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)
144 struct ip6mr_result res;
145 struct fib_lookup_arg arg = { .result = &res, };
148 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
149 flowi6_to_flowi(flp6), 0, &arg);
156 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
157 int flags, struct fib_lookup_arg *arg)
159 struct ip6mr_result *res = arg->result;
160 struct mr6_table *mrt;
162 switch (rule->action) {
165 case FR_ACT_UNREACHABLE:
167 case FR_ACT_PROHIBIT:
169 case FR_ACT_BLACKHOLE:
174 mrt = ip6mr_get_table(rule->fr_net, rule->table);
181 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
186 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
190 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
191 struct fib_rule_hdr *frh, struct nlattr **tb)
196 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
202 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
203 struct fib_rule_hdr *frh)
211 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
212 .family = RTNL_FAMILY_IP6MR,
213 .rule_size = sizeof(struct ip6mr_rule),
214 .addr_size = sizeof(struct in6_addr),
215 .action = ip6mr_rule_action,
216 .match = ip6mr_rule_match,
217 .configure = ip6mr_rule_configure,
218 .compare = ip6mr_rule_compare,
219 .default_pref = fib_default_rule_pref,
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;
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);
268 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(&mrt->ipmr_expire_timer);
337 mroute_clean_tables(mrt);
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)
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;
675 skb->pkt_type = PACKET_HOST;
677 skb_tunnel_rx(skb, reg_dev);
688 static const struct inet6_protocol pim6_protocol = {
692 /* Service routines creating virtual interfaces: PIMREG */
694 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
695 struct net_device *dev)
697 struct net *net = dev_net(dev);
698 struct mr6_table *mrt;
699 struct flowi6 fl6 = {
700 .flowi6_oif = dev->ifindex,
701 .flowi6_iif = skb->skb_iif,
702 .flowi6_mark = skb->mark,
706 err = ip6mr_fib_lookup(net, &fl6, &mrt);
712 read_lock(&mrt_lock);
713 dev->stats.tx_bytes += skb->len;
714 dev->stats.tx_packets++;
715 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
716 read_unlock(&mrt_lock);
721 static const struct net_device_ops reg_vif_netdev_ops = {
722 .ndo_start_xmit = reg_vif_xmit,
725 static void reg_vif_setup(struct net_device *dev)
727 dev->type = ARPHRD_PIMREG;
728 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
729 dev->flags = IFF_NOARP;
730 dev->netdev_ops = ®_vif_netdev_ops;
731 dev->destructor = free_netdev;
732 dev->features |= NETIF_F_NETNS_LOCAL;
735 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
737 struct net_device *dev;
740 if (mrt->id == RT6_TABLE_DFLT)
741 sprintf(name, "pim6reg");
743 sprintf(name, "pim6reg%u", mrt->id);
745 dev = alloc_netdev(0, name, reg_vif_setup);
749 dev_net_set(dev, net);
751 if (register_netdevice(dev)) {
764 /* allow the register to be completed before unregistering. */
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->iflink;
997 /* And finish update writing critical data */
998 write_lock_bh(&mrt_lock);
1000 #ifdef CONFIG_IPV6_PIMSM_V2
1001 if (v->flags & MIFF_REGISTER)
1002 mrt->mroute_reg_vif_num = vifi;
1004 if (vifi + 1 > mrt->maxvif)
1005 mrt->maxvif = vifi + 1;
1006 write_unlock_bh(&mrt_lock);
1010 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1011 const struct in6_addr *origin,
1012 const struct in6_addr *mcastgrp)
1014 int line = MFC6_HASH(mcastgrp, origin);
1015 struct mfc6_cache *c;
1017 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1018 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1019 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1025 /* Look for a (*,*,oif) entry */
1026 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1029 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1030 struct mfc6_cache *c;
1032 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1033 if (ipv6_addr_any(&c->mf6c_origin) &&
