2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
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/uaccess.h>
20 #include <linux/types.h>
21 #include <linux/sched.h>
22 #include <linux/errno.h>
23 #include <linux/timer.h>
25 #include <linux/kernel.h>
26 #include <linux/fcntl.h>
27 #include <linux/stat.h>
28 #include <linux/socket.h>
29 #include <linux/inet.h>
30 #include <linux/netdevice.h>
31 #include <linux/inetdevice.h>
32 #include <linux/proc_fs.h>
33 #include <linux/seq_file.h>
34 #include <linux/init.h>
35 #include <linux/slab.h>
36 #include <linux/compat.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 <linux/export.h>
54 #include <net/ip6_checksum.h>
55 #include <linux/netconf.h>
58 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;
68 bool mroute_do_assert;
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 void 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 void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
119 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
120 struct netlink_callback *cb);
121 static void mroute_clean_tables(struct mr6_table *mrt, bool all);
122 static void ipmr_expire_process(unsigned long arg);
124 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
125 #define ip6mr_for_each_table(mrt, net) \
126 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
128 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
130 struct mr6_table *mrt;
132 ip6mr_for_each_table(mrt, net) {
139 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
140 struct mr6_table **mrt)
143 struct ip6mr_result res;
144 struct fib_lookup_arg arg = {
146 .flags = FIB_LOOKUP_NOREF,
149 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
150 flowi6_to_flowi(flp6), 0, &arg);
157 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
158 int flags, struct fib_lookup_arg *arg)
160 struct ip6mr_result *res = arg->result;
161 struct mr6_table *mrt;
163 switch (rule->action) {
166 case FR_ACT_UNREACHABLE:
168 case FR_ACT_PROHIBIT:
170 case FR_ACT_BLACKHOLE:
175 mrt = ip6mr_get_table(rule->fr_net, rule->table);
182 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
187 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
191 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
192 struct fib_rule_hdr *frh, struct nlattr **tb)
197 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
203 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
204 struct fib_rule_hdr *frh)
212 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
213 .family = RTNL_FAMILY_IP6MR,
214 .rule_size = sizeof(struct ip6mr_rule),
215 .addr_size = sizeof(struct in6_addr),
216 .action = ip6mr_rule_action,
217 .match = ip6mr_rule_match,
218 .configure = ip6mr_rule_configure,
219 .compare = ip6mr_rule_compare,
220 .fill = ip6mr_rule_fill,
221 .nlgroup = RTNLGRP_IPV6_RULE,
222 .policy = ip6mr_rule_policy,
223 .owner = THIS_MODULE,
226 static int __net_init ip6mr_rules_init(struct net *net)
228 struct fib_rules_ops *ops;
229 struct mr6_table *mrt;
232 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
236 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
238 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
244 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
248 net->ipv6.mr6_rules_ops = ops;
252 ip6mr_free_table(mrt);
254 fib_rules_unregister(ops);
258 static void __net_exit ip6mr_rules_exit(struct net *net)
260 struct mr6_table *mrt, *next;
263 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
264 list_del(&mrt->list);
265 ip6mr_free_table(mrt);
267 fib_rules_unregister(net->ipv6.mr6_rules_ops);
271 #define ip6mr_for_each_table(mrt, net) \
272 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
274 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
276 return net->ipv6.mrt6;
279 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
280 struct mr6_table **mrt)
282 *mrt = net->ipv6.mrt6;
286 static int __net_init ip6mr_rules_init(struct net *net)
288 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
289 return net->ipv6.mrt6 ? 0 : -ENOMEM;
292 static void __net_exit ip6mr_rules_exit(struct net *net)
295 ip6mr_free_table(net->ipv6.mrt6);
296 net->ipv6.mrt6 = NULL;
301 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
303 struct mr6_table *mrt;
306 mrt = ip6mr_get_table(net, id);
310 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
314 write_pnet(&mrt->net, net);
316 /* Forwarding cache */
317 for (i = 0; i < MFC6_LINES; i++)
318 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
320 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
322 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
325 #ifdef CONFIG_IPV6_PIMSM_V2
326 mrt->mroute_reg_vif_num = -1;
328 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
329 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
334 static void ip6mr_free_table(struct mr6_table *mrt)
336 del_timer_sync(&mrt->ipmr_expire_timer);
337 mroute_clean_tables(mrt, true);
341 #ifdef CONFIG_PROC_FS
343 struct ipmr_mfc_iter {
344 struct seq_net_private p;
345 struct mr6_table *mrt;
346 struct list_head *cache;
351 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
352 struct ipmr_mfc_iter *it, loff_t pos)
354 struct mr6_table *mrt = it->mrt;
355 struct mfc6_cache *mfc;
357 read_lock(&mrt_lock);
358 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
359 it->cache = &mrt->mfc6_cache_array[it->ct];
360 list_for_each_entry(mfc, it->cache, list)
364 read_unlock(&mrt_lock);
366 spin_lock_bh(&mfc_unres_lock);
367 it->cache = &mrt->mfc6_unres_queue;
368 list_for_each_entry(mfc, it->cache, list)
371 spin_unlock_bh(&mfc_unres_lock);
378 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
381 struct ipmr_vif_iter {
382 struct seq_net_private p;
383 struct mr6_table *mrt;
