1 /*
2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
4 *
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>
11 *
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.
16 *
17 */
18
19 #include <linux/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>
46
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>
56
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
73 };
74
75 struct ip6mr_rule {
76 struct fib_rule common;
77 };
78
79 struct ip6mr_result {
80 struct mr6_table *mrt;
81 };
82
83 /* Big lock, protecting vif table, mrt cache and mroute socket state.
84 Note that the changes are semaphored via rtnl_lock.
85 */
86
87 static DEFINE_RWLOCK(mrt_lock);
88
89 /*
90 * Multicast router control variables
91 */
92
93 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
94
95 /* Special spinlock for queue of unresolved entries */
96 static DEFINE_SPINLOCK(mfc_unres_lock);
97
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.
102
103 In this case data path is free of exclusive locks at all.
104 */
105
106 static struct kmem_cache *mrt_cachep __read_mostly;
107
108 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
109 static void ip6mr_free_table(struct mr6_table *mrt);
110
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 void mrt6msg_netlink_event(struct mr6_table *mrt, struct sk_buff *pkt);
120 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
121 struct netlink_callback *cb);
122 static void mroute_clean_tables(struct mr6_table *mrt, bool all);
123 static void ipmr_expire_process(struct timer_list *t);
124
125 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
126 #define ip6mr_for_each_table(mrt, net) \
127 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
128
129 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
130 {
131 struct mr6_table *mrt;
132
133 ip6mr_for_each_table(mrt, net) {
134 if (mrt->id == id)
135 return mrt;
136 }
137 return NULL;
138 }
139
140 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
141 struct mr6_table **mrt)
142 {
143 int err;
144 struct ip6mr_result res;
145 struct fib_lookup_arg arg = {
146 .result = &res,
147 .flags = FIB_LOOKUP_NOREF,
148 };
149
150 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
151 flowi6_to_flowi(flp6), 0, &arg);
152 if (err < 0)
153 return err;
154 *mrt = res.mrt;
155 return 0;
156 }
157
158 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
159 int flags, struct fib_lookup_arg *arg)
160 {
161 struct ip6mr_result *res = arg->result;
162 struct mr6_table *mrt;
163
164 switch (rule->action) {
165 case FR_ACT_TO_TBL:
166 break;
167 case FR_ACT_UNREACHABLE:
168 return -ENETUNREACH;
169 case FR_ACT_PROHIBIT:
170 return -EACCES;
171 case FR_ACT_BLACKHOLE:
172 default:
173 return -EINVAL;
174 }
175
176 mrt = ip6mr_get_table(rule->fr_net, rule->table);
177 if (!mrt)
178 return -EAGAIN;
179 res->mrt = mrt;
180 return 0;
181 }
182
183 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
184 {
185 return 1;
186 }
187
188 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
189 FRA_GENERIC_POLICY,
190 };
191
192 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
193 struct fib_rule_hdr *frh, struct nlattr **tb)
194 {
195 return 0;
196 }
197
198 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
199 struct nlattr **tb)
200 {
201 return 1;
202 }
203
204 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
205 struct fib_rule_hdr *frh)
206 {
207 frh->dst_len = 0;
208 frh->src_len = 0;
209 frh->tos = 0;
210 return 0;
211 }
212
213 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
214 .family = RTNL_FAMILY_IP6MR,
215 .rule_size = sizeof(struct ip6mr_rule),
216 .addr_size = sizeof(struct in6_addr),
217 .action = ip6mr_rule_action,
218 .match = ip6mr_rule_match,
219 .configure = ip6mr_rule_configure,
220 .compare = ip6mr_rule_compare,
221 .fill = ip6mr_rule_fill,
222 .nlgroup = RTNLGRP_IPV6_RULE,
223 .policy = ip6mr_rule_policy,
224 .owner = THIS_MODULE,
225 };
226
227 static int __net_init ip6mr_rules_init(struct net *net)
228 {
229 struct fib_rules_ops *ops;
230 struct mr6_table *mrt;
231 int err;
232
233 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
234 if (IS_ERR(ops))
235 return PTR_ERR(ops);
236
237 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
238
239 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
240 if (!mrt) {
241 err = -ENOMEM;
242 goto err1;
243 }
244
245 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
246 if (err < 0)
247 goto err2;
248
249 net->ipv6.mr6_rules_ops = ops;
250 return 0;
251
252 err2:
253 ip6mr_free_table(mrt);
254 err1:
255 fib_rules_unregister(ops);
256 return err;
257 }
258
259 static void __net_exit ip6mr_rules_exit(struct net *net)
260 {
261 struct mr6_table *mrt, *next;
262
263 rtnl_lock();
264 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
265 list_del(&mrt->list);
266 ip6mr_free_table(mrt);
267 }
268 fib_rules_unregister(net->ipv6.mr6_rules_ops);
269 rtnl_unlock();
270 }
271 #else
272 #define ip6mr_for_each_table(mrt, net) \
273 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
274
275 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
276 {
277 return net->ipv6.mrt6;
278 }
279
280 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
281 struct mr6_table **mrt)
282 {
283 *mrt = net->ipv6.mrt6;
284 return 0;
285 }
286
287 static int __net_init ip6mr_rules_init(struct net *net)
288 {
289 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
290 return net->ipv6.mrt6 ? 0 : -ENOMEM;
291 }
292
293 static void __net_exit ip6mr_rules_exit(struct net *net)
294 {
295 rtnl_lock();
296 ip6mr_free_table(net->ipv6.mrt6);
297 net->ipv6.mrt6 = NULL;
298 rtnl_unlock();
299 }
300 #endif
301
302 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
303 {
304 struct mr6_table *mrt;
305 unsigned int i;
306
307 mrt = ip6mr_get_table(net, id);
308 if (mrt)
309 return mrt;
310
311 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
312 if (!mrt)
313 return NULL;
314 mrt->id = id;
315 write_pnet(&mrt->net, net);
316
317 /* Forwarding cache */
318 for (i = 0; i < MFC6_LINES; i++)
319 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
320
321 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
322
323 timer_setup(&mrt->ipmr_expire_timer, ipmr_expire_process, 0);
324
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;
332 }
333
334 static void ip6mr_free_table(struct mr6_table *mrt)
335 {
336 del_timer_sync(&mrt->ipmr_expire_timer);
337 mroute_clean_tables(mrt, true);
338 kfree(mrt);
339 }
340
341 #ifdef CONFIG_PROC_FS
342
343 struct ipmr_mfc_iter {
344 struct seq_net_private p;
345 struct mr6_table *mrt;
346 struct list_head *cache;
347 int ct;
348 };
349
350
351 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
352 struct ipmr_mfc_iter *it, loff_t pos)
353 {
354 struct mr6_table *mrt = it->mrt;
355 struct mfc6_cache *mfc;
356
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;
363 }
364 read_unlock(&mrt_lock);
365
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);
372
373 it->cache = NULL;
374 return NULL;
375 }
376
377 /*
378 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
379 */
380
381 struct ipmr_vif_iter {
382 struct seq_net_private p;
383 struct mr6_table *mrt;
384 int ct;
385 };
386
387 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
388 struct ipmr_vif_iter *iter,
389 loff_t pos)
390 {
391 struct mr6_table *mrt = iter->mrt;
392
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];
398 }
399 return NULL;
400 }
401
402 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
403 __acquires(mrt_lock)
404 {
405 struct ipmr_vif_iter *iter = seq->private;
406 struct net *net = seq_file_net(seq);
407 struct mr6_table *mrt;
408
409 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
