?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
kvm tools, setup: Create private directory
[linux-2.6/next.git] / net / ipv6 / ip6mr.c
blob82a809901f8e9d3522f562562edf1e0f5e05e74d
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/system.h>
20 #include <asm/uaccess.h>
21 #include <linux/types.h>
22 #include <linux/sched.h>
23 #include <linux/errno.h>
24 #include <linux/timer.h>
25 #include <linux/mm.h>
26 #include <linux/kernel.h>
27 #include <linux/fcntl.h>
28 #include <linux/stat.h>
29 #include <linux/socket.h>
30 #include <linux/inet.h>
31 #include <linux/netdevice.h>
32 #include <linux/inetdevice.h>
33 #include <linux/proc_fs.h>
34 #include <linux/seq_file.h>
35 #include <linux/init.h>
36 #include <linux/slab.h>
37 #include <linux/compat.h>
38 #include <net/protocol.h>
39 #include <linux/skbuff.h>
40 #include <net/sock.h>
41 #include <net/raw.h>
42 #include <linux/notifier.h>
43 #include <linux/if_arp.h>
44 #include <net/checksum.h>
45 #include <net/netlink.h>
46 #include <net/fib_rules.h>
47
48 #include <net/ipv6.h>
49 #include <net/ip6_route.h>
50 #include <linux/mroute6.h>
51 #include <linux/pim.h>
52 #include <net/addrconf.h>
53 #include <linux/netfilter_ipv6.h>
54 #include <net/ip6_checksum.h>
55
56 struct mr6_table {
57 struct list_head list;
58 #ifdef CONFIG_NET_NS
59 struct net *net;
60 #endif
61 u32 id;
62 struct sock *mroute6_sk;
63 struct timer_list ipmr_expire_timer;
64 struct list_head mfc6_unres_queue;
65 struct list_head mfc6_cache_array[MFC6_LINES];
66 struct mif_device vif6_table[MAXMIFS];
67 int maxvif;
68 atomic_t cache_resolve_queue_len;
69 int mroute_do_assert;
70 int mroute_do_pim;
71 #ifdef CONFIG_IPV6_PIMSM_V2
72 int mroute_reg_vif_num;
73 #endif
74 };
75
76 struct ip6mr_rule {
77 struct fib_rule common;
78 };
79
80 struct ip6mr_result {
81 struct mr6_table *mrt;
82 };
83
84 /* Big lock, protecting vif table, mrt cache and mroute socket state.
85 Note that the changes are semaphored via rtnl_lock.
86 */
87
88 static DEFINE_RWLOCK(mrt_lock);
89
90 /*
91 * Multicast router control variables
92 */
93
94 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
95
96 /* Special spinlock for queue of unresolved entries */
97 static DEFINE_SPINLOCK(mfc_unres_lock);
98
99 /* We return to original Alan's scheme. Hash table of resolved
100 entries is changed only in process context and protected
101 with weak lock mrt_lock. Queue of unresolved entries is protected
102 with strong spinlock mfc_unres_lock.
103
104 In this case data path is free of exclusive locks at all.
105 */
106
107 static struct kmem_cache *mrt_cachep __read_mostly;
108
109 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
110 static void ip6mr_free_table(struct mr6_table *mrt);
111
112 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
113 struct sk_buff *skb, struct mfc6_cache *cache);
114 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
115 mifi_t mifi, int assert);
116 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
117 struct mfc6_cache *c, struct rtmsg *rtm);
118 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
119 struct netlink_callback *cb);
120 static void mroute_clean_tables(struct mr6_table *mrt);
121 static void ipmr_expire_process(unsigned long arg);
122
123 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
124 #define ip6mr_for_each_table(mrt, net) \
125 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
126
127 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
128 {
129 struct mr6_table *mrt;
130
131 ip6mr_for_each_table(mrt, net) {
132 if (mrt->id == id)
133 return mrt;
134 }
135 return NULL;
136 }
137
138 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
139 struct mr6_table **mrt)
140 {
141 struct ip6mr_result res;
142 struct fib_lookup_arg arg = { .result = &res, };
143 int err;
144
145 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
146 flowi6_to_flowi(flp6), 0, &arg);
147 if (err < 0)
148 return err;
149 *mrt = res.mrt;
150 return 0;
151 }
152
153 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
154 int flags, struct fib_lookup_arg *arg)
155 {
156 struct ip6mr_result *res = arg->result;
157 struct mr6_table *mrt;
158
159 switch (rule->action) {
160 case FR_ACT_TO_TBL:
161 break;
162 case FR_ACT_UNREACHABLE:
163 return -ENETUNREACH;
164 case FR_ACT_PROHIBIT:
165 return -EACCES;
166 case FR_ACT_BLACKHOLE:
167 default:
168 return -EINVAL;
169 }
170
171 mrt = ip6mr_get_table(rule->fr_net, rule->table);
172 if (mrt == NULL)
173 return -EAGAIN;
174 res->mrt = mrt;
175 return 0;
176 }
177
178 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
179 {
180 return 1;
181 }
182
183 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
184 FRA_GENERIC_POLICY,
185 };
186
187 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
188 struct fib_rule_hdr *frh, struct nlattr **tb)
189 {
190 return 0;
191 }
192
193 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
194 struct nlattr **tb)
195 {
196 return 1;
197 }
198
199 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
200 struct fib_rule_hdr *frh)
201 {
202 frh->dst_len = 0;
203 frh->src_len = 0;
204 frh->tos = 0;
205 return 0;
206 }
207
208 static const struct fib_rules_ops __net_initdata ip6mr_rules_ops_template = {
209 .family = RTNL_FAMILY_IP6MR,
210 .rule_size = sizeof(struct ip6mr_rule),
211 .addr_size = sizeof(struct in6_addr),
212 .action = ip6mr_rule_action,
213 .match = ip6mr_rule_match,
214 .configure = ip6mr_rule_configure,
215 .compare = ip6mr_rule_compare,
216 .default_pref = fib_default_rule_pref,
217 .fill = ip6mr_rule_fill,
218 .nlgroup = RTNLGRP_IPV6_RULE,
219 .policy = ip6mr_rule_policy,
220 .owner = THIS_MODULE,
221 };
222
223 static int __net_init ip6mr_rules_init(struct net *net)
224 {
225 struct fib_rules_ops *ops;
226 struct mr6_table *mrt;
227 int err;
228
229 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
230 if (IS_ERR(ops))
231 return PTR_ERR(ops);
232
233 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
234
235 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
236 if (mrt == NULL) {
237 err = -ENOMEM;
238 goto err1;
239 }
240
241 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
242 if (err < 0)
243 goto err2;
244
245 net->ipv6.mr6_rules_ops = ops;
246 return 0;
247
248 err2:
249 kfree(mrt);
250 err1:
251 fib_rules_unregister(ops);
252 return err;
253 }
254
255 static void __net_exit ip6mr_rules_exit(struct net *net)
256 {
257 struct mr6_table *mrt, *next;
258
259 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
260 list_del(&mrt->list);
261 ip6mr_free_table(mrt);
262 }
263 fib_rules_unregister(net->ipv6.mr6_rules_ops);
264 }
265 #else
266 #define ip6mr_for_each_table(mrt, net) \
267 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
268
269 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
270 {
271 return net->ipv6.mrt6;