387 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
388 struct ipmr_vif_iter *iter,
391 struct mr6_table *mrt = iter->mrt;
393 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
394 if (!MIF_EXISTS(mrt, iter->ct))
397 return &mrt->vif6_table[iter->ct];
402 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
405 struct ipmr_vif_iter *iter = seq->private;
406 struct net *net = seq_file_net(seq);
407 struct mr6_table *mrt;
409 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
411 return ERR_PTR(-ENOENT);
415 read_lock(&mrt_lock);
416 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
420 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
422 struct ipmr_vif_iter *iter = seq->private;
423 struct net *net = seq_file_net(seq);
424 struct mr6_table *mrt = iter->mrt;
427 if (v == SEQ_START_TOKEN)
428 return ip6mr_vif_seq_idx(net, iter, 0);
430 while (++iter->ct < mrt->maxvif) {
431 if (!MIF_EXISTS(mrt, iter->ct))
433 return &mrt->vif6_table[iter->ct];
438 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
441 read_unlock(&mrt_lock);
444 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
446 struct ipmr_vif_iter *iter = seq->private;
447 struct mr6_table *mrt = iter->mrt;
449 if (v == SEQ_START_TOKEN) {
451 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
453 const struct mif_device *vif = v;
454 const char *name = vif->dev ? vif->dev->name : "none";
457 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
458 vif - mrt->vif6_table,
459 name, vif->bytes_in, vif->pkt_in,
460 vif->bytes_out, vif->pkt_out,
466 static const struct seq_operations ip6mr_vif_seq_ops = {
467 .start = ip6mr_vif_seq_start,
468 .next = ip6mr_vif_seq_next,
469 .stop = ip6mr_vif_seq_stop,
470 .show = ip6mr_vif_seq_show,
473 static int ip6mr_vif_open(struct inode *inode, struct file *file)
475 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
476 sizeof(struct ipmr_vif_iter));
479 static const struct file_operations ip6mr_vif_fops = {
480 .owner = THIS_MODULE,
481 .open = ip6mr_vif_open,
484 .release = seq_release_net,
487 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
489 struct ipmr_mfc_iter *it = seq->private;
490 struct net *net = seq_file_net(seq);
491 struct mr6_table *mrt;
493 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
495 return ERR_PTR(-ENOENT);
498 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
502 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
504 struct mfc6_cache *mfc = v;
505 struct ipmr_mfc_iter *it = seq->private;
506 struct net *net = seq_file_net(seq);
507 struct mr6_table *mrt = it->mrt;
511 if (v == SEQ_START_TOKEN)
512 return ipmr_mfc_seq_idx(net, seq->private, 0);
514 if (mfc->list.next != it->cache)
515 return list_entry(mfc->list.next, struct mfc6_cache, list);
517 if (it->cache == &mrt->mfc6_unres_queue)
520 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
522 while (++it->ct < MFC6_LINES) {
523 it->cache = &mrt->mfc6_cache_array[it->ct];
524 if (list_empty(it->cache))
526 return list_first_entry(it->cache, struct mfc6_cache, list);
529 /* exhausted cache_array, show unresolved */
530 read_unlock(&mrt_lock);
531 it->cache = &mrt->mfc6_unres_queue;
534 spin_lock_bh(&mfc_unres_lock);
535 if (!list_empty(it->cache))
536 return list_first_entry(it->cache, struct mfc6_cache, list);
539 spin_unlock_bh(&mfc_unres_lock);
545 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
547 struct ipmr_mfc_iter *it = seq->private;
548 struct mr6_table *mrt = it->mrt;
550 if (it->cache == &mrt->mfc6_unres_queue)
551 spin_unlock_bh(&mfc_unres_lock);
552 else if (it->cache == &mrt->mfc6_cache_array[it->ct])
553 read_unlock(&mrt_lock);
556 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
560 if (v == SEQ_START_TOKEN) {
564 "Iif Pkts Bytes Wrong Oifs\n");
566 const struct mfc6_cache *mfc = v;
567 const struct ipmr_mfc_iter *it = seq->private;
568 struct mr6_table *mrt = it->mrt;
570 seq_printf(seq, "%pI6 %pI6 %-3hd",
571 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
574 if (it->cache != &mrt->mfc6_unres_queue) {
575 seq_printf(seq, " %8lu %8lu %8lu",
577 mfc->mfc_un.res.bytes,
578 mfc->mfc_un.res.wrong_if);
579 for (n = mfc->mfc_un.res.minvif;
580 n < mfc->mfc_un.res.maxvif; n++) {
581 if (MIF_EXISTS(mrt, n) &&
582 mfc->mfc_un.res.ttls[n] < 255)
585 n, mfc->mfc_un.res.ttls[n]);
588 /* unresolved mfc_caches don't contain
589 * pkt, bytes and wrong_if values
591 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
598 static const struct seq_operations ipmr_mfc_seq_ops = {
599 .start = ipmr_mfc_seq_start,
600 .next = ipmr_mfc_seq_next,
601 .stop = ipmr_mfc_seq_stop,
602 .show = ipmr_mfc_seq_show,
605 static int ipmr_mfc_open(struct inode *inode, struct file *file)
607 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
608 sizeof(struct ipmr_mfc_iter));
611 static const struct file_operations ip6mr_mfc_fops = {
612 .owner = THIS_MODULE,
613 .open = ipmr_mfc_open,
616 .release = seq_release_net,
620 #ifdef CONFIG_IPV6_PIMSM_V2
622 static int pim6_rcv(struct sk_buff *skb)
624 struct pimreghdr *pim;
625 struct ipv6hdr *encap;
626 struct net_device *reg_dev = NULL;
627 struct net *net = dev_net(skb->dev);
628 struct mr6_table *mrt;
629 struct flowi6 fl6 = {
630 .flowi6_iif = skb->dev->ifindex,
631 .flowi6_mark = skb->mark,
635 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
638 pim = (struct pimreghdr *)skb_transport_header(skb);
639 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
640 (pim->flags & PIM_NULL_REGISTER) ||
641 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
642 sizeof(*pim), IPPROTO_PIM,
643 csum_partial((void *)pim, sizeof(*pim), 0)) &&
644 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
647 /* check if the inner packet is destined to mcast group */
