]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/ipv6/ip6mr.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
s390/early: improve machine detection
[mirror_ubuntu-jammy-kernel.git] / net / ipv6 / ip6mr.c
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 <linux/rhashtable.h>
36 #include <net/protocol.h>
37 #include <linux/skbuff.h>
38 #include <net/raw.h>
39 #include <linux/notifier.h>
40 #include <linux/if_arp.h>
41 #include <net/checksum.h>
42 #include <net/netlink.h>
43 #include <net/fib_rules.h>
44
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <linux/mroute6.h>
48 #include <linux/pim.h>
49 #include <net/addrconf.h>
50 #include <linux/netfilter_ipv6.h>
51 #include <linux/export.h>
52 #include <net/ip6_checksum.h>
53 #include <linux/netconf.h>
54 #include <net/ip_tunnels.h>
55
56 #include <linux/nospec.h>
57
58 struct ip6mr_rule {
59 struct fib_rule common;
60 };
61
62 struct ip6mr_result {
63 struct mr_table *mrt;
64 };
65
66 /* Big lock, protecting vif table, mrt cache and mroute socket state.
67 Note that the changes are semaphored via rtnl_lock.
68 */
69
70 static DEFINE_RWLOCK(mrt_lock);
71
72 /* Multicast router control variables */
73
74 /* Special spinlock for queue of unresolved entries */
75 static DEFINE_SPINLOCK(mfc_unres_lock);
76
77 /* We return to original Alan's scheme. Hash table of resolved
78 entries is changed only in process context and protected
79 with weak lock mrt_lock. Queue of unresolved entries is protected
80 with strong spinlock mfc_unres_lock.
81
82 In this case data path is free of exclusive locks at all.
83 */
84
85 static struct kmem_cache *mrt_cachep __read_mostly;
86
87 static struct mr_table *ip6mr_new_table(struct net *net, u32 id);
88 static void ip6mr_free_table(struct mr_table *mrt);
89
90 static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
91 struct net_device *dev, struct sk_buff *skb,
92 struct mfc6_cache *cache);
93 static int ip6mr_cache_report(struct mr_table *mrt, struct sk_buff *pkt,
94 mifi_t mifi, int assert);
95 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc,
96 int cmd);
97 static void mrt6msg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
98 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
99 struct netlink_callback *cb);
100 static void mroute_clean_tables(struct mr_table *mrt, bool all);
101 static void ipmr_expire_process(struct timer_list *t);
102
103 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
104 #define ip6mr_for_each_table(mrt, net) \
105 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
106
107 static struct mr_table *ip6mr_mr_table_iter(struct net *net,
108 struct mr_table *mrt)
109 {
110 struct mr_table *ret;
111
112 if (!mrt)
113 ret = list_entry_rcu(net->ipv6.mr6_tables.next,
114 struct mr_table, list);
115 else
116 ret = list_entry_rcu(mrt->list.next,
117 struct mr_table, list);
118
119 if (&ret->list == &net->ipv6.mr6_tables)
120 return NULL;
121 return ret;
122 }
123
124 static struct mr_table *ip6mr_get_table(struct net *net, u32 id)
125 {
126 struct mr_table *mrt;
127
128 ip6mr_for_each_table(mrt, net) {
129 if (mrt->id == id)
130 return mrt;
131 }
132 return NULL;
133 }
134
135 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
136 struct mr_table **mrt)
137 {
138 int err;
139 struct ip6mr_result res;
140 struct fib_lookup_arg arg = {
141 .result = &res,
142 .flags = FIB_LOOKUP_NOREF,
143 };
144
145 /* update flow if oif or iif point to device enslaved to l3mdev */
146 l3mdev_update_flow(net, flowi6_to_flowi(flp6));
147
148 err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
149 flowi6_to_flowi(flp6), 0, &arg);
150 if (err < 0)
151 return err;
152 *mrt = res.mrt;
153 return 0;
154 }
155
156 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
157 int flags, struct fib_lookup_arg *arg)
158 {
159 struct ip6mr_result *res = arg->result;
160 struct mr_table *mrt;
161
162 switch (rule->action) {
163 case FR_ACT_TO_TBL:
164 break;
165 case FR_ACT_UNREACHABLE:
166 return -ENETUNREACH;
167 case FR_ACT_PROHIBIT:
168 return -EACCES;
169 case FR_ACT_BLACKHOLE:
170 default:
171 return -EINVAL;
172 }
173
174 arg->table = fib_rule_get_table(rule, arg);
175
176 mrt = ip6mr_get_table(rule->fr_net, arg->table);
177 if (!mrt)
178 return -EAGAIN;
179 res->mrt = mrt;
180 return 0;
181 }
182
183 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
184 {
185 return 1;
186 }
187
188 static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
189 FRA_GENERIC_POLICY,
190 };
191
192 static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
193 struct fib_rule_hdr *frh, struct nlattr **tb,
194 struct netlink_ext_ack *extack)
195 {
196 return 0;
197 }
198
199 static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
200 struct nlattr **tb)
201 {
202 return 1;
203 }
204
205 static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
206 struct fib_rule_hdr *frh)
207 {
208 frh->dst_len = 0;
209 frh->src_len = 0;
210 frh->tos = 0;
211 return 0;
212 }
213
214 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template = {
215 .family = RTNL_FAMILY_IP6MR,
216 .rule_size = sizeof(struct ip6mr_rule),
217 .addr_size = sizeof(struct in6_addr),
218 .action = ip6mr_rule_action,
219 .match = ip6mr_rule_match,
220 .configure = ip6mr_rule_configure,
221 .compare = ip6mr_rule_compare,
222 .fill = ip6mr_rule_fill,
223 .nlgroup = RTNLGRP_IPV6_RULE,
224 .policy = ip6mr_rule_policy,
225 .owner = THIS_MODULE,
226 };
227
228 static int __net_init ip6mr_rules_init(struct net *net)
229 {
230 struct fib_rules_ops *ops;
231 struct mr_table *mrt;
232 int err;
233
234 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
235 if (IS_ERR(ops))
236 return PTR_ERR(ops);
237
238 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
239
240 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
241 if (IS_ERR(mrt)) {
242 err = PTR_ERR(mrt);
243 goto err1;
244 }
245
246 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
247 if (err < 0)
248 goto err2;
249
250 net->ipv6.mr6_rules_ops = ops;
251 return 0;
252
253 err2:
254 ip6mr_free_table(mrt);
255 err1:
256 fib_rules_unregister(ops);
257 return err;
258 }
259
260 static void __net_exit ip6mr_rules_exit(struct net *net)
261 {
262 struct mr_table *mrt, *next;
263
264 rtnl_lock();
265 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
266 list_del(&mrt->list);
267 ip6mr_free_table(mrt);
268 }
269 fib_rules_unregister(net->ipv6.mr6_rules_ops);