272 }
273
274 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
275 struct mr6_table **mrt)
276 {
277 *mrt = net->ipv6.mrt6;
278 return 0;
279 }
280
281 static int __net_init ip6mr_rules_init(struct net *net)
282 {
283 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
284 return net->ipv6.mrt6 ? 0 : -ENOMEM;
285 }
286
287 static void __net_exit ip6mr_rules_exit(struct net *net)
288 {
289 ip6mr_free_table(net->ipv6.mrt6);
290 }
291 #endif
292
293 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
294 {
295 struct mr6_table *mrt;
296 unsigned int i;
297
298 mrt = ip6mr_get_table(net, id);
299 if (mrt != NULL)
300 return mrt;
301
302 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
303 if (mrt == NULL)
304 return NULL;
305 mrt->id = id;
306 write_pnet(&mrt->net, net);
307
308 /* Forwarding cache */
309 for (i = 0; i < MFC6_LINES; i++)
310 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
311
312 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
313
314 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
315 (unsigned long)mrt);
316
317 #ifdef CONFIG_IPV6_PIMSM_V2
318 mrt->mroute_reg_vif_num = -1;
319 #endif
320 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
321 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
322 #endif
323 return mrt;
324 }
325
326 static void ip6mr_free_table(struct mr6_table *mrt)
327 {
328 del_timer(&mrt->ipmr_expire_timer);
329 mroute_clean_tables(mrt);
330 kfree(mrt);
331 }
332
333 #ifdef CONFIG_PROC_FS
334
335 struct ipmr_mfc_iter {
336 struct seq_net_private p;
337 struct mr6_table *mrt;
338 struct list_head *cache;
339 int ct;
340 };
341
342
343 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
344 struct ipmr_mfc_iter *it, loff_t pos)
345 {
346 struct mr6_table *mrt = it->mrt;
347 struct mfc6_cache *mfc;
348
349 read_lock(&mrt_lock);
350 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
351 it->cache = &mrt->mfc6_cache_array[it->ct];
352 list_for_each_entry(mfc, it->cache, list)
353 if (pos-- == 0)
354 return mfc;
355 }
356 read_unlock(&mrt_lock);
357
358 spin_lock_bh(&mfc_unres_lock);
359 it->cache = &mrt->mfc6_unres_queue;
360 list_for_each_entry(mfc, it->cache, list)
361 if (pos-- == 0)
362 return mfc;
363 spin_unlock_bh(&mfc_unres_lock);
364
365 it->cache = NULL;
366 return NULL;
367 }
368
369 /*
370 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
371 */
372
373 struct ipmr_vif_iter {
374 struct seq_net_private p;
375 struct mr6_table *mrt;
376 int ct;
377 };
378
379 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
380 struct ipmr_vif_iter *iter,
381 loff_t pos)
382 {
383 struct mr6_table *mrt = iter->mrt;
384
385 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
386 if (!MIF_EXISTS(mrt, iter->ct))
387 continue;
388 if (pos-- == 0)
389 return &mrt->vif6_table[iter->ct];
390 }
391 return NULL;
392 }
393
394 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
395 __acquires(mrt_lock)
396 {
397 struct ipmr_vif_iter *iter = seq->private;
398 struct net *net = seq_file_net(seq);
399 struct mr6_table *mrt;
400
401 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
402 if (mrt == NULL)
403 return ERR_PTR(-ENOENT);
404
405 iter->mrt = mrt;
406
407 read_lock(&mrt_lock);
408 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
409 : SEQ_START_TOKEN;
410 }
411
412 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
413 {
414 struct ipmr_vif_iter *iter = seq->private;
415 struct net *net = seq_file_net(seq);
416 struct mr6_table *mrt = iter->mrt;
417
418 ++*pos;
419 if (v == SEQ_START_TOKEN)
420 return ip6mr_vif_seq_idx(net, iter, 0);
421
422 while (++iter->ct < mrt->maxvif) {
423 if (!MIF_EXISTS(mrt, iter->ct))
424 continue;
425 return &mrt->vif6_table[iter->ct];
426 }
427 return NULL;
428 }
429
430 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
431 __releases(mrt_lock)
432 {
433 read_unlock(&mrt_lock);
434 }
435
436 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
437 {
438 struct ipmr_vif_iter *iter = seq->private;
439 struct mr6_table *mrt = iter->mrt;
440
441 if (v == SEQ_START_TOKEN) {
442 seq_puts(seq,
443 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
444 } else {
445 const struct mif_device *vif = v;
446 const char *name = vif->dev ? vif->dev->name : "none";
447
448 seq_printf(seq,
449 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
450 vif - mrt->vif6_table,
451 name, vif->bytes_in, vif->pkt_in,
452 vif->bytes_out, vif->pkt_out,
453 vif->flags);
454 }
455 return 0;
456 }
457
458 static const struct seq_operations ip6mr_vif_seq_ops = {
459 .start = ip6mr_vif_seq_start,
460 .next = ip6mr_vif_seq_next,
461 .stop = ip6mr_vif_seq_stop,
462 .show = ip6mr_vif_seq_show,
463 };
464
465 static int ip6mr_vif_open(struct inode *inode, struct file *file)
466 {
467 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
468 sizeof(struct ipmr_vif_iter));
469 }
470
471 static const struct file_operations ip6mr_vif_fops = {
472 .owner = THIS_MODULE,
473 .open = ip6mr_vif_open,
474 .read = seq_read,
475 .llseek = seq_lseek,
476 .release = seq_release_net,
477 };
478
479 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
480 {
481 struct ipmr_mfc_iter *it = seq->private;
482 struct net *net = seq_file_net(seq);
483 struct mr6_table *mrt;
484
485 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
486 if (mrt == NULL)
487 return ERR_PTR(-ENOENT);
488
489 it->mrt = mrt;
490 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
491 : SEQ_START_TOKEN;
492 }
493
494 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
495 {
496 struct mfc6_cache *mfc = v;
497 struct ipmr_mfc_iter *it = seq->private;
498 struct net *net = seq_file_net(seq);
499 struct mr6_table *mrt = it->mrt;
500
501 ++*pos;
502
503 if (v == SEQ_START_TOKEN)
504 return ipmr_mfc_seq_idx(net, seq->private, 0);
505
506 if (mfc->list.next != it->cache)
507 return list_entry(mfc->list.next, struct mfc6_cache, list);
508
509 if (it->cache == &mrt->mfc6_unres_queue)
510 goto end_of_list;
511
512 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
513
514 while (++it->ct < MFC6_LINES) {
515 it->cache = &mrt->mfc6_cache_array[it->ct];
516 if (list_empty(it->cache))
517 continue;
518 return list_first_entry(it->cache, struct mfc6_cache, list);
519 }
520
521 /* exhausted cache_array, show unresolved */
522 read_unlock(&mrt_lock);
523 it->cache = &mrt->mfc6_unres_queue;
524 it->ct = 0;
525
526 spin_lock_bh(&mfc_unres_lock);
527 if (!list_empty(it->cache))
528 return list_first_entry(it->cache, struct mfc6_cache, list);
529
530 end_of_list:
531 spin_unlock_bh(&mfc_unres_lock);
532 it->cache = NULL;
533
534 return NULL;
535 }
536
537 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
538 {
539 struct ipmr_mfc_iter *it = seq->private;
540 struct mr6_table *mrt = it->mrt;
541
542 if (it->cache == &mrt->mfc6_unres_queue)
