2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
6 * LSIIT Laboratory, Strasbourg, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
19 #include <asm/system.h>
20 #include <asm/uaccess.h>
21 #include <linux/types.h>
22 #include <linux/sched.h>
23 #include <linux/errno.h>
24 #include <linux/timer.h>
26 #include <linux/kernel.h>
27 #include <linux/fcntl.h>
28 #include <linux/stat.h>
29 #include <linux/socket.h>
30 #include <linux/inet.h>
31 #include <linux/netdevice.h>
32 #include <linux/inetdevice.h>
33 #include <linux/proc_fs.h>
34 #include <linux/seq_file.h>
35 #include <linux/init.h>
36 #include <linux/slab.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
41 #include <linux/notifier.h>
42 #include <linux/if_arp.h>
43 #include <net/checksum.h>
44 #include <net/netlink.h>
45 #include <net/fib_rules.h>
48 #include <net/ip6_route.h>
49 #include <linux/mroute6.h>
50 #include <linux/pim.h>
51 #include <net/addrconf.h>
52 #include <linux/netfilter_ipv6.h>
53 #include <net/ip6_checksum.h>
56 struct list_head list;
61 struct sock *mroute6_sk;
62 struct timer_list ipmr_expire_timer;
63 struct list_head mfc6_unres_queue;
64 struct list_head mfc6_cache_array[MFC6_LINES];
65 struct mif_device vif6_table[MAXMIFS];
67 atomic_t cache_resolve_queue_len;
70 #ifdef CONFIG_IPV6_PIMSM_V2
71 int mroute_reg_vif_num;
76 struct fib_rule common;
80 struct mr6_table *mrt;
83 /* Big lock, protecting vif table, mrt cache and mroute socket state.
84 Note that the changes are semaphored via rtnl_lock.
87 static DEFINE_RWLOCK(mrt_lock);
90 * Multicast router control variables
93 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
95 /* Special spinlock for queue of unresolved entries */
96 static DEFINE_SPINLOCK(mfc_unres_lock);
98 /* We return to original Alan's scheme. Hash table of resolved
99 entries is changed only in process context and protected
100 with weak lock mrt_lock. Queue of unresolved entries is protected
101 with strong spinlock mfc_unres_lock.
103 In this case data path is free of exclusive locks at all.
106 static struct kmem_cache *mrt_cachep __read_mostly;
108 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
109 static void ip6mr_free_table(struct mr6_table *mrt);
111 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
112 struct sk_buff *skb, struct mfc6_cache *cache);
113 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
114 mifi_t mifi, int assert);
115 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
116 struct mfc6_cache *c, struct rtmsg *rtm);
117 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
118 struct netlink_callback *cb);
119 static void mroute_clean_tables(struct mr6_table *mrt);
120 static void ipmr_expire_process(unsigned long arg);
122 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
123 #define ip6mr_for_each_table(mrt, net) \
124 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
126 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
128 struct mr6_table *mrt;
130 ip6mr_for_each_table(mrt, net) {
137 static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
138 struct mr6_table **mrt)
140 struct ip6mr_result res;
141 struct fib_lookup_arg arg = { .result = &res, };
144 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, flp, 0, &arg);
151 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
152 int flags, struct fib_lookup_arg *arg)
154 struct ip6mr_result *res = arg->result;
155 struct mr6_table *mrt;
157 switch (rule->action) {
160 case FR_ACT_UNREACHABLE:
162 case FR_ACT_PROHIBIT:
164 case FR_ACT_BLACKHOLE:
169 mrt = ip6mr_get_table(rule->fr_net, rule->table);
176 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
181 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
185 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
186 struct fib_rule_hdr *frh, struct nlattr **tb)
191 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
197 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
198 struct fib_rule_hdr *frh)
206 static const struct fib_rules_ops __net_initdata ip6mr_rules_ops_template = {
207 .family = RTNL_FAMILY_IP6MR,
208 .rule_size = sizeof(struct ip6mr_rule),
209 .addr_size = sizeof(struct in6_addr),
210 .action = ip6mr_rule_action,
211 .match = ip6mr_rule_match,
212 .configure = ip6mr_rule_configure,
213 .compare = ip6mr_rule_compare,
214 .default_pref = fib_default_rule_pref,
215 .fill = ip6mr_rule_fill,
216 .nlgroup = RTNLGRP_IPV6_RULE,
217 .policy = ip6mr_rule_policy,
218 .owner = THIS_MODULE,
221 static int __net_init ip6mr_rules_init(struct net *net)
223 struct fib_rules_ops *ops;
224 struct mr6_table *mrt;
227 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
231 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
233 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
239 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
243 net->ipv6.mr6_rules_ops = ops;
249 fib_rules_unregister(ops);
253 static void __net_exit ip6mr_rules_exit(struct net *net)
255 struct mr6_table *mrt, *next;
257 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
258 list_del(&mrt->list);
259 ip6mr_free_table(mrt);
261 fib_rules_unregister(net->ipv6.mr6_rules_ops);
264 #define ip6mr_for_each_table(mrt, net) \
265 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
267 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
