xref: /linux/net/ipv6/ip6mr.c (revision 4359a011e259a4608afc7fb3635370c9d4ba5943)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Linux IPv6 multicast routing support for BSD pim6sd
4  *	Based on net/ipv4/ipmr.c.
5  *
6  *	(c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
7  *		LSIIT Laboratory, Strasbourg, France
8  *	(c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
9  *		6WIND, Paris, France
10  *	Copyright (C)2007,2008 USAGI/WIDE Project
11  *		YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
12  */
13 
14 #include <linux/uaccess.h>
15 #include <linux/types.h>
16 #include <linux/sched.h>
17 #include <linux/errno.h>
18 #include <linux/mm.h>
19 #include <linux/kernel.h>
20 #include <linux/fcntl.h>
21 #include <linux/stat.h>
22 #include <linux/socket.h>
23 #include <linux/inet.h>
24 #include <linux/netdevice.h>
25 #include <linux/inetdevice.h>
26 #include <linux/proc_fs.h>
27 #include <linux/seq_file.h>
28 #include <linux/init.h>
29 #include <linux/compat.h>
30 #include <linux/rhashtable.h>
31 #include <net/protocol.h>
32 #include <linux/skbuff.h>
33 #include <net/raw.h>
34 #include <linux/notifier.h>
35 #include <linux/if_arp.h>
36 #include <net/checksum.h>
37 #include <net/netlink.h>
38 #include <net/fib_rules.h>
39 
40 #include <net/ipv6.h>
41 #include <net/ip6_route.h>
42 #include <linux/mroute6.h>
43 #include <linux/pim.h>
44 #include <net/addrconf.h>
45 #include <linux/netfilter_ipv6.h>
46 #include <linux/export.h>
47 #include <net/ip6_checksum.h>
48 #include <linux/netconf.h>
49 #include <net/ip_tunnels.h>
50 
51 #include <linux/nospec.h>
52 
53 struct ip6mr_rule {
54 	struct fib_rule		common;
55 };
56 
57 struct ip6mr_result {
58 	struct mr_table	*mrt;
59 };
60 
61 /* Big lock, protecting vif table, mrt cache and mroute socket state.
62    Note that the changes are semaphored via rtnl_lock.
63  */
64 
65 static DEFINE_SPINLOCK(mrt_lock);
66 
67 static struct net_device *vif_dev_read(const struct vif_device *vif)
68 {
69 	return rcu_dereference(vif->dev);
70 }
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(const 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(const struct mr_table *mrt, struct sk_buff *pkt);
98 static int ip6mr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
99 			      struct netlink_ext_ack *extack);
100 static int ip6mr_rtm_dumproute(struct sk_buff *skb,
101 			       struct netlink_callback *cb);
102 static void mroute_clean_tables(struct mr_table *mrt, int flags);
103 static void ipmr_expire_process(struct timer_list *t);
104 
105 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
106 #define ip6mr_for_each_table(mrt, net) \
107 	list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list, \
108 				lockdep_rtnl_is_held() || \
109 				list_empty(&net->ipv6.mr6_tables))
110 
111 static struct mr_table *ip6mr_mr_table_iter(struct net *net,
112 					    struct mr_table *mrt)
113 {
114 	struct mr_table *ret;
115 
116 	if (!mrt)
117 		ret = list_entry_rcu(net->ipv6.mr6_tables.next,
118 				     struct mr_table, list);
119 	else
120 		ret = list_entry_rcu(mrt->list.next,
121 				     struct mr_table, list);
122 
123 	if (&ret->list == &net->ipv6.mr6_tables)
124 		return NULL;
125 	return ret;
126 }
127 
128 static struct mr_table *ip6mr_get_table(struct net *net, u32 id)
129 {
130 	struct mr_table *mrt;
131 
132 	ip6mr_for_each_table(mrt, net) {
133 		if (mrt->id == id)
134 			return mrt;
135 	}
136 	return NULL;
137 }
138 
139 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
140 			    struct mr_table **mrt)
141 {
142 	int err;
143 	struct ip6mr_result res;
144 	struct fib_lookup_arg arg = {
145 		.result = &res,
146 		.flags = FIB_LOOKUP_NOREF,
147 	};
148 
149 	/* update flow if oif or iif point to device enslaved to l3mdev */
150 	l3mdev_update_flow(net, flowi6_to_flowi(flp6));
151 
152 	err = fib_rules_lookup(net->ipv6.mr6_rules_ops,
153 			       flowi6_to_flowi(flp6), 0, &arg);
154 	if (err < 0)
155 		return err;
156 	*mrt = res.mrt;
157 	return 0;
158 }
159 
160 static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
161 			     int flags, struct fib_lookup_arg *arg)
162 {
163 	struct ip6mr_result *res = arg->result;
164 	struct mr_table *mrt;
165 
166 	switch (rule->action) {
167 	case FR_ACT_TO_TBL:
168 		break;
169 	case FR_ACT_UNREACHABLE:
170 		return -ENETUNREACH;
171 	case FR_ACT_PROHIBIT:
172 		return -EACCES;
173 	case FR_ACT_BLACKHOLE:
174 	default:
175 		return -EINVAL;
176 	}
177 
178 	arg->table = fib_rule_get_table(rule, arg);
179 
180 	mrt = ip6mr_get_table(rule->fr_net, arg->table);
181 	if (!mrt)
182 		return -EAGAIN;
183 	res->mrt = mrt;
184 	return 0;
185 }
186 
187 static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
188 {
189 	return 1;
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 	.owner		= THIS_MODULE,
225 };
226 
227 static int __net_init ip6mr_rules_init(struct net *net)
228 {
229 	struct fib_rules_ops *ops;
230 	struct mr_table *mrt;
231 	int err;
232 
233 	ops = fib_rules_register(&ip6mr_rules_ops_template, net);
234 	if (IS_ERR(ops))
235 		return PTR_ERR(ops);
236 
237 	INIT_LIST_HEAD(&net->ipv6.mr6_tables);
238 
239 	mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
240 	if (IS_ERR(mrt)) {
241 		err = PTR_ERR(mrt);
242 		goto err1;
243 	}
244 
245 	err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
246 	if (err < 0)
247 		goto err2;
248 
249 	net->ipv6.mr6_rules_ops = ops;
250 	return 0;
251 
252 err2:
253 	rtnl_lock();
254 	ip6mr_free_table(mrt);
255 	rtnl_unlock();
256 err1:
257 	fib_rules_unregister(ops);
258 	return err;
259 }
260 
261 static void __net_exit ip6mr_rules_exit(struct net *net)
262 {
263 	struct mr_table *mrt, *next;
264 
265 	ASSERT_RTNL();
266 	list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
267 		list_del(&mrt->list);
268 		ip6mr_free_table(mrt);
269 	}
270 	fib_rules_unregister(net->ipv6.mr6_rules_ops);
271 }
272 
273 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb,
274 			    struct netlink_ext_ack *extack)
275 {
276 	return fib_rules_dump(net, nb, RTNL_FAMILY_IP6MR, extack);
277 }
278 
279 static unsigned int ip6mr_rules_seq_read(struct net *net)
280 {
281 	return fib_rules_seq_read(net, RTNL_FAMILY_IP6MR);
282 }
283 
284 bool ip6mr_rule_default(const struct fib_rule *rule)
285 {
286 	return fib_rule_matchall(rule) && rule->action == FR_ACT_TO_TBL &&
287 	       rule->table == RT6_TABLE_DFLT && !rule->l3mdev;
288 }
289 EXPORT_SYMBOL(ip6mr_rule_default);
290 #else
291 #define ip6mr_for_each_table(mrt, net) \
292 	for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
293 
294 static struct mr_table *ip6mr_mr_table_iter(struct net *net,
295 					    struct mr_table *mrt)
296 {
297 	if (!mrt)
298 		return net->ipv6.mrt6;
299 	return NULL;
300 }
301 
302 static struct mr_table *ip6mr_get_table(struct net *net, u32 id)
303 {
304 	return net->ipv6.mrt6;
305 }
306 
307 static int ip6mr_fib_lookup(struct net *net, struct flowi6 *flp6,
308 			    struct mr_table **mrt)
309 {
310 	*mrt = net->ipv6.mrt6;
311 	return 0;
312 }
313 
314 static int __net_init ip6mr_rules_init(struct net *net)
315 {
316 	struct mr_table *mrt;
317 
318 	mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
319 	if (IS_ERR(mrt))
320 		return PTR_ERR(mrt);
321 	net->ipv6.mrt6 = mrt;
322 	return 0;
323 }
324 
325 static void __net_exit ip6mr_rules_exit(struct net *net)
326 {
327 	ASSERT_RTNL();
328 	ip6mr_free_table(net->ipv6.mrt6);
329 	net->ipv6.mrt6 = NULL;
330 }
331 
332 static int ip6mr_rules_dump(struct net *net, struct notifier_block *nb,
333 			    struct netlink_ext_ack *extack)
334 {
335 	return 0;
336 }
337 
338 static unsigned int ip6mr_rules_seq_read(struct net *net)
339 {
340 	return 0;
341 }
342 #endif
343 
344 static int ip6mr_hash_cmp(struct rhashtable_compare_arg *arg,
345 			  const void *ptr)
346 {
347 	const struct mfc6_cache_cmp_arg *cmparg = arg->key;
348 	struct mfc6_cache *c = (struct mfc6_cache *)ptr;
349 
350 	return !ipv6_addr_equal(&c->mf6c_mcastgrp, &cmparg->mf6c_mcastgrp) ||
351 	       !ipv6_addr_equal(&c->mf6c_origin, &cmparg->mf6c_origin);
352 }
353 
354 static const struct rhashtable_params ip6mr_rht_params = {
355 	.head_offset = offsetof(struct mr_mfc, mnode),
356 	.key_offset = offsetof(struct mfc6_cache, cmparg),
357 	.key_len = sizeof(struct mfc6_cache_cmp_arg),
358 	.nelem_hint = 3,
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, MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC |
397 				 MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC);
398 	rhltable_destroy(&mrt->mfc_hash);
399 	kfree(mrt);
400 }
401 
402 #ifdef CONFIG_PROC_FS
403 /* The /proc interfaces to multicast routing
404  * /proc/ip6_mr_cache /proc/ip6_mr_vif
405  */
406 
407 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
408 	__acquires(RCU)
409 {
410 	struct mr_vif_iter *iter = seq->private;
411 	struct net *net = seq_file_net(seq);
412 	struct mr_table *mrt;
413 
414 	mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
415 	if (!mrt)
416 		return ERR_PTR(-ENOENT);
417 
418 	iter->mrt = mrt;
419 
420 	rcu_read_lock();
421 	return mr_vif_seq_start(seq, pos);
422 }
423 
424 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
425 	__releases(RCU)
426 {
427 	rcu_read_unlock();
428 }
429 
430 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
431 {
432 	struct mr_vif_iter *iter = seq->private;
433 	struct mr_table *mrt = iter->mrt;
434 
435 	if (v == SEQ_START_TOKEN) {
