Merge branch 'tty-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
[linux-2.6/cjktty.git] / net / ipv6 / ip6mr.c
blob9fab274019c0656dd4f01365ac14e46e0f1f764d
1 /*
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>
8 * 6WIND, Paris, France
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/system.h>
20 #include <asm/uaccess.h>
21 #include <linux/types.h>
22 #include <linux/sched.h>
23 #include <linux/errno.h>
24 #include <linux/timer.h>
25 #include <linux/mm.h>
26 #include <linux/kernel.h>
27 #include <linux/fcntl.h>
28 #include <linux/stat.h>
29 #include <linux/socket.h>
30 #include <linux/inet.h>
31 #include <linux/netdevice.h>
32 #include <linux/inetdevice.h>
33 #include <linux/proc_fs.h>
34 #include <linux/seq_file.h>
35 #include <linux/init.h>
36 #include <linux/slab.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
39 #include <net/sock.h>
40 #include <net/raw.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>
47 #include <net/ipv6.h>
48 #include <net/ip6_route.h>
49 #include <linux/mroute6.h>
50 #include <linux/pim.h>
51 #include <net/addrconf.h>
52 #include <linux/netfilter_ipv6.h>
53 #include <net/ip6_checksum.h>
55 struct mr6_table {
56 struct list_head list;
57 #ifdef CONFIG_NET_NS
58 struct net *net;
59 #endif
60 u32 id;
61 struct sock *mroute6_sk;
62 struct timer_list ipmr_expire_timer;
63 struct list_head mfc6_unres_queue;
64 struct list_head mfc6_cache_array[MFC6_LINES];
65 struct mif_device vif6_table[MAXMIFS];
66 int maxvif;
67 atomic_t cache_resolve_queue_len;
68 int mroute_do_assert;
69 int mroute_do_pim;
70 #ifdef CONFIG_IPV6_PIMSM_V2
71 int mroute_reg_vif_num;
72 #endif
75 struct ip6mr_rule {
76 struct fib_rule common;
79 struct ip6mr_result {
80 struct mr6_table *mrt;
83 /* Big lock, protecting vif table, mrt cache and mroute socket state.
84 Note that the changes are semaphored via rtnl_lock.
87 static DEFINE_RWLOCK(mrt_lock);
90 * Multicast router control variables
93 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
95 /* Special spinlock for queue of unresolved entries */
96 static DEFINE_SPINLOCK(mfc_unres_lock);
98 /* We return to original Alan's scheme. Hash table of resolved
99 entries is changed only in process context and protected
100 with weak lock mrt_lock. Queue of unresolved entries is protected
101 with strong spinlock mfc_unres_lock.
103 In this case data path is free of exclusive locks at all.
106 static struct kmem_cache *mrt_cachep __read_mostly;
108 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
109 static void ip6mr_free_table(struct mr6_table *mrt);
111 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
112 struct sk_buff *skb, struct mfc6_cache *cache);
113 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
114 mifi_t mifi, int assert);
115 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
116 struct mfc6_cache *c, struct rtmsg *rtm);
117 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
118 struct netlink_callback *cb);
119 static void mroute_clean_tables(struct mr6_table *mrt);
120 static void ipmr_expire_process(unsigned long arg);
122 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
123 #define ip6mr_for_each_table(mrt, net) \
124 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
126 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
128 struct mr6_table *mrt;
130 ip6mr_for_each_table(mrt, net) {
131 if (mrt->id == id)
132 return mrt;
134 return NULL;
137 static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
138 struct mr6_table **mrt)
140 struct ip6mr_result res;
141 struct fib_lookup_arg arg = { .result = &res, };
142 int err;
144 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, flp, 0, &arg);
145 if (err < 0)
146 return err;
147 *mrt = res.mrt;
148 return 0;
151 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
152 int flags, struct fib_lookup_arg *arg)
154 struct ip6mr_result *res = arg->result;
155 struct mr6_table *mrt;
157 switch (rule->action) {
158 case FR_ACT_TO_TBL:
159 break;
160 case FR_ACT_UNREACHABLE:
161 return -ENETUNREACH;
162 case FR_ACT_PROHIBIT:
163 return -EACCES;
164 case FR_ACT_BLACKHOLE:
165 default:
166 return -EINVAL;
169 mrt = ip6mr_get_table(rule->fr_net, rule->table);
170 if (mrt == NULL)
171 return -EAGAIN;
172 res->mrt = mrt;
173 return 0;
176 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
178 return 1;
181 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
182 FRA_GENERIC_POLICY,
185 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
186 struct fib_rule_hdr *frh, struct nlattr **tb)
188 return 0;
191 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
192 struct nlattr **tb)
194 return 1;
197 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
198 struct fib_rule_hdr *frh)
200 frh->dst_len = 0;
201 frh->src_len = 0;
202 frh->tos = 0;
203 return 0;
206 static const struct fib_rules_ops __net_initdata ip6mr_rules_ops_template = {
207 .family = RTNL_FAMILY_IP6MR,
208 .rule_size = sizeof(struct ip6mr_rule),
209 .addr_size = sizeof(struct in6_addr),
210 .action = ip6mr_rule_action,
211 .match = ip6mr_rule_match,
212 .configure = ip6mr_rule_configure,
213 .compare = ip6mr_rule_compare,
214 .default_pref = fib_default_rule_pref,
215 .fill = ip6mr_rule_fill,
216 .nlgroup = RTNLGRP_IPV6_RULE,
217 .policy = ip6mr_rule_policy,
218 .owner = THIS_MODULE,
221 static int __net_init ip6mr_rules_init(struct net *net)
223 struct fib_rules_ops *ops;
224 struct mr6_table *mrt;
225 int err;
227 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
228 if (IS_ERR(ops))
229 return PTR_ERR(ops);
231 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
233 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
234 if (mrt == NULL) {
235 err = -ENOMEM;
236 goto err1;
239 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
240 if (err < 0)
241 goto err2;
243 net->ipv6.mr6_rules_ops = ops;
244 return 0;
246 err2:
247 kfree(mrt);
248 err1:
249 fib_rules_unregister(ops);
250 return err;
253 static void __net_exit ip6mr_rules_exit(struct net *net)
255 struct mr6_table *mrt, *next;
257 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
258 list_del(&mrt->list);
259 ip6mr_free_table(mrt);
261 fib_rules_unregister(net->ipv6.mr6_rules_ops);
263 #else
264 #define ip6mr_for_each_table(mrt, net) \
265 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
267 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
