xref: /openbmc/linux/net/ipv4/ipmr.c (revision 94b00cd6)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IP multicast routing support for mrouted 3.6/3.8
4  *
5  *		(c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6  *	  Linux Consultancy and Custom Driver Development
7  *
8  *	Fixes:
9  *	Michael Chastain	:	Incorrect size of copying.
10  *	Alan Cox		:	Added the cache manager code
11  *	Alan Cox		:	Fixed the clone/copy bug and device race.
12  *	Mike McLagan		:	Routing by source
13  *	Malcolm Beattie		:	Buffer handling fixes.
14  *	Alexey Kuznetsov	:	Double buffer free and other fixes.
15  *	SVR Anand		:	Fixed several multicast bugs and problems.
16  *	Alexey Kuznetsov	:	Status, optimisations and more.
17  *	Brad Parker		:	Better behaviour on mrouted upcall
18  *					overflow.
19  *      Carlos Picoto           :       PIMv1 Support
20  *	Pavlin Ivanov Radoslavov:	PIMv2 Registers must checksum only PIM header
21  *					Relax this requirement to work with older peers.
22  */
23 
24 #include <linux/uaccess.h>
25 #include <linux/types.h>
26 #include <linux/cache.h>
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/mm.h>
30 #include <linux/kernel.h>
31 #include <linux/fcntl.h>
32 #include <linux/stat.h>
33 #include <linux/socket.h>
34 #include <linux/in.h>
35 #include <linux/inet.h>
36 #include <linux/netdevice.h>
37 #include <linux/inetdevice.h>
38 #include <linux/igmp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/mroute.h>
42 #include <linux/init.h>
43 #include <linux/if_ether.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <linux/skbuff.h>
49 #include <net/route.h>
50 #include <net/icmp.h>
51 #include <net/udp.h>
52 #include <net/raw.h>
53 #include <linux/notifier.h>
54 #include <linux/if_arp.h>
55 #include <linux/netfilter_ipv4.h>
56 #include <linux/compat.h>
57 #include <linux/export.h>
58 #include <linux/rhashtable.h>
59 #include <net/ip_tunnels.h>
60 #include <net/checksum.h>
61 #include <net/netlink.h>
62 #include <net/fib_rules.h>
63 #include <linux/netconf.h>
64 #include <net/rtnh.h>
65 
66 #include <linux/nospec.h>
67 
68 struct ipmr_rule {
69 	struct fib_rule		common;
70 };
71 
72 struct ipmr_result {
73 	struct mr_table		*mrt;
74 };
75 
76 /* Big lock, protecting vif table, mrt cache and mroute socket state.
77  * Note that the changes are semaphored via rtnl_lock.
78  */
79 
80 static DEFINE_SPINLOCK(mrt_lock);
81 
82 static struct net_device *vif_dev_read(const struct vif_device *vif)
83 {
84 	return rcu_dereference(vif->dev);
85 }
86 
87 /* Multicast router control variables */
88 
89 /* Special spinlock for queue of unresolved entries */
90 static DEFINE_SPINLOCK(mfc_unres_lock);
91 
92 /* We return to original Alan's scheme. Hash table of resolved
93  * entries is changed only in process context and protected
94  * with weak lock mrt_lock. Queue of unresolved entries is protected
95  * with strong spinlock mfc_unres_lock.
96  *
97  * In this case data path is free of exclusive locks at all.
98  */
99 
100 static struct kmem_cache *mrt_cachep __ro_after_init;
101 
102 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
103 static void ipmr_free_table(struct mr_table *mrt);
104 
105 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
106 			  struct net_device *dev, struct sk_buff *skb,
107 			  struct mfc_cache *cache, int local);
108 static int ipmr_cache_report(const struct mr_table *mrt,
109 			     struct sk_buff *pkt, vifi_t vifi, int assert);
110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
111 				 int cmd);
112 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
113 static void mroute_clean_tables(struct mr_table *mrt, int flags);
114 static void ipmr_expire_process(struct timer_list *t);
115 
116 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
117 #define ipmr_for_each_table(mrt, net)					\
118 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,	\
119 				lockdep_rtnl_is_held() ||		\
120 				list_empty(&net->ipv4.mr_tables))
121 
122 static struct mr_table *ipmr_mr_table_iter(struct net *net,
123 					   struct mr_table *mrt)
124 {
125 	struct mr_table *ret;
126 
127 	if (!mrt)
128 		ret = list_entry_rcu(net->ipv4.mr_tables.next,
129 				     struct mr_table, list);
130 	else
131 		ret = list_entry_rcu(mrt->list.next,
132 				     struct mr_table, list);
133 
134 	if (&ret->list == &net->ipv4.mr_tables)
135 		return NULL;
136 	return ret;
137 }
138 
139 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
140 {
141 	struct mr_table *mrt;
142 
143 	ipmr_for_each_table(mrt, net) {
144 		if (mrt->id == id)
145 			return mrt;
146 	}
147 	return NULL;
148 }
149 
150 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
151 			   struct mr_table **mrt)
152 {
153 	int err;
154 	struct ipmr_result res;
155 	struct fib_lookup_arg arg = {
156 		.result = &res,
157 		.flags = FIB_LOOKUP_NOREF,
158 	};
159 
160 	/* update flow if oif or iif point to device enslaved to l3mdev */
161 	l3mdev_update_flow(net, flowi4_to_flowi(flp4));
162 
163 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
164 			       flowi4_to_flowi(flp4), 0, &arg);
165 	if (err < 0)
166 		return err;
167 	*mrt = res.mrt;
168 	return 0;
169 }
170 
171 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
172 			    int flags, struct fib_lookup_arg *arg)
173 {
174 	struct ipmr_result *res = arg->result;
175 	struct mr_table *mrt;
176 
177 	switch (rule->action) {
178 	case FR_ACT_TO_TBL:
179 		break;
180 	case FR_ACT_UNREACHABLE:
181 		return -ENETUNREACH;
182 	case FR_ACT_PROHIBIT:
183 		return -EACCES;
184 	case FR_ACT_BLACKHOLE:
185 	default:
186 		return -EINVAL;
187 	}
188 
189 	arg->table = fib_rule_get_table(rule, arg);
190 
191 	mrt = ipmr_get_table(rule->fr_net, arg->table);
192 	if (!mrt)
193 		return -EAGAIN;
194 	res->mrt = mrt;
195 	return 0;
196 }
197 
198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
199 {
200 	return 1;
201 }
202 
203 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
204 			       struct fib_rule_hdr *frh, struct nlattr **tb,
205 			       struct netlink_ext_ack *extack)
206 {
207 	return 0;
208 }
209 
210 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
211 			     struct nlattr **tb)
212 {
213 	return 1;
214 }
215 
216 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
217 			  struct fib_rule_hdr *frh)
218 {
219 	frh->dst_len = 0;
220 	frh->src_len = 0;
221 	frh->tos     = 0;
222 	return 0;
223 }
224 
225 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
226 	.family		= RTNL_FAMILY_IPMR,
227 	.rule_size	= sizeof(struct ipmr_rule),
228 	.addr_size	= sizeof(u32),
229 	.action		= ipmr_rule_action,
230 	.match		= ipmr_rule_match,
231 	.configure	= ipmr_rule_configure,
232 	.compare	= ipmr_rule_compare,
233 	.fill		= ipmr_rule_fill,
234 	.nlgroup	= RTNLGRP_IPV4_RULE,
235 	.owner		= THIS_MODULE,
236 };
237 
238 static int __net_init ipmr_rules_init(struct net *net)
239 {
240 	struct fib_rules_ops *ops;
241 	struct mr_table *mrt;
242 	int err;
243 
244 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 	if (IS_ERR(ops))
246 		return PTR_ERR(ops);
247 
248 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
249 
250 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 	if (IS_ERR(mrt)) {
252 		err = PTR_ERR(mrt);
253 		goto err1;
254 	}
255 
256 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
257 	if (err < 0)
258 		goto err2;
259 
260 	net->ipv4.mr_rules_ops = ops;
261 	return 0;
262 
263 err2:
264 	rtnl_lock();
265 	ipmr_free_table(mrt);
266 	rtnl_unlock();
267 err1:
268 	fib_rules_unregister(ops);
269 	return err;
270 }
271 
272 static void __net_exit ipmr_rules_exit(struct net *net)
273 {
274 	struct mr_table *mrt, *next;
275 
276 	ASSERT_RTNL();
277 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
278 		list_del(&mrt->list);
279 		ipmr_free_table(mrt);
280 	}
281 	fib_rules_unregister(net->ipv4.mr_rules_ops);
282 }
283 
284 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
285 			   struct netlink_ext_ack *extack)
286 {
287 	return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
288 }
289 
290 static unsigned int ipmr_rules_seq_read(struct net *net)
291 {
292 	return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
293 }
294 
295 bool ipmr_rule_default(const struct fib_rule *rule)
296 {
297 	return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
298 }
299 EXPORT_SYMBOL(ipmr_rule_default);
300 #else
301 #define ipmr_for_each_table(mrt, net) \
302 	for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
303 
304 static struct mr_table *ipmr_mr_table_iter(struct net *net,
305 					   struct mr_table *mrt)
306 {
307 	if (!mrt)
308 		return net->ipv4.mrt;
309 	return NULL;
310 }
311 
312 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
313 {
314 	return net->ipv4.mrt;
315 }
316 
317 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
318 			   struct mr_table **mrt)
319 {
320 	*mrt = net->ipv4.mrt;
321 	return 0;
322 }
323 
324 static int __net_init ipmr_rules_init(struct net *net)
325 {
326 	struct mr_table *mrt;
327 
328 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
329 	if (IS_ERR(mrt))
330 		return PTR_ERR(mrt);
331 	net->ipv4.mrt = mrt;
332 	return 0;
333 }
334 
335 static void __net_exit ipmr_rules_exit(struct net *net)
336 {
337 	ASSERT_RTNL();
338 	ipmr_free_table(net->ipv4.mrt);
339 	net->ipv4.mrt = NULL;
340 }
341 
342 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
343 			   struct netlink_ext_ack *extack)
344 {
345 	return 0;
346 }
347 
348 static unsigned int ipmr_rules_seq_read(struct net *net)
349 {
350 	return 0;
351 }
352 
353 bool ipmr_rule_default(const struct fib_rule *rule)
354 {
355 	return true;
356 }
357 EXPORT_SYMBOL(ipmr_rule_default);
358 #endif
359 
360 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
361 				const void *ptr)
362 {
363 	const struct mfc_cache_cmp_arg *cmparg = arg->key;
