dm thin metadata: fix __udivdi3 undefined on 32-bit
[linux/fpc-iii.git] / net / ipv6 / ip6mr.c
blob9b92960f024d7fc6d41822b7c9241a74421038b0
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/uaccess.h>
20 #include <linux/types.h>
21 #include <linux/sched.h>
22 #include <linux/errno.h>
23 #include <linux/timer.h>
24 #include <linux/mm.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>
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 <linux/export.h>
54 #include <net/ip6_checksum.h>
55 #include <linux/netconf.h>
57 struct mr6_table {
58 struct list_head list;
59 possible_net_t net;
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 bool mroute_do_assert;
69 bool 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 void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
112 struct sk_buff *skb, struct mfc6_cache *cache);
113 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
114 mifi_t mifi, int assert);
115 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
116 struct mfc6_cache *c, struct rtmsg *rtm);
117 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
118 int cmd);
119 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
120 struct netlink_callback *cb);
121 static void mroute_clean_tables(struct mr6_table *mrt, bool all);
122 static void ipmr_expire_process(unsigned long arg);
124 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
125 #define ip6mr_for_each_table(mrt, net) \
126 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
128 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
130 struct mr6_table *mrt;
132 ip6mr_for_each_table(mrt, net) {
133 if (mrt->id == id)
134 return mrt;
136 return NULL;
139 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
140 struct mr6_table **mrt)
142 int err;
143 struct ip6mr_result res;
144 struct fib_lookup_arg arg = {
145 .result = &res,
146 .flags = FIB_LOOKUP_NOREF,
149 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
150 flowi6_to_flowi(flp6), 0, &arg);
151 if (err < 0)
152 return err;
153 *mrt = res.mrt;
154 return 0;
157 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
158 int flags, struct fib_lookup_arg *arg)
160 struct ip6mr_result *res = arg->result;
161 struct mr6_table *mrt;
163 switch (rule->action) {
164 case FR_ACT_TO_TBL:
165 break;
166 case FR_ACT_UNREACHABLE:
167 return -ENETUNREACH;
168 case FR_ACT_PROHIBIT:
169 return -EACCES;
170 case FR_ACT_BLACKHOLE:
171 default:
172 return -EINVAL;
175 mrt = ip6mr_get_table(rule->fr_net, rule->table);
176 if (!mrt)
177 return -EAGAIN;
178 res->mrt = mrt;
179 return 0;
182 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
184 return 1;
187 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
188 FRA_GENERIC_POLICY,
191 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
192 struct fib_rule_hdr *frh, struct nlattr **tb)
194 return 0;
197 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
198 struct nlattr **tb)
200 return 1;
203 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
204 struct fib_rule_hdr *frh)
206 frh->dst_len = 0;
207 frh->src_len = 0;
208 frh->tos = 0;
209 return 0;
212 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
213 .family = RTNL_FAMILY_IP6MR,
214 .rule_size = sizeof(struct ip6mr_rule),
215 .addr_size = sizeof(struct in6_addr),
216 .action = ip6mr_rule_action,
217 .match = ip6mr_rule_match,
218 .configure = ip6mr_rule_configure,
219 .compare = ip6mr_rule_compare,
220 .fill = ip6mr_rule_fill,
221 .nlgroup = RTNLGRP_IPV6_RULE,
222 .policy = ip6mr_rule_policy,
223 .owner = THIS_MODULE,
226 static int __net_init ip6mr_rules_init(struct net *net)
228 struct fib_rules_ops *ops;
229 struct mr6_table *mrt;
230 int err;
232 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
233 if (IS_ERR(ops))
234 return PTR_ERR(ops);
236 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
238 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
239 if (!mrt) {
240 err = -ENOMEM;
241 goto err1;
244 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
245 if (err < 0)
246 goto err2;
248 net->ipv6.mr6_rules_ops = ops;
249 return 0;
251 err2:
252 ip6mr_free_table(mrt);
253 err1:
254 fib_rules_unregister(ops);
255 return err;
258 static void __net_exit ip6mr_rules_exit(struct net *net)
260 struct mr6_table *mrt, *next;
262 rtnl_lock();
263 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
264 list_del(&mrt->list);
265 ip6mr_free_table(mrt);
267 fib_rules_unregister(net->ipv6.mr6_rules_ops);
268 rtnl_unlock();
270 #else
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;
283 return 0;
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)
294 rtnl_lock();
295 ip6mr_free_table(net->ipv6.mrt6);
296 net->ipv6.mrt6 = NULL;
297 rtnl_unlock();
299 #endif
301 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
303 struct mr6_table *mrt;
304 unsigned int i;
306 mrt = ip6mr_get_table(net, id);
307 if (mrt)
308 return mrt;
310 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
311 if (!mrt)
312 return NULL;
313 mrt->id = id;
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,
323 (unsigned long)mrt);
325 #ifdef CONFIG_IPV6_PIMSM_V2
326 mrt->mroute_reg_vif_num = -1;
327 #endif
328 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
329 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
330 #endif
331 return mrt;
334 static void ip6mr_free_table(struct mr6_table *mrt)
336 del_timer_sync(&mrt->ipmr_expire_timer);
337 mroute_clean_tables(mrt, true);
338 kfree(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;
347 int ct;
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)
361 if (pos-- == 0)
362 return mfc;
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)
369 if (pos-- == 0)
370 return mfc;
371 spin_unlock_bh(&mfc_unres_lock);
373 it->cache = NULL;
374 return NULL;
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;
384 int ct;
387 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
388 struct ipmr_vif_iter *iter,
389 loff_t pos)
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))
395 continue;
396 if (pos-- == 0)
397 return &mrt->vif6_table[iter->ct];
399 return NULL;
402 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
403 __acquires(mrt_lock)
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);
410 if (!mrt)
411 return ERR_PTR(-ENOENT);
413 iter->mrt = mrt;
415 read_lock(&mrt_lock);
416 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
417 : SEQ_START_TOKEN;
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;
426 ++*pos;
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))
432 continue;
433 return &mrt->vif6_table[iter->ct];
435 return NULL;
438 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
439 __releases(mrt_lock)
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) {
450 seq_puts(seq,
451 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
452 } else {
453 const struct mif_device *vif = v;
454 const char *name = vif->dev ? vif->dev->name : "none";
456 seq_printf(seq,
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,
461 vif->flags);
463 return 0;
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,
482 .read = seq_read,
483 .llseek = seq_lseek,
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);
494 if (!mrt)
495 return ERR_PTR(-ENOENT);
497 it->mrt = mrt;
498 it->cache = NULL;
499 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
500 : SEQ_START_TOKEN;
503 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
505 struct mfc6_cache *mfc = v;