269 return net->ipv6.mrt6;
272 static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
273 struct mr6_table **mrt)
275 *mrt = net->ipv6.mrt6;
279 static int __net_init ip6mr_rules_init(struct net *net)
281 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
282 return net->ipv6.mrt6 ? 0 : -ENOMEM;
285 static void __net_exit ip6mr_rules_exit(struct net *net)
287 ip6mr_free_table(net->ipv6.mrt6);
291 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
293 struct mr6_table *mrt;
296 mrt = ip6mr_get_table(net, id);
300 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
304 write_pnet(&mrt->net, net);
306 /* Forwarding cache */
307 for (i = 0; i < MFC6_LINES; i++)
308 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
310 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
312 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
315 #ifdef CONFIG_IPV6_PIMSM_V2
316 mrt->mroute_reg_vif_num = -1;
318 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
319 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
324 static void ip6mr_free_table(struct mr6_table *mrt)
326 del_timer(&mrt->ipmr_expire_timer);
327 mroute_clean_tables(mrt);
331 #ifdef CONFIG_PROC_FS
333 struct ipmr_mfc_iter {
334 struct seq_net_private p;
335 struct mr6_table *mrt;
336 struct list_head *cache;
341 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
342 struct ipmr_mfc_iter *it, loff_t pos)
344 struct mr6_table *mrt = it->mrt;
345 struct mfc6_cache *mfc;
347 read_lock(&mrt_lock);
348 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
349 it->cache = &mrt->mfc6_cache_array[it->ct];
350 list_for_each_entry(mfc, it->cache, list)
354 read_unlock(&mrt_lock);
356 spin_lock_bh(&mfc_unres_lock);
357 it->cache = &mrt->mfc6_unres_queue;
358 list_for_each_entry(mfc, it->cache, list)
361 spin_unlock_bh(&mfc_unres_lock);
368 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
371 struct ipmr_vif_iter {
372 struct seq_net_private p;
373 struct mr6_table *mrt;
377 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
378 struct ipmr_vif_iter *iter,
381 struct mr6_table *mrt = iter->mrt;
383 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
384 if (!MIF_EXISTS(mrt, iter->ct))
387 return &mrt->vif6_table[iter->ct];
392 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
395 struct ipmr_vif_iter *iter = seq->private;
396 struct net *net = seq_file_net(seq);
397 struct mr6_table *mrt;
399 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
401 return ERR_PTR(-ENOENT);
405 read_lock(&mrt_lock);
406 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
410 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
412 struct ipmr_vif_iter *iter = seq->private;
413 struct net *net = seq_file_net(seq);
414 struct mr6_table *mrt = iter->mrt;
417 if (v == SEQ_START_TOKEN)
418 return ip6mr_vif_seq_idx(net, iter, 0);
420 while (++iter->ct < mrt->maxvif) {
421 if (!MIF_EXISTS(mrt, iter->ct))
423 return &mrt->vif6_table[iter->ct];
428 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
431 read_unlock(&mrt_lock);
434 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
436 struct ipmr_vif_iter *iter = seq->private;
437 struct mr6_table *mrt = iter->mrt;
439 if (v == SEQ_START_TOKEN) {
441 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
443 const struct mif_device *vif = v;
444 const char *name = vif->dev ? vif->dev->name : "none";
447 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
448 vif - mrt->vif6_table,
449 name, vif->bytes_in, vif->pkt_in,
450 vif->bytes_out, vif->pkt_out,
456 static const struct seq_operations ip6mr_vif_seq_ops = {
457 .start = ip6mr_vif_seq_start,
458 .next = ip6mr_vif_seq_next,
459 .stop = ip6mr_vif_seq_stop,
460 .show = ip6mr_vif_seq_show,
463 static int ip6mr_vif_open(struct inode *inode, struct file *file)
465 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
466 sizeof(struct ipmr_vif_iter));
469 static const struct file_operations ip6mr_vif_fops = {
470 .owner = THIS_MODULE,
471 .open = ip6mr_vif_open,
474 .release = seq_release_net,
477 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
479 struct ipmr_mfc_iter *it = seq->private;
480 struct net *net = seq_file_net(seq);
481 struct mr6_table *mrt;
483 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
485 return ERR_PTR(-ENOENT);
488 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
492 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
494 struct mfc6_cache *mfc = v;
495 struct ipmr_mfc_iter *it = seq->private;
496 struct net *net = seq_file_net(seq);
497 struct mr6_table *mrt = it->mrt;
501 if (v == SEQ_START_TOKEN)
502 return ipmr_mfc_seq_idx(net, seq->private, 0);
504 if (mfc->list.next != it->cache)
505 return list_entry(mfc->list.next, struct mfc6_cache, list);
507 if (it->cache == &mrt->mfc6_unres_queue)
510 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
512 while (++it->ct < MFC6_LINES) {
513 it->cache = &mrt->mfc6_cache_array[it->ct];
514 if (list_empty(it->cache))
516 return list_first_entry(it->cache, struct mfc6_cache, list);
519 /* exhausted cache_array, show unresolved */
520 read_unlock(&mrt_lock);
521 it->cache = &mrt->mfc6_unres_queue;
524 spin_lock_bh(&mfc_unres_lock);
525 if (!list_empty(it->cache))
526 return list_first_entry(it->cache, struct mfc6_cache, list);
529 spin_unlock_bh(&mfc_unres_lock);