364 	const struct mfc_cache *c = ptr;
365 
366 	return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
367 	       cmparg->mfc_origin != c->mfc_origin;
368 }
369 
370 static const struct rhashtable_params ipmr_rht_params = {
371 	.head_offset = offsetof(struct mr_mfc, mnode),
372 	.key_offset = offsetof(struct mfc_cache, cmparg),
373 	.key_len = sizeof(struct mfc_cache_cmp_arg),
374 	.nelem_hint = 3,
375 	.obj_cmpfn = ipmr_hash_cmp,
376 	.automatic_shrinking = true,
377 };
378 
379 static void ipmr_new_table_set(struct mr_table *mrt,
380 			       struct net *net)
381 {
382 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
383 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384 #endif
385 }
386 
387 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
388 	.mfc_mcastgrp = htonl(INADDR_ANY),
389 	.mfc_origin = htonl(INADDR_ANY),
390 };
391 
392 static struct mr_table_ops ipmr_mr_table_ops = {
393 	.rht_params = &ipmr_rht_params,
394 	.cmparg_any = &ipmr_mr_table_ops_cmparg_any,
395 };
396 
397 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
398 {
399 	struct mr_table *mrt;
400 
401 	/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
402 	if (id != RT_TABLE_DEFAULT && id >= 1000000000)
403 		return ERR_PTR(-EINVAL);
404 
405 	mrt = ipmr_get_table(net, id);
406 	if (mrt)
407 		return mrt;
408 
409 	return mr_table_alloc(net, id, &ipmr_mr_table_ops,
410 			      ipmr_expire_process, ipmr_new_table_set);
411 }
412 
413 static void ipmr_free_table(struct mr_table *mrt)
414 {
415 	timer_shutdown_sync(&mrt->ipmr_expire_timer);
416 	mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
417 				 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
418 	rhltable_destroy(&mrt->mfc_hash);
419 	kfree(mrt);
420 }
421 
422 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
423 
424 /* Initialize ipmr pimreg/tunnel in_device */
425 static bool ipmr_init_vif_indev(const struct net_device *dev)
426 {
427 	struct in_device *in_dev;
428 
429 	ASSERT_RTNL();
430 
431 	in_dev = __in_dev_get_rtnl(dev);
432 	if (!in_dev)
433 		return false;
434 	ipv4_devconf_setall(in_dev);
435 	neigh_parms_data_state_setall(in_dev->arp_parms);
436 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
437 
438 	return true;
439 }
440 
441 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
442 {
443 	struct net_device *tunnel_dev, *new_dev;
444 	struct ip_tunnel_parm p = { };
445 	int err;
446 
447 	tunnel_dev = __dev_get_by_name(net, "tunl0");
448 	if (!tunnel_dev)
449 		goto out;
450 
451 	p.iph.daddr = v->vifc_rmt_addr.s_addr;
452 	p.iph.saddr = v->vifc_lcl_addr.s_addr;
453 	p.iph.version = 4;
454 	p.iph.ihl = 5;
455 	p.iph.protocol = IPPROTO_IPIP;
456 	sprintf(p.name, "dvmrp%d", v->vifc_vifi);
457 
458 	if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
459 		goto out;
460 	err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
461 			SIOCADDTUNNEL);
462 	if (err)
463 		goto out;
464 
465 	new_dev = __dev_get_by_name(net, p.name);
466 	if (!new_dev)
467 		goto out;
468 
469 	new_dev->flags |= IFF_MULTICAST;
470 	if (!ipmr_init_vif_indev(new_dev))
471 		goto out_unregister;
472 	if (dev_open(new_dev, NULL))
473 		goto out_unregister;
474 	dev_hold(new_dev);
475 	err = dev_set_allmulti(new_dev, 1);
476 	if (err) {
477 		dev_close(new_dev);
478 		tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
479 				SIOCDELTUNNEL);
480 		dev_put(new_dev);
481 		new_dev = ERR_PTR(err);
482 	}
483 	return new_dev;
484 
485 out_unregister:
486 	unregister_netdevice(new_dev);
487 out:
488 	return ERR_PTR(-ENOBUFS);
489 }
490 
491 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
492 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
493 {
494 	struct net *net = dev_net(dev);
495 	struct mr_table *mrt;
496 	struct flowi4 fl4 = {
497 		.flowi4_oif	= dev->ifindex,
498 		.flowi4_iif	= skb->skb_iif ? : LOOPBACK_IFINDEX,
499 		.flowi4_mark	= skb->mark,
500 	};
501 	int err;
502 
503 	err = ipmr_fib_lookup(net, &fl4, &mrt);
504 	if (err < 0) {
505 		kfree_skb(skb);
506 		return err;
507 	}
508 
509 	DEV_STATS_ADD(dev, tx_bytes, skb->len);
510 	DEV_STATS_INC(dev, tx_packets);
511 	rcu_read_lock();
512 
513 	/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
514 	ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
515 			  IGMPMSG_WHOLEPKT);
516 
517 	rcu_read_unlock();
518 	kfree_skb(skb);
519 	return NETDEV_TX_OK;
520 }
521 
522 static int reg_vif_get_iflink(const struct net_device *dev)
523 {
524 	return 0;
525 }
526 
527 static const struct net_device_ops reg_vif_netdev_ops = {
528 	.ndo_start_xmit	= reg_vif_xmit,
529 	.ndo_get_iflink = reg_vif_get_iflink,
530 };
531 
532 static void reg_vif_setup(struct net_device *dev)
533 {
534 	dev->type		= ARPHRD_PIMREG;
535 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
536 	dev->flags		= IFF_NOARP;
537 	dev->netdev_ops		= &reg_vif_netdev_ops;
538 	dev->needs_free_netdev	= true;
539 	dev->features		|= NETIF_F_NETNS_LOCAL;
540 }
541 
542 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
543 {
544 	struct net_device *dev;
545 	char name[IFNAMSIZ];
546 
547 	if (mrt->id == RT_TABLE_DEFAULT)
548 		sprintf(name, "pimreg");
549 	else
550 		sprintf(name, "pimreg%u", mrt->id);
551 
552 	dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
553 
554 	if (!dev)
555 		return NULL;
556 
557 	dev_net_set(dev, net);
558 
559 	if (register_netdevice(dev)) {
560 		free_netdev(dev);
561 		return NULL;
562 	}
563 
564 	if (!ipmr_init_vif_indev(dev))
565 		goto failure;
566 	if (dev_open(dev, NULL))
567 		goto failure;
568 
569 	dev_hold(dev);
570 
571 	return dev;
572 
573 failure:
574 	unregister_netdevice(dev);
575 	return NULL;
576 }
577 
578 /* called with rcu_read_lock() */
579 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
580 		     unsigned int pimlen)
581 {
582 	struct net_device *reg_dev = NULL;
583 	struct iphdr *encap;
584 	int vif_num;
585 
586 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
587 	/* Check that:
588 	 * a. packet is really sent to a multicast group
589 	 * b. packet is not a NULL-REGISTER
590 	 * c. packet is not truncated
591 	 */
592 	if (!ipv4_is_multicast(encap->daddr) ||
593 	    encap->tot_len == 0 ||
594 	    ntohs(encap->tot_len) + pimlen > skb->len)
595 		return 1;
596 
597 	/* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
598 	vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
599 	if (vif_num >= 0)
600 		reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
601 	if (!reg_dev)
602 		return 1;
603 
604 	skb->mac_header = skb->network_header;
605 	skb_pull(skb, (u8 *)encap - skb->data);
606 	skb_reset_network_header(skb);
607 	skb->protocol = htons(ETH_P_IP);
608 	skb->ip_summed = CHECKSUM_NONE;
609 
610 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
611 
612 	netif_rx(skb);
613 
614 	return NET_RX_SUCCESS;
615 }
616 #else
617 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
618 {
619 	return NULL;
620 }
621 #endif
622 
623 static int call_ipmr_vif_entry_notifiers(struct net *net,
624 					 enum fib_event_type event_type,
625 					 struct vif_device *vif,
626 					 struct net_device *vif_dev,
627 					 vifi_t vif_index, u32 tb_id)
628 {
629 	return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
630 				     vif, vif_dev, vif_index, tb_id,
631 				     &net->ipv4.ipmr_seq);
632 }
633 
634 static int call_ipmr_mfc_entry_notifiers(struct net *net,
635 					 enum fib_event_type event_type,
636 					 struct mfc_cache *mfc, u32 tb_id)
637 {
638 	return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
639 				     &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
640 }
641 
642 /**
643  *	vif_delete - Delete a VIF entry
644  *	@mrt: Table to delete from
645  *	@vifi: VIF identifier to delete
646  *	@notify: Set to 1, if the caller is a notifier_call
647  *	@head: if unregistering the VIF, place it on this queue
648  */
649 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
650 		      struct list_head *head)
651 {
652 	struct net *net = read_pnet(&mrt->net);
653 	struct vif_device *v;
654 	struct net_device *dev;
655 	struct in_device *in_dev;
656 
657 	if (vifi < 0 || vifi >= mrt->maxvif)
658 		return -EADDRNOTAVAIL;
659 
660 	v = &mrt->vif_table[vifi];
661 
662 	dev = rtnl_dereference(v->dev);
663 	if (!dev)
664 		return -EADDRNOTAVAIL;
665 
666 	spin_lock(&mrt_lock);
667 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
668 				      vifi, mrt->id);
669 	RCU_INIT_POINTER(v->dev, NULL);
670 
671 	if (vifi == mrt->mroute_reg_vif_num) {
672 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
673 		WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
674 	}
675 	if (vifi + 1 == mrt->maxvif) {
676 		int tmp;
677 
678 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
679 			if (VIF_EXISTS(mrt, tmp))
680 				break;
681 		}
682 		WRITE_ONCE(mrt->maxvif, tmp + 1);
683 	}
684 
685 	spin_unlock(&mrt_lock);
686 
687 	dev_set_allmulti(dev, -1);
688 
689 	in_dev = __in_dev_get_rtnl(dev);
690 	if (in_dev) {
691 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
692 		inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
693 					    NETCONFA_MC_FORWARDING,
694 					    dev->ifindex, &in_dev->cnf);
695 		ip_rt_multicast_event(in_dev);
696 	}
697 
698 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
699 		unregister_netdevice_queue(dev, head);
700 
701 	netdev_put(dev, &v->dev_tracker);
702 	return 0;
703 }
704 
705 static void ipmr_cache_free_rcu(struct rcu_head *head)
706 {
707 	struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
708 
709 	kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
710 }
711 
712 static void ipmr_cache_free(struct mfc_cache *c)
713 {
714 	call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
715 }
716 
717 /* Destroy an unresolved cache entry, killing queued skbs
718  * and reporting error to netlink readers.
