?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/mm: Add TLB purge to free pmd/pte page interfaces
[linux/fpc-iii.git] / net / ipv6 / ip6mr.c
blob4b93ad4fe6d8d3a2b4295e3ef6aadf4b2ebe9d09
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 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 int ip6mr_rtm_dumproute(struct sk_buff *skb,
120 struct netlink_callback *cb);
121 static void mroute_clean_tables(struct mr6_table *mrt, bool all);
122 static void ipmr_expire_process(unsigned long arg);
123
124 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
125 #define ip6mr_for_each_table(mrt, net) \
126 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
127
128 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
129 {
130 struct mr6_table *mrt;
131
132 ip6mr_for_each_table(mrt, net) {
133 if (mrt->id == id)
134 return mrt;
135 }
136 return NULL;
137 }
138
139 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
140 struct mr6_table **mrt)
141 {
142 int err;
143 struct ip6mr_result res;
144 struct fib_lookup_arg arg = {
145 .result = &res,
146 .flags = FIB_LOOKUP_NOREF,
147 };
148
149 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
150 flowi6_to_flowi(flp6), 0, &arg);
151 if (err < 0)
152 return err;
153 *mrt = res.mrt;
154 return 0;
155 }
156
157 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
158 int flags, struct fib_lookup_arg *arg)
159 {
160 struct ip6mr_result *res = arg->result;
161 struct mr6_table *mrt;
162
163 switch (rule->action) {
164 case FR_ACT_TO_TBL:
165 break;
166 case FR_ACT_UNREACHABLE:
167 return -ENETUNREACH;
168 case FR_ACT_PROHIBIT:
169 return -EACCES;
170 case FR_ACT_BLACKHOLE:
171 default:
172 return -EINVAL;
173 }
174
175 mrt = ip6mr_get_table(rule->fr_net, rule->table);
176 if (!mrt)
177 return -EAGAIN;
178 res->mrt = mrt;
179 return 0;
180 }
181
182 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
183 {
184 return 1;
185 }
186
187 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
188 FRA_GENERIC_POLICY,
189 };
190
191 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
192 struct fib_rule_hdr *frh, struct nlattr **tb)
193 {
194 return 0;
195 }
196
197 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
198 struct nlattr **tb)
199 {
200 return 1;
201 }
202
203 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
204 struct fib_rule_hdr *frh)
205 {
206 frh->dst_len = 0;
207 frh->src_len = 0;
208 frh->tos = 0;
209 return 0;
210 }
211
212 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
213 .family = RTNL_FAMILY_IP6MR,
214 .rule_size = sizeof(struct ip6mr_rule),
215 .addr_size = sizeof(struct in6_addr),
216 .action = ip6mr_rule_action,
217 .match = ip6mr_rule_match,
218 .configure = ip6mr_rule_configure,
219 .compare = ip6mr_rule_compare,
220 .fill = ip6mr_rule_fill,
221 .nlgroup = RTNLGRP_IPV6_RULE,
222 .policy = ip6mr_rule_policy,
223 .owner = THIS_MODULE,
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) {
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 ip6mr_free_table(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 fib_rules_unregister(net->ipv6.mr6_rules_ops);
268 rtnl_unlock();
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)
308 return mrt;
309
310 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
311 if (!mrt)
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_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_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 = &reg_vif_netdev_ops;
737 dev->destructor = free_netdev;
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),
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 = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
851 nlh->nlmsg_type = NLMSG_ERROR;
852 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
853 skb_trim(skb, nlh->nlmsg_len);
854 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
855 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
856 } else
857 kfree_skb(skb);
858 }
859
860 ip6mr_cache_free(c);
861 }
862
863
864 /* Timer process for all the unresolved queue. */
865
866 static void ipmr_do_expire_process(struct mr6_table *mrt)
867 {
868 unsigned long now = jiffies;
869 unsigned long expires = 10 * HZ;
870 struct mfc6_cache *c, *next;
871
872 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
873 if (time_after(c->mfc_un.unres.expires, now)) {
874 /* not yet... */
875 unsigned long interval = c->mfc_un.unres.expires - now;
876 if (interval < expires)
877 expires = interval;
878 continue;
879 }
880
881 list_del(&c->list);
882 mr6_netlink_event(mrt, c, RTM_DELROUTE);
883 ip6mr_destroy_unres(mrt, c);
884 }
885
886 if (!list_empty(&mrt->mfc6_unres_queue))
887 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
888 }
889
890 static void ipmr_expire_process(unsigned long arg)
891 {
892 struct mr6_table *mrt = (struct mr6_table *)arg;
893
894 if (!spin_trylock(&mfc_unres_lock)) {
895 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
896 return;
897 }
898
899 if (!list_empty(&mrt->mfc6_unres_queue))
900 ipmr_do_expire_process(mrt);
901
902 spin_unlock(&mfc_unres_lock);
903 }
904
905 /* Fill oifs list. It is called under write locked mrt_lock. */
906
907 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
908 unsigned char *ttls)
909 {
910 int vifi;
911
912 cache->mfc_un.res.minvif = MAXMIFS;
913 cache->mfc_un.res.maxvif = 0;
914 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
915
916 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
917 if (MIF_EXISTS(mrt, vifi) &&
918 ttls[vifi] && ttls[vifi] < 255) {
919 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
920 if (cache->mfc_un.res.minvif > vifi)
921 cache->mfc_un.res.minvif = vifi;
922 if (cache->mfc_un.res.maxvif <= vifi)
923 cache->mfc_un.res.maxvif = vifi + 1;
924 }
925 }
926 cache->mfc_un.res.lastuse = jiffies;
927 }
928
929 static int mif6_add(struct net *net, struct mr6_table *mrt,
930 struct mif6ctl *vifc, int mrtsock)
931 {
932 int vifi = vifc->mif6c_mifi;
933 struct mif_device *v = &mrt->vif6_table[vifi];
