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/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.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1078 c->mfc_un.res.minvif = MAXMIFS;
1082 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1084 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1087 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1088 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1093 * A cache entry has gone into a resolved state from queued
1096 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1097 struct mfc6_cache *uc, struct mfc6_cache *c)
1099 struct sk_buff *skb;
1102 * Play the pending entries through our router
1105 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1106 if (ipv6_hdr(skb)->version == 0) {
1107 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1109 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1110 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1112 nlh->nlmsg_type = NLMSG_ERROR;
1113 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1114 skb_trim(skb, nlh->nlmsg_len);
1115 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1117 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1119 ip6_mr_forward(net, mrt, skb, c);
1124 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1125 * expects the following bizarre scheme.
1127 * Called under mrt_lock.
1130 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1131 mifi_t mifi, int assert)
1133 struct sk_buff *skb;
1134 struct mrt6msg *msg;
1137 #ifdef CONFIG_IPV6_PIMSM_V2
1138 if (assert == MRT6MSG_WHOLEPKT)
1139 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1143 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1148 /* I suppose that internal messages
1149 * do not require checksums */
1151 skb->ip_summed = CHECKSUM_UNNECESSARY;
1153 #ifdef CONFIG_IPV6_PIMSM_V2
1154 if (assert == MRT6MSG_WHOLEPKT) {
1155 /* Ugly, but we have no choice with this interface.
1156 Duplicate old header, fix length etc.
1157 And all this only to mangle msg->im6_msgtype and
1158 to set msg->im6_mbz to "mbz" :-)
1160 skb_push(skb, -skb_network_offset(pkt));
1162 skb_push(skb, sizeof(*msg));
1163 skb_reset_transport_header(skb);
1164 msg = (struct mrt6msg *)skb_transport_header(skb);
1166 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1167 msg->im6_mif = mrt->mroute_reg_vif_num;
1169 msg->im6_src = ipv6_hdr(pkt)->saddr;
1170 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1172 skb->ip_summed = CHECKSUM_UNNECESSARY;
1177 * Copy the IP header
1180 skb_put(skb, sizeof(struct ipv6hdr));
1181 skb_reset_network_header(skb);
1182 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1187 skb_put(skb, sizeof(*msg));
1188 skb_reset_transport_header(skb);
1189 msg = (struct mrt6msg *)skb_transport_header(skb);
1192 msg->im6_msgtype = assert;
1193 msg->im6_mif = mifi;
1195 msg->im6_src = ipv6_hdr(pkt)->saddr;
1196 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1198 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1199 skb->ip_summed = CHECKSUM_UNNECESSARY;
1202 if (!mrt->mroute6_sk) {
1208 * Deliver to user space multicast routing algorithms
1210 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1212 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1220 * Queue a packet for resolution. It gets locked cache entry!
