?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
MIPS: ath79: Use ath79_get_sys_clk_rate to get basic clock rates
[linux-2.6/btrfs-unstable.git] / net / ipv6 / ip6mr.c
blob03986d31fa417a357b72e529e28fea0de0294a9e
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 <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>
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 #ifdef CONFIG_NET_NS
60 struct net *net;
61 #endif
62 u32 id;
63 struct sock *mroute6_sk;
64 struct timer_list ipmr_expire_timer;
65 struct list_head mfc6_unres_queue;
66 struct list_head mfc6_cache_array[MFC6_LINES];
67 struct mif_device vif6_table[MAXMIFS];
68 int maxvif;
69 atomic_t cache_resolve_queue_len;
70 bool mroute_do_assert;
71 bool mroute_do_pim;
72 #ifdef CONFIG_IPV6_PIMSM_V2
73 int mroute_reg_vif_num;
74 #endif
75 };
76
77 struct ip6mr_rule {
78 struct fib_rule common;
79 };
80
81 struct ip6mr_result {
82 struct mr6_table *mrt;
83 };
84
85 /* Big lock, protecting vif table, mrt cache and mroute socket state.
86 Note that the changes are semaphored via rtnl_lock.
87 */
88
89 static DEFINE_RWLOCK(mrt_lock);
90
91 /*
92 * Multicast router control variables
93 */
94
95 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
96
97 /* Special spinlock for queue of unresolved entries */
98 static DEFINE_SPINLOCK(mfc_unres_lock);
99
100 /* We return to original Alan's scheme. Hash table of resolved
101 entries is changed only in process context and protected
102 with weak lock mrt_lock. Queue of unresolved entries is protected
103 with strong spinlock mfc_unres_lock.
104
105 In this case data path is free of exclusive locks at all.
106 */
107
108 static struct kmem_cache *mrt_cachep __read_mostly;
109
110 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
111 static void ip6mr_free_table(struct mr6_table *mrt);
112
113 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
114 struct sk_buff *skb, struct mfc6_cache *cache);
115 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
116 mifi_t mifi, int assert);
117 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
118 struct mfc6_cache *c, struct rtmsg *rtm);
119 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
120 int cmd);
121 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
122 struct netlink_callback *cb);
123 static void mroute_clean_tables(struct mr6_table *mrt);
124 static void ipmr_expire_process(unsigned long arg);
125
126 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
127 #define ip6mr_for_each_table(mrt, net) \
128 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
129
130 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
131 {
132 struct mr6_table *mrt;
133
134 ip6mr_for_each_table(mrt, net) {
135 if (mrt->id == id)
136 return mrt;
137 }
138 return NULL;
139 }
140
141 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
142 struct mr6_table **mrt)
143 {
144 struct ip6mr_result res;
145 struct fib_lookup_arg arg = { .result = &res, };
146 int err;
147
148 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
149 flowi6_to_flowi(flp6), 0, &arg);
150 if (err < 0)
151 return err;
152 *mrt = res.mrt;
153 return 0;
154 }
155
156 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
157 int flags, struct fib_lookup_arg *arg)
158 {
159 struct ip6mr_result *res = arg->result;
160 struct mr6_table *mrt;
161
162 switch (rule->action) {
163 case FR_ACT_TO_TBL:
164 break;
165 case FR_ACT_UNREACHABLE:
166 return -ENETUNREACH;
167 case FR_ACT_PROHIBIT:
168 return -EACCES;
169 case FR_ACT_BLACKHOLE:
170 default:
171 return -EINVAL;
172 }
173
174 mrt = ip6mr_get_table(rule->fr_net, rule->table);
175 if (mrt == NULL)
176 return -EAGAIN;
177 res->mrt = mrt;
178 return 0;
179 }
180
181 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
182 {
183 return 1;
184 }
185
186 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
187 FRA_GENERIC_POLICY,
188 };
189
190 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
191 struct fib_rule_hdr *frh, struct nlattr **tb)
192 {
193 return 0;
194 }
195
196 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
197 struct nlattr **tb)
198 {
199 return 1;
200 }
201
202 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
203 struct fib_rule_hdr *frh)
204 {
205 frh->dst_len = 0;
206 frh->src_len = 0;
207 frh->tos = 0;
208 return 0;
209 }
210
211 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
212 .family = RTNL_FAMILY_IP6MR,
213 .rule_size = sizeof(struct ip6mr_rule),
214 .addr_size = sizeof(struct in6_addr),
215 .action = ip6mr_rule_action,
216 .match = ip6mr_rule_match,
217 .configure = ip6mr_rule_configure,
218 .compare = ip6mr_rule_compare,
219 .default_pref = fib_default_rule_pref,
220 .fill = ip6mr_rule_fill,
221 .nlgroup = RTNLGRP_IPV6_RULE,
222 .policy = ip6mr_rule_policy,
223 .owner = THIS_MODULE,
224 };
225
226 static int __net_init ip6mr_rules_init(struct net *net)
227 {
228 struct fib_rules_ops *ops;
229 struct mr6_table *mrt;
230 int err;
231
232 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
233 if (IS_ERR(ops))
234 return PTR_ERR(ops);
235
236 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
237
238 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
239 if (mrt == NULL) {
240 err = -ENOMEM;
241 goto err1;
242 }
243
244 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
245 if (err < 0)
246 goto err2;
247
248 net->ipv6.mr6_rules_ops = ops;
249 return 0;
250
251 err2:
252 kfree(mrt);
253 err1:
254 fib_rules_unregister(ops);
255 return err;
256 }
257
258 static void __net_exit ip6mr_rules_exit(struct net *net)
259 {
260 struct mr6_table *mrt, *next;
261
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);
266 }
267 rtnl_unlock();
268 fib_rules_unregister(net->ipv6.mr6_rules_ops);
269 }
270 #else
271 #define ip6mr_for_each_table(mrt, net) \
272 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
273
274 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
275 {
276 return net->ipv6.mrt6;
277 }
278
279 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
280 struct mr6_table **mrt)
281 {
282 *mrt = net->ipv6.mrt6;
283 return 0;
284 }
285
286 static int __net_init ip6mr_rules_init(struct net *net)
287 {
288 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
289 return net->ipv6.mrt6 ? 0 : -ENOMEM;
290 }
291
292 static void __net_exit ip6mr_rules_exit(struct net *net)
293 {
294 rtnl_lock();
295 ip6mr_free_table(net->ipv6.mrt6);
296 net->ipv6.mrt6 = NULL;
297 rtnl_unlock();
298 }
299 #endif
300
301 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
302 {
303 struct mr6_table *mrt;
304 unsigned int i;
305
306 mrt = ip6mr_get_table(net, id);
307 if (mrt != NULL)
308 return mrt;
309
310 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
311 if (mrt == NULL)
312 return NULL;
313 mrt->id = id;
314 write_pnet(&mrt->net, net);
315
316 /* Forwarding cache */
317 for (i = 0; i < MFC6_LINES; i++)
318 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
319
320 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
321
322 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
323 (unsigned long)mrt);
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(&mrt->ipmr_expire_timer);
337 mroute_clean_tables(mrt);
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 == NULL)
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 == NULL)
495 return ERR_PTR(-ENOENT);
496
497 it->mrt = mrt;
498 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
499 : SEQ_START_TOKEN;
500 }
501