648 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
651 if (!ipv6_addr_is_multicast(&encap->daddr) ||
652 encap->payload_len == 0 ||
653 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
656 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
658 reg_vif_num = mrt->mroute_reg_vif_num;
660 read_lock(&mrt_lock);
661 if (reg_vif_num >= 0)
662 reg_dev = mrt->vif6_table[reg_vif_num].dev;
665 read_unlock(&mrt_lock);
670 skb->mac_header = skb->network_header;
671 skb_pull(skb, (u8 *)encap - skb->data);
672 skb_reset_network_header(skb);
673 skb->protocol = htons(ETH_P_IPV6);
674 skb->ip_summed = CHECKSUM_NONE;
676 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
687 static const struct inet6_protocol pim6_protocol = {
691 /* Service routines creating virtual interfaces: PIMREG */
693 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
694 struct net_device *dev)
696 struct net *net = dev_net(dev);
697 struct mr6_table *mrt;
698 struct flowi6 fl6 = {
699 .flowi6_oif = dev->ifindex,
700 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
701 .flowi6_mark = skb->mark,
705 err = ip6mr_fib_lookup(net, &fl6, &mrt);
711 read_lock(&mrt_lock);
712 dev->stats.tx_bytes += skb->len;
713 dev->stats.tx_packets++;
714 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
715 read_unlock(&mrt_lock);
720 static int reg_vif_get_iflink(const struct net_device *dev)
725 static const struct net_device_ops reg_vif_netdev_ops = {
726 .ndo_start_xmit = reg_vif_xmit,
727 .ndo_get_iflink = reg_vif_get_iflink,
730 static void reg_vif_setup(struct net_device *dev)
732 dev->type = ARPHRD_PIMREG;
733 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
734 dev->flags = IFF_NOARP;
735 dev->netdev_ops = ®_vif_netdev_ops;
736 dev->destructor = free_netdev;
737 dev->features |= NETIF_F_NETNS_LOCAL;
740 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
742 struct net_device *dev;
745 if (mrt->id == RT6_TABLE_DFLT)
746 sprintf(name, "pim6reg");
748 sprintf(name, "pim6reg%u", mrt->id);
750 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
754 dev_net_set(dev, net);
756 if (register_netdevice(dev)) {
768 unregister_netdevice(dev);
777 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
779 struct mif_device *v;
780 struct net_device *dev;
781 struct inet6_dev *in6_dev;
783 if (vifi < 0 || vifi >= mrt->maxvif)
784 return -EADDRNOTAVAIL;
786 v = &mrt->vif6_table[vifi];
788 write_lock_bh(&mrt_lock);
793 write_unlock_bh(&mrt_lock);
794 return -EADDRNOTAVAIL;
797 #ifdef CONFIG_IPV6_PIMSM_V2
798 if (vifi == mrt->mroute_reg_vif_num)
799 mrt->mroute_reg_vif_num = -1;
802 if (vifi + 1 == mrt->maxvif) {
804 for (tmp = vifi - 1; tmp >= 0; tmp--) {
805 if (MIF_EXISTS(mrt, tmp))
808 mrt->maxvif = tmp + 1;
811 write_unlock_bh(&mrt_lock);
813 dev_set_allmulti(dev, -1);
815 in6_dev = __in6_dev_get(dev);
817 in6_dev->cnf.mc_forwarding--;
818 inet6_netconf_notify_devconf(dev_net(dev),
819 NETCONFA_MC_FORWARDING,
820 dev->ifindex, &in6_dev->cnf);
823 if (v->flags & MIFF_REGISTER)
824 unregister_netdevice_queue(dev, head);
830 static inline void ip6mr_cache_free(struct mfc6_cache *c)
832 kmem_cache_free(mrt_cachep, c);
835 /* Destroy an unresolved cache entry, killing queued skbs
836 and reporting error to netlink readers.
839 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
841 struct net *net = read_pnet(&mrt->net);
844 atomic_dec(&mrt->cache_resolve_queue_len);
846 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
847 if (ipv6_hdr(skb)->version == 0) {
848 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
849 nlh->nlmsg_type = NLMSG_ERROR;
850 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
851 skb_trim(skb, nlh->nlmsg_len);
852 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
853 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
862 /* Timer process for all the unresolved queue. */
864 static void ipmr_do_expire_process(struct mr6_table *mrt)
866 unsigned long now = jiffies;
867 unsigned long expires = 10 * HZ;
868 struct mfc6_cache *c, *next;
870 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
871 if (time_after(c->mfc_un.unres.expires, now)) {
873 unsigned long interval = c->mfc_un.unres.expires - now;
874 if (interval < expires)
880 mr6_netlink_event(mrt, c, RTM_DELROUTE);
881 ip6mr_destroy_unres(mrt, c);
884 if (!list_empty(&mrt->mfc6_unres_queue))
885 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
888 static void ipmr_expire_process(unsigned long arg)
890 struct mr6_table *mrt = (struct mr6_table *)arg;
892 if (!spin_trylock(&mfc_unres_lock)) {
893 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
897 if (!list_empty(&mrt->mfc6_unres_queue))
898 ipmr_do_expire_process(mrt);
900 spin_unlock(&mfc_unres_lock);
903 /* Fill oifs list. It is called under write locked mrt_lock. */
905 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
910 cache->mfc_un.res.minvif = MAXMIFS;
911 cache->mfc_un.res.maxvif = 0;
912 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
914 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
915 if (MIF_EXISTS(mrt, vifi) &&
916 ttls[vifi] && ttls[vifi] < 255) {
917 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
918 if (cache->mfc_un.res.minvif > vifi)
919 cache->mfc_un.res.minvif = vifi;
920 if (cache->mfc_un.res.maxvif <= vifi)
921 cache->mfc_un.res.maxvif = vifi + 1;
926 static int mif6_add(struct net *net, struct mr6_table *mrt,
927 struct mif6ctl *vifc, int mrtsock)
929 int vifi = vifc->mif6c_mifi;
930 struct mif_device *v = &mrt->vif6_table[vifi];
931 struct net_device *dev;
932 struct inet6_dev *in6_dev;
936 if (MIF_EXISTS(mrt, vifi))
939 switch (vifc->mif6c_flags) {
940 #ifdef CONFIG_IPV6_PIMSM_V2
943 * Special Purpose VIF in PIM
944 * All the packets will be sent to the daemon