1224 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1228 struct mfc6_cache *c;
1230 spin_lock_bh(&mfc_unres_lock);
1231 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1232 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1233 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1241 * Create a new entry if allowable
1244 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1245 (c = ip6mr_cache_alloc_unres()) == NULL) {
1246 spin_unlock_bh(&mfc_unres_lock);
1253 * Fill in the new cache entry
1255 c->mf6c_parent = -1;
1256 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1257 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1260 * Reflect first query at pim6sd
1262 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1264 /* If the report failed throw the cache entry
1267 spin_unlock_bh(&mfc_unres_lock);
1269 ip6mr_cache_free(c);
1274 atomic_inc(&mrt->cache_resolve_queue_len);
1275 list_add(&c->list, &mrt->mfc6_unres_queue);
1276 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1278 ipmr_do_expire_process(mrt);
1282 * See if we can append the packet
1284 if (c->mfc_un.unres.unresolved.qlen > 3) {
1288 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1292 spin_unlock_bh(&mfc_unres_lock);
1297 * MFC6 cache manipulation by user space
1300 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1304 struct mfc6_cache *c, *next;
1306 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1308 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1309 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1310 ipv6_addr_equal(&c->mf6c_mcastgrp,
1311 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1312 (parent == -1 || parent == c->mf6c_parent)) {
1313 write_lock_bh(&mrt_lock);
1315 write_unlock_bh(&mrt_lock);
1317 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1318 ip6mr_cache_free(c);
1325 static int ip6mr_device_event(struct notifier_block *this,
1326 unsigned long event, void *ptr)
1328 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1329 struct net *net = dev_net(dev);
1330 struct mr6_table *mrt;
1331 struct mif_device *v;
1335 if (event != NETDEV_UNREGISTER)
1338 ip6mr_for_each_table(mrt, net) {
1339 v = &mrt->vif6_table[0];
1340 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1342 mif6_delete(mrt, ct, &list);
1345 unregister_netdevice_many(&list);
1350 static struct notifier_block ip6_mr_notifier = {
1351 .notifier_call = ip6mr_device_event
1355 * Setup for IP multicast routing
1358 static int __net_init ip6mr_net_init(struct net *net)
1362 err = ip6mr_rules_init(net);
1366 #ifdef CONFIG_PROC_FS
1368 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1370 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1371 goto proc_cache_fail;
1376 #ifdef CONFIG_PROC_FS
1378 remove_proc_entry("ip6_mr_vif", net->proc_net);
1380 ip6mr_rules_exit(net);
1386 static void __net_exit ip6mr_net_exit(struct net *net)
1388 #ifdef CONFIG_PROC_FS
1389 remove_proc_entry("ip6_mr_cache", net->proc_net);
1390 remove_proc_entry("ip6_mr_vif", net->proc_net);
1392 ip6mr_rules_exit(net);
1395 static struct pernet_operations ip6mr_net_ops = {
1396 .init = ip6mr_net_init,
1397 .exit = ip6mr_net_exit,
1400 int __init ip6_mr_init(void)
1404 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1405 sizeof(struct mfc6_cache),
1406 0, SLAB_HWCACHE_ALIGN,
1411 err = register_pernet_subsys(&ip6mr_net_ops);
1413 goto reg_pernet_fail;
1415 err = register_netdevice_notifier(&ip6_mr_notifier);
1417 goto reg_notif_fail;
1418 #ifdef CONFIG_IPV6_PIMSM_V2
1419 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1420 pr_err("%s: can't add PIM protocol\n", __func__);
1422 goto add_proto_fail;
1425 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1426 ip6mr_rtm_dumproute, NULL);
1428 #ifdef CONFIG_IPV6_PIMSM_V2
1430 unregister_netdevice_notifier(&ip6_mr_notifier);
1433 unregister_pernet_subsys(&ip6mr_net_ops);
1435 kmem_cache_destroy(mrt_cachep);
1439 void ip6_mr_cleanup(void)
1441 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1442 #ifdef CONFIG_IPV6_PIMSM_V2
1443 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1445 unregister_netdevice_notifier(&ip6_mr_notifier);
1446 unregister_pernet_subsys(&ip6mr_net_ops);
1447 kmem_cache_destroy(mrt_cachep);
1450 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1451 struct mf6cctl *mfc, int mrtsock, int parent)
1455 struct mfc6_cache *uc, *c;
1456 unsigned char ttls[MAXMIFS];
1459 if (mfc->mf6cc_parent >= MAXMIFS)
1462 memset(ttls, 255, MAXMIFS);
1463 for (i = 0; i < MAXMIFS; i++) {
1464 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1469 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1471 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1472 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1473 ipv6_addr_equal(&c->mf6c_mcastgrp,
1474 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1475 (parent == -1 || parent == mfc->mf6cc_parent)) {
1482 write_lock_bh(&mrt_lock);
1483 c->mf6c_parent = mfc->mf6cc_parent;
1484 ip6mr_update_thresholds(mrt, c, ttls);
1486 c->mfc_flags |= MFC_STATIC;
1487 write_unlock_bh(&mrt_lock);
1488 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1492 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1493 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1496 c = ip6mr_cache_alloc();
1500 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1501 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1502 c->mf6c_parent = mfc->mf6cc_parent;
1503 ip6mr_update_thresholds(mrt, c, ttls);
1505 c->mfc_flags |= MFC_STATIC;
1507 write_lock_bh(&mrt_lock);
1508 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1509 write_unlock_bh(&mrt_lock);
1512 * Check to see if we resolved a queued list. If so we
1513 * need to send on the frames and tidy up.
