xref: /openbmc/linux/net/ipv4/ipmr.c (revision a9a08845)
1 /*
2  *	IP multicast routing support for mrouted 3.6/3.8
3  *
4  *		(c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5  *	  Linux Consultancy and Custom Driver Development
6  *
7  *	This program is free software; you can redistribute it and/or
8  *	modify it under the terms of the GNU General Public License
9  *	as published by the Free Software Foundation; either version
10  *	2 of the License, or (at your option) any later version.
11  *
12  *	Fixes:
13  *	Michael Chastain	:	Incorrect size of copying.
14  *	Alan Cox		:	Added the cache manager code
15  *	Alan Cox		:	Fixed the clone/copy bug and device race.
16  *	Mike McLagan		:	Routing by source
17  *	Malcolm Beattie		:	Buffer handling fixes.
18  *	Alexey Kuznetsov	:	Double buffer free and other fixes.
19  *	SVR Anand		:	Fixed several multicast bugs and problems.
20  *	Alexey Kuznetsov	:	Status, optimisations and more.
21  *	Brad Parker		:	Better behaviour on mrouted upcall
22  *					overflow.
23  *      Carlos Picoto           :       PIMv1 Support
24  *	Pavlin Ivanov Radoslavov:	PIMv2 Registers must checksum only PIM header
25  *					Relax this requirement to work with older peers.
26  *
27  */
28 
29 #include <linux/uaccess.h>
30 #include <linux/types.h>
31 #include <linux/capability.h>
32 #include <linux/errno.h>
33 #include <linux/timer.h>
34 #include <linux/mm.h>
35 #include <linux/kernel.h>
36 #include <linux/fcntl.h>
37 #include <linux/stat.h>
38 #include <linux/socket.h>
39 #include <linux/in.h>
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/inetdevice.h>
43 #include <linux/igmp.h>
44 #include <linux/proc_fs.h>
45 #include <linux/seq_file.h>
46 #include <linux/mroute.h>
47 #include <linux/init.h>
48 #include <linux/if_ether.h>
49 #include <linux/slab.h>
50 #include <net/net_namespace.h>
51 #include <net/ip.h>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.h>
55 #include <net/sock.h>
56 #include <net/icmp.h>
57 #include <net/udp.h>
58 #include <net/raw.h>
59 #include <linux/notifier.h>
60 #include <linux/if_arp.h>
61 #include <linux/netfilter_ipv4.h>
62 #include <linux/compat.h>
63 #include <linux/export.h>
64 #include <net/ip_tunnels.h>
65 #include <net/checksum.h>
66 #include <net/netlink.h>
67 #include <net/fib_rules.h>
68 #include <linux/netconf.h>
69 #include <net/nexthop.h>
70 #include <net/switchdev.h>
71 
72 struct ipmr_rule {
73 	struct fib_rule		common;
74 };
75 
76 struct ipmr_result {
77 	struct mr_table		*mrt;
78 };
79 
80 /* Big lock, protecting vif table, mrt cache and mroute socket state.
81  * Note that the changes are semaphored via rtnl_lock.
82  */
83 
84 static DEFINE_RWLOCK(mrt_lock);
85 
86 /* Multicast router control variables */
87 
88 /* Special spinlock for queue of unresolved entries */
89 static DEFINE_SPINLOCK(mfc_unres_lock);
90 
91 /* We return to original Alan's scheme. Hash table of resolved
92  * entries is changed only in process context and protected
93  * with weak lock mrt_lock. Queue of unresolved entries is protected
94  * with strong spinlock mfc_unres_lock.
95  *
96  * In this case data path is free of exclusive locks at all.
97  */
98 
99 static struct kmem_cache *mrt_cachep __read_mostly;
100 
101 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
102 static void ipmr_free_table(struct mr_table *mrt);
103 
104 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
105 			  struct net_device *dev, struct sk_buff *skb,
106 			  struct mfc_cache *cache, int local);
107 static int ipmr_cache_report(struct mr_table *mrt,
108 			     struct sk_buff *pkt, vifi_t vifi, int assert);
109 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
110 			      struct mfc_cache *c, struct rtmsg *rtm);
111 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
112 				 int cmd);
113 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
114 static void mroute_clean_tables(struct mr_table *mrt, bool all);
115 static void ipmr_expire_process(struct timer_list *t);
116 
117 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
118 #define ipmr_for_each_table(mrt, net) \
119 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
120 
121 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
122 {
123 	struct mr_table *mrt;
124 
125 	ipmr_for_each_table(mrt, net) {
126 		if (mrt->id == id)
127 			return mrt;
128 	}
129 	return NULL;
130 }
131 
132 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
133 			   struct mr_table **mrt)
134 {
135 	int err;
136 	struct ipmr_result res;
137 	struct fib_lookup_arg arg = {
138 		.result = &res,
139 		.flags = FIB_LOOKUP_NOREF,
140 	};
141 
142 	/* update flow if oif or iif point to device enslaved to l3mdev */
143 	l3mdev_update_flow(net, flowi4_to_flowi(flp4));
144 
145 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
146 			       flowi4_to_flowi(flp4), 0, &arg);
147 	if (err < 0)
148 		return err;
149 	*mrt = res.mrt;
150 	return 0;
151 }
152 
153 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
154 			    int flags, struct fib_lookup_arg *arg)
155 {
156 	struct ipmr_result *res = arg->result;
157 	struct mr_table *mrt;
158 
159 	switch (rule->action) {
160 	case FR_ACT_TO_TBL:
161 		break;
162 	case FR_ACT_UNREACHABLE:
163 		return -ENETUNREACH;
164 	case FR_ACT_PROHIBIT:
165 		return -EACCES;
166 	case FR_ACT_BLACKHOLE:
167 	default:
168 		return -EINVAL;
169 	}
170 
171 	arg->table = fib_rule_get_table(rule, arg);
172 
173 	mrt = ipmr_get_table(rule->fr_net, arg->table);
174 	if (!mrt)
175 		return -EAGAIN;
176 	res->mrt = mrt;
177 	return 0;
178 }
179 
180 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
181 {
182 	return 1;
183 }
184 
185 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
186 	FRA_GENERIC_POLICY,
187 };
188 
189 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
190 			       struct fib_rule_hdr *frh, struct nlattr **tb)
191 {
192 	return 0;
193 }
194 
195 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
196 			     struct nlattr **tb)
197 {
198 	return 1;
199 }
200 
201 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
202 			  struct fib_rule_hdr *frh)
203 {
204 	frh->dst_len = 0;
205 	frh->src_len = 0;
206 	frh->tos     = 0;
207 	return 0;
208 }
209 
210 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
211 	.family		= RTNL_FAMILY_IPMR,
212 	.rule_size	= sizeof(struct ipmr_rule),
213 	.addr_size	= sizeof(u32),
214 	.action		= ipmr_rule_action,
215 	.match		= ipmr_rule_match,
216 	.configure	= ipmr_rule_configure,
217 	.compare	= ipmr_rule_compare,
218 	.fill		= ipmr_rule_fill,
219 	.nlgroup	= RTNLGRP_IPV4_RULE,
220 	.policy		= ipmr_rule_policy,
221 	.owner		= THIS_MODULE,
222 };
223 
224 static int __net_init ipmr_rules_init(struct net *net)
225 {
226 	struct fib_rules_ops *ops;
227 	struct mr_table *mrt;
228 	int err;
229 
230 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
231 	if (IS_ERR(ops))
232 		return PTR_ERR(ops);
233 
234 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
235 
236 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
237 	if (IS_ERR(mrt)) {
238 		err = PTR_ERR(mrt);
239 		goto err1;
240 	}
241 
242 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
243 	if (err < 0)
244 		goto err2;
245 
246 	net->ipv4.mr_rules_ops = ops;
247 	return 0;
248 
249 err2:
250 	ipmr_free_table(mrt);
251 err1:
252 	fib_rules_unregister(ops);
253 	return err;
254 }
255 
256 static void __net_exit ipmr_rules_exit(struct net *net)
257 {
258 	struct mr_table *mrt, *next;
259 
260 	rtnl_lock();
261 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
262 		list_del(&mrt->list);
263 		ipmr_free_table(mrt);
264 	}
265 	fib_rules_unregister(net->ipv4.mr_rules_ops);
266 	rtnl_unlock();
267 }
268 
269 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
270 {
271 	return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR);
272 }
273 
274 static unsigned int ipmr_rules_seq_read(struct net *net)
275 {
276 	return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
277 }
278 
279 bool ipmr_rule_default(const struct fib_rule *rule)
280 {
281 	return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
282 }
283 EXPORT_SYMBOL(ipmr_rule_default);
284 #else
285 #define ipmr_for_each_table(mrt, net) \
286 	for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
287 
288 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
289 {
290 	return net->ipv4.mrt;
291 }
292 
293 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
294 			   struct mr_table **mrt)
295 {
296 	*mrt = net->ipv4.mrt;
297 	return 0;
298 }
299 
300 static int __net_init ipmr_rules_init(struct net *net)
301 {
302 	struct mr_table *mrt;
303 
304 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
305 	if (IS_ERR(mrt))
306 		return PTR_ERR(mrt);
307 	net->ipv4.mrt = mrt;
308 	return 0;
309 }
310 
311 static void __net_exit ipmr_rules_exit(struct net *net)
312 {
313 	rtnl_lock();
314 	ipmr_free_table(net->ipv4.mrt);
315 	net->ipv4.mrt = NULL;
316 	rtnl_unlock();
317 }
318 
319 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
320 {
321 	return 0;
322 }
323 
324 static unsigned int ipmr_rules_seq_read(struct net *net)
325 {
326 	return 0;
327 }
328 
329 bool ipmr_rule_default(const struct fib_rule *rule)
330 {
331 	return true;
332 }
333 EXPORT_SYMBOL(ipmr_rule_default);
334 #endif
335 
336 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
337 				const void *ptr)
338 {
339 	const struct mfc_cache_cmp_arg *cmparg = arg->key;
340 	struct mfc_cache *c = (struct mfc_cache *)ptr;
341 
342 	return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
343 	       cmparg->mfc_origin != c->mfc_origin;
344 }
345 
346 static const struct rhashtable_params ipmr_rht_params = {
347 	.head_offset = offsetof(struct mfc_cache, mnode),
348 	.key_offset = offsetof(struct mfc_cache, cmparg),
349 	.key_len = sizeof(struct mfc_cache_cmp_arg),
350 	.nelem_hint = 3,
351 	.locks_mul = 1,
352 	.obj_cmpfn = ipmr_hash_cmp,
353 	.automatic_shrinking = true,
354 };
355 
356 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
357 {
358 	struct mr_table *mrt;
359 
360 	/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
361 	if (id != RT_TABLE_DEFAULT && id >= 1000000000)
362 		return ERR_PTR(-EINVAL);
363 
364 	mrt = ipmr_get_table(net, id);
365 	if (mrt)
366 		return mrt;
367 
368 	mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
369 	if (!mrt)
370 		return ERR_PTR(-ENOMEM);
371 	write_pnet(&mrt->net, net);