934 struct net_device *dev;
935 struct inet6_dev *in6_dev;
936 int err;
937
938 /* Is vif busy ? */
939 if (MIF_EXISTS(mrt, vifi))
940 return -EADDRINUSE;
941
942 switch (vifc->mif6c_flags) {
943 #ifdef CONFIG_IPV6_PIMSM_V2
944 case MIFF_REGISTER:
945 /*
946 * Special Purpose VIF in PIM
947 * All the packets will be sent to the daemon
948 */
949 if (mrt->mroute_reg_vif_num >= 0)
950 return -EADDRINUSE;
951 dev = ip6mr_reg_vif(net, mrt);
952 if (!dev)
953 return -ENOBUFS;
954 err = dev_set_allmulti(dev, 1);
955 if (err) {
956 unregister_netdevice(dev);
957 dev_put(dev);
958 return err;
959 }
960 break;
961 #endif
962 case 0:
963 dev = dev_get_by_index(net, vifc->mif6c_pifi);
964 if (!dev)
965 return -EADDRNOTAVAIL;
966 err = dev_set_allmulti(dev, 1);
967 if (err) {
968 dev_put(dev);
969 return err;
970 }
971 break;
972 default:
973 return -EINVAL;
974 }
975
976 in6_dev = __in6_dev_get(dev);
977 if (in6_dev) {
978 in6_dev->cnf.mc_forwarding++;
979 inet6_netconf_notify_devconf(dev_net(dev),
980 NETCONFA_MC_FORWARDING,
981 dev->ifindex, &in6_dev->cnf);
982 }
983
984 /*
985 * Fill in the VIF structures
986 */
987 v->rate_limit = vifc->vifc_rate_limit;
988 v->flags = vifc->mif6c_flags;
989 if (!mrtsock)
990 v->flags |= VIFF_STATIC;
991 v->threshold = vifc->vifc_threshold;
992 v->bytes_in = 0;
993 v->bytes_out = 0;
994 v->pkt_in = 0;
995 v->pkt_out = 0;
996 v->link = dev->ifindex;
997 if (v->flags & MIFF_REGISTER)
998 v->link = dev_get_iflink(dev);
999
1000 /* And finish update writing critical data */
1001 write_lock_bh(&mrt_lock);
1002 v->dev = dev;
1003 #ifdef CONFIG_IPV6_PIMSM_V2
1004 if (v->flags & MIFF_REGISTER)
1005 mrt->mroute_reg_vif_num = vifi;
1006 #endif
1007 if (vifi + 1 > mrt->maxvif)
1008 mrt->maxvif = vifi + 1;
1009 write_unlock_bh(&mrt_lock);
1010 return 0;
1011 }
1012
1013 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
1014 const struct in6_addr *origin,
1015 const struct in6_addr *mcastgrp)
1016 {
1017 int line = MFC6_HASH(mcastgrp, origin);
1018 struct mfc6_cache *c;
1019
1020 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1021 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1022 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1023 return c;
1024 }
1025 return NULL;
1026 }
1027
1028 /* Look for a (*,*,oif) entry */
1029 static struct mfc6_cache *ip6mr_cache_find_any_parent(struct mr6_table *mrt,
1030 mifi_t mifi)
1031 {
1032 int line = MFC6_HASH(&in6addr_any, &in6addr_any);
1033 struct mfc6_cache *c;
1034
1035 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1036 if (ipv6_addr_any(&c->mf6c_origin) &&
1037 ipv6_addr_any(&c->mf6c_mcastgrp) &&
1038 (c->mfc_un.res.ttls[mifi] < 255))
1039 return c;
1040
1041 return NULL;
1042 }
1043
1044 /* Look for a (*,G) entry */
1045 static struct mfc6_cache *ip6mr_cache_find_any(struct mr6_table *mrt,
1046 struct in6_addr *mcastgrp,
1047 mifi_t mifi)
1048 {
1049 int line = MFC6_HASH(mcastgrp, &in6addr_any);
1050 struct mfc6_cache *c, *proxy;
1051
1052 if (ipv6_addr_any(mcastgrp))
1053 goto skip;
1054
1055 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list)
1056 if (ipv6_addr_any(&c->mf6c_origin) &&
1057 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) {
1058 if (c->mfc_un.res.ttls[mifi] < 255)
1059 return c;
1060
1061 /* It's ok if the mifi is part of the static tree */
1062 proxy = ip6mr_cache_find_any_parent(mrt,
1063 c->mf6c_parent);
1064 if (proxy && proxy->mfc_un.res.ttls[mifi] < 255)
1065 return c;
1066 }
1067
1068 skip:
1069 return ip6mr_cache_find_any_parent(mrt, mifi);
1070 }
1071
1072 /*
1073 * Allocate a multicast cache entry
1074 */
1075 static struct mfc6_cache *ip6mr_cache_alloc(void)
1076 {
1077 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1078 if (!c)
1079 return NULL;
1080 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1081 c->mfc_un.res.minvif = MAXMIFS;
1082 return c;
1083 }
1084
1085 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1086 {
1087 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1088 if (!c)
1089 return NULL;
1090 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1091 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1092 return c;
1093 }
1094
1095 /*
1096 * A cache entry has gone into a resolved state from queued
1097 */
1098
1099 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1100 struct mfc6_cache *uc, struct mfc6_cache *c)
1101 {
1102 struct sk_buff *skb;
1103
1104 /*
1105 * Play the pending entries through our router
1106 */
1107
1108 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1109 if (ipv6_hdr(skb)->version == 0) {
1110 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1111
1112 if (__ip6mr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1113 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1114 } else {
1115 nlh->nlmsg_type = NLMSG_ERROR;
1116 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1117 skb_trim(skb, nlh->nlmsg_len);
1118 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1119 }
1120 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1121 } else
1122 ip6_mr_forward(net, mrt, skb, c);
1123 }
1124 }
1125
1126 /*
1127 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1128 * expects the following bizarre scheme.
1129 *
1130 * Called under mrt_lock.
1131 */
1132
1133 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1134 mifi_t mifi, int assert)
1135 {
1136 struct sk_buff *skb;
1137 struct mrt6msg *msg;
1138 int ret;
1139
1140 #ifdef CONFIG_IPV6_PIMSM_V2
1141 if (assert == MRT6MSG_WHOLEPKT)
1142 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1143 +sizeof(*msg));
1144 else
1145 #endif
1146 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1147
1148 if (!skb)
1149 return -ENOBUFS;
1150
1151 /* I suppose that internal messages
1152 * do not require checksums */
1153
1154 skb->ip_summed = CHECKSUM_UNNECESSARY;
1155
1156 #ifdef CONFIG_IPV6_PIMSM_V2
1157 if (assert == MRT6MSG_WHOLEPKT) {
1158 /* Ugly, but we have no choice with this interface.
