xref: /linux/net/ipv6/ip6mr.c (revision 78c3925c048c752334873f56c3a3d1c9d53e0416)
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);
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 	timer_shutdown_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_ADD(dev, tx_bytes, skb->len);
612 	DEV_STATS_INC(dev, 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_INC(dev, 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_pull(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(mfc6_cache, SLAB_HWCACHE_ALIGN);
1377 	if (!mrt_cachep)
1378 		return -ENOMEM;
1379 
1380 	err = register_pernet_subsys(&ip6mr_net_ops);
1381 	if (err)
1382 		goto reg_pernet_fail;
1383 
1384 	err = register_netdevice_notifier(&ip6_mr_notifier);
1385 	if (err)
1386 		goto reg_notif_fail;
1387 #ifdef CONFIG_IPV6_PIMSM_V2
1388 	if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1389 		pr_err("%s: can't add PIM protocol\n", __func__);
1390 		err = -EAGAIN;
1391 		goto add_proto_fail;
1392 	}
1393 #endif
1394 	err = rtnl_register_module(THIS_MODULE, RTNL_FAMILY_IP6MR, RTM_GETROUTE,
1395 				   ip6mr_rtm_getroute, ip6mr_rtm_dumproute, 0);
1396 	if (err == 0)
1397 		return 0;
1398 
1399 #ifdef CONFIG_IPV6_PIMSM_V2
1400 	inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1401 add_proto_fail:
1402 	unregister_netdevice_notifier(&ip6_mr_notifier);
1403 #endif
1404 reg_notif_fail:
1405 	unregister_pernet_subsys(&ip6mr_net_ops);
1406 reg_pernet_fail:
1407 	kmem_cache_destroy(mrt_cachep);
1408 	return err;
1409 }
1410 
1411 void ip6_mr_cleanup(void)
1412 {
1413 	rtnl_unregister(RTNL_FAMILY_IP6MR, RTM_GETROUTE);
1414 #ifdef CONFIG_IPV6_PIMSM_V2
1415 	inet6_del_protocol(&pim6_protocol, IPPROTO_PIM);
1416 #endif
1417 	unregister_netdevice_notifier(&ip6_mr_notifier);
1418 	unregister_pernet_subsys(&ip6mr_net_ops);
1419 	kmem_cache_destroy(mrt_cachep);
1420 }
1421 
1422 static int ip6mr_mfc_add(struct net *net, struct mr_table *mrt,
1423 			 struct mf6cctl *mfc, int mrtsock, int parent)
1424 {
1425 	unsigned char ttls[MAXMIFS];
1426 	struct mfc6_cache *uc, *c;
1427 	struct mr_mfc *_uc;
1428 	bool found;
1429 	int i, err;
1430 
1431 	if (mfc->mf6cc_parent >= MAXMIFS)
1432 		return -ENFILE;
1433 
1434 	memset(ttls, 255, MAXMIFS);
1435 	for (i = 0; i < MAXMIFS; i++) {
1436 		if (IF_ISSET(i, &mfc->mf6cc_ifset))
1437 			ttls[i] = 1;
1438 	}
1439 
1440 	/* The entries are added/deleted only under RTNL */
1441 	rcu_read_lock();
1442 	c = ip6mr_cache_find_parent(mrt, &mfc->mf6cc_origin.sin6_addr,
1443 				    &mfc->mf6cc_mcastgrp.sin6_addr, parent);
1444 	rcu_read_unlock();
1445 	if (c) {
1446 		spin_lock(&mrt_lock);
1447 		c->_c.mfc_parent = mfc->mf6cc_parent;
1448 		ip6mr_update_thresholds(mrt, &c->_c, ttls);
1449 		if (!mrtsock)
1450 			c->_c.mfc_flags |= MFC_STATIC;
1451 		spin_unlock(&mrt_lock);
1452 		call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE,
1453 					       c, mrt->id);
1454 		mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1455 		return 0;
1456 	}
1457 
1458 	if (!ipv6_addr_any(&mfc->mf6cc_mcastgrp.sin6_addr) &&
1459 	    !ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1460 		return -EINVAL;
1461 
1462 	c = ip6mr_cache_alloc();
1463 	if (!c)
1464 		return -ENOMEM;
1465 
1466 	c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1467 	c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1468 	c->_c.mfc_parent = mfc->mf6cc_parent;
1469 	ip6mr_update_thresholds(mrt, &c->_c, ttls);
1470 	if (!mrtsock)
1471 		c->_c.mfc_flags |= MFC_STATIC;
1472 
1473 	err = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1474 				  ip6mr_rht_params);
1475 	if (err) {
1476 		pr_err("ip6mr: rhtable insert error %d\n", err);
1477 		ip6mr_cache_free(c);
1478 		return err;
1479 	}
1480 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1481 
1482 	/* Check to see if we resolved a queued list. If so we
1483 	 * need to send on the frames and tidy up.
