xref: /linux/net/ipv4/ipmr.c (revision b37042b2bb7cd751f03b73afb90364a418d870f4)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IP multicast routing support for mrouted 3.6/3.8
4  *
5  *		(c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6  *	  Linux Consultancy and Custom Driver Development
7  *
8  *	Fixes:
9  *	Michael Chastain	:	Incorrect size of copying.
10  *	Alan Cox		:	Added the cache manager code
11  *	Alan Cox		:	Fixed the clone/copy bug and device race.
12  *	Mike McLagan		:	Routing by source
13  *	Malcolm Beattie		:	Buffer handling fixes.
14  *	Alexey Kuznetsov	:	Double buffer free and other fixes.
15  *	SVR Anand		:	Fixed several multicast bugs and problems.
16  *	Alexey Kuznetsov	:	Status, optimisations and more.
17  *	Brad Parker		:	Better behaviour on mrouted upcall
18  *					overflow.
19  *      Carlos Picoto           :       PIMv1 Support
20  *	Pavlin Ivanov Radoslavov:	PIMv2 Registers must checksum only PIM header
21  *					Relax this requirement to work with older peers.
22  */
23 
24 #include <linux/uaccess.h>
25 #include <linux/types.h>
26 #include <linux/cache.h>
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/mm.h>
30 #include <linux/kernel.h>
31 #include <linux/fcntl.h>
32 #include <linux/stat.h>
33 #include <linux/socket.h>
34 #include <linux/in.h>
35 #include <linux/inet.h>
36 #include <linux/netdevice.h>
37 #include <linux/inetdevice.h>
38 #include <linux/igmp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/mroute.h>
42 #include <linux/init.h>
43 #include <linux/if_ether.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <linux/skbuff.h>
49 #include <net/route.h>
50 #include <net/icmp.h>
51 #include <net/udp.h>
52 #include <net/raw.h>
53 #include <linux/notifier.h>
54 #include <linux/if_arp.h>
55 #include <linux/netfilter_ipv4.h>
56 #include <linux/compat.h>
57 #include <linux/export.h>
58 #include <linux/rhashtable.h>
59 #include <net/ip_tunnels.h>
60 #include <net/checksum.h>
61 #include <net/netlink.h>
62 #include <net/fib_rules.h>
63 #include <linux/netconf.h>
64 #include <net/rtnh.h>
65 
66 #include <linux/nospec.h>
67 
68 struct ipmr_rule {
69 	struct fib_rule		common;
70 };
71 
72 struct ipmr_result {
73 	struct mr_table		*mrt;
74 };
75 
76 /* Big lock, protecting vif table, mrt cache and mroute socket state.
77  * Note that the changes are semaphored via rtnl_lock.
78  */
79 
80 static DEFINE_SPINLOCK(mrt_lock);
81 
82 static struct net_device *vif_dev_read(const struct vif_device *vif)
83 {
84 	return rcu_dereference(vif->dev);
85 }
86 
87 /* Multicast router control variables */
88 
89 /* Special spinlock for queue of unresolved entries */
90 static DEFINE_SPINLOCK(mfc_unres_lock);
91 
92 /* We return to original Alan's scheme. Hash table of resolved
93  * entries is changed only in process context and protected
94  * with weak lock mrt_lock. Queue of unresolved entries is protected
95  * with strong spinlock mfc_unres_lock.
96  *
97  * In this case data path is free of exclusive locks at all.
98  */
99 
100 static struct kmem_cache *mrt_cachep __ro_after_init;
101 
102 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
103 static void ipmr_free_table(struct mr_table *mrt);
104 
105 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
106 			  struct net_device *dev, struct sk_buff *skb,
107 			  struct mfc_cache *cache, int local);
108 static int ipmr_cache_report(const struct mr_table *mrt,
109 			     struct sk_buff *pkt, vifi_t vifi, int assert);
110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
111 				 int cmd);
112 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
113 static void mroute_clean_tables(struct mr_table *mrt, int flags);
114 static void ipmr_expire_process(struct timer_list *t);
115 
116 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
117 #define ipmr_for_each_table(mrt, net)					\
118 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,	\
119 				lockdep_rtnl_is_held() ||		\
120 				list_empty(&net->ipv4.mr_tables))
121 
122 static struct mr_table *ipmr_mr_table_iter(struct net *net,
123 					   struct mr_table *mrt)
124 {
125 	struct mr_table *ret;
126 
127 	if (!mrt)
128 		ret = list_entry_rcu(net->ipv4.mr_tables.next,
129 				     struct mr_table, list);
130 	else
131 		ret = list_entry_rcu(mrt->list.next,
132 				     struct mr_table, list);
133 
134 	if (&ret->list == &net->ipv4.mr_tables)
135 		return NULL;
136 	return ret;
137 }
138 
139 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
140 {
141 	struct mr_table *mrt;
142 
143 	ipmr_for_each_table(mrt, net) {
144 		if (mrt->id == id)
145 			return mrt;
146 	}
147 	return NULL;
148 }
149 
150 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
151 			   struct mr_table **mrt)
152 {
153 	int err;
154 	struct ipmr_result res;
155 	struct fib_lookup_arg arg = {
156 		.result = &res,
157 		.flags = FIB_LOOKUP_NOREF,
158 	};
159 
160 	/* update flow if oif or iif point to device enslaved to l3mdev */
161 	l3mdev_update_flow(net, flowi4_to_flowi(flp4));
162 
163 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
164 			       flowi4_to_flowi(flp4), 0, &arg);
165 	if (err < 0)
166 		return err;
167 	*mrt = res.mrt;
168 	return 0;
169 }
170 
171 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
172 			    int flags, struct fib_lookup_arg *arg)
173 {
174 	struct ipmr_result *res = arg->result;
175 	struct mr_table *mrt;
176 
177 	switch (rule->action) {
178 	case FR_ACT_TO_TBL:
179 		break;
180 	case FR_ACT_UNREACHABLE:
181 		return -ENETUNREACH;
182 	case FR_ACT_PROHIBIT:
183 		return -EACCES;
184 	case FR_ACT_BLACKHOLE:
185 	default:
186 		return -EINVAL;
187 	}
188 
189 	arg->table = fib_rule_get_table(rule, arg);
190 
191 	mrt = ipmr_get_table(rule->fr_net, arg->table);
192 	if (!mrt)
193 		return -EAGAIN;
194 	res->mrt = mrt;
195 	return 0;
196 }
197 
198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
199 {
200 	return 1;
201 }
202 
203 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
204 			       struct fib_rule_hdr *frh, struct nlattr **tb,
205 			       struct netlink_ext_ack *extack)
206 {
207 	return 0;
208 }
209 
210 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
211 			     struct nlattr **tb)
212 {
213 	return 1;
214 }
215 
216 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
217 			  struct fib_rule_hdr *frh)
218 {
219 	frh->dst_len = 0;
220 	frh->src_len = 0;
221 	frh->tos     = 0;
222 	return 0;
223 }
224 
225 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
226 	.family		= RTNL_FAMILY_IPMR,
227 	.rule_size	= sizeof(struct ipmr_rule),
228 	.addr_size	= sizeof(u32),
229 	.action		= ipmr_rule_action,
230 	.match		= ipmr_rule_match,
231 	.configure	= ipmr_rule_configure,
232 	.compare	= ipmr_rule_compare,
233 	.fill		= ipmr_rule_fill,
234 	.nlgroup	= RTNLGRP_IPV4_RULE,
235 	.owner		= THIS_MODULE,
236 };
237 
238 static int __net_init ipmr_rules_init(struct net *net)
239 {
240 	struct fib_rules_ops *ops;
241 	struct mr_table *mrt;
242 	int err;
243 
244 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 	if (IS_ERR(ops))
246 		return PTR_ERR(ops);
247 
248 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
249 
250 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 	if (IS_ERR(mrt)) {
252 		err = PTR_ERR(mrt);
253 		goto err1;
254 	}
255 
256 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
257 	if (err < 0)
258 		goto err2;
259 
260 	net->ipv4.mr_rules_ops = ops;
261 	return 0;
262 
263 err2:
264 	rtnl_lock();
265 	ipmr_free_table(mrt);
266 	rtnl_unlock();
267 err1:
268 	fib_rules_unregister(ops);
269 	return err;
270 }
271 
272 static void __net_exit ipmr_rules_exit(struct net *net)
273 {
274 	struct mr_table *mrt, *next;
275 
276 	ASSERT_RTNL();
277 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
278 		list_del(&mrt->list);
279 		ipmr_free_table(mrt);
280 	}
281 	fib_rules_unregister(net->ipv4.mr_rules_ops);
282 }
283 
284 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
285 			   struct netlink_ext_ack *extack)
286 {
287 	return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
288 }
289 
290 static unsigned int ipmr_rules_seq_read(struct net *net)
291 {
292 	return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
293 }
294 
295 bool ipmr_rule_default(const struct fib_rule *rule)
296 {
297 	return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
298 }
299 EXPORT_SYMBOL(ipmr_rule_default);
300 #else
301 #define ipmr_for_each_table(mrt, net) \
302 	for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
303 
304 static struct mr_table *ipmr_mr_table_iter(struct net *net,
305 					   struct mr_table *mrt)
306 {
307 	if (!mrt)
308 		return net->ipv4.mrt;
309 	return NULL;
310 }
311 
312 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
313 {
314 	return net->ipv4.mrt;
315 }
316 
317 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
318 			   struct mr_table **mrt)
319 {
320 	*mrt = net->ipv4.mrt;
321 	return 0;
322 }
323 
324 static int __net_init ipmr_rules_init(struct net *net)
325 {
326 	struct mr_table *mrt;
327 
328 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
329 	if (IS_ERR(mrt))
330 		return PTR_ERR(mrt);
331 	net->ipv4.mrt = mrt;
332 	return 0;
333 }
334 
335 static void __net_exit ipmr_rules_exit(struct net *net)
336 {
337 	ASSERT_RTNL();
338 	ipmr_free_table(net->ipv4.mrt);
339 	net->ipv4.mrt = NULL;
340 }
341 
342 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
343 			   struct netlink_ext_ack *extack)
344 {
345 	return 0;
346 }
347 
348 static unsigned int ipmr_rules_seq_read(struct net *net)
349 {
350 	return 0;
351 }
352 
353 bool ipmr_rule_default(const struct fib_rule *rule)
354 {
355 	return true;
356 }
357 EXPORT_SYMBOL(ipmr_rule_default);
358 #endif
359 
360 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
361 				const void *ptr)
362 {
363 	const struct mfc_cache_cmp_arg *cmparg = arg->key;
