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;
63 struct sock *mroute6_sk;
64 struct timer_list ipmr_expire_timer;
65 struct list_head mfc6_unres_queue;
66 struct list_head mfc6_cache_array[MFC6_LINES];
67 struct mif_device vif6_table[MAXMIFS];
69 atomic_t cache_resolve_queue_len;
70 bool mroute_do_assert;
72 #ifdef CONFIG_IPV6_PIMSM_V2
73 int mroute_reg_vif_num;
78 struct fib_rule common;
82 struct mr6_table *mrt;
85 /* Big lock, protecting vif table, mrt cache and mroute socket state.
86 Note that the changes are semaphored via rtnl_lock.
89 static DEFINE_RWLOCK(mrt_lock);
92 * Multicast router control variables
95 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
97 /* Special spinlock for queue of unresolved entries */
98 static DEFINE_SPINLOCK(mfc_unres_lock);
100 /* We return to original Alan's scheme. Hash table of resolved
101 entries is changed only in process context and protected
102 with weak lock mrt_lock. Queue of unresolved entries is protected
103 with strong spinlock mfc_unres_lock.
105 In this case data path is free of exclusive locks at all.
108 static struct kmem_cache *mrt_cachep __read_mostly;
110 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
111 static void ip6mr_free_table(struct mr6_table *mrt);
113 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
114 struct sk_buff *skb, struct mfc6_cache *cache);
115 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
116 mifi_t mifi, int assert);
117 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
118 struct mfc6_cache *c, struct rtmsg *rtm);
119 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
121 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
122 struct netlink_callback *cb);
123 static void mroute_clean_tables(struct mr6_table *mrt);
124 static void ipmr_expire_process(unsigned long arg);
126 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
127 #define ip6mr_for_each_table(mrt, net) \
128 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
130 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
132 struct mr6_table *mrt;
134 ip6mr_for_each_table(mrt, net) {
141 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
142 struct mr6_table **mrt)
145 struct ip6mr_result res;
146 struct fib_lookup_arg arg = {
148 .flags = FIB_LOOKUP_NOREF,
151 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
152 flowi6_to_flowi(flp6), 0, &arg);
159 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
160 int flags, struct fib_lookup_arg *arg)
162 struct ip6mr_result *res = arg->result;
163 struct mr6_table *mrt;
165 switch (rule->action) {
168 case FR_ACT_UNREACHABLE:
170 case FR_ACT_PROHIBIT:
172 case FR_ACT_BLACKHOLE:
177 mrt = ip6mr_get_table(rule->fr_net, rule->table);
184 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
189 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
193 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
194 struct fib_rule_hdr *frh, struct nlattr **tb)
199 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
205 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
206 struct fib_rule_hdr *frh)
214 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
215 .family = RTNL_FAMILY_IP6MR,
216 .rule_size = sizeof(struct ip6mr_rule),
217 .addr_size = sizeof(struct in6_addr),
218 .action = ip6mr_rule_action,
219 .match = ip6mr_rule_match,
220 .configure = ip6mr_rule_configure,
221 .compare = ip6mr_rule_compare,
222 .default_pref = fib_default_rule_pref,
223 .fill = ip6mr_rule_fill,
224 .nlgroup = RTNLGRP_IPV6_RULE,
225 .policy = ip6mr_rule_policy,
226 .owner = THIS_MODULE,
229 static int __net_init ip6mr_rules_init(struct net *net)
231 struct fib_rules_ops *ops;
232 struct mr6_table *mrt;
235 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
239 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
241 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
247 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
251 net->ipv6.mr6_rules_ops = ops;
257 fib_rules_unregister(ops);
261 static void __net_exit ip6mr_rules_exit(struct net *net)
263 struct mr6_table *mrt, *next;
266 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
267 list_del(&mrt->list);
268 ip6mr_free_table(mrt);
271 fib_rules_unregister(net->ipv6.mr6_rules_ops);
274 #define ip6mr_for_each_table(mrt, net) \
275 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
277 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
279 return net->ipv6.mrt6;
282 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
283 struct mr6_table **mrt)
285 *mrt = net->ipv6.mrt6;
289 static int __net_init ip6mr_rules_init(struct net *net)
291 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
292 return net->ipv6.mrt6 ? 0 : -ENOMEM;
295 static void __net_exit ip6mr_rules_exit(struct net *net)
298 ip6mr_free_table(net->ipv6.mrt6);
299 net->ipv6.mrt6 = NULL;
304 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
306 struct mr6_table *mrt;
309 mrt = ip6mr_get_table(net, id);
313 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
317 write_pnet(&mrt->net, net);
319 /* Forwarding cache */
320 for (i = 0; i < MFC6_LINES; i++)
321 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
323 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
325 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
328 #ifdef CONFIG_IPV6_PIMSM_V2
329 mrt->mroute_reg_vif_num = -1;
331 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
332 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
337 static void ip6mr_free_table(struct mr6_table *mrt)
339 del_timer(&mrt->ipmr_expire_timer);
340 mroute_clean_tables(mrt);
344 #ifdef CONFIG_PROC_FS
346 struct ipmr_mfc_iter {
347 struct seq_net_private p;
348 struct mr6_table *mrt;
349 struct list_head *cache;
354 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
355 struct ipmr_mfc_iter *it, loff_t pos)
357 struct mr6_table *mrt = it->mrt;
358 struct mfc6_cache *mfc;
