Loading...
1/*
2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
4 *
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
8 * 6WIND, Paris, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 *
17 */
18
19#include <linux/uaccess.h>
20#include <linux/types.h>
21#include <linux/sched.h>
22#include <linux/errno.h>
23#include <linux/mm.h>
24#include <linux/kernel.h>
25#include <linux/fcntl.h>
26#include <linux/stat.h>
27#include <linux/socket.h>
28#include <linux/inet.h>
29#include <linux/netdevice.h>
30#include <linux/inetdevice.h>
31#include <linux/proc_fs.h>
32#include <linux/seq_file.h>
33#include <linux/init.h>
34#include <linux/compat.h>
35#include <net/protocol.h>
36#include <linux/skbuff.h>
37#include <net/raw.h>
38#include <linux/notifier.h>
39#include <linux/if_arp.h>
40#include <net/checksum.h>
41#include <net/netlink.h>
42#include <net/fib_rules.h>
43
44#include <net/ipv6.h>
45#include <net/ip6_route.h>
46#include <linux/mroute6.h>
47#include <linux/pim.h>
48#include <net/addrconf.h>
49#include <linux/netfilter_ipv6.h>
50#include <linux/export.h>
51#include <net/ip6_checksum.h>
52#include <linux/netconf.h>
53
54struct ip6mr_rule {
55 struct fib_rule common;
56};
57
58struct ip6mr_result {
59 struct mr_table *mrt;
60};
61
62/* Big lock, protecting vif table, mrt cache and mroute socket state.
63 Note that the changes are semaphored via rtnl_lock.
64 */
65
66static DEFINE_RWLOCK(mrt_lock);
67
68/* Multicast router control variables */
69
70/* Special spinlock for queue of unresolved entries */
71static DEFINE_SPINLOCK(mfc_unres_lock);
72
73/* We return to original Alan's scheme. Hash table of resolved
74 entries is changed only in process context and protected
75 with weak lock mrt_lock. Queue of unresolved entries is protected
76 with strong spinlock mfc_unres_lock.
77
78 In this case data path is free of exclusive locks at all.
79 */
80
81static struct kmem_cache *mrt_cachep __read_mostly;
82
83static struct mr_table *ip6mr_new_table(struct net *net, u32 id);
84static void ip6mr_free_table(struct mr_table *mrt);
85
86static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
87 struct sk_buff *skb, struct mfc6_cache *cache);
88static int ip6mr_cache_report(struct mr_table *mrt, struct sk_buff *pkt,
89 mifi_t mifi, int assert);
90static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc,
91 int cmd);
92static void mrt6msg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
93static int ip6mr_rtm_dumproute(struct sk_buff *skb,
94 struct netlink_callback *cb);
95static void mroute_clean_tables(struct mr_table *mrt, bool all);
96static void ipmr_expire_process(struct timer_list *t);
97
98#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
99#define ip6mr_for_each_table(mrt, net) \
100 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
101
102static struct mr_table *ip6mr_mr_table_iter(struct net *net,
103 struct mr_table *mrt)
104{
105 struct mr_table *ret;
106
107 if (!mrt)
108 ret = list_entry_rcu(net->ipv6.mr6_tables.next,
109 struct mr_table, list);
110 else
111 ret = list_entry_rcu(mrt->list.next,
112 struct mr_table, list);
113
114 if (&ret->list == &net->ipv6.mr6_tables)
115 return NULL;
116 return ret;
117}
118
119static struct mr_table *ip6mr_get_table(struct net *net, u32 id)
120{
121 struct mr_table *mrt;
122
123 ip6mr_for_each_table(mrt, net) {
124 if (mrt->id == id)
125 return mrt;
126 }
127 return NULL;
128}
129
130static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
131 struct mr_table **mrt)
132{
133 int err;
134 struct ip6mr_result res;
135 struct fib_lookup_arg arg = {
136 .result = &res,
137 .flags = FIB_LOOKUP_NOREF,
138 };
139
140 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
141 flowi6_to_flowi(flp6), 0, &arg);
142 if (err < 0)
143 return err;
144 *mrt = res.mrt;
145 return 0;
146}
147
148static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
149 int flags, struct fib_lookup_arg *arg)
150{
151 struct ip6mr_result *res = arg->result;
152 struct mr_table *mrt;
153
154 switch (rule->action) {
155 case FR_ACT_TO_TBL:
156 break;
157 case FR_ACT_UNREACHABLE:
158 return -ENETUNREACH;
159 case FR_ACT_PROHIBIT:
160 return -EACCES;
161 case FR_ACT_BLACKHOLE:
162 default:
163 return -EINVAL;
164 }
165
166 mrt = ip6mr_get_table(rule->fr_net, rule->table);
167 if (!mrt)
168 return -EAGAIN;
169 res->mrt = mrt;
170 return 0;
171}
172
173static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
174{
175 return 1;
176}
177
178static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
179 FRA_GENERIC_POLICY,
180};
181
182static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
183 struct fib_rule_hdr *frh, struct nlattr **tb)
184{
185 return 0;
186}
187
188static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
189 struct nlattr **tb)
190{
191 return 1;
192}
193
194static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
195 struct fib_rule_hdr *frh)
196{
197 frh->dst_len = 0;
198 frh->src_len = 0;
199 frh->tos = 0;
200 return 0;
201}
202
203static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
204 .family = RTNL_FAMILY_IP6MR,
205 .rule_size = sizeof(struct ip6mr_rule),
206 .addr_size = sizeof(struct in6_addr),
207 .action = ip6mr_rule_action,
208 .match = ip6mr_rule_match,
209 .configure = ip6mr_rule_configure,
210 .compare = ip6mr_rule_compare,
211 .fill = ip6mr_rule_fill,
212 .nlgroup = RTNLGRP_IPV6_RULE,
213 .policy = ip6mr_rule_policy,
214 .owner = THIS_MODULE,
215};
216
217static int __net_init ip6mr_rules_init(struct net *net)
218{
219 struct fib_rules_ops *ops;
220 struct mr_table *mrt;
221 int err;
222
223 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
224 if (IS_ERR(ops))
225 return PTR_ERR(ops);
226
227 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
228
229 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
230 if (!mrt) {
231 err = -ENOMEM;
232 goto err1;
233 }
234
235 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
236 if (err < 0)
237 goto err2;
238
239 net->ipv6.mr6_rules_ops = ops;
240 return 0;
241
242err2:
243 ip6mr_free_table(mrt);
244err1:
245 fib_rules_unregister(ops);
246 return err;
247}
248
249static void __net_exit ip6mr_rules_exit(struct net *net)
250{
251 struct mr_table *mrt, *next;
252
253 rtnl_lock();
254 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
255 list_del(&mrt->list);
256 ip6mr_free_table(mrt);
257 }
258 fib_rules_unregister(net->ipv6.mr6_rules_ops);
259 rtnl_unlock();
260}
261
262static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb)
263{
264 return fib_rules_dump(net, nb, RTNL_FAMILY_IP6MR);
265}
266
267static unsigned int ip6mr_rules_seq_read(struct net *net)
268{
269 return fib_rules_seq_read(net, RTNL_FAMILY_IP6MR);
270}
271
272bool ip6mr_rule_default(const struct fib_rule *rule)
273{
274 return fib_rule_matchall(rule) && rule->action == FR_ACT_TO_TBL &&
275 rule->table == RT6_TABLE_DFLT && !rule->l3mdev;
276}
277EXPORT_SYMBOL(ip6mr_rule_default);
278#else
279#define ip6mr_for_each_table(mrt, net) \
280 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
281
282static struct mr_table *ip6mr_mr_table_iter(struct net *net,
283 struct mr_table *mrt)
284{
285 if (!mrt)
286 return net->ipv6.mrt6;
287 return NULL;
288}
289
290static struct mr_table *ip6mr_get_table(struct net *net, u32 id)
291{
292 return net->ipv6.mrt6;
293}
294
295static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
296 struct mr_table **mrt)
297{
298 *mrt = net->ipv6.mrt6;
299 return 0;
300}
301
302static int __net_init ip6mr_rules_init(struct net *net)
303{
304 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
305 return net->ipv6.mrt6 ? 0 : -ENOMEM;
306}
307
308static void __net_exit ip6mr_rules_exit(struct net *net)
309{
310 rtnl_lock();
311 ip6mr_free_table(net->ipv6.mrt6);
312 net->ipv6.mrt6 = NULL;
313 rtnl_unlock();
314}
315
316static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb)
317{
318 return 0;
319}
320
321static unsigned int ip6mr_rules_seq_read(struct net *net)
322{
323 return 0;
324}
325#endif
326
327static int ip6mr_hash_cmp(struct rhashtable_compare_arg *arg,
328 const void *ptr)
329{
330 const struct mfc6_cache_cmp_arg *cmparg = arg->key;
331 struct mfc6_cache *c = (struct mfc6_cache *)ptr;
332
333 return !ipv6_addr_equal(&c->mf6c_mcastgrp, &cmparg->mf6c_mcastgrp) ||
334 !ipv6_addr_equal(&c->mf6c_origin, &cmparg->mf6c_origin);
335}
336
337static const struct rhashtable_params ip6mr_rht_params = {
338 .head_offset = offsetof(struct mr_mfc, mnode),
339 .key_offset = offsetof(struct mfc6_cache, cmparg),
340 .key_len = sizeof(struct mfc6_cache_cmp_arg),
341 .nelem_hint = 3,
342 .locks_mul = 1,
343 .obj_cmpfn = ip6mr_hash_cmp,
344 .automatic_shrinking = true,
345};
346
347static void ip6mr_new_table_set(struct mr_table *mrt,
348 struct net *net)
349{
350#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
351 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
352#endif
353}
354
355static struct mfc6_cache_cmp_arg ip6mr_mr_table_ops_cmparg_any = {
356 .mf6c_origin = IN6ADDR_ANY_INIT,
357 .mf6c_mcastgrp = IN6ADDR_ANY_INIT,
358};
359
360static struct mr_table_ops ip6mr_mr_table_ops = {
361 .rht_params = &ip6mr_rht_params,
362 .cmparg_any = &ip6mr_mr_table_ops_cmparg_any,
363};
364
365static struct mr_table *ip6mr_new_table(struct net *net, u32 id)
366{
367 struct mr_table *mrt;
368
369 mrt = ip6mr_get_table(net, id);
370 if (mrt)
371 return mrt;
372
373 return mr_table_alloc(net, id, &ip6mr_mr_table_ops,
374 ipmr_expire_process, ip6mr_new_table_set);
375}
376
377static void ip6mr_free_table(struct mr_table *mrt)
378{
379 del_timer_sync(&mrt->ipmr_expire_timer);
380 mroute_clean_tables(mrt, true);
381 rhltable_destroy(&mrt->mfc_hash);
382 kfree(mrt);
383}
384
385#ifdef CONFIG_PROC_FS
386/* The /proc interfaces to multicast routing
387 * /proc/ip6_mr_cache /proc/ip6_mr_vif
388 */
389
390static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
391 __acquires(mrt_lock)
392{
393 struct mr_vif_iter *iter = seq->private;
394 struct net *net = seq_file_net(seq);
395 struct mr_table *mrt;
396
397 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
398 if (!mrt)
399 return ERR_PTR(-ENOENT);
400
401 iter->mrt = mrt;
402
403 read_lock(&mrt_lock);
404 return mr_vif_seq_start(seq, pos);
405}
406
407static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
408 __releases(mrt_lock)
409{
410 read_unlock(&mrt_lock);
411}
412
413static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
414{
415 struct mr_vif_iter *iter = seq->private;
416 struct mr_table *mrt = iter->mrt;
417
418 if (v == SEQ_START_TOKEN) {
419 seq_puts(seq,
420 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
421 } else {
422 const struct vif_device *vif = v;
423 const char *name = vif->dev ? vif->dev->name : "none";
424
425 seq_printf(seq,
426 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
427 vif - mrt->vif_table,
428 name, vif->bytes_in, vif->pkt_in,
429 vif->bytes_out, vif->pkt_out,
430 vif->flags);
431 }
432 return 0;
433}
434
435static const struct seq_operations ip6mr_vif_seq_ops = {
436 .start = ip6mr_vif_seq_start,
437 .next = mr_vif_seq_next,
438 .stop = ip6mr_vif_seq_stop,
439 .show = ip6mr_vif_seq_show,
440};
441
442static int ip6mr_vif_open(struct inode *inode, struct file *file)
443{
444 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
445 sizeof(struct mr_vif_iter));
446}
447
448static const struct file_operations ip6mr_vif_fops = {
449 .open = ip6mr_vif_open,
450 .read = seq_read,
451 .llseek = seq_lseek,
452 .release = seq_release_net,
453};
454
455static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
456{
457 struct net *net = seq_file_net(seq);
458 struct mr_table *mrt;
459
460 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
461 if (!mrt)
462 return ERR_PTR(-ENOENT);
463
464 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
465}
466
467static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
468{
469 int n;
470
471 if (v == SEQ_START_TOKEN) {
472 seq_puts(seq,
473 "Group "
474 "Origin "
475 "Iif Pkts Bytes Wrong Oifs\n");
476 } else {
477 const struct mfc6_cache *mfc = v;
478 const struct mr_mfc_iter *it = seq->private;
479 struct mr_table *mrt = it->mrt;
480
481 seq_printf(seq, "%pI6 %pI6 %-3hd",
482 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
483 mfc->_c.mfc_parent);
484
485 if (it->cache != &mrt->mfc_unres_queue) {
486 seq_printf(seq, " %8lu %8lu %8lu",
487 mfc->_c.mfc_un.res.pkt,
488 mfc->_c.mfc_un.res.bytes,
489 mfc->_c.mfc_un.res.wrong_if);
490 for (n = mfc->_c.mfc_un.res.minvif;
491 n < mfc->_c.mfc_un.res.maxvif; n++) {
492 if (VIF_EXISTS(mrt, n) &&
493 mfc->_c.mfc_un.res.ttls[n] < 255)
494 seq_printf(seq,
495 " %2d:%-3d", n,
496 mfc->_c.mfc_un.res.ttls[n]);
497 }
498 } else {
499 /* unresolved mfc_caches don't contain
500 * pkt, bytes and wrong_if values
501 */
502 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
503 }
504 seq_putc(seq, '\n');
505 }
506 return 0;
507}
508
509static const struct seq_operations ipmr_mfc_seq_ops = {
510 .start = ipmr_mfc_seq_start,
511 .next = mr_mfc_seq_next,
512 .stop = mr_mfc_seq_stop,
513 .show = ipmr_mfc_seq_show,
514};
515
516static int ipmr_mfc_open(struct inode *inode, struct file *file)
517{
518 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
519 sizeof(struct mr_mfc_iter));
520}
521
522static const struct file_operations ip6mr_mfc_fops = {
523 .open = ipmr_mfc_open,
524 .read = seq_read,
525 .llseek = seq_lseek,
526 .release = seq_release_net,
527};
528#endif
529
530#ifdef CONFIG_IPV6_PIMSM_V2