946 if (mrt->mroute_reg_vif_num >= 0)
948 dev = ip6mr_reg_vif(net, mrt);
951 err = dev_set_allmulti(dev, 1);
953 unregister_netdevice(dev);
960 dev = dev_get_by_index(net, vifc->mif6c_pifi);
962 return -EADDRNOTAVAIL;
963 err = dev_set_allmulti(dev, 1);
973 in6_dev = __in6_dev_get(dev);
975 in6_dev->cnf.mc_forwarding++;
976 inet6_netconf_notify_devconf(dev_net(dev),
977 NETCONFA_MC_FORWARDING,
978 dev->ifindex, &in6_dev->cnf);
982 * Fill in the VIF structures
984 v->rate_limit = vifc->vifc_rate_limit;
985 v->flags = vifc->mif6c_flags;
987 v->flags |= VIFF_STATIC;
988 v->threshold = vifc->vifc_threshold;
993 v->link = dev->ifindex;
994 if (v->flags & MIFF_REGISTER)
995 v->link = dev_get_iflink(dev);
997 /* And finish update writing critical data */
998 write_lock_bh(&mrt_lock);
1000 #ifdef CONFIG_IPV6_PIMSM_V2
1001 if (v->flags & MIFF_REGISTER)
1002 mrt->mroute_reg_vif_num = vifi;
1004 if (vifi + 1 > mrt->maxvif)
1005 mrt->maxvif = vifi + 1;
1006 write_unlock_bh(&mrt_lock);
1010 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1011 const struct in6_addr *origin,
1012 const struct in6_addr *mcastgrp)
1014 int line = MFC6_HASH(mcastgrp, origin);
1015 struct mfc6_cache *c;
1017 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1018 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1019 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1025 /* Look for a (*,*,oif) entry */
1026 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1029 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1030 struct mfc6_cache *c;
1032 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1033 if (ipv6_addr_any(&c->mf6c_origin) &&
1034 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1035 (c->mfc_un.res.ttls[mifi] < 255))
1041 /* Look for a (*,G) entry */
1042 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1043 struct in6_addr *mcastgrp,
1046 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1047 struct mfc6_cache *c, *proxy;
1049 if (ipv6_addr_any(mcastgrp))
1052 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1053 if (ipv6_addr_any(&c->mf6c_origin) &&
1054 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1055 if (c->mfc_un.res.ttls[mifi] < 255)
1058 /* It's ok if the mifi is part of the static tree */
1059 proxy = ip6mr_cache_find_any_parent(mrt,
1061 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1066 return ip6mr_cache_find_any_parent(mrt, mifi);
1070 * Allocate a multicast cache entry
1072 static struct mfc6_cache *ip6mr_cache_alloc(void)
1074 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1077 c->mfc_un.res.minvif = MAXMIFS;
1081 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1083 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1086 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1087 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1092 * A cache entry has gone into a resolved state from queued
1095 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1096 struct mfc6_cache *uc, struct mfc6_cache *c)
1098 struct sk_buff *skb;
1101 * Play the pending entries through our router
1104 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1105 if (ipv6_hdr(skb)->version == 0) {
1106 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1108 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1109 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1111 nlh->nlmsg_type = NLMSG_ERROR;
1112 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1113 skb_trim(skb, nlh->nlmsg_len);
1114 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1116 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1118 ip6_mr_forward(net, mrt, skb, c);
1123 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1124 * expects the following bizarre scheme.
1126 * Called under mrt_lock.
1129 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1130 mifi_t mifi, int assert)
1132 struct sk_buff *skb;
1133 struct mrt6msg *msg;
1136 #ifdef CONFIG_IPV6_PIMSM_V2
1137 if (assert == MRT6MSG_WHOLEPKT)
1138 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1142 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1147 /* I suppose that internal messages
1148 * do not require checksums */
1150 skb->ip_summed = CHECKSUM_UNNECESSARY;
1152 #ifdef CONFIG_IPV6_PIMSM_V2
1153 if (assert == MRT6MSG_WHOLEPKT) {
1154 /* Ugly, but we have no choice with this interface.
1155 Duplicate old header, fix length etc.
1156 And all this only to mangle msg->im6_msgtype and
1157 to set msg->im6_mbz to "mbz" :-)
1159 skb_push(skb, -skb_network_offset(pkt));
1161 skb_push(skb, sizeof(*msg));
1162 skb_reset_transport_header(skb);
1163 msg = (struct mrt6msg *)skb_transport_header(skb);
1165 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1166 msg->im6_mif = mrt->mroute_reg_vif_num;
1168 msg->im6_src = ipv6_hdr(pkt)->saddr;
1169 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1171 skb->ip_summed = CHECKSUM_UNNECESSARY;
1176 * Copy the IP header
1179 skb_put(skb, sizeof(struct ipv6hdr));
1180 skb_reset_network_header(skb);
1181 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1186 skb_put(skb, sizeof(*msg));
1187 skb_reset_transport_header(skb);
1188 msg = (struct mrt6msg *)skb_transport_header(skb);
1191 msg->im6_msgtype = assert;
1192 msg->im6_mif = mifi;
1194 msg->im6_src = ipv6_hdr(pkt)->saddr;
1195 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1197 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1198 skb->ip_summed = CHECKSUM_UNNECESSARY;
1201 if (!mrt->mroute6_sk) {
1207 * Deliver to user space multicast routing algorithms
1209 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1211 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1219 * Queue a packet for resolution. It gets locked cache entry!