372 	mrt->id = id;
373 
374 	rhltable_init(&mrt->mfc_hash, &ipmr_rht_params);
375 	INIT_LIST_HEAD(&mrt->mfc_cache_list);
376 	INIT_LIST_HEAD(&mrt->mfc_unres_queue);
377 
378 	timer_setup(&mrt->ipmr_expire_timer, ipmr_expire_process, 0);
379 
380 	mrt->mroute_reg_vif_num = -1;
381 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
382 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
383 #endif
384 	return mrt;
385 }
386 
387 static void ipmr_free_table(struct mr_table *mrt)
388 {
389 	del_timer_sync(&mrt->ipmr_expire_timer);
390 	mroute_clean_tables(mrt, true);
391 	rhltable_destroy(&mrt->mfc_hash);
392 	kfree(mrt);
393 }
394 
395 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
396 
397 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
398 {
399 	struct net *net = dev_net(dev);
400 
401 	dev_close(dev);
402 
403 	dev = __dev_get_by_name(net, "tunl0");
404 	if (dev) {
405 		const struct net_device_ops *ops = dev->netdev_ops;
406 		struct ifreq ifr;
407 		struct ip_tunnel_parm p;
408 
409 		memset(&p, 0, sizeof(p));
410 		p.iph.daddr = v->vifc_rmt_addr.s_addr;
411 		p.iph.saddr = v->vifc_lcl_addr.s_addr;
412 		p.iph.version = 4;
413 		p.iph.ihl = 5;
414 		p.iph.protocol = IPPROTO_IPIP;
415 		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
416 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
417 
418 		if (ops->ndo_do_ioctl) {
419 			mm_segment_t oldfs = get_fs();
420 
421 			set_fs(KERNEL_DS);
422 			ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
423 			set_fs(oldfs);
424 		}
425 	}
426 }
427 
428 /* Initialize ipmr pimreg/tunnel in_device */
429 static bool ipmr_init_vif_indev(const struct net_device *dev)
430 {
431 	struct in_device *in_dev;
432 
433 	ASSERT_RTNL();
434 
435 	in_dev = __in_dev_get_rtnl(dev);
436 	if (!in_dev)
437 		return false;
438 	ipv4_devconf_setall(in_dev);
439 	neigh_parms_data_state_setall(in_dev->arp_parms);
440 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
441 
442 	return true;
443 }
444 
445 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
446 {
447 	struct net_device  *dev;
448 
449 	dev = __dev_get_by_name(net, "tunl0");
450 
451 	if (dev) {
452 		const struct net_device_ops *ops = dev->netdev_ops;
453 		int err;
454 		struct ifreq ifr;
455 		struct ip_tunnel_parm p;
456 
457 		memset(&p, 0, sizeof(p));
458 		p.iph.daddr = v->vifc_rmt_addr.s_addr;
459 		p.iph.saddr = v->vifc_lcl_addr.s_addr;
460 		p.iph.version = 4;
461 		p.iph.ihl = 5;
462 		p.iph.protocol = IPPROTO_IPIP;
463 		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
464 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
465 
466 		if (ops->ndo_do_ioctl) {
467 			mm_segment_t oldfs = get_fs();
468 
469 			set_fs(KERNEL_DS);
470 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
471 			set_fs(oldfs);
472 		} else {
473 			err = -EOPNOTSUPP;
474 		}
475 		dev = NULL;
476 
477 		if (err == 0 &&
478 		    (dev = __dev_get_by_name(net, p.name)) != NULL) {
479 			dev->flags |= IFF_MULTICAST;
480 			if (!ipmr_init_vif_indev(dev))
481 				goto failure;
482 			if (dev_open(dev))
483 				goto failure;
484 			dev_hold(dev);
485 		}
486 	}
487 	return dev;
488 
489 failure:
490 	unregister_netdevice(dev);
491 	return NULL;
492 }
493 
494 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
495 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
496 {
497 	struct net *net = dev_net(dev);
498 	struct mr_table *mrt;
499 	struct flowi4 fl4 = {
500 		.flowi4_oif	= dev->ifindex,
501 		.flowi4_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
502 		.flowi4_mark	= skb->mark,
503 	};
504 	int err;
505 
506 	err = ipmr_fib_lookup(net, &fl4, &mrt);
507 	if (err < 0) {
508 		kfree_skb(skb);
509 		return err;
510 	}
511 
512 	read_lock(&mrt_lock);
513 	dev->stats.tx_bytes += skb->len;
514 	dev->stats.tx_packets++;
515 	ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
516 	read_unlock(&mrt_lock);
517 	kfree_skb(skb);
518 	return NETDEV_TX_OK;
519 }
520 
521 static int reg_vif_get_iflink(const struct net_device *dev)
522 {
523 	return 0;
524 }
525 
526 static const struct net_device_ops reg_vif_netdev_ops = {
527 	.ndo_start_xmit	= reg_vif_xmit,
528 	.ndo_get_iflink = reg_vif_get_iflink,
529 };
530 
531 static void reg_vif_setup(struct net_device *dev)
532 {
533 	dev->type		= ARPHRD_PIMREG;
534 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
535 	dev->flags		= IFF_NOARP;
536 	dev->netdev_ops		= &reg_vif_netdev_ops;
537 	dev->needs_free_netdev	= true;
538 	dev->features		|= NETIF_F_NETNS_LOCAL;
539 }
540 
541 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
542 {
543 	struct net_device *dev;
544 	char name[IFNAMSIZ];
545 
546 	if (mrt->id == RT_TABLE_DEFAULT)
547 		sprintf(name, "pimreg");
548 	else
549 		sprintf(name, "pimreg%u", mrt->id);
550 
551 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
552 
553 	if (!dev)
554 		return NULL;
555 
556 	dev_net_set(dev, net);
557 
558 	if (register_netdevice(dev)) {
559 		free_netdev(dev);
560 		return NULL;
561 	}
562 
563 	if (!ipmr_init_vif_indev(dev))
564 		goto failure;
565 	if (dev_open(dev))
566 		goto failure;
567 
568 	dev_hold(dev);
569 
570 	return dev;
571 
572 failure:
573 	unregister_netdevice(dev);
574 	return NULL;
575 }
576 
577 /* called with rcu_read_lock() */
578 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
579 		     unsigned int pimlen)
580 {
581 	struct net_device *reg_dev = NULL;
582 	struct iphdr *encap;
583 
584 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
585 	/* Check that:
586 	 * a. packet is really sent to a multicast group
587 	 * b. packet is not a NULL-REGISTER
588 	 * c. packet is not truncated
589 	 */
590 	if (!ipv4_is_multicast(encap->daddr) ||
591 	    encap->tot_len == 0 ||
592 	    ntohs(encap->tot_len) + pimlen > skb->len)
593 		return 1;
594 
595 	read_lock(&mrt_lock);
596 	if (mrt->mroute_reg_vif_num >= 0)
597 		reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
598 	read_unlock(&mrt_lock);
599 
600 	if (!reg_dev)
601 		return 1;
602 
603 	skb->mac_header = skb->network_header;
604 	skb_pull(skb, (u8 *)encap - skb->data);
605 	skb_reset_network_header(skb);
606 	skb->protocol = htons(ETH_P_IP);
607 	skb->ip_summed = CHECKSUM_NONE;
608 
609 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
610 
611 	netif_rx(skb);
612 
613 	return NET_RX_SUCCESS;
614 }
615 #else
616 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
617 {
618 	return NULL;
619 }
620 #endif
621 
622 static int call_ipmr_vif_entry_notifier(struct notifier_block *nb,
623 					struct net *net,
624 					enum fib_event_type event_type,
625 					struct vif_device *vif,
626 					vifi_t vif_index, u32 tb_id)
627 {
628 	struct vif_entry_notifier_info info = {
629 		.info = {
630 			.family = RTNL_FAMILY_IPMR,
631 			.net = net,
632 		},
633 		.dev = vif->dev,
634 		.vif_index = vif_index,
635 		.vif_flags = vif->flags,
636 		.tb_id = tb_id,
637 	};
638 
639 	return call_fib_notifier(nb, net, event_type, &info.info);
640 }
641 
642 static int call_ipmr_vif_entry_notifiers(struct net *net,
643 					 enum fib_event_type event_type,
644 					 struct vif_device *vif,
645 					 vifi_t vif_index, u32 tb_id)
646 {
647 	struct vif_entry_notifier_info info = {
648 		.info = {
649 			.family = RTNL_FAMILY_IPMR,
650 			.net = net,
651 		},
652 		.dev = vif->dev,
653 		.vif_index = vif_index,
654 		.vif_flags = vif->flags,
655 		.tb_id = tb_id,
656 	};
657 
658 	ASSERT_RTNL();
659 	net->ipv4.ipmr_seq++;
660 	return call_fib_notifiers(net, event_type, &info.info);
661 }
662 
663 static int call_ipmr_mfc_entry_notifier(struct notifier_block *nb,
664 					struct net *net,
665 					enum fib_event_type event_type,
666 					struct mfc_cache *mfc, u32 tb_id)
667 {
668 	struct mfc_entry_notifier_info info = {
669 		.info = {
670 			.family = RTNL_FAMILY_IPMR,
671 			.net = net,
672 		},
673 		.mfc = mfc,
674 		.tb_id = tb_id
675 	};
676 
677 	return call_fib_notifier(nb, net, event_type, &info.info);
678 }
679 
680 static int call_ipmr_mfc_entry_notifiers(struct net *net,
681 					 enum fib_event_type event_type,
682 					 struct mfc_cache *mfc, u32 tb_id)
683 {
684 	struct mfc_entry_notifier_info info = {
685 		.info = {
686 			.family = RTNL_FAMILY_IPMR,
687 			.net = net,
688 		},
689 		.mfc = mfc,
690 		.tb_id = tb_id
691 	};
692 
693 	ASSERT_RTNL();
694 	net->ipv4.ipmr_seq++;
695 	return call_fib_notifiers(net, event_type, &info.info);
696 }
697 
698 /**
699  *	vif_delete - Delete a VIF entry
700  *	@notify: Set to 1, if the caller is a notifier_call
701  */
702 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
703 		      struct list_head *head)
704 {
705 	struct net *net = read_pnet(&mrt->net);
706 	struct vif_device *v;
707 	struct net_device *dev;
708 	struct in_device *in_dev;
709 
710 	if (vifi < 0 || vifi >= mrt->maxvif)
711 		return -EADDRNOTAVAIL;
712 
713 	v = &mrt->vif_table[vifi];
714 
715 	if (VIF_EXISTS(mrt, vifi))
716 		call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
717 					      mrt->id);
718 
719 	write_lock_bh(&mrt_lock);
720 	dev = v->dev;
721 	v->dev = NULL;
722 
723 	if (!dev) {
724 		write_unlock_bh(&mrt_lock);
725 		return -EADDRNOTAVAIL;
726 	}
727 
728 	if (vifi == mrt->mroute_reg_vif_num)
729 		mrt->mroute_reg_vif_num = -1;
730 
731 	if (vifi + 1 == mrt->maxvif) {
732 		int tmp;
733 
734 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
735 			if (VIF_EXISTS(mrt, tmp))
736 				break;
737 		}
738 		mrt->maxvif = tmp+1;
739 	}
740 
741 	write_unlock_bh(&mrt_lock);
742 
743 	dev_set_allmulti(dev, -1);
744 
745 	in_dev = __in_dev_get_rtnl(dev);
746 	if (in_dev) {
747 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
748 		inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
749 					    NETCONFA_MC_FORWARDING,
750 					    dev->ifindex, &in_dev->cnf);
751 		ip_rt_multicast_event(in_dev);
752 	}
753 
754 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
755 		unregister_netdevice_queue(dev, head);
756 
757 	dev_put(dev);
758 	return 0;
759 }
760 
761 static void ipmr_cache_free_rcu(struct rcu_head *head)
762 {
763 	struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
764 
765 	kmem_cache_free(mrt_cachep, c);
766 }
767 
768 void ipmr_cache_free(struct mfc_cache *c)
769 {
770 	call_rcu(&c->rcu, ipmr_cache_free_rcu);
771 }
772 EXPORT_SYMBOL(ipmr_cache_free);
773 
774 /* Destroy an unresolved cache entry, killing queued skbs
775  * and reporting error to netlink readers.
