xref: /openbmc/linux/net/ipv4/ipmr.c (revision 37be287c)
1 /*
2  *	IP multicast routing support for mrouted 3.6/3.8
3  *
4  *		(c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5  *	  Linux Consultancy and Custom Driver Development
6  *
7  *	This program is free software; you can redistribute it and/or
8  *	modify it under the terms of the GNU General Public License
9  *	as published by the Free Software Foundation; either version
10  *	2 of the License, or (at your option) any later version.
11  *
12  *	Fixes:
13  *	Michael Chastain	:	Incorrect size of copying.
14  *	Alan Cox		:	Added the cache manager code
15  *	Alan Cox		:	Fixed the clone/copy bug and device race.
16  *	Mike McLagan		:	Routing by source
17  *	Malcolm Beattie		:	Buffer handling fixes.
18  *	Alexey Kuznetsov	:	Double buffer free and other fixes.
19  *	SVR Anand		:	Fixed several multicast bugs and problems.
20  *	Alexey Kuznetsov	:	Status, optimisations and more.
21  *	Brad Parker		:	Better behaviour on mrouted upcall
22  *					overflow.
23  *      Carlos Picoto           :       PIMv1 Support
24  *	Pavlin Ivanov Radoslavov:	PIMv2 Registers must checksum only PIM header
25  *					Relax this requirement to work with older peers.
26  *
27  */
28 
29 #include <asm/uaccess.h>
30 #include <linux/types.h>
31 #include <linux/capability.h>
32 #include <linux/errno.h>
33 #include <linux/timer.h>
34 #include <linux/mm.h>
35 #include <linux/kernel.h>
36 #include <linux/fcntl.h>
37 #include <linux/stat.h>
38 #include <linux/socket.h>
39 #include <linux/in.h>
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/inetdevice.h>
43 #include <linux/igmp.h>
44 #include <linux/proc_fs.h>
45 #include <linux/seq_file.h>
46 #include <linux/mroute.h>
47 #include <linux/init.h>
48 #include <linux/if_ether.h>
49 #include <linux/slab.h>
50 #include <net/net_namespace.h>
51 #include <net/ip.h>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.h>
55 #include <net/sock.h>
56 #include <net/icmp.h>
57 #include <net/udp.h>
58 #include <net/raw.h>
59 #include <linux/notifier.h>
60 #include <linux/if_arp.h>
61 #include <linux/netfilter_ipv4.h>
62 #include <linux/compat.h>
63 #include <linux/export.h>
64 #include <net/ip_tunnels.h>
65 #include <net/checksum.h>
66 #include <net/netlink.h>
67 #include <net/fib_rules.h>
68 #include <linux/netconf.h>
69 
70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
71 #define CONFIG_IP_PIMSM	1
72 #endif
73 
74 struct mr_table {
75 	struct list_head	list;
76 #ifdef CONFIG_NET_NS
77 	struct net		*net;
78 #endif
79 	u32			id;
80 	struct sock __rcu	*mroute_sk;
81 	struct timer_list	ipmr_expire_timer;
82 	struct list_head	mfc_unres_queue;
83 	struct list_head	mfc_cache_array[MFC_LINES];
84 	struct vif_device	vif_table[MAXVIFS];
85 	int			maxvif;
86 	atomic_t		cache_resolve_queue_len;
87 	bool			mroute_do_assert;
88 	bool			mroute_do_pim;
89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
90 	int			mroute_reg_vif_num;
91 #endif
92 };
93 
94 struct ipmr_rule {
95 	struct fib_rule		common;
96 };
97 
98 struct ipmr_result {
99 	struct mr_table		*mrt;
100 };
101 
102 /* Big lock, protecting vif table, mrt cache and mroute socket state.
103  * Note that the changes are semaphored via rtnl_lock.
104  */
105 
106 static DEFINE_RWLOCK(mrt_lock);
107 
108 /*
109  *	Multicast router control variables
110  */
111 
112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
113 
114 /* Special spinlock for queue of unresolved entries */
115 static DEFINE_SPINLOCK(mfc_unres_lock);
116 
117 /* We return to original Alan's scheme. Hash table of resolved
118  * entries is changed only in process context and protected
119  * with weak lock mrt_lock. Queue of unresolved entries is protected
120  * with strong spinlock mfc_unres_lock.
121  *
122  * In this case data path is free of exclusive locks at all.
123  */
124 
125 static struct kmem_cache *mrt_cachep __read_mostly;
126 
127 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
128 static void ipmr_free_table(struct mr_table *mrt);
129 
130 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
131 			  struct sk_buff *skb, struct mfc_cache *cache,
132 			  int local);
133 static int ipmr_cache_report(struct mr_table *mrt,
134 			     struct sk_buff *pkt, vifi_t vifi, int assert);
135 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
136 			      struct mfc_cache *c, struct rtmsg *rtm);
137 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
138 				 int cmd);
139 static void mroute_clean_tables(struct mr_table *mrt);
140 static void ipmr_expire_process(unsigned long arg);
141 
142 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
143 #define ipmr_for_each_table(mrt, net) \
144 	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
145 
146 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
147 {
148 	struct mr_table *mrt;
149 
150 	ipmr_for_each_table(mrt, net) {
151 		if (mrt->id == id)
152 			return mrt;
153 	}
154 	return NULL;
155 }
156 
157 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
158 			   struct mr_table **mrt)
159 {
160 	int err;
161 	struct ipmr_result res;
162 	struct fib_lookup_arg arg = {
163 		.result = &res,
164 		.flags = FIB_LOOKUP_NOREF,
165 	};
166 
167 	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
168 			       flowi4_to_flowi(flp4), 0, &arg);
169 	if (err < 0)
170 		return err;
171 	*mrt = res.mrt;
172 	return 0;
173 }
174 
175 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
176 			    int flags, struct fib_lookup_arg *arg)
177 {
178 	struct ipmr_result *res = arg->result;
179 	struct mr_table *mrt;
180 
181 	switch (rule->action) {
182 	case FR_ACT_TO_TBL:
183 		break;
184 	case FR_ACT_UNREACHABLE:
185 		return -ENETUNREACH;
186 	case FR_ACT_PROHIBIT:
187 		return -EACCES;
188 	case FR_ACT_BLACKHOLE:
189 	default:
190 		return -EINVAL;
191 	}
192 
193 	mrt = ipmr_get_table(rule->fr_net, rule->table);
194 	if (mrt == NULL)
195 		return -EAGAIN;
196 	res->mrt = mrt;
197 	return 0;
198 }
199 
200 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
201 {
202 	return 1;
203 }
204 
205 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
206 	FRA_GENERIC_POLICY,
207 };
208 
209 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
210 			       struct fib_rule_hdr *frh, struct nlattr **tb)
211 {
212 	return 0;
213 }
214 
215 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
216 			     struct nlattr **tb)
217 {
218 	return 1;
219 }
220 
221 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
222 			  struct fib_rule_hdr *frh)
223 {
224 	frh->dst_len = 0;
225 	frh->src_len = 0;
226 	frh->tos     = 0;
227 	return 0;
228 }
229 
230 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
231 	.family		= RTNL_FAMILY_IPMR,
232 	.rule_size	= sizeof(struct ipmr_rule),
233 	.addr_size	= sizeof(u32),
234 	.action		= ipmr_rule_action,
235 	.match		= ipmr_rule_match,
236 	.configure	= ipmr_rule_configure,
237 	.compare	= ipmr_rule_compare,
238 	.default_pref	= fib_default_rule_pref,
239 	.fill		= ipmr_rule_fill,
240 	.nlgroup	= RTNLGRP_IPV4_RULE,
241 	.policy		= ipmr_rule_policy,
242 	.owner		= THIS_MODULE,
243 };
244 
245 static int __net_init ipmr_rules_init(struct net *net)
246 {
247 	struct fib_rules_ops *ops;
248 	struct mr_table *mrt;
249 	int err;
250 
251 	ops = fib_rules_register(&ipmr_rules_ops_template, net);
252 	if (IS_ERR(ops))
253 		return PTR_ERR(ops);
254 
255 	INIT_LIST_HEAD(&net->ipv4.mr_tables);
256 
257 	mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
258 	if (mrt == NULL) {
259 		err = -ENOMEM;
260 		goto err1;
261 	}
262 
263 	err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
264 	if (err < 0)
265 		goto err2;
266 
267 	net->ipv4.mr_rules_ops = ops;
268 	return 0;
269 
270 err2:
271 	kfree(mrt);
272 err1:
273 	fib_rules_unregister(ops);
274 	return err;
275 }
276 
277 static void __net_exit ipmr_rules_exit(struct net *net)
278 {
279 	struct mr_table *mrt, *next;
280 
281 	list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
282 		list_del(&mrt->list);
283 		ipmr_free_table(mrt);
284 	}
285 	fib_rules_unregister(net->ipv4.mr_rules_ops);
286 }
287 #else
288 #define ipmr_for_each_table(mrt, net) \
289 	for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
