xref: /linux/net/ipv4/ipmr.c (revision 787268686c64be60023743b527645a907e38ccca)
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/flow.h>
46 #include <net/net_namespace.h>
47 #include <net/ip.h>
48 #include <net/protocol.h>
49 #include <linux/skbuff.h>
50 #include <net/route.h>
51 #include <net/icmp.h>
52 #include <net/udp.h>
53 #include <net/raw.h>
54 #include <linux/notifier.h>
55 #include <linux/if_arp.h>
56 #include <linux/netfilter_ipv4.h>
57 #include <linux/compat.h>
58 #include <linux/export.h>
59 #include <linux/rhashtable.h>
60 #include <net/ip_tunnels.h>
61 #include <net/checksum.h>
62 #include <net/netlink.h>
63 #include <net/fib_rules.h>
64 #include <linux/netconf.h>
65 #include <net/rtnh.h>
66 #include <net/inet_dscp.h>
67 
68 #include <linux/nospec.h>
69 
70 struct ipmr_rule {
71 	struct fib_rule		common;
72 };
73 
74 struct ipmr_result {
75 	struct mr_table		*mrt;
76 };
77 
78 /* Big lock, protecting vif table, mrt cache and mroute socket state.
79  * Note that the changes are semaphored via rtnl_lock.
80  */
81 
82 static DEFINE_SPINLOCK(mrt_lock);
83 
84 static struct net_device *vif_dev_read(const struct vif_device *vif)
85 {
86 	return rcu_dereference(vif->dev);
87 }
88 
89 /* Multicast router control variables */
90 
91 /* Special spinlock for queue of unresolved entries */
92 static DEFINE_SPINLOCK(mfc_unres_lock);
93 
94 /* We return to original Alan's scheme. Hash table of resolved
95  * entries is changed only in process context and protected
96  * with weak lock mrt_lock. Queue of unresolved entries is protected
97  * with strong spinlock mfc_unres_lock.
98  *
99  * In this case data path is free of exclusive locks at all.
100  */
101 
102 static struct kmem_cache *mrt_cachep __ro_after_init;
103 
104 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
105 static void ipmr_free_table(struct mr_table *mrt,
106 			    struct list_head *dev_kill_list);
107 
108 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
109 			  struct net_device *dev, struct sk_buff *skb,
110 			  struct mfc_cache *cache, int local);
111 static int ipmr_cache_report(const struct mr_table *mrt,
112 			     struct sk_buff *pkt, vifi_t vifi, int assert,
113 			     bool nl_query);
114 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
115 				 int cmd);
116 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
117 static void mroute_clean_tables(struct mr_table *mrt, int flags,
118 				struct list_head *dev_kill_list);
119 static void ipmr_expire_process(struct timer_list *t);
120 
121 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
122 #define ipmr_for_each_table(mrt, net)					\
123 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,	\
124 				lockdep_rtnl_is_held() ||		\
125 				list_empty(&net->ipv4.mr_tables))
126 
127 static struct mr_table *ipmr_mr_table_iter(struct net *net,
128 					   struct mr_table *mrt)
129 {
130 	struct mr_table *ret;
131 
132 	if (!mrt)
133 		ret = list_entry_rcu(net->ipv4.mr_tables.next,
134 				     struct mr_table, list);
135 	else
136 		ret = list_entry_rcu(mrt->list.next,
137 				     struct mr_table, list);
138 
139 	if (&ret->list == &net->ipv4.mr_tables)
140 		return NULL;
141 	return ret;
142 }
143 
144 static struct mr_table *__ipmr_get_table(struct net *net, u32 id)
145 {
146 	struct mr_table *mrt;
147 
148 	ipmr_for_each_table(mrt, net) {
149 		if (mrt->id == id)
150 			return mrt;
151 	}
152 	return NULL;
153 }
154 
155 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
156 			   struct mr_table **mrt)
157 {
158 	int err;
159 	struct ipmr_result res;
160 	struct fib_lookup_arg arg = {
161 		.result = &res,
162 		.flags = FIB_LOOKUP_NOREF,
163 	};
164 
165 	/* update flow if oif or iif point to device enslaved to l3mdev */
166 	l3mdev_update_flow(net, flowi4_to_flowi(flp4));
167 
168 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
169 			       flowi4_to_flowi(flp4), 0, &arg);
170 	if (err < 0)
171 		return err;
172 	*mrt = res.mrt;
173 	return 0;
174 }
175 
176 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
177 			    int flags, struct fib_lookup_arg *arg)
178 {
179 	struct ipmr_result *res = arg->result;
180 	struct mr_table *mrt;
181 
182 	switch (rule->action) {
183 	case FR_ACT_TO_TBL:
184 		break;
185 	case FR_ACT_UNREACHABLE:
186 		return -ENETUNREACH;
187 	case FR_ACT_PROHIBIT:
188 		return -EACCES;
189 	case FR_ACT_BLACKHOLE:
190 	default:
191 		return -EINVAL;
192 	}
193 
194 	arg->table = fib_rule_get_table(rule, arg);
195 
196 	mrt = __ipmr_get_table(rule->fr_net, arg->table);
197 	if (!mrt)
198 		return -EAGAIN;
199 	res->mrt = mrt;
200 	return 0;
201 }
202 
203 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
204 {
205 	return 1;
206 }
207 
208 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
209 			       struct fib_rule_hdr *frh, struct nlattr **tb,
210 			       struct netlink_ext_ack *extack)
211 {
212 	return 0;
213 }
214 
215 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
216 			     struct nlattr **tb)
217 {
218 	return 1;
219 }
220 
221 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
222 			  struct fib_rule_hdr *frh)
223 {
224 	frh->dst_len = 0;
225 	frh->src_len = 0;
226 	frh->tos     = 0;
227 	return 0;
228 }
229 
230 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
231 	.family		= RTNL_FAMILY_IPMR,
232 	.rule_size	= sizeof(struct ipmr_rule),
233 	.addr_size	= sizeof(u32),
234 	.action		= ipmr_rule_action,
235 	.match		= ipmr_rule_match,
236 	.configure	= ipmr_rule_configure,
237 	.compare	= ipmr_rule_compare,
238 	.fill		= ipmr_rule_fill,
239 	.nlgroup	= RTNLGRP_IPV4_RULE,
240 	.owner		= THIS_MODULE,
241 };
242 
243 static int __net_init ipmr_rules_init(struct net *net)
244 {
245 	struct fib_rules_ops *ops;
246 	LIST_HEAD(dev_kill_list);
247 	struct mr_table *mrt;
248 	int err;
249 
250 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
251 	if (IS_ERR(ops))
252 		return PTR_ERR(ops);
253 
254 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
255 
256 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
257 	if (IS_ERR(mrt)) {
258 		err = PTR_ERR(mrt);
259 		goto err1;
260 	}
261 
262 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT);
263 	if (err < 0)
264 		goto err2;
265 
266 	net->ipv4.mr_rules_ops = ops;
267 	return 0;
268 
269 err2:
270 	ipmr_free_table(mrt, &dev_kill_list);
271 err1:
272 	fib_rules_unregister(ops);
273 	return err;
274 }
275 
276 static void __net_exit ipmr_rules_exit(struct net *net)
277 {
278 	fib_rules_unregister(net->ipv4.mr_rules_ops);
279 }
280 
281 static void __net_exit ipmr_rules_exit_rtnl(struct net *net,
282 					    struct list_head *dev_kill_list)
283 {
284 	struct mr_table *mrt, *next;
285 
286 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
287 		list_del_rcu(&mrt->list);
288 		ipmr_free_table(mrt, dev_kill_list);
289 	}
290 }
291 
292 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
293 			   struct netlink_ext_ack *extack)
294 {
295 	return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
296 }
297 
298 static unsigned int ipmr_rules_seq_read(const struct net *net)
299 {
300 	return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
301 }
302 
303 bool ipmr_rule_default(const struct fib_rule *rule)
304 {
305 	return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
306 }
307 EXPORT_SYMBOL(ipmr_rule_default);
308 #else
309 static struct mr_table *ipmr_mr_table_iter(struct net *net,
310 					   struct mr_table *mrt)
311 {
312 	if (!mrt)
313 		return rcu_dereference(net->ipv4.mrt);
314 	return NULL;
315 }
316 
317 static struct mr_table *__ipmr_get_table(struct net *net, u32 id)
318 {
319 	return rcu_dereference_check(net->ipv4.mrt,
320 				     lockdep_rtnl_is_held() ||
321 				     !rcu_access_pointer(net->ipv4.mrt));
322 }
323 
324 #define ipmr_for_each_table(mrt, net)				\
325 	for (mrt = __ipmr_get_table(net, 0); mrt; mrt = NULL)
326 
327 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
328 			   struct mr_table **mrt)
329 {
330 	*mrt = rcu_dereference(net->ipv4.mrt);
331 	if (!*mrt)
332 		return -EAGAIN;
333 	return 0;
334 }
335 
336 static int __net_init ipmr_rules_init(struct net *net)
337 {
338 	struct mr_table *mrt;
339 
340 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
341 	if (IS_ERR(mrt))
342 		return PTR_ERR(mrt);
343 
344 	rcu_assign_pointer(net->ipv4.mrt, mrt);
345 	return 0;
346 }
347 
348 static void __net_exit ipmr_rules_exit(struct net *net)
349 {
350 }
351 
352 static void __net_exit ipmr_rules_exit_rtnl(struct net *net,
353 					    struct list_head *dev_kill_list)
354 {
355 	struct mr_table *mrt = rcu_dereference_protected(net->ipv4.mrt, 1);
356 
357 	RCU_INIT_POINTER(net->ipv4.mrt, NULL);
358 	ipmr_free_table(mrt, dev_kill_list);
359 }
360 
361 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
362 			   struct netlink_ext_ack *extack)
363 {
364 	return 0;
365 }
366 
367 static unsigned int ipmr_rules_seq_read(const struct net *net)
368 {
369 	return 0;
370 }
371 
372 bool ipmr_rule_default(const struct fib_rule *rule)
373 {
374 	return true;
375 }
376 EXPORT_SYMBOL(ipmr_rule_default);
377 #endif
378 
379 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
380 {
381 	struct mr_table *mrt;
382 
383 	rcu_read_lock();
384 	mrt = __ipmr_get_table(net, id);
385 	rcu_read_unlock();
386 
387 	return mrt;
388 }
389 
390 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
391 				const void *ptr)
392 {
393 	const struct mfc_cache_cmp_arg *cmparg = arg->key;
394 	const struct mfc_cache *c = ptr;
395 
396 	return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
397 	       cmparg->mfc_origin != c->mfc_origin;
398 }
399 
400 static const struct rhashtable_params ipmr_rht_params = {
401 	.head_offset = offsetof(struct mr_mfc, mnode),
402 	.key_offset = offsetof(struct mfc_cache, cmparg),
403 	.key_len = sizeof(struct mfc_cache_cmp_arg),
404 	.nelem_hint = 3,
405 	.obj_cmpfn = ipmr_hash_cmp,
406 	.automatic_shrinking = true,
407 };
408 
409 static void ipmr_new_table_set(struct mr_table *mrt,
410 			       struct net *net)
411 {
412 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
413 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
414 #endif
415 }
416 
417 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
418 	.mfc_mcastgrp = htonl(INADDR_ANY),
419 	.mfc_origin = htonl(INADDR_ANY),
420 };
421 
422 static struct mr_table_ops ipmr_mr_table_ops = {
423 	.rht_params = &ipmr_rht_params,
424 	.cmparg_any = &ipmr_mr_table_ops_cmparg_any,
425 };
426 
427 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
428 {
429 	struct mr_table *mrt;
430 
431 	/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
432 	if (id != RT_TABLE_DEFAULT && id >= 1000000000)
433 		return ERR_PTR(-EINVAL);
434 
435 	mrt = __ipmr_get_table(net, id);
436 	if (mrt)
437 		return mrt;
438 
439 	return mr_table_alloc(net, id, &ipmr_mr_table_ops,
440 			      ipmr_expire_process, ipmr_new_table_set);
441 }
442 
443 static void ipmr_free_table(struct mr_table *mrt, struct list_head *dev_kill_list)
444 {
445 	struct net *net = read_pnet(&mrt->net);
446 	LIST_HEAD(ipmr_dev_kill_list);
447 
448 	WARN_ON_ONCE(!mr_can_free_table(net));
449 
450 	mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
451 			    MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC,
452 			    &ipmr_dev_kill_list);
453 	timer_shutdown_sync(&mrt->ipmr_expire_timer);
454 	mr_table_free(mrt);
455 
456 	WARN_ON_ONCE(!net_initialized(net) && !list_empty(&ipmr_dev_kill_list));
457 	list_splice(&ipmr_dev_kill_list, dev_kill_list);
458 }
459 
460 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
461 
462 /* Initialize ipmr pimreg/tunnel in_device */
463 static bool ipmr_init_vif_indev(const struct net_device *dev)
464 {
465 	struct in_device *in_dev;
466 
467 	ASSERT_RTNL();
468 
469 	in_dev = __in_dev_get_rtnl(dev);
470 	if (!in_dev)
471 		return false;
472 	ipv4_devconf_setall(in_dev);
473 	neigh_parms_data_state_setall(in_dev->arp_parms);
474 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
475 
476 	return true;
477 }
478 
479 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
480 {
481 	struct net_device *tunnel_dev, *new_dev;
482 	struct ip_tunnel_parm_kern p = { };
483 	LIST_HEAD(dev_kill_list);
484 	int err;
485 
486 	tunnel_dev = __dev_get_by_name(net, "tunl0");
487 	if (!tunnel_dev)
488 		goto out;
489 
490 	p.iph.daddr = v->vifc_rmt_addr.s_addr;
491 	p.iph.saddr = v->vifc_lcl_addr.s_addr;
492 	p.iph.version = 4;
493 	p.iph.ihl = 5;
494 	p.iph.protocol = IPPROTO_IPIP;
495 	sprintf(p.name, "dvmrp%d", v->vifc_vifi);
496 
497 	if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
498 		goto out;
499 	err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
500 			SIOCADDTUNNEL);
501 	if (err)
502 		goto out;
503 
504 	new_dev = __dev_get_by_name(net, p.name);
505 	if (!new_dev)
506 		goto out;
507 
508 	new_dev->flags |= IFF_MULTICAST;
509 	if (!ipmr_init_vif_indev(new_dev))
510 		goto out_unregister;
511 	if (dev_open(new_dev, NULL))
512 		goto out_unregister;
513 	dev_hold(new_dev);
514 	err = dev_set_allmulti(new_dev, 1);
515 	if (err) {
516 		dev_close(new_dev);
517 		tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
518 				SIOCDELTUNNEL);
519 		dev_put(new_dev);
520 		new_dev = ERR_PTR(err);
521 	}
522 	return new_dev;
523 
524 out_unregister:
525 	new_dev->rtnl_link_ops->dellink(new_dev, &dev_kill_list);
526 	unregister_netdevice_many(&dev_kill_list);
527 out:
528 	return ERR_PTR(-ENOBUFS);
529 }
530 
531 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
532 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
533 {
534 	struct net *net = dev_net(dev);
535 	struct mr_table *mrt;
536 	struct flowi4 fl4 = {
537 		.flowi4_oif	= dev->ifindex,
538 		.flowi4_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
539 		.flowi4_mark	= skb->mark,
540 	};
541 	int err;
542 
543 	if (skb->protocol != htons(ETH_P_IP) || !pskb_inet_may_pull(skb)) {
544 		DEV_STATS_INC(dev, tx_errors);
545 		kfree_skb(skb);
546 		return NETDEV_TX_OK;
547 	}
548 
549 	rcu_read_lock();
550 	err = ipmr_fib_lookup(net, &fl4, &mrt);
551 	if (err < 0) {
552 		rcu_read_unlock();
553 		kfree_skb(skb);
554 		return err;
555 	}
556 
557 	DEV_STATS_ADD(dev, tx_bytes, skb->len);
558 	DEV_STATS_INC(dev, tx_packets);
559 
560 	/* skb->cb[] holds qdisc/tc state at this point, not IPCB.