502 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
503 {
504 struct mfc6_cache *mfc = v;
505 struct ipmr_mfc_iter *it = seq->private;
506 struct net *net = seq_file_net(seq);
507 struct mr6_table *mrt = it->mrt;
508
509 ++*pos;
510
511 if (v == SEQ_START_TOKEN)
512 return ipmr_mfc_seq_idx(net, seq->private, 0);
513
514 if (mfc->list.next != it->cache)
515 return list_entry(mfc->list.next, struct mfc6_cache, list);
516
517 if (it->cache == &mrt->mfc6_unres_queue)
518 goto end_of_list;
519
520 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
521
522 while (++it->ct < MFC6_LINES) {
523 it->cache = &mrt->mfc6_cache_array[it->ct];
524 if (list_empty(it->cache))
525 continue;
526 return list_first_entry(it->cache, struct mfc6_cache, list);
527 }
528
529 /* exhausted cache_array, show unresolved */
530 read_unlock(&mrt_lock);
531 it->cache = &mrt->mfc6_unres_queue;
532 it->ct = 0;
533
534 spin_lock_bh(&mfc_unres_lock);
535 if (!list_empty(it->cache))
536 return list_first_entry(it->cache, struct mfc6_cache, list);
537
538 end_of_list:
539 spin_unlock_bh(&mfc_unres_lock);
540 it->cache = NULL;
541
542 return NULL;
543 }
544
545 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
546 {
547 struct ipmr_mfc_iter *it = seq->private;
548 struct mr6_table *mrt = it->mrt;
549
550 if (it->cache == &mrt->mfc6_unres_queue)
551 spin_unlock_bh(&mfc_unres_lock);
552 else if (it->cache == mrt->mfc6_cache_array)
553 read_unlock(&mrt_lock);
554 }
555
556 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
557 {
558 int n;
559
560 if (v == SEQ_START_TOKEN) {
561 seq_puts(seq,
562 "Group "
563 "Origin "
564 "Iif Pkts Bytes Wrong Oifs\n");
565 } else {
566 const struct mfc6_cache *mfc = v;
567 const struct ipmr_mfc_iter *it = seq->private;
568 struct mr6_table *mrt = it->mrt;
569
570 seq_printf(seq, "%pI6 %pI6 %-3hd",
571 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
572 mfc->mf6c_parent);
573
574 if (it->cache != &mrt->mfc6_unres_queue) {
575 seq_printf(seq, " %8lu %8lu %8lu",
576 mfc->mfc_un.res.pkt,
577 mfc->mfc_un.res.bytes,
578 mfc->mfc_un.res.wrong_if);
579 for (n = mfc->mfc_un.res.minvif;
580 n < mfc->mfc_un.res.maxvif; n++) {
581 if (MIF_EXISTS(mrt, n) &&
582 mfc->mfc_un.res.ttls[n] < 255)
583 seq_printf(seq,
584 " %2d:%-3d",
585 n, mfc->mfc_un.res.ttls[n]);
586 }
587 } else {
588 /* unresolved mfc_caches don't contain
589 * pkt, bytes and wrong_if values
590 */
591 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
592 }
593 seq_putc(seq, '\n');
594 }
595 return 0;
596 }
597
598 static const struct seq_operations ipmr_mfc_seq_ops = {
599 .start = ipmr_mfc_seq_start,
600 .next = ipmr_mfc_seq_next,
601 .stop = ipmr_mfc_seq_stop,
602 .show = ipmr_mfc_seq_show,
603 };
604
605 static int ipmr_mfc_open(struct inode *inode, struct file *file)
606 {
607 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
608 sizeof(struct ipmr_mfc_iter));
609 }
610
611 static const struct file_operations ip6mr_mfc_fops = {
612 .owner = THIS_MODULE,
613 .open = ipmr_mfc_open,
614 .read = seq_read,
615 .llseek = seq_lseek,
616 .release = seq_release_net,
617 };
618 #endif
619
620 #ifdef CONFIG_IPV6_PIMSM_V2
621
622 static int pim6_rcv(struct sk_buff *skb)
623 {
624 struct pimreghdr *pim;
625 struct ipv6hdr *encap;
626 struct net_device *reg_dev = NULL;
627 struct net *net = dev_net(skb->dev);
628 struct mr6_table *mrt;
629 struct flowi6 fl6 = {
630 .flowi6_iif = skb->dev->ifindex,
631 .flowi6_mark = skb->mark,
632 };
633 int reg_vif_num;
634
635 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
636 goto drop;
637
638 pim = (struct pimreghdr *)skb_transport_header(skb);
639 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
640 (pim->flags & PIM_NULL_REGISTER) ||
641 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
642 sizeof(*pim), IPPROTO_PIM,
643 csum_partial((void *)pim, sizeof(*pim), 0)) &&
644 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
645 goto drop;
646
647 /* check if the inner packet is destined to mcast group */
648 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
649 sizeof(*pim));
650
651 if (!ipv6_addr_is_multicast(&encap->daddr) ||
652 encap->payload_len == 0 ||
653 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
654 goto drop;
655
656 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
657 goto drop;
658 reg_vif_num = mrt->mroute_reg_vif_num;
659
660 read_lock(&mrt_lock);
661 if (reg_vif_num >= 0)
662 reg_dev = mrt->vif6_table[reg_vif_num].dev;
663 if (reg_dev)
664 dev_hold(reg_dev);
665 read_unlock(&mrt_lock);
666
667 if (reg_dev == NULL)
668 goto drop;
669
670 skb->mac_header = skb->network_header;
671 skb_pull(skb, (u8 *)encap - skb->data);
672 skb_reset_network_header(skb);
673 skb->protocol = htons(ETH_P_IPV6);
674 skb->ip_summed = CHECKSUM_NONE;
675 skb->pkt_type = PACKET_HOST;
676
677 skb_tunnel_rx(skb, 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,
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 const struct net_device_ops reg_vif_netdev_ops = {
722 .ndo_start_xmit = reg_vif_xmit,
723 };
724
725 static void reg_vif_setup(struct net_device *dev)
726 {
727 dev->type = ARPHRD_PIMREG;
728 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
729 dev->flags = IFF_NOARP;
730 dev->netdev_ops = &reg_vif_netdev_ops;
731 dev->destructor = free_netdev;
732 dev->features |= NETIF_F_NETNS_LOCAL;
733 }
734
735 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
736 {
737 struct net_device *dev;
738 char name[IFNAMSIZ];
739
740 if (mrt->id == RT6_TABLE_DFLT)
741 sprintf(name, "pim6reg");
742 else
743 sprintf(name, "pim6reg%u", mrt->id);
744
745 dev = alloc_netdev(0, name, reg_vif_setup);
746 if (dev == NULL)
747 return NULL;
748
749 dev_net_set(dev, net);
750
751 if (register_netdevice(dev)) {
752 free_netdev(dev);
753 return NULL;
754 }
755 dev->iflink = 0;
756
757 if (dev_open(dev))
758 goto failure;
759
760 dev_hold(dev);
761 return dev;
762
763 failure:
764 /* allow the register to be completed before unregistering. */
765 rtnl_unlock();
766 rtnl_lock();
767
768 unregister_netdevice(dev);
769 return NULL;
770 }
771 #endif
772
773 /*
774 * Delete a VIF entry
775 */
776
777 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
778 {
779 struct mif_device *v;
780 struct net_device *dev;
781 struct inet6_dev *in6_dev;
782
783 if (vifi < 0 || vifi >= mrt->maxvif)
784 return -EADDRNOTAVAIL;
785
786 v = &mrt->vif6_table[vifi];
787
788 write_lock_bh(&mrt_lock);
789 dev = v->dev;
790 v->dev = NULL;
791
792 if (!dev) {
793 write_unlock_bh(&mrt_lock);
794 return -EADDRNOTAVAIL;
795 }
796
797 #ifdef CONFIG_IPV6_PIMSM_V2
798 if (vifi == mrt->mroute_reg_vif_num)
799 mrt->mroute_reg_vif_num = -1;
800 #endif
801
802 if (vifi + 1 == mrt->maxvif) {
803 int tmp;
804 for (tmp = vifi - 1; tmp >= 0; tmp--) {
805 if (MIF_EXISTS(mrt, tmp))
806 break;
807 }
808 mrt->maxvif = tmp + 1;
809 }
810
811 write_unlock_bh(&mrt_lock);
812
813 dev_set_allmulti(dev, -1);
814
815 in6_dev = __in6_dev_get(dev);
816 if (in6_dev) {
817 in6_dev->cnf.mc_forwarding--;
818 inet6_netconf_notify_devconf(dev_net(dev),
819 NETCONFA_MC_FORWARDING,
820 dev->ifindex, &in6_dev->cnf);
821 }
822
823 if (v->flags & MIFF_REGISTER)
824 unregister_netdevice_queue(dev, head);
825
826 dev_put(dev);
827 return 0;
828 }
829
830 static inline void ip6mr_cache_free(struct mfc6_cache *c)
831 {
832 kmem_cache_free(mrt_cachep, c);
833 }
834
835 /* Destroy an unresolved cache entry, killing queued skbs
836 and reporting error to netlink readers.