270 rtnl_unlock();
271 }
272
273 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb)
274 {
275 return fib_rules_dump(net, nb, RTNL_FAMILY_IP6MR);
276 }
277
278 static unsigned int ip6mr_rules_seq_read(struct net *net)
279 {
280 return fib_rules_seq_read(net, RTNL_FAMILY_IP6MR);
281 }
282
283 bool ip6mr_rule_default(const struct fib_rule *rule)
284 {
285 return fib_rule_matchall(rule) && rule->action == FR_ACT_TO_TBL &&
286 rule->table == RT6_TABLE_DFLT && !rule->l3mdev;
287 }
288 EXPORT_SYMBOL(ip6mr_rule_default);
289 #else
290 #define ip6mr_for_each_table(mrt, net) \
291 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
292
293 static struct mr_table *ip6mr_mr_table_iter(struct net *net,
294 struct mr_table *mrt)
295 {
296 if (!mrt)
297 return net->ipv6.mrt6;
298 return NULL;
299 }
300
301 static struct mr_table *ip6mr_get_table(struct net *net, u32 id)
302 {
303 return net->ipv6.mrt6;
304 }
305
306 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
307 struct mr_table **mrt)
308 {
309 *mrt = net->ipv6.mrt6;
310 return 0;
311 }
312
313 static int __net_init ip6mr_rules_init(struct net *net)
314 {
315 struct mr_table *mrt;
316
317 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
318 if (IS_ERR(mrt))
319 return PTR_ERR(mrt);
320 net->ipv6.mrt6 = mrt;
321 return 0;
322 }
323
324 static void __net_exit ip6mr_rules_exit(struct net *net)
325 {
326 rtnl_lock();
327 ip6mr_free_table(net->ipv6.mrt6);
328 net->ipv6.mrt6 = NULL;
329 rtnl_unlock();
330 }
331
332 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb)
333 {
334 return 0;
335 }
336
337 static unsigned int ip6mr_rules_seq_read(struct net *net)
338 {
339 return 0;
340 }
341 #endif
342
343 static int ip6mr_hash_cmp(struct rhashtable_compare_arg *arg,
344 const void *ptr)
345 {
346 const struct mfc6_cache_cmp_arg *cmparg = arg->key;
347 struct mfc6_cache *c = (struct mfc6_cache *)ptr;
348
349 return !ipv6_addr_equal(&c->mf6c_mcastgrp, &cmparg->mf6c_mcastgrp) ||
350 !ipv6_addr_equal(&c->mf6c_origin, &cmparg->mf6c_origin);
351 }
352
353 static const struct rhashtable_params ip6mr_rht_params = {
354 .head_offset = offsetof(struct mr_mfc, mnode),
355 .key_offset = offsetof(struct mfc6_cache, cmparg),
356 .key_len = sizeof(struct mfc6_cache_cmp_arg),
357 .nelem_hint = 3,
358 .locks_mul = 1,
359 .obj_cmpfn = ip6mr_hash_cmp,
360 .automatic_shrinking = true,
361 };
362
363 static void ip6mr_new_table_set(struct mr_table *mrt,
364 struct net *net)
365 {
366 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
367 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
368 #endif
369 }
370
371 static struct mfc6_cache_cmp_arg ip6mr_mr_table_ops_cmparg_any = {
372 .mf6c_origin = IN6ADDR_ANY_INIT,
373 .mf6c_mcastgrp = IN6ADDR_ANY_INIT,
374 };
375
376 static struct mr_table_ops ip6mr_mr_table_ops = {
377 .rht_params = &ip6mr_rht_params,
378 .cmparg_any = &ip6mr_mr_table_ops_cmparg_any,
379 };
380
381 static struct mr_table *ip6mr_new_table(struct net *net, u32 id)
382 {
383 struct mr_table *mrt;
384
385 mrt = ip6mr_get_table(net, id);
386 if (mrt)
387 return mrt;
388
389 return mr_table_alloc(net, id, &ip6mr_mr_table_ops,
390 ipmr_expire_process, ip6mr_new_table_set);
391 }
392
393 static void ip6mr_free_table(struct mr_table *mrt)
394 {
395 del_timer_sync(&mrt->ipmr_expire_timer);
396 mroute_clean_tables(mrt, true);
397 rhltable_destroy(&mrt->mfc_hash);
398 kfree(mrt);
399 }
400
401 #ifdef CONFIG_PROC_FS
402 /* The /proc interfaces to multicast routing
403 * /proc/ip6_mr_cache /proc/ip6_mr_vif
404 */
405
406 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
407 __acquires(mrt_lock)
408 {
409 struct mr_vif_iter *iter = seq->private;
410 struct net *net = seq_file_net(seq);
411 struct mr_table *mrt;
412
413 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
414 if (!mrt)
415 return ERR_PTR(-ENOENT);
416
417 iter->mrt = mrt;
418
419 read_lock(&mrt_lock);
420 return mr_vif_seq_start(seq, pos);
421 }
422
423 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
424 __releases(mrt_lock)
425 {
426 read_unlock(&mrt_lock);
427 }
428
429 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
430 {
431 struct mr_vif_iter *iter = seq->private;
432 struct mr_table *mrt = iter->mrt;
433
434 if (v == SEQ_START_TOKEN) {
435 seq_puts(seq,
436 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
437 } else {
438 const struct vif_device *vif = v;
439 const char *name = vif->dev ? vif->dev->name : "none";
440
441 seq_printf(seq,
442 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
443 vif - mrt->vif_table,
444 name, vif->bytes_in, vif->pkt_in,
445 vif->bytes_out, vif->pkt_out,
446 vif->flags);
447 }
448 return 0;
449 }
450
451 static const struct seq_operations ip6mr_vif_seq_ops = {
452 .start = ip6mr_vif_seq_start,
453 .next = mr_vif_seq_next,
454 .stop = ip6mr_vif_seq_stop,
455 .show = ip6mr_vif_seq_show,
456 };
457
458 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
459 {
460 struct net *net = seq_file_net(seq);
461 struct mr_table *mrt;
462
463 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
464 if (!mrt)
465 return ERR_PTR(-ENOENT);
466
467 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
468 }
469
470 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
471 {
472 int n;
473
474 if (v == SEQ_START_TOKEN) {
475 seq_puts(seq,
476 "Group "
477 "Origin "
478 "Iif Pkts Bytes Wrong Oifs\n");
479 } else {
480 const struct mfc6_cache *mfc = v;
481 const struct mr_mfc_iter *it = seq->private;
482 struct mr_table *mrt = it->mrt;
483
484 seq_printf(seq, "%pI6 %pI6 %-3hd",
485 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
486 mfc->_c.mfc_parent);
487
488 if (it->cache != &mrt->mfc_unres_queue) {
489 seq_printf(seq, " %8lu %8lu %8lu",
490 mfc->_c.mfc_un.res.pkt,
491 mfc->_c.mfc_un.res.bytes,
492 mfc->_c.mfc_un.res.wrong_if);
493 for (n = mfc->_c.mfc_un.res.minvif;
494 n < mfc->_c.mfc_un.res.maxvif; n++) {
495 if (VIF_EXISTS(mrt, n) &&
496 mfc->_c.mfc_un.res.ttls[n] < 255)
497 seq_printf(seq,
498 " %2d:%-3d", n,
499 mfc->_c.mfc_un.res.ttls[n]);
500 }
501 } else {
502 /* unresolved mfc_caches don't contain
503 * pkt, bytes and wrong_if values
504 */
505 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
506 }
507 seq_putc(seq, '\n');
508 }
509 return 0;
510 }
511
512 static const struct seq_operations ipmr_mfc_seq_ops = {
513 .start = ipmr_mfc_seq_start,
514 .next = mr_mfc_seq_next,