719  */
720 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
721 {
722 	struct net *net = read_pnet(&mrt->net);
723 	struct sk_buff *skb;
724 	struct nlmsgerr *e;
725 
726 	atomic_dec(&mrt->cache_resolve_queue_len);
727 
728 	while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
729 		if (ip_hdr(skb)->version == 0) {
730 			struct nlmsghdr *nlh = skb_pull(skb,
731 							sizeof(struct iphdr));
732 			nlh->nlmsg_type = NLMSG_ERROR;
733 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
734 			skb_trim(skb, nlh->nlmsg_len);
735 			e = nlmsg_data(nlh);
736 			e->error = -ETIMEDOUT;
737 			memset(&e->msg, 0, sizeof(e->msg));
738 
739 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
740 		} else {
741 			kfree_skb(skb);
742 		}
743 	}
744 
745 	ipmr_cache_free(c);
746 }
747 
748 /* Timer process for the unresolved queue. */
749 static void ipmr_expire_process(struct timer_list *t)
750 {
751 	struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
752 	struct mr_mfc *c, *next;
753 	unsigned long expires;
754 	unsigned long now;
755 
756 	if (!spin_trylock(&mfc_unres_lock)) {
757 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
758 		return;
759 	}
760 
761 	if (list_empty(&mrt->mfc_unres_queue))
762 		goto out;
763 
764 	now = jiffies;
765 	expires = 10*HZ;
766 
767 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
768 		if (time_after(c->mfc_un.unres.expires, now)) {
769 			unsigned long interval = c->mfc_un.unres.expires - now;
770 			if (interval < expires)
771 				expires = interval;
772 			continue;
773 		}
774 
775 		list_del(&c->list);
776 		mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
777 		ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
778 	}
779 
780 	if (!list_empty(&mrt->mfc_unres_queue))
781 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
782 
783 out:
784 	spin_unlock(&mfc_unres_lock);
785 }
786 
787 /* Fill oifs list. It is called under locked mrt_lock. */
788 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
789 				   unsigned char *ttls)
790 {
791 	int vifi;
792 
793 	cache->mfc_un.res.minvif = MAXVIFS;
794 	cache->mfc_un.res.maxvif = 0;
795 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
796 
797 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
798 		if (VIF_EXISTS(mrt, vifi) &&
799 		    ttls[vifi] && ttls[vifi] < 255) {
800 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
801 			if (cache->mfc_un.res.minvif > vifi)
802 				cache->mfc_un.res.minvif = vifi;
803 			if (cache->mfc_un.res.maxvif <= vifi)
804 				cache->mfc_un.res.maxvif = vifi + 1;
805 		}
806 	}
807 	cache->mfc_un.res.lastuse = jiffies;
808 }
809 
810 static int vif_add(struct net *net, struct mr_table *mrt,
811 		   struct vifctl *vifc, int mrtsock)
812 {
813 	struct netdev_phys_item_id ppid = { };
814 	int vifi = vifc->vifc_vifi;
815 	struct vif_device *v = &mrt->vif_table[vifi];
816 	struct net_device *dev;
817 	struct in_device *in_dev;
818 	int err;
819 
820 	/* Is vif busy ? */
821 	if (VIF_EXISTS(mrt, vifi))
822 		return -EADDRINUSE;
823 
824 	switch (vifc->vifc_flags) {
825 	case VIFF_REGISTER:
826 		if (!ipmr_pimsm_enabled())
827 			return -EINVAL;
828 		/* Special Purpose VIF in PIM
829 		 * All the packets will be sent to the daemon
830 		 */
831 		if (mrt->mroute_reg_vif_num >= 0)
832 			return -EADDRINUSE;
833 		dev = ipmr_reg_vif(net, mrt);
834 		if (!dev)
835 			return -ENOBUFS;
836 		err = dev_set_allmulti(dev, 1);
837 		if (err) {
838 			unregister_netdevice(dev);
839 			dev_put(dev);
840 			return err;
841 		}
842 		break;
843 	case VIFF_TUNNEL:
844 		dev = ipmr_new_tunnel(net, vifc);
845 		if (IS_ERR(dev))
846 			return PTR_ERR(dev);
847 		break;
848 	case VIFF_USE_IFINDEX:
849 	case 0:
850 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
851 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
852 			if (dev && !__in_dev_get_rtnl(dev)) {
853 				dev_put(dev);
854 				return -EADDRNOTAVAIL;
855 			}
856 		} else {
857 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
858 		}
859 		if (!dev)
860 			return -EADDRNOTAVAIL;
861 		err = dev_set_allmulti(dev, 1);
862 		if (err) {
863 			dev_put(dev);
864 			return err;
865 		}
866 		break;
867 	default:
868 		return -EINVAL;
869 	}
870 
871 	in_dev = __in_dev_get_rtnl(dev);
872 	if (!in_dev) {
873 		dev_put(dev);
874 		return -EADDRNOTAVAIL;
875 	}
876 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
877 	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
878 				    dev->ifindex, &in_dev->cnf);
879 	ip_rt_multicast_event(in_dev);
880 
881 	/* Fill in the VIF structures */
882 	vif_device_init(v, dev, vifc->vifc_rate_limit,
883 			vifc->vifc_threshold,
884 			vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
885 			(VIFF_TUNNEL | VIFF_REGISTER));
886 
887 	err = dev_get_port_parent_id(dev, &ppid, true);
888 	if (err == 0) {
889 		memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
890 		v->dev_parent_id.id_len = ppid.id_len;
891 	} else {
892 		v->dev_parent_id.id_len = 0;
893 	}
894 
895 	v->local = vifc->vifc_lcl_addr.s_addr;
896 	v->remote = vifc->vifc_rmt_addr.s_addr;
897 
898 	/* And finish update writing critical data */
899 	spin_lock(&mrt_lock);
900 	rcu_assign_pointer(v->dev, dev);
901 	netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
902 	if (v->flags & VIFF_REGISTER) {
903 		/* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
904 		WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
905 	}
906 	if (vifi+1 > mrt->maxvif)
907 		WRITE_ONCE(mrt->maxvif, vifi + 1);
908 	spin_unlock(&mrt_lock);
909 	call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
910 				      vifi, mrt->id);
911 	return 0;
912 }
913 
914 /* called with rcu_read_lock() */
915 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
916 					 __be32 origin,
917 					 __be32 mcastgrp)
918 {
919 	struct mfc_cache_cmp_arg arg = {
920 			.mfc_mcastgrp = mcastgrp,
921 			.mfc_origin = origin
922 	};
923 
924 	return mr_mfc_find(mrt, &arg);
925 }
926 
927 /* Look for a (*,G) entry */
928 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
929 					     __be32 mcastgrp, int vifi)
930 {
931 	struct mfc_cache_cmp_arg arg = {
932 			.mfc_mcastgrp = mcastgrp,
933 			.mfc_origin = htonl(INADDR_ANY)
934 	};
935 
936 	if (mcastgrp == htonl(INADDR_ANY))
937 		return mr_mfc_find_any_parent(mrt, vifi);
938 	return mr_mfc_find_any(mrt, vifi, &arg);
939 }
940 
941 /* Look for a (S,G,iif) entry if parent != -1 */
942 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
943 						__be32 origin, __be32 mcastgrp,
944 						int parent)
945 {
946 	struct mfc_cache_cmp_arg arg = {
947 			.mfc_mcastgrp = mcastgrp,
948 			.mfc_origin = origin,
949 	};
950 
951 	return mr_mfc_find_parent(mrt, &arg, parent);
952 }
953 
954 /* Allocate a multicast cache entry */
955 static struct mfc_cache *ipmr_cache_alloc(void)
956 {
957 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
958 
959 	if (c) {
960 		c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
961 		c->_c.mfc_un.res.minvif = MAXVIFS;
962 		c->_c.free = ipmr_cache_free_rcu;
963 		refcount_set(&c->_c.mfc_un.res.refcount, 1);
964 	}
965 	return c;
966 }
967 
968 static struct mfc_cache *ipmr_cache_alloc_unres(void)
969 {
970 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
971 
972 	if (c) {
973 		skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
974 		c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
975 	}
976 	return c;
977 }
978 
979 /* A cache entry has gone into a resolved state from queued */
980 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
981 			       struct mfc_cache *uc, struct mfc_cache *c)
982 {
983 	struct sk_buff *skb;
984 	struct nlmsgerr *e;
985 
986 	/* Play the pending entries through our router */
987 	while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
988 		if (ip_hdr(skb)->version == 0) {
989 			struct nlmsghdr *nlh = skb_pull(skb,
990 							sizeof(struct iphdr));
991 
992 			if (mr_fill_mroute(mrt, skb, &c->_c,
993 					   nlmsg_data(nlh)) > 0) {
994 				nlh->nlmsg_len = skb_tail_pointer(skb) -
995 						 (u8 *)nlh;
996 			} else {
997 				nlh->nlmsg_type = NLMSG_ERROR;
998 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
999 				skb_trim(skb, nlh->nlmsg_len);
1000 				e = nlmsg_data(nlh);
1001 				e->error = -EMSGSIZE;
1002 				memset(&e->msg, 0, sizeof(e->msg));
1003 			}
1004 
1005 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1006 		} else {
1007 			rcu_read_lock();
1008 			ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1009 			rcu_read_unlock();
1010 		}
1011 	}
1012 }
1013 
1014 /* Bounce a cache query up to mrouted and netlink.
1015  *
1016  * Called under rcu_read_lock().
1017  */
1018 static int ipmr_cache_report(const struct mr_table *mrt,
1019 			     struct sk_buff *pkt, vifi_t vifi, int assert)
1020 {
1021 	const int ihl = ip_hdrlen(pkt);
1022 	struct sock *mroute_sk;
1023 	struct igmphdr *igmp;
1024 	struct igmpmsg *msg;
1025 	struct sk_buff *skb;
1026 	int ret;
1027 
1028 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1029 	if (!mroute_sk)
1030 		return -EINVAL;
1031 
1032 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1033 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1034 	else
1035 		skb = alloc_skb(128, GFP_ATOMIC);
1036 
1037 	if (!skb)
1038 		return -ENOBUFS;
1039 
1040 	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1041 		/* Ugly, but we have no choice with this interface.
1042 		 * Duplicate old header, fix ihl, length etc.
1043 		 * And all this only to mangle msg->im_msgtype and
1044 		 * to set msg->im_mbz to "mbz" :-)
1045 		 */
1046 		skb_push(skb, sizeof(struct iphdr));
1047 		skb_reset_network_header(skb);
1048 		skb_reset_transport_header(skb);
1049 		msg = (struct igmpmsg *)skb_network_header(skb);
1050 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1051 		msg->im_msgtype = assert;
1052 		msg->im_mbz = 0;
1053 		if (assert == IGMPMSG_WRVIFWHOLE) {
1054 			msg->im_vif = vifi;
1055 			msg->im_vif_hi = vifi >> 8;
1056 		} else {
1057 			/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1058 			int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1059 
1060 			msg->im_vif = vif_num;
1061 			msg->im_vif_hi = vif_num >> 8;
1062 		}
1063 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1064 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1065 					     sizeof(struct iphdr));
1066 	} else {
1067 		/* Copy the IP header */
1068 		skb_set_network_header(skb, skb->len);
1069 		skb_put(skb, ihl);
1070 		skb_copy_to_linear_data(skb, pkt->data, ihl);
1071 		/* Flag to the kernel this is a route add */
1072 		ip_hdr(skb)->protocol = 0;
1073 		msg = (struct igmpmsg *)skb_network_header(skb);
1074 		msg->im_vif = vifi;
1075 		msg->im_vif_hi = vifi >> 8;
1076 		ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1077 		memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1078 		/* Add our header */
1079 		igmp = skb_put(skb, sizeof(struct igmphdr));
1080 		igmp->type = assert;
1081 		msg->im_msgtype = assert;
1082 		igmp->code = 0;
1083 		ip_hdr(skb)->tot_len = htons(skb->len);	/* Fix the length */
1084 		skb->transport_header = skb->network_header;
1085 	}
1086 
1087 	igmpmsg_netlink_event(mrt, skb);
1088 
1089 	/* Deliver to mrouted */
1090 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1091 
1092 	if (ret < 0) {
1093 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1094 		kfree_skb(skb);
1095 	}
1096 
1097 	return ret;
1098 }
1099 
1100 /* Queue a packet for resolution. It gets locked cache entry! */
1101 /* Called under rcu_read_lock() */
1102 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1103 				 struct sk_buff *skb, struct net_device *dev)
1104 {
1105 	const struct iphdr *iph = ip_hdr(skb);
1106 	struct mfc_cache *c;
1107 	bool found = false;
1108 	int err;
1109 
1110 	spin_lock_bh(&mfc_unres_lock);
1111 	list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1112 		if (c->mfc_mcastgrp == iph->daddr &&
1113 		    c->mfc_origin == iph->saddr) {
1114 			found = true;
1115 			break;
1116 		}
1117 	}
1118 
1119 	if (!found) {
1120 		/* Create a new entry if allowable */
1121 		c = ipmr_cache_alloc_unres();
1122 		if (!c) {
1123 			spin_unlock_bh(&mfc_unres_lock);
1124 
1125 			kfree_skb(skb);
1126 			return -ENOBUFS;
1127 		}
1128 
1129 		/* Fill in the new cache entry */
1130 		c->_c.mfc_parent = -1;
1131 		c->mfc_origin	= iph->saddr;
1132 		c->mfc_mcastgrp	= iph->daddr;
1133 
1134 		/* Reflect first query at mrouted. */
1135 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1136 
1137 		if (err < 0) {
1138 			/* If the report failed throw the cache entry
1139 			   out - Brad Parker
1140 			 */
1141 			spin_unlock_bh(&mfc_unres_lock);
1142 
1143 			ipmr_cache_free(c);
1144 			kfree_skb(skb);
1145 			return err;
1146 		}
1147 
1148 		atomic_inc(&mrt->cache_resolve_queue_len);
1149 		list_add(&c->_c.list, &mrt->mfc_unres_queue);
1150 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1151 
1152 		if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1153 			mod_timer(&mrt->ipmr_expire_timer,
1154 				  c->_c.mfc_un.unres.expires);
1155 	}
1156 
1157 	/* See if we can append the packet */
1158 	if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1159 		kfree_skb(skb);
1160 		err = -ENOBUFS;
1161 	} else {
1162 		if (dev) {
1163 			skb->dev = dev;
1164 			skb->skb_iif = dev->ifindex;
1165 		}
1166 		skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1167 		err = 0;
1168 	}
1169 
1170 	spin_unlock_bh(&mfc_unres_lock);
1171 	return err;
1172 }
1173 
1174 /* MFC cache manipulation by user space mroute daemon */
1175 
1176 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1177 {
1178 	struct net *net = read_pnet(&mrt->net);
1179 	struct mfc_cache *c;
1180 
1181 	/* The entries are added/deleted only under RTNL */
1182 	rcu_read_lock();
1183 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1184 				   mfc->mfcc_mcastgrp.s_addr, parent);
1185 	rcu_read_unlock();
1186 	if (!c)
1187 		return -ENOENT;
1188 	rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1189 	list_del_rcu(&c->_c.list);
1190 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1191 	mroute_netlink_event(mrt, c, RTM_DELROUTE);
1192 	mr_cache_put(&c->_c);
1193 
1194 	return 0;
1195 }
1196 
1197 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1198 			struct mfcctl *mfc, int mrtsock, int parent)
1199 {
1200 	struct mfc_cache *uc, *c;
1201 	struct mr_mfc *_uc;
1202 	bool found;
1203 	int ret;
1204 
1205 	if (mfc->mfcc_parent >= MAXVIFS)
1206 		return -ENFILE;
1207 
1208 	/* The entries are added/deleted only under RTNL */
1209 	rcu_read_lock();
1210 	c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1211 				   mfc->mfcc_mcastgrp.s_addr, parent);
1212 	rcu_read_unlock();
1213 	if (c) {
1214 		spin_lock(&mrt_lock);
1215 		c->_c.mfc_parent = mfc->mfcc_parent;
1216 		ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1217 		if (!mrtsock)
1218 			c->_c.mfc_flags |= MFC_STATIC;
1219 		spin_unlock(&mrt_lock);
1220 		call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1221 					      mrt->id);
1222 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1223 		return 0;
1224 	}
1225 
1226 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1227 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1228 		return -EINVAL;
1229 
1230 	c = ipmr_cache_alloc();
1231 	if (!c)
1232 		return -ENOMEM;
1233 
1234 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1235 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1236 	c->_c.mfc_parent = mfc->mfcc_parent;
1237 	ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1238 	if (!mrtsock)
1239 		c->_c.mfc_flags |= MFC_STATIC;
1240 
1241 	ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1242 				  ipmr_rht_params);
1243 	if (ret) {
1244 		pr_err("ipmr: rhtable insert error %d\n", ret);
1245 		ipmr_cache_free(c);
1246 		return ret;
1247 	}
1248 	list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1249 	/* Check to see if we resolved a queued list. If so we
1250 	 * need to send on the frames and tidy up.