1223 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1227 struct mfc6_cache *c;
1229 spin_lock_bh(&mfc_unres_lock);
1230 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1231 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1232 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1240 * Create a new entry if allowable
1243 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1244 (c = ip6mr_cache_alloc_unres()) == NULL) {
1245 spin_unlock_bh(&mfc_unres_lock);
1252 * Fill in the new cache entry
1254 c->mf6c_parent = -1;
1255 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1256 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1259 * Reflect first query at pim6sd
1261 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1263 /* If the report failed throw the cache entry
1266 spin_unlock_bh(&mfc_unres_lock);
1268 ip6mr_cache_free(c);
1273 atomic_inc(&mrt->cache_resolve_queue_len);
1274 list_add(&c->list, &mrt->mfc6_unres_queue);
1275 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1277 ipmr_do_expire_process(mrt);
1281 * See if we can append the packet
1283 if (c->mfc_un.unres.unresolved.qlen > 3) {
1287 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1291 spin_unlock_bh(&mfc_unres_lock);
1296 * MFC6 cache manipulation by user space
1299 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1303 struct mfc6_cache *c, *next;
1305 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1307 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1308 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1309 ipv6_addr_equal(&c->mf6c_mcastgrp,
1310 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1311 (parent == -1 || parent == c->mf6c_parent)) {
1312 write_lock_bh(&mrt_lock);
1314 write_unlock_bh(&mrt_lock);
1316 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1317 ip6mr_cache_free(c);
1324 static int ip6mr_device_event(struct notifier_block *this,
1325 unsigned long event, void *ptr)
1327 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1328 struct net *net = dev_net(dev);
1329 struct mr6_table *mrt;
1330 struct mif_device *v;
1334 if (event != NETDEV_UNREGISTER)
1337 ip6mr_for_each_table(mrt, net) {
1338 v = &mrt->vif6_table[0];
1339 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1341 mif6_delete(mrt, ct, &list);
1344 unregister_netdevice_many(&list);
1349 static struct notifier_block ip6_mr_notifier = {
1350 .notifier_call = ip6mr_device_event
1354 * Setup for IP multicast routing
1357 static int __net_init ip6mr_net_init(struct net *net)
1361 err = ip6mr_rules_init(net);
1365 #ifdef CONFIG_PROC_FS
1367 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1369 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1370 goto proc_cache_fail;
1375 #ifdef CONFIG_PROC_FS
1377 remove_proc_entry("ip6_mr_vif", net->proc_net);
1379 ip6mr_rules_exit(net);
1385 static void __net_exit ip6mr_net_exit(struct net *net)
1387 #ifdef CONFIG_PROC_FS
1388 remove_proc_entry("ip6_mr_cache", net->proc_net);
1389 remove_proc_entry("ip6_mr_vif", net->proc_net);
1391 ip6mr_rules_exit(net);
1394 static struct pernet_operations ip6mr_net_ops = {
1395 .init = ip6mr_net_init,
1396 .exit = ip6mr_net_exit,
1399 int __init ip6_mr_init(void)
1403 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1404 sizeof(struct mfc6_cache),
1405 0, SLAB_HWCACHE_ALIGN,
1410 err = register_pernet_subsys(&ip6mr_net_ops);
1412 goto reg_pernet_fail;
1414 err = register_netdevice_notifier(&ip6_mr_notifier);
1416 goto reg_notif_fail;
1417 #ifdef CONFIG_IPV6_PIMSM_V2
1418 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1419 pr_err("%s: can't add PIM protocol\n", __func__);
1421 goto add_proto_fail;
1424 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1425 ip6mr_rtm_dumproute, NULL);
1427 #ifdef CONFIG_IPV6_PIMSM_V2
1429 unregister_netdevice_notifier(&ip6_mr_notifier);
1432 unregister_pernet_subsys(&ip6mr_net_ops);
1434 kmem_cache_destroy(mrt_cachep);
1438 void ip6_mr_cleanup(void)
1440 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1441 #ifdef CONFIG_IPV6_PIMSM_V2
1442 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1444 unregister_netdevice_notifier(&ip6_mr_notifier);
1445 unregister_pernet_subsys(&ip6mr_net_ops);
1446 kmem_cache_destroy(mrt_cachep);
1449 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1450 struct mf6cctl *mfc, int mrtsock, int parent)
1454 struct mfc6_cache *uc, *c;
1455 unsigned char ttls[MAXMIFS];
1458 if (mfc->mf6cc_parent >= MAXMIFS)
1461 memset(ttls, 255, MAXMIFS);
1462 for (i = 0; i < MAXMIFS; i++) {
1463 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1468 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1470 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1471 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1472 ipv6_addr_equal(&c->mf6c_mcastgrp,
1473 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1474 (parent == -1 || parent == mfc->mf6cc_parent)) {
1481 write_lock_bh(&mrt_lock);
1482 c->mf6c_parent = mfc->mf6cc_parent;
1483 ip6mr_update_thresholds(mrt, c, ttls);
1485 c->mfc_flags |= MFC_STATIC;
1486 write_unlock_bh(&mrt_lock);
1487 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1491 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1492 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1495 c = ip6mr_cache_alloc();
1499 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1500 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1501 c->mf6c_parent = mfc->mf6cc_parent;
1502 ip6mr_update_thresholds(mrt, c, ttls);
1504 c->mfc_flags |= MFC_STATIC;
1506 write_lock_bh(&mrt_lock);
1507 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1508 write_unlock_bh(&mrt_lock);
1511 * Check to see if we resolved a queued list. If so we
1512 * need to send on the frames and tidy up.