515 .stop = mr_mfc_seq_stop,
516 .show = ipmr_mfc_seq_show,
517 };
518 #endif
519
520 #ifdef CONFIG_IPV6_PIMSM_V2
521
522 static int pim6_rcv(struct sk_buff *skb)
523 {
524 struct pimreghdr *pim;
525 struct ipv6hdr *encap;
526 struct net_device *reg_dev = NULL;
527 struct net *net = dev_net(skb->dev);
528 struct mr_table *mrt;
529 struct flowi6 fl6 = {
530 .flowi6_iif = skb->dev->ifindex,
531 .flowi6_mark = skb->mark,
532 };
533 int reg_vif_num;
534
535 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
536 goto drop;
537
538 pim = (struct pimreghdr *)skb_transport_header(skb);
539 if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) ||
540 (pim->flags & PIM_NULL_REGISTER) ||
541 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
542 sizeof(*pim), IPPROTO_PIM,
543 csum_partial((void *)pim, sizeof(*pim), 0)) &&
544 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
545 goto drop;
546
547 /* check if the inner packet is destined to mcast group */
548 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
549 sizeof(*pim));
550
551 if (!ipv6_addr_is_multicast(&encap->daddr) ||
552 encap->payload_len == 0 ||
553 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
554 goto drop;
555
556 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
557 goto drop;
558 reg_vif_num = mrt->mroute_reg_vif_num;
559
560 read_lock(&mrt_lock);
561 if (reg_vif_num >= 0)
562 reg_dev = mrt->vif_table[reg_vif_num].dev;
563 if (reg_dev)
564 dev_hold(reg_dev);
565 read_unlock(&mrt_lock);
566
567 if (!reg_dev)
568 goto drop;
569
570 skb->mac_header = skb->network_header;
571 skb_pull(skb, (u8 *)encap - skb->data);
572 skb_reset_network_header(skb);
573 skb->protocol = htons(ETH_P_IPV6);
574 skb->ip_summed = CHECKSUM_NONE;
575
576 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
577
578 netif_rx(skb);
579
580 dev_put(reg_dev);
581 return 0;
582 drop:
583 kfree_skb(skb);
584 return 0;
585 }
586
587 static const struct inet6_protocol pim6_protocol = {
588 .handler = pim6_rcv,
589 };
590
591 /* Service routines creating virtual interfaces: PIMREG */
592
593 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
594 struct net_device *dev)
595 {
596 struct net *net = dev_net(dev);
597 struct mr_table *mrt;
598 struct flowi6 fl6 = {
599 .flowi6_oif = dev->ifindex,
600 .flowi6_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
601 .flowi6_mark = skb->mark,
602 };
603
604 if (!pskb_inet_may_pull(skb))
605 goto tx_err;
606
607 if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
608 goto tx_err;
609
610 read_lock(&mrt_lock);
611 dev->stats.tx_bytes += skb->len;
612 dev->stats.tx_packets++;
613 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
614 read_unlock(&mrt_lock);
615 kfree_skb(skb);
616 return NETDEV_TX_OK;
617
618 tx_err:
619 dev->stats.tx_errors++;
620 kfree_skb(skb);
621 return NETDEV_TX_OK;
622 }
623
624 static int reg_vif_get_iflink(const struct net_device *dev)
625 {
626 return 0;
627 }
628
629 static const struct net_device_ops reg_vif_netdev_ops = {
630 .ndo_start_xmit = reg_vif_xmit,
631 .ndo_get_iflink = reg_vif_get_iflink,
632 };
633
634 static void reg_vif_setup(struct net_device *dev)
635 {
636 dev->type = ARPHRD_PIMREG;
637 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
638 dev->flags = IFF_NOARP;
639 dev->netdev_ops = &reg_vif_netdev_ops;
640 dev->needs_free_netdev = true;
641 dev->features |= NETIF_F_NETNS_LOCAL;
642 }
643
644 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr_table *mrt)
645 {
646 struct net_device *dev;
647 char name[IFNAMSIZ];
648
649 if (mrt->id == RT6_TABLE_DFLT)
650 sprintf(name, "pim6reg");
651 else
652 sprintf(name, "pim6reg%u", mrt->id);
653
654 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
655 if (!dev)
656 return NULL;
657
658 dev_net_set(dev, net);
659
660 if (register_netdevice(dev)) {
661 free_netdev(dev);
662 return NULL;
663 }
664
665 if (dev_open(dev, NULL))
666 goto failure;
667
668 dev_hold(dev);
669 return dev;
670
671 failure:
672 unregister_netdevice(dev);
673 return NULL;
674 }
675 #endif
676
677 static int call_ip6mr_vif_entry_notifiers(struct net *net,
678 enum fib_event_type event_type,
679 struct vif_device *vif,
680 mifi_t vif_index, u32 tb_id)
681 {
682 return mr_call_vif_notifiers(net, RTNL_FAMILY_IP6MR, event_type,
683 vif, vif_index, tb_id,
684 &net->ipv6.ipmr_seq);
685 }
686
687 static int call_ip6mr_mfc_entry_notifiers(struct net *net,
688 enum fib_event_type event_type,
689 struct mfc6_cache *mfc, u32 tb_id)
690 {
691 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IP6MR, event_type,
692 &mfc->_c, tb_id, &net->ipv6.ipmr_seq);
693 }
694
695 /* Delete a VIF entry */
696 static int mif6_delete(struct mr_table *mrt, int vifi, int notify,
697 struct list_head *head)
698 {
699 struct vif_device *v;
700 struct net_device *dev;
701 struct inet6_dev *in6_dev;
702
703 if (vifi < 0 || vifi >= mrt->maxvif)
704 return -EADDRNOTAVAIL;
705
706 v = &mrt->vif_table[vifi];
707
708 if (VIF_EXISTS(mrt, vifi))
709 call_ip6mr_vif_entry_notifiers(read_pnet(&mrt->net),
710 FIB_EVENT_VIF_DEL, v, vifi,
711 mrt->id);
712
713 write_lock_bh(&mrt_lock);
714 dev = v->dev;
715 v->dev = NULL;
716
717 if (!dev) {
718 write_unlock_bh(&mrt_lock);
719 return -EADDRNOTAVAIL;
720 }
721
722 #ifdef CONFIG_IPV6_PIMSM_V2
723 if (vifi == mrt->mroute_reg_vif_num)
724 mrt->mroute_reg_vif_num = -1;
725 #endif
726
727 if (vifi + 1 == mrt->maxvif) {
728 int tmp;
729 for (tmp = vifi - 1; tmp >= 0; tmp--) {
730 if (VIF_EXISTS(mrt, tmp))
731 break;
732 }
733 mrt->maxvif = tmp + 1;
734 }
735
736 write_unlock_bh(&mrt_lock);
737
738 dev_set_allmulti(dev, -1);
739
740 in6_dev = __in6_dev_get(dev);
741 if (in6_dev) {
742 in6_dev->cnf.mc_forwarding--;
743 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
744 NETCONFA_MC_FORWARDING,
745 dev->ifindex, &in6_dev->cnf);
746 }
747
748 if ((v->flags & MIFF_REGISTER) && !notify)
749 unregister_netdevice_queue(dev, head);
750
751 dev_put(dev);
752 return 0;
753 }
754
755 static inline void ip6mr_cache_free_rcu(struct rcu_head *head)
756 {
757 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
758
759 kmem_cache_free(mrt_cachep, (struct mfc6_cache *)c);
760 }
761
762 static inline void ip6mr_cache_free(struct mfc6_cache *c)
763 {
764 call_rcu(&c->_c.rcu, ip6mr_cache_free_rcu);
765 }
766
767 /* Destroy an unresolved cache entry, killing queued skbs
768 and reporting error to netlink readers.