1251 	 */
1252 	found = false;
1253 	spin_lock_bh(&mfc_unres_lock);
1254 	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1255 		uc = (struct mfc_cache *)_uc;
1256 		if (uc->mfc_origin == c->mfc_origin &&
1257 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1258 			list_del(&_uc->list);
1259 			atomic_dec(&mrt->cache_resolve_queue_len);
1260 			found = true;
1261 			break;
1262 		}
1263 	}
1264 	if (list_empty(&mrt->mfc_unres_queue))
1265 		del_timer(&mrt->ipmr_expire_timer);
1266 	spin_unlock_bh(&mfc_unres_lock);
1267 
1268 	if (found) {
1269 		ipmr_cache_resolve(net, mrt, uc, c);
1270 		ipmr_cache_free(uc);
1271 	}
1272 	call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1273 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1274 	return 0;
1275 }
1276 
1277 /* Close the multicast socket, and clear the vif tables etc */
1278 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1279 {
1280 	struct net *net = read_pnet(&mrt->net);
1281 	struct mr_mfc *c, *tmp;
1282 	struct mfc_cache *cache;
1283 	LIST_HEAD(list);
1284 	int i;
1285 
1286 	/* Shut down all active vif entries */
1287 	if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1288 		for (i = 0; i < mrt->maxvif; i++) {
1289 			if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1290 			     !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1291 			    (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1292 				continue;
1293 			vif_delete(mrt, i, 0, &list);
1294 		}
1295 		unregister_netdevice_many(&list);
1296 	}
1297 
1298 	/* Wipe the cache */
1299 	if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1300 		list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1301 			if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1302 			    (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1303 				continue;
1304 			rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1305 			list_del_rcu(&c->list);
1306 			cache = (struct mfc_cache *)c;
1307 			call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1308 						      mrt->id);
1309 			mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1310 			mr_cache_put(c);
1311 		}
1312 	}
1313 
1314 	if (flags & MRT_FLUSH_MFC) {
1315 		if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1316 			spin_lock_bh(&mfc_unres_lock);
1317 			list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1318 				list_del(&c->list);
1319 				cache = (struct mfc_cache *)c;
1320 				mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1321 				ipmr_destroy_unres(mrt, cache);
1322 			}
1323 			spin_unlock_bh(&mfc_unres_lock);
1324 		}
1325 	}
1326 }
1327 
1328 /* called from ip_ra_control(), before an RCU grace period,
1329  * we don't need to call synchronize_rcu() here
1330  */
1331 static void mrtsock_destruct(struct sock *sk)
1332 {
1333 	struct net *net = sock_net(sk);
1334 	struct mr_table *mrt;
1335 
1336 	rtnl_lock();
1337 	ipmr_for_each_table(mrt, net) {
1338 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1339 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1340 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1341 						    NETCONFA_MC_FORWARDING,
1342 						    NETCONFA_IFINDEX_ALL,
1343 						    net->ipv4.devconf_all);
1344 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1345 			mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1346 		}
1347 	}
1348 	rtnl_unlock();
1349 }
1350 
1351 /* Socket options and virtual interface manipulation. The whole
1352  * virtual interface system is a complete heap, but unfortunately
1353  * that's how BSD mrouted happens to think. Maybe one day with a proper
1354  * MOSPF/PIM router set up we can clean this up.
1355  */
1356 
1357 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1358 			 unsigned int optlen)
1359 {
1360 	struct net *net = sock_net(sk);
1361 	int val, ret = 0, parent = 0;
1362 	struct mr_table *mrt;
1363 	struct vifctl vif;
1364 	struct mfcctl mfc;
1365 	bool do_wrvifwhole;
1366 	u32 uval;
1367 
1368 	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1369 	rtnl_lock();
1370 	if (sk->sk_type != SOCK_RAW ||
1371 	    inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1372 		ret = -EOPNOTSUPP;
1373 		goto out_unlock;
1374 	}
1375 
1376 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1377 	if (!mrt) {
1378 		ret = -ENOENT;
1379 		goto out_unlock;
1380 	}
1381 	if (optname != MRT_INIT) {
1382 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1383 		    !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1384 			ret = -EACCES;
1385 			goto out_unlock;
1386 		}
1387 	}
1388 
1389 	switch (optname) {
1390 	case MRT_INIT:
1391 		if (optlen != sizeof(int)) {
1392 			ret = -EINVAL;
1393 			break;
1394 		}
1395 		if (rtnl_dereference(mrt->mroute_sk)) {
1396 			ret = -EADDRINUSE;
1397 			break;
1398 		}
1399 
1400 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1401 		if (ret == 0) {
1402 			rcu_assign_pointer(mrt->mroute_sk, sk);
1403 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1404 			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1405 						    NETCONFA_MC_FORWARDING,
1406 						    NETCONFA_IFINDEX_ALL,
1407 						    net->ipv4.devconf_all);
1408 		}
1409 		break;
1410 	case MRT_DONE:
1411 		if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1412 			ret = -EACCES;
1413 		} else {
1414 			/* We need to unlock here because mrtsock_destruct takes
1415 			 * care of rtnl itself and we can't change that due to
1416 			 * the IP_ROUTER_ALERT setsockopt which runs without it.
1417 			 */
1418 			rtnl_unlock();
1419 			ret = ip_ra_control(sk, 0, NULL);
1420 			goto out;
1421 		}
1422 		break;
1423 	case MRT_ADD_VIF:
1424 	case MRT_DEL_VIF:
1425 		if (optlen != sizeof(vif)) {
1426 			ret = -EINVAL;
1427 			break;
1428 		}
1429 		if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1430 			ret = -EFAULT;
1431 			break;
1432 		}
1433 		if (vif.vifc_vifi >= MAXVIFS) {
1434 			ret = -ENFILE;
1435 			break;
1436 		}
1437 		if (optname == MRT_ADD_VIF) {
1438 			ret = vif_add(net, mrt, &vif,
1439 				      sk == rtnl_dereference(mrt->mroute_sk));
1440 		} else {
1441 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1442 		}
1443 		break;
1444 	/* Manipulate the forwarding caches. These live
1445 	 * in a sort of kernel/user symbiosis.
1446 	 */
1447 	case MRT_ADD_MFC:
1448 	case MRT_DEL_MFC:
1449 		parent = -1;
1450 		fallthrough;
1451 	case MRT_ADD_MFC_PROXY:
1452 	case MRT_DEL_MFC_PROXY:
1453 		if (optlen != sizeof(mfc)) {
1454 			ret = -EINVAL;
1455 			break;
1456 		}
1457 		if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1458 			ret = -EFAULT;
1459 			break;
1460 		}
1461 		if (parent == 0)
1462 			parent = mfc.mfcc_parent;
1463 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1464 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1465 		else
1466 			ret = ipmr_mfc_add(net, mrt, &mfc,
1467 					   sk == rtnl_dereference(mrt->mroute_sk),
1468 					   parent);
1469 		break;
1470 	case MRT_FLUSH:
1471 		if (optlen != sizeof(val)) {
1472 			ret = -EINVAL;
1473 			break;
1474 		}
1475 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1476 			ret = -EFAULT;
1477 			break;
1478 		}
1479 		mroute_clean_tables(mrt, val);
1480 		break;
1481 	/* Control PIM assert. */
1482 	case MRT_ASSERT:
1483 		if (optlen != sizeof(val)) {
1484 			ret = -EINVAL;
1485 			break;
1486 		}
1487 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1488 			ret = -EFAULT;
1489 			break;
1490 		}
1491 		mrt->mroute_do_assert = val;
1492 		break;
1493 	case MRT_PIM:
1494 		if (!ipmr_pimsm_enabled()) {
1495 			ret = -ENOPROTOOPT;
1496 			break;
1497 		}
1498 		if (optlen != sizeof(val)) {
1499 			ret = -EINVAL;
1500 			break;
1501 		}
1502 		if (copy_from_sockptr(&val, optval, sizeof(val))) {
1503 			ret = -EFAULT;
1504 			break;
1505 		}
1506 
1507 		do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1508 		val = !!val;
1509 		if (val != mrt->mroute_do_pim) {
1510 			mrt->mroute_do_pim = val;
1511 			mrt->mroute_do_assert = val;
1512 			mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1513 		}
1514 		break;
1515 	case MRT_TABLE:
1516 		if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1517 			ret = -ENOPROTOOPT;
1518 			break;
1519 		}
1520 		if (optlen != sizeof(uval)) {
1521 			ret = -EINVAL;
1522 			break;
1523 		}
1524 		if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1525 			ret = -EFAULT;
1526 			break;
1527 		}
1528 
1529 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1530 			ret = -EBUSY;
1531 		} else {
1532 			mrt = ipmr_new_table(net, uval);
1533 			if (IS_ERR(mrt))
1534 				ret = PTR_ERR(mrt);
1535 			else
1536 				raw_sk(sk)->ipmr_table = uval;
1537 		}
1538 		break;
1539 	/* Spurious command, or MRT_VERSION which you cannot set. */
1540 	default:
1541 		ret = -ENOPROTOOPT;
1542 	}
1543 out_unlock:
1544 	rtnl_unlock();
1545 out:
1546 	return ret;
1547 }
1548 
1549 /* Execute if this ioctl is a special mroute ioctl */
1550 int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1551 {
1552 	switch (cmd) {
1553 	/* These userspace buffers will be consumed by ipmr_ioctl() */
1554 	case SIOCGETVIFCNT: {
1555 		struct sioc_vif_req buffer;
1556 
1557 		return sock_ioctl_inout(sk, cmd, arg, &buffer,
1558 				      sizeof(buffer));
1559 		}
1560 	case SIOCGETSGCNT: {
1561 		struct sioc_sg_req buffer;
1562 
1563 		return sock_ioctl_inout(sk, cmd, arg, &buffer,
1564 				      sizeof(buffer));
1565 		}
1566 	}
1567 	/* return code > 0 means that the ioctl was not executed */
1568 	return 1;
1569 }
1570 
1571 /* Getsock opt support for the multicast routing system. */
1572 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1573 			 sockptr_t optlen)
1574 {
1575 	int olr;
1576 	int val;
1577 	struct net *net = sock_net(sk);
1578 	struct mr_table *mrt;
1579 
1580 	if (sk->sk_type != SOCK_RAW ||
1581 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1582 		return -EOPNOTSUPP;
1583 
1584 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1585 	if (!mrt)
1586 		return -ENOENT;
1587 
1588 	switch (optname) {
1589 	case MRT_VERSION:
1590 		val = 0x0305;
1591 		break;
1592 	case MRT_PIM:
1593 		if (!ipmr_pimsm_enabled())
1594 			return -ENOPROTOOPT;
1595 		val = mrt->mroute_do_pim;
1596 		break;
1597 	case MRT_ASSERT:
1598 		val = mrt->mroute_do_assert;
1599 		break;
1600 	default:
1601 		return -ENOPROTOOPT;
1602 	}
1603 
1604 	if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1605 		return -EFAULT;
1606 	if (olr < 0)
1607 		return -EINVAL;
1608 
1609 	olr = min_t(unsigned int, olr, sizeof(int));
1610 
1611 	if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1612 		return -EFAULT;
1613 	if (copy_to_sockptr(optval, &val, olr))
1614 		return -EFAULT;
1615 	return 0;
1616 }
1617 
1618 /* The IP multicast ioctl support routines. */