1516 spin_lock_bh(&mfc_unres_lock);
1517 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1518 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1519 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1520 list_del(&uc->list);
1521 atomic_dec(&mrt->cache_resolve_queue_len);
1526 if (list_empty(&mrt->mfc6_unres_queue))
1527 del_timer(&mrt->ipmr_expire_timer);
1528 spin_unlock_bh(&mfc_unres_lock);
1531 ip6mr_cache_resolve(net, mrt, uc, c);
1532 ip6mr_cache_free(uc);
1534 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1539 * Close the multicast socket, and clear the vif tables etc
1542 static void mroute_clean_tables(struct mr6_table *mrt, bool all)
1546 struct mfc6_cache *c, *next;
1549 * Shut down all active vif entries
1551 for (i = 0; i < mrt->maxvif; i++) {
1552 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC))
1554 mif6_delete(mrt, i, &list);
1556 unregister_netdevice_many(&list);
1561 for (i = 0; i < MFC6_LINES; i++) {
1562 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1563 if (!all && (c->mfc_flags & MFC_STATIC))
1565 write_lock_bh(&mrt_lock);
1567 write_unlock_bh(&mrt_lock);
1569 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1570 ip6mr_cache_free(c);
1574 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1575 spin_lock_bh(&mfc_unres_lock);
1576 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1578 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1579 ip6mr_destroy_unres(mrt, c);
1581 spin_unlock_bh(&mfc_unres_lock);
1585 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1588 struct net *net = sock_net(sk);
1591 write_lock_bh(&mrt_lock);
1592 if (likely(mrt->mroute6_sk == NULL)) {
1593 mrt->mroute6_sk = sk;
1594 net->ipv6.devconf_all->mc_forwarding++;
1595 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1596 NETCONFA_IFINDEX_ALL,
1597 net->ipv6.devconf_all);
1601 write_unlock_bh(&mrt_lock);
1608 int ip6mr_sk_done(struct sock *sk)
1611 struct net *net = sock_net(sk);
1612 struct mr6_table *mrt;
1615 ip6mr_for_each_table(mrt, net) {
1616 if (sk == mrt->mroute6_sk) {
1617 write_lock_bh(&mrt_lock);
1618 mrt->mroute6_sk = NULL;
1619 net->ipv6.devconf_all->mc_forwarding--;
1620 inet6_netconf_notify_devconf(net,
1621 NETCONFA_MC_FORWARDING,
1622 NETCONFA_IFINDEX_ALL,
1623 net->ipv6.devconf_all);
1624 write_unlock_bh(&mrt_lock);
1626 mroute_clean_tables(mrt, false);
1636 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1638 struct mr6_table *mrt;
1639 struct flowi6 fl6 = {
1640 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1641 .flowi6_oif = skb->dev->ifindex,
1642 .flowi6_mark = skb->mark,
1645 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1648 return mrt->mroute6_sk;
1652 * Socket options and virtual interface manipulation. The whole
1653 * virtual interface system is a complete heap, but unfortunately
1654 * that's how BSD mrouted happens to think. Maybe one day with a proper
1655 * MOSPF/PIM router set up we can clean this up.