561 	 * ipmr_cache_report() reads it as IPCB to build IP_PKTINFO.
562 	 */
563 	memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
564 
565 	/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
566 	ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
567 			  IGMPMSG_WHOLEPKT, false);
568 
569 	rcu_read_unlock();
570 	kfree_skb(skb);
571 	return NETDEV_TX_OK;
572 }
573 
574 static int reg_vif_get_iflink(const struct net_device *dev)
575 {
576 	return 0;
577 }
578 
579 static const struct net_device_ops reg_vif_netdev_ops = {
580 	.ndo_start_xmit	= reg_vif_xmit,
581 	.ndo_get_iflink = reg_vif_get_iflink,
582 };
583 
584 static void reg_vif_setup(struct net_device *dev)
585 {
586 	dev->type		= ARPHRD_PIMREG;
587 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
588 	dev->flags		= IFF_NOARP;
589 	dev->netdev_ops		= &reg_vif_netdev_ops;
590 	dev->needs_free_netdev	= true;
591 	dev->netns_immutable	= true;
592 }
593 
594 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
595 {
596 	struct net_device *dev;
597 	char name[IFNAMSIZ];
598 
599 	if (mrt->id == RT_TABLE_DEFAULT)
600 		sprintf(name, "pimreg");
601 	else
602 		sprintf(name, "pimreg%u", mrt->id);
603 
604 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
605 
606 	if (!dev)
607 		return NULL;
608 
609 	dev_net_set(dev, net);
610 
611 	if (register_netdevice(dev)) {
612 		free_netdev(dev);
613 		return NULL;
614 	}
615 
616 	if (!ipmr_init_vif_indev(dev))
617 		goto failure;
618 	if (dev_open(dev, NULL))
619 		goto failure;
620 
621 	dev_hold(dev);
622 
623 	return dev;
624 
625 failure:
626 	unregister_netdevice(dev);
627 	return NULL;
628 }
629 
630 /* called with rcu_read_lock() */
631 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
632 		     unsigned int pimlen)
633 {
634 	struct net_device *reg_dev = NULL;
635 	struct iphdr *encap;
636 	int vif_num;
637 
638 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
639 	/* Check that:
640 	 * a. packet is really sent to a multicast group
641 	 * b. packet is not a NULL-REGISTER
642 	 * c. packet is not truncated
643 	 */
644 	if (!ipv4_is_multicast(encap->daddr) ||
645 	    encap->tot_len == 0 ||
646 	    ntohs(encap->tot_len) + pimlen > skb->len)
647 		return 1;
648 
649 	/* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
650 	vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
651 	if (vif_num >= 0)
652 		reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
653 	if (!reg_dev)
654 		return 1;
655 
656 	skb->mac_header = skb->network_header;
657 	skb_pull(skb, (u8 *)encap - skb->data);
658 	skb_reset_network_header(skb);
659 	skb->protocol = htons(ETH_P_IP);
660 	skb->ip_summed = CHECKSUM_NONE;
661 
662 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
663 
664 	netif_rx(skb);
665 
666 	return NET_RX_SUCCESS;
667 }
668 #else
669 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
670 {
671 	return NULL;
672 }
673 #endif
674 
675 static int call_ipmr_vif_entry_notifiers(struct net *net,
676 					 enum fib_event_type event_type,
677 					 struct vif_device *vif,
678 					 struct net_device *vif_dev,
679 					 vifi_t vif_index, u32 tb_id)
680 {
681 	return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
682 				     vif, vif_dev, vif_index, tb_id,
683 				     &net->ipv4.ipmr_seq);
684 }
685 
686 static int call_ipmr_mfc_entry_notifiers(struct net *net,
687 					 enum fib_event_type event_type,
688 					 struct mfc_cache *mfc, u32 tb_id)
689 {
690 	return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
691 				     &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
692 }
693 
694 /**
695  *	vif_delete - Delete a VIF entry
696  *	@mrt: Table to delete from
697  *	@vifi: VIF identifier to delete
698  *	@notify: Set to 1, if the caller is a notifier_call
699  *	@head: if unregistering the VIF, place it on this queue
700  */
701 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
702 		      struct list_head *head)
703 {
704 	struct net *net = read_pnet(&mrt->net);
705 	struct vif_device *v;
706 	struct net_device *dev;
707 	struct in_device *in_dev;
708 
709 	if (vifi < 0 || vifi >= mrt->maxvif)
710 		return -EADDRNOTAVAIL;
711 
712 	v = &mrt->vif_table[vifi];
713 
714 	dev = rtnl_dereference(v->dev);
715 	if (!dev)
716 		return -EADDRNOTAVAIL;
717 
718 	spin_lock(&mrt_lock);
719 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
720 				      vifi, mrt->id);
721 	RCU_INIT_POINTER(v->dev, NULL);
722 
723 	if (vifi == mrt->mroute_reg_vif_num) {
724 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
725 		WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
726 	}
727 	if (vifi + 1 == mrt->maxvif) {
728 		int tmp;
729 
730 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
731 			if (VIF_EXISTS(mrt, tmp))
732 				break;
733 		}
734 		WRITE_ONCE(mrt->maxvif, tmp + 1);
735 	}
736 
737 	spin_unlock(&mrt_lock);
738 
739 	dev_set_allmulti(dev, -1);
740 
741 	in_dev = __in_dev_get_rtnl(dev);
742 	if (in_dev) {
743 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
744 		inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
745 					    NETCONFA_MC_FORWARDING,
746 					    dev->ifindex, &in_dev->cnf);
747 		ip_rt_multicast_event(in_dev);
748 	}
749 
750 	if (!notify) {
751 		if (v->flags & VIFF_TUNNEL)
752 			dev->rtnl_link_ops->dellink(dev, head);
753 		else if (v->flags & VIFF_REGISTER)
754 			unregister_netdevice_queue(dev, head);
755 	}
756 
757 	netdev_put(dev, &v->dev_tracker);
758 	return 0;
759 }
760 
761 static void ipmr_cache_free_rcu(struct rcu_head *head)
762 {
763 	struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
764 
765 	kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
766 }
767 
768 static void ipmr_cache_free(struct mfc_cache *c)
769 {
770 	call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
771 }
772 
773 /* Destroy an unresolved cache entry, killing queued skbs
774  * and reporting error to netlink readers.
775  */
776 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
777 {
778 	struct net *net = read_pnet(&mrt->net);
779 	struct sk_buff *skb;
780 	struct nlmsgerr *e;
781 
782 	WRITE_ONCE(mrt->cache_resolve_queue_len,
783 		   mrt->cache_resolve_queue_len - 1);
784 
785 	while ((skb = skb_dequeue(&c->_c.mfc_un.unres.nl_queries))) {
786 		struct nlmsghdr *nlh = skb_pull(skb, sizeof(struct iphdr));
787 
788 		nlh->nlmsg_type = NLMSG_ERROR;
789 		nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
790 		skb_trim(skb, nlh->nlmsg_len);
791 		e = nlmsg_data(nlh);
792 		e->error = -ETIMEDOUT;
793 		memset(&e->msg, 0, sizeof(e->msg));
794 
795 		rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
796 	}
797 
798 	skb_queue_purge(&c->_c.mfc_un.unres.unresolved);
799 
800 	ipmr_cache_free(c);
801 }
802 
803 /* Timer process for the unresolved queue. */
804 static void ipmr_expire_process(struct timer_list *t)
805 {
806 	struct mr_table *mrt = timer_container_of(mrt, t, ipmr_expire_timer);
807 	struct mr_mfc *c, *next;
808 	unsigned long expires;
809 	unsigned long now;
810 
811 	if (!spin_trylock(&mfc_unres_lock)) {
812 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
813 		return;
814 	}
815 
816 	if (list_empty(&mrt->mfc_unres_queue))
817 		goto out;
818 
819 	now = jiffies;
820 	expires = 10*HZ;
821 
822 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
823 		if (time_after(c->mfc_un.unres.expires, now)) {
824 			unsigned long interval = c->mfc_un.unres.expires - now;
825 			if (interval < expires)
826 				expires = interval;
827 			continue;
828 		}
829 
830 		list_del(&c->list);
831 		mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
832 		ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
833 	}
834 
835 	if (!list_empty(&mrt->mfc_unres_queue))
836 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
837 
838 out:
839 	spin_unlock(&mfc_unres_lock);
840 }
841 
842 /* Fill oifs list. It is called under locked mrt_lock. */
843 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
844 				   unsigned char *ttls)
845 {
846 	int vifi;
847 
848 	cache->mfc_un.res.minvif = MAXVIFS;
849 	cache->mfc_un.res.maxvif = 0;
850 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
851 
852 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
853 		if (VIF_EXISTS(mrt, vifi) &&
854 		    ttls[vifi] && ttls[vifi] < 255) {
855 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
856 			if (cache->mfc_un.res.minvif > vifi)
857 				cache->mfc_un.res.minvif = vifi;
858 			if (cache->mfc_un.res.maxvif <= vifi)
859 				cache->mfc_un.res.maxvif = vifi + 1;
860 		}
861 	}
862 	WRITE_ONCE(cache->mfc_un.res.lastuse, jiffies);
863 }
864 
865 static int vif_add(struct net *net, struct mr_table *mrt,
866 		   struct vifctl *vifc, int mrtsock)
867 {
868 	struct netdev_phys_item_id ppid = { };
869 	int vifi = vifc->vifc_vifi;
870 	struct vif_device *v = &mrt->vif_table[vifi];
871 	struct net_device *dev;
872 	struct in_device *in_dev;
873 	int err;
874 
875 	/* Is vif busy ? */
876 	if (VIF_EXISTS(mrt, vifi))
877 		return -EADDRINUSE;
878 
879 	switch (vifc->vifc_flags) {
880 	case VIFF_REGISTER:
881 		if (!ipmr_pimsm_enabled())
882 			return -EINVAL;
883 		/* Special Purpose VIF in PIM
884 		 * All the packets will be sent to the daemon
885 		 */
886 		if (mrt->mroute_reg_vif_num >= 0)
887 			return -EADDRINUSE;
888 		dev = ipmr_reg_vif(net, mrt);
889 		if (!dev)
890 			return -ENOBUFS;
891 		err = dev_set_allmulti(dev, 1);
892 		if (err) {
893 			unregister_netdevice(dev);
894 			dev_put(dev);
895 			return err;
896 		}
897 		break;
898 	case VIFF_TUNNEL:
899 		dev = ipmr_new_tunnel(net, vifc);
900 		if (IS_ERR(dev))
901 			return PTR_ERR(dev);
902 		break;
903 	case VIFF_USE_IFINDEX:
904 	case 0:
905 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
906 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
907 			if (dev && !__in_dev_get_rtnl(dev)) {
908 				dev_put(dev);
909 				return -EADDRNOTAVAIL;
910 			}
911 		} else {
912 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
913 		}
914 		if (!dev)
915 			return -EADDRNOTAVAIL;
916 		err = dev_set_allmulti(dev, 1);
917 		if (err) {
918 			dev_put(dev);
919 			return err;
920 		}
921 		break;
922 	default:
923 		return -EINVAL;
924 	}
925 
926 	in_dev = __in_dev_get_rtnl(dev);
927 	if (!in_dev) {
928 		dev_put(dev);
929 		return -EADDRNOTAVAIL;
930 	}
931 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
932 	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
933 				    dev->ifindex, &in_dev->cnf);
934 	ip_rt_multicast_event(in_dev);
935 
936 	/* Fill in the VIF structures */
937 	vif_device_init(v, dev, vifc->vifc_rate_limit,
938 			vifc->vifc_threshold,
939 			vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
940 			(VIFF_TUNNEL | VIFF_REGISTER));
941 
942 	err = netif_get_port_parent_id(dev, &ppid, true);
943 	if (err == 0) {
944 		memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
945 		v->dev_parent_id.id_len = ppid.id_len;
946 	} else {
947 		v->dev_parent_id.id_len = 0;
948 	}
949 
950 	v->local = vifc->vifc_lcl_addr.s_addr;
951 	v->remote = vifc->vifc_rmt_addr.s_addr;
952 
953 	/* And finish update writing critical data */
954 	spin_lock(&mrt_lock);
955 	rcu_assign_pointer(v->dev, dev);
956 	netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
957 	if (v->flags & VIFF_REGISTER) {