1619 int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1620 {
1621 	struct vif_device *vif;
1622 	struct mfc_cache *c;
1623 	struct net *net = sock_net(sk);
1624 	struct sioc_vif_req *vr;
1625 	struct sioc_sg_req *sr;
1626 	struct mr_table *mrt;
1627 
1628 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1629 	if (!mrt)
1630 		return -ENOENT;
1631 
1632 	switch (cmd) {
1633 	case SIOCGETVIFCNT:
1634 		vr = (struct sioc_vif_req *)arg;
1635 		if (vr->vifi >= mrt->maxvif)
1636 			return -EINVAL;
1637 		vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1638 		rcu_read_lock();
1639 		vif = &mrt->vif_table[vr->vifi];
1640 		if (VIF_EXISTS(mrt, vr->vifi)) {
1641 			vr->icount = READ_ONCE(vif->pkt_in);
1642 			vr->ocount = READ_ONCE(vif->pkt_out);
1643 			vr->ibytes = READ_ONCE(vif->bytes_in);
1644 			vr->obytes = READ_ONCE(vif->bytes_out);
1645 			rcu_read_unlock();
1646 
1647 			return 0;
1648 		}
1649 		rcu_read_unlock();
1650 		return -EADDRNOTAVAIL;
1651 	case SIOCGETSGCNT:
1652 		sr = (struct sioc_sg_req *)arg;
1653 
1654 		rcu_read_lock();
1655 		c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1656 		if (c) {
1657 			sr->pktcnt = c->_c.mfc_un.res.pkt;
1658 			sr->bytecnt = c->_c.mfc_un.res.bytes;
1659 			sr->wrong_if = c->_c.mfc_un.res.wrong_if;
1660 			rcu_read_unlock();
1661 			return 0;
1662 		}
1663 		rcu_read_unlock();
1664 		return -EADDRNOTAVAIL;
1665 	default:
1666 		return -ENOIOCTLCMD;
1667 	}
1668 }
1669 
1670 #ifdef CONFIG_COMPAT
1671 struct compat_sioc_sg_req {
1672 	struct in_addr src;
1673 	struct in_addr grp;
1674 	compat_ulong_t pktcnt;
1675 	compat_ulong_t bytecnt;
1676 	compat_ulong_t wrong_if;
1677 };
1678 
1679 struct compat_sioc_vif_req {
1680 	vifi_t	vifi;		/* Which iface */
1681 	compat_ulong_t icount;
1682 	compat_ulong_t ocount;
1683 	compat_ulong_t ibytes;
1684 	compat_ulong_t obytes;
1685 };
1686 
1687 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1688 {
1689 	struct compat_sioc_sg_req sr;
1690 	struct compat_sioc_vif_req vr;
1691 	struct vif_device *vif;
1692 	struct mfc_cache *c;
1693 	struct net *net = sock_net(sk);
1694 	struct mr_table *mrt;
1695 
1696 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1697 	if (!mrt)
1698 		return -ENOENT;
1699 
1700 	switch (cmd) {
1701 	case SIOCGETVIFCNT:
1702 		if (copy_from_user(&vr, arg, sizeof(vr)))
1703 			return -EFAULT;
1704 		if (vr.vifi >= mrt->maxvif)
1705 			return -EINVAL;
1706 		vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1707 		rcu_read_lock();
1708 		vif = &mrt->vif_table[vr.vifi];
1709 		if (VIF_EXISTS(mrt, vr.vifi)) {
1710 			vr.icount = READ_ONCE(vif->pkt_in);
1711 			vr.ocount = READ_ONCE(vif->pkt_out);
1712 			vr.ibytes = READ_ONCE(vif->bytes_in);
1713 			vr.obytes = READ_ONCE(vif->bytes_out);
1714 			rcu_read_unlock();
1715 
1716 			if (copy_to_user(arg, &vr, sizeof(vr)))
1717 				return -EFAULT;
1718 			return 0;
1719 		}
1720 		rcu_read_unlock();
1721 		return -EADDRNOTAVAIL;
1722 	case SIOCGETSGCNT:
1723 		if (copy_from_user(&sr, arg, sizeof(sr)))
1724 			return -EFAULT;
1725 
1726 		rcu_read_lock();
1727 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1728 		if (c) {
1729 			sr.pktcnt = c->_c.mfc_un.res.pkt;
1730 			sr.bytecnt = c->_c.mfc_un.res.bytes;
1731 			sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1732 			rcu_read_unlock();
1733 
1734 			if (copy_to_user(arg, &sr, sizeof(sr)))
1735 				return -EFAULT;
1736 			return 0;
1737 		}
1738 		rcu_read_unlock();
1739 		return -EADDRNOTAVAIL;
1740 	default:
1741 		return -ENOIOCTLCMD;
1742 	}
1743 }
1744 #endif
1745 
1746 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1747 {
1748 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1749 	struct net *net = dev_net(dev);
1750 	struct mr_table *mrt;
1751 	struct vif_device *v;
1752 	int ct;
1753 
1754 	if (event != NETDEV_UNREGISTER)
1755 		return NOTIFY_DONE;
1756 
1757 	ipmr_for_each_table(mrt, net) {
1758 		v = &mrt->vif_table[0];
1759 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1760 			if (rcu_access_pointer(v->dev) == dev)
1761 				vif_delete(mrt, ct, 1, NULL);
1762 		}
1763 	}
1764 	return NOTIFY_DONE;
1765 }
1766 
1767 static struct notifier_block ip_mr_notifier = {
1768 	.notifier_call = ipmr_device_event,
1769 };
1770 
1771 /* Encapsulate a packet by attaching a valid IPIP header to it.
1772  * This avoids tunnel drivers and other mess and gives us the speed so
1773  * important for multicast video.
1774  */
1775 static void ip_encap(struct net *net, struct sk_buff *skb,
1776 		     __be32 saddr, __be32 daddr)
1777 {
1778 	struct iphdr *iph;
1779 	const struct iphdr *old_iph = ip_hdr(skb);
1780 
1781 	skb_push(skb, sizeof(struct iphdr));
1782 	skb->transport_header = skb->network_header;
1783 	skb_reset_network_header(skb);
1784 	iph = ip_hdr(skb);
1785 
1786 	iph->version	=	4;
1787 	iph->tos	=	old_iph->tos;
1788 	iph->ttl	=	old_iph->ttl;
1789 	iph->frag_off	=	0;
1790 	iph->daddr	=	daddr;
1791 	iph->saddr	=	saddr;
1792 	iph->protocol	=	IPPROTO_IPIP;
1793 	iph->ihl	=	5;
1794 	iph->tot_len	=	htons(skb->len);
1795 	ip_select_ident(net, skb, NULL);
1796 	ip_send_check(iph);
1797 
1798 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1799 	nf_reset_ct(skb);
1800 }
1801 
1802 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1803 				      struct sk_buff *skb)
1804 {
1805 	struct ip_options *opt = &(IPCB(skb)->opt);
1806 
1807 	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1808 
1809 	if (unlikely(opt->optlen))
1810 		ip_forward_options(skb);
1811 
1812 	return dst_output(net, sk, skb);
1813 }
1814 
1815 #ifdef CONFIG_NET_SWITCHDEV
1816 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1817 				   int in_vifi, int out_vifi)
1818 {
1819 	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1820 	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1821 
1822 	if (!skb->offload_l3_fwd_mark)
1823 		return false;
1824 	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1825 		return false;
1826 	return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1827 					&in_vif->dev_parent_id);
1828 }
1829 #else
1830 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1831 				   int in_vifi, int out_vifi)
1832 {
1833 	return false;
1834 }
1835 #endif
1836 
1837 /* Processing handlers for ipmr_forward, under rcu_read_lock() */
1838 
1839 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1840 			    int in_vifi, struct sk_buff *skb, int vifi)
1841 {
1842 	const struct iphdr *iph = ip_hdr(skb);
1843 	struct vif_device *vif = &mrt->vif_table[vifi];
1844 	struct net_device *vif_dev;
1845 	struct net_device *dev;
1846 	struct rtable *rt;
1847 	struct flowi4 fl4;
1848 	int    encap = 0;
1849 
1850 	vif_dev = vif_dev_read(vif);
1851 	if (!vif_dev)
1852 		goto out_free;
1853 
1854 	if (vif->flags & VIFF_REGISTER) {
1855 		WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1856 		WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1857 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1858 		DEV_STATS_INC(vif_dev, tx_packets);
1859 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1860 		goto out_free;
1861 	}
1862 
1863 	if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1864 		goto out_free;
1865 
1866 	if (vif->flags & VIFF_TUNNEL) {
1867 		rt = ip_route_output_ports(net, &fl4, NULL,
1868 					   vif->remote, vif->local,
1869 					   0, 0,
1870 					   IPPROTO_IPIP,
1871 					   RT_TOS(iph->tos), vif->link);
1872 		if (IS_ERR(rt))
1873 			goto out_free;
1874 		encap = sizeof(struct iphdr);
1875 	} else {
1876 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1877 					   0, 0,
1878 					   IPPROTO_IPIP,
1879 					   RT_TOS(iph->tos), vif->link);
1880 		if (IS_ERR(rt))
1881 			goto out_free;
1882 	}
1883 
1884 	dev = rt->dst.dev;
1885 
1886 	if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1887 		/* Do not fragment multicasts. Alas, IPv4 does not
1888 		 * allow to send ICMP, so that packets will disappear
1889 		 * to blackhole.
1890 		 */
1891 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1892 		ip_rt_put(rt);
1893 		goto out_free;
1894 	}
1895 
1896 	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1897 
1898 	if (skb_cow(skb, encap)) {
1899 		ip_rt_put(rt);
1900 		goto out_free;
1901 	}
1902 
1903 	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1904 	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1905 
1906 	skb_dst_drop(skb);
1907 	skb_dst_set(skb, &rt->dst);
1908 	ip_decrease_ttl(ip_hdr(skb));
1909 
1910 	/* FIXME: forward and output firewalls used to be called here.
1911 	 * What do we do with netfilter? -- RR
1912 	 */
1913 	if (vif->flags & VIFF_TUNNEL) {
1914 		ip_encap(net, skb, vif->local, vif->remote);
1915 		/* FIXME: extra output firewall step used to be here. --RR */
1916 		DEV_STATS_INC(vif_dev, tx_packets);
1917 		DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1918 	}
1919 
1920 	IPCB(skb)->flags |= IPSKB_FORWARDED;
1921 
1922 	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1923 	 * not only before forwarding, but after forwarding on all output
1924 	 * interfaces. It is clear, if mrouter runs a multicasting
1925 	 * program, it should receive packets not depending to what interface
1926 	 * program is joined.
1927 	 * If we will not make it, the program will have to join on all
1928 	 * interfaces. On the other hand, multihoming host (or router, but
1929 	 * not mrouter) cannot join to more than one interface - it will
1930 	 * result in receiving multiple packets.
1931 	 */
1932 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1933 		net, NULL, skb, skb->dev, dev,
1934 		ipmr_forward_finish);
1935 	return;
1936 
1937 out_free:
1938 	kfree_skb(skb);
1939 }
1940 
1941 /* Called with mrt_lock or rcu_read_lock() */
1942 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1943 {
1944 	int ct;
1945 	/* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1946 	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1947 		if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1948 			break;
1949 	}
1950 	return ct;
1951 }
1952 
1953 /* "local" means that we should preserve one skb (for local delivery) */
1954 /* Called uner rcu_read_lock() */
1955 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1956 			  struct net_device *dev, struct sk_buff *skb,
1957 			  struct mfc_cache *c, int local)
1958 {
1959 	int true_vifi = ipmr_find_vif(mrt, dev);
1960 	int psend = -1;
1961 	int vif, ct;
1962 
1963 	vif = c->_c.mfc_parent;
1964 	c->_c.mfc_un.res.pkt++;
1965 	c->_c.mfc_un.res.bytes += skb->len;
1966 	c->_c.mfc_un.res.lastuse = jiffies;
1967 
1968 	if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1969 		struct mfc_cache *cache_proxy;
1970 
1971 		/* For an (*,G) entry, we only check that the incoming
1972 		 * interface is part of the static tree.