436 		seq_puts(seq,
437 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags\n");
438 	} else {
439 		const struct vif_device *vif = v;
440 		const struct net_device *vif_dev;
441 		const char *name;
442 
443 		vif_dev = vif_dev_read(vif);
444 		name = vif_dev ? vif_dev->name : "none";
445 
446 		seq_printf(seq,
447 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X\n",
448 			   vif - mrt->vif_table,
449 			   name, vif->bytes_in, vif->pkt_in,
450 			   vif->bytes_out, vif->pkt_out,
451 			   vif->flags);
452 	}
453 	return 0;
454 }
455 
456 static const struct seq_operations ip6mr_vif_seq_ops = {
457 	.start = ip6mr_vif_seq_start,
458 	.next  = mr_vif_seq_next,
459 	.stop  = ip6mr_vif_seq_stop,
460 	.show  = ip6mr_vif_seq_show,
461 };
462 
463 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
464 {
465 	struct net *net = seq_file_net(seq);
466 	struct mr_table *mrt;
467 
468 	mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
469 	if (!mrt)
470 		return ERR_PTR(-ENOENT);
471 
472 	return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
473 }
474 
475 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
476 {
477 	int n;
478 
479 	if (v == SEQ_START_TOKEN) {
480 		seq_puts(seq,
481 			 "Group                            "
482 			 "Origin                           "
483 			 "Iif      Pkts  Bytes     Wrong  Oifs\n");
484 	} else {
485 		const struct mfc6_cache *mfc = v;
486 		const struct mr_mfc_iter *it = seq->private;
487 		struct mr_table *mrt = it->mrt;
488 
489 		seq_printf(seq, "%pI6 %pI6 %-3hd",
490 			   &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
491 			   mfc->_c.mfc_parent);
492 
493 		if (it->cache != &mrt->mfc_unres_queue) {
494 			seq_printf(seq, " %8lu %8lu %8lu",
495 				   mfc->_c.mfc_un.res.pkt,
496 				   mfc->_c.mfc_un.res.bytes,
497 				   mfc->_c.mfc_un.res.wrong_if);
498 			for (n = mfc->_c.mfc_un.res.minvif;
499 			     n < mfc->_c.mfc_un.res.maxvif; n++) {
500 				if (VIF_EXISTS(mrt, n) &&
501 				    mfc->_c.mfc_un.res.ttls[n] < 255)
502 					seq_printf(seq,
503 						   " %2d:%-3d", n,
504 						   mfc->_c.mfc_un.res.ttls[n]);
505 			}
506 		} else {
507 			/* unresolved mfc_caches don't contain
508 			 * pkt, bytes and wrong_if values
509 			 */
510 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
511 		}
512 		seq_putc(seq, '\n');
513 	}
514 	return 0;
515 }
516 
517 static const struct seq_operations ipmr_mfc_seq_ops = {
518 	.start = ipmr_mfc_seq_start,
519 	.next  = mr_mfc_seq_next,
520 	.stop  = mr_mfc_seq_stop,
521 	.show  = ipmr_mfc_seq_show,
522 };
523 #endif
524 
525 #ifdef CONFIG_IPV6_PIMSM_V2
526 
527 static int pim6_rcv(struct sk_buff *skb)
528 {
529 	struct pimreghdr *pim;
530 	struct ipv6hdr   *encap;
531 	struct net_device  *reg_dev = NULL;
532 	struct net *net = dev_net(skb->dev);
533 	struct mr_table *mrt;
534 	struct flowi6 fl6 = {
535 		.flowi6_iif	= skb->dev->ifindex,
536 		.flowi6_mark	= skb->mark,
537 	};
538 	int reg_vif_num;
539 
540 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
541 		goto drop;
542 
543 	pim = (struct pimreghdr *)skb_transport_header(skb);
544 	if (pim->type != ((PIM_VERSION << 4) | PIM_TYPE_REGISTER) ||
545 	    (pim->flags & PIM_NULL_REGISTER) ||
546 	    (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
547 			     sizeof(*pim), IPPROTO_PIM,
548 			     csum_partial((void *)pim, sizeof(*pim), 0)) &&
549 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
550 		goto drop;
551 
552 	/* check if the inner packet is destined to mcast group */
553 	encap = (struct ipv6hdr *)(skb_transport_header(skb) +
554 				   sizeof(*pim));
555 
556 	if (!ipv6_addr_is_multicast(&encap->daddr) ||
557 	    encap->payload_len == 0 ||
558 	    ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
559 		goto drop;
560 
561 	if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
562 		goto drop;
563 
564 	/* Pairs with WRITE_ONCE() in mif6_add()/mif6_delete() */
565 	reg_vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
566 	if (reg_vif_num >= 0)
567 		reg_dev = vif_dev_read(&mrt->vif_table[reg_vif_num]);
568 
569 	if (!reg_dev)
570 		goto drop;
571 
572 	skb->mac_header = skb->network_header;
573 	skb_pull(skb, (u8 *)encap - skb->data);
574 	skb_reset_network_header(skb);
575 	skb->protocol = htons(ETH_P_IPV6);
576 	skb->ip_summed = CHECKSUM_NONE;
577 
578 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
579 
580 	netif_rx(skb);
581 
582 	return 0;
583  drop:
584 	kfree_skb(skb);
585 	return 0;
586 }
587 
588 static const struct inet6_protocol pim6_protocol = {
589 	.handler	=	pim6_rcv,
590 };
591 
592 /* Service routines creating virtual interfaces: PIMREG */
593 
594 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
595 				      struct net_device *dev)
596 {
597 	struct net *net = dev_net(dev);
598 	struct mr_table *mrt;
599 	struct flowi6 fl6 = {
600 		.flowi6_oif	= dev->ifindex,
601 		.flowi6_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
602 		.flowi6_mark	= skb->mark,
603 	};
604 
605 	if (!pskb_inet_may_pull(skb))
606 		goto tx_err;
607 
608 	if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
609 		goto tx_err;
610 
611 	dev->stats.tx_bytes += skb->len;
612 	dev->stats.tx_packets++;
613 	rcu_read_lock();
614 	ip6mr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
615 			   MRT6MSG_WHOLEPKT);
616 	rcu_read_unlock();
617 	kfree_skb(skb);
618 	return NETDEV_TX_OK;
619 
620 tx_err:
621 	dev->stats.tx_errors++;
622 	kfree_skb(skb);
623 	return NETDEV_TX_OK;
624 }
625 
626 static int reg_vif_get_iflink(const struct net_device *dev)
627 {
628 	return 0;
629 }
630 
631 static const struct net_device_ops reg_vif_netdev_ops = {
632 	.ndo_start_xmit	= reg_vif_xmit,
633 	.ndo_get_iflink = reg_vif_get_iflink,
634 };
635 
636 static void reg_vif_setup(struct net_device *dev)
637 {
638 	dev->type		= ARPHRD_PIMREG;
639 	dev->mtu		= 1500 - sizeof(struct ipv6hdr) - 8;
640 	dev->flags		= IFF_NOARP;
641 	dev->netdev_ops		= &reg_vif_netdev_ops;
642 	dev->needs_free_netdev	= true;
643 	dev->features		|= NETIF_F_NETNS_LOCAL;
644 }
645 
646 static struct net_device *ip6mr_reg_vif(struct net *net, struct mr_table *mrt)
647 {
648 	struct net_device *dev;
649 	char name[IFNAMSIZ];
650 
651 	if (mrt->id == RT6_TABLE_DFLT)
652 		sprintf(name, "pim6reg");
653 	else
654 		sprintf(name, "pim6reg%u", mrt->id);
655 
656 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
657 	if (!dev)
658 		return NULL;
659 
660 	dev_net_set(dev, net);
661 
662 	if (register_netdevice(dev)) {
663 		free_netdev(dev);
664 		return NULL;
665 	}
666 
667 	if (dev_open(dev, NULL))
668 		goto failure;
669 
670 	dev_hold(dev);
671 	return dev;
672 
673 failure:
674 	unregister_netdevice(dev);
675 	return NULL;
676 }
677 #endif
678 
679 static int call_ip6mr_vif_entry_notifiers(struct net *net,
680 					  enum fib_event_type event_type,
681 					  struct vif_device *vif,
682 					  struct net_device *vif_dev,
683 					  mifi_t vif_index, u32 tb_id)
684 {
685 	return mr_call_vif_notifiers(net, RTNL_FAMILY_IP6MR, event_type,
686 				     vif, vif_dev, vif_index, tb_id,
687 				     &net->ipv6.ipmr_seq);
688 }
689 
690 static int call_ip6mr_mfc_entry_notifiers(struct net *net,
691 					  enum fib_event_type event_type,
692 					  struct mfc6_cache *mfc, u32 tb_id)
693 {
694 	return mr_call_mfc_notifiers(net, RTNL_FAMILY_IP6MR, event_type,
695 				     &mfc->_c, tb_id, &net->ipv6.ipmr_seq);
696 }
697 
698 /* Delete a VIF entry */
699 static int mif6_delete(struct mr_table *mrt, int vifi, int notify,
700 		       struct list_head *head)
701 {
702 	struct vif_device *v;
703 	struct net_device *dev;
704 	struct inet6_dev *in6_dev;
705 
706 	if (vifi < 0 || vifi >= mrt->maxvif)
707 		return -EADDRNOTAVAIL;
708 
709 	v = &mrt->vif_table[vifi];
710 
711 	dev = rtnl_dereference(v->dev);
712 	if (!dev)
713 		return -EADDRNOTAVAIL;
714 
715 	call_ip6mr_vif_entry_notifiers(read_pnet(&mrt->net),
716 				       FIB_EVENT_VIF_DEL, v, dev,
717 				       vifi, mrt->id);
718 	spin_lock(&mrt_lock);
719 	RCU_INIT_POINTER(v->dev, NULL);
720 
721 #ifdef CONFIG_IPV6_PIMSM_V2
722 	if (vifi == mrt->mroute_reg_vif_num) {
723 		/* Pairs with READ_ONCE() in ip6mr_cache_report() and reg_vif_xmit() */
724 		WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
725 	}
726 #endif
727 
728 	if (vifi + 1 == mrt->maxvif) {
729 		int tmp;
730 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
731 			if (VIF_EXISTS(mrt, tmp))
732 				break;
733 		}
734 		WRITE_ONCE(mrt->maxvif, tmp + 1);
735 	}
736 
737 	spin_unlock(&mrt_lock);
738 
739 	dev_set_allmulti(dev, -1);
740 
741 	in6_dev = __in6_dev_get(dev);
742 	if (in6_dev) {
743 		atomic_dec(&in6_dev->cnf.mc_forwarding);
744 		inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
745 					     NETCONFA_MC_FORWARDING,
746 					     dev->ifindex, &in6_dev->cnf);
747 	}
748 
749 	if ((v->flags & MIFF_REGISTER) && !notify)
750 		unregister_netdevice_queue(dev, head);
751 
752 	netdev_put(dev, &v->dev_tracker);
753 	return 0;
754 }
755 
756 static inline void ip6mr_cache_free_rcu(struct rcu_head *head)
757 {
758 	struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
759 
760 	kmem_cache_free(mrt_cachep, (struct mfc6_cache *)c);
761 }
762 
763 static inline void ip6mr_cache_free(struct mfc6_cache *c)
764 {
765 	call_rcu(&c->_c.rcu, ip6mr_cache_free_rcu);
766 }
767 
768 /* Destroy an unresolved cache entry, killing queued skbs
769    and reporting error to netlink readers.
