xref: /openbmc/linux/net/ipv4/ipmr.c (revision 59bd9ded)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IP multicast routing support for mrouted 3.6/3.8
4  *
5  *		(c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6  *	  Linux Consultancy and Custom Driver Development
7  *
8  *	Fixes:
9  *	Michael Chastain	:	Incorrect size of copying.
10  *	Alan Cox		:	Added the cache manager code
11  *	Alan Cox		:	Fixed the clone/copy bug and device race.
12  *	Mike McLagan		:	Routing by source
13  *	Malcolm Beattie		:	Buffer handling fixes.
14  *	Alexey Kuznetsov	:	Double buffer free and other fixes.
15  *	SVR Anand		:	Fixed several multicast bugs and problems.
16  *	Alexey Kuznetsov	:	Status, optimisations and more.
17  *	Brad Parker		:	Better behaviour on mrouted upcall
18  *					overflow.
19  *      Carlos Picoto           :       PIMv1 Support
20  *	Pavlin Ivanov Radoslavov:	PIMv2 Registers must checksum only PIM header
21  *					Relax this requirement to work with older peers.
22  */
23 
24 #include <linux/uaccess.h>
25 #include <linux/types.h>
26 #include <linux/cache.h>
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/mm.h>
30 #include <linux/kernel.h>
31 #include <linux/fcntl.h>
32 #include <linux/stat.h>
33 #include <linux/socket.h>
34 #include <linux/in.h>
35 #include <linux/inet.h>
36 #include <linux/netdevice.h>
37 #include <linux/inetdevice.h>
38 #include <linux/igmp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/mroute.h>
42 #include <linux/init.h>
43 #include <linux/if_ether.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <linux/skbuff.h>
49 #include <net/route.h>
50 #include <net/icmp.h>
51 #include <net/udp.h>
52 #include <net/raw.h>
53 #include <linux/notifier.h>
54 #include <linux/if_arp.h>
55 #include <linux/netfilter_ipv4.h>
56 #include <linux/compat.h>
57 #include <linux/export.h>
58 #include <linux/rhashtable.h>
59 #include <net/ip_tunnels.h>
60 #include <net/checksum.h>
61 #include <net/netlink.h>
62 #include <net/fib_rules.h>
63 #include <linux/netconf.h>
64 #include <net/rtnh.h>
65 
66 #include <linux/nospec.h>
67 
68 struct ipmr_rule {
69 	struct fib_rule		common;
70 };
71 
72 struct ipmr_result {
73 	struct mr_table		*mrt;
74 };
75 
76 /* Big lock, protecting vif table, mrt cache and mroute socket state.
77  * Note that the changes are semaphored via rtnl_lock.
78  */
79 
80 static DEFINE_RWLOCK(mrt_lock);
81 
82 /* Multicast router control variables */
83 
84 /* Special spinlock for queue of unresolved entries */
85 static DEFINE_SPINLOCK(mfc_unres_lock);
86 
87 /* We return to original Alan's scheme. Hash table of resolved
88  * entries is changed only in process context and protected
89  * with weak lock mrt_lock. Queue of unresolved entries is protected
90  * with strong spinlock mfc_unres_lock.
91  *
92  * In this case data path is free of exclusive locks at all.
93  */
94 
95 static struct kmem_cache *mrt_cachep __ro_after_init;
96 
97 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
98 static void ipmr_free_table(struct mr_table *mrt);
99 
100 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
101 			  struct net_device *dev, struct sk_buff *skb,
102 			  struct mfc_cache *cache, int local);
103 static int ipmr_cache_report(struct mr_table *mrt,
104 			     struct sk_buff *pkt, vifi_t vifi, int assert);
105 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
106 				 int cmd);
107 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
108 static void mroute_clean_tables(struct mr_table *mrt, int flags);
109 static void ipmr_expire_process(struct timer_list *t);
110 
111 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
112 #define ipmr_for_each_table(mrt, net) \
113 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
114 
115 static struct mr_table *ipmr_mr_table_iter(struct net *net,
116 					   struct mr_table *mrt)
117 {
118 	struct mr_table *ret;
119 
120 	if (!mrt)
121 		ret = list_entry_rcu(net->ipv4.mr_tables.next,
122 				     struct mr_table, list);
123 	else
124 		ret = list_entry_rcu(mrt->list.next,
125 				     struct mr_table, list);
126 
127 	if (&ret->list == &net->ipv4.mr_tables)
128 		return NULL;
129 	return ret;
130 }
131 
132 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
133 {
134 	struct mr_table *mrt;
135 
136 	ipmr_for_each_table(mrt, net) {
137 		if (mrt->id == id)
138 			return mrt;
139 	}
140 	return NULL;
141 }
142 
143 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
144 			   struct mr_table **mrt)
145 {
146 	int err;
147 	struct ipmr_result res;
148 	struct fib_lookup_arg arg = {
149 		.result = &res,
150 		.flags = FIB_LOOKUP_NOREF,
151 	};
152 
153 	/* update flow if oif or iif point to device enslaved to l3mdev */
154 	l3mdev_update_flow(net, flowi4_to_flowi(flp4));
155 
156 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
157 			       flowi4_to_flowi(flp4), 0, &arg);
158 	if (err < 0)
159 		return err;
160 	*mrt = res.mrt;
161 	return 0;
162 }
163 
164 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
165 			    int flags, struct fib_lookup_arg *arg)
166 {
167 	struct ipmr_result *res = arg->result;
168 	struct mr_table *mrt;
169 
170 	switch (rule->action) {
171 	case FR_ACT_TO_TBL:
172 		break;
173 	case FR_ACT_UNREACHABLE:
174 		return -ENETUNREACH;
175 	case FR_ACT_PROHIBIT:
176 		return -EACCES;
177 	case FR_ACT_BLACKHOLE:
178 	default:
179 		return -EINVAL;
180 	}
181 
182 	arg->table = fib_rule_get_table(rule, arg);
183 
184 	mrt = ipmr_get_table(rule->fr_net, arg->table);
185 	if (!mrt)
186 		return -EAGAIN;
187 	res->mrt = mrt;
188 	return 0;
189 }
190 
191 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
192 {
193 	return 1;
194 }
195 
196 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
197 	FRA_GENERIC_POLICY,
198 };
199 
200 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
201 			       struct fib_rule_hdr *frh, struct nlattr **tb,
202 			       struct netlink_ext_ack *extack)
203 {
204 	return 0;
205 }
206 
207 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
208 			     struct nlattr **tb)
209 {
210 	return 1;
211 }
212 
213 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
214 			  struct fib_rule_hdr *frh)
215 {
216 	frh->dst_len = 0;
217 	frh->src_len = 0;
218 	frh->tos     = 0;
219 	return 0;
220 }
221 
222 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
223 	.family		= RTNL_FAMILY_IPMR,
224 	.rule_size	= sizeof(struct ipmr_rule),
225 	.addr_size	= sizeof(u32),
226 	.action		= ipmr_rule_action,
227 	.match		= ipmr_rule_match,
228 	.configure	= ipmr_rule_configure,
229 	.compare	= ipmr_rule_compare,
230 	.fill		= ipmr_rule_fill,
231 	.nlgroup	= RTNLGRP_IPV4_RULE,
232 	.policy		= ipmr_rule_policy,
233 	.owner		= THIS_MODULE,
234 };
235 
236 static int __net_init ipmr_rules_init(struct net *net)
237 {
238 	struct fib_rules_ops *ops;
239 	struct mr_table *mrt;
240 	int err;
241 
242 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
243 	if (IS_ERR(ops))
244 		return PTR_ERR(ops);
245 
246 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
247 
248 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
249 	if (IS_ERR(mrt)) {
250 		err = PTR_ERR(mrt);
251 		goto err1;
252 	}
253 
254 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
255 	if (err < 0)
256 		goto err2;
257 
258 	net->ipv4.mr_rules_ops = ops;
259 	return 0;
260 
261 err2:
262 	ipmr_free_table(mrt);
263 err1:
264 	fib_rules_unregister(ops);
265 	return err;
266 }
267 
268 static void __net_exit ipmr_rules_exit(struct net *net)
269 {
270 	struct mr_table *mrt, *next;
271 
272 	rtnl_lock();
273 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
274 		list_del(&mrt->list);
275 		ipmr_free_table(mrt);
276 	}
277 	fib_rules_unregister(net->ipv4.mr_rules_ops);
278 	rtnl_unlock();
279 }
280 
281 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
282 {
283 	return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR);
284 }
285 
286 static unsigned int ipmr_rules_seq_read(struct net *net)
287 {
288 	return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
289 }
290 
291 bool ipmr_rule_default(const struct fib_rule *rule)
292 {
293 	return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
294 }
295 EXPORT_SYMBOL(ipmr_rule_default);
296 #else
297 #define ipmr_for_each_table(mrt, net) \
298 	for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
299 
300 static struct mr_table *ipmr_mr_table_iter(struct net *net,
301 					   struct mr_table *mrt)
302 {
303 	if (!mrt)
304 		return net->ipv4.mrt;
305 	return NULL;
306 }
307 
308 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
309 {
310 	return net->ipv4.mrt;
311 }
312 
313 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
314 			   struct mr_table **mrt)
315 {
316 	*mrt = net->ipv4.mrt;
317 	return 0;
318 }
319 
320 static int __net_init ipmr_rules_init(struct net *net)
321 {
322 	struct mr_table *mrt;
323 
324 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
325 	if (IS_ERR(mrt))
326 		return PTR_ERR(mrt);
327 	net->ipv4.mrt = mrt;
328 	return 0;
329 }
330 
331 static void __net_exit ipmr_rules_exit(struct net *net)
332 {
333 	rtnl_lock();
334 	ipmr_free_table(net->ipv4.mrt);
335 	net->ipv4.mrt = NULL;
336 	rtnl_unlock();
337 }
338 
339 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
340 {
341 	return 0;
342 }
343 
344 static unsigned int ipmr_rules_seq_read(struct net *net)
345 {
346 	return 0;
347 }
348 
349 bool ipmr_rule_default(const struct fib_rule *rule)
350 {
351 	return true;
352 }
353 EXPORT_SYMBOL(ipmr_rule_default);
354 #endif
355 
356 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
357 				const void *ptr)
358 {
359 	const struct mfc_cache_cmp_arg *cmparg = arg->key;
360 	struct mfc_cache *c = (struct mfc_cache *)ptr;
361 
362 	return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
363 	       cmparg->mfc_origin != c->mfc_origin;
364 }
365 
366 static const struct rhashtable_params ipmr_rht_params = {
367 	.head_offset = offsetof(struct mr_mfc, mnode),