1658 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1660 int ret, parent = 0;
1664 struct net *net = sock_net(sk);
1665 struct mr6_table *mrt;
1667 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1671 if (optname != MRT6_INIT) {
1672 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1678 if (sk->sk_type != SOCK_RAW ||
1679 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1681 if (optlen < sizeof(int))
1684 return ip6mr_sk_init(mrt, sk);
1687 return ip6mr_sk_done(sk);
1690 if (optlen < sizeof(vif))
1692 if (copy_from_user(&vif, optval, sizeof(vif)))
1694 if (vif.mif6c_mifi >= MAXMIFS)
1697 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1702 if (optlen < sizeof(mifi_t))
1704 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1707 ret = mif6_delete(mrt, mifi, NULL);
1712 * Manipulate the forwarding caches. These live
1713 * in a sort of kernel/user symbiosis.
1718 case MRT6_ADD_MFC_PROXY:
1719 case MRT6_DEL_MFC_PROXY:
1720 if (optlen < sizeof(mfc))
1722 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1725 parent = mfc.mf6cc_parent;
1727 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1728 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1730 ret = ip6mr_mfc_add(net, mrt, &mfc,
1731 sk == mrt->mroute6_sk, parent);
1736 * Control PIM assert (to activate pim will activate assert)
1742 if (optlen != sizeof(v))
1744 if (get_user(v, (int __user *)optval))
1746 mrt->mroute_do_assert = v;
1750 #ifdef CONFIG_IPV6_PIMSM_V2
1755 if (optlen != sizeof(v))
1757 if (get_user(v, (int __user *)optval))
1762 if (v != mrt->mroute_do_pim) {
1763 mrt->mroute_do_pim = v;
1764 mrt->mroute_do_assert = v;
1771 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1776 if (optlen != sizeof(u32))
1778 if (get_user(v, (u32 __user *)optval))
1780 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1781 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1783 if (sk == mrt->mroute6_sk)
1788 if (!ip6mr_new_table(net, v))
1790 raw6_sk(sk)->ip6mr_table = v;
1796 * Spurious command, or MRT6_VERSION which you cannot
1800 return -ENOPROTOOPT;
1805 * Getsock opt support for the multicast routing system.
1808 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1813 struct net *net = sock_net(sk);
1814 struct mr6_table *mrt;
1816 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1824 #ifdef CONFIG_IPV6_PIMSM_V2
1826 val = mrt->mroute_do_pim;
1830 val = mrt->mroute_do_assert;
1833 return -ENOPROTOOPT;
1836 if (get_user(olr, optlen))
1839 olr = min_t(int, olr, sizeof(int));
1843 if (put_user(olr, optlen))
1845 if (copy_to_user(optval, &val, olr))
1851 * The IP multicast ioctl support routines.
1854 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1856 struct sioc_sg_req6 sr;
1857 struct sioc_mif_req6 vr;
1858 struct mif_device *vif;
1859 struct mfc6_cache *c;
1860 struct net *net = sock_net(sk);
1861 struct mr6_table *mrt;
1863 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1868 case SIOCGETMIFCNT_IN6:
1869 if (copy_from_user(&vr, arg, sizeof(vr)))
1871 if (vr.mifi >= mrt->maxvif)
1873 read_lock(&mrt_lock);
1874 vif = &mrt->vif6_table[vr.mifi];
1875 if (MIF_EXISTS(mrt, vr.mifi)) {
1876 vr.icount = vif->pkt_in;
1877 vr.ocount = vif->pkt_out;
1878 vr.ibytes = vif->bytes_in;
1879 vr.obytes = vif->bytes_out;
1880 read_unlock(&mrt_lock);
1882 if (copy_to_user(arg, &vr, sizeof(vr)))
1886 read_unlock(&mrt_lock);
1887 return -EADDRNOTAVAIL;
1888 case SIOCGETSGCNT_IN6:
1889 if (copy_from_user(&sr, arg, sizeof(sr)))