837 */
838
839 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
840 {
841 struct net *net = read_pnet(&mrt->net);
842 struct sk_buff *skb;
843
844 atomic_dec(&mrt->cache_resolve_queue_len);
845
846 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
847 if (ipv6_hdr(skb)->version == 0) {
848 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
849 nlh->nlmsg_type = NLMSG_ERROR;
850 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
851 skb_trim(skb, nlh->nlmsg_len);
852 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
853 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
854 } else
855 kfree_skb(skb);
856 }
857
858 ip6mr_cache_free(c);
859 }
860
861
862 /* Timer process for all the unresolved queue. */
863
864 static void ipmr_do_expire_process(struct mr6_table *mrt)
865 {
866 unsigned long now = jiffies;
867 unsigned long expires = 10 * HZ;
868 struct mfc6_cache *c, *next;
869
870 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
871 if (time_after(c->mfc_un.unres.expires, now)) {
872 /* not yet... */
873 unsigned long interval = c->mfc_un.unres.expires - now;
874 if (interval < expires)
875 expires = interval;
876 continue;
877 }
878
879 list_del(&c->list);
880 mr6_netlink_event(mrt, c, RTM_DELROUTE);
881 ip6mr_destroy_unres(mrt, c);
882 }
883
884 if (!list_empty(&mrt->mfc6_unres_queue))
885 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
886 }
887
888 static void ipmr_expire_process(unsigned long arg)
889 {
890 struct mr6_table *mrt = (struct mr6_table *)arg;
891
892 if (!spin_trylock(&mfc_unres_lock)) {
893 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
894 return;
895 }
896
897 if (!list_empty(&mrt->mfc6_unres_queue))
898 ipmr_do_expire_process(mrt);
899
900 spin_unlock(&mfc_unres_lock);
901 }
902
903 /* Fill oifs list. It is called under write locked mrt_lock. */
904
905 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
906 unsigned char *ttls)
907 {
908 int vifi;
909
910 cache->mfc_un.res.minvif = MAXMIFS;
911 cache->mfc_un.res.maxvif = 0;
912 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
913
914 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
915 if (MIF_EXISTS(mrt, vifi) &&
916 ttls[vifi] && ttls[vifi] < 255) {
917 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
918 if (cache->mfc_un.res.minvif > vifi)
919 cache->mfc_un.res.minvif = vifi;
920 if (cache->mfc_un.res.maxvif <= vifi)
921 cache->mfc_un.res.maxvif = vifi + 1;
922 }
923 }
924 }
925
926 static int mif6_add(struct net *net, struct mr6_table *mrt,
927 struct mif6ctl *vifc, int mrtsock)
928 {
929 int vifi = vifc->mif6c_mifi;
930 struct mif_device *v = &mrt->vif6_table[vifi];
931 struct net_device *dev;
932 struct inet6_dev *in6_dev;
933 int err;
934
935 /* Is vif busy ? */
936 if (MIF_EXISTS(mrt, vifi))
937 return -EADDRINUSE;
938
939 switch (vifc->mif6c_flags) {
940 #ifdef CONFIG_IPV6_PIMSM_V2
941 case MIFF_REGISTER:
942 /*
943 * Special Purpose VIF in PIM
944 * All the packets will be sent to the daemon
945 */
946 if (mrt->mroute_reg_vif_num >= 0)
947 return -EADDRINUSE;
948 dev = ip6mr_reg_vif(net, mrt);
949 if (!dev)
950 return -ENOBUFS;
951 err = dev_set_allmulti(dev, 1);
952 if (err) {
953 unregister_netdevice(dev);
954 dev_put(dev);
955 return err;
956 }
957 break;
958 #endif
959 case 0:
960 dev = dev_get_by_index(net, vifc->mif6c_pifi);
961 if (!dev)
962 return -EADDRNOTAVAIL;
963 err = dev_set_allmulti(dev, 1);
964 if (err) {
965 dev_put(dev);
966 return err;
967 }
968 break;
969 default:
970 return -EINVAL;
971 }
972
973 in6_dev = __in6_dev_get(dev);
974 if (in6_dev) {
975 in6_dev->cnf.mc_forwarding++;
976 inet6_netconf_notify_devconf(dev_net(dev),
977 NETCONFA_MC_FORWARDING,
978 dev->ifindex, &in6_dev->cnf);
979 }
980
981 /*
982 * Fill in the VIF structures
983 */
984 v->rate_limit = vifc->vifc_rate_limit;
985 v->flags = vifc->mif6c_flags;
986 if (!mrtsock)
987 v->flags |= VIFF_STATIC;
988 v->threshold = vifc->vifc_threshold;
989 v->bytes_in = 0;
990 v->bytes_out = 0;
991 v->pkt_in = 0;
992 v->pkt_out = 0;
993 v->link = dev->ifindex;
994 if (v->flags & MIFF_REGISTER)
995 v->link = dev->iflink;
996
997 /* And finish update writing critical data */
998 write_lock_bh(&mrt_lock);
999 v->dev = dev;
1000 #ifdef CONFIG_IPV6_PIMSM_V2
1001 if (v->flags & MIFF_REGISTER)
1002 mrt->mroute_reg_vif_num = vifi;
1003 #endif
1004 if (vifi + 1 > mrt->maxvif)
1005 mrt->maxvif = vifi + 1;
1006 write_unlock_bh(&mrt_lock);
1007 return 0;
1008 }
1009
1010 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1011 const struct in6_addr *origin,
1012 const struct in6_addr *mcastgrp)
1013 {
1014 int line = MFC6_HASH(mcastgrp, origin);
1015 struct mfc6_cache *c;
1016
1017 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1018 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1019 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1020 return c;
1021 }
1022 return NULL;
1023 }
1024
1025 /* Look for a (*,*,oif) entry */
1026 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1027 mifi_t mifi)
1028 {
1029 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1030 struct mfc6_cache *c;
1031
1032 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1033 if (ipv6_addr_any(&c->mf6c_origin) &&
1034 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1035 (c->mfc_un.res.ttls[mifi] < 255))
1036 return c;
1037
1038 return NULL;
1039 }
1040
1041 /* Look for a (*,G) entry */
1042 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1043 struct in6_addr *mcastgrp,
1044 mifi_t mifi)
1045 {
1046 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1047 struct mfc6_cache *c, *proxy;
1048
1049 if (ipv6_addr_any(mcastgrp))
1050 goto skip;
1051
1052 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1053 if (ipv6_addr_any(&c->mf6c_origin) &&
1054 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1055 if (c->mfc_un.res.ttls[mifi] < 255)
1056 return c;
1057
1058 /* It's ok if the mifi is part of the static tree */
1059 proxy = ip6mr_cache_find_any_parent(mrt,
1060 c->mf6c_parent);
1061 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1062 return c;
1063 }
1064
1065 skip:
1066 return ip6mr_cache_find_any_parent(mrt, mifi);
1067 }
1068
1069 /*
1070 * Allocate a multicast cache entry
1071 */
1072 static struct mfc6_cache *ip6mr_cache_alloc(void)
1073 {
1074 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1075 if (c == NULL)
1076 return NULL;
1077 c->mfc_un.res.minvif = MAXMIFS;
1078 return c;
1079 }
1080
1081 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1082 {
1083 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1084 if (c == NULL)
1085 return NULL;
1086 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1087 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1088 return c;
1089 }
1090
1091 /*
1092 * A cache entry has gone into a resolved state from queued
1093 */
1094
1095 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1096 struct mfc6_cache *uc, struct mfc6_cache *c)
1097 {
1098 struct sk_buff *skb;
1099
1100 /*
1101 * Play the pending entries through our router
1102 */
1103
1104 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1105 if (ipv6_hdr(skb)->version == 0) {
1106 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1107
1108 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1109 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1110 } else {
1111 nlh->nlmsg_type = NLMSG_ERROR;
1112 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1113 skb_trim(skb, nlh->nlmsg_len);
1114 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1115 }
1116 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1117 } else
1118 ip6_mr_forward(net, mrt, skb, c);
1119 }
1120 }
1121
1122 /*
1123 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1124 * expects the following bizarre scheme.