290 
291 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
292 {
293 	return net->ipv4.mrt;
294 }
295 
296 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
297 			   struct mr_table **mrt)
298 {
299 	*mrt = net->ipv4.mrt;
300 	return 0;
301 }
302 
303 static int __net_init ipmr_rules_init(struct net *net)
304 {
305 	net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
306 	return net->ipv4.mrt ? 0 : -ENOMEM;
307 }
308 
309 static void __net_exit ipmr_rules_exit(struct net *net)
310 {
311 	ipmr_free_table(net->ipv4.mrt);
312 }
313 #endif
314 
315 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
316 {
317 	struct mr_table *mrt;
318 	unsigned int i;
319 
320 	mrt = ipmr_get_table(net, id);
321 	if (mrt != NULL)
322 		return mrt;
323 
324 	mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
325 	if (mrt == NULL)
326 		return NULL;
327 	write_pnet(&mrt->net, net);
328 	mrt->id = id;
329 
330 	/* Forwarding cache */
331 	for (i = 0; i < MFC_LINES; i++)
332 		INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
333 
334 	INIT_LIST_HEAD(&mrt->mfc_unres_queue);
335 
336 	setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
337 		    (unsigned long)mrt);
338 
339 #ifdef CONFIG_IP_PIMSM
340 	mrt->mroute_reg_vif_num = -1;
341 #endif
342 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
343 	list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
344 #endif
345 	return mrt;
346 }
347 
348 static void ipmr_free_table(struct mr_table *mrt)
349 {
350 	del_timer_sync(&mrt->ipmr_expire_timer);
351 	mroute_clean_tables(mrt);
352 	kfree(mrt);
353 }
354 
355 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
356 
357 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
358 {
359 	struct net *net = dev_net(dev);
360 
361 	dev_close(dev);
362 
363 	dev = __dev_get_by_name(net, "tunl0");
364 	if (dev) {
365 		const struct net_device_ops *ops = dev->netdev_ops;
366 		struct ifreq ifr;
367 		struct ip_tunnel_parm p;
368 
369 		memset(&p, 0, sizeof(p));
370 		p.iph.daddr = v->vifc_rmt_addr.s_addr;
371 		p.iph.saddr = v->vifc_lcl_addr.s_addr;
372 		p.iph.version = 4;
373 		p.iph.ihl = 5;
374 		p.iph.protocol = IPPROTO_IPIP;
375 		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
376 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
377 
378 		if (ops->ndo_do_ioctl) {
379 			mm_segment_t oldfs = get_fs();
380 
381 			set_fs(KERNEL_DS);
382 			ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
383 			set_fs(oldfs);
384 		}
385 	}
386 }
387 
388 static
389 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
390 {
391 	struct net_device  *dev;
392 
393 	dev = __dev_get_by_name(net, "tunl0");
394 
395 	if (dev) {
396 		const struct net_device_ops *ops = dev->netdev_ops;
397 		int err;
398 		struct ifreq ifr;
399 		struct ip_tunnel_parm p;
400 		struct in_device  *in_dev;
401 
402 		memset(&p, 0, sizeof(p));
403 		p.iph.daddr = v->vifc_rmt_addr.s_addr;
404 		p.iph.saddr = v->vifc_lcl_addr.s_addr;
405 		p.iph.version = 4;
406 		p.iph.ihl = 5;
407 		p.iph.protocol = IPPROTO_IPIP;
408 		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
409 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
410 
411 		if (ops->ndo_do_ioctl) {
412 			mm_segment_t oldfs = get_fs();
413 
414 			set_fs(KERNEL_DS);
415 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
416 			set_fs(oldfs);
417 		} else {
418 			err = -EOPNOTSUPP;
419 		}
420 		dev = NULL;
421 
422 		if (err == 0 &&
423 		    (dev = __dev_get_by_name(net, p.name)) != NULL) {
424 			dev->flags |= IFF_MULTICAST;
425 
426 			in_dev = __in_dev_get_rtnl(dev);
427 			if (in_dev == NULL)
428 				goto failure;
429 
430 			ipv4_devconf_setall(in_dev);
431 			neigh_parms_data_state_setall(in_dev->arp_parms);
432 			IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
433 
434 			if (dev_open(dev))
435 				goto failure;
436 			dev_hold(dev);
437 		}
438 	}
439 	return dev;
440 
441 failure:
442 	/* allow the register to be completed before unregistering. */
443 	rtnl_unlock();
444 	rtnl_lock();
445 
446 	unregister_netdevice(dev);
447 	return NULL;
448 }
449 
450 #ifdef CONFIG_IP_PIMSM
451 
452 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
453 {
454 	struct net *net = dev_net(dev);
455 	struct mr_table *mrt;
456 	struct flowi4 fl4 = {
457 		.flowi4_oif	= dev->ifindex,
458 		.flowi4_iif	= skb->skb_iif,
459 		.flowi4_mark	= skb->mark,
460 	};
461 	int err;
462 
463 	err = ipmr_fib_lookup(net, &fl4, &mrt);
464 	if (err < 0) {
465 		kfree_skb(skb);
466 		return err;
467 	}
468 
469 	read_lock(&mrt_lock);
470 	dev->stats.tx_bytes += skb->len;
471 	dev->stats.tx_packets++;
472 	ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
473 	read_unlock(&mrt_lock);
474 	kfree_skb(skb);
475 	return NETDEV_TX_OK;
476 }
477 
478 static const struct net_device_ops reg_vif_netdev_ops = {
479 	.ndo_start_xmit	= reg_vif_xmit,
480 };
481 
482 static void reg_vif_setup(struct net_device *dev)
483 {
484 	dev->type		= ARPHRD_PIMREG;
485 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
486 	dev->flags		= IFF_NOARP;
487 	dev->netdev_ops		= &reg_vif_netdev_ops,
488 	dev->destructor		= free_netdev;
489 	dev->features		|= NETIF_F_NETNS_LOCAL;
490 }
491 
492 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
493 {
494 	struct net_device *dev;
495 	struct in_device *in_dev;
496 	char name[IFNAMSIZ];
497 
498 	if (mrt->id == RT_TABLE_DEFAULT)
499 		sprintf(name, "pimreg");
500 	else
501 		sprintf(name, "pimreg%u", mrt->id);
502 
503 	dev = alloc_netdev(0, name, reg_vif_setup);
504 
505 	if (dev == NULL)
506 		return NULL;
507 
508 	dev_net_set(dev, net);
509 
510 	if (register_netdevice(dev)) {
511 		free_netdev(dev);
512 		return NULL;
513 	}
514 	dev->iflink = 0;
515 
516 	rcu_read_lock();
517 	in_dev = __in_dev_get_rcu(dev);
518 	if (!in_dev) {
519 		rcu_read_unlock();
520 		goto failure;
521 	}
522 
523 	ipv4_devconf_setall(in_dev);
524 	neigh_parms_data_state_setall(in_dev->arp_parms);
525 	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
526 	rcu_read_unlock();
527 
528 	if (dev_open(dev))
529 		goto failure;
530 
531 	dev_hold(dev);
532 
533 	return dev;
534 
535 failure:
536 	/* allow the register to be completed before unregistering. */
537 	rtnl_unlock();
538 	rtnl_lock();
539 
540 	unregister_netdevice(dev);
541 	return NULL;
542 }
543 #endif
544 
545 /**
546  *	vif_delete - Delete a VIF entry
547  *	@notify: Set to 1, if the caller is a notifier_call
548  */
549 
550 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
551 		      struct list_head *head)
552 {
553 	struct vif_device *v;
554 	struct net_device *dev;
555 	struct in_device *in_dev;
556 
557 	if (vifi < 0 || vifi >= mrt->maxvif)
558 		return -EADDRNOTAVAIL;
559 
560 	v = &mrt->vif_table[vifi];
561 
562 	write_lock_bh(&mrt_lock);
563 	dev = v->dev;
564 	v->dev = NULL;
565 
566 	if (!dev) {
567 		write_unlock_bh(&mrt_lock);
568 		return -EADDRNOTAVAIL;
569 	}
570 
571 #ifdef CONFIG_IP_PIMSM
572 	if (vifi == mrt->mroute_reg_vif_num)
573 		mrt->mroute_reg_vif_num = -1;
574 #endif
575 
576 	if (vifi + 1 == mrt->maxvif) {
577 		int tmp;
578 
579 		for (tmp = vifi - 1; tmp >= 0; tmp--) {
580 			if (VIF_EXISTS(mrt, tmp))
581 				break;
582 		}
583 		mrt->maxvif = tmp+1;
584 	}
585 
586 	write_unlock_bh(&mrt_lock);
587 
588 	dev_set_allmulti(dev, -1);
589 
590 	in_dev = __in_dev_get_rtnl(dev);
591 	if (in_dev) {
592 		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
593 		inet_netconf_notify_devconf(dev_net(dev),
594 					    NETCONFA_MC_FORWARDING,
595 					    dev->ifindex, &in_dev->cnf);
596 		ip_rt_multicast_event(in_dev);
597 	}
598 
599 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
600 		unregister_netdevice_queue(dev, head);
601 
602 	dev_put(dev);
603 	return 0;
604 }
605 
606 static void ipmr_cache_free_rcu(struct rcu_head *head)
607 {
608 	struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
609 
610 	kmem_cache_free(mrt_cachep, c);
611 }
612 
613 static inline void ipmr_cache_free(struct mfc_cache *c)
614 {
615 	call_rcu(&c->rcu, ipmr_cache_free_rcu);
616 }
617 
618 /* Destroy an unresolved cache entry, killing queued skbs
619  * and reporting error to netlink readers.