360 read_lock(&mrt_lock);
361 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
362 it->cache = &mrt->mfc6_cache_array[it->ct];
363 list_for_each_entry(mfc, it->cache, list)
367 read_unlock(&mrt_lock);
369 spin_lock_bh(&mfc_unres_lock);
370 it->cache = &mrt->mfc6_unres_queue;
371 list_for_each_entry(mfc, it->cache, list)
374 spin_unlock_bh(&mfc_unres_lock);
381 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
384 struct ipmr_vif_iter {
385 struct seq_net_private p;
386 struct mr6_table *mrt;
390 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
391 struct ipmr_vif_iter *iter,
394 struct mr6_table *mrt = iter->mrt;
396 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
397 if (!MIF_EXISTS(mrt, iter->ct))
400 return &mrt->vif6_table[iter->ct];
405 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
408 struct ipmr_vif_iter *iter = seq->private;
409 struct net *net = seq_file_net(seq);
410 struct mr6_table *mrt;
412 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
414 return ERR_PTR(-ENOENT);
418 read_lock(&mrt_lock);
419 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
423 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
425 struct ipmr_vif_iter *iter = seq->private;
426 struct net *net = seq_file_net(seq);
427 struct mr6_table *mrt = iter->mrt;
430 if (v == SEQ_START_TOKEN)
431 return ip6mr_vif_seq_idx(net, iter, 0);
433 while (++iter->ct < mrt->maxvif) {
434 if (!MIF_EXISTS(mrt, iter->ct))
436 return &mrt->vif6_table[iter->ct];
441 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
444 read_unlock(&mrt_lock);
447 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
449 struct ipmr_vif_iter *iter = seq->private;
450 struct mr6_table *mrt = iter->mrt;
452 if (v == SEQ_START_TOKEN) {
454 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
456 const struct mif_device *vif = v;
457 const char *name = vif->dev ? vif->dev->name : "none";
460 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
461 vif - mrt->vif6_table,
462 name, vif->bytes_in, vif->pkt_in,
463 vif->bytes_out, vif->pkt_out,
469 static const struct seq_operations ip6mr_vif_seq_ops = {
470 .start = ip6mr_vif_seq_start,
471 .next = ip6mr_vif_seq_next,
472 .stop = ip6mr_vif_seq_stop,
473 .show = ip6mr_vif_seq_show,
476 static int ip6mr_vif_open(struct inode *inode, struct file *file)
478 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
479 sizeof(struct ipmr_vif_iter));
482 static const struct file_operations ip6mr_vif_fops = {
483 .owner = THIS_MODULE,
484 .open = ip6mr_vif_open,
487 .release = seq_release_net,
490 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
492 struct ipmr_mfc_iter *it = seq->private;
493 struct net *net = seq_file_net(seq);
494 struct mr6_table *mrt;
496 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
498 return ERR_PTR(-ENOENT);
501 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
505 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
507 struct mfc6_cache *mfc = v;
508 struct ipmr_mfc_iter *it = seq->private;
509 struct net *net = seq_file_net(seq);
510 struct mr6_table *mrt = it->mrt;
514 if (v == SEQ_START_TOKEN)
515 return ipmr_mfc_seq_idx(net, seq->private, 0);
517 if (mfc->list.next != it->cache)
518 return list_entry(mfc->list.next, struct mfc6_cache, list);
520 if (it->cache == &mrt->mfc6_unres_queue)
523 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
525 while (++it->ct < MFC6_LINES) {
526 it->cache = &mrt->mfc6_cache_array[it->ct];
527 if (list_empty(it->cache))
529 return list_first_entry(it->cache, struct mfc6_cache, list);
532 /* exhausted cache_array, show unresolved */
533 read_unlock(&mrt_lock);
534 it->cache = &mrt->mfc6_unres_queue;
537 spin_lock_bh(&mfc_unres_lock);
538 if (!list_empty(it->cache))
539 return list_first_entry(it->cache, struct mfc6_cache, list);
542 spin_unlock_bh(&mfc_unres_lock);
548 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
550 struct ipmr_mfc_iter *it = seq->private;
551 struct mr6_table *mrt = it->mrt;
553 if (it->cache == &mrt->mfc6_unres_queue)
554 spin_unlock_bh(&mfc_unres_lock);
555 else if (it->cache == mrt->mfc6_cache_array)
556 read_unlock(&mrt_lock);
559 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
563 if (v == SEQ_START_TOKEN) {
567 "Iif Pkts Bytes Wrong Oifs\n");
569 const struct mfc6_cache *mfc = v;
570 const struct ipmr_mfc_iter *it = seq->private;
571 struct mr6_table *mrt = it->mrt;
573 seq_printf(seq, "%pI6 %pI6 %-3hd",
574 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
577 if (it->cache != &mrt->mfc6_unres_queue) {
578 seq_printf(seq, " %8lu %8lu %8lu",
580 mfc->mfc_un.res.bytes,
581 mfc->mfc_un.res.wrong_if);
582 for (n = mfc->mfc_un.res.minvif;
583 n < mfc->mfc_un.res.maxvif; n++) {
584 if (MIF_EXISTS(mrt, n) &&
585 mfc->mfc_un.res.ttls[n] < 255)
588 n, mfc->mfc_un.res.ttls[n]);
591 /* unresolved mfc_caches don't contain
592 * pkt, bytes and wrong_if values
594 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
601 static const struct seq_operations ipmr_mfc_seq_ops = {
602 .start = ipmr_mfc_seq_start,
603 .next = ipmr_mfc_seq_next,
604 .stop = ipmr_mfc_seq_stop,
605 .show = ipmr_mfc_seq_show,
608 static int ipmr_mfc_open(struct inode *inode, struct file *file)
610 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
611 sizeof(struct ipmr_mfc_iter));
614 static const struct file_operations ip6mr_mfc_fops = {
615 .owner = THIS_MODULE,
616 .open = ipmr_mfc_open,
619 .release = seq_release_net,
623 #ifdef CONFIG_IPV6_PIMSM_V2
625 static int pim6_rcv(struct sk_buff *skb)
627 struct pimreghdr *pim;
628 struct ipv6hdr *encap;
629 struct net_device *reg_dev = NULL;
630 struct net *net = dev_net(skb->dev);
631 struct mr6_table *mrt;
632 struct flowi6 fl6 = {
633 .flowi6_iif = skb->dev->ifindex,
634 .flowi6_mark = skb->mark,
638 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
641 pim = (struct pimreghdr *)skb_transport_header(skb);
