xref: /linux/net/ipv4/ipmr.c (revision 55d0969c451159cff86949b38c39171cab962069)
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 #include <net/inet_dscp.h>
66 
67 #include <linux/nospec.h>
68 
69 struct ipmr_rule {
70 	struct fib_rule		common;
71 };
72 
73 struct ipmr_result {
74 	struct mr_table		*mrt;
75 };
76 
77 /* Big lock, protecting vif table, mrt cache and mroute socket state.
78  * Note that the changes are semaphored via rtnl_lock.
79  */
80 
81 static DEFINE_SPINLOCK(mrt_lock);
82 
83 static struct net_device *vif_dev_read(const struct vif_device *vif)
84 {
85 	return rcu_dereference(vif->dev);
86 }
87 
88 /* Multicast router control variables */
89 
90 /* Special spinlock for queue of unresolved entries */
91 static DEFINE_SPINLOCK(mfc_unres_lock);
92 
93 /* We return to original Alan's scheme. Hash table of resolved
94  * entries is changed only in process context and protected
95  * with weak lock mrt_lock. Queue of unresolved entries is protected
96  * with strong spinlock mfc_unres_lock.
97  *
98  * In this case data path is free of exclusive locks at all.
99  */
100 
101 static struct kmem_cache *mrt_cachep __ro_after_init;
102 
103 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
104 static void ipmr_free_table(struct mr_table *mrt);
105 
106 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
107 			  struct net_device *dev, struct sk_buff *skb,
108 			  struct mfc_cache *cache, int local);
109 static int ipmr_cache_report(const struct mr_table *mrt,
110 			     struct sk_buff *pkt, vifi_t vifi, int assert);
111 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
112 				 int cmd);
113 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
114 static void mroute_clean_tables(struct mr_table *mrt, int flags);
115 static void ipmr_expire_process(struct timer_list *t);
116 
117 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
118 #define ipmr_for_each_table(mrt, net)					\
119 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,	\
120 				lockdep_rtnl_is_held() ||		\
121 				list_empty(&net->ipv4.mr_tables))
122 
123 static struct mr_table *ipmr_mr_table_iter(struct net *net,
124 					   struct mr_table *mrt)
125 {
126 	struct mr_table *ret;
127 
128 	if (!mrt)
129 		ret = list_entry_rcu(net->ipv4.mr_tables.next,
130 				     struct mr_table, list);
131 	else
132 		ret = list_entry_rcu(mrt->list.next,
133 				     struct mr_table, list);
134 
135 	if (&ret->list == &net->ipv4.mr_tables)
136 		return NULL;
137 	return ret;
138 }
139 
140 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
141 {
142 	struct mr_table *mrt;
143 
144 	ipmr_for_each_table(mrt, net) {
145 		if (mrt->id == id)
146 			return mrt;
147 	}
148 	return NULL;
149 }
150 
151 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
152 			   struct mr_table **mrt)
153 {
154 	int err;
155 	struct ipmr_result res;
156 	struct fib_lookup_arg arg = {
157 		.result = &res,
158 		.flags = FIB_LOOKUP_NOREF,
159 	};
160 
161 	/* update flow if oif or iif point to device enslaved to l3mdev */
162 	l3mdev_update_flow(net, flowi4_to_flowi(flp4));
163 
164 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
165 			       flowi4_to_flowi(flp4), 0, &arg);
166 	if (err < 0)
167 		return err;
168 	*mrt = res.mrt;
169 	return 0;
170 }
171 
172 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
173 			    int flags, struct fib_lookup_arg *arg)
174 {
175 	struct ipmr_result *res = arg->result;
176 	struct mr_table *mrt;
177 
178 	switch (rule->action) {
179 	case FR_ACT_TO_TBL:
180 		break;
181 	case FR_ACT_UNREACHABLE:
182 		return -ENETUNREACH;
183 	case FR_ACT_PROHIBIT:
184 		return -EACCES;
185 	case FR_ACT_BLACKHOLE:
186 	default:
187 		return -EINVAL;
188 	}
189 
190 	arg->table = fib_rule_get_table(rule, arg);
191 
192 	mrt = ipmr_get_table(rule->fr_net, arg->table);
193 	if (!mrt)
194 		return -EAGAIN;
195 	res->mrt = mrt;
196 	return 0;
197 }
198 
199 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
200 {
201 	return 1;
202 }
203 
204 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
205 			       struct fib_rule_hdr *frh, struct nlattr **tb,
206 			       struct netlink_ext_ack *extack)
207 {
208 	return 0;
209 }
210 
211 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
212 			     struct nlattr **tb)
213 {
214 	return 1;
215 }
216 
217 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
218 			  struct fib_rule_hdr *frh)
219 {
220 	frh->dst_len = 0;
221 	frh->src_len = 0;
222 	frh->tos     = 0;
223 	return 0;
224 }
225 
226 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
227 	.family		= RTNL_FAMILY_IPMR,
228 	.rule_size	= sizeof(struct ipmr_rule),
229 	.addr_size	= sizeof(u32),
230 	.action		= ipmr_rule_action,
231 	.match		= ipmr_rule_match,
232 	.configure	= ipmr_rule_configure,
233 	.compare	= ipmr_rule_compare,
234 	.fill		= ipmr_rule_fill,
235 	.nlgroup	= RTNLGRP_IPV4_RULE,
236 	.owner		= THIS_MODULE,
237 };
238 
239 static int __net_init ipmr_rules_init(struct net *net)
240 {
241 	struct fib_rules_ops *ops;
242 	struct mr_table *mrt;
243 	int err;
244 
245 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
246 	if (IS_ERR(ops))
247 		return PTR_ERR(ops);
248 
249 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
250 
251 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
252 	if (IS_ERR(mrt)) {
253 		err = PTR_ERR(mrt);
254 		goto err1;
255 	}
256 
257 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT);
258 	if (err < 0)
259 		goto err2;
260 
261 	net->ipv4.mr_rules_ops = ops;
262 	return 0;
263 
264 err2:
265 	rtnl_lock();
266 	ipmr_free_table(mrt);
267 	rtnl_unlock();
268 err1:
269 	fib_rules_unregister(ops);
270 	return err;
271 }
272 
273 static void __net_exit ipmr_rules_exit(struct net *net)
274 {
275 	struct mr_table *mrt, *next;
276 
277 	ASSERT_RTNL();
278 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
279 		list_del(&mrt->list);
280 		ipmr_free_table(mrt);
281 	}
282 	fib_rules_unregister(net->ipv4.mr_rules_ops);
283 }
284 
285 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
286 			   struct netlink_ext_ack *extack)
287 {
288 	return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
289 }
290 
291 static unsigned int ipmr_rules_seq_read(struct net *net)
292 {
293 	return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
294 }
295 
296 bool ipmr_rule_default(const struct fib_rule *rule)
297 {
298 	return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
299 }
300 EXPORT_SYMBOL(ipmr_rule_default);
301 #else
302 #define ipmr_for_each_table(mrt, net) \
303 	for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
304 
305 static struct mr_table *ipmr_mr_table_iter(struct net *net,
306 					   struct mr_table *mrt)
307 {
308 	if (!mrt)
309 		return net->ipv4.mrt;
310 	return NULL;
311 }
312 
313 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
314 {
315 	return net->ipv4.mrt;
316 }
317 
318 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
319 			   struct mr_table **mrt)
320 {
321 	*mrt = net->ipv4.mrt;
322 	return 0;
323 }
324 
325 static int __net_init ipmr_rules_init(struct net *net)
326 {
327 	struct mr_table *mrt;
328 
329 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
330 	if (IS_ERR(mrt))
331 		return PTR_ERR(mrt);
332 	net->ipv4.mrt = mrt;
333 	return 0;
334 }
335 
336 static void __net_exit ipmr_rules_exit(struct net *net)
337 {
338 	ASSERT_RTNL();
339 	ipmr_free_table(net->ipv4.mrt);
340 	net->ipv4.mrt = NULL;
341 }
342 
343 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
344 			   struct netlink_ext_ack *extack)
345 {
346 	return 0;
347 }
348 
349 static unsigned int ipmr_rules_seq_read(struct net *net)
350 {
351 	return 0;
352 }
353 
354 bool ipmr_rule_default(const struct fib_rule *rule)
355 {
356 	return true;
357 }
358 EXPORT_SYMBOL(ipmr_rule_default);
359 #endif
360 
361 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
362 				const void *ptr)
363 {
364 	const struct mfc_cache_cmp_arg *cmparg = arg->key;
365 	const struct mfc_cache *c = ptr;
366 
367 	return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
368 	       cmparg->mfc_origin != c->mfc_origin;
369 }
370 
371 static const struct rhashtable_params ipmr_rht_params = {
372 	.head_offset = offsetof(struct mr_mfc, mnode),
373 	.key_offset = offsetof(struct mfc_cache, cmparg),
374 	.key_len = sizeof(struct mfc_cache_cmp_arg),
375 	.nelem_hint = 3,
376 	.obj_cmpfn = ipmr_hash_cmp,
377 	.automatic_shrinking = true,
378 };
379 
380 static void ipmr_new_table_set(struct mr_table *mrt,
381 			       struct net *net)
382 {
383 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
384 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
385 #endif
386 }
387 
388 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
389 	.mfc_mcastgrp = htonl(INADDR_ANY),
390 	.mfc_origin = htonl(INADDR_ANY),
391 };
392 
393 static struct mr_table_ops ipmr_mr_table_ops = {
394 	.rht_params = &ipmr_rht_params,
395 	.cmparg_any = &ipmr_mr_table_ops_cmparg_any,
396 };
397 
398 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
399 {
400 	struct mr_table *mrt;
401 
402 	/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
403 	if (id != RT_TABLE_DEFAULT && id >= 1000000000)
404 		return ERR_PTR(-EINVAL);
405 
406 	mrt = ipmr_get_table(net, id);
407 	if (mrt)
408 		return mrt;
409 
410 	return mr_table_alloc(net, id, &ipmr_mr_table_ops,
411 			      ipmr_expire_process, ipmr_new_table_set);
412 }
413 
414 static void ipmr_free_table(struct mr_table *mrt)
415 {
416 	timer_shutdown_sync(&mrt->ipmr_expire_timer);
417 	mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
418 				 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
419 	rhltable_destroy(&mrt->mfc_hash);
420 	kfree(mrt);
421 }
422 
423 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
424 
425 /* Initialize ipmr pimreg/tunnel in_device */
426 static bool ipmr_init_vif_indev(const struct net_device *dev)
427 {
428 	struct in_device *in_dev;
429 
430 	ASSERT_RTNL();
431 
432 	in_dev = __in_dev_get_rtnl(dev);
433 	if (!in_dev)
434 		return false;
435 	ipv4_devconf_setall(in_dev);
436 	neigh_parms_data_state_setall(in_dev->arp_parms);
437 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
438 
439 	return true;
440 }
441 
442 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
443 {
444 	struct net_device *tunnel_dev, *new_dev;
445 	struct ip_tunnel_parm_kern p = { };
446 	int err;
447 
448 	tunnel_dev = __dev_get_by_name(net, "tunl0");
449 	if (!tunnel_dev)
450 		goto out;
451 
452 	p.iph.daddr = v->vifc_rmt_addr.s_addr;
453 	p.iph.saddr = v->vifc_lcl_addr.s_addr;
454 	p.iph.version = 4;
455 	p.iph.ihl = 5;
456 	p.iph.protocol = IPPROTO_IPIP;
457 	sprintf(p.name, "dvmrp%d", v->vifc_vifi);
458 
459 	if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
460 		goto out;
461 	err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
462 			SIOCADDTUNNEL);
463 	if (err)
464 		goto out;
465 
466 	new_dev = __dev_get_by_name(net, p.name);
467 	if (!new_dev)
468 		goto out;
469 
470 	new_dev->flags |= IFF_MULTICAST;
471 	if (!ipmr_init_vif_indev(new_dev))
472 		goto out_unregister;
473 	if (dev_open(new_dev, NULL))
474 		goto out_unregister;
475 	dev_hold(new_dev);
476 	err = dev_set_allmulti(new_dev, 1);
477 	if (err) {
478 		dev_close(new_dev);