958 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
959 		WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
960 	}
961 	if (vifi+1 > mrt->maxvif)
962 		WRITE_ONCE(mrt->maxvif, vifi + 1);
963 	spin_unlock(&mrt_lock);
964 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
965 				      vifi, mrt->id);
966 	return 0;
967 }
968 
969 /* called with rcu_read_lock() */
970 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
971 					 __be32 origin,
972 					 __be32 mcastgrp)
973 {
974 	struct mfc_cache_cmp_arg arg = {
975 			.mfc_mcastgrp = mcastgrp,
976 			.mfc_origin = origin
977 	};
978 
979 	return mr_mfc_find(mrt, &arg);
980 }
981 
982 /* Look for a (*,G) entry */
983 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
984 					     __be32 mcastgrp, int vifi)
985 {
986 	struct mfc_cache_cmp_arg arg = {
987 			.mfc_mcastgrp = mcastgrp,
988 			.mfc_origin = htonl(INADDR_ANY)
989 	};
990 
991 	if (mcastgrp == htonl(INADDR_ANY))
992 		return mr_mfc_find_any_parent(mrt, vifi);
993 	return mr_mfc_find_any(mrt, vifi, &arg);
994 }
995 
996 /* Look for a (S,G,iif) entry if parent != -1 */
997 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
998 						__be32 origin, __be32 mcastgrp,
999 						int parent)
1000 {
1001 	struct mfc_cache_cmp_arg arg = {
1002 			.mfc_mcastgrp = mcastgrp,
1003 			.mfc_origin = origin,
1004 	};
1005 
1006 	return mr_mfc_find_parent(mrt, &arg, parent);
1007 }
1008 
1009 /* Allocate a multicast cache entry */
1010 static struct mfc_cache *ipmr_cache_alloc(void)
1011 {
1012 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1013 
1014 	if (c) {
1015 		c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1016 		c->_c.mfc_un.res.minvif = MAXVIFS;
1017 		c->_c.free = ipmr_cache_free_rcu;
1018 		refcount_set(&c->_c.mfc_un.res.refcount, 1);
1019 	}
1020 	return c;
1021 }
1022 
1023 static struct mfc_cache *ipmr_cache_alloc_unres(void)
1024 {
1025 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1026 
1027 	if (c) {
1028 		skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1029 		skb_queue_head_init(&c->_c.mfc_un.unres.nl_queries);
1030 		c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1031 	}
1032 	return c;
1033 }
1034 
1035 /* A cache entry has gone into a resolved state from queued */
1036 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1037 			       struct mfc_cache *uc, struct mfc_cache *c)
1038 {
1039 	struct sk_buff *skb;
1040 	struct nlmsgerr *e;
1041 
1042 	/* Answer the pending netlink route queries */
1043 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.nl_queries))) {
1044 		struct nlmsghdr *nlh = skb_pull(skb, sizeof(struct iphdr));
1045 
1046 		if (mr_fill_mroute(mrt, skb, &c->_c, nlmsg_data(nlh)) > 0) {
1047 			nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1048 		} else {
1049 			nlh->nlmsg_type = NLMSG_ERROR;
1050 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1051 			skb_trim(skb, nlh->nlmsg_len);
1052 			e = nlmsg_data(nlh);
1053 			e->error = -EMSGSIZE;
1054 			memset(&e->msg, 0, sizeof(e->msg));
1055 		}
1056 
1057 		rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1058 	}
1059 
1060 	/* Play the pending packets through our router */
1061 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1062 		rcu_read_lock();
1063 		ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1064 		rcu_read_unlock();
1065 	}
1066 }
1067 
1068 /* Bounce a cache query up to mrouted and netlink.
1069  *
1070  * Called under rcu_read_lock().
1071  */
1072 static int ipmr_cache_report(const struct mr_table *mrt,
1073 			     struct sk_buff *pkt, vifi_t vifi, int assert,
1074 			     bool nl_query)
1075 {
1076 	unsigned char pkt_cb[sizeof(pkt->cb)];
1077 	const int ihl = ip_hdrlen(pkt);
1078 	struct sock *mroute_sk;
1079 	struct igmphdr *igmp;
1080 	struct igmpmsg *msg;
1081 	struct sk_buff *skb;
1082 	int ret;
1083 
1084 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1085 	if (!mroute_sk)
1086 		return -EINVAL;
1087 
1088 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1089 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1090 	else
1091 		skb = alloc_skb(128, GFP_ATOMIC);
1092 
1093 	if (!skb)
1094 		return -ENOBUFS;
1095 
1096 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1097 		/* Ugly, but we have no choice with this interface.
1098 		 * Duplicate old header, fix ihl, length etc.
1099 		 * And all this only to mangle msg->im_msgtype and
1100 		 * to set msg->im_mbz to "mbz" :-)
1101 		 */
1102 		skb_push(skb, sizeof(struct iphdr));
1103 		skb_reset_network_header(skb);
1104 		skb_reset_transport_header(skb);
1105 		msg = (struct igmpmsg *)skb_network_header(skb);
1106 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1107 		msg->im_msgtype = assert;
1108 		msg->im_mbz = 0;
1109 		if (assert == IGMPMSG_WRVIFWHOLE) {
1110 			msg->im_vif = vifi;
1111 			msg->im_vif_hi = vifi >> 8;
1112 		} else {
1113 			/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1114 			int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1115 
1116 			msg->im_vif = vif_num;
1117 			msg->im_vif_hi = vif_num >> 8;
1118 		}
1119 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1120 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1121 					     sizeof(struct iphdr));
1122 	} else {
1123 		/* Copy the IP header */
1124 		skb_set_network_header(skb, skb->len);
1125 		skb_put(skb, ihl);
1126 		skb_copy_to_linear_data(skb, pkt->data, ihl);
1127 		/* Flag to the kernel this is a route add */
1128 		ip_hdr(skb)->protocol = 0;
1129 		msg = (struct igmpmsg *)skb_network_header(skb);
1130 		msg->im_vif = vifi;
1131 		msg->im_vif_hi = vifi >> 8;
1132 		/* Add our header.
1133 		 * Note that code, csum and group fields are cleared.
1134 		 */
1135 		igmp = skb_put_zero(skb, sizeof(struct igmphdr));
1136 		igmp->type = assert;
1137 		msg->im_msgtype = assert;
1138 		ip_hdr(skb)->tot_len = htons(skb->len);	/* Fix the length */
1139 		skb->transport_header = skb->network_header;
1140 	}
1141 
1142 	memset(skb->cb, 0, sizeof(skb->cb));
1143 	/* Route queries from ipmr_get_route() carry NETLINK_CB, not IPCB. */
1144 	if (!nl_query) {
1145 		memcpy(pkt_cb, pkt->cb, sizeof(pkt_cb));
1146 		ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1147 		*PKTINFO_SKB_CB(skb) = *PKTINFO_SKB_CB(pkt);
1148 		memcpy(pkt->cb, pkt_cb, sizeof(pkt_cb));
1149 	}
1150 
1151 	igmpmsg_netlink_event(mrt, skb);
1152 
1153 	/* Deliver to mrouted */
1154 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1155 
1156 	if (ret < 0) {
1157 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1158 		kfree_skb(skb);
1159 	}
1160 
1161 	return ret;
1162 }
1163 
1164 /* Queue a packet for resolution. It gets locked cache entry! */
1165 /* Called under rcu_read_lock() */
1166 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1167 				 struct sk_buff *skb, struct net_device *dev,
1168 				 bool nl_query)
1169 {
1170 	struct net *net = read_pnet(&mrt->net);
1171 	const struct iphdr *iph = ip_hdr(skb);
1172 	struct mfc_cache *c = NULL;
1173 	bool found = false;
1174 	int err;
1175 
1176 	spin_lock_bh(&mfc_unres_lock);
1177 
1178 	if (!check_net(net)) {
1179 		err = -EINVAL;
1180 		goto err;
1181 	}
1182 
1183 	list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1184 		if (c->mfc_mcastgrp == iph->daddr &&
1185 		    c->mfc_origin == iph->saddr) {
1186 			found = true;
1187 			break;
1188 		}
1189 	}
1190 
1191 	if (!found) {
1192 		/* Create a new entry if allowable */
1193 		c = ipmr_cache_alloc_unres();
1194 		if (!c) {
1195 			err = -ENOBUFS;
1196 			goto err;
1197 		}
1198 
1199 		/* Fill in the new cache entry */
1200 		c->_c.mfc_parent = -1;
1201 		c->mfc_origin	= iph->saddr;
1202 		c->mfc_mcastgrp	= iph->daddr;
1203 
1204 		/* Reflect first query at mrouted. */
1205 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE,
1206 					nl_query);
1207 		if (err < 0)
1208 			goto err;
1209 
1210 		WRITE_ONCE(mrt->cache_resolve_queue_len,
1211 			   mrt->cache_resolve_queue_len + 1);
1212 		list_add(&c->_c.list, &mrt->mfc_unres_queue);
1213 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1214 
1215 		if (mrt->cache_resolve_queue_len == 1)
1216 			mod_timer(&mrt->ipmr_expire_timer,
1217 				  c->_c.mfc_un.unres.expires);
1218 	}
1219 
1220 	/* See if we can append the packet */
1221 	if (skb_queue_len(&c->_c.mfc_un.unres.unresolved) +
1222 	    skb_queue_len(&c->_c.mfc_un.unres.nl_queries) > 3) {
1223 		c = NULL;
1224 		err = -ENOBUFS;
1225 		goto err;
1226 	}
1227 
1228 	if (dev) {
1229 		skb->dev = dev;
1230 		skb->skb_iif = dev->ifindex;
1231 	}
1232 
1233 	skb_queue_tail(nl_query ? &c->_c.mfc_un.unres.nl_queries :
1234 				  &c->_c.mfc_un.unres.unresolved, skb);
1235 
1236 	spin_unlock_bh(&mfc_unres_lock);
1237 	return 0;
1238 
1239 err:
1240 	spin_unlock_bh(&mfc_unres_lock);
1241 	if (c)
1242 		ipmr_cache_free(c);
1243 	kfree_skb(skb);
1244 	return err;
1245 }
1246 
1247 /* MFC cache manipulation by user space mroute daemon */
1248 
1249 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1250 {
1251 	struct net *net = read_pnet(&mrt->net);
1252 	struct mfc_cache *c;
1253 
1254 	rcu_read_lock();
1255 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1256 				   mfc->mfcc_mcastgrp.s_addr, parent);
1257 	rcu_read_unlock();
1258 	if (!c)
1259 		return -ENOENT;
1260 	rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1261 	list_del_rcu(&c->_c.list);
1262 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1263 	mroute_netlink_event(mrt, c, RTM_DELROUTE);
1264 	mr_cache_put(&c->_c);
1265 
1266 	return 0;
1267 }
1268 
1269 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1270 			struct mfcctl *mfc, int mrtsock, int parent)
1271 {
1272 	struct mfc_cache *uc, *c;
1273 	struct mr_mfc *_uc;
1274 	bool found;
1275 	int ret;
1276 
1277 	if (mfc->mfcc_parent >= MAXVIFS)
1278 		return -ENFILE;
1279 
1280 	rcu_read_lock();
1281 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1282 				   mfc->mfcc_mcastgrp.s_addr, parent);
1283 	rcu_read_unlock();
1284 	if (c) {
1285 		spin_lock(&mrt_lock);
1286 		c->_c.mfc_parent = mfc->mfcc_parent;
1287 		ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1288 		if (!mrtsock)
1289 			c->_c.mfc_flags |= MFC_STATIC;
1290 		spin_unlock(&mrt_lock);
1291 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1292 					      mrt->id);
1293 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1294 		return 0;
1295 	}
1296 
1297 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1298 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1299 		return -EINVAL;
1300 
1301 	c = ipmr_cache_alloc();
1302 	if (!c)
1303 		return -ENOMEM;
1304 
1305 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1306 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1307 	c->_c.mfc_parent = mfc->mfcc_parent;
1308 	ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1309 	if (!mrtsock)
1310 		c->_c.mfc_flags |= MFC_STATIC;
1311 
1312 	ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1313 				  ipmr_rht_params);
1314 	if (ret) {
1315 		pr_err("ipmr: rhtable insert error %d\n", ret);
1316 		ipmr_cache_free(c);
1317 		return ret;
1318 	}
1319 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1320 	/* Check to see if we resolved a queued list. If so we
1321 	 * need to send on the frames and tidy up.