1125 *
1126 * Called under mrt_lock.
1127 */
1128
1129 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1130 mifi_t mifi, int assert)
1131 {
1132 struct sk_buff *skb;
1133 struct mrt6msg *msg;
1134 int ret;
1135
1136 #ifdef CONFIG_IPV6_PIMSM_V2
1137 if (assert == MRT6MSG_WHOLEPKT)
1138 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1139 +sizeof(*msg));
1140 else
1141 #endif
1142 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1143
1144 if (!skb)
1145 return -ENOBUFS;
1146
1147 /* I suppose that internal messages
1148 * do not require checksums */
1149
1150 skb->ip_summed = CHECKSUM_UNNECESSARY;
1151
1152 #ifdef CONFIG_IPV6_PIMSM_V2
1153 if (assert == MRT6MSG_WHOLEPKT) {
1154 /* Ugly, but we have no choice with this interface.
1155 Duplicate old header, fix length etc.
1156 And all this only to mangle msg->im6_msgtype and
1157 to set msg->im6_mbz to "mbz" :-)
1158 */
1159 skb_push(skb, -skb_network_offset(pkt));
1160
1161 skb_push(skb, sizeof(*msg));
1162 skb_reset_transport_header(skb);
1163 msg = (struct mrt6msg *)skb_transport_header(skb);
1164 msg->im6_mbz = 0;
1165 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1166 msg->im6_mif = mrt->mroute_reg_vif_num;
1167 msg->im6_pad = 0;
1168 msg->im6_src = ipv6_hdr(pkt)->saddr;
1169 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1170
1171 skb->ip_summed = CHECKSUM_UNNECESSARY;
1172 } else
1173 #endif
1174 {
1175 /*
1176 * Copy the IP header
1177 */
1178
1179 skb_put(skb, sizeof(struct ipv6hdr));
1180 skb_reset_network_header(skb);
1181 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1182
1183 /*
1184 * Add our header
1185 */
1186 skb_put(skb, sizeof(*msg));
1187 skb_reset_transport_header(skb);
1188 msg = (struct mrt6msg *)skb_transport_header(skb);
1189
1190 msg->im6_mbz = 0;
1191 msg->im6_msgtype = assert;
1192 msg->im6_mif = mifi;
1193 msg->im6_pad = 0;
1194 msg->im6_src = ipv6_hdr(pkt)->saddr;
1195 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1196
1197 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1198 skb->ip_summed = CHECKSUM_UNNECESSARY;
1199 }
1200
1201 if (mrt->mroute6_sk == NULL) {
1202 kfree_skb(skb);
1203 return -EINVAL;
1204 }
1205
1206 /*
1207 * Deliver to user space multicast routing algorithms
1208 */
1209 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1210 if (ret < 0) {
1211 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1212 kfree_skb(skb);
1213 }
1214
1215 return ret;
1216 }
1217
1218 /*
1219 * Queue a packet for resolution. It gets locked cache entry!
1220 */
1221
1222 static int
1223 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1224 {
1225 bool found = false;
1226 int err;
1227 struct mfc6_cache *c;
1228
1229 spin_lock_bh(&mfc_unres_lock);
1230 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1231 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1232 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1233 found = true;
1234 break;
1235 }
1236 }
1237
1238 if (!found) {
1239 /*
1240 * Create a new entry if allowable
1241 */
1242
1243 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1244 (c = ip6mr_cache_alloc_unres()) == NULL) {
1245 spin_unlock_bh(&mfc_unres_lock);
1246
1247 kfree_skb(skb);
1248 return -ENOBUFS;
1249 }
1250
1251 /*
1252 * Fill in the new cache entry
1253 */
1254 c->mf6c_parent = -1;
1255 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1256 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1257
1258 /*
1259 * Reflect first query at pim6sd
1260 */
1261 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1262 if (err < 0) {
1263 /* If the report failed throw the cache entry
1264 out - Brad Parker
1265 */
1266 spin_unlock_bh(&mfc_unres_lock);
1267
1268 ip6mr_cache_free(c);
1269 kfree_skb(skb);
1270 return err;
1271 }
1272
1273 atomic_inc(&mrt->cache_resolve_queue_len);
1274 list_add(&c->list, &mrt->mfc6_unres_queue);
1275 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1276
1277 ipmr_do_expire_process(mrt);
1278 }
1279
1280 /*
1281 * See if we can append the packet
1282 */
1283 if (c->mfc_un.unres.unresolved.qlen > 3) {
1284 kfree_skb(skb);
1285 err = -ENOBUFS;
1286 } else {
1287 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1288 err = 0;
1289 }
1290
1291 spin_unlock_bh(&mfc_unres_lock);
1292 return err;
1293 }
1294
1295 /*
1296 * MFC6 cache manipulation by user space
1297 */
1298
1299 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1300 int parent)
1301 {
1302 int line;
1303 struct mfc6_cache *c, *next;
1304
1305 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1306
1307 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1308 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1309 ipv6_addr_equal(&c->mf6c_mcastgrp,
1310 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1311 (parent == -1 || parent == c->mf6c_parent)) {
1312 write_lock_bh(&mrt_lock);
1313 list_del(&c->list);
1314 write_unlock_bh(&mrt_lock);
1315
1316 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1317 ip6mr_cache_free(c);
1318 return 0;
1319 }
1320 }
1321 return -ENOENT;
1322 }
1323
1324 static int ip6mr_device_event(struct notifier_block *this,
1325 unsigned long event, void *ptr)
1326 {
1327 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1328 struct net *net = dev_net(dev);
1329 struct mr6_table *mrt;
1330 struct mif_device *v;
1331 int ct;
1332 LIST_HEAD(list);
1333
1334 if (event != NETDEV_UNREGISTER)
1335 return NOTIFY_DONE;
1336
1337 ip6mr_for_each_table(mrt, net) {
1338 v = &mrt->vif6_table[0];
1339 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1340 if (v->dev == dev)
1341 mif6_delete(mrt, ct, &list);
1342 }
1343 }
1344 unregister_netdevice_many(&list);
1345
1346 return NOTIFY_DONE;
1347 }
1348
1349 static struct notifier_block ip6_mr_notifier = {
1350 .notifier_call = ip6mr_device_event
1351 };
1352
1353 /*
1354 * Setup for IP multicast routing
1355 */
1356
1357 static int __net_init ip6mr_net_init(struct net *net)
1358 {
1359 int err;
1360
1361 err = ip6mr_rules_init(net);
1362 if (err < 0)
1363 goto fail;
1364
1365 #ifdef CONFIG_PROC_FS
1366 err = -ENOMEM;
1367 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1368 goto proc_vif_fail;
1369 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1370 goto proc_cache_fail;
1371 #endif
1372
1373 return 0;
1374
1375 #ifdef CONFIG_PROC_FS
1376 proc_cache_fail:
1377 remove_proc_entry("ip6_mr_vif", net->proc_net);
1378 proc_vif_fail:
1379 ip6mr_rules_exit(net);
1380 #endif
1381 fail:
1382 return err;
1383 }
1384
1385 static void __net_exit ip6mr_net_exit(struct net *net)
1386 {
1387 #ifdef CONFIG_PROC_FS
1388 remove_proc_entry("ip6_mr_cache", net->proc_net);
1389 remove_proc_entry("ip6_mr_vif", net->proc_net);
1390 #endif
1391 ip6mr_rules_exit(net);
1392 }
1393
1394 static struct pernet_operations ip6mr_net_ops = {
1395 .init = ip6mr_net_init,
1396 .exit = ip6mr_net_exit,
1397 };
1398
1399 int __init ip6_mr_init(void)
1400 {
1401 int err;
1402
1403 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1404 sizeof(struct mfc6_cache),
1405 0, SLAB_HWCACHE_ALIGN,
1406 NULL);
1407 if (!mrt_cachep)
1408 return -ENOMEM;
1409
1410 err = register_pernet_subsys(&ip6mr_net_ops);
1411 if (err)
1412 goto reg_pernet_fail;
1413
1414 err = register_netdevice_notifier(&ip6_mr_notifier);
1415 if (err)
1416 goto reg_notif_fail;
1417 #ifdef CONFIG_IPV6_PIMSM_V2
1418 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1419 pr_err("%s: can't add PIM protocol\n", __func__);
1420 err = -EAGAIN;
1421 goto add_proto_fail;
1422 }
1423 #endif
