?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
fs/9p: [fscache] wait for page write in cached mode
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / ip6mr.c
blob0e1d53bcf1e03626061aa75011dd7eccc6a89d6e
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 flowi *flp,
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, flp, 0, &arg);
146 if (err < 0)
147 return err;
148 *mrt = res.mrt;
149 return 0;
150 }
151
152 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
153 int flags, struct fib_lookup_arg *arg)
154 {
155 struct ip6mr_result *res = arg->result;
156 struct mr6_table *mrt;
157
158 switch (rule->action) {
159 case FR_ACT_TO_TBL:
160 break;
161 case FR_ACT_UNREACHABLE:
162 return -ENETUNREACH;
163 case FR_ACT_PROHIBIT:
164 return -EACCES;
165 case FR_ACT_BLACKHOLE:
166 default:
167 return -EINVAL;
168 }
169
170 mrt = ip6mr_get_table(rule->fr_net, rule->table);
171 if (mrt == NULL)
172 return -EAGAIN;
173 res->mrt = mrt;
174 return 0;
175 }
176
177 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
178 {
179 return 1;
180 }
181
182 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
183 FRA_GENERIC_POLICY,
184 };
185
186 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
187 struct fib_rule_hdr *frh, struct nlattr **tb)
188 {
189 return 0;
190 }
191
192 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
193 struct nlattr **tb)
194 {
195 return 1;
196 }
197
198 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
199 struct fib_rule_hdr *frh)
200 {
201 frh->dst_len = 0;
202 frh->src_len = 0;
203 frh->tos = 0;
204 return 0;
205 }
206
207 static const struct fib_rules_ops __net_initdata ip6mr_rules_ops_template = {
208 .family = RTNL_FAMILY_IP6MR,
209 .rule_size = sizeof(struct ip6mr_rule),
210 .addr_size = sizeof(struct in6_addr),
211 .action = ip6mr_rule_action,
212 .match = ip6mr_rule_match,
213 .configure = ip6mr_rule_configure,
214 .compare = ip6mr_rule_compare,
215 .default_pref = fib_default_rule_pref,
216 .fill = ip6mr_rule_fill,
217 .nlgroup = RTNLGRP_IPV6_RULE,
218 .policy = ip6mr_rule_policy,
219 .owner = THIS_MODULE,
220 };
221
222 static int __net_init ip6mr_rules_init(struct net *net)
223 {
224 struct fib_rules_ops *ops;
225 struct mr6_table *mrt;
226 int err;
227
228 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
229 if (IS_ERR(ops))
230 return PTR_ERR(ops);
231
232 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
233
234 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
235 if (mrt == NULL) {
236 err = -ENOMEM;
237 goto err1;
238 }
239
240 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
241 if (err < 0)
242 goto err2;
243
244 net->ipv6.mr6_rules_ops = ops;
245 return 0;
246
247 err2:
248 kfree(mrt);
249 err1:
250 fib_rules_unregister(ops);
251 return err;
252 }
253
254 static void __net_exit ip6mr_rules_exit(struct net *net)
255 {
256 struct mr6_table *mrt, *next;
257
258 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
259 list_del(&mrt->list);
260 ip6mr_free_table(mrt);
261 }
262 fib_rules_unregister(net->ipv6.mr6_rules_ops);
263 }
264 #else
265 #define ip6mr_for_each_table(mrt, net) \
266 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
267
268 static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
269 {
270 return net->ipv6.mrt6;
271 }
272
273 static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
274 struct mr6_table **mrt)
275 {
276 *mrt = net->ipv6.mrt6;
277 return 0;
278 }
279
280 static int __net_init ip6mr_rules_init(struct net *net)
281 {
282 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
283 return net->ipv6.mrt6 ? 0 : -ENOMEM;
284 }
285
286 static void __net_exit ip6mr_rules_exit(struct net *net)
287 {
288 ip6mr_free_table(net->ipv6.mrt6);
289 }
290 #endif
291
292 static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
293 {
294 struct mr6_table *mrt;
295 unsigned int i;
296
297 mrt = ip6mr_get_table(net, id);