1892 read_lock(&mrt_lock);
1893 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1895 sr.pktcnt = c->mfc_un.res.pkt;
1896 sr.bytecnt = c->mfc_un.res.bytes;
1897 sr.wrong_if = c->mfc_un.res.wrong_if;
1898 read_unlock(&mrt_lock);
1900 if (copy_to_user(arg, &sr, sizeof(sr)))
1904 read_unlock(&mrt_lock);
1905 return -EADDRNOTAVAIL;
1907 return -ENOIOCTLCMD;
1911 #ifdef CONFIG_COMPAT
1912 struct compat_sioc_sg_req6 {
1913 struct sockaddr_in6 src;
1914 struct sockaddr_in6 grp;
1915 compat_ulong_t pktcnt;
1916 compat_ulong_t bytecnt;
1917 compat_ulong_t wrong_if;
1920 struct compat_sioc_mif_req6 {
1922 compat_ulong_t icount;
1923 compat_ulong_t ocount;
1924 compat_ulong_t ibytes;
1925 compat_ulong_t obytes;
1928 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1930 struct compat_sioc_sg_req6 sr;
1931 struct compat_sioc_mif_req6 vr;
1932 struct mif_device *vif;
1933 struct mfc6_cache *c;
1934 struct net *net = sock_net(sk);
1935 struct mr6_table *mrt;
1937 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1942 case SIOCGETMIFCNT_IN6:
1943 if (copy_from_user(&vr, arg, sizeof(vr)))
1945 if (vr.mifi >= mrt->maxvif)
1947 read_lock(&mrt_lock);
1948 vif = &mrt->vif6_table[vr.mifi];
1949 if (MIF_EXISTS(mrt, vr.mifi)) {
1950 vr.icount = vif->pkt_in;
1951 vr.ocount = vif->pkt_out;
1952 vr.ibytes = vif->bytes_in;
1953 vr.obytes = vif->bytes_out;
1954 read_unlock(&mrt_lock);
1956 if (copy_to_user(arg, &vr, sizeof(vr)))
1960 read_unlock(&mrt_lock);
1961 return -EADDRNOTAVAIL;
1962 case SIOCGETSGCNT_IN6:
1963 if (copy_from_user(&sr, arg, sizeof(sr)))
1966 read_lock(&mrt_lock);
1967 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1969 sr.pktcnt = c->mfc_un.res.pkt;
1970 sr.bytecnt = c->mfc_un.res.bytes;
1971 sr.wrong_if = c->mfc_un.res.wrong_if;
1972 read_unlock(&mrt_lock);
1974 if (copy_to_user(arg, &sr, sizeof(sr)))
1978 read_unlock(&mrt_lock);
1979 return -EADDRNOTAVAIL;
1981 return -ENOIOCTLCMD;
1986 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1988 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
1989 IPSTATS_MIB_OUTFORWDATAGRAMS);
1990 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)),
1991 IPSTATS_MIB_OUTOCTETS, skb->len);
1992 return dst_output(net, sk, skb);
1996 * Processing handlers for ip6mr_forward
1999 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
2000 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
2002 struct ipv6hdr *ipv6h;
2003 struct mif_device *vif = &mrt->vif6_table[vifi];
2004 struct net_device *dev;
2005 struct dst_entry *dst;
2011 #ifdef CONFIG_IPV6_PIMSM_V2
2012 if (vif->flags & MIFF_REGISTER) {
2014 vif->bytes_out += skb->len;
2015 vif->dev->stats.tx_bytes += skb->len;
2016 vif->dev->stats.tx_packets++;
2017 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2022 ipv6h = ipv6_hdr(skb);
2024 fl6 = (struct flowi6) {
2025 .flowi6_oif = vif->link,
2026 .daddr = ipv6h->daddr,
2029 dst = ip6_route_output(net, NULL, &fl6);
2036 skb_dst_set(skb, dst);
2039 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2040 * not only before forwarding, but after forwarding on all output
2041 * interfaces. It is clear, if mrouter runs a multicasting
2042 * program, it should receive packets not depending to what interface
2043 * program is joined.
2044 * If we will not make it, the program will have to join on all
2045 * interfaces. On the other hand, multihoming host (or router, but
2046 * not mrouter) cannot join to more than one interface - it will
2047 * result in receiving multiple packets.