769 */
770
771 static void ip6mr_destroy_unres(struct mr_table *mrt, struct mfc6_cache *c)
772 {
773 struct net *net = read_pnet(&mrt->net);
774 struct sk_buff *skb;
775
776 atomic_dec(&mrt->cache_resolve_queue_len);
777
778 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved)) != NULL) {
779 if (ipv6_hdr(skb)->version == 0) {
780 struct nlmsghdr *nlh = skb_pull(skb,
781 sizeof(struct ipv6hdr));
782 nlh->nlmsg_type = NLMSG_ERROR;
783 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
784 skb_trim(skb, nlh->nlmsg_len);
785 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
786 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
787 } else
788 kfree_skb(skb);
789 }
790
791 ip6mr_cache_free(c);
792 }
793
794
795 /* Timer process for all the unresolved queue. */
796
797 static void ipmr_do_expire_process(struct mr_table *mrt)
798 {
799 unsigned long now = jiffies;
800 unsigned long expires = 10 * HZ;
801 struct mr_mfc *c, *next;
802
803 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
804 if (time_after(c->mfc_un.unres.expires, now)) {
805 /* not yet... */
806 unsigned long interval = c->mfc_un.unres.expires - now;
807 if (interval < expires)
808 expires = interval;
809 continue;
810 }
811
812 list_del(&c->list);
813 mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE);
814 ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c);
815 }
816
817 if (!list_empty(&mrt->mfc_unres_queue))
818 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
819 }
820
821 static void ipmr_expire_process(struct timer_list *t)
822 {
823 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
824
825 if (!spin_trylock(&mfc_unres_lock)) {
826 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
827 return;
828 }
829
830 if (!list_empty(&mrt->mfc_unres_queue))
831 ipmr_do_expire_process(mrt);
832
833 spin_unlock(&mfc_unres_lock);
834 }
835
836 /* Fill oifs list. It is called under write locked mrt_lock. */
837
838 static void ip6mr_update_thresholds(struct mr_table *mrt,
839 struct mr_mfc *cache,
840 unsigned char *ttls)
841 {
842 int vifi;
843
844 cache->mfc_un.res.minvif = MAXMIFS;
845 cache->mfc_un.res.maxvif = 0;
846 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
847
848 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
849 if (VIF_EXISTS(mrt, vifi) &&
850 ttls[vifi] && ttls[vifi] < 255) {
851 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
852 if (cache->mfc_un.res.minvif > vifi)
853 cache->mfc_un.res.minvif = vifi;
854 if (cache->mfc_un.res.maxvif <= vifi)
855 cache->mfc_un.res.maxvif = vifi + 1;
856 }
857 }
858 cache->mfc_un.res.lastuse = jiffies;
859 }
860
861 static int mif6_add(struct net *net, struct mr_table *mrt,
862 struct mif6ctl *vifc, int mrtsock)
863 {
864 int vifi = vifc->mif6c_mifi;
865 struct vif_device *v = &mrt->vif_table[vifi];
866 struct net_device *dev;
867 struct inet6_dev *in6_dev;
868 int err;
869
870 /* Is vif busy ? */
871 if (VIF_EXISTS(mrt, vifi))
872 return -EADDRINUSE;
873
874 switch (vifc->mif6c_flags) {
875 #ifdef CONFIG_IPV6_PIMSM_V2
876 case MIFF_REGISTER:
877 /*
878 * Special Purpose VIF in PIM
879 * All the packets will be sent to the daemon
880 */
881 if (mrt->mroute_reg_vif_num >= 0)
882 return -EADDRINUSE;
883 dev = ip6mr_reg_vif(net, mrt);
884 if (!dev)
885 return -ENOBUFS;
886 err = dev_set_allmulti(dev, 1);
887 if (err) {
888 unregister_netdevice(dev);
889 dev_put(dev);
890 return err;
891 }
892 break;
893 #endif
894 case 0:
895 dev = dev_get_by_index(net, vifc->mif6c_pifi);
896 if (!dev)
897 return -EADDRNOTAVAIL;
898 err = dev_set_allmulti(dev, 1);
899 if (err) {
900 dev_put(dev);
901 return err;
902 }
903 break;
904 default:
905 return -EINVAL;
906 }
907
908 in6_dev = __in6_dev_get(dev);
909 if (in6_dev) {
910 in6_dev->cnf.mc_forwarding++;
911 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
912 NETCONFA_MC_FORWARDING,
913 dev->ifindex, &in6_dev->cnf);
914 }
915
916 /* Fill in the VIF structures */
917 vif_device_init(v, dev, vifc->vifc_rate_limit, vifc->vifc_threshold,
918 vifc->mif6c_flags | (!mrtsock ? VIFF_STATIC : 0),
919 MIFF_REGISTER);
920
921 /* And finish update writing critical data */
922 write_lock_bh(&mrt_lock);
923 v->dev = dev;
924 #ifdef CONFIG_IPV6_PIMSM_V2
925 if (v->flags & MIFF_REGISTER)
926 mrt->mroute_reg_vif_num = vifi;
927 #endif
928 if (vifi + 1 > mrt->maxvif)
929 mrt->maxvif = vifi + 1;
930 write_unlock_bh(&mrt_lock);
931 call_ip6mr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD,
932 v, vifi, mrt->id);
933 return 0;
934 }
935
936 static struct mfc6_cache *ip6mr_cache_find(struct mr_table *mrt,
937 const struct in6_addr *origin,
938 const struct in6_addr *mcastgrp)
939 {
940 struct mfc6_cache_cmp_arg arg = {
941 .mf6c_origin = *origin,
942 .mf6c_mcastgrp = *mcastgrp,
943 };
944
945 return mr_mfc_find(mrt, &arg);
946 }
947
948 /* Look for a (*,G) entry */
949 static struct mfc6_cache *ip6mr_cache_find_any(struct mr_table *mrt,
950 struct in6_addr *mcastgrp,
951 mifi_t mifi)
952 {
953 struct mfc6_cache_cmp_arg arg = {
954 .mf6c_origin = in6addr_any,
955 .mf6c_mcastgrp = *mcastgrp,
956 };
957
958 if (ipv6_addr_any(mcastgrp))
959 return mr_mfc_find_any_parent(mrt, mifi);
960 return mr_mfc_find_any(mrt, mifi, &arg);
961 }
962
963 /* Look for a (S,G,iif) entry if parent != -1 */
964 static struct mfc6_cache *
965 ip6mr_cache_find_parent(struct mr_table *mrt,
966 const struct in6_addr *origin,
967 const struct in6_addr *mcastgrp,
968 int parent)
969 {
970 struct mfc6_cache_cmp_arg arg = {
971 .mf6c_origin = *origin,
972 .mf6c_mcastgrp = *mcastgrp,
973 };
974
975 return mr_mfc_find_parent(mrt, &arg, parent);
976 }
977
978 /* Allocate a multicast cache entry */
979 static struct mfc6_cache *ip6mr_cache_alloc(void)
980 {
981 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
982 if (!c)
983 return NULL;
984 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
985 c->_c.mfc_un.res.minvif = MAXMIFS;
986 c->_c.free = ip6mr_cache_free_rcu;
987 refcount_set(&c->_c.mfc_un.res.refcount, 1);
988 return c;
989 }
990
991 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
992 {
993 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
994 if (!c)
995 return NULL;
996 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
997 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
998 return c;
999 }
1000
1001 /*
1002 * A cache entry has gone into a resolved state from queued
1003 */
1004
1005 static void ip6mr_cache_resolve(struct net *net, struct mr_table *mrt,
1006 struct mfc6_cache *uc, struct mfc6_cache *c)
1007 {
1008 struct sk_buff *skb;
1009
1010 /*
1011 * Play the pending entries through our router
1012 */
1013
1014 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1015 if (ipv6_hdr(skb)->version == 0) {
1016 struct nlmsghdr *nlh = skb_pull(skb,
1017 sizeof(struct ipv6hdr));
1018
1019 if (mr_fill_mroute(mrt, skb, &c->_c,
1020 nlmsg_data(nlh)) > 0) {
1021 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1022 } else {
1023 nlh->nlmsg_type = NLMSG_ERROR;
1024 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1025 skb_trim(skb, nlh->nlmsg_len);
1026 ((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1027 }
1028 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1029 } else
1030 ip6_mr_forward(net, mrt, skb->dev, skb, c);
1031 }
1032 }
1033
1034 /*
1035 * Bounce a cache query up to pim6sd and netlink.
1036 *
1037 * Called under mrt_lock.
1038 */
1039
1040 static int ip6mr_cache_report(struct mr_table *mrt, struct sk_buff *pkt,
1041 mifi_t mifi, int assert)
1042 {
1043 struct sock *mroute6_sk;
1044 struct sk_buff *skb;
1045 struct mrt6msg *msg;
1046 int ret;
1047
1048 #ifdef CONFIG_IPV6_PIMSM_V2
1049 if (assert == MRT6MSG_WHOLEPKT)
1050 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1051 +sizeof(*msg));
1052 else
1053 #endif
1054 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1055
1056 if (!skb)
1057 return -ENOBUFS;
1058
1059 /* I suppose that internal messages
1060 * do not require checksums */
1061
1062 skb->ip_summed = CHECKSUM_UNNECESSARY;
1063
1064 #ifdef CONFIG_IPV6_PIMSM_V2
1065 if (assert == MRT6MSG_WHOLEPKT) {
1066 /* Ugly, but we have no choice with this interface.