535 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
537 struct ipmr_mfc_iter *it = seq->private;
538 struct mr6_table *mrt = it->mrt;
540 if (it->cache == &mrt->mfc6_unres_queue)
541 spin_unlock_bh(&mfc_unres_lock);
542 else if (it->cache == mrt->mfc6_cache_array)
543 read_unlock(&mrt_lock);
546 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
550 if (v == SEQ_START_TOKEN) {
554 "Iif Pkts Bytes Wrong Oifs\n");
556 const struct mfc6_cache *mfc = v;
557 const struct ipmr_mfc_iter *it = seq->private;
558 struct mr6_table *mrt = it->mrt;
560 seq_printf(seq, "%pI6 %pI6 %-3hd",
561 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
564 if (it->cache != &mrt->mfc6_unres_queue) {
565 seq_printf(seq, " %8lu %8lu %8lu",
567 mfc->mfc_un.res.bytes,
568 mfc->mfc_un.res.wrong_if);
569 for (n = mfc->mfc_un.res.minvif;
570 n < mfc->mfc_un.res.maxvif; n++) {
571 if (MIF_EXISTS(mrt, n) &&
572 mfc->mfc_un.res.ttls[n] < 255)
575 n, mfc->mfc_un.res.ttls[n]);
578 /* unresolved mfc_caches don't contain
579 * pkt, bytes and wrong_if values
581 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
588 static const struct seq_operations ipmr_mfc_seq_ops = {
589 .start = ipmr_mfc_seq_start,
590 .next = ipmr_mfc_seq_next,
591 .stop = ipmr_mfc_seq_stop,
592 .show = ipmr_mfc_seq_show,
595 static int ipmr_mfc_open(struct inode *inode, struct file *file)
597 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
598 sizeof(struct ipmr_mfc_iter));
601 static const struct file_operations ip6mr_mfc_fops = {
602 .owner = THIS_MODULE,
603 .open = ipmr_mfc_open,
606 .release = seq_release_net,
610 #ifdef CONFIG_IPV6_PIMSM_V2
612 static int pim6_rcv(struct sk_buff *skb)
614 struct pimreghdr *pim;
615 struct ipv6hdr *encap;
616 struct net_device *reg_dev = NULL;
617 struct net *net = dev_net(skb->dev);
618 struct mr6_table *mrt;
620 .iif = skb->dev->ifindex,
625 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
628 pim = (struct pimreghdr *)skb_transport_header(skb);
629 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
630 (pim->flags & PIM_NULL_REGISTER) ||
631 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
632 sizeof(*pim), IPPROTO_PIM,
633 csum_partial((void *)pim, sizeof(*pim), 0)) &&
634 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
637 /* check if the inner packet is destined to mcast group */
638 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
641 if (!ipv6_addr_is_multicast(&encap->daddr) ||
642 encap->payload_len == 0 ||
643 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
646 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
648 reg_vif_num = mrt->mroute_reg_vif_num;
650 read_lock(&mrt_lock);
651 if (reg_vif_num >= 0)
652 reg_dev = mrt->vif6_table[reg_vif_num].dev;
655 read_unlock(&mrt_lock);
660 skb->mac_header = skb->network_header;
661 skb_pull(skb, (u8 *)encap - skb->data);
662 skb_reset_network_header(skb);
663 skb->protocol = htons(ETH_P_IPV6);
665 skb->pkt_type = PACKET_HOST;
667 skb_tunnel_rx(skb, reg_dev);
669 if (netif_rx(skb) == NET_RX_DROP)
670 reg_dev->stats.rx_dropped++;
679 static const struct inet6_protocol pim6_protocol = {
683 /* Service routines creating virtual interfaces: PIMREG */
685 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
686 struct net_device *dev)
688 struct net *net = dev_net(dev);
689 struct mr6_table *mrt;
697 err = ip6mr_fib_lookup(net, &fl, &mrt);
701 read_lock(&mrt_lock);
702 dev->stats.tx_bytes += skb->len;
703 dev->stats.tx_packets++;
704 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
705 read_unlock(&mrt_lock);
710 static const struct net_device_ops reg_vif_netdev_ops = {
711 .ndo_start_xmit = reg_vif_xmit,
714 static void reg_vif_setup(struct net_device *dev)
716 dev->type = ARPHRD_PIMREG;
717 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
718 dev->flags = IFF_NOARP;
719 dev->netdev_ops = ®_vif_netdev_ops;
720 dev->destructor = free_netdev;
721 dev->features |= NETIF_F_NETNS_LOCAL;
724 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
726 struct net_device *dev;
729 if (mrt->id == RT6_TABLE_DFLT)
730 sprintf(name, "pim6reg");
732 sprintf(name, "pim6reg%u", mrt->id);
734 dev = alloc_netdev(0, name, reg_vif_setup);
738 dev_net_set(dev, net);
740 if (register_netdevice(dev)) {
753 /* allow the register to be completed before unregistering. */
757 unregister_netdevice(dev);
766 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
768 struct mif_device *v;
769 struct net_device *dev;
770 struct inet6_dev *in6_dev;
772 if (vifi < 0 || vifi >= mrt->maxvif)
773 return -EADDRNOTAVAIL;
775 v = &mrt->vif6_table[vifi];
777 write_lock_bh(&mrt_lock);
782 write_unlock_bh(&mrt_lock);
783 return -EADDRNOTAVAIL;
786 #ifdef CONFIG_IPV6_PIMSM_V2
787 if (vifi == mrt->mroute_reg_vif_num)
788 mrt->mroute_reg_vif_num = -1;
791 if (vifi + 1 == mrt->maxvif) {
793 for (tmp = vifi - 1; tmp >= 0; tmp--) {
794 if (MIF_EXISTS(mrt, tmp))
797 mrt->maxvif = tmp + 1;
800 write_unlock_bh(&mrt_lock);
802 dev_set_allmulti(dev, -1);
804 in6_dev = __in6_dev_get(dev);
806 in6_dev->cnf.mc_forwarding--;
808 if (v->flags & MIFF_REGISTER)
809 unregister_netdevice_queue(dev, head);
815 static inline void ip6mr_cache_free(struct mfc6_cache *c)
817 kmem_cache_free(mrt_cachep, c);
820 /* Destroy an unresolved cache entry, killing queued skbs
821 and reporting error to netlink readers.