642 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
643 (pim->flags & PIM_NULL_REGISTER) ||
644 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
645 sizeof(*pim), IPPROTO_PIM,
646 csum_partial((void *)pim, sizeof(*pim), 0)) &&
647 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
650 /* check if the inner packet is destined to mcast group */
651 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
654 if (!ipv6_addr_is_multicast(&encap->daddr) ||
655 encap->payload_len == 0 ||
656 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
659 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
661 reg_vif_num = mrt->mroute_reg_vif_num;
663 read_lock(&mrt_lock);
664 if (reg_vif_num >= 0)
665 reg_dev = mrt->vif6_table[reg_vif_num].dev;
668 read_unlock(&mrt_lock);
673 skb->mac_header = skb->network_header;
674 skb_pull(skb, (u8 *)encap - skb->data);
675 skb_reset_network_header(skb);
676 skb->protocol = htons(ETH_P_IPV6);
677 skb->ip_summed = CHECKSUM_NONE;
679 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
690 static const struct inet6_protocol pim6_protocol = {
694 /* Service routines creating virtual interfaces: PIMREG */
696 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
697 struct net_device *dev)
699 struct net *net = dev_net(dev);
700 struct mr6_table *mrt;
701 struct flowi6 fl6 = {
702 .flowi6_oif = dev->ifindex,
703 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
704 .flowi6_mark = skb->mark,
708 err = ip6mr_fib_lookup(net, &fl6, &mrt);
714 read_lock(&mrt_lock);
715 dev->stats.tx_bytes += skb->len;
716 dev->stats.tx_packets++;
717 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
718 read_unlock(&mrt_lock);
723 static const struct net_device_ops reg_vif_netdev_ops = {
724 .ndo_start_xmit = reg_vif_xmit,
727 static void reg_vif_setup(struct net_device *dev)
729 dev->type = ARPHRD_PIMREG;
730 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
731 dev->flags = IFF_NOARP;
732 dev->netdev_ops = ®_vif_netdev_ops;
733 dev->destructor = free_netdev;
734 dev->features |= NETIF_F_NETNS_LOCAL;
737 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
739 struct net_device *dev;
742 if (mrt->id == RT6_TABLE_DFLT)
743 sprintf(name, "pim6reg");
745 sprintf(name, "pim6reg%u", mrt->id);
747 dev = alloc_netdev(0, name, reg_vif_setup);
751 dev_net_set(dev, net);
753 if (register_netdevice(dev)) {
766 /* allow the register to be completed before unregistering. */
770 unregister_netdevice(dev);
779 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
781 struct mif_device *v;
782 struct net_device *dev;
783 struct inet6_dev *in6_dev;
785 if (vifi < 0 || vifi >= mrt->maxvif)
786 return -EADDRNOTAVAIL;
788 v = &mrt->vif6_table[vifi];
790 write_lock_bh(&mrt_lock);
795 write_unlock_bh(&mrt_lock);
796 return -EADDRNOTAVAIL;
799 #ifdef CONFIG_IPV6_PIMSM_V2
800 if (vifi == mrt->mroute_reg_vif_num)
801 mrt->mroute_reg_vif_num = -1;
804 if (vifi + 1 == mrt->maxvif) {
806 for (tmp = vifi - 1; tmp >= 0; tmp--) {
807 if (MIF_EXISTS(mrt, tmp))
810 mrt->maxvif = tmp + 1;
813 write_unlock_bh(&mrt_lock);
815 dev_set_allmulti(dev, -1);
817 in6_dev = __in6_dev_get(dev);
819 in6_dev->cnf.mc_forwarding--;
820 inet6_netconf_notify_devconf(dev_net(dev),
821 NETCONFA_MC_FORWARDING,
822 dev->ifindex, &in6_dev->cnf);
825 if (v->flags & MIFF_REGISTER)
826 unregister_netdevice_queue(dev, head);
832 static inline void ip6mr_cache_free(struct mfc6_cache *c)
834 kmem_cache_free(mrt_cachep, c);
837 /* Destroy an unresolved cache entry, killing queued skbs
838 and reporting error to netlink readers.
841 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
843 struct net *net = read_pnet(&mrt->net);
846 atomic_dec(&mrt->cache_resolve_queue_len);
848 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
849 if (ipv6_hdr(skb)->version == 0) {
850 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
851 nlh->nlmsg_type = NLMSG_ERROR;
852 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
853 skb_trim(skb, nlh->nlmsg_len);
854 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
855 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
864 /* Timer process for all the unresolved queue. */
866 static void ipmr_do_expire_process(struct mr6_table *mrt)
868 unsigned long now = jiffies;
869 unsigned long expires = 10 * HZ;
870 struct mfc6_cache *c, *next;
872 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
873 if (time_after(c->mfc_un.unres.expires, now)) {
875 unsigned long interval = c->mfc_un.unres.expires - now;
876 if (interval < expires)
882 mr6_netlink_event(mrt, c, RTM_DELROUTE);
883 ip6mr_destroy_unres(mrt, c);
886 if (!list_empty(&mrt->mfc6_unres_queue))
887 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
890 static void ipmr_expire_process(unsigned long arg)
892 struct mr6_table *mrt = (struct mr6_table *)arg;
894 if (!spin_trylock(&mfc_unres_lock)) {
895 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
899 if (!list_empty(&mrt->mfc6_unres_queue))
900 ipmr_do_expire_process(mrt);
902 spin_unlock(&mfc_unres_lock);
905 /* Fill oifs list. It is called under write locked mrt_lock. */
907 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
912 cache->mfc_un.res.minvif = MAXMIFS;
913 cache->mfc_un.res.maxvif = 0;
914 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
916 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
917 if (MIF_EXISTS(mrt, vifi) &&
918 ttls[vifi] && ttls[vifi] < 255) {
919 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
920 if (cache->mfc_un.res.minvif > vifi)
921 cache->mfc_un.res.minvif = vifi;
922 if (cache->mfc_un.res.maxvif <= vifi)
923 cache->mfc_un.res.maxvif = vifi + 1;
928 static int mif6_add(struct net *net, struct mr6_table *mrt,
929 struct mif6ctl *vifc, int mrtsock)
931 int vifi = vifc->mif6c_mifi;
932 struct mif_device *v = &mrt->vif6_table[vifi];
933 struct net_device *dev;
934 struct inet6_dev *in6_dev;