506 struct ipmr_mfc_iter *it = seq->private;
507 struct net *net = seq_file_net(seq);
508 struct mr6_table *mrt = it->mrt;
510 ++*pos;
512 if (v == SEQ_START_TOKEN)
513 return ipmr_mfc_seq_idx(net, seq->private, 0);
515 if (mfc->list.next != it->cache)
516 return list_entry(mfc->list.next, struct mfc6_cache, list);
518 if (it->cache == &mrt->mfc6_unres_queue)
519 goto end_of_list;
521 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
523 while (++it->ct < MFC6_LINES) {
524 it->cache = &mrt->mfc6_cache_array[it->ct];
525 if (list_empty(it->cache))
526 continue;
527 return list_first_entry(it->cache, struct mfc6_cache, list);
530 /* exhausted cache_array, show unresolved */
531 read_unlock(&mrt_lock);
532 it->cache = &mrt->mfc6_unres_queue;
533 it->ct = 0;
535 spin_lock_bh(&mfc_unres_lock);
536 if (!list_empty(it->cache))
537 return list_first_entry(it->cache, struct mfc6_cache, list);
539 end_of_list:
540 spin_unlock_bh(&mfc_unres_lock);
541 it->cache = NULL;
543 return NULL;
546 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
548 struct ipmr_mfc_iter *it = seq->private;
549 struct mr6_table *mrt = it->mrt;
551 if (it->cache == &mrt->mfc6_unres_queue)
552 spin_unlock_bh(&mfc_unres_lock);
553 else if (it->cache == &mrt->mfc6_cache_array[it->ct])
554 read_unlock(&mrt_lock);
557 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
559 int n;
561 if (v == SEQ_START_TOKEN) {
562 seq_puts(seq,
563 "Group "
564 "Origin "
565 "Iif Pkts Bytes Wrong Oifs\n");
566 } else {
567 const struct mfc6_cache *mfc = v;
568 const struct ipmr_mfc_iter *it = seq->private;
569 struct mr6_table *mrt = it->mrt;
571 seq_printf(seq, "%pI6 %pI6 %-3hd",
572 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
573 mfc->mf6c_parent);
575 if (it->cache != &mrt->mfc6_unres_queue) {
576 seq_printf(seq, " %8lu %8lu %8lu",
577 mfc->mfc_un.res.pkt,
578 mfc->mfc_un.res.bytes,
579 mfc->mfc_un.res.wrong_if);
580 for (n = mfc->mfc_un.res.minvif;
581 n < mfc->mfc_un.res.maxvif; n++) {
582 if (MIF_EXISTS(mrt, n) &&
583 mfc->mfc_un.res.ttls[n] < 255)
584 seq_printf(seq,
585 " %2d:%-3d",
586 n, mfc->mfc_un.res.ttls[n]);
588 } else {
589 /* unresolved mfc_caches don't contain
590 * pkt, bytes and wrong_if values
592 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
594 seq_putc(seq, '\n');
596 return 0;
599 static const struct seq_operations ipmr_mfc_seq_ops = {
600 .start = ipmr_mfc_seq_start,
601 .next = ipmr_mfc_seq_next,
602 .stop = ipmr_mfc_seq_stop,
603 .show = ipmr_mfc_seq_show,
606 static int ipmr_mfc_open(struct inode *inode, struct file *file)
608 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
609 sizeof(struct ipmr_mfc_iter));
612 static const struct file_operations ip6mr_mfc_fops = {
613 .owner = THIS_MODULE,
614 .open = ipmr_mfc_open,
615 .read = seq_read,
616 .llseek = seq_lseek,
617 .release = seq_release_net,
619 #endif
621 #ifdef CONFIG_IPV6_PIMSM_V2
623 static int pim6_rcv(struct sk_buff *skb)
625 struct pimreghdr *pim;
626 struct ipv6hdr *encap;
627 struct net_device *reg_dev = NULL;
628 struct net *net = dev_net(skb->dev);
629 struct mr6_table *mrt;
630 struct flowi6 fl6 = {
631 .flowi6_iif = skb->dev->ifindex,
632 .flowi6_mark = skb->mark,
634 int reg_vif_num;
636 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
637 goto drop;
639 pim = (struct pimreghdr *)skb_transport_header(skb);
640 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
641 (pim->flags & PIM_NULL_REGISTER) ||
642 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
643 sizeof(*pim), IPPROTO_PIM,
644 csum_partial((void *)pim, sizeof(*pim), 0)) &&
645 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
646 goto drop;
648 /* check if the inner packet is destined to mcast group */
649 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
650 sizeof(*pim));
652 if (!ipv6_addr_is_multicast(&encap->daddr) ||
653 encap->payload_len == 0 ||
654 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
655 goto drop;
657 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
658 goto drop;
659 reg_vif_num = mrt->mroute_reg_vif_num;
661 read_lock(&mrt_lock);
662 if (reg_vif_num >= 0)
663 reg_dev = mrt->vif6_table[reg_vif_num].dev;
664 if (reg_dev)
665 dev_hold(reg_dev);
666 read_unlock(&mrt_lock);
668 if (!reg_dev)
669 goto drop;
671 skb->mac_header = skb->network_header;
672 skb_pull(skb, (u8 *)encap - skb->data);
673 skb_reset_network_header(skb);
674 skb->protocol = htons(ETH_P_IPV6);
675 skb->ip_summed = CHECKSUM_NONE;
677 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
679 netif_rx(skb);
681 dev_put(reg_dev);
682 return 0;
683 drop:
684 kfree_skb(skb);
685 return 0;
688 static const struct inet6_protocol pim6_protocol = {
689 .handler = pim6_rcv,
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 ? : LOOPBACK_IFINDEX,
702 .flowi6_mark = skb->mark,
704 int err;
706 err = ip6mr_fib_lookup(net, &fl6, &mrt);
707 if (err < 0) {
708 kfree_skb(skb);
709 return err;
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);
717 kfree_skb(skb);
718 return NETDEV_TX_OK;
721 static int reg_vif_get_iflink(const struct net_device *dev)
723 return 0;
726 static const struct net_device_ops reg_vif_netdev_ops = {
727 .ndo_start_xmit = reg_vif_xmit,
728 .ndo_get_iflink = reg_vif_get_iflink,
731 static void reg_vif_setup(struct net_device *dev)
733 dev->type = ARPHRD_PIMREG;
734 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
735 dev->flags = IFF_NOARP;
736 dev->netdev_ops = &reg_vif_netdev_ops;
737 dev->destructor = free_netdev;
738 dev->features |= NETIF_F_NETNS_LOCAL;
741 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
743 struct net_device *dev;
744 char name[IFNAMSIZ];
746 if (mrt->id == RT6_TABLE_DFLT)
747 sprintf(name, "pim6reg");
748 else
749 sprintf(name, "pim6reg%u", mrt->id);
751 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
752 if (!dev)
753 return NULL;
755 dev_net_set(dev, net);
757 if (register_netdevice(dev)) {
758 free_netdev(dev);
759 return NULL;
762 if (dev_open(dev))
763 goto failure;
765 dev_hold(dev);
766 return dev;
768 failure:
769 unregister_netdevice(dev);
770 return NULL;
772 #endif
775 * Delete a VIF entry
778 static int mif6_delete(struct mr6_table *mrt, int vifi, int notify,
779 struct list_head *head)
781 struct mif_device *v;
782 struct net_device *dev;
783 struct inet6_dev *in6_dev;
785 if (vifi < 0 || vifi >= mrt->maxvif)
786 return -EADDRNOTAVAIL;
788 v = &mrt->vif6_table[vifi];
790 write_lock_bh(&mrt_lock);
791 dev = v->dev;
792 v->dev = NULL;
794 if (!dev) {
795 write_unlock_bh(&mrt_lock);
796 return -EADDRNOTAVAIL;
799 #ifdef CONFIG_IPV6_PIMSM_V2
800 if (vifi == mrt->mroute_reg_vif_num)
801 mrt->mroute_reg_vif_num = -1;
802 #endif
804 if (vifi + 1 == mrt->maxvif) {
805 int tmp;
806 for (tmp = vifi - 1; tmp >= 0; tmp--) {
807 if (MIF_EXISTS(mrt, tmp))
808 break;
810 mrt->maxvif = tmp + 1;
813 write_unlock_bh(&mrt_lock);
815 dev_set_allmulti(dev, -1);
817 in6_dev = __in6_dev_get(dev);
818 if (in6_dev) {
819 in6_dev->cnf.mc_forwarding--;
820 inet6_netconf_notify_devconf(dev_net(dev),
821 NETCONFA_MC_FORWARDING,
822 dev->ifindex, &in6_dev->cnf);
825 if ((v->flags & MIFF_REGISTER) && !notify)
826 unregister_netdevice_queue(dev, head);
828 dev_put(dev);
829 return 0;
832 static inline void ip6mr_cache_free(struct mfc6_cache *c)
834 kmem_cache_free(mrt_cachep, c);
837 /* Destroy an unresolved cache entry, killing queued skbs
838 and reporting error to netlink readers.