1322 	 */
1323 	found = false;
1324 	spin_lock_bh(&mfc_unres_lock);
1325 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1326 		uc = (struct mfc_cache *)_uc;
1327 		if (uc->mfc_origin == c->mfc_origin &&
1328 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1329 			list_del(&_uc->list);
1330 			WRITE_ONCE(mrt->cache_resolve_queue_len,
1331 				   mrt->cache_resolve_queue_len - 1);
1332 			found = true;
1333 			break;
1334 		}
1335 	}
1336 	if (list_empty(&mrt->mfc_unres_queue))
1337 		timer_delete(&mrt->ipmr_expire_timer);
1338 	spin_unlock_bh(&mfc_unres_lock);
1339 
1340 	if (found) {
1341 		ipmr_cache_resolve(net, mrt, uc, c);
1342 		ipmr_cache_free(uc);
1343 	}
1344 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1345 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1346 	return 0;
1347 }
1348 
1349 /* Close the multicast socket, and clear the vif tables etc */
1350 static void mroute_clean_tables(struct mr_table *mrt, int flags,
1351 				struct list_head *dev_kill_list)
1352 {
1353 	struct net *net = read_pnet(&mrt->net);
1354 	struct mfc_cache *cache;
1355 	struct mr_mfc *c, *tmp;
1356 	int i;
1357 
1358 	/* Shut down all active vif entries */
1359 	if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1360 		for (i = 0; i < mrt->maxvif; i++) {
1361 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1362 			     !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1363 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1364 				continue;
1365 			vif_delete(mrt, i, 0, dev_kill_list);
1366 		}
1367 	}
1368 
1369 	/* Wipe the cache */
1370 	if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1371 		mutex_lock(&net->ipv4.mfc_mutex);
1372 
1373 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1374 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1375 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1376 				continue;
1377 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1378 			list_del_rcu(&c->list);
1379 			cache = (struct mfc_cache *)c;
1380 			call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1381 						      mrt->id);
1382 			mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1383 			mr_cache_put(c);
1384 		}
1385 
1386 		mutex_unlock(&net->ipv4.mfc_mutex);
1387 	}
1388 
1389 	if (flags & MRT_FLUSH_MFC) {
1390 		if (READ_ONCE(mrt->cache_resolve_queue_len) || !check_net(net)) {
1391 			spin_lock_bh(&mfc_unres_lock);
1392 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1393 				list_del(&c->list);
1394 				cache = (struct mfc_cache *)c;
1395 				mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1396 				ipmr_destroy_unres(mrt, cache);
1397 			}
1398 			spin_unlock_bh(&mfc_unres_lock);
1399 		}
1400 	}
1401 }
1402 
1403 /* called from ip_ra_control(), before an RCU grace period,
1404  * we don't need to call synchronize_rcu() here
1405  */
1406 static void mrtsock_destruct(struct sock *sk)
1407 {
1408 	struct net *net = sock_net(sk);
1409 	LIST_HEAD(dev_kill_list);
1410 	struct mr_table *mrt;
1411 
1412 	rtnl_lock();
1413 
1414 	ipmr_for_each_table(mrt, net) {
1415 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1416 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1417 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1418 						    NETCONFA_MC_FORWARDING,
1419 						    NETCONFA_IFINDEX_ALL,
1420 						    net->ipv4.devconf_all);
1421 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1422 			mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC,
1423 					    &dev_kill_list);
1424 		}
1425 	}
1426 
1427 	unregister_netdevice_many(&dev_kill_list);
1428 
1429 	rtnl_unlock();
1430 }
1431 
1432 /* Socket options and virtual interface manipulation. The whole
1433  * virtual interface system is a complete heap, but unfortunately
1434  * that's how BSD mrouted happens to think. Maybe one day with a proper
1435  * MOSPF/PIM router set up we can clean this up.
1436  */
1437 
1438 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1439 			 unsigned int optlen)
1440 {
1441 	struct net *net = sock_net(sk);
1442 	int val, ret = 0, parent = 0;
1443 	struct mr_table *mrt;
1444 	struct vifctl vif;
1445 	struct mfcctl mfc;
1446 	bool do_wrvifwhole;
1447 	u32 uval;
1448 
1449 	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1450 	rtnl_lock();
1451 	if (sk->sk_type != SOCK_RAW ||
1452 	    inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1453 		ret = -EOPNOTSUPP;
1454 		goto out_unlock;
1455 	}
1456 
1457 	mrt = __ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1458 	if (!mrt) {
1459 		ret = -ENOENT;
1460 		goto out_unlock;
1461 	}
1462 	if (optname != MRT_INIT) {
1463 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1464 		    !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1465 			ret = -EACCES;
1466 			goto out_unlock;
1467 		}
1468 	}
1469 
1470 	switch (optname) {
1471 	case MRT_INIT:
1472 		if (optlen != sizeof(int)) {
1473 			ret = -EINVAL;
1474 			break;
1475 		}
1476 		if (rtnl_dereference(mrt->mroute_sk)) {
1477 			ret = -EADDRINUSE;
1478 			break;
1479 		}
1480 
1481 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1482 		if (ret == 0) {
1483 			rcu_assign_pointer(mrt->mroute_sk, sk);
1484 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1485 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1486 						    NETCONFA_MC_FORWARDING,
1487 						    NETCONFA_IFINDEX_ALL,
1488 						    net->ipv4.devconf_all);
1489 		}
1490 		break;
1491 	case MRT_DONE:
1492 		if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1493 			ret = -EACCES;
1494 		} else {
1495 			/* We need to unlock here because mrtsock_destruct takes
1496 			 * care of rtnl itself and we can't change that due to
1497 			 * the IP_ROUTER_ALERT setsockopt which runs without it.
1498 			 */
1499 			rtnl_unlock();
1500 			ret = ip_ra_control(sk, 0, NULL);
1501 			goto out;
1502 		}
1503 		break;
1504 	case MRT_ADD_VIF:
1505 	case MRT_DEL_VIF:
1506 		if (optlen != sizeof(vif)) {
1507 			ret = -EINVAL;
1508 			break;
1509 		}
1510 		if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1511 			ret = -EFAULT;
1512 			break;
1513 		}
1514 		if (vif.vifc_vifi >= MAXVIFS) {
1515 			ret = -ENFILE;
1516 			break;
1517 		}
1518 		if (optname == MRT_ADD_VIF) {
1519 			ret = vif_add(net, mrt, &vif,
1520 				      sk == rtnl_dereference(mrt->mroute_sk));
1521 		} else {
1522 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1523 		}
1524 		break;
1525 	/* Manipulate the forwarding caches. These live
1526 	 * in a sort of kernel/user symbiosis.
1527 	 */
1528 	case MRT_ADD_MFC:
1529 	case MRT_DEL_MFC:
1530 		parent = -1;
1531 		fallthrough;
1532 	case MRT_ADD_MFC_PROXY:
1533 	case MRT_DEL_MFC_PROXY:
1534 		if (optlen != sizeof(mfc)) {
1535 			ret = -EINVAL;
1536 			break;
1537 		}
1538 		if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1539 			ret = -EFAULT;
1540 			break;
1541 		}
1542 		if (parent == 0)
1543 			parent = mfc.mfcc_parent;
1544 
1545 		mutex_lock(&net->ipv4.mfc_mutex);
1546 
1547 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1548 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1549 		else
1550 			ret = ipmr_mfc_add(net, mrt, &mfc,
1551 					   sk == rtnl_dereference(mrt->mroute_sk),
1552 					   parent);
1553 
1554 		mutex_unlock(&net->ipv4.mfc_mutex);
1555 		break;
1556 	case MRT_FLUSH: {
1557 		LIST_HEAD(dev_kill_list);
1558 
1559 		if (optlen != sizeof(val)) {
1560 			ret = -EINVAL;
1561 			break;
1562 		}
1563 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1564 			ret = -EFAULT;
1565 			break;
1566 		}
1567 
1568 		mroute_clean_tables(mrt, val, &dev_kill_list);
1569 		unregister_netdevice_many(&dev_kill_list);
1570 		break;
1571 	}
1572 	/* Control PIM assert. */
1573 	case MRT_ASSERT:
1574 		if (optlen != sizeof(val)) {
1575 			ret = -EINVAL;
1576 			break;
1577 		}
1578 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1579 			ret = -EFAULT;
1580 			break;
1581 		}
1582 		WRITE_ONCE(mrt->mroute_do_assert, val);
1583 		break;
1584 	case MRT_PIM:
1585 		if (!ipmr_pimsm_enabled()) {
1586 			ret = -ENOPROTOOPT;
1587 			break;
1588 		}
1589 		if (optlen != sizeof(val)) {
1590 			ret = -EINVAL;
1591 			break;
1592 		}
1593 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1594 			ret = -EFAULT;
1595 			break;
1596 		}
1597 
1598 		do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1599 		val = !!val;
1600 		if (val != mrt->mroute_do_pim) {
1601 			WRITE_ONCE(mrt->mroute_do_pim, val);
1602 			WRITE_ONCE(mrt->mroute_do_assert, val);
1603 			WRITE_ONCE(mrt->mroute_do_wrvifwhole, do_wrvifwhole);
1604 		}
1605 		break;
1606 	case MRT_TABLE:
1607 		if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1608 			ret = -ENOPROTOOPT;
1609 			break;
1610 		}
1611 		if (optlen != sizeof(uval)) {
1612 			ret = -EINVAL;
1613 			break;
1614 		}
1615 		if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1616 			ret = -EFAULT;
1617 			break;
1618 		}
1619 
1620 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1621 			ret = -EBUSY;
1622 		} else {
1623 			mrt = ipmr_new_table(net, uval);
1624 			if (IS_ERR(mrt))
1625 				ret = PTR_ERR(mrt);
1626 			else
1627 				raw_sk(sk)->ipmr_table = uval;
1628 		}
1629 		break;
1630 	/* Spurious command, or MRT_VERSION which you cannot set. */
1631 	default:
1632 		ret = -ENOPROTOOPT;
1633 	}
1634 out_unlock:
1635 	rtnl_unlock();
1636 out:
1637 	return ret;
1638 }
1639 
1640 /* Execute if this ioctl is a special mroute ioctl */
1641 int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1642 {
1643 	switch (cmd) {
1644 	/* These userspace buffers will be consumed by ipmr_ioctl() */
1645 	case SIOCGETVIFCNT: {
1646 		struct sioc_vif_req buffer;
1647 
1648 		return sock_ioctl_inout(sk, cmd, arg, &buffer,
1649 				      sizeof(buffer));
1650 		}
1651 	case SIOCGETSGCNT: {
1652 		struct sioc_sg_req buffer;
1653 
1654 		return sock_ioctl_inout(sk, cmd, arg, &buffer,
1655 				      sizeof(buffer));
1656 		}
1657 	}
1658 	/* return code > 0 means that the ioctl was not executed */
1659 	return 1;
1660 }
1661 
1662 /* Getsock opt support for the multicast routing system. */
1663 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1664 			 sockptr_t optlen)
1665 {
1666 	int olr;
1667 	int val;
1668 	struct net *net = sock_net(sk);
1669 	struct mr_table *mrt;
1670 
1671 	if (sk->sk_type != SOCK_RAW ||
1672 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1673 		return -EOPNOTSUPP;
1674 
1675 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1676 	if (!mrt)
1677 		return -ENOENT;
1678 
1679 	switch (optname) {
1680 	case MRT_VERSION:
1681 		val = 0x0305;
1682 		break;
1683 	case MRT_PIM:
1684 		if (!ipmr_pimsm_enabled())
1685 			return -ENOPROTOOPT;
1686 		val = READ_ONCE(mrt->mroute_do_pim);
1687 		break;
1688 	case MRT_ASSERT:
1689 		val = READ_ONCE(mrt->mroute_do_assert);
1690 		break;
1691 	default:
1692 		return -ENOPROTOOPT;
1693 	}
1694 
1695 	if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1696 		return -EFAULT;
1697 	if (olr < 0)
1698 		return -EINVAL;
1699 
1700 	olr = min_t(unsigned int, olr, sizeof(int));
1701 
1702 	if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1703 		return -EFAULT;
1704 	if (copy_to_sockptr(optval, &val, olr))
1705 		return -EFAULT;
1706 	return 0;
1707 }
1708 
1709 /* The IP multicast ioctl support routines. */
1710 int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1711 {
1712 	struct vif_device *vif;
1713 	struct mfc_cache *c;
1714 	struct net *net = sock_net(sk);
1715 	struct sioc_vif_req *vr;
1716 	struct sioc_sg_req *sr;
1717 	struct mr_table *mrt;
1718 
1719 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1720 	if (!mrt)
1721 		return -ENOENT;
1722 
1723 	switch (cmd) {
1724 	case SIOCGETVIFCNT:
1725 		vr = (struct sioc_vif_req *)arg;
1726 		if (vr->vifi >= mrt->maxvif)
1727 			return -EINVAL;
1728 		vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1729 		rcu_read_lock();
1730 		vif = &mrt->vif_table[vr->vifi];
1731 		if (VIF_EXISTS(mrt, vr->vifi)) {
1732 			vr->icount = READ_ONCE(vif->pkt_in);
1733 			vr->ocount = READ_ONCE(vif->pkt_out);
1734 			vr->ibytes = READ_ONCE(vif->bytes_in);
1735 			vr->obytes = READ_ONCE(vif->bytes_out);
1736 			rcu_read_unlock();
1737 
1738 			return 0;
1739 		}
1740 		rcu_read_unlock();
1741 		return -EADDRNOTAVAIL;
1742 	case SIOCGETSGCNT:
1743 		sr = (struct sioc_sg_req *)arg;
1744 
1745 		rcu_read_lock();
1746 		c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1747 		if (c) {
1748 			sr->pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1749 			sr->bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1750 			sr->wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1751 			rcu_read_unlock();
1752 			return 0;
1753 		}
1754 		rcu_read_unlock();
1755 		return -EADDRNOTAVAIL;
1756 	default:
1757 		return -ENOIOCTLCMD;
1758 	}
1759 }
1760 
1761 #ifdef CONFIG_COMPAT
1762 struct compat_sioc_sg_req {
1763 	struct in_addr src;
1764 	struct in_addr grp;
1765 	compat_ulong_t pktcnt;
1766 	compat_ulong_t bytecnt;
1767 	compat_ulong_t wrong_if;
1768 };
1769 
1770 struct compat_sioc_vif_req {
1771 	vifi_t	vifi;		/* Which iface */
1772 	compat_ulong_t icount;
1773 	compat_ulong_t ocount;
1774 	compat_ulong_t ibytes;
1775 	compat_ulong_t obytes;
1776 };
1777 
1778 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1779 {
1780 	struct compat_sioc_sg_req sr;
1781 	struct compat_sioc_vif_req vr;
1782 	struct vif_device *vif;
1783 	struct mfc_cache *c;
1784 	struct net *net = sock_net(sk);
1785 	struct mr_table *mrt;
1786 
1787 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1788 	if (!mrt)
1789 		return -ENOENT;
1790 
1791 	switch (cmd) {
1792 	case SIOCGETVIFCNT:
1793 		if (copy_from_user(&vr, arg, sizeof(vr)))
1794 			return -EFAULT;
1795 		if (vr.vifi >= mrt->maxvif)
1796 			return -EINVAL;
1797 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1798 		rcu_read_lock();
1799 		vif = &mrt->vif_table[vr.vifi];
1800 		if (VIF_EXISTS(mrt, vr.vifi)) {
1801 			vr.icount = READ_ONCE(vif->pkt_in);
1802 			vr.ocount = READ_ONCE(vif->pkt_out);
1803 			vr.ibytes = READ_ONCE(vif->bytes_in);
1804 			vr.obytes = READ_ONCE(vif->bytes_out);
1805 			rcu_read_unlock();
1806 
1807 			if (copy_to_user(arg, &vr, sizeof(vr)))
1808 				return -EFAULT;
1809 			return 0;
1810 		}
1811 		rcu_read_unlock();
1812 		return -EADDRNOTAVAIL;
1813 	case SIOCGETSGCNT:
1814 		if (copy_from_user(&sr, arg, sizeof(sr)))
1815 			return -EFAULT;
1816 
1817 		rcu_read_lock();
1818 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1819 		if (c) {
1820 			sr.pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1821 			sr.bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1822 			sr.wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1823 			rcu_read_unlock();
1824 
1825 			if (copy_to_user(arg, &sr, sizeof(sr)))
1826 				return -EFAULT;
1827 			return 0;
1828 		}
1829 		rcu_read_unlock();
1830 		return -EADDRNOTAVAIL;
1831 	default:
1832 		return -ENOIOCTLCMD;
1833 	}
1834 }
1835 #endif
1836 
1837 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1838 {
1839 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1840 	struct net *net = dev_net(dev);
1841 	struct mr_table *mrt;
1842 	struct vif_device *v;
1843 	int ct;
1844 
1845 	if (event != NETDEV_UNREGISTER)
1846 		return NOTIFY_DONE;
1847 
1848 	ipmr_for_each_table(mrt, net) {
1849 		v = &mrt->vif_table[0];
1850 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1851 			if (rcu_access_pointer(v->dev) == dev)
1852 				vif_delete(mrt, ct, 1, NULL);
1853 		}
1854 	}
1855 	return NOTIFY_DONE;
1856 }
1857 
1858 static struct notifier_block ip_mr_notifier = {
1859 	.notifier_call = ipmr_device_event,
1860 };
1861 
1862 /* Encapsulate a packet by attaching a valid IPIP header to it.