1484 	 */
1485 	found = false;
1486 	spin_lock_bh(&mfc_unres_lock);
1487 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1488 		uc = (struct mfc6_cache *)_uc;
1489 		if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1490 		    ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1491 			list_del(&_uc->list);
1492 			atomic_dec(&mrt->cache_resolve_queue_len);
1493 			found = true;
1494 			break;
1495 		}
1496 	}
1497 	if (list_empty(&mrt->mfc_unres_queue))
1498 		del_timer(&mrt->ipmr_expire_timer);
1499 	spin_unlock_bh(&mfc_unres_lock);
1500 
1501 	if (found) {
1502 		ip6mr_cache_resolve(net, mrt, uc, c);
1503 		ip6mr_cache_free(uc);
1504 	}
1505 	call_ip6mr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD,
1506 				       c, mrt->id);
1507 	mr6_netlink_event(mrt, c, RTM_NEWROUTE);
1508 	return 0;
1509 }
1510 
1511 /*
1512  *	Close the multicast socket, and clear the vif tables etc
1513  */
1514 
1515 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1516 {
1517 	struct mr_mfc *c, *tmp;
1518 	LIST_HEAD(list);
1519 	int i;
1520 
1521 	/* Shut down all active vif entries */
1522 	if (flags & (MRT6_FLUSH_MIFS | MRT6_FLUSH_MIFS_STATIC)) {
1523 		for (i = 0; i < mrt->maxvif; i++) {
1524 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1525 			     !(flags & MRT6_FLUSH_MIFS_STATIC)) ||
1526 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT6_FLUSH_MIFS)))
1527 				continue;
1528 			mif6_delete(mrt, i, 0, &list);
1529 		}
1530 		unregister_netdevice_many(&list);
1531 	}
1532 
1533 	/* Wipe the cache */
1534 	if (flags & (MRT6_FLUSH_MFC | MRT6_FLUSH_MFC_STATIC)) {
1535 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1536 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC_STATIC)) ||
1537 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT6_FLUSH_MFC)))
1538 				continue;
1539 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ip6mr_rht_params);
1540 			list_del_rcu(&c->list);
1541 			call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt->net),
1542 						       FIB_EVENT_ENTRY_DEL,
1543 						       (struct mfc6_cache *)c, mrt->id);
1544 			mr6_netlink_event(mrt, (struct mfc6_cache *)c, RTM_DELROUTE);
1545 			mr_cache_put(c);
1546 		}
1547 	}
1548 
1549 	if (flags & MRT6_FLUSH_MFC) {
1550 		if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1551 			spin_lock_bh(&mfc_unres_lock);
1552 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1553 				list_del(&c->list);
1554 				mr6_netlink_event(mrt, (struct mfc6_cache *)c,
1555 						  RTM_DELROUTE);
1556 				ip6mr_destroy_unres(mrt, (struct mfc6_cache *)c);
1557 			}
1558 			spin_unlock_bh(&mfc_unres_lock);
1559 		}
1560 	}
1561 }
1562 
1563 static int ip6mr_sk_init(struct mr_table *mrt, struct sock *sk)
1564 {
1565 	int err = 0;
1566 	struct net *net = sock_net(sk);
1567 
1568 	rtnl_lock();
1569 	spin_lock(&mrt_lock);
1570 	if (rtnl_dereference(mrt->mroute_sk)) {
1571 		err = -EADDRINUSE;
1572 	} else {
1573 		rcu_assign_pointer(mrt->mroute_sk, sk);
1574 		sock_set_flag(sk, SOCK_RCU_FREE);
1575 		atomic_inc(&net->ipv6.devconf_all->mc_forwarding);
1576 	}
1577 	spin_unlock(&mrt_lock);
1578 
1579 	if (!err)
1580 		inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1581 					     NETCONFA_MC_FORWARDING,
1582 					     NETCONFA_IFINDEX_ALL,
1583 					     net->ipv6.devconf_all);
1584 	rtnl_unlock();
1585 
1586 	return err;
1587 }
1588 
1589 int ip6mr_sk_done(struct sock *sk)
1590 {
1591 	struct net *net = sock_net(sk);
1592 	struct ipv6_devconf *devconf;
1593 	struct mr_table *mrt;
1594 	int err = -EACCES;
1595 
1596 	if (sk->sk_type != SOCK_RAW ||
1597 	    inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1598 		return err;
1599 
1600 	devconf = net->ipv6.devconf_all;
1601 	if (!devconf || !atomic_read(&devconf->mc_forwarding))
1602 		return err;
1603 
1604 	rtnl_lock();
1605 	ip6mr_for_each_table(mrt, net) {
1606 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1607 			spin_lock(&mrt_lock);
1608 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1609 			/* Note that mroute_sk had SOCK_RCU_FREE set,
1610 			 * so the RCU grace period before sk freeing
1611 			 * is guaranteed by sk_destruct()
1612 			 */
1613 			atomic_dec(&devconf->mc_forwarding);
1614 			spin_unlock(&mrt_lock);
1615 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
1616 						     NETCONFA_MC_FORWARDING,
1617 						     NETCONFA_IFINDEX_ALL,
1618 						     net->ipv6.devconf_all);
1619 
1620 			mroute_clean_tables(mrt, MRT6_FLUSH_MIFS | MRT6_FLUSH_MFC);
1621 			err = 0;
1622 			break;
1623 		}
1624 	}
1625 	rtnl_unlock();
1626 
1627 	return err;
1628 }
1629 
1630 bool mroute6_is_socket(struct net *net, struct sk_buff *skb)
1631 {
1632 	struct mr_table *mrt;
1633 	struct flowi6 fl6 = {
1634 		.flowi6_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
1635 		.flowi6_oif	= skb->dev->ifindex,
1636 		.flowi6_mark	= skb->mark,
1637 	};
1638 
1639 	if (ip6mr_fib_lookup(net, &fl6, &mrt) < 0)
1640 		return NULL;
1641 
1642 	return rcu_access_pointer(mrt->mroute_sk);
1643 }
1644 EXPORT_SYMBOL(mroute6_is_socket);
1645 
1646 /*
1647  *	Socket options and virtual interface manipulation. The whole
1648  *	virtual interface system is a complete heap, but unfortunately
1649  *	that's how BSD mrouted happens to think. Maybe one day with a proper
1650  *	MOSPF/PIM router set up we can clean this up.