776  */
777 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
778 {
779 	struct net *net = read_pnet(&mrt->net);
780 	struct sk_buff *skb;
781 	struct nlmsgerr *e;
782 
783 	atomic_dec(&mrt->cache_resolve_queue_len);
784 
785 	while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
786 		if (ip_hdr(skb)->version == 0) {
787 			struct nlmsghdr *nlh = skb_pull(skb,
788 							sizeof(struct iphdr));
789 			nlh->nlmsg_type = NLMSG_ERROR;
790 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
791 			skb_trim(skb, nlh->nlmsg_len);
792 			e = nlmsg_data(nlh);
793 			e->error = -ETIMEDOUT;
794 			memset(&e->msg, 0, sizeof(e->msg));
795 
796 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
797 		} else {
798 			kfree_skb(skb);
799 		}
800 	}
801 
802 	ipmr_cache_free(c);
803 }
804 
805 /* Timer process for the unresolved queue. */
806 static void ipmr_expire_process(struct timer_list *t)
807 {
808 	struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
809 	unsigned long now;
810 	unsigned long expires;
811 	struct mfc_cache *c, *next;
812 
813 	if (!spin_trylock(&mfc_unres_lock)) {
814 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
815 		return;
816 	}
817 
818 	if (list_empty(&mrt->mfc_unres_queue))
819 		goto out;
820 
821 	now = jiffies;
822 	expires = 10*HZ;
823 
824 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
825 		if (time_after(c->mfc_un.unres.expires, now)) {
826 			unsigned long interval = c->mfc_un.unres.expires - now;
827 			if (interval < expires)
828 				expires = interval;
829 			continue;
830 		}
831 
832 		list_del(&c->list);
833 		mroute_netlink_event(mrt, c, RTM_DELROUTE);
834 		ipmr_destroy_unres(mrt, c);
835 	}
836 
837 	if (!list_empty(&mrt->mfc_unres_queue))
838 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
839 
840 out:
841 	spin_unlock(&mfc_unres_lock);
842 }
843 
844 /* Fill oifs list. It is called under write locked mrt_lock. */
845 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
846 				   unsigned char *ttls)
847 {
848 	int vifi;
849 
850 	cache->mfc_un.res.minvif = MAXVIFS;
851 	cache->mfc_un.res.maxvif = 0;
852 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
853 
854 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
855 		if (VIF_EXISTS(mrt, vifi) &&
856 		    ttls[vifi] && ttls[vifi] < 255) {
857 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
858 			if (cache->mfc_un.res.minvif > vifi)
859 				cache->mfc_un.res.minvif = vifi;
860 			if (cache->mfc_un.res.maxvif <= vifi)
861 				cache->mfc_un.res.maxvif = vifi + 1;
862 		}
863 	}
864 	cache->mfc_un.res.lastuse = jiffies;
865 }
866 
867 static int vif_add(struct net *net, struct mr_table *mrt,
868 		   struct vifctl *vifc, int mrtsock)
869 {
870 	int vifi = vifc->vifc_vifi;
871 	struct switchdev_attr attr = {
872 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
873 	};
874 	struct vif_device *v = &mrt->vif_table[vifi];
875 	struct net_device *dev;
876 	struct in_device *in_dev;
877 	int err;
878 
879 	/* Is vif busy ? */
880 	if (VIF_EXISTS(mrt, vifi))
881 		return -EADDRINUSE;
882 
883 	switch (vifc->vifc_flags) {
884 	case VIFF_REGISTER:
885 		if (!ipmr_pimsm_enabled())
886 			return -EINVAL;
887 		/* Special Purpose VIF in PIM
888 		 * All the packets will be sent to the daemon
889 		 */
890 		if (mrt->mroute_reg_vif_num >= 0)
891 			return -EADDRINUSE;
892 		dev = ipmr_reg_vif(net, mrt);
893 		if (!dev)
894 			return -ENOBUFS;
895 		err = dev_set_allmulti(dev, 1);
896 		if (err) {
897 			unregister_netdevice(dev);
898 			dev_put(dev);
899 			return err;
900 		}
901 		break;
902 	case VIFF_TUNNEL:
903 		dev = ipmr_new_tunnel(net, vifc);
904 		if (!dev)
905 			return -ENOBUFS;
906 		err = dev_set_allmulti(dev, 1);
907 		if (err) {
908 			ipmr_del_tunnel(dev, vifc);
909 			dev_put(dev);
910 			return err;
911 		}
912 		break;
913 	case VIFF_USE_IFINDEX:
914 	case 0:
915 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
916 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
917 			if (dev && !__in_dev_get_rtnl(dev)) {
918 				dev_put(dev);
919 				return -EADDRNOTAVAIL;
920 			}
921 		} else {
922 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
923 		}
924 		if (!dev)
925 			return -EADDRNOTAVAIL;
926 		err = dev_set_allmulti(dev, 1);
927 		if (err) {
928 			dev_put(dev);
929 			return err;
930 		}
931 		break;
932 	default:
933 		return -EINVAL;
934 	}
935 
936 	in_dev = __in_dev_get_rtnl(dev);
937 	if (!in_dev) {
938 		dev_put(dev);
939 		return -EADDRNOTAVAIL;
940 	}
941 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
942 	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
943 				    dev->ifindex, &in_dev->cnf);
944 	ip_rt_multicast_event(in_dev);
945 
946 	/* Fill in the VIF structures */
947 
948 	attr.orig_dev = dev;
949 	if (!switchdev_port_attr_get(dev, &attr)) {
950 		memcpy(v->dev_parent_id.id, attr.u.ppid.id, attr.u.ppid.id_len);
951 		v->dev_parent_id.id_len = attr.u.ppid.id_len;
952 	} else {
953 		v->dev_parent_id.id_len = 0;
954 	}
955 	v->rate_limit = vifc->vifc_rate_limit;
956 	v->local = vifc->vifc_lcl_addr.s_addr;
957 	v->remote = vifc->vifc_rmt_addr.s_addr;
958 	v->flags = vifc->vifc_flags;
959 	if (!mrtsock)
960 		v->flags |= VIFF_STATIC;
961 	v->threshold = vifc->vifc_threshold;
962 	v->bytes_in = 0;
963 	v->bytes_out = 0;
964 	v->pkt_in = 0;
965 	v->pkt_out = 0;
966 	v->link = dev->ifindex;
967 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
968 		v->link = dev_get_iflink(dev);
969 
970 	/* And finish update writing critical data */
971 	write_lock_bh(&mrt_lock);
972 	v->dev = dev;
973 	if (v->flags & VIFF_REGISTER)
974 		mrt->mroute_reg_vif_num = vifi;
975 	if (vifi+1 > mrt->maxvif)
976 		mrt->maxvif = vifi+1;
977 	write_unlock_bh(&mrt_lock);
978 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
979 	return 0;
980 }
981 
982 /* called with rcu_read_lock() */
983 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
984 					 __be32 origin,
985 					 __be32 mcastgrp)
986 {
987 	struct mfc_cache_cmp_arg arg = {
988 			.mfc_mcastgrp = mcastgrp,
989 			.mfc_origin = origin
990 	};
991 	struct rhlist_head *tmp, *list;
992 	struct mfc_cache *c;
993 
994 	list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
995 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
996 		return c;
997 
998 	return NULL;
999 }
1000 
1001 /* Look for a (*,*,oif) entry */
1002 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
1003 						    int vifi)
1004 {
1005 	struct mfc_cache_cmp_arg arg = {
1006 			.mfc_mcastgrp = htonl(INADDR_ANY),
1007 			.mfc_origin = htonl(INADDR_ANY)
1008 	};
1009 	struct rhlist_head *tmp, *list;
1010 	struct mfc_cache *c;
1011 
1012 	list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
1013 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
1014 		if (c->mfc_un.res.ttls[vifi] < 255)
1015 			return c;
1016 
1017 	return NULL;
1018 }
1019 
1020 /* Look for a (*,G) entry */
1021 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
1022 					     __be32 mcastgrp, int vifi)
1023 {
1024 	struct mfc_cache_cmp_arg arg = {
1025 			.mfc_mcastgrp = mcastgrp,
1026 			.mfc_origin = htonl(INADDR_ANY)
1027 	};
1028 	struct rhlist_head *tmp, *list;
1029 	struct mfc_cache *c, *proxy;
1030 
1031 	if (mcastgrp == htonl(INADDR_ANY))
1032 		goto skip;
1033 
1034 	list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
1035 	rhl_for_each_entry_rcu(c, tmp, list, mnode) {
1036 		if (c->mfc_un.res.ttls[vifi] < 255)
1037 			return c;
1038 
1039 		/* It's ok if the vifi is part of the static tree */
1040 		proxy = ipmr_cache_find_any_parent(mrt, c->mfc_parent);
1041 		if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
1042 			return c;
1043 	}
1044 
1045 skip:
1046 	return ipmr_cache_find_any_parent(mrt, vifi);
1047 }
1048 
1049 /* Look for a (S,G,iif) entry if parent != -1 */
1050 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
1051 						__be32 origin, __be32 mcastgrp,
1052 						int parent)
1053 {
1054 	struct mfc_cache_cmp_arg arg = {
1055 			.mfc_mcastgrp = mcastgrp,
1056 			.mfc_origin = origin,
1057 	};
1058 	struct rhlist_head *tmp, *list;
1059 	struct mfc_cache *c;
1060 
1061 	list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
1062 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
1063 		if (parent == -1 || parent == c->mfc_parent)
1064 			return c;
1065 
1066 	return NULL;
1067 }
1068 
1069 /* Allocate a multicast cache entry */
1070 static struct mfc_cache *ipmr_cache_alloc(void)
1071 {
1072 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1073 
1074 	if (c) {
1075 		c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1076 		c->mfc_un.res.minvif = MAXVIFS;
1077 		refcount_set(&c->mfc_un.res.refcount, 1);
1078 	}
1079 	return c;
1080 }
1081 
1082 static struct mfc_cache *ipmr_cache_alloc_unres(void)
1083 {
1084 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1085 
1086 	if (c) {
1087 		skb_queue_head_init(&c->mfc_un.unres.unresolved);
1088 		c->mfc_un.unres.expires = jiffies + 10*HZ;
1089 	}
1090 	return c;
1091 }
1092 
1093 /* A cache entry has gone into a resolved state from queued */
1094 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1095 			       struct mfc_cache *uc, struct mfc_cache *c)
1096 {
1097 	struct sk_buff *skb;
1098 	struct nlmsgerr *e;
1099 
1100 	/* Play the pending entries through our router */
1101 	while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1102 		if (ip_hdr(skb)->version == 0) {
1103 			struct nlmsghdr *nlh = skb_pull(skb,
1104 							sizeof(struct iphdr));
1105 
1106 			if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1107 				nlh->nlmsg_len = skb_tail_pointer(skb) -
1108 						 (u8 *)nlh;
1109 			} else {
1110 				nlh->nlmsg_type = NLMSG_ERROR;
1111 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1112 				skb_trim(skb, nlh->nlmsg_len);
1113 				e = nlmsg_data(nlh);
1114 				e->error = -EMSGSIZE;
1115 				memset(&e->msg, 0, sizeof(e->msg));
1116 			}
1117 
1118 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1119 		} else {
1120 			ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1121 		}
1122 	}
1123 }
1124 
1125 /* Bounce a cache query up to mrouted and netlink.
1126  *
1127  * Called under mrt_lock.
1128  */
1129 static int ipmr_cache_report(struct mr_table *mrt,
1130 			     struct sk_buff *pkt, vifi_t vifi, int assert)
1131 {
1132 	const int ihl = ip_hdrlen(pkt);
1133 	struct sock *mroute_sk;
1134 	struct igmphdr *igmp;
1135 	struct igmpmsg *msg;
1136 	struct sk_buff *skb;
1137 	int ret;
1138 
1139 	if (assert == IGMPMSG_WHOLEPKT)
1140 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1141 	else
1142 		skb = alloc_skb(128, GFP_ATOMIC);
1143 
1144 	if (!skb)
1145 		return -ENOBUFS;
1146 
1147 	if (assert == IGMPMSG_WHOLEPKT) {
1148 		/* Ugly, but we have no choice with this interface.
1149 		 * Duplicate old header, fix ihl, length etc.
1150 		 * And all this only to mangle msg->im_msgtype and
1151 		 * to set msg->im_mbz to "mbz" :-)
1152 		 */
1153 		skb_push(skb, sizeof(struct iphdr));
1154 		skb_reset_network_header(skb);
1155 		skb_reset_transport_header(skb);
1156 		msg = (struct igmpmsg *)skb_network_header(skb);
1157 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1158 		msg->im_msgtype = IGMPMSG_WHOLEPKT;
1159 		msg->im_mbz = 0;
1160 		msg->im_vif = mrt->mroute_reg_vif_num;
1161 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1162 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1163 					     sizeof(struct iphdr));
1164 	} else {
1165 		/* Copy the IP header */
1166 		skb_set_network_header(skb, skb->len);
1167 		skb_put(skb, ihl);
1168 		skb_copy_to_linear_data(skb, pkt->data, ihl);
1169 		/* Flag to the kernel this is a route add */
1170 		ip_hdr(skb)->protocol = 0;
1171 		msg = (struct igmpmsg *)skb_network_header(skb);
1172 		msg->im_vif = vifi;
1173 		skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1174 		/* Add our header */
1175 		igmp = skb_put(skb, sizeof(struct igmphdr));
1176 		igmp->type = assert;
1177 		msg->im_msgtype = assert;
1178 		igmp->code = 0;
1179 		ip_hdr(skb)->tot_len = htons(skb->len);	/* Fix the length */
1180 		skb->transport_header = skb->network_header;
1181 	}
1182 
1183 	rcu_read_lock();
1184 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1185 	if (!mroute_sk) {
1186 		rcu_read_unlock();
1187 		kfree_skb(skb);
1188 		return -EINVAL;
1189 	}
1190 
1191 	igmpmsg_netlink_event(mrt, skb);
1192 
1193 	/* Deliver to mrouted */
1194 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1195 	rcu_read_unlock();
1196 	if (ret < 0) {
1197 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1198 		kfree_skb(skb);
1199 	}
1200 
1201 	return ret;
1202 }
1203 
1204 /* Queue a packet for resolution. It gets locked cache entry! */
1205 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1206 				 struct sk_buff *skb, struct net_device *dev)
1207 {
1208 	const struct iphdr *iph = ip_hdr(skb);
1209 	struct mfc_cache *c;
1210 	bool found = false;
1211 	int err;
1212 
1213 	spin_lock_bh(&mfc_unres_lock);
1214 	list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1215 		if (c->mfc_mcastgrp == iph->daddr &&
1216 		    c->mfc_origin == iph->saddr) {
1217 			found = true;
1218 			break;
1219 		}
1220 	}
1221 
1222 	if (!found) {
1223 		/* Create a new entry if allowable */
1224 		if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1225 		    (c = ipmr_cache_alloc_unres()) == NULL) {
1226 			spin_unlock_bh(&mfc_unres_lock);
1227 
1228 			kfree_skb(skb);
1229 			return -ENOBUFS;
1230 		}
1231 
1232 		/* Fill in the new cache entry */
1233 		c->mfc_parent	= -1;
1234 		c->mfc_origin	= iph->saddr;
1235 		c->mfc_mcastgrp	= iph->daddr;
1236 
1237 		/* Reflect first query at mrouted. */
1238 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1239 		if (err < 0) {
1240 			/* If the report failed throw the cache entry
1241 			   out - Brad Parker
1242 			 */
1243 			spin_unlock_bh(&mfc_unres_lock);
1244 
1245 			ipmr_cache_free(c);
1246 			kfree_skb(skb);
1247 			return err;
1248 		}
1249 
1250 		atomic_inc(&mrt->cache_resolve_queue_len);
1251 		list_add(&c->list, &mrt->mfc_unres_queue);
1252 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1253 
1254 		if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1255 			mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1256 	}
1257 
1258 	/* See if we can append the packet */
1259 	if (c->mfc_un.unres.unresolved.qlen > 3) {
1260 		kfree_skb(skb);
1261 		err = -ENOBUFS;
1262 	} else {
1263 		if (dev) {
1264 			skb->dev = dev;
1265 			skb->skb_iif = dev->ifindex;
1266 		}
1267 		skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1268 		err = 0;
1269 	}
1270 
1271 	spin_unlock_bh(&mfc_unres_lock);
1272 	return err;
1273 }
1274 
1275 /* MFC cache manipulation by user space mroute daemon */
1276 
1277 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1278 {
1279 	struct net *net = read_pnet(&mrt->net);
1280 	struct mfc_cache *c;
1281 
1282 	/* The entries are added/deleted only under RTNL */
1283 	rcu_read_lock();
1284 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1285 				   mfc->mfcc_mcastgrp.s_addr, parent);
1286 	rcu_read_unlock();
1287 	if (!c)
1288 		return -ENOENT;
1289 	rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1290 	list_del_rcu(&c->list);
1291 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1292 	mroute_netlink_event(mrt, c, RTM_DELROUTE);
1293 	ipmr_cache_put(c);
1294 
1295 	return 0;
1296 }
1297 
1298 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1299 			struct mfcctl *mfc, int mrtsock, int parent)
1300 {
1301 	struct mfc_cache *uc, *c;
1302 	bool found;
1303 	int ret;
1304 
1305 	if (mfc->mfcc_parent >= MAXVIFS)
1306 		return -ENFILE;
1307 
1308 	/* The entries are added/deleted only under RTNL */
1309 	rcu_read_lock();
1310 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1311 				   mfc->mfcc_mcastgrp.s_addr, parent);
1312 	rcu_read_unlock();
1313 	if (c) {
1314 		write_lock_bh(&mrt_lock);
1315 		c->mfc_parent = mfc->mfcc_parent;
1316 		ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1317 		if (!mrtsock)
1318 			c->mfc_flags |= MFC_STATIC;
1319 		write_unlock_bh(&mrt_lock);
1320 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1321 					      mrt->id);
1322 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1323 		return 0;
1324 	}
1325 
1326 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1327 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1328 		return -EINVAL;
1329 
1330 	c = ipmr_cache_alloc();
1331 	if (!c)
1332 		return -ENOMEM;
1333 
1334 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1335 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1336 	c->mfc_parent = mfc->mfcc_parent;
1337 	ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1338 	if (!mrtsock)
1339 		c->mfc_flags |= MFC_STATIC;
1340 
1341 	ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->mnode,
1342 				  ipmr_rht_params);
1343 	if (ret) {
1344 		pr_err("ipmr: rhtable insert error %d\n", ret);
1345 		ipmr_cache_free(c);
1346 		return ret;
1347 	}
1348 	list_add_tail_rcu(&c->list, &mrt->mfc_cache_list);
1349 	/* Check to see if we resolved a queued list. If so we
1350 	 * need to send on the frames and tidy up.