1863  * This avoids tunnel drivers and other mess and gives us the speed so
1864  * important for multicast video.
1865  */
1866 static void ip_encap(struct net *net, struct sk_buff *skb,
1867 		     __be32 saddr, __be32 daddr)
1868 {
1869 	struct iphdr *iph;
1870 	const struct iphdr *old_iph = ip_hdr(skb);
1871 
1872 	skb_push(skb, sizeof(struct iphdr));
1873 	skb->transport_header = skb->network_header;
1874 	skb_reset_network_header(skb);
1875 	iph = ip_hdr(skb);
1876 
1877 	iph->version	=	4;
1878 	iph->tos	=	old_iph->tos;
1879 	iph->ttl	=	old_iph->ttl;
1880 	iph->frag_off	=	0;
1881 	iph->daddr	=	daddr;
1882 	iph->saddr	=	saddr;
1883 	iph->protocol	=	IPPROTO_IPIP;
1884 	iph->ihl	=	5;
1885 	iph->tot_len	=	htons(skb->len);
1886 	ip_select_ident(net, skb, NULL);
1887 	ip_send_check(iph);
1888 
1889 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1890 	nf_reset_ct(skb);
1891 }
1892 
1893 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1894 				      struct sk_buff *skb)
1895 {
1896 	struct ip_options *opt = &(IPCB(skb)->opt);
1897 
1898 	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1899 
1900 	if (unlikely(opt->optlen))
1901 		ip_forward_options(skb);
1902 
1903 	return dst_output(net, sk, skb);
1904 }
1905 
1906 #ifdef CONFIG_NET_SWITCHDEV
1907 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1908 				   int in_vifi, int out_vifi)
1909 {
1910 	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1911 	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1912 
1913 	if (!skb->offload_l3_fwd_mark)
1914 		return false;
1915 	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1916 		return false;
1917 	return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1918 					&in_vif->dev_parent_id);
1919 }
1920 #else
1921 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1922 				   int in_vifi, int out_vifi)
1923 {
1924 	return false;
1925 }
1926 #endif
1927 
1928 /* Processing handlers for ipmr_forward, under rcu_read_lock() */
1929 
1930 static int ipmr_prepare_xmit(struct net *net, struct mr_table *mrt,
1931 			     struct sk_buff *skb, int vifi)
1932 {
1933 	const struct iphdr *iph = ip_hdr(skb);
1934 	struct vif_device *vif = &mrt->vif_table[vifi];
1935 	struct net_device *vif_dev;
1936 	struct rtable *rt;
1937 	struct flowi4 fl4;
1938 	int    encap = 0;
1939 
1940 	vif_dev = vif_dev_read(vif);
1941 	if (!vif_dev)
1942 		return -1;
1943 
1944 	if (vif->flags & VIFF_REGISTER) {
1945 		WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1946 		WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1947 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1948 		DEV_STATS_INC(vif_dev, tx_packets);
1949 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT, false);
1950 		return -1;
1951 	}
1952 
1953 	if (vif->flags & VIFF_TUNNEL) {
1954 		rt = ip_route_output_ports(net, &fl4, NULL,
1955 					   vif->remote, vif->local,
1956 					   0, 0,
1957 					   IPPROTO_IPIP,
1958 					   iph->tos & INET_DSCP_MASK, vif->link);
1959 		if (IS_ERR(rt))
1960 			return -1;
1961 		encap = sizeof(struct iphdr);
1962 	} else {
1963 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1964 					   0, 0,
1965 					   IPPROTO_IPIP,
1966 					   iph->tos & INET_DSCP_MASK, vif->link);
1967 		if (IS_ERR(rt))
1968 			return -1;
1969 	}
1970 
1971 	if (skb->len+encap > dst4_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1972 		/* Do not fragment multicasts. Alas, IPv4 does not
1973 		 * allow to send ICMP, so that packets will disappear
1974 		 * to blackhole.
1975 		 */
1976 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1977 		ip_rt_put(rt);
1978 		return -1;
1979 	}
1980 
1981 	encap += LL_RESERVED_SPACE(dst_dev_rcu(&rt->dst)) + rt->dst.header_len;
1982 
1983 	if (skb_cow(skb, encap)) {
1984 		ip_rt_put(rt);
1985 		return -1;
1986 	}
1987 
1988 	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1989 	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1990 
1991 	skb_dst_drop(skb);
1992 	skb_dst_set(skb, &rt->dst);
1993 	ip_decrease_ttl(ip_hdr(skb));
1994 
1995 	/* FIXME: forward and output firewalls used to be called here.
1996 	 * What do we do with netfilter? -- RR
1997 	 */
1998 	if (vif->flags & VIFF_TUNNEL) {
1999 		ip_encap(net, skb, vif->local, vif->remote);
2000 		/* FIXME: extra output firewall step used to be here. --RR */
2001 		DEV_STATS_INC(vif_dev, tx_packets);
2002 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
2003 	}
2004 
2005 	return 0;
2006 }
2007 
2008 static void ipmr_queue_fwd_xmit(struct net *net, struct mr_table *mrt,
2009 				int in_vifi, struct sk_buff *skb, int vifi)
2010 {
2011 	struct rtable *rt;
2012 
2013 	if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
2014 		goto out_free;
2015 
2016 	if (ipmr_prepare_xmit(net, mrt, skb, vifi))
2017 		goto out_free;
2018 
2019 	rt = skb_rtable(skb);
2020 
2021 	IPCB(skb)->flags |= IPSKB_FORWARDED;
2022 
2023 	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2024 	 * not only before forwarding, but after forwarding on all output
2025 	 * interfaces. It is clear, if mrouter runs a multicasting
2026 	 * program, it should receive packets not depending to what interface
2027 	 * program is joined.
2028 	 * If we will not make it, the program will have to join on all
2029 	 * interfaces. On the other hand, multihoming host (or router, but
2030 	 * not mrouter) cannot join to more than one interface - it will
2031 	 * result in receiving multiple packets.
2032 	 */
2033 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
2034 		net, NULL, skb, skb->dev, dst_dev_rcu(&rt->dst),
2035 		ipmr_forward_finish);
2036 	return;
2037 
2038 out_free:
2039 	kfree_skb(skb);
2040 }
2041 
2042 static void ipmr_queue_output_xmit(struct net *net, struct mr_table *mrt,
2043 				   struct sk_buff *skb, int vifi)
2044 {
2045 	if (ipmr_prepare_xmit(net, mrt, skb, vifi))
2046 		goto out_free;
2047 
2048 	ip_mc_output(net, NULL, skb);
2049 	return;
2050 
2051 out_free:
2052 	kfree_skb(skb);
2053 }
2054 
2055 /* Called with mrt_lock or rcu_read_lock() */
2056 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
2057 {
2058 	int ct;
2059 	/* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
2060 	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
2061 		if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
2062 			break;
2063 	}
2064 	return ct;
2065 }
2066 
2067 /* "local" means that we should preserve one skb (for local delivery) */
2068 /* Called uner rcu_read_lock() */
2069 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
2070 			  struct net_device *dev, struct sk_buff *skb,
2071 			  struct mfc_cache *c, int local)
2072 {
2073 	int true_vifi = ipmr_find_vif(mrt, dev);
2074 	int psend = -1;
2075 	int vif, ct;
2076 
2077 	vif = c->_c.mfc_parent;
2078 	atomic_long_inc(&c->_c.mfc_un.res.pkt);
2079 	atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes);
2080 	WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies);
2081 
2082 	if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
2083 		struct mfc_cache *cache_proxy;
2084 
2085 		/* For an (*,G) entry, we only check that the incoming
2086 		 * interface is part of the static tree.
2087 		 */
2088 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2089 		if (cache_proxy &&
2090 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
2091 			goto forward;
2092 	}
2093 
2094 	/* Wrong interface: drop packet and (maybe) send PIM assert. */
2095 	if (true_vifi < 0 ||
2096 	    rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
2097 		if (rt_is_output_route(skb_rtable(skb))) {
2098 			/* It is our own packet, looped back.
2099 			 * Very complicated situation...
2100 			 *
2101 			 * The best workaround until routing daemons will be
2102 			 * fixed is not to redistribute packet, if it was
2103 			 * send through wrong interface. It means, that
2104 			 * multicast applications WILL NOT work for
2105 			 * (S,G), which have default multicast route pointing
2106 			 * to wrong oif. In any case, it is not a good
2107 			 * idea to use multicasting applications on router.
2108 			 */
2109 			goto dont_forward;
2110 		}
2111 
2112 		atomic_long_inc(&c->_c.mfc_un.res.wrong_if);
2113 
2114 		if (true_vifi >= 0 && READ_ONCE(mrt->mroute_do_assert) &&
2115 		    /* pimsm uses asserts, when switching from RPT to SPT,
2116 		     * so that we cannot check that packet arrived on an oif.
2117 		     * It is bad, but otherwise we would need to move pretty
2118 		     * large chunk of pimd to kernel. Ough... --ANK
2119 		     */
2120 		    (READ_ONCE(mrt->mroute_do_pim) ||
2121 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2122 		    time_after(jiffies,
2123 			       c->_c.mfc_un.res.last_assert +
2124 			       MFC_ASSERT_THRESH)) {
2125 			c->_c.mfc_un.res.last_assert = jiffies;
2126 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF,
2127 					  false);
2128 			if (READ_ONCE(mrt->mroute_do_wrvifwhole))
2129 				ipmr_cache_report(mrt, skb, true_vifi,
2130 						  IGMPMSG_WRVIFWHOLE, false);
2131 		}
2132 		goto dont_forward;
2133 	}
2134 
2135 forward:
2136 	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2137 		   mrt->vif_table[vif].pkt_in + 1);
2138 	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2139 		   mrt->vif_table[vif].bytes_in + skb->len);
2140 
2141 	/* Forward the frame */
2142 	if (c->mfc_origin == htonl(INADDR_ANY) &&
2143 	    c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2144 		if (true_vifi >= 0 &&
2145 		    true_vifi != c->_c.mfc_parent &&
2146 		    ip_hdr(skb)->ttl >
2147 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2148 			/* It's an (*,*) entry and the packet is not coming from
2149 			 * the upstream: forward the packet to the upstream
2150 			 * only.