2052 vif->bytes_out += skb->len;
2054 /* We are about to write */
2055 /* XXX: extension headers? */
2056 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2059 ipv6h = ipv6_hdr(skb);
2062 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2064 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2065 net, NULL, skb, skb->dev, dev,
2066 ip6mr_forward2_finish);
2073 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2077 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2078 if (mrt->vif6_table[ct].dev == dev)
2084 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2085 struct sk_buff *skb, struct mfc6_cache *cache)
2089 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2091 vif = cache->mf6c_parent;
2092 cache->mfc_un.res.pkt++;
2093 cache->mfc_un.res.bytes += skb->len;
2095 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2096 struct mfc6_cache *cache_proxy;
2098 /* For an (*,G) entry, we only check that the incoming
2099 * interface is part of the static tree.
2101 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2103 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2108 * Wrong interface: drop packet and (maybe) send PIM assert.
2110 if (mrt->vif6_table[vif].dev != skb->dev) {
2111 cache->mfc_un.res.wrong_if++;
2113 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2114 /* pimsm uses asserts, when switching from RPT to SPT,
2115 so that we cannot check that packet arrived on an oif.
2116 It is bad, but otherwise we would need to move pretty
2117 large chunk of pimd to kernel. Ough... --ANK
2119 (mrt->mroute_do_pim ||
2120 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2122 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2123 cache->mfc_un.res.last_assert = jiffies;
2124 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2130 mrt->vif6_table[vif].pkt_in++;
2131 mrt->vif6_table[vif].bytes_in += skb->len;
2136 if (ipv6_addr_any(&cache->mf6c_origin) &&
2137 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2138 if (true_vifi >= 0 &&
2139 true_vifi != cache->mf6c_parent &&
2140 ipv6_hdr(skb)->hop_limit >
2141 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2142 /* It's an (*,*) entry and the packet is not coming from
2143 * the upstream: forward the packet to the upstream
2146 psend = cache->mf6c_parent;
2151 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2152 /* For (*,G) entry, don't forward to the incoming interface */
2153 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2154 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2156 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2158 ip6mr_forward2(net, mrt, skb2, cache, psend);
2165 ip6mr_forward2(net, mrt, skb, cache, psend);
2175 * Multicast packets for forwarding arrive here
2178 int ip6_mr_input(struct sk_buff *skb)
2180 struct mfc6_cache *cache;
2181 struct net *net = dev_net(skb->dev);
2182 struct mr6_table *mrt;
2183 struct flowi6 fl6 = {
2184 .flowi6_iif = skb->dev->ifindex,
2185 .flowi6_mark = skb->mark,
2189 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2195 read_lock(&mrt_lock);
2196 cache = ip6mr_cache_find(mrt,
2197 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2199 int vif = ip6mr_find_vif(mrt, skb->dev);
2202 cache = ip6mr_cache_find_any(mrt,
2203 &ipv6_hdr(skb)->daddr,
2208 * No usable cache entry
2213 vif = ip6mr_find_vif(mrt, skb->dev);
2215 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2216 read_unlock(&mrt_lock);
2220 read_unlock(&mrt_lock);
2225 ip6_mr_forward(net, mrt, skb, cache);
2227 read_unlock(&mrt_lock);
2233 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2234 struct mfc6_cache *c, struct rtmsg *rtm)
2237 struct rtnexthop *nhp;
2238 struct nlattr *mp_attr;
2239 struct rta_mfc_stats mfcs;
2241 /* If cache is unresolved, don't try to parse IIF and OIF */
2242 if (c->mf6c_parent >= MAXMIFS)
2245 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2246 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2248 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2252 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2253 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2254 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2256 nla_nest_cancel(skb, mp_attr);
2260 nhp->rtnh_flags = 0;
2261 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2262 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2263 nhp->rtnh_len = sizeof(*nhp);
