vif(mrt, skb->dev);
2213 cache = ipmr_cache_find_any(mrt, daddr, vif);
2216 struct sk_buff *skb2;
2218 struct net_device *dev;
2227 read_lock(&mrt_lock);
2229 vif = ipmr_find_vif(mrt, dev);
2231 read_unlock(&mrt_lock);
2235 skb2 = skb_clone(skb, GFP_ATOMIC);
2237 read_unlock(&mrt_lock);
2242 NETLINK_CB(skb2).portid = portid;
2243 skb_push(skb2, sizeof(struct iphdr));
2244 skb_reset_network_header(skb2);
2246 iph->ihl = sizeof(struct iphdr) >> 2;
2250 err = ipmr_cache_unresolved(mrt, vif, skb2);
2251 read_unlock(&mrt_lock);
2256 read_lock(&mrt_lock);
2257 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2258 cache->mfc_flags |= MFC_NOTIFY;
2259 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2260 read_unlock(&mrt_lock);
2265 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2266 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2269 struct nlmsghdr *nlh;
2273 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2277 rtm = nlmsg_data(nlh);
2278 rtm->rtm_family = RTNL_FAMILY_IPMR;
2279 rtm->rtm_dst_len = 32;
2280 rtm->rtm_src_len = 32;
2282 rtm->rtm_table = mrt->id;
2283 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2284 goto nla_put_failure;
2285 rtm->rtm_type = RTN_MULTICAST;
2286 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2287 if (c->mfc_flags & MFC_STATIC)
2288 rtm->rtm_protocol = RTPROT_STATIC;
2290 rtm->rtm_protocol = RTPROT_MROUTED;
2293 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2294 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2295 goto nla_put_failure;
2296 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2297 /* do not break the dump if cache is unresolved */
2298 if (err < 0 && err != -ENOENT)
2299 goto nla_put_failure;
2301 nlmsg_end(skb, nlh);
2305 nlmsg_cancel(skb, nlh);
2309 static size_t mroute_msgsize(bool unresolved, int maxvif)
2312 NLMSG_ALIGN(sizeof(struct rtmsg))
2313 + nla_total_size(4) /* RTA_TABLE */
2314 + nla_total_size(4) /* RTA_SRC */
2315 + nla_total_size(4) /* RTA_DST */
2320 + nla_total_size(4) /* RTA_IIF */
2321 + nla_total_size(0) /* RTA_MULTIPATH */
2322 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2324 + nla_total_size(sizeof(struct rta_mfc_stats))
2330 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2333 struct net *net = read_pnet(&mrt->net);
2334 struct sk_buff *skb;
2337 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2342 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2346 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2352 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2355 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2357 struct net *net = sock_net(skb->sk);
2358 struct mr_table *mrt;
2359 struct mfc_cache *mfc;
2360 unsigned int t = 0, s_t;
2361 unsigned int h = 0, s_h;
2362 unsigned int e = 0, s_e;
2369 ipmr_for_each_table(mrt, net) {
2374 for (h = s_h; h < MFC_LINES; h++) {
2375 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2378 if (ipmr_fill_mroute(mrt, skb,
2379 NETLINK_CB(cb->skb).portid,
2389 spin_lock_bh(&mfc_unres_lock);
2390 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2393 if (ipmr_fill_mroute(mrt, skb,
2394 NETLINK_CB(cb->skb).portid,
2398 spin_unlock_bh(&mfc_unres_lock);
2404 spin_unlock_bh(&mfc_unres_lock);
2420 #ifdef CONFIG_PROC_FS
2422 * The /proc interfaces to multicast routing :
2423 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2425 struct ipmr_vif_iter {
2426 struct seq_net_private p;
2427 struct mr_table *mrt;
2431 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2432 struct ipmr_vif_iter *iter,
2435 struct mr_table *mrt = iter->mrt;