938 if (MIF_EXISTS(mrt, vifi))
941 switch (vifc->mif6c_flags) {
942 #ifdef CONFIG_IPV6_PIMSM_V2
945 * Special Purpose VIF in PIM
946 * All the packets will be sent to the daemon
948 if (mrt->mroute_reg_vif_num >= 0)
950 dev = ip6mr_reg_vif(net, mrt);
953 err = dev_set_allmulti(dev, 1);
955 unregister_netdevice(dev);
962 dev = dev_get_by_index(net, vifc->mif6c_pifi);
964 return -EADDRNOTAVAIL;
965 err = dev_set_allmulti(dev, 1);
975 in6_dev = __in6_dev_get(dev);
977 in6_dev->cnf.mc_forwarding++;
978 inet6_netconf_notify_devconf(dev_net(dev),
979 NETCONFA_MC_FORWARDING,
980 dev->ifindex, &in6_dev->cnf);
984 * Fill in the VIF structures
986 v->rate_limit = vifc->vifc_rate_limit;
987 v->flags = vifc->mif6c_flags;
989 v->flags |= VIFF_STATIC;
990 v->threshold = vifc->vifc_threshold;
995 v->link = dev->ifindex;
996 if (v->flags & MIFF_REGISTER)
997 v->link = dev->iflink;
999 /* And finish update writing critical data */
1000 write_lock_bh(&mrt_lock);
1002 #ifdef CONFIG_IPV6_PIMSM_V2
1003 if (v->flags & MIFF_REGISTER)
1004 mrt->mroute_reg_vif_num = vifi;
1006 if (vifi + 1 > mrt->maxvif)
1007 mrt->maxvif = vifi + 1;
1008 write_unlock_bh(&mrt_lock);
1012 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1013 const struct in6_addr *origin,
1014 const struct in6_addr *mcastgrp)
1016 int line = MFC6_HASH(mcastgrp, origin);
1017 struct mfc6_cache *c;
1019 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1020 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1021 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1027 /* Look for a (*,*,oif) entry */
1028 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1031 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1032 struct mfc6_cache *c;
1034 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1035 if (ipv6_addr_any(&c->mf6c_origin) &&
1036 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1037 (c->mfc_un.res.ttls[mifi] < 255))
1043 /* Look for a (*,G) entry */
1044 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1045 struct in6_addr *mcastgrp,
1048 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1049 struct mfc6_cache *c, *proxy;
1051 if (ipv6_addr_any(mcastgrp))
1054 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1055 if (ipv6_addr_any(&c->mf6c_origin) &&
1056 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1057 if (c->mfc_un.res.ttls[mifi] < 255)
1060 /* It's ok if the mifi is part of the static tree */
1061 proxy = ip6mr_cache_find_any_parent(mrt,
1063 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1068 return ip6mr_cache_find_any_parent(mrt, mifi);
1072 * Allocate a multicast cache entry
1074 static struct mfc6_cache *ip6mr_cache_alloc(void)
1076 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1079 c->mfc_un.res.minvif = MAXMIFS;
1083 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1085 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1088 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1089 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1094 * A cache entry has gone into a resolved state from queued
1097 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1098 struct mfc6_cache *uc, struct mfc6_cache *c)
1100 struct sk_buff *skb;
1103 * Play the pending entries through our router
1106 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1107 if (ipv6_hdr(skb)->version == 0) {
1108 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1110 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1111 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1113 nlh->nlmsg_type = NLMSG_ERROR;
1114 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1115 skb_trim(skb, nlh->nlmsg_len);
1116 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1118 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1120 ip6_mr_forward(net, mrt, skb, c);
1125 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1126 * expects the following bizarre scheme.
1128 * Called under mrt_lock.
1131 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1132 mifi_t mifi, int assert)
1134 struct sk_buff *skb;
1135 struct mrt6msg *msg;
1138 #ifdef CONFIG_IPV6_PIMSM_V2
1139 if (assert == MRT6MSG_WHOLEPKT)
1140 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1144 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1149 /* I suppose that internal messages
1150 * do not require checksums */
1152 skb->ip_summed = CHECKSUM_UNNECESSARY;
1154 #ifdef CONFIG_IPV6_PIMSM_V2
1155 if (assert == MRT6MSG_WHOLEPKT) {
1156 /* Ugly, but we have no choice with this interface.
1157 Duplicate old header, fix length etc.
1158 And all this only to mangle msg->im6_msgtype and
1159 to set msg->im6_mbz to "mbz" :-)
1161 skb_push(skb, -skb_network_offset(pkt));
1163 skb_push(skb, sizeof(*msg));
1164 skb_reset_transport_header(skb);
1165 msg = (struct mrt6msg *)skb_transport_header(skb);
1167 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1168 msg->im6_mif = mrt->mroute_reg_vif_num;
1170 msg->im6_src = ipv6_hdr(pkt)->saddr;
1171 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1173 skb->ip_summed = CHECKSUM_UNNECESSARY;
1178 * Copy the IP header
1181 skb_put(skb, sizeof(struct ipv6hdr));
1182 skb_reset_network_header(skb);
1183 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1188 skb_put(skb, sizeof(*msg));
1189 skb_reset_transport_header(skb);
1190 msg = (struct mrt6msg *)skb_transport_header(skb);
1193 msg->im6_msgtype = assert;
1194 msg->im6_mif = mifi;
1196 msg->im6_src = ipv6_hdr(pkt)->saddr;
1197 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1199 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1200 skb->ip_summed = CHECKSUM_UNNECESSARY;
1203 if (mrt->mroute6_sk == NULL) {
1209 * Deliver to user space multicast routing algorithms
1211 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1213 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1221 * Queue a packet for resolution. It gets locked cache entry!