824 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
826 struct net *net = read_pnet(&mrt->net);
829 atomic_dec(&mrt->cache_resolve_queue_len);
831 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
832 if (ipv6_hdr(skb)->version == 0) {
833 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
834 nlh->nlmsg_type = NLMSG_ERROR;
835 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
836 skb_trim(skb, nlh->nlmsg_len);
837 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
838 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
847 /* Timer process for all the unresolved queue. */
849 static void ipmr_do_expire_process(struct mr6_table *mrt)
851 unsigned long now = jiffies;
852 unsigned long expires = 10 * HZ;
853 struct mfc6_cache *c, *next;
855 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
856 if (time_after(c->mfc_un.unres.expires, now)) {
858 unsigned long interval = c->mfc_un.unres.expires - now;
859 if (interval < expires)
865 ip6mr_destroy_unres(mrt, c);
868 if (!list_empty(&mrt->mfc6_unres_queue))
869 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
872 static void ipmr_expire_process(unsigned long arg)
874 struct mr6_table *mrt = (struct mr6_table *)arg;
876 if (!spin_trylock(&mfc_unres_lock)) {
877 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
881 if (!list_empty(&mrt->mfc6_unres_queue))
882 ipmr_do_expire_process(mrt);
884 spin_unlock(&mfc_unres_lock);
887 /* Fill oifs list. It is called under write locked mrt_lock. */
889 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
894 cache->mfc_un.res.minvif = MAXMIFS;
895 cache->mfc_un.res.maxvif = 0;
896 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
898 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
899 if (MIF_EXISTS(mrt, vifi) &&
900 ttls[vifi] && ttls[vifi] < 255) {
901 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
902 if (cache->mfc_un.res.minvif > vifi)
903 cache->mfc_un.res.minvif = vifi;
904 if (cache->mfc_un.res.maxvif <= vifi)
905 cache->mfc_un.res.maxvif = vifi + 1;
910 static int mif6_add(struct net *net, struct mr6_table *mrt,
911 struct mif6ctl *vifc, int mrtsock)
913 int vifi = vifc->mif6c_mifi;
914 struct mif_device *v = &mrt->vif6_table[vifi];
915 struct net_device *dev;
916 struct inet6_dev *in6_dev;
920 if (MIF_EXISTS(mrt, vifi))
923 switch (vifc->mif6c_flags) {
924 #ifdef CONFIG_IPV6_PIMSM_V2
927 * Special Purpose VIF in PIM
928 * All the packets will be sent to the daemon
930 if (mrt->mroute_reg_vif_num >= 0)
932 dev = ip6mr_reg_vif(net, mrt);
935 err = dev_set_allmulti(dev, 1);
937 unregister_netdevice(dev);
944 dev = dev_get_by_index(net, vifc->mif6c_pifi);
946 return -EADDRNOTAVAIL;
947 err = dev_set_allmulti(dev, 1);
957 in6_dev = __in6_dev_get(dev);
959 in6_dev->cnf.mc_forwarding++;
962 * Fill in the VIF structures
964 v->rate_limit = vifc->vifc_rate_limit;
965 v->flags = vifc->mif6c_flags;
967 v->flags |= VIFF_STATIC;
968 v->threshold = vifc->vifc_threshold;
973 v->link = dev->ifindex;
974 if (v->flags & MIFF_REGISTER)
975 v->link = dev->iflink;
977 /* And finish update writing critical data */
978 write_lock_bh(&mrt_lock);
980 #ifdef CONFIG_IPV6_PIMSM_V2
981 if (v->flags & MIFF_REGISTER)
982 mrt->mroute_reg_vif_num = vifi;
984 if (vifi + 1 > mrt->maxvif)
985 mrt->maxvif = vifi + 1;
986 write_unlock_bh(&mrt_lock);
990 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
991 struct in6_addr *origin,
992 struct in6_addr *mcastgrp)
994 int line = MFC6_HASH(mcastgrp, origin);
995 struct mfc6_cache *c;
997 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
998 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
999 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1006 * Allocate a multicast cache entry
1008 static struct mfc6_cache *ip6mr_cache_alloc(void)
1010 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1013 c->mfc_un.res.minvif = MAXMIFS;
1017 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1019 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1022 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1023 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1028 * A cache entry has gone into a resolved state from queued
1031 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1032 struct mfc6_cache *uc, struct mfc6_cache *c)
1034 struct sk_buff *skb;
1037 * Play the pending entries through our router
1040 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1041 if (ipv6_hdr(skb)->version == 0) {
1043 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1045 if (__ip6mr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) {
1046 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1048 nlh->nlmsg_type = NLMSG_ERROR;
1049 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
1050 skb_trim(skb, nlh->nlmsg_len);
1051 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
1053 err = rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
1055 ip6_mr_forward(net, mrt, skb, c);
1060 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1061 * expects the following bizarre scheme.
1063 * Called under mrt_lock.
