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>
70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
71 #define CONFIG_IP_PIMSM 1
75 struct list_head list;
78 struct sock __rcu *mroute_sk;
79 struct timer_list ipmr_expire_timer;
80 struct list_head mfc_unres_queue;
81 struct list_head mfc_cache_array[MFC_LINES];
82 struct vif_device vif_table[MAXVIFS];
84 atomic_t cache_resolve_queue_len;
85 bool mroute_do_assert;
87 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
88 int mroute_reg_vif_num;
93 struct fib_rule common;
100 /* Big lock, protecting vif table, mrt cache and mroute socket state.
101 * Note that the changes are semaphored via rtnl_lock.
104 static DEFINE_RWLOCK(mrt_lock);
107 * Multicast router control variables
110 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
112 /* Special spinlock for queue of unresolved entries */
113 static DEFINE_SPINLOCK(mfc_unres_lock);
115 /* We return to original Alan's scheme. Hash table of resolved
116 * entries is changed only in process context and protected
117 * with weak lock mrt_lock. Queue of unresolved entries is protected
118 * with strong spinlock mfc_unres_lock.
120 * In this case data path is free of exclusive locks at all.
123 static struct kmem_cache *mrt_cachep __read_mostly;
125 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
126 static void ipmr_free_table(struct mr_table *mrt);
128 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
129 struct sk_buff *skb, struct mfc_cache *cache,
131 static int ipmr_cache_report(struct mr_table *mrt,
132 struct sk_buff *pkt, vifi_t vifi, int assert);
133 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
134 struct mfc_cache *c, struct rtmsg *rtm);
135 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
137 static void mroute_clean_tables(struct mr_table *mrt);
138 static void ipmr_expire_process(unsigned long arg);
140 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
141 #define ipmr_for_each_table(mrt, net) \
142 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
144 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
146 struct mr_table *mrt;
148 ipmr_for_each_table(mrt, net) {
155 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
156 struct mr_table **mrt)
159 struct ipmr_result res;
160 struct fib_lookup_arg arg = {
162 .flags = FIB_LOOKUP_NOREF,
165 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
166 flowi4_to_flowi(flp4), 0, &arg);
173 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
174 int flags, struct fib_lookup_arg *arg)
176 struct ipmr_result *res = arg->result;
177 struct mr_table *mrt;
179 switch (rule->action) {
182 case FR_ACT_UNREACHABLE:
184 case FR_ACT_PROHIBIT:
186 case FR_ACT_BLACKHOLE:
191 mrt = ipmr_get_table(rule->fr_net, rule->table);
198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
203 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
207 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
208 struct fib_rule_hdr *frh, struct nlattr **tb)
213 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
219 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
220 struct fib_rule_hdr *frh)
228 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
229 .family = RTNL_FAMILY_IPMR,
230 .rule_size = sizeof(struct ipmr_rule),
231 .addr_size = sizeof(u32),
232 .action = ipmr_rule_action,
233 .match = ipmr_rule_match,
234 .configure = ipmr_rule_configure,
235 .compare = ipmr_rule_compare,
236 .default_pref = fib_default_rule_pref,
237 .fill = ipmr_rule_fill,
238 .nlgroup = RTNLGRP_IPV4_RULE,
239 .policy = ipmr_rule_policy,
240 .owner = THIS_MODULE,
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;
279 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
280 list_del(&mrt->list);
281 ipmr_free_table(mrt);
283 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 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
304 return net->ipv4.mrt ? 0 : -ENOMEM;
307 static void __net_exit ipmr_rules_exit(struct net *net)
309 ipmr_free_table(net->ipv4.mrt);
313 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
315 struct mr_table *mrt;
318 mrt = ipmr_get_table(net, id);
322 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
325 write_pnet(&mrt->net, net);
328 /* Forwarding cache */
329 for (i = 0; i < MFC_LINES; i++)
330 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
332 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
334 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
337 #ifdef CONFIG_IP_PIMSM
338 mrt->mroute_reg_vif_num = -1;
340 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
341 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
346 static void ipmr_free_table(struct mr_table *mrt)
348 del_timer_sync(&mrt->ipmr_expire_timer);
349 mroute_clean_tables(mrt);
353 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
355 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
357 struct net *net = dev_net(dev);
361 dev = __dev_get_by_name(net, "tunl0");
363 const struct net_device_ops *ops = dev->netdev_ops;
365 struct ip_tunnel_parm p;
367 memset(&p, 0, sizeof(p));
368 p.iph.daddr = v->vifc_rmt_addr.s_addr;
369 p.iph.saddr = v->vifc_lcl_addr.s_addr;
372 p.iph.protocol = IPPROTO_IPIP;
373 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
374 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
376 if (ops->ndo_do_ioctl) {
377 mm_segment_t oldfs = get_fs();
380 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
387 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
389 struct net_device *dev;
391 dev = __dev_get_by_name(net, "tunl0");
394 const struct net_device_ops *ops = dev->netdev_ops;
397 struct ip_tunnel_parm p;
398 struct in_device *in_dev;
400 memset(&p, 0, sizeof(p));
401 p.iph.daddr = v->vifc_rmt_addr.s_addr;
402 p.iph.saddr = v->vifc_lcl_addr.s_addr;
405 p.iph.protocol = IPPROTO_IPIP;
406 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