1351 	 */
1352 	found = false;
1353 	spin_lock_bh(&mfc_unres_lock);
1354 	list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1355 		if (uc->mfc_origin == c->mfc_origin &&
1356 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1357 			list_del(&uc->list);
1358 			atomic_dec(&mrt->cache_resolve_queue_len);
1359 			found = true;
1360 			break;
1361 		}
1362 	}
1363 	if (list_empty(&mrt->mfc_unres_queue))
1364 		del_timer(&mrt->ipmr_expire_timer);
1365 	spin_unlock_bh(&mfc_unres_lock);
1366 
1367 	if (found) {
1368 		ipmr_cache_resolve(net, mrt, uc, c);
1369 		ipmr_cache_free(uc);
1370 	}
1371 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1372 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1373 	return 0;
1374 }
1375 
1376 /* Close the multicast socket, and clear the vif tables etc */
1377 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1378 {
1379 	struct net *net = read_pnet(&mrt->net);
1380 	struct mfc_cache *c, *tmp;
1381 	LIST_HEAD(list);
1382 	int i;
1383 
1384 	/* Shut down all active vif entries */
1385 	for (i = 0; i < mrt->maxvif; i++) {
1386 		if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1387 			continue;
1388 		vif_delete(mrt, i, 0, &list);
1389 	}
1390 	unregister_netdevice_many(&list);
1391 
1392 	/* Wipe the cache */
1393 	list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1394 		if (!all && (c->mfc_flags & MFC_STATIC))
1395 			continue;
1396 		rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1397 		list_del_rcu(&c->list);
1398 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c,
1399 					      mrt->id);
1400 		mroute_netlink_event(mrt, c, RTM_DELROUTE);
1401 		ipmr_cache_put(c);
1402 	}
1403 
1404 	if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1405 		spin_lock_bh(&mfc_unres_lock);
1406 		list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1407 			list_del(&c->list);
1408 			mroute_netlink_event(mrt, c, RTM_DELROUTE);
1409 			ipmr_destroy_unres(mrt, c);
1410 		}
1411 		spin_unlock_bh(&mfc_unres_lock);
1412 	}
1413 }
1414 
1415 /* called from ip_ra_control(), before an RCU grace period,
1416  * we dont need to call synchronize_rcu() here
1417  */
1418 static void mrtsock_destruct(struct sock *sk)
1419 {
1420 	struct net *net = sock_net(sk);
1421 	struct mr_table *mrt;
1422 
1423 	ASSERT_RTNL();
1424 	ipmr_for_each_table(mrt, net) {
1425 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1426 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1427 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1428 						    NETCONFA_MC_FORWARDING,
1429 						    NETCONFA_IFINDEX_ALL,
1430 						    net->ipv4.devconf_all);
1431 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1432 			mroute_clean_tables(mrt, false);
1433 		}
1434 	}
1435 }
1436 
1437 /* Socket options and virtual interface manipulation. The whole
1438  * virtual interface system is a complete heap, but unfortunately
1439  * that's how BSD mrouted happens to think. Maybe one day with a proper
1440  * MOSPF/PIM router set up we can clean this up.
1441  */
1442 
1443 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1444 			 unsigned int optlen)
1445 {
1446 	struct net *net = sock_net(sk);
1447 	int val, ret = 0, parent = 0;
1448 	struct mr_table *mrt;
1449 	struct vifctl vif;
1450 	struct mfcctl mfc;
1451 	u32 uval;
1452 
1453 	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1454 	rtnl_lock();
1455 	if (sk->sk_type != SOCK_RAW ||
1456 	    inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1457 		ret = -EOPNOTSUPP;
1458 		goto out_unlock;
1459 	}
1460 
1461 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1462 	if (!mrt) {
1463 		ret = -ENOENT;
1464 		goto out_unlock;
1465 	}
1466 	if (optname != MRT_INIT) {
1467 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1468 		    !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1469 			ret = -EACCES;
1470 			goto out_unlock;
1471 		}
1472 	}
1473 
1474 	switch (optname) {
1475 	case MRT_INIT:
1476 		if (optlen != sizeof(int)) {
1477 			ret = -EINVAL;
1478 			break;
1479 		}
1480 		if (rtnl_dereference(mrt->mroute_sk)) {
1481 			ret = -EADDRINUSE;
1482 			break;
1483 		}
1484 
1485 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1486 		if (ret == 0) {
1487 			rcu_assign_pointer(mrt->mroute_sk, sk);
1488 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1489 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1490 						    NETCONFA_MC_FORWARDING,
1491 						    NETCONFA_IFINDEX_ALL,
1492 						    net->ipv4.devconf_all);
1493 		}
1494 		break;
1495 	case MRT_DONE:
1496 		if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1497 			ret = -EACCES;
1498 		} else {
1499 			ret = ip_ra_control(sk, 0, NULL);
1500 			goto out_unlock;
1501 		}
1502 		break;
1503 	case MRT_ADD_VIF:
1504 	case MRT_DEL_VIF:
1505 		if (optlen != sizeof(vif)) {
1506 			ret = -EINVAL;
1507 			break;
1508 		}
1509 		if (copy_from_user(&vif, optval, sizeof(vif))) {
1510 			ret = -EFAULT;
1511 			break;
1512 		}
1513 		if (vif.vifc_vifi >= MAXVIFS) {
1514 			ret = -ENFILE;
1515 			break;
1516 		}
1517 		if (optname == MRT_ADD_VIF) {
1518 			ret = vif_add(net, mrt, &vif,
1519 				      sk == rtnl_dereference(mrt->mroute_sk));
1520 		} else {
1521 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1522 		}
1523 		break;
1524 	/* Manipulate the forwarding caches. These live
1525 	 * in a sort of kernel/user symbiosis.
1526 	 */
1527 	case MRT_ADD_MFC:
1528 	case MRT_DEL_MFC:
1529 		parent = -1;
1530 		/* fall through */
1531 	case MRT_ADD_MFC_PROXY:
1532 	case MRT_DEL_MFC_PROXY:
1533 		if (optlen != sizeof(mfc)) {
1534 			ret = -EINVAL;
1535 			break;
1536 		}
1537 		if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1538 			ret = -EFAULT;
1539 			break;
1540 		}
1541 		if (parent == 0)
1542 			parent = mfc.mfcc_parent;
1543 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1544 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1545 		else
1546 			ret = ipmr_mfc_add(net, mrt, &mfc,
1547 					   sk == rtnl_dereference(mrt->mroute_sk),
1548 					   parent);
1549 		break;
1550 	/* Control PIM assert. */
1551 	case MRT_ASSERT:
1552 		if (optlen != sizeof(val)) {
1553 			ret = -EINVAL;
1554 			break;
1555 		}
1556 		if (get_user(val, (int __user *)optval)) {
1557 			ret = -EFAULT;
1558 			break;
1559 		}
1560 		mrt->mroute_do_assert = val;
1561 		break;
1562 	case MRT_PIM:
1563 		if (!ipmr_pimsm_enabled()) {
1564 			ret = -ENOPROTOOPT;
1565 			break;
1566 		}
1567 		if (optlen != sizeof(val)) {
1568 			ret = -EINVAL;
1569 			break;
1570 		}
1571 		if (get_user(val, (int __user *)optval)) {
1572 			ret = -EFAULT;
1573 			break;
1574 		}
1575 
1576 		val = !!val;
1577 		if (val != mrt->mroute_do_pim) {
1578 			mrt->mroute_do_pim = val;
1579 			mrt->mroute_do_assert = val;
1580 		}
1581 		break;
1582 	case MRT_TABLE:
1583 		if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1584 			ret = -ENOPROTOOPT;
1585 			break;
1586 		}
1587 		if (optlen != sizeof(uval)) {
1588 			ret = -EINVAL;
1589 			break;
1590 		}
1591 		if (get_user(uval, (u32 __user *)optval)) {
1592 			ret = -EFAULT;
1593 			break;
1594 		}
1595 
1596 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1597 			ret = -EBUSY;
1598 		} else {
1599 			mrt = ipmr_new_table(net, uval);
1600 			if (IS_ERR(mrt))
1601 				ret = PTR_ERR(mrt);
1602 			else
1603 				raw_sk(sk)->ipmr_table = uval;
1604 		}
1605 		break;
1606 	/* Spurious command, or MRT_VERSION which you cannot set. */
1607 	default:
1608 		ret = -ENOPROTOOPT;
1609 	}
1610 out_unlock:
1611 	rtnl_unlock();
1612 	return ret;
1613 }
1614 
1615 /* Getsock opt support for the multicast routing system. */
1616 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1617 {
1618 	int olr;
1619 	int val;
1620 	struct net *net = sock_net(sk);
1621 	struct mr_table *mrt;
1622 
1623 	if (sk->sk_type != SOCK_RAW ||
1624 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1625 		return -EOPNOTSUPP;
1626 
1627 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1628 	if (!mrt)
1629 		return -ENOENT;
1630 
1631 	switch (optname) {
1632 	case MRT_VERSION:
1633 		val = 0x0305;
1634 		break;
1635 	case MRT_PIM:
1636 		if (!ipmr_pimsm_enabled())
1637 			return -ENOPROTOOPT;
1638 		val = mrt->mroute_do_pim;
1639 		break;
1640 	case MRT_ASSERT:
1641 		val = mrt->mroute_do_assert;
1642 		break;
1643 	default:
1644 		return -ENOPROTOOPT;
1645 	}
1646 
1647 	if (get_user(olr, optlen))
1648 		return -EFAULT;
1649 	olr = min_t(unsigned int, olr, sizeof(int));
1650 	if (olr < 0)
1651 		return -EINVAL;
1652 	if (put_user(olr, optlen))
1653 		return -EFAULT;
1654 	if (copy_to_user(optval, &val, olr))
1655 		return -EFAULT;
1656 	return 0;
1657 }
1658 
1659 /* The IP multicast ioctl support routines. */
1660 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1661 {
1662 	struct sioc_sg_req sr;
1663 	struct sioc_vif_req vr;
1664 	struct vif_device *vif;
1665 	struct mfc_cache *c;
1666 	struct net *net = sock_net(sk);
1667 	struct mr_table *mrt;
1668 
1669 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1670 	if (!mrt)
1671 		return -ENOENT;
1672 
1673 	switch (cmd) {
1674 	case SIOCGETVIFCNT:
1675 		if (copy_from_user(&vr, arg, sizeof(vr)))
1676 			return -EFAULT;
1677 		if (vr.vifi >= mrt->maxvif)
1678 			return -EINVAL;
1679 		read_lock(&mrt_lock);
1680 		vif = &mrt->vif_table[vr.vifi];
1681 		if (VIF_EXISTS(mrt, vr.vifi)) {
1682 			vr.icount = vif->pkt_in;
1683 			vr.ocount = vif->pkt_out;
1684 			vr.ibytes = vif->bytes_in;
1685 			vr.obytes = vif->bytes_out;
1686 			read_unlock(&mrt_lock);
1687 
1688 			if (copy_to_user(arg, &vr, sizeof(vr)))
1689 				return -EFAULT;
1690 			return 0;
1691 		}
1692 		read_unlock(&mrt_lock);
1693 		return -EADDRNOTAVAIL;
1694 	case SIOCGETSGCNT:
1695 		if (copy_from_user(&sr, arg, sizeof(sr)))
1696 			return -EFAULT;
1697 
1698 		rcu_read_lock();
1699 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1700 		if (c) {
1701 			sr.pktcnt = c->mfc_un.res.pkt;
1702 			sr.bytecnt = c->mfc_un.res.bytes;
1703 			sr.wrong_if = c->mfc_un.res.wrong_if;
1704 			rcu_read_unlock();
1705 
1706 			if (copy_to_user(arg, &sr, sizeof(sr)))
1707 				return -EFAULT;
1708 			return 0;
1709 		}
1710 		rcu_read_unlock();
1711 		return -EADDRNOTAVAIL;
1712 	default:
1713 		return -ENOIOCTLCMD;
1714 	}
1715 }
1716 
1717 #ifdef CONFIG_COMPAT
1718 struct compat_sioc_sg_req {
1719 	struct in_addr src;
1720 	struct in_addr grp;
1721 	compat_ulong_t pktcnt;
1722 	compat_ulong_t bytecnt;
1723 	compat_ulong_t wrong_if;
1724 };
1725 
1726 struct compat_sioc_vif_req {
1727 	vifi_t	vifi;		/* Which iface */
1728 	compat_ulong_t icount;
1729 	compat_ulong_t ocount;
1730 	compat_ulong_t ibytes;
1731 	compat_ulong_t obytes;
1732 };
1733 
1734 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1735 {
1736 	struct compat_sioc_sg_req sr;
1737 	struct compat_sioc_vif_req vr;
1738 	struct vif_device *vif;
1739 	struct mfc_cache *c;
1740 	struct net *net = sock_net(sk);
1741 	struct mr_table *mrt;
1742 
1743 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1744 	if (!mrt)
1745 		return -ENOENT;
1746 
1747 	switch (cmd) {
1748 	case SIOCGETVIFCNT:
1749 		if (copy_from_user(&vr, arg, sizeof(vr)))
1750 			return -EFAULT;
1751 		if (vr.vifi >= mrt->maxvif)
1752 			return -EINVAL;
1753 		read_lock(&mrt_lock);
1754 		vif = &mrt->vif_table[vr.vifi];
1755 		if (VIF_EXISTS(mrt, vr.vifi)) {
1756 			vr.icount = vif->pkt_in;
1757 			vr.ocount = vif->pkt_out;
1758 			vr.ibytes = vif->bytes_in;
1759 			vr.obytes = vif->bytes_out;
1760 			read_unlock(&mrt_lock);
1761 
1762 			if (copy_to_user(arg, &vr, sizeof(vr)))
1763 				return -EFAULT;
1764 			return 0;
1765 		}
1766 		read_unlock(&mrt_lock);
1767 		return -EADDRNOTAVAIL;
1768 	case SIOCGETSGCNT:
1769 		if (copy_from_user(&sr, arg, sizeof(sr)))
1770 			return -EFAULT;
1771 
1772 		rcu_read_lock();
1773 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1774 		if (c) {
1775 			sr.pktcnt = c->mfc_un.res.pkt;
1776 			sr.bytecnt = c->mfc_un.res.bytes;
1777 			sr.wrong_if = c->mfc_un.res.wrong_if;
1778 			rcu_read_unlock();
1779 
1780 			if (copy_to_user(arg, &sr, sizeof(sr)))
1781 				return -EFAULT;
1782 			return 0;
1783 		}
1784 		rcu_read_unlock();
1785 		return -EADDRNOTAVAIL;
1786 	default:
1787 		return -ENOIOCTLCMD;
1788 	}
1789 }
1790 #endif
1791 
1792 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1793 {
1794 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1795 	struct net *net = dev_net(dev);
1796 	struct mr_table *mrt;
1797 	struct vif_device *v;
1798 	int ct;
1799 
1800 	if (event != NETDEV_UNREGISTER)
1801 		return NOTIFY_DONE;
1802 
1803 	ipmr_for_each_table(mrt, net) {
1804 		v = &mrt->vif_table[0];
1805 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1806 			if (v->dev == dev)
1807 				vif_delete(mrt, ct, 1, NULL);
1808 		}
1809 	}
1810 	return NOTIFY_DONE;
1811 }
1812 
1813 static struct notifier_block ip_mr_notifier = {
1814 	.notifier_call = ipmr_device_event,
1815 };
1816 
1817 /* Encapsulate a packet by attaching a valid IPIP header to it.