410 if (!mrt)
411 return ERR_PTR(-ENOENT);
412
413 iter->mrt = mrt;
414
415 read_lock(&mrt_lock);
416 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
417 : SEQ_START_TOKEN;
418 }
419
420 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
421 {
422 struct ipmr_vif_iter *iter = seq->private;
423 struct net *net = seq_file_net(seq);
424 struct mr6_table *mrt = iter->mrt;
425
426 ++*pos;
427 if (v == SEQ_START_TOKEN)
428 return ip6mr_vif_seq_idx(net, iter, 0);
429
430 while (++iter->ct < mrt->maxvif) {
431 if (!MIF_EXISTS(mrt, iter->ct))
432 continue;
433 return &mrt->vif6_table[iter->ct];
434 }
435 return NULL;
436 }
437
438 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
439 __releases(mrt_lock)
440 {
441 read_unlock(&mrt_lock);
442 }
443
444 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
445 {
446 struct ipmr_vif_iter *iter = seq->private;
447 struct mr6_table *mrt = iter->mrt;
448
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";
455
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);
462 }
463 return 0;
464 }
465
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,
471 };
472
473 static int ip6mr_vif_open(struct inode *inode, struct file *file)
474 {
475 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
476 sizeof(struct ipmr_vif_iter));
477 }
478
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,
485 };
486
487 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
488 {
489 struct ipmr_mfc_iter *it = seq->private;
490 struct net *net = seq_file_net(seq);
491 struct mr6_table *mrt;
492
493 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
494 if (!mrt)
495 return ERR_PTR(-ENOENT);
496
497 it->mrt = mrt;
498 it->cache = NULL;
499 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
500 : SEQ_START_TOKEN;
501 }
502
503 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
504 {
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;
509
510 ++*pos;
511
512 if (v == SEQ_START_TOKEN)
513 return ipmr_mfc_seq_idx(net, seq->private, 0);
514
515 if (mfc->list.next != it->cache)
516 return list_entry(mfc->list.next, struct mfc6_cache, list);
517
518 if (it->cache == &mrt->mfc6_unres_queue)
519 goto end_of_list;
520
521 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
522
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);
528 }
529
530 /* exhausted cache_array, show unresolved */
531 read_unlock(&mrt_lock);
532 it->cache = &mrt->mfc6_unres_queue;
533 it->ct = 0;
534
535 spin_lock_bh(&mfc_unres_lock);
536 if (!list_empty(it->cache))
537 return list_first_entry(it->cache, struct mfc6_cache, list);
538
539 end_of_list:
540 spin_unlock_bh(&mfc_unres_lock);
541 it->cache = NULL;
542
543 return NULL;
544 }
545
546 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
547 {
548 struct ipmr_mfc_iter *it = seq->private;
549 struct mr6_table *mrt = it->mrt;
550
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);
555 }
556
557 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
558 {
559 int n;
560
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;
570
571 seq_printf(seq, "%pI6 %pI6 %-3hd",
572 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
573 mfc->mf6c_parent);
574
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]);
587 }
588 } else {
589 /* unresolved mfc_caches don't contain
590 * pkt, bytes and wrong_if values
591 */
592 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
593 }
594 seq_putc(seq, '\n');
595 }
596 return 0;
597 }
598
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,
604 };
605
606 static int ipmr_mfc_open(struct inode *inode, struct file *file)
607 {
608 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
609 sizeof(struct ipmr_mfc_iter));
610 }
611
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,
618 };
619 #endif
620
621 #ifdef CONFIG_IPV6_PIMSM_V2
622
623 static int pim6_rcv(struct sk_buff *skb)
624 {
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,
633 };
634 int reg_vif_num;
635
636 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
637 goto drop;
638
639 pim = (struct pimreghdr *)skb_transport_header(skb);
640 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_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;
647
648 /* check if the inner packet is destined to mcast group */
649 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
650 sizeof(*pim));
651
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;
656
657 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
658 goto drop;
659 reg_vif_num = mrt->mroute_reg_vif_num;
660
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);
667
668 if (!reg_dev)
669 goto drop;
670
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;
676
677 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
678
679 netif_rx(skb);
680
681 dev_put(reg_dev);
682 return 0;
683 drop:
684 kfree_skb(skb);
685 return 0;
686 }
687
688 static const struct inet6_protocol pim6_protocol = {
689 .handler = pim6_rcv,
690 };
691
692 /* Service routines creating virtual interfaces: PIMREG */
693
694 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
695 struct net_device *dev)
696 {
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,
703 };
704 int err;
705
706 err = ip6mr_fib_lookup(net, &fl6, &mrt);
707 if (err < 0) {
708 kfree_skb(skb);
709 return err;
710 }
711
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;
719 }
720
721 static int reg_vif_get_iflink(const struct net_device *dev)
722 {
723 return 0;
724 }
725
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,
729 };
730
731 static void reg_vif_setup(struct net_device *dev)
732 {
733 dev->type = ARPHRD_PIMREG;
734 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
735 dev->flags = IFF_NOARP;
736 dev->netdev_ops = ®_vif_netdev_ops;
737 dev->needs_free_netdev = true;
738 dev->features |= NETIF_F_NETNS_LOCAL;
739 }
740
741 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
742 {
743 struct net_device *dev;
744 char name[IFNAMSIZ];
745
746 if (mrt->id == RT6_TABLE_DFLT)
747 sprintf(name, "pim6reg");
748 else
749 sprintf(name, "pim6reg%u", mrt->id);
750
751 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
752 if (!dev)
753 return NULL;
754
755 dev_net_set(dev, net);
756
757 if (register_netdevice(dev)) {
758 free_netdev(dev);
759 return NULL;
760 }
761
762 if (dev_open(dev))
763 goto failure;
764
765 dev_hold(dev);
766 return dev;
767
768 failure:
769 unregister_netdevice(dev);
770 return NULL;
771 }
772 #endif
773
774 /*
775 * Delete a VIF entry
776 */
777
778 static int mif6_delete(struct mr6_table *mrt, int vifi, int notify,
779 struct list_head *head)
780 {
781 struct mif_device *v;
782 struct net_device *dev;
783 struct inet6_dev *in6_dev;
784
785 if (vifi < 0 || vifi >= mrt->maxvif)
786 return -EADDRNOTAVAIL;
787
788 v = &mrt->vif6_table[vifi];
789
790 write_lock_bh(&mrt_lock);
791 dev = v->dev;
792 v->dev = NULL;
793
794 if (!dev) {
795 write_unlock_bh(&mrt_lock);
796 return -EADDRNOTAVAIL;
797 }
798
799 #ifdef CONFIG_IPV6_PIMSM_V2
800 if (vifi == mrt->mroute_reg_vif_num)
801 mrt->mroute_reg_vif_num = -1;
802 #endif
803
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;
809 }
810 mrt->maxvif = tmp + 1;
811 }
812
813 write_unlock_bh(&mrt_lock);
814
815 dev_set_allmulti(dev, -1);
816
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), RTM_NEWNETCONF,
821 NETCONFA_MC_FORWARDING,
822 dev->ifindex, &in6_dev->cnf);
823 }
824
825 if ((v->flags & MIFF_REGISTER) && !notify)
826 unregister_netdevice_queue(dev, head);
827
828 dev_put(dev);
829 return 0;
830 }
831
832 static inline void ip6mr_cache_free(struct mfc6_cache *c)
833 {
834 kmem_cache_free(mrt_cachep, c);
835 }
836
837 /* Destroy an unresolved cache entry, killing queued skbs
838 and reporting error to netlink readers.