531
532static int pim6_rcv(struct sk_buff *skb)
533{
534 struct pimreghdr *pim;
535 struct ipv6hdr *encap;
536 struct net_device *reg_dev = NULL;
537 struct net *net = dev_net(skb->dev);
538 struct mr_table *mrt;
539 struct flowi6 fl6 = {
540 .flowi6_iif = skb->dev->ifindex,
541 .flowi6_mark = skb->mark,
542 };
543 int reg_vif_num;
544
545 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
546 goto drop;
547
548 pim = (struct pimreghdr *)skb_transport_header(skb);
549 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) ||
550 (pim->flags & PIM_NULL_REGISTER) ||
551 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
552 sizeof(*pim), IPPROTO_PIM,
553 csum_partial((void *)pim, sizeof(*pim), 0)) &&
554 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
555 goto drop;
556
557 /* check if the inner packet is destined to mcast group */
558 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
559 sizeof(*pim));
560
561 if (!ipv6_addr_is_multicast(&encap->daddr) ||
562 encap->payload_len == 0 ||
563 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
564 goto drop;
565
566 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
567 goto drop;
568 reg_vif_num = mrt->mroute_reg_vif_num;
569
570 read_lock(&mrt_lock);
571 if (reg_vif_num >= 0)
572 reg_dev = mrt->vif_table[reg_vif_num].dev;
573 if (reg_dev)
574 dev_hold(reg_dev);
575 read_unlock(&mrt_lock);
576
577 if (!reg_dev)
578 goto drop;
579
580 skb->mac_header = skb->network_header;
581 skb_pull(skb, (u8 *)encap - skb->data);
582 skb_reset_network_header(skb);
583 skb->protocol = htons(ETH_P_IPV6);
584 skb->ip_summed = CHECKSUM_NONE;
585
586 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
587
588 netif_rx(skb);
589
590 dev_put(reg_dev);
591 return 0;
592 drop:
593 kfree_skb(skb);
594 return 0;
595}
596
597static const struct inet6_protocol pim6_protocol = {
598 .handler = pim6_rcv,
599};
600
601/* Service routines creating virtual interfaces: PIMREG */
602
603static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
604 struct net_device *dev)
605{
606 struct net *net = dev_net(dev);
607 struct mr_table *mrt;
608 struct flowi6 fl6 = {
609 .flowi6_oif = dev->ifindex,
610 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
611 .flowi6_mark = skb->mark,
612 };
613 int err;
614
615 err = ip6mr_fib_lookup(net, &fl6, &mrt);
616 if (err < 0) {
617 kfree_skb(skb);
618 return err;
619 }
620
621 read_lock(&mrt_lock);
622 dev->stats.tx_bytes += skb->len;
623 dev->stats.tx_packets++;
624 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
625 read_unlock(&mrt_lock);
626 kfree_skb(skb);
627 return NETDEV_TX_OK;
628}
629
630static int reg_vif_get_iflink(const struct net_device *dev)
631{
632 return 0;
633}
634
635static const struct net_device_ops reg_vif_netdev_ops = {
636 .ndo_start_xmit = reg_vif_xmit,
637 .ndo_get_iflink = reg_vif_get_iflink,
638};
639
640static void reg_vif_setup(struct net_device *dev)
641{
642 dev->type = ARPHRD_PIMREG;
643 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
644 dev->flags = IFF_NOARP;
645 dev->netdev_ops = ®_vif_netdev_ops;
646 dev->needs_free_netdev = true;
647 dev->features |= NETIF_F_NETNS_LOCAL;
648}
649
650static struct net_device *ip6mr_reg_vif(struct net *net, struct mr_table *mrt)
651{
652 struct net_device *dev;
653 char name[IFNAMSIZ];
654
655 if (mrt->id == RT6_TABLE_DFLT)
656 sprintf(name, "pim6reg");
657 else
658 sprintf(name, "pim6reg%u", mrt->id);
659
660 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
661 if (!dev)
662 return NULL;
663
664 dev_net_set(dev, net);
665
666 if (register_netdevice(dev)) {
667 free_netdev(dev);
668 return NULL;
669 }
670
671 if (dev_open(dev))
672 goto failure;
673
674 dev_hold(dev);
675 return dev;
676
677failure:
678 unregister_netdevice(dev);
679 return NULL;
680}
681#endif
682
683static int call_ip6mr_vif_entry_notifiers(struct net *net,
684 enum fib_event_type event_type,
685 struct vif_device *vif,
686 mifi_t vif_index, u32 tb_id)
687{
688 return mr_call_vif_notifiers(net, RTNL_FAMILY_IP6MR, event_type,
689 vif, vif_index, tb_id,
690 &net->ipv6.ipmr_seq);
691}
692
693static int call_ip6mr_mfc_entry_notifiers(struct net *net,
694 enum fib_event_type event_type,
695 struct mfc6_cache *mfc, u32 tb_id)
696{
697 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IP6MR, event_type,
698 &mfc->_c, tb_id, &net->ipv6.ipmr_seq);
699}
700
701/* Delete a VIF entry */
702static int mif6_delete(struct mr_table *mrt, int vifi, int notify,
703 struct list_head *head)
704{
705 struct vif_device *v;
706 struct net_device *dev;
707 struct inet6_dev *in6_dev;
708
709 if (vifi < 0 || vifi >= mrt->maxvif)
710 return -EADDRNOTAVAIL;
711
712 v = &mrt->vif_table[vifi];
713
714 if (VIF_EXISTS(mrt, vifi))
715 call_ip6mr_vif_entry_notifiers(read_pnet(&mrt->net),
716 FIB_EVENT_VIF_DEL, v, vifi,
717 mrt->id);
718
719 write_lock_bh(&mrt_lock);
720 dev = v->dev;
721 v->dev = NULL;
722
723 if (!dev) {
724 write_unlock_bh(&mrt_lock);
725 return -EADDRNOTAVAIL;
726 }
727
728#ifdef CONFIG_IPV6_PIMSM_V2
729 if (vifi == mrt->mroute_reg_vif_num)
730 mrt->mroute_reg_vif_num = -1;
731#endif
732
733 if (vifi + 1 == mrt->maxvif) {
734 int tmp;
735 for (tmp = vifi - 1; tmp >= 0; tmp--) {
736 if (VIF_EXISTS(mrt, tmp))
737 break;
738 }
739 mrt->maxvif = tmp + 1;
740 }
741
742 write_unlock_bh(&mrt_lock);
743
744 dev_set_allmulti(dev, -1);
745
746 in6_dev = __in6_dev_get(dev);
747 if (in6_dev) {
748 in6_dev->cnf.mc_forwarding--;
749 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
750 NETCONFA_MC_FORWARDING,
751 dev->ifindex, &in6_dev->cnf);
752 }
753
754 if ((v->flags & MIFF_REGISTER) && !notify)
755 unregister_netdevice_queue(dev, head);
756
757 dev_put(dev);
758 return 0;
759}
760
761static inline void ip6mr_cache_free_rcu(struct rcu_head *head)
762{
763 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
764
765 kmem_cache_free(mrt_cachep, (struct mfc6_cache *)c);
766}
767
768static inline void ip6mr_cache_free(struct mfc6_cache *c)
769{
770 call_rcu(&c->_c.rcu, ip6mr_cache_free_rcu);
771}
772
773/* Destroy an unresolved cache entry, killing queued skbs
774 and reporting error to netlink readers.
775 */
776
777static void ip6mr_destroy_unres(struct mr_table *mrt, struct mfc6_cache *c)
778{
779 struct net *net = read_pnet(&mrt->net);
780 struct sk_buff *skb;
781
782 atomic_dec(&mrt->cache_resolve_queue_len);
783
784 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved)) != NULL) {
785 if (ipv6_hdr(skb)->version == 0) {
786 struct nlmsghdr *nlh = skb_pull(skb,
787 sizeof(struct ipv6hdr));
788 nlh->nlmsg_type = NLMSG_ERROR;
789 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
790 skb_trim(skb, nlh->nlmsg_len);
791 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
792 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
793 } else
794 kfree_skb(skb);
795 }
796
797 ip6mr_cache_free(c);
798}
799
800
801/* Timer process for all the unresolved queue. */
802
803static void ipmr_do_expire_process(struct mr_table *mrt)
804{
805 unsigned long now = jiffies;
806 unsigned long expires = 10 * HZ;
807 struct mr_mfc *c, *next;
808
809 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
810 if (time_after(c->mfc_un.unres.expires, now)) {
811 /* not yet... */
812 unsigned long interval = c->mfc_un.unres.expires - now;
813 if (interval < expires)
814 expires = interval;
815 continue;
816 }
817
818 list_del(&c->list);
819 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE);
820 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c);
821 }
822
823 if (!list_empty(&mrt->mfc_unres_queue))
824 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
825}
826
827static void ipmr_expire_process(struct timer_list *t)
828{
829 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
830
831 if (!spin_trylock(&mfc_unres_lock)) {
832 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
833 return;
834 }
835
836 if (!list_empty(&mrt->mfc_unres_queue))
837 ipmr_do_expire_process(mrt);
838
839 spin_unlock(&mfc_unres_lock);
840}
841
842/* Fill oifs list. It is called under write locked mrt_lock. */
843
844static void ip6mr_update_thresholds(struct mr_table *mrt,
845 struct mr_mfc *cache,
846 unsigned char *ttls)
847{
848 int vifi;
849
850 cache->mfc_un.res.minvif = MAXMIFS;
851 cache->mfc_un.res.maxvif = 0;
852 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
853
854 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
855 if (VIF_EXISTS(mrt, vifi) &&
856 ttls[vifi] && ttls[vifi] < 255) {
857 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
858 if (cache->mfc_un.res.minvif > vifi)
859 cache->mfc_un.res.minvif = vifi;
860 if (cache->mfc_un.res.maxvif <= vifi)
861 cache->mfc_un.res.maxvif = vifi + 1;
862 }
863 }
864 cache->mfc_un.res.lastuse = jiffies;
865}
866
867static int mif6_add(struct net *net, struct mr_table *mrt,
868 struct mif6ctl *vifc, int mrtsock)
869{
870 int vifi = vifc->mif6c_mifi;
871 struct vif_device *v = &mrt->vif_table[vifi];
872 struct net_device *dev;
873 struct inet6_dev *in6_dev;
874 int err;
875
876 /* Is vif busy ? */
877 if (VIF_EXISTS(mrt, vifi))
878 return -EADDRINUSE;
879
880 switch (vifc->mif6c_flags) {
881#ifdef CONFIG_IPV6_PIMSM_V2
882 case MIFF_REGISTER:
883 /*
884 * Special Purpose VIF in PIM
885 * All the packets will be sent to the daemon
886 */
887 if (mrt->mroute_reg_vif_num >= 0)
888 return -EADDRINUSE;
889 dev = ip6mr_reg_vif(net, mrt);
890 if (!dev)
891 return -ENOBUFS;
892 err = dev_set_allmulti(dev, 1);
893 if (err) {
894 unregister_netdevice(dev);
895 dev_put(dev);
896 return err;
897 }
898 break;
899#endif
900 case 0:
901 dev = dev_get_by_index(net, vifc->mif6c_pifi);
902 if (!dev)
903 return -EADDRNOTAVAIL;
904 err = dev_set_allmulti(dev, 1);
905 if (err) {
906 dev_put(dev);
907 return err;
908 }
909 break;
910 default:
911 return -EINVAL;
912 }
913
914 in6_dev = __in6_dev_get(dev);
915 if (in6_dev) {
916 in6_dev->cnf.mc_forwarding++;
917 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
918 NETCONFA_MC_FORWARDING,
919 dev->ifindex, &in6_dev->cnf);
920 }
921
922 /* Fill in the VIF structures */
923 vif_device_init(v, dev, vifc->vifc_rate_limit, vifc->vifc_threshold,
924 vifc->mif6c_flags | (!mrtsock ? VIFF_STATIC : 0),
925 MIFF_REGISTER);
926
927 /* And finish update writing critical data */
928 write_lock_bh(&mrt_lock);
929 v->dev = dev;
930#ifdef CONFIG_IPV6_PIMSM_V2
931 if (v->flags & MIFF_REGISTER)
932 mrt->mroute_reg_vif_num = vifi;
933#endif
934 if (vifi + 1 > mrt->maxvif)
935 mrt->maxvif = vifi + 1;
936 write_unlock_bh(&mrt_lock);
937 call_ip6mr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD,
938 v, vifi, mrt->id);
939 return 0;
940}
941
942static struct mfc6_cache *ip6mr_cache_find(struct mr_table *mrt,
943 const struct in6_addr *origin,
944 const struct in6_addr *mcastgrp)
945{
946 struct mfc6_cache_cmp_arg arg = {
947 .mf6c_origin = *origin,
948 .mf6c_mcastgrp = *mcastgrp,
949 };
950
951 return mr_mfc_find(mrt, &arg);
952}
953
954/* Look for a (*,G) entry */
955static struct mfc6_cache *ip6mr_cache_find_any(struct mr_table *mrt,
956 struct in6_addr *mcastgrp,
957 mifi_t mifi)
958{
959 struct mfc6_cache_cmp_arg arg = {
960 .mf6c_origin = in6addr_any,
961 .mf6c_mcastgrp = *mcastgrp,
962 };
963
964 if (ipv6_addr_any(mcastgrp))
965 return mr_mfc_find_any_parent(mrt, mifi);
966 return mr_mfc_find_any(mrt, mifi, &arg);
967}
968
969/* Look for a (S,G,iif) entry if parent != -1 */
970static struct mfc6_cache *
971ip6mr_cache_find_parent(struct mr_table *mrt,
972 const struct in6_addr *origin,
973 const struct in6_addr *mcastgrp,
974 int parent)
975{
976 struct mfc6_cache_cmp_arg arg = {
977 .mf6c_origin = *origin,
978 .mf6c_mcastgrp = *mcastgrp,
979 };
980
981 return mr_mfc_find_parent(mrt, &arg, parent);
982}
983
984/* Allocate a multicast cache entry */
985static struct mfc6_cache *ip6mr_cache_alloc(void)
986{
987 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
988 if (!c)
989 return NULL;
990 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
991 c->_c.mfc_un.res.minvif = MAXMIFS;
992 c->_c.free = ip6mr_cache_free_rcu;
993 refcount_set(&c->_c.mfc_un.res.refcount, 1);
994 return c;
995}
996
997static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
998{
999 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1000 if (!c)
1001 return NULL;
1002 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1003 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1004 return c;
1005}
1006
1007/*
1008 * A cache entry has gone into a resolved state from queued
1009 */
1010
1011static void ip6mr_cache_resolve(struct net *net, struct mr_table *mrt,
1012 struct mfc6_cache *uc, struct mfc6_cache *c)
1013{
1014 struct sk_buff *skb;
1015
1016 /*
1017 * Play the pending entries through our router
1018 */
1019
1020 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1021 if (ipv6_hdr(skb)->version == 0) {
1022 struct nlmsghdr *nlh = skb_pull(skb,
1023 sizeof(struct ipv6hdr));
1024
1025 if (mr_fill_mroute(mrt, skb, &c->_c,
1026 nlmsg_data(nlh)) > 0) {
1027 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1028 } else {
1029 nlh->nlmsg_type = NLMSG_ERROR;
1030 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1031 skb_trim(skb, nlh->nlmsg_len);
1032 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1033 }
1034 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1035 } else
1036 ip6_mr_forward(net, mrt, skb, c);
1037 }
1038}
1039
1040/*
1041 * Bounce a cache query up to pim6sd and netlink.
1042 *
1043 * Called under mrt_lock.