1066 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1067 mifi_t mifi, int assert)
1069 struct sk_buff *skb;
1070 struct mrt6msg *msg;
1073 #ifdef CONFIG_IPV6_PIMSM_V2
1074 if (assert == MRT6MSG_WHOLEPKT)
1075 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1079 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1084 /* I suppose that internal messages
1085 * do not require checksums */
1087 skb->ip_summed = CHECKSUM_UNNECESSARY;
1089 #ifdef CONFIG_IPV6_PIMSM_V2
1090 if (assert == MRT6MSG_WHOLEPKT) {
1091 /* Ugly, but we have no choice with this interface.
1092 Duplicate old header, fix length etc.
1093 And all this only to mangle msg->im6_msgtype and
1094 to set msg->im6_mbz to "mbz" :-)
1096 skb_push(skb, -skb_network_offset(pkt));
1098 skb_push(skb, sizeof(*msg));
1099 skb_reset_transport_header(skb);
1100 msg = (struct mrt6msg *)skb_transport_header(skb);
1102 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1103 msg->im6_mif = mrt->mroute_reg_vif_num;
1105 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1106 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1108 skb->ip_summed = CHECKSUM_UNNECESSARY;
1113 * Copy the IP header
1116 skb_put(skb, sizeof(struct ipv6hdr));
1117 skb_reset_network_header(skb);
1118 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1123 skb_put(skb, sizeof(*msg));
1124 skb_reset_transport_header(skb);
1125 msg = (struct mrt6msg *)skb_transport_header(skb);
1128 msg->im6_msgtype = assert;
1129 msg->im6_mif = mifi;
1131 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1132 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1134 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1135 skb->ip_summed = CHECKSUM_UNNECESSARY;
1138 if (mrt->mroute6_sk == NULL) {
1144 * Deliver to user space multicast routing algorithms
1146 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1148 if (net_ratelimit())
1149 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
1157 * Queue a packet for resolution. It gets locked cache entry!
1161 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1165 struct mfc6_cache *c;
1167 spin_lock_bh(&mfc_unres_lock);
1168 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1169 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1170 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1178 * Create a new entry if allowable
1181 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1182 (c = ip6mr_cache_alloc_unres()) == NULL) {
1183 spin_unlock_bh(&mfc_unres_lock);
1190 * Fill in the new cache entry
1192 c->mf6c_parent = -1;
1193 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1194 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1197 * Reflect first query at pim6sd
1199 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1201 /* If the report failed throw the cache entry
1204 spin_unlock_bh(&mfc_unres_lock);
1206 ip6mr_cache_free(c);
1211 atomic_inc(&mrt->cache_resolve_queue_len);
1212 list_add(&c->list, &mrt->mfc6_unres_queue);
1214 ipmr_do_expire_process(mrt);
1218 * See if we can append the packet
1220 if (c->mfc_un.unres.unresolved.qlen > 3) {
1224 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1228 spin_unlock_bh(&mfc_unres_lock);
1233 * MFC6 cache manipulation by user space
1236 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc)
1239 struct mfc6_cache *c, *next;
1241 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1243 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1244 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1245 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1246 write_lock_bh(&mrt_lock);
1248 write_unlock_bh(&mrt_lock);
1250 ip6mr_cache_free(c);
1257 static int ip6mr_device_event(struct notifier_block *this,
1258 unsigned long event, void *ptr)
1260 struct net_device *dev = ptr;
1261 struct net *net = dev_net(dev);
1262 struct mr6_table *mrt;
1263 struct mif_device *v;
1267 if (event != NETDEV_UNREGISTER)
1270 ip6mr_for_each_table(mrt, net) {
1271 v = &mrt->vif6_table[0];
1272 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1274 mif6_delete(mrt, ct, &list);
1277 unregister_netdevice_many(&list);
1282 static struct notifier_block ip6_mr_notifier = {
1283 .notifier_call = ip6mr_device_event
1287 * Setup for IP multicast routing
1290 static int __net_init ip6mr_net_init(struct net *net)
1294 err = ip6mr_rules_init(net);
1298 #ifdef CONFIG_PROC_FS
1300 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
1302 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops))
1303 goto proc_cache_fail;
1308 #ifdef CONFIG_PROC_FS
1310 proc_net_remove(net, "ip6_mr_vif");
1312 ip6mr_rules_exit(net);
1318 static void __net_exit ip6mr_net_exit(struct net *net)
1320 #ifdef CONFIG_PROC_FS
1321 proc_net_remove(net, "ip6_mr_cache");
1322 proc_net_remove(net, "ip6_mr_vif");
1324 ip6mr_rules_exit(net);
1327 static struct pernet_operations ip6mr_net_ops = {
1328 .init = ip6mr_net_init,
1329 .exit = ip6mr_net_exit,
1332 int __init ip6_mr_init(void)
1336 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1337 sizeof(struct mfc6_cache),
1338 0, SLAB_HWCACHE_ALIGN,
1343 err = register_pernet_subsys(&ip6mr_net_ops);
1345 goto reg_pernet_fail;
1347 err = register_netdevice_notifier(&ip6_mr_notifier);
1349 goto reg_notif_fail;
1350 #ifdef CONFIG_IPV6_PIMSM_V2
1351 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1352 printk(KERN_ERR "ip6_mr_init: can't add PIM protocol\n");
1354 goto add_proto_fail;
1357 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, ip6mr_rtm_dumproute);
1359 #ifdef CONFIG_IPV6_PIMSM_V2
1361 unregister_netdevice_notifier(&ip6_mr_notifier);
1364 unregister_pernet_subsys(&ip6mr_net_ops);
1366 kmem_cache_destroy(mrt_cachep);
1370 void ip6_mr_cleanup(void)
1372 unregister_netdevice_notifier(&ip6_mr_notifier);
1373 unregister_pernet_subsys(&ip6mr_net_ops);
1374 kmem_cache_destroy(mrt_cachep);
1377 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1378 struct mf6cctl *mfc, int mrtsock)
1382 struct mfc6_cache *uc, *c;
1383 unsigned char ttls[MAXMIFS];
1386 if (mfc->mf6cc_parent >= MAXMIFS)
1389 memset(ttls, 255, MAXMIFS);
1390 for (i = 0; i < MAXMIFS; i++) {
1391 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1396 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1398 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1399 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1400 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1407 write_lock_bh(&mrt_lock);
1408 c->mf6c_parent = mfc->mf6cc_parent;
1409 ip6mr_update_thresholds(mrt, c, ttls);
1411 c->mfc_flags |= MFC_STATIC;
1412 write_unlock_bh(&mrt_lock);
1416 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1419 c = ip6mr_cache_alloc();
1423 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1424 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1425 c->mf6c_parent = mfc->mf6cc_parent;
1426 ip6mr_update_thresholds(mrt, c, ttls);
1428 c->mfc_flags |= MFC_STATIC;
1430 write_lock_bh(&mrt_lock);
1431 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1432 write_unlock_bh(&mrt_lock);
1435 * Check to see if we resolved a queued list. If so we
1436 * need to send on the frames and tidy up.