1651  */
1652 
1653 int ip6_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1654 			  unsigned int optlen)
1655 {
1656 	int ret, parent = 0;
1657 	struct mif6ctl vif;
1658 	struct mf6cctl mfc;
1659 	mifi_t mifi;
1660 	struct net *net = sock_net(sk);
1661 	struct mr_table *mrt;
1662 
1663 	if (sk->sk_type != SOCK_RAW ||
1664 	    inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1665 		return -EOPNOTSUPP;
1666 
1667 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1668 	if (!mrt)
1669 		return -ENOENT;
1670 
1671 	if (optname != MRT6_INIT) {
1672 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1673 		    !ns_capable(net->user_ns, CAP_NET_ADMIN))
1674 			return -EACCES;
1675 	}
1676 
1677 	switch (optname) {
1678 	case MRT6_INIT:
1679 		if (optlen < sizeof(int))
1680 			return -EINVAL;
1681 
1682 		return ip6mr_sk_init(mrt, sk);
1683 
1684 	case MRT6_DONE:
1685 		return ip6mr_sk_done(sk);
1686 
1687 	case MRT6_ADD_MIF:
1688 		if (optlen < sizeof(vif))
1689 			return -EINVAL;
1690 		if (copy_from_sockptr(&vif, optval, sizeof(vif)))
1691 			return -EFAULT;
1692 		if (vif.mif6c_mifi >= MAXMIFS)
1693 			return -ENFILE;
1694 		rtnl_lock();
1695 		ret = mif6_add(net, mrt, &vif,
1696 			       sk == rtnl_dereference(mrt->mroute_sk));
1697 		rtnl_unlock();
1698 		return ret;
1699 
1700 	case MRT6_DEL_MIF:
1701 		if (optlen < sizeof(mifi_t))
1702 			return -EINVAL;
1703 		if (copy_from_sockptr(&mifi, optval, sizeof(mifi_t)))
1704 			return -EFAULT;
1705 		rtnl_lock();
1706 		ret = mif6_delete(mrt, mifi, 0, NULL);
1707 		rtnl_unlock();
1708 		return ret;
1709 
1710 	/*
1711 	 *	Manipulate the forwarding caches. These live
1712 	 *	in a sort of kernel/user symbiosis.
1713 	 */
1714 	case MRT6_ADD_MFC:
1715 	case MRT6_DEL_MFC:
1716 		parent = -1;
1717 		fallthrough;
1718 	case MRT6_ADD_MFC_PROXY:
1719 	case MRT6_DEL_MFC_PROXY:
1720 		if (optlen < sizeof(mfc))
1721 			return -EINVAL;
1722 		if (copy_from_sockptr(&mfc, optval, sizeof(mfc)))
1723 			return -EFAULT;
1724 		if (parent == 0)
1725 			parent = mfc.mf6cc_parent;
1726 		rtnl_lock();
1727 		if (optname == MRT6_DEL_MFC || optname == MRT6_DEL_MFC_PROXY)
1728 			ret = ip6mr_mfc_delete(mrt, &mfc, parent);
1729 		else
1730 			ret = ip6mr_mfc_add(net, mrt, &mfc,
1731 					    sk ==
1732 					    rtnl_dereference(mrt->mroute_sk),
1733 					    parent);
1734 		rtnl_unlock();
1735 		return ret;
1736 
1737 	case MRT6_FLUSH:
1738 	{
1739 		int flags;
1740 
1741 		if (optlen != sizeof(flags))
1742 			return -EINVAL;
1743 		if (copy_from_sockptr(&flags, optval, sizeof(flags)))
1744 			return -EFAULT;
1745 		rtnl_lock();
1746 		mroute_clean_tables(mrt, flags);
1747 		rtnl_unlock();
1748 		return 0;
1749 	}
1750 
1751 	/*
1752 	 *	Control PIM assert (to activate pim will activate assert)
1753 	 */
1754 	case MRT6_ASSERT:
1755 	{
1756 		int v;
1757 
1758 		if (optlen != sizeof(v))
1759 			return -EINVAL;
1760 		if (copy_from_sockptr(&v, optval, sizeof(v)))
1761 			return -EFAULT;
1762 		mrt->mroute_do_assert = v;
1763 		return 0;
1764 	}
1765 
1766 #ifdef CONFIG_IPV6_PIMSM_V2
1767 	case MRT6_PIM:
1768 	{
1769 		bool do_wrmifwhole;
1770 		int v;
1771 
1772 		if (optlen != sizeof(v))
1773 			return -EINVAL;
1774 		if (copy_from_sockptr(&v, optval, sizeof(v)))
1775 			return -EFAULT;
1776 
1777 		do_wrmifwhole = (v == MRT6MSG_WRMIFWHOLE);
1778 		v = !!v;
1779 		rtnl_lock();
1780 		ret = 0;
1781 		if (v != mrt->mroute_do_pim) {
1782 			mrt->mroute_do_pim = v;
1783 			mrt->mroute_do_assert = v;
1784 			mrt->mroute_do_wrvifwhole = do_wrmifwhole;
1785 		}
1786 		rtnl_unlock();
1787 		return ret;
1788 	}
1789 
1790 #endif
1791 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1792 	case MRT6_TABLE:
1793 	{
1794 		u32 v;
1795 
1796 		if (optlen != sizeof(u32))
1797 			return -EINVAL;
1798 		if (copy_from_sockptr(&v, optval, sizeof(v)))
1799 			return -EFAULT;
1800 		/* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1801 		if (v != RT_TABLE_DEFAULT && v >= 100000000)
1802 			return -EINVAL;
1803 		if (sk == rcu_access_pointer(mrt->mroute_sk))
1804 			return -EBUSY;
1805 
1806 		rtnl_lock();
1807 		ret = 0;
1808 		mrt = ip6mr_new_table(net, v);
1809 		if (IS_ERR(mrt))
1810 			ret = PTR_ERR(mrt);
1811 		else
1812 			raw6_sk(sk)->ip6mr_table = v;
1813 		rtnl_unlock();
1814 		return ret;
1815 	}
1816 #endif
1817 	/*
1818 	 *	Spurious command, or MRT6_VERSION which you cannot
1819 	 *	set.