269 return net->ipv6.mrt6;
272 static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
273 struct mr6_table **mrt)
275 *mrt = net->ipv6.mrt6;
276 return 0;
279 static int __net_init ip6mr_rules_init(struct net *net)
281 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
282 return net->ipv6.mrt6 ? 0 : -ENOMEM;
285 static void __net_exit ip6mr_rules_exit(struct net *net)
287 ip6mr_free_table(net->ipv6.mrt6);
289 #endif
291 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
293 struct mr6_table *mrt;
294 unsigned int i;
296 mrt = ip6mr_get_table(net, id);
297 if (mrt != NULL)
298 return mrt;
300 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
301 if (mrt == NULL)
302 return NULL;
303 mrt->id = id;
304 write_pnet(&mrt->net, net);
306 /* Forwarding cache */
307 for (i = 0; i < MFC6_LINES; i++)
308 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
310 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
312 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
313 (unsigned long)mrt);
315 #ifdef CONFIG_IPV6_PIMSM_V2
316 mrt->mroute_reg_vif_num = -1;
317 #endif
318 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
319 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
320 #endif
321 return mrt;
324 static void ip6mr_free_table(struct mr6_table *mrt)
326 del_timer(&mrt->ipmr_expire_timer);
327 mroute_clean_tables(mrt);
328 kfree(mrt);
331 #ifdef CONFIG_PROC_FS
333 struct ipmr_mfc_iter {
334 struct seq_net_private p;
335 struct mr6_table *mrt;
336 struct list_head *cache;
337 int ct;
341 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
342 struct ipmr_mfc_iter *it, loff_t pos)
344 struct mr6_table *mrt = it->mrt;
345 struct mfc6_cache *mfc;
347 read_lock(&mrt_lock);
348 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
349 it->cache = &mrt->mfc6_cache_array[it->ct];
350 list_for_each_entry(mfc, it->cache, list)
351 if (pos-- == 0)
352 return mfc;
354 read_unlock(&mrt_lock);
356 spin_lock_bh(&mfc_unres_lock);
357 it->cache = &mrt->mfc6_unres_queue;
358 list_for_each_entry(mfc, it->cache, list)
359 if (pos-- == 0)
360 return mfc;
361 spin_unlock_bh(&mfc_unres_lock);
363 it->cache = NULL;
364 return NULL;
368 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
371 struct ipmr_vif_iter {
372 struct seq_net_private p;
373 struct mr6_table *mrt;
374 int ct;
377 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
378 struct ipmr_vif_iter *iter,
379 loff_t pos)
381 struct mr6_table *mrt = iter->mrt;
383 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
384 if (!MIF_EXISTS(mrt, iter->ct))
385 continue;
386 if (pos-- == 0)
387 return &mrt->vif6_table[iter->ct];
389 return NULL;
392 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
393 __acquires(mrt_lock)
395 struct ipmr_vif_iter *iter = seq->private;
396 struct net *net = seq_file_net(seq);
397 struct mr6_table *mrt;
399 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
400 if (mrt == NULL)
401 return ERR_PTR(-ENOENT);
403 iter->mrt = mrt;
405 read_lock(&mrt_lock);
406 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
407 : SEQ_START_TOKEN;
410 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
412 struct ipmr_vif_iter *iter = seq->private;
413 struct net *net = seq_file_net(seq);
414 struct mr6_table *mrt = iter->mrt;
416 ++*pos;
417 if (v == SEQ_START_TOKEN)
418 return ip6mr_vif_seq_idx(net, iter, 0);
420 while (++iter->ct < mrt->maxvif) {
421 if (!MIF_EXISTS(mrt, iter->ct))
422 continue;
423 return &mrt->vif6_table[iter->ct];
425 return NULL;
428 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
429 __releases(mrt_lock)
431 read_unlock(&mrt_lock);
434 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
436 struct ipmr_vif_iter *iter = seq->private;
437 struct mr6_table *mrt = iter->mrt;
439 if (v == SEQ_START_TOKEN) {
440 seq_puts(seq,
441 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
442 } else {
443 const struct mif_device *vif = v;
444 const char *name = vif->dev ? vif->dev->name : "none";
446 seq_printf(seq,
447 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
448 vif - mrt->vif6_table,
449 name, vif->bytes_in, vif->pkt_in,
450 vif->bytes_out, vif->pkt_out,
451 vif->flags);
453 return 0;
456 static const struct seq_operations ip6mr_vif_seq_ops = {
457 .start = ip6mr_vif_seq_start,
458 .next = ip6mr_vif_seq_next,
459 .stop = ip6mr_vif_seq_stop,
460 .show = ip6mr_vif_seq_show,
463 static int ip6mr_vif_open(struct inode *inode, struct file *file)
465 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
466 sizeof(struct ipmr_vif_iter));
469 static const struct file_operations ip6mr_vif_fops = {
470 .owner = THIS_MODULE,
471 .open = ip6mr_vif_open,
472 .read = seq_read,
473 .llseek = seq_lseek,
474 .release = seq_release_net,
477 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
479 struct ipmr_mfc_iter *it = seq->private;
480 struct net *net = seq_file_net(seq);
481 struct mr6_table *mrt;
483 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
484 if (mrt == NULL)
485 return ERR_PTR(-ENOENT);
487 it->mrt = mrt;
488 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
489 : SEQ_START_TOKEN;
492 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
494 struct mfc6_cache *mfc = v;
495 struct ipmr_mfc_iter *it = seq->private;
496 struct net *net = seq_file_net(seq);
497 struct mr6_table *mrt = it->mrt;
499 ++*pos;
501 if (v == SEQ_START_TOKEN)
502 return ipmr_mfc_seq_idx(net, seq->private, 0);
504 if (mfc->list.next != it->cache)
505 return list_entry(mfc->list.next, struct mfc6_cache, list);
507 if (it->cache == &mrt->mfc6_unres_queue)
508 goto end_of_list;
510 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
512 while (++it->ct < MFC6_LINES) {
513 it->cache = &mrt->mfc6_cache_array[it->ct];
514 if (list_empty(it->cache))
515 continue;
516 return list_first_entry(it->cache, struct mfc6_cache, list);
519 /* exhausted cache_array, show unresolved */
520 read_unlock(&mrt_lock);
521 it->cache = &mrt->mfc6_unres_queue;
522 it->ct = 0;
524 spin_lock_bh(&mfc_unres_lock);
525 if (!list_empty(it->cache))
526 return list_first_entry(it->cache, struct mfc6_cache, list);
528 end_of_list:
529 spin_unlock_bh(&mfc_unres_lock);
530 it->cache = NULL;
532 return NULL;
535 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
537 struct ipmr_mfc_iter *it = seq->private;
538 struct mr6_table *mrt = it->mrt;
540 if (it->cache == &mrt->mfc6_unres_queue)