841 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
843 struct net *net = read_pnet(&mrt->net);
844 struct sk_buff *skb;
846 atomic_dec(&mrt->cache_resolve_queue_len);
848 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
849 if (ipv6_hdr(skb)->version == 0) {
850 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
851 nlh->nlmsg_type = NLMSG_ERROR;
852 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
853 skb_trim(skb, nlh->nlmsg_len);
854 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
855 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
856 } else
857 kfree_skb(skb);
860 ip6mr_cache_free(c);
864 /* Timer process for all the unresolved queue. */
866 static void ipmr_do_expire_process(struct mr6_table *mrt)
868 unsigned long now = jiffies;
869 unsigned long expires = 10 * HZ;
870 struct mfc6_cache *c, *next;
872 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
873 if (time_after(c->mfc_un.unres.expires, now)) {
874 /* not yet... */
875 unsigned long interval = c->mfc_un.unres.expires - now;
876 if (interval < expires)
877 expires = interval;
878 continue;
881 list_del(&c->list);
882 mr6_netlink_event(mrt, c, RTM_DELROUTE);
883 ip6mr_destroy_unres(mrt, c);
886 if (!list_empty(&mrt->mfc6_unres_queue))
887 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
890 static void ipmr_expire_process(unsigned long arg)
892 struct mr6_table *mrt = (struct mr6_table *)arg;
894 if (!spin_trylock(&mfc_unres_lock)) {
895 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
896 return;
899 if (!list_empty(&mrt->mfc6_unres_queue))
900 ipmr_do_expire_process(mrt);
902 spin_unlock(&mfc_unres_lock);
905 /* Fill oifs list. It is called under write locked mrt_lock. */
907 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
908 unsigned char *ttls)
910 int vifi;
912 cache->mfc_un.res.minvif = MAXMIFS;
913 cache->mfc_un.res.maxvif = 0;
914 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
916 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
917 if (MIF_EXISTS(mrt, vifi) &&
918 ttls[vifi] && ttls[vifi] < 255) {
919 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
920 if (cache->mfc_un.res.minvif > vifi)
921 cache->mfc_un.res.minvif = vifi;
922 if (cache->mfc_un.res.maxvif <= vifi)
923 cache->mfc_un.res.maxvif = vifi + 1;
928 static int mif6_add(struct net *net, struct mr6_table *mrt,
929 struct mif6ctl *vifc, int mrtsock)
931 int vifi = vifc->mif6c_mifi;
932 struct mif_device *v = &mrt->vif6_table[vifi];
933 struct net_device *dev;
934 struct inet6_dev *in6_dev;
935 int err;
937 /* Is vif busy ? */
938 if (MIF_EXISTS(mrt, vifi))
939 return -EADDRINUSE;
941 switch (vifc->mif6c_flags) {
942 #ifdef CONFIG_IPV6_PIMSM_V2
943 case MIFF_REGISTER:
945 * Special Purpose VIF in PIM
946 * All the packets will be sent to the daemon
948 if (mrt->mroute_reg_vif_num >= 0)
949 return -EADDRINUSE;
950 dev = ip6mr_reg_vif(net, mrt);
951 if (!dev)
952 return -ENOBUFS;
953 err = dev_set_allmulti(dev, 1);
954 if (err) {
955 unregister_netdevice(dev);
956 dev_put(dev);
957 return err;
959 break;
960 #endif
961 case 0:
962 dev = dev_get_by_index(net, vifc->mif6c_pifi);
963 if (!dev)
964 return -EADDRNOTAVAIL;
965 err = dev_set_allmulti(dev, 1);
966 if (err) {
967 dev_put(dev);
968 return err;
970 break;
971 default:
972 return -EINVAL;
975 in6_dev = __in6_dev_get(dev);
976 if (in6_dev) {
977 in6_dev->cnf.mc_forwarding++;
978 inet6_netconf_notify_devconf(dev_net(dev),
979 NETCONFA_MC_FORWARDING,
980 dev->ifindex, &in6_dev->cnf);
984 * Fill in the VIF structures
986 v->rate_limit = vifc->vifc_rate_limit;
987 v->flags = vifc->mif6c_flags;
988 if (!mrtsock)
989 v->flags |= VIFF_STATIC;
990 v->threshold = vifc->vifc_threshold;
991 v->bytes_in = 0;
992 v->bytes_out = 0;
993 v->pkt_in = 0;
994 v->pkt_out = 0;
995 v->link = dev->ifindex;
996 if (v->flags & MIFF_REGISTER)
997 v->link = dev_get_iflink(dev);
999 /* And finish update writing critical data */
1000 write_lock_bh(&mrt_lock);
1001 v->dev = dev;
1002 #ifdef CONFIG_IPV6_PIMSM_V2
1003 if (v->flags & MIFF_REGISTER)
1004 mrt->mroute_reg_vif_num = vifi;
1005 #endif
1006 if (vifi + 1 > mrt->maxvif)
1007 mrt->maxvif = vifi + 1;
1008 write_unlock_bh(&mrt_lock);
1009 return 0;
1012 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1013 const struct in6_addr *origin,
1014 const struct in6_addr *mcastgrp)
1016 int line = MFC6_HASH(mcastgrp, origin);
1017 struct mfc6_cache *c;
1019 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1020 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1021 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1022 return c;
1024 return NULL;
1027 /* Look for a (*,*,oif) entry */
1028 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1029 mifi_t mifi)
1031 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1032 struct mfc6_cache *c;
1034 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1035 if (ipv6_addr_any(&c->mf6c_origin) &&
1036 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1037 (c->mfc_un.res.ttls[mifi] < 255))
1038 return c;
1040 return NULL;
1043 /* Look for a (*,G) entry */
1044 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1045 struct in6_addr *mcastgrp,
1046 mifi_t mifi)
1048 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1049 struct mfc6_cache *c, *proxy;
1051 if (ipv6_addr_any(mcastgrp))
1052 goto skip;
1054 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1055 if (ipv6_addr_any(&c->mf6c_origin) &&
1056 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1057 if (c->mfc_un.res.ttls[mifi] < 255)
1058 return c;
1060 /* It's ok if the mifi is part of the static tree */
1061 proxy = ip6mr_cache_find_any_parent(mrt,
1062 c->mf6c_parent);
1063 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1064 return c;
1067 skip:
1068 return ip6mr_cache_find_any_parent(mrt, mifi);
1072 * Allocate a multicast cache entry
1074 static struct mfc6_cache *ip6mr_cache_alloc(void)
1076 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1077 if (!c)
1078 return NULL;
1079 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1080 c->mfc_un.res.minvif = MAXMIFS;
1081 return c;
1084 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1086 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1087 if (!c)
1088 return NULL;
1089 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1090 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1091 return c;
1095 * A cache entry has gone into a resolved state from queued
1098 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1099 struct mfc6_cache *uc, struct mfc6_cache *c)
1101 struct sk_buff *skb;
1104 * Play the pending entries through our router
1107 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1108 if (ipv6_hdr(skb)->version == 0) {
1109 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1111 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1112 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1113 } else {
1114 nlh->nlmsg_type = NLMSG_ERROR;
1115 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1116 skb_trim(skb, nlh->nlmsg_len);
1117 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1119 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1120 } else
1121 ip6_mr_forward(net, mrt, skb, c);
1126 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1127 * expects the following bizarre scheme.