364 	const struct mfc_cache *c = ptr;
365 
366 	return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
367 	       cmparg->mfc_origin != c->mfc_origin;
368 }
369 
370 static const struct rhashtable_params ipmr_rht_params = {
371 	.head_offset = offsetof(struct mr_mfc, mnode),
372 	.key_offset = offsetof(struct mfc_cache, cmparg),
373 	.key_len = sizeof(struct mfc_cache_cmp_arg),
374 	.nelem_hint = 3,
375 	.obj_cmpfn = ipmr_hash_cmp,
376 	.automatic_shrinking = true,
377 };
378 
379 static void ipmr_new_table_set(struct mr_table *mrt,
380 			       struct net *net)
381 {
382 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
383 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384 #endif
385 }
386 
387 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
388 	.mfc_mcastgrp = htonl(INADDR_ANY),
389 	.mfc_origin = htonl(INADDR_ANY),
390 };
391 
392 static struct mr_table_ops ipmr_mr_table_ops = {
393 	.rht_params = &ipmr_rht_params,
394 	.cmparg_any = &ipmr_mr_table_ops_cmparg_any,
395 };
396 
397 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
398 {
399 	struct mr_table *mrt;
400 
401 	/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
402 	if (id != RT_TABLE_DEFAULT && id >= 1000000000)
403 		return ERR_PTR(-EINVAL);
404 
405 	mrt = ipmr_get_table(net, id);
406 	if (mrt)
407 		return mrt;
408 
409 	return mr_table_alloc(net, id, &ipmr_mr_table_ops,
410 			      ipmr_expire_process, ipmr_new_table_set);
411 }
412 
413 static void ipmr_free_table(struct mr_table *mrt)
414 {
415 	del_timer_sync(&mrt->ipmr_expire_timer);
416 	mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
417 				 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
418 	rhltable_destroy(&mrt->mfc_hash);
419 	kfree(mrt);
420 }
421 
422 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
423 
424 /* Initialize ipmr pimreg/tunnel in_device */
425 static bool ipmr_init_vif_indev(const struct net_device *dev)
426 {
427 	struct in_device *in_dev;
428 
429 	ASSERT_RTNL();
430 
431 	in_dev = __in_dev_get_rtnl(dev);
432 	if (!in_dev)
433 		return false;
434 	ipv4_devconf_setall(in_dev);
435 	neigh_parms_data_state_setall(in_dev->arp_parms);
436 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
437 
438 	return true;
439 }
440 
441 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
442 {
443 	struct net_device *tunnel_dev, *new_dev;
444 	struct ip_tunnel_parm p = { };
445 	int err;
446 
447 	tunnel_dev = __dev_get_by_name(net, "tunl0");
448 	if (!tunnel_dev)
449 		goto out;
450 
451 	p.iph.daddr = v->vifc_rmt_addr.s_addr;
452 	p.iph.saddr = v->vifc_lcl_addr.s_addr;
453 	p.iph.version = 4;
454 	p.iph.ihl = 5;
455 	p.iph.protocol = IPPROTO_IPIP;
456 	sprintf(p.name, "dvmrp%d", v->vifc_vifi);
457 
458 	if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
459 		goto out;
460 	err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
461 			SIOCADDTUNNEL);
462 	if (err)
463 		goto out;
464 
465 	new_dev = __dev_get_by_name(net, p.name);
466 	if (!new_dev)
467 		goto out;
468 
469 	new_dev->flags |= IFF_MULTICAST;
470 	if (!ipmr_init_vif_indev(new_dev))
471 		goto out_unregister;
472 	if (dev_open(new_dev, NULL))
473 		goto out_unregister;
474 	dev_hold(new_dev);
475 	err = dev_set_allmulti(new_dev, 1);
476 	if (err) {
477 		dev_close(new_dev);
478 		tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
479 				SIOCDELTUNNEL);
480 		dev_put(new_dev);
481 		new_dev = ERR_PTR(err);
482 	}
483 	return new_dev;
484 
485 out_unregister:
486 	unregister_netdevice(new_dev);
487 out:
488 	return ERR_PTR(-ENOBUFS);
489 }
490 
491 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
492 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
493 {
494 	struct net *net = dev_net(dev);
495 	struct mr_table *mrt;
496 	struct flowi4 fl4 = {
497 		.flowi4_oif	= dev->ifindex,
498 		.flowi4_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
499 		.flowi4_mark	= skb->mark,
500 	};
501 	int err;
502 
503 	err = ipmr_fib_lookup(net, &fl4, &mrt);
504 	if (err < 0) {
505 		kfree_skb(skb);
506 		return err;
507 	}
508 
509 	dev->stats.tx_bytes += skb->len;
510 	dev->stats.tx_packets++;
511 	rcu_read_lock();
512 
513 	/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
514 	ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
515 			  IGMPMSG_WHOLEPKT);
516 
517 	rcu_read_unlock();
518 	kfree_skb(skb);
519 	return NETDEV_TX_OK;
520 }
521 
522 static int reg_vif_get_iflink(const struct net_device *dev)
523 {
524 	return 0;
525 }
526 
527 static const struct net_device_ops reg_vif_netdev_ops = {
528 	.ndo_start_xmit	= reg_vif_xmit,
529 	.ndo_get_iflink = reg_vif_get_iflink,
530 };
531 
532 static void reg_vif_setup(struct net_device *dev)
533 {
534 	dev->type		= ARPHRD_PIMREG;
535 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
536 	dev->flags		= IFF_NOARP;
537 	dev->netdev_ops		= &reg_vif_netdev_ops;
538 	dev->needs_free_netdev	= true;
539 	dev->features		|= NETIF_F_NETNS_LOCAL;
540 }
541 
542 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
543 {
544 	struct net_device *dev;
545 	char name[IFNAMSIZ];
546 
547 	if (mrt->id == RT_TABLE_DEFAULT)
548 		sprintf(name, "pimreg");
549 	else
550 		sprintf(name, "pimreg%u", mrt->id);
551 
552 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
553 
554 	if (!dev)
555 		return NULL;
556 
557 	dev_net_set(dev, net);
558 
559 	if (register_netdevice(dev)) {
560 		free_netdev(dev);
561 		return NULL;
562 	}
563 
564 	if (!ipmr_init_vif_indev(dev))
565 		goto failure;
566 	if (dev_open(dev, NULL))
567 		goto failure;
568 
569 	dev_hold(dev);
570 
571 	return dev;
572 
573 failure:
574 	unregister_netdevice(dev);
575 	return NULL;
576 }
577 
578 /* called with rcu_read_lock() */
579 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
580 		     unsigned int pimlen)
581 {
582 	struct net_device *reg_dev = NULL;
583 	struct iphdr *encap;
584 	int vif_num;
585 
586 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
587 	/* Check that:
588 	 * a. packet is really sent to a multicast group
589 	 * b. packet is not a NULL-REGISTER
590 	 * c. packet is not truncated
591 	 */
592 	if (!ipv4_is_multicast(encap->daddr) ||
593 	    encap->tot_len == 0 ||
594 	    ntohs(encap->tot_len) + pimlen > skb->len)
595 		return 1;
596 
597 	/* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
598 	vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
599 	if (vif_num >= 0)
600 		reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
601 	if (!reg_dev)
602 		return 1;
603 
604 	skb->mac_header = skb->network_header;
605 	skb_pull(skb, (u8 *)encap - skb->data);
606 	skb_reset_network_header(skb);
607 	skb->protocol = htons(ETH_P_IP);
608 	skb->ip_summed = CHECKSUM_NONE;
609 
610 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
611 
612 	netif_rx(skb);
613 
614 	return NET_RX_SUCCESS;
615 }
616 #else
617 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
618 {
619 	return NULL;
620 }
621 #endif
622 
623 static int call_ipmr_vif_entry_notifiers(struct net *net,
624 					 enum fib_event_type event_type,
625 					 struct vif_device *vif,
626 					 struct net_device *vif_dev,
627 					 vifi_t vif_index, u32 tb_id)
628 {
629 	return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
630 				     vif, vif_dev, vif_index, tb_id,
631 				     &net->ipv4.ipmr_seq);
632 }
633 
634 static int call_ipmr_mfc_entry_notifiers(struct net *net,
635 					 enum fib_event_type event_type,
636 					 struct mfc_cache *mfc, u32 tb_id)
637 {
638 	return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
639 				     &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
640 }
641 
642 /**
643  *	vif_delete - Delete a VIF entry
644  *	@mrt: Table to delete from
645  *	@vifi: VIF identifier to delete
646  *	@notify: Set to 1, if the caller is a notifier_call
647  *	@head: if unregistering the VIF, place it on this queue
648  */
649 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
650 		      struct list_head *head)
651 {
652 	struct net *net = read_pnet(&mrt->net);
653 	struct vif_device *v;
654 	struct net_device *dev;
655 	struct in_device *in_dev;
656 
657 	if (vifi < 0 || vifi >= mrt->maxvif)
658 		return -EADDRNOTAVAIL;
659 
660 	v = &mrt->vif_table[vifi];
661 
662 	dev = rtnl_dereference(v->dev);
663 	if (!dev)
664 		return -EADDRNOTAVAIL;
665 
666 	spin_lock(&mrt_lock);
667 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
668 				      vifi, mrt->id);
669 	RCU_INIT_POINTER(v->dev, NULL);
670 
671 	if (vifi == mrt->mroute_reg_vif_num) {
672 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
673 		WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
674 	}
675 	if (vifi + 1 == mrt->maxvif) {
676 		int tmp;
677 
678 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
679 			if (VIF_EXISTS(mrt, tmp))
680 				break;
681 		}
682 		WRITE_ONCE(mrt->maxvif, tmp + 1);
683 	}
684 
685 	spin_unlock(&mrt_lock);
686 
687 	dev_set_allmulti(dev, -1);
688 
689 	in_dev = __in_dev_get_rtnl(dev);
690 	if (in_dev) {
691 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
692 		inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
693 					    NETCONFA_MC_FORWARDING,
694 					    dev->ifindex, &in_dev->cnf);
695 		ip_rt_multicast_event(in_dev);
696 	}
697 
698 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
699 		unregister_netdevice_queue(dev, head);
700 
701 	netdev_put(dev, &v->dev_tracker);
702 	return 0;
703 }
704 
705 static void ipmr_cache_free_rcu(struct rcu_head *head)
706 {
707 	struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
708 
709 	kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
710 }
711 
712 static void ipmr_cache_free(struct mfc_cache *c)
713 {
714 	call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
715 }
716 
717 /* Destroy an unresolved cache entry, killing queued skbs
718  * and reporting error to netlink readers.