1424 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1425 ip6mr_rtm_dumproute, NULL);
1426 return 0;
1427 #ifdef CONFIG_IPV6_PIMSM_V2
1428 add_proto_fail:
1429 unregister_netdevice_notifier(&ip6_mr_notifier);
1430 #endif
1431 reg_notif_fail:
1432 unregister_pernet_subsys(&ip6mr_net_ops);
1433 reg_pernet_fail:
1434 kmem_cache_destroy(mrt_cachep);
1435 return err;
1436 }
1437
1438 void ip6_mr_cleanup(void)
1439 {
1440 unregister_netdevice_notifier(&ip6_mr_notifier);
1441 unregister_pernet_subsys(&ip6mr_net_ops);
1442 kmem_cache_destroy(mrt_cachep);
1443 }
1444
1445 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1446 struct mf6cctl *mfc, int mrtsock, int parent)
1447 {
1448 bool found = false;
1449 int line;
1450 struct mfc6_cache *uc, *c;
1451 unsigned char ttls[MAXMIFS];
1452 int i;
1453
1454 if (mfc->mf6cc_parent >= MAXMIFS)
1455 return -ENFILE;
1456
1457 memset(ttls, 255, MAXMIFS);
1458 for (i = 0; i < MAXMIFS; i++) {
1459 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1460 ttls[i] = 1;
1461
1462 }
1463
1464 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1465
1466 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1467 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1468 ipv6_addr_equal(&c->mf6c_mcastgrp,
1469 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1470 (parent == -1 || parent == mfc->mf6cc_parent)) {
1471 found = true;
1472 break;
1473 }
1474 }
1475
1476 if (found) {
1477 write_lock_bh(&mrt_lock);
1478 c->mf6c_parent = mfc->mf6cc_parent;
1479 ip6mr_update_thresholds(mrt, c, ttls);
1480 if (!mrtsock)
1481 c->mfc_flags |= MFC_STATIC;
1482 write_unlock_bh(&mrt_lock);
1483 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1484 return 0;
1485 }
1486
1487 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1488 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1489 return -EINVAL;
1490
1491 c = ip6mr_cache_alloc();
1492 if (c == NULL)
1493 return -ENOMEM;
1494
1495 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1496 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1497 c->mf6c_parent = mfc->mf6cc_parent;
1498 ip6mr_update_thresholds(mrt, c, ttls);
1499 if (!mrtsock)
1500 c->mfc_flags |= MFC_STATIC;
1501
1502 write_lock_bh(&mrt_lock);
1503 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1504 write_unlock_bh(&mrt_lock);
1505
1506 /*
1507 * Check to see if we resolved a queued list. If so we
1508 * need to send on the frames and tidy up.
1509 */
1510 found = false;
1511 spin_lock_bh(&mfc_unres_lock);
1512 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1513 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1514 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1515 list_del(&uc->list);
1516 atomic_dec(&mrt->cache_resolve_queue_len);
1517 found = true;
1518 break;
1519 }
1520 }
1521 if (list_empty(&mrt->mfc6_unres_queue))
1522 del_timer(&mrt->ipmr_expire_timer);
1523 spin_unlock_bh(&mfc_unres_lock);
1524
1525 if (found) {
1526 ip6mr_cache_resolve(net, mrt, uc, c);
1527 ip6mr_cache_free(uc);
1528 }
1529 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1530 return 0;
1531 }
1532
1533 /*
1534 * Close the multicast socket, and clear the vif tables etc
1535 */
1536
1537 static void mroute_clean_tables(struct mr6_table *mrt)
1538 {
1539 int i;
1540 LIST_HEAD(list);
1541 struct mfc6_cache *c, *next;
1542
1543 /*
1544 * Shut down all active vif entries
1545 */
1546 for (i = 0; i < mrt->maxvif; i++) {
1547 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1548 mif6_delete(mrt, i, &list);
1549 }
1550 unregister_netdevice_many(&list);
1551
1552 /*
1553 * Wipe the cache
1554 */
1555 for (i = 0; i < MFC6_LINES; i++) {
1556 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1557 if (c->mfc_flags & MFC_STATIC)
1558 continue;
1559 write_lock_bh(&mrt_lock);
1560 list_del(&c->list);
1561 write_unlock_bh(&mrt_lock);
1562
1563 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1564 ip6mr_cache_free(c);
1565 }
1566 }
1567
1568 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1569 spin_lock_bh(&mfc_unres_lock);
1570 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1571 list_del(&c->list);
1572 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1573 ip6mr_destroy_unres(mrt, c);
1574 }
1575 spin_unlock_bh(&mfc_unres_lock);
1576 }
1577 }
1578
1579 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1580 {
1581 int err = 0;
1582 struct net *net = sock_net(sk);
1583
1584 rtnl_lock();
1585 write_lock_bh(&mrt_lock);
1586 if (likely(mrt->mroute6_sk == NULL)) {
1587 mrt->mroute6_sk = sk;
1588 net->ipv6.devconf_all->mc_forwarding++;
1589 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1590 NETCONFA_IFINDEX_ALL,
1591 net->ipv6.devconf_all);
1592 }
1593 else
1594 err = -EADDRINUSE;
1595 write_unlock_bh(&mrt_lock);
1596
1597 rtnl_unlock();
1598
1599 return err;
1600 }
1601
1602 int ip6mr_sk_done(struct sock *sk)
1603 {
1604 int err = -EACCES;
1605 struct net *net = sock_net(sk);
1606 struct mr6_table *mrt;
1607
1608 rtnl_lock();
1609 ip6mr_for_each_table(mrt, net) {
1610 if (sk == mrt->mroute6_sk) {
1611 write_lock_bh(&mrt_lock);
1612 mrt->mroute6_sk = NULL;
1613 net->ipv6.devconf_all->mc_forwarding--;
1614 inet6_netconf_notify_devconf(net,
1615 NETCONFA_MC_FORWARDING,
1616 NETCONFA_IFINDEX_ALL,
1617 net->ipv6.devconf_all);
1618 write_unlock_bh(&mrt_lock);
1619
1620 mroute_clean_tables(mrt);
1621 err = 0;
1622 break;
1623 }
1624 }
1625 rtnl_unlock();
1626
1627 return err;
1628 }
1629
1630 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1631 {
1632 struct mr6_table *mrt;
1633 struct flowi6 fl6 = {
1634 .flowi6_iif = skb->skb_iif,
1635 .flowi6_oif = skb->dev->ifindex,
1636 .flowi6_mark = skb->mark,
1637 };
1638
1639 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1640 return NULL;
1641
1642 return mrt->mroute6_sk;
1643 }
1644
1645 /*
1646 * Socket options and virtual interface manipulation. The whole
1647 * virtual interface system is a complete heap, but unfortunately
1648 * that's how BSD mrouted happens to think. Maybe one day with a proper
1649 * MOSPF/PIM router set up we can clean this up.
1650 */
1651
1652 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1653 {
1654 int ret, parent = 0;
1655 struct mif6ctl vif;
1656 struct mf6cctl mfc;
1657 mifi_t mifi;
1658 struct net *net = sock_net(sk);
1659 struct mr6_table *mrt;
1660
1661 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1662 if (mrt == NULL)
1663 return -ENOENT;
1664
1665 if (optname != MRT6_INIT) {
1666 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1667 return -EACCES;
1668 }
1669
1670 switch (optname) {
1671 case MRT6_INIT:
1672 if (sk->sk_type != SOCK_RAW ||
1673 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1674 return -EOPNOTSUPP;
1675 if (optlen < sizeof(int))
1676 return -EINVAL;
1677
1678 return ip6mr_sk_init(mrt, sk);
1679
1680 case MRT6_DONE:
1681 return ip6mr_sk_done(sk);
1682
1683 case MRT6_ADD_MIF:
1684 if (optlen < sizeof(vif))
1685 return -EINVAL;
1686 if (copy_from_user(&vif, optval, sizeof(vif)))
1687 return -EFAULT;
1688 if (vif.mif6c_mifi >= MAXMIFS)
1689 return -ENFILE;
1690 rtnl_lock();
1691 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1692 rtnl_unlock();
1693 return ret;
1694
1695 case MRT6_DEL_MIF:
1696 if (optlen < sizeof(mifi_t))
1697 return -EINVAL;
1698 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1699 return -EFAULT;
1700 rtnl_lock();
1701 ret = mif6_delete(mrt, mifi, NULL);
1702 rtnl_unlock();
1703 return ret;
1704
1705 /*
1706 * Manipulate the forwarding caches. These live
1707 * in a sort of kernel/user symbiosis.