368 	.key_offset = offsetof(struct mfc_cache, cmparg),
369 	.key_len = sizeof(struct mfc_cache_cmp_arg),
370 	.nelem_hint = 3,
371 	.obj_cmpfn = ipmr_hash_cmp,
372 	.automatic_shrinking = true,
373 };
374 
375 static void ipmr_new_table_set(struct mr_table *mrt,
376 			       struct net *net)
377 {
378 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
379 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
380 #endif
381 }
382 
383 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
384 	.mfc_mcastgrp = htonl(INADDR_ANY),
385 	.mfc_origin = htonl(INADDR_ANY),
386 };
387 
388 static struct mr_table_ops ipmr_mr_table_ops = {
389 	.rht_params = &ipmr_rht_params,
390 	.cmparg_any = &ipmr_mr_table_ops_cmparg_any,
391 };
392 
393 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
394 {
395 	struct mr_table *mrt;
396 
397 	/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
398 	if (id != RT_TABLE_DEFAULT && id >= 1000000000)
399 		return ERR_PTR(-EINVAL);
400 
401 	mrt = ipmr_get_table(net, id);
402 	if (mrt)
403 		return mrt;
404 
405 	return mr_table_alloc(net, id, &ipmr_mr_table_ops,
406 			      ipmr_expire_process, ipmr_new_table_set);
407 }
408 
409 static void ipmr_free_table(struct mr_table *mrt)
410 {
411 	del_timer_sync(&mrt->ipmr_expire_timer);
412 	mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
413 				 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
414 	rhltable_destroy(&mrt->mfc_hash);
415 	kfree(mrt);
416 }
417 
418 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
419 
420 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
421 {
422 	struct net *net = dev_net(dev);
423 
424 	dev_close(dev);
425 
426 	dev = __dev_get_by_name(net, "tunl0");
427 	if (dev) {
428 		const struct net_device_ops *ops = dev->netdev_ops;
429 		struct ifreq ifr;
430 		struct ip_tunnel_parm p;
431 
432 		memset(&p, 0, sizeof(p));
433 		p.iph.daddr = v->vifc_rmt_addr.s_addr;
434 		p.iph.saddr = v->vifc_lcl_addr.s_addr;
435 		p.iph.version = 4;
436 		p.iph.ihl = 5;
437 		p.iph.protocol = IPPROTO_IPIP;
438 		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
439 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
440 
441 		if (ops->ndo_do_ioctl) {
442 			mm_segment_t oldfs = get_fs();
443 
444 			set_fs(KERNEL_DS);
445 			ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
446 			set_fs(oldfs);
447 		}
448 	}
449 }
450 
451 /* Initialize ipmr pimreg/tunnel in_device */
452 static bool ipmr_init_vif_indev(const struct net_device *dev)
453 {
454 	struct in_device *in_dev;
455 
456 	ASSERT_RTNL();
457 
458 	in_dev = __in_dev_get_rtnl(dev);
459 	if (!in_dev)
460 		return false;
461 	ipv4_devconf_setall(in_dev);
462 	neigh_parms_data_state_setall(in_dev->arp_parms);
463 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
464 
465 	return true;
466 }
467 
468 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
469 {
470 	struct net_device  *dev;
471 
472 	dev = __dev_get_by_name(net, "tunl0");
473 
474 	if (dev) {
475 		const struct net_device_ops *ops = dev->netdev_ops;
476 		int err;
477 		struct ifreq ifr;
478 		struct ip_tunnel_parm p;
479 
480 		memset(&p, 0, sizeof(p));
481 		p.iph.daddr = v->vifc_rmt_addr.s_addr;
482 		p.iph.saddr = v->vifc_lcl_addr.s_addr;
483 		p.iph.version = 4;
484 		p.iph.ihl = 5;
485 		p.iph.protocol = IPPROTO_IPIP;
486 		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
487 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
488 
489 		if (ops->ndo_do_ioctl) {
490 			mm_segment_t oldfs = get_fs();
491 
492 			set_fs(KERNEL_DS);
493 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
494 			set_fs(oldfs);
495 		} else {
496 			err = -EOPNOTSUPP;
497 		}
498 		dev = NULL;
499 
500 		if (err == 0 &&
501 		    (dev = __dev_get_by_name(net, p.name)) != NULL) {
502 			dev->flags |= IFF_MULTICAST;
503 			if (!ipmr_init_vif_indev(dev))
504 				goto failure;
505 			if (dev_open(dev, NULL))
506 				goto failure;
507 			dev_hold(dev);
508 		}
509 	}
510 	return dev;
511 
512 failure:
513 	unregister_netdevice(dev);
514 	return NULL;
515 }
516 
517 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
518 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
519 {
520 	struct net *net = dev_net(dev);
521 	struct mr_table *mrt;
522 	struct flowi4 fl4 = {
523 		.flowi4_oif	= dev->ifindex,
524 		.flowi4_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
525 		.flowi4_mark	= skb->mark,
526 	};
527 	int err;
528 
529 	err = ipmr_fib_lookup(net, &fl4, &mrt);
530 	if (err < 0) {
531 		kfree_skb(skb);
532 		return err;
533 	}
534 
535 	read_lock(&mrt_lock);
536 	dev->stats.tx_bytes += skb->len;
537 	dev->stats.tx_packets++;
538 	ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
539 	read_unlock(&mrt_lock);
540 	kfree_skb(skb);
541 	return NETDEV_TX_OK;
542 }
543 
544 static int reg_vif_get_iflink(const struct net_device *dev)
545 {
546 	return 0;
547 }
548 
549 static const struct net_device_ops reg_vif_netdev_ops = {
550 	.ndo_start_xmit	= reg_vif_xmit,
551 	.ndo_get_iflink = reg_vif_get_iflink,
552 };
553 
554 static void reg_vif_setup(struct net_device *dev)
555 {
556 	dev->type		= ARPHRD_PIMREG;
557 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
558 	dev->flags		= IFF_NOARP;
559 	dev->netdev_ops		= &reg_vif_netdev_ops;
560 	dev->needs_free_netdev	= true;
561 	dev->features		|= NETIF_F_NETNS_LOCAL;
562 }
563 
564 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
565 {
566 	struct net_device *dev;
567 	char name[IFNAMSIZ];
568 
569 	if (mrt->id == RT_TABLE_DEFAULT)
570 		sprintf(name, "pimreg");
571 	else
572 		sprintf(name, "pimreg%u", mrt->id);
573 
574 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
575 
576 	if (!dev)
577 		return NULL;
578 
579 	dev_net_set(dev, net);
580 
581 	if (register_netdevice(dev)) {
582 		free_netdev(dev);
583 		return NULL;
584 	}
585 
586 	if (!ipmr_init_vif_indev(dev))
587 		goto failure;
588 	if (dev_open(dev, NULL))
589 		goto failure;
590 
591 	dev_hold(dev);
592 
593 	return dev;
594 
595 failure:
596 	unregister_netdevice(dev);
597 	return NULL;
598 }
599 
600 /* called with rcu_read_lock() */
601 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
602 		     unsigned int pimlen)
603 {
604 	struct net_device *reg_dev = NULL;
605 	struct iphdr *encap;
606 
607 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
608 	/* Check that:
609 	 * a. packet is really sent to a multicast group
610 	 * b. packet is not a NULL-REGISTER
611 	 * c. packet is not truncated
612 	 */
613 	if (!ipv4_is_multicast(encap->daddr) ||
614 	    encap->tot_len == 0 ||
615 	    ntohs(encap->tot_len) + pimlen > skb->len)
616 		return 1;
617 
618 	read_lock(&mrt_lock);
619 	if (mrt->mroute_reg_vif_num >= 0)
620 		reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
621 	read_unlock(&mrt_lock);
622 
623 	if (!reg_dev)
624 		return 1;
625 
626 	skb->mac_header = skb->network_header;
627 	skb_pull(skb, (u8 *)encap - skb->data);
628 	skb_reset_network_header(skb);
629 	skb->protocol = htons(ETH_P_IP);
630 	skb->ip_summed = CHECKSUM_NONE;
631 
632 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
633 
634 	netif_rx(skb);
635 
636 	return NET_RX_SUCCESS;
637 }
638 #else
639 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
640 {
641 	return NULL;
642 }
643 #endif
644 
645 static int call_ipmr_vif_entry_notifiers(struct net *net,
646 					 enum fib_event_type event_type,
647 					 struct vif_device *vif,
648 					 vifi_t vif_index, u32 tb_id)
649 {
650 	return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
651 				     vif, vif_index, tb_id,
652 				     &net->ipv4.ipmr_seq);
653 }
654 
655 static int call_ipmr_mfc_entry_notifiers(struct net *net,
656 					 enum fib_event_type event_type,
657 					 struct mfc_cache *mfc, u32 tb_id)
658 {
659 	return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
660 				     &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
661 }
662 
663 /**
664  *	vif_delete - Delete a VIF entry
665  *	@notify: Set to 1, if the caller is a notifier_call
666  */
667 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
668 		      struct list_head *head)
669 {
670 	struct net *net = read_pnet(&mrt->net);
671 	struct vif_device *v;
672 	struct net_device *dev;
673 	struct in_device *in_dev;
674 
675 	if (vifi < 0 || vifi >= mrt->maxvif)
676 		return -EADDRNOTAVAIL;
677 
678 	v = &mrt->vif_table[vifi];
679 
680 	if (VIF_EXISTS(mrt, vifi))
681 		call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
682 					      mrt->id);
683 
684 	write_lock_bh(&mrt_lock);
685 	dev = v->dev;
686 	v->dev = NULL;
687 
688 	if (!dev) {
689 		write_unlock_bh(&mrt_lock);
690 		return -EADDRNOTAVAIL;
691 	}
692 
693 	if (vifi == mrt->mroute_reg_vif_num)
694 		mrt->mroute_reg_vif_num = -1;
695 
696 	if (vifi + 1 == mrt->maxvif) {
697 		int tmp;
698 
699 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
700 			if (VIF_EXISTS(mrt, tmp))
701 				break;
702 		}
703 		mrt->maxvif = tmp+1;
704 	}
705 
706 	write_unlock_bh(&mrt_lock);
707 
708 	dev_set_allmulti(dev, -1);
709 
710 	in_dev = __in_dev_get_rtnl(dev);
711 	if (in_dev) {
712 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
713 		inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
714 					    NETCONFA_MC_FORWARDING,
715 					    dev->ifindex, &in_dev->cnf);
716 		ip_rt_multicast_event(in_dev);
717 	}
718 
719 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
720 		unregister_netdevice_queue(dev, head);
721 
722 	dev_put(dev);
723 	return 0;
724 }
725 
726 static void ipmr_cache_free_rcu(struct rcu_head *head)
727 {
728 	struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
729 
730 	kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
731 }
732 
733 static void ipmr_cache_free(struct mfc_cache *c)
734 {
735 	call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
736 }
737 
738 /* Destroy an unresolved cache entry, killing queued skbs
739  * and reporting error to netlink readers.