719  */
720 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
721 {
722 	struct net *net = read_pnet(&mrt->net);
723 	struct sk_buff *skb;
724 	struct nlmsgerr *e;
725 
726 	atomic_dec(&mrt->cache_resolve_queue_len);
727 
728 	while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
729 		if (ip_hdr(skb)->version == 0) {
730 			struct nlmsghdr *nlh = skb_pull(skb,
731 							sizeof(struct iphdr));
732 			nlh->nlmsg_type = NLMSG_ERROR;
733 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
734 			skb_trim(skb, nlh->nlmsg_len);
735 			e = nlmsg_data(nlh);
736 			e->error = -ETIMEDOUT;
737 			memset(&e->msg, 0, sizeof(e->msg));
738 
739 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
740 		} else {
741 			kfree_skb(skb);
742 		}
743 	}
744 
745 	ipmr_cache_free(c);
746 }
747 
748 /* Timer process for the unresolved queue. */
749 static void ipmr_expire_process(struct timer_list *t)
750 {
751 	struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
752 	struct mr_mfc *c, *next;
753 	unsigned long expires;
754 	unsigned long now;
755 
756 	if (!spin_trylock(&mfc_unres_lock)) {
757 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
758 		return;
759 	}
760 
761 	if (list_empty(&mrt->mfc_unres_queue))
762 		goto out;
763 
764 	now = jiffies;
765 	expires = 10*HZ;
766 
767 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
768 		if (time_after(c->mfc_un.unres.expires, now)) {
769 			unsigned long interval = c->mfc_un.unres.expires - now;
770 			if (interval < expires)
771 				expires = interval;
772 			continue;
773 		}
774 
775 		list_del(&c->list);
776 		mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
777 		ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
778 	}
779 
780 	if (!list_empty(&mrt->mfc_unres_queue))
781 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
782 
783 out:
784 	spin_unlock(&mfc_unres_lock);
785 }
786 
787 /* Fill oifs list. It is called under locked mrt_lock. */
788 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
789 				   unsigned char *ttls)
790 {
791 	int vifi;
792 
793 	cache->mfc_un.res.minvif = MAXVIFS;
794 	cache->mfc_un.res.maxvif = 0;
795 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
796 
797 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
798 		if (VIF_EXISTS(mrt, vifi) &&
799 		    ttls[vifi] && ttls[vifi] < 255) {
800 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
801 			if (cache->mfc_un.res.minvif > vifi)
802 				cache->mfc_un.res.minvif = vifi;
803 			if (cache->mfc_un.res.maxvif <= vifi)
804 				cache->mfc_un.res.maxvif = vifi + 1;
805 		}
806 	}
807 	cache->mfc_un.res.lastuse = jiffies;
808 }
809 
810 static int vif_add(struct net *net, struct mr_table *mrt,
811 		   struct vifctl *vifc, int mrtsock)
812 {
813 	struct netdev_phys_item_id ppid = { };
814 	int vifi = vifc->vifc_vifi;
815 	struct vif_device *v = &mrt->vif_table[vifi];
816 	struct net_device *dev;
817 	struct in_device *in_dev;
818 	int err;
819 
820 	/* Is vif busy ? */
821 	if (VIF_EXISTS(mrt, vifi))
822 		return -EADDRINUSE;
823 
824 	switch (vifc->vifc_flags) {
825 	case VIFF_REGISTER:
826 		if (!ipmr_pimsm_enabled())
827 			return -EINVAL;
828 		/* Special Purpose VIF in PIM
829 		 * All the packets will be sent to the daemon
830 		 */
831 		if (mrt->mroute_reg_vif_num >= 0)
832 			return -EADDRINUSE;
833 		dev = ipmr_reg_vif(net, mrt);
834 		if (!dev)
835 			return -ENOBUFS;
836 		err = dev_set_allmulti(dev, 1);
837 		if (err) {
838 			unregister_netdevice(dev);
839 			dev_put(dev);
840 			return err;
841 		}
842 		break;
843 	case VIFF_TUNNEL:
844 		dev = ipmr_new_tunnel(net, vifc);
845 		if (IS_ERR(dev))
846 			return PTR_ERR(dev);
847 		break;
848 	case VIFF_USE_IFINDEX:
849 	case 0:
850 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
851 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
852 			if (dev && !__in_dev_get_rtnl(dev)) {
853 				dev_put(dev);
854 				return -EADDRNOTAVAIL;
855 			}
856 		} else {
857 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
858 		}
859 		if (!dev)
860 			return -EADDRNOTAVAIL;
861 		err = dev_set_allmulti(dev, 1);
862 		if (err) {
863 			dev_put(dev);
864 			return err;
865 		}
866 		break;
867 	default:
868 		return -EINVAL;
869 	}
870 
871 	in_dev = __in_dev_get_rtnl(dev);
872 	if (!in_dev) {
873 		dev_put(dev);
874 		return -EADDRNOTAVAIL;
875 	}
876 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
877 	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
878 				    dev->ifindex, &in_dev->cnf);
879 	ip_rt_multicast_event(in_dev);
880 
881 	/* Fill in the VIF structures */
882 	vif_device_init(v, dev, vifc->vifc_rate_limit,
883 			vifc->vifc_threshold,
884 			vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
885 			(VIFF_TUNNEL | VIFF_REGISTER));
886 
887 	err = dev_get_port_parent_id(dev, &ppid, true);
888 	if (err == 0) {
889 		memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
890 		v->dev_parent_id.id_len = ppid.id_len;
891 	} else {
892 		v->dev_parent_id.id_len = 0;
893 	}
894 
895 	v->local = vifc->vifc_lcl_addr.s_addr;
896 	v->remote = vifc->vifc_rmt_addr.s_addr;
897 
898 	/* And finish update writing critical data */
899 	spin_lock(&mrt_lock);
900 	rcu_assign_pointer(v->dev, dev);
901 	netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
902 	if (v->flags & VIFF_REGISTER) {
903 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
904 		WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
905 	}
906 	if (vifi+1 > mrt->maxvif)
907 		WRITE_ONCE(mrt->maxvif, vifi + 1);
908 	spin_unlock(&mrt_lock);
909 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
910 				      vifi, mrt->id);
911 	return 0;
912 }
913 
914 /* called with rcu_read_lock() */
915 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
916 					 __be32 origin,
917 					 __be32 mcastgrp)
918 {
919 	struct mfc_cache_cmp_arg arg = {
920 			.mfc_mcastgrp = mcastgrp,
921 			.mfc_origin = origin
922 	};
923 
924 	return mr_mfc_find(mrt, &arg);
925 }
926 
927 /* Look for a (*,G) entry */
928 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
929 					     __be32 mcastgrp, int vifi)
930 {
931 	struct mfc_cache_cmp_arg arg = {
932 			.mfc_mcastgrp = mcastgrp,
933 			.mfc_origin = htonl(INADDR_ANY)
934 	};
935 
936 	if (mcastgrp == htonl(INADDR_ANY))
937 		return mr_mfc_find_any_parent(mrt, vifi);
938 	return mr_mfc_find_any(mrt, vifi, &arg);
939 }
940 
941 /* Look for a (S,G,iif) entry if parent != -1 */
942 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
943 						__be32 origin, __be32 mcastgrp,
944 						int parent)
945 {
946 	struct mfc_cache_cmp_arg arg = {
947 			.mfc_mcastgrp = mcastgrp,
948 			.mfc_origin = origin,
949 	};
950 
951 	return mr_mfc_find_parent(mrt, &arg, parent);
952 }
953 
954 /* Allocate a multicast cache entry */
955 static struct mfc_cache *ipmr_cache_alloc(void)
956 {
957 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
958 
959 	if (c) {
960 		c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
961 		c->_c.mfc_un.res.minvif = MAXVIFS;
962 		c->_c.free = ipmr_cache_free_rcu;
963 		refcount_set(&c->_c.mfc_un.res.refcount, 1);
964 	}
965 	return c;
966 }
967 
968 static struct mfc_cache *ipmr_cache_alloc_unres(void)
969 {
970 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
971 
972 	if (c) {
973 		skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
974 		c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
975 	}
976 	return c;
977 }
978 
979 /* A cache entry has gone into a resolved state from queued */
980 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
981 			       struct mfc_cache *uc, struct mfc_cache *c)
982 {
983 	struct sk_buff *skb;
984 	struct nlmsgerr *e;
985 
986 	/* Play the pending entries through our router */
987 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
988 		if (ip_hdr(skb)->version == 0) {
989 			struct nlmsghdr *nlh = skb_pull(skb,
990 							sizeof(struct iphdr));
991 
992 			if (mr_fill_mroute(mrt, skb, &c->_c,
993 					   nlmsg_data(nlh)) > 0) {
994 				nlh->nlmsg_len = skb_tail_pointer(skb) -
995 						 (u8 *)nlh;
996 			} else {
997 				nlh->nlmsg_type = NLMSG_ERROR;
998 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
999 				skb_trim(skb, nlh->nlmsg_len);
1000 				e = nlmsg_data(nlh);
1001 				e->error = -EMSGSIZE;
1002 				memset(&e->msg, 0, sizeof(e->msg));
1003 			}
1004 
1005 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1006 		} else {
1007 			rcu_read_lock();
1008 			ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1009 			rcu_read_unlock();
1010 		}
1011 	}
1012 }
1013 
1014 /* Bounce a cache query up to mrouted and netlink.
1015  *
1016  * Called under rcu_read_lock().
1017  */
1018 static int ipmr_cache_report(const struct mr_table *mrt,
1019 			     struct sk_buff *pkt, vifi_t vifi, int assert)
1020 {
1021 	const int ihl = ip_hdrlen(pkt);
1022 	struct sock *mroute_sk;
1023 	struct igmphdr *igmp;
1024 	struct igmpmsg *msg;
1025 	struct sk_buff *skb;
1026 	int ret;
1027 
1028 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1029 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1030 	else
1031 		skb = alloc_skb(128, GFP_ATOMIC);
1032 
1033 	if (!skb)
1034 		return -ENOBUFS;
1035 
1036 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1037 		/* Ugly, but we have no choice with this interface.
1038 		 * Duplicate old header, fix ihl, length etc.
1039 		 * And all this only to mangle msg->im_msgtype and
1040 		 * to set msg->im_mbz to "mbz" :-)
1041 		 */
1042 		skb_push(skb, sizeof(struct iphdr));
1043 		skb_reset_network_header(skb);
1044 		skb_reset_transport_header(skb);
1045 		msg = (struct igmpmsg *)skb_network_header(skb);
1046 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1047 		msg->im_msgtype = assert;
1048 		msg->im_mbz = 0;
1049 		if (assert == IGMPMSG_WRVIFWHOLE) {
1050 			msg->im_vif = vifi;
1051 			msg->im_vif_hi = vifi >> 8;
1052 		} else {
1053 			/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1054 			int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1055 
1056 			msg->im_vif = vif_num;
1057 			msg->im_vif_hi = vif_num >> 8;
1058 		}
1059 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1060 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1061 					     sizeof(struct iphdr));
1062 	} else {
1063 		/* Copy the IP header */
1064 		skb_set_network_header(skb, skb->len);
1065 		skb_put(skb, ihl);
1066 		skb_copy_to_linear_data(skb, pkt->data, ihl);
1067 		/* Flag to the kernel this is a route add */
1068 		ip_hdr(skb)->protocol = 0;
1069 		msg = (struct igmpmsg *)skb_network_header(skb);
1070 		msg->im_vif = vifi;
1071 		msg->im_vif_hi = vifi >> 8;
1072 		skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1073 		/* Add our header */
1074 		igmp = skb_put(skb, sizeof(struct igmphdr));
1075 		igmp->type = assert;
1076 		msg->im_msgtype = assert;
1077 		igmp->code = 0;
1078 		ip_hdr(skb)->tot_len = htons(skb->len);	/* Fix the length */
1079 		skb->transport_header = skb->network_header;
1080 	}
1081 
1082 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1083 	if (!mroute_sk) {
1084 		kfree_skb(skb);
1085 		return -EINVAL;
1086 	}
1087 
1088 	igmpmsg_netlink_event(mrt, skb);
1089 
1090 	/* Deliver to mrouted */
1091 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1092 
1093 	if (ret < 0) {
1094 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1095 		kfree_skb(skb);
1096 	}
1097 
1098 	return ret;
1099 }
1100 
1101 /* Queue a packet for resolution. It gets locked cache entry! */
1102 /* Called under rcu_read_lock() */
1103 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1104 				 struct sk_buff *skb, struct net_device *dev)
1105 {
1106 	const struct iphdr *iph = ip_hdr(skb);
1107 	struct mfc_cache *c;
1108 	bool found = false;
1109 	int err;
1110 
1111 	spin_lock_bh(&mfc_unres_lock);
1112 	list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1113 		if (c->mfc_mcastgrp == iph->daddr &&
1114 		    c->mfc_origin == iph->saddr) {
1115 			found = true;
1116 			break;
1117 		}
1118 	}
1119 
1120 	if (!found) {
1121 		/* Create a new entry if allowable */
1122 		c = ipmr_cache_alloc_unres();
1123 		if (!c) {
1124 			spin_unlock_bh(&mfc_unres_lock);
1125 
1126 			kfree_skb(skb);
1127 			return -ENOBUFS;
1128 		}
1129 
1130 		/* Fill in the new cache entry */
1131 		c->_c.mfc_parent = -1;
1132 		c->mfc_origin	= iph->saddr;
1133 		c->mfc_mcastgrp	= iph->daddr;
1134 
1135 		/* Reflect first query at mrouted. */
1136 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1137 
1138 		if (err < 0) {
1139 			/* If the report failed throw the cache entry
1140 			   out - Brad Parker
1141 			 */
1142 			spin_unlock_bh(&mfc_unres_lock);
1143 
1144 			ipmr_cache_free(c);
1145 			kfree_skb(skb);
1146 			return err;
1147 		}
1148 
1149 		atomic_inc(&mrt->cache_resolve_queue_len);
1150 		list_add(&c->_c.list, &mrt->mfc_unres_queue);
1151 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1152 
1153 		if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1154 			mod_timer(&mrt->ipmr_expire_timer,
1155 				  c->_c.mfc_un.unres.expires);
1156 	}
1157 
1158 	/* See if we can append the packet */
1159 	if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1160 		kfree_skb(skb);
1161 		err = -ENOBUFS;
1162 	} else {
1163 		if (dev) {
1164 			skb->dev = dev;
1165 			skb->skb_iif = dev->ifindex;
1166 		}
1167 		skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1168 		err = 0;
1169 	}
1170 
1171 	spin_unlock_bh(&mfc_unres_lock);
1172 	return err;
1173 }
1174 
1175 /* MFC cache manipulation by user space mroute daemon */
1176 
1177 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1178 {
1179 	struct net *net = read_pnet(&mrt->net);
1180 	struct mfc_cache *c;
1181 
1182 	/* The entries are added/deleted only under RTNL */
1183 	rcu_read_lock();
1184 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1185 				   mfc->mfcc_mcastgrp.s_addr, parent);
1186 	rcu_read_unlock();
1187 	if (!c)
1188 		return -ENOENT;
1189 	rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1190 	list_del_rcu(&c->_c.list);
1191 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1192 	mroute_netlink_event(mrt, c, RTM_DELROUTE);
1193 	mr_cache_put(&c->_c);
1194 
1195 	return 0;
1196 }
1197 
1198 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1199 			struct mfcctl *mfc, int mrtsock, int parent)
1200 {
1201 	struct mfc_cache *uc, *c;
1202 	struct mr_mfc *_uc;
1203 	bool found;
1204 	int ret;
1205 
1206 	if (mfc->mfcc_parent >= MAXVIFS)
1207 		return -ENFILE;
1208 
1209 	/* The entries are added/deleted only under RTNL */
1210 	rcu_read_lock();
1211 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1212 				   mfc->mfcc_mcastgrp.s_addr, parent);
1213 	rcu_read_unlock();
1214 	if (c) {
1215 		spin_lock(&mrt_lock);
1216 		c->_c.mfc_parent = mfc->mfcc_parent;
1217 		ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1218 		if (!mrtsock)
1219 			c->_c.mfc_flags |= MFC_STATIC;
1220 		spin_unlock(&mrt_lock);
1221 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1222 					      mrt->id);
1223 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1224 		return 0;
1225 	}
1226 
1227 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1228 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1229 		return -EINVAL;
1230 
1231 	c = ipmr_cache_alloc();
1232 	if (!c)
1233 		return -ENOMEM;
1234 
1235 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1236 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1237 	c->_c.mfc_parent = mfc->mfcc_parent;
1238 	ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1239 	if (!mrtsock)
1240 		c->_c.mfc_flags |= MFC_STATIC;
1241 
1242 	ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1243 				  ipmr_rht_params);
1244 	if (ret) {
1245 		pr_err("ipmr: rhtable insert error %d\n", ret);
1246 		ipmr_cache_free(c);
1247 		return ret;
1248 	}
1249 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1250 	/* Check to see if we resolved a queued list. If so we
1251 	 * need to send on the frames and tidy up.