1225 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1229 struct mfc6_cache *c;
1231 spin_lock_bh(&mfc_unres_lock);
1232 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1233 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1234 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1242 * Create a new entry if allowable
1245 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1246 (c = ip6mr_cache_alloc_unres()) == NULL) {
1247 spin_unlock_bh(&mfc_unres_lock);
1254 * Fill in the new cache entry
1256 c->mf6c_parent = -1;
1257 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1258 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1261 * Reflect first query at pim6sd
1263 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1265 /* If the report failed throw the cache entry
1268 spin_unlock_bh(&mfc_unres_lock);
1270 ip6mr_cache_free(c);
1275 atomic_inc(&mrt->cache_resolve_queue_len);
1276 list_add(&c->list, &mrt->mfc6_unres_queue);
1277 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1279 ipmr_do_expire_process(mrt);
1283 * See if we can append the packet
1285 if (c->mfc_un.unres.unresolved.qlen > 3) {
1289 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1293 spin_unlock_bh(&mfc_unres_lock);
1298 * MFC6 cache manipulation by user space
1301 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1305 struct mfc6_cache *c, *next;
1307 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1309 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1310 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1311 ipv6_addr_equal(&c->mf6c_mcastgrp,
1312 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1313 (parent == -1 || parent == c->mf6c_parent)) {
1314 write_lock_bh(&mrt_lock);
1316 write_unlock_bh(&mrt_lock);
1318 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1319 ip6mr_cache_free(c);
1326 static int ip6mr_device_event(struct notifier_block *this,
1327 unsigned long event, void *ptr)
1329 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1330 struct net *net = dev_net(dev);
1331 struct mr6_table *mrt;
1332 struct mif_device *v;
1336 if (event != NETDEV_UNREGISTER)
1339 ip6mr_for_each_table(mrt, net) {
1340 v = &mrt->vif6_table[0];
1341 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1343 mif6_delete(mrt, ct, &list);
1346 unregister_netdevice_many(&list);
1351 static struct notifier_block ip6_mr_notifier = {
1352 .notifier_call = ip6mr_device_event
1356 * Setup for IP multicast routing
1359 static int __net_init ip6mr_net_init(struct net *net)
1363 err = ip6mr_rules_init(net);
1367 #ifdef CONFIG_PROC_FS
1369 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1371 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1372 goto proc_cache_fail;
1377 #ifdef CONFIG_PROC_FS
1379 remove_proc_entry("ip6_mr_vif", net->proc_net);
1381 ip6mr_rules_exit(net);
1387 static void __net_exit ip6mr_net_exit(struct net *net)
1389 #ifdef CONFIG_PROC_FS
1390 remove_proc_entry("ip6_mr_cache", net->proc_net);
1391 remove_proc_entry("ip6_mr_vif", net->proc_net);
1393 ip6mr_rules_exit(net);
1396 static struct pernet_operations ip6mr_net_ops = {
1397 .init = ip6mr_net_init,
1398 .exit = ip6mr_net_exit,
1401 int __init ip6_mr_init(void)
1405 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1406 sizeof(struct mfc6_cache),
1407 0, SLAB_HWCACHE_ALIGN,
1412 err = register_pernet_subsys(&ip6mr_net_ops);
1414 goto reg_pernet_fail;
1416 err = register_netdevice_notifier(&ip6_mr_notifier);
1418 goto reg_notif_fail;
1419 #ifdef CONFIG_IPV6_PIMSM_V2
1420 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1421 pr_err("%s: can't add PIM protocol\n", __func__);
1423 goto add_proto_fail;
1426 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1427 ip6mr_rtm_dumproute, NULL);
1429 #ifdef CONFIG_IPV6_PIMSM_V2
1431 unregister_netdevice_notifier(&ip6_mr_notifier);
1434 unregister_pernet_subsys(&ip6mr_net_ops);
1436 kmem_cache_destroy(mrt_cachep);
1440 void ip6_mr_cleanup(void)
1442 unregister_netdevice_notifier(&ip6_mr_notifier);
1443 unregister_pernet_subsys(&ip6mr_net_ops);
1444 kmem_cache_destroy(mrt_cachep);
1447 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1448 struct mf6cctl *mfc, int mrtsock, int parent)
1452 struct mfc6_cache *uc, *c;
1453 unsigned char ttls[MAXMIFS];
1456 if (mfc->mf6cc_parent >= MAXMIFS)
1459 memset(ttls, 255, MAXMIFS);
1460 for (i = 0; i < MAXMIFS; i++) {
1461 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1466 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1468 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1469 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1470 ipv6_addr_equal(&c->mf6c_mcastgrp,
1471 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1472 (parent == -1 || parent == mfc->mf6cc_parent)) {
1479 write_lock_bh(&mrt_lock);
1480 c->mf6c_parent = mfc->mf6cc_parent;
1481 ip6mr_update_thresholds(mrt, c, ttls);
1483 c->mfc_flags |= MFC_STATIC;
1484 write_unlock_bh(&mrt_lock);
1485 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1489 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1490 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1493 c = ip6mr_cache_alloc();
1497 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1498 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1499 c->mf6c_parent = mfc->mf6cc_parent;
1500 ip6mr_update_thresholds(mrt, c, ttls);
1502 c->mfc_flags |= MFC_STATIC;
1504 write_lock_bh(&mrt_lock);
1505 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1506 write_unlock_bh(&mrt_lock);
1509 * Check to see if we resolved a queued list. If so we
1510 * need to send on the frames and tidy up.