1159 Duplicate old header, fix length etc.
1160 And all this only to mangle msg->im6_msgtype and
1161 to set msg->im6_mbz to "mbz" :-)
1162 */
1163 skb_push(skb, -skb_network_offset(pkt));
1164
1165 skb_push(skb, sizeof(*msg));
1166 skb_reset_transport_header(skb);
1167 msg = (struct mrt6msg *)skb_transport_header(skb);
1168 msg->im6_mbz = 0;
1169 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1170 msg->im6_mif = mrt->mroute_reg_vif_num;
1171 msg->im6_pad = 0;
1172 msg->im6_src = ipv6_hdr(pkt)->saddr;
1173 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1174
1175 skb->ip_summed = CHECKSUM_UNNECESSARY;
1176 } else
1177 #endif
1178 {
1179 /*
1180 * Copy the IP header
1181 */
1182
1183 skb_put(skb, sizeof(struct ipv6hdr));
1184 skb_reset_network_header(skb);
1185 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1186
1187 /*
1188 * Add our header
1189 */
1190 skb_put(skb, sizeof(*msg));
1191 skb_reset_transport_header(skb);
1192 msg = (struct mrt6msg *)skb_transport_header(skb);
1193
1194 msg->im6_mbz = 0;
1195 msg->im6_msgtype = assert;
1196 msg->im6_mif = mifi;
1197 msg->im6_pad = 0;
1198 msg->im6_src = ipv6_hdr(pkt)->saddr;
1199 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1200
1201 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1202 skb->ip_summed = CHECKSUM_UNNECESSARY;
1203 }
1204
1205 if (!mrt->mroute6_sk) {
1206 kfree_skb(skb);
1207 return -EINVAL;
1208 }
1209
1210 /*
1211 * Deliver to user space multicast routing algorithms
1212 */
1213 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1214 if (ret < 0) {
1215 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1216 kfree_skb(skb);
1217 }
1218
1219 return ret;
1220 }
1221
1222 /*
1223 * Queue a packet for resolution. It gets locked cache entry!
1224 */
1225
1226 static int
1227 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1228 {
1229 bool found = false;
1230 int err;
1231 struct mfc6_cache *c;
1232
1233 spin_lock_bh(&mfc_unres_lock);
1234 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1235 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1236 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1237 found = true;
1238 break;
1239 }
1240 }
1241
1242 if (!found) {
1243 /*
1244 * Create a new entry if allowable
1245 */
1246
1247 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1248 (c = ip6mr_cache_alloc_unres()) == NULL) {
1249 spin_unlock_bh(&mfc_unres_lock);
1250
1251 kfree_skb(skb);
1252 return -ENOBUFS;
1253 }
1254
1255 /*
1256 * Fill in the new cache entry
1257 */
1258 c->mf6c_parent = -1;
1259 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1260 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1261
1262 /*
1263 * Reflect first query at pim6sd
1264 */
1265 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1266 if (err < 0) {
1267 /* If the report failed throw the cache entry
1268 out - Brad Parker
1269 */
1270 spin_unlock_bh(&mfc_unres_lock);
1271
1272 ip6mr_cache_free(c);
1273 kfree_skb(skb);
1274 return err;
1275 }
1276
1277 atomic_inc(&mrt->cache_resolve_queue_len);
1278 list_add(&c->list, &mrt->mfc6_unres_queue);
1279 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1280
1281 ipmr_do_expire_process(mrt);
1282 }
1283
1284 /*
1285 * See if we can append the packet
1286 */
1287 if (c->mfc_un.unres.unresolved.qlen > 3) {
1288 kfree_skb(skb);
1289 err = -ENOBUFS;
1290 } else {
1291 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1292 err = 0;
1293 }
1294
1295 spin_unlock_bh(&mfc_unres_lock);
1296 return err;
1297 }
1298
1299 /*
1300 * MFC6 cache manipulation by user space
1301 */
1302
1303 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc,
1304 int parent)
1305 {
1306 int line;
1307 struct mfc6_cache *c, *next;
1308
1309 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1310
1311 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1312 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1313 ipv6_addr_equal(&c->mf6c_mcastgrp,
1314 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1315 (parent == -1 || parent == c->mf6c_parent)) {
1316 write_lock_bh(&mrt_lock);
1317 list_del(&c->list);
1318 write_unlock_bh(&mrt_lock);
1319
1320 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1321 ip6mr_cache_free(c);
1322 return 0;
1323 }
1324 }
1325 return -ENOENT;
1326 }
1327
1328 static int ip6mr_device_event(struct notifier_block *this,
1329 unsigned long event, void *ptr)
1330 {
1331 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1332 struct net *net = dev_net(dev);
1333 struct mr6_table *mrt;
1334 struct mif_device *v;
1335 int ct;
1336
1337 if (event != NETDEV_UNREGISTER)
1338 return NOTIFY_DONE;
1339
1340 ip6mr_for_each_table(mrt, net) {
1341 v = &mrt->vif6_table[0];
1342 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1343 if (v->dev == dev)
1344 mif6_delete(mrt, ct, 1, NULL);
1345 }
1346 }
1347
1348 return NOTIFY_DONE;
1349 }
1350
1351 static struct notifier_block ip6_mr_notifier = {
1352 .notifier_call = ip6mr_device_event
1353 };
1354
1355 /*
1356 * Setup for IP multicast routing
1357 */
1358
1359 static int __net_init ip6mr_net_init(struct net *net)
1360 {
1361 int err;
1362
1363 err = ip6mr_rules_init(net);
1364 if (err < 0)
1365 goto fail;
1366
1367 #ifdef CONFIG_PROC_FS
1368 err = -ENOMEM;
1369 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1370 goto proc_vif_fail;
1371 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1372 goto proc_cache_fail;
1373 #endif
1374
1375 return 0;
1376
1377 #ifdef CONFIG_PROC_FS
1378 proc_cache_fail:
1379 remove_proc_entry("ip6_mr_vif", net->proc_net);
1380 proc_vif_fail:
1381 ip6mr_rules_exit(net);
1382 #endif
1383 fail:
1384 return err;
1385 }
1386
1387 static void __net_exit ip6mr_net_exit(struct net *net)
1388 {
1389 #ifdef CONFIG_PROC_FS
1390 remove_proc_entry("ip6_mr_cache", net->proc_net);
1391 remove_proc_entry("ip6_mr_vif", net->proc_net);
1392 #endif
1393 ip6mr_rules_exit(net);
1394 }
1395
1396 static struct pernet_operations ip6mr_net_ops = {
1397 .init = ip6mr_net_init,
1398 .exit = ip6mr_net_exit,
1399 };
1400
1401 int __init ip6_mr_init(void)
1402 {
1403 int err;
1404
1405 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1406 sizeof(struct mfc6_cache),
1407 0, SLAB_HWCACHE_ALIGN,
1408 NULL);
1409 if (!mrt_cachep)
1410 return -ENOMEM;
1411
1412 err = register_pernet_subsys(&ip6mr_net_ops);
1413 if (err)
1414 goto reg_pernet_fail;
1415
1416 err = register_netdevice_notifier(&ip6_mr_notifier);
1417 if (err)
1418 goto reg_notif_fail;
1419 #ifdef CONFIG_IPV6_PIMSM_V2