479 		tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
480 				SIOCDELTUNNEL);
481 		dev_put(new_dev);
482 		new_dev = ERR_PTR(err);
483 	}
484 	return new_dev;
485 
486 out_unregister:
487 	unregister_netdevice(new_dev);
488 out:
489 	return ERR_PTR(-ENOBUFS);
490 }
491 
492 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
493 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
494 {
495 	struct net *net = dev_net(dev);
496 	struct mr_table *mrt;
497 	struct flowi4 fl4 = {
498 		.flowi4_oif	= dev->ifindex,
499 		.flowi4_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
500 		.flowi4_mark	= skb->mark,
501 	};
502 	int err;
503 
504 	err = ipmr_fib_lookup(net, &fl4, &mrt);
505 	if (err < 0) {
506 		kfree_skb(skb);
507 		return err;
508 	}
509 
510 	DEV_STATS_ADD(dev, tx_bytes, skb->len);
511 	DEV_STATS_INC(dev, tx_packets);
512 	rcu_read_lock();
513 
514 	/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
515 	ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
516 			  IGMPMSG_WHOLEPKT);
517 
518 	rcu_read_unlock();
519 	kfree_skb(skb);
520 	return NETDEV_TX_OK;
521 }
522 
523 static int reg_vif_get_iflink(const struct net_device *dev)
524 {
525 	return 0;
526 }
527 
528 static const struct net_device_ops reg_vif_netdev_ops = {
529 	.ndo_start_xmit	= reg_vif_xmit,
530 	.ndo_get_iflink = reg_vif_get_iflink,
531 };
532 
533 static void reg_vif_setup(struct net_device *dev)
534 {
535 	dev->type		= ARPHRD_PIMREG;
536 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
537 	dev->flags		= IFF_NOARP;
538 	dev->netdev_ops		= &reg_vif_netdev_ops;
539 	dev->needs_free_netdev	= true;
540 	dev->netns_local	= true;
541 }
542 
543 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
544 {
545 	struct net_device *dev;
546 	char name[IFNAMSIZ];
547 
548 	if (mrt->id == RT_TABLE_DEFAULT)
549 		sprintf(name, "pimreg");
550 	else
551 		sprintf(name, "pimreg%u", mrt->id);
552 
553 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
554 
555 	if (!dev)
556 		return NULL;
557 
558 	dev_net_set(dev, net);
559 
560 	if (register_netdevice(dev)) {
561 		free_netdev(dev);
562 		return NULL;
563 	}
564 
565 	if (!ipmr_init_vif_indev(dev))
566 		goto failure;
567 	if (dev_open(dev, NULL))
568 		goto failure;
569 
570 	dev_hold(dev);
571 
572 	return dev;
573 
574 failure:
575 	unregister_netdevice(dev);
576 	return NULL;
577 }
578 
579 /* called with rcu_read_lock() */
580 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
581 		     unsigned int pimlen)
582 {
583 	struct net_device *reg_dev = NULL;
584 	struct iphdr *encap;
585 	int vif_num;
586 
587 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
588 	/* Check that:
589 	 * a. packet is really sent to a multicast group
590 	 * b. packet is not a NULL-REGISTER
591 	 * c. packet is not truncated
592 	 */
593 	if (!ipv4_is_multicast(encap->daddr) ||
594 	    encap->tot_len == 0 ||
595 	    ntohs(encap->tot_len) + pimlen > skb->len)
596 		return 1;
597 
598 	/* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
599 	vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
600 	if (vif_num >= 0)
601 		reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
602 	if (!reg_dev)
603 		return 1;
604 
605 	skb->mac_header = skb->network_header;
606 	skb_pull(skb, (u8 *)encap - skb->data);
607 	skb_reset_network_header(skb);
608 	skb->protocol = htons(ETH_P_IP);
609 	skb->ip_summed = CHECKSUM_NONE;
610 
611 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
612 
613 	netif_rx(skb);
614 
615 	return NET_RX_SUCCESS;
616 }
617 #else
618 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
619 {
620 	return NULL;
621 }
622 #endif
623 
624 static int call_ipmr_vif_entry_notifiers(struct net *net,
625 					 enum fib_event_type event_type,
626 					 struct vif_device *vif,
627 					 struct net_device *vif_dev,
628 					 vifi_t vif_index, u32 tb_id)
629 {
630 	return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
631 				     vif, vif_dev, vif_index, tb_id,
632 				     &net->ipv4.ipmr_seq);
633 }
634 
635 static int call_ipmr_mfc_entry_notifiers(struct net *net,
636 					 enum fib_event_type event_type,
637 					 struct mfc_cache *mfc, u32 tb_id)
638 {
639 	return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
640 				     &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
641 }
642 
643 /**
644  *	vif_delete - Delete a VIF entry
645  *	@mrt: Table to delete from
646  *	@vifi: VIF identifier to delete
647  *	@notify: Set to 1, if the caller is a notifier_call
648  *	@head: if unregistering the VIF, place it on this queue
649  */
650 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
651 		      struct list_head *head)
652 {
653 	struct net *net = read_pnet(&mrt->net);
654 	struct vif_device *v;
655 	struct net_device *dev;
656 	struct in_device *in_dev;
657 
658 	if (vifi < 0 || vifi >= mrt->maxvif)
659 		return -EADDRNOTAVAIL;
660 
661 	v = &mrt->vif_table[vifi];
662 
663 	dev = rtnl_dereference(v->dev);
664 	if (!dev)
665 		return -EADDRNOTAVAIL;
666 
667 	spin_lock(&mrt_lock);
668 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
669 				      vifi, mrt->id);
670 	RCU_INIT_POINTER(v->dev, NULL);
671 
672 	if (vifi == mrt->mroute_reg_vif_num) {
673 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
674 		WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
675 	}
676 	if (vifi + 1 == mrt->maxvif) {
677 		int tmp;
678 
679 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
680 			if (VIF_EXISTS(mrt, tmp))
681 				break;
682 		}
683 		WRITE_ONCE(mrt->maxvif, tmp + 1);
684 	}
685 
686 	spin_unlock(&mrt_lock);
687 
688 	dev_set_allmulti(dev, -1);
689 
690 	in_dev = __in_dev_get_rtnl(dev);
691 	if (in_dev) {
692 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
693 		inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
694 					    NETCONFA_MC_FORWARDING,
695 					    dev->ifindex, &in_dev->cnf);
696 		ip_rt_multicast_event(in_dev);
697 	}
698 
699 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
700 		unregister_netdevice_queue(dev, head);
701 
702 	netdev_put(dev, &v->dev_tracker);
703 	return 0;
704 }
705 
706 static void ipmr_cache_free_rcu(struct rcu_head *head)
707 {
708 	struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
709 
710 	kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
711 }
712 
713 static void ipmr_cache_free(struct mfc_cache *c)
714 {
715 	call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
716 }
717 
718 /* Destroy an unresolved cache entry, killing queued skbs
719  * and reporting error to netlink readers.
720  */
721 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
722 {
723 	struct net *net = read_pnet(&mrt->net);
724 	struct sk_buff *skb;
725 	struct nlmsgerr *e;
726 
727 	atomic_dec(&mrt->cache_resolve_queue_len);
728 
729 	while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
730 		if (ip_hdr(skb)->version == 0) {
731 			struct nlmsghdr *nlh = skb_pull(skb,
732 							sizeof(struct iphdr));
733 			nlh->nlmsg_type = NLMSG_ERROR;
734 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
735 			skb_trim(skb, nlh->nlmsg_len);
736 			e = nlmsg_data(nlh);
737 			e->error = -ETIMEDOUT;
738 			memset(&e->msg, 0, sizeof(e->msg));
739 
740 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
741 		} else {
742 			kfree_skb(skb);
743 		}
744 	}
745 
746 	ipmr_cache_free(c);
747 }
748 
749 /* Timer process for the unresolved queue. */
750 static void ipmr_expire_process(struct timer_list *t)
751 {
752 	struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
753 	struct mr_mfc *c, *next;
754 	unsigned long expires;
755 	unsigned long now;
756 
757 	if (!spin_trylock(&mfc_unres_lock)) {
758 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
759 		return;
760 	}
761 
762 	if (list_empty(&mrt->mfc_unres_queue))
763 		goto out;
764 
765 	now = jiffies;
766 	expires = 10*HZ;
767 
768 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
769 		if (time_after(c->mfc_un.unres.expires, now)) {
770 			unsigned long interval = c->mfc_un.unres.expires - now;
771 			if (interval < expires)
772 				expires = interval;
773 			continue;
774 		}
775 
776 		list_del(&c->list);
777 		mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
778 		ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
779 	}
780 
781 	if (!list_empty(&mrt->mfc_unres_queue))
782 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
783 
784 out:
785 	spin_unlock(&mfc_unres_lock);
786 }
787 
788 /* Fill oifs list. It is called under locked mrt_lock. */
789 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
790 				   unsigned char *ttls)
791 {
792 	int vifi;
793 
794 	cache->mfc_un.res.minvif = MAXVIFS;
795 	cache->mfc_un.res.maxvif = 0;
796 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
797 
798 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
799 		if (VIF_EXISTS(mrt, vifi) &&
800 		    ttls[vifi] && ttls[vifi] < 255) {
801 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
802 			if (cache->mfc_un.res.minvif > vifi)
803 				cache->mfc_un.res.minvif = vifi;
804 			if (cache->mfc_un.res.maxvif <= vifi)
805 				cache->mfc_un.res.maxvif = vifi + 1;
806 		}
807 	}
808 	cache->mfc_un.res.lastuse = jiffies;
809 }
810 
811 static int vif_add(struct net *net, struct mr_table *mrt,
812 		   struct vifctl *vifc, int mrtsock)
813 {
814 	struct netdev_phys_item_id ppid = { };
815 	int vifi = vifc->vifc_vifi;
816 	struct vif_device *v = &mrt->vif_table[vifi];
817 	struct net_device *dev;
818 	struct in_device *in_dev;
819 	int err;
820 
821 	/* Is vif busy ? */
822 	if (VIF_EXISTS(mrt, vifi))
823 		return -EADDRINUSE;
824 
825 	switch (vifc->vifc_flags) {
826 	case VIFF_REGISTER:
827 		if (!ipmr_pimsm_enabled())
828 			return -EINVAL;
829 		/* Special Purpose VIF in PIM
830 		 * All the packets will be sent to the daemon
831 		 */
832 		if (mrt->mroute_reg_vif_num >= 0)
833 			return -EADDRINUSE;
834 		dev = ipmr_reg_vif(net, mrt);
835 		if (!dev)
836 			return -ENOBUFS;
837 		err = dev_set_allmulti(dev, 1);
838 		if (err) {
839 			unregister_netdevice(dev);
840 			dev_put(dev);
841 			return err;
842 		}
843 		break;
844 	case VIFF_TUNNEL:
845 		dev = ipmr_new_tunnel(net, vifc);
846 		if (IS_ERR(dev))
847 			return PTR_ERR(dev);
848 		break;
849 	case VIFF_USE_IFINDEX:
850 	case 0:
851 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
852 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
853 			if (dev && !__in_dev_get_rtnl(dev)) {
854 				dev_put(dev);
855 				return -EADDRNOTAVAIL;
856 			}
857 		} else {
858 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
859 		}
860 		if (!dev)
861 			return -EADDRNOTAVAIL;
862 		err = dev_set_allmulti(dev, 1);
863 		if (err) {
864 			dev_put(dev);
865 			return err;
866 		}
867 		break;
868 	default:
869 		return -EINVAL;
870 	}
871 
872 	in_dev = __in_dev_get_rtnl(dev);
873 	if (!in_dev) {
874 		dev_put(dev);
875 		return -EADDRNOTAVAIL;
876 	}
877 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
878 	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
879 				    dev->ifindex, &in_dev->cnf);
880 	ip_rt_multicast_event(in_dev);
881 
882 	/* Fill in the VIF structures */
883 	vif_device_init(v, dev, vifc->vifc_rate_limit,
884 			vifc->vifc_threshold,
885 			vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
886 			(VIFF_TUNNEL | VIFF_REGISTER));