1034 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1035 (c->mfc_un.res.ttls[mifi] < 255))
1041 /* Look for a (*,G) entry */
1042 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1043 struct in6_addr *mcastgrp,
1046 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1047 struct mfc6_cache *c, *proxy;
1049 if (ipv6_addr_any(mcastgrp))
1052 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1053 if (ipv6_addr_any(&c->mf6c_origin) &&
1054 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1055 if (c->mfc_un.res.ttls[mifi] < 255)
1058 /* It's ok if the mifi is part of the static tree */
1059 proxy = ip6mr_cache_find_any_parent(mrt,
1061 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1066 return ip6mr_cache_find_any_parent(mrt, mifi);
1070 * Allocate a multicast cache entry
1072 static struct mfc6_cache *ip6mr_cache_alloc(void)
1074 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1077 c->mfc_un.res.minvif = MAXMIFS;
1081 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1083 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1086 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1087 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1092 * A cache entry has gone into a resolved state from queued
1095 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1096 struct mfc6_cache *uc, struct mfc6_cache *c)
1098 struct sk_buff *skb;
1101 * Play the pending entries through our router
1104 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1105 if (ipv6_hdr(skb)->version == 0) {
1106 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1108 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1109 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1111 nlh->nlmsg_type = NLMSG_ERROR;
1112 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1113 skb_trim(skb, nlh->nlmsg_len);
1114 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1116 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1118 ip6_mr_forward(net, mrt, skb, c);
1123 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1124 * expects the following bizarre scheme.
1126 * Called under mrt_lock.
1129 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1130 mifi_t mifi, int assert)
1132 struct sk_buff *skb;
1133 struct mrt6msg *msg;
1136 #ifdef CONFIG_IPV6_PIMSM_V2
1137 if (assert == MRT6MSG_WHOLEPKT)
1138 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1142 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1147 /* I suppose that internal messages
1148 * do not require checksums */
1150 skb->ip_summed = CHECKSUM_UNNECESSARY;
1152 #ifdef CONFIG_IPV6_PIMSM_V2
1153 if (assert == MRT6MSG_WHOLEPKT) {
1154 /* Ugly, but we have no choice with this interface.
1155 Duplicate old header, fix length etc.
1156 And all this only to mangle msg->im6_msgtype and
1157 to set msg->im6_mbz to "mbz" :-)
1159 skb_push(skb, -skb_network_offset(pkt));
1161 skb_push(skb, sizeof(*msg));
1162 skb_reset_transport_header(skb);
1163 msg = (struct mrt6msg *)skb_transport_header(skb);
1165 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1166 msg->im6_mif = mrt->mroute_reg_vif_num;
1168 msg->im6_src = ipv6_hdr(pkt)->saddr;
1169 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1171 skb->ip_summed = CHECKSUM_UNNECESSARY;
1176 * Copy the IP header
1179 skb_put(skb, sizeof(struct ipv6hdr));
1180 skb_reset_network_header(skb);
1181 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1186 skb_put(skb, sizeof(*msg));
1187 skb_reset_transport_header(skb);
1188 msg = (struct mrt6msg *)skb_transport_header(skb);
1191 msg->im6_msgtype = assert;
1192 msg->im6_mif = mifi;
1194 msg->im6_src = ipv6_hdr(pkt)->saddr;
1195 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1197 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1198 skb->ip_summed = CHECKSUM_UNNECESSARY;
1201 if (mrt->mroute6_sk == NULL) {
1207 * Deliver to user space multicast routing algorithms
1209 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1211 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1219 * Queue a packet for resolution. It gets locked cache entry!