1420 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1421 pr_err("%s: can't add PIM protocol\n", __func__);
1422 err = -EAGAIN;
1423 goto add_proto_fail;
1424 }
1425 #endif
1426 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1427 ip6mr_rtm_dumproute, NULL);
1428 return 0;
1429 #ifdef CONFIG_IPV6_PIMSM_V2
1430 add_proto_fail:
1431 unregister_netdevice_notifier(&ip6_mr_notifier);
1432 #endif
1433 reg_notif_fail:
1434 unregister_pernet_subsys(&ip6mr_net_ops);
1435 reg_pernet_fail:
1436 kmem_cache_destroy(mrt_cachep);
1437 return err;
1438 }
1439
1440 void ip6_mr_cleanup(void)
1441 {
1442 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1443 #ifdef CONFIG_IPV6_PIMSM_V2
1444 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1445 #endif
1446 unregister_netdevice_notifier(&ip6_mr_notifier);
1447 unregister_pernet_subsys(&ip6mr_net_ops);
1448 kmem_cache_destroy(mrt_cachep);
1449 }
1450
1451 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1452 struct mf6cctl *mfc, int mrtsock, int parent)
1453 {
1454 bool found = false;
1455 int line;
1456 struct mfc6_cache *uc, *c;
1457 unsigned char ttls[MAXMIFS];
1458 int i;
1459
1460 if (mfc->mf6cc_parent >= MAXMIFS)
1461 return -ENFILE;
1462
1463 memset(ttls, 255, MAXMIFS);
1464 for (i = 0; i < MAXMIFS; i++) {
1465 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1466 ttls[i] = 1;
1467
1468 }
1469
1470 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1471
1472 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1473 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1474 ipv6_addr_equal(&c->mf6c_mcastgrp,
1475 &mfc->mf6cc_mcastgrp.sin6_addr) &&
1476 (parent == -1 || parent == mfc->mf6cc_parent)) {
1477 found = true;
1478 break;
1479 }
1480 }
1481
1482 if (found) {
1483 write_lock_bh(&mrt_lock);
1484 c->mf6c_parent = mfc->mf6cc_parent;
1485 ip6mr_update_thresholds(mrt, c, ttls);
1486 if (!mrtsock)
1487 c->mfc_flags |= MFC_STATIC;
1488 write_unlock_bh(&mrt_lock);
1489 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1490 return 0;
1491 }
1492
1493 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1494 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1495 return -EINVAL;
1496
1497 c = ip6mr_cache_alloc();
1498 if (!c)
1499 return -ENOMEM;
1500
1501 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1502 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1503 c->mf6c_parent = mfc->mf6cc_parent;
1504 ip6mr_update_thresholds(mrt, c, ttls);
1505 if (!mrtsock)
1506 c->mfc_flags |= MFC_STATIC;
1507
1508 write_lock_bh(&mrt_lock);
1509 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1510 write_unlock_bh(&mrt_lock);
1511
1512 /*
1513 * Check to see if we resolved a queued list. If so we
1514 * need to send on the frames and tidy up.
1515 */
1516 found = false;
1517 spin_lock_bh(&mfc_unres_lock);
1518 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1519 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1520 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1521 list_del(&uc->list);
1522 atomic_dec(&mrt->cache_resolve_queue_len);
1523 found = true;
1524 break;
1525 }
1526 }
1527 if (list_empty(&mrt->mfc6_unres_queue))
1528 del_timer(&mrt->ipmr_expire_timer);
1529 spin_unlock_bh(&mfc_unres_lock);
1530
1531 if (found) {
1532 ip6mr_cache_resolve(net, mrt, uc, c);
1533 ip6mr_cache_free(uc);
1534 }
1535 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1536 return 0;
1537 }
1538
1539 /*
1540 * Close the multicast socket, and clear the vif tables etc
1541 */
1542
1543 static void mroute_clean_tables(struct mr6_table *mrt, bool all)
1544 {
1545 int i;
1546 LIST_HEAD(list);
1547 struct mfc6_cache *c, *next;
1548
1549 /*
1550 * Shut down all active vif entries
1551 */
1552 for (i = 0; i < mrt->maxvif; i++) {
1553 if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC))
1554 continue;
1555 mif6_delete(mrt, i, 0, &list);
1556 }
1557 unregister_netdevice_many(&list);
1558
1559 /*
1560 * Wipe the cache
1561 */
1562 for (i = 0; i < MFC6_LINES; i++) {
1563 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1564 if (!all && (c->mfc_flags & MFC_STATIC))
1565 continue;
1566 write_lock_bh(&mrt_lock);
1567 list_del(&c->list);
1568 write_unlock_bh(&mrt_lock);
1569
1570 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1571 ip6mr_cache_free(c);
1572 }
1573 }
1574
1575 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1576 spin_lock_bh(&mfc_unres_lock);
1577 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1578 list_del(&c->list);
1579 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1580 ip6mr_destroy_unres(mrt, c);
1581 }
1582 spin_unlock_bh(&mfc_unres_lock);
1583 }
1584 }
1585
1586 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1587 {
1588 int err = 0;
1589 struct net *net = sock_net(sk);
1590
1591 rtnl_lock();
1592 write_lock_bh(&mrt_lock);
1593 if (likely(mrt->mroute6_sk == NULL)) {
1594 mrt->mroute6_sk = sk;
1595 net->ipv6.devconf_all->mc_forwarding++;
1596 } else {
1597 err = -EADDRINUSE;
1598 }
1599 write_unlock_bh(&mrt_lock);
1600
1601 if (!err)
1602 inet6_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1603 NETCONFA_IFINDEX_ALL,
1604 net->ipv6.devconf_all);
1605 rtnl_unlock();
1606
1607 return err;
1608 }
1609
1610 int ip6mr_sk_done(struct sock *sk)
1611 {
1612 int err = -EACCES;
1613 struct net *net = sock_net(sk);
1614 struct mr6_table *mrt;
1615
1616 rtnl_lock();
1617 ip6mr_for_each_table(mrt, net) {
1618 if (sk == mrt->mroute6_sk) {
1619 write_lock_bh(&mrt_lock);
1620 mrt->mroute6_sk = NULL;
1621 net->ipv6.devconf_all->mc_forwarding--;
1622 write_unlock_bh(&mrt_lock);
1623 inet6_netconf_notify_devconf(net,
1624 NETCONFA_MC_FORWARDING,
1625 NETCONFA_IFINDEX_ALL,
1626 net->ipv6.devconf_all);
1627
1628 mroute_clean_tables(mrt, false);
1629 err = 0;
1630 break;
1631 }
1632 }
1633 rtnl_unlock();
1634
1635 return err;
1636 }
1637
1638 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1639 {
1640 struct mr6_table *mrt;
1641 struct flowi6 fl6 = {
1642 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1643 .flowi6_oif = skb->dev->ifindex,
1644 .flowi6_mark = skb->mark,
1645 };
1646
1647 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1648 return NULL;
1649
1650 return mrt->mroute6_sk;
1651 }
1652
1653 /*
1654 * Socket options and virtual interface manipulation. The whole
1655 * virtual interface system is a complete heap, but unfortunately
1656 * that's how BSD mrouted happens to think. Maybe one day with a proper
1657 * MOSPF/PIM router set up we can clean this up.