1515 spin_lock_bh(&mfc_unres_lock);
1516 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1517 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1518 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1519 list_del(&uc->list);
1520 atomic_dec(&mrt->cache_resolve_queue_len);
1525 if (list_empty(&mrt->mfc6_unres_queue))
1526 del_timer(&mrt->ipmr_expire_timer);
1527 spin_unlock_bh(&mfc_unres_lock);
1530 ip6mr_cache_resolve(net, mrt, uc, c);
1531 ip6mr_cache_free(uc);
1533 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1538 * Close the multicast socket, and clear the vif tables etc
1541 static void mroute_clean_tables(struct mr6_table *mrt, bool all)
1545 struct mfc6_cache *c, *next;
1548 * Shut down all active vif entries
1550 for (i = 0; i < mrt->maxvif; i++) {
1551 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC))
1553 mif6_delete(mrt, i, &list);
1555 unregister_netdevice_many(&list);
1560 for (i = 0; i < MFC6_LINES; i++) {
1561 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1562 if (!all && (c->mfc_flags & MFC_STATIC))
1564 write_lock_bh(&mrt_lock);
1566 write_unlock_bh(&mrt_lock);
1568 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1569 ip6mr_cache_free(c);
1573 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1574 spin_lock_bh(&mfc_unres_lock);
1575 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1577 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1578 ip6mr_destroy_unres(mrt, c);
1580 spin_unlock_bh(&mfc_unres_lock);
1584 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1587 struct net *net = sock_net(sk);
1590 write_lock_bh(&mrt_lock);
1591 if (likely(mrt->mroute6_sk == NULL)) {
1592 mrt->mroute6_sk = sk;
1593 net->ipv6.devconf_all->mc_forwarding++;
1594 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1595 NETCONFA_IFINDEX_ALL,
1596 net->ipv6.devconf_all);
1600 write_unlock_bh(&mrt_lock);
1607 int ip6mr_sk_done(struct sock *sk)
1610 struct net *net = sock_net(sk);
1611 struct mr6_table *mrt;
1614 ip6mr_for_each_table(mrt, net) {
1615 if (sk == mrt->mroute6_sk) {
1616 write_lock_bh(&mrt_lock);
1617 mrt->mroute6_sk = NULL;
1618 net->ipv6.devconf_all->mc_forwarding--;
1619 inet6_netconf_notify_devconf(net,
1620 NETCONFA_MC_FORWARDING,
1621 NETCONFA_IFINDEX_ALL,
1622 net->ipv6.devconf_all);
1623 write_unlock_bh(&mrt_lock);
1625 mroute_clean_tables(mrt, false);
1635 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1637 struct mr6_table *mrt;
1638 struct flowi6 fl6 = {
1639 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1640 .flowi6_oif = skb->dev->ifindex,
1641 .flowi6_mark = skb->mark,
1644 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1647 return mrt->mroute6_sk;
1651 * Socket options and virtual interface manipulation. The whole
1652 * virtual interface system is a complete heap, but unfortunately
1653 * that's how BSD mrouted happens to think. Maybe one day with a proper
1654 * MOSPF/PIM router set up we can clean this up.
1657 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1659 int ret, parent = 0;
1663 struct net *net = sock_net(sk);
1664 struct mr6_table *mrt;
1666 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1670 if (optname != MRT6_INIT) {
1671 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1677 if (sk->sk_type != SOCK_RAW ||
1678 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1680 if (optlen < sizeof(int))
1683 return ip6mr_sk_init(mrt, sk);
1686 return ip6mr_sk_done(sk);
1689 if (optlen < sizeof(vif))
1691 if (copy_from_user(&vif, optval, sizeof(vif)))
1693 if (vif.mif6c_mifi >= MAXMIFS)
1696 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1701 if (optlen < sizeof(mifi_t))
1703 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1706 ret = mif6_delete(mrt, mifi, NULL);
1711 * Manipulate the forwarding caches. These live
1712 * in a sort of kernel/user symbiosis.
1717 case MRT6_ADD_MFC_PROXY:
1718 case MRT6_DEL_MFC_PROXY:
1719 if (optlen < sizeof(mfc))
1721 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1724 parent = mfc.mf6cc_parent;
1726 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1727 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1729 ret = ip6mr_mfc_add(net, mrt, &mfc,
1730 sk == mrt->mroute6_sk, parent);
1735 * Control PIM assert (to activate pim will activate assert)
1741 if (optlen != sizeof(v))
1743 if (get_user(v, (int __user *)optval))
1745 mrt->mroute_do_assert = v;
1749 #ifdef CONFIG_IPV6_PIMSM_V2
1754 if (optlen != sizeof(v))
1756 if (get_user(v, (int __user *)optval))
1761 if (v != mrt->mroute_do_pim) {
1762 mrt->mroute_do_pim = v;
1763 mrt->mroute_do_assert = v;
1770 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1775 if (optlen != sizeof(u32))
1777 if (get_user(v, (u32 __user *)optval))
1779 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1780 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1782 if (sk == mrt->mroute6_sk)
1787 if (!ip6mr_new_table(net, v))
1789 raw6_sk(sk)->ip6mr_table = v;
1795 * Spurious command, or MRT6_VERSION which you cannot
1799 return -ENOPROTOOPT;
1804 * Getsock opt support for the multicast routing system.
1807 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1812 struct net *net = sock_net(sk);
1813 struct mr6_table *mrt;
1815 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1823 #ifdef CONFIG_IPV6_PIMSM_V2
1825 val = mrt->mroute_do_pim;
1829 val = mrt->mroute_do_assert;
1832 return -ENOPROTOOPT;
1835 if (get_user(olr, optlen))
1838 olr = min_t(int, olr, sizeof(int));
1842 if (put_user(olr, optlen))
1844 if (copy_to_user(optval, &val, olr))
1850 * The IP multicast ioctl support routines.