1067 Duplicate old header, fix length etc.
1068 And all this only to mangle msg->im6_msgtype and
1069 to set msg->im6_mbz to "mbz" :-)
1070 */
1071 skb_push(skb, -skb_network_offset(pkt));
1072
1073 skb_push(skb, sizeof(*msg));
1074 skb_reset_transport_header(skb);
1075 msg = (struct mrt6msg *)skb_transport_header(skb);
1076 msg->im6_mbz = 0;
1077 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1078 msg->im6_mif = mrt->mroute_reg_vif_num;
1079 msg->im6_pad = 0;
1080 msg->im6_src = ipv6_hdr(pkt)->saddr;
1081 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1082
1083 skb->ip_summed = CHECKSUM_UNNECESSARY;
1084 } else
1085 #endif
1086 {
1087 /*
1088 * Copy the IP header
1089 */
1090
1091 skb_put(skb, sizeof(struct ipv6hdr));
1092 skb_reset_network_header(skb);
1093 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1094
1095 /*
1096 * Add our header
1097 */
1098 skb_put(skb, sizeof(*msg));
1099 skb_reset_transport_header(skb);
1100 msg = (struct mrt6msg *)skb_transport_header(skb);
1101
1102 msg->im6_mbz = 0;
1103 msg->im6_msgtype = assert;
1104 msg->im6_mif = mifi;
1105 msg->im6_pad = 0;
1106 msg->im6_src = ipv6_hdr(pkt)->saddr;
1107 msg->im6_dst = ipv6_hdr(pkt)->daddr;
1108
1109 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1110 skb->ip_summed = CHECKSUM_UNNECESSARY;
1111 }
1112
1113 rcu_read_lock();
1114 mroute6_sk = rcu_dereference(mrt->mroute_sk);
1115 if (!mroute6_sk) {
1116 rcu_read_unlock();
1117 kfree_skb(skb);
1118 return -EINVAL;
1119 }
1120
1121 mrt6msg_netlink_event(mrt, skb);
1122
1123 /* Deliver to user space multicast routing algorithms */
1124 ret = sock_queue_rcv_skb(mroute6_sk, skb);
1125 rcu_read_unlock();
1126 if (ret < 0) {
1127 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1128 kfree_skb(skb);
1129 }
1130
1131 return ret;
1132 }
1133
1134 /* Queue a packet for resolution. It gets locked cache entry! */
1135 static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi,
1136 struct sk_buff *skb, struct net_device *dev)
1137 {
1138 struct mfc6_cache *c;
1139 bool found = false;
1140 int err;
1141
1142 spin_lock_bh(&mfc_unres_lock);
1143 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1144 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1145 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1146 found = true;
1147 break;
1148 }
1149 }
1150
1151 if (!found) {
1152 /*
1153 * Create a new entry if allowable
1154 */
1155
1156 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1157 (c = ip6mr_cache_alloc_unres()) == NULL) {
1158 spin_unlock_bh(&mfc_unres_lock);
1159
1160 kfree_skb(skb);
1161 return -ENOBUFS;
1162 }
1163
1164 /* Fill in the new cache entry */
1165 c->_c.mfc_parent = -1;
1166 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1167 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1168
1169 /*
1170 * Reflect first query at pim6sd
1171 */
1172 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1173 if (err < 0) {
1174 /* If the report failed throw the cache entry
1175 out - Brad Parker
1176 */
1177 spin_unlock_bh(&mfc_unres_lock);
1178
1179 ip6mr_cache_free(c);
1180 kfree_skb(skb);
1181 return err;
1182 }
1183
1184 atomic_inc(&mrt->cache_resolve_queue_len);
1185 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1186 mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1187
1188 ipmr_do_expire_process(mrt);
1189 }
1190
1191 /* See if we can append the packet */
1192 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1193 kfree_skb(skb);
1194 err = -ENOBUFS;
1195 } else {
1196 if (dev) {
1197 skb->dev = dev;
1198 skb->skb_iif = dev->ifindex;
1199 }
1200 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1201 err = 0;
1202 }
1203
1204 spin_unlock_bh(&mfc_unres_lock);
1205 return err;
1206 }
1207
1208 /*
1209 * MFC6 cache manipulation by user space
1210 */
1211
1212 static int ip6mr_mfc_delete(struct mr_table *mrt, struct mf6cctl *mfc,
1213 int parent)
1214 {
1215 struct mfc6_cache *c;
1216
1217 /* The entries are added/deleted only under RTNL */
1218 rcu_read_lock();
1219 c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr,
1220 &mfc->mf6cc_mcastgrp.sin6_addr, parent);
1221 rcu_read_unlock();
1222 if (!c)
1223 return -ENOENT;
1224 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ip6mr_rht_params);
1225 list_del_rcu(&c->_c.list);
1226
1227 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net),
1228 FIB_EVENT_ENTRY_DEL, c, mrt->id);
1229 mr6_netlink_event(mrt, c, RTM_DELROUTE);
1230 mr_cache_put(&c->_c);
1231 return 0;
1232 }
1233
1234 static int ip6mr_device_event(struct notifier_block *this,
1235 unsigned long event, void *ptr)
1236 {
1237 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1238 struct net *net = dev_net(dev);
1239 struct mr_table *mrt;
1240 struct vif_device *v;
1241 int ct;
1242
1243 if (event != NETDEV_UNREGISTER)
1244 return NOTIFY_DONE;
1245
1246 ip6mr_for_each_table(mrt, net) {
1247 v = &mrt->vif_table[0];
1248 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1249 if (v->dev == dev)
1250 mif6_delete(mrt, ct, 1, NULL);
1251 }
1252 }
1253
1254 return NOTIFY_DONE;
1255 }
1256
1257 static unsigned int ip6mr_seq_read(struct net *net)
1258 {
1259 ASSERT_RTNL();
1260
1261 return net->ipv6.ipmr_seq + ip6mr_rules_seq_read(net);
1262 }
1263
1264 static int ip6mr_dump(struct net *net, struct notifier_block *nb)
1265 {
1266 return mr_dump(net, nb, RTNL_FAMILY_IP6MR, ip6mr_rules_dump,
1267 ip6mr_mr_table_iter, &mrt_lock);
1268 }
1269
1270 static struct notifier_block ip6_mr_notifier = {
1271 .notifier_call = ip6mr_device_event
1272 };
1273
1274 static const struct fib_notifier_ops ip6mr_notifier_ops_template = {
1275 .family = RTNL_FAMILY_IP6MR,
1276 .fib_seq_read = ip6mr_seq_read,
1277 .fib_dump = ip6mr_dump,
1278 .owner = THIS_MODULE,
1279 };
1280
1281 static int __net_init ip6mr_notifier_init(struct net *net)
1282 {
1283 struct fib_notifier_ops *ops;
1284
1285 net->ipv6.ipmr_seq = 0;
1286
1287 ops = fib_notifier_ops_register(&ip6mr_notifier_ops_template, net);
1288 if (IS_ERR(ops))
1289 return PTR_ERR(ops);
1290
1291 net->ipv6.ip6mr_notifier_ops = ops;
1292
1293 return 0;
1294 }
1295
1296 static void __net_exit ip6mr_notifier_exit(struct net *net)
1297 {
1298 fib_notifier_ops_unregister(net->ipv6.ip6mr_notifier_ops);
1299 net->ipv6.ip6mr_notifier_ops = NULL;
1300 }
1301
1302 /* Setup for IP multicast routing */
1303 static int __net_init ip6mr_net_init(struct net *net)
1304 {
1305 int err;
1306
1307 err = ip6mr_notifier_init(net);
1308 if (err)
1309 return err;
1310
1311 err = ip6mr_rules_init(net);
1312 if (err < 0)
1313 goto ip6mr_rules_fail;
1314
1315 #ifdef CONFIG_PROC_FS
1316 err = -ENOMEM;
1317 if (!proc_create_net("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_seq_ops,
1318 sizeof(struct mr_vif_iter)))
1319 goto proc_vif_fail;
1320 if (!proc_create_net("ip6_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
1321 sizeof(struct mr_mfc_iter)))
1322 goto proc_cache_fail;
1323 #endif
1324
1325 return 0;
1326
1327 #ifdef CONFIG_PROC_FS
1328 proc_cache_fail:
1329 remove_proc_entry("ip6_mr_vif", net->proc_net);
1330 proc_vif_fail:
1331 ip6mr_rules_exit(net);
1332 #endif
1333 ip6mr_rules_fail:
1334 ip6mr_notifier_exit(net);
1335 return err;
1336 }
1337
1338 static 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
1348 static struct pernet_operations ip6mr_net_ops = {
1349 .init = ip6mr_net_init,
1350 .exit = ip6mr_net_exit,
1351 };
1352
1353 int __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);
1385 add_proto_fail:
1386 unregister_netdevice_notifier(&ip6_mr_notifier);
1387 #endif
1388 reg_notif_fail:
1389 unregister_pernet_subsys(&ip6mr_net_ops);
1390 reg_pernet_fail:
1391 kmem_cache_destroy(mrt_cachep);
1392 return err;
1393 }
1394
1395 void 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
1406 static 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
1499 static 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
1539 static 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
1565 int 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
1601 bool 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 }
1615 EXPORT_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
1624 int 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 mrt = ip6mr_new_table(net, v);
1761 if (IS_ERR(mrt))
1762 ret = PTR_ERR(mrt);
1763 else
1764 raw6_sk(sk)->ip6mr_table = v;
1765 rtnl_unlock();
1766 return ret;
1767 }
1768 #endif
1769 /*
1770 * Spurious command, or MRT6_VERSION which you cannot
1771 * set.