543 spin_unlock_bh(&mfc_unres_lock);
544 else if (it->cache == mrt->mfc6_cache_array)
545 read_unlock(&mrt_lock);
546 }
547
548 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
549 {
550 int n;
551
552 if (v == SEQ_START_TOKEN) {
553 seq_puts(seq,
554 "Group "
555 "Origin "
556 "Iif Pkts Bytes Wrong Oifs\n");
557 } else {
558 const struct mfc6_cache *mfc = v;
559 const struct ipmr_mfc_iter *it = seq->private;
560 struct mr6_table *mrt = it->mrt;
561
562 seq_printf(seq, "%pI6 %pI6 %-3hd",
563 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
564 mfc->mf6c_parent);
565
566 if (it->cache != &mrt->mfc6_unres_queue) {
567 seq_printf(seq, " %8lu %8lu %8lu",
568 mfc->mfc_un.res.pkt,
569 mfc->mfc_un.res.bytes,
570 mfc->mfc_un.res.wrong_if);
571 for (n = mfc->mfc_un.res.minvif;
572 n < mfc->mfc_un.res.maxvif; n++) {
573 if (MIF_EXISTS(mrt, n) &&
574 mfc->mfc_un.res.ttls[n] < 255)
575 seq_printf(seq,
576 " %2d:%-3d",
577 n, mfc->mfc_un.res.ttls[n]);
578 }
579 } else {
580 /* unresolved mfc_caches don't contain
581 * pkt, bytes and wrong_if values
582 */
583 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
584 }
585 seq_putc(seq, '\n');
586 }
587 return 0;
588 }
589
590 static const struct seq_operations ipmr_mfc_seq_ops = {
591 .start = ipmr_mfc_seq_start,
592 .next = ipmr_mfc_seq_next,
593 .stop = ipmr_mfc_seq_stop,
594 .show = ipmr_mfc_seq_show,
595 };
596
597 static int ipmr_mfc_open(struct inode *inode, struct file *file)
598 {
599 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
600 sizeof(struct ipmr_mfc_iter));
601 }
602
603 static const struct file_operations ip6mr_mfc_fops = {
604 .owner = THIS_MODULE,
605 .open = ipmr_mfc_open,
606 .read = seq_read,
607 .llseek = seq_lseek,
608 .release = seq_release_net,
609 };
610 #endif
611
612 #ifdef CONFIG_IPV6_PIMSM_V2
613
614 static int pim6_rcv(struct sk_buff *skb)
615 {
616 struct pimreghdr *pim;
617 struct ipv6hdr *encap;
618 struct net_device *reg_dev = NULL;
619 struct net *net = dev_net(skb->dev);
620 struct mr6_table *mrt;
621 struct flowi6 fl6 = {
622 .flowi6_iif = skb->dev->ifindex,
623 .flowi6_mark = skb->mark,
624 };
625 int reg_vif_num;
626
627 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
628 goto drop;
629
630 pim = (struct pimreghdr *)skb_transport_header(skb);
631 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
632 (pim->flags & PIM_NULL_REGISTER) ||
633 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
634 sizeof(*pim), IPPROTO_PIM,
635 csum_partial((void *)pim, sizeof(*pim), 0)) &&
636 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
637 goto drop;
638
639 /* check if the inner packet is destined to mcast group */
640 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
641 sizeof(*pim));
642
643 if (!ipv6_addr_is_multicast(&encap->daddr) ||
644 encap->payload_len == 0 ||
645 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
646 goto drop;
647
648 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
649 goto drop;
650 reg_vif_num = mrt->mroute_reg_vif_num;
651
652 read_lock(&mrt_lock);
653 if (reg_vif_num >= 0)
654 reg_dev = mrt->vif6_table[reg_vif_num].dev;
655 if (reg_dev)
656 dev_hold(reg_dev);
657 read_unlock(&mrt_lock);
658
659 if (reg_dev == NULL)
660 goto drop;
661
662 skb->mac_header = skb->network_header;
663 skb_pull(skb, (u8 *)encap - skb->data);
664 skb_reset_network_header(skb);
665 skb->protocol = htons(ETH_P_IPV6);
666 skb->ip_summed = CHECKSUM_NONE;
667 skb->pkt_type = PACKET_HOST;
668
669 skb_tunnel_rx(skb, reg_dev);
670
671 netif_rx(skb);
672
673 dev_put(reg_dev);
674 return 0;
675 drop:
676 kfree_skb(skb);
677 return 0;
678 }
679
680 static const struct inet6_protocol pim6_protocol = {
681 .handler = pim6_rcv,
682 };
683
684 /* Service routines creating virtual interfaces: PIMREG */
685
686 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
687 struct net_device *dev)
688 {
689 struct net *net = dev_net(dev);
690 struct mr6_table *mrt;
691 struct flowi6 fl6 = {
692 .flowi6_oif = dev->ifindex,
693 .flowi6_iif = skb->skb_iif,
694 .flowi6_mark = skb->mark,
695 };
696 int err;
697
698 err = ip6mr_fib_lookup(net, &fl6, &mrt);
699 if (err < 0)
700 return err;
701
702 read_lock(&mrt_lock);
703 dev->stats.tx_bytes += skb->len;
704 dev->stats.tx_packets++;
705 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
706 read_unlock(&mrt_lock);
707 kfree_skb(skb);
708 return NETDEV_TX_OK;
709 }
710
711 static const struct net_device_ops reg_vif_netdev_ops = {
712 .ndo_start_xmit = reg_vif_xmit,
713 };
714
715 static void reg_vif_setup(struct net_device *dev)
716 {
717 dev->type = ARPHRD_PIMREG;
718 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
719 dev->flags = IFF_NOARP;
720 dev->netdev_ops = &reg_vif_netdev_ops;
721 dev->destructor = free_netdev;
722 dev->features |= NETIF_F_NETNS_LOCAL;
723 }
724
725 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
726 {
727 struct net_device *dev;
728 char name[IFNAMSIZ];
729
730 if (mrt->id == RT6_TABLE_DFLT)
731 sprintf(name, "pim6reg");
732 else
733 sprintf(name, "pim6reg%u", mrt->id);
734
735 dev = alloc_netdev(0, name, reg_vif_setup);
736 if (dev == NULL)
737 return NULL;
738
739 dev_net_set(dev, net);
740
741 if (register_netdevice(dev)) {
742 free_netdev(dev);
743 return NULL;
744 }
745 dev->iflink = 0;
746
747 if (dev_open(dev))
748 goto failure;
749
750 dev_hold(dev);
751 return dev;
752
753 failure:
754 /* allow the register to be completed before unregistering. */
755 rtnl_unlock();
756 rtnl_lock();
757
758 unregister_netdevice(dev);
759 return NULL;
760 }
761 #endif
762
763 /*
764 * Delete a VIF entry
765 */
766
767 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
768 {
769 struct mif_device *v;
770 struct net_device *dev;
771 struct inet6_dev *in6_dev;
772
773 if (vifi < 0 || vifi >= mrt->maxvif)
774 return -EADDRNOTAVAIL;
775
776 v = &mrt->vif6_table[vifi];
777
778 write_lock_bh(&mrt_lock);
779 dev = v->dev;
780 v->dev = NULL;
781
782 if (!dev) {
783 write_unlock_bh(&mrt_lock);
784 return -EADDRNOTAVAIL;
785 }
786
787 #ifdef CONFIG_IPV6_PIMSM_V2
788 if (vifi == mrt->mroute_reg_vif_num)
789 mrt->mroute_reg_vif_num = -1;
790 #endif
791
792 if (vifi + 1 == mrt->maxvif) {
793 int tmp;
794 for (tmp = vifi - 1; tmp >= 0; tmp--) {
795 if (MIF_EXISTS(mrt, tmp))
796 break;
797 }
798 mrt->maxvif = tmp + 1;
799 }
800
801 write_unlock_bh(&mrt_lock);
802
803 dev_set_allmulti(dev, -1);
804
805 in6_dev = __in6_dev_get(dev);
806 if (in6_dev)
807 in6_dev->cnf.mc_forwarding--;
808
809 if (v->flags & MIFF_REGISTER)
810 unregister_netdevice_queue(dev, head);
811
812 dev_put(dev);
813 return 0;
814 }
815
816 static inline void ip6mr_cache_free(struct mfc6_cache *c)
817 {
818 kmem_cache_free(mrt_cachep, c);
819 }
820
821 /* Destroy an unresolved cache entry, killing queued skbs
822 and reporting error to netlink readers.