1820 	 */
1821 	default:
1822 		return -ENOPROTOOPT;
1823 	}
1824 }
1825 
1826 /*
1827  *	Getsock opt support for the multicast routing system.
1828  */
1829 
1830 int ip6_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1831 			  sockptr_t optlen)
1832 {
1833 	int olr;
1834 	int val;
1835 	struct net *net = sock_net(sk);
1836 	struct mr_table *mrt;
1837 
1838 	if (sk->sk_type != SOCK_RAW ||
1839 	    inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1840 		return -EOPNOTSUPP;
1841 
1842 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1843 	if (!mrt)
1844 		return -ENOENT;
1845 
1846 	switch (optname) {
1847 	case MRT6_VERSION:
1848 		val = 0x0305;
1849 		break;
1850 #ifdef CONFIG_IPV6_PIMSM_V2
1851 	case MRT6_PIM:
1852 		val = mrt->mroute_do_pim;
1853 		break;
1854 #endif
1855 	case MRT6_ASSERT:
1856 		val = mrt->mroute_do_assert;
1857 		break;
1858 	default:
1859 		return -ENOPROTOOPT;
1860 	}
1861 
1862 	if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1863 		return -EFAULT;
1864 
1865 	olr = min_t(int, olr, sizeof(int));
1866 	if (olr < 0)
1867 		return -EINVAL;
1868 
1869 	if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1870 		return -EFAULT;
1871 	if (copy_to_sockptr(optval, &val, olr))
1872 		return -EFAULT;
1873 	return 0;
1874 }
1875 
1876 /*
1877  *	The IP multicast ioctl support routines.
1878  */
1879 int ip6mr_ioctl(struct sock *sk, int cmd, void *arg)
1880 {
1881 	struct sioc_sg_req6 *sr;
1882 	struct sioc_mif_req6 *vr;
1883 	struct vif_device *vif;
1884 	struct mfc6_cache *c;
1885 	struct net *net = sock_net(sk);
1886 	struct mr_table *mrt;
1887 
1888 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1889 	if (!mrt)
1890 		return -ENOENT;
1891 
1892 	switch (cmd) {
1893 	case SIOCGETMIFCNT_IN6:
1894 		vr = (struct sioc_mif_req6 *)arg;
1895 		if (vr->mifi >= mrt->maxvif)
1896 			return -EINVAL;
1897 		vr->mifi = array_index_nospec(vr->mifi, mrt->maxvif);
1898 		rcu_read_lock();
1899 		vif = &mrt->vif_table[vr->mifi];
1900 		if (VIF_EXISTS(mrt, vr->mifi)) {
1901 			vr->icount = READ_ONCE(vif->pkt_in);
1902 			vr->ocount = READ_ONCE(vif->pkt_out);
1903 			vr->ibytes = READ_ONCE(vif->bytes_in);
1904 			vr->obytes = READ_ONCE(vif->bytes_out);
1905 			rcu_read_unlock();
1906 			return 0;
1907 		}
1908 		rcu_read_unlock();
1909 		return -EADDRNOTAVAIL;
1910 	case SIOCGETSGCNT_IN6:
1911 		sr = (struct sioc_sg_req6 *)arg;
1912 
1913 		rcu_read_lock();
1914 		c = ip6mr_cache_find(mrt, &sr->src.sin6_addr,
1915 				     &sr->grp.sin6_addr);
1916 		if (c) {
1917 			sr->pktcnt = c->_c.mfc_un.res.pkt;
1918 			sr->bytecnt = c->_c.mfc_un.res.bytes;
1919 			sr->wrong_if = c->_c.mfc_un.res.wrong_if;
1920 			rcu_read_unlock();
1921 			return 0;
1922 		}
1923 		rcu_read_unlock();
1924 		return -EADDRNOTAVAIL;
1925 	default:
1926 		return -ENOIOCTLCMD;
1927 	}
1928 }
1929 
1930 #ifdef CONFIG_COMPAT
1931 struct compat_sioc_sg_req6 {
1932 	struct sockaddr_in6 src;
1933 	struct sockaddr_in6 grp;
1934 	compat_ulong_t pktcnt;
1935 	compat_ulong_t bytecnt;
1936 	compat_ulong_t wrong_if;
1937 };
1938 
1939 struct compat_sioc_mif_req6 {
1940 	mifi_t	mifi;
1941 	compat_ulong_t icount;
1942 	compat_ulong_t ocount;
1943 	compat_ulong_t ibytes;
1944 	compat_ulong_t obytes;
1945 };
1946 
1947 int ip6mr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1948 {
1949 	struct compat_sioc_sg_req6 sr;
1950 	struct compat_sioc_mif_req6 vr;
1951 	struct vif_device *vif;
1952 	struct mfc6_cache *c;
1953 	struct net *net = sock_net(sk);
1954 	struct mr_table *mrt;
1955 
1956 	mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1957 	if (!mrt)
1958 		return -ENOENT;
1959 
1960 	switch (cmd) {
1961 	case SIOCGETMIFCNT_IN6:
1962 		if (copy_from_user(&vr, arg, sizeof(vr)))
1963 			return -EFAULT;
1964 		if (vr.mifi >= mrt->maxvif)
1965 			return -EINVAL;
1966 		vr.mifi = array_index_nospec(vr.mifi, mrt->maxvif);
1967 		rcu_read_lock();
1968 		vif = &mrt->vif_table[vr.mifi];
1969 		if (VIF_EXISTS(mrt, vr.mifi)) {
1970 			vr.icount = READ_ONCE(vif->pkt_in);
1971 			vr.ocount = READ_ONCE(vif->pkt_out);
1972 			vr.ibytes = READ_ONCE(vif->bytes_in);
1973 			vr.obytes = READ_ONCE(vif->bytes_out);
1974 			rcu_read_unlock();
1975 
1976 			if (copy_to_user(arg, &vr, sizeof(vr)))
1977 				return -EFAULT;
1978 			return 0;
1979 		}
1980 		rcu_read_unlock();
1981 		return -EADDRNOTAVAIL;
1982 	case SIOCGETSGCNT_IN6:
1983 		if (copy_from_user(&sr, arg, sizeof(sr)))
1984 			return -EFAULT;
1985 
1986 		rcu_read_lock();
1987 		c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1988 		if (c) {