1818  * This avoids tunnel drivers and other mess and gives us the speed so
1819  * important for multicast video.
1820  */
1821 static void ip_encap(struct net *net, struct sk_buff *skb,
1822 		     __be32 saddr, __be32 daddr)
1823 {
1824 	struct iphdr *iph;
1825 	const struct iphdr *old_iph = ip_hdr(skb);
1826 
1827 	skb_push(skb, sizeof(struct iphdr));
1828 	skb->transport_header = skb->network_header;
1829 	skb_reset_network_header(skb);
1830 	iph = ip_hdr(skb);
1831 
1832 	iph->version	=	4;
1833 	iph->tos	=	old_iph->tos;
1834 	iph->ttl	=	old_iph->ttl;
1835 	iph->frag_off	=	0;
1836 	iph->daddr	=	daddr;
1837 	iph->saddr	=	saddr;
1838 	iph->protocol	=	IPPROTO_IPIP;
1839 	iph->ihl	=	5;
1840 	iph->tot_len	=	htons(skb->len);
1841 	ip_select_ident(net, skb, NULL);
1842 	ip_send_check(iph);
1843 
1844 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1845 	nf_reset(skb);
1846 }
1847 
1848 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1849 				      struct sk_buff *skb)
1850 {
1851 	struct ip_options *opt = &(IPCB(skb)->opt);
1852 
1853 	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1854 	IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1855 
1856 	if (unlikely(opt->optlen))
1857 		ip_forward_options(skb);
1858 
1859 	return dst_output(net, sk, skb);
1860 }
1861 
1862 #ifdef CONFIG_NET_SWITCHDEV
1863 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1864 				   int in_vifi, int out_vifi)
1865 {
1866 	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1867 	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1868 
1869 	if (!skb->offload_mr_fwd_mark)
1870 		return false;
1871 	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1872 		return false;
1873 	return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1874 					&in_vif->dev_parent_id);
1875 }
1876 #else
1877 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1878 				   int in_vifi, int out_vifi)
1879 {
1880 	return false;
1881 }
1882 #endif
1883 
1884 /* Processing handlers for ipmr_forward */
1885 
1886 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1887 			    int in_vifi, struct sk_buff *skb,
1888 			    struct mfc_cache *c, int vifi)
1889 {
1890 	const struct iphdr *iph = ip_hdr(skb);
1891 	struct vif_device *vif = &mrt->vif_table[vifi];
1892 	struct net_device *dev;
1893 	struct rtable *rt;
1894 	struct flowi4 fl4;
1895 	int    encap = 0;
1896 
1897 	if (!vif->dev)
1898 		goto out_free;
1899 
1900 	if (vif->flags & VIFF_REGISTER) {
1901 		vif->pkt_out++;
1902 		vif->bytes_out += skb->len;
1903 		vif->dev->stats.tx_bytes += skb->len;
1904 		vif->dev->stats.tx_packets++;
1905 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1906 		goto out_free;
1907 	}
1908 
1909 	if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1910 		goto out_free;
1911 
1912 	if (vif->flags & VIFF_TUNNEL) {
1913 		rt = ip_route_output_ports(net, &fl4, NULL,
1914 					   vif->remote, vif->local,
1915 					   0, 0,
1916 					   IPPROTO_IPIP,
1917 					   RT_TOS(iph->tos), vif->link);
1918 		if (IS_ERR(rt))
1919 			goto out_free;
1920 		encap = sizeof(struct iphdr);
1921 	} else {
1922 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1923 					   0, 0,
1924 					   IPPROTO_IPIP,
1925 					   RT_TOS(iph->tos), vif->link);
1926 		if (IS_ERR(rt))
1927 			goto out_free;
1928 	}
1929 
1930 	dev = rt->dst.dev;
1931 
1932 	if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1933 		/* Do not fragment multicasts. Alas, IPv4 does not
1934 		 * allow to send ICMP, so that packets will disappear
1935 		 * to blackhole.
1936 		 */
1937 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1938 		ip_rt_put(rt);
1939 		goto out_free;
1940 	}
1941 
1942 	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1943 
1944 	if (skb_cow(skb, encap)) {
1945 		ip_rt_put(rt);
1946 		goto out_free;
1947 	}
1948 
1949 	vif->pkt_out++;
1950 	vif->bytes_out += skb->len;
1951 
1952 	skb_dst_drop(skb);
1953 	skb_dst_set(skb, &rt->dst);
1954 	ip_decrease_ttl(ip_hdr(skb));
1955 
1956 	/* FIXME: forward and output firewalls used to be called here.
1957 	 * What do we do with netfilter? -- RR
1958 	 */
1959 	if (vif->flags & VIFF_TUNNEL) {
1960 		ip_encap(net, skb, vif->local, vif->remote);
1961 		/* FIXME: extra output firewall step used to be here. --RR */
1962 		vif->dev->stats.tx_packets++;
1963 		vif->dev->stats.tx_bytes += skb->len;
1964 	}
1965 
1966 	IPCB(skb)->flags |= IPSKB_FORWARDED;
1967 
1968 	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1969 	 * not only before forwarding, but after forwarding on all output
1970 	 * interfaces. It is clear, if mrouter runs a multicasting
1971 	 * program, it should receive packets not depending to what interface
1972 	 * program is joined.
1973 	 * If we will not make it, the program will have to join on all
1974 	 * interfaces. On the other hand, multihoming host (or router, but
1975 	 * not mrouter) cannot join to more than one interface - it will
1976 	 * result in receiving multiple packets.
1977 	 */
1978 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1979 		net, NULL, skb, skb->dev, dev,
1980 		ipmr_forward_finish);
1981 	return;
1982 
1983 out_free:
1984 	kfree_skb(skb);
1985 }
1986 
1987 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1988 {
1989 	int ct;
1990 
1991 	for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1992 		if (mrt->vif_table[ct].dev == dev)
1993 			break;
1994 	}
1995 	return ct;
1996 }
1997 
1998 /* "local" means that we should preserve one skb (for local delivery) */
1999 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
2000 			  struct net_device *dev, struct sk_buff *skb,
2001 			  struct mfc_cache *cache, int local)
2002 {
2003 	int true_vifi = ipmr_find_vif(mrt, dev);
2004 	int psend = -1;
2005 	int vif, ct;
2006 
2007 	vif = cache->mfc_parent;
2008 	cache->mfc_un.res.pkt++;
2009 	cache->mfc_un.res.bytes += skb->len;
2010 	cache->mfc_un.res.lastuse = jiffies;
2011 
2012 	if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
2013 		struct mfc_cache *cache_proxy;
2014 
2015 		/* For an (*,G) entry, we only check that the incomming
2016 		 * interface is part of the static tree.
2017 		 */
2018 		cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
2019 		if (cache_proxy &&
2020 		    cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
2021 			goto forward;
2022 	}
2023 
2024 	/* Wrong interface: drop packet and (maybe) send PIM assert. */
2025 	if (mrt->vif_table[vif].dev != dev) {
2026 		if (rt_is_output_route(skb_rtable(skb))) {
2027 			/* It is our own packet, looped back.
2028 			 * Very complicated situation...
2029 			 *
2030 			 * The best workaround until routing daemons will be
2031 			 * fixed is not to redistribute packet, if it was
2032 			 * send through wrong interface. It means, that
2033 			 * multicast applications WILL NOT work for
2034 			 * (S,G), which have default multicast route pointing
2035 			 * to wrong oif. In any case, it is not a good
2036 			 * idea to use multicasting applications on router.
2037 			 */
2038 			goto dont_forward;
2039 		}
2040 
2041 		cache->mfc_un.res.wrong_if++;
2042 
2043 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
2044 		    /* pimsm uses asserts, when switching from RPT to SPT,
2045 		     * so that we cannot check that packet arrived on an oif.
2046 		     * It is bad, but otherwise we would need to move pretty
2047 		     * large chunk of pimd to kernel. Ough... --ANK
2048 		     */
2049 		    (mrt->mroute_do_pim ||
2050 		     cache->mfc_un.res.ttls[true_vifi] < 255) &&
2051 		    time_after(jiffies,
2052 			       cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2053 			cache->mfc_un.res.last_assert = jiffies;
2054 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2055 		}
2056 		goto dont_forward;
2057 	}
2058 
2059 forward:
2060 	mrt->vif_table[vif].pkt_in++;
2061 	mrt->vif_table[vif].bytes_in += skb->len;
2062 
2063 	/* Forward the frame */
2064 	if (cache->mfc_origin == htonl(INADDR_ANY) &&
2065 	    cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
2066 		if (true_vifi >= 0 &&
2067 		    true_vifi != cache->mfc_parent &&
2068 		    ip_hdr(skb)->ttl >
2069 				cache->mfc_un.res.ttls[cache->mfc_parent]) {
2070 			/* It's an (*,*) entry and the packet is not coming from
2071 			 * the upstream: forward the packet to the upstream
2072 			 * only.
2073 			 */
2074 			psend = cache->mfc_parent;
2075 			goto last_forward;
2076 		}
2077 		goto dont_forward;
2078 	}
2079 	for (ct = cache->mfc_un.res.maxvif - 1;
2080 	     ct >= cache->mfc_un.res.minvif; ct--) {
2081 		/* For (*,G) entry, don't forward to the incoming interface */
2082 		if ((cache->mfc_origin != htonl(INADDR_ANY) ||
2083 		     ct != true_vifi) &&
2084 		    ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
2085 			if (psend != -1) {
2086 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2087 
2088 				if (skb2)
2089 					ipmr_queue_xmit(net, mrt, true_vifi,
2090 							skb2, cache, psend);
2091 			}
2092 			psend = ct;
2093 		}
2094 	}
2095 last_forward:
2096 	if (psend != -1) {
2097 		if (local) {
2098 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2099 
2100 			if (skb2)
2101 				ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2102 						cache, psend);
2103 		} else {
2104 			ipmr_queue_xmit(net, mrt, true_vifi, skb, cache, psend);
2105 			return;
2106 		}
2107 	}
2108 
2109 dont_forward:
2110 	if (!local)
2111 		kfree_skb(skb);
2112 }
2113 
2114 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2115 {
2116 	struct rtable *rt = skb_rtable(skb);
2117 	struct iphdr *iph = ip_hdr(skb);
2118 	struct flowi4 fl4 = {
2119 		.daddr = iph->daddr,
2120 		.saddr = iph->saddr,
2121 		.flowi4_tos = RT_TOS(iph->tos),
2122 		.flowi4_oif = (rt_is_output_route(rt) ?