839 */
840
841 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
842 {
843 struct net *net = read_pnet(&mrt->net);
844 struct sk_buff *skb;
845
846 atomic_dec(&mrt->cache_resolve_queue_len);
847
848 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
849 if (ipv6_hdr(skb)->version == 0) {
850 struct nlmsghdr *nlh = skb_pull(skb,
851 sizeof(struct ipv6hdr));
852 nlh->nlmsg_type = NLMSG_ERROR;
853 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
854 skb_trim(skb, nlh->nlmsg_len);
855 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
856 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
857 } else
858 kfree_skb(skb);
859 }
860
861 ip6mr_cache_free(c);
862 }
863
864
865 /* Timer process for all the unresolved queue. */
866
867 static void ipmr_do_expire_process(struct mr6_table *mrt)
868 {
869 unsigned long now = jiffies;
870 unsigned long expires = 10 * HZ;
871 struct mfc6_cache *c, *next;
872
873 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
874 if (time_after(c->mfc_un.unres.expires, now)) {
875 /* not yet... */
876 unsigned long interval = c->mfc_un.unres.expires - now;
877 if (interval < expires)
878 expires = interval;
879 continue;
880 }
881
882 list_del(&c->list);
883 mr6_netlink_event(mrt, c, RTM_DELROUTE);
884 ip6mr_destroy_unres(mrt, c);
885 }
886
887 if (!list_empty(&mrt->mfc6_unres_queue))
888 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
889 }
890
891 static void ipmr_expire_process(struct timer_list *t)
892 {
893 struct mr6_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
894
895 if (!spin_trylock(&mfc_unres_lock)) {
896 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
897 return;
898 }
899
900 if (!list_empty(&mrt->mfc6_unres_queue))
901 ipmr_do_expire_process(mrt);
902
903 spin_unlock(&mfc_unres_lock);
904 }
905
906 /* Fill oifs list. It is called under write locked mrt_lock. */
907
908 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
909 unsigned char *ttls)
910 {
911 int vifi;
912
913 cache->mfc_un.res.minvif = MAXMIFS;
914 cache->mfc_un.res.maxvif = 0;
915 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
916
917 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
918 if (MIF_EXISTS(mrt, vifi) &&
919 ttls[vifi] && ttls[vifi] < 255) {
920 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
921 if (cache->mfc_un.res.minvif > vifi)
922 cache->mfc_un.res.minvif = vifi;
923 if (cache->mfc_un.res.maxvif <= vifi)
924 cache->mfc_un.res.maxvif = vifi + 1;
925 }
926 }
927 cache->mfc_un.res.lastuse = jiffies;
928 }
929
930 static int mif6_add(struct net *net, struct mr6_table *mrt,
931 struct mif6ctl *vifc, int mrtsock)
932 {
933 int vifi = vifc->mif6c_mifi;
934 struct mif_device *v = &mrt->vif6_table[vifi];
935 struct net_device *dev;
936 struct inet6_dev *in6_dev;
937 int err;
938
939 /* Is vif busy ? */
940 if (MIF_EXISTS(mrt, vifi))
941 return -EADDRINUSE;
942
943 switch (vifc->mif6c_flags) {
944 #ifdef CONFIG_IPV6_PIMSM_V2
945 case MIFF_REGISTER:
946 /*
947 * Special Purpose VIF in PIM
948 * All the packets will be sent to the daemon
949 */
950 if (mrt->mroute_reg_vif_num >= 0)
951 return -EADDRINUSE;
952 dev = ip6mr_reg_vif(net, mrt);
953 if (!dev)
954 return -ENOBUFS;
955 err = dev_set_allmulti(dev, 1);
956 if (err) {
957 unregister_netdevice(dev);
958 dev_put(dev);
959 return err;
960 }
961 break;
962 #endif
963 case 0:
964 dev = dev_get_by_index(net, vifc->mif6c_pifi);
965 if (!dev)
966 return -EADDRNOTAVAIL;
967 err = dev_set_allmulti(dev, 1);
968 if (err) {
969 dev_put(dev);
970 return err;
971 }
972 break;
973 default:
974 return -EINVAL;
975 }
976
977 in6_dev = __in6_dev_get(dev);
978 if (in6_dev) {
979 in6_dev->cnf.mc_forwarding++;
980 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
981 NETCONFA_MC_FORWARDING,
982 dev->ifindex, &in6_dev->cnf);
983 }
984
985 /*
986 * Fill in the VIF structures
987 */
988 v->rate_limit = vifc->vifc_rate_limit;
989 v->flags = vifc->mif6c_flags;
990 if (!mrtsock)
991 v->flags |= VIFF_STATIC;
992 v->threshold = vifc->vifc_threshold;
993 v->bytes_in = 0;
994 v->bytes_out = 0;
995 v->pkt_in = 0;
996 v->pkt_out = 0;
997 v->link = dev->ifindex;
998 if (v->flags & MIFF_REGISTER)
999 v->link = dev_get_iflink(dev);
1000
1001 /* And finish update writing critical data */
1002 write_lock_bh(&mrt_lock);
1003 v->dev = dev;
1004 #ifdef CONFIG_IPV6_PIMSM_V2
1005 if (v->flags & MIFF_REGISTER)
1006 mrt->mroute_reg_vif_num = vifi;
1007 #endif
1008 if (vifi + 1 > mrt->maxvif)
1009 mrt->maxvif = vifi + 1;
1010 write_unlock_bh(&mrt_lock);
1011 return 0;
1012 }
1013
1014 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1015 const struct in6_addr *origin,
1016 const struct in6_addr *mcastgrp)
1017 {
1018 int line = MFC6_HASH(mcastgrp, origin);
1019 struct mfc6_cache *c;
1020
1021 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1022 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1023 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1024 return c;
1025 }
1026 return NULL;
1027 }
1028
1029 /* Look for a (*,*,oif) entry */
1030 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1031 mifi_t mifi)
1032 {
1033 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1034 struct mfc6_cache *c;
1035
1036 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1037 if (ipv6_addr_any(&c->mf6c_origin) &&
1038 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1039 (c->mfc_un.res.ttls[mifi] < 255))
1040 return c;
1041
1042 return NULL;
1043 }
1044
1045 /* Look for a (*,G) entry */
1046 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1047 struct in6_addr *mcastgrp,
1048 mifi_t mifi)
1049 {
1050 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1051 struct mfc6_cache *c, *proxy;
1052
1053 if (ipv6_addr_any(mcastgrp))
1054 goto skip;
1055
1056 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1057 if (ipv6_addr_any(&c->mf6c_origin) &&
1058 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1059 if (c->mfc_un.res.ttls[mifi] < 255)
1060 return c;
1061
1062 /* It's ok if the mifi is part of the static tree */
1063 proxy = ip6mr_cache_find_any_parent(mrt,
1064 c->mf6c_parent);
1065 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1066 return c;
1067 }
1068
1069 skip:
1070 return ip6mr_cache_find_any_parent(mrt, mifi);
1071 }
1072
1073 /*
1074 * Allocate a multicast cache entry
1075 */
1076 static struct mfc6_cache *ip6mr_cache_alloc(void)
1077 {
1078 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1079 if (!c)
1080 return NULL;
1081 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1082 c->mfc_un.res.minvif = MAXMIFS;
1083 return c;
1084 }
1085
1086 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1087 {
1088 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1089 if (!c)
1090 return NULL;
1091 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1092 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1093 return c;
1094 }
1095
1096 /*
1097 * A cache entry has gone into a resolved state from queued
1098 */
1099
1100 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1101 struct mfc6_cache *uc, struct mfc6_cache *c)
1102 {
1103 struct sk_buff *skb;
1104
1105 /*
1106 * Play the pending entries through our router
1107 */
1108
1109 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1110 if (ipv6_hdr(skb)->version == 0) {
1111 struct nlmsghdr *nlh = skb_pull(skb,
1112 sizeof(struct ipv6hdr));
1113
1114 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1115 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1116 } else {
1117 nlh->nlmsg_type = NLMSG_ERROR;
1118 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1119 skb_trim(skb, nlh->nlmsg_len);
1120 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1121 }
1122 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1123 } else
1124 ip6_mr_forward(net, mrt, skb, c);
1125 }
1126 }
1127
1128 /*
1129 * Bounce a cache query up to pim6sd and netlink.
1130 *
1131 * Called under mrt_lock.
1132 */
1133
1134 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1135 mifi_t mifi, int assert)
1136 {
1137 struct sk_buff *skb;
1138 struct mrt6msg *msg;
1139 int ret;
1140
1141 #ifdef CONFIG_IPV6_PIMSM_V2
1142 if (assert == MRT6MSG_WHOLEPKT)
1143 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1144 +sizeof(*msg));
1145 else
1146 #endif
1147 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1148
1149 if (!skb)
1150 return -ENOBUFS;
1151
1152 /* I suppose that internal messages
1153 * do not require checksums */
1154
1155 skb->ip_summed = CHECKSUM_UNNECESSARY;
1156
1157 #ifdef CONFIG_IPV6_PIMSM_V2
1158 if (assert == MRT6MSG_WHOLEPKT) {
1159 /* Ugly, but we have no choice with this interface.