620  */
621 
622 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
623 {
624 	struct net *net = read_pnet(&mrt->net);
625 	struct sk_buff *skb;
626 	struct nlmsgerr *e;
627 
628 	atomic_dec(&mrt->cache_resolve_queue_len);
629 
630 	while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
631 		if (ip_hdr(skb)->version == 0) {
632 			struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
633 			nlh->nlmsg_type = NLMSG_ERROR;
634 			nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
635 			skb_trim(skb, nlh->nlmsg_len);
636 			e = nlmsg_data(nlh);
637 			e->error = -ETIMEDOUT;
638 			memset(&e->msg, 0, sizeof(e->msg));
639 
640 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
641 		} else {
642 			kfree_skb(skb);
643 		}
644 	}
645 
646 	ipmr_cache_free(c);
647 }
648 
649 
650 /* Timer process for the unresolved queue. */
651 
652 static void ipmr_expire_process(unsigned long arg)
653 {
654 	struct mr_table *mrt = (struct mr_table *)arg;
655 	unsigned long now;
656 	unsigned long expires;
657 	struct mfc_cache *c, *next;
658 
659 	if (!spin_trylock(&mfc_unres_lock)) {
660 		mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
661 		return;
662 	}
663 
664 	if (list_empty(&mrt->mfc_unres_queue))
665 		goto out;
666 
667 	now = jiffies;
668 	expires = 10*HZ;
669 
670 	list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
671 		if (time_after(c->mfc_un.unres.expires, now)) {
672 			unsigned long interval = c->mfc_un.unres.expires - now;
673 			if (interval < expires)
674 				expires = interval;
675 			continue;
676 		}
677 
678 		list_del(&c->list);
679 		mroute_netlink_event(mrt, c, RTM_DELROUTE);
680 		ipmr_destroy_unres(mrt, c);
681 	}
682 
683 	if (!list_empty(&mrt->mfc_unres_queue))
684 		mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
685 
686 out:
687 	spin_unlock(&mfc_unres_lock);
688 }
689 
690 /* Fill oifs list. It is called under write locked mrt_lock. */
691 
692 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
693 				   unsigned char *ttls)
694 {
695 	int vifi;
696 
697 	cache->mfc_un.res.minvif = MAXVIFS;
698 	cache->mfc_un.res.maxvif = 0;
699 	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
700 
701 	for (vifi = 0; vifi < mrt->maxvif; vifi++) {
702 		if (VIF_EXISTS(mrt, vifi) &&
703 		    ttls[vifi] && ttls[vifi] < 255) {
704 			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
705 			if (cache->mfc_un.res.minvif > vifi)
706 				cache->mfc_un.res.minvif = vifi;
707 			if (cache->mfc_un.res.maxvif <= vifi)
708 				cache->mfc_un.res.maxvif = vifi + 1;
709 		}
710 	}
711 }
712 
713 static int vif_add(struct net *net, struct mr_table *mrt,
714 		   struct vifctl *vifc, int mrtsock)
715 {
716 	int vifi = vifc->vifc_vifi;
717 	struct vif_device *v = &mrt->vif_table[vifi];
718 	struct net_device *dev;
719 	struct in_device *in_dev;
720 	int err;
721 
722 	/* Is vif busy ? */
723 	if (VIF_EXISTS(mrt, vifi))
724 		return -EADDRINUSE;
725 
726 	switch (vifc->vifc_flags) {
727 #ifdef CONFIG_IP_PIMSM
728 	case VIFF_REGISTER:
729 		/*
730 		 * Special Purpose VIF in PIM
731 		 * All the packets will be sent to the daemon
732 		 */
733 		if (mrt->mroute_reg_vif_num >= 0)
734 			return -EADDRINUSE;
735 		dev = ipmr_reg_vif(net, mrt);
736 		if (!dev)
737 			return -ENOBUFS;
738 		err = dev_set_allmulti(dev, 1);
739 		if (err) {
740 			unregister_netdevice(dev);
741 			dev_put(dev);
742 			return err;
743 		}
744 		break;
745 #endif
746 	case VIFF_TUNNEL:
747 		dev = ipmr_new_tunnel(net, vifc);
748 		if (!dev)
749 			return -ENOBUFS;
750 		err = dev_set_allmulti(dev, 1);
751 		if (err) {
752 			ipmr_del_tunnel(dev, vifc);
753 			dev_put(dev);
754 			return err;
755 		}
756 		break;
757 
758 	case VIFF_USE_IFINDEX:
759 	case 0:
760 		if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
761 			dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
762 			if (dev && __in_dev_get_rtnl(dev) == NULL) {
763 				dev_put(dev);
764 				return -EADDRNOTAVAIL;
765 			}
766 		} else {
767 			dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
768 		}
769 		if (!dev)
770 			return -EADDRNOTAVAIL;
771 		err = dev_set_allmulti(dev, 1);
772 		if (err) {
773 			dev_put(dev);
774 			return err;
775 		}
776 		break;
777 	default:
778 		return -EINVAL;
779 	}
780 
781 	in_dev = __in_dev_get_rtnl(dev);
782 	if (!in_dev) {
783 		dev_put(dev);
784 		return -EADDRNOTAVAIL;
785 	}
786 	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
787 	inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
788 				    &in_dev->cnf);
789 	ip_rt_multicast_event(in_dev);
790 
791 	/* Fill in the VIF structures */
792 
793 	v->rate_limit = vifc->vifc_rate_limit;
794 	v->local = vifc->vifc_lcl_addr.s_addr;
795 	v->remote = vifc->vifc_rmt_addr.s_addr;
796 	v->flags = vifc->vifc_flags;
797 	if (!mrtsock)
798 		v->flags |= VIFF_STATIC;
799 	v->threshold = vifc->vifc_threshold;
800 	v->bytes_in = 0;
801 	v->bytes_out = 0;
802 	v->pkt_in = 0;
803 	v->pkt_out = 0;
804 	v->link = dev->ifindex;
805 	if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
806 		v->link = dev->iflink;
807 
808 	/* And finish update writing critical data */
809 	write_lock_bh(&mrt_lock);
810 	v->dev = dev;
811 #ifdef CONFIG_IP_PIMSM
812 	if (v->flags & VIFF_REGISTER)
813 		mrt->mroute_reg_vif_num = vifi;
814 #endif
815 	if (vifi+1 > mrt->maxvif)
816 		mrt->maxvif = vifi+1;
817 	write_unlock_bh(&mrt_lock);
818 	return 0;
819 }
820 
821 /* called with rcu_read_lock() */
822 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
823 					 __be32 origin,
824 					 __be32 mcastgrp)
825 {
826 	int line = MFC_HASH(mcastgrp, origin);
827 	struct mfc_cache *c;
828 
829 	list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
830 		if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
831 			return c;
832 	}
833 	return NULL;
834 }
835 
836 /* Look for a (*,*,oif) entry */
837 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
838 						    int vifi)
839 {
840 	int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
841 	struct mfc_cache *c;
842 
843 	list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
844 		if (c->mfc_origin == htonl(INADDR_ANY) &&
845 		    c->mfc_mcastgrp == htonl(INADDR_ANY) &&
846 		    c->mfc_un.res.ttls[vifi] < 255)
847 			return c;
848 
849 	return NULL;
850 }
851 
852 /* Look for a (*,G) entry */
853 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
854 					     __be32 mcastgrp, int vifi)
855 {
856 	int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
857 	struct mfc_cache *c, *proxy;
858 
859 	if (mcastgrp == htonl(INADDR_ANY))
860 		goto skip;
861 
862 	list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
863 		if (c->mfc_origin == htonl(INADDR_ANY) &&
864 		    c->mfc_mcastgrp == mcastgrp) {
865 			if (c->mfc_un.res.ttls[vifi] < 255)
866 				return c;
867 
868 			/* It's ok if the vifi is part of the static tree */
869 			proxy = ipmr_cache_find_any_parent(mrt,
870 							   c->mfc_parent);
871 			if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
872 				return c;
873 		}
874 
875 skip:
876 	return ipmr_cache_find_any_parent(mrt, vifi);
877 }
878 
879 /*
880  *	Allocate a multicast cache entry
881  */
882 static struct mfc_cache *ipmr_cache_alloc(void)
883 {
884 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
885 
886 	if (c)
887 		c->mfc_un.res.minvif = MAXVIFS;
888 	return c;
889 }
890 
891 static struct mfc_cache *ipmr_cache_alloc_unres(void)
892 {
893 	struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
894 
895 	if (c) {
896 		skb_queue_head_init(&c->mfc_un.unres.unresolved);
897 		c->mfc_un.unres.expires = jiffies + 10*HZ;
898 	}
899 	return c;
900 }
901 
902 /*
903  *	A cache entry has gone into a resolved state from queued
904  */
905 
906 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
907 			       struct mfc_cache *uc, struct mfc_cache *c)
908 {
909 	struct sk_buff *skb;
910 	struct nlmsgerr *e;
911 
912 	/* Play the pending entries through our router */
913 
914 	while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
915 		if (ip_hdr(skb)->version == 0) {
916 			struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
917 
918 			if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
919 				nlh->nlmsg_len = skb_tail_pointer(skb) -
920 						 (u8 *)nlh;
921 			} else {
922 				nlh->nlmsg_type = NLMSG_ERROR;
923 				nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
924 				skb_trim(skb, nlh->nlmsg_len);
925 				e = nlmsg_data(nlh);
926 				e->error = -EMSGSIZE;
927 				memset(&e->msg, 0, sizeof(e->msg));
928 			}
929 
930 			rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
931 		} else {
932 			ip_mr_forward(net, mrt, skb, c, 0);
933 		}
934 	}
935 }
936 
937 /*
938  *	Bounce a cache query up to mrouted. We could use netlink for this but mrouted
939  *	expects the following bizarre scheme.
940  *
941  *	Called under mrt_lock.