887 
888 	err = dev_get_port_parent_id(dev, &ppid, true);
889 	if (err == 0) {
890 		memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
891 		v->dev_parent_id.id_len = ppid.id_len;
892 	} else {
893 		v->dev_parent_id.id_len = 0;
894 	}
895 
896 	v->local = vifc->vifc_lcl_addr.s_addr;
897 	v->remote = vifc->vifc_rmt_addr.s_addr;
898 
899 	/* And finish update writing critical data */
900 	spin_lock(&mrt_lock);
901 	rcu_assign_pointer(v->dev, dev);
902 	netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
903 	if (v->flags & VIFF_REGISTER) {
904 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
905 		WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
906 	}
907 	if (vifi+1 > mrt->maxvif)
908 		WRITE_ONCE(mrt->maxvif, vifi + 1);
909 	spin_unlock(&mrt_lock);
910 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
911 				      vifi, mrt->id);
912 	return 0;
913 }
914 
915 /* called with rcu_read_lock() */
916 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
917 					 __be32 origin,
918 					 __be32 mcastgrp)
919 {
920 	struct mfc_cache_cmp_arg arg = {
921 			.mfc_mcastgrp = mcastgrp,
922 			.mfc_origin = origin
923 	};
924 
925 	return mr_mfc_find(mrt, &arg);
926 }
927 
928 /* Look for a (*,G) entry */
929 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
930 					     __be32 mcastgrp, int vifi)
931 {
932 	struct mfc_cache_cmp_arg arg = {
933 			.mfc_mcastgrp = mcastgrp,
934 			.mfc_origin = htonl(INADDR_ANY)
935 	};
936 
937 	if (mcastgrp == htonl(INADDR_ANY))
938 		return mr_mfc_find_any_parent(mrt, vifi);
939 	return mr_mfc_find_any(mrt, vifi, &arg);
940 }
941 
942 /* Look for a (S,G,iif) entry if parent != -1 */
943 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
944 						__be32 origin, __be32 mcastgrp,
945 						int parent)
946 {
947 	struct mfc_cache_cmp_arg arg = {
948 			.mfc_mcastgrp = mcastgrp,
949 			.mfc_origin = origin,
950 	};
951 
952 	return mr_mfc_find_parent(mrt, &arg, parent);
953 }
954 
955 /* Allocate a multicast cache entry */
956 static struct mfc_cache *ipmr_cache_alloc(void)
957 {
958 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
959 
960 	if (c) {
961 		c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
962 		c->_c.mfc_un.res.minvif = MAXVIFS;
963 		c->_c.free = ipmr_cache_free_rcu;
964 		refcount_set(&c->_c.mfc_un.res.refcount, 1);
965 	}
966 	return c;
967 }
968 
969 static struct mfc_cache *ipmr_cache_alloc_unres(void)
970 {
971 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
972 
973 	if (c) {
974 		skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
975 		c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
976 	}
977 	return c;
978 }
979 
980 /* A cache entry has gone into a resolved state from queued */
981 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
982 			       struct mfc_cache *uc, struct mfc_cache *c)
983 {
984 	struct sk_buff *skb;
985 	struct nlmsgerr *e;
986 
987 	/* Play the pending entries through our router */
988 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
989 		if (ip_hdr(skb)->version == 0) {
990 			struct nlmsghdr *nlh = skb_pull(skb,
991 							sizeof(struct iphdr));
992 
993 			if (mr_fill_mroute(mrt, skb, &c->_c,
994 					   nlmsg_data(nlh)) > 0) {
995 				nlh->nlmsg_len = skb_tail_pointer(skb) -
996 						 (u8 *)nlh;
997 			} else {
998 				nlh->nlmsg_type = NLMSG_ERROR;
999 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1000 				skb_trim(skb, nlh->nlmsg_len);
1001 				e = nlmsg_data(nlh);
1002 				e->error = -EMSGSIZE;
1003 				memset(&e->msg, 0, sizeof(e->msg));
1004 			}
1005 
1006 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1007 		} else {
1008 			rcu_read_lock();
1009 			ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1010 			rcu_read_unlock();
1011 		}
1012 	}
1013 }
1014 
1015 /* Bounce a cache query up to mrouted and netlink.
1016  *
1017  * Called under rcu_read_lock().
1018  */
1019 static int ipmr_cache_report(const struct mr_table *mrt,
1020 			     struct sk_buff *pkt, vifi_t vifi, int assert)
1021 {
1022 	const int ihl = ip_hdrlen(pkt);
1023 	struct sock *mroute_sk;
1024 	struct igmphdr *igmp;
1025 	struct igmpmsg *msg;
1026 	struct sk_buff *skb;
1027 	int ret;
1028 
1029 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1030 	if (!mroute_sk)
1031 		return -EINVAL;
1032 
1033 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1034 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1035 	else
1036 		skb = alloc_skb(128, GFP_ATOMIC);
1037 
1038 	if (!skb)
1039 		return -ENOBUFS;
1040 
1041 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1042 		/* Ugly, but we have no choice with this interface.
1043 		 * Duplicate old header, fix ihl, length etc.
1044 		 * And all this only to mangle msg->im_msgtype and
1045 		 * to set msg->im_mbz to "mbz" :-)
1046 		 */
1047 		skb_push(skb, sizeof(struct iphdr));
1048 		skb_reset_network_header(skb);
1049 		skb_reset_transport_header(skb);
1050 		msg = (struct igmpmsg *)skb_network_header(skb);
1051 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1052 		msg->im_msgtype = assert;
1053 		msg->im_mbz = 0;
1054 		if (assert == IGMPMSG_WRVIFWHOLE) {
1055 			msg->im_vif = vifi;
1056 			msg->im_vif_hi = vifi >> 8;
1057 		} else {
1058 			/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1059 			int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1060 
1061 			msg->im_vif = vif_num;
1062 			msg->im_vif_hi = vif_num >> 8;
1063 		}
1064 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1065 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1066 					     sizeof(struct iphdr));
1067 	} else {
1068 		/* Copy the IP header */
1069 		skb_set_network_header(skb, skb->len);
1070 		skb_put(skb, ihl);
1071 		skb_copy_to_linear_data(skb, pkt->data, ihl);
1072 		/* Flag to the kernel this is a route add */
1073 		ip_hdr(skb)->protocol = 0;
1074 		msg = (struct igmpmsg *)skb_network_header(skb);
1075 		msg->im_vif = vifi;
1076 		msg->im_vif_hi = vifi >> 8;
1077 		ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1078 		memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1079 		/* Add our header */
1080 		igmp = skb_put(skb, sizeof(struct igmphdr));
1081 		igmp->type = assert;
1082 		msg->im_msgtype = assert;
1083 		igmp->code = 0;
1084 		ip_hdr(skb)->tot_len = htons(skb->len);	/* Fix the length */
1085 		skb->transport_header = skb->network_header;
1086 	}
1087 
1088 	igmpmsg_netlink_event(mrt, skb);
1089 
1090 	/* Deliver to mrouted */
1091 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1092 
1093 	if (ret < 0) {
1094 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1095 		kfree_skb(skb);
1096 	}
1097 
1098 	return ret;
1099 }
1100 
1101 /* Queue a packet for resolution. It gets locked cache entry! */
1102 /* Called under rcu_read_lock() */
1103 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1104 				 struct sk_buff *skb, struct net_device *dev)
1105 {
1106 	const struct iphdr *iph = ip_hdr(skb);
1107 	struct mfc_cache *c;
1108 	bool found = false;
1109 	int err;
1110 
1111 	spin_lock_bh(&mfc_unres_lock);
1112 	list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1113 		if (c->mfc_mcastgrp == iph->daddr &&
1114 		    c->mfc_origin == iph->saddr) {
1115 			found = true;
1116 			break;
1117 		}
1118 	}
1119 
1120 	if (!found) {
1121 		/* Create a new entry if allowable */
1122 		c = ipmr_cache_alloc_unres();
1123 		if (!c) {
1124 			spin_unlock_bh(&mfc_unres_lock);
1125 
1126 			kfree_skb(skb);
1127 			return -ENOBUFS;
1128 		}
1129 
1130 		/* Fill in the new cache entry */
1131 		c->_c.mfc_parent = -1;
1132 		c->mfc_origin	= iph->saddr;
1133 		c->mfc_mcastgrp	= iph->daddr;
1134 
1135 		/* Reflect first query at mrouted. */
1136 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1137 
1138 		if (err < 0) {
1139 			/* If the report failed throw the cache entry
1140 			   out - Brad Parker
1141 			 */
1142 			spin_unlock_bh(&mfc_unres_lock);
1143 
1144 			ipmr_cache_free(c);
1145 			kfree_skb(skb);
1146 			return err;
1147 		}
1148 
1149 		atomic_inc(&mrt->cache_resolve_queue_len);
1150 		list_add(&c->_c.list, &mrt->mfc_unres_queue);
1151 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1152 
1153 		if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1154 			mod_timer(&mrt->ipmr_expire_timer,
1155 				  c->_c.mfc_un.unres.expires);
1156 	}
1157 
1158 	/* See if we can append the packet */
1159 	if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1160 		kfree_skb(skb);
1161 		err = -ENOBUFS;
1162 	} else {
1163 		if (dev) {
1164 			skb->dev = dev;
1165 			skb->skb_iif = dev->ifindex;
1166 		}
1167 		skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1168 		err = 0;
1169 	}
1170 
1171 	spin_unlock_bh(&mfc_unres_lock);
1172 	return err;
1173 }
1174 
1175 /* MFC cache manipulation by user space mroute daemon */
1176 
1177 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1178 {
1179 	struct net *net = read_pnet(&mrt->net);
1180 	struct mfc_cache *c;
1181 
1182 	/* The entries are added/deleted only under RTNL */
1183 	rcu_read_lock();
1184 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1185 				   mfc->mfcc_mcastgrp.s_addr, parent);
1186 	rcu_read_unlock();
1187 	if (!c)
1188 		return -ENOENT;
1189 	rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1190 	list_del_rcu(&c->_c.list);
1191 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1192 	mroute_netlink_event(mrt, c, RTM_DELROUTE);
1193 	mr_cache_put(&c->_c);
1194 
1195 	return 0;
1196 }
1197 
1198 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1199 			struct mfcctl *mfc, int mrtsock, int parent)
1200 {
1201 	struct mfc_cache *uc, *c;
1202 	struct mr_mfc *_uc;
1203 	bool found;
1204 	int ret;
1205 
1206 	if (mfc->mfcc_parent >= MAXVIFS)
1207 		return -ENFILE;
1208 
1209 	/* The entries are added/deleted only under RTNL */
1210 	rcu_read_lock();
1211 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1212 				   mfc->mfcc_mcastgrp.s_addr, parent);
1213 	rcu_read_unlock();
1214 	if (c) {
1215 		spin_lock(&mrt_lock);
1216 		c->_c.mfc_parent = mfc->mfcc_parent;
1217 		ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1218 		if (!mrtsock)
1219 			c->_c.mfc_flags |= MFC_STATIC;
1220 		spin_unlock(&mrt_lock);
1221 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1222 					      mrt->id);
1223 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1224 		return 0;
1225 	}
1226 
1227 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1228 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1229 		return -EINVAL;
1230 
1231 	c = ipmr_cache_alloc();
1232 	if (!c)
1233 		return -ENOMEM;
1234 
1235 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1236 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1237 	c->_c.mfc_parent = mfc->mfcc_parent;
1238 	ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1239 	if (!mrtsock)
1240 		c->_c.mfc_flags |= MFC_STATIC;
1241 
1242 	ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1243 				  ipmr_rht_params);
1244 	if (ret) {
1245 		pr_err("ipmr: rhtable insert error %d\n", ret);
1246 		ipmr_cache_free(c);
1247 		return ret;
1248 	}
1249 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1250 	/* Check to see if we resolved a queued list. If so we
1251 	 * need to send on the frames and tidy up.