1772 */
1773 default:
1774 return -ENOPROTOOPT;
1775 }
1776 }
1777
1778 /*
1779 * Getsock opt support for the multicast routing system.
1780 */
1781
1782 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1783 int __user *optlen)
1784 {
1785 int olr;
1786 int val;
1787 struct net *net = sock_net(sk);
1788 struct mr_table *mrt;
1789
1790 if (sk->sk_type != SOCK_RAW ||
1791 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1792 return -EOPNOTSUPP;
1793
1794 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1795 if (!mrt)
1796 return -ENOENT;
1797
1798 switch (optname) {
1799 case MRT6_VERSION:
1800 val = 0x0305;
1801 break;
1802 #ifdef CONFIG_IPV6_PIMSM_V2
1803 case MRT6_PIM:
1804 val = mrt->mroute_do_pim;
1805 break;
1806 #endif
1807 case MRT6_ASSERT:
1808 val = mrt->mroute_do_assert;
1809 break;
1810 default:
1811 return -ENOPROTOOPT;
1812 }
1813
1814 if (get_user(olr, optlen))
1815 return -EFAULT;
1816
1817 olr = min_t(int, olr, sizeof(int));
1818 if (olr < 0)
1819 return -EINVAL;
1820
1821 if (put_user(olr, optlen))
1822 return -EFAULT;
1823 if (copy_to_user(optval, &val, olr))
1824 return -EFAULT;
1825 return 0;
1826 }
1827
1828 /*
1829 * The IP multicast ioctl support routines.
1830 */
1831
1832 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1833 {
1834 struct sioc_sg_req6 sr;
1835 struct sioc_mif_req6 vr;
1836 struct vif_device *vif;
1837 struct mfc6_cache *c;
1838 struct net *net = sock_net(sk);
1839 struct mr_table *mrt;
1840
1841 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1842 if (!mrt)
1843 return -ENOENT;
1844
1845 switch (cmd) {
1846 case SIOCGETMIFCNT_IN6:
1847 if (copy_from_user(&vr, arg, sizeof(vr)))
1848 return -EFAULT;
1849 if (vr.mifi >= mrt->maxvif)
1850 return -EINVAL;
1851 vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif);
1852 read_lock(&mrt_lock);
1853 vif = &mrt->vif_table[vr.mifi];
1854 if (VIF_EXISTS(mrt, vr.mifi)) {
1855 vr.icount = vif->pkt_in;
1856 vr.ocount = vif->pkt_out;
1857 vr.ibytes = vif->bytes_in;
1858 vr.obytes = vif->bytes_out;
1859 read_unlock(&mrt_lock);
1860
1861 if (copy_to_user(arg, &vr, sizeof(vr)))
1862 return -EFAULT;
1863 return 0;
1864 }
1865 read_unlock(&mrt_lock);
1866 return -EADDRNOTAVAIL;
1867 case SIOCGETSGCNT_IN6:
1868 if (copy_from_user(&sr, arg, sizeof(sr)))
1869 return -EFAULT;
1870
1871 rcu_read_lock();
1872 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1873 if (c) {
1874 sr.pktcnt = c->_c.mfc_un.res.pkt;
1875 sr.bytecnt = c->_c.mfc_un.res.bytes;
1876 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1877 rcu_read_unlock();
1878
1879 if (copy_to_user(arg, &sr, sizeof(sr)))
1880 return -EFAULT;
1881 return 0;
1882 }
1883 rcu_read_unlock();
1884 return -EADDRNOTAVAIL;
1885 default:
1886 return -ENOIOCTLCMD;
1887 }
1888 }
1889
1890 #ifdef CONFIG_COMPAT
1891 struct compat_sioc_sg_req6 {
1892 struct sockaddr_in6 src;
1893 struct sockaddr_in6 grp;
1894 compat_ulong_t pktcnt;
1895 compat_ulong_t bytecnt;
1896 compat_ulong_t wrong_if;
1897 };
1898
1899 struct compat_sioc_mif_req6 {
1900 mifi_t mifi;
1901 compat_ulong_t icount;
1902 compat_ulong_t ocount;
1903 compat_ulong_t ibytes;
1904 compat_ulong_t obytes;
1905 };
1906
1907 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1908 {
1909 struct compat_sioc_sg_req6 sr;
1910 struct compat_sioc_mif_req6 vr;
1911 struct vif_device *vif;
1912 struct mfc6_cache *c;
1913 struct net *net = sock_net(sk);
1914 struct mr_table *mrt;
1915
1916 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1917 if (!mrt)
1918 return -ENOENT;
1919
1920 switch (cmd) {
1921 case SIOCGETMIFCNT_IN6:
1922 if (copy_from_user(&vr, arg, sizeof(vr)))
1923 return -EFAULT;
1924 if (vr.mifi >= mrt->maxvif)
1925 return -EINVAL;
1926 vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif);
1927 read_lock(&mrt_lock);
1928 vif = &mrt->vif_table[vr.mifi];
1929 if (VIF_EXISTS(mrt, vr.mifi)) {
1930 vr.icount = vif->pkt_in;
1931 vr.ocount = vif->pkt_out;
1932 vr.ibytes = vif->bytes_in;
1933 vr.obytes = vif->bytes_out;
1934 read_unlock(&mrt_lock);
1935
1936 if (copy_to_user(arg, &vr, sizeof(vr)))
1937 return -EFAULT;
1938 return 0;
1939 }
1940 read_unlock(&mrt_lock);
1941 return -EADDRNOTAVAIL;
1942 case SIOCGETSGCNT_IN6:
1943 if (copy_from_user(&sr, arg, sizeof(sr)))
1944 return -EFAULT;
1945
1946 rcu_read_lock();
1947 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1948 if (c) {
1949 sr.pktcnt = c->_c.mfc_un.res.pkt;
1950 sr.bytecnt = c->_c.mfc_un.res.bytes;
1951 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1952 rcu_read_unlock();
1953
1954 if (copy_to_user(arg, &sr, sizeof(sr)))
1955 return -EFAULT;
1956 return 0;
1957 }
1958 rcu_read_unlock();
1959 return -EADDRNOTAVAIL;
1960 default:
1961 return -ENOIOCTLCMD;
1962 }
1963 }
1964 #endif
1965
1966 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1967 {
1968 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
1969 IPSTATS_MIB_OUTFORWDATAGRAMS);
1970 __IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)),
1971 IPSTATS_MIB_OUTOCTETS, skb->len);
1972 return dst_output(net, sk, skb);
1973 }
1974
1975 /*
1976 * Processing handlers for ip6mr_forward
1977 */
1978
1979 static int ip6mr_forward2(struct net *net, struct mr_table *mrt,
1980 struct sk_buff *skb, int vifi)
1981 {
1982 struct ipv6hdr *ipv6h;
1983 struct vif_device *vif = &mrt->vif_table[vifi];
1984 struct net_device *dev;
1985 struct dst_entry *dst;
1986 struct flowi6 fl6;
1987
1988 if (!vif->dev)
1989 goto out_free;
1990
1991 #ifdef CONFIG_IPV6_PIMSM_V2
1992 if (vif->flags & MIFF_REGISTER) {
1993 vif->pkt_out++;
1994 vif->bytes_out += skb->len;
1995 vif->dev->stats.tx_bytes += skb->len;
1996 vif->dev->stats.tx_packets++;
1997 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1998 goto out_free;
1999 }
2000 #endif
2001
2002 ipv6h = ipv6_hdr(skb);
2003
2004 fl6 = (struct flowi6) {
2005 .flowi6_oif = vif->link,
2006 .daddr = ipv6h->daddr,
2007 };
2008
2009 dst = ip6_route_output(net, NULL, &fl6);
2010 if (dst->error) {
2011 dst_release(dst);
2012 goto out_free;
2013 }
2014
2015 skb_dst_drop(skb);
2016 skb_dst_set(skb, dst);
2017
2018 /*
2019 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2020 * not only before forwarding, but after forwarding on all output
2021 * interfaces. It is clear, if mrouter runs a multicasting
2022 * program, it should receive packets not depending to what interface
2023 * program is joined.