407 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
409 if (ops->ndo_do_ioctl) {
410 mm_segment_t oldfs = get_fs();
413 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
421 (dev = __dev_get_by_name(net, p.name)) != NULL) {
422 dev->flags |= IFF_MULTICAST;
424 in_dev = __in_dev_get_rtnl(dev);
428 ipv4_devconf_setall(in_dev);
429 neigh_parms_data_state_setall(in_dev->arp_parms);
430 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
440 /* allow the register to be completed before unregistering. */
444 unregister_netdevice(dev);
448 #ifdef CONFIG_IP_PIMSM
450 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
452 struct net *net = dev_net(dev);
453 struct mr_table *mrt;
454 struct flowi4 fl4 = {
455 .flowi4_oif = dev->ifindex,
456 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
457 .flowi4_mark = skb->mark,
461 err = ipmr_fib_lookup(net, &fl4, &mrt);
467 read_lock(&mrt_lock);
468 dev->stats.tx_bytes += skb->len;
469 dev->stats.tx_packets++;
470 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
471 read_unlock(&mrt_lock);
476 static int reg_vif_get_iflink(const struct net_device *dev)
481 static const struct net_device_ops reg_vif_netdev_ops = {
482 .ndo_start_xmit = reg_vif_xmit,
483 .ndo_get_iflink = reg_vif_get_iflink,
486 static void reg_vif_setup(struct net_device *dev)
488 dev->type = ARPHRD_PIMREG;
489 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
490 dev->flags = IFF_NOARP;
491 dev->netdev_ops = ®_vif_netdev_ops;
492 dev->destructor = free_netdev;
493 dev->features |= NETIF_F_NETNS_LOCAL;
496 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
498 struct net_device *dev;
499 struct in_device *in_dev;
502 if (mrt->id == RT_TABLE_DEFAULT)
503 sprintf(name, "pimreg");
505 sprintf(name, "pimreg%u", mrt->id);
507 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
512 dev_net_set(dev, net);
514 if (register_netdevice(dev)) {
520 in_dev = __in_dev_get_rcu(dev);
526 ipv4_devconf_setall(in_dev);
527 neigh_parms_data_state_setall(in_dev->arp_parms);
528 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
539 /* allow the register to be completed before unregistering. */
543 unregister_netdevice(dev);
549 * vif_delete - Delete a VIF entry
550 * @notify: Set to 1, if the caller is a notifier_call
553 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
554 struct list_head *head)
556 struct vif_device *v;
557 struct net_device *dev;
558 struct in_device *in_dev;
560 if (vifi < 0 || vifi >= mrt->maxvif)
561 return -EADDRNOTAVAIL;
563 v = &mrt->vif_table[vifi];
565 write_lock_bh(&mrt_lock);
570 write_unlock_bh(&mrt_lock);
571 return -EADDRNOTAVAIL;
574 #ifdef CONFIG_IP_PIMSM
575 if (vifi == mrt->mroute_reg_vif_num)
576 mrt->mroute_reg_vif_num = -1;
579 if (vifi + 1 == mrt->maxvif) {
582 for (tmp = vifi - 1; tmp >= 0; tmp--) {
583 if (VIF_EXISTS(mrt, tmp))
589 write_unlock_bh(&mrt_lock);
591 dev_set_allmulti(dev, -1);
593 in_dev = __in_dev_get_rtnl(dev);
595 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
596 inet_netconf_notify_devconf(dev_net(dev),
597 NETCONFA_MC_FORWARDING,
598 dev->ifindex, &in_dev->cnf);
599 ip_rt_multicast_event(in_dev);
602 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
603 unregister_netdevice_queue(dev, head);
609 static void ipmr_cache_free_rcu(struct rcu_head *head)
611 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
613 kmem_cache_free(mrt_cachep, c);
616 static inline void ipmr_cache_free(struct mfc_cache *c)
618 call_rcu(&c->rcu, ipmr_cache_free_rcu);
621 /* Destroy an unresolved cache entry, killing queued skbs
622 * and reporting error to netlink readers.
625 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
627 struct net *net = read_pnet(&mrt->net);
631 atomic_dec(&mrt->cache_resolve_queue_len);
633 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
634 if (ip_hdr(skb)->version == 0) {
635 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
636 nlh->nlmsg_type = NLMSG_ERROR;
637 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
638 skb_trim(skb, nlh->nlmsg_len);
640 e->error = -ETIMEDOUT;
641 memset(&e->msg, 0, sizeof(e->msg));
643 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
653 /* Timer process for the unresolved queue. */
655 static void ipmr_expire_process(unsigned long arg)
657 struct mr_table *mrt = (struct mr_table *)arg;
659 unsigned long expires;
660 struct mfc_cache *c, *next;
662 if (!spin_trylock(&mfc_unres_lock)) {
663 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
667 if (list_empty(&mrt->mfc_unres_queue))
673 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
674 if (time_after(c->mfc_un.unres.expires, now)) {
675 unsigned long interval = c->mfc_un.unres.expires - now;
676 if (interval < expires)
682 mroute_netlink_event(mrt, c, RTM_DELROUTE);
683 ipmr_destroy_unres(mrt, c);
686 if (!list_empty(&mrt->mfc_unres_queue))
687 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
690 spin_unlock(&mfc_unres_lock);
693 /* Fill oifs list. It is called under write locked mrt_lock. */
695 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
700 cache->mfc_un.res.minvif = MAXVIFS;
701 cache->mfc_un.res.maxvif = 0;
702 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
704 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
705 if (VIF_EXISTS(mrt, vifi) &&
706 ttls[vifi] && ttls[vifi] < 255) {
707 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
708 if (cache->mfc_un.res.minvif > vifi)
709 cache->mfc_un.res.minvif = vifi;
710 if (cache->mfc_un.res.maxvif <= vifi)
711 cache->mfc_un.res.maxvif = vifi + 1;
716 static int vif_add(struct net *net, struct mr_table *mrt,
717 struct vifctl *vifc, int mrtsock)
719 int vifi = vifc->vifc_vifi;
720 struct vif_device *v = &mrt->vif_table[vifi];
721 struct net_device *dev;
722 struct in_device *in_dev;
726 if (VIF_EXISTS(mrt, vifi))
729 switch (vifc->vifc_flags) {
730 #ifdef CONFIG_IP_PIMSM