823 */
824
825 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
826 {
827 struct net *net = read_pnet(&mrt->net);
828 struct sk_buff *skb;
829
830 atomic_dec(&mrt->cache_resolve_queue_len);
831
832 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
833 if (ipv6_hdr(skb)->version == 0) {
834 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
835 nlh->nlmsg_type = NLMSG_ERROR;
836 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
837 skb_trim(skb, nlh->nlmsg_len);
838 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
839 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
840 } else
841 kfree_skb(skb);
842 }
843
844 ip6mr_cache_free(c);
845 }
846
847
848 /* Timer process for all the unresolved queue. */
849
850 static void ipmr_do_expire_process(struct mr6_table *mrt)
851 {
852 unsigned long now = jiffies;
853 unsigned long expires = 10 * HZ;
854 struct mfc6_cache *c, *next;
855
856 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
857 if (time_after(c->mfc_un.unres.expires, now)) {
858 /* not yet... */
859 unsigned long interval = c->mfc_un.unres.expires - now;
860 if (interval < expires)
861 expires = interval;
862 continue;
863 }
864
865 list_del(&c->list);
866 ip6mr_destroy_unres(mrt, c);
867 }
868
869 if (!list_empty(&mrt->mfc6_unres_queue))
870 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
871 }
872
873 static void ipmr_expire_process(unsigned long arg)
874 {
875 struct mr6_table *mrt = (struct mr6_table *)arg;
876
877 if (!spin_trylock(&mfc_unres_lock)) {
878 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
879 return;
880 }
881
882 if (!list_empty(&mrt->mfc6_unres_queue))
883 ipmr_do_expire_process(mrt);
884
885 spin_unlock(&mfc_unres_lock);
886 }
887
888 /* Fill oifs list. It is called under write locked mrt_lock. */
889
890 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
891 unsigned char *ttls)
892 {
893 int vifi;
894
895 cache->mfc_un.res.minvif = MAXMIFS;
896 cache->mfc_un.res.maxvif = 0;
897 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
898
899 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
900 if (MIF_EXISTS(mrt, vifi) &&
901 ttls[vifi] && ttls[vifi] < 255) {
902 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
903 if (cache->mfc_un.res.minvif > vifi)
904 cache->mfc_un.res.minvif = vifi;
905 if (cache->mfc_un.res.maxvif <= vifi)
906 cache->mfc_un.res.maxvif = vifi + 1;
907 }
908 }
909 }
910
911 static int mif6_add(struct net *net, struct mr6_table *mrt,
912 struct mif6ctl *vifc, int mrtsock)
913 {
914 int vifi = vifc->mif6c_mifi;
915 struct mif_device *v = &mrt->vif6_table[vifi];
916 struct net_device *dev;
917 struct inet6_dev *in6_dev;
918 int err;
919
920 /* Is vif busy ? */
921 if (MIF_EXISTS(mrt, vifi))
922 return -EADDRINUSE;
923
924 switch (vifc->mif6c_flags) {
925 #ifdef CONFIG_IPV6_PIMSM_V2
926 case MIFF_REGISTER:
927 /*
928 * Special Purpose VIF in PIM
929 * All the packets will be sent to the daemon
930 */
931 if (mrt->mroute_reg_vif_num >= 0)
932 return -EADDRINUSE;
933 dev = ip6mr_reg_vif(net, mrt);
934 if (!dev)
935 return -ENOBUFS;
936 err = dev_set_allmulti(dev, 1);
937 if (err) {
938 unregister_netdevice(dev);
939 dev_put(dev);
940 return err;
941 }
942 break;
943 #endif
944 case 0:
945 dev = dev_get_by_index(net, vifc->mif6c_pifi);
946 if (!dev)
947 return -EADDRNOTAVAIL;
948 err = dev_set_allmulti(dev, 1);
949 if (err) {
950 dev_put(dev);
951 return err;
952 }
953 break;
954 default:
955 return -EINVAL;
956 }
957
958 in6_dev = __in6_dev_get(dev);
959 if (in6_dev)
960 in6_dev->cnf.mc_forwarding++;
961
962 /*
963 * Fill in the VIF structures
964 */
965 v->rate_limit = vifc->vifc_rate_limit;
966 v->flags = vifc->mif6c_flags;
967 if (!mrtsock)
968 v->flags |= VIFF_STATIC;
969 v->threshold = vifc->vifc_threshold;
970 v->bytes_in = 0;
971 v->bytes_out = 0;
972 v->pkt_in = 0;
973 v->pkt_out = 0;
974 v->link = dev->ifindex;
975 if (v->flags & MIFF_REGISTER)
976 v->link = dev->iflink;
977
978 /* And finish update writing critical data */
979 write_lock_bh(&mrt_lock);
980 v->dev = dev;
981 #ifdef CONFIG_IPV6_PIMSM_V2
982 if (v->flags & MIFF_REGISTER)
983 mrt->mroute_reg_vif_num = vifi;
984 #endif
985 if (vifi + 1 > mrt->maxvif)
986 mrt->maxvif = vifi + 1;
987 write_unlock_bh(&mrt_lock);
988 return 0;
989 }
990
991 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
992 const struct in6_addr *origin,
993 const struct in6_addr *mcastgrp)
994 {
995 int line = MFC6_HASH(mcastgrp, origin);
996 struct mfc6_cache *c;
997
998 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
999 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1000 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1001 return c;
1002 }
1003 return NULL;
1004 }
1005
1006 /*
1007 * Allocate a multicast cache entry
1008 */
1009 static struct mfc6_cache *ip6mr_cache_alloc(void)
1010 {
1011 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1012 if (c == NULL)
1013 return NULL;
1014 c->mfc_un.res.minvif = MAXMIFS;
1015 return c;
1016 }
1017
1018 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1019 {
1020 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1021 if (c == NULL)
1022 return NULL;
1023 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1024 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1025 return c;
1026 }
1027
1028 /*
1029 * A cache entry has gone into a resolved state from queued
1030 */
1031
1032 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1033 struct mfc6_cache *uc, struct mfc6_cache *c)
1034 {
1035 struct sk_buff *skb;
1036
1037 /*
1038 * Play the pending entries through our router
1039 */
1040
1041 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1042 if (ipv6_hdr(skb)->version == 0) {
1043 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1044
1045 if (__ip6mr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) {
1046 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1047 } else {
1048 nlh->nlmsg_type = NLMSG_ERROR;
1049 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
1050 skb_trim(skb, nlh->nlmsg_len);
1051 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
1052 }
1053 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
1054 } else
1055 ip6_mr_forward(net, mrt, skb, c);
1056 }
1057 }
1058
1059 /*
1060 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1061 * expects the following bizarre scheme.
1062 *
1063 * Called under mrt_lock.
1064 */
1065
1066 static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1067 mifi_t mifi, int assert)
1068 {
1069 struct sk_buff *skb;
1070 struct mrt6msg *msg;
1071 int ret;
1072
1073 #ifdef CONFIG_IPV6_PIMSM_V2
1074 if (assert == MRT6MSG_WHOLEPKT)
1075 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1076 +sizeof(*msg));
1077 else
1078 #endif
1079 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1080
1081 if (!skb)
1082 return -ENOBUFS;
1083
1084 /* I suppose that internal messages
1085 * do not require checksums */
1086
1087 skb->ip_summed = CHECKSUM_UNNECESSARY;
1088
1089 #ifdef CONFIG_IPV6_PIMSM_V2
1090 if (assert == MRT6MSG_WHOLEPKT) {
1091 /* Ugly, but we have no choice with this interface.
1092 Duplicate old header, fix length etc.
1093 And all this only to mangle msg->im6_msgtype and
1094 to set msg->im6_mbz to "mbz" :-)
1095 */
1096 skb_push(skb, -skb_network_offset(pkt));
1097
1098 skb_push(skb, sizeof(*msg));
1099 skb_reset_transport_header(skb);
1100 msg = (struct mrt6msg *)skb_transport_header(skb);
1101 msg->im6_mbz = 0;
1102 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1103 msg->im6_mif = mrt->mroute_reg_vif_num;
1104 msg->im6_pad = 0;
1105 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1106 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1107
1108 skb->ip_summed = CHECKSUM_UNNECESSARY;
1109 } else
1110 #endif
1111 {
1112 /*
1113 * Copy the IP header
1114 */
1115
1116 skb_put(skb, sizeof(struct ipv6hdr));
1117 skb_reset_network_header(skb);
1118 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1119
1120 /*
1121 * Add our header
1122 */
1123 skb_put(skb, sizeof(*msg));
1124 skb_reset_transport_header(skb);
1125 msg = (struct mrt6msg *)skb_transport_header(skb);
1126
1127 msg->im6_mbz = 0;
1128 msg->im6_msgtype = assert;
1129 msg->im6_mif = mifi;
1130 msg->im6_pad = 0;
1131 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1132 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1133
1134 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1135 skb->ip_summed = CHECKSUM_UNNECESSARY;
1136 }
1137
1138 if (mrt->mroute6_sk == NULL) {
1139 kfree_skb(skb);
1140 return -EINVAL;
1141 }
1142
1143 /*
1144 * Deliver to user space multicast routing algorithms
1145 */
1146 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1147 if (ret < 0) {
1148 if (net_ratelimit())
1149 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
1150 kfree_skb(skb);
1151 }
1152
1153 return ret;
1154 }
1155
1156 /*
1157 * Queue a packet for resolution. It gets locked cache entry!