942  */
943 
944 static int ipmr_cache_report(struct mr_table *mrt,
945 			     struct sk_buff *pkt, vifi_t vifi, int assert)
946 {
947 	struct sk_buff *skb;
948 	const int ihl = ip_hdrlen(pkt);
949 	struct igmphdr *igmp;
950 	struct igmpmsg *msg;
951 	struct sock *mroute_sk;
952 	int ret;
953 
954 #ifdef CONFIG_IP_PIMSM
955 	if (assert == IGMPMSG_WHOLEPKT)
956 		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
957 	else
958 #endif
959 		skb = alloc_skb(128, GFP_ATOMIC);
960 
961 	if (!skb)
962 		return -ENOBUFS;
963 
964 #ifdef CONFIG_IP_PIMSM
965 	if (assert == IGMPMSG_WHOLEPKT) {
966 		/* Ugly, but we have no choice with this interface.
967 		 * Duplicate old header, fix ihl, length etc.
968 		 * And all this only to mangle msg->im_msgtype and
969 		 * to set msg->im_mbz to "mbz" :-)
970 		 */
971 		skb_push(skb, sizeof(struct iphdr));
972 		skb_reset_network_header(skb);
973 		skb_reset_transport_header(skb);
974 		msg = (struct igmpmsg *)skb_network_header(skb);
975 		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
976 		msg->im_msgtype = IGMPMSG_WHOLEPKT;
977 		msg->im_mbz = 0;
978 		msg->im_vif = mrt->mroute_reg_vif_num;
979 		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
980 		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
981 					     sizeof(struct iphdr));
982 	} else
983 #endif
984 	{
985 
986 	/* Copy the IP header */
987 
988 	skb_set_network_header(skb, skb->len);
989 	skb_put(skb, ihl);
990 	skb_copy_to_linear_data(skb, pkt->data, ihl);
991 	ip_hdr(skb)->protocol = 0;	/* Flag to the kernel this is a route add */
992 	msg = (struct igmpmsg *)skb_network_header(skb);
993 	msg->im_vif = vifi;
994 	skb_dst_set(skb, dst_clone(skb_dst(pkt)));
995 
996 	/* Add our header */
997 
998 	igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
999 	igmp->type	=
1000 	msg->im_msgtype = assert;
1001 	igmp->code	= 0;
1002 	ip_hdr(skb)->tot_len = htons(skb->len);		/* Fix the length */
1003 	skb->transport_header = skb->network_header;
1004 	}
1005 
1006 	rcu_read_lock();
1007 	mroute_sk = rcu_dereference(mrt->mroute_sk);
1008 	if (mroute_sk == NULL) {
1009 		rcu_read_unlock();
1010 		kfree_skb(skb);
1011 		return -EINVAL;
1012 	}
1013 
1014 	/* Deliver to mrouted */
1015 
1016 	ret = sock_queue_rcv_skb(mroute_sk, skb);
1017 	rcu_read_unlock();
1018 	if (ret < 0) {
1019 		net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1020 		kfree_skb(skb);
1021 	}
1022 
1023 	return ret;
1024 }
1025 
1026 /*
1027  *	Queue a packet for resolution. It gets locked cache entry!
1028  */
1029 
1030 static int
1031 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1032 {
1033 	bool found = false;
1034 	int err;
1035 	struct mfc_cache *c;
1036 	const struct iphdr *iph = ip_hdr(skb);
1037 
1038 	spin_lock_bh(&mfc_unres_lock);
1039 	list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1040 		if (c->mfc_mcastgrp == iph->daddr &&
1041 		    c->mfc_origin == iph->saddr) {
1042 			found = true;
1043 			break;
1044 		}
1045 	}
1046 
1047 	if (!found) {
1048 		/* Create a new entry if allowable */
1049 
1050 		if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1051 		    (c = ipmr_cache_alloc_unres()) == NULL) {
1052 			spin_unlock_bh(&mfc_unres_lock);
1053 
1054 			kfree_skb(skb);
1055 			return -ENOBUFS;
1056 		}
1057 
1058 		/* Fill in the new cache entry */
1059 
1060 		c->mfc_parent	= -1;
1061 		c->mfc_origin	= iph->saddr;
1062 		c->mfc_mcastgrp	= iph->daddr;
1063 
1064 		/* Reflect first query at mrouted. */
1065 
1066 		err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1067 		if (err < 0) {
1068 			/* If the report failed throw the cache entry
1069 			   out - Brad Parker
1070 			 */
1071 			spin_unlock_bh(&mfc_unres_lock);
1072 
1073 			ipmr_cache_free(c);
1074 			kfree_skb(skb);
1075 			return err;
1076 		}
1077 
1078 		atomic_inc(&mrt->cache_resolve_queue_len);
1079 		list_add(&c->list, &mrt->mfc_unres_queue);
1080 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1081 
1082 		if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1083 			mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1084 	}
1085 
1086 	/* See if we can append the packet */
1087 
1088 	if (c->mfc_un.unres.unresolved.qlen > 3) {
1089 		kfree_skb(skb);
1090 		err = -ENOBUFS;
1091 	} else {
1092 		skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1093 		err = 0;
1094 	}
1095 
1096 	spin_unlock_bh(&mfc_unres_lock);
1097 	return err;
1098 }
1099 
1100 /*
1101  *	MFC cache manipulation by user space mroute daemon
1102  */
1103 
1104 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1105 {
1106 	int line;
1107 	struct mfc_cache *c, *next;
1108 
1109 	line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1110 
1111 	list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1112 		if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1113 		    c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1114 		    (parent == -1 || parent == c->mfc_parent)) {
1115 			list_del_rcu(&c->list);
1116 			mroute_netlink_event(mrt, c, RTM_DELROUTE);
1117 			ipmr_cache_free(c);
1118 			return 0;
1119 		}
1120 	}
1121 	return -ENOENT;
1122 }
1123 
1124 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1125 			struct mfcctl *mfc, int mrtsock, int parent)
1126 {
1127 	bool found = false;
1128 	int line;
1129 	struct mfc_cache *uc, *c;
1130 
1131 	if (mfc->mfcc_parent >= MAXVIFS)
1132 		return -ENFILE;
1133 
1134 	line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1135 
1136 	list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1137 		if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1138 		    c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1139 		    (parent == -1 || parent == c->mfc_parent)) {
1140 			found = true;
1141 			break;
1142 		}
1143 	}
1144 
1145 	if (found) {
1146 		write_lock_bh(&mrt_lock);
1147 		c->mfc_parent = mfc->mfcc_parent;
1148 		ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1149 		if (!mrtsock)
1150 			c->mfc_flags |= MFC_STATIC;
1151 		write_unlock_bh(&mrt_lock);
1152 		mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1153 		return 0;
1154 	}
1155 
1156 	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1157 	    !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1158 		return -EINVAL;
1159 
1160 	c = ipmr_cache_alloc();
1161 	if (c == NULL)
1162 		return -ENOMEM;
1163 
1164 	c->mfc_origin = mfc->mfcc_origin.s_addr;
1165 	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1166 	c->mfc_parent = mfc->mfcc_parent;
1167 	ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1168 	if (!mrtsock)
1169 		c->mfc_flags |= MFC_STATIC;
1170 
1171 	list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1172 
1173 	/*
1174 	 *	Check to see if we resolved a queued list. If so we
1175 	 *	need to send on the frames and tidy up.
1176 	 */
1177 	found = false;
1178 	spin_lock_bh(&mfc_unres_lock);
1179 	list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1180 		if (uc->mfc_origin == c->mfc_origin &&
1181 		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1182 			list_del(&uc->list);
1183 			atomic_dec(&mrt->cache_resolve_queue_len);
1184 			found = true;
1185 			break;
1186 		}
1187 	}
1188 	if (list_empty(&mrt->mfc_unres_queue))
1189 		del_timer(&mrt->ipmr_expire_timer);
1190 	spin_unlock_bh(&mfc_unres_lock);
1191 
1192 	if (found) {
1193 		ipmr_cache_resolve(net, mrt, uc, c);
1194 		ipmr_cache_free(uc);
1195 	}
1196 	mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1197 	return 0;
1198 }
1199 
1200 /*
1201  *	Close the multicast socket, and clear the vif tables etc
1202  */
1203 
1204 static void mroute_clean_tables(struct mr_table *mrt)
1205 {
1206 	int i;
1207 	LIST_HEAD(list);
1208 	struct mfc_cache *c, *next;
1209 
1210 	/* Shut down all active vif entries */
1211 
1212 	for (i = 0; i < mrt->maxvif; i++) {
1213 		if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1214 			vif_delete(mrt, i, 0, &list);
1215 	}
1216 	unregister_netdevice_many(&list);
1217 
1218 	/* Wipe the cache */
1219 
1220 	for (i = 0; i < MFC_LINES; i++) {
1221 		list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1222 			if (c->mfc_flags & MFC_STATIC)
1223 				continue;
1224 			list_del_rcu(&c->list);
1225 			mroute_netlink_event(mrt, c, RTM_DELROUTE);
1226 			ipmr_cache_free(c);
1227 		}
1228 	}
1229 
1230 	if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1231 		spin_lock_bh(&mfc_unres_lock);
1232 		list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1233 			list_del(&c->list);
1234 			mroute_netlink_event(mrt, c, RTM_DELROUTE);
1235 			ipmr_destroy_unres(mrt, c);
1236 		}
1237 		spin_unlock_bh(&mfc_unres_lock);
1238 	}
1239 }
1240 
1241 /* called from ip_ra_control(), before an RCU grace period,
1242  * we dont need to call synchronize_rcu() here
1243  */
1244 static void mrtsock_destruct(struct sock *sk)
1245 {
1246 	struct net *net = sock_net(sk);
1247 	struct mr_table *mrt;
1248 
1249 	rtnl_lock();
1250 	ipmr_for_each_table(mrt, net) {
1251 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1252 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1253 			inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1254 						    NETCONFA_IFINDEX_ALL,
1255 						    net->ipv4.devconf_all);
1256 			RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1257 			mroute_clean_tables(mrt);
1258 		}
1259 	}
1260 	rtnl_unlock();
1261 }
1262 
1263 /*
1264  *	Socket options and virtual interface manipulation. The whole
1265  *	virtual interface system is a complete heap, but unfortunately
1266  *	that's how BSD mrouted happens to think. Maybe one day with a proper
1267  *	MOSPF/PIM router set up we can clean this up.