1129 * Called under mrt_lock.
1132 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1133 mifi_t mifi, int assert)
1135 struct sk_buff *skb;
1136 struct mrt6msg *msg;
1137 int ret;
1139 #ifdef CONFIG_IPV6_PIMSM_V2
1140 if (assert == MRT6MSG_WHOLEPKT)
1141 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1142 +sizeof(*msg));
1143 else
1144 #endif
1145 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1147 if (!skb)
1148 return -ENOBUFS;
1150 /* I suppose that internal messages
1151 * do not require checksums */
1153 skb->ip_summed = CHECKSUM_UNNECESSARY;
1155 #ifdef CONFIG_IPV6_PIMSM_V2
1156 if (assert == MRT6MSG_WHOLEPKT) {
1157 /* Ugly, but we have no choice with this interface.
1158 Duplicate old header, fix length etc.
1159 And all this only to mangle msg->im6_msgtype and
1160 to set msg->im6_mbz to "mbz" :-)
1162 skb_push(skb, -skb_network_offset(pkt));
1164 skb_push(skb, sizeof(*msg));
1165 skb_reset_transport_header(skb);
1166 msg = (struct mrt6msg *)skb_transport_header(skb);
1167 msg->im6_mbz = 0;
1168 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1169 msg->im6_mif = mrt->mroute_reg_vif_num;
1170 msg->im6_pad = 0;
1171 msg->im6_src = ipv6_hdr(pkt)->saddr;
1172 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1174 skb->ip_summed = CHECKSUM_UNNECESSARY;
1175 } else
1176 #endif
1179 * Copy the IP header
1182 skb_put(skb, sizeof(struct ipv6hdr));
1183 skb_reset_network_header(skb);
1184 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1187 * Add our header
1189 skb_put(skb, sizeof(*msg));
1190 skb_reset_transport_header(skb);
1191 msg = (struct mrt6msg *)skb_transport_header(skb);
1193 msg->im6_mbz = 0;
1194 msg->im6_msgtype = assert;
1195 msg->im6_mif = mifi;
1196 msg->im6_pad = 0;
1197 msg->im6_src = ipv6_hdr(pkt)->saddr;
1198 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1200 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1201 skb->ip_summed = CHECKSUM_UNNECESSARY;
1204 if (!mrt->mroute6_sk) {
1205 kfree_skb(skb);
1206 return -EINVAL;
1210 * Deliver to user space multicast routing algorithms
1212 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1213 if (ret < 0) {
1214 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1215 kfree_skb(skb);
1218 return ret;
1222 * Queue a packet for resolution. It gets locked cache entry!
1225 static int
1226 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1228 bool found = false;
1229 int err;
1230 struct mfc6_cache *c;
1232 spin_lock_bh(&mfc_unres_lock);
1233 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1234 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1235 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1236 found = true;
1237 break;
1241 if (!found) {
1243 * Create a new entry if allowable
1246 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1247 (c = ip6mr_cache_alloc_unres()) == NULL) {
1248 spin_unlock_bh(&mfc_unres_lock);
1250 kfree_skb(skb);
1251 return -ENOBUFS;
1255 * Fill in the new cache entry
1257 c->mf6c_parent = -1;
1258 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1259 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1262 * Reflect first query at pim6sd
1264 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1265 if (err < 0) {
1266 /* If the report failed throw the cache entry
1267 out - Brad Parker
1269 spin_unlock_bh(&mfc_unres_lock);
1271 ip6mr_cache_free(c);
1272 kfree_skb(skb);
1273 return err;
1276 atomic_inc(&mrt->cache_resolve_queue_len);
1277 list_add(&c->list, &mrt->mfc6_unres_queue);
1278 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1280 ipmr_do_expire_process(mrt);
1284 * See if we can append the packet
1286 if (c->mfc_un.unres.unresolved.qlen > 3) {
1287 kfree_skb(skb);
1288 err = -ENOBUFS;
1289 } else {
1290 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1291 err = 0;
1294 spin_unlock_bh(&mfc_unres_lock);
1295 return err;
1299 * MFC6 cache manipulation by user space
1302 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1303 int parent)
1305 int line;
1306 struct mfc6_cache *c, *next;
1308 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1310 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1311 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1312 ipv6_addr_equal(&c->mf6c_mcastgrp,
1313 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1314 (parent == -1 || parent == c->mf6c_parent)) {
1315 write_lock_bh(&mrt_lock);
1316 list_del(&c->list);
1317 write_unlock_bh(&mrt_lock);
1319 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1320 ip6mr_cache_free(c);
1321 return 0;
1324 return -ENOENT;
1327 static int ip6mr_device_event(struct notifier_block *this,
1328 unsigned long event, void *ptr)
1330 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1331 struct net *net = dev_net(dev);
1332 struct mr6_table *mrt;
1333 struct mif_device *v;
1334 int ct;
1336 if (event != NETDEV_UNREGISTER)
1337 return NOTIFY_DONE;
1339 ip6mr_for_each_table(mrt, net) {
1340 v = &mrt->vif6_table[0];
1341 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1342 if (v->dev == dev)
1343 mif6_delete(mrt, ct, 1, NULL);
1347 return NOTIFY_DONE;
1350 static struct notifier_block ip6_mr_notifier = {
1351 .notifier_call = ip6mr_device_event
1355 * Setup for IP multicast routing
1358 static int __net_init ip6mr_net_init(struct net *net)
1360 int err;
1362 err = ip6mr_rules_init(net);
1363 if (err < 0)
1364 goto fail;
1366 #ifdef CONFIG_PROC_FS
1367 err = -ENOMEM;
1368 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1369 goto proc_vif_fail;
1370 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1371 goto proc_cache_fail;
1372 #endif
1374 return 0;
1376 #ifdef CONFIG_PROC_FS
1377 proc_cache_fail:
1378 remove_proc_entry("ip6_mr_vif", net->proc_net);
1379 proc_vif_fail:
1380 ip6mr_rules_exit(net);
1381 #endif
1382 fail:
1383 return err;
1386 static void __net_exit ip6mr_net_exit(struct net *net)
1388 #ifdef CONFIG_PROC_FS
1389 remove_proc_entry("ip6_mr_cache", net->proc_net);
1390 remove_proc_entry("ip6_mr_vif", net->proc_net);
1391 #endif
1392 ip6mr_rules_exit(net);
1395 static struct pernet_operations ip6mr_net_ops = {
1396 .init = ip6mr_net_init,
1397 .exit = ip6mr_net_exit,
1400 int __init ip6_mr_init(void)
1402 int err;
1404 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1405 sizeof(struct mfc6_cache),
1406 0, SLAB_HWCACHE_ALIGN,