1223 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1227 struct mfc6_cache *c;
1229 spin_lock_bh(&mfc_unres_lock);
1230 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1231 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1232 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1240 * Create a new entry if allowable
1243 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1244 (c = ip6mr_cache_alloc_unres()) == NULL) {
1245 spin_unlock_bh(&mfc_unres_lock);
1252 * Fill in the new cache entry
1254 c->mf6c_parent = -1;
1255 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1256 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1259 * Reflect first query at pim6sd
1261 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1263 /* If the report failed throw the cache entry
1266 spin_unlock_bh(&mfc_unres_lock);
1268 ip6mr_cache_free(c);
1273 atomic_inc(&mrt->cache_resolve_queue_len);
1274 list_add(&c->list, &mrt->mfc6_unres_queue);
1275 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1277 ipmr_do_expire_process(mrt);
1281 * See if we can append the packet
1283 if (c->mfc_un.unres.unresolved.qlen > 3) {
1287 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1291 spin_unlock_bh(&mfc_unres_lock);
1296 * MFC6 cache manipulation by user space
1299 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1303 struct mfc6_cache *c, *next;
1305 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1307 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1308 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1309 ipv6_addr_equal(&c->mf6c_mcastgrp,
1310 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1311 (parent == -1 || parent == c->mf6c_parent)) {
1312 write_lock_bh(&mrt_lock);
1314 write_unlock_bh(&mrt_lock);
1316 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1317 ip6mr_cache_free(c);
1324 static int ip6mr_device_event(struct notifier_block *this,
1325 unsigned long event, void *ptr)
1327 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1328 struct net *net = dev_net(dev);
1329 struct mr6_table *mrt;
1330 struct mif_device *v;
1334 if (event != NETDEV_UNREGISTER)
1337 ip6mr_for_each_table(mrt, net) {
1338 v = &mrt->vif6_table[0];
1339 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1341 mif6_delete(mrt, ct, &list);
1344 unregister_netdevice_many(&list);
1349 static struct notifier_block ip6_mr_notifier = {
1350 .notifier_call = ip6mr_device_event
1354 * Setup for IP multicast routing
1357 static int __net_init ip6mr_net_init(struct net *net)
1361 err = ip6mr_rules_init(net);
1365 #ifdef CONFIG_PROC_FS
1367 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1369 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1370 goto proc_cache_fail;
1375 #ifdef CONFIG_PROC_FS
1377 remove_proc_entry("ip6_mr_vif", net->proc_net);
1379 ip6mr_rules_exit(net);
1385 static void __net_exit ip6mr_net_exit(struct net *net)
1387 #ifdef CONFIG_PROC_FS
1388 remove_proc_entry("ip6_mr_cache", net->proc_net);
1389 remove_proc_entry("ip6_mr_vif", net->proc_net);
1391 ip6mr_rules_exit(net);
1394 static struct pernet_operations ip6mr_net_ops = {
1395 .init = ip6mr_net_init,
1396 .exit = ip6mr_net_exit,
1399 int __init ip6_mr_init(void)
1403 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1404 sizeof(struct mfc6_cache),
1405 0, SLAB_HWCACHE_ALIGN,
1410 err = register_pernet_subsys(&ip6mr_net_ops);
1412 goto reg_pernet_fail;
1414 err = register_netdevice_notifier(&ip6_mr_notifier);
1416 goto reg_notif_fail;
1417 #ifdef CONFIG_IPV6_PIMSM_V2
1418 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1419 pr_err("%s: can't add PIM protocol\n", __func__);
1421 goto add_proto_fail;
1424 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1425 ip6mr_rtm_dumproute, NULL);
1427 #ifdef CONFIG_IPV6_PIMSM_V2
1429 unregister_netdevice_notifier(&ip6_mr_notifier);
1432 unregister_pernet_subsys(&ip6mr_net_ops);
1434 kmem_cache_destroy(mrt_cachep);
1438 void ip6_mr_cleanup(void)
1440 unregister_netdevice_notifier(&ip6_mr_notifier);
1441 unregister_pernet_subsys(&ip6mr_net_ops);
1442 kmem_cache_destroy(mrt_cachep);
1445 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1446 struct mf6cctl *mfc, int mrtsock, int parent)
1450 struct mfc6_cache *uc, *c;
1451 unsigned char ttls[MAXMIFS];
1454 if (mfc->mf6cc_parent >= MAXMIFS)
1457 memset(ttls, 255, MAXMIFS);
1458 for (i = 0; i < MAXMIFS; i++) {
1459 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1464 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1466 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1467 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1468 ipv6_addr_equal(&c->mf6c_mcastgrp,
1469 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1470 (parent == -1 || parent == mfc->mf6cc_parent)) {
1477 write_lock_bh(&mrt_lock);
1478 c->mf6c_parent = mfc->mf6cc_parent;
1479 ip6mr_update_thresholds(mrt, c, ttls);
1481 c->mfc_flags |= MFC_STATIC;
1482 write_unlock_bh(&mrt_lock);
1483 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1487 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1488 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1491 c = ip6mr_cache_alloc();
1495 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1496 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1497 c->mf6c_parent = mfc->mf6cc_parent;
1498 ip6mr_update_thresholds(mrt, c, ttls);
1500 c->mfc_flags |= MFC_STATIC;
1502 write_lock_bh(&mrt_lock);
1503 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1504 write_unlock_bh(&mrt_lock);
1507 * Check to see if we resolved a queued list. If so we
1508 * need to send on the frames and tidy up.