733 * Special Purpose VIF in PIM
734 * All the packets will be sent to the daemon
736 if (mrt->mroute_reg_vif_num >= 0)
738 dev = ipmr_reg_vif(net, mrt);
741 err = dev_set_allmulti(dev, 1);
743 unregister_netdevice(dev);
750 dev = ipmr_new_tunnel(net, vifc);
753 err = dev_set_allmulti(dev, 1);
755 ipmr_del_tunnel(dev, vifc);
761 case VIFF_USE_IFINDEX:
763 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
764 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
765 if (dev && !__in_dev_get_rtnl(dev)) {
767 return -EADDRNOTAVAIL;
770 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
773 return -EADDRNOTAVAIL;
774 err = dev_set_allmulti(dev, 1);
784 in_dev = __in_dev_get_rtnl(dev);
787 return -EADDRNOTAVAIL;
789 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
790 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
792 ip_rt_multicast_event(in_dev);
794 /* Fill in the VIF structures */
796 v->rate_limit = vifc->vifc_rate_limit;
797 v->local = vifc->vifc_lcl_addr.s_addr;
798 v->remote = vifc->vifc_rmt_addr.s_addr;
799 v->flags = vifc->vifc_flags;
801 v->flags |= VIFF_STATIC;
802 v->threshold = vifc->vifc_threshold;
807 v->link = dev->ifindex;
808 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
809 v->link = dev_get_iflink(dev);
811 /* And finish update writing critical data */
812 write_lock_bh(&mrt_lock);
814 #ifdef CONFIG_IP_PIMSM
815 if (v->flags & VIFF_REGISTER)
816 mrt->mroute_reg_vif_num = vifi;
818 if (vifi+1 > mrt->maxvif)
819 mrt->maxvif = vifi+1;
820 write_unlock_bh(&mrt_lock);
824 /* called with rcu_read_lock() */
825 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
829 int line = MFC_HASH(mcastgrp, origin);
832 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
833 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
839 /* Look for a (*,*,oif) entry */
840 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
843 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
846 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
847 if (c->mfc_origin == htonl(INADDR_ANY) &&
848 c->mfc_mcastgrp == htonl(INADDR_ANY) &&
849 c->mfc_un.res.ttls[vifi] < 255)
855 /* Look for a (*,G) entry */
856 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
857 __be32 mcastgrp, int vifi)
859 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
860 struct mfc_cache *c, *proxy;
862 if (mcastgrp == htonl(INADDR_ANY))
865 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
866 if (c->mfc_origin == htonl(INADDR_ANY) &&
867 c->mfc_mcastgrp == mcastgrp) {
868 if (c->mfc_un.res.ttls[vifi] < 255)
871 /* It's ok if the vifi is part of the static tree */
872 proxy = ipmr_cache_find_any_parent(mrt,
874 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
879 return ipmr_cache_find_any_parent(mrt, vifi);
883 * Allocate a multicast cache entry
885 static struct mfc_cache *ipmr_cache_alloc(void)
887 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
890 c->mfc_un.res.minvif = MAXVIFS;
894 static struct mfc_cache *ipmr_cache_alloc_unres(void)
896 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
899 skb_queue_head_init(&c->mfc_un.unres.unresolved);
900 c->mfc_un.unres.expires = jiffies + 10*HZ;
906 * A cache entry has gone into a resolved state from queued
909 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
910 struct mfc_cache *uc, struct mfc_cache *c)
915 /* Play the pending entries through our router */
917 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
918 if (ip_hdr(skb)->version == 0) {
919 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
921 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
922 nlh->nlmsg_len = skb_tail_pointer(skb) -
925 nlh->nlmsg_type = NLMSG_ERROR;
926 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
927 skb_trim(skb, nlh->nlmsg_len);
929 e->error = -EMSGSIZE;
930 memset(&e->msg, 0, sizeof(e->msg));
933 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
935 ip_mr_forward(net, mrt, skb, c, 0);
941 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
942 * expects the following bizarre scheme.
944 * Called under mrt_lock.
947 static int ipmr_cache_report(struct mr_table *mrt,
948 struct sk_buff *pkt, vifi_t vifi, int assert)
951 const int ihl = ip_hdrlen(pkt);
952 struct igmphdr *igmp;
954 struct sock *mroute_sk;
957 #ifdef CONFIG_IP_PIMSM
958 if (assert == IGMPMSG_WHOLEPKT)
959 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
962 skb = alloc_skb(128, GFP_ATOMIC);
967 #ifdef CONFIG_IP_PIMSM
968 if (assert == IGMPMSG_WHOLEPKT) {
969 /* Ugly, but we have no choice with this interface.
970 * Duplicate old header, fix ihl, length etc.
971 * And all this only to mangle msg->im_msgtype and
972 * to set msg->im_mbz to "mbz" :-)
974 skb_push(skb, sizeof(struct iphdr));
975 skb_reset_network_header(skb);
976 skb_reset_transport_header(skb);
977 msg = (struct igmpmsg *)skb_network_header(skb);
978 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
979 msg->im_msgtype = IGMPMSG_WHOLEPKT;
981 msg->im_vif = mrt->mroute_reg_vif_num;
982 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
983 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
984 sizeof(struct iphdr));
989 /* Copy the IP header */
991 skb_set_network_header(skb, skb->len);
993 skb_copy_to_linear_data(skb, pkt->data, ihl);
994 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
995 msg = (struct igmpmsg *)skb_network_header(skb);
997 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1001 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
1003 msg->im_msgtype = assert;
1005 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1006 skb->transport_header = skb->network_header;
1010 mroute_sk = rcu_dereference(mrt->mroute_sk);
1017 /* Deliver to mrouted */
1019 ret = sock_queue_rcv_skb(mroute_sk, skb);
1022 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1030 * Queue a packet for resolution. It gets locked cache entry!