2123 			       skb->dev->ifindex : 0),
2124 		.flowi4_iif = (rt_is_output_route(rt) ?
2125 			       LOOPBACK_IFINDEX :
2126 			       skb->dev->ifindex),
2127 		.flowi4_mark = skb->mark,
2128 	};
2129 	struct mr_table *mrt;
2130 	int err;
2131 
2132 	err = ipmr_fib_lookup(net, &fl4, &mrt);
2133 	if (err)
2134 		return ERR_PTR(err);
2135 	return mrt;
2136 }
2137 
2138 /* Multicast packets for forwarding arrive here
2139  * Called with rcu_read_lock();
2140  */
2141 int ip_mr_input(struct sk_buff *skb)
2142 {
2143 	struct mfc_cache *cache;
2144 	struct net *net = dev_net(skb->dev);
2145 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2146 	struct mr_table *mrt;
2147 	struct net_device *dev;
2148 
2149 	/* skb->dev passed in is the loX master dev for vrfs.
2150 	 * As there are no vifs associated with loopback devices,
2151 	 * get the proper interface that does have a vif associated with it.
2152 	 */
2153 	dev = skb->dev;
2154 	if (netif_is_l3_master(skb->dev)) {
2155 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2156 		if (!dev) {
2157 			kfree_skb(skb);
2158 			return -ENODEV;
2159 		}
2160 	}
2161 
2162 	/* Packet is looped back after forward, it should not be
2163 	 * forwarded second time, but still can be delivered locally.
2164 	 */
2165 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2166 		goto dont_forward;
2167 
2168 	mrt = ipmr_rt_fib_lookup(net, skb);
2169 	if (IS_ERR(mrt)) {
2170 		kfree_skb(skb);
2171 		return PTR_ERR(mrt);
2172 	}
2173 	if (!local) {
2174 		if (IPCB(skb)->opt.router_alert) {
2175 			if (ip_call_ra_chain(skb))
2176 				return 0;
2177 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2178 			/* IGMPv1 (and broken IGMPv2 implementations sort of
2179 			 * Cisco IOS <= 11.2(8)) do not put router alert
2180 			 * option to IGMP packets destined to routable
2181 			 * groups. It is very bad, because it means
2182 			 * that we can forward NO IGMP messages.
2183 			 */
2184 			struct sock *mroute_sk;
2185 
2186 			mroute_sk = rcu_dereference(mrt->mroute_sk);
2187 			if (mroute_sk) {
2188 				nf_reset(skb);
2189 				raw_rcv(mroute_sk, skb);
2190 				return 0;
2191 			}
2192 		    }
2193 	}
2194 
2195 	/* already under rcu_read_lock() */
2196 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2197 	if (!cache) {
2198 		int vif = ipmr_find_vif(mrt, dev);
2199 
2200 		if (vif >= 0)
2201 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2202 						    vif);
2203 	}
2204 
2205 	/* No usable cache entry */
2206 	if (!cache) {
2207 		int vif;
2208 
2209 		if (local) {
2210 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2211 			ip_local_deliver(skb);
2212 			if (!skb2)
2213 				return -ENOBUFS;
2214 			skb = skb2;
2215 		}
2216 
2217 		read_lock(&mrt_lock);
2218 		vif = ipmr_find_vif(mrt, dev);
2219 		if (vif >= 0) {
2220 			int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2221 			read_unlock(&mrt_lock);
2222 
2223 			return err2;
2224 		}
2225 		read_unlock(&mrt_lock);
2226 		kfree_skb(skb);
2227 		return -ENODEV;
2228 	}
2229 
2230 	read_lock(&mrt_lock);
2231 	ip_mr_forward(net, mrt, dev, skb, cache, local);
2232 	read_unlock(&mrt_lock);
2233 
2234 	if (local)
2235 		return ip_local_deliver(skb);
2236 
2237 	return 0;
2238 
2239 dont_forward:
2240 	if (local)
2241 		return ip_local_deliver(skb);
2242 	kfree_skb(skb);
2243 	return 0;
2244 }
2245 
2246 #ifdef CONFIG_IP_PIMSM_V1
2247 /* Handle IGMP messages of PIMv1 */
2248 int pim_rcv_v1(struct sk_buff *skb)
2249 {
2250 	struct igmphdr *pim;
2251 	struct net *net = dev_net(skb->dev);
2252 	struct mr_table *mrt;
2253 
2254 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2255 		goto drop;
2256 
2257 	pim = igmp_hdr(skb);
2258 
2259 	mrt = ipmr_rt_fib_lookup(net, skb);
2260 	if (IS_ERR(mrt))
2261 		goto drop;
2262 	if (!mrt->mroute_do_pim ||
2263 	    pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2264 		goto drop;
2265 
2266 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2267 drop:
2268 		kfree_skb(skb);
2269 	}
2270 	return 0;
2271 }
2272 #endif
2273 
2274 #ifdef CONFIG_IP_PIMSM_V2
2275 static int pim_rcv(struct sk_buff *skb)
2276 {
2277 	struct pimreghdr *pim;
2278 	struct net *net = dev_net(skb->dev);
2279 	struct mr_table *mrt;
2280 
2281 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2282 		goto drop;
2283 
2284 	pim = (struct pimreghdr *)skb_transport_header(skb);
2285 	if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2286 	    (pim->flags & PIM_NULL_REGISTER) ||
2287 	    (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2288 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2289 		goto drop;
2290 
2291 	mrt = ipmr_rt_fib_lookup(net, skb);
2292 	if (IS_ERR(mrt))
2293 		goto drop;
2294 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2295 drop:
2296 		kfree_skb(skb);
2297 	}
2298 	return 0;
2299 }
2300 #endif
2301 
2302 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2303 			      struct mfc_cache *c, struct rtmsg *rtm)
2304 {
2305 	struct rta_mfc_stats mfcs;
2306 	struct nlattr *mp_attr;
2307 	struct rtnexthop *nhp;
2308 	unsigned long lastuse;
2309 	int ct;
2310 
2311 	/* If cache is unresolved, don't try to parse IIF and OIF */
2312 	if (c->mfc_parent >= MAXVIFS) {
2313 		rtm->rtm_flags |= RTNH_F_UNRESOLVED;
2314 		return -ENOENT;
2315 	}
2316 
2317 	if (VIF_EXISTS(mrt, c->mfc_parent) &&
2318 	    nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2319 		return -EMSGSIZE;
2320 
2321 	if (c->mfc_flags & MFC_OFFLOAD)
2322 		rtm->rtm_flags |= RTNH_F_OFFLOAD;
2323 
2324 	if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2325 		return -EMSGSIZE;
2326 
2327 	for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2328 		if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2329 			if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2330 				nla_nest_cancel(skb, mp_attr);
2331 				return -EMSGSIZE;
2332 			}
2333 
2334 			nhp->rtnh_flags = 0;
2335 			nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2336 			nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2337 			nhp->rtnh_len = sizeof(*nhp);
2338 		}
2339 	}
2340 
2341 	nla_nest_end(skb, mp_attr);
2342 
2343 	lastuse = READ_ONCE(c->mfc_un.res.lastuse);
2344 	lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
2345 
2346 	mfcs.mfcs_packets = c->mfc_un.res.pkt;
2347 	mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2348 	mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2349 	if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
2350 	    nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
2351 			      RTA_PAD))
2352 		return -EMSGSIZE;
2353 
2354 	rtm->rtm_type = RTN_MULTICAST;
2355 	return 1;
2356 }
2357 
2358 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2359 		   __be32 saddr, __be32 daddr,
2360 		   struct rtmsg *rtm, u32 portid)
2361 {
2362 	struct mfc_cache *cache;
2363 	struct mr_table *mrt;
2364 	int err;
2365 
2366 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2367 	if (!mrt)
2368 		return -ENOENT;
2369 
2370 	rcu_read_lock();
2371 	cache = ipmr_cache_find(mrt, saddr, daddr);
2372 	if (!cache && skb->dev) {
2373 		int vif = ipmr_find_vif(mrt, skb->dev);
2374 
2375 		if (vif >= 0)
2376 			cache = ipmr_cache_find_any(mrt, daddr, vif);
2377 	}
2378 	if (!cache) {
2379 		struct sk_buff *skb2;
2380 		struct iphdr *iph;
2381 		struct net_device *dev;
2382 		int vif = -1;
2383 
2384 		dev = skb->dev;
2385 		read_lock(&mrt_lock);
2386 		if (dev)
2387 			vif = ipmr_find_vif(mrt, dev);
2388 		if (vif < 0) {
2389 			read_unlock(&mrt_lock);
2390 			rcu_read_unlock();
2391 			return -ENODEV;
2392 		}
2393 		skb2 = skb_clone(skb, GFP_ATOMIC);
2394 		if (!skb2) {
2395 			read_unlock(&mrt_lock);
2396 			rcu_read_unlock();
2397 			return -ENOMEM;
2398 		}
2399 
2400 		NETLINK_CB(skb2).portid = portid;
2401 		skb_push(skb2, sizeof(struct iphdr));
2402 		skb_reset_network_header(skb2);
2403 		iph = ip_hdr(skb2);
2404 		iph->ihl = sizeof(struct iphdr) >> 2;
2405 		iph->saddr = saddr;
2406 		iph->daddr = daddr;
2407 		iph->version = 0;
2408 		err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2409 		read_unlock(&mrt_lock);
2410 		rcu_read_unlock();
2411 		return err;
2412 	}
2413 
2414 	read_lock(&mrt_lock);
2415 	err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2416 	read_unlock(&mrt_lock);
2417 	rcu_read_unlock();
2418 	return err;
2419 }
2420 
2421 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2422 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2423 			    int flags)
2424 {
2425 	struct nlmsghdr *nlh;
2426 	struct rtmsg *rtm;
2427 	int err;
2428 
2429 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2430 	if (!nlh)
2431 		return -EMSGSIZE;
2432 
2433 	rtm = nlmsg_data(nlh);
2434 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2435 	rtm->rtm_dst_len  = 32;
2436 	rtm->rtm_src_len  = 32;
2437 	rtm->rtm_tos      = 0;
2438 	rtm->rtm_table    = mrt->id;
2439 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2440 		goto nla_put_failure;
2441 	rtm->rtm_type     = RTN_MULTICAST;
2442 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2443 	if (c->mfc_flags & MFC_STATIC)
2444 		rtm->rtm_protocol = RTPROT_STATIC;
2445 	else
2446 		rtm->rtm_protocol = RTPROT_MROUTED;
2447 	rtm->rtm_flags    = 0;
2448 
2449 	if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2450 	    nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2451 		goto nla_put_failure;
2452 	err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2453 	/* do not break the dump if cache is unresolved */
2454 	if (err < 0 && err != -ENOENT)
2455 		goto nla_put_failure;
2456 
2457 	nlmsg_end(skb, nlh);
2458 	return 0;
2459 
2460 nla_put_failure:
2461 	nlmsg_cancel(skb, nlh);
2462 	return -EMSGSIZE;
2463 }
2464 
2465 static size_t mroute_msgsize(bool unresolved, int maxvif)
2466 {
2467 	size_t len =
2468 		NLMSG_ALIGN(sizeof(struct rtmsg))
2469 		+ nla_total_size(4)	/* RTA_TABLE */
2470 		+ nla_total_size(4)	/* RTA_SRC */
2471 		+ nla_total_size(4)	/* RTA_DST */
2472 		;
2473 
2474 	if (!unresolved)
2475 		len = len
2476 		      + nla_total_size(4)	/* RTA_IIF */
2477 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2478 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2479 						/* RTA_MFC_STATS */
2480 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2481 		;
2482 
2483 	return len;
2484 }
2485 
2486 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2487 				 int cmd)
2488 {
2489 	struct net *net = read_pnet(&mrt->net);
2490 	struct sk_buff *skb;
2491 	int err = -ENOBUFS;
2492 
2493 	skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2494 			GFP_ATOMIC);
2495 	if (!skb)
2496 		goto errout;
2497 
2498 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2499 	if (err < 0)
2500 		goto errout;
2501 
2502 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2503 	return;
2504 
2505 errout:
2506 	kfree_skb(skb);
2507 	if (err < 0)
2508 		rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2509 }
2510 
2511 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2512 {
2513 	size_t len =
2514 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2515 		+ nla_total_size(1)	/* IPMRA_CREPORT_MSGTYPE */
2516 		+ nla_total_size(4)	/* IPMRA_CREPORT_VIF_ID */
2517 		+ nla_total_size(4)	/* IPMRA_CREPORT_SRC_ADDR */
2518 		+ nla_total_size(4)	/* IPMRA_CREPORT_DST_ADDR */
2519 					/* IPMRA_CREPORT_PKT */
2520 		+ nla_total_size(payloadlen)
2521 		;
2522 
2523 	return len;
2524 }
2525 
2526 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2527 {
2528 	struct net *net = read_pnet(&mrt->net);
2529 	struct nlmsghdr *nlh;
2530 	struct rtgenmsg *rtgenm;
2531 	struct igmpmsg *msg;
2532 	struct sk_buff *skb;
2533 	struct nlattr *nla;
2534 	int payloadlen;
2535 