541 spin_unlock_bh(&mfc_unres_lock);
542 else if (it->cache == mrt->mfc6_cache_array)
543 read_unlock(&mrt_lock);
546 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
548 int n;
550 if (v == SEQ_START_TOKEN) {
551 seq_puts(seq,
552 "Group "
553 "Origin "
554 "Iif Pkts Bytes Wrong Oifs\n");
555 } else {
556 const struct mfc6_cache *mfc = v;
557 const struct ipmr_mfc_iter *it = seq->private;
558 struct mr6_table *mrt = it->mrt;
560 seq_printf(seq, "%pI6 %pI6 %-3hd",
561 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
562 mfc->mf6c_parent);
564 if (it->cache != &mrt->mfc6_unres_queue) {
565 seq_printf(seq, " %8lu %8lu %8lu",
566 mfc->mfc_un.res.pkt,
567 mfc->mfc_un.res.bytes,
568 mfc->mfc_un.res.wrong_if);
569 for (n = mfc->mfc_un.res.minvif;
570 n < mfc->mfc_un.res.maxvif; n++) {
571 if (MIF_EXISTS(mrt, n) &&
572 mfc->mfc_un.res.ttls[n] < 255)
573 seq_printf(seq,
574 " %2d:%-3d",
575 n, mfc->mfc_un.res.ttls[n]);
577 } else {
578 /* unresolved mfc_caches don't contain
579 * pkt, bytes and wrong_if values
581 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
583 seq_putc(seq, '\n');
585 return 0;
588 static const struct seq_operations ipmr_mfc_seq_ops = {
589 .start = ipmr_mfc_seq_start,
590 .next = ipmr_mfc_seq_next,
591 .stop = ipmr_mfc_seq_stop,
592 .show = ipmr_mfc_seq_show,
595 static int ipmr_mfc_open(struct inode *inode, struct file *file)
597 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
598 sizeof(struct ipmr_mfc_iter));
601 static const struct file_operations ip6mr_mfc_fops = {
602 .owner = THIS_MODULE,
603 .open = ipmr_mfc_open,
604 .read = seq_read,
605 .llseek = seq_lseek,
606 .release = seq_release_net,
608 #endif
610 #ifdef CONFIG_IPV6_PIMSM_V2
612 static int pim6_rcv(struct sk_buff *skb)
614 struct pimreghdr *pim;
615 struct ipv6hdr *encap;
616 struct net_device *reg_dev = NULL;
617 struct net *net = dev_net(skb->dev);
618 struct mr6_table *mrt;
619 struct flowi fl = {
620 .iif = skb->dev->ifindex,
621 .mark = skb->mark,
623 int reg_vif_num;
625 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
626 goto drop;
628 pim = (struct pimreghdr *)skb_transport_header(skb);
629 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
630 (pim->flags & PIM_NULL_REGISTER) ||
631 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
632 sizeof(*pim), IPPROTO_PIM,
633 csum_partial((void *)pim, sizeof(*pim), 0)) &&
634 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
635 goto drop;
637 /* check if the inner packet is destined to mcast group */
638 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
639 sizeof(*pim));
641 if (!ipv6_addr_is_multicast(&encap->daddr) ||
642 encap->payload_len == 0 ||
643 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
644 goto drop;
646 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
647 goto drop;
648 reg_vif_num = mrt->mroute_reg_vif_num;
650 read_lock(&mrt_lock);
651 if (reg_vif_num >= 0)
652 reg_dev = mrt->vif6_table[reg_vif_num].dev;
653 if (reg_dev)
654 dev_hold(reg_dev);
655 read_unlock(&mrt_lock);
657 if (reg_dev == NULL)
658 goto drop;
660 skb->mac_header = skb->network_header;
661 skb_pull(skb, (u8 *)encap - skb->data);
662 skb_reset_network_header(skb);
663 skb->protocol = htons(ETH_P_IPV6);
664 skb->ip_summed = 0;
665 skb->pkt_type = PACKET_HOST;
667 skb_tunnel_rx(skb, reg_dev);
669 netif_rx(skb);
671 dev_put(reg_dev);
672 return 0;
673 drop:
674 kfree_skb(skb);
675 return 0;
678 static const struct inet6_protocol pim6_protocol = {
679 .handler = pim6_rcv,
682 /* Service routines creating virtual interfaces: PIMREG */
684 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
685 struct net_device *dev)
687 struct net *net = dev_net(dev);
688 struct mr6_table *mrt;
689 struct flowi fl = {
690 .oif = dev->ifindex,
691 .iif = skb->skb_iif,
692 .mark = skb->mark,
694 int err;
696 err = ip6mr_fib_lookup(net, &fl, &mrt);
697 if (err < 0)
698 return err;
700 read_lock(&mrt_lock);
701 dev->stats.tx_bytes += skb->len;
702 dev->stats.tx_packets++;
703 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
704 read_unlock(&mrt_lock);
705 kfree_skb(skb);
706 return NETDEV_TX_OK;
709 static const struct net_device_ops reg_vif_netdev_ops = {
710 .ndo_start_xmit = reg_vif_xmit,
713 static void reg_vif_setup(struct net_device *dev)
715 dev->type = ARPHRD_PIMREG;
716 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
717 dev->flags = IFF_NOARP;
718 dev->netdev_ops = &reg_vif_netdev_ops;
719 dev->destructor = free_netdev;
720 dev->features |= NETIF_F_NETNS_LOCAL;
723 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
725 struct net_device *dev;
726 char name[IFNAMSIZ];
728 if (mrt->id == RT6_TABLE_DFLT)
729 sprintf(name, "pim6reg");
730 else
731 sprintf(name, "pim6reg%u", mrt->id);
733 dev = alloc_netdev(0, name, reg_vif_setup);
734 if (dev == NULL)
735 return NULL;
737 dev_net_set(dev, net);
739 if (register_netdevice(dev)) {
740 free_netdev(dev);
741 return NULL;
743 dev->iflink = 0;
745 if (dev_open(dev))
746 goto failure;
748 dev_hold(dev);
749 return dev;
751 failure:
752 /* allow the register to be completed before unregistering. */
753 rtnl_unlock();
754 rtnl_lock();
756 unregister_netdevice(dev);
757 return NULL;
759 #endif
762 * Delete a VIF entry
765 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
767 struct mif_device *v;
768 struct net_device *dev;
769 struct inet6_dev *in6_dev;
771 if (vifi < 0 || vifi >= mrt->maxvif)
772 return -EADDRNOTAVAIL;
774 v = &mrt->vif6_table[vifi];
776 write_lock_bh(&mrt_lock);
777 dev = v->dev;
778 v->dev = NULL;
780 if (!dev) {
781 write_unlock_bh(&mrt_lock);
782 return -EADDRNOTAVAIL;
785 #ifdef CONFIG_IPV6_PIMSM_V2
786 if (vifi == mrt->mroute_reg_vif_num)
787 mrt->mroute_reg_vif_num = -1;
788 #endif
790 if (vifi + 1 == mrt->maxvif) {
791 int tmp;
792 for (tmp = vifi - 1; tmp >= 0; tmp--) {
793 if (MIF_EXISTS(mrt, tmp))
794 break;
796 mrt->maxvif = tmp + 1;
799 write_unlock_bh(&mrt_lock);
801 dev_set_allmulti(dev, -1);
803 in6_dev = __in6_dev_get(dev);
804 if (in6_dev)
805 in6_dev->cnf.mc_forwarding--;
807 if (v->flags & MIFF_REGISTER)
808 unregister_netdevice_queue(dev, head);
810 dev_put(dev);
811 return 0;
814 static inline void ip6mr_cache_free(struct mfc6_cache *c)
816 kmem_cache_free(mrt_cachep, c);
819 /* Destroy an unresolved cache entry, killing queued skbs
820 and reporting error to netlink readers.