1439 spin_lock_bh(&mfc_unres_lock);
1440 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1441 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1442 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1443 list_del(&uc->list);
1444 atomic_dec(&mrt->cache_resolve_queue_len);
1449 if (list_empty(&mrt->mfc6_unres_queue))
1450 del_timer(&mrt->ipmr_expire_timer);
1451 spin_unlock_bh(&mfc_unres_lock);
1454 ip6mr_cache_resolve(net, mrt, uc, c);
1455 ip6mr_cache_free(uc);
1461 * Close the multicast socket, and clear the vif tables etc
1464 static void mroute_clean_tables(struct mr6_table *mrt)
1468 struct mfc6_cache *c, *next;
1471 * Shut down all active vif entries
1473 for (i = 0; i < mrt->maxvif; i++) {
1474 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1475 mif6_delete(mrt, i, &list);
1477 unregister_netdevice_many(&list);
1482 for (i = 0; i < MFC6_LINES; i++) {
1483 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1484 if (c->mfc_flags & MFC_STATIC)
1486 write_lock_bh(&mrt_lock);
1488 write_unlock_bh(&mrt_lock);
1490 ip6mr_cache_free(c);
1494 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1495 spin_lock_bh(&mfc_unres_lock);
1496 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1498 ip6mr_destroy_unres(mrt, c);
1500 spin_unlock_bh(&mfc_unres_lock);
1504 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1507 struct net *net = sock_net(sk);
1510 write_lock_bh(&mrt_lock);
1511 if (likely(mrt->mroute6_sk == NULL)) {
1512 mrt->mroute6_sk = sk;
1513 net->ipv6.devconf_all->mc_forwarding++;
1517 write_unlock_bh(&mrt_lock);
1524 int ip6mr_sk_done(struct sock *sk)
1527 struct net *net = sock_net(sk);
1528 struct mr6_table *mrt;
1531 ip6mr_for_each_table(mrt, net) {
1532 if (sk == mrt->mroute6_sk) {
1533 write_lock_bh(&mrt_lock);
1534 mrt->mroute6_sk = NULL;
1535 net->ipv6.devconf_all->mc_forwarding--;
1536 write_unlock_bh(&mrt_lock);
1538 mroute_clean_tables(mrt);
1548 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1550 struct mr6_table *mrt;
1552 .iif = skb->skb_iif,
1553 .oif = skb->dev->ifindex,
1557 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
1560 return mrt->mroute6_sk;
1564 * Socket options and virtual interface manipulation. The whole
1565 * virtual interface system is a complete heap, but unfortunately
1566 * that's how BSD mrouted happens to think. Maybe one day with a proper
1567 * MOSPF/PIM router set up we can clean this up.
1570 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1576 struct net *net = sock_net(sk);
1577 struct mr6_table *mrt;
1579 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1583 if (optname != MRT6_INIT) {
1584 if (sk != mrt->mroute6_sk && !capable(CAP_NET_ADMIN))
1590 if (sk->sk_type != SOCK_RAW ||
1591 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1593 if (optlen < sizeof(int))
1596 return ip6mr_sk_init(mrt, sk);
1599 return ip6mr_sk_done(sk);
1602 if (optlen < sizeof(vif))
1604 if (copy_from_user(&vif, optval, sizeof(vif)))
1606 if (vif.mif6c_mifi >= MAXMIFS)
1609 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1614 if (optlen < sizeof(mifi_t))
1616 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1619 ret = mif6_delete(mrt, mifi, NULL);
1624 * Manipulate the forwarding caches. These live
1625 * in a sort of kernel/user symbiosis.