1268  */
1269 
1270 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1271 {
1272 	int ret, parent = 0;
1273 	struct vifctl vif;
1274 	struct mfcctl mfc;
1275 	struct net *net = sock_net(sk);
1276 	struct mr_table *mrt;
1277 
1278 	if (sk->sk_type != SOCK_RAW ||
1279 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1280 		return -EOPNOTSUPP;
1281 
1282 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1283 	if (mrt == NULL)
1284 		return -ENOENT;
1285 
1286 	if (optname != MRT_INIT) {
1287 		if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1288 		    !ns_capable(net->user_ns, CAP_NET_ADMIN))
1289 			return -EACCES;
1290 	}
1291 
1292 	switch (optname) {
1293 	case MRT_INIT:
1294 		if (optlen != sizeof(int))
1295 			return -EINVAL;
1296 
1297 		rtnl_lock();
1298 		if (rtnl_dereference(mrt->mroute_sk)) {
1299 			rtnl_unlock();
1300 			return -EADDRINUSE;
1301 		}
1302 
1303 		ret = ip_ra_control(sk, 1, mrtsock_destruct);
1304 		if (ret == 0) {
1305 			rcu_assign_pointer(mrt->mroute_sk, sk);
1306 			IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1307 			inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1308 						    NETCONFA_IFINDEX_ALL,
1309 						    net->ipv4.devconf_all);
1310 		}
1311 		rtnl_unlock();
1312 		return ret;
1313 	case MRT_DONE:
1314 		if (sk != rcu_access_pointer(mrt->mroute_sk))
1315 			return -EACCES;
1316 		return ip_ra_control(sk, 0, NULL);
1317 	case MRT_ADD_VIF:
1318 	case MRT_DEL_VIF:
1319 		if (optlen != sizeof(vif))
1320 			return -EINVAL;
1321 		if (copy_from_user(&vif, optval, sizeof(vif)))
1322 			return -EFAULT;
1323 		if (vif.vifc_vifi >= MAXVIFS)
1324 			return -ENFILE;
1325 		rtnl_lock();
1326 		if (optname == MRT_ADD_VIF) {
1327 			ret = vif_add(net, mrt, &vif,
1328 				      sk == rtnl_dereference(mrt->mroute_sk));
1329 		} else {
1330 			ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1331 		}
1332 		rtnl_unlock();
1333 		return ret;
1334 
1335 		/*
1336 		 *	Manipulate the forwarding caches. These live
1337 		 *	in a sort of kernel/user symbiosis.
1338 		 */
1339 	case MRT_ADD_MFC:
1340 	case MRT_DEL_MFC:
1341 		parent = -1;
1342 	case MRT_ADD_MFC_PROXY:
1343 	case MRT_DEL_MFC_PROXY:
1344 		if (optlen != sizeof(mfc))
1345 			return -EINVAL;
1346 		if (copy_from_user(&mfc, optval, sizeof(mfc)))
1347 			return -EFAULT;
1348 		if (parent == 0)
1349 			parent = mfc.mfcc_parent;
1350 		rtnl_lock();
1351 		if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1352 			ret = ipmr_mfc_delete(mrt, &mfc, parent);
1353 		else
1354 			ret = ipmr_mfc_add(net, mrt, &mfc,
1355 					   sk == rtnl_dereference(mrt->mroute_sk),
1356 					   parent);
1357 		rtnl_unlock();
1358 		return ret;
1359 		/*
1360 		 *	Control PIM assert.
1361 		 */
1362 	case MRT_ASSERT:
1363 	{
1364 		int v;
1365 		if (optlen != sizeof(v))
1366 			return -EINVAL;
1367 		if (get_user(v, (int __user *)optval))
1368 			return -EFAULT;
1369 		mrt->mroute_do_assert = v;
1370 		return 0;
1371 	}
1372 #ifdef CONFIG_IP_PIMSM
1373 	case MRT_PIM:
1374 	{
1375 		int v;
1376 
1377 		if (optlen != sizeof(v))
1378 			return -EINVAL;
1379 		if (get_user(v, (int __user *)optval))
1380 			return -EFAULT;
1381 		v = !!v;
1382 
1383 		rtnl_lock();
1384 		ret = 0;
1385 		if (v != mrt->mroute_do_pim) {
1386 			mrt->mroute_do_pim = v;
1387 			mrt->mroute_do_assert = v;
1388 		}
1389 		rtnl_unlock();
1390 		return ret;
1391 	}
1392 #endif
1393 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1394 	case MRT_TABLE:
1395 	{
1396 		u32 v;
1397 
1398 		if (optlen != sizeof(u32))
1399 			return -EINVAL;
1400 		if (get_user(v, (u32 __user *)optval))
1401 			return -EFAULT;
1402 
1403 		/* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1404 		if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1405 			return -EINVAL;
1406 
1407 		rtnl_lock();
1408 		ret = 0;
1409 		if (sk == rtnl_dereference(mrt->mroute_sk)) {
1410 			ret = -EBUSY;
1411 		} else {
1412 			if (!ipmr_new_table(net, v))
1413 				ret = -ENOMEM;
1414 			else
1415 				raw_sk(sk)->ipmr_table = v;
1416 		}
1417 		rtnl_unlock();
1418 		return ret;
1419 	}
1420 #endif
1421 	/*
1422 	 *	Spurious command, or MRT_VERSION which you cannot
1423 	 *	set.
1424 	 */
1425 	default:
1426 		return -ENOPROTOOPT;
1427 	}
1428 }
1429 
1430 /*
1431  *	Getsock opt support for the multicast routing system.
1432  */
1433 
1434 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1435 {
1436 	int olr;
1437 	int val;
1438 	struct net *net = sock_net(sk);
1439 	struct mr_table *mrt;
1440 
1441 	if (sk->sk_type != SOCK_RAW ||
1442 	    inet_sk(sk)->inet_num != IPPROTO_IGMP)
1443 		return -EOPNOTSUPP;
1444 
1445 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1446 	if (mrt == NULL)
1447 		return -ENOENT;
1448 
1449 	if (optname != MRT_VERSION &&
1450 #ifdef CONFIG_IP_PIMSM
1451 	   optname != MRT_PIM &&
1452 #endif
1453 	   optname != MRT_ASSERT)
1454 		return -ENOPROTOOPT;
1455 
1456 	if (get_user(olr, optlen))
1457 		return -EFAULT;
1458 
1459 	olr = min_t(unsigned int, olr, sizeof(int));
1460 	if (olr < 0)
1461 		return -EINVAL;
1462 
1463 	if (put_user(olr, optlen))
1464 		return -EFAULT;
1465 	if (optname == MRT_VERSION)
1466 		val = 0x0305;
1467 #ifdef CONFIG_IP_PIMSM
1468 	else if (optname == MRT_PIM)
1469 		val = mrt->mroute_do_pim;
1470 #endif
1471 	else
1472 		val = mrt->mroute_do_assert;
1473 	if (copy_to_user(optval, &val, olr))
1474 		return -EFAULT;
1475 	return 0;
1476 }
1477 
1478 /*
1479  *	The IP multicast ioctl support routines.