2151 			 */
2152 			psend = c->_c.mfc_parent;
2153 			goto last_forward;
2154 		}
2155 		goto dont_forward;
2156 	}
2157 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2158 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2159 		/* For (*,G) entry, don't forward to the incoming interface */
2160 		if ((c->mfc_origin != htonl(INADDR_ANY) ||
2161 		     ct != true_vifi) &&
2162 		    ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2163 			if (psend != -1) {
2164 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2165 
2166 				if (skb2)
2167 					ipmr_queue_fwd_xmit(net, mrt, true_vifi,
2168 							    skb2, psend);
2169 			}
2170 			psend = ct;
2171 		}
2172 	}
2173 last_forward:
2174 	if (psend != -1) {
2175 		if (local) {
2176 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2177 
2178 			if (skb2)
2179 				ipmr_queue_fwd_xmit(net, mrt, true_vifi, skb2,
2180 						    psend);
2181 		} else {
2182 			ipmr_queue_fwd_xmit(net, mrt, true_vifi, skb, psend);
2183 			return;
2184 		}
2185 	}
2186 
2187 dont_forward:
2188 	if (!local)
2189 		kfree_skb(skb);
2190 }
2191 
2192 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2193 {
2194 	struct rtable *rt = skb_rtable(skb);
2195 	struct iphdr *iph = ip_hdr(skb);
2196 	struct flowi4 fl4 = {
2197 		.daddr = iph->daddr,
2198 		.saddr = iph->saddr,
2199 		.flowi4_dscp = ip4h_dscp(iph),
2200 		.flowi4_oif = (rt_is_output_route(rt) ?
2201 			       skb->dev->ifindex : 0),
2202 		.flowi4_iif = (rt_is_output_route(rt) ?
2203 			       LOOPBACK_IFINDEX :
2204 			       skb->dev->ifindex),
2205 		.flowi4_mark = skb->mark,
2206 	};
2207 	struct mr_table *mrt;
2208 	int err;
2209 
2210 	err = ipmr_fib_lookup(net, &fl4, &mrt);
2211 	if (err)
2212 		return ERR_PTR(err);
2213 	return mrt;
2214 }
2215 
2216 /* Multicast packets for forwarding arrive here
2217  * Called with rcu_read_lock();
2218  */
2219 int ip_mr_input(struct sk_buff *skb)
2220 {
2221 	struct mfc_cache *cache;
2222 	struct net *net = dev_net(skb->dev);
2223 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2224 	struct mr_table *mrt;
2225 	struct net_device *dev;
2226 
2227 	/* skb->dev passed in is the loX master dev for vrfs.
2228 	 * As there are no vifs associated with loopback devices,
2229 	 * get the proper interface that does have a vif associated with it.
2230 	 */
2231 	dev = skb->dev;
2232 	if (netif_is_l3_master(skb->dev)) {
2233 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2234 		if (!dev) {
2235 			kfree_skb(skb);
2236 			return -ENODEV;
2237 		}
2238 	}
2239 
2240 	/* Packet is looped back after forward, it should not be
2241 	 * forwarded second time, but still can be delivered locally.
2242 	 */
2243 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2244 		goto dont_forward;
2245 
2246 	if (!local)
2247 		skb_orphan(skb);
2248 
2249 	mrt = ipmr_rt_fib_lookup(net, skb);
2250 	if (IS_ERR(mrt)) {
2251 		kfree_skb(skb);
2252 		return PTR_ERR(mrt);
2253 	}
2254 	if (!local) {
2255 		if (IPCB(skb)->opt.router_alert) {
2256 			if (ip_call_ra_chain(skb))
2257 				return 0;
2258 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2259 			/* IGMPv1 (and broken IGMPv2 implementations sort of
2260 			 * Cisco IOS <= 11.2(8)) do not put router alert
2261 			 * option to IGMP packets destined to routable
2262 			 * groups. It is very bad, because it means
2263 			 * that we can forward NO IGMP messages.
2264 			 */
2265 			struct sock *mroute_sk;
2266 
2267 			mroute_sk = rcu_dereference(mrt->mroute_sk);
2268 			if (mroute_sk) {
2269 				nf_reset_ct(skb);
2270 				raw_rcv(mroute_sk, skb);
2271 				return 0;
2272 			}
2273 		}
2274 	}
2275 
2276 	/* already under rcu_read_lock() */
2277 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2278 	if (!cache) {
2279 		int vif = ipmr_find_vif(mrt, dev);
2280 
2281 		if (vif >= 0)
2282 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2283 						    vif);
2284 	}
2285 
2286 	/* No usable cache entry */
2287 	if (!cache) {
2288 		int vif;
2289 
2290 		if (local) {
2291 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2292 			ip_local_deliver(skb);
2293 			if (!skb2)
2294 				return -ENOBUFS;
2295 			skb = skb2;
2296 		}
2297 
2298 		vif = ipmr_find_vif(mrt, dev);
2299 		if (vif >= 0)
2300 			return ipmr_cache_unresolved(mrt, vif, skb, dev, false);
2301 		kfree_skb(skb);
2302 		return -ENODEV;
2303 	}
2304 
2305 	ip_mr_forward(net, mrt, dev, skb, cache, local);
2306 
2307 	if (local)
2308 		return ip_local_deliver(skb);
2309 
2310 	return 0;
2311 
2312 dont_forward:
2313 	if (local)
2314 		return ip_local_deliver(skb);
2315 	kfree_skb(skb);
2316 	return 0;
2317 }
2318 
2319 static void ip_mr_output_finish(struct net *net, struct mr_table *mrt,
2320 				struct net_device *dev, struct sk_buff *skb,
2321 				struct mfc_cache *c)
2322 {
2323 	int psend = -1;
2324 	int ct;
2325 
2326 	atomic_long_inc(&c->_c.mfc_un.res.pkt);
2327 	atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes);
2328 	WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies);
2329 
2330 	/* Forward the frame */
2331 	if (c->mfc_origin == htonl(INADDR_ANY) &&
2332 	    c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2333 		if (ip_hdr(skb)->ttl >
2334 		    c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2335 			/* It's an (*,*) entry and the packet is not coming from
2336 			 * the upstream: forward the packet to the upstream
2337 			 * only.
2338 			 */
2339 			psend = c->_c.mfc_parent;
2340 			goto last_xmit;
2341 		}
2342 		goto dont_xmit;
2343 	}
2344 
2345 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2346 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2347 		if (ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2348 			if (psend != -1) {
2349 				struct sk_buff *skb2;
2350 
2351 				skb2 = skb_clone(skb, GFP_ATOMIC);
2352 				if (skb2)
2353 					ipmr_queue_output_xmit(net, mrt,
2354 							       skb2, psend);
2355 			}
2356 			psend = ct;
2357 		}
2358 	}
2359 
2360 last_xmit:
2361 	if (psend != -1) {
2362 		ipmr_queue_output_xmit(net, mrt, skb, psend);
2363 		return;
2364 	}
2365 
2366 dont_xmit:
2367 	kfree_skb(skb);
2368 }
2369 
2370 /* Multicast packets for forwarding arrive here
2371  * Called with rcu_read_lock();
2372  */
2373 int ip_mr_output(struct net *net, struct sock *sk, struct sk_buff *skb)
2374 {
2375 	struct rtable *rt = skb_rtable(skb);
2376 	struct mfc_cache *cache;
2377 	struct net_device *dev;
2378 	struct mr_table *mrt;
2379 	int vif;
2380 
2381 	guard(rcu)();
2382 
2383 	dev = dst_dev_rcu(&rt->dst);
2384 
2385 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2386 		goto mc_output;
2387 	if (!(IPCB(skb)->flags & IPSKB_MCROUTE))
2388 		goto mc_output;
2389 
2390 	skb->dev = dev;
2391 
2392 	mrt = ipmr_rt_fib_lookup(net, skb);
2393 	if (IS_ERR(mrt))
2394 		goto mc_output;
2395 
2396 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2397 	if (!cache) {
2398 		vif = ipmr_find_vif(mrt, dev);
2399 		if (vif >= 0)
2400 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2401 						    vif);
2402 	}
2403 
2404 	/* No usable cache entry */
2405 	if (!cache) {
2406 		vif = ipmr_find_vif(mrt, dev);
2407 		if (vif >= 0)
2408 			return ipmr_cache_unresolved(mrt, vif, skb, dev, false);
2409 		goto mc_output;
2410 	}
2411 
2412 	vif = cache->_c.mfc_parent;
2413 	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev)
2414 		goto mc_output;
2415 
2416 	ip_mr_output_finish(net, mrt, dev, skb, cache);
2417 	return 0;
2418 
2419 mc_output:
2420 	return ip_mc_output(net, sk, skb);
2421 }
2422 
2423 #ifdef CONFIG_IP_PIMSM_V1
2424 /* Handle IGMP messages of PIMv1 */
2425 int pim_rcv_v1(struct sk_buff *skb)
2426 {
2427 	struct igmphdr *pim;
2428 	struct net *net = dev_net(skb->dev);
2429 	struct mr_table *mrt;
2430 
2431 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2432 		goto drop;
2433 
2434 	pim = igmp_hdr(skb);
2435 
2436 	mrt = ipmr_rt_fib_lookup(net, skb);
2437 	if (IS_ERR(mrt))
2438 		goto drop;
2439 	if (!READ_ONCE(mrt->mroute_do_pim) ||
2440 	    pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2441 		goto drop;
2442 
2443 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2444 drop:
2445 		kfree_skb(skb);
2446 	}
2447 	return 0;
2448 }
2449 #endif
2450 
2451 #ifdef CONFIG_IP_PIMSM_V2
2452 static int pim_rcv(struct sk_buff *skb)
2453 {
2454 	struct pimreghdr *pim;
2455 	struct net *net = dev_net(skb->dev);
2456 	struct mr_table *mrt;
2457 
2458 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2459 		goto drop;
2460 
2461 	pim = (struct pimreghdr *)skb_transport_header(skb);
2462 	if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2463 	    (pim->flags & PIM_NULL_REGISTER) ||
2464 	    (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2465 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2466 		goto drop;
2467 
2468 	mrt = ipmr_rt_fib_lookup(net, skb);
2469 	if (IS_ERR(mrt))
2470 		goto drop;
2471 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2472 drop:
2473 		kfree_skb(skb);
2474 	}
2475 	return 0;
2476 }
2477 #endif
2478 
2479 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2480 		   __be32 saddr, __be32 daddr,
2481 		   struct rtmsg *rtm, u32 portid)
2482 {
2483 	struct mfc_cache *cache;
2484 	struct mr_table *mrt;
2485 	int err;
2486 
2487 	rcu_read_lock();
2488 	mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
2489 	if (!mrt) {
2490 		rcu_read_unlock();
2491 		return -ENOENT;
2492 	}
2493 
2494 	cache = ipmr_cache_find(mrt, saddr, daddr);
2495 	if (!cache && skb->dev) {
2496 		int vif = ipmr_find_vif(mrt, skb->dev);
2497 
2498 		if (vif >= 0)
2499 			cache = ipmr_cache_find_any(mrt, daddr, vif);
2500 	}
2501 	if (!cache) {
2502 		struct sk_buff *skb2;
2503 		struct iphdr *iph;
2504 		struct net_device *dev;
2505 		int vif = -1;
2506 
2507 		dev = skb->dev;
2508 		if (dev)
2509 			vif = ipmr_find_vif(mrt, dev);
2510 		if (vif < 0) {
2511 			rcu_read_unlock();
2512 			return -ENODEV;
2513 		}
2514 
2515 		skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2516 		if (!skb2) {
2517 			rcu_read_unlock();
2518 			return -ENOMEM;
2519 		}
2520 
2521 		NETLINK_CB(skb2).portid = portid;
2522 		skb_push(skb2, sizeof(struct iphdr));
2523 		skb_reset_network_header(skb2);
2524 		iph = ip_hdr(skb2);
2525 		memset(iph, 0, sizeof(*iph));
2526 		iph->ihl = sizeof(struct iphdr) >> 2;
2527 		iph->saddr = saddr;
2528 		iph->daddr = daddr;
2529 		iph->version = 0;
2530 		err = ipmr_cache_unresolved(mrt, vif, skb2, dev, true);
2531 		rcu_read_unlock();
2532 		return err;
2533 	}
2534 
2535 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2536 	rcu_read_unlock();
2537 	return err;
2538 }
2539 
2540 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2541 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2542 			    int flags)
2543 {
2544 	struct nlmsghdr *nlh;
2545 	struct rtmsg *rtm;
2546 	int err;
2547 
2548 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2549 	if (!nlh)
2550 		return -EMSGSIZE;
2551 
2552 	rtm = nlmsg_data(nlh);
2553 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2554 	rtm->rtm_dst_len  = 32;
2555 	rtm->rtm_src_len  = 32;
2556 	rtm->rtm_tos      = 0;
2557 	rtm->rtm_table    = mrt->id;
2558 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2559 		goto nla_put_failure;
2560 	rtm->rtm_type     = RTN_MULTICAST;
2561 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2562 	if (c->_c.mfc_flags & MFC_STATIC)
2563 		rtm->rtm_protocol = RTPROT_STATIC;
2564 	else
2565 		rtm->rtm_protocol = RTPROT_MROUTED;
2566 	rtm->rtm_flags    = 0;
2567 
2568 	if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2569 	    nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2570 		goto nla_put_failure;
2571 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2572 	/* do not break the dump if cache is unresolved */
2573 	if (err < 0 && err != -ENOENT)
2574 		goto nla_put_failure;
2575 
2576 	nlmsg_end(skb, nlh);
2577 	return 0;
2578 
2579 nla_put_failure:
2580 	nlmsg_cancel(skb, nlh);
2581 	return -EMSGSIZE;
2582 }
2583 
2584 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2585 			     u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2586 			     int flags)
2587 {