1629 if (optlen < sizeof(mfc))
1631 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1634 if (optname == MRT6_DEL_MFC)
1635 ret = ip6mr_mfc_delete(mrt, &mfc);
1637 ret = ip6mr_mfc_add(net, mrt, &mfc, sk == mrt->mroute6_sk);
1642 * Control PIM assert (to activate pim will activate assert)
1647 if (get_user(v, (int __user *)optval))
1649 mrt->mroute_do_assert = !!v;
1653 #ifdef CONFIG_IPV6_PIMSM_V2
1657 if (get_user(v, (int __user *)optval))
1662 if (v != mrt->mroute_do_pim) {
1663 mrt->mroute_do_pim = v;
1664 mrt->mroute_do_assert = v;
1671 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1676 if (optlen != sizeof(u32))
1678 if (get_user(v, (u32 __user *)optval))
1680 if (sk == mrt->mroute6_sk)
1685 if (!ip6mr_new_table(net, v))
1687 raw6_sk(sk)->ip6mr_table = v;
1693 * Spurious command, or MRT6_VERSION which you cannot
1697 return -ENOPROTOOPT;
1702 * Getsock opt support for the multicast routing system.
1705 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1710 struct net *net = sock_net(sk);
1711 struct mr6_table *mrt;
1713 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1721 #ifdef CONFIG_IPV6_PIMSM_V2
1723 val = mrt->mroute_do_pim;
1727 val = mrt->mroute_do_assert;
1730 return -ENOPROTOOPT;
1733 if (get_user(olr, optlen))
1736 olr = min_t(int, olr, sizeof(int));
1740 if (put_user(olr, optlen))
1742 if (copy_to_user(optval, &val, olr))
1748 * The IP multicast ioctl support routines.
1751 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1753 struct sioc_sg_req6 sr;
1754 struct sioc_mif_req6 vr;
1755 struct mif_device *vif;
1756 struct mfc6_cache *c;
1757 struct net *net = sock_net(sk);
1758 struct mr6_table *mrt;
1760 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1765 case SIOCGETMIFCNT_IN6:
1766 if (copy_from_user(&vr, arg, sizeof(vr)))
1768 if (vr.mifi >= mrt->maxvif)
1770 read_lock(&mrt_lock);
1771 vif = &mrt->vif6_table[vr.mifi];
1772 if (MIF_EXISTS(mrt, vr.mifi)) {
1773 vr.icount = vif->pkt_in;
1774 vr.ocount = vif->pkt_out;
1775 vr.ibytes = vif->bytes_in;
1776 vr.obytes = vif->bytes_out;
1777 read_unlock(&mrt_lock);
1779 if (copy_to_user(arg, &vr, sizeof(vr)))
1783 read_unlock(&mrt_lock);
1784 return -EADDRNOTAVAIL;
1785 case SIOCGETSGCNT_IN6:
1786 if (copy_from_user(&sr, arg, sizeof(sr)))
1789 read_lock(&mrt_lock);
1790 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1792 sr.pktcnt = c->mfc_un.res.pkt;
1793 sr.bytecnt = c->mfc_un.res.bytes;
1794 sr.wrong_if = c->mfc_un.res.wrong_if;
1795 read_unlock(&mrt_lock);
1797 if (copy_to_user(arg, &sr, sizeof(sr)))
1801 read_unlock(&mrt_lock);
1802 return -EADDRNOTAVAIL;
1804 return -ENOIOCTLCMD;
1809 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1811 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1812 IPSTATS_MIB_OUTFORWDATAGRAMS);
1813 return dst_output(skb);
1817 * Processing handlers for ip6mr_forward
1820 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1821 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1823 struct ipv6hdr *ipv6h;
1824 struct mif_device *vif = &mrt->vif6_table[vifi];
1825 struct net_device *dev;
1826 struct dst_entry *dst;
1829 if (vif->dev == NULL)
1832 #ifdef CONFIG_IPV6_PIMSM_V2
1833 if (vif->flags & MIFF_REGISTER) {
1835 vif->bytes_out += skb->len;
1836 vif->dev->stats.tx_bytes += skb->len;
1837 vif->dev->stats.tx_packets++;
1838 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1843 ipv6h = ipv6_hdr(skb);
1845 fl = (struct flowi) {
1848 { .daddr = ipv6h->daddr, }
1852 dst = ip6_route_output(net, NULL, &fl);
1857 skb_dst_set(skb, dst);
1860 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1861 * not only before forwarding, but after forwarding on all output
1862 * interfaces. It is clear, if mrouter runs a multicasting
1863 * program, it should receive packets not depending to what interface
1864 * program is joined.
1865 * If we will not make it, the program will have to join on all
1866 * interfaces. On the other hand, multihoming host (or router, but
1867 * not mrouter) cannot join to more than one interface - it will
1868 * result in receiving multiple packets.
1873 vif->bytes_out += skb->len;
1875 /* We are about to write */
1876 /* XXX: extension headers? */
1877 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1880 ipv6h = ipv6_hdr(skb);
1883 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1885 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
1886 ip6mr_forward2_finish);
1893 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
1897 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
1898 if (mrt->vif6_table[ct].dev == dev)
1904 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
1905 struct sk_buff *skb, struct mfc6_cache *cache)
1910 vif = cache->mf6c_parent;
1911 cache->mfc_un.res.pkt++;
1912 cache->mfc_un.res.bytes += skb->len;
1915 * Wrong interface: drop packet and (maybe) send PIM assert.
1917 if (mrt->vif6_table[vif].dev != skb->dev) {
1920 cache->mfc_un.res.wrong_if++;
1921 true_vifi = ip6mr_find_vif(mrt, skb->dev);
1923 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1924 /* pimsm uses asserts, when switching from RPT to SPT,
1925 so that we cannot check that packet arrived on an oif.