1160 Duplicate old header, fix length etc.
1161 And all this only to mangle msg->im6_msgtype and
1162 to set msg->im6_mbz to "mbz" :-)
1163 */
1164 skb_push(skb, -skb_network_offset(pkt));
1165
1166 skb_push(skb, sizeof(*msg));
1167 skb_reset_transport_header(skb);
1168 msg = (struct mrt6msg *)skb_transport_header(skb);
1169 msg->im6_mbz = 0;
1170 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1171 msg->im6_mif = mrt->mroute_reg_vif_num;
1172 msg->im6_pad = 0;
1173 msg->im6_src = ipv6_hdr(pkt)->saddr;
1174 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1175
1176 skb->ip_summed = CHECKSUM_UNNECESSARY;
1177 } else
1178 #endif
1179 {
1180 /*
1181 * Copy the IP header
1182 */
1183
1184 skb_put(skb, sizeof(struct ipv6hdr));
1185 skb_reset_network_header(skb);
1186 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1187
1188 /*
1189 * Add our header
1190 */
1191 skb_put(skb, sizeof(*msg));
1192 skb_reset_transport_header(skb);
1193 msg = (struct mrt6msg *)skb_transport_header(skb);
1194
1195 msg->im6_mbz = 0;
1196 msg->im6_msgtype = assert;
1197 msg->im6_mif = mifi;
1198 msg->im6_pad = 0;
1199 msg->im6_src = ipv6_hdr(pkt)->saddr;
1200 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1201
1202 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1203 skb->ip_summed = CHECKSUM_UNNECESSARY;
1204 }
1205
1206 if (!mrt->mroute6_sk) {
1207 kfree_skb(skb);
1208 return -EINVAL;
1209 }
1210
1211 mrt6msg_netlink_event(mrt, skb);
1212
1213 /*
1214 * Deliver to user space multicast routing algorithms
1215 */
1216 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1217 if (ret < 0) {
1218 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1219 kfree_skb(skb);
1220 }
1221
1222 return ret;
1223 }
1224
1225 /*
1226 * Queue a packet for resolution. It gets locked cache entry!
1227 */
1228
1229 static int
1230 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1231 {
1232 bool found = false;
1233 int err;
1234 struct mfc6_cache *c;
1235
1236 spin_lock_bh(&mfc_unres_lock);
1237 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1238 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1239 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1240 found = true;
1241 break;
1242 }
1243 }
1244
1245 if (!found) {
1246 /*
1247 * Create a new entry if allowable
1248 */
1249
1250 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1251 (c = ip6mr_cache_alloc_unres()) == NULL) {
1252 spin_unlock_bh(&mfc_unres_lock);
1253
1254 kfree_skb(skb);
1255 return -ENOBUFS;
1256 }
1257
1258 /*
1259 * Fill in the new cache entry
1260 */
1261 c->mf6c_parent = -1;
1262 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1263 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1264
1265 /*
1266 * Reflect first query at pim6sd
1267 */
1268 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1269 if (err < 0) {
1270 /* If the report failed throw the cache entry
1271 out - Brad Parker
1272 */
1273 spin_unlock_bh(&mfc_unres_lock);
1274
1275 ip6mr_cache_free(c);
1276 kfree_skb(skb);
1277 return err;
1278 }
1279
1280 atomic_inc(&mrt->cache_resolve_queue_len);
1281 list_add(&c->list, &mrt->mfc6_unres_queue);
1282 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1283
1284 ipmr_do_expire_process(mrt);
1285 }
1286
1287 /*
1288 * See if we can append the packet
1289 */
1290 if (c->mfc_un.unres.unresolved.qlen > 3) {
1291 kfree_skb(skb);
1292 err = -ENOBUFS;
1293 } else {
1294 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1295 err = 0;
1296 }
1297
1298 spin_unlock_bh(&mfc_unres_lock);
1299 return err;
1300 }
1301
1302 /*
1303 * MFC6 cache manipulation by user space
1304 */
1305
1306 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1307 int parent)
1308 {
1309 int line;
1310 struct mfc6_cache *c, *next;
1311
1312 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1313
1314 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1315 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1316 ipv6_addr_equal(&c->mf6c_mcastgrp,
1317 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1318 (parent == -1 || parent == c->mf6c_parent)) {
1319 write_lock_bh(&mrt_lock);
1320 list_del(&c->list);
1321 write_unlock_bh(&mrt_lock);
1322
1323 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1324 ip6mr_cache_free(c);
1325 return 0;
1326 }
1327 }
1328 return -ENOENT;
1329 }
1330
1331 static int ip6mr_device_event(struct notifier_block *this,
1332 unsigned long event, void *ptr)
1333 {
1334 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1335 struct net *net = dev_net(dev);
1336 struct mr6_table *mrt;
1337 struct mif_device *v;
1338 int ct;
1339
1340 if (event != NETDEV_UNREGISTER)
1341 return NOTIFY_DONE;
1342
1343 ip6mr_for_each_table(mrt, net) {
1344 v = &mrt->vif6_table[0];
1345 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1346 if (v->dev == dev)
1347 mif6_delete(mrt, ct, 1, NULL);
1348 }
1349 }
1350
1351 return NOTIFY_DONE;
1352 }
1353
1354 static struct notifier_block ip6_mr_notifier = {
1355 .notifier_call = ip6mr_device_event
1356 };
1357
1358 /*
1359 * Setup for IP multicast routing
1360 */
1361
1362 static int __net_init ip6mr_net_init(struct net *net)
1363 {
1364 int err;
1365
1366 err = ip6mr_rules_init(net);
1367 if (err < 0)
1368 goto fail;
1369
1370 #ifdef CONFIG_PROC_FS
1371 err = -ENOMEM;
1372 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1373 goto proc_vif_fail;
1374 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1375 goto proc_cache_fail;
1376 #endif
1377
1378 return 0;
1379
1380 #ifdef CONFIG_PROC_FS
1381 proc_cache_fail:
1382 remove_proc_entry("ip6_mr_vif", net->proc_net);
1383 proc_vif_fail:
1384 ip6mr_rules_exit(net);
1385 #endif
1386 fail:
1387 return err;
1388 }
1389
1390 static void __net_exit ip6mr_net_exit(struct net *net)
1391 {
1392 #ifdef CONFIG_PROC_FS
1393 remove_proc_entry("ip6_mr_cache", net->proc_net);
1394 remove_proc_entry("ip6_mr_vif", net->proc_net);
1395 #endif
1396 ip6mr_rules_exit(net);
1397 }
1398
1399 static struct pernet_operations ip6mr_net_ops = {
1400 .init = ip6mr_net_init,
1401 .exit = ip6mr_net_exit,
1402 };
1403
1404 int __init ip6_mr_init(void)
1405 {
1406 int err;
1407
1408 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1409 sizeof(struct mfc6_cache),
1410 0, SLAB_HWCACHE_ALIGN,
1411 NULL);
1412 if (!mrt_cachep)
1413 return -ENOMEM;
1414
1415 err = register_pernet_subsys(&ip6mr_net_ops);
1416 if (err)
1417 goto reg_pernet_fail;
1418
1419 err = register_netdevice_notifier(&ip6_mr_notifier);
1420 if (err)
1421 goto reg_notif_fail;
1422 #ifdef CONFIG_IPV6_PIMSM_V2
1423 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1424 pr_err("%s: can't add PIM protocol\n", __func__);
1425 err = -EAGAIN;
1426 goto add_proto_fail;
1427 }
1428 #endif
1429 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1430 ip6mr_rtm_dumproute, 0);
1431 return 0;
1432 #ifdef CONFIG_IPV6_PIMSM_V2
1433 add_proto_fail:
1434 unregister_netdevice_notifier(&ip6_mr_notifier);
1435 #endif
1436 reg_notif_fail:
1437 unregister_pernet_subsys(&ip6mr_net_ops);
1438 reg_pernet_fail:
1439 kmem_cache_destroy(mrt_cachep);
1440 return err;
1441 }
1442
1443 void ip6_mr_cleanup(void)
1444 {
1445 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1446 #ifdef CONFIG_IPV6_PIMSM_V2
1447 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1448 #endif
1449 unregister_netdevice_notifier(&ip6_mr_notifier);
1450 unregister_pernet_subsys(&ip6mr_net_ops);
1451 kmem_cache_destroy(mrt_cachep);
1452 }
1453
1454 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1455 struct mf6cctl *mfc, int mrtsock, int parent)
1456 {
1457 bool found = false;
1458 int line;
1459 struct mfc6_cache *uc, *c;
1460 unsigned char ttls[MAXMIFS];
1461 int i;
1462
1463 if (mfc->mf6cc_parent >= MAXMIFS)
1464 return -ENFILE;
1465
1466 memset(ttls, 255, MAXMIFS);
1467 for (i = 0; i < MAXMIFS; i++) {
1468 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1469 ttls[i] = 1;
1470
1471 }
1472
1473 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1474
1475 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1476 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1477 ipv6_addr_equal(&c->mf6c_mcastgrp,
1478 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1479 (parent == -1 || parent == mfc->mf6cc_parent)) {
1480 found = true;
1481 break;
1482 }
1483 }
1484
1485 if (found) {
1486 write_lock_bh(&mrt_lock);
1487 c->mf6c_parent = mfc->mf6cc_parent;
1488 ip6mr_update_thresholds(mrt, c, ttls);
1489 if (!mrtsock)
1490 c->mfc_flags |= MFC_STATIC;
1491 write_unlock_bh(&mrt_lock);
1492 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1493 return 0;
1494 }
1495
1496 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1497 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1498 return -EINVAL;
1499
1500 c = ip6mr_cache_alloc();
1501 if (!c)
1502 return -ENOMEM;
1503
1504 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1505 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1506 c->mf6c_parent = mfc->mf6cc_parent;
1507 ip6mr_update_thresholds(mrt, c, ttls);
1508 if (!mrtsock)
1509 c->mfc_flags |= MFC_STATIC;
1510
1511 write_lock_bh(&mrt_lock);
1512 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1513 write_unlock_bh(&mrt_lock);
1514
1515 /*
1516 * Check to see if we resolved a queued list. If so we
1517 * need to send on the frames and tidy up.