740  */
741 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
742 {
743 	struct net *net = read_pnet(&mrt->net);
744 	struct sk_buff *skb;
745 	struct nlmsgerr *e;
746 
747 	atomic_dec(&mrt->cache_resolve_queue_len);
748 
749 	while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
750 		if (ip_hdr(skb)->version == 0) {
751 			struct nlmsghdr *nlh = skb_pull(skb,
752 							sizeof(struct iphdr));
753 			nlh->nlmsg_type = NLMSG_ERROR;
754 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
755 			skb_trim(skb, nlh->nlmsg_len);
756 			e = nlmsg_data(nlh);
757 			e->error = -ETIMEDOUT;
758 			memset(&e->msg, 0, sizeof(e->msg));
759 
760 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
761 		} else {
762 			kfree_skb(skb);
763 		}
764 	}
765 
766 	ipmr_cache_free(c);
767 }
768 
769 /* Timer process for the unresolved queue. */
770 static void ipmr_expire_process(struct timer_list *t)
771 {
772 	struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
773 	struct mr_mfc *c, *next;
774 	unsigned long expires;
775 	unsigned long now;
776 
777 	if (!spin_trylock(&mfc_unres_lock)) {
778 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
779 		return;
780 	}
781 
782 	if (list_empty(&mrt->mfc_unres_queue))
783 		goto out;
784 
785 	now = jiffies;
786 	expires = 10*HZ;
787 
788 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
789 		if (time_after(c->mfc_un.unres.expires, now)) {
790 			unsigned long interval = c->mfc_un.unres.expires - now;
791 			if (interval < expires)
792 				expires = interval;
793 			continue;
794 		}
795 
796 		list_del(&c->list);
797 		mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
798 		ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
799 	}
800 
801 	if (!list_empty(&mrt->mfc_unres_queue))
802 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
803 
804 out:
805 	spin_unlock(&mfc_unres_lock);
806 }
807 
808 /* Fill oifs list. It is called under write locked mrt_lock. */
809 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
810 				   unsigned char *ttls)
811 {
812 	int vifi;
813 
814 	cache->mfc_un.res.minvif = MAXVIFS;
815 	cache->mfc_un.res.maxvif = 0;
816 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
817 
818 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
819 		if (VIF_EXISTS(mrt, vifi) &&
820 		    ttls[vifi] && ttls[vifi] < 255) {
821 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
822 			if (cache->mfc_un.res.minvif > vifi)
823 				cache->mfc_un.res.minvif = vifi;
824 			if (cache->mfc_un.res.maxvif <= vifi)
825 				cache->mfc_un.res.maxvif = vifi + 1;
826 		}
827 	}
828 	cache->mfc_un.res.lastuse = jiffies;
829 }
830 
831 static int vif_add(struct net *net, struct mr_table *mrt,
832 		   struct vifctl *vifc, int mrtsock)
833 {
834 	struct netdev_phys_item_id ppid = { };
835 	int vifi = vifc->vifc_vifi;
836 	struct vif_device *v = &mrt->vif_table[vifi];
837 	struct net_device *dev;
838 	struct in_device *in_dev;
839 	int err;
840 
841 	/* Is vif busy ? */
842 	if (VIF_EXISTS(mrt, vifi))
843 		return -EADDRINUSE;
844 
845 	switch (vifc->vifc_flags) {
846 	case VIFF_REGISTER:
847 		if (!ipmr_pimsm_enabled())
848 			return -EINVAL;
849 		/* Special Purpose VIF in PIM
850 		 * All the packets will be sent to the daemon
851 		 */
852 		if (mrt->mroute_reg_vif_num >= 0)
853 			return -EADDRINUSE;
854 		dev = ipmr_reg_vif(net, mrt);
855 		if (!dev)
856 			return -ENOBUFS;
857 		err = dev_set_allmulti(dev, 1);
858 		if (err) {
859 			unregister_netdevice(dev);
860 			dev_put(dev);
861 			return err;
862 		}
863 		break;
864 	case VIFF_TUNNEL:
865 		dev = ipmr_new_tunnel(net, vifc);
866 		if (!dev)
867 			return -ENOBUFS;
868 		err = dev_set_allmulti(dev, 1);
869 		if (err) {
870 			ipmr_del_tunnel(dev, vifc);
871 			dev_put(dev);
872 			return err;
873 		}
874 		break;
875 	case VIFF_USE_IFINDEX:
876 	case 0:
877 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
878 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
879 			if (dev && !__in_dev_get_rtnl(dev)) {
880 				dev_put(dev);
881 				return -EADDRNOTAVAIL;
882 			}
883 		} else {
884 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
885 		}
886 		if (!dev)
887 			return -EADDRNOTAVAIL;
888 		err = dev_set_allmulti(dev, 1);
889 		if (err) {
890 			dev_put(dev);
891 			return err;
892 		}
893 		break;
894 	default:
895 		return -EINVAL;
896 	}
897 
898 	in_dev = __in_dev_get_rtnl(dev);
899 	if (!in_dev) {
900 		dev_put(dev);
901 		return -EADDRNOTAVAIL;
902 	}
903 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
904 	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
905 				    dev->ifindex, &in_dev->cnf);
906 	ip_rt_multicast_event(in_dev);
907 
908 	/* Fill in the VIF structures */
909 	vif_device_init(v, dev, vifc->vifc_rate_limit,
910 			vifc->vifc_threshold,
911 			vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
912 			(VIFF_TUNNEL | VIFF_REGISTER));
913 
914 	err = dev_get_port_parent_id(dev, &ppid, true);
915 	if (err == 0) {
916 		memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
917 		v->dev_parent_id.id_len = ppid.id_len;
918 	} else {
919 		v->dev_parent_id.id_len = 0;
920 	}
921 
922 	v->local = vifc->vifc_lcl_addr.s_addr;
923 	v->remote = vifc->vifc_rmt_addr.s_addr;
924 
925 	/* And finish update writing critical data */
926 	write_lock_bh(&mrt_lock);
927 	v->dev = dev;
928 	if (v->flags & VIFF_REGISTER)
929 		mrt->mroute_reg_vif_num = vifi;
930 	if (vifi+1 > mrt->maxvif)
931 		mrt->maxvif = vifi+1;
932 	write_unlock_bh(&mrt_lock);
933 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
934 	return 0;
935 }
936 
937 /* called with rcu_read_lock() */
938 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
939 					 __be32 origin,
940 					 __be32 mcastgrp)
941 {
942 	struct mfc_cache_cmp_arg arg = {
943 			.mfc_mcastgrp = mcastgrp,
944 			.mfc_origin = origin
945 	};
946 
947 	return mr_mfc_find(mrt, &arg);
948 }
949 
950 /* Look for a (*,G) entry */
951 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
952 					     __be32 mcastgrp, int vifi)
953 {
954 	struct mfc_cache_cmp_arg arg = {
955 			.mfc_mcastgrp = mcastgrp,
956 			.mfc_origin = htonl(INADDR_ANY)
957 	};
958 
959 	if (mcastgrp == htonl(INADDR_ANY))
960 		return mr_mfc_find_any_parent(mrt, vifi);
961 	return mr_mfc_find_any(mrt, vifi, &arg);
962 }
963 
964 /* Look for a (S,G,iif) entry if parent != -1 */
965 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
966 						__be32 origin, __be32 mcastgrp,
967 						int parent)
968 {
969 	struct mfc_cache_cmp_arg arg = {
970 			.mfc_mcastgrp = mcastgrp,
971 			.mfc_origin = origin,
972 	};
973 
974 	return mr_mfc_find_parent(mrt, &arg, parent);
975 }
976 
977 /* Allocate a multicast cache entry */
978 static struct mfc_cache *ipmr_cache_alloc(void)
979 {
980 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
981 
982 	if (c) {
983 		c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
984 		c->_c.mfc_un.res.minvif = MAXVIFS;
985 		c->_c.free = ipmr_cache_free_rcu;
986 		refcount_set(&c->_c.mfc_un.res.refcount, 1);
987 	}
988 	return c;
989 }
990 
991 static struct mfc_cache *ipmr_cache_alloc_unres(void)
992 {
993 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
994 
995 	if (c) {
996 		skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
997 		c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
998 	}
999 	return c;
1000 }
1001 
1002 /* A cache entry has gone into a resolved state from queued */
1003 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1004 			       struct mfc_cache *uc, struct mfc_cache *c)
1005 {
1006 	struct sk_buff *skb;
1007 	struct nlmsgerr *e;
1008 
1009 	/* Play the pending entries through our router */
1010 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1011 		if (ip_hdr(skb)->version == 0) {
1012 			struct nlmsghdr *nlh = skb_pull(skb,
1013 							sizeof(struct iphdr));
1014 
1015 			if (mr_fill_mroute(mrt, skb, &c->_c,
1016 					   nlmsg_data(nlh)) > 0) {
1017 				nlh->nlmsg_len = skb_tail_pointer(skb) -
1018 						 (u8 *)nlh;
1019 			} else {
1020 				nlh->nlmsg_type = NLMSG_ERROR;
1021 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1022 				skb_trim(skb, nlh->nlmsg_len);
1023 				e = nlmsg_data(nlh);
1024 				e->error = -EMSGSIZE;
1025 				memset(&e->msg, 0, sizeof(e->msg));
1026 			}
1027 
1028 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1029 		} else {
1030 			ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1031 		}
1032 	}
1033 }
1034 
1035 /* Bounce a cache query up to mrouted and netlink.
1036  *
1037  * Called under mrt_lock.
1038  */
1039 static int ipmr_cache_report(struct mr_table *mrt,
1040 			     struct sk_buff *pkt, vifi_t vifi, int assert)
1041 {
1042 	const int ihl = ip_hdrlen(pkt);
1043 	struct sock *mroute_sk;
1044 	struct igmphdr *igmp;
1045 	struct igmpmsg *msg;
1046 	struct sk_buff *skb;
1047 	int ret;
1048 
1049 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1050 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1051 	else
1052 		skb = alloc_skb(128, GFP_ATOMIC);
1053 
1054 	if (!skb)
1055 		return -ENOBUFS;
1056 
1057 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1058 		/* Ugly, but we have no choice with this interface.
1059 		 * Duplicate old header, fix ihl, length etc.
1060 		 * And all this only to mangle msg->im_msgtype and
1061 		 * to set msg->im_mbz to "mbz" :-)
1062 		 */
1063 		skb_push(skb, sizeof(struct iphdr));
1064 		skb_reset_network_header(skb);
1065 		skb_reset_transport_header(skb);
1066 		msg = (struct igmpmsg *)skb_network_header(skb);
1067 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1068 		msg->im_msgtype = assert;
1069 		msg->im_mbz = 0;
1070 		if (assert == IGMPMSG_WRVIFWHOLE)
1071 			msg->im_vif = vifi;
1072 		else
1073 			msg->im_vif = mrt->mroute_reg_vif_num;
1074 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1075 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1076 					     sizeof(struct iphdr));
1077 	} else {
1078 		/* Copy the IP header */
1079 		skb_set_network_header(skb, skb->len);
1080 		skb_put(skb, ihl);
1081 		skb_copy_to_linear_data(skb, pkt->data, ihl);
1082 		/* Flag to the kernel this is a route add */
1083 		ip_hdr(skb)->protocol = 0;
1084 		msg = (struct igmpmsg *)skb_network_header(skb);
1085 		msg->im_vif = vifi;
1086 		skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1087 		/* Add our header */
1088 		igmp = skb_put(skb, sizeof(struct igmphdr));
1089 		igmp->type = assert;
1090 		msg->im_msgtype = assert;
1091 		igmp->code = 0;
1092 		ip_hdr(skb)->tot_len = htons(skb->len);	/* Fix the length */
1093 		skb->transport_header = skb->network_header;
1094 	}
1095 
1096 	rcu_read_lock();
1097 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1098 	if (!mroute_sk) {
1099 		rcu_read_unlock();
1100 		kfree_skb(skb);
1101 		return -EINVAL;
1102 	}
1103 
1104 	igmpmsg_netlink_event(mrt, skb);
1105 
1106 	/* Deliver to mrouted */
1107 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1108 	rcu_read_unlock();
1109 	if (ret < 0) {
1110 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1111 		kfree_skb(skb);
1112 	}
1113 
1114 	return ret;
1115 }
1116 
1117 /* Queue a packet for resolution. It gets locked cache entry! */
1118 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1119 				 struct sk_buff *skb, struct net_device *dev)
1120 {
1121 	const struct iphdr *iph = ip_hdr(skb);
1122 	struct mfc_cache *c;
1123 	bool found = false;
1124 	int err;
1125 
1126 	spin_lock_bh(&mfc_unres_lock);
1127 	list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1128 		if (c->mfc_mcastgrp == iph->daddr &&
1129 		    c->mfc_origin == iph->saddr) {
1130 			found = true;
1131 			break;
1132 		}
1133 	}
1134 
1135 	if (!found) {
1136 		/* Create a new entry if allowable */
1137 		if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1138 		    (c = ipmr_cache_alloc_unres()) == NULL) {
1139 			spin_unlock_bh(&mfc_unres_lock);
1140 
1141 			kfree_skb(skb);
1142 			return -ENOBUFS;
1143 		}
1144 
1145 		/* Fill in the new cache entry */
1146 		c->_c.mfc_parent = -1;
1147 		c->mfc_origin	= iph->saddr;
1148 		c->mfc_mcastgrp	= iph->daddr;
1149 
1150 		/* Reflect first query at mrouted. */
1151 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1152 
1153 		if (err < 0) {
1154 			/* If the report failed throw the cache entry
1155 			   out - Brad Parker
1156 			 */
1157 			spin_unlock_bh(&mfc_unres_lock);
1158 
1159 			ipmr_cache_free(c);
1160 			kfree_skb(skb);
1161 			return err;
1162 		}
1163 
1164 		atomic_inc(&mrt->cache_resolve_queue_len);
1165 		list_add(&c->_c.list, &mrt->mfc_unres_queue);
1166 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1167 
1168 		if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1169 			mod_timer(&mrt->ipmr_expire_timer,
1170 				  c->_c.mfc_un.unres.expires);
1171 	}
1172 
1173 	/* See if we can append the packet */
1174 	if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1175 		kfree_skb(skb);
1176 		err = -ENOBUFS;
1177 	} else {
1178 		if (dev) {
1179 			skb->dev = dev;
1180 			skb->skb_iif = dev->ifindex;
1181 		}
1182 		skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1183 		err = 0;
1184 	}
1185 
1186 	spin_unlock_bh(&mfc_unres_lock);
1187 	return err;
1188 }
1189 
1190 /* MFC cache manipulation by user space mroute daemon */
1191 
1192 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1193 {
1194 	struct net *net = read_pnet(&mrt->net);
1195 	struct mfc_cache *c;
1196 
1197 	/* The entries are added/deleted only under RTNL */
1198 	rcu_read_lock();
1199 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1200 				   mfc->mfcc_mcastgrp.s_addr, parent);
1201 	rcu_read_unlock();
1202 	if (!c)
1203 		return -ENOENT;
1204 	rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1205 	list_del_rcu(&c->_c.list);
1206 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1207 	mroute_netlink_event(mrt, c, RTM_DELROUTE);
1208 	mr_cache_put(&c->_c);
1209 
1210 	return 0;
1211 }
1212 
1213 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1214 			struct mfcctl *mfc, int mrtsock, int parent)
1215 {
1216 	struct mfc_cache *uc, *c;
1217 	struct mr_mfc *_uc;
1218 	bool found;
1219 	int ret;
1220 
1221 	if (mfc->mfcc_parent >= MAXVIFS)
1222 		return -ENFILE;
1223 
1224 	/* The entries are added/deleted only under RTNL */
1225 	rcu_read_lock();
1226 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1227 				   mfc->mfcc_mcastgrp.s_addr, parent);
1228 	rcu_read_unlock();
1229 	if (c) {
1230 		write_lock_bh(&mrt_lock);
1231 		c->_c.mfc_parent = mfc->mfcc_parent;
1232 		ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1233 		if (!mrtsock)
1234 			c->_c.mfc_flags |= MFC_STATIC;
1235 		write_unlock_bh(&mrt_lock);
1236 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1237 					      mrt->id);
1238 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1239 		return 0;
1240 	}
1241 
1242 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1243 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1244 		return -EINVAL;
1245 
1246 	c = ipmr_cache_alloc();
1247 	if (!c)
1248 		return -ENOMEM;
1249 
1250 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1251 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1252 	c->_c.mfc_parent = mfc->mfcc_parent;
1253 	ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1254 	if (!mrtsock)
1255 		c->_c.mfc_flags |= MFC_STATIC;
1256 
1257 	ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1258 				  ipmr_rht_params);
1259 	if (ret) {
1260 		pr_err("ipmr: rhtable insert error %d\n", ret);
1261 		ipmr_cache_free(c);
1262 		return ret;
1263 	}
1264 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1265 	/* Check to see if we resolved a queued list. If so we
1266 	 * need to send on the frames and tidy up.