1989 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1990 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1991 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1992 			rcu_read_unlock();
1993 
1994 			if (copy_to_user(arg, &sr, sizeof(sr)))
1995 				return -EFAULT;
1996 			return 0;
1997 		}
1998 		rcu_read_unlock();
1999 		return -EADDRNOTAVAIL;
2000 	default:
2001 		return -ENOIOCTLCMD;
2002 	}
2003 }
2004 #endif
2005 
2006 static inline int ip6mr_forward2_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
2007 {
2008 	IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
2009 		      IPSTATS_MIB_OUTFORWDATAGRAMS);
2010 	return dst_output(net, sk, skb);
2011 }
2012 
2013 /*
2014  *	Processing handlers for ip6mr_forward
2015  */
2016 
2017 static int ip6mr_forward2(struct net *net, struct mr_table *mrt,
2018 			  struct sk_buff *skb, int vifi)
2019 {
2020 	struct vif_device *vif = &mrt->vif_table[vifi];
2021 	struct net_device *vif_dev;
2022 	struct ipv6hdr *ipv6h;
2023 	struct dst_entry *dst;
2024 	struct flowi6 fl6;
2025 
2026 	vif_dev = vif_dev_read(vif);
2027 	if (!vif_dev)
2028 		goto out_free;
2029 
2030 #ifdef CONFIG_IPV6_PIMSM_V2
2031 	if (vif->flags & MIFF_REGISTER) {
2032 		WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
2033 		WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
2034 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
2035 		DEV_STATS_INC(vif_dev, tx_packets);
2036 		ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
2037 		goto out_free;
2038 	}
2039 #endif
2040 
2041 	ipv6h = ipv6_hdr(skb);
2042 
2043 	fl6 = (struct flowi6) {
2044 		.flowi6_oif = vif->link,
2045 		.daddr = ipv6h->daddr,
2046 	};
2047 
2048 	dst = ip6_route_output(net, NULL, &fl6);
2049 	if (dst->error) {
2050 		dst_release(dst);
2051 		goto out_free;
2052 	}
2053 
2054 	skb_dst_drop(skb);
2055 	skb_dst_set(skb, dst);
2056 
2057 	/*
2058 	 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2059 	 * not only before forwarding, but after forwarding on all output
2060 	 * interfaces. It is clear, if mrouter runs a multicasting
2061 	 * program, it should receive packets not depending to what interface
2062 	 * program is joined.
2063 	 * If we will not make it, the program will have to join on all
2064 	 * interfaces. On the other hand, multihoming host (or router, but
2065 	 * not mrouter) cannot join to more than one interface - it will
2066 	 * result in receiving multiple packets.
2067 	 */
2068 	skb->dev = vif_dev;
2069 	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
2070 	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
2071 
2072 	/* We are about to write */
2073 	/* XXX: extension headers? */
2074 	if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(vif_dev)))
2075 		goto out_free;
2076 
2077 	ipv6h = ipv6_hdr(skb);
2078 	ipv6h->hop_limit--;
2079 
2080 	IP6CB(skb)->flags |= IP6SKB_FORWARDED;
2081 
2082 	return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
2083 		       net, NULL, skb, skb->dev, vif_dev,
2084 		       ip6mr_forward2_finish);
2085 
2086 out_free:
2087 	kfree_skb(skb);
2088 	return 0;
2089 }
2090 
2091 /* Called with rcu_read_lock() */
2092 static int ip6mr_find_vif(struct mr_table *mrt, struct net_device *dev)
2093 {
2094 	int ct;
2095 
2096 	/* Pairs with WRITE_ONCE() in mif6_delete()/mif6_add() */
2097 	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
2098 		if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
2099 			break;
2100 	}
2101 	return ct;
2102 }
2103 
2104 /* Called under rcu_read_lock() */
2105 static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
2106 			   struct net_device *dev, struct sk_buff *skb,
2107 			   struct mfc6_cache *c)
2108 {
2109 	int psend = -1;
2110 	int vif, ct;
2111 	int true_vifi = ip6mr_find_vif(mrt, dev);
2112 
2113 	vif = c->_c.mfc_parent;
2114 	c->_c.mfc_un.res.pkt++;
2115 	c->_c.mfc_un.res.bytes += skb->len;
2116 	c->_c.mfc_un.res.lastuse = jiffies;
2117 
2118 	if (ipv6_addr_any(&c->mf6c_origin) && true_vifi >= 0) {
2119 		struct mfc6_cache *cache_proxy;
2120 
2121 		/* For an (*,G) entry, we only check that the incoming
2122 		 * interface is part of the static tree.
2123 		 */
2124 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2125 		if (cache_proxy &&
2126 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
2127 			goto forward;
2128 	}
2129 
2130 	/*
2131 	 * Wrong interface: drop packet and (maybe) send PIM assert.
2132 	 */
2133 	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
2134 		c->_c.mfc_un.res.wrong_if++;
2135 
2136 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
2137 		    /* pimsm uses asserts, when switching from RPT to SPT,
2138 		       so that we cannot check that packet arrived on an oif.