1252 	 */
1253 	found = false;
1254 	spin_lock_bh(&mfc_unres_lock);
1255 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1256 		uc = (struct mfc_cache *)_uc;
1257 		if (uc->mfc_origin == c->mfc_origin &&
1258 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1259 			list_del(&_uc->list);
1260 			atomic_dec(&mrt->cache_resolve_queue_len);
1261 			found = true;
1262 			break;
1263 		}
1264 	}
1265 	if (list_empty(&mrt->mfc_unres_queue))
1266 		del_timer(&mrt->ipmr_expire_timer);
1267 	spin_unlock_bh(&mfc_unres_lock);
1268 
1269 	if (found) {
1270 		ipmr_cache_resolve(net, mrt, uc, c);
1271 		ipmr_cache_free(uc);
1272 	}
1273 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1274 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1275 	return 0;
1276 }
1277 
1278 /* Close the multicast socket, and clear the vif tables etc */
1279 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1280 {
1281 	struct net *net = read_pnet(&mrt->net);
1282 	struct mr_mfc *c, *tmp;
1283 	struct mfc_cache *cache;
1284 	LIST_HEAD(list);
1285 	int i;
1286 
1287 	/* Shut down all active vif entries */
1288 	if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1289 		for (i = 0; i < mrt->maxvif; i++) {
1290 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1291 			     !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1292 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1293 				continue;
1294 			vif_delete(mrt, i, 0, &list);
1295 		}
1296 		unregister_netdevice_many(&list);
1297 	}
1298 
1299 	/* Wipe the cache */
1300 	if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1301 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1302 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1303 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1304 				continue;
1305 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1306 			list_del_rcu(&c->list);
1307 			cache = (struct mfc_cache *)c;
1308 			call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1309 						      mrt->id);
1310 			mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1311 			mr_cache_put(c);
1312 		}
1313 	}
1314 
1315 	if (flags & MRT_FLUSH_MFC) {
1316 		if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1317 			spin_lock_bh(&mfc_unres_lock);
1318 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1319 				list_del(&c->list);
1320 				cache = (struct mfc_cache *)c;
1321 				mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1322 				ipmr_destroy_unres(mrt, cache);
1323 			}
1324 			spin_unlock_bh(&mfc_unres_lock);
1325 		}
1326 	}
1327 }
1328 
1329 /* called from ip_ra_control(), before an RCU grace period,
1330  * we don't need to call synchronize_rcu() here
1331  */
1332 static void mrtsock_destruct(struct sock *sk)
1333 {
1334 	struct net *net = sock_net(sk);
1335 	struct mr_table *mrt;
1336 
1337 	rtnl_lock();
1338 	ipmr_for_each_table(mrt, net) {
1339 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1340 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1341 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1342 						    NETCONFA_MC_FORWARDING,
1343 						    NETCONFA_IFINDEX_ALL,
1344 						    net->ipv4.devconf_all);
1345 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1346 			mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1347 		}
1348 	}
1349 	rtnl_unlock();
1350 }
1351 
1352 /* Socket options and virtual interface manipulation. The whole
1353  * virtual interface system is a complete heap, but unfortunately
1354  * that's how BSD mrouted happens to think. Maybe one day with a proper
1355  * MOSPF/PIM router set up we can clean this up.
1356  */
1357 
1358 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1359 			 unsigned int optlen)
1360 {
1361 	struct net *net = sock_net(sk);
1362 	int val, ret = 0, parent = 0;
1363 	struct mr_table *mrt;
1364 	struct vifctl vif;
1365 	struct mfcctl mfc;
1366 	bool do_wrvifwhole;
1367 	u32 uval;
1368 
1369 	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1370 	rtnl_lock();
1371 	if (sk->sk_type != SOCK_RAW ||
1372 	    inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1373 		ret = -EOPNOTSUPP;
1374 		goto out_unlock;
1375 	}
1376 
1377 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1378 	if (!mrt) {
1379 		ret = -ENOENT;
1380 		goto out_unlock;
1381 	}
1382 	if (optname != MRT_INIT) {
1383 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1384 		    !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1385 			ret = -EACCES;
1386 			goto out_unlock;
1387 		}
1388 	}
1389 
1390 	switch (optname) {
1391 	case MRT_INIT:
1392 		if (optlen != sizeof(int)) {
1393 			ret = -EINVAL;
1394 			break;
1395 		}
1396 		if (rtnl_dereference(mrt->mroute_sk)) {
1397 			ret = -EADDRINUSE;
1398 			break;
1399 		}
1400 
1401 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1402 		if (ret == 0) {
1403 			rcu_assign_pointer(mrt->mroute_sk, sk);
1404 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1405 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1406 						    NETCONFA_MC_FORWARDING,
1407 						    NETCONFA_IFINDEX_ALL,
1408 						    net->ipv4.devconf_all);
1409 		}
1410 		break;
1411 	case MRT_DONE:
1412 		if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1413 			ret = -EACCES;
1414 		} else {
1415 			/* We need to unlock here because mrtsock_destruct takes
1416 			 * care of rtnl itself and we can't change that due to
1417 			 * the IP_ROUTER_ALERT setsockopt which runs without it.
1418 			 */
1419 			rtnl_unlock();
1420 			ret = ip_ra_control(sk, 0, NULL);
1421 			goto out;
1422 		}
1423 		break;
1424 	case MRT_ADD_VIF:
1425 	case MRT_DEL_VIF:
1426 		if (optlen != sizeof(vif)) {
1427 			ret = -EINVAL;
1428 			break;
1429 		}
1430 		if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1431 			ret = -EFAULT;
1432 			break;
1433 		}
1434 		if (vif.vifc_vifi >= MAXVIFS) {
1435 			ret = -ENFILE;
1436 			break;
1437 		}
1438 		if (optname == MRT_ADD_VIF) {
1439 			ret = vif_add(net, mrt, &vif,
1440 				      sk == rtnl_dereference(mrt->mroute_sk));
1441 		} else {
1442 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1443 		}
1444 		break;
1445 	/* Manipulate the forwarding caches. These live
1446 	 * in a sort of kernel/user symbiosis.
1447 	 */
1448 	case MRT_ADD_MFC:
1449 	case MRT_DEL_MFC:
1450 		parent = -1;
1451 		fallthrough;
1452 	case MRT_ADD_MFC_PROXY:
1453 	case MRT_DEL_MFC_PROXY:
1454 		if (optlen != sizeof(mfc)) {
1455 			ret = -EINVAL;
1456 			break;
1457 		}
1458 		if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1459 			ret = -EFAULT;
1460 			break;
1461 		}
1462 		if (parent == 0)
1463 			parent = mfc.mfcc_parent;
1464 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1465 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1466 		else
1467 			ret = ipmr_mfc_add(net, mrt, &mfc,
1468 					   sk == rtnl_dereference(mrt->mroute_sk),
1469 					   parent);
1470 		break;
1471 	case MRT_FLUSH:
1472 		if (optlen != sizeof(val)) {
1473 			ret = -EINVAL;
1474 			break;
1475 		}
1476 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1477 			ret = -EFAULT;
1478 			break;
1479 		}
1480 		mroute_clean_tables(mrt, val);
1481 		break;
1482 	/* Control PIM assert. */
1483 	case MRT_ASSERT:
1484 		if (optlen != sizeof(val)) {
1485 			ret = -EINVAL;
1486 			break;
1487 		}
1488 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1489 			ret = -EFAULT;
1490 			break;
1491 		}
1492 		mrt->mroute_do_assert = val;
1493 		break;
1494 	case MRT_PIM:
1495 		if (!ipmr_pimsm_enabled()) {
1496 			ret = -ENOPROTOOPT;
1497 			break;
1498 		}
1499 		if (optlen != sizeof(val)) {
1500 			ret = -EINVAL;
1501 			break;
1502 		}
1503 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1504 			ret = -EFAULT;
1505 			break;
1506 		}
1507 
1508 		do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1509 		val = !!val;
1510 		if (val != mrt->mroute_do_pim) {
1511 			mrt->mroute_do_pim = val;
1512 			mrt->mroute_do_assert = val;
1513 			mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1514 		}
1515 		break;
1516 	case MRT_TABLE:
1517 		if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1518 			ret = -ENOPROTOOPT;
1519 			break;
1520 		}
1521 		if (optlen != sizeof(uval)) {
1522 			ret = -EINVAL;
1523 			break;
1524 		}
1525 		if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1526 			ret = -EFAULT;
1527 			break;
1528 		}
1529 
1530 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1531 			ret = -EBUSY;
1532 		} else {
1533 			mrt = ipmr_new_table(net, uval);
1534 			if (IS_ERR(mrt))
1535 				ret = PTR_ERR(mrt);
1536 			else
1537 				raw_sk(sk)->ipmr_table = uval;
1538 		}
1539 		break;
1540 	/* Spurious command, or MRT_VERSION which you cannot set. */
1541 	default:
1542 		ret = -ENOPROTOOPT;
1543 	}
1544 out_unlock:
1545 	rtnl_unlock();
1546 out:
1547 	return ret;
1548 }
1549 
1550 /* Execute if this ioctl is a special mroute ioctl */
1551 int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1552 {
1553 	switch (cmd) {
1554 	/* These userspace buffers will be consumed by ipmr_ioctl() */
1555 	case SIOCGETVIFCNT: {
1556 		struct sioc_vif_req buffer;
1557 
1558 		return sock_ioctl_inout(sk, cmd, arg, &buffer,
1559 				      sizeof(buffer));
1560 		}
1561 	case SIOCGETSGCNT: {
1562 		struct sioc_sg_req buffer;
1563 
1564 		return sock_ioctl_inout(sk, cmd, arg, &buffer,
1565 				      sizeof(buffer));
1566 		}
1567 	}
1568 	/* return code > 0 means that the ioctl was not executed */
1569 	return 1;
1570 }
1571 
1572 /* Getsock opt support for the multicast routing system. */
1573 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1574 			 sockptr_t optlen)
1575 {
1576 	int olr;
1577 	int val;
1578 	struct net *net = sock_net(sk);
1579 	struct mr_table *mrt;
1580 
1581 	if (sk->sk_type != SOCK_RAW ||
1582 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1583 		return -EOPNOTSUPP;
1584 
1585 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1586 	if (!mrt)
1587 		return -ENOENT;
1588 
1589 	switch (optname) {
1590 	case MRT_VERSION:
1591 		val = 0x0305;
1592 		break;
1593 	case MRT_PIM:
1594 		if (!ipmr_pimsm_enabled())
1595 			return -ENOPROTOOPT;