823 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
825 struct net *net = read_pnet(&mrt->net);
826 struct sk_buff *skb;
828 atomic_dec(&mrt->cache_resolve_queue_len);
830 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
831 if (ipv6_hdr(skb)->version == 0) {
832 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
833 nlh->nlmsg_type = NLMSG_ERROR;
834 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
835 skb_trim(skb, nlh->nlmsg_len);
836 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
837 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
838 } else
839 kfree_skb(skb);
842 ip6mr_cache_free(c);
846 /* Timer process for all the unresolved queue. */
848 static void ipmr_do_expire_process(struct mr6_table *mrt)
850 unsigned long now = jiffies;
851 unsigned long expires = 10 * HZ;
852 struct mfc6_cache *c, *next;
854 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
855 if (time_after(c->mfc_un.unres.expires, now)) {
856 /* not yet... */
857 unsigned long interval = c->mfc_un.unres.expires - now;
858 if (interval < expires)
859 expires = interval;
860 continue;
863 list_del(&c->list);
864 ip6mr_destroy_unres(mrt, c);
867 if (!list_empty(&mrt->mfc6_unres_queue))
868 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
871 static void ipmr_expire_process(unsigned long arg)
873 struct mr6_table *mrt = (struct mr6_table *)arg;
875 if (!spin_trylock(&mfc_unres_lock)) {
876 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
877 return;
880 if (!list_empty(&mrt->mfc6_unres_queue))
881 ipmr_do_expire_process(mrt);
883 spin_unlock(&mfc_unres_lock);
886 /* Fill oifs list. It is called under write locked mrt_lock. */
888 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
889 unsigned char *ttls)
891 int vifi;
893 cache->mfc_un.res.minvif = MAXMIFS;
894 cache->mfc_un.res.maxvif = 0;
895 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
897 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
898 if (MIF_EXISTS(mrt, vifi) &&
899 ttls[vifi] && ttls[vifi] < 255) {
900 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
901 if (cache->mfc_un.res.minvif > vifi)
902 cache->mfc_un.res.minvif = vifi;
903 if (cache->mfc_un.res.maxvif <= vifi)
904 cache->mfc_un.res.maxvif = vifi + 1;
909 static int mif6_add(struct net *net, struct mr6_table *mrt,
910 struct mif6ctl *vifc, int mrtsock)
912 int vifi = vifc->mif6c_mifi;
913 struct mif_device *v = &mrt->vif6_table[vifi];
914 struct net_device *dev;
915 struct inet6_dev *in6_dev;
916 int err;
918 /* Is vif busy ? */
919 if (MIF_EXISTS(mrt, vifi))
920 return -EADDRINUSE;
922 switch (vifc->mif6c_flags) {
923 #ifdef CONFIG_IPV6_PIMSM_V2
924 case MIFF_REGISTER:
926 * Special Purpose VIF in PIM
927 * All the packets will be sent to the daemon
929 if (mrt->mroute_reg_vif_num >= 0)
930 return -EADDRINUSE;
931 dev = ip6mr_reg_vif(net, mrt);
932 if (!dev)
933 return -ENOBUFS;
934 err = dev_set_allmulti(dev, 1);
935 if (err) {
936 unregister_netdevice(dev);
937 dev_put(dev);
938 return err;
940 break;
941 #endif
942 case 0:
943 dev = dev_get_by_index(net, vifc->mif6c_pifi);
944 if (!dev)
945 return -EADDRNOTAVAIL;
946 err = dev_set_allmulti(dev, 1);
947 if (err) {
948 dev_put(dev);
949 return err;
951 break;
952 default:
953 return -EINVAL;
956 in6_dev = __in6_dev_get(dev);
957 if (in6_dev)
958 in6_dev->cnf.mc_forwarding++;
961 * Fill in the VIF structures
963 v->rate_limit = vifc->vifc_rate_limit;
964 v->flags = vifc->mif6c_flags;
965 if (!mrtsock)
966 v->flags |= VIFF_STATIC;
967 v->threshold = vifc->vifc_threshold;
968 v->bytes_in = 0;
969 v->bytes_out = 0;
970 v->pkt_in = 0;
971 v->pkt_out = 0;
972 v->link = dev->ifindex;
973 if (v->flags & MIFF_REGISTER)
974 v->link = dev->iflink;
976 /* And finish update writing critical data */
977 write_lock_bh(&mrt_lock);
978 v->dev = dev;
979 #ifdef CONFIG_IPV6_PIMSM_V2
980 if (v->flags & MIFF_REGISTER)
981 mrt->mroute_reg_vif_num = vifi;
982 #endif
983 if (vifi + 1 > mrt->maxvif)
984 mrt->maxvif = vifi + 1;
985 write_unlock_bh(&mrt_lock);
986 return 0;
989 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
990 struct in6_addr *origin,
991 struct in6_addr *mcastgrp)
993 int line = MFC6_HASH(mcastgrp, origin);
994 struct mfc6_cache *c;
996 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
997 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
998 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
999 return c;
1001 return NULL;
1005 * Allocate a multicast cache entry
1007 static struct mfc6_cache *ip6mr_cache_alloc(void)
1009 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1010 if (c == NULL)
1011 return NULL;
1012 c->mfc_un.res.minvif = MAXMIFS;
1013 return c;
1016 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1018 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1019 if (c == NULL)
1020 return NULL;
1021 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1022 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1023 return c;
1027 * A cache entry has gone into a resolved state from queued
1030 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1031 struct mfc6_cache *uc, struct mfc6_cache *c)
1033 struct sk_buff *skb;
1036 * Play the pending entries through our router
1039 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1040 if (ipv6_hdr(skb)->version == 0) {
1041 int err;
1042 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1044 if (__ip6mr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) {
1045 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1046 } else {
1047 nlh->nlmsg_type = NLMSG_ERROR;
1048 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
1049 skb_trim(skb, nlh->nlmsg_len);
1050 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
1052 err = rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
1053 } else
1054 ip6_mr_forward(net, mrt, skb, c);
1059 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1060 * expects the following bizarre scheme.
1062 * Called under mrt_lock.
1065 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1066 mifi_t mifi, int assert)
1068 struct sk_buff *skb;
1069 struct mrt6msg *msg;
1070 int ret;
1072 #ifdef CONFIG_IPV6_PIMSM_V2
1073 if (assert == MRT6MSG_WHOLEPKT)
1074 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1075 +sizeof(*msg));
1076 else
1077 #endif
1078 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1080 if (!skb)
1081 return -ENOBUFS;
1083 /* I suppose that internal messages
1084 * do not require checksums */
1086 skb->ip_summed = CHECKSUM_UNNECESSARY;
1088 #ifdef CONFIG_IPV6_PIMSM_V2
1089 if (assert == MRT6MSG_WHOLEPKT) {
1090 /* Ugly, but we have no choice with this interface.