1044 */
1045
1046static int ip6mr_cache_report(struct mr_table *mrt, struct sk_buff *pkt,
1047 mifi_t mifi, int assert)
1048{
1049 struct sock *mroute6_sk;
1050 struct sk_buff *skb;
1051 struct mrt6msg *msg;
1052 int ret;
1053
1054#ifdef CONFIG_IPV6_PIMSM_V2
1055 if (assert == MRT6MSG_WHOLEPKT)
1056 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1057 +sizeof(*msg));
1058 else
1059#endif
1060 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1061
1062 if (!skb)
1063 return -ENOBUFS;
1064
1065 /* I suppose that internal messages
1066 * do not require checksums */
1067
1068 skb->ip_summed = CHECKSUM_UNNECESSARY;
1069
1070#ifdef CONFIG_IPV6_PIMSM_V2
1071 if (assert == MRT6MSG_WHOLEPKT) {
1072 /* Ugly, but we have no choice with this interface.
1073 Duplicate old header, fix length etc.
1074 And all this only to mangle msg->im6_msgtype and
1075 to set msg->im6_mbz to "mbz" :-)
1076 */
1077 skb_push(skb, -skb_network_offset(pkt));
1078
1079 skb_push(skb, sizeof(*msg));
1080 skb_reset_transport_header(skb);
1081 msg = (struct mrt6msg *)skb_transport_header(skb);
1082 msg->im6_mbz = 0;
1083 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1084 msg->im6_mif = mrt->mroute_reg_vif_num;
1085 msg->im6_pad = 0;
1086 msg->im6_src = ipv6_hdr(pkt)->saddr;
1087 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1088
1089 skb->ip_summed = CHECKSUM_UNNECESSARY;
1090 } else
1091#endif
1092 {
1093 /*
1094 * Copy the IP header
1095 */
1096
1097 skb_put(skb, sizeof(struct ipv6hdr));
1098 skb_reset_network_header(skb);
1099 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1100
1101 /*
1102 * Add our header
1103 */
1104 skb_put(skb, sizeof(*msg));
1105 skb_reset_transport_header(skb);
1106 msg = (struct mrt6msg *)skb_transport_header(skb);
1107
1108 msg->im6_mbz = 0;
1109 msg->im6_msgtype = assert;
1110 msg->im6_mif = mifi;
1111 msg->im6_pad = 0;
1112 msg->im6_src = ipv6_hdr(pkt)->saddr;
1113 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1114
1115 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1116 skb->ip_summed = CHECKSUM_UNNECESSARY;
1117 }
1118
1119 rcu_read_lock();
1120 mroute6_sk = rcu_dereference(mrt->mroute_sk);
1121 if (!mroute6_sk) {
1122 rcu_read_unlock();
1123 kfree_skb(skb);
1124 return -EINVAL;
1125 }
1126
1127 mrt6msg_netlink_event(mrt, skb);
1128
1129 /* Deliver to user space multicast routing algorithms */
1130 ret = sock_queue_rcv_skb(mroute6_sk, skb);
1131 rcu_read_unlock();
1132 if (ret < 0) {
1133 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1134 kfree_skb(skb);
1135 }
1136
1137 return ret;
1138}
1139
1140/* Queue a packet for resolution. It gets locked cache entry! */
1141static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi,
1142 struct sk_buff *skb)
1143{
1144 struct mfc6_cache *c;
1145 bool found = false;
1146 int err;
1147
1148 spin_lock_bh(&mfc_unres_lock);
1149 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1150 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1151 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1152 found = true;
1153 break;
1154 }
1155 }
1156
1157 if (!found) {
1158 /*
1159 * Create a new entry if allowable
1160 */
1161
1162 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1163 (c = ip6mr_cache_alloc_unres()) == NULL) {
1164 spin_unlock_bh(&mfc_unres_lock);
1165
1166 kfree_skb(skb);
1167 return -ENOBUFS;
1168 }
1169
1170 /* Fill in the new cache entry */
1171 c->_c.mfc_parent = -1;
1172 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1173 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1174
1175 /*
1176 * Reflect first query at pim6sd
1177 */
1178 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1179 if (err < 0) {
1180 /* If the report failed throw the cache entry
1181 out - Brad Parker
1182 */
1183 spin_unlock_bh(&mfc_unres_lock);
1184
1185 ip6mr_cache_free(c);
1186 kfree_skb(skb);
1187 return err;
1188 }
1189
1190 atomic_inc(&mrt->cache_resolve_queue_len);
1191 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1192 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1193
1194 ipmr_do_expire_process(mrt);
1195 }
1196
1197 /* See if we can append the packet */
1198 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1199 kfree_skb(skb);
1200 err = -ENOBUFS;
1201 } else {
1202 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1203 err = 0;
1204 }
1205
1206 spin_unlock_bh(&mfc_unres_lock);
1207 return err;
1208}
1209
1210/*
1211 * MFC6 cache manipulation by user space
1212 */
1213
1214static int ip6mr_mfc_delete(struct mr_table *mrt, struct mf6cctl *mfc,
1215 int parent)
1216{
1217 struct mfc6_cache *c;
1218
1219 /* The entries are added/deleted only under RTNL */
1220 rcu_read_lock();
1221 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr,
1222 &mfc->mf6cc_mcastgrp.sin6_addr, parent);
1223 rcu_read_unlock();
1224 if (!c)
1225 return -ENOENT;
1226 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ip6mr_rht_params);
1227 list_del_rcu(&c->_c.list);
1228
1229 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net),
1230 FIB_EVENT_ENTRY_DEL, c, mrt->id);
1231 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1232 mr_cache_put(&c->_c);
1233 return 0;
1234}
1235
1236static int ip6mr_device_event(struct notifier_block *this,
1237 unsigned long event, void *ptr)
1238{
1239 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1240 struct net *net = dev_net(dev);
1241 struct mr_table *mrt;
1242 struct vif_device *v;
1243 int ct;
1244
1245 if (event != NETDEV_UNREGISTER)
1246 return NOTIFY_DONE;
1247
1248 ip6mr_for_each_table(mrt, net) {
1249 v = &mrt->vif_table[0];
1250 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1251 if (v->dev == dev)
1252 mif6_delete(mrt, ct, 1, NULL);
1253 }
1254 }
1255
1256 return NOTIFY_DONE;
1257}
1258
1259static unsigned int ip6mr_seq_read(struct net *net)
1260{
1261 ASSERT_RTNL();
1262
1263 return net->ipv6.ipmr_seq + ip6mr_rules_seq_read(net);
1264}
1265
1266static int ip6mr_dump(struct net *net, struct notifier_block *nb)
1267{
1268 return mr_dump(net, nb, RTNL_FAMILY_IP6MR, ip6mr_rules_dump,
1269 ip6mr_mr_table_iter, &mrt_lock);
1270}
1271
1272static struct notifier_block ip6_mr_notifier = {
1273 .notifier_call = ip6mr_device_event
1274};
1275
1276static const struct fib_notifier_ops ip6mr_notifier_ops_template = {
1277 .family = RTNL_FAMILY_IP6MR,
1278 .fib_seq_read = ip6mr_seq_read,
1279 .fib_dump = ip6mr_dump,
1280 .owner = THIS_MODULE,
1281};
1282
1283static int __net_init ip6mr_notifier_init(struct net *net)
1284{
1285 struct fib_notifier_ops *ops;
1286
1287 net->ipv6.ipmr_seq = 0;
1288
1289 ops = fib_notifier_ops_register(&ip6mr_notifier_ops_template, net);
1290 if (IS_ERR(ops))
1291 return PTR_ERR(ops);
1292
1293 net->ipv6.ip6mr_notifier_ops = ops;
1294
1295 return 0;
1296}
1297
1298static void __net_exit ip6mr_notifier_exit(struct net *net)
1299{
1300 fib_notifier_ops_unregister(net->ipv6.ip6mr_notifier_ops);
1301 net->ipv6.ip6mr_notifier_ops = NULL;
1302}
1303
1304/* Setup for IP multicast routing */
1305static int __net_init ip6mr_net_init(struct net *net)
1306{
1307 int err;
1308
1309 err = ip6mr_notifier_init(net);
1310 if (err)
1311 return err;
1312
1313 err = ip6mr_rules_init(net);
1314 if (err < 0)
1315 goto ip6mr_rules_fail;
1316
1317#ifdef CONFIG_PROC_FS
1318 err = -ENOMEM;
1319 if (!proc_create("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_fops))
1320 goto proc_vif_fail;
1321 if (!proc_create("ip6_mr_cache", 0, net->proc_net, &ip6mr_mfc_fops))
1322 goto proc_cache_fail;
1323#endif
1324
1325 return 0;
1326
1327#ifdef CONFIG_PROC_FS
1328proc_cache_fail:
1329 remove_proc_entry("ip6_mr_vif", net->proc_net);
1330proc_vif_fail:
1331 ip6mr_rules_exit(net);
1332#endif
1333ip6mr_rules_fail:
1334 ip6mr_notifier_exit(net);
1335 return err;
1336}
1337
1338static void __net_exit ip6mr_net_exit(struct net *net)
1339{
1340#ifdef CONFIG_PROC_FS
1341 remove_proc_entry("ip6_mr_cache", net->proc_net);
1342 remove_proc_entry("ip6_mr_vif", net->proc_net);
1343#endif
1344 ip6mr_rules_exit(net);
1345 ip6mr_notifier_exit(net);
1346}
1347
1348static struct pernet_operations ip6mr_net_ops = {
1349 .init = ip6mr_net_init,
1350 .exit = ip6mr_net_exit,
1351};
1352
1353int __init ip6_mr_init(void)
1354{
1355 int err;
1356
1357 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1358 sizeof(struct mfc6_cache),
1359 0, SLAB_HWCACHE_ALIGN,
1360 NULL);
1361 if (!mrt_cachep)
1362 return -ENOMEM;
1363
1364 err = register_pernet_subsys(&ip6mr_net_ops);
1365 if (err)
1366 goto reg_pernet_fail;
1367
1368 err = register_netdevice_notifier(&ip6_mr_notifier);
1369 if (err)
1370 goto reg_notif_fail;
1371#ifdef CONFIG_IPV6_PIMSM_V2
1372 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1373 pr_err("%s: can't add PIM protocol\n", __func__);
1374 err = -EAGAIN;
1375 goto add_proto_fail;
1376 }
1377#endif
1378 err = rtnl_register_module(THIS_MODULE, RTNL_FAMILY_IP6MR, RTM_GETROUTE,
1379 NULL, ip6mr_rtm_dumproute, 0);
1380 if (err == 0)
1381 return 0;
1382
1383#ifdef CONFIG_IPV6_PIMSM_V2
1384 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1385add_proto_fail:
1386 unregister_netdevice_notifier(&ip6_mr_notifier);
1387#endif
1388reg_notif_fail:
1389 unregister_pernet_subsys(&ip6mr_net_ops);
1390reg_pernet_fail:
1391 kmem_cache_destroy(mrt_cachep);
1392 return err;
1393}
1394
1395void ip6_mr_cleanup(void)
1396{
1397 rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1398#ifdef CONFIG_IPV6_PIMSM_V2
1399 inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1400#endif
1401 unregister_netdevice_notifier(&ip6_mr_notifier);
1402 unregister_pernet_subsys(&ip6mr_net_ops);
1403 kmem_cache_destroy(mrt_cachep);
1404}
1405
1406static int ip6mr_mfc_add(struct net *net, struct mr_table *mrt,
1407 struct mf6cctl *mfc, int mrtsock, int parent)
1408{
1409 unsigned char ttls[MAXMIFS];
1410 struct mfc6_cache *uc, *c;
1411 struct mr_mfc *_uc;
1412 bool found;
1413 int i, err;
1414
1415 if (mfc->mf6cc_parent >= MAXMIFS)
1416 return -ENFILE;
1417
1418 memset(ttls, 255, MAXMIFS);
1419 for (i = 0; i < MAXMIFS; i++) {
1420 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1421 ttls[i] = 1;
1422 }
1423
1424 /* The entries are added/deleted only under RTNL */
1425 rcu_read_lock();
1426 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr,
1427 &mfc->mf6cc_mcastgrp.sin6_addr, parent);
1428 rcu_read_unlock();
1429 if (c) {
1430 write_lock_bh(&mrt_lock);
1431 c->_c.mfc_parent = mfc->mf6cc_parent;
1432 ip6mr_update_thresholds(mrt, &c->_c, ttls);
1433 if (!mrtsock)
1434 c->_c.mfc_flags |= MFC_STATIC;
1435 write_unlock_bh(&mrt_lock);
1436 call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE,
1437 c, mrt->id);
1438 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1439 return 0;
1440 }
1441
1442 if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1443 !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1444 return -EINVAL;
1445
1446 c = ip6mr_cache_alloc();
1447 if (!c)
1448 return -ENOMEM;
1449
1450 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1451 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1452 c->_c.mfc_parent = mfc->mf6cc_parent;
1453 ip6mr_update_thresholds(mrt, &c->_c, ttls);
1454 if (!mrtsock)
1455 c->_c.mfc_flags |= MFC_STATIC;
1456
1457 err = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1458 ip6mr_rht_params);
1459 if (err) {
1460 pr_err("ip6mr: rhtable insert error %d\n", err);
1461 ip6mr_cache_free(c);
1462 return err;
1463 }
1464 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1465
1466 /* Check to see if we resolved a queued list. If so we
1467 * need to send on the frames and tidy up.