1267 	 */
1268 	found = false;
1269 	spin_lock_bh(&mfc_unres_lock);
1270 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1271 		uc = (struct mfc_cache *)_uc;
1272 		if (uc->mfc_origin == c->mfc_origin &&
1273 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1274 			list_del(&_uc->list);
1275 			atomic_dec(&mrt->cache_resolve_queue_len);
1276 			found = true;
1277 			break;
1278 		}
1279 	}
1280 	if (list_empty(&mrt->mfc_unres_queue))
1281 		del_timer(&mrt->ipmr_expire_timer);
1282 	spin_unlock_bh(&mfc_unres_lock);
1283 
1284 	if (found) {
1285 		ipmr_cache_resolve(net, mrt, uc, c);
1286 		ipmr_cache_free(uc);
1287 	}
1288 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1289 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1290 	return 0;
1291 }
1292 
1293 /* Close the multicast socket, and clear the vif tables etc */
1294 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1295 {
1296 	struct net *net = read_pnet(&mrt->net);
1297 	struct mr_mfc *c, *tmp;
1298 	struct mfc_cache *cache;
1299 	LIST_HEAD(list);
1300 	int i;
1301 
1302 	/* Shut down all active vif entries */
1303 	if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1304 		for (i = 0; i < mrt->maxvif; i++) {
1305 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1306 			     !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1307 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1308 				continue;
1309 			vif_delete(mrt, i, 0, &list);
1310 		}
1311 		unregister_netdevice_many(&list);
1312 	}
1313 
1314 	/* Wipe the cache */
1315 	if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1316 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1317 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1318 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1319 				continue;
1320 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1321 			list_del_rcu(&c->list);
1322 			cache = (struct mfc_cache *)c;
1323 			call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1324 						      mrt->id);
1325 			mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1326 			mr_cache_put(c);
1327 		}
1328 	}
1329 
1330 	if (flags & MRT_FLUSH_MFC) {
1331 		if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1332 			spin_lock_bh(&mfc_unres_lock);
1333 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1334 				list_del(&c->list);
1335 				cache = (struct mfc_cache *)c;
1336 				mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1337 				ipmr_destroy_unres(mrt, cache);
1338 			}
1339 			spin_unlock_bh(&mfc_unres_lock);
1340 		}
1341 	}
1342 }
1343 
1344 /* called from ip_ra_control(), before an RCU grace period,
1345  * we dont need to call synchronize_rcu() here
1346  */
1347 static void mrtsock_destruct(struct sock *sk)
1348 {
1349 	struct net *net = sock_net(sk);
1350 	struct mr_table *mrt;
1351 
1352 	rtnl_lock();
1353 	ipmr_for_each_table(mrt, net) {
1354 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1355 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1356 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1357 						    NETCONFA_MC_FORWARDING,
1358 						    NETCONFA_IFINDEX_ALL,
1359 						    net->ipv4.devconf_all);
1360 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1361 			mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1362 		}
1363 	}
1364 	rtnl_unlock();
1365 }
1366 
1367 /* Socket options and virtual interface manipulation. The whole
1368  * virtual interface system is a complete heap, but unfortunately
1369  * that's how BSD mrouted happens to think. Maybe one day with a proper
1370  * MOSPF/PIM router set up we can clean this up.
1371  */
1372 
1373 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1374 			 unsigned int optlen)
1375 {
1376 	struct net *net = sock_net(sk);
1377 	int val, ret = 0, parent = 0;
1378 	struct mr_table *mrt;
1379 	struct vifctl vif;
1380 	struct mfcctl mfc;
1381 	bool do_wrvifwhole;
1382 	u32 uval;
1383 
1384 	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1385 	rtnl_lock();
1386 	if (sk->sk_type != SOCK_RAW ||
1387 	    inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1388 		ret = -EOPNOTSUPP;
1389 		goto out_unlock;
1390 	}
1391 
1392 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1393 	if (!mrt) {
1394 		ret = -ENOENT;
1395 		goto out_unlock;
1396 	}
1397 	if (optname != MRT_INIT) {
1398 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1399 		    !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1400 			ret = -EACCES;
1401 			goto out_unlock;
1402 		}
1403 	}
1404 
1405 	switch (optname) {
1406 	case MRT_INIT:
1407 		if (optlen != sizeof(int)) {
1408 			ret = -EINVAL;
1409 			break;
1410 		}
1411 		if (rtnl_dereference(mrt->mroute_sk)) {
1412 			ret = -EADDRINUSE;
1413 			break;
1414 		}
1415 
1416 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1417 		if (ret == 0) {
1418 			rcu_assign_pointer(mrt->mroute_sk, sk);
1419 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1420 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1421 						    NETCONFA_MC_FORWARDING,
1422 						    NETCONFA_IFINDEX_ALL,
1423 						    net->ipv4.devconf_all);
1424 		}
1425 		break;
1426 	case MRT_DONE:
1427 		if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1428 			ret = -EACCES;
1429 		} else {
1430 			/* We need to unlock here because mrtsock_destruct takes
1431 			 * care of rtnl itself and we can't change that due to
1432 			 * the IP_ROUTER_ALERT setsockopt which runs without it.
1433 			 */
1434 			rtnl_unlock();
1435 			ret = ip_ra_control(sk, 0, NULL);
1436 			goto out;
1437 		}
1438 		break;
1439 	case MRT_ADD_VIF:
1440 	case MRT_DEL_VIF:
1441 		if (optlen != sizeof(vif)) {
1442 			ret = -EINVAL;
1443 			break;
1444 		}
1445 		if (copy_from_user(&vif, optval, sizeof(vif))) {
1446 			ret = -EFAULT;
1447 			break;
1448 		}
1449 		if (vif.vifc_vifi >= MAXVIFS) {
1450 			ret = -ENFILE;
1451 			break;
1452 		}
1453 		if (optname == MRT_ADD_VIF) {
1454 			ret = vif_add(net, mrt, &vif,
1455 				      sk == rtnl_dereference(mrt->mroute_sk));
1456 		} else {
1457 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1458 		}
1459 		break;
1460 	/* Manipulate the forwarding caches. These live
1461 	 * in a sort of kernel/user symbiosis.
1462 	 */
1463 	case MRT_ADD_MFC:
1464 	case MRT_DEL_MFC:
1465 		parent = -1;
1466 		/* fall through */
1467 	case MRT_ADD_MFC_PROXY:
1468 	case MRT_DEL_MFC_PROXY:
1469 		if (optlen != sizeof(mfc)) {
1470 			ret = -EINVAL;
1471 			break;
1472 		}
1473 		if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1474 			ret = -EFAULT;
1475 			break;
1476 		}
1477 		if (parent == 0)
1478 			parent = mfc.mfcc_parent;
1479 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1480 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1481 		else
1482 			ret = ipmr_mfc_add(net, mrt, &mfc,
1483 					   sk == rtnl_dereference(mrt->mroute_sk),
1484 					   parent);
1485 		break;
1486 	case MRT_FLUSH:
1487 		if (optlen != sizeof(val)) {
1488 			ret = -EINVAL;
1489 			break;
1490 		}
1491 		if (get_user(val, (int __user *)optval)) {
1492 			ret = -EFAULT;
1493 			break;
1494 		}
1495 		mroute_clean_tables(mrt, val);
1496 		break;
1497 	/* Control PIM assert. */
1498 	case MRT_ASSERT:
1499 		if (optlen != sizeof(val)) {
1500 			ret = -EINVAL;
1501 			break;
1502 		}
1503 		if (get_user(val, (int __user *)optval)) {
1504 			ret = -EFAULT;
1505 			break;
1506 		}
1507 		mrt->mroute_do_assert = val;
1508 		break;
1509 	case MRT_PIM:
1510 		if (!ipmr_pimsm_enabled()) {
1511 			ret = -ENOPROTOOPT;
1512 			break;
1513 		}
1514 		if (optlen != sizeof(val)) {
1515 			ret = -EINVAL;
1516 			break;
1517 		}
1518 		if (get_user(val, (int __user *)optval)) {
1519 			ret = -EFAULT;
1520 			break;
1521 		}
1522 
1523 		do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1524 		val = !!val;
1525 		if (val != mrt->mroute_do_pim) {
1526 			mrt->mroute_do_pim = val;
1527 			mrt->mroute_do_assert = val;
1528 			mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1529 		}
1530 		break;
1531 	case MRT_TABLE:
1532 		if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1533 			ret = -ENOPROTOOPT;
1534 			break;
1535 		}
1536 		if (optlen != sizeof(uval)) {
1537 			ret = -EINVAL;
1538 			break;
1539 		}
1540 		if (get_user(uval, (u32 __user *)optval)) {
1541 			ret = -EFAULT;
1542 			break;
1543 		}
1544 
1545 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1546 			ret = -EBUSY;
1547 		} else {
1548 			mrt = ipmr_new_table(net, uval);
1549 			if (IS_ERR(mrt))
1550 				ret = PTR_ERR(mrt);
1551 			else
1552 				raw_sk(sk)->ipmr_table = uval;
1553 		}
1554 		break;
1555 	/* Spurious command, or MRT_VERSION which you cannot set. */
1556 	default:
1557 		ret = -ENOPROTOOPT;
1558 	}
1559 out_unlock:
1560 	rtnl_unlock();
1561 out:
1562 	return ret;
1563 }
1564 
1565 /* Getsock opt support for the multicast routing system. */
1566 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1567 {
1568 	int olr;
1569 	int val;
1570 	struct net *net = sock_net(sk);
1571 	struct mr_table *mrt;
1572 
1573 	if (sk->sk_type != SOCK_RAW ||
1574 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1575 		return -EOPNOTSUPP;
1576 
1577 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1578 	if (!mrt)
1579 		return -ENOENT;
1580 
1581 	switch (optname) {
1582 	case MRT_VERSION:
1583 		val = 0x0305;
1584 		break;
1585 	case MRT_PIM:
1586 		if (!ipmr_pimsm_enabled())
1587 			return -ENOPROTOOPT;
1588 		val = mrt->mroute_do_pim;
1589 		break;
1590 	case MRT_ASSERT:
1591 		val = mrt->mroute_do_assert;
1592 		break;
1593 	default:
1594 		return -ENOPROTOOPT;
1595 	}
1596 
1597 	if (get_user(olr, optlen))
1598 		return -EFAULT;
1599 	olr = min_t(unsigned int, olr, sizeof(int));
1600 	if (olr < 0)
1601 		return -EINVAL;
1602 	if (put_user(olr, optlen))
1603 		return -EFAULT;
1604 	if (copy_to_user(optval, &val, olr))
1605 		return -EFAULT;
1606 	return 0;
1607 }
1608 
1609 /* The IP multicast ioctl support routines. */
1610 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1611 {
1612 	struct sioc_sg_req sr;
1613 	struct sioc_vif_req vr;
1614 	struct vif_device *vif;
1615 	struct mfc_cache *c;
1616 	struct net *net = sock_net(sk);
1617 	struct mr_table *mrt;
1618 
1619 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1620 	if (!mrt)