1252 	 */
1253 	found = false;
1254 	spin_lock_bh(&mfc_unres_lock);
1255 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1256 		uc = (struct mfc_cache *)_uc;
1257 		if (uc->mfc_origin == c->mfc_origin &&
1258 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1259 			list_del(&_uc->list);
1260 			atomic_dec(&mrt->cache_resolve_queue_len);
1261 			found = true;
1262 			break;
1263 		}
1264 	}
1265 	if (list_empty(&mrt->mfc_unres_queue))
1266 		del_timer(&mrt->ipmr_expire_timer);
1267 	spin_unlock_bh(&mfc_unres_lock);
1268 
1269 	if (found) {
1270 		ipmr_cache_resolve(net, mrt, uc, c);
1271 		ipmr_cache_free(uc);
1272 	}
1273 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1274 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1275 	return 0;
1276 }
1277 
1278 /* Close the multicast socket, and clear the vif tables etc */
1279 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1280 {
1281 	struct net *net = read_pnet(&mrt->net);
1282 	struct mr_mfc *c, *tmp;
1283 	struct mfc_cache *cache;
1284 	LIST_HEAD(list);
1285 	int i;
1286 
1287 	/* Shut down all active vif entries */
1288 	if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1289 		for (i = 0; i < mrt->maxvif; i++) {
1290 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1291 			     !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1292 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1293 				continue;
1294 			vif_delete(mrt, i, 0, &list);
1295 		}
1296 		unregister_netdevice_many(&list);
1297 	}
1298 
1299 	/* Wipe the cache */
1300 	if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1301 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1302 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1303 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1304 				continue;
1305 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1306 			list_del_rcu(&c->list);
1307 			cache = (struct mfc_cache *)c;
1308 			call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1309 						      mrt->id);
1310 			mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1311 			mr_cache_put(c);
1312 		}
1313 	}
1314 
1315 	if (flags & MRT_FLUSH_MFC) {
1316 		if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1317 			spin_lock_bh(&mfc_unres_lock);
1318 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1319 				list_del(&c->list);
1320 				cache = (struct mfc_cache *)c;
1321 				mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1322 				ipmr_destroy_unres(mrt, cache);
1323 			}
1324 			spin_unlock_bh(&mfc_unres_lock);
1325 		}
1326 	}
1327 }
1328 
1329 /* called from ip_ra_control(), before an RCU grace period,
1330  * we don't need to call synchronize_rcu() here
1331  */
1332 static void mrtsock_destruct(struct sock *sk)
1333 {
1334 	struct net *net = sock_net(sk);
1335 	struct mr_table *mrt;
1336 
1337 	rtnl_lock();
1338 	ipmr_for_each_table(mrt, net) {
1339 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1340 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1341 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1342 						    NETCONFA_MC_FORWARDING,
1343 						    NETCONFA_IFINDEX_ALL,
1344 						    net->ipv4.devconf_all);
1345 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1346 			mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1347 		}
1348 	}
1349 	rtnl_unlock();
1350 }
1351 
1352 /* Socket options and virtual interface manipulation. The whole
1353  * virtual interface system is a complete heap, but unfortunately
1354  * that's how BSD mrouted happens to think. Maybe one day with a proper
1355  * MOSPF/PIM router set up we can clean this up.
1356  */
1357 
1358 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1359 			 unsigned int optlen)
1360 {
1361 	struct net *net = sock_net(sk);
1362 	int val, ret = 0, parent = 0;
1363 	struct mr_table *mrt;
1364 	struct vifctl vif;
1365 	struct mfcctl mfc;
1366 	bool do_wrvifwhole;
1367 	u32 uval;
1368 
1369 	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1370 	rtnl_lock();
1371 	if (sk->sk_type != SOCK_RAW ||
1372 	    inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1373 		ret = -EOPNOTSUPP;
1374 		goto out_unlock;
1375 	}
1376 
1377 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1378 	if (!mrt) {
1379 		ret = -ENOENT;
1380 		goto out_unlock;
1381 	}
1382 	if (optname != MRT_INIT) {
1383 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1384 		    !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1385 			ret = -EACCES;
1386 			goto out_unlock;
1387 		}
1388 	}
1389 
1390 	switch (optname) {
1391 	case MRT_INIT:
1392 		if (optlen != sizeof(int)) {
1393 			ret = -EINVAL;
1394 			break;
1395 		}
1396 		if (rtnl_dereference(mrt->mroute_sk)) {
1397 			ret = -EADDRINUSE;
1398 			break;
1399 		}
1400 
1401 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1402 		if (ret == 0) {
1403 			rcu_assign_pointer(mrt->mroute_sk, sk);
1404 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1405 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1406 						    NETCONFA_MC_FORWARDING,
1407 						    NETCONFA_IFINDEX_ALL,
1408 						    net->ipv4.devconf_all);
1409 		}
1410 		break;
1411 	case MRT_DONE:
1412 		if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1413 			ret = -EACCES;
1414 		} else {
1415 			/* We need to unlock here because mrtsock_destruct takes
1416 			 * care of rtnl itself and we can't change that due to
1417 			 * the IP_ROUTER_ALERT setsockopt which runs without it.
1418 			 */
1419 			rtnl_unlock();
1420 			ret = ip_ra_control(sk, 0, NULL);
1421 			goto out;
1422 		}
1423 		break;
1424 	case MRT_ADD_VIF:
1425 	case MRT_DEL_VIF:
1426 		if (optlen != sizeof(vif)) {
1427 			ret = -EINVAL;
1428 			break;
1429 		}
1430 		if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1431 			ret = -EFAULT;
1432 			break;
1433 		}
1434 		if (vif.vifc_vifi >= MAXVIFS) {
1435 			ret = -ENFILE;
1436 			break;
1437 		}
1438 		if (optname == MRT_ADD_VIF) {
1439 			ret = vif_add(net, mrt, &vif,
1440 				      sk == rtnl_dereference(mrt->mroute_sk));
1441 		} else {
1442 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1443 		}
1444 		break;
1445 	/* Manipulate the forwarding caches. These live
1446 	 * in a sort of kernel/user symbiosis.
1447 	 */
1448 	case MRT_ADD_MFC:
1449 	case MRT_DEL_MFC:
1450 		parent = -1;
1451 		fallthrough;
1452 	case MRT_ADD_MFC_PROXY:
1453 	case MRT_DEL_MFC_PROXY:
1454 		if (optlen != sizeof(mfc)) {
1455 			ret = -EINVAL;
1456 			break;
1457 		}
1458 		if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1459 			ret = -EFAULT;
1460 			break;
1461 		}
1462 		if (parent == 0)
1463 			parent = mfc.mfcc_parent;
1464 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1465 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1466 		else
1467 			ret = ipmr_mfc_add(net, mrt, &mfc,
1468 					   sk == rtnl_dereference(mrt->mroute_sk),
1469 					   parent);
1470 		break;
1471 	case MRT_FLUSH:
1472 		if (optlen != sizeof(val)) {
1473 			ret = -EINVAL;
1474 			break;
1475 		}
1476 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1477 			ret = -EFAULT;
1478 			break;
1479 		}
1480 		mroute_clean_tables(mrt, val);
1481 		break;
1482 	/* Control PIM assert. */
1483 	case MRT_ASSERT:
1484 		if (optlen != sizeof(val)) {
1485 			ret = -EINVAL;
1486 			break;
1487 		}
1488 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1489 			ret = -EFAULT;
1490 			break;
1491 		}
1492 		mrt->mroute_do_assert = val;
1493 		break;
1494 	case MRT_PIM:
1495 		if (!ipmr_pimsm_enabled()) {
1496 			ret = -ENOPROTOOPT;
1497 			break;
1498 		}
1499 		if (optlen != sizeof(val)) {
1500 			ret = -EINVAL;
1501 			break;
1502 		}
1503 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1504 			ret = -EFAULT;
1505 			break;
1506 		}
1507 
1508 		do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1509 		val = !!val;
1510 		if (val != mrt->mroute_do_pim) {
1511 			mrt->mroute_do_pim = val;
1512 			mrt->mroute_do_assert = val;
1513 			mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1514 		}
1515 		break;
1516 	case MRT_TABLE:
1517 		if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1518 			ret = -ENOPROTOOPT;
1519 			break;
1520 		}
1521 		if (optlen != sizeof(uval)) {
1522 			ret = -EINVAL;
1523 			break;
1524 		}
1525 		if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1526 			ret = -EFAULT;
1527 			break;
1528 		}
1529 
1530 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1531 			ret = -EBUSY;
1532 		} else {
1533 			mrt = ipmr_new_table(net, uval);
1534 			if (IS_ERR(mrt))
1535 				ret = PTR_ERR(mrt);
1536 			else
1537 				raw_sk(sk)->ipmr_table = uval;
1538 		}
1539 		break;
1540 	/* Spurious command, or MRT_VERSION which you cannot set. */
1541 	default:
1542 		ret = -ENOPROTOOPT;
1543 	}
1544 out_unlock:
1545 	rtnl_unlock();
1546 out:
1547 	return ret;
1548 }
1549 
1550 /* Getsock opt support for the multicast routing system. */
1551 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1552 {
1553 	int olr;
1554 	int val;
1555 	struct net *net = sock_net(sk);
1556 	struct mr_table *mrt;
1557 
1558 	if (sk->sk_type != SOCK_RAW ||
1559 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1560 		return -EOPNOTSUPP;
1561 
1562 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1563 	if (!mrt)
1564 		return -ENOENT;
1565 
1566 	switch (optname) {
1567 	case MRT_VERSION:
1568 		val = 0x0305;
1569 		break;
1570 	case MRT_PIM:
1571 		if (!ipmr_pimsm_enabled())
1572 			return -ENOPROTOOPT;
1573 		val = mrt->mroute_do_pim;
1574 		break;
1575 	case MRT_ASSERT:
1576 		val = mrt->mroute_do_assert;
1577 		break;
1578 	default:
1579 		return -ENOPROTOOPT;
1580 	}
1581 
1582 	if (get_user(olr, optlen))
1583 		return -EFAULT;
1584 	olr = min_t(unsigned int, olr, sizeof(int));
1585 	if (olr < 0)
1586 		return -EINVAL;
1587 	if (put_user(olr, optlen))
1588 		return -EFAULT;
1589 	if (copy_to_user(optval, &val, olr))
1590 		return -EFAULT;
1591 	return 0;
1592 }
1593 
1594 /* The IP multicast ioctl support routines. */
1595 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1596 {
1597 	struct sioc_sg_req sr;
1598 	struct sioc_vif_req vr;
1599 	struct vif_device *vif;
1600 	struct mfc_cache *c;
1601 	struct net *net = sock_net(sk);
1602 	struct mr_table *mrt;
1603 
1604 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1605 	if (!mrt)
1606 		return -ENOENT;
1607 
1608 	switch (cmd) {
1609 	case SIOCGETVIFCNT:
1610 		if (copy_from_user(&vr, arg, sizeof(vr)))
1611 			return -EFAULT;
1612 		if (vr.vifi >= mrt->maxvif)
1613 			return -EINVAL;