770  */
771 
772 static void ip6mr_destroy_unres(struct mr_table *mrt, struct mfc6_cache *c)
773 {
774 	struct net *net = read_pnet(&mrt->net);
775 	struct sk_buff *skb;
776 
777 	atomic_dec(&mrt->cache_resolve_queue_len);
778 
779 	while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved)) != NULL) {
780 		if (ipv6_hdr(skb)->version == 0) {
781 			struct nlmsghdr *nlh = skb_pull(skb,
782 							sizeof(struct ipv6hdr));
783 			nlh->nlmsg_type = NLMSG_ERROR;
784 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
785 			skb_trim(skb, nlh->nlmsg_len);
786 			((struct nlmsgerr *)nlmsg_data(nlh))->error = -ETIMEDOUT;
787 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
788 		} else
789 			kfree_skb(skb);
790 	}
791 
792 	ip6mr_cache_free(c);
793 }
794 
795 
796 /* Timer process for all the unresolved queue. */
797 
798 static void ipmr_do_expire_process(struct mr_table *mrt)
799 {
800 	unsigned long now = jiffies;
801 	unsigned long expires = 10 * HZ;
802 	struct mr_mfc *c, *next;
803 
804 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
805 		if (time_after(c->mfc_un.unres.expires, now)) {
806 			/* not yet... */
807 			unsigned long interval = c->mfc_un.unres.expires - now;
808 			if (interval < expires)
809 				expires = interval;
810 			continue;
811 		}
812 
813 		list_del(&c->list);
814 		mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE);
815 		ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c);
816 	}
817 
818 	if (!list_empty(&mrt->mfc_unres_queue))
819 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
820 }
821 
822 static void ipmr_expire_process(struct timer_list *t)
823 {
824 	struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
825 
826 	if (!spin_trylock(&mfc_unres_lock)) {
827 		mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
828 		return;
829 	}
830 
831 	if (!list_empty(&mrt->mfc_unres_queue))
832 		ipmr_do_expire_process(mrt);
833 
834 	spin_unlock(&mfc_unres_lock);
835 }
836 
837 /* Fill oifs list. It is called under locked mrt_lock. */
838 
839 static void ip6mr_update_thresholds(struct mr_table *mrt,
840 				    struct mr_mfc *cache,
841 				    unsigned char *ttls)
842 {
843 	int vifi;
844 
845 	cache->mfc_un.res.minvif = MAXMIFS;
846 	cache->mfc_un.res.maxvif = 0;
847 	memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
848 
849 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
850 		if (VIF_EXISTS(mrt, vifi) &&
851 		    ttls[vifi] && ttls[vifi] < 255) {
852 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
853 			if (cache->mfc_un.res.minvif > vifi)
854 				cache->mfc_un.res.minvif = vifi;
855 			if (cache->mfc_un.res.maxvif <= vifi)
856 				cache->mfc_un.res.maxvif = vifi + 1;
857 		}
858 	}
859 	cache->mfc_un.res.lastuse = jiffies;
860 }
861 
862 static int mif6_add(struct net *net, struct mr_table *mrt,
863 		    struct mif6ctl *vifc, int mrtsock)
864 {
865 	int vifi = vifc->mif6c_mifi;
866 	struct vif_device *v = &mrt->vif_table[vifi];
867 	struct net_device *dev;
868 	struct inet6_dev *in6_dev;
869 	int err;
870 
871 	/* Is vif busy ? */
872 	if (VIF_EXISTS(mrt, vifi))
873 		return -EADDRINUSE;
874 
875 	switch (vifc->mif6c_flags) {
876 #ifdef CONFIG_IPV6_PIMSM_V2
877 	case MIFF_REGISTER:
878 		/*
879 		 * Special Purpose VIF in PIM
880 		 * All the packets will be sent to the daemon
881 		 */
882 		if (mrt->mroute_reg_vif_num >= 0)
883 			return -EADDRINUSE;
884 		dev = ip6mr_reg_vif(net, mrt);
885 		if (!dev)
886 			return -ENOBUFS;
887 		err = dev_set_allmulti(dev, 1);
888 		if (err) {
889 			unregister_netdevice(dev);
890 			dev_put(dev);
891 			return err;
892 		}
893 		break;
894 #endif
895 	case 0:
896 		dev = dev_get_by_index(net, vifc->mif6c_pifi);
897 		if (!dev)
898 			return -EADDRNOTAVAIL;
899 		err = dev_set_allmulti(dev, 1);
900 		if (err) {
901 			dev_put(dev);
902 			return err;
903 		}
904 		break;
905 	default:
906 		return -EINVAL;
907 	}
908 
909 	in6_dev = __in6_dev_get(dev);
910 	if (in6_dev) {
911 		atomic_inc(&in6_dev->cnf.mc_forwarding);
912 		inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
913 					     NETCONFA_MC_FORWARDING,
914 					     dev->ifindex, &in6_dev->cnf);
915 	}
916 
917 	/* Fill in the VIF structures */
918 	vif_device_init(v, dev, vifc->vifc_rate_limit, vifc->vifc_threshold,
919 			vifc->mif6c_flags | (!mrtsock ? VIFF_STATIC : 0),
920 			MIFF_REGISTER);
921 
922 	/* And finish update writing critical data */
923 	spin_lock(&mrt_lock);
924 	rcu_assign_pointer(v->dev, dev);
925 	netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
926 #ifdef CONFIG_IPV6_PIMSM_V2
927 	if (v->flags & MIFF_REGISTER)
928 		WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
929 #endif
930 	if (vifi + 1 > mrt->maxvif)
931 		WRITE_ONCE(mrt->maxvif, vifi + 1);
932 	spin_unlock(&mrt_lock);
933 	call_ip6mr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD,
934 				       v, dev, vifi, mrt->id);
935 	return 0;
936 }
937 
938 static struct mfc6_cache *ip6mr_cache_find(struct mr_table *mrt,
939 					   const struct in6_addr *origin,
940 					   const struct in6_addr *mcastgrp)
941 {
942 	struct mfc6_cache_cmp_arg arg = {
943 		.mf6c_origin = *origin,
944 		.mf6c_mcastgrp = *mcastgrp,
945 	};
946 
947 	return mr_mfc_find(mrt, &arg);
948 }
949 
950 /* Look for a (*,G) entry */
951 static struct mfc6_cache *ip6mr_cache_find_any(struct mr_table *mrt,
952 					       struct in6_addr *mcastgrp,
953 					       mifi_t mifi)
954 {
955 	struct mfc6_cache_cmp_arg arg = {
956 		.mf6c_origin = in6addr_any,
957 		.mf6c_mcastgrp = *mcastgrp,
958 	};
959 
960 	if (ipv6_addr_any(mcastgrp))
961 		return mr_mfc_find_any_parent(mrt, mifi);
962 	return mr_mfc_find_any(mrt, mifi, &arg);
963 }
964 
965 /* Look for a (S,G,iif) entry if parent != -1 */
966 static struct mfc6_cache *
967 ip6mr_cache_find_parent(struct mr_table *mrt,
968 			const struct in6_addr *origin,
969 			const struct in6_addr *mcastgrp,
970 			int parent)
971 {
972 	struct mfc6_cache_cmp_arg arg = {
973 		.mf6c_origin = *origin,
974 		.mf6c_mcastgrp = *mcastgrp,
975 	};
976 
977 	return mr_mfc_find_parent(mrt, &arg, parent);
978 }
979 
980 /* Allocate a multicast cache entry */
981 static struct mfc6_cache *ip6mr_cache_alloc(void)
982 {
983 	struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
984 	if (!c)
985 		return NULL;
986 	c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
987 	c->_c.mfc_un.res.minvif = MAXMIFS;
988 	c->_c.free = ip6mr_cache_free_rcu;
989 	refcount_set(&c->_c.mfc_un.res.refcount, 1);
990 	return c;
991 }
992 
993 static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
994 {
995 	struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
996 	if (!c)
997 		return NULL;
998 	skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
999 	c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1000 	return c;
1001 }
1002 
1003 /*
1004  *	A cache entry has gone into a resolved state from queued
1005  */
1006 
1007 static void ip6mr_cache_resolve(struct net *net, struct mr_table *mrt,
1008 				struct mfc6_cache *uc, struct mfc6_cache *c)
1009 {
1010 	struct sk_buff *skb;
1011 
1012 	/*
1013 	 *	Play the pending entries through our router
1014 	 */
1015 
1016 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1017 		if (ipv6_hdr(skb)->version == 0) {
1018 			struct nlmsghdr *nlh = skb_pull(skb,
1019 							sizeof(struct ipv6hdr));
1020 
1021 			if (mr_fill_mroute(mrt, skb, &c->_c,
1022 					   nlmsg_data(nlh)) > 0) {
1023 				nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1024 			} else {
1025 				nlh->nlmsg_type = NLMSG_ERROR;
1026 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1027 				skb_trim(skb, nlh->nlmsg_len);
1028 				((struct nlmsgerr *)nlmsg_data(nlh))->error = -EMSGSIZE;
1029 			}
1030 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1031 		} else {
1032 			rcu_read_lock();
1033 			ip6_mr_forward(net, mrt, skb->dev, skb, c);
1034 			rcu_read_unlock();
1035 		}
1036 	}
1037 }
1038 
1039 /*
1040  *	Bounce a cache query up to pim6sd and netlink.