1621 		return -ENOENT;
1622 
1623 	switch (cmd) {
1624 	case SIOCGETVIFCNT:
1625 		if (copy_from_user(&vr, arg, sizeof(vr)))
1626 			return -EFAULT;
1627 		if (vr.vifi >= mrt->maxvif)
1628 			return -EINVAL;
1629 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1630 		read_lock(&mrt_lock);
1631 		vif = &mrt->vif_table[vr.vifi];
1632 		if (VIF_EXISTS(mrt, vr.vifi)) {
1633 			vr.icount = vif->pkt_in;
1634 			vr.ocount = vif->pkt_out;
1635 			vr.ibytes = vif->bytes_in;
1636 			vr.obytes = vif->bytes_out;
1637 			read_unlock(&mrt_lock);
1638 
1639 			if (copy_to_user(arg, &vr, sizeof(vr)))
1640 				return -EFAULT;
1641 			return 0;
1642 		}
1643 		read_unlock(&mrt_lock);
1644 		return -EADDRNOTAVAIL;
1645 	case SIOCGETSGCNT:
1646 		if (copy_from_user(&sr, arg, sizeof(sr)))
1647 			return -EFAULT;
1648 
1649 		rcu_read_lock();
1650 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1651 		if (c) {
1652 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1653 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1654 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1655 			rcu_read_unlock();
1656 
1657 			if (copy_to_user(arg, &sr, sizeof(sr)))
1658 				return -EFAULT;
1659 			return 0;
1660 		}
1661 		rcu_read_unlock();
1662 		return -EADDRNOTAVAIL;
1663 	default:
1664 		return -ENOIOCTLCMD;
1665 	}
1666 }
1667 
1668 #ifdef CONFIG_COMPAT
1669 struct compat_sioc_sg_req {
1670 	struct in_addr src;
1671 	struct in_addr grp;
1672 	compat_ulong_t pktcnt;
1673 	compat_ulong_t bytecnt;
1674 	compat_ulong_t wrong_if;
1675 };
1676 
1677 struct compat_sioc_vif_req {
1678 	vifi_t	vifi;		/* Which iface */
1679 	compat_ulong_t icount;
1680 	compat_ulong_t ocount;
1681 	compat_ulong_t ibytes;
1682 	compat_ulong_t obytes;
1683 };
1684 
1685 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1686 {
1687 	struct compat_sioc_sg_req sr;
1688 	struct compat_sioc_vif_req vr;
1689 	struct vif_device *vif;
1690 	struct mfc_cache *c;
1691 	struct net *net = sock_net(sk);
1692 	struct mr_table *mrt;
1693 
1694 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1695 	if (!mrt)
1696 		return -ENOENT;
1697 
1698 	switch (cmd) {
1699 	case SIOCGETVIFCNT:
1700 		if (copy_from_user(&vr, arg, sizeof(vr)))
1701 			return -EFAULT;
1702 		if (vr.vifi >= mrt->maxvif)
1703 			return -EINVAL;
1704 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1705 		read_lock(&mrt_lock);
1706 		vif = &mrt->vif_table[vr.vifi];
1707 		if (VIF_EXISTS(mrt, vr.vifi)) {
1708 			vr.icount = vif->pkt_in;
1709 			vr.ocount = vif->pkt_out;
1710 			vr.ibytes = vif->bytes_in;
1711 			vr.obytes = vif->bytes_out;
1712 			read_unlock(&mrt_lock);
1713 
1714 			if (copy_to_user(arg, &vr, sizeof(vr)))
1715 				return -EFAULT;
1716 			return 0;
1717 		}
1718 		read_unlock(&mrt_lock);
1719 		return -EADDRNOTAVAIL;
1720 	case SIOCGETSGCNT:
1721 		if (copy_from_user(&sr, arg, sizeof(sr)))
1722 			return -EFAULT;
1723 
1724 		rcu_read_lock();
1725 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1726 		if (c) {
1727 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1728 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1729 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1730 			rcu_read_unlock();
1731 
1732 			if (copy_to_user(arg, &sr, sizeof(sr)))
1733 				return -EFAULT;
1734 			return 0;
1735 		}
1736 		rcu_read_unlock();
1737 		return -EADDRNOTAVAIL;
1738 	default:
1739 		return -ENOIOCTLCMD;
1740 	}
1741 }
1742 #endif
1743 
1744 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1745 {
1746 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1747 	struct net *net = dev_net(dev);
1748 	struct mr_table *mrt;
1749 	struct vif_device *v;
1750 	int ct;
1751 
1752 	if (event != NETDEV_UNREGISTER)
1753 		return NOTIFY_DONE;
1754 
1755 	ipmr_for_each_table(mrt, net) {
1756 		v = &mrt->vif_table[0];
1757 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1758 			if (v->dev == dev)
1759 				vif_delete(mrt, ct, 1, NULL);
1760 		}
1761 	}
1762 	return NOTIFY_DONE;
1763 }
1764 
1765 static struct notifier_block ip_mr_notifier = {
1766 	.notifier_call = ipmr_device_event,
1767 };
1768 
1769 /* Encapsulate a packet by attaching a valid IPIP header to it.
1770  * This avoids tunnel drivers and other mess and gives us the speed so
1771  * important for multicast video.
1772  */
1773 static void ip_encap(struct net *net, struct sk_buff *skb,
1774 		     __be32 saddr, __be32 daddr)
1775 {
1776 	struct iphdr *iph;
1777 	const struct iphdr *old_iph = ip_hdr(skb);
1778 
1779 	skb_push(skb, sizeof(struct iphdr));
1780 	skb->transport_header = skb->network_header;
1781 	skb_reset_network_header(skb);
1782 	iph = ip_hdr(skb);
1783 
1784 	iph->version	=	4;
1785 	iph->tos	=	old_iph->tos;
1786 	iph->ttl	=	old_iph->ttl;
1787 	iph->frag_off	=	0;
1788 	iph->daddr	=	daddr;
1789 	iph->saddr	=	saddr;
1790 	iph->protocol	=	IPPROTO_IPIP;
1791 	iph->ihl	=	5;
1792 	iph->tot_len	=	htons(skb->len);
1793 	ip_select_ident(net, skb, NULL);
1794 	ip_send_check(iph);
1795 
1796 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1797 	nf_reset(skb);
1798 }
1799 
1800 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1801 				      struct sk_buff *skb)
1802 {
1803 	struct ip_options *opt = &(IPCB(skb)->opt);
1804 
1805 	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1806 	IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1807 
1808 	if (unlikely(opt->optlen))
1809 		ip_forward_options(skb);
1810 
1811 	return dst_output(net, sk, skb);
1812 }
1813 
1814 #ifdef CONFIG_NET_SWITCHDEV
1815 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1816 				   int in_vifi, int out_vifi)
1817 {
1818 	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1819 	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1820 
1821 	if (!skb->offload_l3_fwd_mark)
1822 		return false;
1823 	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1824 		return false;
1825 	return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1826 					&in_vif->dev_parent_id);
1827 }
1828 #else
1829 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1830 				   int in_vifi, int out_vifi)
1831 {
1832 	return false;
1833 }
1834 #endif
1835 
1836 /* Processing handlers for ipmr_forward */
1837 
1838 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1839 			    int in_vifi, struct sk_buff *skb, int vifi)
1840 {
1841 	const struct iphdr *iph = ip_hdr(skb);
1842 	struct vif_device *vif = &mrt->vif_table[vifi];
1843 	struct net_device *dev;
1844 	struct rtable *rt;
1845 	struct flowi4 fl4;
1846 	int    encap = 0;
1847 
1848 	if (!vif->dev)
1849 		goto out_free;
1850 
1851 	if (vif->flags & VIFF_REGISTER) {
1852 		vif->pkt_out++;
1853 		vif->bytes_out += skb->len;
1854 		vif->dev->stats.tx_bytes += skb->len;
1855 		vif->dev->stats.tx_packets++;
1856 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1857 		goto out_free;
1858 	}
1859 
1860 	if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1861 		goto out_free;
1862 
1863 	if (vif->flags & VIFF_TUNNEL) {
1864 		rt = ip_route_output_ports(net, &fl4, NULL,
1865 					   vif->remote, vif->local,
1866 					   0, 0,
1867 					   IPPROTO_IPIP,
1868 					   RT_TOS(iph->tos), vif->link);
1869 		if (IS_ERR(rt))
1870 			goto out_free;
1871 		encap = sizeof(struct iphdr);
1872 	} else {
1873 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1874 					   0, 0,
1875 					   IPPROTO_IPIP,
1876 					   RT_TOS(iph->tos), vif->link);
1877 		if (IS_ERR(rt))
1878 			goto out_free;
1879 	}
1880 
1881 	dev = rt->dst.dev;
1882 
1883 	if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1884 		/* Do not fragment multicasts. Alas, IPv4 does not
1885 		 * allow to send ICMP, so that packets will disappear
1886 		 * to blackhole.
1887 		 */
1888 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1889 		ip_rt_put(rt);
1890 		goto out_free;
1891 	}
1892 
1893 	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1894 
1895 	if (skb_cow(skb, encap)) {
1896 		ip_rt_put(rt);
1897 		goto out_free;
1898 	}
1899 
1900 	vif->pkt_out++;
1901 	vif->bytes_out += skb->len;
1902 
1903 	skb_dst_drop(skb);
1904 	skb_dst_set(skb, &rt->dst);
1905 	ip_decrease_ttl(ip_hdr(skb));
1906 
1907 	/* FIXME: forward and output firewalls used to be called here.
1908 	 * What do we do with netfilter? -- RR
1909 	 */
1910 	if (vif->flags & VIFF_TUNNEL) {
1911 		ip_encap(net, skb, vif->local, vif->remote);
1912 		/* FIXME: extra output firewall step used to be here. --RR */
1913 		vif->dev->stats.tx_packets++;
1914 		vif->dev->stats.tx_bytes += skb->len;
1915 	}
1916 
1917 	IPCB(skb)->flags |= IPSKB_FORWARDED;
1918 
1919 	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1920 	 * not only before forwarding, but after forwarding on all output
1921 	 * interfaces. It is clear, if mrouter runs a multicasting
1922 	 * program, it should receive packets not depending to what interface
1923 	 * program is joined.
1924 	 * If we will not make it, the program will have to join on all
1925 	 * interfaces. On the other hand, multihoming host (or router, but
1926 	 * not mrouter) cannot join to more than one interface - it will
1927 	 * result in receiving multiple packets.
1928 	 */
1929 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1930 		net, NULL, skb, skb->dev, dev,
1931 		ipmr_forward_finish);
1932 	return;
1933 
1934 out_free:
1935 	kfree_skb(skb);
1936 }
1937 
1938 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1939 {
1940 	int ct;
1941 
1942 	for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1943 		if (mrt->vif_table[ct].dev == dev)
1944 			break;
1945 	}
1946 	return ct;
1947 }
1948 
1949 /* "local" means that we should preserve one skb (for local delivery) */
1950 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1951 			  struct net_device *dev, struct sk_buff *skb,
1952 			  struct mfc_cache *c, int local)
1953 {
1954 	int true_vifi = ipmr_find_vif(mrt, dev);
1955 	int psend = -1;
1956 	int vif, ct;
1957 
1958 	vif = c->_c.mfc_parent;
1959 	c->_c.mfc_un.res.pkt++;
1960 	c->_c.mfc_un.res.bytes += skb->len;
1961 	c->_c.mfc_un.res.lastuse = jiffies;
1962 
1963 	if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1964 		struct mfc_cache *cache_proxy;
1965 
1966 		/* For an (*,G) entry, we only check that the incomming
1967 		 * interface is part of the static tree.