1658 */
1659
1660 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1661 {
1662 int ret, parent = 0;
1663 struct mif6ctl vif;
1664 struct mf6cctl mfc;
1665 mifi_t mifi;
1666 struct net *net = sock_net(sk);
1667 struct mr6_table *mrt;
1668
1669 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1670 if (!mrt)
1671 return -ENOENT;
1672
1673 if (optname != MRT6_INIT) {
1674 if (sk != mrt->mroute6_sk && !ns_capable(net->user_ns, CAP_NET_ADMIN))
1675 return -EACCES;
1676 }
1677
1678 switch (optname) {
1679 case MRT6_INIT:
1680 if (sk->sk_type != SOCK_RAW ||
1681 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1682 return -EOPNOTSUPP;
1683 if (optlen < sizeof(int))
1684 return -EINVAL;
1685
1686 return ip6mr_sk_init(mrt, sk);
1687
1688 case MRT6_DONE:
1689 return ip6mr_sk_done(sk);
1690
1691 case MRT6_ADD_MIF:
1692 if (optlen < sizeof(vif))
1693 return -EINVAL;
1694 if (copy_from_user(&vif, optval, sizeof(vif)))
1695 return -EFAULT;
1696 if (vif.mif6c_mifi >= MAXMIFS)
1697 return -ENFILE;
1698 rtnl_lock();
1699 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1700 rtnl_unlock();
1701 return ret;
1702
1703 case MRT6_DEL_MIF:
1704 if (optlen < sizeof(mifi_t))
1705 return -EINVAL;
1706 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1707 return -EFAULT;
1708 rtnl_lock();
1709 ret = mif6_delete(mrt, mifi, 0, NULL);
1710 rtnl_unlock();
1711 return ret;
1712
1713 /*
1714 * Manipulate the forwarding caches. These live
1715 * in a sort of kernel/user symbiosis.
1716 */
1717 case MRT6_ADD_MFC:
1718 case MRT6_DEL_MFC:
1719 parent = -1;
1720 case MRT6_ADD_MFC_PROXY:
1721 case MRT6_DEL_MFC_PROXY:
1722 if (optlen < sizeof(mfc))
1723 return -EINVAL;
1724 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1725 return -EFAULT;
1726 if (parent == 0)
1727 parent = mfc.mf6cc_parent;
1728 rtnl_lock();
1729 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1730 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1731 else
1732 ret = ip6mr_mfc_add(net, mrt, &mfc,
1733 sk == mrt->mroute6_sk, parent);
1734 rtnl_unlock();
1735 return ret;
1736
1737 /*
1738 * Control PIM assert (to activate pim will activate assert)
1739 */
1740 case MRT6_ASSERT:
1741 {
1742 int v;
1743
1744 if (optlen != sizeof(v))
1745 return -EINVAL;
1746 if (get_user(v, (int __user *)optval))
1747 return -EFAULT;
1748 mrt->mroute_do_assert = v;
1749 return 0;
1750 }
1751
1752 #ifdef CONFIG_IPV6_PIMSM_V2
1753 case MRT6_PIM:
1754 {
1755 int v;
1756
1757 if (optlen != sizeof(v))
1758 return -EINVAL;
1759 if (get_user(v, (int __user *)optval))
1760 return -EFAULT;
1761 v = !!v;
1762 rtnl_lock();
1763 ret = 0;
1764 if (v != mrt->mroute_do_pim) {
1765 mrt->mroute_do_pim = v;
1766 mrt->mroute_do_assert = v;
1767 }
1768 rtnl_unlock();
1769 return ret;
1770 }
1771
1772 #endif
1773 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1774 case MRT6_TABLE:
1775 {
1776 u32 v;
1777
1778 if (optlen != sizeof(u32))
1779 return -EINVAL;
1780 if (get_user(v, (u32 __user *)optval))
1781 return -EFAULT;
1782 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1783 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1784 return -EINVAL;
1785 if (sk == mrt->mroute6_sk)
1786 return -EBUSY;
1787
1788 rtnl_lock();
1789 ret = 0;
1790 if (!ip6mr_new_table(net, v))
1791 ret = -ENOMEM;
1792 else
1793 raw6_sk(sk)->ip6mr_table = v;
1794 rtnl_unlock();
1795 return ret;
1796 }
1797 #endif
1798 /*
1799 * Spurious command, or MRT6_VERSION which you cannot
1800 * set.
1801 */
1802 default:
1803 return -ENOPROTOOPT;
1804 }
1805 }
1806
1807 /*
1808 * Getsock opt support for the multicast routing system.
1809 */
1810
1811 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1812 int __user *optlen)
1813 {
1814 int olr;
1815 int val;
1816 struct net *net = sock_net(sk);
1817 struct mr6_table *mrt;
1818
1819 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1820 if (!mrt)
1821 return -ENOENT;
1822
1823 switch (optname) {
1824 case MRT6_VERSION:
1825 val = 0x0305;
1826 break;
1827 #ifdef CONFIG_IPV6_PIMSM_V2
1828 case MRT6_PIM:
1829 val = mrt->mroute_do_pim;
1830 break;
1831 #endif
1832 case MRT6_ASSERT:
1833 val = mrt->mroute_do_assert;
1834 break;
1835 default:
1836 return -ENOPROTOOPT;
1837 }
1838
1839 if (get_user(olr, optlen))
1840 return -EFAULT;
1841
1842 olr = min_t(int, olr, sizeof(int));
1843 if (olr < 0)
1844 return -EINVAL;
1845
1846 if (put_user(olr, optlen))
1847 return -EFAULT;
1848 if (copy_to_user(optval, &val, olr))
1849 return -EFAULT;
1850 return 0;
1851 }
1852
1853 /*
1854 * The IP multicast ioctl support routines.