1034 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1038 struct mfc_cache *c;
1039 const struct iphdr *iph = ip_hdr(skb);
1041 spin_lock_bh(&mfc_unres_lock);
1042 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1043 if (c->mfc_mcastgrp == iph->daddr &&
1044 c->mfc_origin == iph->saddr) {
1051 /* Create a new entry if allowable */
1053 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1054 (c = ipmr_cache_alloc_unres()) == NULL) {
1055 spin_unlock_bh(&mfc_unres_lock);
1061 /* Fill in the new cache entry */
1064 c->mfc_origin = iph->saddr;
1065 c->mfc_mcastgrp = iph->daddr;
1067 /* Reflect first query at mrouted. */
1069 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1071 /* If the report failed throw the cache entry
1074 spin_unlock_bh(&mfc_unres_lock);
1081 atomic_inc(&mrt->cache_resolve_queue_len);
1082 list_add(&c->list, &mrt->mfc_unres_queue);
1083 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1085 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1086 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1089 /* See if we can append the packet */
1091 if (c->mfc_un.unres.unresolved.qlen > 3) {
1095 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1099 spin_unlock_bh(&mfc_unres_lock);
1104 * MFC cache manipulation by user space mroute daemon
1107 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1110 struct mfc_cache *c, *next;
1112 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1114 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1115 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1116 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1117 (parent == -1 || parent == c->mfc_parent)) {
1118 list_del_rcu(&c->list);
1119 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1127 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1128 struct mfcctl *mfc, int mrtsock, int parent)
1132 struct mfc_cache *uc, *c;
1134 if (mfc->mfcc_parent >= MAXVIFS)
1137 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1139 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1140 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1141 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1142 (parent == -1 || parent == c->mfc_parent)) {
1149 write_lock_bh(&mrt_lock);
1150 c->mfc_parent = mfc->mfcc_parent;
1151 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1153 c->mfc_flags |= MFC_STATIC;
1154 write_unlock_bh(&mrt_lock);
1155 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1159 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1160 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1163 c = ipmr_cache_alloc();
1167 c->mfc_origin = mfc->mfcc_origin.s_addr;
1168 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1169 c->mfc_parent = mfc->mfcc_parent;
1170 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1172 c->mfc_flags |= MFC_STATIC;
1174 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1177 * Check to see if we resolved a queued list. If so we
1178 * need to send on the frames and tidy up.
1181 spin_lock_bh(&mfc_unres_lock);
1182 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1183 if (uc->mfc_origin == c->mfc_origin &&
1184 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1185 list_del(&uc->list);
1186 atomic_dec(&mrt->cache_resolve_queue_len);
1191 if (list_empty(&mrt->mfc_unres_queue))
1192 del_timer(&mrt->ipmr_expire_timer);
1193 spin_unlock_bh(&mfc_unres_lock);
1196 ipmr_cache_resolve(net, mrt, uc, c);
1197 ipmr_cache_free(uc);
1199 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1204 * Close the multicast socket, and clear the vif tables etc
1207 static void mroute_clean_tables(struct mr_table *mrt)
1211 struct mfc_cache *c, *next;
1213 /* Shut down all active vif entries */
1215 for (i = 0; i < mrt->maxvif; i++) {
1216 if (!(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 (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);
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 read_lock(&mrt_lock);
1579 vif = &mrt->vif_table[vr.vifi];
1580 if (VIF_EXISTS(mrt, vr.vifi)) {
1581 vr.icount = vif->pkt_in;
1582 vr.ocount = vif->pkt_out;
1583 vr.ibytes = vif->bytes_in;
1584 vr.obytes = vif->bytes_out;
1585 read_unlock(&mrt_lock);
1587 if (copy_to_user(arg, &vr, sizeof(vr)))
1591 read_unlock(&mrt_lock);
1592 return -EADDRNOTAVAIL;
1594 if (copy_from_user(&sr, arg, sizeof(sr)))
1598 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1600 sr.pktcnt = c->mfc_un.res.pkt;
1601 sr.bytecnt = c->mfc_un.res.bytes;
1602 sr.wrong_if = c->mfc_un.res.wrong_if;
1605 if (copy_to_user(arg, &sr, sizeof(sr)))
1610 return -EADDRNOTAVAIL;
1612 return -ENOIOCTLCMD;
1618 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1620 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1621 struct net *net = dev_net(dev);
1622 struct mr_table *mrt;
1623 struct vif_device *v;
1626 if (event != NETDEV_UNREGISTER)
1629 ipmr_for_each_table(mrt, net) {
1630 v = &mrt->vif_table[0];
1631 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1633 vif_delete(mrt, ct, 1, NULL);
1640 static struct notifier_block ip_mr_notifier = {
1641 .notifier_call = ipmr_device_event,
1645 * Encapsulate a packet by attaching a valid IPIP header to it.
1646 * This avoids tunnel drivers and other mess and gives us the speed so
1647 * important for multicast video.