2139 		       It is bad, but otherwise we would need to move pretty
2140 		       large chunk of pimd to kernel. Ough... --ANK
2141 		     */
2142 		    (mrt->mroute_do_pim ||
2143 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2144 		    time_after(jiffies,
2145 			       c->_c.mfc_un.res.last_assert +
2146 			       MFC_ASSERT_THRESH)) {
2147 			c->_c.mfc_un.res.last_assert = jiffies;
2148 			ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
2149 			if (mrt->mroute_do_wrvifwhole)
2150 				ip6mr_cache_report(mrt, skb, true_vifi,
2151 						   MRT6MSG_WRMIFWHOLE);
2152 		}
2153 		goto dont_forward;
2154 	}
2155 
2156 forward:
2157 	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2158 		   mrt->vif_table[vif].pkt_in + 1);
2159 	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2160 		   mrt->vif_table[vif].bytes_in + skb->len);
2161 
2162 	/*
2163 	 *	Forward the frame
2164 	 */
2165 	if (ipv6_addr_any(&c->mf6c_origin) &&
2166 	    ipv6_addr_any(&c->mf6c_mcastgrp)) {
2167 		if (true_vifi >= 0 &&
2168 		    true_vifi != c->_c.mfc_parent &&
2169 		    ipv6_hdr(skb)->hop_limit >
2170 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2171 			/* It's an (*,*) entry and the packet is not coming from
2172 			 * the upstream: forward the packet to the upstream
2173 			 * only.
2174 			 */
2175 			psend = c->_c.mfc_parent;
2176 			goto last_forward;
2177 		}
2178 		goto dont_forward;
2179 	}
2180 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2181 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2182 		/* For (*,G) entry, don't forward to the incoming interface */
2183 		if ((!ipv6_addr_any(&c->mf6c_origin) || ct != true_vifi) &&
2184 		    ipv6_hdr(skb)->hop_limit > c->_c.mfc_un.res.ttls[ct]) {
2185 			if (psend != -1) {
2186 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2187 				if (skb2)
2188 					ip6mr_forward2(net, mrt, skb2, psend);
2189 			}
2190 			psend = ct;
2191 		}
2192 	}
2193 last_forward:
2194 	if (psend != -1) {
2195 		ip6mr_forward2(net, mrt, skb, psend);
2196 		return;
2197 	}
2198 
2199 dont_forward:
2200 	kfree_skb(skb);
2201 }
2202 
2203 
2204 /*
2205  *	Multicast packets for forwarding arrive here
2206  */
2207 
2208 int ip6_mr_input(struct sk_buff *skb)
2209 {
2210 	struct mfc6_cache *cache;
2211 	struct net *net = dev_net(skb->dev);
2212 	struct mr_table *mrt;
2213 	struct flowi6 fl6 = {
2214 		.flowi6_iif	= skb->dev->ifindex,
2215 		.flowi6_mark	= skb->mark,
2216 	};
2217 	int err;
2218 	struct net_device *dev;
2219 
2220 	/* skb->dev passed in is the master dev for vrfs.
2221 	 * Get the proper interface that does have a vif associated with it.
2222 	 */
2223 	dev = skb->dev;
2224 	if (netif_is_l3_master(skb->dev)) {
2225 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2226 		if (!dev) {
2227 			kfree_skb(skb);
2228 			return -ENODEV;
2229 		}
2230 	}
2231 
2232 	err = ip6mr_fib_lookup(net, &fl6, &mrt);
2233 	if (err < 0) {
2234 		kfree_skb(skb);
2235 		return err;
2236 	}
2237 
2238 	cache = ip6mr_cache_find(mrt,
2239 				 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
2240 	if (!cache) {
2241 		int vif = ip6mr_find_vif(mrt, dev);
2242 
2243 		if (vif >= 0)
2244 			cache = ip6mr_cache_find_any(mrt,
2245 						     &ipv6_hdr(skb)->daddr,
2246 						     vif);
2247 	}
2248 
2249 	/*
2250 	 *	No usable cache entry
2251 	 */
2252 	if (!cache) {
2253 		int vif;
2254 
2255 		vif = ip6mr_find_vif(mrt, dev);
2256 		if (vif >= 0) {
2257 			int err = ip6mr_cache_unresolved(mrt, vif, skb, dev);
2258 
2259 			return err;
2260 		}
2261 		kfree_skb(skb);
2262 		return -ENODEV;
2263 	}
2264 
2265 	ip6_mr_forward(net, mrt, dev, skb, cache);
2266 
2267 	return 0;
2268 }
2269 
2270 int ip6mr_get_route(struct net *net, struct sk_buff *skb, struct rtmsg *rtm,
2271 		    u32 portid)
2272 {
2273 	int err;
2274 	struct mr_table *mrt;
2275 	struct mfc6_cache *cache;
2276 	struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2277 
2278 	mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2279 	if (!mrt)
2280 		return -ENOENT;
2281 
2282 	rcu_read_lock();
2283 	cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2284 	if (!cache && skb->dev) {
2285 		int vif = ip6mr_find_vif(mrt, skb->dev);
2286 
2287 		if (vif >= 0)
2288 			cache = ip6mr_cache_find_any(mrt, &rt->rt6i_dst.addr,
2289 						     vif);
2290 	}
2291 
2292 	if (!cache) {
2293 		struct sk_buff *skb2;
2294 		struct ipv6hdr *iph;
2295 		struct net_device *dev;
2296 		int vif;
2297 
2298 		dev = skb->dev;
2299 		if (!dev || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2300 			rcu_read_unlock();
2301 			return -ENODEV;
2302 		}
2303 
2304 		/* really correct? */
2305 		skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2306 		if (!skb2) {
2307 			rcu_read_unlock();
2308 			return -ENOMEM;
2309 		}
2310 
2311 		NETLINK_CB(skb2).portid = portid;