1511 spin_lock_bh(&mfc_unres_lock);
1512 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1513 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1514 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1515 list_del(&uc->list);
1516 atomic_dec(&mrt->cache_resolve_queue_len);
1521 if (list_empty(&mrt->mfc6_unres_queue))
1522 del_timer(&mrt->ipmr_expire_timer);
1523 spin_unlock_bh(&mfc_unres_lock);
1526 ip6mr_cache_resolve(net, mrt, uc, c);
1527 ip6mr_cache_free(uc);
1529 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1534 * Close the multicast socket, and clear the vif tables etc
1537 static void mroute_clean_tables(struct mr6_table *mrt)
1541 struct mfc6_cache *c, *next;
1544 * Shut down all active vif entries
1546 for (i = 0; i < mrt->maxvif; i++) {
1547 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1548 mif6_delete(mrt, i, &list);
1550 unregister_netdevice_many(&list);
1555 for (i = 0; i < MFC6_LINES; i++) {
1556 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1557 if (c->mfc_flags & MFC_STATIC)
1559 write_lock_bh(&mrt_lock);
1561 write_unlock_bh(&mrt_lock);
1563 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1564 ip6mr_cache_free(c);
1568 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1569 spin_lock_bh(&mfc_unres_lock);
1570 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1572 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1573 ip6mr_destroy_unres(mrt, c);
1575 spin_unlock_bh(&mfc_unres_lock);
1579 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1582 struct net *net = sock_net(sk);
1585 write_lock_bh(&mrt_lock);
1586 if (likely(mrt->mroute6_sk == NULL)) {
1587 mrt->mroute6_sk = sk;
1588 net->ipv6.devconf_all->mc_forwarding++;
1589 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1590 NETCONFA_IFINDEX_ALL,
1591 net->ipv6.devconf_all);
1595 write_unlock_bh(&mrt_lock);
1602 int ip6mr_sk_done(struct sock *sk)
1605 struct net *net = sock_net(sk);
1606 struct mr6_table *mrt;
1609 ip6mr_for_each_table(mrt, net) {
1610 if (sk == mrt->mroute6_sk) {
1611 write_lock_bh(&mrt_lock);
1612 mrt->mroute6_sk = NULL;
1613 net->ipv6.devconf_all->mc_forwarding--;
1614 inet6_netconf_notify_devconf(net,
1615 NETCONFA_MC_FORWARDING,
1616 NETCONFA_IFINDEX_ALL,
1617 net->ipv6.devconf_all);
1618 write_unlock_bh(&mrt_lock);
1620 mroute_clean_tables(mrt);
1630 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1632 struct mr6_table *mrt;
1633 struct flowi6 fl6 = {
1634 .flowi6_iif = skb->skb_iif,
1635 .flowi6_oif = skb->dev->ifindex,
1636 .flowi6_mark = skb->mark,
1639 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1642 return mrt->mroute6_sk;
1646 * Socket options and virtual interface manipulation. The whole
1647 * virtual interface system is a complete heap, but unfortunately
1648 * that's how BSD mrouted happens to think. Maybe one day with a proper
1649 * MOSPF/PIM router set up we can clean this up.