1158 */
1159
1160 static int
1161 ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1162 {
1163 bool found = false;
1164 int err;
1165 struct mfc6_cache *c;
1166
1167 spin_lock_bh(&mfc_unres_lock);
1168 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1169 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1170 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1171 found = true;
1172 break;
1173 }
1174 }
1175
1176 if (!found) {
1177 /*
1178 * Create a new entry if allowable
1179 */
1180
1181 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1182 (c = ip6mr_cache_alloc_unres()) == NULL) {
1183 spin_unlock_bh(&mfc_unres_lock);
1184
1185 kfree_skb(skb);
1186 return -ENOBUFS;
1187 }
1188
1189 /*
1190 * Fill in the new cache entry
1191 */
1192 c->mf6c_parent = -1;
1193 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1194 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1195
1196 /*
1197 * Reflect first query at pim6sd
1198 */
1199 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1200 if (err < 0) {
1201 /* If the report failed throw the cache entry
1202 out - Brad Parker
1203 */
1204 spin_unlock_bh(&mfc_unres_lock);
1205
1206 ip6mr_cache_free(c);
1207 kfree_skb(skb);
1208 return err;
1209 }
1210
1211 atomic_inc(&mrt->cache_resolve_queue_len);
1212 list_add(&c->list, &mrt->mfc6_unres_queue);
1213
1214 ipmr_do_expire_process(mrt);
1215 }
1216
1217 /*
1218 * See if we can append the packet
1219 */
1220 if (c->mfc_un.unres.unresolved.qlen > 3) {
1221 kfree_skb(skb);
1222 err = -ENOBUFS;
1223 } else {
1224 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1225 err = 0;
1226 }
1227
1228 spin_unlock_bh(&mfc_unres_lock);
1229 return err;
1230 }
1231
1232 /*
1233 * MFC6 cache manipulation by user space
1234 */
1235
1236 static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc)
1237 {
1238 int line;
1239 struct mfc6_cache *c, *next;
1240
1241 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1242
1243 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1244 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1245 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1246 write_lock_bh(&mrt_lock);
1247 list_del(&c->list);
1248 write_unlock_bh(&mrt_lock);
1249
1250 ip6mr_cache_free(c);
1251 return 0;
1252 }
1253 }
1254 return -ENOENT;
1255 }
1256
1257 static int ip6mr_device_event(struct notifier_block *this,
1258 unsigned long event, void *ptr)
1259 {
1260 struct net_device *dev = ptr;
1261 struct net *net = dev_net(dev);
1262 struct mr6_table *mrt;
1263 struct mif_device *v;
1264 int ct;
1265 LIST_HEAD(list);
1266
1267 if (event != NETDEV_UNREGISTER)
1268 return NOTIFY_DONE;
1269
1270 ip6mr_for_each_table(mrt, net) {
1271 v = &mrt->vif6_table[0];
1272 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1273 if (v->dev == dev)
1274 mif6_delete(mrt, ct, &list);
1275 }
1276 }
1277 unregister_netdevice_many(&list);
1278
1279 return NOTIFY_DONE;
1280 }
1281
1282 static struct notifier_block ip6_mr_notifier = {
1283 .notifier_call = ip6mr_device_event
1284 };
1285
1286 /*
1287 * Setup for IP multicast routing
1288 */
1289
1290 static int __net_init ip6mr_net_init(struct net *net)
1291 {
1292 int err;
1293
1294 err = ip6mr_rules_init(net);
1295 if (err < 0)
1296 goto fail;
1297
1298 #ifdef CONFIG_PROC_FS
1299 err = -ENOMEM;
1300 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
1301 goto proc_vif_fail;
1302 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops))
1303 goto proc_cache_fail;
1304 #endif
1305
1306 return 0;
1307
1308 #ifdef CONFIG_PROC_FS
1309 proc_cache_fail:
1310 proc_net_remove(net, "ip6_mr_vif");
1311 proc_vif_fail:
1312 ip6mr_rules_exit(net);
1313 #endif
1314 fail:
1315 return err;
1316 }
1317
1318 static void __net_exit ip6mr_net_exit(struct net *net)
1319 {
1320 #ifdef CONFIG_PROC_FS
1321 proc_net_remove(net, "ip6_mr_cache");
1322 proc_net_remove(net, "ip6_mr_vif");
1323 #endif
1324 ip6mr_rules_exit(net);
1325 }
1326
1327 static struct pernet_operations ip6mr_net_ops = {
1328 .init = ip6mr_net_init,
1329 .exit = ip6mr_net_exit,
1330 };
1331
1332 int __init ip6_mr_init(void)
1333 {
1334 int err;
1335
1336 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1337 sizeof(struct mfc6_cache),
1338 0, SLAB_HWCACHE_ALIGN,
1339 NULL);
1340 if (!mrt_cachep)
1341 return -ENOMEM;
1342
1343 err = register_pernet_subsys(&ip6mr_net_ops);
1344 if (err)
1345 goto reg_pernet_fail;
1346
1347 err = register_netdevice_notifier(&ip6_mr_notifier);
1348 if (err)
1349 goto reg_notif_fail;
1350 #ifdef CONFIG_IPV6_PIMSM_V2
1351 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1352 printk(KERN_ERR "ip6_mr_init: can't add PIM protocol\n");
1353 err = -EAGAIN;
1354 goto add_proto_fail;
1355 }
1356 #endif
1357 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, ip6mr_rtm_dumproute);
1358 return 0;
1359 #ifdef CONFIG_IPV6_PIMSM_V2
1360 add_proto_fail:
1361 unregister_netdevice_notifier(&ip6_mr_notifier);
1362 #endif
1363 reg_notif_fail:
1364 unregister_pernet_subsys(&ip6mr_net_ops);
1365 reg_pernet_fail:
1366 kmem_cache_destroy(mrt_cachep);
1367 return err;
1368 }
1369
1370 void ip6_mr_cleanup(void)
1371 {
1372 unregister_netdevice_notifier(&ip6_mr_notifier);
1373 unregister_pernet_subsys(&ip6mr_net_ops);
1374 kmem_cache_destroy(mrt_cachep);
1375 }
1376
1377 static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1378 struct mf6cctl *mfc, int mrtsock)
1379 {
1380 bool found = false;
1381 int line;
1382 struct mfc6_cache *uc, *c;
1383 unsigned char ttls[MAXMIFS];
1384 int i;
1385
1386 if (mfc->mf6cc_parent >= MAXMIFS)
1387 return -ENFILE;
1388
1389 memset(ttls, 255, MAXMIFS);
1390 for (i = 0; i < MAXMIFS; i++) {
1391 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1392 ttls[i] = 1;
1393
1394 }
1395
1396 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1397
1398 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1399 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1400 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1401 found = true;
1402 break;
1403 }
1404 }
1405
1406 if (found) {
1407 write_lock_bh(&mrt_lock);
1408 c->mf6c_parent = mfc->mf6cc_parent;
1409 ip6mr_update_thresholds(mrt, c, ttls);
1410 if (!mrtsock)
1411 c->mfc_flags |= MFC_STATIC;
1412 write_unlock_bh(&mrt_lock);
1413 return 0;
1414 }
1415
1416 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1417 return -EINVAL;
1418
1419 c = ip6mr_cache_alloc();
1420 if (c == NULL)
1421 return -ENOMEM;
1422
1423 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1424 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1425 c->mf6c_parent = mfc->mf6cc_parent;
1426 ip6mr_update_thresholds(mrt, c, ttls);
1427 if (!mrtsock)
1428 c->mfc_flags |= MFC_STATIC;
1429
1430 write_lock_bh(&mrt_lock);
1431 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1432 write_unlock_bh(&mrt_lock);
1433
1434 /*
1435 * Check to see if we resolved a queued list. If so we
1436 * need to send on the frames and tidy up.