2536 	payloadlen = pkt->len - sizeof(struct igmpmsg);
2537 	msg = (struct igmpmsg *)skb_network_header(pkt);
2538 
2539 	skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2540 	if (!skb)
2541 		goto errout;
2542 
2543 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2544 			sizeof(struct rtgenmsg), 0);
2545 	if (!nlh)
2546 		goto errout;
2547 	rtgenm = nlmsg_data(nlh);
2548 	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2549 	if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2550 	    nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2551 	    nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2552 			    msg->im_src.s_addr) ||
2553 	    nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2554 			    msg->im_dst.s_addr))
2555 		goto nla_put_failure;
2556 
2557 	nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2558 	if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2559 				  nla_data(nla), payloadlen))
2560 		goto nla_put_failure;
2561 
2562 	nlmsg_end(skb, nlh);
2563 
2564 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2565 	return;
2566 
2567 nla_put_failure:
2568 	nlmsg_cancel(skb, nlh);
2569 errout:
2570 	kfree_skb(skb);
2571 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2572 }
2573 
2574 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2575 			     struct netlink_ext_ack *extack)
2576 {
2577 	struct net *net = sock_net(in_skb->sk);
2578 	struct nlattr *tb[RTA_MAX + 1];
2579 	struct sk_buff *skb = NULL;
2580 	struct mfc_cache *cache;
2581 	struct mr_table *mrt;
2582 	struct rtmsg *rtm;
2583 	__be32 src, grp;
2584 	u32 tableid;
2585 	int err;
2586 
2587 	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX,
2588 			  rtm_ipv4_policy, extack);
2589 	if (err < 0)
2590 		goto errout;
2591 
2592 	rtm = nlmsg_data(nlh);
2593 
2594 	src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2595 	grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2596 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2597 
2598 	mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2599 	if (!mrt) {
2600 		err = -ENOENT;
2601 		goto errout_free;
2602 	}
2603 
2604 	/* entries are added/deleted only under RTNL */
2605 	rcu_read_lock();
2606 	cache = ipmr_cache_find(mrt, src, grp);
2607 	rcu_read_unlock();
2608 	if (!cache) {
2609 		err = -ENOENT;
2610 		goto errout_free;
2611 	}
2612 
2613 	skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2614 	if (!skb) {
2615 		err = -ENOBUFS;
2616 		goto errout_free;
2617 	}
2618 
2619 	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2620 			       nlh->nlmsg_seq, cache,
2621 			       RTM_NEWROUTE, 0);
2622 	if (err < 0)
2623 		goto errout_free;
2624 
2625 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2626 
2627 errout:
2628 	return err;
2629 
2630 errout_free:
2631 	kfree_skb(skb);
2632 	goto errout;
2633 }
2634 
2635 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2636 {
2637 	struct net *net = sock_net(skb->sk);
2638 	struct mr_table *mrt;
2639 	struct mfc_cache *mfc;
2640 	unsigned int t = 0, s_t;
2641 	unsigned int e = 0, s_e;
2642 
2643 	s_t = cb->args[0];
2644 	s_e = cb->args[1];
2645 
2646 	rcu_read_lock();
2647 	ipmr_for_each_table(mrt, net) {
2648 		if (t < s_t)
2649 			goto next_table;
2650 		list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
2651 			if (e < s_e)
2652 				goto next_entry;
2653 			if (ipmr_fill_mroute(mrt, skb,
2654 					     NETLINK_CB(cb->skb).portid,
2655 					     cb->nlh->nlmsg_seq,
2656 					     mfc, RTM_NEWROUTE,
2657 					     NLM_F_MULTI) < 0)
2658 				goto done;
2659 next_entry:
2660 			e++;
2661 		}
2662 		e = 0;
2663 		s_e = 0;
2664 
2665 		spin_lock_bh(&mfc_unres_lock);
2666 		list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2667 			if (e < s_e)
2668 				goto next_entry2;
2669 			if (ipmr_fill_mroute(mrt, skb,
2670 					     NETLINK_CB(cb->skb).portid,
2671 					     cb->nlh->nlmsg_seq,
2672 					     mfc, RTM_NEWROUTE,
2673 					     NLM_F_MULTI) < 0) {
2674 				spin_unlock_bh(&mfc_unres_lock);
2675 				goto done;
2676 			}
2677 next_entry2:
2678 			e++;
2679 		}
2680 		spin_unlock_bh(&mfc_unres_lock);
2681 		e = 0;
2682 		s_e = 0;
2683 next_table:
2684 		t++;
2685 	}
2686 done:
2687 	rcu_read_unlock();
2688 
2689 	cb->args[1] = e;
2690 	cb->args[0] = t;
2691 
2692 	return skb->len;
2693 }
2694 
2695 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2696 	[RTA_SRC]	= { .type = NLA_U32 },
2697 	[RTA_DST]	= { .type = NLA_U32 },
2698 	[RTA_IIF]	= { .type = NLA_U32 },
2699 	[RTA_TABLE]	= { .type = NLA_U32 },
2700 	[RTA_MULTIPATH]	= { .len = sizeof(struct rtnexthop) },
2701 };
2702 
2703 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2704 {
2705 	switch (rtm_protocol) {
2706 	case RTPROT_STATIC:
2707 	case RTPROT_MROUTED:
2708 		return true;
2709 	}
2710 	return false;
2711 }
2712 
2713 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2714 {
2715 	struct rtnexthop *rtnh = nla_data(nla);
2716 	int remaining = nla_len(nla), vifi = 0;
2717 
2718 	while (rtnh_ok(rtnh, remaining)) {
2719 		mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2720 		if (++vifi == MAXVIFS)
2721 			break;
2722 		rtnh = rtnh_next(rtnh, &remaining);
2723 	}
2724 
2725 	return remaining > 0 ? -EINVAL : vifi;
2726 }
2727 
2728 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2729 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2730 			    struct mfcctl *mfcc, int *mrtsock,
2731 			    struct mr_table **mrtret,
2732 			    struct netlink_ext_ack *extack)
2733 {
2734 	struct net_device *dev = NULL;
2735 	u32 tblid = RT_TABLE_DEFAULT;
2736 	struct mr_table *mrt;
2737 	struct nlattr *attr;
2738 	struct rtmsg *rtm;
2739 	int ret, rem;
2740 
2741 	ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy,
2742 			     extack);
2743 	if (ret < 0)
2744 		goto out;
2745 	rtm = nlmsg_data(nlh);
2746 
2747 	ret = -EINVAL;
2748 	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2749 	    rtm->rtm_type != RTN_MULTICAST ||
2750 	    rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2751 	    !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2752 		goto out;
2753 
2754 	memset(mfcc, 0, sizeof(*mfcc));
2755 	mfcc->mfcc_parent = -1;
2756 	ret = 0;
2757 	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2758 		switch (nla_type(attr)) {
2759 		case RTA_SRC:
2760 			mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2761 			break;
2762 		case RTA_DST:
2763 			mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2764 			break;
2765 		case RTA_IIF:
2766 			dev = __dev_get_by_index(net, nla_get_u32(attr));
2767 			if (!dev) {
2768 				ret = -ENODEV;
2769 				goto out;
2770 			}
2771 			break;
2772 		case RTA_MULTIPATH:
2773 			if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2774 				ret = -EINVAL;
2775 				goto out;
2776 			}
2777 			break;
2778 		case RTA_PREFSRC:
2779 			ret = 1;
2780 			break;
2781 		case RTA_TABLE:
2782 			tblid = nla_get_u32(attr);
2783 			break;
2784 		}
2785 	}
2786 	mrt = ipmr_get_table(net, tblid);
2787 	if (!mrt) {
2788 		ret = -ENOENT;
2789 		goto out;
2790 	}
2791 	*mrtret = mrt;
2792 	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2793 	if (dev)
2794 		mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2795 
2796 out:
2797 	return ret;
2798 }
2799 
2800 /* takes care of both newroute and delroute */
2801 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2802 			  struct netlink_ext_ack *extack)
2803 {
2804 	struct net *net = sock_net(skb->sk);
2805 	int ret, mrtsock, parent;
2806 	struct mr_table *tbl;
2807 	struct mfcctl mfcc;
2808 
2809 	mrtsock = 0;
2810 	tbl = NULL;
2811 	ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2812 	if (ret < 0)
2813 		return ret;
2814 
2815 	parent = ret ? mfcc.mfcc_parent : -1;
2816 	if (nlh->nlmsg_type == RTM_NEWROUTE)
2817 		return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2818 	else
2819 		return ipmr_mfc_delete(tbl, &mfcc, parent);
2820 }
2821 
2822 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2823 {
2824 	u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2825 
2826 	if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2827 	    nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2828 	    nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2829 			mrt->mroute_reg_vif_num) ||
2830 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2831 		       mrt->mroute_do_assert) ||
2832 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim))
2833 		return false;
2834 
2835 	return true;
2836 }
2837 
2838 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2839 {
2840 	struct nlattr *vif_nest;
2841 	struct vif_device *vif;
2842 
2843 	/* if the VIF doesn't exist just continue */
2844 	if (!VIF_EXISTS(mrt, vifid))
2845 		return true;
2846 
2847 	vif = &mrt->vif_table[vifid];
2848 	vif_nest = nla_nest_start(skb, IPMRA_VIF);
2849 	if (!vif_nest)
2850 		return false;
2851 	if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2852 	    nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2853 	    nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2854 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2855 			      IPMRA_VIFA_PAD) ||
2856 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2857 			      IPMRA_VIFA_PAD) ||
2858 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2859 			      IPMRA_VIFA_PAD) ||
2860 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2861 			      IPMRA_VIFA_PAD) ||
2862 	    nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2863 	    nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2864 		nla_nest_cancel(skb, vif_nest);
2865 		return false;
2866 	}
2867 	nla_nest_end(skb, vif_nest);
2868 
2869 	return true;
2870 }
2871 
2872 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2873 {
2874 	struct net *net = sock_net(skb->sk);
2875 	struct nlmsghdr *nlh = NULL;
2876 	unsigned int t = 0, s_t;
2877 	unsigned int e = 0, s_e;
2878 	struct mr_table *mrt;
2879 
2880 	s_t = cb->args[0];
2881 	s_e = cb->args[1];
2882 
2883 	ipmr_for_each_table(mrt, net) {
2884 		struct nlattr *vifs, *af;
2885 		struct ifinfomsg *hdr;
2886 		u32 i;
2887 
2888 		if (t < s_t)
2889 			goto skip_table;
2890 		nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2891 				cb->nlh->nlmsg_seq, RTM_NEWLINK,
2892 				sizeof(*hdr), NLM_F_MULTI);
2893 		if (!nlh)
2894 			break;
2895 
2896 		hdr = nlmsg_data(nlh);
2897 		memset(hdr, 0, sizeof(*hdr));
2898 		hdr->ifi_family = RTNL_FAMILY_IPMR;
2899 
2900 		af = nla_nest_start(skb, IFLA_AF_SPEC);
2901 		if (!af) {
2902 			nlmsg_cancel(skb, nlh);
2903 			goto out;
2904 		}
2905 
2906 		if (!ipmr_fill_table(mrt, skb)) {
2907 			nlmsg_cancel(skb, nlh);
2908 			goto out;
2909 		}
2910 
2911 		vifs = nla_nest_start(skb, IPMRA_TABLE_VIFS);
2912 		if (!vifs) {
2913 			nla_nest_end(skb, af);
2914 			nlmsg_end(skb, nlh);
2915 			goto out;
2916 		}
2917 		for (i = 0; i < mrt->maxvif; i++) {
2918 			if (e < s_e)
2919 				goto skip_entry;
2920 			if (!ipmr_fill_vif(mrt, i, skb)) {
2921 				nla_nest_end(skb, vifs);
2922 				nla_nest_end(skb, af);
2923 				nlmsg_end(skb, nlh);
2924 				goto out;
2925 			}
2926 skip_entry:
2927 			e++;
2928 		}
2929 		s_e = 0;
2930 		e = 0;
2931 		nla_nest_end(skb, vifs);
2932 		nla_nest_end(skb, af);
2933 		nlmsg_end(skb, nlh);
2934 skip_table:
2935 		t++;
2936 	}
2937 
2938 out:
2939 	cb->args[1] = e;
2940 	cb->args[0] = t;
2941 
2942 	return skb->len;
2943 }
2944 
2945 #ifdef CONFIG_PROC_FS
2946 /* The /proc interfaces to multicast routing :
2947  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2948  */
2949 struct ipmr_vif_iter {
2950 	struct seq_net_private p;
2951 	struct mr_table *mrt;
2952 	int ct;
2953 };
2954 
2955 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2956 					   struct ipmr_vif_iter *iter,