298 if (mrt != NULL)
299 return mrt;
300
301 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
302 if (mrt == NULL)
303 return NULL;
304 mrt->id = id;
305 write_pnet(&mrt->net, net);
306
307 /* Forwarding cache */
308 for (i = 0; i < MFC6_LINES; i++)
309 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
310
311 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
312
313 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
314 (unsigned long)mrt);
315
316 #ifdef CONFIG_IPV6_PIMSM_V2
317 mrt->mroute_reg_vif_num = -1;
318 #endif
319 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
320 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
321 #endif
322 return mrt;
323 }
324
325 static void ip6mr_free_table(struct mr6_table *mrt)
326 {
327 del_timer(&mrt->ipmr_expire_timer);
328 mroute_clean_tables(mrt);
329 kfree(mrt);
330 }
331
332 #ifdef CONFIG_PROC_FS
333
334 struct ipmr_mfc_iter {
335 struct seq_net_private p;
336 struct mr6_table *mrt;
337 struct list_head *cache;
338 int ct;
339 };
340
341
342 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
343 struct ipmr_mfc_iter *it, loff_t pos)
344 {
345 struct mr6_table *mrt = it->mrt;
346 struct mfc6_cache *mfc;
347
348 read_lock(&mrt_lock);
349 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
350 it->cache = &mrt->mfc6_cache_array[it->ct];
351 list_for_each_entry(mfc, it->cache, list)
352 if (pos-- == 0)
353 return mfc;
354 }
355 read_unlock(&mrt_lock);
356
357 spin_lock_bh(&mfc_unres_lock);
358 it->cache = &mrt->mfc6_unres_queue;
359 list_for_each_entry(mfc, it->cache, list)
360 if (pos-- == 0)
361 return mfc;
362 spin_unlock_bh(&mfc_unres_lock);
363
364 it->cache = NULL;
365 return NULL;
366 }
367
368 /*
369 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
370 */
371
372 struct ipmr_vif_iter {
373 struct seq_net_private p;
374 struct mr6_table *mrt;
375 int ct;
376 };
377
378 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
379 struct ipmr_vif_iter *iter,
380 loff_t pos)
381 {
382 struct mr6_table *mrt = iter->mrt;
383
384 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
385 if (!MIF_EXISTS(mrt, iter->ct))
386 continue;
387 if (pos-- == 0)
388 return &mrt->vif6_table[iter->ct];
389 }
390 return NULL;
391 }
392
393 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
394 __acquires(mrt_lock)
395 {
396 struct ipmr_vif_iter *iter = seq->private;
397 struct net *net = seq_file_net(seq);
398 struct mr6_table *mrt;
399
400 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
401 if (mrt == NULL)
402 return ERR_PTR(-ENOENT);
403
404 iter->mrt = mrt;
405
406 read_lock(&mrt_lock);
407 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
408 : SEQ_START_TOKEN;
409 }
410
411 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
412 {
413 struct ipmr_vif_iter *iter = seq->private;
414 struct net *net = seq_file_net(seq);
415 struct mr6_table *mrt = iter->mrt;
416
417 ++*pos;
418 if (v == SEQ_START_TOKEN)
419 return ip6mr_vif_seq_idx(net, iter, 0);
420
421 while (++iter->ct < mrt->maxvif) {
422 if (!MIF_EXISTS(mrt, iter->ct))
423 continue;
424 return &mrt->vif6_table[iter->ct];
425 }
426 return NULL;
427 }
428
429 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
430 __releases(mrt_lock)
431 {
432 read_unlock(&mrt_lock);
433 }
434
435 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
436 {
437 struct ipmr_vif_iter *iter = seq->private;
438 struct mr6_table *mrt = iter->mrt;
439
440 if (v == SEQ_START_TOKEN) {
441 seq_puts(seq,
442 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
443 } else {
444 const struct mif_device *vif = v;
445 const char *name = vif->dev ? vif->dev->name : "none";
446
447 seq_printf(seq,
448 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
449 vif - mrt->vif6_table,
450 name, vif->bytes_in, vif->pkt_in,
451 vif->bytes_out, vif->pkt_out,
452 vif->flags);
453 }
454 return 0;
455 }
456