1968 		 */
1969 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1970 		if (cache_proxy &&
1971 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1972 			goto forward;
1973 	}
1974 
1975 	/* Wrong interface: drop packet and (maybe) send PIM assert. */
1976 	if (mrt->vif_table[vif].dev != dev) {
1977 		if (rt_is_output_route(skb_rtable(skb))) {
1978 			/* It is our own packet, looped back.
1979 			 * Very complicated situation...
1980 			 *
1981 			 * The best workaround until routing daemons will be
1982 			 * fixed is not to redistribute packet, if it was
1983 			 * send through wrong interface. It means, that
1984 			 * multicast applications WILL NOT work for
1985 			 * (S,G), which have default multicast route pointing
1986 			 * to wrong oif. In any case, it is not a good
1987 			 * idea to use multicasting applications on router.
1988 			 */
1989 			goto dont_forward;
1990 		}
1991 
1992 		c->_c.mfc_un.res.wrong_if++;
1993 
1994 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
1995 		    /* pimsm uses asserts, when switching from RPT to SPT,
1996 		     * so that we cannot check that packet arrived on an oif.
1997 		     * It is bad, but otherwise we would need to move pretty
1998 		     * large chunk of pimd to kernel. Ough... --ANK
1999 		     */
2000 		    (mrt->mroute_do_pim ||
2001 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2002 		    time_after(jiffies,
2003 			       c->_c.mfc_un.res.last_assert +
2004 			       MFC_ASSERT_THRESH)) {
2005 			c->_c.mfc_un.res.last_assert = jiffies;
2006 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2007 			if (mrt->mroute_do_wrvifwhole)
2008 				ipmr_cache_report(mrt, skb, true_vifi,
2009 						  IGMPMSG_WRVIFWHOLE);
2010 		}
2011 		goto dont_forward;
2012 	}
2013 
2014 forward:
2015 	mrt->vif_table[vif].pkt_in++;
2016 	mrt->vif_table[vif].bytes_in += skb->len;
2017 
2018 	/* Forward the frame */
2019 	if (c->mfc_origin == htonl(INADDR_ANY) &&
2020 	    c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2021 		if (true_vifi >= 0 &&
2022 		    true_vifi != c->_c.mfc_parent &&
2023 		    ip_hdr(skb)->ttl >
2024 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2025 			/* It's an (*,*) entry and the packet is not coming from
2026 			 * the upstream: forward the packet to the upstream
2027 			 * only.
2028 			 */
2029 			psend = c->_c.mfc_parent;
2030 			goto last_forward;
2031 		}
2032 		goto dont_forward;
2033 	}
2034 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2035 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2036 		/* For (*,G) entry, don't forward to the incoming interface */
2037 		if ((c->mfc_origin != htonl(INADDR_ANY) ||
2038 		     ct != true_vifi) &&
2039 		    ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2040 			if (psend != -1) {
2041 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2042 
2043 				if (skb2)
2044 					ipmr_queue_xmit(net, mrt, true_vifi,
2045 							skb2, psend);
2046 			}
2047 			psend = ct;
2048 		}
2049 	}
2050 last_forward:
2051 	if (psend != -1) {
2052 		if (local) {
2053 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2054 
2055 			if (skb2)
2056 				ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2057 						psend);
2058 		} else {
2059 			ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2060 			return;
2061 		}
2062 	}
2063 
2064 dont_forward:
2065 	if (!local)
2066 		kfree_skb(skb);
2067 }
2068 
2069 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2070 {
2071 	struct rtable *rt = skb_rtable(skb);
2072 	struct iphdr *iph = ip_hdr(skb);
2073 	struct flowi4 fl4 = {
2074 		.daddr = iph->daddr,
2075 		.saddr = iph->saddr,
2076 		.flowi4_tos = RT_TOS(iph->tos),
2077 		.flowi4_oif = (rt_is_output_route(rt) ?
2078 			       skb->dev->ifindex : 0),
2079 		.flowi4_iif = (rt_is_output_route(rt) ?
2080 			       LOOPBACK_IFINDEX :
2081 			       skb->dev->ifindex),
2082 		.flowi4_mark = skb->mark,
2083 	};
2084 	struct mr_table *mrt;
2085 	int err;
2086 
2087 	err = ipmr_fib_lookup(net, &fl4, &mrt);
2088 	if (err)
2089 		return ERR_PTR(err);
2090 	return mrt;
2091 }
2092 
2093 /* Multicast packets for forwarding arrive here
2094  * Called with rcu_read_lock();
2095  */
2096 int ip_mr_input(struct sk_buff *skb)
2097 {
2098 	struct mfc_cache *cache;
2099 	struct net *net = dev_net(skb->dev);
2100 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2101 	struct mr_table *mrt;
2102 	struct net_device *dev;
2103 
2104 	/* skb->dev passed in is the loX master dev for vrfs.
2105 	 * As there are no vifs associated with loopback devices,
2106 	 * get the proper interface that does have a vif associated with it.
2107 	 */
2108 	dev = skb->dev;
2109 	if (netif_is_l3_master(skb->dev)) {
2110 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2111 		if (!dev) {
2112 			kfree_skb(skb);
2113 			return -ENODEV;
2114 		}
2115 	}
2116 
2117 	/* Packet is looped back after forward, it should not be
2118 	 * forwarded second time, but still can be delivered locally.
2119 	 */
2120 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2121 		goto dont_forward;
2122 
2123 	mrt = ipmr_rt_fib_lookup(net, skb);
2124 	if (IS_ERR(mrt)) {
2125 		kfree_skb(skb);
2126 		return PTR_ERR(mrt);
2127 	}
2128 	if (!local) {
2129 		if (IPCB(skb)->opt.router_alert) {
2130 			if (ip_call_ra_chain(skb))
2131 				return 0;
2132 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2133 			/* IGMPv1 (and broken IGMPv2 implementations sort of
2134 			 * Cisco IOS <= 11.2(8)) do not put router alert
2135 			 * option to IGMP packets destined to routable
2136 			 * groups. It is very bad, because it means
2137 			 * that we can forward NO IGMP messages.
2138 			 */
2139 			struct sock *mroute_sk;
2140 
2141 			mroute_sk = rcu_dereference(mrt->mroute_sk);
2142 			if (mroute_sk) {
2143 				nf_reset(skb);
2144 				raw_rcv(mroute_sk, skb);
2145 				return 0;
2146 			}
2147 		    }
2148 	}
2149 
2150 	/* already under rcu_read_lock() */
2151 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2152 	if (!cache) {
2153 		int vif = ipmr_find_vif(mrt, dev);
2154 
2155 		if (vif >= 0)
2156 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2157 						    vif);
2158 	}
2159 
2160 	/* No usable cache entry */
2161 	if (!cache) {
2162 		int vif;
2163 
2164 		if (local) {
2165 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2166 			ip_local_deliver(skb);
2167 			if (!skb2)
2168 				return -ENOBUFS;
2169 			skb = skb2;
2170 		}
2171 
2172 		read_lock(&mrt_lock);
2173 		vif = ipmr_find_vif(mrt, dev);
2174 		if (vif >= 0) {
2175 			int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2176 			read_unlock(&mrt_lock);
2177 
2178 			return err2;
2179 		}
2180 		read_unlock(&mrt_lock);
2181 		kfree_skb(skb);
2182 		return -ENODEV;
2183 	}
2184 
2185 	read_lock(&mrt_lock);
2186 	ip_mr_forward(net, mrt, dev, skb, cache, local);
2187 	read_unlock(&mrt_lock);
2188 
2189 	if (local)
2190 		return ip_local_deliver(skb);
2191 
2192 	return 0;
2193 
2194 dont_forward:
2195 	if (local)
2196 		return ip_local_deliver(skb);
2197 	kfree_skb(skb);
2198 	return 0;
2199 }
2200 
2201 #ifdef CONFIG_IP_PIMSM_V1
2202 /* Handle IGMP messages of PIMv1 */
2203 int pim_rcv_v1(struct sk_buff *skb)
2204 {
2205 	struct igmphdr *pim;
2206 	struct net *net = dev_net(skb->dev);
2207 	struct mr_table *mrt;
2208 
2209 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2210 		goto drop;
2211 
2212 	pim = igmp_hdr(skb);
2213 
2214 	mrt = ipmr_rt_fib_lookup(net, skb);
2215 	if (IS_ERR(mrt))
2216 		goto drop;
2217 	if (!mrt->mroute_do_pim ||
2218 	    pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2219 		goto drop;
2220 
2221 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2222 drop:
2223 		kfree_skb(skb);
2224 	}
2225 	return 0;
2226 }
2227 #endif
2228 
2229 #ifdef CONFIG_IP_PIMSM_V2
2230 static int pim_rcv(struct sk_buff *skb)
2231 {
2232 	struct pimreghdr *pim;
2233 	struct net *net = dev_net(skb->dev);
2234 	struct mr_table *mrt;
2235 
2236 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2237 		goto drop;
2238 
2239 	pim = (struct pimreghdr *)skb_transport_header(skb);
2240 	if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2241 	    (pim->flags & PIM_NULL_REGISTER) ||
2242 	    (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2243 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2244 		goto drop;
2245 
2246 	mrt = ipmr_rt_fib_lookup(net, skb);
2247 	if (IS_ERR(mrt))
2248 		goto drop;
2249 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2250 drop:
2251 		kfree_skb(skb);
2252 	}
2253 	return 0;
2254 }
2255 #endif
2256 
2257 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2258 		   __be32 saddr, __be32 daddr,
2259 		   struct rtmsg *rtm, u32 portid)
2260 {
2261 	struct mfc_cache *cache;
2262 	struct mr_table *mrt;
2263 	int err;
2264 
2265 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2266 	if (!mrt)
2267 		return -ENOENT;
2268 
2269 	rcu_read_lock();
2270 	cache = ipmr_cache_find(mrt, saddr, daddr);
2271 	if (!cache && skb->dev) {
2272 		int vif = ipmr_find_vif(mrt, skb->dev);
2273 
2274 		if (vif >= 0)
2275 			cache = ipmr_cache_find_any(mrt, daddr, vif);
2276 	}
2277 	if (!cache) {
2278 		struct sk_buff *skb2;
2279 		struct iphdr *iph;
2280 		struct net_device *dev;
2281 		int vif = -1;
2282 
2283 		dev = skb->dev;
2284 		read_lock(&mrt_lock);
2285 		if (dev)
2286 			vif = ipmr_find_vif(mrt, dev);
2287 		if (vif < 0) {
2288 			read_unlock(&mrt_lock);
2289 			rcu_read_unlock();
2290 			return -ENODEV;
2291 		}
2292 		skb2 = skb_clone(skb, GFP_ATOMIC);
2293 		if (!skb2) {
2294 			read_unlock(&mrt_lock);
2295 			rcu_read_unlock();
2296 			return -ENOMEM;
2297 		}
2298 
2299 		NETLINK_CB(skb2).portid = portid;
2300 		skb_push(skb2, sizeof(struct iphdr));
2301 		skb_reset_network_header(skb2);
2302 		iph = ip_hdr(skb2);
2303 		iph->ihl = sizeof(struct iphdr) >> 2;
2304 		iph->saddr = saddr;
2305 		iph->daddr = daddr;
2306 		iph->version = 0;
2307 		err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2308 		read_unlock(&mrt_lock);
2309 		rcu_read_unlock();
2310 		return err;
2311 	}
2312 
2313 	read_lock(&mrt_lock);
2314 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2315 	read_unlock(&mrt_lock);
2316 	rcu_read_unlock();
2317 	return err;
2318 }
2319 
2320 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2321 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2322 			    int flags)
2323 {
2324 	struct nlmsghdr *nlh;
2325 	struct rtmsg *rtm;
2326 	int err;
2327 
2328 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2329 	if (!nlh)
2330 		return -EMSGSIZE;
2331 
2332 	rtm = nlmsg_data(nlh);
2333 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2334 	rtm->rtm_dst_len  = 32;
2335 	rtm->rtm_src_len  = 32;
2336 	rtm->rtm_tos      = 0;
2337 	rtm->rtm_table    = mrt->id;
2338 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2339 		goto nla_put_failure;
2340 	rtm->rtm_type     = RTN_MULTICAST;
2341 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2342 	if (c->_c.mfc_flags & MFC_STATIC)
2343 		rtm->rtm_protocol = RTPROT_STATIC;
2344 	else
2345 		rtm->rtm_protocol = RTPROT_MROUTED;
2346 	rtm->rtm_flags    = 0;
2347 
2348 	if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2349 	    nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2350 		goto nla_put_failure;
2351 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2352 	/* do not break the dump if cache is unresolved */
2353 	if (err < 0 && err != -ENOENT)
2354 		goto nla_put_failure;
2355 
2356 	nlmsg_end(skb, nlh);
2357 	return 0;
2358 