2588 	return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2589 				cmd, flags);
2590 }
2591 
2592 static size_t mroute_msgsize(bool unresolved)
2593 {
2594 	size_t len =
2595 		NLMSG_ALIGN(sizeof(struct rtmsg))
2596 		+ nla_total_size(4)	/* RTA_TABLE */
2597 		+ nla_total_size(4)	/* RTA_SRC */
2598 		+ nla_total_size(4)	/* RTA_DST */
2599 		;
2600 
2601 	if (!unresolved)
2602 		len = len
2603 		      + nla_total_size(4)	/* RTA_IIF */
2604 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2605 		      + MAXVIFS * NLA_ALIGN(sizeof(struct rtnexthop))
2606 						/* RTA_MFC_STATS */
2607 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2608 		;
2609 
2610 	return len;
2611 }
2612 
2613 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2614 				 int cmd)
2615 {
2616 	struct net *net = read_pnet(&mrt->net);
2617 	struct sk_buff *skb;
2618 	int err = -ENOBUFS;
2619 
2620 	skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS),
2621 			GFP_ATOMIC);
2622 	if (!skb)
2623 		goto errout;
2624 
2625 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2626 	if (err < 0)
2627 		goto errout;
2628 
2629 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2630 	return;
2631 
2632 errout:
2633 	kfree_skb(skb);
2634 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2635 }
2636 
2637 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2638 {
2639 	size_t len =
2640 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2641 		+ nla_total_size(1)	/* IPMRA_CREPORT_MSGTYPE */
2642 		+ nla_total_size(4)	/* IPMRA_CREPORT_VIF_ID */
2643 		+ nla_total_size(4)	/* IPMRA_CREPORT_SRC_ADDR */
2644 		+ nla_total_size(4)	/* IPMRA_CREPORT_DST_ADDR */
2645 		+ nla_total_size(4)	/* IPMRA_CREPORT_TABLE */
2646 					/* IPMRA_CREPORT_PKT */
2647 		+ nla_total_size(payloadlen)
2648 		;
2649 
2650 	return len;
2651 }
2652 
2653 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2654 {
2655 	struct net *net = read_pnet(&mrt->net);
2656 	struct nlmsghdr *nlh;
2657 	struct rtgenmsg *rtgenm;
2658 	struct igmpmsg *msg;
2659 	struct sk_buff *skb;
2660 	struct nlattr *nla;
2661 	int payloadlen;
2662 
2663 	payloadlen = pkt->len - sizeof(struct igmpmsg);
2664 	msg = (struct igmpmsg *)skb_network_header(pkt);
2665 
2666 	skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2667 	if (!skb)
2668 		goto errout;
2669 
2670 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2671 			sizeof(struct rtgenmsg), 0);
2672 	if (!nlh)
2673 		goto errout;
2674 	rtgenm = nlmsg_data(nlh);
2675 	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2676 	if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2677 	    nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2678 	    nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2679 			    msg->im_src.s_addr) ||
2680 	    nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2681 			    msg->im_dst.s_addr) ||
2682 	    nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2683 		goto nla_put_failure;
2684 
2685 	nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2686 	if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2687 				  nla_data(nla), payloadlen))
2688 		goto nla_put_failure;
2689 
2690 	nlmsg_end(skb, nlh);
2691 
2692 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2693 	return;
2694 
2695 nla_put_failure:
2696 	nlmsg_cancel(skb, nlh);
2697 errout:
2698 	kfree_skb(skb);
2699 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2700 }
2701 
2702 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2703 				       const struct nlmsghdr *nlh,
2704 				       struct nlattr **tb,
2705 				       struct netlink_ext_ack *extack)
2706 {
2707 	struct rtmsg *rtm;
2708 	int i, err;
2709 
2710 	rtm = nlmsg_payload(nlh, sizeof(*rtm));
2711 	if (!rtm) {
2712 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2713 		return -EINVAL;
2714 	}
2715 
2716 	if (!netlink_strict_get_check(skb))
2717 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2718 					      rtm_ipv4_policy, extack);
2719 
2720 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2721 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2722 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2723 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2724 		NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2725 		return -EINVAL;
2726 	}
2727 
2728 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2729 					    rtm_ipv4_policy, extack);
2730 	if (err)
2731 		return err;
2732 
2733 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2734 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2735 		NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2736 		return -EINVAL;
2737 	}
2738 
2739 	for (i = 0; i <= RTA_MAX; i++) {
2740 		if (!tb[i])
2741 			continue;
2742 
2743 		switch (i) {
2744 		case RTA_SRC:
2745 		case RTA_DST:
2746 		case RTA_TABLE:
2747 			break;
2748 		default:
2749 			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2750 			return -EINVAL;
2751 		}
2752 	}
2753 
2754 	return 0;
2755 }
2756 
2757 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2758 			     struct netlink_ext_ack *extack)
2759 {
2760 	struct net *net = sock_net(in_skb->sk);
2761 	struct nlattr *tb[RTA_MAX + 1];
2762 	struct mfc_cache *cache;
2763 	struct mr_table *mrt;
2764 	struct sk_buff *skb;
2765 	__be32 src, grp;
2766 	u32 tableid;
2767 	int err;
2768 
2769 	err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2770 	if (err < 0)
2771 		goto errout;
2772 
2773 	src = nla_get_in_addr_default(tb[RTA_SRC], 0);
2774 	grp = nla_get_in_addr_default(tb[RTA_DST], 0);
2775 	tableid = nla_get_u32_default(tb[RTA_TABLE], 0);
2776 
2777 	skb = nlmsg_new(mroute_msgsize(false), GFP_KERNEL);
2778 	if (!skb) {
2779 		err = -ENOBUFS;
2780 		goto errout;
2781 	}
2782 
2783 	rcu_read_lock();
2784 
2785 	mrt = __ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2786 	if (!mrt) {
2787 		err = -ENOENT;
2788 		goto errout_unlock;
2789 	}
2790 
2791 	cache = ipmr_cache_find(mrt, src, grp);
2792 	if (!cache) {
2793 		err = -ENOENT;
2794 		goto errout_unlock;
2795 	}
2796 
2797 	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2798 			       nlh->nlmsg_seq, cache,
2799 			       RTM_NEWROUTE, 0);
2800 	if (err < 0)
2801 		goto errout_unlock;
2802 
2803 	rcu_read_unlock();
2804 
2805 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2806 errout:
2807 	return err;
2808 
2809 errout_unlock:
2810 	rcu_read_unlock();
2811 	kfree_skb(skb);
2812 	goto errout;
2813 }
2814 
2815 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2816 {
2817 	struct fib_dump_filter filter = {};
2818 	int err;
2819 
2820 	rcu_read_lock();
2821 
2822 	if (cb->strict_check) {
2823 		err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2824 					    &filter, cb);
2825 		if (err < 0)
2826 			goto out;
2827 	}
2828 
2829 	if (filter.table_id) {
2830 		struct mr_table *mrt;
2831 
2832 		mrt = __ipmr_get_table(sock_net(skb->sk), filter.table_id);
2833 		if (!mrt) {
2834 			if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR) {
2835 				err = skb->len;
2836 				goto out;
2837 			}
2838 
2839 			NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2840 			err = -ENOENT;
2841 			goto out;
2842 		}
2843 
2844 		err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2845 				    &mfc_unres_lock, &filter);
2846 		err = skb->len ? : err;
2847 		goto out;
2848 	}
2849 
2850 	err = mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2851 			       _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2852 out:
2853 	rcu_read_unlock();
2854 
2855 	return err;
2856 }
2857 
2858 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2859 	[RTA_SRC]	= { .type = NLA_U32 },
2860 	[RTA_DST]	= { .type = NLA_U32 },
2861 	[RTA_IIF]	= { .type = NLA_U32 },
2862 	[RTA_TABLE]	= { .type = NLA_U32 },
2863 	[RTA_MULTIPATH]	= { .len = sizeof(struct rtnexthop) },
2864 };
2865 
2866 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2867 {
2868 	switch (rtm_protocol) {
2869 	case RTPROT_STATIC:
2870 	case RTPROT_MROUTED:
2871 		return true;
2872 	}
2873 	return false;
2874 }
2875 
2876 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2877 {
2878 	struct rtnexthop *rtnh = nla_data(nla);
2879 	int remaining = nla_len(nla), vifi = 0;
2880 
2881 	while (rtnh_ok(rtnh, remaining)) {
2882 		mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2883 		if (++vifi == MAXVIFS)
2884 			break;
2885 		rtnh = rtnh_next(rtnh, &remaining);
2886 	}
2887 
2888 	return remaining > 0 ? -EINVAL : vifi;
2889 }
2890 
2891 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2892 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2893 			    struct mfcctl *mfcc, int *mrtsock,
2894 			    struct mr_table **mrtret,
2895 			    struct netlink_ext_ack *extack)
2896 {
2897 	struct net_device *dev = NULL;
2898 	u32 tblid = RT_TABLE_DEFAULT;
2899 	int ret, rem, iif = 0;
2900 	struct mr_table *mrt;
2901 	struct nlattr *attr;
2902 	struct rtmsg *rtm;
2903 
2904 	ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2905 					rtm_ipmr_policy, extack);
2906 	if (ret < 0)
2907 		goto out;
2908 	rtm = nlmsg_data(nlh);
2909 
2910 	ret = -EINVAL;
2911 	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2912 	    rtm->rtm_type != RTN_MULTICAST ||
2913 	    rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2914 	    !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2915 		goto out;
2916 
2917 	memset(mfcc, 0, sizeof(*mfcc));
2918 	mfcc->mfcc_parent = -1;
2919 	ret = 0;
2920 	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2921 		switch (nla_type(attr)) {
2922 		case RTA_SRC:
2923 			mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2924 			break;
2925 		case RTA_DST:
2926 			mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2927 			break;
2928 		case RTA_IIF:
2929 			iif = nla_get_u32(attr);
2930 			break;
2931 		case RTA_MULTIPATH:
2932 			if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2933 				ret = -EINVAL;
2934 				goto out;
2935 			}
2936 			break;
2937 		case RTA_PREFSRC:
2938 			ret = 1;
2939 			break;
2940 		case RTA_TABLE:
2941 			tblid = nla_get_u32(attr);
2942 			break;
2943 		}
2944 	}
2945 
2946 	rcu_read_lock();
2947 
2948 	mrt = __ipmr_get_table(net, tblid);
2949 	if (!mrt) {
2950 		ret = -ENOENT;
2951 		goto unlock;
2952 	}
2953 
2954 	if (iif) {
2955 		dev = dev_get_by_index_rcu(net, iif);
2956 		if (!dev) {
2957 			ret = -ENODEV;
2958 			goto unlock;
2959 		}
2960 
2961 		mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2962 	}
2963 
2964 	*mrtret = mrt;
2965 	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2966 
2967 unlock:
2968 	rcu_read_unlock();
2969 out:
2970 	return ret;
2971 }
2972 
2973 /* takes care of both newroute and delroute */
2974 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2975 			  struct netlink_ext_ack *extack)
2976 {
2977 	struct net *net = sock_net(skb->sk);
2978 	int ret, mrtsock = 0, parent;
2979 	struct mr_table *tbl = NULL;
2980 	struct mfcctl mfcc;
2981 
2982 	ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2983 	if (ret < 0)
2984 		return ret;
2985 
2986 	parent = ret ? mfcc.mfcc_parent : -1;
2987 
2988 	mutex_lock(&net->ipv4.mfc_mutex);
2989 
2990 	if (nlh->nlmsg_type == RTM_NEWROUTE)
2991 		ret = ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2992 	else
2993 		ret = ipmr_mfc_delete(tbl, &mfcc, parent);
2994 
2995 	mutex_unlock(&net->ipv4.mfc_mutex);
2996 
2997 	return ret;
2998 }
2999 
3000 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
3001 {
3002 	if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
3003 	    nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN,
3004 			READ_ONCE(mrt->cache_resolve_queue_len)) ||
3005 	    nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
3006 			READ_ONCE(mrt->mroute_reg_vif_num)) ||
3007 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
3008 		       READ_ONCE(mrt->mroute_do_assert)) ||
3009 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM,
3010 		       READ_ONCE(mrt->mroute_do_pim)) ||
3011 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
3012 		       READ_ONCE(mrt->mroute_do_wrvifwhole)))
3013 		return false;
3014 
3015 	return true;
3016 }
3017 
3018 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
3019 {
3020 	struct net_device *vif_dev;