2312 		skb_reset_transport_header(skb2);
2313 
2314 		skb_put(skb2, sizeof(struct ipv6hdr));
2315 		skb_reset_network_header(skb2);
2316 
2317 		iph = ipv6_hdr(skb2);
2318 		iph->version = 0;
2319 		iph->priority = 0;
2320 		iph->flow_lbl[0] = 0;
2321 		iph->flow_lbl[1] = 0;
2322 		iph->flow_lbl[2] = 0;
2323 		iph->payload_len = 0;
2324 		iph->nexthdr = IPPROTO_NONE;
2325 		iph->hop_limit = 0;
2326 		iph->saddr = rt->rt6i_src.addr;
2327 		iph->daddr = rt->rt6i_dst.addr;
2328 
2329 		err = ip6mr_cache_unresolved(mrt, vif, skb2, dev);
2330 		rcu_read_unlock();
2331 
2332 		return err;
2333 	}
2334 
2335 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2336 	rcu_read_unlock();
2337 	return err;
2338 }
2339 
2340 static int ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2341 			     u32 portid, u32 seq, struct mfc6_cache *c, int cmd,
2342 			     int flags)
2343 {
2344 	struct nlmsghdr *nlh;
2345 	struct rtmsg *rtm;
2346 	int err;
2347 
2348 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2349 	if (!nlh)
2350 		return -EMSGSIZE;
2351 
2352 	rtm = nlmsg_data(nlh);
2353 	rtm->rtm_family   = RTNL_FAMILY_IP6MR;
2354 	rtm->rtm_dst_len  = 128;
2355 	rtm->rtm_src_len  = 128;
2356 	rtm->rtm_tos      = 0;
2357 	rtm->rtm_table    = mrt->id;
2358 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2359 		goto nla_put_failure;
2360 	rtm->rtm_type = RTN_MULTICAST;
2361 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2362 	if (c->_c.mfc_flags & MFC_STATIC)
2363 		rtm->rtm_protocol = RTPROT_STATIC;
2364 	else
2365 		rtm->rtm_protocol = RTPROT_MROUTED;
2366 	rtm->rtm_flags    = 0;
2367 
2368 	if (nla_put_in6_addr(skb, RTA_SRC, &c->mf6c_origin) ||
2369 	    nla_put_in6_addr(skb, RTA_DST, &c->mf6c_mcastgrp))
2370 		goto nla_put_failure;
2371 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2372 	/* do not break the dump if cache is unresolved */
2373 	if (err < 0 && err != -ENOENT)
2374 		goto nla_put_failure;
2375 
2376 	nlmsg_end(skb, nlh);
2377 	return 0;
2378 
2379 nla_put_failure:
2380 	nlmsg_cancel(skb, nlh);
2381 	return -EMSGSIZE;
2382 }
2383 
2384 static int _ip6mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2385 			      u32 portid, u32 seq, struct mr_mfc *c,
2386 			      int cmd, int flags)
2387 {
2388 	return ip6mr_fill_mroute(mrt, skb, portid, seq, (struct mfc6_cache *)c,
2389 				 cmd, flags);
2390 }
2391 
2392 static int mr6_msgsize(bool unresolved, int maxvif)
2393 {
2394 	size_t len =
2395 		NLMSG_ALIGN(sizeof(struct rtmsg))
2396 		+ nla_total_size(4)	/* RTA_TABLE */
2397 		+ nla_total_size(sizeof(struct in6_addr))	/* RTA_SRC */
2398 		+ nla_total_size(sizeof(struct in6_addr))	/* RTA_DST */
2399 		;
2400 
2401 	if (!unresolved)
2402 		len = len
2403 		      + nla_total_size(4)	/* RTA_IIF */
2404 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2405 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2406 						/* RTA_MFC_STATS */
2407 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2408 		;
2409 
2410 	return len;
2411 }
2412 
2413 static void mr6_netlink_event(struct mr_table *mrt, struct mfc6_cache *mfc,
2414 			      int cmd)
2415 {
2416 	struct net *net = read_pnet(&mrt->net);
2417 	struct sk_buff *skb;
2418 	int err = -ENOBUFS;
2419 
2420 	skb = nlmsg_new(mr6_msgsize(mfc->_c.mfc_parent >= MAXMIFS, mrt->maxvif),
2421 			GFP_ATOMIC);
2422 	if (!skb)
2423 		goto errout;
2424 
2425 	err = ip6mr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2426 	if (err < 0)
2427 		goto errout;
2428 
2429 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE, NULL, GFP_ATOMIC);
2430 	return;
2431 
2432 errout:
2433 	kfree_skb(skb);
2434 	if (err < 0)
2435 		rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE, err);
2436 }
2437 
2438 static size_t mrt6msg_netlink_msgsize(size_t payloadlen)
2439 {
2440 	size_t len =
2441 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2442 		+ nla_total_size(1)	/* IP6MRA_CREPORT_MSGTYPE */
2443 		+ nla_total_size(4)	/* IP6MRA_CREPORT_MIF_ID */
2444 					/* IP6MRA_CREPORT_SRC_ADDR */
2445 		+ nla_total_size(sizeof(struct in6_addr))
2446 					/* IP6MRA_CREPORT_DST_ADDR */
2447 		+ nla_total_size(sizeof(struct in6_addr))
2448 					/* IP6MRA_CREPORT_PKT */
2449 		+ nla_total_size(payloadlen)
2450 		;
2451 
2452 	return len;
2453 }
2454 
2455 static void mrt6msg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2456 {
2457 	struct net *net = read_pnet(&mrt->net);
2458 	struct nlmsghdr *nlh;
2459 	struct rtgenmsg *rtgenm;
2460 	struct mrt6msg *msg;
2461 	struct sk_buff *skb;
2462 	struct nlattr *nla;
2463 	int payloadlen;
2464 
2465 	payloadlen = pkt->len - sizeof(struct mrt6msg);
2466 	msg = (struct mrt6msg *)skb_transport_header(pkt);
2467 
2468 	skb = nlmsg_new(mrt6msg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2469 	if (!skb)
2470 		goto errout;
2471 