1650 static void ip_encap(struct net *net, struct sk_buff *skb,
1651 __be32 saddr, __be32 daddr)
1654 const struct iphdr *old_iph = ip_hdr(skb);
1656 skb_push(skb, sizeof(struct iphdr));
1657 skb->transport_header = skb->network_header;
1658 skb_reset_network_header(skb);
1662 iph->tos = old_iph->tos;
1663 iph->ttl = old_iph->ttl;
1667 iph->protocol = IPPROTO_IPIP;
1669 iph->tot_len = htons(skb->len);
1670 ip_select_ident(net, skb, NULL);
1673 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1677 static inline int ipmr_forward_finish(struct sk_buff *skb)
1679 struct ip_options *opt = &(IPCB(skb)->opt);
1681 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1682 IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1684 if (unlikely(opt->optlen))
1685 ip_forward_options(skb);
1687 return dst_output(skb);
1691 * Processing handlers for ipmr_forward
1694 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1695 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1697 const struct iphdr *iph = ip_hdr(skb);
1698 struct vif_device *vif = &mrt->vif_table[vifi];
1699 struct net_device *dev;
1707 #ifdef CONFIG_IP_PIMSM
1708 if (vif->flags & VIFF_REGISTER) {
1710 vif->bytes_out += skb->len;
1711 vif->dev->stats.tx_bytes += skb->len;
1712 vif->dev->stats.tx_packets++;
1713 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1718 if (vif->flags & VIFF_TUNNEL) {
1719 rt = ip_route_output_ports(net, &fl4, NULL,
1720 vif->remote, vif->local,
1723 RT_TOS(iph->tos), vif->link);
1726 encap = sizeof(struct iphdr);
1728 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1731 RT_TOS(iph->tos), vif->link);
1738 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1739 /* Do not fragment multicasts. Alas, IPv4 does not
1740 * allow to send ICMP, so that packets will disappear
1744 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1749 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1751 if (skb_cow(skb, encap)) {
1757 vif->bytes_out += skb->len;
1760 skb_dst_set(skb, &rt->dst);
1761 ip_decrease_ttl(ip_hdr(skb));
1763 /* FIXME: forward and output firewalls used to be called here.
1764 * What do we do with netfilter? -- RR
1766 if (vif->flags & VIFF_TUNNEL) {
1767 ip_encap(net, skb, vif->local, vif->remote);
1768 /* FIXME: extra output firewall step used to be here. --RR */
1769 vif->dev->stats.tx_packets++;
1770 vif->dev->stats.tx_bytes += skb->len;
1773 IPCB(skb)->flags |= IPSKB_FORWARDED;
1776 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1777 * not only before forwarding, but after forwarding on all output
1778 * interfaces. It is clear, if mrouter runs a multicasting
1779 * program, it should receive packets not depending to what interface
1780 * program is joined.
1781 * If we will not make it, the program will have to join on all
1782 * interfaces. On the other hand, multihoming host (or router, but
1783 * not mrouter) cannot join to more than one interface - it will
1784 * result in receiving multiple packets.
1786 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1787 ipmr_forward_finish);
1794 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1798 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1799 if (mrt->vif_table[ct].dev == dev)
1805 /* "local" means that we should preserve one skb (for local delivery) */
1807 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1808 struct sk_buff *skb, struct mfc_cache *cache,
1813 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1815 vif = cache->mfc_parent;
1816 cache->mfc_un.res.pkt++;
1817 cache->mfc_un.res.bytes += skb->len;
1819 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1820 struct mfc_cache *cache_proxy;
1822 /* For an (*,G) entry, we only check that the incomming
1823 * interface is part of the static tree.
1825 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1827 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1832 * Wrong interface: drop packet and (maybe) send PIM assert.
1834 if (mrt->vif_table[vif].dev != skb->dev) {
1835 if (rt_is_output_route(skb_rtable(skb))) {
1836 /* It is our own packet, looped back.
1837 * Very complicated situation...
1839 * The best workaround until routing daemons will be
1840 * fixed is not to redistribute packet, if it was
1841 * send through wrong interface. It means, that
1842 * multicast applications WILL NOT work for
1843 * (S,G), which have default multicast route pointing
1844 * to wrong oif. In any case, it is not a good
1845 * idea to use multicasting applications on router.
1850 cache->mfc_un.res.wrong_if++;
1852 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1853 /* pimsm uses asserts, when switching from RPT to SPT,
1854 * so that we cannot check that packet arrived on an oif.
1855 * It is bad, but otherwise we would need to move pretty
1856 * large chunk of pimd to kernel. Ough... --ANK
1858 (mrt->mroute_do_pim ||
1859 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1861 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1862 cache->mfc_un.res.last_assert = jiffies;
1863 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1869 mrt->vif_table[vif].pkt_in++;
1870 mrt->vif_table[vif].bytes_in += skb->len;
1875 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1876 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1877 if (true_vifi >= 0 &&
1878 true_vifi != cache->mfc_parent &&
1880 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1881 /* It's an (*,*) entry and the packet is not coming from
1882 * the upstream: forward the packet to the upstream
1885 psend = cache->mfc_parent;
1890 for (ct = cache->mfc_un.res.maxvif - 1;
1891 ct >= cache->mfc_un.res.minvif; ct--) {
1892 /* For (*,G) entry, don't forward to the incoming interface */
1893 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1895 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1897 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1900 ipmr_queue_xmit(net, mrt, skb2, cache,
1909 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1912 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1914 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1924 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1926 struct rtable *rt = skb_rtable(skb);
1927 struct iphdr *iph = ip_hdr(skb);
1928 struct flowi4 fl4 = {
1929 .daddr = iph->daddr,
1930 .saddr = iph->saddr,
1931 .flowi4_tos = RT_TOS(iph->tos),
1932 .flowi4_oif = (rt_is_output_route(rt) ?
1933 skb->dev->ifindex : 0),
1934 .flowi4_iif = (rt_is_output_route(rt) ?