457 static const struct seq_operations ip6mr_vif_seq_ops = {
458 .start = ip6mr_vif_seq_start,
459 .next = ip6mr_vif_seq_next,
460 .stop = ip6mr_vif_seq_stop,
461 .show = ip6mr_vif_seq_show,
462 };
463
464 static int ip6mr_vif_open(struct inode *inode, struct file *file)
465 {
466 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
467 sizeof(struct ipmr_vif_iter));
468 }
469
470 static const struct file_operations ip6mr_vif_fops = {
471 .owner = THIS_MODULE,
472 .open = ip6mr_vif_open,
473 .read = seq_read,
474 .llseek = seq_lseek,
475 .release = seq_release_net,
476 };
477
478 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
479 {
480 struct ipmr_mfc_iter *it = seq->private;
481 struct net *net = seq_file_net(seq);
482 struct mr6_table *mrt;
483
484 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
485 if (mrt == NULL)
486 return ERR_PTR(-ENOENT);
487
488 it->mrt = mrt;
489 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
490 : SEQ_START_TOKEN;
491 }
492
493 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
494 {
495 struct mfc6_cache *mfc = v;
496 struct ipmr_mfc_iter *it = seq->private;
497 struct net *net = seq_file_net(seq);
498 struct mr6_table *mrt = it->mrt;
499
500 ++*pos;
501
502 if (v == SEQ_START_TOKEN)
503 return ipmr_mfc_seq_idx(net, seq->private, 0);
504
505 if (mfc->list.next != it->cache)
506 return list_entry(mfc->list.next, struct mfc6_cache, list);
507
508 if (it->cache == &mrt->mfc6_unres_queue)
509 goto end_of_list;
510
511 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
512
513 while (++it->ct < MFC6_LINES) {
514 it->cache = &mrt->mfc6_cache_array[it->ct];
515 if (list_empty(it->cache))
516 continue;
517 return list_first_entry(it->cache, struct mfc6_cache, list);
518 }
519
520 /* exhausted cache_array, show unresolved */
521 read_unlock(&mrt_lock);
522 it->cache = &mrt->mfc6_unres_queue;
523 it->ct = 0;
524
525 spin_lock_bh(&mfc_unres_lock);
526 if (!list_empty(it->cache))
527 return list_first_entry(it->cache, struct mfc6_cache, list);
528
529 end_of_list:
530 spin_unlock_bh(&mfc_unres_lock);
531 it->cache = NULL;
532
533 return NULL;
534 }
535
536 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
537 {
538 struct ipmr_mfc_iter *it = seq->private;
539 struct mr6_table *mrt = it->mrt;
540
541 if (it->cache == &mrt->mfc6_unres_queue)
542 spin_unlock_bh(&mfc_unres_lock);
543 else if (it->cache == mrt->mfc6_cache_array)
544 read_unlock(&mrt_lock);
545 }
546
547 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
548 {
549 int n;
550
551 if (v == SEQ_START_TOKEN) {
552 seq_puts(seq,
553 "Group "
554 "Origin "
555 "Iif Pkts Bytes Wrong Oifs\n");
556 } else {
557 const struct mfc6_cache *mfc = v;
558 const struct ipmr_mfc_iter *it = seq->private;
559 struct mr6_table *mrt = it->mrt;
560
561 seq_printf(seq, "%pI6 %pI6 %-3hd",
562 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
563 mfc->mf6c_parent);
564
565 if (it->cache != &mrt->mfc6_unres_queue) {
566 seq_printf(seq, " %8lu %8lu %8lu",
567 mfc->mfc_un.res.pkt,
568 mfc->mfc_un.res.bytes,
569 mfc->mfc_un.res.wrong_if);
570 for (n = mfc->mfc_un.res.minvif;
571 n < mfc->mfc_un.res.maxvif; n++) {
572 if (MIF_EXISTS(mrt, n) &&
573 mfc->mfc_un.res.ttls[n] < 255)
574 seq_printf(seq,
575 " %2d:%-3d",
576 n, mfc->mfc_un.res.ttls[n]);
577 }
578 } else {
579 /* unresolved mfc_caches don't contain
580 * pkt, bytes and wrong_if values
581 */
582 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
583 }
584 seq_putc(seq, '\n');
585 }
586 return 0;
587 }
588
589 static const struct seq_operations ipmr_mfc_seq_ops = {
590 .start = ipmr_mfc_seq_start,
591 .next = ipmr_mfc_seq_next,
592 .stop = ipmr_mfc_seq_stop,
593 .show = ipmr_mfc_seq_show,
594 };
595
596 static int ipmr_mfc_open(struct inode *inode, struct file *file)
597 {
598 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
599 sizeof(struct ipmr_mfc_iter));
600 }
601