1708 */
1709 case MRT6_ADD_MFC:
1710 case MRT6_DEL_MFC:
1711 parent = -1;
1712 case MRT6_ADD_MFC_PROXY:
1713 case MRT6_DEL_MFC_PROXY:
1714 if (optlen < sizeof(mfc))
1715 return -EINVAL;
1716 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1717 return -EFAULT;
1718 if (parent == 0)
1719 parent = mfc.mf6cc_parent;
1720 rtnl_lock();
1721 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1722 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1723 else
1724 ret = ip6mr_mfc_add(net, mrt, &mfc,
1725 sk == mrt->mroute6_sk, parent);
1726 rtnl_unlock();
1727 return ret;
1728
1729 /*
1730 * Control PIM assert (to activate pim will activate assert)
1731 */
1732 case MRT6_ASSERT:
1733 {
1734 int v;
1735
1736 if (optlen != sizeof(v))
1737 return -EINVAL;
1738 if (get_user(v, (int __user *)optval))
1739 return -EFAULT;
1740 mrt->mroute_do_assert = v;
1741 return 0;
1742 }
1743
1744 #ifdef CONFIG_IPV6_PIMSM_V2
1745 case MRT6_PIM:
1746 {
1747 int v;
1748
1749 if (optlen != sizeof(v))
1750 return -EINVAL;
1751 if (get_user(v, (int __user *)optval))
1752 return -EFAULT;
1753 v = !!v;
1754 rtnl_lock();
1755 ret = 0;
1756 if (v != mrt->mroute_do_pim) {
1757 mrt->mroute_do_pim = v;
1758 mrt->mroute_do_assert = v;
1759 }
1760 rtnl_unlock();
1761 return ret;
1762 }
1763
1764 #endif
1765 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1766 case MRT6_TABLE:
1767 {
1768 u32 v;
1769
1770 if (optlen != sizeof(u32))
1771 return -EINVAL;
1772 if (get_user(v, (u32 __user *)optval))
1773 return -EFAULT;
1774 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1775 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1776 return -EINVAL;
1777 if (sk == mrt->mroute6_sk)
1778 return -EBUSY;
1779
1780 rtnl_lock();
1781 ret = 0;
1782 if (!ip6mr_new_table(net, v))
1783 ret = -ENOMEM;
1784 raw6_sk(sk)->ip6mr_table = v;
1785 rtnl_unlock();
1786 return ret;
1787 }
1788 #endif
1789 /*
1790 * Spurious command, or MRT6_VERSION which you cannot
1791 * set.
1792 */
1793 default:
1794 return -ENOPROTOOPT;
1795 }
1796 }
1797
1798 /*
1799 * Getsock opt support for the multicast routing system.
1800 */
1801
1802 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1803 int __user *optlen)
1804 {
1805 int olr;
1806 int val;
1807 struct net *net = sock_net(sk);
1808 struct mr6_table *mrt;
1809
1810 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1811 if (mrt == NULL)
1812 return -ENOENT;
1813
1814 switch (optname) {
1815 case MRT6_VERSION:
1816 val = 0x0305;
1817 break;
1818 #ifdef CONFIG_IPV6_PIMSM_V2
1819 case MRT6_PIM:
1820 val = mrt->mroute_do_pim;
1821 break;
1822 #endif
1823 case MRT6_ASSERT:
1824 val = mrt->mroute_do_assert;
1825 break;
1826 default:
1827 return -ENOPROTOOPT;
1828 }
1829
1830 if (get_user(olr, optlen))
1831 return -EFAULT;
1832
1833 olr = min_t(int, olr, sizeof(int));
1834 if (olr < 0)
1835 return -EINVAL;
1836
1837 if (put_user(olr, optlen))
1838 return -EFAULT;
1839 if (copy_to_user(optval, &val, olr))
1840 return -EFAULT;
1841 return 0;
1842 }
1843
1844 /*
1845 * The IP multicast ioctl support routines.
1846 */
1847
1848 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1849 {
1850 struct sioc_sg_req6 sr;
1851 struct sioc_mif_req6 vr;
1852 struct mif_device *vif;
1853 struct mfc6_cache *c;
1854 struct net *net = sock_net(sk);
1855 struct mr6_table *mrt;
1856
1857 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1858 if (mrt == NULL)
1859 return -ENOENT;
1860
1861 switch (cmd) {
1862 case SIOCGETMIFCNT_IN6:
1863 if (copy_from_user(&vr, arg, sizeof(vr)))
1864 return -EFAULT;
1865 if (vr.mifi >= mrt->maxvif)
1866 return -EINVAL;
1867 read_lock(&mrt_lock);
1868 vif = &mrt->vif6_table[vr.mifi];
1869 if (MIF_EXISTS(mrt, vr.mifi)) {
1870 vr.icount = vif->pkt_in;
1871 vr.ocount = vif->pkt_out;
1872 vr.ibytes = vif->bytes_in;
1873 vr.obytes = vif->bytes_out;
1874 read_unlock(&mrt_lock);
1875
1876 if (copy_to_user(arg, &vr, sizeof(vr)))
1877 return -EFAULT;
1878 return 0;
1879 }
1880 read_unlock(&mrt_lock);
1881 return -EADDRNOTAVAIL;
1882 case SIOCGETSGCNT_IN6:
1883 if (copy_from_user(&sr, arg, sizeof(sr)))
1884 return -EFAULT;
1885
1886 read_lock(&mrt_lock);
1887 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1888 if (c) {
1889 sr.pktcnt = c->mfc_un.res.pkt;
1890 sr.bytecnt = c->mfc_un.res.bytes;
1891 sr.wrong_if = c->mfc_un.res.wrong_if;
1892 read_unlock(&mrt_lock);
1893
1894 if (copy_to_user(arg, &sr, sizeof(sr)))
1895 return -EFAULT;
1896 return 0;
1897 }
1898 read_unlock(&mrt_lock);
1899 return -EADDRNOTAVAIL;
1900 default:
1901 return -ENOIOCTLCMD;
1902 }
1903 }
1904
1905 #ifdef CONFIG_COMPAT
1906 struct compat_sioc_sg_req6 {
1907 struct sockaddr_in6 src;
1908 struct sockaddr_in6 grp;
1909 compat_ulong_t pktcnt;
1910 compat_ulong_t bytecnt;
1911 compat_ulong_t wrong_if;
1912 };
1913
1914 struct compat_sioc_mif_req6 {
1915 mifi_t mifi;
1916 compat_ulong_t icount;
1917 compat_ulong_t ocount;
1918 compat_ulong_t ibytes;
1919 compat_ulong_t obytes;
1920 };
1921
1922 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1923 {
1924 struct compat_sioc_sg_req6 sr;
1925 struct compat_sioc_mif_req6 vr;
1926 struct mif_device *vif;
1927 struct mfc6_cache *c;
1928 struct net *net = sock_net(sk);
1929 struct mr6_table *mrt;
1930
1931 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1932 if (mrt == NULL)
1933 return -ENOENT;
1934
1935 switch (cmd) {
1936 case SIOCGETMIFCNT_IN6:
1937 if (copy_from_user(&vr, arg, sizeof(vr)))
1938 return -EFAULT;
1939 if (vr.mifi >= mrt->maxvif)
1940 return -EINVAL;
1941 read_lock(&mrt_lock);
1942 vif = &mrt->vif6_table[vr.mifi];