2437 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2438 if (!VIF_EXISTS(mrt, iter->ct))
2441 return &mrt->vif_table[iter->ct];
2446 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2447 __acquires(mrt_lock)
2449 struct ipmr_vif_iter *iter = seq->private;
2450 struct net *net = seq_file_net(seq);
2451 struct mr_table *mrt;
2453 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2455 return ERR_PTR(-ENOENT);
2459 read_lock(&mrt_lock);
2460 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2464 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2466 struct ipmr_vif_iter *iter = seq->private;
2467 struct net *net = seq_file_net(seq);
2468 struct mr_table *mrt = iter->mrt;
2471 if (v == SEQ_START_TOKEN)
2472 return ipmr_vif_seq_idx(net, iter, 0);
2474 while (++iter->ct < mrt->maxvif) {
2475 if (!VIF_EXISTS(mrt, iter->ct))
2477 return &mrt->vif_table[iter->ct];
2482 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2483 __releases(mrt_lock)
2485 read_unlock(&mrt_lock);
2488 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2490 struct ipmr_vif_iter *iter = seq->private;
2491 struct mr_table *mrt = iter->mrt;
2493 if (v == SEQ_START_TOKEN) {
2495 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2497 const struct vif_device *vif = v;
2498 const char *name = vif->dev ? vif->dev->name : "none";
2501 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2502 vif - mrt->vif_table,
2503 name, vif->bytes_in, vif->pkt_in,
2504 vif->bytes_out, vif->pkt_out,
2505 vif->flags, vif->local, vif->remote);
2510 static const struct seq_operations ipmr_vif_seq_ops = {
2511 .start = ipmr_vif_seq_start,
2512 .next = ipmr_vif_seq_next,
2513 .stop = ipmr_vif_seq_stop,
2514 .show = ipmr_vif_seq_show,
2517 static int ipmr_vif_open(struct inode *inode, struct file *file)
2519 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2520 sizeof(struct ipmr_vif_iter));
2523 static const struct file_operations ipmr_vif_fops = {
2524 .owner = THIS_MODULE,
2525 .open = ipmr_vif_open,
2527 .llseek = seq_lseek,
2528 .release = seq_release_net,
2531 struct ipmr_mfc_iter {
2532 struct seq_net_private p;
2533 struct mr_table *mrt;
2534 struct list_head *cache;
2539 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2540 struct ipmr_mfc_iter *it, loff_t pos)
2542 struct mr_table *mrt = it->mrt;
2543 struct mfc_cache *mfc;
2546 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2547 it->cache = &mrt->mfc_cache_array[it->ct];
2548 list_for_each_entry_rcu(mfc, it->cache, list)
2554 spin_lock_bh(&mfc_unres_lock);
2555 it->cache = &mrt->mfc_unres_queue;
2556 list_for_each_entry(mfc, it->cache, list)
2559 spin_unlock_bh(&mfc_unres_lock);
2566 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2568 struct ipmr_mfc_iter *it = seq->private;
2569 struct net *net = seq_file_net(seq);
2570 struct mr_table *mrt;
2572 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2574 return ERR_PTR(-ENOENT);
2579 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2583 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2585 struct mfc_cache *mfc = v;
2586 struct ipmr_mfc_iter *it = seq->private;
2587 struct net *net = seq_file_net(seq);
2588 struct mr_table *mrt = it->mrt;
2592 if (v == SEQ_START_TOKEN)
2593 return ipmr_mfc_seq_idx(net, seq->private, 0);
2595 if (mfc->list.next != it->cache)
2596 return list_entry(mfc->list.next, struct mfc_cache, list);
2598 if (it->cache == &mrt->mfc_unres_queue)
2601 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2603 while (++it->ct < MFC_LINES) {
2604 it->cache = &mrt->mfc_cache_array[it->ct];
2605 if (list_empty(it->cache))