602 static const struct file_operations ip6mr_mfc_fops = {
603 .owner = THIS_MODULE,
604 .open = ipmr_mfc_open,
605 .read = seq_read,
606 .llseek = seq_lseek,
607 .release = seq_release_net,
608 };
609 #endif
610
611 #ifdef CONFIG_IPV6_PIMSM_V2
612
613 static int pim6_rcv(struct sk_buff *skb)
614 {
615 struct pimreghdr *pim;
616 struct ipv6hdr *encap;
617 struct net_device *reg_dev = NULL;
618 struct net *net = dev_net(skb->dev);
619 struct mr6_table *mrt;
620 struct flowi fl = {
621 .iif = skb->dev->ifindex,
622 .mark = skb->mark,
623 };
624 int reg_vif_num;
625
626 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
627 goto drop;
628
629 pim = (struct pimreghdr *)skb_transport_header(skb);
630 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
631 (pim->flags & PIM_NULL_REGISTER) ||
632 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
633 sizeof(*pim), IPPROTO_PIM,
634 csum_partial((void *)pim, sizeof(*pim), 0)) &&
635 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
636 goto drop;
637
638 /* check if the inner packet is destined to mcast group */
639 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
640 sizeof(*pim));
641
642 if (!ipv6_addr_is_multicast(&encap->daddr) ||
643 encap->payload_len == 0 ||
644 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
645 goto drop;
646
647 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
648 goto drop;
649 reg_vif_num = mrt->mroute_reg_vif_num;
650
651 read_lock(&mrt_lock);
652 if (reg_vif_num >= 0)
653 reg_dev = mrt->vif6_table[reg_vif_num].dev;
654 if (reg_dev)
655 dev_hold(reg_dev);
656 read_unlock(&mrt_lock);
657
658 if (reg_dev == NULL)
659 goto drop;
660
661 skb->mac_header = skb->network_header;
662 skb_pull(skb, (u8 *)encap - skb->data);
663 skb_reset_network_header(skb);
664 skb->protocol = htons(ETH_P_IPV6);
665 skb->ip_summed = 0;
666 skb->pkt_type = PACKET_HOST;
667
668 skb_tunnel_rx(skb, reg_dev);
669
670 netif_rx(skb);
671
672 dev_put(reg_dev);
673 return 0;
674 drop:
675 kfree_skb(skb);
676 return 0;
677 }
678
679 static const struct inet6_protocol pim6_protocol = {
680 .handler = pim6_rcv,
681 };
682
683 /* Service routines creating virtual interfaces: PIMREG */
684
685 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
686 struct net_device *dev)
687 {
688 struct net *net = dev_net(dev);
689 struct mr6_table *mrt;
690 struct flowi fl = {
691 .oif = dev->ifindex,
692 .iif = skb->skb_iif,
693 .mark = skb->mark,
694 };
695 int err;
696
697 err = ip6mr_fib_lookup(net, &fl, &mrt);
698 if (err < 0)
699 return err;
700
701 read_lock(&mrt_lock);
702 dev->stats.tx_bytes += skb->len;
703 dev->stats.tx_packets++;
704 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
705 read_unlock(&mrt_lock);
706 kfree_skb(skb);
707 return NETDEV_TX_OK;
708 }
709
710 static const struct net_device_ops reg_vif_netdev_ops = {
711 .ndo_start_xmit = reg_vif_xmit,
712 };
713
714 static void reg_vif_setup(struct net_device *dev)
715 {
716 dev->type = ARPHRD_PIMREG;
717 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
718 dev->flags = IFF_NOARP;
719 dev->netdev_ops = &reg_vif_netdev_ops;
720 dev->destructor = free_netdev;
721 dev->features |= NETIF_F_NETNS_LOCAL;
722 }
723
724 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
725 {
726 struct net_device *dev;
727 char name[IFNAMSIZ];
728
729 if (mrt->id == RT6_TABLE_DFLT)
730 sprintf(name, "pim6reg");
731 else
732 sprintf(name, "pim6reg%u", mrt->id);
733
734 dev = alloc_netdev(0, name, reg_vif_setup);
735 if (dev == NULL)
736 return NULL;
737
738 dev_net_set(dev, net);
739
740 if (register_netdevice(dev)) {
741 free_netdev(dev);
742 return NULL;
743 }
744 dev->iflink = 0;
745
746 if (dev_open(dev))
747 goto failure;
748
749 dev_hold(dev);
750 return dev;
751
752 failure:
753 /* allow the register to be completed before unregistering. */
754 rtnl_unlock();
755 rtnl_lock();
756
757 unregister_netdevice(dev);
758 return NULL;
759 }