1041  *
1042  *	Called under rcu_read_lock()
1043  */
1044 
1045 static int ip6mr_cache_report(const struct mr_table *mrt, struct sk_buff *pkt,
1046 			      mifi_t mifi, int assert)
1047 {
1048 	struct sock *mroute6_sk;
1049 	struct sk_buff *skb;
1050 	struct mrt6msg *msg;
1051 	int ret;
1052 
1053 #ifdef CONFIG_IPV6_PIMSM_V2
1054 	if (assert == MRT6MSG_WHOLEPKT || assert == MRT6MSG_WRMIFWHOLE)
1055 		skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1056 						+sizeof(*msg));
1057 	else
1058 #endif
1059 		skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1060 
1061 	if (!skb)
1062 		return -ENOBUFS;
1063 
1064 	/* I suppose that internal messages
1065 	 * do not require checksums */
1066 
1067 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1068 
1069 #ifdef CONFIG_IPV6_PIMSM_V2
1070 	if (assert == MRT6MSG_WHOLEPKT || assert == MRT6MSG_WRMIFWHOLE) {
1071 		/* Ugly, but we have no choice with this interface.
1072 		   Duplicate old header, fix length etc.
1073 		   And all this only to mangle msg->im6_msgtype and
1074 		   to set msg->im6_mbz to "mbz" :-)
1075 		 */
1076 		skb_push(skb, -skb_network_offset(pkt));
1077 
1078 		skb_push(skb, sizeof(*msg));
1079 		skb_reset_transport_header(skb);
1080 		msg = (struct mrt6msg *)skb_transport_header(skb);
1081 		msg->im6_mbz = 0;
1082 		msg->im6_msgtype = assert;
1083 		if (assert == MRT6MSG_WRMIFWHOLE)
1084 			msg->im6_mif = mifi;
1085 		else
1086 			msg->im6_mif = READ_ONCE(mrt->mroute_reg_vif_num);
1087 		msg->im6_pad = 0;
1088 		msg->im6_src = ipv6_hdr(pkt)->saddr;
1089 		msg->im6_dst = ipv6_hdr(pkt)->daddr;
1090 
1091 		skb->ip_summed = CHECKSUM_UNNECESSARY;
1092 	} else
1093 #endif
1094 	{
1095 	/*
1096 	 *	Copy the IP header
1097 	 */
1098 
1099 	skb_put(skb, sizeof(struct ipv6hdr));
1100 	skb_reset_network_header(skb);
1101 	skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1102 
1103 	/*
1104 	 *	Add our header
1105 	 */
1106 	skb_put(skb, sizeof(*msg));
1107 	skb_reset_transport_header(skb);
1108 	msg = (struct mrt6msg *)skb_transport_header(skb);
1109 
1110 	msg->im6_mbz = 0;
1111 	msg->im6_msgtype = assert;
1112 	msg->im6_mif = mifi;
1113 	msg->im6_pad = 0;
1114 	msg->im6_src = ipv6_hdr(pkt)->saddr;
1115 	msg->im6_dst = ipv6_hdr(pkt)->daddr;
1116 
1117 	skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1118 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1119 	}
1120 
1121 	mroute6_sk = rcu_dereference(mrt->mroute_sk);
1122 	if (!mroute6_sk) {
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 
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! */
1141 static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi,
1142 				  struct sk_buff *skb, struct net_device *dev)
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 		c = ip6mr_cache_alloc_unres();
1163 		if (!c) {
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 		if (dev) {
1203 			skb->dev = dev;
1204 			skb->skb_iif = dev->ifindex;
1205 		}
1206 		skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1207 		err = 0;
1208 	}
1209 
1210 	spin_unlock_bh(&mfc_unres_lock);
1211 	return err;
1212 }
1213 
1214 /*
1215  *	MFC6 cache manipulation by user space
1216  */
1217 
1218 static int ip6mr_mfc_delete(struct mr_table *mrt, struct mf6cctl *mfc,
1219 			    int parent)
1220 {
1221 	struct mfc6_cache *c;
1222 
1223 	/* The entries are added/deleted only under RTNL */
1224 	rcu_read_lock();
1225 	c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr,
1226 				    &mfc->mf6cc_mcastgrp.sin6_addr, parent);
1227 	rcu_read_unlock();
1228 	if (!c)
1229 		return -ENOENT;
1230 	rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ip6mr_rht_params);
1231 	list_del_rcu(&c->_c.list);
1232 
1233 	call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net),
1234 				       FIB_EVENT_ENTRY_DEL, c, mrt->id);
1235 	mr6_netlink_event(mrt, c, RTM_DELROUTE);
1236 	mr_cache_put(&c->_c);
1237 	return 0;
1238 }
1239 
1240 static int ip6mr_device_event(struct notifier_block *this,
1241 			      unsigned long event, void *ptr)
1242 {
1243 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1244 	struct net *net = dev_net(dev);
1245 	struct mr_table *mrt;
1246 	struct vif_device *v;
1247 	int ct;
1248 
1249 	if (event != NETDEV_UNREGISTER)
1250 		return NOTIFY_DONE;
1251 
1252 	ip6mr_for_each_table(mrt, net) {
1253 		v = &mrt->vif_table[0];
1254 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1255 			if (rcu_access_pointer(v->dev) == dev)
1256 				mif6_delete(mrt, ct, 1, NULL);
1257 		}
1258 	}
1259 
1260 	return NOTIFY_DONE;
1261 }
1262 
1263 static unsigned int ip6mr_seq_read(struct net *net)
1264 {
1265 	ASSERT_RTNL();
1266 
1267 	return net->ipv6.ipmr_seq + ip6mr_rules_seq_read(net);
1268 }
1269 
1270 static int ip6mr_dump(struct net *net, struct notifier_block *nb,
1271 		      struct netlink_ext_ack *extack)
1272 {
1273 	return mr_dump(net, nb, RTNL_FAMILY_IP6MR, ip6mr_rules_dump,
1274 		       ip6mr_mr_table_iter, extack);
1275 }
1276 
1277 static struct notifier_block ip6_mr_notifier = {
1278 	.notifier_call = ip6mr_device_event
1279 };
1280 
1281 static const struct fib_notifier_ops ip6mr_notifier_ops_template = {
1282 	.family		= RTNL_FAMILY_IP6MR,
1283 	.fib_seq_read	= ip6mr_seq_read,
1284 	.fib_dump	= ip6mr_dump,
1285 	.owner		= THIS_MODULE,
1286 };
1287 
1288 static int __net_init ip6mr_notifier_init(struct net *net)
1289 {
1290 	struct fib_notifier_ops *ops;
1291 
1292 	net->ipv6.ipmr_seq = 0;
1293 
1294 	ops = fib_notifier_ops_register(&ip6mr_notifier_ops_template, net);
1295 	if (IS_ERR(ops))
1296 		return PTR_ERR(ops);
1297 
1298 	net->ipv6.ip6mr_notifier_ops = ops;
1299 
1300 	return 0;
1301 }
1302 
1303 static void __net_exit ip6mr_notifier_exit(struct net *net)
1304 {
1305 	fib_notifier_ops_unregister(net->ipv6.ip6mr_notifier_ops);
1306 	net->ipv6.ip6mr_notifier_ops = NULL;
1307 }
1308 
1309 /* Setup for IP multicast routing */
1310 static int __net_init ip6mr_net_init(struct net *net)
1311 {
1312 	int err;
1313 
1314 	err = ip6mr_notifier_init(net);
1315 	if (err)
1316 		return err;
1317 
1318 	err = ip6mr_rules_init(net);
1319 	if (err < 0)
1320 		goto ip6mr_rules_fail;
1321 
1322 #ifdef CONFIG_PROC_FS
1323 	err = -ENOMEM;
1324 	if (!proc_create_net("ip6_mr_vif", 0, net->proc_net, &ip6mr_vif_seq_ops,
1325 			sizeof(struct mr_vif_iter)))
1326 		goto proc_vif_fail;
1327 	if (!proc_create_net("ip6_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
1328 			sizeof(struct mr_mfc_iter)))
1329 		goto proc_cache_fail;
1330 #endif
1331 
1332 	return 0;
1333 
1334 #ifdef CONFIG_PROC_FS
1335 proc_cache_fail:
1336 	remove_proc_entry("ip6_mr_vif", net->proc_net);
1337 proc_vif_fail:
1338 	rtnl_lock();
1339 	ip6mr_rules_exit(net);
1340 	rtnl_unlock();
1341 #endif
1342 ip6mr_rules_fail:
1343 	ip6mr_notifier_exit(net);
1344 	return err;
1345 }
1346 
1347 static void __net_exit ip6mr_net_exit(struct net *net)
1348 {
1349 #ifdef CONFIG_PROC_FS
1350 	remove_proc_entry("ip6_mr_cache", net->proc_net);
1351 	remove_proc_entry("ip6_mr_vif", net->proc_net);
1352 #endif
1353 	ip6mr_notifier_exit(net);
1354 }
1355 
1356 static void __net_exit ip6mr_net_exit_batch(struct list_head *net_list)
1357 {
1358 	struct net *net;
1359 
1360 	rtnl_lock();
1361 	list_for_each_entry(net, net_list, exit_list)
1362 		ip6mr_rules_exit(net);
1363 	rtnl_unlock();
1364 }
1365 
1366 static struct pernet_operations ip6mr_net_ops = {
1367 	.init = ip6mr_net_init,
1368 	.exit = ip6mr_net_exit,
1369 	.exit_batch = ip6mr_net_exit_batch,
1370 };
1371 
1372 int __init ip6_mr_init(void)
1373 {
1374 	int err;
1375 
1376 	mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1377 				       sizeof(struct mfc6_cache),
1378 				       0, SLAB_HWCACHE_ALIGN,
1379 				       NULL);
1380 	if (!mrt_cachep)
1381 		return -ENOMEM;
1382 
1383 	err = register_pernet_subsys(&ip6mr_net_ops);
1384 	if (err)
1385 		goto reg_pernet_fail;
1386 
1387 	err = register_netdevice_notifier(&ip6_mr_notifier);
1388 	if (err)
1389 		goto reg_notif_fail;
1390 #ifdef CONFIG_IPV6_PIMSM_V2
1391 	if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1392 		pr_err("%s: can't add PIM protocol\n", __func__);
1393 		err = -EAGAIN;
1394 		goto add_proto_fail;
1395 	}
1396 #endif
1397 	err = rtnl_register_module(THIS_MODULE, RTNL_FAMILY_IP6MR, RTM_GETROUTE,
1398 				   ip6mr_rtm_getroute, ip6mr_rtm_dumproute, 0);
1399 	if (err == 0)
1400 		return 0;
1401 
1402 #ifdef CONFIG_IPV6_PIMSM_V2
1403 	inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1404 add_proto_fail:
1405 	unregister_netdevice_notifier(&ip6_mr_notifier);
1406 #endif
1407 reg_notif_fail:
1408 	unregister_pernet_subsys(&ip6mr_net_ops);
1409 reg_pernet_fail:
1410 	kmem_cache_destroy(mrt_cachep);
1411 	return err;
1412 }
1413 
1414 void ip6_mr_cleanup(void)
1415 {
1416 	rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1417 #ifdef CONFIG_IPV6_PIMSM_V2
1418 	inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1419 #endif
1420 	unregister_netdevice_notifier(&ip6_mr_notifier);
1421 	unregister_pernet_subsys(&ip6mr_net_ops);
1422 	kmem_cache_destroy(mrt_cachep);
1423 }
1424 