2957 					   loff_t pos)
2958 {
2959 	struct mr_table *mrt = iter->mrt;
2960 
2961 	for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2962 		if (!VIF_EXISTS(mrt, iter->ct))
2963 			continue;
2964 		if (pos-- == 0)
2965 			return &mrt->vif_table[iter->ct];
2966 	}
2967 	return NULL;
2968 }
2969 
2970 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2971 	__acquires(mrt_lock)
2972 {
2973 	struct ipmr_vif_iter *iter = seq->private;
2974 	struct net *net = seq_file_net(seq);
2975 	struct mr_table *mrt;
2976 
2977 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2978 	if (!mrt)
2979 		return ERR_PTR(-ENOENT);
2980 
2981 	iter->mrt = mrt;
2982 
2983 	read_lock(&mrt_lock);
2984 	return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2985 		: SEQ_START_TOKEN;
2986 }
2987 
2988 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2989 {
2990 	struct ipmr_vif_iter *iter = seq->private;
2991 	struct net *net = seq_file_net(seq);
2992 	struct mr_table *mrt = iter->mrt;
2993 
2994 	++*pos;
2995 	if (v == SEQ_START_TOKEN)
2996 		return ipmr_vif_seq_idx(net, iter, 0);
2997 
2998 	while (++iter->ct < mrt->maxvif) {
2999 		if (!VIF_EXISTS(mrt, iter->ct))
3000 			continue;
3001 		return &mrt->vif_table[iter->ct];
3002 	}
3003 	return NULL;
3004 }
3005 
3006 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
3007 	__releases(mrt_lock)
3008 {
3009 	read_unlock(&mrt_lock);
3010 }
3011 
3012 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
3013 {
3014 	struct ipmr_vif_iter *iter = seq->private;
3015 	struct mr_table *mrt = iter->mrt;
3016 
3017 	if (v == SEQ_START_TOKEN) {
3018 		seq_puts(seq,
3019 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
3020 	} else {
3021 		const struct vif_device *vif = v;
3022 		const char *name =  vif->dev ? vif->dev->name : "none";
3023 
3024 		seq_printf(seq,
3025 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
3026 			   vif - mrt->vif_table,
3027 			   name, vif->bytes_in, vif->pkt_in,
3028 			   vif->bytes_out, vif->pkt_out,
3029 			   vif->flags, vif->local, vif->remote);
3030 	}
3031 	return 0;
3032 }
3033 
3034 static const struct seq_operations ipmr_vif_seq_ops = {
3035 	.start = ipmr_vif_seq_start,
3036 	.next  = ipmr_vif_seq_next,
3037 	.stop  = ipmr_vif_seq_stop,
3038 	.show  = ipmr_vif_seq_show,
3039 };
3040 
3041 static int ipmr_vif_open(struct inode *inode, struct file *file)
3042 {
3043 	return seq_open_net(inode, file, &ipmr_vif_seq_ops,
3044 			    sizeof(struct ipmr_vif_iter));
3045 }
3046 
3047 static const struct file_operations ipmr_vif_fops = {
3048 	.open    = ipmr_vif_open,
3049 	.read    = seq_read,
3050 	.llseek  = seq_lseek,
3051 	.release = seq_release_net,
3052 };
3053 
3054 struct ipmr_mfc_iter {
3055 	struct seq_net_private p;
3056 	struct mr_table *mrt;
3057 	struct list_head *cache;
3058 };
3059 
3060 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
3061 					  struct ipmr_mfc_iter *it, loff_t pos)
3062 {
3063 	struct mr_table *mrt = it->mrt;
3064 	struct mfc_cache *mfc;
3065 
3066 	rcu_read_lock();
3067 	it->cache = &mrt->mfc_cache_list;
3068 	list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
3069 		if (pos-- == 0)
3070 			return mfc;
3071 	rcu_read_unlock();
3072 
3073 	spin_lock_bh(&mfc_unres_lock);
3074 	it->cache = &mrt->mfc_unres_queue;
3075 	list_for_each_entry(mfc, it->cache, list)
3076 		if (pos-- == 0)
3077 			return mfc;
3078 	spin_unlock_bh(&mfc_unres_lock);
3079 
3080 	it->cache = NULL;
3081 	return NULL;
3082 }
3083 
3084 
3085 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
3086 {
3087 	struct ipmr_mfc_iter *it = seq->private;
3088 	struct net *net = seq_file_net(seq);
3089 	struct mr_table *mrt;
3090 
3091 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
3092 	if (!mrt)
3093 		return ERR_PTR(-ENOENT);
3094 
3095 	it->mrt = mrt;
3096 	it->cache = NULL;
3097 	return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
3098 		: SEQ_START_TOKEN;
3099 }
3100 
3101 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3102 {
3103 	struct ipmr_mfc_iter *it = seq->private;
3104 	struct net *net = seq_file_net(seq);
3105 	struct mr_table *mrt = it->mrt;
3106 	struct mfc_cache *mfc = v;
3107 
3108 	++*pos;
3109 
3110 	if (v == SEQ_START_TOKEN)
3111 		return ipmr_mfc_seq_idx(net, seq->private, 0);
3112 
3113 	if (mfc->list.next != it->cache)
3114 		return list_entry(mfc->list.next, struct mfc_cache, list);
3115 
3116 	if (it->cache == &mrt->mfc_unres_queue)
3117 		goto end_of_list;
3118 
3119 	/* exhausted cache_array, show unresolved */
3120 	rcu_read_unlock();
3121 	it->cache = &mrt->mfc_unres_queue;
3122 
3123 	spin_lock_bh(&mfc_unres_lock);
3124 	if (!list_empty(it->cache))
3125 		return list_first_entry(it->cache, struct mfc_cache, list);
3126 
3127 end_of_list:
3128 	spin_unlock_bh(&mfc_unres_lock);
3129 	it->cache = NULL;
3130 
3131 	return NULL;
3132 }
3133 
3134 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
3135 {
3136 	struct ipmr_mfc_iter *it = seq->private;
3137 	struct mr_table *mrt = it->mrt;
3138 
3139 	if (it->cache == &mrt->mfc_unres_queue)
3140 		spin_unlock_bh(&mfc_unres_lock);
3141 	else if (it->cache == &mrt->mfc_cache_list)
3142 		rcu_read_unlock();
3143 }
3144 
3145 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
3146 {
3147 	int n;
3148 
3149 	if (v == SEQ_START_TOKEN) {
3150 		seq_puts(seq,
3151 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
3152 	} else {
3153 		const struct mfc_cache *mfc = v;
3154 		const struct ipmr_mfc_iter *it = seq->private;
3155 		const struct mr_table *mrt = it->mrt;
3156 
3157 		seq_printf(seq, "%08X %08X %-3hd",
3158 			   (__force u32) mfc->mfc_mcastgrp,
3159 			   (__force u32) mfc->mfc_origin,
3160 			   mfc->mfc_parent);
3161 
3162 		if (it->cache != &mrt->mfc_unres_queue) {
3163 			seq_printf(seq, " %8lu %8lu %8lu",
3164 				   mfc->mfc_un.res.pkt,
3165 				   mfc->mfc_un.res.bytes,
3166 				   mfc->mfc_un.res.wrong_if);
3167 			for (n = mfc->mfc_un.res.minvif;
3168 			     n < mfc->mfc_un.res.maxvif; n++) {
3169 				if (VIF_EXISTS(mrt, n) &&
3170 				    mfc->mfc_un.res.ttls[n] < 255)
3171 					seq_printf(seq,
3172 					   " %2d:%-3d",
3173 					   n, mfc->mfc_un.res.ttls[n]);
3174 			}
3175 		} else {
3176 			/* unresolved mfc_caches don't contain
3177 			 * pkt, bytes and wrong_if values
3178 			 */
3179 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3180 		}
3181 		seq_putc(seq, '\n');
3182 	}
3183 	return 0;
3184 }
3185 
3186 static const struct seq_operations ipmr_mfc_seq_ops = {
3187 	.start = ipmr_mfc_seq_start,
3188 	.next  = ipmr_mfc_seq_next,
3189 	.stop  = ipmr_mfc_seq_stop,
3190 	.show  = ipmr_mfc_seq_show,
3191 };
3192 
3193 static int ipmr_mfc_open(struct inode *inode, struct file *file)
3194 {
3195 	return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
3196 			    sizeof(struct ipmr_mfc_iter));
3197 }
3198 
3199 static const struct file_operations ipmr_mfc_fops = {
3200 	.open    = ipmr_mfc_open,
3201 	.read    = seq_read,
3202 	.llseek  = seq_lseek,
3203 	.release = seq_release_net,
3204 };
3205 #endif
3206 
3207 #ifdef CONFIG_IP_PIMSM_V2
3208 static const struct net_protocol pim_protocol = {
3209 	.handler	=	pim_rcv,
3210 	.netns_ok	=	1,
3211 };
3212 #endif
3213 
3214 static unsigned int ipmr_seq_read(struct net *net)
3215 {
3216 	ASSERT_RTNL();
3217 
3218 	return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3219 }
3220 
3221 static int ipmr_dump(struct net *net, struct notifier_block *nb)
3222 {
3223 	struct mr_table *mrt;
3224 	int err;
3225 
3226 	err = ipmr_rules_dump(net, nb);
3227 	if (err)
3228 		return err;
3229 
3230 	ipmr_for_each_table(mrt, net) {
3231 		struct vif_device *v = &mrt->vif_table[0];
3232 		struct mfc_cache *mfc;
3233 		int vifi;
3234 
3235 		/* Notifiy on table VIF entries */
3236 		read_lock(&mrt_lock);
3237 		for (vifi = 0; vifi < mrt->maxvif; vifi++, v++) {
3238 			if (!v->dev)
3239 				continue;
3240 
3241 			call_ipmr_vif_entry_notifier(nb, net, FIB_EVENT_VIF_ADD,
3242 						     v, vifi, mrt->id);
3243 		}
3244 		read_unlock(&mrt_lock);
3245 
3246 		/* Notify on table MFC entries */
3247 		list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
3248 			call_ipmr_mfc_entry_notifier(nb, net,
3249 						     FIB_EVENT_ENTRY_ADD, mfc,
3250 						     mrt->id);
3251 	}
3252 
3253 	return 0;
3254 }
3255 
3256 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3257 	.family		= RTNL_FAMILY_IPMR,
3258 	.fib_seq_read	= ipmr_seq_read,
3259 	.fib_dump	= ipmr_dump,
3260 	.owner		= THIS_MODULE,
3261 };
3262 
3263 static int __net_init ipmr_notifier_init(struct net *net)
3264 {
3265 	struct fib_notifier_ops *ops;
3266 
3267 	net->ipv4.ipmr_seq = 0;
3268 
3269 	ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3270 	if (IS_ERR(ops))
3271 		return PTR_ERR(ops);
3272 	net->ipv4.ipmr_notifier_ops = ops;
3273 
3274 	return 0;
3275 }
3276 
3277 static void __net_exit ipmr_notifier_exit(struct net *net)
3278 {
3279 	fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3280 	net->ipv4.ipmr_notifier_ops = NULL;
3281 }
3282 
3283 /* Setup for IP multicast routing */
3284 static int __net_init ipmr_net_init(struct net *net)
3285 {
3286 	int err;
3287 
3288 	err = ipmr_notifier_init(net);
3289 	if (err)
3290 		goto ipmr_notifier_fail;
3291 
3292 	err = ipmr_rules_init(net);
3293 	if (err < 0)
3294 		goto ipmr_rules_fail;
3295 
3296 #ifdef CONFIG_PROC_FS
3297 	err = -ENOMEM;
3298 	if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
3299 		goto proc_vif_fail;
3300 	if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
3301 		goto proc_cache_fail;
3302 #endif
3303 	return 0;
3304 
3305 #ifdef CONFIG_PROC_FS
3306 proc_cache_fail:
3307 	remove_proc_entry("ip_mr_vif", net->proc_net);
3308 proc_vif_fail:
3309 	ipmr_rules_exit(net);
3310 #endif
3311 ipmr_rules_fail:
3312 	ipmr_notifier_exit(net);
3313 ipmr_notifier_fail:
3314 	return err;
3315 }
3316 
3317 static void __net_exit ipmr_net_exit(struct net *net)
3318 {
3319 #ifdef CONFIG_PROC_FS
3320 	remove_proc_entry("ip_mr_cache", net->proc_net);
3321 	remove_proc_entry("ip_mr_vif", net->proc_net);
3322 #endif
3323 	ipmr_notifier_exit(net);
3324 	ipmr_rules_exit(net);
3325 }
3326 
3327 static struct pernet_operations ipmr_net_ops = {
3328 	.init = ipmr_net_init,
3329 	.exit = ipmr_net_exit,
3330 };
3331 
3332 int __init ip_mr_init(void)
3333 {
3334 	int err;
3335 
3336 	mrt_cachep = kmem_cache_create("ip_mrt_cache",
3337 				       sizeof(struct mfc_cache),
3338 				       0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3339 				       NULL);
3340 
3341 	err = register_pernet_subsys(&ipmr_net_ops);
3342 	if (err)
3343 		goto reg_pernet_fail;
3344 
3345 	err = register_netdevice_notifier(&ip_mr_notifier);
3346 	if (err)
3347 		goto reg_notif_fail;
3348 #ifdef CONFIG_IP_PIMSM_V2
3349 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3350 		pr_err("%s: can't add PIM protocol\n", __func__);
3351 		err = -EAGAIN;
3352 		goto add_proto_fail;
3353 	}
3354 #endif
3355 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3356 		      ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3357 	rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3358 		      ipmr_rtm_route, NULL, 0);
3359 	rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3360 		      ipmr_rtm_route, NULL, 0);
3361 
3362 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3363 		      NULL, ipmr_rtm_dumplink, 0);
3364 	return 0;
3365 
3366 #ifdef CONFIG_IP_PIMSM_V2
3367 add_proto_fail:
3368 	unregister_netdevice_notifier(&ip_mr_notifier);
3369 #endif
3370 reg_notif_fail:
3371 	unregister_pernet_subsys(&ipmr_net_ops);
3372 reg_pernet_fail:
3373 	kmem_cache_destroy(mrt_cachep);
3374 	return err;
3375 }
3376