1926 It is bad, but otherwise we would need to move pretty
1927 large chunk of pimd to kernel. Ough... --ANK
1929 (mrt->mroute_do_pim ||
1930 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1932 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1933 cache->mfc_un.res.last_assert = jiffies;
1934 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
1939 mrt->vif6_table[vif].pkt_in++;
1940 mrt->vif6_table[vif].bytes_in += skb->len;
1945 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1946 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1948 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1950 ip6mr_forward2(net, mrt, skb2, cache, psend);
1956 ip6mr_forward2(net, mrt, skb, cache, psend);
1967 * Multicast packets for forwarding arrive here
1970 int ip6_mr_input(struct sk_buff *skb)
1972 struct mfc6_cache *cache;
1973 struct net *net = dev_net(skb->dev);
1974 struct mr6_table *mrt;
1976 .iif = skb->dev->ifindex,
1981 err = ip6mr_fib_lookup(net, &fl, &mrt);
1985 read_lock(&mrt_lock);
1986 cache = ip6mr_cache_find(mrt,
1987 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1990 * No usable cache entry
1992 if (cache == NULL) {
1995 vif = ip6mr_find_vif(mrt, skb->dev);
1997 int err = ip6mr_cache_unresolved(mrt, vif, skb);
1998 read_unlock(&mrt_lock);
2002 read_unlock(&mrt_lock);
2007 ip6_mr_forward(net, mrt, skb, cache);
2009 read_unlock(&mrt_lock);
2015 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2016 struct mfc6_cache *c, struct rtmsg *rtm)
2019 struct rtnexthop *nhp;
2020 u8 *b = skb_tail_pointer(skb);
2021 struct rtattr *mp_head;
2023 /* If cache is unresolved, don't try to parse IIF and OIF */
2024 if (c->mf6c_parent >= MAXMIFS)
2027 if (MIF_EXISTS(mrt, c->mf6c_parent))
2028 RTA_PUT(skb, RTA_IIF, 4, &mrt->vif6_table[c->mf6c_parent].dev->ifindex);
2030 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
2032 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2033 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2034 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
2035 goto rtattr_failure;
2036 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
2037 nhp->rtnh_flags = 0;
2038 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2039 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2040 nhp->rtnh_len = sizeof(*nhp);
2043 mp_head->rta_type = RTA_MULTIPATH;
2044 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
2045 rtm->rtm_type = RTN_MULTICAST;
2053 int ip6mr_get_route(struct net *net,
2054 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2057 struct mr6_table *mrt;
2058 struct mfc6_cache *cache;
2059 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2061 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2065 read_lock(&mrt_lock);
2066 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2069 struct sk_buff *skb2;
2070 struct ipv6hdr *iph;
2071 struct net_device *dev;
2075 read_unlock(&mrt_lock);
2080 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2081 read_unlock(&mrt_lock);
2085 /* really correct? */
2086 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2088 read_unlock(&mrt_lock);
2092 skb_reset_transport_header(skb2);
2094 skb_put(skb2, sizeof(struct ipv6hdr));
2095 skb_reset_network_header(skb2);
2097 iph = ipv6_hdr(skb2);
2100 iph->flow_lbl[0] = 0;
2101 iph->flow_lbl[1] = 0;
2102 iph->flow_lbl[2] = 0;
2103 iph->payload_len = 0;
2104 iph->nexthdr = IPPROTO_NONE;
2106 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
2107 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
2109 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2110 read_unlock(&mrt_lock);
2115 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2116 cache->mfc_flags |= MFC_NOTIFY;
2118 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2119 read_unlock(&mrt_lock);
2123 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2124 u32 pid, u32 seq, struct mfc6_cache *c)
2126 struct nlmsghdr *nlh;
2129 nlh = nlmsg_put(skb, pid, seq, RTM_NEWROUTE, sizeof(*rtm), NLM_F_MULTI);
2133 rtm = nlmsg_data(nlh);
2134 rtm->rtm_family = RTNL_FAMILY_IPMR;
2135 rtm->rtm_dst_len = 128;
2136 rtm->rtm_src_len = 128;
2138 rtm->rtm_table = mrt->id;
2139 NLA_PUT_U32(skb, RTA_TABLE, mrt->id);
2140 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2141 rtm->rtm_protocol = RTPROT_UNSPEC;
2144 NLA_PUT(skb, RTA_SRC, 16, &c->mf6c_origin);
2145 NLA_PUT(skb, RTA_DST, 16, &c->mf6c_mcastgrp);
2147 if (__ip6mr_fill_mroute(mrt, skb, c, rtm) < 0)
2148 goto nla_put_failure;
2150 return nlmsg_end(skb, nlh);
2153 nlmsg_cancel(skb, nlh);
2157 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2159 struct net *net = sock_net(skb->sk);
2160 struct mr6_table *mrt;
2161 struct mfc6_cache *mfc;
2162 unsigned int t = 0, s_t;
2163 unsigned int h = 0, s_h;
2164 unsigned int e = 0, s_e;
2170 read_lock(&mrt_lock);
2171 ip6mr_for_each_table(mrt, net) {
2176 for (h = s_h; h < MFC6_LINES; h++) {
2177 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2180 if (ip6mr_fill_mroute(mrt, skb,
2181 NETLINK_CB(cb->skb).pid,
2195 read_unlock(&mrt_lock);