1614 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1615 		rcu_read_lock();
1616 		vif = &mrt->vif_table[vr.vifi];
1617 		if (VIF_EXISTS(mrt, vr.vifi)) {
1618 			vr.icount = READ_ONCE(vif->pkt_in);
1619 			vr.ocount = READ_ONCE(vif->pkt_out);
1620 			vr.ibytes = READ_ONCE(vif->bytes_in);
1621 			vr.obytes = READ_ONCE(vif->bytes_out);
1622 			rcu_read_unlock();
1623 
1624 			if (copy_to_user(arg, &vr, sizeof(vr)))
1625 				return -EFAULT;
1626 			return 0;
1627 		}
1628 		rcu_read_unlock();
1629 		return -EADDRNOTAVAIL;
1630 	case SIOCGETSGCNT:
1631 		if (copy_from_user(&sr, arg, sizeof(sr)))
1632 			return -EFAULT;
1633 
1634 		rcu_read_lock();
1635 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1636 		if (c) {
1637 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1638 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1639 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1640 			rcu_read_unlock();
1641 
1642 			if (copy_to_user(arg, &sr, sizeof(sr)))
1643 				return -EFAULT;
1644 			return 0;
1645 		}
1646 		rcu_read_unlock();
1647 		return -EADDRNOTAVAIL;
1648 	default:
1649 		return -ENOIOCTLCMD;
1650 	}
1651 }
1652 
1653 #ifdef CONFIG_COMPAT
1654 struct compat_sioc_sg_req {
1655 	struct in_addr src;
1656 	struct in_addr grp;
1657 	compat_ulong_t pktcnt;
1658 	compat_ulong_t bytecnt;
1659 	compat_ulong_t wrong_if;
1660 };
1661 
1662 struct compat_sioc_vif_req {
1663 	vifi_t	vifi;		/* Which iface */
1664 	compat_ulong_t icount;
1665 	compat_ulong_t ocount;
1666 	compat_ulong_t ibytes;
1667 	compat_ulong_t obytes;
1668 };
1669 
1670 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1671 {
1672 	struct compat_sioc_sg_req sr;
1673 	struct compat_sioc_vif_req vr;
1674 	struct vif_device *vif;
1675 	struct mfc_cache *c;
1676 	struct net *net = sock_net(sk);
1677 	struct mr_table *mrt;
1678 
1679 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1680 	if (!mrt)
1681 		return -ENOENT;
1682 
1683 	switch (cmd) {
1684 	case SIOCGETVIFCNT:
1685 		if (copy_from_user(&vr, arg, sizeof(vr)))
1686 			return -EFAULT;
1687 		if (vr.vifi >= mrt->maxvif)
1688 			return -EINVAL;
1689 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1690 		rcu_read_lock();
1691 		vif = &mrt->vif_table[vr.vifi];
1692 		if (VIF_EXISTS(mrt, vr.vifi)) {
1693 			vr.icount = READ_ONCE(vif->pkt_in);
1694 			vr.ocount = READ_ONCE(vif->pkt_out);
1695 			vr.ibytes = READ_ONCE(vif->bytes_in);
1696 			vr.obytes = READ_ONCE(vif->bytes_out);
1697 			rcu_read_unlock();
1698 
1699 			if (copy_to_user(arg, &vr, sizeof(vr)))
1700 				return -EFAULT;
1701 			return 0;
1702 		}
1703 		rcu_read_unlock();
1704 		return -EADDRNOTAVAIL;
1705 	case SIOCGETSGCNT:
1706 		if (copy_from_user(&sr, arg, sizeof(sr)))
1707 			return -EFAULT;
1708 
1709 		rcu_read_lock();
1710 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1711 		if (c) {
1712 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1713 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1714 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1715 			rcu_read_unlock();
1716 
1717 			if (copy_to_user(arg, &sr, sizeof(sr)))
1718 				return -EFAULT;
1719 			return 0;
1720 		}
1721 		rcu_read_unlock();
1722 		return -EADDRNOTAVAIL;
1723 	default:
1724 		return -ENOIOCTLCMD;
1725 	}
1726 }
1727 #endif
1728 
1729 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1730 {
1731 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1732 	struct net *net = dev_net(dev);
1733 	struct mr_table *mrt;
1734 	struct vif_device *v;
1735 	int ct;
1736 
1737 	if (event != NETDEV_UNREGISTER)
1738 		return NOTIFY_DONE;
1739 
1740 	ipmr_for_each_table(mrt, net) {
1741 		v = &mrt->vif_table[0];
1742 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1743 			if (rcu_access_pointer(v->dev) == dev)
1744 				vif_delete(mrt, ct, 1, NULL);
1745 		}
1746 	}
1747 	return NOTIFY_DONE;
1748 }
1749 
1750 static struct notifier_block ip_mr_notifier = {
1751 	.notifier_call = ipmr_device_event,
1752 };
1753 
1754 /* Encapsulate a packet by attaching a valid IPIP header to it.
1755  * This avoids tunnel drivers and other mess and gives us the speed so
1756  * important for multicast video.
1757  */
1758 static void ip_encap(struct net *net, struct sk_buff *skb,
1759 		     __be32 saddr, __be32 daddr)
1760 {
1761 	struct iphdr *iph;
1762 	const struct iphdr *old_iph = ip_hdr(skb);
1763 
1764 	skb_push(skb, sizeof(struct iphdr));
1765 	skb->transport_header = skb->network_header;
1766 	skb_reset_network_header(skb);
1767 	iph = ip_hdr(skb);
1768 
1769 	iph->version	=	4;
1770 	iph->tos	=	old_iph->tos;
1771 	iph->ttl	=	old_iph->ttl;
1772 	iph->frag_off	=	0;
1773 	iph->daddr	=	daddr;
1774 	iph->saddr	=	saddr;
1775 	iph->protocol	=	IPPROTO_IPIP;
1776 	iph->ihl	=	5;
1777 	iph->tot_len	=	htons(skb->len);
1778 	ip_select_ident(net, skb, NULL);
1779 	ip_send_check(iph);
1780 
1781 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1782 	nf_reset_ct(skb);
1783 }
1784 
1785 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1786 				      struct sk_buff *skb)
1787 {
1788 	struct ip_options *opt = &(IPCB(skb)->opt);
1789 
1790 	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1791 	IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1792 
1793 	if (unlikely(opt->optlen))
1794 		ip_forward_options(skb);
1795 
1796 	return dst_output(net, sk, skb);
1797 }
1798 
1799 #ifdef CONFIG_NET_SWITCHDEV
1800 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1801 				   int in_vifi, int out_vifi)
1802 {
1803 	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1804 	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1805 
1806 	if (!skb->offload_l3_fwd_mark)
1807 		return false;
1808 	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1809 		return false;
1810 	return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1811 					&in_vif->dev_parent_id);
1812 }
1813 #else
1814 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1815 				   int in_vifi, int out_vifi)
1816 {
1817 	return false;
1818 }
1819 #endif
1820 
1821 /* Processing handlers for ipmr_forward, under rcu_read_lock() */
1822 
1823 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1824 			    int in_vifi, struct sk_buff *skb, int vifi)
1825 {
1826 	const struct iphdr *iph = ip_hdr(skb);
1827 	struct vif_device *vif = &mrt->vif_table[vifi];
1828 	struct net_device *vif_dev;
1829 	struct net_device *dev;
1830 	struct rtable *rt;
1831 	struct flowi4 fl4;
1832 	int    encap = 0;
1833 
1834 	vif_dev = vif_dev_read(vif);
1835 	if (!vif_dev)
1836 		goto out_free;
1837 
1838 	if (vif->flags & VIFF_REGISTER) {
1839 		WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1840 		WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1841 		vif_dev->stats.tx_bytes += skb->len;
1842 		vif_dev->stats.tx_packets++;
1843 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1844 		goto out_free;
1845 	}
1846 
1847 	if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1848 		goto out_free;
1849 
1850 	if (vif->flags & VIFF_TUNNEL) {
1851 		rt = ip_route_output_ports(net, &fl4, NULL,
1852 					   vif->remote, vif->local,
1853 					   0, 0,
1854 					   IPPROTO_IPIP,
1855 					   RT_TOS(iph->tos), vif->link);
1856 		if (IS_ERR(rt))
1857 			goto out_free;
1858 		encap = sizeof(struct iphdr);
1859 	} else {
1860 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1861 					   0, 0,
1862 					   IPPROTO_IPIP,
1863 					   RT_TOS(iph->tos), vif->link);
1864 		if (IS_ERR(rt))
1865 			goto out_free;
1866 	}
1867 
1868 	dev = rt->dst.dev;
1869 
1870 	if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1871 		/* Do not fragment multicasts. Alas, IPv4 does not
1872 		 * allow to send ICMP, so that packets will disappear
1873 		 * to blackhole.
1874 		 */
1875 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1876 		ip_rt_put(rt);
1877 		goto out_free;
1878 	}
1879 
1880 	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1881 
1882 	if (skb_cow(skb, encap)) {
1883 		ip_rt_put(rt);
1884 		goto out_free;
1885 	}
1886 
1887 	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1888 	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1889 
1890 	skb_dst_drop(skb);
1891 	skb_dst_set(skb, &rt->dst);
1892 	ip_decrease_ttl(ip_hdr(skb));
1893 
1894 	/* FIXME: forward and output firewalls used to be called here.
1895 	 * What do we do with netfilter? -- RR
1896 	 */
1897 	if (vif->flags & VIFF_TUNNEL) {
1898 		ip_encap(net, skb, vif->local, vif->remote);
1899 		/* FIXME: extra output firewall step used to be here. --RR */
1900 		vif_dev->stats.tx_packets++;
1901 		vif_dev->stats.tx_bytes += skb->len;
1902 	}
1903 
1904 	IPCB(skb)->flags |= IPSKB_FORWARDED;
1905 
1906 	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1907 	 * not only before forwarding, but after forwarding on all output
1908 	 * interfaces. It is clear, if mrouter runs a multicasting
1909 	 * program, it should receive packets not depending to what interface
1910 	 * program is joined.
1911 	 * If we will not make it, the program will have to join on all
1912 	 * interfaces. On the other hand, multihoming host (or router, but
1913 	 * not mrouter) cannot join to more than one interface - it will
1914 	 * result in receiving multiple packets.
1915 	 */
1916 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1917 		net, NULL, skb, skb->dev, dev,
1918 		ipmr_forward_finish);
1919 	return;
1920 
1921 out_free:
1922 	kfree_skb(skb);
1923 }
1924 
1925 /* Called with mrt_lock or rcu_read_lock() */
1926 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1927 {
1928 	int ct;
1929 	/* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1930 	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1931 		if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1932 			break;
1933 	}
1934 	return ct;
1935 }
1936 
1937 /* "local" means that we should preserve one skb (for local delivery) */
1938 /* Called uner rcu_read_lock() */
1939 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1940 			  struct net_device *dev, struct sk_buff *skb,
1941 			  struct mfc_cache *c, int local)
1942 {
1943 	int true_vifi = ipmr_find_vif(mrt, dev);
1944 	int psend = -1;
1945 	int vif, ct;
1946 
1947 	vif = c->_c.mfc_parent;
1948 	c->_c.mfc_un.res.pkt++;
1949 	c->_c.mfc_un.res.bytes += skb->len;
1950 	c->_c.mfc_un.res.lastuse = jiffies;
1951 
1952 	if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1953 		struct mfc_cache *cache_proxy;
1954 
1955 		/* For an (*,G) entry, we only check that the incoming
1956 		 * interface is part of the static tree.