1652 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1654 int ret, parent = 0;
1658 struct net *net = sock_net(sk);
1659 struct mr6_table *mrt;
1661 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1665 if (optname != MRT6_INIT) {
1666 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1672 if (sk->sk_type != SOCK_RAW ||
1673 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1675 if (optlen < sizeof(int))
1678 return ip6mr_sk_init(mrt, sk);
1681 return ip6mr_sk_done(sk);
1684 if (optlen < sizeof(vif))
1686 if (copy_from_user(&vif, optval, sizeof(vif)))
1688 if (vif.mif6c_mifi >= MAXMIFS)
1691 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1696 if (optlen < sizeof(mifi_t))
1698 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1701 ret = mif6_delete(mrt, mifi, NULL);
1706 * Manipulate the forwarding caches. These live
1707 * in a sort of kernel/user symbiosis.
1712 case MRT6_ADD_MFC_PROXY:
1713 case MRT6_DEL_MFC_PROXY:
1714 if (optlen < sizeof(mfc))
1716 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1719 parent = mfc.mf6cc_parent;
1721 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1722 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1724 ret = ip6mr_mfc_add(net, mrt, &mfc,
1725 sk == mrt->mroute6_sk, parent);
1730 * Control PIM assert (to activate pim will activate assert)
1736 if (optlen != sizeof(v))
1738 if (get_user(v, (int __user *)optval))
1740 mrt->mroute_do_assert = v;
1744 #ifdef CONFIG_IPV6_PIMSM_V2
1749 if (optlen != sizeof(v))
1751 if (get_user(v, (int __user *)optval))
1756 if (v != mrt->mroute_do_pim) {
1757 mrt->mroute_do_pim = v;
1758 mrt->mroute_do_assert = v;
1765 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1770 if (optlen != sizeof(u32))
1772 if (get_user(v, (u32 __user *)optval))
1774 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1775 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1777 if (sk == mrt->mroute6_sk)
1782 if (!ip6mr_new_table(net, v))
1784 raw6_sk(sk)->ip6mr_table = v;
1790 * Spurious command, or MRT6_VERSION which you cannot
1794 return -ENOPROTOOPT;
1799 * Getsock opt support for the multicast routing system.
1802 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1807 struct net *net = sock_net(sk);
1808 struct mr6_table *mrt;
1810 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1818 #ifdef CONFIG_IPV6_PIMSM_V2
1820 val = mrt->mroute_do_pim;
1824 val = mrt->mroute_do_assert;
1827 return -ENOPROTOOPT;
1830 if (get_user(olr, optlen))
1833 olr = min_t(int, olr, sizeof(int));
1837 if (put_user(olr, optlen))
1839 if (copy_to_user(optval, &val, olr))
1845 * The IP multicast ioctl support routines.
1848 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1850 struct sioc_sg_req6 sr;
1851 struct sioc_mif_req6 vr;
1852 struct mif_device *vif;
1853 struct mfc6_cache *c;
1854 struct net *net = sock_net(sk);
1855 struct mr6_table *mrt;
1857 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1862 case SIOCGETMIFCNT_IN6:
1863 if (copy_from_user(&vr, arg, sizeof(vr)))
1865 if (vr.mifi >= mrt->maxvif)
1867 read_lock(&mrt_lock);
1868 vif = &mrt->vif6_table[vr.mifi];
1869 if (MIF_EXISTS(mrt, vr.mifi)) {
1870 vr.icount = vif->pkt_in;
1871 vr.ocount = vif->pkt_out;
1872 vr.ibytes = vif->bytes_in;
1873 vr.obytes = vif->bytes_out;
1874 read_unlock(&mrt_lock);
1876 if (copy_to_user(arg, &vr, sizeof(vr)))
1880 read_unlock(&mrt_lock);
1881 return -EADDRNOTAVAIL;
1882 case SIOCGETSGCNT_IN6:
1883 if (copy_from_user(&sr, arg, sizeof(sr)))
1886 read_lock(&mrt_lock);