1468 */
1469 found = false;
1470 spin_lock_bh(&mfc_unres_lock);
1471 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1472 uc = (struct mfc6_cache *)_uc;
1473 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1474 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1475 list_del(&_uc->list);
1476 atomic_dec(&mrt->cache_resolve_queue_len);
1477 found = true;
1478 break;
1479 }
1480 }
1481 if (list_empty(&mrt->mfc_unres_queue))
1482 del_timer(&mrt->ipmr_expire_timer);
1483 spin_unlock_bh(&mfc_unres_lock);
1484
1485 if (found) {
1486 ip6mr_cache_resolve(net, mrt, uc, c);
1487 ip6mr_cache_free(uc);
1488 }
1489 call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD,
1490 c, mrt->id);
1491 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1492 return 0;
1493}
1494
1495/*
1496 * Close the multicast socket, and clear the vif tables etc
1497 */
1498
1499static void mroute_clean_tables(struct mr_table *mrt, bool all)
1500{
1501 struct mr_mfc *c, *tmp;
1502 LIST_HEAD(list);
1503 int i;
1504
1505 /* Shut down all active vif entries */
1506 for (i = 0; i < mrt->maxvif; i++) {
1507 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1508 continue;
1509 mif6_delete(mrt, i, 0, &list);
1510 }
1511 unregister_netdevice_many(&list);
1512
1513 /* Wipe the cache */
1514 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1515 if (!all && (c->mfc_flags & MFC_STATIC))
1516 continue;
1517 rhltable_remove(&mrt->mfc_hash, &c->mnode, ip6mr_rht_params);
1518 list_del_rcu(&c->list);
1519 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE);
1520 mr_cache_put(c);
1521 }
1522
1523 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1524 spin_lock_bh(&mfc_unres_lock);
1525 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1526 list_del(&c->list);
1527 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net),
1528 FIB_EVENT_ENTRY_DEL,
1529 (struct mfc6_cache *)c,
1530 mrt->id);
1531 mr6_netlink_event(mrt, (struct mfc6_cache *)c,
1532 RTM_DELROUTE);
1533 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c);
1534 }
1535 spin_unlock_bh(&mfc_unres_lock);
1536 }
1537}
1538
1539static int ip6mr_sk_init(struct mr_table *mrt, struct sock *sk)
1540{
1541 int err = 0;
1542 struct net *net = sock_net(sk);
1543
1544 rtnl_lock();
1545 write_lock_bh(&mrt_lock);
1546 if (rtnl_dereference(mrt->mroute_sk)) {
1547 err = -EADDRINUSE;
1548 } else {
1549 rcu_assign_pointer(mrt->mroute_sk, sk);
1550 sock_set_flag(sk, SOCK_RCU_FREE);
1551 net->ipv6.devconf_all->mc_forwarding++;
1552 }
1553 write_unlock_bh(&mrt_lock);
1554
1555 if (!err)
1556 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1557 NETCONFA_MC_FORWARDING,
1558 NETCONFA_IFINDEX_ALL,
1559 net->ipv6.devconf_all);
1560 rtnl_unlock();
1561
1562 return err;
1563}
1564
1565int ip6mr_sk_done(struct sock *sk)
1566{
1567 int err = -EACCES;
1568 struct net *net = sock_net(sk);
1569 struct mr_table *mrt;
1570
1571 if (sk->sk_type != SOCK_RAW ||
1572 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1573 return err;
1574
1575 rtnl_lock();
1576 ip6mr_for_each_table(mrt, net) {
1577 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1578 write_lock_bh(&mrt_lock);
1579 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1580 /* Note that mroute_sk had SOCK_RCU_FREE set,
1581 * so the RCU grace period before sk freeing
1582 * is guaranteed by sk_destruct()
1583 */
1584 net->ipv6.devconf_all->mc_forwarding--;
1585 write_unlock_bh(&mrt_lock);
1586 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1587 NETCONFA_MC_FORWARDING,
1588 NETCONFA_IFINDEX_ALL,
1589 net->ipv6.devconf_all);
1590
1591 mroute_clean_tables(mrt, false);
1592 err = 0;
1593 break;
1594 }
1595 }
1596 rtnl_unlock();
1597
1598 return err;
1599}
1600
1601bool mroute6_is_socket(struct net *net, struct sk_buff *skb)
1602{
1603 struct mr_table *mrt;
1604 struct flowi6 fl6 = {
1605 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
1606 .flowi6_oif = skb->dev->ifindex,
1607 .flowi6_mark = skb->mark,
1608 };
1609
1610 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1611 return NULL;
1612
1613 return rcu_access_pointer(mrt->mroute_sk);
1614}
1615EXPORT_SYMBOL(mroute6_is_socket);
1616
1617/*
1618 * Socket options and virtual interface manipulation. The whole
1619 * virtual interface system is a complete heap, but unfortunately
1620 * that's how BSD mrouted happens to think. Maybe one day with a proper
1621 * MOSPF/PIM router set up we can clean this up.
1622 */
1623
1624int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1625{
1626 int ret, parent = 0;
1627 struct mif6ctl vif;
1628 struct mf6cctl mfc;
1629 mifi_t mifi;
1630 struct net *net = sock_net(sk);
1631 struct mr_table *mrt;
1632
1633 if (sk->sk_type != SOCK_RAW ||
1634 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1635 return -EOPNOTSUPP;
1636
1637 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1638 if (!mrt)
1639 return -ENOENT;
1640
1641 if (optname != MRT6_INIT) {
1642 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1643 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1644 return -EACCES;
1645 }
1646
1647 switch (optname) {
1648 case MRT6_INIT:
1649 if (optlen < sizeof(int))
1650 return -EINVAL;
1651
1652 return ip6mr_sk_init(mrt, sk);
1653
1654 case MRT6_DONE:
1655 return ip6mr_sk_done(sk);
1656
1657 case MRT6_ADD_MIF:
1658 if (optlen < sizeof(vif))
1659 return -EINVAL;
1660 if (copy_from_user(&vif, optval, sizeof(vif)))
1661 return -EFAULT;
1662 if (vif.mif6c_mifi >= MAXMIFS)
1663 return -ENFILE;
1664 rtnl_lock();
1665 ret = mif6_add(net, mrt, &vif,
1666 sk == rtnl_dereference(mrt->mroute_sk));
1667 rtnl_unlock();
1668 return ret;
1669
1670 case MRT6_DEL_MIF:
1671 if (optlen < sizeof(mifi_t))
1672 return -EINVAL;
1673 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1674 return -EFAULT;
1675 rtnl_lock();
1676 ret = mif6_delete(mrt, mifi, 0, NULL);
1677 rtnl_unlock();
1678 return ret;
1679
1680 /*
1681 * Manipulate the forwarding caches. These live
1682 * in a sort of kernel/user symbiosis.
1683 */
1684 case MRT6_ADD_MFC:
1685 case MRT6_DEL_MFC:
1686 parent = -1;
1687 /* fall through */
1688 case MRT6_ADD_MFC_PROXY:
1689 case MRT6_DEL_MFC_PROXY:
1690 if (optlen < sizeof(mfc))
1691 return -EINVAL;
1692 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1693 return -EFAULT;
1694 if (parent == 0)
1695 parent = mfc.mf6cc_parent;
1696 rtnl_lock();
1697 if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1698 ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1699 else
1700 ret = ip6mr_mfc_add(net, mrt, &mfc,
1701 sk ==
1702 rtnl_dereference(mrt->mroute_sk),
1703 parent);
1704 rtnl_unlock();
1705 return ret;
1706
1707 /*
1708 * Control PIM assert (to activate pim will activate assert)
1709 */
1710 case MRT6_ASSERT:
1711 {
1712 int v;
1713
1714 if (optlen != sizeof(v))
1715 return -EINVAL;
1716 if (get_user(v, (int __user *)optval))
1717 return -EFAULT;
1718 mrt->mroute_do_assert = v;
1719 return 0;
1720 }
1721
1722#ifdef CONFIG_IPV6_PIMSM_V2
1723 case MRT6_PIM:
1724 {
1725 int v;
1726
1727 if (optlen != sizeof(v))
1728 return -EINVAL;
1729 if (get_user(v, (int __user *)optval))
1730 return -EFAULT;
1731 v = !!v;
1732 rtnl_lock();
1733 ret = 0;
1734 if (v != mrt->mroute_do_pim) {
1735 mrt->mroute_do_pim = v;
1736 mrt->mroute_do_assert = v;
1737 }
1738 rtnl_unlock();
1739 return ret;
1740 }
1741
1742#endif
1743#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1744 case MRT6_TABLE:
1745 {
1746 u32 v;
1747
1748 if (optlen != sizeof(u32))
1749 return -EINVAL;
1750 if (get_user(v, (u32 __user *)optval))
1751 return -EFAULT;
1752 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1753 if (v != RT_TABLE_DEFAULT && v >= 100000000)
1754 return -EINVAL;
1755 if (sk == rcu_access_pointer(mrt->mroute_sk))
1756 return -EBUSY;
1757
1758 rtnl_lock();
1759 ret = 0;
1760 if (!ip6mr_new_table(net, v))
1761 ret = -ENOMEM;
1762 raw6_sk(sk)->ip6mr_table = v;
1763 rtnl_unlock();
1764 return ret;
1765 }
1766#endif
1767 /*
1768 * Spurious command, or MRT6_VERSION which you cannot
1769 * set.
1770 */
1771 default:
1772 return -ENOPROTOOPT;
1773 }
1774}
1775
1776/*
1777 * Getsock opt support for the multicast routing system.
1778 */
1779
1780int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1781 int __user *optlen)
1782{
1783 int olr;
1784 int val;
1785 struct net *net = sock_net(sk);
1786 struct mr_table *mrt;
1787
1788 if (sk->sk_type != SOCK_RAW ||
1789 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1790 return -EOPNOTSUPP;
1791
1792 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1793 if (!mrt)
1794 return -ENOENT;
1795
1796 switch (optname) {
1797 case MRT6_VERSION:
1798 val = 0x0305;
1799 break;
1800#ifdef CONFIG_IPV6_PIMSM_V2
1801 case MRT6_PIM:
1802 val = mrt->mroute_do_pim;
1803 break;
1804#endif
1805 case MRT6_ASSERT:
1806 val = mrt->mroute_do_assert;
1807 break;
1808 default:
1809 return -ENOPROTOOPT;
1810 }
1811
1812 if (get_user(olr, optlen))
1813 return -EFAULT;
1814
1815 olr = min_t(int, olr, sizeof(int));
1816 if (olr < 0)
1817 return -EINVAL;
1818
1819 if (put_user(olr, optlen))
1820 return -EFAULT;
1821 if (copy_to_user(optval, &val, olr))
1822 return -EFAULT;
1823 return 0;
1824}
1825
1826/*
1827 * The IP multicast ioctl support routines.
1828 */
1829
1830int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1831{
1832 struct sioc_sg_req6 sr;
1833 struct sioc_mif_req6 vr;
1834 struct vif_device *vif;
1835 struct mfc6_cache *c;
1836 struct net *net = sock_net(sk);
1837 struct mr_table *mrt;
1838
1839 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1840 if (!mrt)
1841 return -ENOENT;
1842
1843 switch (cmd) {
1844 case SIOCGETMIFCNT_IN6:
1845 if (copy_from_user(&vr, arg, sizeof(vr)))
1846 return -EFAULT;
1847 if (vr.mifi >= mrt->maxvif)
1848 return -EINVAL;
1849 read_lock(&mrt_lock);
1850 vif = &mrt->vif_table[vr.mifi];
1851 if (VIF_EXISTS(mrt, vr.mifi)) {
1852 vr.icount = vif->pkt_in;
1853 vr.ocount = vif->pkt_out;
1854 vr.ibytes = vif->bytes_in;
1855 vr.obytes = vif->bytes_out;
1856 read_unlock(&mrt_lock);
1857
1858 if (copy_to_user(arg, &vr, sizeof(vr)))
1859 return -EFAULT;
1860 return 0;
1861 }
1862 read_unlock(&mrt_lock);
1863 return -EADDRNOTAVAIL;
1864 case SIOCGETSGCNT_IN6:
1865 if (copy_from_user(&sr, arg, sizeof(sr)))
1866 return -EFAULT;
1867
1868 rcu_read_lock();
1869 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1870 if (c) {
1871 sr.pktcnt = c->_c.mfc_un.res.pkt;
1872 sr.bytecnt = c->_c.mfc_un.res.bytes;
1873 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1874 rcu_read_unlock();
1875
1876 if (copy_to_user(arg, &sr, sizeof(sr)))
1877 return -EFAULT;
1878 return 0;
1879 }
1880 rcu_read_unlock();
1881 return -EADDRNOTAVAIL;
1882 default:
1883 return -ENOIOCTLCMD;
1884 }
1885}
1886
1887#ifdef CONFIG_COMPAT
1888struct compat_sioc_sg_req6 {
1889 struct sockaddr_in6 src;
1890 struct sockaddr_in6 grp;
1891 compat_ulong_t pktcnt;
1892 compat_ulong_t bytecnt;
1893 compat_ulong_t wrong_if;
1894};
1895
1896struct compat_sioc_mif_req6 {
1897 mifi_t mifi;
1898 compat_ulong_t icount;
1899 compat_ulong_t ocount;
1900 compat_ulong_t ibytes;
1901 compat_ulong_t obytes;
1902};
1903
1904int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1905{
1906 struct compat_sioc_sg_req6 sr;
1907 struct compat_sioc_mif_req6 vr;
1908 struct vif_device *vif;
1909 struct mfc6_cache *c;
1910 struct net *net = sock_net(sk);
1911 struct mr_table *mrt;
1912
1913 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1914 if (!mrt)
1915 return -ENOENT;
1916
1917 switch (cmd) {
1918 case SIOCGETMIFCNT_IN6:
1919 if (copy_from_user(&vr, arg, sizeof(vr)))
1920 return -EFAULT;
1921 if (vr.mifi >= mrt->maxvif)
1922 return -EINVAL;
1923 read_lock(&mrt_lock);
1924 vif = &mrt->vif_table[vr.mifi];
1925 if (VIF_EXISTS(mrt, vr.mifi)) {
1926 vr.icount = vif->pkt_in;
1927 vr.ocount = vif->pkt_out;
1928 vr.ibytes = vif->bytes_in;
1929 vr.obytes = vif->bytes_out;
1930 read_unlock(&mrt_lock);
1931
1932 if (copy_to_user(arg, &vr, sizeof(vr)))
1933 return -EFAULT;
1934 return 0;
1935 }
1936 read_unlock(&mrt_lock);
1937 return -EADDRNOTAVAIL;
1938 case SIOCGETSGCNT_IN6:
1939 if (copy_from_user(&sr, arg, sizeof(sr)))
1940 return -EFAULT;
1941
1942 rcu_read_lock();
1943 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1944 if (c) {
1945 sr.pktcnt = c->_c.mfc_un.res.pkt;
1946 sr.bytecnt = c->_c.mfc_un.res.bytes;
1947 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1948 rcu_read_unlock();
1949
1950 if (copy_to_user(arg, &sr, sizeof(sr)))
1951 return -EFAULT;
1952 return 0;
1953 }
1954 rcu_read_unlock();
1955 return -EADDRNOTAVAIL;
1956 default:
1957 return -ENOIOCTLCMD;
1958 }
1959}
1960#endif
1961
1962static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1963{
1964 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
1965 IPSTATS_MIB_OUTFORWDATAGRAMS);
1966 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)),
1967 IPSTATS_MIB_OUTOCTETS, skb->len);
1968 return dst_output(net, sk, skb);
1969}
1970
1971/*
1972 * Processing handlers for ip6mr_forward
1973 */
1974
1975static int ip6mr_forward2(struct net *net, struct mr_table *mrt,
1976 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1977{
1978 struct ipv6hdr *ipv6h;
1979 struct vif_device *vif = &mrt->vif_table[vifi];
1980 struct net_device *dev;
1981 struct dst_entry *dst;
1982 struct flowi6 fl6;
1983
1984 if (!vif->dev)
1985 goto out_free;
1986
1987#ifdef CONFIG_IPV6_PIMSM_V2
1988 if (vif->flags & MIFF_REGISTER) {
1989 vif->pkt_out++;
1990 vif->bytes_out += skb->len;
1991 vif->dev->stats.tx_bytes += skb->len;
1992 vif->dev->stats.tx_packets++;
1993 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1994 goto out_free;
1995 }
1996#endif
1997
1998 ipv6h = ipv6_hdr(skb);
1999
2000 fl6 = (struct flowi6) {
2001 .flowi6_oif = vif->link,
2002 .daddr = ipv6h->daddr,
2003 };
2004
2005 dst = ip6_route_output(net, NULL, &fl6);
2006 if (dst->error) {
2007 dst_release(dst);
2008 goto out_free;
2009 }
2010
2011 skb_dst_drop(skb);
2012 skb_dst_set(skb, dst);
2013
2014 /*
2015 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2016 * not only before forwarding, but after forwarding on all output
2017 * interfaces. It is clear, if mrouter runs a multicasting
2018 * program, it should receive packets not depending to what interface
2019 * program is joined.
2020 * If we will not make it, the program will have to join on all
2021 * interfaces. On the other hand, multihoming host (or router, but
2022 * not mrouter) cannot join to more than one interface - it will
2023 * result in receiving multiple packets.