1091 Duplicate old header, fix length etc.
1092 And all this only to mangle msg->im6_msgtype and
1093 to set msg->im6_mbz to "mbz" :-)
1095 skb_push(skb, -skb_network_offset(pkt));
1097 skb_push(skb, sizeof(*msg));
1098 skb_reset_transport_header(skb);
1099 msg = (struct mrt6msg *)skb_transport_header(skb);
1100 msg->im6_mbz = 0;
1101 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1102 msg->im6_mif = mrt->mroute_reg_vif_num;
1103 msg->im6_pad = 0;
1104 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1105 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1107 skb->ip_summed = CHECKSUM_UNNECESSARY;
1108 } else
1109 #endif
1112 * Copy the IP header
1115 skb_put(skb, sizeof(struct ipv6hdr));
1116 skb_reset_network_header(skb);
1117 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1120 * Add our header
1122 skb_put(skb, sizeof(*msg));
1123 skb_reset_transport_header(skb);
1124 msg = (struct mrt6msg *)skb_transport_header(skb);
1126 msg->im6_mbz = 0;
1127 msg->im6_msgtype = assert;
1128 msg->im6_mif = mifi;
1129 msg->im6_pad = 0;
1130 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1131 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1133 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1134 skb->ip_summed = CHECKSUM_UNNECESSARY;
1137 if (mrt->mroute6_sk == NULL) {
1138 kfree_skb(skb);
1139 return -EINVAL;
1143 * Deliver to user space multicast routing algorithms
1145 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1146 if (ret < 0) {
1147 if (net_ratelimit())
1148 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
1149 kfree_skb(skb);
1152 return ret;
1156 * Queue a packet for resolution. It gets locked cache entry!
1159 static int
1160 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1162 bool found = false;
1163 int err;
1164 struct mfc6_cache *c;
1166 spin_lock_bh(&mfc_unres_lock);
1167 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1168 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1169 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1170 found = true;
1171 break;
1175 if (!found) {
1177 * Create a new entry if allowable
1180 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1181 (c = ip6mr_cache_alloc_unres()) == NULL) {
1182 spin_unlock_bh(&mfc_unres_lock);
1184 kfree_skb(skb);
1185 return -ENOBUFS;
1189 * Fill in the new cache entry
1191 c->mf6c_parent = -1;
1192 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1193 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1196 * Reflect first query at pim6sd
1198 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1199 if (err < 0) {
1200 /* If the report failed throw the cache entry
1201 out - Brad Parker
1203 spin_unlock_bh(&mfc_unres_lock);
1205 ip6mr_cache_free(c);
1206 kfree_skb(skb);
1207 return err;
1210 atomic_inc(&mrt->cache_resolve_queue_len);
1211 list_add(&c->list, &mrt->mfc6_unres_queue);
1213 ipmr_do_expire_process(mrt);
1217 * See if we can append the packet
1219 if (c->mfc_un.unres.unresolved.qlen > 3) {
1220 kfree_skb(skb);
1221 err = -ENOBUFS;
1222 } else {
1223 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1224 err = 0;
1227 spin_unlock_bh(&mfc_unres_lock);
1228 return err;
1232 * MFC6 cache manipulation by user space
1235 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc)
1237 int line;
1238 struct mfc6_cache *c, *next;
1240 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1242 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1243 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1244 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1245 write_lock_bh(&mrt_lock);
1246 list_del(&c->list);
1247 write_unlock_bh(&mrt_lock);
1249 ip6mr_cache_free(c);
1250 return 0;
1253 return -ENOENT;
1256 static int ip6mr_device_event(struct notifier_block *this,
1257 unsigned long event, void *ptr)
1259 struct net_device *dev = ptr;
1260 struct net *net = dev_net(dev);
1261 struct mr6_table *mrt;
1262 struct mif_device *v;
1263 int ct;
1264 LIST_HEAD(list);
1266 if (event != NETDEV_UNREGISTER)
1267 return NOTIFY_DONE;
1269 ip6mr_for_each_table(mrt, net) {
1270 v = &mrt->vif6_table[0];
1271 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1272 if (v->dev == dev)
1273 mif6_delete(mrt, ct, &list);
1276 unregister_netdevice_many(&list);
1278 return NOTIFY_DONE;
1281 static struct notifier_block ip6_mr_notifier = {
1282 .notifier_call = ip6mr_device_event
1286 * Setup for IP multicast routing
1289 static int __net_init ip6mr_net_init(struct net *net)
1291 int err;
1293 err = ip6mr_rules_init(net);
1294 if (err < 0)
1295 goto fail;
1297 #ifdef CONFIG_PROC_FS
1298 err = -ENOMEM;
1299 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
1300 goto proc_vif_fail;
1301 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops))
1302 goto proc_cache_fail;
1303 #endif
1305 return 0;
1307 #ifdef CONFIG_PROC_FS
1308 proc_cache_fail:
1309 proc_net_remove(net, "ip6_mr_vif");
1310 proc_vif_fail:
1311 ip6mr_rules_exit(net);
1312 #endif
1313 fail:
1314 return err;
1317 static void __net_exit ip6mr_net_exit(struct net *net)
1319 #ifdef CONFIG_PROC_FS
1320 proc_net_remove(net, "ip6_mr_cache");
1321 proc_net_remove(net, "ip6_mr_vif");
1322 #endif
1323 ip6mr_rules_exit(net);
1326 static struct pernet_operations ip6mr_net_ops = {
1327 .init = ip6mr_net_init,
1328 .exit = ip6mr_net_exit,
1331 int __init ip6_mr_init(void)
1333 int err;
1335 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1336 sizeof(struct mfc6_cache),
1337 0, SLAB_HWCACHE_ALIGN,
1338 NULL);
1339 if (!mrt_cachep)
1340 return -ENOMEM;
1342 err = register_pernet_subsys(&ip6mr_net_ops);
1343 if (err)
1344 goto reg_pernet_fail;
1346 err = register_netdevice_notifier(&ip6_mr_notifier);
1347 if (err)
1348 goto reg_notif_fail;
1349 #ifdef CONFIG_IPV6_PIMSM_V2
1350 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1351 printk(KERN_ERR "ip6_mr_init: can't add PIM protocol\n");
1352 err = -EAGAIN;
1353 goto add_proto_fail;
1355 #endif
1356 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, ip6mr_rtm_dumproute);
1357 return 0;
1358 #ifdef CONFIG_IPV6_PIMSM_V2
1359 add_proto_fail:
1360 unregister_netdevice_notifier(&ip6_mr_notifier);
1361 #endif
1362 reg_notif_fail:
1363 unregister_pernet_subsys(&ip6mr_net_ops);
1364 reg_pernet_fail:
1365 kmem_cache_destroy(mrt_cachep);
1366 return err;
1369 void ip6_mr_cleanup(void)
1371 unregister_netdevice_notifier(&ip6_mr_notifier);
1372 unregister_pernet_subsys(&ip6mr_net_ops);
1373 kmem_cache_destroy(mrt_cachep);
1376 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1377 struct mf6cctl *mfc, int mrtsock)
1379 bool found = false;
1380 int line;
1381 struct mfc6_cache *uc, *c;
1382 unsigned char ttls[MAXMIFS];
1383 int i;
1385 if (mfc->mf6cc_parent >= MAXMIFS)
1386 return -ENFILE;
1388 memset(ttls, 255, MAXMIFS);
1389 for (i = 0; i < MAXMIFS; i++) {
1390 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1391 ttls[i] = 1;
1395 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1397 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1398 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1399 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1400 found = true;
1401 break;
1405 if (found) {
1406 write_lock_bh(&mrt_lock);
1407 c->mf6c_parent = mfc->mf6cc_parent;
1408 ip6mr_update_thresholds(mrt, c, ttls);
1409 if (!mrtsock)
1410 c->mfc_flags |= MFC_STATIC;
1411 write_unlock_bh(&mrt_lock);
1412 return 0;
1415 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1416 return -EINVAL;
1418 c = ip6mr_cache_alloc();
1419 if (c == NULL)
1420 return -ENOMEM;
1422 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1423 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1424 c->mf6c_parent = mfc->mf6cc_parent;
1425 ip6mr_update_thresholds(mrt, c, ttls);
1426 if (!mrtsock)
1427 c->mfc_flags |= MFC_STATIC;
1429 write_lock_bh(&mrt_lock);
1430 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1431 write_unlock_bh(&mrt_lock);
1434 * Check to see if we resolved a queued list. If so we
1435 * need to send on the frames and tidy up.