1853 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1855 struct sioc_sg_req6 sr;
1856 struct sioc_mif_req6 vr;
1857 struct mif_device *vif;
1858 struct mfc6_cache *c;
1859 struct net *net = sock_net(sk);
1860 struct mr6_table *mrt;
1862 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1867 case SIOCGETMIFCNT_IN6:
1868 if (copy_from_user(&vr, arg, sizeof(vr)))
1870 if (vr.mifi >= mrt->maxvif)
1872 read_lock(&mrt_lock);
1873 vif = &mrt->vif6_table[vr.mifi];
1874 if (MIF_EXISTS(mrt, vr.mifi)) {
1875 vr.icount = vif->pkt_in;
1876 vr.ocount = vif->pkt_out;
1877 vr.ibytes = vif->bytes_in;
1878 vr.obytes = vif->bytes_out;
1879 read_unlock(&mrt_lock);
1881 if (copy_to_user(arg, &vr, sizeof(vr)))
1885 read_unlock(&mrt_lock);
1886 return -EADDRNOTAVAIL;
1887 case SIOCGETSGCNT_IN6:
1888 if (copy_from_user(&sr, arg, sizeof(sr)))
1891 read_lock(&mrt_lock);
1892 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1894 sr.pktcnt = c->mfc_un.res.pkt;
1895 sr.bytecnt = c->mfc_un.res.bytes;
1896 sr.wrong_if = c->mfc_un.res.wrong_if;
1897 read_unlock(&mrt_lock);
1899 if (copy_to_user(arg, &sr, sizeof(sr)))
1903 read_unlock(&mrt_lock);
1904 return -EADDRNOTAVAIL;
1906 return -ENOIOCTLCMD;
1910 #ifdef CONFIG_COMPAT
1911 struct compat_sioc_sg_req6 {
1912 struct sockaddr_in6 src;
1913 struct sockaddr_in6 grp;
1914 compat_ulong_t pktcnt;
1915 compat_ulong_t bytecnt;
1916 compat_ulong_t wrong_if;
1919 struct compat_sioc_mif_req6 {
1921 compat_ulong_t icount;
1922 compat_ulong_t ocount;
1923 compat_ulong_t ibytes;
1924 compat_ulong_t obytes;
1927 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1929 struct compat_sioc_sg_req6 sr;
1930 struct compat_sioc_mif_req6 vr;
1931 struct mif_device *vif;
1932 struct mfc6_cache *c;
1933 struct net *net = sock_net(sk);
1934 struct mr6_table *mrt;
1936 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1941 case SIOCGETMIFCNT_IN6:
1942 if (copy_from_user(&vr, arg, sizeof(vr)))
1944 if (vr.mifi >= mrt->maxvif)
1946 read_lock(&mrt_lock);
1947 vif = &mrt->vif6_table[vr.mifi];
1948 if (MIF_EXISTS(mrt, vr.mifi)) {
1949 vr.icount = vif->pkt_in;
1950 vr.ocount = vif->pkt_out;
1951 vr.ibytes = vif->bytes_in;
1952 vr.obytes = vif->bytes_out;
1953 read_unlock(&mrt_lock);
1955 if (copy_to_user(arg, &vr, sizeof(vr)))
1959 read_unlock(&mrt_lock);
1960 return -EADDRNOTAVAIL;
1961 case SIOCGETSGCNT_IN6:
1962 if (copy_from_user(&sr, arg, sizeof(sr)))
1965 read_lock(&mrt_lock);
1966 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1968 sr.pktcnt = c->mfc_un.res.pkt;
1969 sr.bytecnt = c->mfc_un.res.bytes;
1970 sr.wrong_if = c->mfc_un.res.wrong_if;
1971 read_unlock(&mrt_lock);
1973 if (copy_to_user(arg, &sr, sizeof(sr)))
1977 read_unlock(&mrt_lock);
1978 return -EADDRNOTAVAIL;
1980 return -ENOIOCTLCMD;
1985 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1987 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
1988 IPSTATS_MIB_OUTFORWDATAGRAMS);
1989 IP6_ADD_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
1990 IPSTATS_MIB_OUTOCTETS, skb->len);
1991 return dst_output(net, sk, skb);
1995 * Processing handlers for ip6mr_forward
1998 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1999 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
2001 struct ipv6hdr *ipv6h;
2002 struct mif_device *vif = &mrt->vif6_table[vifi];
2003 struct net_device *dev;
2004 struct dst_entry *dst;
2010 #ifdef CONFIG_IPV6_PIMSM_V2
2011 if (vif->flags & MIFF_REGISTER) {
2013 vif->bytes_out += skb->len;
2014 vif->dev->stats.tx_bytes += skb->len;
2015 vif->dev->stats.tx_packets++;
2016 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2021 ipv6h = ipv6_hdr(skb);
2023 fl6 = (struct flowi6) {
2024 .flowi6_oif = vif->link,
2025 .daddr = ipv6h->daddr,
2028 dst = ip6_route_output(net, NULL, &fl6);
2035 skb_dst_set(skb, dst);
2038 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2039 * not only before forwarding, but after forwarding on all output
2040 * interfaces. It is clear, if mrouter runs a multicasting
2041 * program, it should receive packets not depending to what interface
2042 * program is joined.
2043 * If we will not make it, the program will have to join on all
2044 * interfaces. On the other hand, multihoming host (or router, but
2045 * not mrouter) cannot join to more than one interface - it will
2046 * result in receiving multiple packets.
2051 vif->bytes_out += skb->len;
2053 /* We are about to write */
2054 /* XXX: extension headers? */
2055 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2058 ipv6h = ipv6_hdr(skb);
2061 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2063 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2064 net, NULL, skb, skb->dev, dev,
2065 ip6mr_forward2_finish);
2072 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2076 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2077 if (mrt->vif6_table[ct].dev == dev)
2083 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2084 struct sk_buff *skb, struct mfc6_cache *cache)
2088 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2090 vif = cache->mf6c_parent;
2091 cache->mfc_un.res.pkt++;
2092 cache->mfc_un.res.bytes += skb->len;
2094 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2095 struct mfc6_cache *cache_proxy;
2097 /* For an (*,G) entry, we only check that the incoming
2098 * interface is part of the static tree.
2100 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2102 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2107 * Wrong interface: drop packet and (maybe) send PIM assert.
2109 if (mrt->vif6_table[vif].dev != skb->dev) {
2110 cache->mfc_un.res.wrong_if++;
2112 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2113 /* pimsm uses asserts, when switching from RPT to SPT,
2114 so that we cannot check that packet arrived on an oif.