1407 NULL);
1408 if (!mrt_cachep)
1409 return -ENOMEM;
1411 err = register_pernet_subsys(&ip6mr_net_ops);
1412 if (err)
1413 goto reg_pernet_fail;
1415 err = register_netdevice_notifier(&ip6_mr_notifier);
1416 if (err)
1417 goto reg_notif_fail;
1418 #ifdef CONFIG_IPV6_PIMSM_V2
1419 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1420 pr_err("%s: can't add PIM protocol\n", __func__);
1421 err = -EAGAIN;
1422 goto add_proto_fail;
1424 #endif
1425 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1426 ip6mr_rtm_dumproute, NULL);
1427 return 0;
1428 #ifdef CONFIG_IPV6_PIMSM_V2
1429 add_proto_fail:
1430 unregister_netdevice_notifier(&ip6_mr_notifier);
1431 #endif
1432 reg_notif_fail:
1433 unregister_pernet_subsys(&ip6mr_net_ops);
1434 reg_pernet_fail:
1435 kmem_cache_destroy(mrt_cachep);
1436 return err;
1439 void ip6_mr_cleanup(void)
1441 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1442 #ifdef CONFIG_IPV6_PIMSM_V2
1443 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1444 #endif
1445 unregister_netdevice_notifier(&ip6_mr_notifier);
1446 unregister_pernet_subsys(&ip6mr_net_ops);
1447 kmem_cache_destroy(mrt_cachep);
1450 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1451 struct mf6cctl *mfc, int mrtsock, int parent)
1453 bool found = false;
1454 int line;
1455 struct mfc6_cache *uc, *c;
1456 unsigned char ttls[MAXMIFS];
1457 int i;
1459 if (mfc->mf6cc_parent >= MAXMIFS)
1460 return -ENFILE;
1462 memset(ttls, 255, MAXMIFS);
1463 for (i = 0; i < MAXMIFS; i++) {
1464 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1465 ttls[i] = 1;
1469 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1471 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1472 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1473 ipv6_addr_equal(&c->mf6c_mcastgrp,
1474 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1475 (parent == -1 || parent == mfc->mf6cc_parent)) {
1476 found = true;
1477 break;
1481 if (found) {
1482 write_lock_bh(&mrt_lock);
1483 c->mf6c_parent = mfc->mf6cc_parent;
1484 ip6mr_update_thresholds(mrt, c, ttls);
1485 if (!mrtsock)
1486 c->mfc_flags |= MFC_STATIC;
1487 write_unlock_bh(&mrt_lock);
1488 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1489 return 0;
1492 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1493 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1494 return -EINVAL;
1496 c = ip6mr_cache_alloc();
1497 if (!c)
1498 return -ENOMEM;
1500 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1501 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1502 c->mf6c_parent = mfc->mf6cc_parent;
1503 ip6mr_update_thresholds(mrt, c, ttls);
1504 if (!mrtsock)
1505 c->mfc_flags |= MFC_STATIC;
1507 write_lock_bh(&mrt_lock);
1508 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1509 write_unlock_bh(&mrt_lock);
1512 * Check to see if we resolved a queued list. If so we
1513 * need to send on the frames and tidy up.
1515 found = false;
1516 spin_lock_bh(&mfc_unres_lock);
1517 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1518 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1519 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1520 list_del(&uc->list);
1521 atomic_dec(&mrt->cache_resolve_queue_len);
1522 found = true;
1523 break;
1526 if (list_empty(&mrt->mfc6_unres_queue))
1527 del_timer(&mrt->ipmr_expire_timer);
1528 spin_unlock_bh(&mfc_unres_lock);
1530 if (found) {
1531 ip6mr_cache_resolve(net, mrt, uc, c);
1532 ip6mr_cache_free(uc);
1534 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1535 return 0;
1539 * Close the multicast socket, and clear the vif tables etc
1542 static void mroute_clean_tables(struct mr6_table *mrt, bool all)
1544 int i;
1545 LIST_HEAD(list);
1546 struct mfc6_cache *c, *next;
1549 * Shut down all active vif entries
1551 for (i = 0; i < mrt->maxvif; i++) {
1552 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC))
1553 continue;
1554 mif6_delete(mrt, i, 0, &list);
1556 unregister_netdevice_many(&list);
1559 * Wipe the cache
1561 for (i = 0; i < MFC6_LINES; i++) {
1562 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1563 if (!all && (c->mfc_flags & MFC_STATIC))
1564 continue;
1565 write_lock_bh(&mrt_lock);
1566 list_del(&c->list);
1567 write_unlock_bh(&mrt_lock);
1569 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1570 ip6mr_cache_free(c);
1574 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1575 spin_lock_bh(&mfc_unres_lock);
1576 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1577 list_del(&c->list);
1578 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1579 ip6mr_destroy_unres(mrt, c);
1581 spin_unlock_bh(&mfc_unres_lock);
1585 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1587 int err = 0;
1588 struct net *net = sock_net(sk);
1590 rtnl_lock();
1591 write_lock_bh(&mrt_lock);
1592 if (likely(mrt->mroute6_sk == NULL)) {
1593 mrt->mroute6_sk = sk;
1594 net->ipv6.devconf_all->mc_forwarding++;
1595 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1596 NETCONFA_IFINDEX_ALL,
1597 net->ipv6.devconf_all);
1599 else
1600 err = -EADDRINUSE;
1601 write_unlock_bh(&mrt_lock);
1603 rtnl_unlock();
1605 return err;
1608 int ip6mr_sk_done(struct sock *sk)
1610 int err = -EACCES;
1611 struct net *net = sock_net(sk);
1612 struct mr6_table *mrt;
1614 rtnl_lock();
1615 ip6mr_for_each_table(mrt, net) {
1616 if (sk == mrt->mroute6_sk) {
1617 write_lock_bh(&mrt_lock);
1618 mrt->mroute6_sk = NULL;
1619 net->ipv6.devconf_all->mc_forwarding--;
1620 inet6_netconf_notify_devconf(net,
1621 NETCONFA_MC_FORWARDING,
1622 NETCONFA_IFINDEX_ALL,
1623 net->ipv6.devconf_all);
1624 write_unlock_bh(&mrt_lock);
1626 mroute_clean_tables(mrt, false);
1627 err = 0;
1628 break;
1631 rtnl_unlock();
1633 return err;
1636 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1638 struct mr6_table *mrt;
1639 struct flowi6 fl6 = {
1640 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1641 .flowi6_oif = skb->dev->ifindex,
1642 .flowi6_mark = skb->mark,
1645 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1646 return NULL;
1648 return mrt->mroute6_sk;
1652 * Socket options and virtual interface manipulation. The whole
1653 * virtual interface system is a complete heap, but unfortunately
1654 * that's how BSD mrouted happens to think. Maybe one day with a proper
1655 * MOSPF/PIM router set up we can clean this up.