1437 found = false;
1438 spin_lock_bh(&mfc_unres_lock);
1439 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1440 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1441 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1442 list_del(&uc->list);
1443 atomic_dec(&mrt->cache_resolve_queue_len);
1444 found = true;
1445 break;
1448 if (list_empty(&mrt->mfc6_unres_queue))
1449 del_timer(&mrt->ipmr_expire_timer);
1450 spin_unlock_bh(&mfc_unres_lock);
1452 if (found) {
1453 ip6mr_cache_resolve(net, mrt, uc, c);
1454 ip6mr_cache_free(uc);
1456 return 0;
1460 * Close the multicast socket, and clear the vif tables etc
1463 static void mroute_clean_tables(struct mr6_table *mrt)
1465 int i;
1466 LIST_HEAD(list);
1467 struct mfc6_cache *c, *next;
1470 * Shut down all active vif entries
1472 for (i = 0; i < mrt->maxvif; i++) {
1473 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1474 mif6_delete(mrt, i, &list);
1476 unregister_netdevice_many(&list);
1479 * Wipe the cache
1481 for (i = 0; i < MFC6_LINES; i++) {
1482 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1483 if (c->mfc_flags & MFC_STATIC)
1484 continue;
1485 write_lock_bh(&mrt_lock);
1486 list_del(&c->list);
1487 write_unlock_bh(&mrt_lock);
1489 ip6mr_cache_free(c);
1493 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1494 spin_lock_bh(&mfc_unres_lock);
1495 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1496 list_del(&c->list);
1497 ip6mr_destroy_unres(mrt, c);
1499 spin_unlock_bh(&mfc_unres_lock);
1503 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1505 int err = 0;
1506 struct net *net = sock_net(sk);
1508 rtnl_lock();
1509 write_lock_bh(&mrt_lock);
1510 if (likely(mrt->mroute6_sk == NULL)) {
1511 mrt->mroute6_sk = sk;
1512 net->ipv6.devconf_all->mc_forwarding++;
1514 else
1515 err = -EADDRINUSE;
1516 write_unlock_bh(&mrt_lock);
1518 rtnl_unlock();
1520 return err;
1523 int ip6mr_sk_done(struct sock *sk)
1525 int err = -EACCES;
1526 struct net *net = sock_net(sk);
1527 struct mr6_table *mrt;
1529 rtnl_lock();
1530 ip6mr_for_each_table(mrt, net) {
1531 if (sk == mrt->mroute6_sk) {
1532 write_lock_bh(&mrt_lock);
1533 mrt->mroute6_sk = NULL;
1534 net->ipv6.devconf_all->mc_forwarding--;
1535 write_unlock_bh(&mrt_lock);
1537 mroute_clean_tables(mrt);
1538 err = 0;
1539 break;
1542 rtnl_unlock();
1544 return err;
1547 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1549 struct mr6_table *mrt;
1550 struct flowi fl = {
1551 .iif = skb->skb_iif,
1552 .oif = skb->dev->ifindex,
1553 .mark = skb->mark,
1556 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
1557 return NULL;
1559 return mrt->mroute6_sk;
1563 * Socket options and virtual interface manipulation. The whole
1564 * virtual interface system is a complete heap, but unfortunately
1565 * that's how BSD mrouted happens to think. Maybe one day with a proper
1566 * MOSPF/PIM router set up we can clean this up.
1569 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1571 int ret;
1572 struct mif6ctl vif;
1573 struct mf6cctl mfc;
1574 mifi_t mifi;
1575 struct net *net = sock_net(sk);
1576 struct mr6_table *mrt;
1578 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1579 if (mrt == NULL)
1580 return -ENOENT;
1582 if (optname != MRT6_INIT) {
1583 if (sk != mrt->mroute6_sk && !capable(CAP_NET_ADMIN))
1584 return -EACCES;
1587 switch (optname) {
1588 case MRT6_INIT:
1589 if (sk->sk_type != SOCK_RAW ||
1590 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1591 return -EOPNOTSUPP;
1592 if (optlen < sizeof(int))
1593 return -EINVAL;
1595 return ip6mr_sk_init(mrt, sk);
1597 case MRT6_DONE:
1598 return ip6mr_sk_done(sk);
1600 case MRT6_ADD_MIF:
1601 if (optlen < sizeof(vif))
1602 return -EINVAL;
1603 if (copy_from_user(&vif, optval, sizeof(vif)))
1604 return -EFAULT;
1605 if (vif.mif6c_mifi >= MAXMIFS)
1606 return -ENFILE;
1607 rtnl_lock();
1608 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1609 rtnl_unlock();
1610 return ret;
1612 case MRT6_DEL_MIF:
1613 if (optlen < sizeof(mifi_t))
1614 return -EINVAL;
1615 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1616 return -EFAULT;
1617 rtnl_lock();
1618 ret = mif6_delete(mrt, mifi, NULL);
1619 rtnl_unlock();
1620 return ret;
1623 * Manipulate the forwarding caches. These live
1624 * in a sort of kernel/user symbiosis.
1626 case MRT6_ADD_MFC:
1627 case MRT6_DEL_MFC:
1628 if (optlen < sizeof(mfc))
1629 return -EINVAL;
1630 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1631 return -EFAULT;
1632 rtnl_lock();
1633 if (optname == MRT6_DEL_MFC)
1634 ret = ip6mr_mfc_delete(mrt, &mfc);
1635 else
1636 ret = ip6mr_mfc_add(net, mrt, &mfc, sk == mrt->mroute6_sk);
1637 rtnl_unlock();
1638 return ret;
1641 * Control PIM assert (to activate pim will activate assert)
1643 case MRT6_ASSERT:
1645 int v;
1646 if (get_user(v, (int __user *)optval))
1647 return -EFAULT;
1648 mrt->mroute_do_assert = !!v;
1649 return 0;
1652 #ifdef CONFIG_IPV6_PIMSM_V2
1653 case MRT6_PIM:
1655 int v;
1656 if (get_user(v, (int __user *)optval))
1657 return -EFAULT;
1658 v = !!v;
1659 rtnl_lock();
1660 ret = 0;
1661 if (v != mrt->mroute_do_pim) {
1662 mrt->mroute_do_pim = v;
1663 mrt->mroute_do_assert = v;
1665 rtnl_unlock();
1666 return ret;
1669 #endif
1670 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1671 case MRT6_TABLE:
1673 u32 v;
1675 if (optlen != sizeof(u32))
1676 return -EINVAL;
1677 if (get_user(v, (u32 __user *)optval))
1678 return -EFAULT;
1679 if (sk == mrt->mroute6_sk)
1680 return -EBUSY;
1682 rtnl_lock();
1683 ret = 0;
1684 if (!ip6mr_new_table(net, v))
1685 ret = -ENOMEM;
1686 raw6_sk(sk)->ip6mr_table = v;
1687 rtnl_unlock();
1688 return ret;
1690 #endif
1692 * Spurious command, or MRT6_VERSION which you cannot
1693 * set.