1437 */
1438 found = false;
1439 spin_lock_bh(&mfc_unres_lock);
1440 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1441 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1442 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1443 list_del(&uc->list);
1444 atomic_dec(&mrt->cache_resolve_queue_len);
1445 found = true;
1446 break;
1447 }
1448 }
1449 if (list_empty(&mrt->mfc6_unres_queue))
1450 del_timer(&mrt->ipmr_expire_timer);
1451 spin_unlock_bh(&mfc_unres_lock);
1452
1453 if (found) {
1454 ip6mr_cache_resolve(net, mrt, uc, c);
1455 ip6mr_cache_free(uc);
1456 }
1457 return 0;
1458 }
1459
1460 /*
1461 * Close the multicast socket, and clear the vif tables etc
1462 */
1463
1464 static void mroute_clean_tables(struct mr6_table *mrt)
1465 {
1466 int i;
1467 LIST_HEAD(list);
1468 struct mfc6_cache *c, *next;
1469
1470 /*
1471 * Shut down all active vif entries
1472 */
1473 for (i = 0; i < mrt->maxvif; i++) {
1474 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1475 mif6_delete(mrt, i, &list);
1476 }
1477 unregister_netdevice_many(&list);
1478
1479 /*
1480 * Wipe the cache
1481 */
1482 for (i = 0; i < MFC6_LINES; i++) {
1483 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1484 if (c->mfc_flags & MFC_STATIC)
1485 continue;
1486 write_lock_bh(&mrt_lock);
1487 list_del(&c->list);
1488 write_unlock_bh(&mrt_lock);
1489
1490 ip6mr_cache_free(c);
1491 }
1492 }
1493
1494 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1495 spin_lock_bh(&mfc_unres_lock);
1496 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1497 list_del(&c->list);
1498 ip6mr_destroy_unres(mrt, c);
1499 }
1500 spin_unlock_bh(&mfc_unres_lock);
1501 }
1502 }
1503
1504 static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1505 {
1506 int err = 0;
1507 struct net *net = sock_net(sk);
1508
1509 rtnl_lock();
1510 write_lock_bh(&mrt_lock);
1511 if (likely(mrt->mroute6_sk == NULL)) {
1512 mrt->mroute6_sk = sk;
1513 net->ipv6.devconf_all->mc_forwarding++;
1514 }
1515 else
1516 err = -EADDRINUSE;
1517 write_unlock_bh(&mrt_lock);
1518
1519 rtnl_unlock();
1520
1521 return err;
1522 }
1523
1524 int ip6mr_sk_done(struct sock *sk)
1525 {
1526 int err = -EACCES;
1527 struct net *net = sock_net(sk);
1528 struct mr6_table *mrt;
1529
1530 rtnl_lock();
1531 ip6mr_for_each_table(mrt, net) {
1532 if (sk == mrt->mroute6_sk) {
1533 write_lock_bh(&mrt_lock);
1534 mrt->mroute6_sk = NULL;
1535 net->ipv6.devconf_all->mc_forwarding--;
1536 write_unlock_bh(&mrt_lock);
1537
1538 mroute_clean_tables(mrt);
1539 err = 0;
1540 break;
1541 }
1542 }
1543 rtnl_unlock();
1544
1545 return err;
1546 }
1547
1548 struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1549 {
1550 struct mr6_table *mrt;
1551 struct flowi6 fl6 = {
1552 .flowi6_iif = skb->skb_iif,
1553 .flowi6_oif = skb->dev->ifindex,
1554 .flowi6_mark = skb->mark,
1555 };
1556
1557 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1558 return NULL;
1559
1560 return mrt->mroute6_sk;
1561 }
1562
1563 /*
1564 * Socket options and virtual interface manipulation. The whole
1565 * virtual interface system is a complete heap, but unfortunately
1566 * that's how BSD mrouted happens to think. Maybe one day with a proper
1567 * MOSPF/PIM router set up we can clean this up.
1568 */
1569
1570 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1571 {
1572 int ret;
1573 struct mif6ctl vif;
1574 struct mf6cctl mfc;
1575 mifi_t mifi;
1576 struct net *net = sock_net(sk);
1577 struct mr6_table *mrt;
1578
1579 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1580 if (mrt == NULL)
1581 return -ENOENT;
1582
1583 if (optname != MRT6_INIT) {
1584 if (sk != mrt->mroute6_sk && !capable(CAP_NET_ADMIN))
1585 return -EACCES;
1586 }
1587
1588 switch (optname) {
1589 case MRT6_INIT:
1590 if (sk->sk_type != SOCK_RAW ||
1591 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1592 return -EOPNOTSUPP;
1593 if (optlen < sizeof(int))
1594 return -EINVAL;
1595
1596 return ip6mr_sk_init(mrt, sk);
1597
1598 case MRT6_DONE:
1599 return ip6mr_sk_done(sk);
1600
1601 case MRT6_ADD_MIF:
1602 if (optlen < sizeof(vif))
1603 return -EINVAL;
1604 if (copy_from_user(&vif, optval, sizeof(vif)))
1605 return -EFAULT;
1606 if (vif.mif6c_mifi >= MAXMIFS)
1607 return -ENFILE;
1608 rtnl_lock();
1609 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1610 rtnl_unlock();
1611 return ret;
1612
1613 case MRT6_DEL_MIF:
1614 if (optlen < sizeof(mifi_t))
1615 return -EINVAL;
1616 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1617 return -EFAULT;
1618 rtnl_lock();
1619 ret = mif6_delete(mrt, mifi, NULL);
1620 rtnl_unlock();
1621 return ret;
1622
1623 /*
1624 * Manipulate the forwarding caches. These live
1625 * in a sort of kernel/user symbiosis.
1626 */
1627 case MRT6_ADD_MFC:
1628 case MRT6_DEL_MFC:
1629 if (optlen < sizeof(mfc))
1630 return -EINVAL;
1631 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1632 return -EFAULT;
1633 rtnl_lock();
1634 if (optname == MRT6_DEL_MFC)
1635 ret = ip6mr_mfc_delete(mrt, &mfc);
1636 else
1637 ret = ip6mr_mfc_add(net, mrt, &mfc, sk == mrt->mroute6_sk);
1638 rtnl_unlock();
1639 return ret;
1640
1641 /*
1642 * Control PIM assert (to activate pim will activate assert)
1643 */
1644 case MRT6_ASSERT:
1645 {
1646 int v;
1647 if (get_user(v, (int __user *)optval))
1648 return -EFAULT;
1649 mrt->mroute_do_assert = !!v;
1650 return 0;
1651 }
1652
1653 #ifdef CONFIG_IPV6_PIMSM_V2
1654 case MRT6_PIM:
1655 {
1656 int v;
1657 if (get_user(v, (int __user *)optval))
1658 return -EFAULT;
1659 v = !!v;
1660 rtnl_lock();
1661 ret = 0;
1662 if (v != mrt->mroute_do_pim) {
1663 mrt->mroute_do_pim = v;
1664 mrt->mroute_do_assert = v;
1665 }
1666 rtnl_unlock();
1667 return ret;
1668 }
1669
1670 #endif
1671 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1672 case MRT6_TABLE:
1673 {
1674 u32 v;
1675
1676 if (optlen != sizeof(u32))
1677 return -EINVAL;
1678 if (get_user(v, (u32 __user *)optval))
1679 return -EFAULT;
1680 if (sk == mrt->mroute6_sk)
1681 return -EBUSY;
1682
1683 rtnl_lock();
1684 ret = 0;
1685 if (!ip6mr_new_table(net, v))
1686 ret = -ENOMEM;
1687 raw6_sk(sk)->ip6mr_table = v;
1688 rtnl_unlock();
1689 return ret;
1690 }
1691 #endif
1692 /*
1693 * Spurious command, or MRT6_VERSION which you cannot
1694 * set.
1695 */
1696 default:
1697 return -ENOPROTOOPT;
1698 }
1699 }
1700
1701 /*
1702 * Getsock opt support for the multicast routing system.
1703 */
1704
1705 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1706 int __user *optlen)
1707 {
1708 int olr;
1709 int val;
1710 struct net *net = sock_net(sk);
1711 struct mr6_table *mrt;
1712
1713 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1714 if (mrt == NULL)
1715 return -ENOENT;
1716
1717 switch (optname) {
1718 case MRT6_VERSION:
1719 val = 0x0305;
1720 break;
1721 #ifdef CONFIG_IPV6_PIMSM_V2
1722 case MRT6_PIM:
1723 val = mrt->mroute_do_pim;
1724 break;
1725 #endif
1726 case MRT6_ASSERT:
1727 val = mrt->mroute_do_assert;
1728 break;
1729 default:
1730 return -ENOPROTOOPT;
1731 }
1732
1733 if (get_user(olr, optlen))
1734 return -EFAULT;
1735
1736 olr = min_t(int, olr, sizeof(int));
1737 if (olr < 0)
1738 return -EINVAL;
1739
1740 if (put_user(olr, optlen))
1741 return -EFAULT;
1742 if (copy_to_user(optval, &val, olr))
1743 return -EFAULT;
1744 return 0;
1745 }
1746
1747 /*
1748 * The IP multicast ioctl support routines.