1480  */
1481 
1482 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1483 {
1484 	struct sioc_sg_req sr;
1485 	struct sioc_vif_req vr;
1486 	struct vif_device *vif;
1487 	struct mfc_cache *c;
1488 	struct net *net = sock_net(sk);
1489 	struct mr_table *mrt;
1490 
1491 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1492 	if (mrt == NULL)
1493 		return -ENOENT;
1494 
1495 	switch (cmd) {
1496 	case SIOCGETVIFCNT:
1497 		if (copy_from_user(&vr, arg, sizeof(vr)))
1498 			return -EFAULT;
1499 		if (vr.vifi >= mrt->maxvif)
1500 			return -EINVAL;
1501 		read_lock(&mrt_lock);
1502 		vif = &mrt->vif_table[vr.vifi];
1503 		if (VIF_EXISTS(mrt, vr.vifi)) {
1504 			vr.icount = vif->pkt_in;
1505 			vr.ocount = vif->pkt_out;
1506 			vr.ibytes = vif->bytes_in;
1507 			vr.obytes = vif->bytes_out;
1508 			read_unlock(&mrt_lock);
1509 
1510 			if (copy_to_user(arg, &vr, sizeof(vr)))
1511 				return -EFAULT;
1512 			return 0;
1513 		}
1514 		read_unlock(&mrt_lock);
1515 		return -EADDRNOTAVAIL;
1516 	case SIOCGETSGCNT:
1517 		if (copy_from_user(&sr, arg, sizeof(sr)))
1518 			return -EFAULT;
1519 
1520 		rcu_read_lock();
1521 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1522 		if (c) {
1523 			sr.pktcnt = c->mfc_un.res.pkt;
1524 			sr.bytecnt = c->mfc_un.res.bytes;
1525 			sr.wrong_if = c->mfc_un.res.wrong_if;
1526 			rcu_read_unlock();
1527 
1528 			if (copy_to_user(arg, &sr, sizeof(sr)))
1529 				return -EFAULT;
1530 			return 0;
1531 		}
1532 		rcu_read_unlock();
1533 		return -EADDRNOTAVAIL;
1534 	default:
1535 		return -ENOIOCTLCMD;
1536 	}
1537 }
1538 
1539 #ifdef CONFIG_COMPAT
1540 struct compat_sioc_sg_req {
1541 	struct in_addr src;
1542 	struct in_addr grp;
1543 	compat_ulong_t pktcnt;
1544 	compat_ulong_t bytecnt;
1545 	compat_ulong_t wrong_if;
1546 };
1547 
1548 struct compat_sioc_vif_req {
1549 	vifi_t	vifi;		/* Which iface */
1550 	compat_ulong_t icount;
1551 	compat_ulong_t ocount;
1552 	compat_ulong_t ibytes;
1553 	compat_ulong_t obytes;
1554 };
1555 
1556 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1557 {
1558 	struct compat_sioc_sg_req sr;
1559 	struct compat_sioc_vif_req vr;
1560 	struct vif_device *vif;
1561 	struct mfc_cache *c;
1562 	struct net *net = sock_net(sk);
1563 	struct mr_table *mrt;
1564 
1565 	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1566 	if (mrt == NULL)
1567 		return -ENOENT;
1568 
1569 	switch (cmd) {
1570 	case SIOCGETVIFCNT:
1571 		if (copy_from_user(&vr, arg, sizeof(vr)))
1572 			return -EFAULT;
1573 		if (vr.vifi >= mrt->maxvif)
1574 			return -EINVAL;
1575 		read_lock(&mrt_lock);
1576 		vif = &mrt->vif_table[vr.vifi];
1577 		if (VIF_EXISTS(mrt, vr.vifi)) {
1578 			vr.icount = vif->pkt_in;
1579 			vr.ocount = vif->pkt_out;
1580 			vr.ibytes = vif->bytes_in;
1581 			vr.obytes = vif->bytes_out;
1582 			read_unlock(&mrt_lock);
1583 
1584 			if (copy_to_user(arg, &vr, sizeof(vr)))
1585 				return -EFAULT;
1586 			return 0;
1587 		}
1588 		read_unlock(&mrt_lock);
1589 		return -EADDRNOTAVAIL;
1590 	case SIOCGETSGCNT:
1591 		if (copy_from_user(&sr, arg, sizeof(sr)))
1592 			return -EFAULT;
1593 
1594 		rcu_read_lock();
1595 		c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1596 		if (c) {
1597 			sr.pktcnt = c->mfc_un.res.pkt;
1598 			sr.bytecnt = c->mfc_un.res.bytes;
1599 			sr.wrong_if = c->mfc_un.res.wrong_if;
1600 			rcu_read_unlock();
1601 
1602 			if (copy_to_user(arg, &sr, sizeof(sr)))
1603 				return -EFAULT;
1604 			return 0;
1605 		}
1606 		rcu_read_unlock();
1607 		return -EADDRNOTAVAIL;
1608 	default:
1609 		return -ENOIOCTLCMD;
1610 	}
1611 }
1612 #endif
1613 
1614 
1615 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1616 {
1617 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1618 	struct net *net = dev_net(dev);
1619 	struct mr_table *mrt;
1620 	struct vif_device *v;
1621 	int ct;
1622 
1623 	if (event != NETDEV_UNREGISTER)
1624 		return NOTIFY_DONE;
1625 
1626 	ipmr_for_each_table(mrt, net) {
1627 		v = &mrt->vif_table[0];
1628 		for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1629 			if (v->dev == dev)
1630 				vif_delete(mrt, ct, 1, NULL);
1631 		}
1632 	}
1633 	return NOTIFY_DONE;
1634 }
1635 
1636 
1637 static struct notifier_block ip_mr_notifier = {
1638 	.notifier_call = ipmr_device_event,
1639 };
1640 
1641 /*
1642  *	Encapsulate a packet by attaching a valid IPIP header to it.
1643  *	This avoids tunnel drivers and other mess and gives us the speed so
1644  *	important for multicast video.
1645  */
1646 
1647 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1648 {
1649 	struct iphdr *iph;
1650 	const struct iphdr *old_iph = ip_hdr(skb);
1651 
1652 	skb_push(skb, sizeof(struct iphdr));
1653 	skb->transport_header = skb->network_header;
1654 	skb_reset_network_header(skb);
1655 	iph = ip_hdr(skb);
1656 
1657 	iph->version	=	4;
1658 	iph->tos	=	old_iph->tos;
1659 	iph->ttl	=	old_iph->ttl;
1660 	iph->frag_off	=	0;
1661 	iph->daddr	=	daddr;
1662 	iph->saddr	=	saddr;
1663 	iph->protocol	=	IPPROTO_IPIP;
1664 	iph->ihl	=	5;
1665 	iph->tot_len	=	htons(skb->len);
1666 	ip_select_ident(skb, skb_dst(skb), NULL);
1667 	ip_send_check(iph);
1668 
1669 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1670 	nf_reset(skb);
1671 }
1672 
1673 static inline int ipmr_forward_finish(struct sk_buff *skb)
1674 {
1675 	struct ip_options *opt = &(IPCB(skb)->opt);
1676 
1677 	IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1678 	IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1679 
1680 	if (unlikely(opt->optlen))
1681 		ip_forward_options(skb);
1682 
1683 	return dst_output(skb);
1684 }
1685 
1686 /*
1687  *	Processing handlers for ipmr_forward
1688  */
1689 
1690 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1691 			    struct sk_buff *skb, struct mfc_cache *c, int vifi)
1692 {
1693 	const struct iphdr *iph = ip_hdr(skb);
1694 	struct vif_device *vif = &mrt->vif_table[vifi];
1695 	struct net_device *dev;
1696 	struct rtable *rt;
1697 	struct flowi4 fl4;
1698 	int    encap = 0;
1699 
1700 	if (vif->dev == NULL)
1701 		goto out_free;
1702 
1703 #ifdef CONFIG_IP_PIMSM
1704 	if (vif->flags & VIFF_REGISTER) {
1705 		vif->pkt_out++;
1706 		vif->bytes_out += skb->len;
1707 		vif->dev->stats.tx_bytes += skb->len;
1708 		vif->dev->stats.tx_packets++;
1709 		ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1710 		goto out_free;
1711 	}
1712 #endif
1713 
1714 	if (vif->flags & VIFF_TUNNEL) {
1715 		rt = ip_route_output_ports(net, &fl4, NULL,
1716 					   vif->remote, vif->local,
1717 					   0, 0,
1718 					   IPPROTO_IPIP,
1719 					   RT_TOS(iph->tos), vif->link);
1720 		if (IS_ERR(rt))
1721 			goto out_free;
1722 		encap = sizeof(struct iphdr);
1723 	} else {
1724 		rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1725 					   0, 0,
1726 					   IPPROTO_IPIP,
1727 					   RT_TOS(iph->tos), vif->link);
1728 		if (IS_ERR(rt))
1729 			goto out_free;
1730 	}
1731 
1732 	dev = rt->dst.dev;
1733 
1734 	if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1735 		/* Do not fragment multicasts. Alas, IPv4 does not
1736 		 * allow to send ICMP, so that packets will disappear
1737 		 * to blackhole.
1738 		 */
1739 
1740 		IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1741 		ip_rt_put(rt);
1742 		goto out_free;
1743 	}
1744 
1745 	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1746 
1747 	if (skb_cow(skb, encap)) {
1748 		ip_rt_put(rt);
1749 		goto out_free;
1750 	}
1751 
1752 	vif->pkt_out++;
1753 	vif->bytes_out += skb->len;
1754 
1755 	skb_dst_drop(skb);
1756 	skb_dst_set(skb, &rt->dst);
1757 	ip_decrease_ttl(ip_hdr(skb));
1758 
1759 	/* FIXME: forward and output firewalls used to be called here.
1760 	 * What do we do with netfilter? -- RR
1761 	 */
1762 	if (vif->flags & VIFF_TUNNEL) {
1763 		ip_encap(skb, vif->local, vif->remote);
1764 		/* FIXME: extra output firewall step used to be here. --RR */
1765 		vif->dev->stats.tx_packets++;
1766 		vif->dev->stats.tx_bytes += skb->len;
1767 	}
1768 
1769 	IPCB(skb)->flags |= IPSKB_FORWARDED;
1770 
1771 	/*
1772 	 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1773 	 * not only before forwarding, but after forwarding on all output
1774 	 * interfaces. It is clear, if mrouter runs a multicasting
1775 	 * program, it should receive packets not depending to what interface
1776 	 * program is joined.
1777 	 * If we will not make it, the program will have to join on all
1778 	 * interfaces. On the other hand, multihoming host (or router, but
1779 	 * not mrouter) cannot join to more than one interface - it will
1780 	 * result in receiving multiple packets.