1695 default:
1696 return -ENOPROTOOPT;
1701 * Getsock opt support for the multicast routing system.
1704 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1705 int __user *optlen)
1707 int olr;
1708 int val;
1709 struct net *net = sock_net(sk);
1710 struct mr6_table *mrt;
1712 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1713 if (mrt == NULL)
1714 return -ENOENT;
1716 switch (optname) {
1717 case MRT6_VERSION:
1718 val = 0x0305;
1719 break;
1720 #ifdef CONFIG_IPV6_PIMSM_V2
1721 case MRT6_PIM:
1722 val = mrt->mroute_do_pim;
1723 break;
1724 #endif
1725 case MRT6_ASSERT:
1726 val = mrt->mroute_do_assert;
1727 break;
1728 default:
1729 return -ENOPROTOOPT;
1732 if (get_user(olr, optlen))
1733 return -EFAULT;
1735 olr = min_t(int, olr, sizeof(int));
1736 if (olr < 0)
1737 return -EINVAL;
1739 if (put_user(olr, optlen))
1740 return -EFAULT;
1741 if (copy_to_user(optval, &val, olr))
1742 return -EFAULT;
1743 return 0;
1747 * The IP multicast ioctl support routines.
1750 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1752 struct sioc_sg_req6 sr;
1753 struct sioc_mif_req6 vr;
1754 struct mif_device *vif;
1755 struct mfc6_cache *c;
1756 struct net *net = sock_net(sk);
1757 struct mr6_table *mrt;
1759 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1760 if (mrt == NULL)
1761 return -ENOENT;
1763 switch (cmd) {
1764 case SIOCGETMIFCNT_IN6:
1765 if (copy_from_user(&vr, arg, sizeof(vr)))
1766 return -EFAULT;
1767 if (vr.mifi >= mrt->maxvif)
1768 return -EINVAL;
1769 read_lock(&mrt_lock);
1770 vif = &mrt->vif6_table[vr.mifi];
1771 if (MIF_EXISTS(mrt, vr.mifi)) {
1772 vr.icount = vif->pkt_in;
1773 vr.ocount = vif->pkt_out;
1774 vr.ibytes = vif->bytes_in;
1775 vr.obytes = vif->bytes_out;
1776 read_unlock(&mrt_lock);
1778 if (copy_to_user(arg, &vr, sizeof(vr)))
1779 return -EFAULT;
1780 return 0;
1782 read_unlock(&mrt_lock);
1783 return -EADDRNOTAVAIL;
1784 case SIOCGETSGCNT_IN6:
1785 if (copy_from_user(&sr, arg, sizeof(sr)))
1786 return -EFAULT;
1788 read_lock(&mrt_lock);
1789 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1790 if (c) {
1791 sr.pktcnt = c->mfc_un.res.pkt;
1792 sr.bytecnt = c->mfc_un.res.bytes;
1793 sr.wrong_if = c->mfc_un.res.wrong_if;
1794 read_unlock(&mrt_lock);
1796 if (copy_to_user(arg, &sr, sizeof(sr)))
1797 return -EFAULT;
1798 return 0;
1800 read_unlock(&mrt_lock);
1801 return -EADDRNOTAVAIL;
1802 default:
1803 return -ENOIOCTLCMD;
1808 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1810 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1811 IPSTATS_MIB_OUTFORWDATAGRAMS);
1812 return dst_output(skb);
1816 * Processing handlers for ip6mr_forward
1819 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1820 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1822 struct ipv6hdr *ipv6h;
1823 struct mif_device *vif = &mrt->vif6_table[vifi];
1824 struct net_device *dev;
1825 struct dst_entry *dst;
1826 struct flowi fl;
1828 if (vif->dev == NULL)
1829 goto out_free;
1831 #ifdef CONFIG_IPV6_PIMSM_V2
1832 if (vif->flags & MIFF_REGISTER) {
1833 vif->pkt_out++;
1834 vif->bytes_out += skb->len;
1835 vif->dev->stats.tx_bytes += skb->len;
1836 vif->dev->stats.tx_packets++;
1837 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1838 goto out_free;
1840 #endif
1842 ipv6h = ipv6_hdr(skb);
1844 fl = (struct flowi) {
1845 .oif = vif->link,
1846 .fl6_dst = ipv6h->daddr,
1849 dst = ip6_route_output(net, NULL, &fl);
1850 if (!dst)
1851 goto out_free;
1853 skb_dst_drop(skb);
1854 skb_dst_set(skb, dst);
1857 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1858 * not only before forwarding, but after forwarding on all output
1859 * interfaces. It is clear, if mrouter runs a multicasting
1860 * program, it should receive packets not depending to what interface
1861 * program is joined.
1862 * If we will not make it, the program will have to join on all
1863 * interfaces. On the other hand, multihoming host (or router, but
1864 * not mrouter) cannot join to more than one interface - it will
1865 * result in receiving multiple packets.
1867 dev = vif->dev;
1868 skb->dev = dev;
1869 vif->pkt_out++;
1870 vif->bytes_out += skb->len;
1872 /* We are about to write */
1873 /* XXX: extension headers? */
1874 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1875 goto out_free;
1877 ipv6h = ipv6_hdr(skb);
1878 ipv6h->hop_limit--;
1880 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1882 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
1883 ip6mr_forward2_finish);
1885 out_free:
1886 kfree_skb(skb);
1887 return 0;
1890 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
1892 int ct;
1894 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
1895 if (mrt->vif6_table[ct].dev == dev)
1896 break;
1898 return ct;
1901 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
1902 struct sk_buff *skb, struct mfc6_cache *cache)
1904 int psend = -1;
1905 int vif, ct;
1907 vif = cache->mf6c_parent;
1908 cache->mfc_un.res.pkt++;
1909 cache->mfc_un.res.bytes += skb->len;
1912 * Wrong interface: drop packet and (maybe) send PIM assert.
1914 if (mrt->vif6_table[vif].dev != skb->dev) {
1915 int true_vifi;
1917 cache->mfc_un.res.wrong_if++;
1918 true_vifi = ip6mr_find_vif(mrt, skb->dev);
1920 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1921 /* pimsm uses asserts, when switching from RPT to SPT,
1922 so that we cannot check that packet arrived on an oif.