1937 .flowi4_mark = skb->mark,
1939 struct mr_table *mrt;
1942 err = ipmr_fib_lookup(net, &fl4, &mrt);
1944 return ERR_PTR(err);
1949 * Multicast packets for forwarding arrive here
1950 * Called with rcu_read_lock();
1953 int ip_mr_input(struct sk_buff *skb)
1955 struct mfc_cache *cache;
1956 struct net *net = dev_net(skb->dev);
1957 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1958 struct mr_table *mrt;
1960 /* Packet is looped back after forward, it should not be
1961 * forwarded second time, but still can be delivered locally.
1963 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1966 mrt = ipmr_rt_fib_lookup(net, skb);
1969 return PTR_ERR(mrt);
1972 if (IPCB(skb)->opt.router_alert) {
1973 if (ip_call_ra_chain(skb))
1975 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1976 /* IGMPv1 (and broken IGMPv2 implementations sort of
1977 * Cisco IOS <= 11.2(8)) do not put router alert
1978 * option to IGMP packets destined to routable
1979 * groups. It is very bad, because it means
1980 * that we can forward NO IGMP messages.
1982 struct sock *mroute_sk;
1984 mroute_sk = rcu_dereference(mrt->mroute_sk);
1987 raw_rcv(mroute_sk, skb);
1993 /* already under rcu_read_lock() */
1994 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1996 int vif = ipmr_find_vif(mrt, skb->dev);
1999 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2004 * No usable cache entry
2010 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2011 ip_local_deliver(skb);
2017 read_lock(&mrt_lock);
2018 vif = ipmr_find_vif(mrt, skb->dev);
2020 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2021 read_unlock(&mrt_lock);
2025 read_unlock(&mrt_lock);
2030 read_lock(&mrt_lock);
2031 ip_mr_forward(net, mrt, skb, cache, local);
2032 read_unlock(&mrt_lock);
2035 return ip_local_deliver(skb);
2041 return ip_local_deliver(skb);
2046 #ifdef CONFIG_IP_PIMSM
2047 /* called with rcu_read_lock() */
2048 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2049 unsigned int pimlen)
2051 struct net_device *reg_dev = NULL;
2052 struct iphdr *encap;
2054 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2057 * a. packet is really sent to a multicast group
2058 * b. packet is not a NULL-REGISTER
2059 * c. packet is not truncated
2061 if (!ipv4_is_multicast(encap->daddr) ||
2062 encap->tot_len == 0 ||
2063 ntohs(encap->tot_len) + pimlen > skb->len)
2066 read_lock(&mrt_lock);
2067 if (mrt->mroute_reg_vif_num >= 0)
2068 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2069 read_unlock(&mrt_lock);
2074 skb->mac_header = skb->network_header;
2075 skb_pull(skb, (u8 *)encap - skb->data);
2076 skb_reset_network_header(skb);
2077 skb->protocol = htons(ETH_P_IP);
2078 skb->ip_summed = CHECKSUM_NONE;
2080 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2084 return NET_RX_SUCCESS;
2088 #ifdef CONFIG_IP_PIMSM_V1
2090 * Handle IGMP messages of PIMv1
2093 int pim_rcv_v1(struct sk_buff *skb)
2095 struct igmphdr *pim;
2096 struct net *net = dev_net(skb->dev);
2097 struct mr_table *mrt;
2099 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2102 pim = igmp_hdr(skb);
2104 mrt = ipmr_rt_fib_lookup(net, skb);
2107 if (!mrt->mroute_do_pim ||
2108 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2111 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2119 #ifdef CONFIG_IP_PIMSM_V2
2120 static int pim_rcv(struct sk_buff *skb)
2122 struct pimreghdr *pim;
2123 struct net *net = dev_net(skb->dev);
2124 struct mr_table *mrt;
2126 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2129 pim = (struct pimreghdr *)skb_transport_header(skb);
2130 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2131 (pim->flags & PIM_NULL_REGISTER) ||
2132 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2133 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2136 mrt = ipmr_rt_fib_lookup(net, skb);
2139 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2147 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2148 struct mfc_cache *c, struct rtmsg *rtm)
2151 struct rtnexthop *nhp;
2152 struct nlattr *mp_attr;
2153 struct rta_mfc_stats mfcs;
2155 /* If cache is unresolved, don't try to parse IIF and OIF */
2156 if (c->mfc_parent >= MAXVIFS)
2159 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2160 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2163 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2166 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2167 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2168 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2169 nla_nest_cancel(skb, mp_attr);
2173 nhp->rtnh_flags = 0;
2174 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2175 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2176 nhp->rtnh_len = sizeof(*nhp);
2180 nla_nest_end(skb, mp_attr);
2182 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2183 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2184 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2185 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2188 rtm->rtm_type = RTN_MULTICAST;
2192 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2193 __be32 saddr, __be32 daddr,
2194 struct rtmsg *rtm, int nowait)
2196 struct mfc_cache *cache;
2197 struct mr_table *mrt;
2200 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2205 cache = ipmr_cache_find(mrt, saddr, daddr);
2206 if (!cache && skb->dev) {
2207 int vif = ipmr_find_vif(mrt, skb->dev);