2359 nla_put_failure:
2360 	nlmsg_cancel(skb, nlh);
2361 	return -EMSGSIZE;
2362 }
2363 
2364 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2365 			     u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2366 			     int flags)
2367 {
2368 	return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2369 				cmd, flags);
2370 }
2371 
2372 static size_t mroute_msgsize(bool unresolved, int maxvif)
2373 {
2374 	size_t len =
2375 		NLMSG_ALIGN(sizeof(struct rtmsg))
2376 		+ nla_total_size(4)	/* RTA_TABLE */
2377 		+ nla_total_size(4)	/* RTA_SRC */
2378 		+ nla_total_size(4)	/* RTA_DST */
2379 		;
2380 
2381 	if (!unresolved)
2382 		len = len
2383 		      + nla_total_size(4)	/* RTA_IIF */
2384 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2385 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2386 						/* RTA_MFC_STATS */
2387 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2388 		;
2389 
2390 	return len;
2391 }
2392 
2393 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2394 				 int cmd)
2395 {
2396 	struct net *net = read_pnet(&mrt->net);
2397 	struct sk_buff *skb;
2398 	int err = -ENOBUFS;
2399 
2400 	skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2401 				       mrt->maxvif),
2402 			GFP_ATOMIC);
2403 	if (!skb)
2404 		goto errout;
2405 
2406 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2407 	if (err < 0)
2408 		goto errout;
2409 
2410 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2411 	return;
2412 
2413 errout:
2414 	kfree_skb(skb);
2415 	if (err < 0)
2416 		rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2417 }
2418 
2419 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2420 {
2421 	size_t len =
2422 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2423 		+ nla_total_size(1)	/* IPMRA_CREPORT_MSGTYPE */
2424 		+ nla_total_size(4)	/* IPMRA_CREPORT_VIF_ID */
2425 		+ nla_total_size(4)	/* IPMRA_CREPORT_SRC_ADDR */
2426 		+ nla_total_size(4)	/* IPMRA_CREPORT_DST_ADDR */
2427 					/* IPMRA_CREPORT_PKT */
2428 		+ nla_total_size(payloadlen)
2429 		;
2430 
2431 	return len;
2432 }
2433 
2434 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2435 {
2436 	struct net *net = read_pnet(&mrt->net);
2437 	struct nlmsghdr *nlh;
2438 	struct rtgenmsg *rtgenm;
2439 	struct igmpmsg *msg;
2440 	struct sk_buff *skb;
2441 	struct nlattr *nla;
2442 	int payloadlen;
2443 
2444 	payloadlen = pkt->len - sizeof(struct igmpmsg);
2445 	msg = (struct igmpmsg *)skb_network_header(pkt);
2446 
2447 	skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2448 	if (!skb)
2449 		goto errout;
2450 
2451 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2452 			sizeof(struct rtgenmsg), 0);
2453 	if (!nlh)
2454 		goto errout;
2455 	rtgenm = nlmsg_data(nlh);
2456 	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2457 	if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2458 	    nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2459 	    nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2460 			    msg->im_src.s_addr) ||
2461 	    nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2462 			    msg->im_dst.s_addr))
2463 		goto nla_put_failure;
2464 
2465 	nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2466 	if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2467 				  nla_data(nla), payloadlen))
2468 		goto nla_put_failure;
2469 
2470 	nlmsg_end(skb, nlh);
2471 
2472 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2473 	return;
2474 
2475 nla_put_failure:
2476 	nlmsg_cancel(skb, nlh);
2477 errout:
2478 	kfree_skb(skb);
2479 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2480 }
2481 
2482 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2483 				       const struct nlmsghdr *nlh,
2484 				       struct nlattr **tb,
2485 				       struct netlink_ext_ack *extack)
2486 {
2487 	struct rtmsg *rtm;
2488 	int i, err;
2489 
2490 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2491 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2492 		return -EINVAL;
2493 	}
2494 
2495 	if (!netlink_strict_get_check(skb))
2496 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2497 					      rtm_ipv4_policy, extack);
2498 
2499 	rtm = nlmsg_data(nlh);
2500 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2501 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2502 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2503 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2504 		NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2505 		return -EINVAL;
2506 	}
2507 
2508 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2509 					    rtm_ipv4_policy, extack);
2510 	if (err)
2511 		return err;
2512 
2513 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2514 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2515 		NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2516 		return -EINVAL;
2517 	}
2518 
2519 	for (i = 0; i <= RTA_MAX; i++) {
2520 		if (!tb[i])
2521 			continue;
2522 
2523 		switch (i) {
2524 		case RTA_SRC:
2525 		case RTA_DST:
2526 		case RTA_TABLE:
2527 			break;
2528 		default:
2529 			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2530 			return -EINVAL;
2531 		}
2532 	}
2533 
2534 	return 0;
2535 }
2536 
2537 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2538 			     struct netlink_ext_ack *extack)
2539 {
2540 	struct net *net = sock_net(in_skb->sk);
2541 	struct nlattr *tb[RTA_MAX + 1];
2542 	struct sk_buff *skb = NULL;
2543 	struct mfc_cache *cache;
2544 	struct mr_table *mrt;
2545 	__be32 src, grp;
2546 	u32 tableid;
2547 	int err;
2548 
2549 	err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2550 	if (err < 0)
2551 		goto errout;
2552 
2553 	src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2554 	grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2555 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2556 
2557 	mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2558 	if (!mrt) {
2559 		err = -ENOENT;
2560 		goto errout_free;
2561 	}
2562 
2563 	/* entries are added/deleted only under RTNL */
2564 	rcu_read_lock();
2565 	cache = ipmr_cache_find(mrt, src, grp);
2566 	rcu_read_unlock();
2567 	if (!cache) {
2568 		err = -ENOENT;
2569 		goto errout_free;
2570 	}
2571 
2572 	skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2573 	if (!skb) {
2574 		err = -ENOBUFS;
2575 		goto errout_free;
2576 	}
2577 
2578 	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2579 			       nlh->nlmsg_seq, cache,
2580 			       RTM_NEWROUTE, 0);
2581 	if (err < 0)
2582 		goto errout_free;
2583 
2584 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2585 
2586 errout:
2587 	return err;
2588 
2589 errout_free:
2590 	kfree_skb(skb);
2591 	goto errout;
2592 }
2593 
2594 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2595 {
2596 	struct fib_dump_filter filter = {};
2597 	int err;
2598 
2599 	if (cb->strict_check) {
2600 		err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2601 					    &filter, cb);
2602 		if (err < 0)
2603 			return err;
2604 	}
2605 
2606 	if (filter.table_id) {
2607 		struct mr_table *mrt;
2608 
2609 		mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2610 		if (!mrt) {
2611 			if (filter.dump_all_families)
2612 				return skb->len;
2613 
2614 			NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2615 			return -ENOENT;
2616 		}
2617 		err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2618 				    &mfc_unres_lock, &filter);
2619 		return skb->len ? : err;
2620 	}
2621 
2622 	return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2623 				_ipmr_fill_mroute, &mfc_unres_lock, &filter);
2624 }
2625 
2626 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2627 	[RTA_SRC]	= { .type = NLA_U32 },
2628 	[RTA_DST]	= { .type = NLA_U32 },
2629 	[RTA_IIF]	= { .type = NLA_U32 },
2630 	[RTA_TABLE]	= { .type = NLA_U32 },
2631 	[RTA_MULTIPATH]	= { .len = sizeof(struct rtnexthop) },
2632 };
2633 
2634 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2635 {
2636 	switch (rtm_protocol) {
2637 	case RTPROT_STATIC:
2638 	case RTPROT_MROUTED:
2639 		return true;
2640 	}
2641 	return false;
2642 }
2643 
2644 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2645 {
2646 	struct rtnexthop *rtnh = nla_data(nla);
2647 	int remaining = nla_len(nla), vifi = 0;
2648 
2649 	while (rtnh_ok(rtnh, remaining)) {
2650 		mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2651 		if (++vifi == MAXVIFS)
2652 			break;
2653 		rtnh = rtnh_next(rtnh, &remaining);
2654 	}
2655 
2656 	return remaining > 0 ? -EINVAL : vifi;
2657 }
2658 
2659 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2660 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2661 			    struct mfcctl *mfcc, int *mrtsock,
2662 			    struct mr_table **mrtret,
2663 			    struct netlink_ext_ack *extack)
2664 {
2665 	struct net_device *dev = NULL;
2666 	u32 tblid = RT_TABLE_DEFAULT;
2667 	struct mr_table *mrt;
2668 	struct nlattr *attr;
2669 	struct rtmsg *rtm;
2670 	int ret, rem;
2671 
2672 	ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2673 					rtm_ipmr_policy, extack);
2674 	if (ret < 0)
2675 		goto out;
2676 	rtm = nlmsg_data(nlh);
2677 
2678 	ret = -EINVAL;
2679 	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2680 	    rtm->rtm_type != RTN_MULTICAST ||
2681 	    rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2682 	    !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2683 		goto out;
2684 
2685 	memset(mfcc, 0, sizeof(*mfcc));
2686 	mfcc->mfcc_parent = -1;
2687 	ret = 0;
2688 	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2689 		switch (nla_type(attr)) {
2690 		case RTA_SRC:
2691 			mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2692 			break;
2693 		case RTA_DST:
2694 			mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2695 			break;
2696 		case RTA_IIF:
2697 			dev = __dev_get_by_index(net, nla_get_u32(attr));
2698 			if (!dev) {
2699 				ret = -ENODEV;
2700 				goto out;
2701 			}
2702 			break;
2703 		case RTA_MULTIPATH:
2704 			if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2705 				ret = -EINVAL;
2706 				goto out;
2707 			}
2708 			break;
2709 		case RTA_PREFSRC:
2710 			ret = 1;
2711 			break;
2712 		case RTA_TABLE:
2713 			tblid = nla_get_u32(attr);
2714 			break;
2715 		}
2716 	}
2717 	mrt = ipmr_get_table(net, tblid);
2718 	if (!mrt) {
2719 		ret = -ENOENT;
2720 		goto out;
2721 	}
2722 	*mrtret = mrt;
2723 	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2724 	if (dev)
2725 		mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2726 
2727 out:
2728 	return ret;
2729 }
2730 
2731 /* takes care of both newroute and delroute */
2732 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2733 			  struct netlink_ext_ack *extack)
2734 {
2735 	struct net *net = sock_net(skb->sk);
2736 	int ret, mrtsock, parent;
2737 	struct mr_table *tbl;
2738 	struct mfcctl mfcc;
2739 
2740 	mrtsock = 0;
2741 	tbl = NULL;
2742 	ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2743 	if (ret < 0)
2744 		return ret;
2745 
2746 	parent = ret ? mfcc.mfcc_parent : -1;
2747 	if (nlh->nlmsg_type == RTM_NEWROUTE)
2748 		return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2749 	else
2750 		return ipmr_mfc_delete(tbl, &mfcc, parent);
2751 }
2752 
2753 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2754 {
2755 	u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2756 
2757 	if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2758 	    nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2759 	    nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2760 			mrt->mroute_reg_vif_num) ||