2267 nla_nest_end(skb, mp_attr);
2269 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2270 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2271 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2272 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) < 0)
2275 rtm->rtm_type = RTN_MULTICAST;
2279 int ip6mr_get_route(struct net *net,
2280 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2283 struct mr6_table *mrt;
2284 struct mfc6_cache *cache;
2285 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2287 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2291 read_lock(&mrt_lock);
2292 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2293 if (!cache && skb->dev) {
2294 int vif = ip6mr_find_vif(mrt, skb->dev);
2297 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2302 struct sk_buff *skb2;
2303 struct ipv6hdr *iph;
2304 struct net_device *dev;
2308 read_unlock(&mrt_lock);
2313 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2314 read_unlock(&mrt_lock);
2318 /* really correct? */
2319 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2321 read_unlock(&mrt_lock);
2325 skb_reset_transport_header(skb2);
2327 skb_put(skb2, sizeof(struct ipv6hdr));
2328 skb_reset_network_header(skb2);
2330 iph = ipv6_hdr(skb2);
2333 iph->flow_lbl[0] = 0;
2334 iph->flow_lbl[1] = 0;
2335 iph->flow_lbl[2] = 0;
2336 iph->payload_len = 0;
2337 iph->nexthdr = IPPROTO_NONE;
2339 iph->saddr = rt->rt6i_src.addr;
2340 iph->daddr = rt->rt6i_dst.addr;
2342 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2343 read_unlock(&mrt_lock);
2348 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2349 cache->mfc_flags |= MFC_NOTIFY;
2351 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2352 read_unlock(&mrt_lock);
2356 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2357 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2360 struct nlmsghdr *nlh;
2364 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2368 rtm = nlmsg_data(nlh);
2369 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2370 rtm->rtm_dst_len = 128;
2371 rtm->rtm_src_len = 128;
2373 rtm->rtm_table = mrt->id;
2374 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2375 goto nla_put_failure;
2376 rtm->rtm_type = RTN_MULTICAST;
2377 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2378 if (c->mfc_flags & MFC_STATIC)
2379 rtm->rtm_protocol = RTPROT_STATIC;
2381 rtm->rtm_protocol = RTPROT_MROUTED;
2384 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2385 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2386 goto nla_put_failure;
2387 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2388 /* do not break the dump if cache is unresolved */
2389 if (err < 0 && err != -ENOENT)
2390 goto nla_put_failure;
2392 nlmsg_end(skb, nlh);
2396 nlmsg_cancel(skb, nlh);
2400 static int mr6_msgsize(bool unresolved, int maxvif)
2403 NLMSG_ALIGN(sizeof(struct rtmsg))
2404 + nla_total_size(4) /* RTA_TABLE */
2405 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2406 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2411 + nla_total_size(4) /* RTA_IIF */
2412 + nla_total_size(0) /* RTA_MULTIPATH */
2413 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2415 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2421 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2424 struct net *net = read_pnet(&mrt->net);
2425 struct sk_buff *skb;
2428 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2433 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2437 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2443 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2446 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2448 struct net *net = sock_net(skb->sk);
2449 struct mr6_table *mrt;
2450 struct mfc6_cache *mfc;
2451 unsigned int t = 0, s_t;
2452 unsigned int h = 0, s_h;
2453 unsigned int e = 0, s_e;
2459 read_lock(&mrt_lock);
2460 ip6mr_for_each_table(mrt, net) {
2465 for (h = s_h; h < MFC6_LINES; h++) {
2466 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2469 if (ip6mr_fill_mroute(mrt, skb,
2470 NETLINK_CB(cb->skb).portid,
2480 spin_lock_bh(&mfc_unres_lock);
2481 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2484 if (ip6mr_fill_mroute(mrt, skb,
2485 NETLINK_CB(cb->skb).portid,
2489 spin_unlock_bh(&mfc_unres_lock);
2495 spin_unlock_bh(&mfc_unres_lock);
2502 read_unlock(&mrt_lock);