2024 */
2025 dev = vif->dev;
2026 skb->dev = dev;
2027 vif->pkt_out++;
2028 vif->bytes_out += skb->len;
2029
2030 /* We are about to write */
2031 /* XXX: extension headers? */
2032 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2033 goto out_free;
2034
2035 ipv6h = ipv6_hdr(skb);
2036 ipv6h->hop_limit--;
2037
2038 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2039
2040 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2041 net, NULL, skb, skb->dev, dev,
2042 ip6mr_forward2_finish);
2043
2044out_free:
2045 kfree_skb(skb);
2046 return 0;
2047}
2048
2049static int ip6mr_find_vif(struct mr_table *mrt, struct net_device *dev)
2050{
2051 int ct;
2052
2053 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2054 if (mrt->vif_table[ct].dev == dev)
2055 break;
2056 }
2057 return ct;
2058}
2059
2060static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
2061 struct sk_buff *skb, struct mfc6_cache *c)
2062{
2063 int psend = -1;
2064 int vif, ct;
2065 int true_vifi = ip6mr_find_vif(mrt, skb->dev);
2066
2067 vif = c->_c.mfc_parent;
2068 c->_c.mfc_un.res.pkt++;
2069 c->_c.mfc_un.res.bytes += skb->len;
2070 c->_c.mfc_un.res.lastuse = jiffies;
2071
2072 if (ipv6_addr_any(&c->mf6c_origin) && true_vifi >= 0) {
2073 struct mfc6_cache *cache_proxy;
2074
2075 /* For an (*,G) entry, we only check that the incoming
2076 * interface is part of the static tree.
2077 */
2078 rcu_read_lock();
2079 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2080 if (cache_proxy &&
2081 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) {
2082 rcu_read_unlock();
2083 goto forward;
2084 }
2085 rcu_read_unlock();
2086 }
2087
2088 /*
2089 * Wrong interface: drop packet and (maybe) send PIM assert.
2090 */
2091 if (mrt->vif_table[vif].dev != skb->dev) {
2092 c->_c.mfc_un.res.wrong_if++;
2093
2094 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2095 /* pimsm uses asserts, when switching from RPT to SPT,
2096 so that we cannot check that packet arrived on an oif.
2097 It is bad, but otherwise we would need to move pretty
2098 large chunk of pimd to kernel. Ough... --ANK
2099 */
2100 (mrt->mroute_do_pim ||
2101 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2102 time_after(jiffies,
2103 c->_c.mfc_un.res.last_assert +
2104 MFC_ASSERT_THRESH)) {
2105 c->_c.mfc_un.res.last_assert = jiffies;
2106 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2107 }
2108 goto dont_forward;
2109 }
2110
2111forward:
2112 mrt->vif_table[vif].pkt_in++;
2113 mrt->vif_table[vif].bytes_in += skb->len;
2114
2115 /*
2116 * Forward the frame
2117 */
2118 if (ipv6_addr_any(&c->mf6c_origin) &&
2119 ipv6_addr_any(&c->mf6c_mcastgrp)) {
2120 if (true_vifi >= 0 &&
2121 true_vifi != c->_c.mfc_parent &&
2122 ipv6_hdr(skb)->hop_limit >
2123 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2124 /* It's an (*,*) entry and the packet is not coming from
2125 * the upstream: forward the packet to the upstream
2126 * only.
2127 */
2128 psend = c->_c.mfc_parent;
2129 goto last_forward;
2130 }
2131 goto dont_forward;
2132 }
2133 for (ct = c->_c.mfc_un.res.maxvif - 1;
2134 ct >= c->_c.mfc_un.res.minvif; ct--) {
2135 /* For (*,G) entry, don't forward to the incoming interface */
2136 if ((!ipv6_addr_any(&c->mf6c_origin) || ct != true_vifi) &&
2137 ipv6_hdr(skb)->hop_limit > c->_c.mfc_un.res.ttls[ct]) {
2138 if (psend != -1) {
2139 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2140 if (skb2)
2141 ip6mr_forward2(net, mrt, skb2,
2142 c, psend);
2143 }
2144 psend = ct;
2145 }
2146 }
2147last_forward:
2148 if (psend != -1) {
2149 ip6mr_forward2(net, mrt, skb, c, psend);
2150 return;
2151 }
2152
2153dont_forward:
2154 kfree_skb(skb);
2155}
2156
2157
2158/*
2159 * Multicast packets for forwarding arrive here
2160 */
2161
2162int ip6_mr_input(struct sk_buff *skb)
2163{
2164 struct mfc6_cache *cache;
2165 struct net *net = dev_net(skb->dev);
2166 struct mr_table *mrt;
2167 struct flowi6 fl6 = {
2168 .flowi6_iif = skb->dev->ifindex,
2169 .flowi6_mark = skb->mark,
2170 };
2171 int err;
2172
2173 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2174 if (err < 0) {
2175 kfree_skb(skb);
2176 return err;
2177 }
2178
2179 read_lock(&mrt_lock);
2180 cache = ip6mr_cache_find(mrt,
2181 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2182 if (!cache) {
2183 int vif = ip6mr_find_vif(mrt, skb->dev);
2184
2185 if (vif >= 0)
2186 cache = ip6mr_cache_find_any(mrt,
2187 &ipv6_hdr(skb)->daddr,
2188 vif);
2189 }
2190
2191 /*
2192 * No usable cache entry
2193 */
2194 if (!cache) {
2195 int vif;
2196
2197 vif = ip6mr_find_vif(mrt, skb->dev);
2198 if (vif >= 0) {
2199 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2200 read_unlock(&mrt_lock);
2201
2202 return err;
2203 }
2204 read_unlock(&mrt_lock);
2205 kfree_skb(skb);
2206 return -ENODEV;
2207 }
2208
2209 ip6_mr_forward(net, mrt, skb, cache);
2210
2211 read_unlock(&mrt_lock);
2212
2213 return 0;
2214}
2215
2216int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2217 u32 portid)
2218{
2219 int err;
2220 struct mr_table *mrt;
2221 struct mfc6_cache *cache;
2222 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2223
2224 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2225 if (!mrt)
2226 return -ENOENT;
2227
2228 read_lock(&mrt_lock);
2229 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2230 if (!cache && skb->dev) {
2231 int vif = ip6mr_find_vif(mrt, skb->dev);
2232
2233 if (vif >= 0)
2234 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2235 vif);
2236 }
2237
2238 if (!cache) {
2239 struct sk_buff *skb2;
2240 struct ipv6hdr *iph;
2241 struct net_device *dev;
2242 int vif;
2243
2244 dev = skb->dev;
2245 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2246 read_unlock(&mrt_lock);
2247 return -ENODEV;
2248 }
2249
2250 /* really correct? */
2251 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2252 if (!skb2) {
2253 read_unlock(&mrt_lock);
2254 return -ENOMEM;
2255 }
2256
2257 NETLINK_CB(skb2).portid = portid;
2258 skb_reset_transport_header(skb2);
2259
2260 skb_put(skb2, sizeof(struct ipv6hdr));
2261 skb_reset_network_header(skb2);
2262
2263 iph = ipv6_hdr(skb2);
2264 iph->version = 0;
2265 iph->priority = 0;
2266 iph->flow_lbl[0] = 0;
2267 iph->flow_lbl[1] = 0;
2268 iph->flow_lbl[2] = 0;
2269 iph->payload_len = 0;
2270 iph->nexthdr = IPPROTO_NONE;
2271 iph->hop_limit = 0;
2272 iph->saddr = rt->rt6i_src.addr;
2273 iph->daddr = rt->rt6i_dst.addr;
2274
2275 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2276 read_unlock(&mrt_lock);
2277
2278 return err;
2279 }
2280
2281 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2282 read_unlock(&mrt_lock);
2283 return err;
2284}
2285
2286static int ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2287 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2288 int flags)
2289{
2290 struct nlmsghdr *nlh;
2291 struct rtmsg *rtm;
2292 int err;
2293
2294 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2295 if (!nlh)
2296 return -EMSGSIZE;
2297
2298 rtm = nlmsg_data(nlh);
2299 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2300 rtm->rtm_dst_len = 128;
2301 rtm->rtm_src_len = 128;
2302 rtm->rtm_tos = 0;
2303 rtm->rtm_table = mrt->id;
2304 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2305 goto nla_put_failure;
2306 rtm->rtm_type = RTN_MULTICAST;
2307 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2308 if (c->_c.mfc_flags & MFC_STATIC)
2309 rtm->rtm_protocol = RTPROT_STATIC;
2310 else
2311 rtm->rtm_protocol = RTPROT_MROUTED;
2312 rtm->rtm_flags = 0;
2313
2314 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2315 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2316 goto nla_put_failure;
2317 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2318 /* do not break the dump if cache is unresolved */
2319 if (err < 0 && err != -ENOENT)
2320 goto nla_put_failure;
2321
2322 nlmsg_end(skb, nlh);
2323 return 0;
2324
2325nla_put_failure:
2326 nlmsg_cancel(skb, nlh);
2327 return -EMSGSIZE;
2328}
2329
2330static int _ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2331 u32 portid, u32 seq, struct mr_mfc *c,
2332 int cmd, int flags)
2333{
2334 return ip6mr_fill_mroute(mrt, skb, portid, seq, (struct mfc6_cache *)c,
2335 cmd, flags);
2336}
2337
2338static int mr6_msgsize(bool unresolved, int maxvif)
2339{
2340 size_t len =
2341 NLMSG_ALIGN(sizeof(struct rtmsg))
2342 + nla_total_size(4) /* RTA_TABLE */
2343 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2344 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2345 ;
2346
2347 if (!unresolved)
2348 len = len
2349 + nla_total_size(4) /* RTA_IIF */
2350 + nla_total_size(0) /* RTA_MULTIPATH */
2351 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2352 /* RTA_MFC_STATS */
2353 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2354 ;
2355
2356 return len;
2357}
2358
2359static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc,
2360 int cmd)
2361{
2362 struct net *net = read_pnet(&mrt->net);
2363 struct sk_buff *skb;
2364 int err = -ENOBUFS;
2365
2366 skb = nlmsg_new(mr6_msgsize(mfc->_c.mfc_parent >= MAXMIFS, mrt->maxvif),
2367 GFP_ATOMIC);
2368 if (!skb)
2369 goto errout;
2370
2371 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2372 if (err < 0)
2373 goto errout;
2374
2375 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2376 return;
2377
2378errout:
2379 kfree_skb(skb);
2380 if (err < 0)
2381 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2382}
2383
2384static size_t mrt6msg_netlink_msgsize(size_t payloadlen)
2385{
2386 size_t len =
2387 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2388 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */
2389 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */
2390 /* IP6MRA_CREPORT_SRC_ADDR */
2391 + nla_total_size(sizeof(struct in6_addr))
2392 /* IP6MRA_CREPORT_DST_ADDR */
2393 + nla_total_size(sizeof(struct in6_addr))
2394 /* IP6MRA_CREPORT_PKT */
2395 + nla_total_size(payloadlen)
2396 ;
2397
2398 return len;
2399}
2400
2401static void mrt6msg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2402{
2403 struct net *net = read_pnet(&mrt->net);
2404 struct nlmsghdr *nlh;
2405 struct rtgenmsg *rtgenm;
2406 struct mrt6msg *msg;
2407 struct sk_buff *skb;
2408 struct nlattr *nla;
2409 int payloadlen;
2410
2411 payloadlen = pkt->len - sizeof(struct mrt6msg);
2412 msg = (struct mrt6msg *)skb_transport_header(pkt);
2413
2414 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2415 if (!skb)
2416 goto errout;
2417
2418 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2419 sizeof(struct rtgenmsg), 0);
2420 if (!nlh)
2421 goto errout;
2422 rtgenm = nlmsg_data(nlh);
2423 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR;
2424 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) ||
2425 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) ||
2426 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR,
2427 &msg->im6_src) ||
2428 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR,
2429 &msg->im6_dst))
2430 goto nla_put_failure;
2431
2432 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen);
2433 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg),
2434 nla_data(nla), payloadlen))
2435 goto nla_put_failure;
2436
2437 nlmsg_end(skb, nlh);
2438
2439 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC);
2440 return;
2441
2442nla_put_failure:
2443 nlmsg_cancel(skb, nlh);
2444errout:
2445 kfree_skb(skb);
2446 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS);
2447}
2448
2449static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2450{
2451 return mr_rtm_dumproute(skb, cb, ip6mr_mr_table_iter,
2452 _ip6mr_fill_mroute, &mfc_unres_lock);
2453}
1/*
2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
4 *
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
8 * 6WIND, Paris, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 *
17 */
18
19#include <asm/uaccess.h>
20#include <linux/types.h>
21#include <linux/sched.h>
22#include <linux/errno.h>
23#include <linux/timer.h>
24#include <linux/mm.h>
25#include <linux/kernel.h>
26#include <linux/fcntl.h>
27#include <linux/stat.h>
28#include <linux/socket.h>
29#include <linux/inet.h>
30#include <linux/netdevice.h>
31#include <linux/inetdevice.h>
32#include <linux/proc_fs.h>
33#include <linux/seq_file.h>
34#include <linux/init.h>
35#include <linux/slab.h>
36#include <linux/compat.h>
37#include <net/protocol.h>
38#include <linux/skbuff.h>
39#include <net/sock.h>
40#include <net/raw.h>
41#include <linux/notifier.h>
42#include <linux/if_arp.h>
43#include <net/checksum.h>
44#include <net/netlink.h>
45#include <net/fib_rules.h>
46
47#include <net/ipv6.h>
48#include <net/ip6_route.h>
49#include <linux/mroute6.h>
50#include <linux/pim.h>
51#include <net/addrconf.h>
52#include <linux/netfilter_ipv6.h>
53#include <linux/export.h>
54#include <net/ip6_checksum.h>
55
56struct 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
76struct ip6mr_rule {
77 struct fib_rule common;
78};
79
80struct 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
88static 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 */
97static 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
107static struct kmem_cache *mrt_cachep __read_mostly;
108
109static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
110static void ip6mr_free_table(struct mr6_table *mrt);
111
112static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
113 struct sk_buff *skb, struct mfc6_cache *cache);
114static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
115 mifi_t mifi, int assert);
116static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
117 struct mfc6_cache *c, struct rtmsg *rtm);
118static int ip6mr_rtm_dumproute(struct sk_buff *skb,
119 struct netlink_callback *cb);
120static void mroute_clean_tables(struct mr6_table *mrt);
121static 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
127static 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
138static 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
153static 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
178static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
179{
180 return 1;
181}
182
183static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
184 FRA_GENERIC_POLICY,
185};
186
187static 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
193static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
194 struct nlattr **tb)
195{
196 return 1;
197}
198
199static 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
208static 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
223static 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
248err2:
249 kfree(mrt);
250err1:
251 fib_rules_unregister(ops);
252 return err;
253}
254
255static 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
269static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
270{
271 return net->ipv6.mrt6;
272}
273
274static 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
281static 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
287static void __net_exit ip6mr_rules_exit(struct net *net)
288{
289 ip6mr_free_table(net->ipv6.mrt6);
290}
291#endif
292
293static 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
326static 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
335struct 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
343static 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
373struct ipmr_vif_iter {
374 struct seq_net_private p;
375 struct mr6_table *mrt;
376 int ct;
377};
378
379static 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
394static 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
412static 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
430static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
431 __releases(mrt_lock)
432{
433 read_unlock(&mrt_lock);
434}
435
436static 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
458static 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
465static 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
471static 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
479static 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
494static 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
537static 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
548static 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
590static 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
597static 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
603static 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
614static 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
680static const struct inet6_protocol pim6_protocol = {
681 .handler = pim6_rcv,
682};
683
684/* Service routines creating virtual interfaces: PIMREG */
685
686static 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 kfree_skb(skb);
701 return err;
702 }
703
704 read_lock(&mrt_lock);
705 dev->stats.tx_bytes += skb->len;
706 dev->stats.tx_packets++;
707 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
708 read_unlock(&mrt_lock);
709 kfree_skb(skb);
710 return NETDEV_TX_OK;
711}
712
713static const struct net_device_ops reg_vif_netdev_ops = {
714 .ndo_start_xmit = reg_vif_xmit,
715};
716
717static void reg_vif_setup(struct net_device *dev)
718{
719 dev->type = ARPHRD_PIMREG;
720 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
721 dev->flags = IFF_NOARP;
722 dev->netdev_ops = ®_vif_netdev_ops;
723 dev->destructor = free_netdev;
724 dev->features |= NETIF_F_NETNS_LOCAL;
725}
726
727static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
728{
729 struct net_device *dev;
730 char name[IFNAMSIZ];
731
732 if (mrt->id == RT6_TABLE_DFLT)
733 sprintf(name, "pim6reg");
734 else
735 sprintf(name, "pim6reg%u", mrt->id);
736
737 dev = alloc_netdev(0, name, reg_vif_setup);
738 if (dev == NULL)
739 return NULL;
740
741 dev_net_set(dev, net);
742
743 if (register_netdevice(dev)) {
744 free_netdev(dev);
745 return NULL;
746 }
747 dev->iflink = 0;
748
749 if (dev_open(dev))
750 goto failure;
751
752 dev_hold(dev);
753 return dev;
754
755failure:
756 /* allow the register to be completed before unregistering. */
757 rtnl_unlock();
758 rtnl_lock();
759
760 unregister_netdevice(dev);
761 return NULL;
762}
763#endif
764
765/*
766 * Delete a VIF entry
767 */
768
769static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
770{
771 struct mif_device *v;
772 struct net_device *dev;
773 struct inet6_dev *in6_dev;
774
775 if (vifi < 0 || vifi >= mrt->maxvif)
776 return -EADDRNOTAVAIL;
777
778 v = &mrt->vif6_table[vifi];
779
780 write_lock_bh(&mrt_lock);
781 dev = v->dev;
782 v->dev = NULL;
783
784 if (!dev) {
785 write_unlock_bh(&mrt_lock);
786 return -EADDRNOTAVAIL;
787 }
788
789#ifdef CONFIG_IPV6_PIMSM_V2
790 if (vifi == mrt->mroute_reg_vif_num)
791 mrt->mroute_reg_vif_num = -1;
792#endif
793
794 if (vifi + 1 == mrt->maxvif) {
795 int tmp;
796 for (tmp = vifi - 1; tmp >= 0; tmp--) {
797 if (MIF_EXISTS(mrt, tmp))
798 break;
799 }
800 mrt->maxvif = tmp + 1;
801 }
802
803 write_unlock_bh(&mrt_lock);
804
805 dev_set_allmulti(dev, -1);
806
807 in6_dev = __in6_dev_get(dev);
808 if (in6_dev)
809 in6_dev->cnf.mc_forwarding--;
810
811 if (v->flags & MIFF_REGISTER)
812 unregister_netdevice_queue(dev, head);
813
814 dev_put(dev);
815 return 0;
816}
817
818static inline void ip6mr_cache_free(struct mfc6_cache *c)
819{
820 kmem_cache_free(mrt_cachep, c);
821}
822
823/* Destroy an unresolved cache entry, killing queued skbs
824 and reporting error to netlink readers.