1781 	 */
1782 	NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1783 		ipmr_forward_finish);
1784 	return;
1785 
1786 out_free:
1787 	kfree_skb(skb);
1788 }
1789 
1790 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1791 {
1792 	int ct;
1793 
1794 	for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1795 		if (mrt->vif_table[ct].dev == dev)
1796 			break;
1797 	}
1798 	return ct;
1799 }
1800 
1801 /* "local" means that we should preserve one skb (for local delivery) */
1802 
1803 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1804 			  struct sk_buff *skb, struct mfc_cache *cache,
1805 			  int local)
1806 {
1807 	int psend = -1;
1808 	int vif, ct;
1809 	int true_vifi = ipmr_find_vif(mrt, skb->dev);
1810 
1811 	vif = cache->mfc_parent;
1812 	cache->mfc_un.res.pkt++;
1813 	cache->mfc_un.res.bytes += skb->len;
1814 
1815 	if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1816 		struct mfc_cache *cache_proxy;
1817 
1818 		/* For an (*,G) entry, we only check that the incomming
1819 		 * interface is part of the static tree.
1820 		 */
1821 		cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1822 		if (cache_proxy &&
1823 		    cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1824 			goto forward;
1825 	}
1826 
1827 	/*
1828 	 * Wrong interface: drop packet and (maybe) send PIM assert.
1829 	 */
1830 	if (mrt->vif_table[vif].dev != skb->dev) {
1831 		if (rt_is_output_route(skb_rtable(skb))) {
1832 			/* It is our own packet, looped back.
1833 			 * Very complicated situation...
1834 			 *
1835 			 * The best workaround until routing daemons will be
1836 			 * fixed is not to redistribute packet, if it was
1837 			 * send through wrong interface. It means, that
1838 			 * multicast applications WILL NOT work for
1839 			 * (S,G), which have default multicast route pointing
1840 			 * to wrong oif. In any case, it is not a good
1841 			 * idea to use multicasting applications on router.
1842 			 */
1843 			goto dont_forward;
1844 		}
1845 
1846 		cache->mfc_un.res.wrong_if++;
1847 
1848 		if (true_vifi >= 0 && mrt->mroute_do_assert &&
1849 		    /* pimsm uses asserts, when switching from RPT to SPT,
1850 		     * so that we cannot check that packet arrived on an oif.
1851 		     * It is bad, but otherwise we would need to move pretty
1852 		     * large chunk of pimd to kernel. Ough... --ANK
1853 		     */
1854 		    (mrt->mroute_do_pim ||
1855 		     cache->mfc_un.res.ttls[true_vifi] < 255) &&
1856 		    time_after(jiffies,
1857 			       cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1858 			cache->mfc_un.res.last_assert = jiffies;
1859 			ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1860 		}
1861 		goto dont_forward;
1862 	}
1863 
1864 forward:
1865 	mrt->vif_table[vif].pkt_in++;
1866 	mrt->vif_table[vif].bytes_in += skb->len;
1867 
1868 	/*
1869 	 *	Forward the frame
1870 	 */
1871 	if (cache->mfc_origin == htonl(INADDR_ANY) &&
1872 	    cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1873 		if (true_vifi >= 0 &&
1874 		    true_vifi != cache->mfc_parent &&
1875 		    ip_hdr(skb)->ttl >
1876 				cache->mfc_un.res.ttls[cache->mfc_parent]) {
1877 			/* It's an (*,*) entry and the packet is not coming from
1878 			 * the upstream: forward the packet to the upstream
1879 			 * only.
1880 			 */
1881 			psend = cache->mfc_parent;
1882 			goto last_forward;
1883 		}
1884 		goto dont_forward;
1885 	}
1886 	for (ct = cache->mfc_un.res.maxvif - 1;
1887 	     ct >= cache->mfc_un.res.minvif; ct--) {
1888 		/* For (*,G) entry, don't forward to the incoming interface */
1889 		if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1890 		     ct != true_vifi) &&
1891 		    ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1892 			if (psend != -1) {
1893 				struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1894 
1895 				if (skb2)
1896 					ipmr_queue_xmit(net, mrt, skb2, cache,
1897 							psend);
1898 			}
1899 			psend = ct;
1900 		}
1901 	}
1902 last_forward:
1903 	if (psend != -1) {
1904 		if (local) {
1905 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1906 
1907 			if (skb2)
1908 				ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1909 		} else {
1910 			ipmr_queue_xmit(net, mrt, skb, cache, psend);
1911 			return;
1912 		}
1913 	}
1914 
1915 dont_forward:
1916 	if (!local)
1917 		kfree_skb(skb);
1918 }
1919 
1920 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1921 {
1922 	struct rtable *rt = skb_rtable(skb);
1923 	struct iphdr *iph = ip_hdr(skb);
1924 	struct flowi4 fl4 = {
1925 		.daddr = iph->daddr,
1926 		.saddr = iph->saddr,
1927 		.flowi4_tos = RT_TOS(iph->tos),
1928 		.flowi4_oif = (rt_is_output_route(rt) ?
1929 			       skb->dev->ifindex : 0),
1930 		.flowi4_iif = (rt_is_output_route(rt) ?
1931 			       LOOPBACK_IFINDEX :
1932 			       skb->dev->ifindex),
1933 		.flowi4_mark = skb->mark,
1934 	};
1935 	struct mr_table *mrt;
1936 	int err;
1937 
1938 	err = ipmr_fib_lookup(net, &fl4, &mrt);
1939 	if (err)
1940 		return ERR_PTR(err);
1941 	return mrt;
1942 }
1943 
1944 /*
1945  *	Multicast packets for forwarding arrive here
1946  *	Called with rcu_read_lock();
1947  */
1948 
1949 int ip_mr_input(struct sk_buff *skb)
1950 {
1951 	struct mfc_cache *cache;
1952 	struct net *net = dev_net(skb->dev);
1953 	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1954 	struct mr_table *mrt;
1955 
1956 	/* Packet is looped back after forward, it should not be
1957 	 * forwarded second time, but still can be delivered locally.
1958 	 */
1959 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
1960 		goto dont_forward;
1961 
1962 	mrt = ipmr_rt_fib_lookup(net, skb);
1963 	if (IS_ERR(mrt)) {
1964 		kfree_skb(skb);
1965 		return PTR_ERR(mrt);
1966 	}
1967 	if (!local) {
1968 		if (IPCB(skb)->opt.router_alert) {
1969 			if (ip_call_ra_chain(skb))
1970 				return 0;
1971 		} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1972 			/* IGMPv1 (and broken IGMPv2 implementations sort of
1973 			 * Cisco IOS <= 11.2(8)) do not put router alert
1974 			 * option to IGMP packets destined to routable
1975 			 * groups. It is very bad, because it means
1976 			 * that we can forward NO IGMP messages.
1977 			 */
1978 			struct sock *mroute_sk;
1979 
1980 			mroute_sk = rcu_dereference(mrt->mroute_sk);
1981 			if (mroute_sk) {
1982 				nf_reset(skb);
1983 				raw_rcv(mroute_sk, skb);
1984 				return 0;
1985 			}
1986 		    }
1987 	}
1988 
1989 	/* already under rcu_read_lock() */
1990 	cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1991 	if (cache == NULL) {
1992 		int vif = ipmr_find_vif(mrt, skb->dev);
1993 
1994 		if (vif >= 0)
1995 			cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1996 						    vif);
1997 	}
1998 
1999 	/*
2000 	 *	No usable cache entry
2001 	 */
2002 	if (cache == NULL) {
2003 		int vif;
2004 
2005 		if (local) {
2006 			struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2007 			ip_local_deliver(skb);
2008 			if (skb2 == NULL)
2009 				return -ENOBUFS;
2010 			skb = skb2;
2011 		}
2012 
2013 		read_lock(&mrt_lock);
2014 		vif = ipmr_find_vif(mrt, skb->dev);
2015 		if (vif >= 0) {
2016 			int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2017 			read_unlock(&mrt_lock);
2018 
2019 			return err2;
2020 		}
2021 		read_unlock(&mrt_lock);
2022 		kfree_skb(skb);
2023 		return -ENODEV;
2024 	}
2025 
2026 	read_lock(&mrt_lock);
2027 	ip_mr_forward(net, mrt, skb, cache, local);
2028 	read_unlock(&mrt_lock);
2029 
2030 	if (local)
2031 		return ip_local_deliver(skb);
2032 
2033 	return 0;
2034 
2035 dont_forward:
2036 	if (local)
2037 		return ip_local_deliver(skb);
2038 	kfree_skb(skb);
2039 	return 0;
2040 }
2041 
2042 #ifdef CONFIG_IP_PIMSM
2043 /* called with rcu_read_lock() */
2044 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2045 		     unsigned int pimlen)
2046 {
2047 	struct net_device *reg_dev = NULL;
2048 	struct iphdr *encap;
2049 
2050 	encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2051 	/*
2052 	 * Check that:
2053 	 * a. packet is really sent to a multicast group
2054 	 * b. packet is not a NULL-REGISTER
2055 	 * c. packet is not truncated
2056 	 */
2057 	if (!ipv4_is_multicast(encap->daddr) ||
2058 	    encap->tot_len == 0 ||
2059 	    ntohs(encap->tot_len) + pimlen > skb->len)
2060 		return 1;
2061 
2062 	read_lock(&mrt_lock);
2063 	if (mrt->mroute_reg_vif_num >= 0)
2064 		reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2065 	read_unlock(&mrt_lock);
2066 
2067 	if (reg_dev == NULL)
2068 		return 1;
2069 