2210 cache = ipmr_cache_find_any(mrt, daddr, vif);
2213 struct sk_buff *skb2;
2215 struct net_device *dev;
2224 read_lock(&mrt_lock);
2226 vif = ipmr_find_vif(mrt, dev);
2228 read_unlock(&mrt_lock);
2232 skb2 = skb_clone(skb, GFP_ATOMIC);
2234 read_unlock(&mrt_lock);
2239 skb_push(skb2, sizeof(struct iphdr));
2240 skb_reset_network_header(skb2);
2242 iph->ihl = sizeof(struct iphdr) >> 2;
2246 err = ipmr_cache_unresolved(mrt, vif, skb2);
2247 read_unlock(&mrt_lock);
2252 read_lock(&mrt_lock);
2253 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2254 cache->mfc_flags |= MFC_NOTIFY;
2255 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2256 read_unlock(&mrt_lock);
2261 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2262 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2265 struct nlmsghdr *nlh;
2269 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2273 rtm = nlmsg_data(nlh);
2274 rtm->rtm_family = RTNL_FAMILY_IPMR;
2275 rtm->rtm_dst_len = 32;
2276 rtm->rtm_src_len = 32;
2278 rtm->rtm_table = mrt->id;
2279 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2280 goto nla_put_failure;
2281 rtm->rtm_type = RTN_MULTICAST;
2282 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2283 if (c->mfc_flags & MFC_STATIC)
2284 rtm->rtm_protocol = RTPROT_STATIC;
2286 rtm->rtm_protocol = RTPROT_MROUTED;
2289 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2290 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2291 goto nla_put_failure;
2292 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2293 /* do not break the dump if cache is unresolved */
2294 if (err < 0 && err != -ENOENT)
2295 goto nla_put_failure;
2297 nlmsg_end(skb, nlh);
2301 nlmsg_cancel(skb, nlh);
2305 static size_t mroute_msgsize(bool unresolved, int maxvif)
2308 NLMSG_ALIGN(sizeof(struct rtmsg))
2309 + nla_total_size(4) /* RTA_TABLE */
2310 + nla_total_size(4) /* RTA_SRC */
2311 + nla_total_size(4) /* RTA_DST */
2316 + nla_total_size(4) /* RTA_IIF */
2317 + nla_total_size(0) /* RTA_MULTIPATH */
2318 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2320 + nla_total_size(sizeof(struct rta_mfc_stats))
2326 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2329 struct net *net = read_pnet(&mrt->net);
2330 struct sk_buff *skb;
2333 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2338 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2342 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2348 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2351 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2353 struct net *net = sock_net(skb->sk);
2354 struct mr_table *mrt;
2355 struct mfc_cache *mfc;
2356 unsigned int t = 0, s_t;
2357 unsigned int h = 0, s_h;
2358 unsigned int e = 0, s_e;
2365 ipmr_for_each_table(mrt, net) {
2370 for (h = s_h; h < MFC_LINES; h++) {
2371 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2374 if (ipmr_fill_mroute(mrt, skb,
2375 NETLINK_CB(cb->skb).portid,
2385 spin_lock_bh(&mfc_unres_lock);
2386 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2389 if (ipmr_fill_mroute(mrt, skb,
2390 NETLINK_CB(cb->skb).portid,
2394 spin_unlock_bh(&mfc_unres_lock);
2400 spin_unlock_bh(&mfc_unres_lock);
2416 #ifdef CONFIG_PROC_FS
2418 * The /proc interfaces to multicast routing :
2419 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2421 struct ipmr_vif_iter {
2422 struct seq_net_private p;
2423 struct mr_table *mrt;
2427 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2428 struct ipmr_vif_iter *iter,
2431 struct mr_table *mrt = iter->mrt;
2433 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2434 if (!VIF_EXISTS(mrt, iter->ct))
2437 return &mrt->vif_table[iter->ct];
2442 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2443 __acquires(mrt_lock)
2445 struct ipmr_vif_iter *iter = seq->private;
2446 struct net *net = seq_file_net(seq);
2447 struct mr_table *mrt;
2449 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2451 return ERR_PTR(-ENOENT);
2455 read_lock(&mrt_lock);
2456 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2460 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2462 struct ipmr_vif_iter *iter = seq->private;
2463 struct net *net = seq_file_net(seq);
2464 struct mr_table *mrt = iter->mrt;
2467 if (v == SEQ_START_TOKEN)
2468 return ipmr_vif_seq_idx(net, iter, 0);
2470 while (++iter->ct < mrt->maxvif) {
2471 if (!VIF_EXISTS(mrt, iter->ct))
2473 return &mrt->vif_table[iter->ct];
2478 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2479 __releases(mrt_lock)
2481 read_unlock(&mrt_lock);
2484 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2486 struct ipmr_vif_iter *iter = seq->private;
2487 struct mr_table *mrt = iter->mrt;
2489 if (v == SEQ_START_TOKEN) {
2491 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2493 const struct vif_device *vif = v;
2494 const char *name = vif->dev ? vif->dev->name : "none";
2497 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2498 vif - mrt->vif_table,
2499 name, vif->bytes_in, vif->pkt_in,
2500 vif->bytes_out, vif->pkt_out,
2501 vif->flags, vif->local, vif->remote);
2506 static const struct seq_operations ipmr_vif_seq_ops = {
2507 .start = ipmr_vif_seq_start,
2508 .next = ipmr_vif_seq_next,
2509 .stop = ipmr_vif_seq_stop,
2510 .show = ipmr_vif_seq_show,
2513 static int ipmr_vif_open(struct inode *inode, struct file *file)
2515 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2516 sizeof(struct ipmr_vif_iter));