760 #endif
761
762 /*
763 * Delete a VIF entry
764 */
765
766 static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
767 {
768 struct mif_device *v;
769 struct net_device *dev;
770 struct inet6_dev *in6_dev;
771
772 if (vifi < 0 || vifi >= mrt->maxvif)
773 return -EADDRNOTAVAIL;
774
775 v = &mrt->vif6_table[vifi];
776
777 write_lock_bh(&mrt_lock);
778 dev = v->dev;
779 v->dev = NULL;
780
781 if (!dev) {
782 write_unlock_bh(&mrt_lock);
783 return -EADDRNOTAVAIL;
784 }
785
786 #ifdef CONFIG_IPV6_PIMSM_V2
787 if (vifi == mrt->mroute_reg_vif_num)
788 mrt->mroute_reg_vif_num = -1;
789 #endif
790
791 if (vifi + 1 == mrt->maxvif) {
792 int tmp;
793 for (tmp = vifi - 1; tmp >= 0; tmp--) {
794 if (MIF_EXISTS(mrt, tmp))
795 break;
796 }
797 mrt->maxvif = tmp + 1;
798 }
799
800 write_unlock_bh(&mrt_lock);
801
802 dev_set_allmulti(dev, -1);
803
804 in6_dev = __in6_dev_get(dev);
805 if (in6_dev)
806 in6_dev->cnf.mc_forwarding--;
807
808 if (v->flags & MIFF_REGISTER)
809 unregister_netdevice_queue(dev, head);
810
811 dev_put(dev);
812 return 0;
813 }
814
815 static inline void ip6mr_cache_free(struct mfc6_cache *c)
816 {
817 kmem_cache_free(mrt_cachep, c);
818 }
819
820 /* Destroy an unresolved cache entry, killing queued skbs
821 and reporting error to netlink readers.
822 */
823
824 static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
825 {
826 struct net *net = read_pnet(&mrt->net);
827 struct sk_buff *skb;
828
829 atomic_dec(&mrt->cache_resolve_queue_len);
830
831 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
832 if (ipv6_hdr(skb)->version == 0) {
833 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
834 nlh->nlmsg_type = NLMSG_ERROR;
835 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
836 skb_trim(skb, nlh->nlmsg_len);
837 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
838 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
839 } else
840 kfree_skb(skb);
841 }
842
843 ip6mr_cache_free(c);
844 }
845
846
847 /* Timer process for all the unresolved queue. */
848
849 static void ipmr_do_expire_process(struct mr6_table *mrt)
850 {
851 unsigned long now = jiffies;
852 unsigned long expires = 10 * HZ;
853 struct mfc6_cache *c, *next;
854
855 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
856 if (time_after(c->mfc_un.unres.expires, now)) {
857 /* not yet... */
858 unsigned long interval = c->mfc_un.unres.expires - now;
859 if (interval < expires)
860 expires = interval;
861 continue;
862 }
863
864 list_del(&c->list);
865 ip6mr_destroy_unres(mrt, c);
866 }
867
868 if (!list_empty(&mrt->mfc6_unres_queue))
869 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
870 }
871
872 static void ipmr_expire_process(unsigned long arg)
873 {
874 struct mr6_table *mrt = (struct mr6_table *)arg;
875
876 if (!spin_trylock(&mfc_unres_lock)) {
877 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
878 return;
879 }
880
881 if (!list_empty(&mrt->mfc6_unres_queue))
882 ipmr_do_expire_process(mrt);
883
884 spin_unlock(&mfc_unres_lock);
885 }
886
887 /* Fill oifs list. It is called under write locked mrt_lock. */
888
889 static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
890 unsigned char *ttls)
891 {
892 int vifi;
893
894 cache->mfc_un.res.minvif = MAXMIFS;
895 cache->mfc_un.res.maxvif = 0;
896 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
897
898 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
899 if (MIF_EXISTS(mrt, vifi) &&
900 ttls[vifi] && ttls[vifi] < 255) {
901 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
902 if (cache->mfc_un.res.minvif > vifi)
903 cache->mfc_un.res.minvif = vifi;
904 if (cache->mfc_un.res.maxvif <= vifi)
905 cache->mfc_un.res.maxvif = vifi + 1;
906 }
907 }
908 }
909
910 static int mif6_add(struct net *net, struct mr6_table *mrt,
911 struct mif6ctl *vifc, int mrtsock)
912 {
913 int vifi = vifc->mif6c_mifi;
914 struct mif_device *v = &mrt->vif6_table[vifi];