2607 return list_first_entry(it->cache, struct mfc_cache, list);
2610 /* exhausted cache_array, show unresolved */
2612 it->cache = &mrt->mfc_unres_queue;
2615 spin_lock_bh(&mfc_unres_lock);
2616 if (!list_empty(it->cache))
2617 return list_first_entry(it->cache, struct mfc_cache, list);
2620 spin_unlock_bh(&mfc_unres_lock);
2626 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2628 struct ipmr_mfc_iter *it = seq->private;
2629 struct mr_table *mrt = it->mrt;
2631 if (it->cache == &mrt->mfc_unres_queue)
2632 spin_unlock_bh(&mfc_unres_lock);
2633 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2637 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2641 if (v == SEQ_START_TOKEN) {
2643 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2645 const struct mfc_cache *mfc = v;
2646 const struct ipmr_mfc_iter *it = seq->private;
2647 const struct mr_table *mrt = it->mrt;
2649 seq_printf(seq, "%08X %08X %-3hd",
2650 (__force u32) mfc->mfc_mcastgrp,
2651 (__force u32) mfc->mfc_origin,
2654 if (it->cache != &mrt->mfc_unres_queue) {
2655 seq_printf(seq, " %8lu %8lu %8lu",
2656 mfc->mfc_un.res.pkt,
2657 mfc->mfc_un.res.bytes,
2658 mfc->mfc_un.res.wrong_if);
2659 for (n = mfc->mfc_un.res.minvif;
2660 n < mfc->mfc_un.res.maxvif; n++) {
2661 if (VIF_EXISTS(mrt, n) &&
2662 mfc->mfc_un.res.ttls[n] < 255)
2665 n, mfc->mfc_un.res.ttls[n]);
2668 /* unresolved mfc_caches don't contain
2669 * pkt, bytes and wrong_if values
2671 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2673 seq_putc(seq, '\n');
2678 static const struct seq_operations ipmr_mfc_seq_ops = {
2679 .start = ipmr_mfc_seq_start,
2680 .next = ipmr_mfc_seq_next,
2681 .stop = ipmr_mfc_seq_stop,
2682 .show = ipmr_mfc_seq_show,
2685 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2687 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2688 sizeof(struct ipmr_mfc_iter));
2691 static const struct file_operations ipmr_mfc_fops = {
2692 .owner = THIS_MODULE,
2693 .open = ipmr_mfc_open,
2695 .llseek = seq_lseek,
2696 .release = seq_release_net,
2700 #ifdef CONFIG_IP_PIMSM_V2
2701 static const struct net_protocol pim_protocol = {
2709 * Setup for IP multicast routing
2711 static int __net_init ipmr_net_init(struct net *net)
2715 err = ipmr_rules_init(net);
2719 #ifdef CONFIG_PROC_FS
2721 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2723 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2724 goto proc_cache_fail;
2728 #ifdef CONFIG_PROC_FS
2730 remove_proc_entry("ip_mr_vif", net->proc_net);
2732 ipmr_rules_exit(net);
2738 static void __net_exit ipmr_net_exit(struct net *net)
2740 #ifdef CONFIG_PROC_FS
2741 remove_proc_entry("ip_mr_cache", net->proc_net);
2742 remove_proc_entry("ip_mr_vif", net->proc_net);
2744 ipmr_rules_exit(net);
2747 static struct pernet_operations ipmr_net_ops = {
2748 .init = ipmr_net_init,
2749 .exit = ipmr_net_exit,
2752 int __init ip_mr_init(void)
2756 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2757 sizeof(struct mfc_cache),
2758 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2763 err = register_pernet_subsys(&ipmr_net_ops);
2765 goto reg_pernet_fail;
2767 err = register_netdevice_notifier(&ip_mr_notifier);
2769 goto reg_notif_fail;
2770 #ifdef CONFIG_IP_PIMSM_V2
2771 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2772 pr_err("%s: can't add PIM protocol\n", __func__);
2774 goto add_proto_fail;
2777 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2778 NULL, ipmr_rtm_dumproute, NULL);
2781 #ifdef CONFIG_IP_PIMSM_V2
2783 unregister_netdevice_notifier(&ip_mr_notifier);
2786 unregister_pernet_subsys(&ipmr_net_ops);
2788 kmem_cache_destroy(mrt_cachep);