2024 * If we will not make it, the program will have to join on all
2025 * interfaces. On the other hand, multihoming host (or router, but
2026 * not mrouter) cannot join to more than one interface - it will
2027 * result in receiving multiple packets.
2028 */
2029 dev = vif->dev;
2030 skb->dev = dev;
2031 vif->pkt_out++;
2032 vif->bytes_out += skb->len;
2033
2034 /* We are about to write */
2035 /* XXX: extension headers? */
2036 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
2037 goto out_free;
2038
2039 ipv6h = ipv6_hdr(skb);
2040 ipv6h->hop_limit--;
2041
2042 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2043
2044 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2045 net, NULL, skb, skb->dev, dev,
2046 ip6mr_forward2_finish);
2047
2048 out_free:
2049 kfree_skb(skb);
2050 return 0;
2051 }
2052
2053 static int ip6mr_find_vif(struct mr_table *mrt, struct net_device *dev)
2054 {
2055 int ct;
2056
2057 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
2058 if (mrt->vif_table[ct].dev == dev)
2059 break;
2060 }
2061 return ct;
2062 }
2063
2064 static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
2065 struct net_device *dev, struct sk_buff *skb,
2066 struct mfc6_cache *c)
2067 {
2068 int psend = -1;
2069 int vif, ct;
2070 int true_vifi = ip6mr_find_vif(mrt, dev);
2071
2072 vif = c->_c.mfc_parent;
2073 c->_c.mfc_un.res.pkt++;
2074 c->_c.mfc_un.res.bytes += skb->len;
2075 c->_c.mfc_un.res.lastuse = jiffies;
2076
2077 if (ipv6_addr_any(&c->mf6c_origin) && true_vifi >= 0) {
2078 struct mfc6_cache *cache_proxy;
2079
2080 /* For an (*,G) entry, we only check that the incoming
2081 * interface is part of the static tree.
2082 */
2083 rcu_read_lock();
2084 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2085 if (cache_proxy &&
2086 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255) {
2087 rcu_read_unlock();
2088 goto forward;
2089 }
2090 rcu_read_unlock();
2091 }
2092
2093 /*
2094 * Wrong interface: drop packet and (maybe) send PIM assert.
2095 */
2096 if (mrt->vif_table[vif].dev != dev) {
2097 c->_c.mfc_un.res.wrong_if++;
2098
2099 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2100 /* pimsm uses asserts, when switching from RPT to SPT,
2101 so that we cannot check that packet arrived on an oif.
2102 It is bad, but otherwise we would need to move pretty
2103 large chunk of pimd to kernel. Ough... --ANK
2104 */
2105 (mrt->mroute_do_pim ||
2106 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2107 time_after(jiffies,
2108 c->_c.mfc_un.res.last_assert +
2109 MFC_ASSERT_THRESH)) {
2110 c->_c.mfc_un.res.last_assert = jiffies;
2111 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2112 }
2113 goto dont_forward;
2114 }
2115
2116 forward:
2117 mrt->vif_table[vif].pkt_in++;
2118 mrt->vif_table[vif].bytes_in += skb->len;
2119
2120 /*
2121 * Forward the frame
2122 */
2123 if (ipv6_addr_any(&c->mf6c_origin) &&
2124 ipv6_addr_any(&c->mf6c_mcastgrp)) {
2125 if (true_vifi >= 0 &&
2126 true_vifi != c->_c.mfc_parent &&
2127 ipv6_hdr(skb)->hop_limit >
2128 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2129 /* It's an (*,*) entry and the packet is not coming from
2130 * the upstream: forward the packet to the upstream
2131 * only.
2132 */
2133 psend = c->_c.mfc_parent;
2134 goto last_forward;
2135 }
2136 goto dont_forward;
2137 }
2138 for (ct = c->_c.mfc_un.res.maxvif - 1;
2139 ct >= c->_c.mfc_un.res.minvif; ct--) {
2140 /* For (*,G) entry, don't forward to the incoming interface */
2141 if ((!ipv6_addr_any(&c->mf6c_origin) || ct != true_vifi) &&
2142 ipv6_hdr(skb)->hop_limit > c->_c.mfc_un.res.ttls[ct]) {
2143 if (psend != -1) {
2144 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2145 if (skb2)
2146 ip6mr_forward2(net, mrt, skb2, psend);
2147 }
2148 psend = ct;
2149 }
2150 }
2151 last_forward:
2152 if (psend != -1) {
2153 ip6mr_forward2(net, mrt, skb, psend);
2154 return;
2155 }
2156
2157 dont_forward:
2158 kfree_skb(skb);
2159 }
2160
2161
2162 /*
2163 * Multicast packets for forwarding arrive here
2164 */
2165
2166 int ip6_mr_input(struct sk_buff *skb)
2167 {
2168 struct mfc6_cache *cache;
2169 struct net *net = dev_net(skb->dev);
2170 struct mr_table *mrt;
2171 struct flowi6 fl6 = {
2172 .flowi6_iif = skb->dev->ifindex,
2173 .flowi6_mark = skb->mark,
2174 };
2175 int err;
2176 struct net_device *dev;
2177
2178 /* skb->dev passed in is the master dev for vrfs.
2179 * Get the proper interface that does have a vif associated with it.