1596 		val = mrt->mroute_do_pim;
1597 		break;
1598 	case MRT_ASSERT:
1599 		val = mrt->mroute_do_assert;
1600 		break;
1601 	default:
1602 		return -ENOPROTOOPT;
1603 	}
1604 
1605 	if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1606 		return -EFAULT;
1607 	if (olr < 0)
1608 		return -EINVAL;
1609 
1610 	olr = min_t(unsigned int, olr, sizeof(int));
1611 
1612 	if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1613 		return -EFAULT;
1614 	if (copy_to_sockptr(optval, &val, olr))
1615 		return -EFAULT;
1616 	return 0;
1617 }
1618 
1619 /* The IP multicast ioctl support routines. */
1620 int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1621 {
1622 	struct vif_device *vif;
1623 	struct mfc_cache *c;
1624 	struct net *net = sock_net(sk);
1625 	struct sioc_vif_req *vr;
1626 	struct sioc_sg_req *sr;
1627 	struct mr_table *mrt;
1628 
1629 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1630 	if (!mrt)
1631 		return -ENOENT;
1632 
1633 	switch (cmd) {
1634 	case SIOCGETVIFCNT:
1635 		vr = (struct sioc_vif_req *)arg;
1636 		if (vr->vifi >= mrt->maxvif)
1637 			return -EINVAL;
1638 		vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1639 		rcu_read_lock();
1640 		vif = &mrt->vif_table[vr->vifi];
1641 		if (VIF_EXISTS(mrt, vr->vifi)) {
1642 			vr->icount = READ_ONCE(vif->pkt_in);
1643 			vr->ocount = READ_ONCE(vif->pkt_out);
1644 			vr->ibytes = READ_ONCE(vif->bytes_in);
1645 			vr->obytes = READ_ONCE(vif->bytes_out);
1646 			rcu_read_unlock();
1647 
1648 			return 0;
1649 		}
1650 		rcu_read_unlock();
1651 		return -EADDRNOTAVAIL;
1652 	case SIOCGETSGCNT:
1653 		sr = (struct sioc_sg_req *)arg;
1654 
1655 		rcu_read_lock();
1656 		c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1657 		if (c) {
1658 			sr->pktcnt = c->_c.mfc_un.res.pkt;
1659 			sr->bytecnt = c->_c.mfc_un.res.bytes;
1660 			sr->wrong_if = c->_c.mfc_un.res.wrong_if;
1661 			rcu_read_unlock();
1662 			return 0;
1663 		}
1664 		rcu_read_unlock();
1665 		return -EADDRNOTAVAIL;
1666 	default:
1667 		return -ENOIOCTLCMD;
1668 	}
1669 }
1670 
1671 #ifdef CONFIG_COMPAT
1672 struct compat_sioc_sg_req {
1673 	struct in_addr src;
1674 	struct in_addr grp;
1675 	compat_ulong_t pktcnt;
1676 	compat_ulong_t bytecnt;
1677 	compat_ulong_t wrong_if;
1678 };
1679 
1680 struct compat_sioc_vif_req {
1681 	vifi_t	vifi;		/* Which iface */
1682 	compat_ulong_t icount;
1683 	compat_ulong_t ocount;
1684 	compat_ulong_t ibytes;
1685 	compat_ulong_t obytes;
1686 };
1687 
1688 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1689 {
1690 	struct compat_sioc_sg_req sr;
1691 	struct compat_sioc_vif_req vr;
1692 	struct vif_device *vif;
1693 	struct mfc_cache *c;
1694 	struct net *net = sock_net(sk);
1695 	struct mr_table *mrt;
1696 
1697 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1698 	if (!mrt)
1699 		return -ENOENT;
1700 
1701 	switch (cmd) {
1702 	case SIOCGETVIFCNT:
1703 		if (copy_from_user(&vr, arg, sizeof(vr)))
1704 			return -EFAULT;
1705 		if (vr.vifi >= mrt->maxvif)
1706 			return -EINVAL;
1707 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1708 		rcu_read_lock();
1709 		vif = &mrt->vif_table[vr.vifi];
1710 		if (VIF_EXISTS(mrt, vr.vifi)) {
1711 			vr.icount = READ_ONCE(vif->pkt_in);
1712 			vr.ocount = READ_ONCE(vif->pkt_out);
1713 			vr.ibytes = READ_ONCE(vif->bytes_in);
1714 			vr.obytes = READ_ONCE(vif->bytes_out);
1715 			rcu_read_unlock();
1716 
1717 			if (copy_to_user(arg, &vr, sizeof(vr)))
1718 				return -EFAULT;
1719 			return 0;
1720 		}
1721 		rcu_read_unlock();
1722 		return -EADDRNOTAVAIL;
1723 	case SIOCGETSGCNT:
1724 		if (copy_from_user(&sr, arg, sizeof(sr)))
1725 			return -EFAULT;
1726 
1727 		rcu_read_lock();
1728 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1729 		if (c) {
1730 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1731 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1732 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1733 			rcu_read_unlock();
1734 
1735 			if (copy_to_user(arg, &sr, sizeof(sr)))
1736 				return -EFAULT;
1737 			return 0;
1738 		}
1739 		rcu_read_unlock();
1740 		return -EADDRNOTAVAIL;
1741 	default:
1742 		return -ENOIOCTLCMD;
1743 	}
1744 }
1745 #endif
1746 
1747 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1748 {
1749 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1750 	struct net *net = dev_net(dev);
1751 	struct mr_table *mrt;
1752 	struct vif_device *v;
1753 	int ct;
1754 
1755 	if (event != NETDEV_UNREGISTER)
1756 		return NOTIFY_DONE;
1757 
1758 	ipmr_for_each_table(mrt, net) {
1759 		v = &mrt->vif_table[0];
1760 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1761 			if (rcu_access_pointer(v->dev) == dev)
1762 				vif_delete(mrt, ct, 1, NULL);
1763 		}
1764 	}
1765 	return NOTIFY_DONE;
1766 }
1767 
1768 static struct notifier_block ip_mr_notifier = {
1769 	.notifier_call = ipmr_device_event,
1770 };
1771 
1772 /* Encapsulate a packet by attaching a valid IPIP header to it.
1773  * This avoids tunnel drivers and other mess and gives us the speed so
1774  * important for multicast video.
1775  */
1776 static void ip_encap(struct net *net, struct sk_buff *skb,
1777 		     __be32 saddr, __be32 daddr)
1778 {
1779 	struct iphdr *iph;
1780 	const struct iphdr *old_iph = ip_hdr(skb);
1781 
1782 	skb_push(skb, sizeof(struct iphdr));
1783 	skb->transport_header = skb->network_header;
1784 	skb_reset_network_header(skb);
1785 	iph = ip_hdr(skb);
1786 
1787 	iph->version	=	4;
1788 	iph->tos	=	old_iph->tos;
1789 	iph->ttl	=	old_iph->ttl;
1790 	iph->frag_off	=	0;
1791 	iph->daddr	=	daddr;
1792 	iph->saddr	=	saddr;
1793 	iph->protocol	=	IPPROTO_IPIP;
1794 	iph->ihl	=	5;
1795 	iph->tot_len	=	htons(skb->len);
1796 	ip_select_ident(net, skb, NULL);
1797 	ip_send_check(iph);
1798 
1799 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1800 	nf_reset_ct(skb);
1801 }
1802 
1803 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1804 				      struct sk_buff *skb)
1805 {
1806 	struct ip_options *opt = &(IPCB(skb)->opt);
1807 
1808 	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1809 
1810 	if (unlikely(opt->optlen))
1811 		ip_forward_options(skb);
1812 
1813 	return dst_output(net, sk, skb);
1814 }
1815 
1816 #ifdef CONFIG_NET_SWITCHDEV
1817 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1818 				   int in_vifi, int out_vifi)
1819 {
1820 	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1821 	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1822 
1823 	if (!skb->offload_l3_fwd_mark)
1824 		return false;
1825 	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1826 		return false;
1827 	return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1828 					&in_vif->dev_parent_id);
1829 }
1830 #else
1831 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1832 				   int in_vifi, int out_vifi)
1833 {
1834 	return false;
1835 }
1836 #endif
1837 
1838 /* Processing handlers for ipmr_forward, under rcu_read_lock() */
1839 
1840 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1841 			    int in_vifi, struct sk_buff *skb, int vifi)
1842 {
1843 	const struct iphdr *iph = ip_hdr(skb);
1844 	struct vif_device *vif = &mrt->vif_table[vifi];
1845 	struct net_device *vif_dev;
1846 	struct net_device *dev;
1847 	struct rtable *rt;
1848 	struct flowi4 fl4;
1849 	int    encap = 0;
1850 
1851 	vif_dev = vif_dev_read(vif);
1852 	if (!vif_dev)
1853 		goto out_free;
1854 
1855 	if (vif->flags & VIFF_REGISTER) {
1856 		WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1857 		WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1858 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1859 		DEV_STATS_INC(vif_dev, tx_packets);
1860 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1861 		goto out_free;
1862 	}
1863 
1864 	if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1865 		goto out_free;
1866 
1867 	if (vif->flags & VIFF_TUNNEL) {
1868 		rt = ip_route_output_ports(net, &fl4, NULL,
1869 					   vif->remote, vif->local,
1870 					   0, 0,
1871 					   IPPROTO_IPIP,
1872 					   iph->tos & INET_DSCP_MASK, vif->link);
1873 		if (IS_ERR(rt))
1874 			goto out_free;
1875 		encap = sizeof(struct iphdr);
1876 	} else {
1877 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1878 					   0, 0,
1879 					   IPPROTO_IPIP,
1880 					   iph->tos & INET_DSCP_MASK, vif->link);
1881 		if (IS_ERR(rt))
1882 			goto out_free;
1883 	}
1884 
1885 	dev = rt->dst.dev;
1886 
1887 	if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1888 		/* Do not fragment multicasts. Alas, IPv4 does not
1889 		 * allow to send ICMP, so that packets will disappear
1890 		 * to blackhole.
1891 		 */
1892 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1893 		ip_rt_put(rt);
1894 		goto out_free;
1895 	}
1896 
1897 	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1898 
1899 	if (skb_cow(skb, encap)) {
1900 		ip_rt_put(rt);
1901 		goto out_free;
1902 	}
1903 
1904 	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1905 	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1906 
1907 	skb_dst_drop(skb);
1908 	skb_dst_set(skb, &rt->dst);
1909 	ip_decrease_ttl(ip_hdr(skb));
1910 
1911 	/* FIXME: forward and output firewalls used to be called here.
1912 	 * What do we do with netfilter? -- RR
1913 	 */
1914 	if (vif->flags & VIFF_TUNNEL) {
1915 		ip_encap(net, skb, vif->local, vif->remote);
1916 		/* FIXME: extra output firewall step used to be here. --RR */
1917 		DEV_STATS_INC(vif_dev, tx_packets);
1918 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1919 	}
1920 
1921 	IPCB(skb)->flags |= IPSKB_FORWARDED;
1922 
1923 	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1924 	 * not only before forwarding, but after forwarding on all output
1925 	 * interfaces. It is clear, if mrouter runs a multicasting
1926 	 * program, it should receive packets not depending to what interface
1927 	 * program is joined.
1928 	 * If we will not make it, the program will have to join on all
1929 	 * interfaces. On the other hand, multihoming host (or router, but
1930 	 * not mrouter) cannot join to more than one interface - it will
1931 	 * result in receiving multiple packets.