1513 spin_lock_bh(&mfc_unres_lock);
1514 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1515 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1516 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1517 list_del(&uc->list);
1518 atomic_dec(&mrt->cache_resolve_queue_len);
1523 if (list_empty(&mrt->mfc6_unres_queue))
1524 del_timer(&mrt->ipmr_expire_timer);
1525 spin_unlock_bh(&mfc_unres_lock);
1528 ip6mr_cache_resolve(net, mrt, uc, c);
1529 ip6mr_cache_free(uc);
1531 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1536 * Close the multicast socket, and clear the vif tables etc
1539 static void mroute_clean_tables(struct mr6_table *mrt)
1543 struct mfc6_cache *c, *next;
1546 * Shut down all active vif entries
1548 for (i = 0; i < mrt->maxvif; i++) {
1549 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1550 mif6_delete(mrt, i, &list);
1552 unregister_netdevice_many(&list);
1557 for (i = 0; i < MFC6_LINES; i++) {
1558 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1559 if (c->mfc_flags & MFC_STATIC)
1561 write_lock_bh(&mrt_lock);
1563 write_unlock_bh(&mrt_lock);
1565 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1566 ip6mr_cache_free(c);
1570 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1571 spin_lock_bh(&mfc_unres_lock);
1572 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1574 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1575 ip6mr_destroy_unres(mrt, c);
1577 spin_unlock_bh(&mfc_unres_lock);
1581 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1584 struct net *net = sock_net(sk);
1587 write_lock_bh(&mrt_lock);
1588 if (likely(mrt->mroute6_sk == NULL)) {
1589 mrt->mroute6_sk = sk;
1590 net->ipv6.devconf_all->mc_forwarding++;
1591 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1592 NETCONFA_IFINDEX_ALL,
1593 net->ipv6.devconf_all);
1597 write_unlock_bh(&mrt_lock);
1604 int ip6mr_sk_done(struct sock *sk)
1607 struct net *net = sock_net(sk);
1608 struct mr6_table *mrt;
1611 ip6mr_for_each_table(mrt, net) {
1612 if (sk == mrt->mroute6_sk) {
1613 write_lock_bh(&mrt_lock);
1614 mrt->mroute6_sk = NULL;
1615 net->ipv6.devconf_all->mc_forwarding--;
1616 inet6_netconf_notify_devconf(net,
1617 NETCONFA_MC_FORWARDING,
1618 NETCONFA_IFINDEX_ALL,
1619 net->ipv6.devconf_all);
1620 write_unlock_bh(&mrt_lock);
1622 mroute_clean_tables(mrt);
1632 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1634 struct mr6_table *mrt;
1635 struct flowi6 fl6 = {
1636 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1637 .flowi6_oif = skb->dev->ifindex,
1638 .flowi6_mark = skb->mark,
1641 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1644 return mrt->mroute6_sk;
1648 * Socket options and virtual interface manipulation. The whole
1649 * virtual interface system is a complete heap, but unfortunately
1650 * that's how BSD mrouted happens to think. Maybe one day with a proper
1651 * MOSPF/PIM router set up we can clean this up.
1654 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1656 int ret, parent = 0;
1660 struct net *net = sock_net(sk);
1661 struct mr6_table *mrt;
1663 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1667 if (optname != MRT6_INIT) {
1668 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1674 if (sk->sk_type != SOCK_RAW ||
1675 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1677 if (optlen < sizeof(int))
1680 return ip6mr_sk_init(mrt, sk);
1683 return ip6mr_sk_done(sk);
1686 if (optlen < sizeof(vif))
1688 if (copy_from_user(&vif, optval, sizeof(vif)))
1690 if (vif.mif6c_mifi >= MAXMIFS)
1693 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1698 if (optlen < sizeof(mifi_t))
1700 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1703 ret = mif6_delete(mrt, mifi, NULL);
1708 * Manipulate the forwarding caches. These live
1709 * in a sort of kernel/user symbiosis.
1714 case MRT6_ADD_MFC_PROXY:
1715 case MRT6_DEL_MFC_PROXY:
1716 if (optlen < sizeof(mfc))
1718 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1721 parent = mfc.mf6cc_parent;
1723 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1724 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1726 ret = ip6mr_mfc_add(net, mrt, &mfc,
1727 sk == mrt->mroute6_sk, parent);
1732 * Control PIM assert (to activate pim will activate assert)
1738 if (optlen != sizeof(v))
1740 if (get_user(v, (int __user *)optval))
1742 mrt->mroute_do_assert = v;
1746 #ifdef CONFIG_IPV6_PIMSM_V2
1751 if (optlen != sizeof(v))
1753 if (get_user(v, (int __user *)optval))
1758 if (v != mrt->mroute_do_pim) {
1759 mrt->mroute_do_pim = v;
1760 mrt->mroute_do_assert = v;
1767 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1772 if (optlen != sizeof(u32))
1774 if (get_user(v, (u32 __user *)optval))
1776 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1777 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1779 if (sk == mrt->mroute6_sk)
1784 if (!ip6mr_new_table(net, v))
1786 raw6_sk(sk)->ip6mr_table = v;
1792 * Spurious command, or MRT6_VERSION which you cannot
1796 return -ENOPROTOOPT;
1801 * Getsock opt support for the multicast routing system.
1804 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1809 struct net *net = sock_net(sk);
1810 struct mr6_table *mrt;
1812 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1820 #ifdef CONFIG_IPV6_PIMSM_V2
1822 val = mrt->mroute_do_pim;
1826 val = mrt->mroute_do_assert;
1829 return -ENOPROTOOPT;
1832 if (get_user(olr, optlen))
1835 olr = min_t(int, olr, sizeof(int));
1839 if (put_user(olr, optlen))
1841 if (copy_to_user(optval, &val, olr))
1847 * The IP multicast ioctl support routines.