1943 if (MIF_EXISTS(mrt, vr.mifi)) {
1944 vr.icount = vif->pkt_in;
1945 vr.ocount = vif->pkt_out;
1946 vr.ibytes = vif->bytes_in;
1947 vr.obytes = vif->bytes_out;
1948 read_unlock(&mrt_lock);
1949
1950 if (copy_to_user(arg, &vr, sizeof(vr)))
1951 return -EFAULT;
1952 return 0;
1953 }
1954 read_unlock(&mrt_lock);
1955 return -EADDRNOTAVAIL;
1956 case SIOCGETSGCNT_IN6:
1957 if (copy_from_user(&sr, arg, sizeof(sr)))
1958 return -EFAULT;
1959
1960 read_lock(&mrt_lock);
1961 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1962 if (c) {
1963 sr.pktcnt = c->mfc_un.res.pkt;
1964 sr.bytecnt = c->mfc_un.res.bytes;
1965 sr.wrong_if = c->mfc_un.res.wrong_if;
1966 read_unlock(&mrt_lock);
1967
1968 if (copy_to_user(arg, &sr, sizeof(sr)))
1969 return -EFAULT;
1970 return 0;
1971 }
1972 read_unlock(&mrt_lock);
1973 return -EADDRNOTAVAIL;
1974 default:
1975 return -ENOIOCTLCMD;
1976 }
1977 }
1978 #endif
1979
1980 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1981 {
1982 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1983 IPSTATS_MIB_OUTFORWDATAGRAMS);
1984 IP6_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1985 IPSTATS_MIB_OUTOCTETS, skb->len);
1986 return dst_output(skb);
1987 }
1988
1989 /*
1990 * Processing handlers for ip6mr_forward
1991 */
1992
1993 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1994 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1995 {
1996 struct ipv6hdr *ipv6h;
1997 struct mif_device *vif = &mrt->vif6_table[vifi];
1998 struct net_device *dev;
1999 struct dst_entry *dst;
2000 struct flowi6 fl6;
2001
2002 if (vif->dev == NULL)
2003 goto out_free;
2004
2005 #ifdef CONFIG_IPV6_PIMSM_V2
2006 if (vif->flags & MIFF_REGISTER) {
2007 vif->pkt_out++;
2008 vif->bytes_out += skb->len;
2009 vif->dev->stats.tx_bytes += skb->len;
2010 vif->dev->stats.tx_packets++;
2011 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2012 goto out_free;
2013 }
2014 #endif
2015
2016 ipv6h = ipv6_hdr(skb);
2017
2018 fl6 = (struct flowi6) {
2019 .flowi6_oif = vif->link,
2020 .daddr = ipv6h->daddr,
2021 };
2022
2023 dst = ip6_route_output(net, NULL, &fl6);
2024 if (dst->error) {
2025 dst_release(dst);
2026 goto out_free;
2027 }
2028
2029 skb_dst_drop(skb);
2030 skb_dst_set(skb, dst);
2031
2032 /*
2033 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2034 * not only before forwarding, but after forwarding on all output
2035 * interfaces. It is clear, if mrouter runs a multicasting
2036 * program, it should receive packets not depending to what interface
2037 * program is joined.
2038 * If we will not make it, the program will have to join on all
2039 * interfaces. On the other hand, multihoming host (or router, but
2040 * not mrouter) cannot join to more than one interface - it will
2041 * result in receiving multiple packets.
2042 */
2043 dev = vif->dev;
2044 skb->dev = dev;
2045 vif->pkt_out++;
2046 vif->bytes_out += skb->len;
2047
2048 /* We are about to write */
2049 /* XXX: extension headers? */
2050 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2051 goto out_free;
2052
2053 ipv6h = ipv6_hdr(skb);
2054 ipv6h->hop_limit--;
2055
2056 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2057
2058 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
2059 ip6mr_forward2_finish);
2060
2061 out_free:
2062 kfree_skb(skb);
2063 return 0;
2064 }
2065
2066 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2067 {
2068 int ct;
2069
2070 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2071 if (mrt->vif6_table[ct].dev == dev)
2072 break;
2073 }
2074 return ct;
2075 }
2076
2077 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2078 struct sk_buff *skb, struct mfc6_cache *cache)
2079 {
2080 int psend = -1;
2081 int vif, ct;
2082 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2083
2084 vif = cache->mf6c_parent;
2085 cache->mfc_un.res.pkt++;
2086 cache->mfc_un.res.bytes += skb->len;
2087
2088 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2089 struct mfc6_cache *cache_proxy;
2090
2091 /* For an (*,G) entry, we only check that the incomming
2092 * interface is part of the static tree.
2093 */
2094 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2095 if (cache_proxy &&
2096 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2097 goto forward;
2098 }
2099
2100 /*
2101 * Wrong interface: drop packet and (maybe) send PIM assert.
2102 */
2103 if (mrt->vif6_table[vif].dev != skb->dev) {
2104 cache->mfc_un.res.wrong_if++;
2105
2106 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2107 /* pimsm uses asserts, when switching from RPT to SPT,
2108 so that we cannot check that packet arrived on an oif.
2109 It is bad, but otherwise we would need to move pretty
2110 large chunk of pimd to kernel. Ough... --ANK
2111 */
2112 (mrt->mroute_do_pim ||
2113 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2114 time_after(jiffies,
2115 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2116 cache->mfc_un.res.last_assert = jiffies;
2117 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2118 }
2119 goto dont_forward;
2120 }
2121
2122 forward:
2123 mrt->vif6_table[vif].pkt_in++;
2124 mrt->vif6_table[vif].bytes_in += skb->len;
2125
2126 /*
2127 * Forward the frame
2128 */
2129 if (ipv6_addr_any(&cache->mf6c_origin) &&
2130 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2131 if (true_vifi >= 0 &&
2132 true_vifi != cache->mf6c_parent &&
2133 ipv6_hdr(skb)->hop_limit >
2134 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2135 /* It's an (*,*) entry and the packet is not coming from
2136 * the upstream: forward the packet to the upstream
2137 * only.