1425 static int ip6mr_mfc_add(struct net *net, struct mr_table *mrt,
1426 			 struct mf6cctl *mfc, int mrtsock, int parent)
1427 {
1428 	unsigned char ttls[MAXMIFS];
1429 	struct mfc6_cache *uc, *c;
1430 	struct mr_mfc *_uc;
1431 	bool found;
1432 	int i, err;
1433 
1434 	if (mfc->mf6cc_parent >= MAXMIFS)
1435 		return -ENFILE;
1436 
1437 	memset(ttls, 255, MAXMIFS);
1438 	for (i = 0; i < MAXMIFS; i++) {
1439 		if (IF_ISSET(i, &mfc->mf6cc_ifset))
1440 			ttls[i] = 1;
1441 	}
1442 
1443 	/* The entries are added/deleted only under RTNL */
1444 	rcu_read_lock();
1445 	c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr,
1446 				    &mfc->mf6cc_mcastgrp.sin6_addr, parent);
1447 	rcu_read_unlock();
1448 	if (c) {
1449 		spin_lock(&mrt_lock);
1450 		c->_c.mfc_parent = mfc->mf6cc_parent;
1451 		ip6mr_update_thresholds(mrt, &c->_c, ttls);
1452 		if (!mrtsock)
1453 			c->_c.mfc_flags |= MFC_STATIC;
1454 		spin_unlock(&mrt_lock);
1455 		call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE,
1456 					       c, mrt->id);
1457 		mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1458 		return 0;
1459 	}
1460 
1461 	if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1462 	    !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1463 		return -EINVAL;
1464 
1465 	c = ip6mr_cache_alloc();
1466 	if (!c)
1467 		return -ENOMEM;
1468 
1469 	c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1470 	c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1471 	c->_c.mfc_parent = mfc->mf6cc_parent;
1472 	ip6mr_update_thresholds(mrt, &c->_c, ttls);
1473 	if (!mrtsock)
1474 		c->_c.mfc_flags |= MFC_STATIC;
1475 
1476 	err = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1477 				  ip6mr_rht_params);
1478 	if (err) {
1479 		pr_err("ip6mr: rhtable insert error %d\n", err);
1480 		ip6mr_cache_free(c);
1481 		return err;
1482 	}
1483 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1484 
1485 	/* Check to see if we resolved a queued list. If so we
1486 	 * need to send on the frames and tidy up.
1487 	 */
1488 	found = false;
1489 	spin_lock_bh(&mfc_unres_lock);
1490 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1491 		uc = (struct mfc6_cache *)_uc;
1492 		if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1493 		    ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1494 			list_del(&_uc->list);
1495 			atomic_dec(&mrt->cache_resolve_queue_len);
1496 			found = true;
1497 			break;
1498 		}
1499 	}
1500 	if (list_empty(&mrt->mfc_unres_queue))
1501 		del_timer(&mrt->ipmr_expire_timer);
1502 	spin_unlock_bh(&mfc_unres_lock);
1503 
1504 	if (found) {
1505 		ip6mr_cache_resolve(net, mrt, uc, c);
1506 		ip6mr_cache_free(uc);
1507 	}
1508 	call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD,
1509 				       c, mrt->id);
1510 	mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1511 	return 0;
1512 }
1513 
1514 /*
1515  *	Close the multicast socket, and clear the vif tables etc
1516  */
1517 
1518 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1519 {
1520 	struct mr_mfc *c, *tmp;
1521 	LIST_HEAD(list);
1522 	int i;
1523 
1524 	/* Shut down all active vif entries */
1525 	if (flags & (MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC)) {
1526 		for (i = 0; i < mrt->maxvif; i++) {
1527 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1528 			     !(flags & MRT6_FLUSH_MIFS_STATIC)) ||
1529 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT6_FLUSH_MIFS)))
1530 				continue;
1531 			mif6_delete(mrt, i, 0, &list);
1532 		}
1533 		unregister_netdevice_many(&list);
1534 	}
1535 
1536 	/* Wipe the cache */
1537 	if (flags & (MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC)) {
1538 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1539 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC_STATIC)) ||
1540 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC)))
1541 				continue;
1542 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ip6mr_rht_params);
1543 			list_del_rcu(&c->list);
1544 			call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net),
1545 						       FIB_EVENT_ENTRY_DEL,
1546 						       (struct mfc6_cache *)c, mrt->id);
1547 			mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE);
1548 			mr_cache_put(c);
1549 		}
1550 	}
1551 
1552 	if (flags & MRT6_FLUSH_MFC) {
1553 		if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1554 			spin_lock_bh(&mfc_unres_lock);
1555 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1556 				list_del(&c->list);
1557 				mr6_netlink_event(mrt, (struct mfc6_cache *)c,
1558 						  RTM_DELROUTE);
1559 				ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c);
1560 			}
1561 			spin_unlock_bh(&mfc_unres_lock);
1562 		}
1563 	}
1564 }
1565 
1566 static int ip6mr_sk_init(struct mr_table *mrt, struct sock *sk)
1567 {
1568 	int err = 0;
1569 	struct net *net = sock_net(sk);
1570 
1571 	rtnl_lock();
1572 	spin_lock(&mrt_lock);
1573 	if (rtnl_dereference(mrt->mroute_sk)) {
1574 		err = -EADDRINUSE;
1575 	} else {
1576 		rcu_assign_pointer(mrt->mroute_sk, sk);
1577 		sock_set_flag(sk, SOCK_RCU_FREE);
1578 		atomic_inc(&net->ipv6.devconf_all->mc_forwarding);
1579 	}
1580 	spin_unlock(&mrt_lock);
1581 
1582 	if (!err)
1583 		inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1584 					     NETCONFA_MC_FORWARDING,
1585 					     NETCONFA_IFINDEX_ALL,
1586 					     net->ipv6.devconf_all);
1587 	rtnl_unlock();
1588 
1589 	return err;
1590 }
1591 
1592 int ip6mr_sk_done(struct sock *sk)
1593 {
1594 	struct net *net = sock_net(sk);
1595 	struct ipv6_devconf *devconf;
1596 	struct mr_table *mrt;
1597 	int err = -EACCES;
1598 
1599 	if (sk->sk_type != SOCK_RAW ||
1600 	    inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1601 		return err;
1602 
1603 	devconf = net->ipv6.devconf_all;
1604 	if (!devconf || !atomic_read(&devconf->mc_forwarding))
1605 		return err;
1606 
1607 	rtnl_lock();
1608 	ip6mr_for_each_table(mrt, net) {
1609 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1610 			spin_lock(&mrt_lock);
1611 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1612 			/* Note that mroute_sk had SOCK_RCU_FREE set,
1613 			 * so the RCU grace period before sk freeing
1614 			 * is guaranteed by sk_destruct()
1615 			 */
1616 			atomic_dec(&devconf->mc_forwarding);
1617 			spin_unlock(&mrt_lock);
1618 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1619 						     NETCONFA_MC_FORWARDING,
1620 						     NETCONFA_IFINDEX_ALL,
1621 						     net->ipv6.devconf_all);
1622 
1623 			mroute_clean_tables(mrt, MRT6_FLUSH_MIFS | MRT6_FLUSH_MFC);
1624 			err = 0;
1625 			break;
1626 		}
1627 	}
1628 	rtnl_unlock();
1629 
1630 	return err;
1631 }
1632 
1633 bool mroute6_is_socket(struct net *net, struct sk_buff *skb)
1634 {
1635 	struct mr_table *mrt;
1636 	struct flowi6 fl6 = {
1637 		.flowi6_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
1638 		.flowi6_oif	= skb->dev->ifindex,
1639 		.flowi6_mark	= skb->mark,
1640 	};
1641 
1642 	if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1643 		return NULL;
1644 
1645 	return rcu_access_pointer(mrt->mroute_sk);
1646 }
1647 EXPORT_SYMBOL(mroute6_is_socket);
1648 
1649 /*
1650  *	Socket options and virtual interface manipulation. The whole
1651  *	virtual interface system is a complete heap, but unfortunately
1652  *	that's how BSD mrouted happens to think. Maybe one day with a proper
1653  *	MOSPF/PIM router set up we can clean this up.
1654  */
1655 
1656 int ip6_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1657 			  unsigned int optlen)
1658 {
1659 	int ret, parent = 0;
1660 	struct mif6ctl vif;
1661 	struct mf6cctl mfc;
1662 	mifi_t mifi;
1663 	struct net *net = sock_net(sk);
1664 	struct mr_table *mrt;
1665 
1666 	if (sk->sk_type != SOCK_RAW ||
1667 	    inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1668 		return -EOPNOTSUPP;
1669 
1670 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1671 	if (!mrt)
1672 		return -ENOENT;
1673 
1674 	if (optname != MRT6_INIT) {
1675 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1676 		    !ns_capable(net->user_ns, CAP_NET_ADMIN))
1677 			return -EACCES;
1678 	}
1679 
1680 	switch (optname) {
1681 	case MRT6_INIT:
1682 		if (optlen < sizeof(int))
1683 			return -EINVAL;
1684 
1685 		return ip6mr_sk_init(mrt, sk);
1686 
1687 	case MRT6_DONE:
1688 		return ip6mr_sk_done(sk);
1689 
1690 	case MRT6_ADD_MIF:
1691 		if (optlen < sizeof(vif))
1692 			return -EINVAL;
1693 		if (copy_from_sockptr(&vif, optval, sizeof(vif)))
1694 			return -EFAULT;
1695 		if (vif.mif6c_mifi >= MAXMIFS)
1696 			return -ENFILE;
1697 		rtnl_lock();
1698 		ret = mif6_add(net, mrt, &vif,
1699 			       sk == rtnl_dereference(mrt->mroute_sk));
1700 		rtnl_unlock();
1701 		return ret;
1702 
1703 	case MRT6_DEL_MIF:
1704 		if (optlen < sizeof(mifi_t))
1705 			return -EINVAL;
1706 		if (copy_from_sockptr(&mifi, optval, sizeof(mifi_t)))
1707 			return -EFAULT;
1708 		rtnl_lock();
1709 		ret = mif6_delete(mrt, mifi, 0, NULL);
1710 		rtnl_unlock();
1711 		return ret;
1712 
1713 	/*
1714 	 *	Manipulate the forwarding caches. These live
1715 	 *	in a sort of kernel/user symbiosis.