1850 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1852 struct sioc_sg_req6 sr;
1853 struct sioc_mif_req6 vr;
1854 struct mif_device *vif;
1855 struct mfc6_cache *c;
1856 struct net *net = sock_net(sk);
1857 struct mr6_table *mrt;
1859 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1864 case SIOCGETMIFCNT_IN6:
1865 if (copy_from_user(&vr, arg, sizeof(vr)))
1867 if (vr.mifi >= mrt->maxvif)
1869 read_lock(&mrt_lock);
1870 vif = &mrt->vif6_table[vr.mifi];
1871 if (MIF_EXISTS(mrt, vr.mifi)) {
1872 vr.icount = vif->pkt_in;
1873 vr.ocount = vif->pkt_out;
1874 vr.ibytes = vif->bytes_in;
1875 vr.obytes = vif->bytes_out;
1876 read_unlock(&mrt_lock);
1878 if (copy_to_user(arg, &vr, sizeof(vr)))
1882 read_unlock(&mrt_lock);
1883 return -EADDRNOTAVAIL;
1884 case SIOCGETSGCNT_IN6:
1885 if (copy_from_user(&sr, arg, sizeof(sr)))
1888 read_lock(&mrt_lock);
1889 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1891 sr.pktcnt = c->mfc_un.res.pkt;
1892 sr.bytecnt = c->mfc_un.res.bytes;
1893 sr.wrong_if = c->mfc_un.res.wrong_if;
1894 read_unlock(&mrt_lock);
1896 if (copy_to_user(arg, &sr, sizeof(sr)))
1900 read_unlock(&mrt_lock);
1901 return -EADDRNOTAVAIL;
1903 return -ENOIOCTLCMD;
1907 #ifdef CONFIG_COMPAT
1908 struct compat_sioc_sg_req6 {
1909 struct sockaddr_in6 src;
1910 struct sockaddr_in6 grp;
1911 compat_ulong_t pktcnt;
1912 compat_ulong_t bytecnt;
1913 compat_ulong_t wrong_if;
1916 struct compat_sioc_mif_req6 {
1918 compat_ulong_t icount;
1919 compat_ulong_t ocount;
1920 compat_ulong_t ibytes;
1921 compat_ulong_t obytes;
1924 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1926 struct compat_sioc_sg_req6 sr;
1927 struct compat_sioc_mif_req6 vr;
1928 struct mif_device *vif;
1929 struct mfc6_cache *c;
1930 struct net *net = sock_net(sk);
1931 struct mr6_table *mrt;
1933 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1938 case SIOCGETMIFCNT_IN6:
1939 if (copy_from_user(&vr, arg, sizeof(vr)))
1941 if (vr.mifi >= mrt->maxvif)
1943 read_lock(&mrt_lock);
1944 vif = &mrt->vif6_table[vr.mifi];
1945 if (MIF_EXISTS(mrt, vr.mifi)) {
1946 vr.icount = vif->pkt_in;
1947 vr.ocount = vif->pkt_out;
1948 vr.ibytes = vif->bytes_in;
1949 vr.obytes = vif->bytes_out;
1950 read_unlock(&mrt_lock);
1952 if (copy_to_user(arg, &vr, sizeof(vr)))
1956 read_unlock(&mrt_lock);
1957 return -EADDRNOTAVAIL;
1958 case SIOCGETSGCNT_IN6:
1959 if (copy_from_user(&sr, arg, sizeof(sr)))
1962 read_lock(&mrt_lock);
1963 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1965 sr.pktcnt = c->mfc_un.res.pkt;
1966 sr.bytecnt = c->mfc_un.res.bytes;
1967 sr.wrong_if = c->mfc_un.res.wrong_if;
1968 read_unlock(&mrt_lock);
1970 if (copy_to_user(arg, &sr, sizeof(sr)))
1974 read_unlock(&mrt_lock);
1975 return -EADDRNOTAVAIL;
1977 return -ENOIOCTLCMD;
1982 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1984 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1985 IPSTATS_MIB_OUTFORWDATAGRAMS);
1986 IP6_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1987 IPSTATS_MIB_OUTOCTETS, skb->len);
1988 return dst_output(skb);
1992 * Processing handlers for ip6mr_forward
1995 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1996 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1998 struct ipv6hdr *ipv6h;
1999 struct mif_device *vif = &mrt->vif6_table[vifi];
2000 struct net_device *dev;
2001 struct dst_entry *dst;
2004 if (vif->dev == NULL)
2007 #ifdef CONFIG_IPV6_PIMSM_V2
2008 if (vif->flags & MIFF_REGISTER) {
2010 vif->bytes_out += skb->len;
2011 vif->dev->stats.tx_bytes += skb->len;
2012 vif->dev->stats.tx_packets++;
2013 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2018 ipv6h = ipv6_hdr(skb);
2020 fl6 = (struct flowi6) {
2021 .flowi6_oif = vif->link,
2022 .daddr = ipv6h->daddr,
2025 dst = ip6_route_output(net, NULL, &fl6);
2032 skb_dst_set(skb, dst);
2035 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2036 * not only before forwarding, but after forwarding on all output
2037 * interfaces. It is clear, if mrouter runs a multicasting
2038 * program, it should receive packets not depending to what interface
2039 * program is joined.
2040 * If we will not make it, the program will have to join on all
2041 * interfaces. On the other hand, multihoming host (or router, but
2042 * not mrouter) cannot join to more than one interface - it will
2043 * result in receiving multiple packets.
2048 vif->bytes_out += skb->len;
2050 /* We are about to write */
2051 /* XXX: extension headers? */
2052 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2055 ipv6h = ipv6_hdr(skb);
2058 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2060 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
2061 ip6mr_forward2_finish);
2068 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2072 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2073 if (mrt->vif6_table[ct].dev == dev)
2079 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2080 struct sk_buff *skb, struct mfc6_cache *cache)
2084 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2086 vif = cache->mf6c_parent;
2087 cache->mfc_un.res.pkt++;
2088 cache->mfc_un.res.bytes += skb->len;
2090 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2091 struct mfc6_cache *cache_proxy;
2093 /* For an (*,G) entry, we only check that the incomming
2094 * interface is part of the static tree.
2096 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2098 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2103 * Wrong interface: drop packet and (maybe) send PIM assert.
2105 if (mrt->vif6_table[vif].dev != skb->dev) {
2106 cache->mfc_un.res.wrong_if++;
2108 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2109 /* pimsm uses asserts, when switching from RPT to SPT,
2110 so that we cannot check that packet arrived on an oif.