1855 */
1856
1857 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1858 {
1859 struct sioc_sg_req6 sr;
1860 struct sioc_mif_req6 vr;
1861 struct mif_device *vif;
1862 struct mfc6_cache *c;
1863 struct net *net = sock_net(sk);
1864 struct mr6_table *mrt;
1865
1866 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1867 if (!mrt)
1868 return -ENOENT;
1869
1870 switch (cmd) {
1871 case SIOCGETMIFCNT_IN6:
1872 if (copy_from_user(&vr, arg, sizeof(vr)))
1873 return -EFAULT;
1874 if (vr.mifi >= mrt->maxvif)
1875 return -EINVAL;
1876 read_lock(&mrt_lock);
1877 vif = &mrt->vif6_table[vr.mifi];
1878 if (MIF_EXISTS(mrt, vr.mifi)) {
1879 vr.icount = vif->pkt_in;
1880 vr.ocount = vif->pkt_out;
1881 vr.ibytes = vif->bytes_in;
1882 vr.obytes = vif->bytes_out;
1883 read_unlock(&mrt_lock);
1884
1885 if (copy_to_user(arg, &vr, sizeof(vr)))
1886 return -EFAULT;
1887 return 0;
1888 }
1889 read_unlock(&mrt_lock);
1890 return -EADDRNOTAVAIL;
1891 case SIOCGETSGCNT_IN6:
1892 if (copy_from_user(&sr, arg, sizeof(sr)))
1893 return -EFAULT;
1894
1895 read_lock(&mrt_lock);
1896 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1897 if (c) {
1898 sr.pktcnt = c->mfc_un.res.pkt;
1899 sr.bytecnt = c->mfc_un.res.bytes;
1900 sr.wrong_if = c->mfc_un.res.wrong_if;
1901 read_unlock(&mrt_lock);
1902
1903 if (copy_to_user(arg, &sr, sizeof(sr)))
1904 return -EFAULT;
1905 return 0;
1906 }
1907 read_unlock(&mrt_lock);
1908 return -EADDRNOTAVAIL;
1909 default:
1910 return -ENOIOCTLCMD;
1911 }
1912 }
1913
1914 #ifdef CONFIG_COMPAT
1915 struct compat_sioc_sg_req6 {
1916 struct sockaddr_in6 src;
1917 struct sockaddr_in6 grp;
1918 compat_ulong_t pktcnt;
1919 compat_ulong_t bytecnt;
1920 compat_ulong_t wrong_if;
1921 };
1922
1923 struct compat_sioc_mif_req6 {
1924 mifi_t mifi;
1925 compat_ulong_t icount;
1926 compat_ulong_t ocount;
1927 compat_ulong_t ibytes;
1928 compat_ulong_t obytes;
1929 };
1930
1931 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1932 {
1933 struct compat_sioc_sg_req6 sr;
1934 struct compat_sioc_mif_req6 vr;
1935 struct mif_device *vif;
1936 struct mfc6_cache *c;
1937 struct net *net = sock_net(sk);
1938 struct mr6_table *mrt;
1939
1940 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1941 if (!mrt)
1942 return -ENOENT;
1943
1944 switch (cmd) {
1945 case SIOCGETMIFCNT_IN6:
1946 if (copy_from_user(&vr, arg, sizeof(vr)))
1947 return -EFAULT;
1948 if (vr.mifi >= mrt->maxvif)
1949 return -EINVAL;
1950 read_lock(&mrt_lock);
1951 vif = &mrt->vif6_table[vr.mifi];
1952 if (MIF_EXISTS(mrt, vr.mifi)) {
1953 vr.icount = vif->pkt_in;
1954 vr.ocount = vif->pkt_out;
1955 vr.ibytes = vif->bytes_in;
1956 vr.obytes = vif->bytes_out;
1957 read_unlock(&mrt_lock);
1958
1959 if (copy_to_user(arg, &vr, sizeof(vr)))
1960 return -EFAULT;
1961 return 0;
1962 }
1963 read_unlock(&mrt_lock);
1964 return -EADDRNOTAVAIL;
1965 case SIOCGETSGCNT_IN6:
1966 if (copy_from_user(&sr, arg, sizeof(sr)))
1967 return -EFAULT;
1968
1969 read_lock(&mrt_lock);
1970 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1971 if (c) {
1972 sr.pktcnt = c->mfc_un.res.pkt;
1973 sr.bytecnt = c->mfc_un.res.bytes;
1974 sr.wrong_if = c->mfc_un.res.wrong_if;
1975 read_unlock(&mrt_lock);
1976
1977 if (copy_to_user(arg, &sr, sizeof(sr)))
1978 return -EFAULT;
1979 return 0;
1980 }
1981 read_unlock(&mrt_lock);
1982 return -EADDRNOTAVAIL;
1983 default:
1984 return -ENOIOCTLCMD;
1985 }
1986 }
1987 #endif
1988
1989 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1990 {
1991 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
1992 IPSTATS_MIB_OUTFORWDATAGRAMS);
1993 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)),
1994 IPSTATS_MIB_OUTOCTETS, skb->len);
1995 return dst_output(net, sk, skb);
1996 }
1997
1998 /*
1999 * Processing handlers for ip6mr_forward
2000 */
2001
2002 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
2003 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
2004 {
2005 struct ipv6hdr *ipv6h;
2006 struct mif_device *vif = &mrt->vif6_table[vifi];
2007 struct net_device *dev;
2008 struct dst_entry *dst;
2009 struct flowi6 fl6;
2010
2011 if (!vif->dev)
2012 goto out_free;
2013
2014 #ifdef CONFIG_IPV6_PIMSM_V2
2015 if (vif->flags & MIFF_REGISTER) {
2016 vif->pkt_out++;
2017 vif->bytes_out += skb->len;
2018 vif->dev->stats.tx_bytes += skb->len;
2019 vif->dev->stats.tx_packets++;
2020 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2021 goto out_free;
2022 }
2023 #endif
2024
2025 ipv6h = ipv6_hdr(skb);
2026
2027 fl6 = (struct flowi6) {
2028 .flowi6_oif = vif->link,
2029 .daddr = ipv6h->daddr,
2030 };
2031
2032 dst = ip6_route_output(net, NULL, &fl6);
2033 if (dst->error) {
2034 dst_release(dst);
2035 goto out_free;
2036 }
2037
2038 skb_dst_drop(skb);
2039 skb_dst_set(skb, dst);
2040
2041 /*
2042 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2043 * not only before forwarding, but after forwarding on all output
2044 * interfaces. It is clear, if mrouter runs a multicasting
2045 * program, it should receive packets not depending to what interface
2046 * program is joined.
2047 * If we will not make it, the program will have to join on all
2048 * interfaces. On the other hand, multihoming host (or router, but
2049 * not mrouter) cannot join to more than one interface - it will
2050 * result in receiving multiple packets.