1749 */
1750
1751 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1752 {
1753 struct sioc_sg_req6 sr;
1754 struct sioc_mif_req6 vr;
1755 struct mif_device *vif;
1756 struct mfc6_cache *c;
1757 struct net *net = sock_net(sk);
1758 struct mr6_table *mrt;
1759
1760 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1761 if (mrt == NULL)
1762 return -ENOENT;
1763
1764 switch (cmd) {
1765 case SIOCGETMIFCNT_IN6:
1766 if (copy_from_user(&vr, arg, sizeof(vr)))
1767 return -EFAULT;
1768 if (vr.mifi >= mrt->maxvif)
1769 return -EINVAL;
1770 read_lock(&mrt_lock);
1771 vif = &mrt->vif6_table[vr.mifi];
1772 if (MIF_EXISTS(mrt, vr.mifi)) {
1773 vr.icount = vif->pkt_in;
1774 vr.ocount = vif->pkt_out;
1775 vr.ibytes = vif->bytes_in;
1776 vr.obytes = vif->bytes_out;
1777 read_unlock(&mrt_lock);
1778
1779 if (copy_to_user(arg, &vr, sizeof(vr)))
1780 return -EFAULT;
1781 return 0;
1782 }
1783 read_unlock(&mrt_lock);
1784 return -EADDRNOTAVAIL;
1785 case SIOCGETSGCNT_IN6:
1786 if (copy_from_user(&sr, arg, sizeof(sr)))
1787 return -EFAULT;
1788
1789 read_lock(&mrt_lock);
1790 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1791 if (c) {
1792 sr.pktcnt = c->mfc_un.res.pkt;
1793 sr.bytecnt = c->mfc_un.res.bytes;
1794 sr.wrong_if = c->mfc_un.res.wrong_if;
1795 read_unlock(&mrt_lock);
1796
1797 if (copy_to_user(arg, &sr, sizeof(sr)))
1798 return -EFAULT;
1799 return 0;
1800 }
1801 read_unlock(&mrt_lock);
1802 return -EADDRNOTAVAIL;
1803 default:
1804 return -ENOIOCTLCMD;
1805 }
1806 }
1807
1808 #ifdef CONFIG_COMPAT
1809 struct compat_sioc_sg_req6 {
1810 struct sockaddr_in6 src;
1811 struct sockaddr_in6 grp;
1812 compat_ulong_t pktcnt;
1813 compat_ulong_t bytecnt;
1814 compat_ulong_t wrong_if;
1815 };
1816
1817 struct compat_sioc_mif_req6 {
1818 mifi_t mifi;
1819 compat_ulong_t icount;
1820 compat_ulong_t ocount;
1821 compat_ulong_t ibytes;
1822 compat_ulong_t obytes;
1823 };
1824
1825 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1826 {
1827 struct compat_sioc_sg_req6 sr;
1828 struct compat_sioc_mif_req6 vr;
1829 struct mif_device *vif;
1830 struct mfc6_cache *c;
1831 struct net *net = sock_net(sk);
1832 struct mr6_table *mrt;
1833
1834 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1835 if (mrt == NULL)
1836 return -ENOENT;
1837
1838 switch (cmd) {
1839 case SIOCGETMIFCNT_IN6:
1840 if (copy_from_user(&vr, arg, sizeof(vr)))
1841 return -EFAULT;
1842 if (vr.mifi >= mrt->maxvif)
1843 return -EINVAL;
1844 read_lock(&mrt_lock);
1845 vif = &mrt->vif6_table[vr.mifi];
1846 if (MIF_EXISTS(mrt, vr.mifi)) {
1847 vr.icount = vif->pkt_in;
1848 vr.ocount = vif->pkt_out;
1849 vr.ibytes = vif->bytes_in;
1850 vr.obytes = vif->bytes_out;
1851 read_unlock(&mrt_lock);
1852
1853 if (copy_to_user(arg, &vr, sizeof(vr)))
1854 return -EFAULT;
1855 return 0;
1856 }
1857 read_unlock(&mrt_lock);
1858 return -EADDRNOTAVAIL;
1859 case SIOCGETSGCNT_IN6:
1860 if (copy_from_user(&sr, arg, sizeof(sr)))
1861 return -EFAULT;
1862
1863 read_lock(&mrt_lock);
1864 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1865 if (c) {
1866 sr.pktcnt = c->mfc_un.res.pkt;
1867 sr.bytecnt = c->mfc_un.res.bytes;
1868 sr.wrong_if = c->mfc_un.res.wrong_if;
1869 read_unlock(&mrt_lock);
1870
1871 if (copy_to_user(arg, &sr, sizeof(sr)))
1872 return -EFAULT;
1873 return 0;
1874 }
1875 read_unlock(&mrt_lock);
1876 return -EADDRNOTAVAIL;
1877 default:
1878 return -ENOIOCTLCMD;
1879 }
1880 }
1881 #endif
1882
1883 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1884 {
1885 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1886 IPSTATS_MIB_OUTFORWDATAGRAMS);
1887 return dst_output(skb);
1888 }
1889
1890 /*
1891 * Processing handlers for ip6mr_forward
1892 */
1893
1894 static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1895 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1896 {
1897 struct ipv6hdr *ipv6h;
1898 struct mif_device *vif = &mrt->vif6_table[vifi];
1899 struct net_device *dev;
1900 struct dst_entry *dst;
1901 struct flowi6 fl6;
1902
1903 if (vif->dev == NULL)
1904 goto out_free;
1905
1906 #ifdef CONFIG_IPV6_PIMSM_V2
1907 if (vif->flags & MIFF_REGISTER) {
1908 vif->pkt_out++;
1909 vif->bytes_out += skb->len;
1910 vif->dev->stats.tx_bytes += skb->len;
1911 vif->dev->stats.tx_packets++;
1912 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1913 goto out_free;
1914 }
1915 #endif
1916
1917 ipv6h = ipv6_hdr(skb);
1918
1919 fl6 = (struct flowi6) {
1920 .flowi6_oif = vif->link,
1921 .daddr = ipv6h->daddr,
1922 };
1923
1924 dst = ip6_route_output(net, NULL, &fl6);
1925 if (!dst)
1926 goto out_free;
1927
1928 skb_dst_drop(skb);
1929 skb_dst_set(skb, dst);
1930
1931 /*
1932 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1933 * not only before forwarding, but after forwarding on all output
1934 * interfaces. It is clear, if mrouter runs a multicasting
1935 * program, it should receive packets not depending to what interface
1936 * program is joined.
1937 * If we will not make it, the program will have to join on all
1938 * interfaces. On the other hand, multihoming host (or router, but
1939 * not mrouter) cannot join to more than one interface - it will
1940 * result in receiving multiple packets.
1941 */
1942 dev = vif->dev;
1943 skb->dev = dev;
1944 vif->pkt_out++;
1945 vif->bytes_out += skb->len;
1946
1947 /* We are about to write */
1948 /* XXX: extension headers? */
1949 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1950 goto out_free;
1951
1952 ipv6h = ipv6_hdr(skb);
1953 ipv6h->hop_limit--;
1954
1955 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1956
1957 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
1958 ip6mr_forward2_finish);
1959
1960 out_free:
1961 kfree_skb(skb);
1962 return 0;
1963 }
1964
1965 static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
1966 {
1967 int ct;
1968
1969 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
1970 if (mrt->vif6_table[ct].dev == dev)
1971 break;
1972 }
1973 return ct;
1974 }
1975
1976 static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
1977 struct sk_buff *skb, struct mfc6_cache *cache)
1978 {
1979 int psend = -1;
1980 int vif, ct;
1981
1982 vif = cache->mf6c_parent;
1983 cache->mfc_un.res.pkt++;
1984 cache->mfc_un.res.bytes += skb->len;
1985
1986 /*
1987 * Wrong interface: drop packet and (maybe) send PIM assert.
1988 */
1989 if (mrt->vif6_table[vif].dev != skb->dev) {
1990 int true_vifi;
1991
1992 cache->mfc_un.res.wrong_if++;
1993 true_vifi = ip6mr_find_vif(mrt, skb->dev);
1994
1995 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1996 /* pimsm uses asserts, when switching from RPT to SPT,
1997 so that we cannot check that packet arrived on an oif.