1973 		 */
1974 		cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1975 		if (cache_proxy &&
1976 		    cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1977 			goto forward;
1978 	}
1979 
1980 	/* Wrong interface: drop packet and (maybe) send PIM assert. */
1981 	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
1982 		if (rt_is_output_route(skb_rtable(skb))) {
1983 			/* It is our own packet, looped back.
1984 			 * Very complicated situation...
1985 			 *
1986 			 * The best workaround until routing daemons will be
1987 			 * fixed is not to redistribute packet, if it was
1988 			 * send through wrong interface. It means, that
1989 			 * multicast applications WILL NOT work for
1990 			 * (S,G), which have default multicast route pointing
1991 			 * to wrong oif. In any case, it is not a good
1992 			 * idea to use multicasting applications on router.
1993 			 */
1994 			goto dont_forward;
1995 		}
1996 
1997 		c->_c.mfc_un.res.wrong_if++;
1998 
1999 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
2000 		    /* pimsm uses asserts, when switching from RPT to SPT,
2001 		     * so that we cannot check that packet arrived on an oif.
2002 		     * It is bad, but otherwise we would need to move pretty
2003 		     * large chunk of pimd to kernel. Ough... --ANK
2004 		     */
2005 		    (mrt->mroute_do_pim ||
2006 		     c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2007 		    time_after(jiffies,
2008 			       c->_c.mfc_un.res.last_assert +
2009 			       MFC_ASSERT_THRESH)) {
2010 			c->_c.mfc_un.res.last_assert = jiffies;
2011 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2012 			if (mrt->mroute_do_wrvifwhole)
2013 				ipmr_cache_report(mrt, skb, true_vifi,
2014 						  IGMPMSG_WRVIFWHOLE);
2015 		}
2016 		goto dont_forward;
2017 	}
2018 
2019 forward:
2020 	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2021 		   mrt->vif_table[vif].pkt_in + 1);
2022 	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2023 		   mrt->vif_table[vif].bytes_in + skb->len);
2024 
2025 	/* Forward the frame */
2026 	if (c->mfc_origin == htonl(INADDR_ANY) &&
2027 	    c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2028 		if (true_vifi >= 0 &&
2029 		    true_vifi != c->_c.mfc_parent &&
2030 		    ip_hdr(skb)->ttl >
2031 				c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2032 			/* It's an (*,*) entry and the packet is not coming from
2033 			 * the upstream: forward the packet to the upstream
2034 			 * only.
2035 			 */
2036 			psend = c->_c.mfc_parent;
2037 			goto last_forward;
2038 		}
2039 		goto dont_forward;
2040 	}
2041 	for (ct = c->_c.mfc_un.res.maxvif - 1;
2042 	     ct >= c->_c.mfc_un.res.minvif; ct--) {
2043 		/* For (*,G) entry, don't forward to the incoming interface */
2044 		if ((c->mfc_origin != htonl(INADDR_ANY) ||
2045 		     ct != true_vifi) &&
2046 		    ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2047 			if (psend != -1) {
2048 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2049 
2050 				if (skb2)
2051 					ipmr_queue_xmit(net, mrt, true_vifi,
2052 							skb2, psend);
2053 			}
2054 			psend = ct;
2055 		}
2056 	}
2057 last_forward:
2058 	if (psend != -1) {
2059 		if (local) {
2060 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2061 
2062 			if (skb2)
2063 				ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2064 						psend);
2065 		} else {
2066 			ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2067 			return;
2068 		}
2069 	}
2070 
2071 dont_forward:
2072 	if (!local)
2073 		kfree_skb(skb);
2074 }
2075 
2076 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2077 {
2078 	struct rtable *rt = skb_rtable(skb);
2079 	struct iphdr *iph = ip_hdr(skb);
2080 	struct flowi4 fl4 = {
2081 		.daddr = iph->daddr,
2082 		.saddr = iph->saddr,
2083 		.flowi4_tos = RT_TOS(iph->tos),
2084 		.flowi4_oif = (rt_is_output_route(rt) ?
2085 			       skb->dev->ifindex : 0),
2086 		.flowi4_iif = (rt_is_output_route(rt) ?
2087 			       LOOPBACK_IFINDEX :
2088 			       skb->dev->ifindex),
2089 		.flowi4_mark = skb->mark,
2090 	};
2091 	struct mr_table *mrt;
2092 	int err;
2093 
2094 	err = ipmr_fib_lookup(net, &fl4, &mrt);
2095 	if (err)
2096 		return ERR_PTR(err);
2097 	return mrt;
2098 }
2099 
2100 /* Multicast packets for forwarding arrive here
2101  * Called with rcu_read_lock();
2102  */
2103 int ip_mr_input(struct sk_buff *skb)
2104 {
2105 	struct mfc_cache *cache;
2106 	struct net *net = dev_net(skb->dev);
2107 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2108 	struct mr_table *mrt;
2109 	struct net_device *dev;
2110 
2111 	/* skb->dev passed in is the loX master dev for vrfs.
2112 	 * As there are no vifs associated with loopback devices,
2113 	 * get the proper interface that does have a vif associated with it.
2114 	 */
2115 	dev = skb->dev;
2116 	if (netif_is_l3_master(skb->dev)) {
2117 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2118 		if (!dev) {
2119 			kfree_skb(skb);
2120 			return -ENODEV;
2121 		}
2122 	}
2123 
2124 	/* Packet is looped back after forward, it should not be
2125 	 * forwarded second time, but still can be delivered locally.
2126 	 */
2127 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2128 		goto dont_forward;
2129 
2130 	mrt = ipmr_rt_fib_lookup(net, skb);
2131 	if (IS_ERR(mrt)) {
2132 		kfree_skb(skb);
2133 		return PTR_ERR(mrt);
2134 	}
2135 	if (!local) {
2136 		if (IPCB(skb)->opt.router_alert) {
2137 			if (ip_call_ra_chain(skb))
2138 				return 0;
2139 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2140 			/* IGMPv1 (and broken IGMPv2 implementations sort of
2141 			 * Cisco IOS <= 11.2(8)) do not put router alert
2142 			 * option to IGMP packets destined to routable
2143 			 * groups. It is very bad, because it means
2144 			 * that we can forward NO IGMP messages.
2145 			 */
2146 			struct sock *mroute_sk;
2147 
2148 			mroute_sk = rcu_dereference(mrt->mroute_sk);
2149 			if (mroute_sk) {
2150 				nf_reset_ct(skb);
2151 				raw_rcv(mroute_sk, skb);
2152 				return 0;
2153 			}
2154 		}
2155 	}
2156 
2157 	/* already under rcu_read_lock() */
2158 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2159 	if (!cache) {
2160 		int vif = ipmr_find_vif(mrt, dev);
2161 
2162 		if (vif >= 0)
2163 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2164 						    vif);
2165 	}
2166 
2167 	/* No usable cache entry */
2168 	if (!cache) {
2169 		int vif;
2170 
2171 		if (local) {
2172 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2173 			ip_local_deliver(skb);
2174 			if (!skb2)
2175 				return -ENOBUFS;
2176 			skb = skb2;
2177 		}
2178 
2179 		vif = ipmr_find_vif(mrt, dev);
2180 		if (vif >= 0)
2181 			return ipmr_cache_unresolved(mrt, vif, skb, dev);
2182 		kfree_skb(skb);
2183 		return -ENODEV;
2184 	}
2185 
2186 	ip_mr_forward(net, mrt, dev, skb, cache, local);
2187 
2188 	if (local)
2189 		return ip_local_deliver(skb);
2190 
2191 	return 0;
2192 
2193 dont_forward:
2194 	if (local)
2195 		return ip_local_deliver(skb);
2196 	kfree_skb(skb);
2197 	return 0;
2198 }
2199 
2200 #ifdef CONFIG_IP_PIMSM_V1
2201 /* Handle IGMP messages of PIMv1 */
2202 int pim_rcv_v1(struct sk_buff *skb)
2203 {
2204 	struct igmphdr *pim;
2205 	struct net *net = dev_net(skb->dev);
2206 	struct mr_table *mrt;
2207 
2208 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2209 		goto drop;
2210 
2211 	pim = igmp_hdr(skb);
2212 
2213 	mrt = ipmr_rt_fib_lookup(net, skb);
2214 	if (IS_ERR(mrt))
2215 		goto drop;
2216 	if (!mrt->mroute_do_pim ||
2217 	    pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2218 		goto drop;
2219 
2220 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2221 drop:
2222 		kfree_skb(skb);
2223 	}
2224 	return 0;
2225 }
2226 #endif
2227 
2228 #ifdef CONFIG_IP_PIMSM_V2
2229 static int pim_rcv(struct sk_buff *skb)
2230 {
2231 	struct pimreghdr *pim;
2232 	struct net *net = dev_net(skb->dev);
2233 	struct mr_table *mrt;
2234 
2235 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2236 		goto drop;
2237 
2238 	pim = (struct pimreghdr *)skb_transport_header(skb);
2239 	if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2240 	    (pim->flags & PIM_NULL_REGISTER) ||
2241 	    (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2242 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2243 		goto drop;
2244 
2245 	mrt = ipmr_rt_fib_lookup(net, skb);
2246 	if (IS_ERR(mrt))
2247 		goto drop;
2248 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2249 drop:
2250 		kfree_skb(skb);
2251 	}
2252 	return 0;
2253 }
2254 #endif
2255 
2256 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2257 		   __be32 saddr, __be32 daddr,
2258 		   struct rtmsg *rtm, u32 portid)
2259 {
2260 	struct mfc_cache *cache;
2261 	struct mr_table *mrt;
2262 	int err;
2263 
2264 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2265 	if (!mrt)
2266 		return -ENOENT;
2267 
2268 	rcu_read_lock();
2269 	cache = ipmr_cache_find(mrt, saddr, daddr);
2270 	if (!cache && skb->dev) {
2271 		int vif = ipmr_find_vif(mrt, skb->dev);
2272 
2273 		if (vif >= 0)
2274 			cache = ipmr_cache_find_any(mrt, daddr, vif);
2275 	}
2276 	if (!cache) {
2277 		struct sk_buff *skb2;
2278 		struct iphdr *iph;
2279 		struct net_device *dev;
2280 		int vif = -1;
2281 
2282 		dev = skb->dev;
2283 		if (dev)
2284 			vif = ipmr_find_vif(mrt, dev);
2285 		if (vif < 0) {
2286 			rcu_read_unlock();
2287 			return -ENODEV;
2288 		}
2289 
2290 		skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2291 		if (!skb2) {
2292 			rcu_read_unlock();
2293 			return -ENOMEM;
2294 		}
2295 
2296 		NETLINK_CB(skb2).portid = portid;
2297 		skb_push(skb2, sizeof(struct iphdr));
2298 		skb_reset_network_header(skb2);
2299 		iph = ip_hdr(skb2);
2300 		iph->ihl = sizeof(struct iphdr) >> 2;
2301 		iph->saddr = saddr;
2302 		iph->daddr = daddr;
2303 		iph->version = 0;
2304 		err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2305 		rcu_read_unlock();
2306 		return err;
2307 	}
2308 
2309 	err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2310 	rcu_read_unlock();
2311 	return err;
2312 }
2313 
2314 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2315 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2316 			    int flags)
2317 {
2318 	struct nlmsghdr *nlh;
2319 	struct rtmsg *rtm;
2320 	int err;
2321 
2322 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2323 	if (!nlh)
2324 		return -EMSGSIZE;
2325 
2326 	rtm = nlmsg_data(nlh);
2327 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2328 	rtm->rtm_dst_len  = 32;
2329 	rtm->rtm_src_len  = 32;
2330 	rtm->rtm_tos      = 0;
2331 	rtm->rtm_table    = mrt->id;
2332 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2333 		goto nla_put_failure;
2334 	rtm->rtm_type     = RTN_MULTICAST;
2335 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2336 	if (c->_c.mfc_flags & MFC_STATIC)
2337 		rtm->rtm_protocol = RTPROT_STATIC;
2338 	else
2339 		rtm->rtm_protocol = RTPROT_MROUTED;
2340 	rtm->rtm_flags    = 0;
2341 
2342 	if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2343 	    nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2344 		goto nla_put_failure;
2345 	err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2346 	/* do not break the dump if cache is unresolved */
2347 	if (err < 0 && err != -ENOENT)
2348 		goto nla_put_failure;
2349 
2350 	nlmsg_end(skb, nlh);
2351 	return 0;
2352 
2353 nla_put_failure:
2354 	nlmsg_cancel(skb, nlh);
2355 	return -EMSGSIZE;
2356 }
2357 
2358 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2359 			     u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2360 			     int flags)
2361 {
2362 	return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2363 				cmd, flags);