1932 	 */
1933 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1934 		net, NULL, skb, skb->dev, dev,
1935 		ipmr_forward_finish);
1936 	return;
1937 
1938 out_free:
1939 	kfree_skb(skb);
1940 }
1941 
1942 /* Called with mrt_lock or rcu_read_lock() */
1943 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1944 {
1945 	int ct;
1946 	/* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1947 	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1948 		if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1949 			break;
1950 	}
1951 	return ct;
1952 }
1953 
1954 /* "local" means that we should preserve one skb (for local delivery) */
1955 /* Called uner rcu_read_lock() */
1956 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1957 			  struct net_device *dev, struct sk_buff *skb,
1958 			  struct mfc_cache *c, int local)
1959 {
1960 	int true_vifi = ipmr_find_vif(mrt, dev);
1961 	int psend = -1;
1962 	int vif, ct;
1963 
1964 	vif = c->_c.mfc_parent;
1965 	c->_c.mfc_un.res.pkt++;
1966 	c->_c.mfc_un.res.bytes += skb->len;
1967 	c->_c.mfc_un.res.lastuse = jiffies;
1968 
1969 	if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1970 		struct mfc_cache *cache_proxy;
1971 
1972 		/* For an (*,G) entry, we only check that the incoming
1973 		 * interface is part of the static tree.
1974 		 */
1975 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1976 		if (cache_proxy &&
1977 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1978 			goto forward;
1979 	}
1980 
1981 	/* Wrong interface: drop packet and (maybe) send PIM assert. */
1982 	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
1983 		if (rt_is_output_route(skb_rtable(skb))) {
1984 			/* It is our own packet, looped back.
1985 			 * Very complicated situation...
1986 			 *
1987 			 * The best workaround until routing daemons will be
1988 			 * fixed is not to redistribute packet, if it was
1989 			 * send through wrong interface. It means, that
1990 			 * multicast applications WILL NOT work for
1991 			 * (S,G), which have default multicast route pointing
1992 			 * to wrong oif. In any case, it is not a good
1993 			 * idea to use multicasting applications on router.
1994 			 */
1995 			goto dont_forward;
1996 		}
1997 
1998 		c->_c.mfc_un.res.wrong_if++;
1999 
2000 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
2001 		    /* pimsm uses asserts, when switching from RPT to SPT,
2002 		     * so that we cannot check that packet arrived on an oif.
2003 		     * It is bad, but otherwise we would need to move pretty
2004 		     * large chunk of pimd to kernel. Ough... --ANK
2005 		     */
2006 		    (mrt->mroute_do_pim ||
2007 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2008 		    time_after(jiffies,
2009 			       c->_c.mfc_un.res.last_assert +
2010 			       MFC_ASSERT_THRESH)) {
2011 			c->_c.mfc_un.res.last_assert = jiffies;
2012 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2013 			if (mrt->mroute_do_wrvifwhole)
2014 				ipmr_cache_report(mrt, skb, true_vifi,
2015 						  IGMPMSG_WRVIFWHOLE);
2016 		}
2017 		goto dont_forward;
2018 	}
2019 
2020 forward:
2021 	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2022 		   mrt->vif_table[vif].pkt_in + 1);
2023 	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2024 		   mrt->vif_table[vif].bytes_in + skb->len);
2025 
2026 	/* Forward the frame */
2027 	if (c->mfc_origin == htonl(INADDR_ANY) &&
2028 	    c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2029 		if (true_vifi >= 0 &&
2030 		    true_vifi != c->_c.mfc_parent &&
2031 		    ip_hdr(skb)->ttl >
2032 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2033 			/* It's an (*,*) entry and the packet is not coming from
2034 			 * the upstream: forward the packet to the upstream
2035 			 * only.
2036 			 */
2037 			psend = c->_c.mfc_parent;
2038 			goto last_forward;
2039 		}
2040 		goto dont_forward;
2041 	}
2042 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2043 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2044 		/* For (*,G) entry, don't forward to the incoming interface */
2045 		if ((c->mfc_origin != htonl(INADDR_ANY) ||
2046 		     ct != true_vifi) &&
2047 		    ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2048 			if (psend != -1) {
2049 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2050 
2051 				if (skb2)
2052 					ipmr_queue_xmit(net, mrt, true_vifi,
2053 							skb2, psend);
2054 			}
2055 			psend = ct;
2056 		}
2057 	}
2058 last_forward:
2059 	if (psend != -1) {
2060 		if (local) {
2061 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2062 
2063 			if (skb2)
2064 				ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2065 						psend);
2066 		} else {
2067 			ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2068 			return;
2069 		}
2070 	}
2071 
2072 dont_forward:
2073 	if (!local)
2074 		kfree_skb(skb);
2075 }
2076 
2077 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2078 {
2079 	struct rtable *rt = skb_rtable(skb);
2080 	struct iphdr *iph = ip_hdr(skb);
2081 	struct flowi4 fl4 = {
2082 		.daddr = iph->daddr,
2083 		.saddr = iph->saddr,
2084 		.flowi4_tos = iph->tos & INET_DSCP_MASK,
2085 		.flowi4_oif = (rt_is_output_route(rt) ?
2086 			       skb->dev->ifindex : 0),
2087 		.flowi4_iif = (rt_is_output_route(rt) ?
2088 			       LOOPBACK_IFINDEX :
2089 			       skb->dev->ifindex),
2090 		.flowi4_mark = skb->mark,
2091 	};
2092 	struct mr_table *mrt;
2093 	int err;
2094 
2095 	err = ipmr_fib_lookup(net, &fl4, &mrt);
2096 	if (err)
2097 		return ERR_PTR(err);
2098 	return mrt;
2099 }
2100 
2101 /* Multicast packets for forwarding arrive here
2102  * Called with rcu_read_lock();
2103  */
2104 int ip_mr_input(struct sk_buff *skb)
2105 {
2106 	struct mfc_cache *cache;
2107 	struct net *net = dev_net(skb->dev);
2108 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2109 	struct mr_table *mrt;
2110 	struct net_device *dev;
2111 
2112 	/* skb->dev passed in is the loX master dev for vrfs.
2113 	 * As there are no vifs associated with loopback devices,
2114 	 * get the proper interface that does have a vif associated with it.
2115 	 */
2116 	dev = skb->dev;
2117 	if (netif_is_l3_master(skb->dev)) {
2118 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2119 		if (!dev) {
2120 			kfree_skb(skb);
2121 			return -ENODEV;
2122 		}
2123 	}
2124 
2125 	/* Packet is looped back after forward, it should not be
2126 	 * forwarded second time, but still can be delivered locally.
2127 	 */
2128 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2129 		goto dont_forward;
2130 
2131 	mrt = ipmr_rt_fib_lookup(net, skb);
2132 	if (IS_ERR(mrt)) {
2133 		kfree_skb(skb);
2134 		return PTR_ERR(mrt);
2135 	}
2136 	if (!local) {
2137 		if (IPCB(skb)->opt.router_alert) {
2138 			if (ip_call_ra_chain(skb))
2139 				return 0;
2140 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2141 			/* IGMPv1 (and broken IGMPv2 implementations sort of
2142 			 * Cisco IOS <= 11.2(8)) do not put router alert
2143 			 * option to IGMP packets destined to routable
2144 			 * groups. It is very bad, because it means
2145 			 * that we can forward NO IGMP messages.
2146 			 */
2147 			struct sock *mroute_sk;
2148 
2149 			mroute_sk = rcu_dereference(mrt->mroute_sk);
2150 			if (mroute_sk) {
2151 				nf_reset_ct(skb);
2152 				raw_rcv(mroute_sk, skb);
2153 				return 0;
2154 			}
2155 		}
2156 	}
2157 
2158 	/* already under rcu_read_lock() */
2159 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2160 	if (!cache) {
2161 		int vif = ipmr_find_vif(mrt, dev);
2162 
2163 		if (vif >= 0)
2164 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2165 						    vif);
2166 	}
2167 
2168 	/* No usable cache entry */
2169 	if (!cache) {
2170 		int vif;
2171 
2172 		if (local) {
2173 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2174 			ip_local_deliver(skb);
2175 			if (!skb2)
2176 				return -ENOBUFS;
2177 			skb = skb2;
2178 		}
2179 
2180 		vif = ipmr_find_vif(mrt, dev);
2181 		if (vif >= 0)
2182 			return ipmr_cache_unresolved(mrt, vif, skb, dev);
2183 		kfree_skb(skb);
2184 		return -ENODEV;
2185 	}
2186 
2187 	ip_mr_forward(net, mrt, dev, skb, cache, local);
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 		if (dev)
2285 			vif = ipmr_find_vif(mrt, dev);
2286 		if (vif < 0) {
2287 			rcu_read_unlock();
2288 			return -ENODEV;
2289 		}
2290 
2291 		skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2292 		if (!skb2) {
2293 			rcu_read_unlock();
2294 			return -ENOMEM;
2295 		}
2296 
2297 		NETLINK_CB(skb2).portid = portid;
2298 		skb_push(skb2, sizeof(struct iphdr));
2299 		skb_reset_network_header(skb2);
2300 		iph = ip_hdr(skb2);
2301 		iph->ihl = sizeof(struct iphdr) >> 2;
2302 		iph->saddr = saddr;
2303 		iph->daddr = daddr;
2304 		iph->version = 0;
2305 		err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2306 		rcu_read_unlock();
2307 		return err;
2308 	}
2309 
2310 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2311 	rcu_read_unlock();
2312 	return err;
2313 }
2314 
2315 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2316 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2317 			    int flags)
2318 {
2319 	struct nlmsghdr *nlh;
2320 	struct rtmsg *rtm;
2321 	int err;
2322 
2323 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2324 	if (!nlh)
2325 		return -EMSGSIZE;
2326 
2327 	rtm = nlmsg_data(nlh);
2328 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2329 	rtm->rtm_dst_len  = 32;
2330 	rtm->rtm_src_len  = 32;
2331 	rtm->rtm_tos      = 0;
2332 	rtm->rtm_table    = mrt->id;
2333 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2334 		goto nla_put_failure;
2335 	rtm->rtm_type     = RTN_MULTICAST;
2336 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2337 	if (c->_c.mfc_flags & MFC_STATIC)
2338 		rtm->rtm_protocol = RTPROT_STATIC;
2339 	else
2340 		rtm->rtm_protocol = RTPROT_MROUTED;
2341 	rtm->rtm_flags    = 0;
2342 
2343 	if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2344 	    nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2345 		goto nla_put_failure;
2346 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2347 	/* do not break the dump if cache is unresolved */
2348 	if (err < 0 && err != -ENOENT)
2349 		goto nla_put_failure;
2350 
2351 	nlmsg_end(skb, nlh);
2352 	return 0;
2353 
2354 nla_put_failure:
2355 	nlmsg_cancel(skb, nlh);
2356 	return -EMSGSIZE;
2357 }
2358 
2359 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2360 			     u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2361 			     int flags)
2362 {
2363 	return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2364 				cmd, flags);
2365 }
2366 
2367 static size_t mroute_msgsize(bool unresolved, int maxvif)
2368 {
2369 	size_t len =
2370 		NLMSG_ALIGN(sizeof(struct rtmsg))
2371 		+ nla_total_size(4)	/* RTA_TABLE */
2372 		+ nla_total_size(4)	/* RTA_SRC */
2373 		+ nla_total_size(4)	/* RTA_DST */
2374 		;
2375 
2376 	if (!unresolved)
2377 		len = len
2378 		      + nla_total_size(4)	/* RTA_IIF */
2379 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2380 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2381 						/* RTA_MFC_STATS */
2382 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2383 		;
2384 
2385 	return len;
2386 }
2387 
2388 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2389 				 int cmd)