825 */
826
827static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
828{
829 struct net *net = read_pnet(&mrt->net);
830 struct sk_buff *skb;
831
832 atomic_dec(&mrt->cache_resolve_queue_len);
833
834 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
835 if (ipv6_hdr(skb)->version == 0) {
836 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
837 nlh->nlmsg_type = NLMSG_ERROR;
838 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
839 skb_trim(skb, nlh->nlmsg_len);
840 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
841 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
842 } else
843 kfree_skb(skb);
844 }
845
846 ip6mr_cache_free(c);
847}
848
849
850/* Timer process for all the unresolved queue. */
851
852static void ipmr_do_expire_process(struct mr6_table *mrt)
853{
854 unsigned long now = jiffies;
855 unsigned long expires = 10 * HZ;
856 struct mfc6_cache *c, *next;
857
858 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
859 if (time_after(c->mfc_un.unres.expires, now)) {
860 /* not yet... */
861 unsigned long interval = c->mfc_un.unres.expires - now;
862 if (interval < expires)
863 expires = interval;
864 continue;
865 }
866
867 list_del(&c->list);
868 ip6mr_destroy_unres(mrt, c);
869 }
870
871 if (!list_empty(&mrt->mfc6_unres_queue))
872 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
873}
874
875static void ipmr_expire_process(unsigned long arg)
876{
877 struct mr6_table *mrt = (struct mr6_table *)arg;
878
879 if (!spin_trylock(&mfc_unres_lock)) {
880 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
881 return;
882 }
883
884 if (!list_empty(&mrt->mfc6_unres_queue))
885 ipmr_do_expire_process(mrt);
886
887 spin_unlock(&mfc_unres_lock);
888}
889
890/* Fill oifs list. It is called under write locked mrt_lock. */
891
892static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
893 unsigned char *ttls)
894{
895 int vifi;
896
897 cache->mfc_un.res.minvif = MAXMIFS;
898 cache->mfc_un.res.maxvif = 0;
899 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
900
901 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
902 if (MIF_EXISTS(mrt, vifi) &&
903 ttls[vifi] && ttls[vifi] < 255) {
904 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
905 if (cache->mfc_un.res.minvif > vifi)
906 cache->mfc_un.res.minvif = vifi;
907 if (cache->mfc_un.res.maxvif <= vifi)
908 cache->mfc_un.res.maxvif = vifi + 1;
909 }
910 }
911}
912
913static int mif6_add(struct net *net, struct mr6_table *mrt,
914 struct mif6ctl *vifc, int mrtsock)
915{
916 int vifi = vifc->mif6c_mifi;
917 struct mif_device *v = &mrt->vif6_table[vifi];
918 struct net_device *dev;
919 struct inet6_dev *in6_dev;
920 int err;
921
922 /* Is vif busy ? */
923 if (MIF_EXISTS(mrt, vifi))
924 return -EADDRINUSE;
925
926 switch (vifc->mif6c_flags) {
927#ifdef CONFIG_IPV6_PIMSM_V2
928 case MIFF_REGISTER:
929 /*
930 * Special Purpose VIF in PIM
931 * All the packets will be sent to the daemon
932 */
933 if (mrt->mroute_reg_vif_num >= 0)
934 return -EADDRINUSE;
935 dev = ip6mr_reg_vif(net, mrt);
936 if (!dev)
937 return -ENOBUFS;
938 err = dev_set_allmulti(dev, 1);
939 if (err) {
940 unregister_netdevice(dev);
941 dev_put(dev);
942 return err;
943 }
944 break;
945#endif
946 case 0:
947 dev = dev_get_by_index(net, vifc->mif6c_pifi);
948 if (!dev)
949 return -EADDRNOTAVAIL;
950 err = dev_set_allmulti(dev, 1);
951 if (err) {
952 dev_put(dev);
953 return err;
954 }
955 break;
956 default:
957 return -EINVAL;
958 }
959
960 in6_dev = __in6_dev_get(dev);
961 if (in6_dev)
962 in6_dev->cnf.mc_forwarding++;
963
964 /*
965 * Fill in the VIF structures
966 */
967 v->rate_limit = vifc->vifc_rate_limit;
968 v->flags = vifc->mif6c_flags;
969 if (!mrtsock)
970 v->flags |= VIFF_STATIC;
971 v->threshold = vifc->vifc_threshold;
972 v->bytes_in = 0;
973 v->bytes_out = 0;
974 v->pkt_in = 0;
975 v->pkt_out = 0;
976 v->link = dev->ifindex;
977 if (v->flags & MIFF_REGISTER)
978 v->link = dev->iflink;
979
980 /* And finish update writing critical data */
981 write_lock_bh(&mrt_lock);
982 v->dev = dev;
983#ifdef CONFIG_IPV6_PIMSM_V2
984 if (v->flags & MIFF_REGISTER)
985 mrt->mroute_reg_vif_num = vifi;
986#endif
987 if (vifi + 1 > mrt->maxvif)
988 mrt->maxvif = vifi + 1;
989 write_unlock_bh(&mrt_lock);
990 return 0;
991}
992
993static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
994 const struct in6_addr *origin,
995 const struct in6_addr *mcastgrp)
996{
997 int line = MFC6_HASH(mcastgrp, origin);
998 struct mfc6_cache *c;
999
1000 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1001 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
1002 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
1003 return c;
1004 }
1005 return NULL;
1006}
1007
1008/*
1009 * Allocate a multicast cache entry
1010 */
1011static struct mfc6_cache *ip6mr_cache_alloc(void)
1012{
1013 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1014 if (c == NULL)
1015 return NULL;
1016 c->mfc_un.res.minvif = MAXMIFS;
1017 return c;
1018}
1019
1020static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1021{
1022 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1023 if (c == NULL)
1024 return NULL;
1025 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1026 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1027 return c;
1028}
1029
1030/*
1031 * A cache entry has gone into a resolved state from queued
1032 */
1033
1034static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1035 struct mfc6_cache *uc, struct mfc6_cache *c)
1036{
1037 struct sk_buff *skb;
1038
1039 /*
1040 * Play the pending entries through our router
1041 */
1042
1043 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1044 if (ipv6_hdr(skb)->version == 0) {
1045 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1046
1047 if (__ip6mr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) {
1048 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1049 } else {
1050 nlh->nlmsg_type = NLMSG_ERROR;
1051 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
1052 skb_trim(skb, nlh->nlmsg_len);
1053 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
1054 }
1055 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
1056 } else
1057 ip6_mr_forward(net, mrt, skb, c);
1058 }
1059}
1060
1061/*
1062 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1063 * expects the following bizarre scheme.
1064 *
1065 * Called under mrt_lock.
1066 */
1067
1068static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1069 mifi_t mifi, int assert)
1070{
1071 struct sk_buff *skb;
1072 struct mrt6msg *msg;
1073 int ret;
1074
1075#ifdef CONFIG_IPV6_PIMSM_V2
1076 if (assert == MRT6MSG_WHOLEPKT)
1077 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1078 +sizeof(*msg));
1079 else
1080#endif
1081 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1082
1083 if (!skb)
1084 return -ENOBUFS;
1085
1086 /* I suppose that internal messages
1087 * do not require checksums */
1088
1089 skb->ip_summed = CHECKSUM_UNNECESSARY;
1090
1091#ifdef CONFIG_IPV6_PIMSM_V2
1092 if (assert == MRT6MSG_WHOLEPKT) {
1093 /* Ugly, but we have no choice with this interface.
1094 Duplicate old header, fix length etc.
1095 And all this only to mangle msg->im6_msgtype and
1096 to set msg->im6_mbz to "mbz" :-)
1097 */
1098 skb_push(skb, -skb_network_offset(pkt));
1099
1100 skb_push(skb, sizeof(*msg));
1101 skb_reset_transport_header(skb);
1102 msg = (struct mrt6msg *)skb_transport_header(skb);
1103 msg->im6_mbz = 0;
1104 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1105 msg->im6_mif = mrt->mroute_reg_vif_num;
1106 msg->im6_pad = 0;
1107 msg->im6_src = ipv6_hdr(pkt)->saddr;
1108 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1109
1110 skb->ip_summed = CHECKSUM_UNNECESSARY;
1111 } else
1112#endif
1113 {
1114 /*
1115 * Copy the IP header
1116 */
1117
1118 skb_put(skb, sizeof(struct ipv6hdr));
1119 skb_reset_network_header(skb);
1120 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1121
1122 /*
1123 * Add our header
1124 */
1125 skb_put(skb, sizeof(*msg));
1126 skb_reset_transport_header(skb);
1127 msg = (struct mrt6msg *)skb_transport_header(skb);
1128
1129 msg->im6_mbz = 0;
1130 msg->im6_msgtype = assert;
1131 msg->im6_mif = mifi;
1132 msg->im6_pad = 0;
1133 msg->im6_src = ipv6_hdr(pkt)->saddr;
1134 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1135
1136 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1137 skb->ip_summed = CHECKSUM_UNNECESSARY;
1138 }
1139
1140 if (mrt->mroute6_sk == NULL) {
1141 kfree_skb(skb);
1142 return -EINVAL;
1143 }
1144
1145 /*
1146 * Deliver to user space multicast routing algorithms
1147 */
1148 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1149 if (ret < 0) {
1150 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1151 kfree_skb(skb);
1152 }
1153
1154 return ret;
1155}
1156
1157/*
1158 * Queue a packet for resolution. It gets locked cache entry!