1518 */
1519 found = false;
1520 spin_lock_bh(&mfc_unres_lock);
1521 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1522 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1523 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1524 list_del(&uc->list);
1525 atomic_dec(&mrt->cache_resolve_queue_len);
1526 found = true;
1527 break;
1528 }
1529 }
1530 if (list_empty(&mrt->mfc6_unres_queue))
1531 del_timer(&mrt->ipmr_expire_timer);
1532 spin_unlock_bh(&mfc_unres_lock);
1533
1534 if (found) {
1535 ip6mr_cache_resolve(net, mrt, uc, c);
1536 ip6mr_cache_free(uc);
1537 }
1538 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1539 return 0;
1540 }
1541
1542 /*
1543 * Close the multicast socket, and clear the vif tables etc
1544 */
1545
1546 static void mroute_clean_tables(struct mr6_table *mrt, bool all)
1547 {
1548 int i;
1549 LIST_HEAD(list);
1550 struct mfc6_cache *c, *next;
1551
1552 /*
1553 * Shut down all active vif entries
1554 */
1555 for (i = 0; i < mrt->maxvif; i++) {
1556 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC))
1557 continue;
1558 mif6_delete(mrt, i, 0, &list);
1559 }
1560 unregister_netdevice_many(&list);
1561
1562 /*
1563 * Wipe the cache
1564 */
1565 for (i = 0; i < MFC6_LINES; i++) {
1566 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1567 if (!all && (c->mfc_flags & MFC_STATIC))
1568 continue;
1569 write_lock_bh(&mrt_lock);
1570 list_del(&c->list);
1571 write_unlock_bh(&mrt_lock);
1572
1573 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1574 ip6mr_cache_free(c);
1575 }
1576 }
1577
1578 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1579 spin_lock_bh(&mfc_unres_lock);
1580 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1581 list_del(&c->list);
1582 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1583 ip6mr_destroy_unres(mrt, c);
1584 }
1585 spin_unlock_bh(&mfc_unres_lock);
1586 }
1587 }
1588
1589 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1590 {
1591 int err = 0;
1592 struct net *net = sock_net(sk);
1593
1594 rtnl_lock();
1595 write_lock_bh(&mrt_lock);
1596 if (likely(mrt->mroute6_sk == NULL)) {
1597 mrt->mroute6_sk = sk;
1598 net->ipv6.devconf_all->mc_forwarding++;
1599 } else {
1600 err = -EADDRINUSE;
1601 }
1602 write_unlock_bh(&mrt_lock);
1603
1604 if (!err)
1605 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1606 NETCONFA_MC_FORWARDING,
1607 NETCONFA_IFINDEX_ALL,
1608 net->ipv6.devconf_all);
1609 rtnl_unlock();
1610
1611 return err;
1612 }
1613
1614 int ip6mr_sk_done(struct sock *sk)
1615 {
1616 int err = -EACCES;
1617 struct net *net = sock_net(sk);
1618 struct mr6_table *mrt;
1619
1620 if (sk->sk_type != SOCK_RAW ||
1621 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1622 return err;
1623
1624 rtnl_lock();
1625 ip6mr_for_each_table(mrt, net) {
1626 if (sk == mrt->mroute6_sk) {
1627 write_lock_bh(&mrt_lock);
1628 mrt->mroute6_sk = NULL;
1629 net->ipv6.devconf_all->mc_forwarding--;
1630 write_unlock_bh(&mrt_lock);
1631 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1632 NETCONFA_MC_FORWARDING,
1633 NETCONFA_IFINDEX_ALL,
1634 net->ipv6.devconf_all);
1635
1636 mroute_clean_tables(mrt, false);
1637 err = 0;
1638 break;
1639 }
1640 }
1641 rtnl_unlock();
1642
1643 return err;
1644 }
1645
1646 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1647 {
1648 struct mr6_table *mrt;
1649 struct flowi6 fl6 = {
1650 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1651 .flowi6_oif = skb->dev->ifindex,
1652 .flowi6_mark = skb->mark,
1653 };
1654
1655 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1656 return NULL;
1657
1658 return mrt->mroute6_sk;
1659 }
1660
1661 /*
1662 * Socket options and virtual interface manipulation. The whole
1663 * virtual interface system is a complete heap, but unfortunately
1664 * that's how BSD mrouted happens to think. Maybe one day with a proper
1665 * MOSPF/PIM router set up we can clean this up.
1666 */
1667
1668 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1669 {
1670 int ret, parent = 0;
1671 struct mif6ctl vif;
1672 struct mf6cctl mfc;
1673 mifi_t mifi;
1674 struct net *net = sock_net(sk);
1675 struct mr6_table *mrt;
1676
1677 if (sk->sk_type != SOCK_RAW ||
1678 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1679 return -EOPNOTSUPP;
1680
1681 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1682 if (!mrt)
1683 return -ENOENT;
1684
1685 if (optname != MRT6_INIT) {
1686 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1687 return -EACCES;
1688 }
1689
1690 switch (optname) {
1691 case MRT6_INIT:
1692 if (optlen < sizeof(int))
1693 return -EINVAL;
1694
1695 return ip6mr_sk_init(mrt, sk);
1696
1697 case MRT6_DONE:
1698 return ip6mr_sk_done(sk);
1699
1700 case MRT6_ADD_MIF:
1701 if (optlen < sizeof(vif))
1702 return -EINVAL;
1703 if (copy_from_user(&vif, optval, sizeof(vif)))
1704 return -EFAULT;
1705 if (vif.mif6c_mifi >= MAXMIFS)
1706 return -ENFILE;
1707 rtnl_lock();
1708 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1709 rtnl_unlock();
1710 return ret;
1711
1712 case MRT6_DEL_MIF:
1713 if (optlen < sizeof(mifi_t))
1714 return -EINVAL;
1715 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1716 return -EFAULT;
1717 rtnl_lock();
1718 ret = mif6_delete(mrt, mifi, 0, NULL);
1719 rtnl_unlock();
1720 return ret;
1721
1722 /*
1723 * Manipulate the forwarding caches. These live
1724 * in a sort of kernel/user symbiosis.