2390 {
2391 	struct net *net = read_pnet(&mrt->net);
2392 	struct sk_buff *skb;
2393 	int err = -ENOBUFS;
2394 
2395 	skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2396 				       mrt->maxvif),
2397 			GFP_ATOMIC);
2398 	if (!skb)
2399 		goto errout;
2400 
2401 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2402 	if (err < 0)
2403 		goto errout;
2404 
2405 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2406 	return;
2407 
2408 errout:
2409 	kfree_skb(skb);
2410 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2411 }
2412 
2413 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2414 {
2415 	size_t len =
2416 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2417 		+ nla_total_size(1)	/* IPMRA_CREPORT_MSGTYPE */
2418 		+ nla_total_size(4)	/* IPMRA_CREPORT_VIF_ID */
2419 		+ nla_total_size(4)	/* IPMRA_CREPORT_SRC_ADDR */
2420 		+ nla_total_size(4)	/* IPMRA_CREPORT_DST_ADDR */
2421 		+ nla_total_size(4)	/* IPMRA_CREPORT_TABLE */
2422 					/* IPMRA_CREPORT_PKT */
2423 		+ nla_total_size(payloadlen)
2424 		;
2425 
2426 	return len;
2427 }
2428 
2429 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2430 {
2431 	struct net *net = read_pnet(&mrt->net);
2432 	struct nlmsghdr *nlh;
2433 	struct rtgenmsg *rtgenm;
2434 	struct igmpmsg *msg;
2435 	struct sk_buff *skb;
2436 	struct nlattr *nla;
2437 	int payloadlen;
2438 
2439 	payloadlen = pkt->len - sizeof(struct igmpmsg);
2440 	msg = (struct igmpmsg *)skb_network_header(pkt);
2441 
2442 	skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2443 	if (!skb)
2444 		goto errout;
2445 
2446 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2447 			sizeof(struct rtgenmsg), 0);
2448 	if (!nlh)
2449 		goto errout;
2450 	rtgenm = nlmsg_data(nlh);
2451 	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2452 	if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2453 	    nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2454 	    nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2455 			    msg->im_src.s_addr) ||
2456 	    nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2457 			    msg->im_dst.s_addr) ||
2458 	    nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2459 		goto nla_put_failure;
2460 
2461 	nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2462 	if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2463 				  nla_data(nla), payloadlen))
2464 		goto nla_put_failure;
2465 
2466 	nlmsg_end(skb, nlh);
2467 
2468 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2469 	return;
2470 
2471 nla_put_failure:
2472 	nlmsg_cancel(skb, nlh);
2473 errout:
2474 	kfree_skb(skb);
2475 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2476 }
2477 
2478 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2479 				       const struct nlmsghdr *nlh,
2480 				       struct nlattr **tb,
2481 				       struct netlink_ext_ack *extack)
2482 {
2483 	struct rtmsg *rtm;
2484 	int i, err;
2485 
2486 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2487 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2488 		return -EINVAL;
2489 	}
2490 
2491 	if (!netlink_strict_get_check(skb))
2492 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2493 					      rtm_ipv4_policy, extack);
2494 
2495 	rtm = nlmsg_data(nlh);
2496 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2497 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2498 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2499 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2500 		NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2501 		return -EINVAL;
2502 	}
2503 
2504 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2505 					    rtm_ipv4_policy, extack);
2506 	if (err)
2507 		return err;
2508 
2509 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2510 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2511 		NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2512 		return -EINVAL;
2513 	}
2514 
2515 	for (i = 0; i <= RTA_MAX; i++) {
2516 		if (!tb[i])
2517 			continue;
2518 
2519 		switch (i) {
2520 		case RTA_SRC:
2521 		case RTA_DST:
2522 		case RTA_TABLE:
2523 			break;
2524 		default:
2525 			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2526 			return -EINVAL;
2527 		}
2528 	}
2529 
2530 	return 0;
2531 }
2532 
2533 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2534 			     struct netlink_ext_ack *extack)
2535 {
2536 	struct net *net = sock_net(in_skb->sk);
2537 	struct nlattr *tb[RTA_MAX + 1];
2538 	struct sk_buff *skb = NULL;
2539 	struct mfc_cache *cache;
2540 	struct mr_table *mrt;
2541 	__be32 src, grp;
2542 	u32 tableid;
2543 	int err;
2544 
2545 	err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2546 	if (err < 0)
2547 		goto errout;
2548 
2549 	src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2550 	grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2551 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2552 
2553 	mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2554 	if (!mrt) {
2555 		err = -ENOENT;
2556 		goto errout_free;
2557 	}
2558 
2559 	/* entries are added/deleted only under RTNL */
2560 	rcu_read_lock();
2561 	cache = ipmr_cache_find(mrt, src, grp);
2562 	rcu_read_unlock();
2563 	if (!cache) {
2564 		err = -ENOENT;
2565 		goto errout_free;
2566 	}
2567 
2568 	skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2569 	if (!skb) {
2570 		err = -ENOBUFS;
2571 		goto errout_free;
2572 	}
2573 
2574 	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2575 			       nlh->nlmsg_seq, cache,
2576 			       RTM_NEWROUTE, 0);
2577 	if (err < 0)
2578 		goto errout_free;
2579 
2580 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2581 
2582 errout:
2583 	return err;
2584 
2585 errout_free:
2586 	kfree_skb(skb);
2587 	goto errout;
2588 }
2589 
2590 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2591 {
2592 	struct fib_dump_filter filter = {
2593 		.rtnl_held = true,
2594 	};
2595 	int err;
2596 
2597 	if (cb->strict_check) {
2598 		err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2599 					    &filter, cb);
2600 		if (err < 0)
2601 			return err;
2602 	}
2603 
2604 	if (filter.table_id) {
2605 		struct mr_table *mrt;
2606 
2607 		mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2608 		if (!mrt) {
2609 			if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2610 				return skb->len;
2611 
2612 			NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2613 			return -ENOENT;
2614 		}
2615 		err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2616 				    &mfc_unres_lock, &filter);
2617 		return skb->len ? : err;
2618 	}
2619 
2620 	return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2621 				_ipmr_fill_mroute, &mfc_unres_lock, &filter);
2622 }
2623 
2624 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2625 	[RTA_SRC]	= { .type = NLA_U32 },
2626 	[RTA_DST]	= { .type = NLA_U32 },
2627 	[RTA_IIF]	= { .type = NLA_U32 },
2628 	[RTA_TABLE]	= { .type = NLA_U32 },
2629 	[RTA_MULTIPATH]	= { .len = sizeof(struct rtnexthop) },
2630 };
2631 
2632 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2633 {
2634 	switch (rtm_protocol) {
2635 	case RTPROT_STATIC:
2636 	case RTPROT_MROUTED:
2637 		return true;
2638 	}
2639 	return false;
2640 }
2641 
2642 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2643 {
2644 	struct rtnexthop *rtnh = nla_data(nla);
2645 	int remaining = nla_len(nla), vifi = 0;
2646 
2647 	while (rtnh_ok(rtnh, remaining)) {
2648 		mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2649 		if (++vifi == MAXVIFS)
2650 			break;
2651 		rtnh = rtnh_next(rtnh, &remaining);
2652 	}
2653 
2654 	return remaining > 0 ? -EINVAL : vifi;
2655 }
2656 
2657 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2658 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2659 			    struct mfcctl *mfcc, int *mrtsock,
2660 			    struct mr_table **mrtret,
2661 			    struct netlink_ext_ack *extack)
2662 {
2663 	struct net_device *dev = NULL;
2664 	u32 tblid = RT_TABLE_DEFAULT;
2665 	struct mr_table *mrt;
2666 	struct nlattr *attr;
2667 	struct rtmsg *rtm;
2668 	int ret, rem;
2669 
2670 	ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2671 					rtm_ipmr_policy, extack);
2672 	if (ret < 0)
2673 		goto out;
2674 	rtm = nlmsg_data(nlh);
2675 
2676 	ret = -EINVAL;
2677 	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2678 	    rtm->rtm_type != RTN_MULTICAST ||
2679 	    rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2680 	    !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2681 		goto out;
2682 
2683 	memset(mfcc, 0, sizeof(*mfcc));
2684 	mfcc->mfcc_parent = -1;
2685 	ret = 0;
2686 	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2687 		switch (nla_type(attr)) {
2688 		case RTA_SRC:
2689 			mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2690 			break;
2691 		case RTA_DST:
2692 			mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2693 			break;
2694 		case RTA_IIF:
2695 			dev = __dev_get_by_index(net, nla_get_u32(attr));
2696 			if (!dev) {
2697 				ret = -ENODEV;
2698 				goto out;
2699 			}
2700 			break;
2701 		case RTA_MULTIPATH:
2702 			if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2703 				ret = -EINVAL;
2704 				goto out;
2705 			}
2706 			break;
2707 		case RTA_PREFSRC:
2708 			ret = 1;
2709 			break;
2710 		case RTA_TABLE:
2711 			tblid = nla_get_u32(attr);
2712 			break;
2713 		}
2714 	}
2715 	mrt = ipmr_get_table(net, tblid);
2716 	if (!mrt) {
2717 		ret = -ENOENT;
2718 		goto out;
2719 	}
2720 	*mrtret = mrt;
2721 	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2722 	if (dev)
2723 		mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2724 
2725 out:
2726 	return ret;
2727 }
2728 
2729 /* takes care of both newroute and delroute */
2730 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2731 			  struct netlink_ext_ack *extack)
2732 {
2733 	struct net *net = sock_net(skb->sk);
2734 	int ret, mrtsock, parent;
2735 	struct mr_table *tbl;
2736 	struct mfcctl mfcc;
2737 
2738 	mrtsock = 0;
2739 	tbl = NULL;
2740 	ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2741 	if (ret < 0)
2742 		return ret;
2743 
2744 	parent = ret ? mfcc.mfcc_parent : -1;
2745 	if (nlh->nlmsg_type == RTM_NEWROUTE)
2746 		return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2747 	else
2748 		return ipmr_mfc_delete(tbl, &mfcc, parent);
2749 }
2750 
2751 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2752 {
2753 	u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2754 
2755 	if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2756 	    nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2757 	    nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2758 			mrt->mroute_reg_vif_num) ||
2759 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2760 		       mrt->mroute_do_assert) ||
2761 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2762 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2763 		       mrt->mroute_do_wrvifwhole))
2764 		return false;
2765 
2766 	return true;
2767 }
2768 
2769 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2770 {
2771 	struct net_device *vif_dev;
2772 	struct nlattr *vif_nest;
2773 	struct vif_device *vif;
2774 
2775 	vif = &mrt->vif_table[vifid];
2776 	vif_dev = rtnl_dereference(vif->dev);