2364 }
2365 
2366 static size_t mroute_msgsize(bool unresolved, int maxvif)
2367 {
2368 	size_t len =
2369 		NLMSG_ALIGN(sizeof(struct rtmsg))
2370 		+ nla_total_size(4)	/* RTA_TABLE */
2371 		+ nla_total_size(4)	/* RTA_SRC */
2372 		+ nla_total_size(4)	/* RTA_DST */
2373 		;
2374 
2375 	if (!unresolved)
2376 		len = len
2377 		      + nla_total_size(4)	/* RTA_IIF */
2378 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2379 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2380 						/* RTA_MFC_STATS */
2381 		      + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2382 		;
2383 
2384 	return len;
2385 }
2386 
2387 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2388 				 int cmd)
2389 {
2390 	struct net *net = read_pnet(&mrt->net);
2391 	struct sk_buff *skb;
2392 	int err = -ENOBUFS;
2393 
2394 	skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2395 				       mrt->maxvif),
2396 			GFP_ATOMIC);
2397 	if (!skb)
2398 		goto errout;
2399 
2400 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2401 	if (err < 0)
2402 		goto errout;
2403 
2404 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2405 	return;
2406 
2407 errout:
2408 	kfree_skb(skb);
2409 	if (err < 0)
2410 		rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2411 }
2412 
2413 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2414 {
2415 	size_t len =
2416 		NLMSG_ALIGN(sizeof(struct rtgenmsg))
2417 		+ nla_total_size(1)	/* IPMRA_CREPORT_MSGTYPE */
2418 		+ nla_total_size(4)	/* IPMRA_CREPORT_VIF_ID */
2419 		+ nla_total_size(4)	/* IPMRA_CREPORT_SRC_ADDR */
2420 		+ nla_total_size(4)	/* IPMRA_CREPORT_DST_ADDR */
2421 		+ nla_total_size(4)	/* IPMRA_CREPORT_TABLE */
2422 					/* IPMRA_CREPORT_PKT */
2423 		+ nla_total_size(payloadlen)
2424 		;
2425 
2426 	return len;
2427 }
2428 
2429 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2430 {
2431 	struct net *net = read_pnet(&mrt->net);
2432 	struct nlmsghdr *nlh;
2433 	struct rtgenmsg *rtgenm;
2434 	struct igmpmsg *msg;
2435 	struct sk_buff *skb;
2436 	struct nlattr *nla;
2437 	int payloadlen;
2438 
2439 	payloadlen = pkt->len - sizeof(struct igmpmsg);
2440 	msg = (struct igmpmsg *)skb_network_header(pkt);
2441 
2442 	skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2443 	if (!skb)
2444 		goto errout;
2445 
2446 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2447 			sizeof(struct rtgenmsg), 0);
2448 	if (!nlh)
2449 		goto errout;
2450 	rtgenm = nlmsg_data(nlh);
2451 	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2452 	if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2453 	    nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2454 	    nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2455 			    msg->im_src.s_addr) ||
2456 	    nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2457 			    msg->im_dst.s_addr) ||
2458 	    nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2459 		goto nla_put_failure;
2460 
2461 	nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2462 	if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2463 				  nla_data(nla), payloadlen))
2464 		goto nla_put_failure;
2465 
2466 	nlmsg_end(skb, nlh);
2467 
2468 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2469 	return;
2470 
2471 nla_put_failure:
2472 	nlmsg_cancel(skb, nlh);
2473 errout:
2474 	kfree_skb(skb);
2475 	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2476 }
2477 
2478 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2479 				       const struct nlmsghdr *nlh,
2480 				       struct nlattr **tb,
2481 				       struct netlink_ext_ack *extack)
2482 {
2483 	struct rtmsg *rtm;
2484 	int i, err;
2485 
2486 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2487 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2488 		return -EINVAL;
2489 	}
2490 
2491 	if (!netlink_strict_get_check(skb))
2492 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2493 					      rtm_ipv4_policy, extack);
2494 
2495 	rtm = nlmsg_data(nlh);
2496 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2497 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2498 	    rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2499 	    rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2500 		NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2501 		return -EINVAL;
2502 	}
2503 
2504 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2505 					    rtm_ipv4_policy, extack);
2506 	if (err)
2507 		return err;
2508 
2509 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2510 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2511 		NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2512 		return -EINVAL;
2513 	}
2514 
2515 	for (i = 0; i <= RTA_MAX; i++) {
2516 		if (!tb[i])
2517 			continue;
2518 
2519 		switch (i) {
2520 		case RTA_SRC:
2521 		case RTA_DST:
2522 		case RTA_TABLE:
2523 			break;
2524 		default:
2525 			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2526 			return -EINVAL;
2527 		}
2528 	}
2529 
2530 	return 0;
2531 }
2532 
2533 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2534 			     struct netlink_ext_ack *extack)
2535 {
2536 	struct net *net = sock_net(in_skb->sk);
2537 	struct nlattr *tb[RTA_MAX + 1];
2538 	struct sk_buff *skb = NULL;
2539 	struct mfc_cache *cache;
2540 	struct mr_table *mrt;
2541 	__be32 src, grp;
2542 	u32 tableid;
2543 	int err;
2544 
2545 	err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2546 	if (err < 0)
2547 		goto errout;
2548 
2549 	src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2550 	grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2551 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2552 
2553 	mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2554 	if (!mrt) {
2555 		err = -ENOENT;
2556 		goto errout_free;
2557 	}
2558 
2559 	/* entries are added/deleted only under RTNL */
2560 	rcu_read_lock();
2561 	cache = ipmr_cache_find(mrt, src, grp);
2562 	rcu_read_unlock();
2563 	if (!cache) {
2564 		err = -ENOENT;
2565 		goto errout_free;
2566 	}
2567 
2568 	skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2569 	if (!skb) {
2570 		err = -ENOBUFS;
2571 		goto errout_free;
2572 	}
2573 
2574 	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2575 			       nlh->nlmsg_seq, cache,
2576 			       RTM_NEWROUTE, 0);
2577 	if (err < 0)
2578 		goto errout_free;
2579 
2580 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2581 
2582 errout:
2583 	return err;
2584 
2585 errout_free:
2586 	kfree_skb(skb);
2587 	goto errout;
2588 }
2589 
2590 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2591 {
2592 	struct fib_dump_filter filter = {};
2593 	int err;
2594 
2595 	if (cb->strict_check) {
2596 		err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2597 					    &filter, cb);
2598 		if (err < 0)
2599 			return err;
2600 	}
2601 
2602 	if (filter.table_id) {
2603 		struct mr_table *mrt;
2604 
2605 		mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2606 		if (!mrt) {
2607 			if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2608 				return skb->len;
2609 
2610 			NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2611 			return -ENOENT;
2612 		}
2613 		err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2614 				    &mfc_unres_lock, &filter);
2615 		return skb->len ? : err;
2616 	}
2617 
2618 	return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2619 				_ipmr_fill_mroute, &mfc_unres_lock, &filter);
2620 }
2621 
2622 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2623 	[RTA_SRC]	= { .type = NLA_U32 },
2624 	[RTA_DST]	= { .type = NLA_U32 },
2625 	[RTA_IIF]	= { .type = NLA_U32 },
2626 	[RTA_TABLE]	= { .type = NLA_U32 },
2627 	[RTA_MULTIPATH]	= { .len = sizeof(struct rtnexthop) },
2628 };
2629 
2630 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2631 {
2632 	switch (rtm_protocol) {
2633 	case RTPROT_STATIC:
2634 	case RTPROT_MROUTED:
2635 		return true;
2636 	}
2637 	return false;
2638 }
2639 
2640 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2641 {
2642 	struct rtnexthop *rtnh = nla_data(nla);
2643 	int remaining = nla_len(nla), vifi = 0;
2644 
2645 	while (rtnh_ok(rtnh, remaining)) {
2646 		mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2647 		if (++vifi == MAXVIFS)
2648 			break;
2649 		rtnh = rtnh_next(rtnh, &remaining);
2650 	}
2651 
2652 	return remaining > 0 ? -EINVAL : vifi;
2653 }
2654 
2655 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2656 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2657 			    struct mfcctl *mfcc, int *mrtsock,
2658 			    struct mr_table **mrtret,
2659 			    struct netlink_ext_ack *extack)
2660 {
2661 	struct net_device *dev = NULL;
2662 	u32 tblid = RT_TABLE_DEFAULT;
2663 	struct mr_table *mrt;
2664 	struct nlattr *attr;
2665 	struct rtmsg *rtm;
2666 	int ret, rem;
2667 
2668 	ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2669 					rtm_ipmr_policy, extack);
2670 	if (ret < 0)
2671 		goto out;
2672 	rtm = nlmsg_data(nlh);
2673 
2674 	ret = -EINVAL;
2675 	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2676 	    rtm->rtm_type != RTN_MULTICAST ||
2677 	    rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2678 	    !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2679 		goto out;
2680 
2681 	memset(mfcc, 0, sizeof(*mfcc));
2682 	mfcc->mfcc_parent = -1;
2683 	ret = 0;
2684 	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2685 		switch (nla_type(attr)) {
2686 		case RTA_SRC:
2687 			mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2688 			break;
2689 		case RTA_DST:
2690 			mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2691 			break;
2692 		case RTA_IIF:
2693 			dev = __dev_get_by_index(net, nla_get_u32(attr));
2694 			if (!dev) {
2695 				ret = -ENODEV;
2696 				goto out;
2697 			}
2698 			break;
2699 		case RTA_MULTIPATH:
2700 			if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2701 				ret = -EINVAL;
2702 				goto out;
2703 			}
2704 			break;
2705 		case RTA_PREFSRC:
2706 			ret = 1;
2707 			break;
2708 		case RTA_TABLE:
2709 			tblid = nla_get_u32(attr);
2710 			break;
2711 		}
2712 	}
2713 	mrt = ipmr_get_table(net, tblid);
2714 	if (!mrt) {
2715 		ret = -ENOENT;
2716 		goto out;
2717 	}
2718 	*mrtret = mrt;
2719 	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2720 	if (dev)
2721 		mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2722 
2723 out:
2724 	return ret;
2725 }
2726 
2727 /* takes care of both newroute and delroute */
2728 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2729 			  struct netlink_ext_ack *extack)
2730 {
2731 	struct net *net = sock_net(skb->sk);
2732 	int ret, mrtsock, parent;
2733 	struct mr_table *tbl;
2734 	struct mfcctl mfcc;
2735 
2736 	mrtsock = 0;
2737 	tbl = NULL;
2738 	ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2739 	if (ret < 0)
2740 		return ret;
2741 
2742 	parent = ret ? mfcc.mfcc_parent : -1;
2743 	if (nlh->nlmsg_type == RTM_NEWROUTE)
2744 		return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2745 	else
2746 		return ipmr_mfc_delete(tbl, &mfcc, parent);
2747 }
2748 
2749 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2750 {
2751 	u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2752 
2753 	if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2754 	    nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2755 	    nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2756 			mrt->mroute_reg_vif_num) ||
2757 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2758 		       mrt->mroute_do_assert) ||
2759 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2760 	    nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2761 		       mrt->mroute_do_wrvifwhole))
2762 		return false;
2763 