1658 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1660 int ret, parent = 0;
1661 struct mif6ctl vif;
1662 struct mf6cctl mfc;
1663 mifi_t mifi;
1664 struct net *net = sock_net(sk);
1665 struct mr6_table *mrt;
1667 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1668 if (!mrt)
1669 return -ENOENT;
1671 if (optname != MRT6_INIT) {
1672 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1673 return -EACCES;
1676 switch (optname) {
1677 case MRT6_INIT:
1678 if (sk->sk_type != SOCK_RAW ||
1679 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1680 return -EOPNOTSUPP;
1681 if (optlen < sizeof(int))
1682 return -EINVAL;
1684 return ip6mr_sk_init(mrt, sk);
1686 case MRT6_DONE:
1687 return ip6mr_sk_done(sk);
1689 case MRT6_ADD_MIF:
1690 if (optlen < sizeof(vif))
1691 return -EINVAL;
1692 if (copy_from_user(&vif, optval, sizeof(vif)))
1693 return -EFAULT;
1694 if (vif.mif6c_mifi >= MAXMIFS)
1695 return -ENFILE;
1696 rtnl_lock();
1697 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1698 rtnl_unlock();
1699 return ret;
1701 case MRT6_DEL_MIF:
1702 if (optlen < sizeof(mifi_t))
1703 return -EINVAL;
1704 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1705 return -EFAULT;
1706 rtnl_lock();
1707 ret = mif6_delete(mrt, mifi, 0, NULL);
1708 rtnl_unlock();
1709 return ret;
1712 * Manipulate the forwarding caches. These live
1713 * in a sort of kernel/user symbiosis.
1715 case MRT6_ADD_MFC:
1716 case MRT6_DEL_MFC:
1717 parent = -1;
1718 case MRT6_ADD_MFC_PROXY:
1719 case MRT6_DEL_MFC_PROXY:
1720 if (optlen < sizeof(mfc))
1721 return -EINVAL;
1722 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1723 return -EFAULT;
1724 if (parent == 0)
1725 parent = mfc.mf6cc_parent;
1726 rtnl_lock();
1727 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1728 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1729 else
1730 ret = ip6mr_mfc_add(net, mrt, &mfc,
1731 sk == mrt->mroute6_sk, parent);
1732 rtnl_unlock();
1733 return ret;
1736 * Control PIM assert (to activate pim will activate assert)
1738 case MRT6_ASSERT:
1740 int v;
1742 if (optlen != sizeof(v))
1743 return -EINVAL;
1744 if (get_user(v, (int __user *)optval))
1745 return -EFAULT;
1746 mrt->mroute_do_assert = v;
1747 return 0;
1750 #ifdef CONFIG_IPV6_PIMSM_V2
1751 case MRT6_PIM:
1753 int v;
1755 if (optlen != sizeof(v))
1756 return -EINVAL;
1757 if (get_user(v, (int __user *)optval))
1758 return -EFAULT;
1759 v = !!v;
1760 rtnl_lock();
1761 ret = 0;
1762 if (v != mrt->mroute_do_pim) {
1763 mrt->mroute_do_pim = v;
1764 mrt->mroute_do_assert = v;
1766 rtnl_unlock();
1767 return ret;
1770 #endif
1771 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1772 case MRT6_TABLE:
1774 u32 v;
1776 if (optlen != sizeof(u32))
1777 return -EINVAL;
1778 if (get_user(v, (u32 __user *)optval))
1779 return -EFAULT;
1780 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1781 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1782 return -EINVAL;
1783 if (sk == mrt->mroute6_sk)
1784 return -EBUSY;
1786 rtnl_lock();
1787 ret = 0;
1788 if (!ip6mr_new_table(net, v))
1789 ret = -ENOMEM;
1790 else
1791 raw6_sk(sk)->ip6mr_table = v;
1792 rtnl_unlock();
1793 return ret;
1795 #endif
1797 * Spurious command, or MRT6_VERSION which you cannot
1798 * set.
1800 default:
1801 return -ENOPROTOOPT;
1806 * Getsock opt support for the multicast routing system.
1809 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1810 int __user *optlen)
1812 int olr;
1813 int val;
1814 struct net *net = sock_net(sk);
1815 struct mr6_table *mrt;
1817 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1818 if (!mrt)
1819 return -ENOENT;
1821 switch (optname) {
1822 case MRT6_VERSION:
1823 val = 0x0305;
1824 break;
1825 #ifdef CONFIG_IPV6_PIMSM_V2
1826 case MRT6_PIM:
1827 val = mrt->mroute_do_pim;
1828 break;
1829 #endif
1830 case MRT6_ASSERT:
1831 val = mrt->mroute_do_assert;
1832 break;
1833 default:
1834 return -ENOPROTOOPT;
1837 if (get_user(olr, optlen))
1838 return -EFAULT;
1840 olr = min_t(int, olr, sizeof(int));
1841 if (olr < 0)
1842 return -EINVAL;
1844 if (put_user(olr, optlen))
1845 return -EFAULT;
1846 if (copy_to_user(optval, &val, olr))
1847 return -EFAULT;
1848 return 0;
1852 * The IP multicast ioctl support routines.
1855 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1857 struct sioc_sg_req6 sr;
1858 struct sioc_mif_req6 vr;
1859 struct mif_device *vif;
1860 struct mfc6_cache *c;
1861 struct net *net = sock_net(sk);
1862 struct mr6_table *mrt;
1864 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1865 if (!mrt)
1866 return -ENOENT;
1868 switch (cmd) {
1869 case SIOCGETMIFCNT_IN6:
1870 if (copy_from_user(&vr, arg, sizeof(vr)))
1871 return -EFAULT;
1872 if (vr.mifi >= mrt->maxvif)
1873 return -EINVAL;
1874 read_lock(&mrt_lock);
1875 vif = &mrt->vif6_table[vr.mifi];
1876 if (MIF_EXISTS(mrt, vr.mifi)) {
1877 vr.icount = vif->pkt_in;
1878 vr.ocount = vif->pkt_out;
1879 vr.ibytes = vif->bytes_in;
1880 vr.obytes = vif->bytes_out;
1881 read_unlock(&mrt_lock);
1883 if (copy_to_user(arg, &vr, sizeof(vr)))
1884 return -EFAULT;
1885 return 0;
1887 read_unlock(&mrt_lock);
1888 return -EADDRNOTAVAIL;
1889 case SIOCGETSGCNT_IN6:
1890 if (copy_from_user(&sr, arg, sizeof(sr)))
1891 return -EFAULT;
1893 read_lock(&mrt_lock);
1894 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1895 if (c) {
1896 sr.pktcnt = c->mfc_un.res.pkt;
1897 sr.bytecnt = c->mfc_un.res.bytes;
1898 sr.wrong_if = c->mfc_un.res.wrong_if;
1899 read_unlock(&mrt_lock);
1901 if (copy_to_user(arg, &sr, sizeof(sr)))
1902 return -EFAULT;
1903 return 0;
1905 read_unlock(&mrt_lock);
1906 return -EADDRNOTAVAIL;
1907 default:
1908 return -ENOIOCTLCMD;
1912 #ifdef CONFIG_COMPAT
1913 struct compat_sioc_sg_req6 {
1914 struct sockaddr_in6 src;
1915 struct sockaddr_in6 grp;
1916 compat_ulong_t pktcnt;
1917 compat_ulong_t bytecnt;
1918 compat_ulong_t wrong_if;
1921 struct compat_sioc_mif_req6 {
1922 mifi_t mifi;
1923 compat_ulong_t icount;