1159 */
1160
1161static int
1162ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1163{
1164 bool found = false;
1165 int err;
1166 struct mfc6_cache *c;
1167
1168 spin_lock_bh(&mfc_unres_lock);
1169 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1170 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1171 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1172 found = true;
1173 break;
1174 }
1175 }
1176
1177 if (!found) {
1178 /*
1179 * Create a new entry if allowable
1180 */
1181
1182 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1183 (c = ip6mr_cache_alloc_unres()) == NULL) {
1184 spin_unlock_bh(&mfc_unres_lock);
1185
1186 kfree_skb(skb);
1187 return -ENOBUFS;
1188 }
1189
1190 /*
1191 * Fill in the new cache entry
1192 */
1193 c->mf6c_parent = -1;
1194 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1195 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1196
1197 /*
1198 * Reflect first query at pim6sd
1199 */
1200 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1201 if (err < 0) {
1202 /* If the report failed throw the cache entry
1203 out - Brad Parker
1204 */
1205 spin_unlock_bh(&mfc_unres_lock);
1206
1207 ip6mr_cache_free(c);
1208 kfree_skb(skb);
1209 return err;
1210 }
1211
1212 atomic_inc(&mrt->cache_resolve_queue_len);
1213 list_add(&c->list, &mrt->mfc6_unres_queue);
1214
1215 ipmr_do_expire_process(mrt);
1216 }
1217
1218 /*
1219 * See if we can append the packet
1220 */
1221 if (c->mfc_un.unres.unresolved.qlen > 3) {
1222 kfree_skb(skb);
1223 err = -ENOBUFS;
1224 } else {
1225 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1226 err = 0;
1227 }
1228
1229 spin_unlock_bh(&mfc_unres_lock);
1230 return err;
1231}
1232
1233/*
1234 * MFC6 cache manipulation by user space
1235 */
1236
1237static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc)
1238{
1239 int line;
1240 struct mfc6_cache *c, *next;
1241
1242 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1243
1244 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1245 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1246 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1247 write_lock_bh(&mrt_lock);
1248 list_del(&c->list);
1249 write_unlock_bh(&mrt_lock);
1250
1251 ip6mr_cache_free(c);
1252 return 0;
1253 }
1254 }
1255 return -ENOENT;
1256}
1257
1258static int ip6mr_device_event(struct notifier_block *this,
1259 unsigned long event, void *ptr)
1260{
1261 struct net_device *dev = ptr;
1262 struct net *net = dev_net(dev);
1263 struct mr6_table *mrt;
1264 struct mif_device *v;
1265 int ct;
1266 LIST_HEAD(list);
1267
1268 if (event != NETDEV_UNREGISTER)
1269 return NOTIFY_DONE;
1270
1271 ip6mr_for_each_table(mrt, net) {
1272 v = &mrt->vif6_table[0];
1273 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1274 if (v->dev == dev)
1275 mif6_delete(mrt, ct, &list);
1276 }
1277 }
1278 unregister_netdevice_many(&list);
1279
1280 return NOTIFY_DONE;
1281}
1282
1283static struct notifier_block ip6_mr_notifier = {
1284 .notifier_call = ip6mr_device_event
1285};
1286
1287/*
1288 * Setup for IP multicast routing
1289 */
1290
1291static int __net_init ip6mr_net_init(struct net *net)
1292{
1293 int err;
1294
1295 err = ip6mr_rules_init(net);
1296 if (err < 0)
1297 goto fail;
1298
1299#ifdef CONFIG_PROC_FS
1300 err = -ENOMEM;
1301 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
1302 goto proc_vif_fail;
1303 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops))
1304 goto proc_cache_fail;
1305#endif
1306
1307 return 0;
1308
1309#ifdef CONFIG_PROC_FS
1310proc_cache_fail:
1311 proc_net_remove(net, "ip6_mr_vif");
1312proc_vif_fail:
1313 ip6mr_rules_exit(net);
1314#endif
1315fail:
1316 return err;
1317}
1318
1319static void __net_exit ip6mr_net_exit(struct net *net)
1320{
1321#ifdef CONFIG_PROC_FS
1322 proc_net_remove(net, "ip6_mr_cache");
1323 proc_net_remove(net, "ip6_mr_vif");
1324#endif
1325 ip6mr_rules_exit(net);
1326}
1327
1328static struct pernet_operations ip6mr_net_ops = {
1329 .init = ip6mr_net_init,
1330 .exit = ip6mr_net_exit,
1331};
1332
1333int __init ip6_mr_init(void)
1334{
1335 int err;
1336
1337 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1338 sizeof(struct mfc6_cache),
1339 0, SLAB_HWCACHE_ALIGN,
1340 NULL);
1341 if (!mrt_cachep)
1342 return -ENOMEM;
1343
1344 err = register_pernet_subsys(&ip6mr_net_ops);
1345 if (err)
1346 goto reg_pernet_fail;
1347
1348 err = register_netdevice_notifier(&ip6_mr_notifier);
1349 if (err)
1350 goto reg_notif_fail;
1351#ifdef CONFIG_IPV6_PIMSM_V2
1352 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1353 pr_err("%s: can't add PIM protocol\n", __func__);
1354 err = -EAGAIN;
1355 goto add_proto_fail;
1356 }
1357#endif
1358 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL,
1359 ip6mr_rtm_dumproute, NULL);
1360 return 0;
1361#ifdef CONFIG_IPV6_PIMSM_V2
1362add_proto_fail:
1363 unregister_netdevice_notifier(&ip6_mr_notifier);
1364#endif
1365reg_notif_fail:
1366 unregister_pernet_subsys(&ip6mr_net_ops);
1367reg_pernet_fail:
1368 kmem_cache_destroy(mrt_cachep);
1369 return err;
1370}
1371
1372void ip6_mr_cleanup(void)
1373{
1374 unregister_netdevice_notifier(&ip6_mr_notifier);
1375 unregister_pernet_subsys(&ip6mr_net_ops);
1376 kmem_cache_destroy(mrt_cachep);
1377}
1378
1379static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1380 struct mf6cctl *mfc, int mrtsock)
1381{
1382 bool found = false;
1383 int line;
1384 struct mfc6_cache *uc, *c;
1385 unsigned char ttls[MAXMIFS];
1386 int i;
1387
1388 if (mfc->mf6cc_parent >= MAXMIFS)
1389 return -ENFILE;
1390
1391 memset(ttls, 255, MAXMIFS);
1392 for (i = 0; i < MAXMIFS; i++) {
1393 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1394 ttls[i] = 1;
1395
1396 }
1397
1398 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1399
1400 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1401 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1402 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1403 found = true;
1404 break;
1405 }
1406 }
1407
1408 if (found) {
1409 write_lock_bh(&mrt_lock);
1410 c->mf6c_parent = mfc->mf6cc_parent;
1411 ip6mr_update_thresholds(mrt, c, ttls);
1412 if (!mrtsock)
1413 c->mfc_flags |= MFC_STATIC;
1414 write_unlock_bh(&mrt_lock);
1415 return 0;
1416 }
1417
1418 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1419 return -EINVAL;
1420
1421 c = ip6mr_cache_alloc();
1422 if (c == NULL)
1423 return -ENOMEM;
1424
1425 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1426 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1427 c->mf6c_parent = mfc->mf6cc_parent;
1428 ip6mr_update_thresholds(mrt, c, ttls);
1429 if (!mrtsock)
1430 c->mfc_flags |= MFC_STATIC;
1431
1432 write_lock_bh(&mrt_lock);
1433 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1434 write_unlock_bh(&mrt_lock);
1435
1436 /*
1437 * Check to see if we resolved a queued list. If so we
1438 * need to send on the frames and tidy up.
1439 */
1440 found = false;
1441 spin_lock_bh(&mfc_unres_lock);
1442 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1443 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1444 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1445 list_del(&uc->list);
1446 atomic_dec(&mrt->cache_resolve_queue_len);
1447 found = true;
1448 break;
1449 }
1450 }
1451 if (list_empty(&mrt->mfc6_unres_queue))
1452 del_timer(&mrt->ipmr_expire_timer);
1453 spin_unlock_bh(&mfc_unres_lock);
1454
1455 if (found) {
1456 ip6mr_cache_resolve(net, mrt, uc, c);
1457 ip6mr_cache_free(uc);
1458 }
1459 return 0;
1460}
1461
1462/*
1463 * Close the multicast socket, and clear the vif tables etc
1464 */
1465
1466static void mroute_clean_tables(struct mr6_table *mrt)
1467{
1468 int i;
1469 LIST_HEAD(list);
1470 struct mfc6_cache *c, *next;
1471
1472 /*
1473 * Shut down all active vif entries
1474 */
1475 for (i = 0; i < mrt->maxvif; i++) {
1476 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1477 mif6_delete(mrt, i, &list);
1478 }
1479 unregister_netdevice_many(&list);
1480
1481 /*
1482 * Wipe the cache
1483 */
1484 for (i = 0; i < MFC6_LINES; i++) {
1485 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1486 if (c->mfc_flags & MFC_STATIC)
1487 continue;
1488 write_lock_bh(&mrt_lock);
1489 list_del(&c->list);
1490 write_unlock_bh(&mrt_lock);
1491
1492 ip6mr_cache_free(c);
1493 }
1494 }
1495
1496 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1497 spin_lock_bh(&mfc_unres_lock);
1498 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1499 list_del(&c->list);
1500 ip6mr_destroy_unres(mrt, c);
1501 }
1502 spin_unlock_bh(&mfc_unres_lock);
1503 }
1504}
1505
1506static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1507{
1508 int err = 0;
1509 struct net *net = sock_net(sk);
1510
1511 rtnl_lock();
1512 write_lock_bh(&mrt_lock);
1513 if (likely(mrt->mroute6_sk == NULL)) {
1514 mrt->mroute6_sk = sk;
1515 net->ipv6.devconf_all->mc_forwarding++;
1516 }
1517 else
1518 err = -EADDRINUSE;
1519 write_unlock_bh(&mrt_lock);
1520
1521 rtnl_unlock();
1522
1523 return err;
1524}
1525
1526int ip6mr_sk_done(struct sock *sk)
1527{
1528 int err = -EACCES;
1529 struct net *net = sock_net(sk);
1530 struct mr6_table *mrt;
1531
1532 rtnl_lock();
1533 ip6mr_for_each_table(mrt, net) {
1534 if (sk == mrt->mroute6_sk) {
1535 write_lock_bh(&mrt_lock);
1536 mrt->mroute6_sk = NULL;
1537 net->ipv6.devconf_all->mc_forwarding--;
1538 write_unlock_bh(&mrt_lock);
1539
1540 mroute_clean_tables(mrt);
1541 err = 0;
1542 break;
1543 }
1544 }
1545 rtnl_unlock();
1546
1547 return err;
1548}
1549
1550struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1551{
1552 struct mr6_table *mrt;
1553 struct flowi6 fl6 = {
1554 .flowi6_iif = skb->skb_iif,
1555 .flowi6_oif = skb->dev->ifindex,
1556 .flowi6_mark = skb->mark,
1557 };
1558
1559 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1560 return NULL;
1561
1562 return mrt->mroute6_sk;
1563}
1564
1565/*
1566 * Socket options and virtual interface manipulation. The whole
1567 * virtual interface system is a complete heap, but unfortunately
1568 * that's how BSD mrouted happens to think. Maybe one day with a proper
1569 * MOSPF/PIM router set up we can clean this up.
1570 */
1571
1572int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1573{
1574 int ret;
1575 struct mif6ctl vif;
1576 struct mf6cctl mfc;
1577 mifi_t mifi;
1578 struct net *net = sock_net(sk);
1579 struct mr6_table *mrt;
1580
1581 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1582 if (mrt == NULL)
1583 return -ENOENT;
1584
1585 if (optname != MRT6_INIT) {
1586 if (sk != mrt->mroute6_sk && !capable(CAP_NET_ADMIN))
1587 return -EACCES;
1588 }
1589
1590 switch (optname) {
1591 case MRT6_INIT:
1592 if (sk->sk_type != SOCK_RAW ||
1593 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1594 return -EOPNOTSUPP;
1595 if (optlen < sizeof(int))
1596 return -EINVAL;
1597
1598 return ip6mr_sk_init(mrt, sk);
1599
1600 case MRT6_DONE:
1601 return ip6mr_sk_done(sk);
1602
1603 case MRT6_ADD_MIF:
1604 if (optlen < sizeof(vif))
1605 return -EINVAL;
1606 if (copy_from_user(&vif, optval, sizeof(vif)))
1607 return -EFAULT;
1608 if (vif.mif6c_mifi >= MAXMIFS)
1609 return -ENFILE;
1610 rtnl_lock();
1611 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1612 rtnl_unlock();
1613 return ret;
1614
1615 case MRT6_DEL_MIF:
1616 if (optlen < sizeof(mifi_t))
1617 return -EINVAL;
1618 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1619 return -EFAULT;
1620 rtnl_lock();
1621 ret = mif6_delete(mrt, mifi, NULL);
1622 rtnl_unlock();
1623 return ret;
1624
1625 /*
1626 * Manipulate the forwarding caches. These live
1627 * in a sort of kernel/user symbiosis.
1628 */
1629 case MRT6_ADD_MFC:
1630 case MRT6_DEL_MFC:
1631 if (optlen < sizeof(mfc))
1632 return -EINVAL;
1633 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1634 return -EFAULT;
1635 rtnl_lock();
1636 if (optname == MRT6_DEL_MFC)
1637 ret = ip6mr_mfc_delete(mrt, &mfc);
1638 else
1639 ret = ip6mr_mfc_add(net, mrt, &mfc, sk == mrt->mroute6_sk);
1640 rtnl_unlock();
1641 return ret;
1642
1643 /*
1644 * Control PIM assert (to activate pim will activate assert)
1645 */
1646 case MRT6_ASSERT:
1647 {
1648 int v;
1649 if (get_user(v, (int __user *)optval))
1650 return -EFAULT;
1651 mrt->mroute_do_assert = !!v;
1652 return 0;
1653 }
1654
1655#ifdef CONFIG_IPV6_PIMSM_V2
1656 case MRT6_PIM:
1657 {
1658 int v;
1659 if (get_user(v, (int __user *)optval))
1660 return -EFAULT;
1661 v = !!v;
1662 rtnl_lock();
1663 ret = 0;
1664 if (v != mrt->mroute_do_pim) {
1665 mrt->mroute_do_pim = v;
1666 mrt->mroute_do_assert = v;
1667 }
1668 rtnl_unlock();
1669 return ret;
1670 }
1671
1672#endif
1673#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1674 case MRT6_TABLE:
1675 {
1676 u32 v;
1677
1678 if (optlen != sizeof(u32))
1679 return -EINVAL;
1680 if (get_user(v, (u32 __user *)optval))
1681 return -EFAULT;
1682 if (sk == mrt->mroute6_sk)
1683 return -EBUSY;
1684
1685 rtnl_lock();
1686 ret = 0;
1687 if (!ip6mr_new_table(net, v))
1688 ret = -ENOMEM;
1689 raw6_sk(sk)->ip6mr_table = v;
1690 rtnl_unlock();
1691 return ret;
1692 }
1693#endif
1694 /*
1695 * Spurious command, or MRT6_VERSION which you cannot
1696 * set.
1697 */
1698 default:
1699 return -ENOPROTOOPT;
1700 }
1701}
1702
1703/*
1704 * Getsock opt support for the multicast routing system.
1705 */
1706
1707int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1708 int __user *optlen)
1709{
1710 int olr;
1711 int val;
1712 struct net *net = sock_net(sk);
1713 struct mr6_table *mrt;
1714
1715 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1716 if (mrt == NULL)
1717 return -ENOENT;
1718
1719 switch (optname) {
1720 case MRT6_VERSION:
1721 val = 0x0305;
1722 break;
1723#ifdef CONFIG_IPV6_PIMSM_V2
1724 case MRT6_PIM:
1725 val = mrt->mroute_do_pim;
1726 break;
1727#endif
1728 case MRT6_ASSERT:
1729 val = mrt->mroute_do_assert;
1730 break;
1731 default:
1732 return -ENOPROTOOPT;
1733 }
1734
1735 if (get_user(olr, optlen))
1736 return -EFAULT;
1737
1738 olr = min_t(int, olr, sizeof(int));
1739 if (olr < 0)
1740 return -EINVAL;
1741
1742 if (put_user(olr, optlen))
1743 return -EFAULT;
1744 if (copy_to_user(optval, &val, olr))
1745 return -EFAULT;
1746 return 0;
1747}
1748
1749/*
1750 * The IP multicast ioctl support routines.