2761 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2762 		       mrt->mroute_do_assert) ||
2763 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2764 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2765 		       mrt->mroute_do_wrvifwhole))
2766 		return false;
2767 
2768 	return true;
2769 }
2770 
2771 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2772 {
2773 	struct nlattr *vif_nest;
2774 	struct vif_device *vif;
2775 
2776 	/* if the VIF doesn't exist just continue */
2777 	if (!VIF_EXISTS(mrt, vifid))
2778 		return true;
2779 
2780 	vif = &mrt->vif_table[vifid];
2781 	vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2782 	if (!vif_nest)
2783 		return false;
2784 	if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2785 	    nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2786 	    nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2787 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2788 			      IPMRA_VIFA_PAD) ||
2789 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2790 			      IPMRA_VIFA_PAD) ||
2791 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2792 			      IPMRA_VIFA_PAD) ||
2793 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2794 			      IPMRA_VIFA_PAD) ||
2795 	    nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2796 	    nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2797 		nla_nest_cancel(skb, vif_nest);
2798 		return false;
2799 	}
2800 	nla_nest_end(skb, vif_nest);
2801 
2802 	return true;
2803 }
2804 
2805 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2806 			       struct netlink_ext_ack *extack)
2807 {
2808 	struct ifinfomsg *ifm;
2809 
2810 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2811 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2812 		return -EINVAL;
2813 	}
2814 
2815 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2816 		NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2817 		return -EINVAL;
2818 	}
2819 
2820 	ifm = nlmsg_data(nlh);
2821 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2822 	    ifm->ifi_change || ifm->ifi_index) {
2823 		NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2824 		return -EINVAL;
2825 	}
2826 
2827 	return 0;
2828 }
2829 
2830 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2831 {
2832 	struct net *net = sock_net(skb->sk);
2833 	struct nlmsghdr *nlh = NULL;
2834 	unsigned int t = 0, s_t;
2835 	unsigned int e = 0, s_e;
2836 	struct mr_table *mrt;
2837 
2838 	if (cb->strict_check) {
2839 		int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2840 
2841 		if (err < 0)
2842 			return err;
2843 	}
2844 
2845 	s_t = cb->args[0];
2846 	s_e = cb->args[1];
2847 
2848 	ipmr_for_each_table(mrt, net) {
2849 		struct nlattr *vifs, *af;
2850 		struct ifinfomsg *hdr;
2851 		u32 i;
2852 
2853 		if (t < s_t)
2854 			goto skip_table;
2855 		nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2856 				cb->nlh->nlmsg_seq, RTM_NEWLINK,
2857 				sizeof(*hdr), NLM_F_MULTI);
2858 		if (!nlh)
2859 			break;
2860 
2861 		hdr = nlmsg_data(nlh);
2862 		memset(hdr, 0, sizeof(*hdr));
2863 		hdr->ifi_family = RTNL_FAMILY_IPMR;
2864 
2865 		af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2866 		if (!af) {
2867 			nlmsg_cancel(skb, nlh);
2868 			goto out;
2869 		}
2870 
2871 		if (!ipmr_fill_table(mrt, skb)) {
2872 			nlmsg_cancel(skb, nlh);
2873 			goto out;
2874 		}
2875 
2876 		vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2877 		if (!vifs) {
2878 			nla_nest_end(skb, af);
2879 			nlmsg_end(skb, nlh);
2880 			goto out;
2881 		}
2882 		for (i = 0; i < mrt->maxvif; i++) {
2883 			if (e < s_e)
2884 				goto skip_entry;
2885 			if (!ipmr_fill_vif(mrt, i, skb)) {
2886 				nla_nest_end(skb, vifs);
2887 				nla_nest_end(skb, af);
2888 				nlmsg_end(skb, nlh);
2889 				goto out;
2890 			}
2891 skip_entry:
2892 			e++;
2893 		}
2894 		s_e = 0;
2895 		e = 0;
2896 		nla_nest_end(skb, vifs);
2897 		nla_nest_end(skb, af);
2898 		nlmsg_end(skb, nlh);
2899 skip_table:
2900 		t++;
2901 	}
2902 
2903 out:
2904 	cb->args[1] = e;
2905 	cb->args[0] = t;
2906 
2907 	return skb->len;
2908 }
2909 
2910 #ifdef CONFIG_PROC_FS
2911 /* The /proc interfaces to multicast routing :
2912  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2913  */
2914 
2915 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2916 	__acquires(mrt_lock)
2917 {
2918 	struct mr_vif_iter *iter = seq->private;
2919 	struct net *net = seq_file_net(seq);
2920 	struct mr_table *mrt;
2921 
2922 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2923 	if (!mrt)
2924 		return ERR_PTR(-ENOENT);
2925 
2926 	iter->mrt = mrt;
2927 
2928 	read_lock(&mrt_lock);
2929 	return mr_vif_seq_start(seq, pos);
2930 }
2931 
2932 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2933 	__releases(mrt_lock)
2934 {
2935 	read_unlock(&mrt_lock);
2936 }
2937 
2938 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2939 {
2940 	struct mr_vif_iter *iter = seq->private;
2941 	struct mr_table *mrt = iter->mrt;
2942 
2943 	if (v == SEQ_START_TOKEN) {
2944 		seq_puts(seq,
2945 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2946 	} else {
2947 		const struct vif_device *vif = v;
2948 		const char *name =  vif->dev ?
2949 				    vif->dev->name : "none";
2950 
2951 		seq_printf(seq,
2952 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2953 			   vif - mrt->vif_table,
2954 			   name, vif->bytes_in, vif->pkt_in,
2955 			   vif->bytes_out, vif->pkt_out,
2956 			   vif->flags, vif->local, vif->remote);
2957 	}
2958 	return 0;
2959 }
2960 
2961 static const struct seq_operations ipmr_vif_seq_ops = {
2962 	.start = ipmr_vif_seq_start,
2963 	.next  = mr_vif_seq_next,
2964 	.stop  = ipmr_vif_seq_stop,
2965 	.show  = ipmr_vif_seq_show,
2966 };
2967 
2968 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2969 {
2970 	struct net *net = seq_file_net(seq);
2971 	struct mr_table *mrt;
2972 
2973 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2974 	if (!mrt)
2975 		return ERR_PTR(-ENOENT);
2976 
2977 	return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2978 }
2979 
2980 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2981 {
2982 	int n;
2983 
2984 	if (v == SEQ_START_TOKEN) {
2985 		seq_puts(seq,
2986 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2987 	} else {
2988 		const struct mfc_cache *mfc = v;
2989 		const struct mr_mfc_iter *it = seq->private;
2990 		const struct mr_table *mrt = it->mrt;
2991 
2992 		seq_printf(seq, "%08X %08X %-3hd",
2993 			   (__force u32) mfc->mfc_mcastgrp,
2994 			   (__force u32) mfc->mfc_origin,
2995 			   mfc->_c.mfc_parent);
2996 
2997 		if (it->cache != &mrt->mfc_unres_queue) {
2998 			seq_printf(seq, " %8lu %8lu %8lu",
2999 				   mfc->_c.mfc_un.res.pkt,
3000 				   mfc->_c.mfc_un.res.bytes,
3001 				   mfc->_c.mfc_un.res.wrong_if);
3002 			for (n = mfc->_c.mfc_un.res.minvif;
3003 			     n < mfc->_c.mfc_un.res.maxvif; n++) {
3004 				if (VIF_EXISTS(mrt, n) &&
3005 				    mfc->_c.mfc_un.res.ttls[n] < 255)
3006 					seq_printf(seq,
3007 					   " %2d:%-3d",
3008 					   n, mfc->_c.mfc_un.res.ttls[n]);
3009 			}
3010 		} else {
3011 			/* unresolved mfc_caches don't contain
3012 			 * pkt, bytes and wrong_if values
3013 			 */
3014 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3015 		}
3016 		seq_putc(seq, '\n');
3017 	}
3018 	return 0;
3019 }
3020 
3021 static const struct seq_operations ipmr_mfc_seq_ops = {
3022 	.start = ipmr_mfc_seq_start,
3023 	.next  = mr_mfc_seq_next,
3024 	.stop  = mr_mfc_seq_stop,
3025 	.show  = ipmr_mfc_seq_show,
3026 };
3027 #endif
3028 
3029 #ifdef CONFIG_IP_PIMSM_V2
3030 static const struct net_protocol pim_protocol = {
3031 	.handler	=	pim_rcv,
3032 	.netns_ok	=	1,
3033 };
3034 #endif
3035 
3036 static unsigned int ipmr_seq_read(struct net *net)
3037 {
3038 	ASSERT_RTNL();
3039 
3040 	return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3041 }
3042 
3043 static int ipmr_dump(struct net *net, struct notifier_block *nb)
3044 {
3045 	return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3046 		       ipmr_mr_table_iter, &mrt_lock);
3047 }
3048 
3049 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3050 	.family		= RTNL_FAMILY_IPMR,
3051 	.fib_seq_read	= ipmr_seq_read,
3052 	.fib_dump	= ipmr_dump,
3053 	.owner		= THIS_MODULE,
3054 };
3055 
3056 static int __net_init ipmr_notifier_init(struct net *net)
3057 {
3058 	struct fib_notifier_ops *ops;
3059 
3060 	net->ipv4.ipmr_seq = 0;
3061 
3062 	ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3063 	if (IS_ERR(ops))
3064 		return PTR_ERR(ops);
3065 	net->ipv4.ipmr_notifier_ops = ops;
3066 
3067 	return 0;
3068 }
3069 
3070 static void __net_exit ipmr_notifier_exit(struct net *net)
3071 {
3072 	fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3073 	net->ipv4.ipmr_notifier_ops = NULL;
3074 }
3075 
3076 /* Setup for IP multicast routing */
3077 static int __net_init ipmr_net_init(struct net *net)
3078 {
3079 	int err;
3080 
3081 	err = ipmr_notifier_init(net);
3082 	if (err)
3083 		goto ipmr_notifier_fail;
3084 
3085 	err = ipmr_rules_init(net);
3086 	if (err < 0)
3087 		goto ipmr_rules_fail;
3088 
3089 #ifdef CONFIG_PROC_FS
3090 	err = -ENOMEM;
3091 	if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3092 			sizeof(struct mr_vif_iter)))
3093 		goto proc_vif_fail;
3094 	if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3095 			sizeof(struct mr_mfc_iter)))
3096 		goto proc_cache_fail;
3097 #endif
3098 	return 0;
3099 
3100 #ifdef CONFIG_PROC_FS
3101 proc_cache_fail:
3102 	remove_proc_entry("ip_mr_vif", net->proc_net);
3103 proc_vif_fail:
3104 	ipmr_rules_exit(net);
3105 #endif
3106 ipmr_rules_fail:
3107 	ipmr_notifier_exit(net);
3108 ipmr_notifier_fail:
3109 	return err;
3110 }
3111 
3112 static void __net_exit ipmr_net_exit(struct net *net)
3113 {
3114 #ifdef CONFIG_PROC_FS
3115 	remove_proc_entry("ip_mr_cache", net->proc_net);
3116 	remove_proc_entry("ip_mr_vif", net->proc_net);
3117 #endif
3118 	ipmr_notifier_exit(net);
3119 	ipmr_rules_exit(net);
3120 }
3121 
3122 static struct pernet_operations ipmr_net_ops = {
3123 	.init = ipmr_net_init,
3124 	.exit = ipmr_net_exit,
3125 };
3126 
3127 int __init ip_mr_init(void)
3128 {
3129 	int err;
3130 
3131 	mrt_cachep = kmem_cache_create("ip_mrt_cache",
3132 				       sizeof(struct mfc_cache),
3133 				       0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3134 				       NULL);
3135 
3136 	err = register_pernet_subsys(&ipmr_net_ops);
3137 	if (err)
3138 		goto reg_pernet_fail;
3139 
3140 	err = register_netdevice_notifier(&ip_mr_notifier);
3141 	if (err)
3142 		goto reg_notif_fail;
3143 #ifdef CONFIG_IP_PIMSM_V2
3144 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3145 		pr_err("%s: can't add PIM protocol\n", __func__);
3146 		err = -EAGAIN;
3147 		goto add_proto_fail;
3148 	}
3149 #endif
3150 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3151 		      ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3152 	rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3153 		      ipmr_rtm_route, NULL, 0);
3154 	rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3155 		      ipmr_rtm_route, NULL, 0);
3156 
3157 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3158 		      NULL, ipmr_rtm_dumplink, 0);
3159 	return 0;
3160 
3161 #ifdef CONFIG_IP_PIMSM_V2
3162 add_proto_fail:
3163 	unregister_netdevice_notifier(&ip_mr_notifier);
3164 #endif
3165 reg_notif_fail:
3166 	unregister_pernet_subsys(&ipmr_net_ops);
3167 reg_pernet_fail:
3168 	kmem_cache_destroy(mrt_cachep);
3169 	return err;
3170 }
3171