2051 */
2052 dev = vif->dev;
2053 skb->dev = dev;
2054 vif->pkt_out++;
2055 vif->bytes_out += skb->len;
2056
2057 /* We are about to write */
2058 /* XXX: extension headers? */
2059 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2060 goto out_free;
2061
2062 ipv6h = ipv6_hdr(skb);
2063 ipv6h->hop_limit--;
2064
2065 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2066
2067 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2068 net, NULL, skb, skb->dev, dev,
2069 ip6mr_forward2_finish);
2070
2071 out_free:
2072 kfree_skb(skb);
2073 return 0;
2074 }
2075
2076 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
2077 {
2078 int ct;
2079
2080 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2081 if (mrt->vif6_table[ct].dev == dev)
2082 break;
2083 }
2084 return ct;
2085 }
2086
2087 static void ip6_mr_forward(struct net *net, struct mr6_table *mrt,
2088 struct sk_buff *skb, struct mfc6_cache *cache)
2089 {
2090 int psend = -1;
2091 int vif, ct;
2092 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2093
2094 vif = cache->mf6c_parent;
2095 cache->mfc_un.res.pkt++;
2096 cache->mfc_un.res.bytes += skb->len;
2097 cache->mfc_un.res.lastuse = jiffies;
2098
2099 if (ipv6_addr_any(&cache->mf6c_origin) && true_vifi >= 0) {
2100 struct mfc6_cache *cache_proxy;
2101
2102 /* For an (*,G) entry, we only check that the incoming
2103 * interface is part of the static tree.
2104 */
2105 cache_proxy = ip6mr_cache_find_any_parent(mrt, vif);
2106 if (cache_proxy &&
2107 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2108 goto forward;
2109 }
2110
2111 /*
2112 * Wrong interface: drop packet and (maybe) send PIM assert.
2113 */
2114 if (mrt->vif6_table[vif].dev != skb->dev) {
2115 cache->mfc_un.res.wrong_if++;
2116
2117 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2118 /* pimsm uses asserts, when switching from RPT to SPT,
2119 so that we cannot check that packet arrived on an oif.
2120 It is bad, but otherwise we would need to move pretty
2121 large chunk of pimd to kernel. Ough... --ANK
2122 */
2123 (mrt->mroute_do_pim ||
2124 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2125 time_after(jiffies,
2126 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2127 cache->mfc_un.res.last_assert = jiffies;
2128 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2129 }
2130 goto dont_forward;
2131 }
2132
2133 forward:
2134 mrt->vif6_table[vif].pkt_in++;
2135 mrt->vif6_table[vif].bytes_in += skb->len;
2136
2137 /*
2138 * Forward the frame
2139 */
2140 if (ipv6_addr_any(&cache->mf6c_origin) &&
2141 ipv6_addr_any(&cache->mf6c_mcastgrp)) {
2142 if (true_vifi >= 0 &&
2143 true_vifi != cache->mf6c_parent &&
2144 ipv6_hdr(skb)->hop_limit >
2145 cache->mfc_un.res.ttls[cache->mf6c_parent]) {
2146 /* It's an (*,*) entry and the packet is not coming from
2147 * the upstream: forward the packet to the upstream
2148 * only.
2149 */
2150 psend = cache->mf6c_parent;
2151 goto last_forward;
2152 }
2153 goto dont_forward;
2154 }
2155 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2156 /* For (*,G) entry, don't forward to the incoming interface */
2157 if ((!ipv6_addr_any(&cache->mf6c_origin) || ct != true_vifi) &&
2158 ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2159 if (psend != -1) {
2160 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2161 if (skb2)
2162 ip6mr_forward2(net, mrt, skb2, cache, psend);
2163 }
2164 psend = ct;
2165 }
2166 }
2167 last_forward:
2168 if (psend != -1) {
2169 ip6mr_forward2(net, mrt, skb, cache, psend);
2170 return;
2171 }
2172
2173 dont_forward:
2174 kfree_skb(skb);
2175 }
2176
2177
2178 /*
2179 * Multicast packets for forwarding arrive here
2180 */
2181
2182 int ip6_mr_input(struct sk_buff *skb)
2183 {
2184 struct mfc6_cache *cache;
2185 struct net *net = dev_net(skb->dev);
2186 struct mr6_table *mrt;
2187 struct flowi6 fl6 = {
2188 .flowi6_iif = skb->dev->ifindex,
2189 .flowi6_mark = skb->mark,
2190 };
2191 int err;
2192
2193 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2194 if (err < 0) {
2195 kfree_skb(skb);
2196 return err;
2197 }
2198
2199 read_lock(&mrt_lock);
2200 cache = ip6mr_cache_find(mrt,
2201 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2202 if (!cache) {
2203 int vif = ip6mr_find_vif(mrt, skb->dev);
2204
2205 if (vif >= 0)
2206 cache = ip6mr_cache_find_any(mrt,
2207 &ipv6_hdr(skb)->daddr,
2208 vif);
2209 }
2210
2211 /*
2212 * No usable cache entry
2213 */
2214 if (!cache) {
2215 int vif;
2216
2217 vif = ip6mr_find_vif(mrt, skb->dev);
2218 if (vif >= 0) {
2219 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2220 read_unlock(&mrt_lock);
2221
2222 return err;
2223 }
2224 read_unlock(&mrt_lock);
2225 kfree_skb(skb);
2226 return -ENODEV;
2227 }
2228
2229 ip6_mr_forward(net, mrt, skb, cache);
2230
2231 read_unlock(&mrt_lock);
2232
2233 return 0;
2234 }
2235
2236
2237 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2238 struct mfc6_cache *c, struct rtmsg *rtm)
2239 {
2240 struct rta_mfc_stats mfcs;
2241 struct nlattr *mp_attr;
2242 struct rtnexthop *nhp;
2243 unsigned long lastuse;
2244 int ct;
2245
2246 /* If cache is unresolved, don't try to parse IIF and OIF */
2247 if (c->mf6c_parent >= MAXMIFS)
2248 return -ENOENT;
2249
2250 if (MIF_EXISTS(mrt, c->mf6c_parent) &&
2251 nla_put_u32(skb, RTA_IIF, mrt->vif6_table[c->mf6c_parent].dev->ifindex) < 0)
2252 return -EMSGSIZE;
2253 mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
2254 if (!mp_attr)
2255 return -EMSGSIZE;
2256
2257 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2258 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2259 nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
2260 if (!nhp) {
2261 nla_nest_cancel(skb, mp_attr);
2262 return -EMSGSIZE;
2263 }
2264
2265 nhp->rtnh_flags = 0;
2266 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2267 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2268 nhp->rtnh_len = sizeof(*nhp);
2269 }
2270 }
2271
2272 nla_nest_end(skb, mp_attr);
2273
2274 lastuse = READ_ONCE(c->mfc_un.res.lastuse);
2275 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
2276
2277 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2278 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2279 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2280 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