1751 */
1752
1753int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1754{
1755 struct sioc_sg_req6 sr;
1756 struct sioc_mif_req6 vr;
1757 struct mif_device *vif;
1758 struct mfc6_cache *c;
1759 struct net *net = sock_net(sk);
1760 struct mr6_table *mrt;
1761
1762 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1763 if (mrt == NULL)
1764 return -ENOENT;
1765
1766 switch (cmd) {
1767 case SIOCGETMIFCNT_IN6:
1768 if (copy_from_user(&vr, arg, sizeof(vr)))
1769 return -EFAULT;
1770 if (vr.mifi >= mrt->maxvif)
1771 return -EINVAL;
1772 read_lock(&mrt_lock);
1773 vif = &mrt->vif6_table[vr.mifi];
1774 if (MIF_EXISTS(mrt, vr.mifi)) {
1775 vr.icount = vif->pkt_in;
1776 vr.ocount = vif->pkt_out;
1777 vr.ibytes = vif->bytes_in;
1778 vr.obytes = vif->bytes_out;
1779 read_unlock(&mrt_lock);
1780
1781 if (copy_to_user(arg, &vr, sizeof(vr)))
1782 return -EFAULT;
1783 return 0;
1784 }
1785 read_unlock(&mrt_lock);
1786 return -EADDRNOTAVAIL;
1787 case SIOCGETSGCNT_IN6:
1788 if (copy_from_user(&sr, arg, sizeof(sr)))
1789 return -EFAULT;
1790
1791 read_lock(&mrt_lock);
1792 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1793 if (c) {
1794 sr.pktcnt = c->mfc_un.res.pkt;
1795 sr.bytecnt = c->mfc_un.res.bytes;
1796 sr.wrong_if = c->mfc_un.res.wrong_if;
1797 read_unlock(&mrt_lock);
1798
1799 if (copy_to_user(arg, &sr, sizeof(sr)))
1800 return -EFAULT;
1801 return 0;
1802 }
1803 read_unlock(&mrt_lock);
1804 return -EADDRNOTAVAIL;
1805 default:
1806 return -ENOIOCTLCMD;
1807 }
1808}
1809
1810#ifdef CONFIG_COMPAT
1811struct compat_sioc_sg_req6 {
1812 struct sockaddr_in6 src;
1813 struct sockaddr_in6 grp;
1814 compat_ulong_t pktcnt;
1815 compat_ulong_t bytecnt;
1816 compat_ulong_t wrong_if;
1817};
1818
1819struct compat_sioc_mif_req6 {
1820 mifi_t mifi;
1821 compat_ulong_t icount;
1822 compat_ulong_t ocount;
1823 compat_ulong_t ibytes;
1824 compat_ulong_t obytes;
1825};
1826
1827int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1828{
1829 struct compat_sioc_sg_req6 sr;
1830 struct compat_sioc_mif_req6 vr;
1831 struct mif_device *vif;
1832 struct mfc6_cache *c;
1833 struct net *net = sock_net(sk);
1834 struct mr6_table *mrt;
1835
1836 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1837 if (mrt == NULL)
1838 return -ENOENT;
1839
1840 switch (cmd) {
1841 case SIOCGETMIFCNT_IN6:
1842 if (copy_from_user(&vr, arg, sizeof(vr)))
1843 return -EFAULT;
1844 if (vr.mifi >= mrt->maxvif)
1845 return -EINVAL;
1846 read_lock(&mrt_lock);
1847 vif = &mrt->vif6_table[vr.mifi];
1848 if (MIF_EXISTS(mrt, vr.mifi)) {
1849 vr.icount = vif->pkt_in;
1850 vr.ocount = vif->pkt_out;
1851 vr.ibytes = vif->bytes_in;
1852 vr.obytes = vif->bytes_out;
1853 read_unlock(&mrt_lock);
1854
1855 if (copy_to_user(arg, &vr, sizeof(vr)))
1856 return -EFAULT;
1857 return 0;
1858 }
1859 read_unlock(&mrt_lock);
1860 return -EADDRNOTAVAIL;
1861 case SIOCGETSGCNT_IN6:
1862 if (copy_from_user(&sr, arg, sizeof(sr)))
1863 return -EFAULT;
1864
1865 read_lock(&mrt_lock);
1866 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1867 if (c) {
1868 sr.pktcnt = c->mfc_un.res.pkt;
1869 sr.bytecnt = c->mfc_un.res.bytes;
1870 sr.wrong_if = c->mfc_un.res.wrong_if;
1871 read_unlock(&mrt_lock);
1872
1873 if (copy_to_user(arg, &sr, sizeof(sr)))
1874 return -EFAULT;
1875 return 0;
1876 }
1877 read_unlock(&mrt_lock);
1878 return -EADDRNOTAVAIL;
1879 default:
1880 return -ENOIOCTLCMD;
1881 }
1882}
1883#endif
1884
1885static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1886{
1887 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1888 IPSTATS_MIB_OUTFORWDATAGRAMS);
1889 IP6_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1890 IPSTATS_MIB_OUTOCTETS, skb->len);
1891 return dst_output(skb);
1892}
1893
1894/*
1895 * Processing handlers for ip6mr_forward
1896 */
1897
1898static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1899 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1900{
1901 struct ipv6hdr *ipv6h;
1902 struct mif_device *vif = &mrt->vif6_table[vifi];
1903 struct net_device *dev;
1904 struct dst_entry *dst;
1905 struct flowi6 fl6;
1906
1907 if (vif->dev == NULL)
1908 goto out_free;
1909
1910#ifdef CONFIG_IPV6_PIMSM_V2
1911 if (vif->flags & MIFF_REGISTER) {
1912 vif->pkt_out++;
1913 vif->bytes_out += skb->len;
1914 vif->dev->stats.tx_bytes += skb->len;
1915 vif->dev->stats.tx_packets++;
1916 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1917 goto out_free;
1918 }
1919#endif
1920
1921 ipv6h = ipv6_hdr(skb);
1922
1923 fl6 = (struct flowi6) {
1924 .flowi6_oif = vif->link,
1925 .daddr = ipv6h->daddr,
1926 };
1927
1928 dst = ip6_route_output(net, NULL, &fl6);
1929 if (dst->error) {
1930 dst_release(dst);
1931 goto out_free;
1932 }
1933
1934 skb_dst_drop(skb);
1935 skb_dst_set(skb, dst);
1936
1937 /*
1938 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1939 * not only before forwarding, but after forwarding on all output
1940 * interfaces. It is clear, if mrouter runs a multicasting
1941 * program, it should receive packets not depending to what interface
1942 * program is joined.
1943 * If we will not make it, the program will have to join on all
1944 * interfaces. On the other hand, multihoming host (or router, but
1945 * not mrouter) cannot join to more than one interface - it will
1946 * result in receiving multiple packets.
1947 */
1948 dev = vif->dev;
1949 skb->dev = dev;
1950 vif->pkt_out++;
1951 vif->bytes_out += skb->len;
1952
1953 /* We are about to write */
1954 /* XXX: extension headers? */
1955 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1956 goto out_free;
1957
1958 ipv6h = ipv6_hdr(skb);
1959 ipv6h->hop_limit--;
1960
1961 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1962
1963 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
1964 ip6mr_forward2_finish);
1965
1966out_free:
1967 kfree_skb(skb);
1968 return 0;
1969}
1970
1971static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
1972{
1973 int ct;
1974
1975 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
1976 if (mrt->vif6_table[ct].dev == dev)
1977 break;
1978 }
1979 return ct;
1980}
1981
1982static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
1983 struct sk_buff *skb, struct mfc6_cache *cache)
1984{
1985 int psend = -1;
1986 int vif, ct;
1987
1988 vif = cache->mf6c_parent;
1989 cache->mfc_un.res.pkt++;
1990 cache->mfc_un.res.bytes += skb->len;
1991
1992 /*
1993 * Wrong interface: drop packet and (maybe) send PIM assert.
1994 */
1995 if (mrt->vif6_table[vif].dev != skb->dev) {
1996 int true_vifi;
1997
1998 cache->mfc_un.res.wrong_if++;
1999 true_vifi = ip6mr_find_vif(mrt, skb->dev);
2000
2001 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2002 /* pimsm uses asserts, when switching from RPT to SPT,
2003 so that we cannot check that packet arrived on an oif.
2004 It is bad, but otherwise we would need to move pretty
2005 large chunk of pimd to kernel. Ough... --ANK
2006 */
2007 (mrt->mroute_do_pim ||
2008 cache->mfc_un.res.ttls[true_vifi] < 255) &&
2009 time_after(jiffies,
2010 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2011 cache->mfc_un.res.last_assert = jiffies;
2012 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2013 }
2014 goto dont_forward;
2015 }
2016
2017 mrt->vif6_table[vif].pkt_in++;
2018 mrt->vif6_table[vif].bytes_in += skb->len;
2019
2020 /*
2021 * Forward the frame
2022 */
2023 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
2024 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
2025 if (psend != -1) {
2026 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2027 if (skb2)
2028 ip6mr_forward2(net, mrt, skb2, cache, psend);
2029 }
2030 psend = ct;
2031 }
2032 }
2033 if (psend != -1) {
2034 ip6mr_forward2(net, mrt, skb, cache, psend);
2035 return 0;
2036 }
2037
2038dont_forward:
2039 kfree_skb(skb);
2040 return 0;
2041}
2042
2043
2044/*
2045 * Multicast packets for forwarding arrive here
2046 */
2047
2048int ip6_mr_input(struct sk_buff *skb)
2049{
2050 struct mfc6_cache *cache;
2051 struct net *net = dev_net(skb->dev);
2052 struct mr6_table *mrt;
2053 struct flowi6 fl6 = {
2054 .flowi6_iif = skb->dev->ifindex,
2055 .flowi6_mark = skb->mark,
2056 };
2057 int err;
2058
2059 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2060 if (err < 0) {
2061 kfree_skb(skb);
2062 return err;
2063 }
2064
2065 read_lock(&mrt_lock);
2066 cache = ip6mr_cache_find(mrt,
2067 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2068
2069 /*
2070 * No usable cache entry
2071 */
2072 if (cache == NULL) {
2073 int vif;
2074
2075 vif = ip6mr_find_vif(mrt, skb->dev);
2076 if (vif >= 0) {
2077 int err = ip6mr_cache_unresolved(mrt, vif, skb);
2078 read_unlock(&mrt_lock);
2079
2080 return err;
2081 }
2082 read_unlock(&mrt_lock);
2083 kfree_skb(skb);
2084 return -ENODEV;
2085 }
2086
2087 ip6_mr_forward(net, mrt, skb, cache);
2088
2089 read_unlock(&mrt_lock);
2090
2091 return 0;
2092}
2093
2094
2095static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2096 struct mfc6_cache *c, struct rtmsg *rtm)
2097{
2098 int ct;
2099 struct rtnexthop *nhp;
2100 u8 *b = skb_tail_pointer(skb);
2101 struct rtattr *mp_head;
2102
2103 /* If cache is unresolved, don't try to parse IIF and OIF */
2104 if (c->mf6c_parent >= MAXMIFS)
2105 return -ENOENT;
2106
2107 if (MIF_EXISTS(mrt, c->mf6c_parent))
2108 RTA_PUT(skb, RTA_IIF, 4, &mrt->vif6_table[c->mf6c_parent].dev->ifindex);
2109
2110 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
2111
2112 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2113 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2114 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
2115 goto rtattr_failure;
2116 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
2117 nhp->rtnh_flags = 0;
2118 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2119 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2120 nhp->rtnh_len = sizeof(*nhp);
2121 }
2122 }
2123 mp_head->rta_type = RTA_MULTIPATH;
2124 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
2125 rtm->rtm_type = RTN_MULTICAST;
2126 return 1;
2127
2128rtattr_failure:
2129 nlmsg_trim(skb, b);
2130 return -EMSGSIZE;
2131}
2132
2133int ip6mr_get_route(struct net *net,
2134 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2135{
2136 int err;
2137 struct mr6_table *mrt;
2138 struct mfc6_cache *cache;
2139 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2140
2141 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2142 if (mrt == NULL)
2143 return -ENOENT;
2144
2145 read_lock(&mrt_lock);
2146 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2147
2148 if (!cache) {
2149 struct sk_buff *skb2;
2150 struct ipv6hdr *iph;
2151 struct net_device *dev;
2152 int vif;
2153
2154 if (nowait) {
2155 read_unlock(&mrt_lock);
2156 return -EAGAIN;
2157 }
2158
2159 dev = skb->dev;
2160 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2161 read_unlock(&mrt_lock);
2162 return -ENODEV;
2163 }
2164
2165 /* really correct? */
2166 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2167 if (!skb2) {
2168 read_unlock(&mrt_lock);
2169 return -ENOMEM;
2170 }
2171
2172 skb_reset_transport_header(skb2);
2173
2174 skb_put(skb2, sizeof(struct ipv6hdr));
2175 skb_reset_network_header(skb2);
2176
2177 iph = ipv6_hdr(skb2);
2178 iph->version = 0;
2179 iph->priority = 0;
2180 iph->flow_lbl[0] = 0;
2181 iph->flow_lbl[1] = 0;
2182 iph->flow_lbl[2] = 0;
2183 iph->payload_len = 0;
2184 iph->nexthdr = IPPROTO_NONE;
2185 iph->hop_limit = 0;
2186 iph->saddr = rt->rt6i_src.addr;
2187 iph->daddr = rt->rt6i_dst.addr;
2188
2189 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2190 read_unlock(&mrt_lock);
2191
2192 return err;
2193 }
2194
2195 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2196 cache->mfc_flags |= MFC_NOTIFY;
2197
2198 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2199 read_unlock(&mrt_lock);
2200 return err;
2201}
2202
2203static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2204 u32 pid, u32 seq, struct mfc6_cache *c)
2205{
2206 struct nlmsghdr *nlh;
2207 struct rtmsg *rtm;
2208
2209 nlh = nlmsg_put(skb, pid, seq, RTM_NEWROUTE, sizeof(*rtm), NLM_F_MULTI);
2210 if (nlh == NULL)
2211 return -EMSGSIZE;
2212
2213 rtm = nlmsg_data(nlh);
2214 rtm->rtm_family = RTNL_FAMILY_IPMR;
2215 rtm->rtm_dst_len = 128;
2216 rtm->rtm_src_len = 128;
2217 rtm->rtm_tos = 0;
2218 rtm->rtm_table = mrt->id;
2219 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2220 goto nla_put_failure;
2221 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2222 rtm->rtm_protocol = RTPROT_UNSPEC;
2223 rtm->rtm_flags = 0;
2224
2225 if (nla_put(skb, RTA_SRC, 16, &c->mf6c_origin) ||
2226 nla_put(skb, RTA_DST, 16, &c->mf6c_mcastgrp))
2227 goto nla_put_failure;
2228 if (__ip6mr_fill_mroute(mrt, skb, c, rtm) < 0)
2229 goto nla_put_failure;
2230
2231 return nlmsg_end(skb, nlh);
2232
2233nla_put_failure:
2234 nlmsg_cancel(skb, nlh);
2235 return -EMSGSIZE;
2236}
2237
2238static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2239{
2240 struct net *net = sock_net(skb->sk);
2241 struct mr6_table *mrt;
2242 struct mfc6_cache *mfc;
2243 unsigned int t = 0, s_t;
2244 unsigned int h = 0, s_h;
2245 unsigned int e = 0, s_e;
2246
2247 s_t = cb->args[0];
2248 s_h = cb->args[1];
2249 s_e = cb->args[2];
2250
2251 read_lock(&mrt_lock);
2252 ip6mr_for_each_table(mrt, net) {
2253 if (t < s_t)
2254 goto next_table;
2255 if (t > s_t)
2256 s_h = 0;
2257 for (h = s_h; h < MFC6_LINES; h++) {
2258 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2259 if (e < s_e)
2260 goto next_entry;
2261 if (ip6mr_fill_mroute(mrt, skb,
2262 NETLINK_CB(cb->skb).pid,
2263 cb->nlh->nlmsg_seq,
2264 mfc) < 0)
2265 goto done;
2266next_entry:
2267 e++;
2268 }
2269 e = s_e = 0;
2270 }
2271 s_h = 0;
2272next_table:
2273 t++;
2274 }
2275done:
2276 read_unlock(&mrt_lock);
2277
2278 cb->args[2] = e;
2279 cb->args[1] = h;
2280 cb->args[0] = t;
2281
2282 return skb->len;
2283}