2180 */
2181 dev = skb->dev;
2182 if (netif_is_l3_master(skb->dev)) {
2183 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2184 if (!dev) {
2185 kfree_skb(skb);
2186 return -ENODEV;
2187 }
2188 }
2189
2190 err = ip6mr_fib_lookup(net, &fl6, &mrt);
2191 if (err < 0) {
2192 kfree_skb(skb);
2193 return err;
2194 }
2195
2196 read_lock(&mrt_lock);
2197 cache = ip6mr_cache_find(mrt,
2198 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2199 if (!cache) {
2200 int vif = ip6mr_find_vif(mrt, dev);
2201
2202 if (vif >= 0)
2203 cache = ip6mr_cache_find_any(mrt,
2204 &ipv6_hdr(skb)->daddr,
2205 vif);
2206 }
2207
2208 /*
2209 * No usable cache entry
2210 */
2211 if (!cache) {
2212 int vif;
2213
2214 vif = ip6mr_find_vif(mrt, dev);
2215 if (vif >= 0) {
2216 int err = ip6mr_cache_unresolved(mrt, vif, skb, dev);
2217 read_unlock(&mrt_lock);
2218
2219 return err;
2220 }
2221 read_unlock(&mrt_lock);
2222 kfree_skb(skb);
2223 return -ENODEV;
2224 }
2225
2226 ip6_mr_forward(net, mrt, dev, skb, cache);
2227
2228 read_unlock(&mrt_lock);
2229
2230 return 0;
2231 }
2232
2233 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2234 u32 portid)
2235 {
2236 int err;
2237 struct mr_table *mrt;
2238 struct mfc6_cache *cache;
2239 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2240
2241 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2242 if (!mrt)
2243 return -ENOENT;
2244
2245 read_lock(&mrt_lock);
2246 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2247 if (!cache && skb->dev) {
2248 int vif = ip6mr_find_vif(mrt, skb->dev);
2249
2250 if (vif >= 0)
2251 cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2252 vif);
2253 }
2254
2255 if (!cache) {
2256 struct sk_buff *skb2;
2257 struct ipv6hdr *iph;
2258 struct net_device *dev;
2259 int vif;
2260
2261 dev = skb->dev;
2262 if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2263 read_unlock(&mrt_lock);
2264 return -ENODEV;
2265 }
2266
2267 /* really correct? */
2268 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2269 if (!skb2) {
2270 read_unlock(&mrt_lock);
2271 return -ENOMEM;
2272 }
2273
2274 NETLINK_CB(skb2).portid = portid;
2275 skb_reset_transport_header(skb2);
2276
2277 skb_put(skb2, sizeof(struct ipv6hdr));
2278 skb_reset_network_header(skb2);
2279
2280 iph = ipv6_hdr(skb2);
2281 iph->version = 0;
2282 iph->priority = 0;
2283 iph->flow_lbl[0] = 0;
2284 iph->flow_lbl[1] = 0;
2285 iph->flow_lbl[2] = 0;
2286 iph->payload_len = 0;
2287 iph->nexthdr = IPPROTO_NONE;
2288 iph->hop_limit = 0;
2289 iph->saddr = rt->rt6i_src.addr;
2290 iph->daddr = rt->rt6i_dst.addr;
2291
2292 err = ip6mr_cache_unresolved(mrt, vif, skb2, dev);
2293 read_unlock(&mrt_lock);
2294
2295 return err;
2296 }
2297
2298 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2299 read_unlock(&mrt_lock);
2300 return err;
2301 }
2302
2303 static int ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2304 u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2305 int flags)
2306 {
2307 struct nlmsghdr *nlh;
2308 struct rtmsg *rtm;
2309 int err;
2310
2311 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2312 if (!nlh)
2313 return -EMSGSIZE;
2314
2315 rtm = nlmsg_data(nlh);
2316 rtm->rtm_family = RTNL_FAMILY_IP6MR;
2317 rtm->rtm_dst_len = 128;
2318 rtm->rtm_src_len = 128;
2319 rtm->rtm_tos = 0;
2320 rtm->rtm_table = mrt->id;
2321 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2322 goto nla_put_failure;
2323 rtm->rtm_type = RTN_MULTICAST;
2324 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2325 if (c->_c.mfc_flags & MFC_STATIC)
2326 rtm->rtm_protocol = RTPROT_STATIC;
2327 else
2328 rtm->rtm_protocol = RTPROT_MROUTED;
2329 rtm->rtm_flags = 0;
2330
2331 if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2332 nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2333 goto nla_put_failure;
2334 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2335 /* do not break the dump if cache is unresolved */
2336 if (err < 0 && err != -ENOENT)
2337 goto nla_put_failure;
2338
2339 nlmsg_end(skb, nlh);
2340 return 0;
2341
2342 nla_put_failure:
2343 nlmsg_cancel(skb, nlh);
2344 return -EMSGSIZE;
2345 }
2346
2347 static int _ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2348 u32 portid, u32 seq, struct mr_mfc *c,
2349 int cmd, int flags)
2350 {
2351 return ip6mr_fill_mroute(mrt, skb, portid, seq, (struct mfc6_cache *)c,
2352 cmd, flags);
2353 }
2354
2355 static int mr6_msgsize(bool unresolved, int maxvif)
2356 {
2357 size_t len =
2358 NLMSG_ALIGN(sizeof(struct rtmsg))
2359 + nla_total_size(4) /* RTA_TABLE */
2360 + nla_total_size(sizeof(struct in6_addr)) /* RTA_SRC */
2361 + nla_total_size(sizeof(struct in6_addr)) /* RTA_DST */
2362 ;
2363
2364 if (!unresolved)
2365 len = len
2366 + nla_total_size(4) /* RTA_IIF */
2367 + nla_total_size(0) /* RTA_MULTIPATH */
2368 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2369 /* RTA_MFC_STATS */
2370 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2371 ;
2372
2373 return len;
2374 }
2375
2376 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc,
2377 int cmd)
2378 {
2379 struct net *net = read_pnet(&mrt->net);
2380 struct sk_buff *skb;
2381 int err = -ENOBUFS;
2382
2383 skb = nlmsg_new(mr6_msgsize(mfc->_c.mfc_parent >= MAXMIFS, mrt->maxvif),
2384 GFP_ATOMIC);
2385 if (!skb)
2386 goto errout;
2387
2388 err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2389 if (err < 0)
2390 goto errout;
2391
2392 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2393 return;
2394
2395 errout:
2396 kfree_skb(skb);
2397 if (err < 0)
2398 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2399 }
2400
2401 static size_t mrt6msg_netlink_msgsize(size_t payloadlen)
2402 {
2403 size_t len =
2404 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2405 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */
2406 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */
2407 /* IP6MRA_CREPORT_SRC_ADDR */
2408 + nla_total_size(sizeof(struct in6_addr))
2409 /* IP6MRA_CREPORT_DST_ADDR */
2410 + nla_total_size(sizeof(struct in6_addr))
2411 /* IP6MRA_CREPORT_PKT */
2412 + nla_total_size(payloadlen)
2413 ;
2414
2415 return len;
2416 }
2417
2418 static void mrt6msg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2419 {
2420 struct net *net = read_pnet(&mrt->net);
2421 struct nlmsghdr *nlh;
2422 struct rtgenmsg *rtgenm;
2423 struct mrt6msg *msg;
2424 struct sk_buff *skb;
2425 struct nlattr *nla;
2426 int payloadlen;
2427
2428 payloadlen = pkt->len - sizeof(struct mrt6msg);
2429 msg = (struct mrt6msg *)skb_transport_header(pkt);
2430
2431 skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2432 if (!skb)
2433 goto errout;
2434
2435 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2436 sizeof(struct rtgenmsg), 0);
2437 if (!nlh)
2438 goto errout;
2439 rtgenm = nlmsg_data(nlh);
2440 rtgenm->rtgen_family = RTNL_FAMILY_IP6MR;
2441 if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) ||
2442 nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) ||
2443 nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR,
2444 &msg->im6_src) ||
2445 nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR,
2446 &msg->im6_dst))
2447 goto nla_put_failure;
2448
2449 nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen);
2450 if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg),
2451 nla_data(nla), payloadlen))
2452 goto nla_put_failure;
2453
2454 nlmsg_end(skb, nlh);
2455
2456 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC);
2457 return;
2458
2459 nla_put_failure:
2460 nlmsg_cancel(skb, nlh);
2461 errout:
2462 kfree_skb(skb);
2463 rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS);
2464 }
2465
2466 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2467 {
2468 const struct nlmsghdr *nlh = cb->nlh;
2469 struct fib_dump_filter filter = {};
2470 int err;
2471
2472 if (cb->strict_check) {
2473 err = ip_valid_fib_dump_req(sock_net(skb->sk), nlh,
2474 &filter, cb);
2475 if (err < 0)
2476 return err;
2477 }
2478
2479 if (filter.table_id) {
2480 struct mr_table *mrt;
2481
2482 mrt = ip6mr_get_table(sock_net(skb->sk), filter.table_id);
2483 if (!mrt) {
2484 if (filter.dump_all_families)
2485 return skb->len;
2486
2487 NL_SET_ERR_MSG_MOD(cb->extack, "MR table does not exist");
2488 return -ENOENT;
2489 }
2490 err = mr_table_dump(mrt, skb, cb, _ip6mr_fill_mroute,
2491 &mfc_unres_lock, &filter);
2492 return skb->len ? : err;
2493 }
2494
2495 return mr_rtm_dumproute(skb, cb, ip6mr_mr_table_iter,
2496 _ip6mr_fill_mroute, &mfc_unres_lock, &filter);
2497 }
Ubuntu Jammy Linux kernel mirror