2138 */
2139 psend = cache->mf6c_parent;
2140 goto last_forward;
2141 }
2142 goto dont_forward;
2143 }
2144 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2145 /* For (*,G) entry, don't forward to the incoming interface */
2146 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2147 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2148 if (psend != -1) {
2149 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2150 if (skb2)
2151 ip6mr_forward2(net, mrt, skb2, cache, psend);
2152 }
2153 psend = ct;
2154 }
2155 }
2156 last_forward:
2157 if (psend != -1) {
2158 ip6mr_forward2(net, mrt, skb, cache, psend);
2159 return 0;
2160 }
2161
2162 dont_forward:
2163 kfree_skb(skb);
2164 return 0;
2165 }
2166
2167
2168 /*
2169 * Multicast packets for forwarding arrive here
2170 */
2171
2172 int ip6_mr_input(struct sk_buff *skb)
2173 {
2174 struct mfc6_cache *cache;
2175 struct net *net = dev_net(skb->dev);
2176 struct mr6_table *mrt;
2177 struct flowi6 fl6 = {
2178 .flowi6_iif = skb->dev->ifindex,
2179 .flowi6_mark = skb->mark,
2180 };
2181 int err;
2182
2183 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2184 if (err < 0) {
2185 kfree_skb(skb);
2186 return err;
2187 }
2188
2189 read_lock(&mrt_lock);
2190 cache = ip6mr_cache_find(mrt,
2191 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2192 if (cache == NULL) {
2193 int vif = ip6mr_find_vif(mrt, skb->dev);
2194
2195 if (vif >= 0)
2196 cache = ip6mr_cache_find_any(mrt,
2197 &ipv6_hdr(skb)->daddr,
2198 vif);
2199 }
2200
2201 /*
2202 * No usable cache entry
2203 */
2204 if (cache == NULL) {
2205 int vif;
2206
2207 vif = ip6mr_find_vif(mrt, skb->dev);
2208 if (vif >= 0) {
2209 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2210 read_unlock(&mrt_lock);
2211
2212 return err;
2213 }
2214 read_unlock(&mrt_lock);
2215 kfree_skb(skb);
2216 return -ENODEV;
2217 }
2218
2219 ip6_mr_forward(net, mrt, skb, cache);
2220
2221 read_unlock(&mrt_lock);
2222
2223 return 0;
2224 }
2225
2226
2227 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2228 struct mfc6_cache *c, struct rtmsg *rtm)
2229 {
2230 int ct;
2231 struct rtnexthop *nhp;
2232 struct nlattr *mp_attr;
2233 struct rta_mfc_stats mfcs;
2234
2235 /* If cache is unresolved, don't try to parse IIF and OIF */
2236 if (c->mf6c_parent >= MAXMIFS)
2237 return -ENOENT;
2238
2239 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2240 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2241 return -EMSGSIZE;
2242 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2243 if (mp_attr == NULL)
2244 return -EMSGSIZE;
2245
2246 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2247 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2248 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2249 if (nhp == NULL) {
2250 nla_nest_cancel(skb, mp_attr);
2251 return -EMSGSIZE;
2252 }
2253
2254 nhp->rtnh_flags = 0;
2255 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2256 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2257 nhp->rtnh_len = sizeof(*nhp);
2258 }
2259 }
2260
2261 nla_nest_end(skb, mp_attr);
2262
2263 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2264 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2265 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2266 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2267 return -EMSGSIZE;
2268
2269 rtm->rtm_type = RTN_MULTICAST;
2270 return 1;
2271 }
2272
2273 int ip6mr_get_route(struct net *net,
2274 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2275 {
2276 int err;
2277 struct mr6_table *mrt;
2278 struct mfc6_cache *cache;
2279 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2280
2281 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2282 if (mrt == NULL)
2283 return -ENOENT;
2284
2285 read_lock(&mrt_lock);
2286 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2287 if (!cache && skb->dev) {
2288 int vif = ip6mr_find_vif(mrt, skb->dev);
2289
2290 if (vif >= 0)
2291 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2292 vif);
2293 }
2294
2295 if (!cache) {
2296 struct sk_buff *skb2;
2297 struct ipv6hdr *iph;
2298 struct net_device *dev;
2299 int vif;
2300
2301 if (nowait) {
2302 read_unlock(&mrt_lock);
2303 return -EAGAIN;
2304 }
2305
2306 dev = skb->dev;
2307 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2308 read_unlock(&mrt_lock);
2309 return -ENODEV;
2310 }
2311
2312 /* really correct? */
2313 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2314 if (!skb2) {
2315 read_unlock(&mrt_lock);
2316 return -ENOMEM;
2317 }
2318
2319 skb_reset_transport_header(skb2);
2320
2321 skb_put(skb2, sizeof(struct ipv6hdr));
2322 skb_reset_network_header(skb2);
2323
2324 iph = ipv6_hdr(skb2);
2325 iph->version = 0;
2326 iph->priority = 0;
2327 iph->flow_lbl[0] = 0;
2328 iph->flow_lbl[1] = 0;
2329 iph->flow_lbl[2] = 0;
2330 iph->payload_len = 0;
2331 iph->nexthdr = IPPROTO_NONE;
2332 iph->hop_limit = 0;
2333 iph->saddr = rt->rt6i_src.addr;
2334 iph->daddr = rt->rt6i_dst.addr;
2335
2336 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2337 read_unlock(&mrt_lock);
2338
2339 return err;
2340 }
2341
2342 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2343 cache->mfc_flags |= MFC_NOTIFY;
2344
2345 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2346 read_unlock(&mrt_lock);
2347 return err;
2348 }
2349
2350 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2351 u32 portid, u32 seq, struct mfc6_cache *c, int cmd)
2352 {
2353 struct nlmsghdr *nlh;
2354 struct rtmsg *rtm;
2355 int err;
2356
2357 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI);
2358 if (nlh == NULL)
2359 return -EMSGSIZE;
2360
2361 rtm = nlmsg_data(nlh);
2362 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2363 rtm->rtm_dst_len = 128;
2364 rtm->rtm_src_len = 128;
2365 rtm->rtm_tos = 0;
2366 rtm->rtm_table = mrt->id;
2367 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2368 goto nla_put_failure;
2369 rtm->rtm_type = RTN_MULTICAST;
2370 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2371 if (c->mfc_flags & MFC_STATIC)
2372 rtm->rtm_protocol = RTPROT_STATIC;
2373 else
2374 rtm->rtm_protocol = RTPROT_MROUTED;
2375 rtm->rtm_flags = 0;
2376
2377 if (nla_put(skb, RTA_SRC, 16, &c->mf6c_origin) ||
2378 nla_put(skb, RTA_DST, 16, &c->mf6c_mcastgrp))
2379 goto nla_put_failure;
2380 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2381 /* do not break the dump if cache is unresolved */
2382 if (err < 0 && err != -ENOENT)
2383 goto nla_put_failure;
2384
2385 return nlmsg_end(skb, nlh);
2386
2387 nla_put_failure:
2388 nlmsg_cancel(skb, nlh);
2389 return -EMSGSIZE;
2390 }
2391
2392 static int mr6_msgsize(bool unresolved, int maxvif)
2393 {
2394 size_t len =
2395 NLMSG_ALIGN(sizeof(struct rtmsg))
2396 + nla_total_size(4) /* RTA_TABLE */
2397 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2398 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2399 ;
2400
2401 if (!unresolved)
2402 len = len
2403 + nla_total_size(4) /* RTA_IIF */
2404 + nla_total_size(0) /* RTA_MULTIPATH */
2405 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2406 /* RTA_MFC_STATS */
2407 + nla_total_size(sizeof(struct rta_mfc_stats))
2408 ;
2409
2410 return len;
2411 }
2412
2413 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2414 int cmd)
2415 {
2416 struct net *net = read_pnet(&mrt->net);
2417 struct sk_buff *skb;
2418 int err = -ENOBUFS;
2419
2420 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2421 GFP_ATOMIC);
2422 if (skb == NULL)
2423 goto errout;
2424
2425 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd);
2426 if (err < 0)
2427 goto errout;
2428
2429 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2430 return;
2431
2432 errout:
2433 kfree_skb(skb);
2434 if (err < 0)
2435 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2436 }
2437
2438 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2439 {
2440 struct net *net = sock_net(skb->sk);
2441 struct mr6_table *mrt;
2442 struct mfc6_cache *mfc;
2443 unsigned int t = 0, s_t;
2444 unsigned int h = 0, s_h;
2445 unsigned int e = 0, s_e;
2446
2447 s_t = cb->args[0];
2448 s_h = cb->args[1];
2449 s_e = cb->args[2];
2450
2451 read_lock(&mrt_lock);
2452 ip6mr_for_each_table(mrt, net) {
2453 if (t < s_t)
2454 goto next_table;
2455 if (t > s_t)
2456 s_h = 0;
2457 for (h = s_h; h < MFC6_LINES; h++) {
2458 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2459 if (e < s_e)
2460 goto next_entry;
2461 if (ip6mr_fill_mroute(mrt, skb,
2462 NETLINK_CB(cb->skb).portid,
2463 cb->nlh->nlmsg_seq,
2464 mfc, RTM_NEWROUTE) < 0)
2465 goto done;
2466 next_entry:
2467 e++;
2468 }
2469 e = s_e = 0;
2470 }
2471 spin_lock_bh(&mfc_unres_lock);
2472 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2473 if (e < s_e)
2474 goto next_entry2;
2475 if (ip6mr_fill_mroute(mrt, skb,
2476 NETLINK_CB(cb->skb).portid,
2477 cb->nlh->nlmsg_seq,
2478 mfc, RTM_NEWROUTE) < 0) {
2479 spin_unlock_bh(&mfc_unres_lock);
2480 goto done;
2481 }
2482 next_entry2:
2483 e++;
2484 }
2485 spin_unlock_bh(&mfc_unres_lock);
2486 e = s_e = 0;
2487 s_h = 0;
2488 next_table:
2489 t++;
2490 }
2491 done:
2492 read_unlock(&mrt_lock);
2493
2494 cb->args[2] = e;
2495 cb->args[1] = h;
2496 cb->args[0] = t;
2497
2498 return skb->len;
2499 }
(deprecated) Btrfs devel tree