1924 compat_ulong_t ocount;
1925 compat_ulong_t ibytes;
1926 compat_ulong_t obytes;
1929 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1931 struct compat_sioc_sg_req6 sr;
1932 struct compat_sioc_mif_req6 vr;
1933 struct mif_device *vif;
1934 struct mfc6_cache *c;
1935 struct net *net = sock_net(sk);
1936 struct mr6_table *mrt;
1938 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1939 if (!mrt)
1940 return -ENOENT;
1942 switch (cmd) {
1943 case SIOCGETMIFCNT_IN6:
1944 if (copy_from_user(&vr, arg, sizeof(vr)))
1945 return -EFAULT;
1946 if (vr.mifi >= mrt->maxvif)
1947 return -EINVAL;
1948 read_lock(&mrt_lock);
1949 vif = &mrt->vif6_table[vr.mifi];
1950 if (MIF_EXISTS(mrt, vr.mifi)) {
1951 vr.icount = vif->pkt_in;
1952 vr.ocount = vif->pkt_out;
1953 vr.ibytes = vif->bytes_in;
1954 vr.obytes = vif->bytes_out;
1955 read_unlock(&mrt_lock);
1957 if (copy_to_user(arg, &vr, sizeof(vr)))
1958 return -EFAULT;
1959 return 0;
1961 read_unlock(&mrt_lock);
1962 return -EADDRNOTAVAIL;
1963 case SIOCGETSGCNT_IN6:
1964 if (copy_from_user(&sr, arg, sizeof(sr)))
1965 return -EFAULT;
1967 read_lock(&mrt_lock);
1968 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1969 if (c) {
1970 sr.pktcnt = c->mfc_un.res.pkt;
1971 sr.bytecnt = c->mfc_un.res.bytes;
1972 sr.wrong_if = c->mfc_un.res.wrong_if;
1973 read_unlock(&mrt_lock);
1975 if (copy_to_user(arg, &sr, sizeof(sr)))
1976 return -EFAULT;
1977 return 0;
1979 read_unlock(&mrt_lock);
1980 return -EADDRNOTAVAIL;
1981 default:
1982 return -ENOIOCTLCMD;
1985 #endif
1987 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1989 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
1990 IPSTATS_MIB_OUTFORWDATAGRAMS);
1991 IP6_ADD_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
1992 IPSTATS_MIB_OUTOCTETS, skb->len);
1993 return dst_output(net, sk, skb);
1997 * Processing handlers for ip6mr_forward
2000 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
2001 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
2003 struct ipv6hdr *ipv6h;
2004 struct mif_device *vif = &mrt->vif6_table[vifi];
2005 struct net_device *dev;
2006 struct dst_entry *dst;
2007 struct flowi6 fl6;
2009 if (!vif->dev)
2010 goto out_free;
2012 #ifdef CONFIG_IPV6_PIMSM_V2
2013 if (vif->flags & MIFF_REGISTER) {
2014 vif->pkt_out++;
2015 vif->bytes_out += skb->len;
2016 vif->dev->stats.tx_bytes += skb->len;
2017 vif->dev->stats.tx_packets++;
2018 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2019 goto out_free;
2021 #endif
2023 ipv6h = ipv6_hdr(skb);
2025 fl6 = (struct flowi6) {
2026 .flowi6_oif = vif->link,
2027 .daddr = ipv6h->daddr,
2030 dst = ip6_route_output(net, NULL, &fl6);
2031 if (dst->error) {
2032 dst_release(dst);
2033 goto out_free;
2036 skb_dst_drop(skb);
2037 skb_dst_set(skb, dst);
2040 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2041 * not only before forwarding, but after forwarding on all output
2042 * interfaces. It is clear, if mrouter runs a multicasting
2043 * program, it should receive packets not depending to what interface
2044 * program is joined.
2045 * If we will not make it, the program will have to join on all
2046 * interfaces. On the other hand, multihoming host (or router, but
2047 * not mrouter) cannot join to more than one interface - it will
2048 * result in receiving multiple packets.
2050 dev = vif->dev;
2051 skb->dev = dev;
2052 vif->pkt_out++;
2053 vif->bytes_out += skb->len;
2055 /* We are about to write */
2056 /* XXX: extension headers? */
2057 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2058 goto out_free;
2060 ipv6h = ipv6_hdr(skb);
2061 ipv6h->hop_limit--;
2063 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2065 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2066 net, NULL, skb, skb->dev, dev,
2067 ip6mr_forward2_finish);
2069 out_free:
2070 kfree_skb(skb);
2071 return 0;
2074 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2076 int ct;
2078 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2079 if (mrt->vif6_table[ct].dev == dev)
2080 break;
2082 return ct;
2085 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2086 struct sk_buff *skb, struct mfc6_cache *cache)
2088 int psend = -1;
2089 int vif, ct;
2090 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2092 vif = cache->mf6c_parent;
2093 cache->mfc_un.res.pkt++;
2094 cache->mfc_un.res.bytes += skb->len;
2096 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2097 struct mfc6_cache *cache_proxy;
2099 /* For an (*,G) entry, we only check that the incoming
2100 * interface is part of the static tree.
2102 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2103 if (cache_proxy &&
2104 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2105 goto forward;
2109 * Wrong interface: drop packet and (maybe) send PIM assert.
2111 if (mrt->vif6_table[vif].dev != skb->dev) {
2112 cache->mfc_un.res.wrong_if++;
2114 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2115 /* pimsm uses asserts, when switching from RPT to SPT,
2116 so that we cannot check that packet arrived on an oif.
2117 It is bad, but otherwise we would need to move pretty
2118 large chunk of pimd to kernel. Ough... --ANK
2120 (mrt->mroute_do_pim ||
2121 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2122 time_after(jiffies,
2123 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2124 cache->mfc_un.res.last_assert = jiffies;
2125 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2127 goto dont_forward;
2130 forward:
2131 mrt->vif6_table[vif].pkt_in++;
2132 mrt->vif6_table[vif].bytes_in += skb->len;
2135 * Forward the frame
2137 if (ipv6_addr_any(&cache->mf6c_origin) &&
2138 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2139 if (true_vifi >= 0 &&
2140 true_vifi != cache->mf6c_parent &&
2141 ipv6_hdr(skb)->hop_limit >
2142 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2143 /* It's an (*,*) entry and the packet is not coming from
2144 * the upstream: forward the packet to the upstream
2145 * only.