1725 */
1726 case MRT6_ADD_MFC:
1727 case MRT6_DEL_MFC:
1728 parent = -1;
1729 /* fall through */
1730 case MRT6_ADD_MFC_PROXY:
1731 case MRT6_DEL_MFC_PROXY:
1732 if (optlen < sizeof(mfc))
1733 return -EINVAL;
1734 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1735 return -EFAULT;
1736 if (parent == 0)
1737 parent = mfc.mf6cc_parent;
1738 rtnl_lock();
1739 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1740 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1741 else
1742 ret = ip6mr_mfc_add(net, mrt, &mfc,
1743 sk == mrt->mroute6_sk, parent);
1744 rtnl_unlock();
1745 return ret;
1746
1747 /*
1748 * Control PIM assert (to activate pim will activate assert)
1749 */
1750 case MRT6_ASSERT:
1751 {
1752 int v;
1753
1754 if (optlen != sizeof(v))
1755 return -EINVAL;
1756 if (get_user(v, (int __user *)optval))
1757 return -EFAULT;
1758 mrt->mroute_do_assert = v;
1759 return 0;
1760 }
1761
1762 #ifdef CONFIG_IPV6_PIMSM_V2
1763 case MRT6_PIM:
1764 {
1765 int v;
1766
1767 if (optlen != sizeof(v))
1768 return -EINVAL;
1769 if (get_user(v, (int __user *)optval))
1770 return -EFAULT;
1771 v = !!v;
1772 rtnl_lock();
1773 ret = 0;
1774 if (v != mrt->mroute_do_pim) {
1775 mrt->mroute_do_pim = v;
1776 mrt->mroute_do_assert = v;
1777 }
1778 rtnl_unlock();
1779 return ret;
1780 }
1781
1782 #endif
1783 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1784 case MRT6_TABLE:
1785 {
1786 u32 v;
1787
1788 if (optlen != sizeof(u32))
1789 return -EINVAL;
1790 if (get_user(v, (u32 __user *)optval))
1791 return -EFAULT;
1792 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1793 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1794 return -EINVAL;
1795 if (sk == mrt->mroute6_sk)
1796 return -EBUSY;
1797
1798 rtnl_lock();
1799 ret = 0;
1800 if (!ip6mr_new_table(net, v))
1801 ret = -ENOMEM;
1802 raw6_sk(sk)->ip6mr_table = v;
1803 rtnl_unlock();
1804 return ret;
1805 }
1806 #endif
1807 /*
1808 * Spurious command, or MRT6_VERSION which you cannot
1809 * set.
1810 */
1811 default:
1812 return -ENOPROTOOPT;
1813 }
1814 }
1815
1816 /*
1817 * Getsock opt support for the multicast routing system.
1818 */
1819
1820 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1821 int __user *optlen)
1822 {
1823 int olr;
1824 int val;
1825 struct net *net = sock_net(sk);
1826 struct mr6_table *mrt;
1827
1828 if (sk->sk_type != SOCK_RAW ||
1829 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1830 return -EOPNOTSUPP;
1831
1832 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1833 if (!mrt)
1834 return -ENOENT;
1835
1836 switch (optname) {
1837 case MRT6_VERSION:
1838 val = 0x0305;
1839 break;
1840 #ifdef CONFIG_IPV6_PIMSM_V2
1841 case MRT6_PIM:
1842 val = mrt->mroute_do_pim;
1843 break;
1844 #endif
1845 case MRT6_ASSERT:
1846 val = mrt->mroute_do_assert;
1847 break;
1848 default:
1849 return -ENOPROTOOPT;
1850 }
1851
1852 if (get_user(olr, optlen))
1853 return -EFAULT;
1854
1855 olr = min_t(int, olr, sizeof(int));
1856 if (olr < 0)
1857 return -EINVAL;
1858
1859 if (put_user(olr, optlen))
1860 return -EFAULT;
1861 if (copy_to_user(optval, &val, olr))
1862 return -EFAULT;
1863 return 0;
1864 }
1865
1866 /*
1867 * The IP multicast ioctl support routines.
1868 */
1869
1870 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1871 {
1872 struct sioc_sg_req6 sr;
1873 struct sioc_mif_req6 vr;
1874 struct mif_device *vif;
1875 struct mfc6_cache *c;
1876 struct net *net = sock_net(sk);
1877 struct mr6_table *mrt;
1878
1879 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1880 if (!mrt)
1881 return -ENOENT;
1882
1883 switch (cmd) {
1884 case SIOCGETMIFCNT_IN6:
1885 if (copy_from_user(&vr, arg, sizeof(vr)))
1886 return -EFAULT;
1887 if (vr.mifi >= mrt->maxvif)
1888 return -EINVAL;
1889 read_lock(&mrt_lock);
1890 vif = &mrt->vif6_table[vr.mifi];
1891 if (MIF_EXISTS(mrt, vr.mifi)) {
1892 vr.icount = vif->pkt_in;
1893 vr.ocount = vif->pkt_out;
1894 vr.ibytes = vif->bytes_in;
1895 vr.obytes = vif->bytes_out;
1896 read_unlock(&mrt_lock);
1897
1898 if (copy_to_user(arg, &vr, sizeof(vr)))
1899 return -EFAULT;
1900 return 0;
1901 }
1902 read_unlock(&mrt_lock);
1903 return -EADDRNOTAVAIL;
1904 case SIOCGETSGCNT_IN6:
1905 if (copy_from_user(&sr, arg, sizeof(sr)))
1906 return -EFAULT;
1907
1908 read_lock(&mrt_lock);
1909 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1910 if (c) {
1911 sr.pktcnt = c->mfc_un.res.pkt;
1912 sr.bytecnt = c->mfc_un.res.bytes;
1913 sr.wrong_if = c->mfc_un.res.wrong_if;
1914 read_unlock(&mrt_lock);
1915
1916 if (copy_to_user(arg, &sr, sizeof(sr)))
1917 return -EFAULT;
1918 return 0;
1919 }
1920 read_unlock(&mrt_lock);
1921 return -EADDRNOTAVAIL;
1922 default:
1923 return -ENOIOCTLCMD;
1924 }
1925 }
1926
1927 #ifdef CONFIG_COMPAT
1928 struct compat_sioc_sg_req6 {
1929 struct sockaddr_in6 src;
1930 struct sockaddr_in6 grp;
1931 compat_ulong_t pktcnt;
1932 compat_ulong_t bytecnt;
1933 compat_ulong_t wrong_if;
1934 };
1935
1936 struct compat_sioc_mif_req6 {
1937 mifi_t mifi;
1938 compat_ulong_t icount;
1939 compat_ulong_t ocount;
1940 compat_ulong_t ibytes;
1941 compat_ulong_t obytes;
1942 };
1943
1944 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1945 {
1946 struct compat_sioc_sg_req6 sr;
1947 struct compat_sioc_mif_req6 vr;
1948 struct mif_device *vif;
1949 struct mfc6_cache *c;
1950 struct net *net = sock_net(sk);
1951 struct mr6_table *mrt;
1952
1953 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1954 if (!mrt)
1955 return -ENOENT;
1956
1957 switch (cmd) {
1958 case SIOCGETMIFCNT_IN6:
1959 if (copy_from_user(&vr, arg, sizeof(vr)))
1960 return -EFAULT;
1961 if (vr.mifi >= mrt->maxvif)
1962 return -EINVAL;
1963 read_lock(&mrt_lock);
1964 vif = &mrt->vif6_table[vr.mifi];
1965 if (MIF_EXISTS(mrt, vr.mifi)) {
1966 vr.icount = vif->pkt_in;
1967 vr.ocount = vif->pkt_out;
1968 vr.ibytes = vif->bytes_in;
1969 vr.obytes = vif->bytes_out;
1970 read_unlock(&mrt_lock);
1971
1972 if (copy_to_user(arg, &vr, sizeof(vr)))
1973 return -EFAULT;
1974 return 0;
1975 }
1976 read_unlock(&mrt_lock);
1977 return -EADDRNOTAVAIL;
1978 case SIOCGETSGCNT_IN6:
1979 if (copy_from_user(&sr, arg, sizeof(sr)))
1980 return -EFAULT;
1981
1982 read_lock(&mrt_lock);
1983 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1984 if (c) {
1985 sr.pktcnt = c->mfc_un.res.pkt;
1986 sr.bytecnt = c->mfc_un.res.bytes;
1987 sr.wrong_if = c->mfc_un.res.wrong_if;
1988 read_unlock(&mrt_lock);
1989
1990 if (copy_to_user(arg, &sr, sizeof(sr)))
1991 return -EFAULT;
1992 return 0;
1993 }
1994 read_unlock(&mrt_lock);
1995 return -EADDRNOTAVAIL;
1996 default:
1997 return -ENOIOCTLCMD;
1998 }
1999 }
2000 #endif
2001
2002 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
2003 {
2004 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
2005 IPSTATS_MIB_OUTFORWDATAGRAMS);
2006 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)),
2007 IPSTATS_MIB_OUTOCTETS, skb->len);
2008 return dst_output(net, sk, skb);
2009 }
2010
2011 /*
2012 * Processing handlers for ip6mr_forward
2013 */
2014
2015 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
2016 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
2017 {
2018 struct ipv6hdr *ipv6h;
2019 struct mif_device *vif = &mrt->vif6_table[vifi];
2020 struct net_device *dev;
2021 struct dst_entry *dst;
2022 struct flowi6 fl6;
2023
2024 if (!vif->dev)
2025 goto out_free;
2026
2027 #ifdef CONFIG_IPV6_PIMSM_V2
2028 if (vif->flags & MIFF_REGISTER) {
2029 vif->pkt_out++;
2030 vif->bytes_out += skb->len;
2031 vif->dev->stats.tx_bytes += skb->len;
2032 vif->dev->stats.tx_packets++;
2033 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2034 goto out_free;
2035 }
2036 #endif
2037
2038 ipv6h = ipv6_hdr(skb);
2039
2040 fl6 = (struct flowi6) {
2041 .flowi6_oif = vif->link,
2042 .daddr = ipv6h->daddr,
2043 };
2044
2045 dst = ip6_route_output(net, NULL, &fl6);
2046 if (dst->error) {
2047 dst_release(dst);
2048 goto out_free;
2049 }
2050
2051 skb_dst_drop(skb);
2052 skb_dst_set(skb, dst);
2053
2054 /*
2055 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2056 * not only before forwarding, but after forwarding on all output
2057 * interfaces. It is clear, if mrouter runs a multicasting
2058 * program, it should receive packets not depending to what interface
2059 * program is joined.