2519 static const struct file_operations ipmr_vif_fops = {
2520 .owner = THIS_MODULE,
2521 .open = ipmr_vif_open,
2523 .llseek = seq_lseek,
2524 .release = seq_release_net,
2527 struct ipmr_mfc_iter {
2528 struct seq_net_private p;
2529 struct mr_table *mrt;
2530 struct list_head *cache;
2535 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2536 struct ipmr_mfc_iter *it, loff_t pos)
2538 struct mr_table *mrt = it->mrt;
2539 struct mfc_cache *mfc;
2542 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2543 it->cache = &mrt->mfc_cache_array[it->ct];
2544 list_for_each_entry_rcu(mfc, it->cache, list)
2550 spin_lock_bh(&mfc_unres_lock);
2551 it->cache = &mrt->mfc_unres_queue;
2552 list_for_each_entry(mfc, it->cache, list)
2555 spin_unlock_bh(&mfc_unres_lock);
2562 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2564 struct ipmr_mfc_iter *it = seq->private;
2565 struct net *net = seq_file_net(seq);
2566 struct mr_table *mrt;
2568 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2570 return ERR_PTR(-ENOENT);
2575 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2579 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2581 struct mfc_cache *mfc = v;
2582 struct ipmr_mfc_iter *it = seq->private;
2583 struct net *net = seq_file_net(seq);
2584 struct mr_table *mrt = it->mrt;
2588 if (v == SEQ_START_TOKEN)
2589 return ipmr_mfc_seq_idx(net, seq->private, 0);
2591 if (mfc->list.next != it->cache)
2592 return list_entry(mfc->list.next, struct mfc_cache, list);
2594 if (it->cache == &mrt->mfc_unres_queue)
2597 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2599 while (++it->ct < MFC_LINES) {
2600 it->cache = &mrt->mfc_cache_array[it->ct];
2601 if (list_empty(it->cache))
2603 return list_first_entry(it->cache, struct mfc_cache, list);
2606 /* exhausted cache_array, show unresolved */
2608 it->cache = &mrt->mfc_unres_queue;
2611 spin_lock_bh(&mfc_unres_lock);
2612 if (!list_empty(it->cache))
2613 return list_first_entry(it->cache, struct mfc_cache, list);
2616 spin_unlock_bh(&mfc_unres_lock);
2622 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2624 struct ipmr_mfc_iter *it = seq->private;
2625 struct mr_table *mrt = it->mrt;
2627 if (it->cache == &mrt->mfc_unres_queue)
2628 spin_unlock_bh(&mfc_unres_lock);
2629 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2633 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2637 if (v == SEQ_START_TOKEN) {
2639 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2641 const struct mfc_cache *mfc = v;
2642 const struct ipmr_mfc_iter *it = seq->private;
2643 const struct mr_table *mrt = it->mrt;
2645 seq_printf(seq, "%08X %08X %-3hd",
2646 (__force u32) mfc->mfc_mcastgrp,
2647 (__force u32) mfc->mfc_origin,
2650 if (it->cache != &mrt->mfc_unres_queue) {
2651 seq_printf(seq, " %8lu %8lu %8lu",
2652 mfc->mfc_un.res.pkt,
2653 mfc->mfc_un.res.bytes,
2654 mfc->mfc_un.res.wrong_if);
2655 for (n = mfc->mfc_un.res.minvif;
2656 n < mfc->mfc_un.res.maxvif; n++) {
2657 if (VIF_EXISTS(mrt, n) &&
2658 mfc->mfc_un.res.ttls[n] < 255)
2661 n, mfc->mfc_un.res.ttls[n]);
2664 /* unresolved mfc_caches don't contain
2665 * pkt, bytes and wrong_if values
2667 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2669 seq_putc(seq, '\n');
2674 static const struct seq_operations ipmr_mfc_seq_ops = {
2675 .start = ipmr_mfc_seq_start,
2676 .next = ipmr_mfc_seq_next,
2677 .stop = ipmr_mfc_seq_stop,
2678 .show = ipmr_mfc_seq_show,
2681 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2683 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2684 sizeof(struct ipmr_mfc_iter));
2687 static const struct file_operations ipmr_mfc_fops = {
2688 .owner = THIS_MODULE,
2689 .open = ipmr_mfc_open,
2691 .llseek = seq_lseek,
2692 .release = seq_release_net,
2696 #ifdef CONFIG_IP_PIMSM_V2
2697 static const struct net_protocol pim_protocol = {
2705 * Setup for IP multicast routing
2707 static int __net_init ipmr_net_init(struct net *net)
2711 err = ipmr_rules_init(net);
2715 #ifdef CONFIG_PROC_FS
2717 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2719 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2720 goto proc_cache_fail;
2724 #ifdef CONFIG_PROC_FS
2726 remove_proc_entry("ip_mr_vif", net->proc_net);
2728 ipmr_rules_exit(net);
2734 static void __net_exit ipmr_net_exit(struct net *net)
2736 #ifdef CONFIG_PROC_FS
2737 remove_proc_entry("ip_mr_cache", net->proc_net);
2738 remove_proc_entry("ip_mr_vif", net->proc_net);
2740 ipmr_rules_exit(net);
2743 static struct pernet_operations ipmr_net_ops = {
2744 .init = ipmr_net_init,
2745 .exit = ipmr_net_exit,
2748 int __init ip_mr_init(void)
2752 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2753 sizeof(struct mfc_cache),
2754 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2759 err = register_pernet_subsys(&ipmr_net_ops);
2761 goto reg_pernet_fail;
2763 err = register_netdevice_notifier(&ip_mr_notifier);
2765 goto reg_notif_fail;
2766 #ifdef CONFIG_IP_PIMSM_V2
2767 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2768 pr_err("%s: can't add PIM protocol\n", __func__);
2770 goto add_proto_fail;
2773 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2774 NULL, ipmr_rtm_dumproute, NULL);
2777 #ifdef CONFIG_IP_PIMSM_V2
2779 unregister_netdevice_notifier(&ip_mr_notifier);
2782 unregister_pernet_subsys(&ipmr_net_ops);
2784 kmem_cache_destroy(mrt_cachep);