3021 	struct nlattr *vif_nest;
3022 	struct vif_device *vif;
3023 
3024 	vif = &mrt->vif_table[vifid];
3025 	vif_dev = vif_dev_read(vif);
3026 	/* if the VIF doesn't exist just continue */
3027 	if (!vif_dev)
3028 		return true;
3029 
3030 	vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
3031 	if (!vif_nest)
3032 		return false;
3033 
3034 	if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, READ_ONCE(vif_dev->ifindex)) ||
3035 	    nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
3036 	    nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
3037 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, READ_ONCE(vif->bytes_in),
3038 			      IPMRA_VIFA_PAD) ||
3039 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, READ_ONCE(vif->bytes_out),
3040 			      IPMRA_VIFA_PAD) ||
3041 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, READ_ONCE(vif->pkt_in),
3042 			      IPMRA_VIFA_PAD) ||
3043 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, READ_ONCE(vif->pkt_out),
3044 			      IPMRA_VIFA_PAD) ||
3045 	    nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
3046 	    nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
3047 		nla_nest_cancel(skb, vif_nest);
3048 		return false;
3049 	}
3050 	nla_nest_end(skb, vif_nest);
3051 
3052 	return true;
3053 }
3054 
3055 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
3056 			       struct netlink_ext_ack *extack)
3057 {
3058 	struct ifinfomsg *ifm;
3059 
3060 	ifm = nlmsg_payload(nlh, sizeof(*ifm));
3061 	if (!ifm) {
3062 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
3063 		return -EINVAL;
3064 	}
3065 
3066 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
3067 		NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
3068 		return -EINVAL;
3069 	}
3070 
3071 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3072 	    ifm->ifi_change || ifm->ifi_index) {
3073 		NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
3074 		return -EINVAL;
3075 	}
3076 
3077 	return 0;
3078 }
3079 
3080 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
3081 {
3082 	struct net *net = sock_net(skb->sk);
3083 	struct nlmsghdr *nlh = NULL;
3084 	unsigned int t = 0, s_t;
3085 	unsigned int e = 0, s_e;
3086 	struct mr_table *mrt;
3087 
3088 	if (cb->strict_check) {
3089 		int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
3090 
3091 		if (err < 0)
3092 			return err;
3093 	}
3094 
3095 	s_t = cb->args[0];
3096 	s_e = cb->args[1];
3097 
3098 	rcu_read_lock();
3099 
3100 	ipmr_for_each_table(mrt, net) {
3101 		struct nlattr *vifs, *af;
3102 		struct ifinfomsg *hdr;
3103 		u32 i;
3104 
3105 		if (t < s_t)
3106 			goto skip_table;
3107 		nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
3108 				cb->nlh->nlmsg_seq, RTM_NEWLINK,
3109 				sizeof(*hdr), NLM_F_MULTI);
3110 		if (!nlh)
3111 			break;
3112 
3113 		hdr = nlmsg_data(nlh);
3114 		memset(hdr, 0, sizeof(*hdr));
3115 		hdr->ifi_family = RTNL_FAMILY_IPMR;
3116 
3117 		af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
3118 		if (!af) {
3119 			nlmsg_cancel(skb, nlh);
3120 			goto out;
3121 		}
3122 
3123 		if (!ipmr_fill_table(mrt, skb)) {
3124 			nlmsg_cancel(skb, nlh);
3125 			goto out;
3126 		}
3127 
3128 		vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
3129 		if (!vifs) {
3130 			nla_nest_end(skb, af);
3131 			nlmsg_end(skb, nlh);
3132 			goto out;
3133 		}
3134 		for (i = 0; i < READ_ONCE(mrt->maxvif); i++) {
3135 			if (e < s_e)
3136 				goto skip_entry;
3137 			if (!ipmr_fill_vif(mrt, i, skb)) {
3138 				nla_nest_end(skb, vifs);
3139 				nla_nest_end(skb, af);
3140 				nlmsg_end(skb, nlh);
3141 				goto out;
3142 			}
3143 skip_entry:
3144 			e++;
3145 		}
3146 		s_e = 0;
3147 		e = 0;
3148 		nla_nest_end(skb, vifs);
3149 		nla_nest_end(skb, af);
3150 		nlmsg_end(skb, nlh);
3151 skip_table:
3152 		t++;
3153 	}
3154 
3155 out:
3156 	rcu_read_unlock();
3157 
3158 	cb->args[1] = e;
3159 	cb->args[0] = t;
3160 
3161 	return skb->len;
3162 }
3163 
3164 #ifdef CONFIG_PROC_FS
3165 /* The /proc interfaces to multicast routing :
3166  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
3167  */
3168 
3169 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
3170 	__acquires(RCU)
3171 {
3172 	struct mr_vif_iter *iter = seq->private;
3173 	struct net *net = seq_file_net(seq);
3174 	struct mr_table *mrt;
3175 
3176 	rcu_read_lock();
3177 	mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
3178 	if (!mrt) {
3179 		rcu_read_unlock();
3180 		return ERR_PTR(-ENOENT);
3181 	}
3182 
3183 	iter->mrt = mrt;
3184 
3185 	return mr_vif_seq_start(seq, pos);
3186 }
3187 
3188 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
3189 	__releases(RCU)
3190 {
3191 	rcu_read_unlock();
3192 }
3193 
3194 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
3195 {
3196 	struct mr_vif_iter *iter = seq->private;
3197 	struct mr_table *mrt = iter->mrt;
3198 
3199 	if (v == SEQ_START_TOKEN) {
3200 		seq_puts(seq,
3201 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
3202 	} else {
3203 		const struct vif_device *vif = v;
3204 		const struct net_device *vif_dev;
3205 		const char *name;
3206 
3207 		vif_dev = vif_dev_read(vif);
3208 		name = vif_dev ? vif_dev->name : "none";
3209 		seq_printf(seq,
3210 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
3211 			   vif - mrt->vif_table,
3212 			   name,
3213 			   READ_ONCE(vif->bytes_in), READ_ONCE(vif->pkt_in),
3214 			   READ_ONCE(vif->bytes_out), READ_ONCE(vif->pkt_out),
3215 			   vif->flags, vif->local, vif->remote);
3216 	}
3217 	return 0;
3218 }
3219 
3220 static const struct seq_operations ipmr_vif_seq_ops = {
3221 	.start = ipmr_vif_seq_start,
3222 	.next  = mr_vif_seq_next,
3223 	.stop  = ipmr_vif_seq_stop,
3224 	.show  = ipmr_vif_seq_show,
3225 };
3226 
3227 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
3228 {
3229 	struct net *net = seq_file_net(seq);
3230 	struct mr_table *mrt;
3231 
3232 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
3233 	if (!mrt)
3234 		return ERR_PTR(-ENOENT);
3235 
3236 	return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
3237 }
3238 
3239 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
3240 {
3241 	int n;
3242 
3243 	if (v == SEQ_START_TOKEN) {
3244 		seq_puts(seq,
3245 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
3246 	} else {
3247 		const struct mfc_cache *mfc = v;
3248 		const struct mr_mfc_iter *it = seq->private;
3249 		const struct mr_table *mrt = it->mrt;
3250 
3251 		seq_printf(seq, "%08X %08X %-3hd",
3252 			   (__force u32) mfc->mfc_mcastgrp,
3253 			   (__force u32) mfc->mfc_origin,
3254 			   mfc->_c.mfc_parent);
3255 
3256 		if (it->cache != &mrt->mfc_unres_queue) {
3257 			seq_printf(seq, " %8lu %8lu %8lu",
3258 				   atomic_long_read(&mfc->_c.mfc_un.res.pkt),
3259 				   atomic_long_read(&mfc->_c.mfc_un.res.bytes),
3260 				   atomic_long_read(&mfc->_c.mfc_un.res.wrong_if));
3261 			for (n = mfc->_c.mfc_un.res.minvif;
3262 			     n < mfc->_c.mfc_un.res.maxvif; n++) {
3263 				if (VIF_EXISTS(mrt, n) &&
3264 				    mfc->_c.mfc_un.res.ttls[n] < 255)
3265 					seq_printf(seq,
3266 					   " %2d:%-3d",
3267 					   n, mfc->_c.mfc_un.res.ttls[n]);
3268 			}
3269 		} else {
3270 			/* unresolved mfc_caches don't contain
3271 			 * pkt, bytes and wrong_if values
3272 			 */
3273 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3274 		}
3275 		seq_putc(seq, '\n');
3276 	}
3277 	return 0;
3278 }
3279 
3280 static const struct seq_operations ipmr_mfc_seq_ops = {
3281 	.start = ipmr_mfc_seq_start,
3282 	.next  = mr_mfc_seq_next,
3283 	.stop  = mr_mfc_seq_stop,
3284 	.show  = ipmr_mfc_seq_show,
3285 };
3286 #endif
3287 
3288 #ifdef CONFIG_IP_PIMSM_V2
3289 static const struct net_protocol pim_protocol = {
3290 	.handler	=	pim_rcv,
3291 };
3292 #endif
3293 
3294 static unsigned int ipmr_seq_read(const struct net *net)
3295 {
3296 	return atomic_read(&net->ipv4.ipmr_seq) + ipmr_rules_seq_read(net);
3297 }
3298 
3299 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3300 		     struct netlink_ext_ack *extack)
3301 {
3302 	return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3303 		       ipmr_mr_table_iter, extack);
3304 }
3305 
3306 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3307 	.family		= RTNL_FAMILY_IPMR,
3308 	.fib_seq_read	= ipmr_seq_read,
3309 	.fib_dump	= ipmr_dump,
3310 	.owner		= THIS_MODULE,
3311 };
3312 
3313 static int __net_init ipmr_notifier_init(struct net *net)
3314 {
3315 	struct fib_notifier_ops *ops;
3316 
3317 	atomic_set(&net->ipv4.ipmr_seq, 0);
3318 
3319 	ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3320 	if (IS_ERR(ops))
3321 		return PTR_ERR(ops);
3322 	net->ipv4.ipmr_notifier_ops = ops;
3323 
3324 	return 0;
3325 }
3326 
3327 static void __net_exit ipmr_notifier_exit(struct net *net)
3328 {
3329 	fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3330 	net->ipv4.ipmr_notifier_ops = NULL;
3331 }
3332 
3333 /* Setup for IP multicast routing */
3334 static int __net_init ipmr_net_init(struct net *net)
3335 {
3336 	LIST_HEAD(dev_kill_list);
3337 	int err;
3338 
3339 	mutex_init(&net->ipv4.mfc_mutex);
3340 
3341 	err = ipmr_notifier_init(net);
3342 	if (err)
3343 		goto ipmr_notifier_fail;
3344 
3345 	err = ipmr_rules_init(net);
3346 	if (err < 0)
3347 		goto ipmr_rules_fail;
3348 
3349 #ifdef CONFIG_PROC_FS
3350 	err = -ENOMEM;
3351 	if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3352 			sizeof(struct mr_vif_iter)))
3353 		goto proc_vif_fail;
3354 	if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3355 			sizeof(struct mr_mfc_iter)))
3356 		goto proc_cache_fail;
3357 #endif
3358 	return 0;
3359 
3360 #ifdef CONFIG_PROC_FS
3361 proc_cache_fail:
3362 	remove_proc_entry("ip_mr_vif", net->proc_net);
3363 proc_vif_fail:
3364 	ipmr_rules_exit_rtnl(net, &dev_kill_list);
3365 	ipmr_rules_exit(net);
3366 #endif
3367 ipmr_rules_fail:
3368 	ipmr_notifier_exit(net);
3369 ipmr_notifier_fail:
3370 	return err;
3371 }
3372 
3373 static void __net_exit ipmr_net_exit(struct net *net)
3374 {
3375 #ifdef CONFIG_PROC_FS
3376 	remove_proc_entry("ip_mr_cache", net->proc_net);
3377 	remove_proc_entry("ip_mr_vif", net->proc_net);
3378 #endif
3379 	ipmr_rules_exit(net);
3380 	ipmr_notifier_exit(net);
3381 }
3382 
3383 static void __net_exit ipmr_net_exit_rtnl(struct net *net,
3384 					  struct list_head *dev_kill_list)
3385 {
3386 	ipmr_rules_exit_rtnl(net, dev_kill_list);
3387 }
3388 
3389 static struct pernet_operations ipmr_net_ops = {
3390 	.init = ipmr_net_init,
3391 	.exit = ipmr_net_exit,
3392 	.exit_rtnl = ipmr_net_exit_rtnl,
3393 };
3394 
3395 static const struct rtnl_msg_handler ipmr_rtnl_msg_handlers[] __initconst = {
3396 	{.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETLINK,
3397 	 .dumpit = ipmr_rtm_dumplink, .flags = RTNL_FLAG_DUMP_UNLOCKED},
3398 	{.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_NEWROUTE,
3399 	 .doit = ipmr_rtm_route, .flags = RTNL_FLAG_DOIT_UNLOCKED},
3400 	{.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_DELROUTE,
3401 	 .doit = ipmr_rtm_route, .flags = RTNL_FLAG_DOIT_UNLOCKED},
3402 	{.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETROUTE,
3403 	 .doit = ipmr_rtm_getroute, .dumpit = ipmr_rtm_dumproute,
3404 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
3405 };
3406 
3407 int __init ip_mr_init(void)
3408 {
3409 	int err;
3410 
3411 	mrt_cachep = KMEM_CACHE(mfc_cache,
3412 				SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT);
3413 
3414 	err = register_pernet_subsys(&ipmr_net_ops);
3415 	if (err)
3416 		goto reg_pernet_fail;
3417 
3418 	err = register_netdevice_notifier(&ip_mr_notifier);
3419 	if (err)
3420 		goto reg_notif_fail;
3421 #ifdef CONFIG_IP_PIMSM_V2
3422 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3423 		pr_err("%s: can't add PIM protocol\n", __func__);
3424 		err = -EAGAIN;
3425 		goto add_proto_fail;
3426 	}
3427 #endif
3428 	rtnl_register_many(ipmr_rtnl_msg_handlers);
3429 
3430 	return 0;
3431 
3432 #ifdef CONFIG_IP_PIMSM_V2
3433 add_proto_fail:
3434 	unregister_netdevice_notifier(&ip_mr_notifier);
3435 #endif
3436 reg_notif_fail:
3437 	unregister_pernet_subsys(&ipmr_net_ops);
3438 reg_pernet_fail:
3439 	kmem_cache_destroy(mrt_cachep);
3440 	return err;
3441 }
3442