2764 	return true;
2765 }
2766 
2767 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2768 {
2769 	struct net_device *vif_dev;
2770 	struct nlattr *vif_nest;
2771 	struct vif_device *vif;
2772 
2773 	vif = &mrt->vif_table[vifid];
2774 	vif_dev = rtnl_dereference(vif->dev);
2775 	/* if the VIF doesn't exist just continue */
2776 	if (!vif_dev)
2777 		return true;
2778 
2779 	vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2780 	if (!vif_nest)
2781 		return false;
2782 
2783 	if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2784 	    nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2785 	    nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2786 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2787 			      IPMRA_VIFA_PAD) ||
2788 	    nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2789 			      IPMRA_VIFA_PAD) ||
2790 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2791 			      IPMRA_VIFA_PAD) ||
2792 	    nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2793 			      IPMRA_VIFA_PAD) ||
2794 	    nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2795 	    nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2796 		nla_nest_cancel(skb, vif_nest);
2797 		return false;
2798 	}
2799 	nla_nest_end(skb, vif_nest);
2800 
2801 	return true;
2802 }
2803 
2804 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2805 			       struct netlink_ext_ack *extack)
2806 {
2807 	struct ifinfomsg *ifm;
2808 
2809 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2810 		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2811 		return -EINVAL;
2812 	}
2813 
2814 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2815 		NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2816 		return -EINVAL;
2817 	}
2818 
2819 	ifm = nlmsg_data(nlh);
2820 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2821 	    ifm->ifi_change || ifm->ifi_index) {
2822 		NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2823 		return -EINVAL;
2824 	}
2825 
2826 	return 0;
2827 }
2828 
2829 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2830 {
2831 	struct net *net = sock_net(skb->sk);
2832 	struct nlmsghdr *nlh = NULL;
2833 	unsigned int t = 0, s_t;
2834 	unsigned int e = 0, s_e;
2835 	struct mr_table *mrt;
2836 
2837 	if (cb->strict_check) {
2838 		int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2839 
2840 		if (err < 0)
2841 			return err;
2842 	}
2843 
2844 	s_t = cb->args[0];
2845 	s_e = cb->args[1];
2846 
2847 	ipmr_for_each_table(mrt, net) {
2848 		struct nlattr *vifs, *af;
2849 		struct ifinfomsg *hdr;
2850 		u32 i;
2851 
2852 		if (t < s_t)
2853 			goto skip_table;
2854 		nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2855 				cb->nlh->nlmsg_seq, RTM_NEWLINK,
2856 				sizeof(*hdr), NLM_F_MULTI);
2857 		if (!nlh)
2858 			break;
2859 
2860 		hdr = nlmsg_data(nlh);
2861 		memset(hdr, 0, sizeof(*hdr));
2862 		hdr->ifi_family = RTNL_FAMILY_IPMR;
2863 
2864 		af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2865 		if (!af) {
2866 			nlmsg_cancel(skb, nlh);
2867 			goto out;
2868 		}
2869 
2870 		if (!ipmr_fill_table(mrt, skb)) {
2871 			nlmsg_cancel(skb, nlh);
2872 			goto out;
2873 		}
2874 
2875 		vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2876 		if (!vifs) {
2877 			nla_nest_end(skb, af);
2878 			nlmsg_end(skb, nlh);
2879 			goto out;
2880 		}
2881 		for (i = 0; i < mrt->maxvif; i++) {
2882 			if (e < s_e)
2883 				goto skip_entry;
2884 			if (!ipmr_fill_vif(mrt, i, skb)) {
2885 				nla_nest_end(skb, vifs);
2886 				nla_nest_end(skb, af);
2887 				nlmsg_end(skb, nlh);
2888 				goto out;
2889 			}
2890 skip_entry:
2891 			e++;
2892 		}
2893 		s_e = 0;
2894 		e = 0;
2895 		nla_nest_end(skb, vifs);
2896 		nla_nest_end(skb, af);
2897 		nlmsg_end(skb, nlh);
2898 skip_table:
2899 		t++;
2900 	}
2901 
2902 out:
2903 	cb->args[1] = e;
2904 	cb->args[0] = t;
2905 
2906 	return skb->len;
2907 }
2908 
2909 #ifdef CONFIG_PROC_FS
2910 /* The /proc interfaces to multicast routing :
2911  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2912  */
2913 
2914 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2915 	__acquires(RCU)
2916 {
2917 	struct mr_vif_iter *iter = seq->private;
2918 	struct net *net = seq_file_net(seq);
2919 	struct mr_table *mrt;
2920 
2921 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2922 	if (!mrt)
2923 		return ERR_PTR(-ENOENT);
2924 
2925 	iter->mrt = mrt;
2926 
2927 	rcu_read_lock();
2928 	return mr_vif_seq_start(seq, pos);
2929 }
2930 
2931 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2932 	__releases(RCU)
2933 {
2934 	rcu_read_unlock();
2935 }
2936 
2937 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2938 {
2939 	struct mr_vif_iter *iter = seq->private;
2940 	struct mr_table *mrt = iter->mrt;
2941 
2942 	if (v == SEQ_START_TOKEN) {
2943 		seq_puts(seq,
2944 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2945 	} else {
2946 		const struct vif_device *vif = v;
2947 		const struct net_device *vif_dev;
2948 		const char *name;
2949 
2950 		vif_dev = vif_dev_read(vif);
2951 		name = vif_dev ? vif_dev->name : "none";
2952 		seq_printf(seq,
2953 			   "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2954 			   vif - mrt->vif_table,
2955 			   name, vif->bytes_in, vif->pkt_in,
2956 			   vif->bytes_out, vif->pkt_out,
2957 			   vif->flags, vif->local, vif->remote);
2958 	}
2959 	return 0;
2960 }
2961 
2962 static const struct seq_operations ipmr_vif_seq_ops = {
2963 	.start = ipmr_vif_seq_start,
2964 	.next  = mr_vif_seq_next,
2965 	.stop  = ipmr_vif_seq_stop,
2966 	.show  = ipmr_vif_seq_show,
2967 };
2968 
2969 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2970 {
2971 	struct net *net = seq_file_net(seq);
2972 	struct mr_table *mrt;
2973 
2974 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2975 	if (!mrt)
2976 		return ERR_PTR(-ENOENT);
2977 
2978 	return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2979 }
2980 
2981 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2982 {
2983 	int n;
2984 
2985 	if (v == SEQ_START_TOKEN) {
2986 		seq_puts(seq,
2987 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2988 	} else {
2989 		const struct mfc_cache *mfc = v;
2990 		const struct mr_mfc_iter *it = seq->private;
2991 		const struct mr_table *mrt = it->mrt;
2992 
2993 		seq_printf(seq, "%08X %08X %-3hd",
2994 			   (__force u32) mfc->mfc_mcastgrp,
2995 			   (__force u32) mfc->mfc_origin,
2996 			   mfc->_c.mfc_parent);
2997 
2998 		if (it->cache != &mrt->mfc_unres_queue) {
2999 			seq_printf(seq, " %8lu %8lu %8lu",
3000 				   mfc->_c.mfc_un.res.pkt,
3001 				   mfc->_c.mfc_un.res.bytes,
3002 				   mfc->_c.mfc_un.res.wrong_if);
3003 			for (n = mfc->_c.mfc_un.res.minvif;
3004 			     n < mfc->_c.mfc_un.res.maxvif; n++) {
3005 				if (VIF_EXISTS(mrt, n) &&
3006 				    mfc->_c.mfc_un.res.ttls[n] < 255)
3007 					seq_printf(seq,
3008 					   " %2d:%-3d",
3009 					   n, mfc->_c.mfc_un.res.ttls[n]);
3010 			}
3011 		} else {
3012 			/* unresolved mfc_caches don't contain
3013 			 * pkt, bytes and wrong_if values
3014 			 */
3015 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3016 		}
3017 		seq_putc(seq, '\n');
3018 	}
3019 	return 0;
3020 }
3021 
3022 static const struct seq_operations ipmr_mfc_seq_ops = {
3023 	.start = ipmr_mfc_seq_start,
3024 	.next  = mr_mfc_seq_next,
3025 	.stop  = mr_mfc_seq_stop,
3026 	.show  = ipmr_mfc_seq_show,
3027 };
3028 #endif
3029 
3030 #ifdef CONFIG_IP_PIMSM_V2
3031 static const struct net_protocol pim_protocol = {
3032 	.handler	=	pim_rcv,
3033 };
3034 #endif
3035 
3036 static unsigned int ipmr_seq_read(struct net *net)
3037 {
3038 	ASSERT_RTNL();
3039 
3040 	return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3041 }
3042 
3043 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3044 		     struct netlink_ext_ack *extack)
3045 {
3046 	return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3047 		       ipmr_mr_table_iter, extack);
3048 }
3049 
3050 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3051 	.family		= RTNL_FAMILY_IPMR,
3052 	.fib_seq_read	= ipmr_seq_read,
3053 	.fib_dump	= ipmr_dump,
3054 	.owner		= THIS_MODULE,
3055 };
3056 
3057 static int __net_init ipmr_notifier_init(struct net *net)
3058 {
3059 	struct fib_notifier_ops *ops;
3060 
3061 	net->ipv4.ipmr_seq = 0;
3062 
3063 	ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3064 	if (IS_ERR(ops))
3065 		return PTR_ERR(ops);
3066 	net->ipv4.ipmr_notifier_ops = ops;
3067 
3068 	return 0;
3069 }
3070 
3071 static void __net_exit ipmr_notifier_exit(struct net *net)
3072 {
3073 	fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3074 	net->ipv4.ipmr_notifier_ops = NULL;
3075 }
3076 
3077 /* Setup for IP multicast routing */
3078 static int __net_init ipmr_net_init(struct net *net)
3079 {
3080 	int err;
3081 
3082 	err = ipmr_notifier_init(net);
3083 	if (err)
3084 		goto ipmr_notifier_fail;
3085 
3086 	err = ipmr_rules_init(net);
3087 	if (err < 0)
3088 		goto ipmr_rules_fail;
3089 
3090 #ifdef CONFIG_PROC_FS
3091 	err = -ENOMEM;
3092 	if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3093 			sizeof(struct mr_vif_iter)))
3094 		goto proc_vif_fail;
3095 	if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3096 			sizeof(struct mr_mfc_iter)))
3097 		goto proc_cache_fail;
3098 #endif
3099 	return 0;
3100 
3101 #ifdef CONFIG_PROC_FS
3102 proc_cache_fail:
3103 	remove_proc_entry("ip_mr_vif", net->proc_net);
3104 proc_vif_fail:
3105 	rtnl_lock();
3106 	ipmr_rules_exit(net);
3107 	rtnl_unlock();
3108 #endif
3109 ipmr_rules_fail:
3110 	ipmr_notifier_exit(net);
3111 ipmr_notifier_fail:
3112 	return err;
3113 }
3114 
3115 static void __net_exit ipmr_net_exit(struct net *net)
3116 {
3117 #ifdef CONFIG_PROC_FS
3118 	remove_proc_entry("ip_mr_cache", net->proc_net);
3119 	remove_proc_entry("ip_mr_vif", net->proc_net);
3120 #endif
3121 	ipmr_notifier_exit(net);
3122 }
3123 
3124 static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3125 {
3126 	struct net *net;
3127 
3128 	rtnl_lock();
3129 	list_for_each_entry(net, net_list, exit_list)
3130 		ipmr_rules_exit(net);
3131 	rtnl_unlock();
3132 }
3133 
3134 static struct pernet_operations ipmr_net_ops = {
3135 	.init = ipmr_net_init,
3136 	.exit = ipmr_net_exit,
3137 	.exit_batch = ipmr_net_exit_batch,
3138 };
3139 
3140 int __init ip_mr_init(void)
3141 {
3142 	int err;
3143 
3144 	mrt_cachep = kmem_cache_create("ip_mrt_cache",
3145 				       sizeof(struct mfc_cache),
3146 				       0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3147 				       NULL);
3148 
3149 	err = register_pernet_subsys(&ipmr_net_ops);
3150 	if (err)
3151 		goto reg_pernet_fail;
3152 
3153 	err = register_netdevice_notifier(&ip_mr_notifier);
3154 	if (err)
3155 		goto reg_notif_fail;
3156 #ifdef CONFIG_IP_PIMSM_V2
3157 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3158 		pr_err("%s: can't add PIM protocol\n", __func__);
3159 		err = -EAGAIN;
3160 		goto add_proto_fail;
3161 	}
3162 #endif
3163 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3164 		      ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3165 	rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3166 		      ipmr_rtm_route, NULL, 0);
3167 	rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3168 		      ipmr_rtm_route, NULL, 0);
3169 
3170 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3171 		      NULL, ipmr_rtm_dumplink, 0);
3172 	return 0;
3173 
3174 #ifdef CONFIG_IP_PIMSM_V2
3175 add_proto_fail:
3176 	unregister_netdevice_notifier(&ip_mr_notifier);
3177 #endif
3178 reg_notif_fail:
3179 	unregister_pernet_subsys(&ipmr_net_ops);
3180 reg_pernet_fail:
3181 	kmem_cache_destroy(mrt_cachep);
3182 	return err;
3183 }
3184