1716 	 */
1717 	case MRT6_ADD_MFC:
1718 	case MRT6_DEL_MFC:
1719 		parent = -1;
1720 		fallthrough;
1721 	case MRT6_ADD_MFC_PROXY:
1722 	case MRT6_DEL_MFC_PROXY:
1723 		if (optlen < sizeof(mfc))
1724 			return -EINVAL;
1725 		if (copy_from_sockptr(&mfc, optval, sizeof(mfc)))
1726 			return -EFAULT;
1727 		if (parent == 0)
1728 			parent = mfc.mf6cc_parent;
1729 		rtnl_lock();
1730 		if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1731 			ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1732 		else
1733 			ret = ip6mr_mfc_add(net, mrt, &mfc,
1734 					    sk ==
1735 					    rtnl_dereference(mrt->mroute_sk),
1736 					    parent);
1737 		rtnl_unlock();
1738 		return ret;
1739 
1740 	case MRT6_FLUSH:
1741 	{
1742 		int flags;
1743 
1744 		if (optlen != sizeof(flags))
1745 			return -EINVAL;
1746 		if (copy_from_sockptr(&flags, optval, sizeof(flags)))
1747 			return -EFAULT;
1748 		rtnl_lock();
1749 		mroute_clean_tables(mrt, flags);
1750 		rtnl_unlock();
1751 		return 0;
1752 	}
1753 
1754 	/*
1755 	 *	Control PIM assert (to activate pim will activate assert)
1756 	 */
1757 	case MRT6_ASSERT:
1758 	{
1759 		int v;
1760 
1761 		if (optlen != sizeof(v))
1762 			return -EINVAL;
1763 		if (copy_from_sockptr(&v, optval, sizeof(v)))
1764 			return -EFAULT;
1765 		mrt->mroute_do_assert = v;
1766 		return 0;
1767 	}
1768 
1769 #ifdef CONFIG_IPV6_PIMSM_V2
1770 	case MRT6_PIM:
1771 	{
1772 		bool do_wrmifwhole;
1773 		int v;
1774 
1775 		if (optlen != sizeof(v))
1776 			return -EINVAL;
1777 		if (copy_from_sockptr(&v, optval, sizeof(v)))
1778 			return -EFAULT;
1779 
1780 		do_wrmifwhole = (v == MRT6MSG_WRMIFWHOLE);
1781 		v = !!v;
1782 		rtnl_lock();
1783 		ret = 0;
1784 		if (v != mrt->mroute_do_pim) {
1785 			mrt->mroute_do_pim = v;
1786 			mrt->mroute_do_assert = v;
1787 			mrt->mroute_do_wrvifwhole = do_wrmifwhole;
1788 		}
1789 		rtnl_unlock();
1790 		return ret;
1791 	}
1792 
1793 #endif
1794 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1795 	case MRT6_TABLE:
1796 	{
1797 		u32 v;
1798 
1799 		if (optlen != sizeof(u32))
1800 			return -EINVAL;
1801 		if (copy_from_sockptr(&v, optval, sizeof(v)))
1802 			return -EFAULT;
1803 		/* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1804 		if (v != RT_TABLE_DEFAULT && v >= 100000000)
1805 			return -EINVAL;
1806 		if (sk == rcu_access_pointer(mrt->mroute_sk))
1807 			return -EBUSY;
1808 
1809 		rtnl_lock();
1810 		ret = 0;
1811 		mrt = ip6mr_new_table(net, v);
1812 		if (IS_ERR(mrt))
1813 			ret = PTR_ERR(mrt);
1814 		else
1815 			raw6_sk(sk)->ip6mr_table = v;
1816 		rtnl_unlock();
1817 		return ret;
1818 	}
1819 #endif
1820 	/*
1821 	 *	Spurious command, or MRT6_VERSION which you cannot
1822 	 *	set.
1823 	 */
1824 	default:
1825 		return -ENOPROTOOPT;
1826 	}
1827 }
1828 
1829 /*
1830  *	Getsock opt support for the multicast routing system.
1831  */
1832 
1833 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1834 			  int __user *optlen)
1835 {
1836 	int olr;
1837 	int val;
1838 	struct net *net = sock_net(sk);
1839 	struct mr_table *mrt;
1840 
1841 	if (sk->sk_type != SOCK_RAW ||
1842 	    inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1843 		return -EOPNOTSUPP;
1844 
1845 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1846 	if (!mrt)
1847 		return -ENOENT;
1848 
1849 	switch (optname) {
1850 	case MRT6_VERSION:
1851 		val = 0x0305;
1852 		break;
1853 #ifdef CONFIG_IPV6_PIMSM_V2
1854 	case MRT6_PIM:
1855 		val = mrt->mroute_do_pim;
1856 		break;
1857 #endif
1858 	case MRT6_ASSERT:
1859 		val = mrt->mroute_do_assert;
1860 		break;
1861 	default:
1862 		return -ENOPROTOOPT;
1863 	}
1864 
1865 	if (get_user(olr, optlen))
1866 		return -EFAULT;
1867 
1868 	olr = min_t(int, olr, sizeof(int));
1869 	if (olr < 0)
1870 		return -EINVAL;
1871 
1872 	if (put_user(olr, optlen))
1873 		return -EFAULT;
1874 	if (copy_to_user(optval, &val, olr))
1875 		return -EFAULT;
1876 	return 0;
1877 }
1878 
1879 /*
1880  *	The IP multicast ioctl support routines.
1881  */
1882 
1883 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1884 {
1885 	struct sioc_sg_req6 sr;
1886 	struct sioc_mif_req6 vr;
1887 	struct vif_device *vif;
1888 	struct mfc6_cache *c;
1889 	struct net *net = sock_net(sk);
1890 	struct mr_table *mrt;
1891 
1892 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1893 	if (!mrt)
1894 		return -ENOENT;
1895 
1896 	switch (cmd) {
1897 	case SIOCGETMIFCNT_IN6:
1898 		if (copy_from_user(&vr, arg, sizeof(vr)))
1899 			return -EFAULT;
1900 		if (vr.mifi >= mrt->maxvif)
1901 			return -EINVAL;
1902 		vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif);
1903 		rcu_read_lock();
1904 		vif = &mrt->vif_table[vr.mifi];
1905 		if (VIF_EXISTS(mrt, vr.mifi)) {
1906 			vr.icount = READ_ONCE(vif->pkt_in);
1907 			vr.ocount = READ_ONCE(vif->pkt_out);
1908 			vr.ibytes = READ_ONCE(vif->bytes_in);
1909 			vr.obytes = READ_ONCE(vif->bytes_out);
1910 			rcu_read_unlock();
1911 
1912 			if (copy_to_user(arg, &vr, sizeof(vr)))
1913 				return -EFAULT;
1914 			return 0;
1915 		}
1916 		rcu_read_unlock();
1917 		return -EADDRNOTAVAIL;
1918 	case SIOCGETSGCNT_IN6:
1919 		if (copy_from_user(&sr, arg, sizeof(sr)))
1920 			return -EFAULT;
1921 
1922 		rcu_read_lock();
1923 		c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1924 		if (c) {
1925 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1926 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1927 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1928 			rcu_read_unlock();
1929 
1930 			if (copy_to_user(arg, &sr, sizeof(sr)))
1931 				return -EFAULT;
1932 			return 0;
1933 		}
1934 		rcu_read_unlock();
1935 		return -EADDRNOTAVAIL;
1936 	default:
1937 		return -ENOIOCTLCMD;
1938 	}
1939 }
1940 
1941 #ifdef CONFIG_COMPAT
1942 struct compat_sioc_sg_req6 {
1943 	struct sockaddr_in6 src;
1944 	struct sockaddr_in6 grp;
1945 	compat_ulong_t pktcnt;
1946 	compat_ulong_t bytecnt;
1947 	compat_ulong_t wrong_if;
1948 };
1949 
1950 struct compat_sioc_mif_req6 {
1951 	mifi_t	mifi;
1952 	compat_ulong_t icount;
1953 	compat_ulong_t ocount;
1954 	compat_ulong_t ibytes;
1955 	compat_ulong_t obytes;
1956 };
1957 
1958 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1959 {
1960 	struct compat_sioc_sg_req6 sr;
1961 	struct compat_sioc_mif_req6 vr;
1962 	struct vif_device *vif;
1963 	struct mfc6_cache *c;
1964 	struct net *net = sock_net(sk);
1965 	struct mr_table *mrt;
1966 
1967 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1968 	if (!mrt)
1969 		return -ENOENT;
1970 
1971 	switch (cmd) {
1972 	case SIOCGETMIFCNT_IN6:
1973 		if (copy_from_user(&vr, arg, sizeof(vr)))
1974 			return -EFAULT;
1975 		if (vr.mifi >= mrt->maxvif)
1976 			return -EINVAL;
1977 		vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif);
1978 		rcu_read_lock();
1979 		vif = &mrt->vif_table[vr.mifi];
1980 		if (VIF_EXISTS(mrt, vr.mifi)) {
1981 			vr.icount = READ_ONCE(vif->pkt_in);
1982 			vr.ocount = READ_ONCE(vif->pkt_out);
1983 			vr.ibytes = READ_ONCE(vif->bytes_in);
1984 			vr.obytes = READ_ONCE(vif->bytes_out);
1985 			rcu_read_unlock();
1986 
1987 			if (copy_to_user(arg, &vr, sizeof(vr)))
1988 				return -EFAULT;
1989 			return 0;
1990 		}
1991 		rcu_read_unlock();
1992 		return -EADDRNOTAVAIL;
1993 	case SIOCGETSGCNT_IN6:
1994 		if (copy_from_user(&sr, arg, sizeof(sr)))
1995 			return -EFAULT;
1996 
1997 		rcu_read_lock();
1998 		c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1999 		if (c) {
2000 			sr.pktcnt = c->_c.mfc_un.res.pkt;
2001 			sr.bytecnt = c->_c.mfc_un.res.bytes;
2002 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
2003 			rcu_read_unlock();
2004 
2005 			if (copy_to_user(arg, &sr, sizeof(sr)))
2006 				return -EFAULT;
2007 			return 0;
2008 		}
2009 		rcu_read_unlock();
2010 		return -EADDRNOTAVAIL;
2011 	default:
2012 		return -ENOIOCTLCMD;
2013 	}
2014 }
2015 #endif
2016 
2017 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
2018 {
2019 	IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
2020 		      IPSTATS_MIB_OUTFORWDATAGRAMS);
2021 	IP6_ADD_STATS(net, ip6_dst_idev(skb_dst(skb)),
2022 		      IPSTATS_MIB_OUTOCTETS, skb->len);
2023 	return dst_output(net, sk, skb);
2024 }
2025 
2026 /*
2027  *	Processing handlers for ip6mr_forward
2028  */
2029 
2030 static int ip6mr_forward2(struct net *net, struct mr_table *mrt,
2031 			  struct sk_buff *skb, int vifi)
2032 {
2033 	struct vif_device *vif = &mrt->vif_table[vifi];
2034 	struct net_device *vif_dev;
2035 	struct ipv6hdr *ipv6h;
2036 	struct dst_entry *dst;
2037 	struct flowi6 fl6;
2038 
2039 	vif_dev = vif_dev_read(vif);
2040 	if (!vif_dev)
2041 		goto out_free;
2042 
2043 #ifdef CONFIG_IPV6_PIMSM_V2
2044 	if (vif->flags & MIFF_REGISTER) {
2045 		WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
2046 		WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
2047 		vif_dev->stats.tx_bytes += skb->len;
2048 		vif_dev->stats.tx_packets++;
2049 		ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2050 		goto out_free;
2051 	}
2052 #endif
2053 
2054 	ipv6h = ipv6_hdr(skb);
2055 
2056 	fl6 = (struct flowi6) {
2057 		.flowi6_oif = vif->link,
2058 		.daddr = ipv6h->daddr,
2059 	};
2060 
2061 	dst = ip6_route_output(net, NULL, &fl6);
2062 	if (dst->error) {
2063 		dst_release(dst);
2064 		goto out_free;
2065 	}
2066 
2067 	skb_dst_drop(skb);
2068 	skb_dst_set(skb, dst);
2069 
2070 	/*
2071 	 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2072 	 * not only before forwarding, but after forwarding on all output
2073 	 * interfaces. It is clear, if mrouter runs a multicasting
2074 	 * program, it should receive packets not depending to what interface
2075 	 * program is joined.