1887 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1889 sr.pktcnt = c->mfc_un.res.pkt;
1890 sr.bytecnt = c->mfc_un.res.bytes;
1891 sr.wrong_if = c->mfc_un.res.wrong_if;
1892 read_unlock(&mrt_lock);
1894 if (copy_to_user(arg, &sr, sizeof(sr)))
1898 read_unlock(&mrt_lock);
1899 return -EADDRNOTAVAIL;
1901 return -ENOIOCTLCMD;
1905 #ifdef CONFIG_COMPAT
1906 struct compat_sioc_sg_req6 {
1907 struct sockaddr_in6 src;
1908 struct sockaddr_in6 grp;
1909 compat_ulong_t pktcnt;
1910 compat_ulong_t bytecnt;
1911 compat_ulong_t wrong_if;
1914 struct compat_sioc_mif_req6 {
1916 compat_ulong_t icount;
1917 compat_ulong_t ocount;
1918 compat_ulong_t ibytes;
1919 compat_ulong_t obytes;
1922 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1924 struct compat_sioc_sg_req6 sr;
1925 struct compat_sioc_mif_req6 vr;
1926 struct mif_device *vif;
1927 struct mfc6_cache *c;
1928 struct net *net = sock_net(sk);
1929 struct mr6_table *mrt;
1931 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1936 case SIOCGETMIFCNT_IN6:
1937 if (copy_from_user(&vr, arg, sizeof(vr)))
1939 if (vr.mifi >= mrt->maxvif)
1941 read_lock(&mrt_lock);
1942 vif = &mrt->vif6_table[vr.mifi];
1943 if (MIF_EXISTS(mrt, vr.mifi)) {
1944 vr.icount = vif->pkt_in;
1945 vr.ocount = vif->pkt_out;
1946 vr.ibytes = vif->bytes_in;
1947 vr.obytes = vif->bytes_out;
1948 read_unlock(&mrt_lock);
1950 if (copy_to_user(arg, &vr, sizeof(vr)))
1954 read_unlock(&mrt_lock);
1955 return -EADDRNOTAVAIL;
1956 case SIOCGETSGCNT_IN6:
1957 if (copy_from_user(&sr, arg, sizeof(sr)))
1960 read_lock(&mrt_lock);
1961 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1963 sr.pktcnt = c->mfc_un.res.pkt;
1964 sr.bytecnt = c->mfc_un.res.bytes;
1965 sr.wrong_if = c->mfc_un.res.wrong_if;
1966 read_unlock(&mrt_lock);
1968 if (copy_to_user(arg, &sr, sizeof(sr)))
1972 read_unlock(&mrt_lock);
1973 return -EADDRNOTAVAIL;
1975 return -ENOIOCTLCMD;
1980 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1982 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1983 IPSTATS_MIB_OUTFORWDATAGRAMS);
1984 IP6_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1985 IPSTATS_MIB_OUTOCTETS, skb->len);
1986 return dst_output(skb);
1990 * Processing handlers for ip6mr_forward
1993 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1994 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1996 struct ipv6hdr *ipv6h;
1997 struct mif_device *vif = &mrt->vif6_table[vifi];
1998 struct net_device *dev;
1999 struct dst_entry *dst;
2002 if (vif->dev == NULL)
2005 #ifdef CONFIG_IPV6_PIMSM_V2
2006 if (vif->flags & MIFF_REGISTER) {
2008 vif->bytes_out += skb->len;
2009 vif->dev->stats.tx_bytes += skb->len;
2010 vif->dev->stats.tx_packets++;
2011 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2016 ipv6h = ipv6_hdr(skb);
2018 fl6 = (struct flowi6) {
2019 .flowi6_oif = vif->link,
2020 .daddr = ipv6h->daddr,
2023 dst = ip6_route_output(net, NULL, &fl6);
2030 skb_dst_set(skb, dst);
2033 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2034 * not only before forwarding, but after forwarding on all output
2035 * interfaces. It is clear, if mrouter runs a multicasting
2036 * program, it should receive packets not depending to what interface
2037 * program is joined.
2038 * If we will not make it, the program will have to join on all
2039 * interfaces. On the other hand, multihoming host (or router, but
2040 * not mrouter) cannot join to more than one interface - it will
2041 * result in receiving multiple packets.