2281 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
2282 RTA_PAD))
2283 return -EMSGSIZE;
2284
2285 rtm->rtm_type = RTN_MULTICAST;
2286 return 1;
2287 }
2288
2289 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2290 int nowait, u32 portid)
2291 {
2292 int err;
2293 struct mr6_table *mrt;
2294 struct mfc6_cache *cache;
2295 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2296
2297 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2298 if (!mrt)
2299 return -ENOENT;
2300
2301 read_lock(&mrt_lock);
2302 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2303 if (!cache && skb->dev) {
2304 int vif = ip6mr_find_vif(mrt, skb->dev);
2305
2306 if (vif >= 0)
2307 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2308 vif);
2309 }
2310
2311 if (!cache) {
2312 struct sk_buff *skb2;
2313 struct ipv6hdr *iph;
2314 struct net_device *dev;
2315 int vif;
2316
2317 if (nowait) {
2318 read_unlock(&mrt_lock);
2319 return -EAGAIN;
2320 }
2321
2322 dev = skb->dev;
2323 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2324 read_unlock(&mrt_lock);
2325 return -ENODEV;
2326 }
2327
2328 /* really correct? */
2329 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2330 if (!skb2) {
2331 read_unlock(&mrt_lock);
2332 return -ENOMEM;
2333 }
2334
2335 NETLINK_CB(skb2).portid = portid;
2336 skb_reset_transport_header(skb2);
2337
2338 skb_put(skb2, sizeof(struct ipv6hdr));
2339 skb_reset_network_header(skb2);
2340
2341 iph = ipv6_hdr(skb2);
2342 iph->version = 0;
2343 iph->priority = 0;
2344 iph->flow_lbl[0] = 0;
2345 iph->flow_lbl[1] = 0;
2346 iph->flow_lbl[2] = 0;
2347 iph->payload_len = 0;
2348 iph->nexthdr = IPPROTO_NONE;
2349 iph->hop_limit = 0;
2350 iph->saddr = rt->rt6i_src.addr;
2351 iph->daddr = rt->rt6i_dst.addr;
2352
2353 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2354 read_unlock(&mrt_lock);
2355
2356 return err;
2357 }
2358
2359 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2360 cache->mfc_flags |= MFC_NOTIFY;
2361
2362 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2363 read_unlock(&mrt_lock);
2364 return err;
2365 }
2366
2367 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2368 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2369 int flags)
2370 {
2371 struct nlmsghdr *nlh;
2372 struct rtmsg *rtm;
2373 int err;
2374
2375 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2376 if (!nlh)
2377 return -EMSGSIZE;
2378
2379 rtm = nlmsg_data(nlh);
2380 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2381 rtm->rtm_dst_len = 128;
2382 rtm->rtm_src_len = 128;
2383 rtm->rtm_tos = 0;
2384 rtm->rtm_table = mrt->id;
2385 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2386 goto nla_put_failure;
2387 rtm->rtm_type = RTN_MULTICAST;
2388 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2389 if (c->mfc_flags & MFC_STATIC)
2390 rtm->rtm_protocol = RTPROT_STATIC;
2391 else
2392 rtm->rtm_protocol = RTPROT_MROUTED;
2393 rtm->rtm_flags = 0;
2394
2395 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2396 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2397 goto nla_put_failure;
2398 err = __ip6mr_fill_mroute(mrt, skb, c, rtm);
2399 /* do not break the dump if cache is unresolved */
2400 if (err < 0 && err != -ENOENT)
2401 goto nla_put_failure;
2402
2403 nlmsg_end(skb, nlh);
2404 return 0;
2405
2406 nla_put_failure:
2407 nlmsg_cancel(skb, nlh);
2408 return -EMSGSIZE;
2409 }
2410
2411 static int mr6_msgsize(bool unresolved, int maxvif)
2412 {
2413 size_t len =
2414 NLMSG_ALIGN(sizeof(struct rtmsg))
2415 + nla_total_size(4) /* RTA_TABLE */
2416 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2417 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2418 ;
2419
2420 if (!unresolved)
2421 len = len
2422 + nla_total_size(4) /* RTA_IIF */
2423 + nla_total_size(0) /* RTA_MULTIPATH */
2424 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2425 /* RTA_MFC_STATS */
2426 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2427 ;
2428
2429 return len;
2430 }
2431
2432 static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc,
2433 int cmd)
2434 {
2435 struct net *net = read_pnet(&mrt->net);
2436 struct sk_buff *skb;
2437 int err = -ENOBUFS;
2438
2439 skb = nlmsg_new(mr6_msgsize(mfc->mf6c_parent >= MAXMIFS, mrt->maxvif),
2440 GFP_ATOMIC);
2441 if (!skb)
2442 goto errout;
2443
2444 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2445 if (err < 0)
2446 goto errout;
2447
2448 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2449 return;
2450
2451 errout:
2452 kfree_skb(skb);
2453 if (err < 0)
2454 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2455 }
2456
2457 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2458 {
2459 struct net *net = sock_net(skb->sk);
2460 struct mr6_table *mrt;
2461 struct mfc6_cache *mfc;
2462 unsigned int t = 0, s_t;
2463 unsigned int h = 0, s_h;
2464 unsigned int e = 0, s_e;
2465
2466 s_t = cb->args[0];
2467 s_h = cb->args[1];
2468 s_e = cb->args[2];
2469
2470 read_lock(&mrt_lock);
2471 ip6mr_for_each_table(mrt, net) {
2472 if (t < s_t)
2473 goto next_table;
2474 if (t > s_t)
2475 s_h = 0;
2476 for (h = s_h; h < MFC6_LINES; h++) {
2477 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2478 if (e < s_e)
2479 goto next_entry;
2480 if (ip6mr_fill_mroute(mrt, skb,
2481 NETLINK_CB(cb->skb).portid,
2482 cb->nlh->nlmsg_seq,
2483 mfc, RTM_NEWROUTE,
2484 NLM_F_MULTI) < 0)
2485 goto done;
2486 next_entry:
2487 e++;
2488 }
2489 e = s_e = 0;
2490 }
2491 spin_lock_bh(&mfc_unres_lock);
2492 list_for_each_entry(mfc, &mrt->mfc6_unres_queue, list) {
2493 if (e < s_e)
2494 goto next_entry2;
2495 if (ip6mr_fill_mroute(mrt, skb,
2496 NETLINK_CB(cb->skb).portid,
2497 cb->nlh->nlmsg_seq,
2498 mfc, RTM_NEWROUTE,
2499 NLM_F_MULTI) < 0) {
2500 spin_unlock_bh(&mfc_unres_lock);
2501 goto done;
2502 }
2503 next_entry2:
2504 e++;
2505 }
2506 spin_unlock_bh(&mfc_unres_lock);
2507 e = s_e = 0;
2508 s_h = 0;
2509 next_table:
2510 t++;
2511 }
2512 done:
2513 read_unlock(&mrt_lock);
2514
2515 cb->args[2] = e;
2516 cb->args[1] = h;
2517 cb->args[0] = t;
2518
2519 return skb->len;
2520 }
Linux with added FPC-III support