1957 		 */
1958 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1959 		if (cache_proxy &&
1960 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1961 			goto forward;
1962 	}
1963 
1964 	/* Wrong interface: drop packet and (maybe) send PIM assert. */
1965 	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
1966 		if (rt_is_output_route(skb_rtable(skb))) {
1967 			/* It is our own packet, looped back.
1968 			 * Very complicated situation...
1969 			 *
1970 			 * The best workaround until routing daemons will be
1971 			 * fixed is not to redistribute packet, if it was
1972 			 * send through wrong interface. It means, that
1973 			 * multicast applications WILL NOT work for
1974 			 * (S,G), which have default multicast route pointing
1975 			 * to wrong oif. In any case, it is not a good
1976 			 * idea to use multicasting applications on router.
1977 			 */
1978 			goto dont_forward;
1979 		}
1980 
1981 		c->_c.mfc_un.res.wrong_if++;
1982 
1983 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
1984 		    /* pimsm uses asserts, when switching from RPT to SPT,
1985 		     * so that we cannot check that packet arrived on an oif.
1986 		     * It is bad, but otherwise we would need to move pretty
1987 		     * large chunk of pimd to kernel. Ough... --ANK
1988 		     */
1989 		    (mrt->mroute_do_pim ||
1990 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1991 		    time_after(jiffies,
1992 			       c->_c.mfc_un.res.last_assert +
1993 			       MFC_ASSERT_THRESH)) {
1994 			c->_c.mfc_un.res.last_assert = jiffies;
1995 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1996 			if (mrt->mroute_do_wrvifwhole)
1997 				ipmr_cache_report(mrt, skb, true_vifi,
1998 						  IGMPMSG_WRVIFWHOLE);
1999 		}
2000 		goto dont_forward;
2001 	}
2002 
2003 forward:
2004 	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2005 		   mrt->vif_table[vif].pkt_in + 1);
2006 	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2007 		   mrt->vif_table[vif].bytes_in + skb->len);
2008 
2009 	/* Forward the frame */
2010 	if (c->mfc_origin == htonl(INADDR_ANY) &&
2011 	    c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2012 		if (true_vifi >= 0 &&
2013 		    true_vifi != c->_c.mfc_parent &&
2014 		    ip_hdr(skb)->ttl >
2015 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2016 			/* It's an (*,*) entry and the packet is not coming from
2017 			 * the upstream: forward the packet to the upstream
2018 			 * only.
2019 			 */
2020 			psend = c->_c.mfc_parent;
2021 			goto last_forward;
2022 		}
2023 		goto dont_forward;
2024 	}
2025 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2026 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2027 		/* For (*,G) entry, don't forward to the incoming interface */
2028 		if ((c->mfc_origin != htonl(INADDR_ANY) ||
2029 		     ct != true_vifi) &&
2030 		    ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2031 			if (psend != -1) {
2032 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2033 
2034 				if (skb2)
2035 					ipmr_queue_xmit(net, mrt, true_vifi,
2036 							skb2, psend);
2037 			}
2038 			psend = ct;
2039 		}
2040 	}
2041 last_forward:
2042 	if (psend != -1) {
2043 		if (local) {
2044 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2045 
2046 			if (skb2)
2047 				ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2048 						psend);
2049 		} else {
2050 			ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2051 			return;
2052 		}
2053 	}
2054 
2055 dont_forward:
2056 	if (!local)
2057 		kfree_skb(skb);
2058 }
2059 
2060 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2061 {
2062 	struct rtable *rt = skb_rtable(skb);
2063 	struct iphdr *iph = ip_hdr(skb);
2064 	struct flowi4 fl4 = {
2065 		.daddr = iph->daddr,
2066 		.saddr = iph->saddr,
2067 		.flowi4_tos = RT_TOS(iph->tos),
2068 		.flowi4_oif = (rt_is_output_route(rt) ?
2069 			       skb->dev->ifindex : 0),
2070 		.flowi4_iif = (rt_is_output_route(rt) ?
2071 			       LOOPBACK_IFINDEX :
2072 			       skb->dev->ifindex),
2073 		.flowi4_mark = skb->mark,
2074 	};
2075 	struct mr_table *mrt;
2076 	int err;
2077 
2078 	err = ipmr_fib_lookup(net, &fl4, &mrt);
2079 	if (err)
2080 		return ERR_PTR(err);
2081 	return mrt;
2082 }
2083 
2084 /* Multicast packets for forwarding arrive here
2085  * Called with rcu_read_lock();
2086  */
2087 int ip_mr_input(struct sk_buff *skb)
2088 {
2089 	struct mfc_cache *cache;
2090 	struct net *net = dev_net(skb->dev);
2091 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2092 	struct mr_table *mrt;
2093 	struct net_device *dev;
2094 
2095 	/* skb->dev passed in is the loX master dev for vrfs.
2096 	 * As there are no vifs associated with loopback devices,
2097 	 * get the proper interface that does have a vif associated with it.
2098 	 */
2099 	dev = skb->dev;
2100 	if (netif_is_l3_master(skb->dev)) {
2101 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2102 		if (!dev) {
2103 			kfree_skb(skb);
2104 			return -ENODEV;
2105 		}
2106 	}
2107 
2108 	/* Packet is looped back after forward, it should not be
2109 	 * forwarded second time, but still can be delivered locally.
2110 	 */
2111 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2112 		goto dont_forward;
2113 
2114 	mrt = ipmr_rt_fib_lookup(net, skb);
2115 	if (IS_ERR(mrt)) {
2116 		kfree_skb(skb);
2117 		return PTR_ERR(mrt);
2118 	}
2119 	if (!local) {
2120 		if (IPCB(skb)->opt.router_alert) {
2121 			if (ip_call_ra_chain(skb))
2122 				return 0;
2123 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2124 			/* IGMPv1 (and broken IGMPv2 implementations sort of
2125 			 * Cisco IOS <= 11.2(8)) do not put router alert
2126 			 * option to IGMP packets destined to routable
2127 			 * groups. It is very bad, because it means
2128 			 * that we can forward NO IGMP messages.
2129 			 */
2130 			struct sock *mroute_sk;
2131 
2132 			mroute_sk = rcu_dereference(mrt->mroute_sk);
2133 			if (mroute_sk) {
2134 				nf_reset_ct(skb);
2135 				raw_rcv(mroute_sk, skb);
2136 				return 0;
2137 			}
2138 		}
2139 	}
2140 
2141 	/* already under rcu_read_lock() */
2142 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2143 	if (!cache) {
2144 		int vif = ipmr_find_vif(mrt, dev);
2145 
2146 		if (vif >= 0)
2147 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2148 						    vif);
2149 	}
2150 
2151 	/* No usable cache entry */
2152 	if (!cache) {
2153 		int vif;
2154 
2155 		if (local) {
2156 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2157 			ip_local_deliver(skb);
2158 			if (!skb2)
2159 				return -ENOBUFS;
2160 			skb = skb2;
2161 		}
2162 
2163 		vif = ipmr_find_vif(mrt, dev);
2164 		if (vif >= 0)
2165 			return ipmr_cache_unresolved(mrt, vif, skb, dev);
2166 		kfree_skb(skb);
2167 		return -ENODEV;
2168 	}
2169 
2170 	ip_mr_forward(net, mrt, dev, skb, cache, local);
2171 
2172 	if (local)
2173 		return ip_local_deliver(skb);
2174 
2175 	return 0;
2176 
2177 dont_forward:
2178 	if (local)
2179 		return ip_local_deliver(skb);
2180 	kfree_skb(skb);
2181 	return 0;
2182 }
2183 
2184 #ifdef CONFIG_IP_PIMSM_V1
2185 /* Handle IGMP messages of PIMv1 */
2186 int pim_rcv_v1(struct sk_buff *skb)
2187 {
2188 	struct igmphdr *pim;
2189 	struct net *net = dev_net(skb->dev);
2190 	struct mr_table *mrt;
2191 
2192 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2193 		goto drop;
2194 
2195 	pim = igmp_hdr(skb);
2196 
2197 	mrt = ipmr_rt_fib_lookup(net, skb);
2198 	if (IS_ERR(mrt))
2199 		goto drop;
2200 	if (!mrt->mroute_do_pim ||
2201 	    pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2202 		goto drop;
2203 
2204 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2205 drop:
2206 		kfree_skb(skb);
2207 	}
2208 	return 0;
2209 }
2210 #endif
2211 
2212 #ifdef CONFIG_IP_PIMSM_V2
2213 static int pim_rcv(struct sk_buff *skb)
2214 {
2215 	struct pimreghdr *pim;
2216 	struct net *net = dev_net(skb->dev);
2217 	struct mr_table *mrt;
2218 
2219 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2220 		goto drop;
2221 
2222 	pim = (struct pimreghdr *)skb_transport_header(skb);
2223 	if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2224 	    (pim->flags & PIM_NULL_REGISTER) ||
2225 	    (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2226 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2227 		goto drop;
2228 
2229 	mrt = ipmr_rt_fib_lookup(net, skb);
2230 	if (IS_ERR(mrt))
2231 		goto drop;
2232 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2233 drop:
2234 		kfree_skb(skb);
2235 	}
2236 	return 0;
2237 }
2238 #endif
2239 
2240 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2241 		   __be32 saddr, __be32 daddr,
2242 		   struct rtmsg *rtm, u32 portid)
2243 {
2244 	struct mfc_cache *cache;
2245 	struct mr_table *mrt;
2246 	int err;
2247 
2248 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2249 	if (!mrt)
2250 		return -ENOENT;
2251 
2252 	rcu_read_lock();
2253 	cache = ipmr_cache_find(mrt, saddr, daddr);
2254 	if (!cache && skb->dev) {
2255 		int vif = ipmr_find_vif(mrt, skb->dev);
2256 
2257 		if (vif >= 0)
2258 			cache = ipmr_cache_find_any(mrt, daddr, vif);
2259 	}
2260 	if (!cache) {
2261 		struct sk_buff *skb2;
2262 		struct iphdr *iph;
2263 		struct net_device *dev;
2264 		int vif = -1;
2265 
2266 		dev = skb->dev;
2267 		if (dev)
2268 			vif = ipmr_find_vif(mrt, dev);
2269 		if (vif < 0) {
2270 			rcu_read_unlock();
2271 			return -ENODEV;
2272 		}
2273 
2274 		skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2275 		if (!skb2) {
2276 			rcu_read_unlock();
2277 			return -ENOMEM;
2278 		}
2279 
2280 		NETLINK_CB(skb2).portid = portid;
2281 		skb_push(skb2, sizeof(struct iphdr));
2282 		skb_reset_network_header(skb2);
2283 		iph = ip_hdr(skb2);
2284 		iph->ihl = sizeof(struct iphdr) >> 2;
2285 		iph->saddr = saddr;
2286 		iph->daddr = daddr;
2287 		iph->version = 0;
2288 		err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2289 		rcu_read_unlock();
2290 		return err;
2291 	}
2292 
2293 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2294 	rcu_read_unlock();
2295 	return err;
2296 }
2297 
2298 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2299 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2300 			    int flags)
2301 {
2302 	struct nlmsghdr *nlh;
2303 	struct rtmsg *rtm;
2304 	int err;
2305 
2306 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2307 	if (!nlh)
2308 		return -EMSGSIZE;
2309 
2310 	rtm = nlmsg_data(nlh);
2311 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2312 	rtm->rtm_dst_len  = 32;
2313 	rtm->rtm_src_len  = 32;
2314 	rtm->rtm_tos      = 0;
2315 	rtm->rtm_table    = mrt->id;
2316 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2317 		goto nla_put_failure;
2318 	rtm->rtm_type     = RTN_MULTICAST;
2319 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2320 	if (c->_c.mfc_flags & MFC_STATIC)
2321 		rtm->rtm_protocol = RTPROT_STATIC;
2322 	else
2323 		rtm->rtm_protocol = RTPROT_MROUTED;
2324 	rtm->rtm_flags    = 0;
2325 
2326 	if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2327 	    nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2328 		goto nla_put_failure;
2329 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2330 	/* do not break the dump if cache is unresolved */
2331 	if (err < 0 && err != -ENOENT)
2332 		goto nla_put_failure;
2333 
2334 	nlmsg_end(skb, nlh);
2335 	return 0;
2336 
2337 nla_put_failure:
2338 	nlmsg_cancel(skb, nlh);
2339 	return -EMSGSIZE;
2340 }
2341 
2342 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2343 			     u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2344 			     int flags)
2345 {
2346 	return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2347 				cmd, flags);