2060 * If we will not make it, the program will have to join on all
2061 * interfaces. On the other hand, multihoming host (or router, but
2062 * not mrouter) cannot join to more than one interface - it will
2063 * result in receiving multiple packets.
2064 */
2065 dev = vif->dev;
2066 skb->dev = dev;
2067 vif->pkt_out++;
2068 vif->bytes_out += skb->len;
2069
2070 /* We are about to write */
2071 /* XXX: extension headers? */
2072 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2073 goto out_free;
2074
2075 ipv6h = ipv6_hdr(skb);
2076 ipv6h->hop_limit--;
2077
2078 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2079
2080 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2081 net, NULL, skb, skb->dev, dev,
2082 ip6mr_forward2_finish);
2083
2084 out_free:
2085 kfree_skb(skb);
2086 return 0;
2087 }
2088
2089 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2090 {
2091 int ct;
2092
2093 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2094 if (mrt->vif6_table[ct].dev == dev)
2095 break;
2096 }
2097 return ct;
2098 }
2099
2100 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2101 struct sk_buff *skb, struct mfc6_cache *cache)
2102 {
2103 int psend = -1;
2104 int vif, ct;
2105 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2106
2107 vif = cache->mf6c_parent;
2108 cache->mfc_un.res.pkt++;
2109 cache->mfc_un.res.bytes += skb->len;
2110 cache->mfc_un.res.lastuse = jiffies;
2111
2112 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2113 struct mfc6_cache *cache_proxy;
2114
2115 /* For an (*,G) entry, we only check that the incoming
2116 * interface is part of the static tree.
2117 */
2118 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2119 if (cache_proxy &&
2120 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2121 goto forward;
2122 }
2123
2124 /*
2125 * Wrong interface: drop packet and (maybe) send PIM assert.
2126 */
2127 if (mrt->vif6_table[vif].dev != skb->dev) {
2128 cache->mfc_un.res.wrong_if++;
2129
2130 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2131 /* pimsm uses asserts, when switching from RPT to SPT,
2132 so that we cannot check that packet arrived on an oif.
2133 It is bad, but otherwise we would need to move pretty
2134 large chunk of pimd to kernel. Ough... --ANK
2135 */
2136 (mrt->mroute_do_pim ||
2137 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2138 time_after(jiffies,
2139 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2140 cache->mfc_un.res.last_assert = jiffies;
2141 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2142 }
2143 goto dont_forward;
2144 }
2145
2146 forward:
2147 mrt->vif6_table[vif].pkt_in++;
2148 mrt->vif6_table[vif].bytes_in += skb->len;
2149
2150 /*
2151 * Forward the frame
2152 */
2153 if (ipv6_addr_any(&cache->mf6c_origin) &&
2154 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2155 if (true_vifi >= 0 &&
2156 true_vifi != cache->mf6c_parent &&
2157 ipv6_hdr(skb)->hop_limit >
2158 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2159 /* It's an (*,*) entry and the packet is not coming from
2160 * the upstream: forward the packet to the upstream
2161 * only.
2162 */
2163 psend = cache->mf6c_parent;
2164 goto last_forward;
2165 }
2166 goto dont_forward;
2167 }
2168 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2169 /* For (*,G) entry, don't forward to the incoming interface */
2170 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2171 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2172 if (psend != -1) {
2173 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2174 if (skb2)
2175 ip6mr_forward2(net, mrt, skb2, cache, psend);
2176 }
2177 psend = ct;
2178 }
2179 }
2180 last_forward:
2181 if (psend != -1) {
2182 ip6mr_forward2(net, mrt, skb, cache, psend);
2183 return;
2184 }
2185
2186 dont_forward:
2187 kfree_skb(skb);
2188 }
2189
2190
2191 /*
2192 * Multicast packets for forwarding arrive here
2193 */
2194
2195 int ip6_mr_input(struct sk_buff *skb)
2196 {
2197 struct mfc6_cache *cache;
2198 struct net *net = dev_net(skb->dev);
2199 struct mr6_table *mrt;
2200 struct flowi6 fl6 = {
2201 .flowi6_iif = skb->dev->ifindex,
2202 .flowi6_mark = skb->mark,
2203 };
2204 int err;
2205
2206 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2207 if (err < 0) {
2208 kfree_skb(skb);
2209 return err;
2210 }
2211
2212 read_lock(&mrt_lock);
2213 cache = ip6mr_cache_find(mrt,
2214 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2215 if (!cache) {
2216 int vif = ip6mr_find_vif(mrt, skb->dev);
2217
2218 if (vif >= 0)
2219 cache = ip6mr_cache_find_any(mrt,
2220 &ipv6_hdr(skb)->daddr,
2221 vif);
2222 }
2223
2224 /*
2225 * No usable cache entry
2226 */
2227 if (!cache) {
2228 int vif;
2229
2230 vif = ip6mr_find_vif(mrt, skb->dev);
2231 if (vif >= 0) {
2232 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2233 read_unlock(&mrt_lock);
2234
2235 return err;
2236 }
2237 read_unlock(&mrt_lock);
2238 kfree_skb(skb);
2239 return -ENODEV;
2240 }
2241
2242 ip6_mr_forward(net, mrt, skb, cache);
2243
2244 read_unlock(&mrt_lock);
2245
2246 return 0;
2247 }
2248
2249
2250 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2251 struct mfc6_cache *c, struct rtmsg *rtm)
2252 {
2253 struct rta_mfc_stats mfcs;
2254 struct nlattr *mp_attr;
2255 struct rtnexthop *nhp;
2256 unsigned long lastuse;
2257 int ct;
2258
2259 /* If cache is unresolved, don't try to parse IIF and OIF */
2260 if (c->mf6c_parent >= MAXMIFS) {
2261 rtm->rtm_flags |= RTNH_F_UNRESOLVED;
2262 return -ENOENT;
2263 }
2264
2265 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2266 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2267 return -EMSGSIZE;
2268 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2269 if (!mp_attr)
2270 return -EMSGSIZE;
2271
2272 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2273 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2274 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2275 if (!nhp) {
2276 nla_nest_cancel(skb, mp_attr);
2277 return -EMSGSIZE;
2278 }
2279
2280 nhp->rtnh_flags = 0;
2281 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2282 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2283 nhp->rtnh_len = sizeof(*nhp);
2284 }
2285 }
2286
2287 nla_nest_end(skb, mp_attr);
2288
2289 lastuse = READ_ONCE(c->mfc_un.res.lastuse);
2290 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
2291
2292 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2293 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2294 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2295 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
2296 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
2297 RTA_PAD))
2298 return -EMSGSIZE;
2299
2300 rtm->rtm_type = RTN_MULTICAST;
2301 return 1;
2302 }
2303
2304 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2305 u32 portid)
2306 {
2307 int err;
2308 struct mr6_table *mrt;
2309 struct mfc6_cache *cache;
2310 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2311
2312 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2313 if (!mrt)
2314 return -ENOENT;
2315
2316 read_lock(&mrt_lock);
2317 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2318 if (!cache && skb->dev) {
2319 int vif = ip6mr_find_vif(mrt, skb->dev);
2320
2321 if (vif >= 0)
2322 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2323 vif);