915 struct net_device *dev;
916 struct inet6_dev *in6_dev;
917 int err;
918
919 /* Is vif busy ? */
920 if (MIF_EXISTS(mrt, vifi))
921 return -EADDRINUSE;
922
923 switch (vifc->mif6c_flags) {
924 #ifdef CONFIG_IPV6_PIMSM_V2
925 case MIFF_REGISTER:
926 /*
927 * Special Purpose VIF in PIM
928 * All the packets will be sent to the daemon
929 */
930 if (mrt->mroute_reg_vif_num >= 0)
931 return -EADDRINUSE;
932 dev = ip6mr_reg_vif(net, mrt);
933 if (!dev)
934 return -ENOBUFS;
935 err = dev_set_allmulti(dev, 1);
936 if (err) {
937 unregister_netdevice(dev);
938 dev_put(dev);
939 return err;
940 }
941 break;
942 #endif
943 case 0:
944 dev = dev_get_by_index(net, vifc->mif6c_pifi);
945 if (!dev)
946 return -EADDRNOTAVAIL;
947 err = dev_set_allmulti(dev, 1);
948 if (err) {
949 dev_put(dev);
950 return err;
951 }
952 break;
953 default:
954 return -EINVAL;
955 }
956
957 in6_dev = __in6_dev_get(dev);
958 if (in6_dev)
959 in6_dev->cnf.mc_forwarding++;
960
961 /*
962 * Fill in the VIF structures
963 */
964 v->rate_limit = vifc->vifc_rate_limit;
965 v->flags = vifc->mif6c_flags;
966 if (!mrtsock)
967 v->flags |= VIFF_STATIC;
968 v->threshold = vifc->vifc_threshold;
969 v->bytes_in = 0;
970 v->bytes_out = 0;
971 v->pkt_in = 0;
972 v->pkt_out = 0;
973 v->link = dev->ifindex;
974 if (v->flags & MIFF_REGISTER)
975 v->link = dev->iflink;
976
977 /* And finish update writing critical data */
978 write_lock_bh(&mrt_lock);
979 v->dev = dev;
980 #ifdef CONFIG_IPV6_PIMSM_V2
981 if (v->flags & MIFF_REGISTER)
982 mrt->mroute_reg_vif_num = vifi;
983 #endif
984 if (vifi + 1 > mrt->maxvif)
985 mrt->maxvif = vifi + 1;
986 write_unlock_bh(&mrt_lock);
987 return 0;
988 }
989
990 static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
991 struct in6_addr *origin,
992 struct in6_addr *mcastgrp)
993 {
994 int line = MFC6_HASH(mcastgrp, origin);
995 struct mfc6_cache *c;
996
997 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
998 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
999 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1000 return c;
1001 }
1002 return NULL;
1003 }
1004
1005 /*
1006 * Allocate a multicast cache entry
1007 */
1008 static struct mfc6_cache *ip6mr_cache_alloc(void)
1009 {
1010 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1011 if (c == NULL)
1012 return NULL;
1013 c->mfc_un.res.minvif = MAXMIFS;
1014 return c;
1015 }
1016
1017 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1018 {
1019 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1020 if (c == NULL)
1021 return NULL;
1022 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1023 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1024 return c;
1025 }
1026
1027 /*
1028 * A cache entry has gone into a resolved state from queued
1029 */
1030
1031 static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1032 struct mfc6_cache *uc, struct mfc6_cache *c)
1033 {
1034 struct sk_buff *skb;
1035
1036 /*
1037 * Play the pending entries through our router
1038 */
1039
1040 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1041 if (ipv6_hdr(skb)->version == 0) {
1042 int err;
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 err = 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 flowi fl = {
1552 .iif = skb->skb_iif,
1553 .oif = skb->dev->ifindex,
1554 .mark = skb->mark,
1555 };
1556
1557 if (ip6mr_fib_lookup(net, &fl, &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 flowi fl;
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 fl = (struct flowi) {
1920 .oif = vif->link,
1921 .fl6_dst = ipv6h->daddr,
1922 };
1923
1924 dst = ip6_route_output(net, NULL, &fl);
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 flowi fl = {
2048 .iif = skb->dev->ifindex,
2049 .mark = skb->mark,
2050 };
2051 int err;
2052
2053 err = ip6mr_fib_lookup(net, &fl, &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 2.6 ibm-acpi/thinkpad-acpi driver git tree