2348 }
2349 
2350 static size_t mroute_msgsize(bool unresolved, int maxvif)
2351 {
2352 	size_t len =
2353 		NLMSG_ALIGN(sizeof(struct rtmsg))
2354 		+ nla_total_size(4)	/* RTA_TABLE */
2355 		+ nla_total_size(4)	/* RTA_SRC */
2356 		+ nla_total_size(4)	/* RTA_DST */
2357 		;
2358 
2359 	if (!unresolved)
2360 		len = len
2361 		      + nla_total_size(4)	/* RTA_IIF */
2362 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2363 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2364 						/* RTA_MFC_STATS */
2365 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2366 		;
2367 
2368 	return len;
2369 }
2370 
2371 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2372 				 int cmd)
2373 {
2374 	struct net *net = read_pnet(&mrt->net);
2375 	struct sk_buff *skb;
2376 	int err = -ENOBUFS;
2377 
2378 	skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2379 				       mrt->maxvif),
2380 			GFP_ATOMIC);
2381 	if (!skb)
2382 		goto errout;
2383 
2384 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2385 	if (err < 0)
2386 		goto errout;
2387 
2388 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2389 	return;
2390 
2391 errout:
2392 	kfree_skb(skb);
2393 	if (err < 0)
2394 		rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2395 }
2396 
2397 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2398 {
2399 	size_t len =
2400 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2401 		+ nla_total_size(1)	/* IPMRA_CREPORT_MSGTYPE */
2402 		+ nla_total_size(4)	/* IPMRA_CREPORT_VIF_ID */
2403 		+ nla_total_size(4)	/* IPMRA_CREPORT_SRC_ADDR */
2404 		+ nla_total_size(4)	/* IPMRA_CREPORT_DST_ADDR */
2405 		+ nla_total_size(4)	/* IPMRA_CREPORT_TABLE */
2406 					/* IPMRA_CREPORT_PKT */
2407 		+ nla_total_size(payloadlen)
2408 		;
2409 
2410 	return len;
2411 }
2412 
2413 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2414 {
2415 	struct net *net = read_pnet(&mrt->net);
2416 	struct nlmsghdr *nlh;
2417 	struct rtgenmsg *rtgenm;
2418 	struct igmpmsg *msg;
2419 	struct sk_buff *skb;
2420 	struct nlattr *nla;
2421 	int payloadlen;
2422 
2423 	payloadlen = pkt->len - sizeof(struct igmpmsg);
2424 	msg = (struct igmpmsg *)skb_network_header(pkt);
2425 
2426 	skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2427 	if (!skb)
2428 		goto errout;
2429 
2430 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2431 			sizeof(struct rtgenmsg), 0);
2432 	if (!nlh)
2433 		goto errout;
2434 	rtgenm = nlmsg_data(nlh);
2435 	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2436 	if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2437 	    nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2438 	    nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2439 			    msg->im_src.s_addr) ||
2440 	    nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2441 			    msg->im_dst.s_addr) ||
2442 	    nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2443 		goto nla_put_failure;
2444 
2445 	nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2446 	if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2447 				  nla_data(nla), payloadlen))
2448 		goto nla_put_failure;
2449 
2450 	nlmsg_end(skb, nlh);
2451 
2452 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2453 	return;
2454 
2455 nla_put_failure:
2456 	nlmsg_cancel(skb, nlh);
2457 errout:
2458 	kfree_skb(skb);
2459 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2460 }
2461 
2462 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2463 				       const struct nlmsghdr *nlh,
2464 				       struct nlattr **tb,
2465 				       struct netlink_ext_ack *extack)
2466 {
2467 	struct rtmsg *rtm;
2468 	int i, err;
2469 
2470 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2471 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2472 		return -EINVAL;
2473 	}
2474 
2475 	if (!netlink_strict_get_check(skb))
2476 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2477 					      rtm_ipv4_policy, extack);
2478 
2479 	rtm = nlmsg_data(nlh);
2480 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2481 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2482 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2483 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2484 		NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2485 		return -EINVAL;
2486 	}
2487 
2488 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2489 					    rtm_ipv4_policy, extack);
2490 	if (err)
2491 		return err;
2492 
2493 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2494 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2495 		NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2496 		return -EINVAL;
2497 	}
2498 
2499 	for (i = 0; i <= RTA_MAX; i++) {
2500 		if (!tb[i])
2501 			continue;
2502 
2503 		switch (i) {
2504 		case RTA_SRC:
2505 		case RTA_DST:
2506 		case RTA_TABLE:
2507 			break;
2508 		default:
2509 			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2510 			return -EINVAL;
2511 		}
2512 	}
2513 
2514 	return 0;
2515 }
2516 
2517 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2518 			     struct netlink_ext_ack *extack)
2519 {
2520 	struct net *net = sock_net(in_skb->sk);
2521 	struct nlattr *tb[RTA_MAX + 1];
2522 	struct sk_buff *skb = NULL;
2523 	struct mfc_cache *cache;
2524 	struct mr_table *mrt;
2525 	__be32 src, grp;
2526 	u32 tableid;
2527 	int err;
2528 
2529 	err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2530 	if (err < 0)
2531 		goto errout;
2532 
2533 	src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2534 	grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2535 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2536 
2537 	mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2538 	if (!mrt) {
2539 		err = -ENOENT;
2540 		goto errout_free;
2541 	}
2542 
2543 	/* entries are added/deleted only under RTNL */
2544 	rcu_read_lock();
2545 	cache = ipmr_cache_find(mrt, src, grp);
2546 	rcu_read_unlock();
2547 	if (!cache) {
2548 		err = -ENOENT;
2549 		goto errout_free;
2550 	}
2551 
2552 	skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2553 	if (!skb) {
2554 		err = -ENOBUFS;
2555 		goto errout_free;
2556 	}
2557 
2558 	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2559 			       nlh->nlmsg_seq, cache,
2560 			       RTM_NEWROUTE, 0);
2561 	if (err < 0)
2562 		goto errout_free;
2563 
2564 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2565 
2566 errout:
2567 	return err;
2568 
2569 errout_free:
2570 	kfree_skb(skb);
2571 	goto errout;
2572 }
2573 
2574 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2575 {
2576 	struct fib_dump_filter filter = {};
2577 	int err;
2578 
2579 	if (cb->strict_check) {
2580 		err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2581 					    &filter, cb);
2582 		if (err < 0)
2583 			return err;
2584 	}
2585 
2586 	if (filter.table_id) {
2587 		struct mr_table *mrt;
2588 
2589 		mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2590 		if (!mrt) {
2591 			if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2592 				return skb->len;
2593 
2594 			NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2595 			return -ENOENT;
2596 		}
2597 		err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2598 				    &mfc_unres_lock, &filter);
2599 		return skb->len ? : err;
2600 	}
2601 
2602 	return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2603 				_ipmr_fill_mroute, &mfc_unres_lock, &filter);
2604 }
2605 
2606 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2607 	[RTA_SRC]	= { .type = NLA_U32 },
2608 	[RTA_DST]	= { .type = NLA_U32 },
2609 	[RTA_IIF]	= { .type = NLA_U32 },
2610 	[RTA_TABLE]	= { .type = NLA_U32 },
2611 	[RTA_MULTIPATH]	= { .len = sizeof(struct rtnexthop) },
2612 };
2613 
2614 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2615 {
2616 	switch (rtm_protocol) {
2617 	case RTPROT_STATIC:
2618 	case RTPROT_MROUTED:
2619 		return true;
2620 	}
2621 	return false;
2622 }
2623 
2624 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2625 {
2626 	struct rtnexthop *rtnh = nla_data(nla);
2627 	int remaining = nla_len(nla), vifi = 0;
2628 
2629 	while (rtnh_ok(rtnh, remaining)) {
2630 		mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2631 		if (++vifi == MAXVIFS)
2632 			break;
2633 		rtnh = rtnh_next(rtnh, &remaining);
2634 	}
2635 
2636 	return remaining > 0 ? -EINVAL : vifi;
2637 }
2638 
2639 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2640 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2641 			    struct mfcctl *mfcc, int *mrtsock,
2642 			    struct mr_table **mrtret,
2643 			    struct netlink_ext_ack *extack)
2644 {
2645 	struct net_device *dev = NULL;
2646 	u32 tblid = RT_TABLE_DEFAULT;
2647 	struct mr_table *mrt;
2648 	struct nlattr *attr;
2649 	struct rtmsg *rtm;
2650 	int ret, rem;
2651 
2652 	ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2653 					rtm_ipmr_policy, extack);
2654 	if (ret < 0)
2655 		goto out;
2656 	rtm = nlmsg_data(nlh);
2657 
2658 	ret = -EINVAL;
2659 	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2660 	    rtm->rtm_type != RTN_MULTICAST ||
2661 	    rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2662 	    !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2663 		goto out;
2664 
2665 	memset(mfcc, 0, sizeof(*mfcc));
2666 	mfcc->mfcc_parent = -1;
2667 	ret = 0;
2668 	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2669 		switch (nla_type(attr)) {
2670 		case RTA_SRC:
2671 			mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2672 			break;
2673 		case RTA_DST:
2674 			mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2675 			break;
2676 		case RTA_IIF:
2677 			dev = __dev_get_by_index(net, nla_get_u32(attr));
2678 			if (!dev) {
2679 				ret = -ENODEV;
2680 				goto out;
2681 			}
2682 			break;
2683 		case RTA_MULTIPATH:
2684 			if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2685 				ret = -EINVAL;
2686 				goto out;
2687 			}
2688 			break;
2689 		case RTA_PREFSRC:
2690 			ret = 1;
2691 			break;
2692 		case RTA_TABLE:
2693 			tblid = nla_get_u32(attr);
2694 			break;
2695 		}
2696 	}
2697 	mrt = ipmr_get_table(net, tblid);
2698 	if (!mrt) {
2699 		ret = -ENOENT;
2700 		goto out;
2701 	}
2702 	*mrtret = mrt;
2703 	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2704 	if (dev)
2705 		mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2706 
2707 out:
2708 	return ret;
2709 }
2710 
2711 /* takes care of both newroute and delroute */
2712 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2713 			  struct netlink_ext_ack *extack)
2714 {
2715 	struct net *net = sock_net(skb->sk);
2716 	int ret, mrtsock, parent;
2717 	struct mr_table *tbl;
2718 	struct mfcctl mfcc;
2719 
2720 	mrtsock = 0;
2721 	tbl = NULL;
2722 	ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2723 	if (ret < 0)
2724 		return ret;
2725 
2726 	parent = ret ? mfcc.mfcc_parent : -1;
2727 	if (nlh->nlmsg_type == RTM_NEWROUTE)
2728 		return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2729 	else
2730 		return ipmr_mfc_delete(tbl, &mfcc, parent);
2731 }
2732 
2733 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2734 {
2735 	u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2736 
2737 	if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2738 	    nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2739 	    nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2740 			mrt->mroute_reg_vif_num) ||
2741 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2742 		       mrt->mroute_do_assert) ||
2743 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2744 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2745 		       mrt->mroute_do_wrvifwhole))
2746 		return false;
2747 