2076 	 * If we will not make it, the program will have to join on all
2077 	 * interfaces. On the other hand, multihoming host (or router, but
2078 	 * not mrouter) cannot join to more than one interface - it will
2079 	 * result in receiving multiple packets.
2080 	 */
2081 	skb->dev = vif_dev;
2082 	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
2083 	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
2084 
2085 	/* We are about to write */
2086 	/* XXX: extension headers? */
2087 	if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(vif_dev)))
2088 		goto out_free;
2089 
2090 	ipv6h = ipv6_hdr(skb);
2091 	ipv6h->hop_limit--;
2092 
2093 	IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2094 
2095 	return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2096 		       net, NULL, skb, skb->dev, vif_dev,
2097 		       ip6mr_forward2_finish);
2098 
2099 out_free:
2100 	kfree_skb(skb);
2101 	return 0;
2102 }
2103 
2104 /* Called with rcu_read_lock() */
2105 static int ip6mr_find_vif(struct mr_table *mrt, struct net_device *dev)
2106 {
2107 	int ct;
2108 
2109 	/* Pairs with WRITE_ONCE() in mif6_delete()/mif6_add() */
2110 	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
2111 		if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
2112 			break;
2113 	}
2114 	return ct;
2115 }
2116 
2117 /* Called under rcu_read_lock() */
2118 static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
2119 			   struct net_device *dev, struct sk_buff *skb,
2120 			   struct mfc6_cache *c)
2121 {
2122 	int psend = -1;
2123 	int vif, ct;
2124 	int true_vifi = ip6mr_find_vif(mrt, dev);
2125 
2126 	vif = c->_c.mfc_parent;
2127 	c->_c.mfc_un.res.pkt++;
2128 	c->_c.mfc_un.res.bytes += skb->len;
2129 	c->_c.mfc_un.res.lastuse = jiffies;
2130 
2131 	if (ipv6_addr_any(&c->mf6c_origin) && true_vifi >= 0) {
2132 		struct mfc6_cache *cache_proxy;
2133 
2134 		/* For an (*,G) entry, we only check that the incoming
2135 		 * interface is part of the static tree.
2136 		 */
2137 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2138 		if (cache_proxy &&
2139 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
2140 			goto forward;
2141 	}
2142 
2143 	/*
2144 	 * Wrong interface: drop packet and (maybe) send PIM assert.
2145 	 */
2146 	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
2147 		c->_c.mfc_un.res.wrong_if++;
2148 
2149 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
2150 		    /* pimsm uses asserts, when switching from RPT to SPT,
2151 		       so that we cannot check that packet arrived on an oif.
2152 		       It is bad, but otherwise we would need to move pretty
2153 		       large chunk of pimd to kernel. Ough... --ANK
2154 		     */
2155 		    (mrt->mroute_do_pim ||
2156 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2157 		    time_after(jiffies,
2158 			       c->_c.mfc_un.res.last_assert +
2159 			       MFC_ASSERT_THRESH)) {
2160 			c->_c.mfc_un.res.last_assert = jiffies;
2161 			ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2162 			if (mrt->mroute_do_wrvifwhole)
2163 				ip6mr_cache_report(mrt, skb, true_vifi,
2164 						   MRT6MSG_WRMIFWHOLE);
2165 		}
2166 		goto dont_forward;
2167 	}
2168 
2169 forward:
2170 	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2171 		   mrt->vif_table[vif].pkt_in + 1);
2172 	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2173 		   mrt->vif_table[vif].bytes_in + skb->len);
2174 
2175 	/*
2176 	 *	Forward the frame
2177 	 */
2178 	if (ipv6_addr_any(&c->mf6c_origin) &&
2179 	    ipv6_addr_any(&c->mf6c_mcastgrp)) {
2180 		if (true_vifi >= 0 &&
2181 		    true_vifi != c->_c.mfc_parent &&
2182 		    ipv6_hdr(skb)->hop_limit >
2183 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2184 			/* It's an (*,*) entry and the packet is not coming from
2185 			 * the upstream: forward the packet to the upstream
2186 			 * only.
2187 			 */
2188 			psend = c->_c.mfc_parent;
2189 			goto last_forward;
2190 		}
2191 		goto dont_forward;
2192 	}
2193 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2194 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2195 		/* For (*,G) entry, don't forward to the incoming interface */
2196 		if ((!ipv6_addr_any(&c->mf6c_origin) || ct != true_vifi) &&
2197 		    ipv6_hdr(skb)->hop_limit > c->_c.mfc_un.res.ttls[ct]) {
2198 			if (psend != -1) {
2199 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2200 				if (skb2)
2201 					ip6mr_forward2(net, mrt, skb2, psend);
2202 			}
2203 			psend = ct;
2204 		}
2205 	}
2206 last_forward:
2207 	if (psend != -1) {
2208 		ip6mr_forward2(net, mrt, skb, psend);
2209 		return;
2210 	}
2211 
2212 dont_forward:
2213 	kfree_skb(skb);
2214 }
2215 
2216 
2217 /*
2218  *	Multicast packets for forwarding arrive here
2219  */
2220 
2221 int ip6_mr_input(struct sk_buff *skb)
2222 {
2223 	struct mfc6_cache *cache;
2224 	struct net *net = dev_net(skb->dev);
2225 	struct mr_table *mrt;
2226 	struct flowi6 fl6 = {
2227 		.flowi6_iif	= skb->dev->ifindex,
2228 		.flowi6_mark	= skb->mark,
2229 	};
2230 	int err;
2231 	struct net_device *dev;
2232 
2233 	/* skb->dev passed in is the master dev for vrfs.
2234 	 * Get the proper interface that does have a vif associated with it.