2115 It is bad, but otherwise we would need to move pretty
2116 large chunk of pimd to kernel. Ough... --ANK
2118 (mrt->mroute_do_pim ||
2119 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2121 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2122 cache->mfc_un.res.last_assert = jiffies;
2123 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2129 mrt->vif6_table[vif].pkt_in++;
2130 mrt->vif6_table[vif].bytes_in += skb->len;
2135 if (ipv6_addr_any(&cache->mf6c_origin) &&
2136 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2137 if (true_vifi >= 0 &&
2138 true_vifi != cache->mf6c_parent &&
2139 ipv6_hdr(skb)->hop_limit >
2140 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2141 /* It's an (*,*) entry and the packet is not coming from
2142 * the upstream: forward the packet to the upstream
2145 psend = cache->mf6c_parent;
2150 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2151 /* For (*,G) entry, don't forward to the incoming interface */
2152 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2153 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2155 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2157 ip6mr_forward2(net, mrt, skb2, cache, psend);
2164 ip6mr_forward2(net, mrt, skb, cache, psend);
2174 * Multicast packets for forwarding arrive here
2177 int ip6_mr_input(struct sk_buff *skb)
2179 struct mfc6_cache *cache;
2180 struct net *net = dev_net(skb->dev);
2181 struct mr6_table *mrt;
2182 struct flowi6 fl6 = {
2183 .flowi6_iif = skb->dev->ifindex,
2184 .flowi6_mark = skb->mark,
2188 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2194 read_lock(&mrt_lock);
2195 cache = ip6mr_cache_find(mrt,
2196 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2198 int vif = ip6mr_find_vif(mrt, skb->dev);
2201 cache = ip6mr_cache_find_any(mrt,
2202 &ipv6_hdr(skb)->daddr,
2207 * No usable cache entry
2212 vif = ip6mr_find_vif(mrt, skb->dev);
2214 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2215 read_unlock(&mrt_lock);
2219 read_unlock(&mrt_lock);
2224 ip6_mr_forward(net, mrt, skb, cache);
2226 read_unlock(&mrt_lock);
2232 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2233 struct mfc6_cache *c, struct rtmsg *rtm)
2236 struct rtnexthop *nhp;
2237 struct nlattr *mp_attr;
2238 struct rta_mfc_stats mfcs;
2240 /* If cache is unresolved, don't try to parse IIF and OIF */
2241 if (c->mf6c_parent >= MAXMIFS)
2244 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2245 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2247 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2251 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2252 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2253 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2255 nla_nest_cancel(skb, mp_attr);
2259 nhp->rtnh_flags = 0;
2260 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2261 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2262 nhp->rtnh_len = sizeof(*nhp);
2266 nla_nest_end(skb, mp_attr);
2268 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2269 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2270 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2271 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2274 rtm->rtm_type = RTN_MULTICAST;
2278 int ip6mr_get_route(struct net *net,
2279 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2282 struct mr6_table *mrt;
2283 struct mfc6_cache *cache;
2284 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2286 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2290 read_lock(&mrt_lock);
2291 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2292 if (!cache && skb->dev) {
2293 int vif = ip6mr_find_vif(mrt, skb->dev);
2296 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2301 struct sk_buff *skb2;
2302 struct ipv6hdr *iph;
2303 struct net_device *dev;
2307 read_unlock(&mrt_lock);
2312 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2313 read_unlock(&mrt_lock);
2317 /* really correct? */
2318 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2320 read_unlock(&mrt_lock);
2324 skb_reset_transport_header(skb2);
2326 skb_put(skb2, sizeof(struct ipv6hdr));
2327 skb_reset_network_header(skb2);
2329 iph = ipv6_hdr(skb2);
2332 iph->flow_lbl[0] = 0;
2333 iph->flow_lbl[1] = 0;
2334 iph->flow_lbl[2] = 0;
2335 iph->payload_len = 0;
2336 iph->nexthdr = IPPROTO_NONE;
2338 iph->saddr = rt->rt6i_src.addr;
2339 iph->daddr = rt->rt6i_dst.addr;
2341 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2342 read_unlock(&mrt_lock);
2347 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2348 cache->mfc_flags |= MFC_NOTIFY;
2350 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2351 read_unlock(&mrt_lock);
2355 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2356 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2359 struct nlmsghdr *nlh;
2363 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2367 rtm = nlmsg_data(nlh);
2368 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2369 rtm->rtm_dst_len = 128;
2370 rtm->rtm_src_len = 128;
2372 rtm->rtm_table = mrt->id;
2373 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2374 goto nla_put_failure;
2375 rtm->rtm_type = RTN_MULTICAST;
2376 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2377 if (c->mfc_flags & MFC_STATIC)
2378 rtm->rtm_protocol = RTPROT_STATIC;
2380 rtm->rtm_protocol = RTPROT_MROUTED;
2383 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2384 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2385 goto nla_put_failure;
2386 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2387 /* do not break the dump if cache is unresolved */
2388 if (err < 0 && err != -ENOENT)
2389 goto nla_put_failure;
2391 nlmsg_end(skb, nlh);
2395 nlmsg_cancel(skb, nlh);
2399 static int mr6_msgsize(bool unresolved, int maxvif)
2402 NLMSG_ALIGN(sizeof(struct rtmsg))
2403 + nla_total_size(4) /* RTA_TABLE */
2404 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2405 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2410 + nla_total_size(4) /* RTA_IIF */
2411 + nla_total_size(0) /* RTA_MULTIPATH */
2412 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2414 + nla_total_size(sizeof(struct rta_mfc_stats))
2420 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2423 struct net *net = read_pnet(&mrt->net);
2424 struct sk_buff *skb;
2427 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2432 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2436 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2442 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2445 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2447 struct net *net = sock_net(skb->sk);
2448 struct mr6_table *mrt;
2449 struct mfc6_cache *mfc;
2450 unsigned int t = 0, s_t;
2451 unsigned int h = 0, s_h;
2452 unsigned int e = 0, s_e;
2458 read_lock(&mrt_lock);
2459 ip6mr_for_each_table(mrt, net) {
2464 for (h = s_h; h < MFC6_LINES; h++) {
2465 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2468 if (ip6mr_fill_mroute(mrt, skb,
2469 NETLINK_CB(cb->skb).portid,
2479 spin_lock_bh(&mfc_unres_lock);
2480 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2483 if (ip6mr_fill_mroute(mrt, skb,
2484 NETLINK_CB(cb->skb).portid,
2488 spin_unlock_bh(&mfc_unres_lock);
2494 spin_unlock_bh(&mfc_unres_lock);
2501 read_unlock(&mrt_lock);