2046 vif->bytes_out += skb->len;
2048 /* We are about to write */
2049 /* XXX: extension headers? */
2050 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2053 ipv6h = ipv6_hdr(skb);
2056 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2058 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
2059 ip6mr_forward2_finish);
2066 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2070 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2071 if (mrt->vif6_table[ct].dev == dev)
2077 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2078 struct sk_buff *skb, struct mfc6_cache *cache)
2082 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2084 vif = cache->mf6c_parent;
2085 cache->mfc_un.res.pkt++;
2086 cache->mfc_un.res.bytes += skb->len;
2088 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2089 struct mfc6_cache *cache_proxy;
2091 /* For an (*,G) entry, we only check that the incomming
2092 * interface is part of the static tree.
2094 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2096 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2101 * Wrong interface: drop packet and (maybe) send PIM assert.
2103 if (mrt->vif6_table[vif].dev != skb->dev) {
2104 cache->mfc_un.res.wrong_if++;
2106 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2107 /* pimsm uses asserts, when switching from RPT to SPT,
2108 so that we cannot check that packet arrived on an oif.
2109 It is bad, but otherwise we would need to move pretty
2110 large chunk of pimd to kernel. Ough... --ANK
2112 (mrt->mroute_do_pim ||
2113 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2115 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2116 cache->mfc_un.res.last_assert = jiffies;
2117 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2123 mrt->vif6_table[vif].pkt_in++;
2124 mrt->vif6_table[vif].bytes_in += skb->len;
2129 if (ipv6_addr_any(&cache->mf6c_origin) &&
2130 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2131 if (true_vifi >= 0 &&
2132 true_vifi != cache->mf6c_parent &&
2133 ipv6_hdr(skb)->hop_limit >
2134 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2135 /* It's an (*,*) entry and the packet is not coming from
2136 * the upstream: forward the packet to the upstream
2139 psend = cache->mf6c_parent;
2144 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2145 /* For (*,G) entry, don't forward to the incoming interface */
2146 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2147 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2149 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2151 ip6mr_forward2(net, mrt, skb2, cache, psend);
2158 ip6mr_forward2(net, mrt, skb, cache, psend);
2168 * Multicast packets for forwarding arrive here
2171 int ip6_mr_input(struct sk_buff *skb)
2173 struct mfc6_cache *cache;
2174 struct net *net = dev_net(skb->dev);
2175 struct mr6_table *mrt;
2176 struct flowi6 fl6 = {
2177 .flowi6_iif = skb->dev->ifindex,
2178 .flowi6_mark = skb->mark,
2182 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2188 read_lock(&mrt_lock);
2189 cache = ip6mr_cache_find(mrt,
2190 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2191 if (cache == NULL) {
2192 int vif = ip6mr_find_vif(mrt, skb->dev);
2195 cache = ip6mr_cache_find_any(mrt,
2196 &ipv6_hdr(skb)->daddr,
2201 * No usable cache entry
2203 if (cache == NULL) {
2206 vif = ip6mr_find_vif(mrt, skb->dev);
2208 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2209 read_unlock(&mrt_lock);
2213 read_unlock(&mrt_lock);
2218 ip6_mr_forward(net, mrt, skb, cache);
2220 read_unlock(&mrt_lock);
2226 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2227 struct mfc6_cache *c, struct rtmsg *rtm)
2230 struct rtnexthop *nhp;
2231 struct nlattr *mp_attr;
2232 struct rta_mfc_stats mfcs;
2234 /* If cache is unresolved, don't try to parse IIF and OIF */
2235 if (c->mf6c_parent >= MAXMIFS)
2238 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2239 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2241 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2242 if (mp_attr == NULL)
2245 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2246 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2247 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2249 nla_nest_cancel(skb, mp_attr);
2253 nhp->rtnh_flags = 0;
2254 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2255 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2256 nhp->rtnh_len = sizeof(*nhp);