2324 }
2325
2326 if (!cache) {
2327 struct sk_buff *skb2;
2328 struct ipv6hdr *iph;
2329 struct net_device *dev;
2330 int vif;
2331
2332 dev = skb->dev;
2333 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2334 read_unlock(&mrt_lock);
2335 return -ENODEV;
2336 }
2337
2338 /* really correct? */
2339 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2340 if (!skb2) {
2341 read_unlock(&mrt_lock);
2342 return -ENOMEM;
2343 }
2344
2345 NETLINK_CB(skb2).portid = portid;
2346 skb_reset_transport_header(skb2);
2347
2348 skb_put(skb2, sizeof(struct ipv6hdr));
2349 skb_reset_network_header(skb2);
2350
2351 iph = ipv6_hdr(skb2);
2352 iph->version = 0;
2353 iph->priority = 0;
2354 iph->flow_lbl[0] = 0;
2355 iph->flow_lbl[1] = 0;
2356 iph->flow_lbl[2] = 0;
2357 iph->payload_len = 0;
2358 iph->nexthdr = IPPROTO_NONE;
2359 iph->hop_limit = 0;
2360 iph->saddr = rt->rt6i_src.addr;
2361 iph->daddr = rt->rt6i_dst.addr;
2362
2363 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2364 read_unlock(&mrt_lock);
2365
2366 return err;
2367 }
2368
2369 if (rtm->rtm_flags & RTM_F_NOTIFY)
2370 cache->mfc_flags |= MFC_NOTIFY;
2371
2372 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2373 read_unlock(&mrt_lock);
2374 return err;
2375 }
2376
2377 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2378 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2379 int flags)
2380 {
2381 struct nlmsghdr *nlh;
2382 struct rtmsg *rtm;
2383 int err;
2384
2385 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2386 if (!nlh)
2387 return -EMSGSIZE;
2388
2389 rtm = nlmsg_data(nlh);
2390 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2391 rtm->rtm_dst_len = 128;
2392 rtm->rtm_src_len = 128;
2393 rtm->rtm_tos = 0;
2394 rtm->rtm_table = mrt->id;
2395 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2396 goto nla_put_failure;
2397 rtm->rtm_type = RTN_MULTICAST;
2398 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2399 if (c->mfc_flags & MFC_STATIC)
2400 rtm->rtm_protocol = RTPROT_STATIC;
2401 else
2402 rtm->rtm_protocol = RTPROT_MROUTED;
2403 rtm->rtm_flags = 0;
2404
2405 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2406 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2407 goto nla_put_failure;
2408 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2409 /* do not break the dump if cache is unresolved */
2410 if (err < 0 && err != -ENOENT)
2411 goto nla_put_failure;
2412
2413 nlmsg_end(skb, nlh);
2414 return 0;
2415
2416 nla_put_failure:
2417 nlmsg_cancel(skb, nlh);
2418 return -EMSGSIZE;
2419 }
2420
2421 static int mr6_msgsize(bool unresolved, int maxvif)
2422 {
2423 size_t len =
2424 NLMSG_ALIGN(sizeof(struct rtmsg))
2425 + nla_total_size(4) /* RTA_TABLE */
2426 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2427 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2428 ;
2429
2430 if (!unresolved)
2431 len = len
2432 + nla_total_size(4) /* RTA_IIF */
2433 + nla_total_size(0) /* RTA_MULTIPATH */
2434 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2435 /* RTA_MFC_STATS */
2436 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2437 ;
2438
2439 return len;
2440 }
2441
2442 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2443 int cmd)
2444 {
2445 struct net *net = read_pnet(&mrt->net);
2446 struct sk_buff *skb;
2447 int err = -ENOBUFS;
2448
2449 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2450 GFP_ATOMIC);
2451 if (!skb)
2452 goto errout;
2453
2454 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2455 if (err < 0)
2456 goto errout;
2457
2458 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2459 return;
2460
2461 errout:
2462 kfree_skb(skb);
2463 if (err < 0)
2464 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2465 }
2466
2467 static size_t mrt6msg_netlink_msgsize(size_t payloadlen)
2468 {
2469 size_t len =
2470 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2471 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */
2472 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */
2473 /* IP6MRA_CREPORT_SRC_ADDR */
2474 + nla_total_size(sizeof(struct in6_addr))
2475 /* IP6MRA_CREPORT_DST_ADDR */
2476 + nla_total_size(sizeof(struct in6_addr))
2477 /* IP6MRA_CREPORT_PKT */
2478 + nla_total_size(payloadlen)
2479 ;
2480
2481 return len;
2482 }
2483
2484 static void mrt6msg_netlink_event(struct mr6_table *mrt, struct sk_buff *pkt)
2485 {
2486 struct net *net = read_pnet(&mrt->net);
2487 struct nlmsghdr *nlh;
2488 struct rtgenmsg *rtgenm;
2489 struct mrt6msg *msg;
2490 struct sk_buff *skb;
2491 struct nlattr *nla;
2492 int payloadlen;
2493
2494 payloadlen = pkt->len - sizeof(struct mrt6msg);
2495 msg = (struct mrt6msg *)skb_transport_header(pkt);
2496
2497 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2498 if (!skb)
2499 goto errout;
2500
2501 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2502 sizeof(struct rtgenmsg), 0);
2503 if (!nlh)
2504 goto errout;
2505 rtgenm = nlmsg_data(nlh);
2506 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR;
2507 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) ||
2508 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) ||
2509 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR,
2510 &msg->im6_src) ||
2511 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR,
2512 &msg->im6_dst))
2513 goto nla_put_failure;
2514
2515 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen);
2516 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg),
2517 nla_data(nla), payloadlen))
2518 goto nla_put_failure;
2519
2520 nlmsg_end(skb, nlh);
2521
2522 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC);
2523 return;
2524
2525 nla_put_failure:
2526 nlmsg_cancel(skb, nlh);
2527 errout:
2528 kfree_skb(skb);
2529 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS);
2530 }
2531
2532 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2533 {
2534 struct net *net = sock_net(skb->sk);
2535 struct mr6_table *mrt;
2536 struct mfc6_cache *mfc;
2537 unsigned int t = 0, s_t;
2538 unsigned int h = 0, s_h;
2539 unsigned int e = 0, s_e;
2540
2541 s_t = cb->args[0];
2542 s_h = cb->args[1];
2543 s_e = cb->args[2];
2544
2545 read_lock(&mrt_lock);
2546 ip6mr_for_each_table(mrt, net) {
2547 if (t < s_t)
2548 goto next_table;
2549 if (t > s_t)
2550 s_h = 0;
2551 for (h = s_h; h < MFC6_LINES; h++) {
2552 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2553 if (e < s_e)
2554 goto next_entry;
2555 if (ip6mr_fill_mroute(mrt, skb,
2556 NETLINK_CB(cb->skb).portid,
2557 cb->nlh->nlmsg_seq,
2558 mfc, RTM_NEWROUTE,
2559 NLM_F_MULTI) < 0)
2560 goto done;
2561 next_entry:
2562 e++;
2563 }
2564 e = s_e = 0;
2565 }
2566 spin_lock_bh(&mfc_unres_lock);
2567 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2568 if (e < s_e)
2569 goto next_entry2;
2570 if (ip6mr_fill_mroute(mrt, skb,
2571 NETLINK_CB(cb->skb).portid,
2572 cb->nlh->nlmsg_seq,
2573 mfc, RTM_NEWROUTE,
2574 NLM_F_MULTI) < 0) {
2575 spin_unlock_bh(&mfc_unres_lock);
2576 goto done;
2577 }
2578 next_entry2:
2579 e++;
2580 }
2581 spin_unlock_bh(&mfc_unres_lock);
2582 e = s_e = 0;
2583 s_h = 0;
2584 next_table:
2585 t++;
2586 }
2587 done:
2588 read_unlock(&mrt_lock);
2589
2590 cb->args[2] = e;
2591 cb->args[1] = h;
2592 cb->args[0] = t;
2593
2594 return skb->len;
2595 }
2596
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.