2111 It is bad, but otherwise we would need to move pretty
2112 large chunk of pimd to kernel. Ough... --ANK
2114 (mrt->mroute_do_pim ||
2115 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2117 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2118 cache->mfc_un.res.last_assert = jiffies;
2119 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2125 mrt->vif6_table[vif].pkt_in++;
2126 mrt->vif6_table[vif].bytes_in += skb->len;
2131 if (ipv6_addr_any(&cache->mf6c_origin) &&
2132 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2133 if (true_vifi >= 0 &&
2134 true_vifi != cache->mf6c_parent &&
2135 ipv6_hdr(skb)->hop_limit >
2136 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2137 /* It's an (*,*) entry and the packet is not coming from
2138 * the upstream: forward the packet to the upstream
2141 psend = cache->mf6c_parent;
2146 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2147 /* For (*,G) entry, don't forward to the incoming interface */
2148 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2149 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2151 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2153 ip6mr_forward2(net, mrt, skb2, cache, psend);
2160 ip6mr_forward2(net, mrt, skb, cache, psend);
2170 * Multicast packets for forwarding arrive here
2173 int ip6_mr_input(struct sk_buff *skb)
2175 struct mfc6_cache *cache;
2176 struct net *net = dev_net(skb->dev);
2177 struct mr6_table *mrt;
2178 struct flowi6 fl6 = {
2179 .flowi6_iif = skb->dev->ifindex,
2180 .flowi6_mark = skb->mark,
2184 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2190 read_lock(&mrt_lock);
2191 cache = ip6mr_cache_find(mrt,
2192 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2193 if (cache == NULL) {
2194 int vif = ip6mr_find_vif(mrt, skb->dev);
2197 cache = ip6mr_cache_find_any(mrt,
2198 &ipv6_hdr(skb)->daddr,
2203 * No usable cache entry
2205 if (cache == NULL) {
2208 vif = ip6mr_find_vif(mrt, skb->dev);
2210 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2211 read_unlock(&mrt_lock);
2215 read_unlock(&mrt_lock);
2220 ip6_mr_forward(net, mrt, skb, cache);
2222 read_unlock(&mrt_lock);
2228 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2229 struct mfc6_cache *c, struct rtmsg *rtm)
2232 struct rtnexthop *nhp;
2233 struct nlattr *mp_attr;
2234 struct rta_mfc_stats mfcs;
2236 /* If cache is unresolved, don't try to parse IIF and OIF */
2237 if (c->mf6c_parent >= MAXMIFS)
2240 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2241 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2243 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2244 if (mp_attr == NULL)
2247 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2248 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2249 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2251 nla_nest_cancel(skb, mp_attr);
2255 nhp->rtnh_flags = 0;
2256 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2257 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2258 nhp->rtnh_len = sizeof(*nhp);
2262 nla_nest_end(skb, mp_attr);
2264 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2265 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2266 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2267 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2270 rtm->rtm_type = RTN_MULTICAST;
2274 int ip6mr_get_route(struct net *net,
2275 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2278 struct mr6_table *mrt;
2279 struct mfc6_cache *cache;
2280 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2282 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2286 read_lock(&mrt_lock);
2287 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2288 if (!cache && skb->dev) {
2289 int vif = ip6mr_find_vif(mrt, skb->dev);
2292 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2297 struct sk_buff *skb2;
2298 struct ipv6hdr *iph;
2299 struct net_device *dev;
2303 read_unlock(&mrt_lock);
2308 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2309 read_unlock(&mrt_lock);
2313 /* really correct? */
2314 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2316 read_unlock(&mrt_lock);
2320 skb_reset_transport_header(skb2);
2322 skb_put(skb2, sizeof(struct ipv6hdr));
2323 skb_reset_network_header(skb2);
2325 iph = ipv6_hdr(skb2);
2328 iph->flow_lbl[0] = 0;
2329 iph->flow_lbl[1] = 0;
2330 iph->flow_lbl[2] = 0;
2331 iph->payload_len = 0;
2332 iph->nexthdr = IPPROTO_NONE;
2334 iph->saddr = rt->rt6i_src.addr;
2335 iph->daddr = rt->rt6i_dst.addr;
2337 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2338 read_unlock(&mrt_lock);
2343 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2344 cache->mfc_flags |= MFC_NOTIFY;
2346 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2347 read_unlock(&mrt_lock);
2351 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2352 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2355 struct nlmsghdr *nlh;
2359 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2363 rtm = nlmsg_data(nlh);
2364 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2365 rtm->rtm_dst_len = 128;
2366 rtm->rtm_src_len = 128;
2368 rtm->rtm_table = mrt->id;
2369 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2370 goto nla_put_failure;
2371 rtm->rtm_type = RTN_MULTICAST;
2372 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2373 if (c->mfc_flags & MFC_STATIC)
2374 rtm->rtm_protocol = RTPROT_STATIC;
2376 rtm->rtm_protocol = RTPROT_MROUTED;
2379 if (nla_put(skb, RTA_SRC, 16, &c->mf6c_origin) ||
2380 nla_put(skb, RTA_DST, 16, &c->mf6c_mcastgrp))
2381 goto nla_put_failure;
2382 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2383 /* do not break the dump if cache is unresolved */
2384 if (err < 0 && err != -ENOENT)
2385 goto nla_put_failure;
2387 return nlmsg_end(skb, nlh);
2390 nlmsg_cancel(skb, nlh);
2394 static int mr6_msgsize(bool unresolved, int maxvif)
2397 NLMSG_ALIGN(sizeof(struct rtmsg))
2398 + nla_total_size(4) /* RTA_TABLE */
2399 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2400 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2405 + nla_total_size(4) /* RTA_IIF */
2406 + nla_total_size(0) /* RTA_MULTIPATH */
2407 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2409 + nla_total_size(sizeof(struct rta_mfc_stats))
2415 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2418 struct net *net = read_pnet(&mrt->net);
2419 struct sk_buff *skb;
2422 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2427 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2431 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2437 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2440 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2442 struct net *net = sock_net(skb->sk);
2443 struct mr6_table *mrt;
2444 struct mfc6_cache *mfc;
2445 unsigned int t = 0, s_t;
2446 unsigned int h = 0, s_h;
2447 unsigned int e = 0, s_e;
2453 read_lock(&mrt_lock);
2454 ip6mr_for_each_table(mrt, net) {
2459 for (h = s_h; h < MFC6_LINES; h++) {
2460 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2463 if (ip6mr_fill_mroute(mrt, skb,
2464 NETLINK_CB(cb->skb).portid,
2474 spin_lock_bh(&mfc_unres_lock);
2475 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2478 if (ip6mr_fill_mroute(mrt, skb,
2479 NETLINK_CB(cb->skb).portid,
2483 spin_unlock_bh(&mfc_unres_lock);
2489 spin_unlock_bh(&mfc_unres_lock);
2496 read_unlock(&mrt_lock);