1998 It is bad, but otherwise we would need to move pretty
1999 large chunk of pimd to kernel. Ough... --ANK
2000 */
2001 (mrt->mroute_do_pim ||
2002 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2003 time_after(jiffies,
2004 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2005 cache->mfc_un.res.last_assert = jiffies;
2006 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2007 }
2008 goto dont_forward;
2009 }
2010
2011 mrt->vif6_table[vif].pkt_in++;
2012 mrt->vif6_table[vif].bytes_in += skb->len;
2013
2014 /*
2015 * Forward the frame
2016 */
2017 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2018 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2019 if (psend != -1) {
2020 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2021 if (skb2)
2022 ip6mr_forward2(net, mrt, skb2, cache, psend);
2023 }
2024 psend = ct;
2025 }
2026 }
2027 if (psend != -1) {
2028 ip6mr_forward2(net, mrt, skb, cache, psend);
2029 return 0;
2030 }
2031
2032 dont_forward:
2033 kfree_skb(skb);
2034 return 0;
2035 }
2036
2037
2038 /*
2039 * Multicast packets for forwarding arrive here
2040 */
2041
2042 int ip6_mr_input(struct sk_buff *skb)
2043 {
2044 struct mfc6_cache *cache;
2045 struct net *net = dev_net(skb->dev);
2046 struct mr6_table *mrt;
2047 struct flowi6 fl6 = {
2048 .flowi6_iif = skb->dev->ifindex,
2049 .flowi6_mark = skb->mark,
2050 };
2051 int err;
2052
2053 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2054 if (err < 0)
2055 return err;
2056
2057 read_lock(&mrt_lock);
2058 cache = ip6mr_cache_find(mrt,
2059 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2060
2061 /*
2062 * No usable cache entry
2063 */
2064 if (cache == NULL) {
2065 int vif;
2066
2067 vif = ip6mr_find_vif(mrt, skb->dev);
2068 if (vif >= 0) {
2069 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2070 read_unlock(&mrt_lock);
2071
2072 return err;
2073 }
2074 read_unlock(&mrt_lock);
2075 kfree_skb(skb);
2076 return -ENODEV;
2077 }
2078
2079 ip6_mr_forward(net, mrt, skb, cache);
2080
2081 read_unlock(&mrt_lock);
2082
2083 return 0;
2084 }
2085
2086
2087 static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2088 struct mfc6_cache *c, struct rtmsg *rtm)
2089 {
2090 int ct;
2091 struct rtnexthop *nhp;
2092 u8 *b = skb_tail_pointer(skb);
2093 struct rtattr *mp_head;
2094
2095 /* If cache is unresolved, don't try to parse IIF and OIF */
2096 if (c->mf6c_parent >= MAXMIFS)
2097 return -ENOENT;
2098
2099 if (MIF_EXISTS(mrt, c->mf6c_parent))
2100 RTA_PUT(skb, RTA_IIF, 4, &mrt->vif6_table[c->mf6c_parent].dev->ifindex);
2101
2102 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
2103
2104 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2105 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2106 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
2107 goto rtattr_failure;
2108 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
2109 nhp->rtnh_flags = 0;
2110 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2111 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2112 nhp->rtnh_len = sizeof(*nhp);
2113 }
2114 }
2115 mp_head->rta_type = RTA_MULTIPATH;
2116 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
2117 rtm->rtm_type = RTN_MULTICAST;
2118 return 1;
2119
2120 rtattr_failure:
2121 nlmsg_trim(skb, b);
2122 return -EMSGSIZE;
2123 }
2124
2125 int ip6mr_get_route(struct net *net,
2126 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2127 {
2128 int err;
2129 struct mr6_table *mrt;
2130 struct mfc6_cache *cache;
2131 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2132
2133 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2134 if (mrt == NULL)
2135 return -ENOENT;
2136
2137 read_lock(&mrt_lock);
2138 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2139
2140 if (!cache) {
2141 struct sk_buff *skb2;
2142 struct ipv6hdr *iph;
2143 struct net_device *dev;
2144 int vif;
2145
2146 if (nowait) {
2147 read_unlock(&mrt_lock);
2148 return -EAGAIN;
2149 }
2150
2151 dev = skb->dev;
2152 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2153 read_unlock(&mrt_lock);
2154 return -ENODEV;
2155 }
2156
2157 /* really correct? */
2158 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2159 if (!skb2) {
2160 read_unlock(&mrt_lock);
2161 return -ENOMEM;
2162 }
2163
2164 skb_reset_transport_header(skb2);
2165
2166 skb_put(skb2, sizeof(struct ipv6hdr));
2167 skb_reset_network_header(skb2);
2168
2169 iph = ipv6_hdr(skb2);
2170 iph->version = 0;
2171 iph->priority = 0;
2172 iph->flow_lbl[0] = 0;
2173 iph->flow_lbl[1] = 0;
2174 iph->flow_lbl[2] = 0;
2175 iph->payload_len = 0;
2176 iph->nexthdr = IPPROTO_NONE;
2177 iph->hop_limit = 0;
2178 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
2179 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
2180
2181 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2182 read_unlock(&mrt_lock);
2183
2184 return err;
2185 }
2186
2187 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2188 cache->mfc_flags |= MFC_NOTIFY;
2189
2190 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2191 read_unlock(&mrt_lock);
2192 return err;
2193 }
2194
2195 static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2196 u32 pid, u32 seq, struct mfc6_cache *c)
2197 {
2198 struct nlmsghdr *nlh;
2199 struct rtmsg *rtm;
2200
2201 nlh = nlmsg_put(skb, pid, seq, RTM_NEWROUTE, sizeof(*rtm), NLM_F_MULTI);
2202 if (nlh == NULL)
2203 return -EMSGSIZE;
2204
2205 rtm = nlmsg_data(nlh);
2206 rtm->rtm_family = RTNL_FAMILY_IPMR;
2207 rtm->rtm_dst_len = 128;
2208 rtm->rtm_src_len = 128;
2209 rtm->rtm_tos = 0;
2210 rtm->rtm_table = mrt->id;
2211 NLA_PUT_U32(skb, RTA_TABLE, mrt->id);
2212 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2213 rtm->rtm_protocol = RTPROT_UNSPEC;
2214 rtm->rtm_flags = 0;
2215
2216 NLA_PUT(skb, RTA_SRC, 16, &c->mf6c_origin);
2217 NLA_PUT(skb, RTA_DST, 16, &c->mf6c_mcastgrp);
2218
2219 if (__ip6mr_fill_mroute(mrt, skb, c, rtm) < 0)
2220 goto nla_put_failure;
2221
2222 return nlmsg_end(skb, nlh);
2223
2224 nla_put_failure:
2225 nlmsg_cancel(skb, nlh);
2226 return -EMSGSIZE;
2227 }
2228
2229 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2230 {
2231 struct net *net = sock_net(skb->sk);
2232 struct mr6_table *mrt;
2233 struct mfc6_cache *mfc;
2234 unsigned int t = 0, s_t;
2235 unsigned int h = 0, s_h;
2236 unsigned int e = 0, s_e;
2237
2238 s_t = cb->args[0];
2239 s_h = cb->args[1];
2240 s_e = cb->args[2];
2241
2242 read_lock(&mrt_lock);
2243 ip6mr_for_each_table(mrt, net) {
2244 if (t < s_t)
2245 goto next_table;
2246 if (t > s_t)
2247 s_h = 0;
2248 for (h = s_h; h < MFC6_LINES; h++) {
2249 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2250 if (e < s_e)
2251 goto next_entry;
2252 if (ip6mr_fill_mroute(mrt, skb,
2253 NETLINK_CB(cb->skb).pid,
2254 cb->nlh->nlmsg_seq,
2255 mfc) < 0)
2256 goto done;
2257 next_entry:
2258 e++;
2259 }
2260 e = s_e = 0;
2261 }
2262 s_h = 0;
2263 next_table:
2264 t++;
2265 }
2266 done:
2267 read_unlock(&mrt_lock);
2268
2269 cb->args[2] = e;
2270 cb->args[1] = h;
2271 cb->args[0] = t;
2272
2273 return skb->len;
2274 }
linux-next