2777 	/* if the VIF doesn't exist just continue */
2778 	if (!vif_dev)
2779 		return true;
2780 
2781 	vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2782 	if (!vif_nest)
2783 		return false;
2784 
2785 	if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2786 	    nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2787 	    nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2788 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2789 			      IPMRA_VIFA_PAD) ||
2790 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2791 			      IPMRA_VIFA_PAD) ||
2792 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2793 			      IPMRA_VIFA_PAD) ||
2794 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2795 			      IPMRA_VIFA_PAD) ||
2796 	    nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2797 	    nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2798 		nla_nest_cancel(skb, vif_nest);
2799 		return false;
2800 	}
2801 	nla_nest_end(skb, vif_nest);
2802 
2803 	return true;
2804 }
2805 
2806 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2807 			       struct netlink_ext_ack *extack)
2808 {
2809 	struct ifinfomsg *ifm;
2810 
2811 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2812 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2813 		return -EINVAL;
2814 	}
2815 
2816 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2817 		NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2818 		return -EINVAL;
2819 	}
2820 
2821 	ifm = nlmsg_data(nlh);
2822 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2823 	    ifm->ifi_change || ifm->ifi_index) {
2824 		NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2825 		return -EINVAL;
2826 	}
2827 
2828 	return 0;
2829 }
2830 
2831 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2832 {
2833 	struct net *net = sock_net(skb->sk);
2834 	struct nlmsghdr *nlh = NULL;
2835 	unsigned int t = 0, s_t;
2836 	unsigned int e = 0, s_e;
2837 	struct mr_table *mrt;
2838 
2839 	if (cb->strict_check) {
2840 		int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2841 
2842 		if (err < 0)
2843 			return err;
2844 	}
2845 
2846 	s_t = cb->args[0];
2847 	s_e = cb->args[1];
2848 
2849 	ipmr_for_each_table(mrt, net) {
2850 		struct nlattr *vifs, *af;
2851 		struct ifinfomsg *hdr;
2852 		u32 i;
2853 
2854 		if (t < s_t)
2855 			goto skip_table;
2856 		nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2857 				cb->nlh->nlmsg_seq, RTM_NEWLINK,
2858 				sizeof(*hdr), NLM_F_MULTI);
2859 		if (!nlh)
2860 			break;
2861 
2862 		hdr = nlmsg_data(nlh);
2863 		memset(hdr, 0, sizeof(*hdr));
2864 		hdr->ifi_family = RTNL_FAMILY_IPMR;
2865 
2866 		af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2867 		if (!af) {
2868 			nlmsg_cancel(skb, nlh);
2869 			goto out;
2870 		}
2871 
2872 		if (!ipmr_fill_table(mrt, skb)) {
2873 			nlmsg_cancel(skb, nlh);
2874 			goto out;
2875 		}
2876 
2877 		vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2878 		if (!vifs) {
2879 			nla_nest_end(skb, af);
2880 			nlmsg_end(skb, nlh);
2881 			goto out;
2882 		}
2883 		for (i = 0; i < mrt->maxvif; i++) {
2884 			if (e < s_e)
2885 				goto skip_entry;
2886 			if (!ipmr_fill_vif(mrt, i, skb)) {
2887 				nla_nest_end(skb, vifs);
2888 				nla_nest_end(skb, af);
2889 				nlmsg_end(skb, nlh);
2890 				goto out;
2891 			}
2892 skip_entry:
2893 			e++;
2894 		}
2895 		s_e = 0;
2896 		e = 0;
2897 		nla_nest_end(skb, vifs);
2898 		nla_nest_end(skb, af);
2899 		nlmsg_end(skb, nlh);
2900 skip_table:
2901 		t++;
2902 	}
2903 
2904 out:
2905 	cb->args[1] = e;
2906 	cb->args[0] = t;
2907 
2908 	return skb->len;
2909 }
2910 
2911 #ifdef CONFIG_PROC_FS
2912 /* The /proc interfaces to multicast routing :
2913  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2914  */
2915 
2916 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2917 	__acquires(RCU)
2918 {
2919 	struct mr_vif_iter *iter = seq->private;
2920 	struct net *net = seq_file_net(seq);
2921 	struct mr_table *mrt;
2922 
2923 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2924 	if (!mrt)
2925 		return ERR_PTR(-ENOENT);
2926 
2927 	iter->mrt = mrt;
2928 
2929 	rcu_read_lock();
2930 	return mr_vif_seq_start(seq, pos);
2931 }
2932 
2933 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2934 	__releases(RCU)
2935 {
2936 	rcu_read_unlock();
2937 }
2938 
2939 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2940 {
2941 	struct mr_vif_iter *iter = seq->private;
2942 	struct mr_table *mrt = iter->mrt;
2943 
2944 	if (v == SEQ_START_TOKEN) {
2945 		seq_puts(seq,
2946 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2947 	} else {
2948 		const struct vif_device *vif = v;
2949 		const struct net_device *vif_dev;
2950 		const char *name;
2951 
2952 		vif_dev = vif_dev_read(vif);
2953 		name = vif_dev ? vif_dev->name : "none";
2954 		seq_printf(seq,
2955 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2956 			   vif - mrt->vif_table,
2957 			   name, vif->bytes_in, vif->pkt_in,
2958 			   vif->bytes_out, vif->pkt_out,
2959 			   vif->flags, vif->local, vif->remote);
2960 	}
2961 	return 0;
2962 }
2963 
2964 static const struct seq_operations ipmr_vif_seq_ops = {
2965 	.start = ipmr_vif_seq_start,
2966 	.next  = mr_vif_seq_next,
2967 	.stop  = ipmr_vif_seq_stop,
2968 	.show  = ipmr_vif_seq_show,
2969 };
2970 
2971 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2972 {
2973 	struct net *net = seq_file_net(seq);
2974 	struct mr_table *mrt;
2975 
2976 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2977 	if (!mrt)
2978 		return ERR_PTR(-ENOENT);
2979 
2980 	return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2981 }
2982 
2983 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2984 {
2985 	int n;
2986 
2987 	if (v == SEQ_START_TOKEN) {
2988 		seq_puts(seq,
2989 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2990 	} else {
2991 		const struct mfc_cache *mfc = v;
2992 		const struct mr_mfc_iter *it = seq->private;
2993 		const struct mr_table *mrt = it->mrt;
2994 
2995 		seq_printf(seq, "%08X %08X %-3hd",
2996 			   (__force u32) mfc->mfc_mcastgrp,
2997 			   (__force u32) mfc->mfc_origin,
2998 			   mfc->_c.mfc_parent);
2999 
3000 		if (it->cache != &mrt->mfc_unres_queue) {
3001 			seq_printf(seq, " %8lu %8lu %8lu",
3002 				   mfc->_c.mfc_un.res.pkt,
3003 				   mfc->_c.mfc_un.res.bytes,
3004 				   mfc->_c.mfc_un.res.wrong_if);
3005 			for (n = mfc->_c.mfc_un.res.minvif;
3006 			     n < mfc->_c.mfc_un.res.maxvif; n++) {
3007 				if (VIF_EXISTS(mrt, n) &&
3008 				    mfc->_c.mfc_un.res.ttls[n] < 255)
3009 					seq_printf(seq,
3010 					   " %2d:%-3d",
3011 					   n, mfc->_c.mfc_un.res.ttls[n]);
3012 			}
3013 		} else {
3014 			/* unresolved mfc_caches don't contain
3015 			 * pkt, bytes and wrong_if values
3016 			 */
3017 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3018 		}
3019 		seq_putc(seq, '\n');
3020 	}
3021 	return 0;
3022 }
3023 
3024 static const struct seq_operations ipmr_mfc_seq_ops = {
3025 	.start = ipmr_mfc_seq_start,
3026 	.next  = mr_mfc_seq_next,
3027 	.stop  = mr_mfc_seq_stop,
3028 	.show  = ipmr_mfc_seq_show,
3029 };
3030 #endif
3031 
3032 #ifdef CONFIG_IP_PIMSM_V2
3033 static const struct net_protocol pim_protocol = {
3034 	.handler	=	pim_rcv,
3035 };
3036 #endif
3037 
3038 static unsigned int ipmr_seq_read(struct net *net)
3039 {
3040 	ASSERT_RTNL();
3041 
3042 	return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3043 }
3044 
3045 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3046 		     struct netlink_ext_ack *extack)
3047 {
3048 	return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3049 		       ipmr_mr_table_iter, extack);
3050 }
3051 
3052 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3053 	.family		= RTNL_FAMILY_IPMR,
3054 	.fib_seq_read	= ipmr_seq_read,
3055 	.fib_dump	= ipmr_dump,
3056 	.owner		= THIS_MODULE,
3057 };
3058 
3059 static int __net_init ipmr_notifier_init(struct net *net)
3060 {
3061 	struct fib_notifier_ops *ops;
3062 
3063 	net->ipv4.ipmr_seq = 0;
3064 
3065 	ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3066 	if (IS_ERR(ops))
3067 		return PTR_ERR(ops);
3068 	net->ipv4.ipmr_notifier_ops = ops;
3069 
3070 	return 0;
3071 }
3072 
3073 static void __net_exit ipmr_notifier_exit(struct net *net)
3074 {
3075 	fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3076 	net->ipv4.ipmr_notifier_ops = NULL;
3077 }
3078 
3079 /* Setup for IP multicast routing */
3080 static int __net_init ipmr_net_init(struct net *net)
3081 {
3082 	int err;
3083 
3084 	err = ipmr_notifier_init(net);
3085 	if (err)
3086 		goto ipmr_notifier_fail;
3087 
3088 	err = ipmr_rules_init(net);
3089 	if (err < 0)
3090 		goto ipmr_rules_fail;
3091 
3092 #ifdef CONFIG_PROC_FS
3093 	err = -ENOMEM;
3094 	if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3095 			sizeof(struct mr_vif_iter)))
3096 		goto proc_vif_fail;
3097 	if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3098 			sizeof(struct mr_mfc_iter)))
3099 		goto proc_cache_fail;
3100 #endif
3101 	return 0;
3102 
3103 #ifdef CONFIG_PROC_FS
3104 proc_cache_fail:
3105 	remove_proc_entry("ip_mr_vif", net->proc_net);
3106 proc_vif_fail:
3107 	rtnl_lock();
3108 	ipmr_rules_exit(net);
3109 	rtnl_unlock();
3110 #endif
3111 ipmr_rules_fail:
3112 	ipmr_notifier_exit(net);
3113 ipmr_notifier_fail:
3114 	return err;
3115 }
3116 
3117 static void __net_exit ipmr_net_exit(struct net *net)
3118 {
3119 #ifdef CONFIG_PROC_FS
3120 	remove_proc_entry("ip_mr_cache", net->proc_net);
3121 	remove_proc_entry("ip_mr_vif", net->proc_net);
3122 #endif
3123 	ipmr_notifier_exit(net);
3124 }
3125 
3126 static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3127 {
3128 	struct net *net;
3129 
3130 	rtnl_lock();
3131 	list_for_each_entry(net, net_list, exit_list)
3132 		ipmr_rules_exit(net);
3133 	rtnl_unlock();
3134 }
3135 
3136 static struct pernet_operations ipmr_net_ops = {
3137 	.init = ipmr_net_init,
3138 	.exit = ipmr_net_exit,
3139 	.exit_batch = ipmr_net_exit_batch,
3140 };
3141 
3142 int __init ip_mr_init(void)
3143 {
3144 	int err;
3145 
3146 	mrt_cachep = KMEM_CACHE(mfc_cache, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3147 
3148 	err = register_pernet_subsys(&ipmr_net_ops);
3149 	if (err)
3150 		goto reg_pernet_fail;
3151 
3152 	err = register_netdevice_notifier(&ip_mr_notifier);
3153 	if (err)
3154 		goto reg_notif_fail;
3155 #ifdef CONFIG_IP_PIMSM_V2
3156 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3157 		pr_err("%s: can't add PIM protocol\n", __func__);
3158 		err = -EAGAIN;
3159 		goto add_proto_fail;
3160 	}
3161 #endif
3162 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3163 		      ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3164 	rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3165 		      ipmr_rtm_route, NULL, 0);
3166 	rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3167 		      ipmr_rtm_route, NULL, 0);
3168 
3169 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3170 		      NULL, ipmr_rtm_dumplink, 0);
3171 	return 0;
3172 
3173 #ifdef CONFIG_IP_PIMSM_V2
3174 add_proto_fail:
3175 	unregister_netdevice_notifier(&ip_mr_notifier);
3176 #endif
3177 reg_notif_fail:
3178 	unregister_pernet_subsys(&ipmr_net_ops);
3179 reg_pernet_fail:
3180 	kmem_cache_destroy(mrt_cachep);
3181 	return err;
3182 }
3183