2260 nla_nest_end(skb, mp_attr);
2262 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2263 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2264 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2265 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2268 rtm->rtm_type = RTN_MULTICAST;
2272 int ip6mr_get_route(struct net *net,
2273 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2276 struct mr6_table *mrt;
2277 struct mfc6_cache *cache;
2278 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2280 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2284 read_lock(&mrt_lock);
2285 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2286 if (!cache && skb->dev) {
2287 int vif = ip6mr_find_vif(mrt, skb->dev);
2290 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2295 struct sk_buff *skb2;
2296 struct ipv6hdr *iph;
2297 struct net_device *dev;
2301 read_unlock(&mrt_lock);
2306 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2307 read_unlock(&mrt_lock);
2311 /* really correct? */
2312 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2314 read_unlock(&mrt_lock);
2318 skb_reset_transport_header(skb2);
2320 skb_put(skb2, sizeof(struct ipv6hdr));
2321 skb_reset_network_header(skb2);
2323 iph = ipv6_hdr(skb2);
2326 iph->flow_lbl[0] = 0;
2327 iph->flow_lbl[1] = 0;
2328 iph->flow_lbl[2] = 0;
2329 iph->payload_len = 0;
2330 iph->nexthdr = IPPROTO_NONE;
2332 iph->saddr = rt->rt6i_src.addr;
2333 iph->daddr = rt->rt6i_dst.addr;
2335 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2336 read_unlock(&mrt_lock);
2341 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2342 cache->mfc_flags |= MFC_NOTIFY;
2344 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2345 read_unlock(&mrt_lock);
2349 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2350 u32 portid, u32 seq, struct mfc6_cache *c, int cmd)
2352 struct nlmsghdr *nlh;
2356 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI);
2360 rtm = nlmsg_data(nlh);
2361 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2362 rtm->rtm_dst_len = 128;
2363 rtm->rtm_src_len = 128;
2365 rtm->rtm_table = mrt->id;
2366 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2367 goto nla_put_failure;
2368 rtm->rtm_type = RTN_MULTICAST;
2369 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2370 if (c->mfc_flags & MFC_STATIC)
2371 rtm->rtm_protocol = RTPROT_STATIC;
2373 rtm->rtm_protocol = RTPROT_MROUTED;
2376 if (nla_put(skb, RTA_SRC, 16, &c->mf6c_origin) ||
2377 nla_put(skb, RTA_DST, 16, &c->mf6c_mcastgrp))
2378 goto nla_put_failure;
2379 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2380 /* do not break the dump if cache is unresolved */
2381 if (err < 0 && err != -ENOENT)
2382 goto nla_put_failure;
2384 return nlmsg_end(skb, nlh);
2387 nlmsg_cancel(skb, nlh);
2391 static int mr6_msgsize(bool unresolved, int maxvif)
2394 NLMSG_ALIGN(sizeof(struct rtmsg))
2395 + nla_total_size(4) /* RTA_TABLE */
2396 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2397 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2402 + nla_total_size(4) /* RTA_IIF */
2403 + nla_total_size(0) /* RTA_MULTIPATH */
2404 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2406 + nla_total_size(sizeof(struct rta_mfc_stats))
2412 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2415 struct net *net = read_pnet(&mrt->net);
2416 struct sk_buff *skb;
2419 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2424 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd);
2428 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2434 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2437 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2439 struct net *net = sock_net(skb->sk);
2440 struct mr6_table *mrt;
2441 struct mfc6_cache *mfc;
2442 unsigned int t = 0, s_t;
2443 unsigned int h = 0, s_h;
2444 unsigned int e = 0, s_e;
2450 read_lock(&mrt_lock);
2451 ip6mr_for_each_table(mrt, net) {
2456 for (h = s_h; h < MFC6_LINES; h++) {
2457 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2460 if (ip6mr_fill_mroute(mrt, skb,
2461 NETLINK_CB(cb->skb).portid,
2463 mfc, RTM_NEWROUTE) < 0)
2470 spin_lock_bh(&mfc_unres_lock);
2471 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2474 if (ip6mr_fill_mroute(mrt, skb,
2475 NETLINK_CB(cb->skb).portid,
2477 mfc, RTM_NEWROUTE) < 0) {
2478 spin_unlock_bh(&mfc_unres_lock);
2484 spin_unlock_bh(&mfc_unres_lock);
2491 read_unlock(&mrt_lock);