2147 psend = cache->mf6c_parent;
2148 goto last_forward;
2150 goto dont_forward;
2152 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2153 /* For (*,G) entry, don't forward to the incoming interface */
2154 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2155 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2156 if (psend != -1) {
2157 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2158 if (skb2)
2159 ip6mr_forward2(net, mrt, skb2, cache, psend);
2161 psend = ct;
2164 last_forward:
2165 if (psend != -1) {
2166 ip6mr_forward2(net, mrt, skb, cache, psend);
2167 return;
2170 dont_forward:
2171 kfree_skb(skb);
2176 * Multicast packets for forwarding arrive here
2179 int ip6_mr_input(struct sk_buff *skb)
2181 struct mfc6_cache *cache;
2182 struct net *net = dev_net(skb->dev);
2183 struct mr6_table *mrt;
2184 struct flowi6 fl6 = {
2185 .flowi6_iif = skb->dev->ifindex,
2186 .flowi6_mark = skb->mark,
2188 int err;
2190 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2191 if (err < 0) {
2192 kfree_skb(skb);
2193 return err;
2196 read_lock(&mrt_lock);
2197 cache = ip6mr_cache_find(mrt,
2198 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2199 if (!cache) {
2200 int vif = ip6mr_find_vif(mrt, skb->dev);
2202 if (vif >= 0)
2203 cache = ip6mr_cache_find_any(mrt,
2204 &ipv6_hdr(skb)->daddr,
2205 vif);
2209 * No usable cache entry
2211 if (!cache) {
2212 int vif;
2214 vif = ip6mr_find_vif(mrt, skb->dev);
2215 if (vif >= 0) {
2216 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2217 read_unlock(&mrt_lock);
2219 return err;
2221 read_unlock(&mrt_lock);
2222 kfree_skb(skb);
2223 return -ENODEV;
2226 ip6_mr_forward(net, mrt, skb, cache);
2228 read_unlock(&mrt_lock);
2230 return 0;
2234 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2235 struct mfc6_cache *c, struct rtmsg *rtm)
2237 int ct;
2238 struct rtnexthop *nhp;
2239 struct nlattr *mp_attr;
2240 struct rta_mfc_stats mfcs;
2242 /* If cache is unresolved, don't try to parse IIF and OIF */
2243 if (c->mf6c_parent >= MAXMIFS)
2244 return -ENOENT;
2246 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2247 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2248 return -EMSGSIZE;
2249 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2250 if (!mp_attr)
2251 return -EMSGSIZE;
2253 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2254 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2255 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2256 if (!nhp) {
2257 nla_nest_cancel(skb, mp_attr);
2258 return -EMSGSIZE;
2261 nhp->rtnh_flags = 0;
2262 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2263 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2264 nhp->rtnh_len = sizeof(*nhp);
2268 nla_nest_end(skb, mp_attr);
2270 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2271 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2272 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2273 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2274 return -EMSGSIZE;
2276 rtm->rtm_type = RTN_MULTICAST;
2277 return 1;
2280 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2281 int nowait, u32 portid)
2283 int err;
2284 struct mr6_table *mrt;
2285 struct mfc6_cache *cache;
2286 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2288 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2289 if (!mrt)
2290 return -ENOENT;
2292 read_lock(&mrt_lock);
2293 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2294 if (!cache && skb->dev) {
2295 int vif = ip6mr_find_vif(mrt, skb->dev);
2297 if (vif >= 0)
2298 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2299 vif);
2302 if (!cache) {
2303 struct sk_buff *skb2;
2304 struct ipv6hdr *iph;
2305 struct net_device *dev;
2306 int vif;
2308 if (nowait) {
2309 read_unlock(&mrt_lock);
2310 return -EAGAIN;
2313 dev = skb->dev;
2314 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2315 read_unlock(&mrt_lock);
2316 return -ENODEV;
2319 /* really correct? */
2320 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2321 if (!skb2) {
2322 read_unlock(&mrt_lock);
2323 return -ENOMEM;
2326 NETLINK_CB(skb2).portid = portid;
2327 skb_reset_transport_header(skb2);
2329 skb_put(skb2, sizeof(struct ipv6hdr));
2330 skb_reset_network_header(skb2);
2332 iph = ipv6_hdr(skb2);
2333 iph->version = 0;
2334 iph->priority = 0;
2335 iph->flow_lbl[0] = 0;
2336 iph->flow_lbl[1] = 0;
2337 iph->flow_lbl[2] = 0;
2338 iph->payload_len = 0;
2339 iph->nexthdr = IPPROTO_NONE;
2340 iph->hop_limit = 0;
2341 iph->saddr = rt->rt6i_src.addr;
2342 iph->daddr = rt->rt6i_dst.addr;
2344 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2345 read_unlock(&mrt_lock);
2347 return err;
2350 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2351 cache->mfc_flags |= MFC_NOTIFY;
2353 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2354 read_unlock(&mrt_lock);
2355 return err;
2358 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2359 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2360 int flags)
2362 struct nlmsghdr *nlh;
2363 struct rtmsg *rtm;
2364 int err;
2366 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2367 if (!nlh)
2368 return -EMSGSIZE;
2370 rtm = nlmsg_data(nlh);
2371 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2372 rtm->rtm_dst_len = 128;
2373 rtm->rtm_src_len = 128;
2374 rtm->rtm_tos = 0;
2375 rtm->rtm_table = mrt->id;
2376 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2377 goto nla_put_failure;
2378 rtm->rtm_type = RTN_MULTICAST;
2379 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2380 if (c->mfc_flags & MFC_STATIC)
2381 rtm->rtm_protocol = RTPROT_STATIC;
2382 else
2383 rtm->rtm_protocol = RTPROT_MROUTED;
2384 rtm->rtm_flags = 0;
2386 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2387 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2388 goto nla_put_failure;
2389 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2390 /* do not break the dump if cache is unresolved */
2391 if (err < 0 && err != -ENOENT)
2392 goto nla_put_failure;
2394 nlmsg_end(skb, nlh);
2395 return 0;
2397 nla_put_failure:
2398 nlmsg_cancel(skb, nlh);
2399 return -EMSGSIZE;
2402 static int mr6_msgsize(bool unresolved, int maxvif)
2404 size_t len =
2405 NLMSG_ALIGN(sizeof(struct rtmsg))
2406 + nla_total_size(4) /* RTA_TABLE */
2407 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2408 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2411 if (!unresolved)
2412 len = len
2413 + nla_total_size(4) /* RTA_IIF */
2414 + nla_total_size(0) /* RTA_MULTIPATH */
2415 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2416 /* RTA_MFC_STATS */
2417 + nla_total_size(sizeof(struct rta_mfc_stats))
2420 return len;
2423 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2424 int cmd)
2426 struct net *net = read_pnet(&mrt->net);
2427 struct sk_buff *skb;
2428 int err = -ENOBUFS;
2430 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2431 GFP_ATOMIC);
2432 if (!skb)
2433 goto errout;
2435 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2436 if (err < 0)
2437 goto errout;
2439 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2440 return;
2442 errout:
2443 kfree_skb(skb);
2444 if (err < 0)
2445 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2448 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2450 struct net *net = sock_net(skb->sk);
2451 struct mr6_table *mrt;
2452 struct mfc6_cache *mfc;
2453 unsigned int t = 0, s_t;
2454 unsigned int h = 0, s_h;
2455 unsigned int e = 0, s_e;
2457 s_t = cb->args[0];
2458 s_h = cb->args[1];
2459 s_e = cb->args[2];
2461 read_lock(&mrt_lock);
2462 ip6mr_for_each_table(mrt, net) {
2463 if (t < s_t)
2464 goto next_table;
2465 if (t > s_t)
2466 s_h = 0;
2467 for (h = s_h; h < MFC6_LINES; h++) {
2468 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2469 if (e < s_e)
2470 goto next_entry;
2471 if (ip6mr_fill_mroute(mrt, skb,
2472 NETLINK_CB(cb->skb).portid,
2473 cb->nlh->nlmsg_seq,
2474 mfc, RTM_NEWROUTE,
2475 NLM_F_MULTI) < 0)
2476 goto done;
2477 next_entry:
2478 e++;
2480 e = s_e = 0;
2482 spin_lock_bh(&mfc_unres_lock);
2483 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2484 if (e < s_e)
2485 goto next_entry2;
2486 if (ip6mr_fill_mroute(mrt, skb,
2487 NETLINK_CB(cb->skb).portid,
2488 cb->nlh->nlmsg_seq,
2489 mfc, RTM_NEWROUTE,
2490 NLM_F_MULTI) < 0) {
2491 spin_unlock_bh(&mfc_unres_lock);
2492 goto done;
2494 next_entry2:
2495 e++;
2497 spin_unlock_bh(&mfc_unres_lock);
2498 e = s_e = 0;
2499 s_h = 0;
2500 next_table:
2501 t++;
2503 done:
2504 read_unlock(&mrt_lock);
2506 cb->args[2] = e;
2507 cb->args[1] = h;
2508 cb->args[0] = t;
2510 return skb->len;