1923 It is bad, but otherwise we would need to move pretty
1924 large chunk of pimd to kernel. Ough... --ANK
1926 (mrt->mroute_do_pim ||
1927 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1928 time_after(jiffies,
1929 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1930 cache->mfc_un.res.last_assert = jiffies;
1931 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
1933 goto dont_forward;
1936 mrt->vif6_table[vif].pkt_in++;
1937 mrt->vif6_table[vif].bytes_in += skb->len;
1940 * Forward the frame
1942 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1943 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1944 if (psend != -1) {
1945 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1946 if (skb2)
1947 ip6mr_forward2(net, mrt, skb2, cache, psend);
1949 psend = ct;
1952 if (psend != -1) {
1953 ip6mr_forward2(net, mrt, skb, cache, psend);
1954 return 0;
1957 dont_forward:
1958 kfree_skb(skb);
1959 return 0;
1964 * Multicast packets for forwarding arrive here
1967 int ip6_mr_input(struct sk_buff *skb)
1969 struct mfc6_cache *cache;
1970 struct net *net = dev_net(skb->dev);
1971 struct mr6_table *mrt;
1972 struct flowi fl = {
1973 .iif = skb->dev->ifindex,
1974 .mark = skb->mark,
1976 int err;
1978 err = ip6mr_fib_lookup(net, &fl, &mrt);
1979 if (err < 0)
1980 return err;
1982 read_lock(&mrt_lock);
1983 cache = ip6mr_cache_find(mrt,
1984 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1987 * No usable cache entry
1989 if (cache == NULL) {
1990 int vif;
1992 vif = ip6mr_find_vif(mrt, skb->dev);
1993 if (vif >= 0) {
1994 int err = ip6mr_cache_unresolved(mrt, vif, skb);
1995 read_unlock(&mrt_lock);
1997 return err;
1999 read_unlock(&mrt_lock);
2000 kfree_skb(skb);
2001 return -ENODEV;
2004 ip6_mr_forward(net, mrt, skb, cache);
2006 read_unlock(&mrt_lock);
2008 return 0;
2012 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2013 struct mfc6_cache *c, struct rtmsg *rtm)
2015 int ct;
2016 struct rtnexthop *nhp;
2017 u8 *b = skb_tail_pointer(skb);
2018 struct rtattr *mp_head;
2020 /* If cache is unresolved, don't try to parse IIF and OIF */
2021 if (c->mf6c_parent >= MAXMIFS)
2022 return -ENOENT;
2024 if (MIF_EXISTS(mrt, c->mf6c_parent))
2025 RTA_PUT(skb, RTA_IIF, 4, &mrt->vif6_table[c->mf6c_parent].dev->ifindex);
2027 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
2029 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2030 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2031 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
2032 goto rtattr_failure;
2033 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
2034 nhp->rtnh_flags = 0;
2035 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2036 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2037 nhp->rtnh_len = sizeof(*nhp);
2040 mp_head->rta_type = RTA_MULTIPATH;
2041 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
2042 rtm->rtm_type = RTN_MULTICAST;
2043 return 1;
2045 rtattr_failure:
2046 nlmsg_trim(skb, b);
2047 return -EMSGSIZE;
2050 int ip6mr_get_route(struct net *net,
2051 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2053 int err;
2054 struct mr6_table *mrt;
2055 struct mfc6_cache *cache;
2056 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2058 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2059 if (mrt == NULL)
2060 return -ENOENT;
2062 read_lock(&mrt_lock);
2063 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2065 if (!cache) {
2066 struct sk_buff *skb2;
2067 struct ipv6hdr *iph;
2068 struct net_device *dev;
2069 int vif;
2071 if (nowait) {
2072 read_unlock(&mrt_lock);
2073 return -EAGAIN;
2076 dev = skb->dev;
2077 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2078 read_unlock(&mrt_lock);
2079 return -ENODEV;
2082 /* really correct? */
2083 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2084 if (!skb2) {
2085 read_unlock(&mrt_lock);
2086 return -ENOMEM;
2089 skb_reset_transport_header(skb2);
2091 skb_put(skb2, sizeof(struct ipv6hdr));
2092 skb_reset_network_header(skb2);
2094 iph = ipv6_hdr(skb2);
2095 iph->version = 0;
2096 iph->priority = 0;
2097 iph->flow_lbl[0] = 0;
2098 iph->flow_lbl[1] = 0;
2099 iph->flow_lbl[2] = 0;
2100 iph->payload_len = 0;
2101 iph->nexthdr = IPPROTO_NONE;
2102 iph->hop_limit = 0;
2103 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
2104 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
2106 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2107 read_unlock(&mrt_lock);
2109 return err;
2112 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2113 cache->mfc_flags |= MFC_NOTIFY;
2115 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2116 read_unlock(&mrt_lock);
2117 return err;
2120 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2121 u32 pid, u32 seq, struct mfc6_cache *c)
2123 struct nlmsghdr *nlh;
2124 struct rtmsg *rtm;
2126 nlh = nlmsg_put(skb, pid, seq, RTM_NEWROUTE, sizeof(*rtm), NLM_F_MULTI);
2127 if (nlh == NULL)
2128 return -EMSGSIZE;
2130 rtm = nlmsg_data(nlh);
2131 rtm->rtm_family = RTNL_FAMILY_IPMR;
2132 rtm->rtm_dst_len = 128;
2133 rtm->rtm_src_len = 128;
2134 rtm->rtm_tos = 0;
2135 rtm->rtm_table = mrt->id;
2136 NLA_PUT_U32(skb, RTA_TABLE, mrt->id);
2137 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2138 rtm->rtm_protocol = RTPROT_UNSPEC;
2139 rtm->rtm_flags = 0;
2141 NLA_PUT(skb, RTA_SRC, 16, &c->mf6c_origin);
2142 NLA_PUT(skb, RTA_DST, 16, &c->mf6c_mcastgrp);
2144 if (__ip6mr_fill_mroute(mrt, skb, c, rtm) < 0)
2145 goto nla_put_failure;
2147 return nlmsg_end(skb, nlh);
2149 nla_put_failure:
2150 nlmsg_cancel(skb, nlh);
2151 return -EMSGSIZE;
2154 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2156 struct net *net = sock_net(skb->sk);
2157 struct mr6_table *mrt;
2158 struct mfc6_cache *mfc;
2159 unsigned int t = 0, s_t;
2160 unsigned int h = 0, s_h;
2161 unsigned int e = 0, s_e;
2163 s_t = cb->args[0];
2164 s_h = cb->args[1];
2165 s_e = cb->args[2];
2167 read_lock(&mrt_lock);
2168 ip6mr_for_each_table(mrt, net) {
2169 if (t < s_t)
2170 goto next_table;
2171 if (t > s_t)
2172 s_h = 0;
2173 for (h = s_h; h < MFC6_LINES; h++) {
2174 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2175 if (e < s_e)
2176 goto next_entry;
2177 if (ip6mr_fill_mroute(mrt, skb,
2178 NETLINK_CB(cb->skb).pid,
2179 cb->nlh->nlmsg_seq,
2180 mfc) < 0)
2181 goto done;
2182 next_entry:
2183 e++;
2185 e = s_e = 0;
2187 s_h = 0;
2188 next_table:
2189 t++;
2191 done:
2192 read_unlock(&mrt_lock);
2194 cb->args[2] = e;
2195 cb->args[1] = h;
2196 cb->args[0] = t;
2198 return skb->len;