2235 	 */
2236 	dev = skb->dev;
2237 	if (netif_is_l3_master(skb->dev)) {
2238 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2239 		if (!dev) {
2240 			kfree_skb(skb);
2241 			return -ENODEV;
2242 		}
2243 	}
2244 
2245 	err = ip6mr_fib_lookup(net, &fl6, &mrt);
2246 	if (err < 0) {
2247 		kfree_skb(skb);
2248 		return err;
2249 	}
2250 
2251 	cache = ip6mr_cache_find(mrt,
2252 				 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2253 	if (!cache) {
2254 		int vif = ip6mr_find_vif(mrt, dev);
2255 
2256 		if (vif >= 0)
2257 			cache = ip6mr_cache_find_any(mrt,
2258 						     &ipv6_hdr(skb)->daddr,
2259 						     vif);
2260 	}
2261 
2262 	/*
2263 	 *	No usable cache entry
2264 	 */
2265 	if (!cache) {
2266 		int vif;
2267 
2268 		vif = ip6mr_find_vif(mrt, dev);
2269 		if (vif >= 0) {
2270 			int err = ip6mr_cache_unresolved(mrt, vif, skb, dev);
2271 
2272 			return err;
2273 		}
2274 		kfree_skb(skb);
2275 		return -ENODEV;
2276 	}
2277 
2278 	ip6_mr_forward(net, mrt, dev, skb, cache);
2279 
2280 	return 0;
2281 }
2282 
2283 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2284 		    u32 portid)
2285 {
2286 	int err;
2287 	struct mr_table *mrt;
2288 	struct mfc6_cache *cache;
2289 	struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2290 
2291 	mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2292 	if (!mrt)
2293 		return -ENOENT;
2294 
2295 	rcu_read_lock();
2296 	cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2297 	if (!cache && skb->dev) {
2298 		int vif = ip6mr_find_vif(mrt, skb->dev);
2299 
2300 		if (vif >= 0)
2301 			cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2302 						     vif);
2303 	}
2304 
2305 	if (!cache) {
2306 		struct sk_buff *skb2;
2307 		struct ipv6hdr *iph;
2308 		struct net_device *dev;
2309 		int vif;
2310 
2311 		dev = skb->dev;
2312 		if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2313 			rcu_read_unlock();
2314 			return -ENODEV;
2315 		}
2316 
2317 		/* really correct? */
2318 		skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2319 		if (!skb2) {
2320 			rcu_read_unlock();
2321 			return -ENOMEM;
2322 		}
2323 
2324 		NETLINK_CB(skb2).portid = portid;
2325 		skb_reset_transport_header(skb2);
2326 
2327 		skb_put(skb2, sizeof(struct ipv6hdr));
2328 		skb_reset_network_header(skb2);
2329 
2330 		iph = ipv6_hdr(skb2);
2331 		iph->version = 0;
2332 		iph->priority = 0;
2333 		iph->flow_lbl[0] = 0;
2334 		iph->flow_lbl[1] = 0;
2335 		iph->flow_lbl[2] = 0;
2336 		iph->payload_len = 0;
2337 		iph->nexthdr = IPPROTO_NONE;
2338 		iph->hop_limit = 0;
2339 		iph->saddr = rt->rt6i_src.addr;
2340 		iph->daddr = rt->rt6i_dst.addr;
2341 
2342 		err = ip6mr_cache_unresolved(mrt, vif, skb2, dev);
2343 		rcu_read_unlock();
2344 
2345 		return err;
2346 	}
2347 
2348 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2349 	rcu_read_unlock();
2350 	return err;
2351 }
2352 
2353 static int ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2354 			     u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2355 			     int flags)
2356 {
2357 	struct nlmsghdr *nlh;
2358 	struct rtmsg *rtm;
2359 	int err;
2360 
2361 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2362 	if (!nlh)
2363 		return -EMSGSIZE;
2364 
2365 	rtm = nlmsg_data(nlh);
2366 	rtm->rtm_family   = RTNL_FAMILY_IP6MR;
2367 	rtm->rtm_dst_len  = 128;
2368 	rtm->rtm_src_len  = 128;
2369 	rtm->rtm_tos      = 0;
2370 	rtm->rtm_table    = mrt->id;
2371 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2372 		goto nla_put_failure;
2373 	rtm->rtm_type = RTN_MULTICAST;
2374 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2375 	if (c->_c.mfc_flags & MFC_STATIC)
2376 		rtm->rtm_protocol = RTPROT_STATIC;
2377 	else
2378 		rtm->rtm_protocol = RTPROT_MROUTED;
2379 	rtm->rtm_flags    = 0;
2380 
2381 	if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2382 	    nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2383 		goto nla_put_failure;
2384 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2385 	/* do not break the dump if cache is unresolved */
2386 	if (err < 0 && err != -ENOENT)
2387 		goto nla_put_failure;
2388 
2389 	nlmsg_end(skb, nlh);
2390 	return 0;
2391 
2392 nla_put_failure:
2393 	nlmsg_cancel(skb, nlh);
2394 	return -EMSGSIZE;
2395 }
2396 
2397 static int _ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2398 			      u32 portid, u32 seq, struct mr_mfc *c,
2399 			      int cmd, int flags)
2400 {
2401 	return ip6mr_fill_mroute(mrt, skb, portid, seq, (struct mfc6_cache *)c,
2402 				 cmd, flags);
2403 }
2404 
2405 static int mr6_msgsize(bool unresolved, int maxvif)
2406 {
2407 	size_t len =
2408 		NLMSG_ALIGN(sizeof(struct rtmsg))
2409 		+ nla_total_size(4)	/* RTA_TABLE */
2410 		+ nla_total_size(sizeof(struct in6_addr))	/* RTA_SRC */
2411 		+ nla_total_size(sizeof(struct in6_addr))	/* RTA_DST */
2412 		;
2413 
2414 	if (!unresolved)
2415 		len = len
2416 		      + nla_total_size(4)	/* RTA_IIF */
2417 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2418 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2419 						/* RTA_MFC_STATS */
2420 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2421 		;
2422 
2423 	return len;
2424 }
2425 
2426 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc,
2427 			      int cmd)
2428 {
2429 	struct net *net = read_pnet(&mrt->net);
2430 	struct sk_buff *skb;
2431 	int err = -ENOBUFS;
2432 
2433 	skb = nlmsg_new(mr6_msgsize(mfc->_c.mfc_parent >= MAXMIFS, mrt->maxvif),
2434 			GFP_ATOMIC);
2435 	if (!skb)
2436 		goto errout;
2437 
2438 	err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2439 	if (err < 0)
2440 		goto errout;
2441 
2442 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2443 	return;
2444 
2445 errout:
2446 	kfree_skb(skb);
2447 	if (err < 0)
2448 		rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2449 }
2450 
2451 static size_t mrt6msg_netlink_msgsize(size_t payloadlen)
2452 {
2453 	size_t len =
2454 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2455 		+ nla_total_size(1)	/* IP6MRA_CREPORT_MSGTYPE */
2456 		+ nla_total_size(4)	/* IP6MRA_CREPORT_MIF_ID */
2457 					/* IP6MRA_CREPORT_SRC_ADDR */
2458 		+ nla_total_size(sizeof(struct in6_addr))
2459 					/* IP6MRA_CREPORT_DST_ADDR */
2460 		+ nla_total_size(sizeof(struct in6_addr))
2461 					/* IP6MRA_CREPORT_PKT */
2462 		+ nla_total_size(payloadlen)
2463 		;
2464 
2465 	return len;
2466 }
2467 
2468 static void mrt6msg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2469 {
2470 	struct net *net = read_pnet(&mrt->net);
2471 	struct nlmsghdr *nlh;
2472 	struct rtgenmsg *rtgenm;
2473 	struct mrt6msg *msg;
2474 	struct sk_buff *skb;
2475 	struct nlattr *nla;
2476 	int payloadlen;
2477 
2478 	payloadlen = pkt->len - sizeof(struct mrt6msg);
2479 	msg = (struct mrt6msg *)skb_transport_header(pkt);
2480 
2481 	skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2482 	if (!skb)
2483 		goto errout;
2484 
2485 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2486 			sizeof(struct rtgenmsg), 0);
2487 	if (!nlh)
2488 		goto errout;
2489 	rtgenm = nlmsg_data(nlh);
2490 	rtgenm->rtgen_family = RTNL_FAMILY_IP6MR;
2491 	if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) ||
2492 	    nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) ||
2493 	    nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR,
2494 			     &msg->im6_src) ||
2495 	    nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR,
2496 			     &msg->im6_dst))
2497 		goto nla_put_failure;
2498 
2499 	nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen);
2500 	if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg),
2501 				  nla_data(nla), payloadlen))
2502 		goto nla_put_failure;
2503 
2504 	nlmsg_end(skb, nlh);
2505 
2506 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC);
2507 	return;
2508 
2509 nla_put_failure:
2510 	nlmsg_cancel(skb, nlh);
2511 errout:
2512 	kfree_skb(skb);
2513 	rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS);
2514 }
2515 
2516 static const struct nla_policy ip6mr_getroute_policy[RTA_MAX + 1] = {
2517 	[RTA_SRC]		= NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
2518 	[RTA_DST]		= NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
2519 	[RTA_TABLE]		= { .type = NLA_U32 },
2520 };
2521 
2522 static int ip6mr_rtm_valid_getroute_req(struct sk_buff *skb,
2523 					const struct nlmsghdr *nlh,
2524 					struct nlattr **tb,
2525 					struct netlink_ext_ack *extack)
2526 {
2527 	struct rtmsg *rtm;
2528 	int err;
2529 
2530 	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, ip6mr_getroute_policy,
2531 			  extack);
2532 	if (err)
2533 		return err;
2534 
2535 	rtm = nlmsg_data(nlh);
2536 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
2537 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
2538 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2539 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2540 		NL_SET_ERR_MSG_MOD(extack,
2541 				   "Invalid values in header for multicast route get request");
2542 		return -EINVAL;
2543 	}
2544 
2545 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2546 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2547 		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
2548 		return -EINVAL;
2549 	}
2550 
2551 	return 0;
2552 }
2553 
2554 static int ip6mr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2555 			      struct netlink_ext_ack *extack)
2556 {
2557 	struct net *net = sock_net(in_skb->sk);
2558 	struct in6_addr src = {}, grp = {};
2559 	struct nlattr *tb[RTA_MAX + 1];
2560 	struct mfc6_cache *cache;
2561 	struct mr_table *mrt;
2562 	struct sk_buff *skb;
2563 	u32 tableid;
2564 	int err;
2565 
2566 	err = ip6mr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2567 	if (err < 0)
2568 		return err;
2569 
2570 	if (tb[RTA_SRC])
2571 		src = nla_get_in6_addr(tb[RTA_SRC]);
2572 	if (tb[RTA_DST])
2573 		grp = nla_get_in6_addr(tb[RTA_DST]);
2574 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2575 
2576 	mrt = ip6mr_get_table(net, tableid ?: RT_TABLE_DEFAULT);
2577 	if (!mrt) {
2578 		NL_SET_ERR_MSG_MOD(extack, "MR table does not exist");
2579 		return -ENOENT;
2580 	}
2581 
2582 	/* entries are added/deleted only under RTNL */
2583 	rcu_read_lock();
2584 	cache = ip6mr_cache_find(mrt, &src, &grp);
2585 	rcu_read_unlock();
2586 	if (!cache) {
2587 		NL_SET_ERR_MSG_MOD(extack, "MR cache entry not found");
2588 		return -ENOENT;
2589 	}
2590 
2591 	skb = nlmsg_new(mr6_msgsize(false, mrt->maxvif), GFP_KERNEL);
2592 	if (!skb)
2593 		return -ENOBUFS;
2594 
2595 	err = ip6mr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2596 				nlh->nlmsg_seq, cache, RTM_NEWROUTE, 0);
2597 	if (err < 0) {
2598 		kfree_skb(skb);
2599 		return err;
2600 	}
2601 
2602 	return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2603 }
2604 
2605 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2606 {
2607 	const struct nlmsghdr *nlh = cb->nlh;
2608 	struct fib_dump_filter filter = {};
2609 	int err;
2610 
2611 	if (cb->strict_check) {
2612 		err = ip_valid_fib_dump_req(sock_net(skb->sk), nlh,
2613 					    &filter, cb);
2614 		if (err < 0)
2615 			return err;
2616 	}
2617 
2618 	if (filter.table_id) {
2619 		struct mr_table *mrt;
2620 
2621 		mrt = ip6mr_get_table(sock_net(skb->sk), filter.table_id);
2622 		if (!mrt) {
2623 			if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IP6MR)
2624 				return skb->len;
2625 
2626 			NL_SET_ERR_MSG_MOD(cb->extack, "MR table does not exist");
2627 			return -ENOENT;
2628 		}
2629 		err = mr_table_dump(mrt, skb, cb, _ip6mr_fill_mroute,
2630 				    &mfc_unres_lock, &filter);
2631 		return skb->len ? : err;
2632 	}
2633 
2634 	return mr_rtm_dumproute(skb, cb, ip6mr_mr_table_iter,
2635 				_ip6mr_fill_mroute, &mfc_unres_lock, &filter);
2636 }
2637