2472 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2473 			sizeof(struct rtgenmsg), 0);
2474 	if (!nlh)
2475 		goto errout;
2476 	rtgenm = nlmsg_data(nlh);
2477 	rtgenm->rtgen_family = RTNL_FAMILY_IP6MR;
2478 	if (nla_put_u8(skb, IP6MRA_CREPORT_MSGTYPE, msg->im6_msgtype) ||
2479 	    nla_put_u32(skb, IP6MRA_CREPORT_MIF_ID, msg->im6_mif) ||
2480 	    nla_put_in6_addr(skb, IP6MRA_CREPORT_SRC_ADDR,
2481 			     &msg->im6_src) ||
2482 	    nla_put_in6_addr(skb, IP6MRA_CREPORT_DST_ADDR,
2483 			     &msg->im6_dst))
2484 		goto nla_put_failure;
2485 
2486 	nla = nla_reserve(skb, IP6MRA_CREPORT_PKT, payloadlen);
2487 	if (!nla || skb_copy_bits(pkt, sizeof(struct mrt6msg),
2488 				  nla_data(nla), payloadlen))
2489 		goto nla_put_failure;
2490 
2491 	nlmsg_end(skb, nlh);
2492 
2493 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MROUTE_R, NULL, GFP_ATOMIC);
2494 	return;
2495 
2496 nla_put_failure:
2497 	nlmsg_cancel(skb, nlh);
2498 errout:
2499 	kfree_skb(skb);
2500 	rtnl_set_sk_err(net, RTNLGRP_IPV6_MROUTE_R, -ENOBUFS);
2501 }
2502 
2503 static const struct nla_policy ip6mr_getroute_policy[RTA_MAX + 1] = {
2504 	[RTA_SRC]		= NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
2505 	[RTA_DST]		= NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
2506 	[RTA_TABLE]		= { .type = NLA_U32 },
2507 };
2508 
2509 static int ip6mr_rtm_valid_getroute_req(struct sk_buff *skb,
2510 					const struct nlmsghdr *nlh,
2511 					struct nlattr **tb,
2512 					struct netlink_ext_ack *extack)
2513 {
2514 	struct rtmsg *rtm;
2515 	int err;
2516 
2517 	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, ip6mr_getroute_policy,
2518 			  extack);
2519 	if (err)
2520 		return err;
2521 
2522 	rtm = nlmsg_data(nlh);
2523 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
2524 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
2525 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2526 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2527 		NL_SET_ERR_MSG_MOD(extack,
2528 				   "Invalid values in header for multicast route get request");
2529 		return -EINVAL;
2530 	}
2531 
2532 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2533 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2534 		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
2535 		return -EINVAL;
2536 	}
2537 
2538 	return 0;
2539 }
2540 
2541 static int ip6mr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2542 			      struct netlink_ext_ack *extack)
2543 {
2544 	struct net *net = sock_net(in_skb->sk);
2545 	struct in6_addr src = {}, grp = {};
2546 	struct nlattr *tb[RTA_MAX + 1];
2547 	struct mfc6_cache *cache;
2548 	struct mr_table *mrt;
2549 	struct sk_buff *skb;
2550 	u32 tableid;
2551 	int err;
2552 
2553 	err = ip6mr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2554 	if (err < 0)
2555 		return err;
2556 
2557 	if (tb[RTA_SRC])
2558 		src = nla_get_in6_addr(tb[RTA_SRC]);
2559 	if (tb[RTA_DST])
2560 		grp = nla_get_in6_addr(tb[RTA_DST]);
2561 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2562 
2563 	mrt = ip6mr_get_table(net, tableid ?: RT_TABLE_DEFAULT);
2564 	if (!mrt) {
2565 		NL_SET_ERR_MSG_MOD(extack, "MR table does not exist");
2566 		return -ENOENT;
2567 	}
2568 
2569 	/* entries are added/deleted only under RTNL */
2570 	rcu_read_lock();
2571 	cache = ip6mr_cache_find(mrt, &src, &grp);
2572 	rcu_read_unlock();
2573 	if (!cache) {
2574 		NL_SET_ERR_MSG_MOD(extack, "MR cache entry not found");
2575 		return -ENOENT;
2576 	}
2577 
2578 	skb = nlmsg_new(mr6_msgsize(false, mrt->maxvif), GFP_KERNEL);
2579 	if (!skb)
2580 		return -ENOBUFS;
2581 
2582 	err = ip6mr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2583 				nlh->nlmsg_seq, cache, RTM_NEWROUTE, 0);
2584 	if (err < 0) {
2585 		kfree_skb(skb);
2586 		return err;
2587 	}
2588 
2589 	return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2590 }
2591 
2592 static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2593 {
2594 	const struct nlmsghdr *nlh = cb->nlh;
2595 	struct fib_dump_filter filter = {
2596 		.rtnl_held = true,
2597 	};
2598 	int err;
2599 
2600 	if (cb->strict_check) {
2601 		err = ip_valid_fib_dump_req(sock_net(skb->sk), nlh,
2602 					    &filter, cb);
2603 		if (err < 0)
2604 			return err;
2605 	}
2606 
2607 	if (filter.table_id) {
2608 		struct mr_table *mrt;
2609 
2610 		mrt = ip6mr_get_table(sock_net(skb->sk), filter.table_id);
2611 		if (!mrt) {
2612 			if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IP6MR)
2613 				return skb->len;
2614 
2615 			NL_SET_ERR_MSG_MOD(cb->extack, "MR table does not exist");
2616 			return -ENOENT;
2617 		}
2618 		err = mr_table_dump(mrt, skb, cb, _ip6mr_fill_mroute,
2619 				    &mfc_unres_lock, &filter);
2620 		return skb->len ? : err;
2621 	}
2622 
2623 	return mr_rtm_dumproute(skb, cb, ip6mr_mr_table_iter,
2624 				_ip6mr_fill_mroute, &mfc_unres_lock, &filter);
2625 }
2626