2748 	return true;
2749 }
2750 
2751 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2752 {
2753 	struct net_device *vif_dev;
2754 	struct nlattr *vif_nest;
2755 	struct vif_device *vif;
2756 
2757 	vif = &mrt->vif_table[vifid];
2758 	vif_dev = rtnl_dereference(vif->dev);
2759 	/* if the VIF doesn't exist just continue */
2760 	if (!vif_dev)
2761 		return true;
2762 
2763 	vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2764 	if (!vif_nest)
2765 		return false;
2766 
2767 	if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2768 	    nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2769 	    nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2770 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2771 			      IPMRA_VIFA_PAD) ||
2772 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2773 			      IPMRA_VIFA_PAD) ||
2774 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2775 			      IPMRA_VIFA_PAD) ||
2776 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2777 			      IPMRA_VIFA_PAD) ||
2778 	    nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2779 	    nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2780 		nla_nest_cancel(skb, vif_nest);
2781 		return false;
2782 	}
2783 	nla_nest_end(skb, vif_nest);
2784 
2785 	return true;
2786 }
2787 
2788 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2789 			       struct netlink_ext_ack *extack)
2790 {
2791 	struct ifinfomsg *ifm;
2792 
2793 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2794 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2795 		return -EINVAL;
2796 	}
2797 
2798 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2799 		NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2800 		return -EINVAL;
2801 	}
2802 
2803 	ifm = nlmsg_data(nlh);
2804 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2805 	    ifm->ifi_change || ifm->ifi_index) {
2806 		NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2807 		return -EINVAL;
2808 	}
2809 
2810 	return 0;
2811 }
2812 
2813 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2814 {
2815 	struct net *net = sock_net(skb->sk);
2816 	struct nlmsghdr *nlh = NULL;
2817 	unsigned int t = 0, s_t;
2818 	unsigned int e = 0, s_e;
2819 	struct mr_table *mrt;
2820 
2821 	if (cb->strict_check) {
2822 		int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2823 
2824 		if (err < 0)
2825 			return err;
2826 	}
2827 
2828 	s_t = cb->args[0];
2829 	s_e = cb->args[1];
2830 
2831 	ipmr_for_each_table(mrt, net) {
2832 		struct nlattr *vifs, *af;
2833 		struct ifinfomsg *hdr;
2834 		u32 i;
2835 
2836 		if (t < s_t)
2837 			goto skip_table;
2838 		nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2839 				cb->nlh->nlmsg_seq, RTM_NEWLINK,
2840 				sizeof(*hdr), NLM_F_MULTI);
2841 		if (!nlh)
2842 			break;
2843 
2844 		hdr = nlmsg_data(nlh);
2845 		memset(hdr, 0, sizeof(*hdr));
2846 		hdr->ifi_family = RTNL_FAMILY_IPMR;
2847 
2848 		af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2849 		if (!af) {
2850 			nlmsg_cancel(skb, nlh);
2851 			goto out;
2852 		}
2853 
2854 		if (!ipmr_fill_table(mrt, skb)) {
2855 			nlmsg_cancel(skb, nlh);
2856 			goto out;
2857 		}
2858 
2859 		vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2860 		if (!vifs) {
2861 			nla_nest_end(skb, af);
2862 			nlmsg_end(skb, nlh);
2863 			goto out;
2864 		}
2865 		for (i = 0; i < mrt->maxvif; i++) {
2866 			if (e < s_e)
2867 				goto skip_entry;
2868 			if (!ipmr_fill_vif(mrt, i, skb)) {
2869 				nla_nest_end(skb, vifs);
2870 				nla_nest_end(skb, af);
2871 				nlmsg_end(skb, nlh);
2872 				goto out;
2873 			}
2874 skip_entry:
2875 			e++;
2876 		}
2877 		s_e = 0;
2878 		e = 0;
2879 		nla_nest_end(skb, vifs);
2880 		nla_nest_end(skb, af);
2881 		nlmsg_end(skb, nlh);
2882 skip_table:
2883 		t++;
2884 	}
2885 
2886 out:
2887 	cb->args[1] = e;
2888 	cb->args[0] = t;
2889 
2890 	return skb->len;
2891 }
2892 
2893 #ifdef CONFIG_PROC_FS
2894 /* The /proc interfaces to multicast routing :
2895  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2896  */
2897 
2898 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2899 	__acquires(RCU)
2900 {
2901 	struct mr_vif_iter *iter = seq->private;
2902 	struct net *net = seq_file_net(seq);
2903 	struct mr_table *mrt;
2904 
2905 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2906 	if (!mrt)
2907 		return ERR_PTR(-ENOENT);
2908 
2909 	iter->mrt = mrt;
2910 
2911 	rcu_read_lock();
2912 	return mr_vif_seq_start(seq, pos);
2913 }
2914 
2915 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2916 	__releases(RCU)
2917 {
2918 	rcu_read_unlock();
2919 }
2920 
2921 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2922 {
2923 	struct mr_vif_iter *iter = seq->private;
2924 	struct mr_table *mrt = iter->mrt;
2925 
2926 	if (v == SEQ_START_TOKEN) {
2927 		seq_puts(seq,
2928 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2929 	} else {
2930 		const struct vif_device *vif = v;
2931 		const struct net_device *vif_dev;
2932 		const char *name;
2933 
2934 		vif_dev = vif_dev_read(vif);
2935 		name = vif_dev ? vif_dev->name : "none";
2936 		seq_printf(seq,
2937 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2938 			   vif - mrt->vif_table,
2939 			   name, vif->bytes_in, vif->pkt_in,
2940 			   vif->bytes_out, vif->pkt_out,
2941 			   vif->flags, vif->local, vif->remote);
2942 	}
2943 	return 0;
2944 }
2945 
2946 static const struct seq_operations ipmr_vif_seq_ops = {
2947 	.start = ipmr_vif_seq_start,
2948 	.next  = mr_vif_seq_next,
2949 	.stop  = ipmr_vif_seq_stop,
2950 	.show  = ipmr_vif_seq_show,
2951 };
2952 
2953 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2954 {
2955 	struct net *net = seq_file_net(seq);
2956 	struct mr_table *mrt;
2957 
2958 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2959 	if (!mrt)
2960 		return ERR_PTR(-ENOENT);
2961 
2962 	return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2963 }
2964 
2965 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2966 {
2967 	int n;
2968 
2969 	if (v == SEQ_START_TOKEN) {
2970 		seq_puts(seq,
2971 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2972 	} else {
2973 		const struct mfc_cache *mfc = v;
2974 		const struct mr_mfc_iter *it = seq->private;
2975 		const struct mr_table *mrt = it->mrt;
2976 
2977 		seq_printf(seq, "%08X %08X %-3hd",
2978 			   (__force u32) mfc->mfc_mcastgrp,
2979 			   (__force u32) mfc->mfc_origin,
2980 			   mfc->_c.mfc_parent);
2981 
2982 		if (it->cache != &mrt->mfc_unres_queue) {
2983 			seq_printf(seq, " %8lu %8lu %8lu",
2984 				   mfc->_c.mfc_un.res.pkt,
2985 				   mfc->_c.mfc_un.res.bytes,
2986 				   mfc->_c.mfc_un.res.wrong_if);
2987 			for (n = mfc->_c.mfc_un.res.minvif;
2988 			     n < mfc->_c.mfc_un.res.maxvif; n++) {
2989 				if (VIF_EXISTS(mrt, n) &&
2990 				    mfc->_c.mfc_un.res.ttls[n] < 255)
2991 					seq_printf(seq,
2992 					   " %2d:%-3d",
2993 					   n, mfc->_c.mfc_un.res.ttls[n]);
2994 			}
2995 		} else {
2996 			/* unresolved mfc_caches don't contain
2997 			 * pkt, bytes and wrong_if values
2998 			 */
2999 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3000 		}
3001 		seq_putc(seq, '\n');
3002 	}
3003 	return 0;
3004 }
3005 
3006 static const struct seq_operations ipmr_mfc_seq_ops = {
3007 	.start = ipmr_mfc_seq_start,
3008 	.next  = mr_mfc_seq_next,
3009 	.stop  = mr_mfc_seq_stop,
3010 	.show  = ipmr_mfc_seq_show,
3011 };
3012 #endif
3013 
3014 #ifdef CONFIG_IP_PIMSM_V2
3015 static const struct net_protocol pim_protocol = {
3016 	.handler	=	pim_rcv,
3017 };
3018 #endif
3019 
3020 static unsigned int ipmr_seq_read(struct net *net)
3021 {
3022 	ASSERT_RTNL();
3023 
3024 	return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3025 }
3026 
3027 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3028 		     struct netlink_ext_ack *extack)
3029 {
3030 	return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3031 		       ipmr_mr_table_iter, extack);
3032 }
3033 
3034 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3035 	.family		= RTNL_FAMILY_IPMR,
3036 	.fib_seq_read	= ipmr_seq_read,
3037 	.fib_dump	= ipmr_dump,
3038 	.owner		= THIS_MODULE,
3039 };
3040 
3041 static int __net_init ipmr_notifier_init(struct net *net)
3042 {
3043 	struct fib_notifier_ops *ops;
3044 
3045 	net->ipv4.ipmr_seq = 0;
3046 
3047 	ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3048 	if (IS_ERR(ops))
3049 		return PTR_ERR(ops);
3050 	net->ipv4.ipmr_notifier_ops = ops;
3051 
3052 	return 0;
3053 }
3054 
3055 static void __net_exit ipmr_notifier_exit(struct net *net)
3056 {
3057 	fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3058 	net->ipv4.ipmr_notifier_ops = NULL;
3059 }
3060 
3061 /* Setup for IP multicast routing */
3062 static int __net_init ipmr_net_init(struct net *net)
3063 {
3064 	int err;
3065 
3066 	err = ipmr_notifier_init(net);
3067 	if (err)
3068 		goto ipmr_notifier_fail;
3069 
3070 	err = ipmr_rules_init(net);
3071 	if (err < 0)
3072 		goto ipmr_rules_fail;
3073 
3074 #ifdef CONFIG_PROC_FS
3075 	err = -ENOMEM;
3076 	if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3077 			sizeof(struct mr_vif_iter)))
3078 		goto proc_vif_fail;
3079 	if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3080 			sizeof(struct mr_mfc_iter)))
3081 		goto proc_cache_fail;
3082 #endif
3083 	return 0;
3084 
3085 #ifdef CONFIG_PROC_FS
3086 proc_cache_fail:
3087 	remove_proc_entry("ip_mr_vif", net->proc_net);
3088 proc_vif_fail:
3089 	rtnl_lock();
3090 	ipmr_rules_exit(net);
3091 	rtnl_unlock();
3092 #endif
3093 ipmr_rules_fail:
3094 	ipmr_notifier_exit(net);
3095 ipmr_notifier_fail:
3096 	return err;
3097 }
3098 
3099 static void __net_exit ipmr_net_exit(struct net *net)
3100 {
3101 #ifdef CONFIG_PROC_FS
3102 	remove_proc_entry("ip_mr_cache", net->proc_net);
3103 	remove_proc_entry("ip_mr_vif", net->proc_net);
3104 #endif
3105 	ipmr_notifier_exit(net);
3106 }
3107 
3108 static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3109 {
3110 	struct net *net;
3111 
3112 	rtnl_lock();
3113 	list_for_each_entry(net, net_list, exit_list)
3114 		ipmr_rules_exit(net);
3115 	rtnl_unlock();
3116 }
3117 
3118 static struct pernet_operations ipmr_net_ops = {
3119 	.init = ipmr_net_init,
3120 	.exit = ipmr_net_exit,
3121 	.exit_batch = ipmr_net_exit_batch,
3122 };
3123 
3124 int __init ip_mr_init(void)
3125 {
3126 	int err;
3127 
3128 	mrt_cachep = kmem_cache_create("ip_mrt_cache",
3129 				       sizeof(struct mfc_cache),
3130 				       0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3131 				       NULL);
3132 
3133 	err = register_pernet_subsys(&ipmr_net_ops);
3134 	if (err)
3135 		goto reg_pernet_fail;
3136 
3137 	err = register_netdevice_notifier(&ip_mr_notifier);
3138 	if (err)
3139 		goto reg_notif_fail;
3140 #ifdef CONFIG_IP_PIMSM_V2
3141 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3142 		pr_err("%s: can't add PIM protocol\n", __func__);
3143 		err = -EAGAIN;
3144 		goto add_proto_fail;
3145 	}
3146 #endif
3147 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3148 		      ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3149 	rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3150 		      ipmr_rtm_route, NULL, 0);
3151 	rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3152 		      ipmr_rtm_route, NULL, 0);
3153 
3154 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3155 		      NULL, ipmr_rtm_dumplink, 0);
3156 	return 0;
3157 
3158 #ifdef CONFIG_IP_PIMSM_V2
3159 add_proto_fail:
3160 	unregister_netdevice_notifier(&ip_mr_notifier);
3161 #endif
3162 reg_notif_fail:
3163 	unregister_pernet_subsys(&ipmr_net_ops);
3164 reg_pernet_fail:
3165 	kmem_cache_destroy(mrt_cachep);
3166 	return err;
3167 }
3168