2070 	skb->mac_header = skb->network_header;
2071 	skb_pull(skb, (u8 *)encap - skb->data);
2072 	skb_reset_network_header(skb);
2073 	skb->protocol = htons(ETH_P_IP);
2074 	skb->ip_summed = CHECKSUM_NONE;
2075 
2076 	skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2077 
2078 	netif_rx(skb);
2079 
2080 	return NET_RX_SUCCESS;
2081 }
2082 #endif
2083 
2084 #ifdef CONFIG_IP_PIMSM_V1
2085 /*
2086  * Handle IGMP messages of PIMv1
2087  */
2088 
2089 int pim_rcv_v1(struct sk_buff *skb)
2090 {
2091 	struct igmphdr *pim;
2092 	struct net *net = dev_net(skb->dev);
2093 	struct mr_table *mrt;
2094 
2095 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2096 		goto drop;
2097 
2098 	pim = igmp_hdr(skb);
2099 
2100 	mrt = ipmr_rt_fib_lookup(net, skb);
2101 	if (IS_ERR(mrt))
2102 		goto drop;
2103 	if (!mrt->mroute_do_pim ||
2104 	    pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2105 		goto drop;
2106 
2107 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2108 drop:
2109 		kfree_skb(skb);
2110 	}
2111 	return 0;
2112 }
2113 #endif
2114 
2115 #ifdef CONFIG_IP_PIMSM_V2
2116 static int pim_rcv(struct sk_buff *skb)
2117 {
2118 	struct pimreghdr *pim;
2119 	struct net *net = dev_net(skb->dev);
2120 	struct mr_table *mrt;
2121 
2122 	if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2123 		goto drop;
2124 
2125 	pim = (struct pimreghdr *)skb_transport_header(skb);
2126 	if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2127 	    (pim->flags & PIM_NULL_REGISTER) ||
2128 	    (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2129 	     csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2130 		goto drop;
2131 
2132 	mrt = ipmr_rt_fib_lookup(net, skb);
2133 	if (IS_ERR(mrt))
2134 		goto drop;
2135 	if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2136 drop:
2137 		kfree_skb(skb);
2138 	}
2139 	return 0;
2140 }
2141 #endif
2142 
2143 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2144 			      struct mfc_cache *c, struct rtmsg *rtm)
2145 {
2146 	int ct;
2147 	struct rtnexthop *nhp;
2148 	struct nlattr *mp_attr;
2149 	struct rta_mfc_stats mfcs;
2150 
2151 	/* If cache is unresolved, don't try to parse IIF and OIF */
2152 	if (c->mfc_parent >= MAXVIFS)
2153 		return -ENOENT;
2154 
2155 	if (VIF_EXISTS(mrt, c->mfc_parent) &&
2156 	    nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2157 		return -EMSGSIZE;
2158 
2159 	if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2160 		return -EMSGSIZE;
2161 
2162 	for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2163 		if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2164 			if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2165 				nla_nest_cancel(skb, mp_attr);
2166 				return -EMSGSIZE;
2167 			}
2168 
2169 			nhp->rtnh_flags = 0;
2170 			nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2171 			nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2172 			nhp->rtnh_len = sizeof(*nhp);
2173 		}
2174 	}
2175 
2176 	nla_nest_end(skb, mp_attr);
2177 
2178 	mfcs.mfcs_packets = c->mfc_un.res.pkt;
2179 	mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2180 	mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2181 	if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2182 		return -EMSGSIZE;
2183 
2184 	rtm->rtm_type = RTN_MULTICAST;
2185 	return 1;
2186 }
2187 
2188 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2189 		   __be32 saddr, __be32 daddr,
2190 		   struct rtmsg *rtm, int nowait)
2191 {
2192 	struct mfc_cache *cache;
2193 	struct mr_table *mrt;
2194 	int err;
2195 
2196 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2197 	if (mrt == NULL)
2198 		return -ENOENT;
2199 
2200 	rcu_read_lock();
2201 	cache = ipmr_cache_find(mrt, saddr, daddr);
2202 	if (cache == NULL && skb->dev) {
2203 		int vif = ipmr_find_vif(mrt, skb->dev);
2204 
2205 		if (vif >= 0)
2206 			cache = ipmr_cache_find_any(mrt, daddr, vif);
2207 	}
2208 	if (cache == NULL) {
2209 		struct sk_buff *skb2;
2210 		struct iphdr *iph;
2211 		struct net_device *dev;
2212 		int vif = -1;
2213 
2214 		if (nowait) {
2215 			rcu_read_unlock();
2216 			return -EAGAIN;
2217 		}
2218 
2219 		dev = skb->dev;
2220 		read_lock(&mrt_lock);
2221 		if (dev)
2222 			vif = ipmr_find_vif(mrt, dev);
2223 		if (vif < 0) {
2224 			read_unlock(&mrt_lock);
2225 			rcu_read_unlock();
2226 			return -ENODEV;
2227 		}
2228 		skb2 = skb_clone(skb, GFP_ATOMIC);
2229 		if (!skb2) {
2230 			read_unlock(&mrt_lock);
2231 			rcu_read_unlock();
2232 			return -ENOMEM;
2233 		}
2234 
2235 		skb_push(skb2, sizeof(struct iphdr));
2236 		skb_reset_network_header(skb2);
2237 		iph = ip_hdr(skb2);
2238 		iph->ihl = sizeof(struct iphdr) >> 2;
2239 		iph->saddr = saddr;
2240 		iph->daddr = daddr;
2241 		iph->version = 0;
2242 		err = ipmr_cache_unresolved(mrt, vif, skb2);
2243 		read_unlock(&mrt_lock);
2244 		rcu_read_unlock();
2245 		return err;
2246 	}
2247 
2248 	read_lock(&mrt_lock);
2249 	if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2250 		cache->mfc_flags |= MFC_NOTIFY;
2251 	err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2252 	read_unlock(&mrt_lock);
2253 	rcu_read_unlock();
2254 	return err;
2255 }
2256 
2257 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2258 			    u32 portid, u32 seq, struct mfc_cache *c, int cmd)
2259 {
2260 	struct nlmsghdr *nlh;
2261 	struct rtmsg *rtm;
2262 	int err;
2263 
2264 	nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI);
2265 	if (nlh == NULL)
2266 		return -EMSGSIZE;
2267 
2268 	rtm = nlmsg_data(nlh);
2269 	rtm->rtm_family   = RTNL_FAMILY_IPMR;
2270 	rtm->rtm_dst_len  = 32;
2271 	rtm->rtm_src_len  = 32;
2272 	rtm->rtm_tos      = 0;
2273 	rtm->rtm_table    = mrt->id;
2274 	if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2275 		goto nla_put_failure;
2276 	rtm->rtm_type     = RTN_MULTICAST;
2277 	rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2278 	if (c->mfc_flags & MFC_STATIC)
2279 		rtm->rtm_protocol = RTPROT_STATIC;
2280 	else
2281 		rtm->rtm_protocol = RTPROT_MROUTED;
2282 	rtm->rtm_flags    = 0;
2283 
2284 	if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2285 	    nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2286 		goto nla_put_failure;
2287 	err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2288 	/* do not break the dump if cache is unresolved */
2289 	if (err < 0 && err != -ENOENT)
2290 		goto nla_put_failure;
2291 
2292 	return nlmsg_end(skb, nlh);
2293 
2294 nla_put_failure:
2295 	nlmsg_cancel(skb, nlh);
2296 	return -EMSGSIZE;
2297 }
2298 
2299 static size_t mroute_msgsize(bool unresolved, int maxvif)
2300 {
2301 	size_t len =
2302 		NLMSG_ALIGN(sizeof(struct rtmsg))
2303 		+ nla_total_size(4)	/* RTA_TABLE */
2304 		+ nla_total_size(4)	/* RTA_SRC */
2305 		+ nla_total_size(4)	/* RTA_DST */
2306 		;
2307 
2308 	if (!unresolved)
2309 		len = len
2310 		      + nla_total_size(4)	/* RTA_IIF */
2311 		      + nla_total_size(0)	/* RTA_MULTIPATH */
2312 		      + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2313 						/* RTA_MFC_STATS */
2314 		      + nla_total_size(sizeof(struct rta_mfc_stats))
2315 		;
2316 
2317 	return len;
2318 }
2319 
2320 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2321 				 int cmd)
2322 {
2323 	struct net *net = read_pnet(&mrt->net);
2324 	struct sk_buff *skb;
2325 	int err = -ENOBUFS;
2326 
2327 	skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2328 			GFP_ATOMIC);
2329 	if (skb == NULL)
2330 		goto errout;
2331 
2332 	err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd);
2333 	if (err < 0)
2334 		goto errout;
2335 
2336 	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2337 	return;
2338 
2339 errout:
2340 	kfree_skb(skb);
2341 	if (err < 0)
2342 		rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2343 }
2344 
2345 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2346 {
2347 	struct net *net = sock_net(skb->sk);
2348 	struct mr_table *mrt;
2349 	struct mfc_cache *mfc;
2350 	unsigned int t = 0, s_t;
2351 	unsigned int h = 0, s_h;
2352 	unsigned int e = 0, s_e;
2353 
2354 	s_t = cb->args[0];
2355 	s_h = cb->args[1];
2356 	s_e = cb->args[2];
2357 
2358 	rcu_read_lock();
2359 	ipmr_for_each_table(mrt, net) {
2360 		if (t < s_t)
2361 			goto next_table;
2362 		if (t > s_t)
2363 			s_h = 0;
2364 		for (h = s_h; h < MFC_LINES; h++) {
2365 			list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2366 				if (e < s_e)
2367 					goto next_entry;
2368 				if (ipmr_fill_mroute(mrt, skb,
2369 						     NETLINK_CB(cb->skb).portid,
2370 						     cb->nlh->nlmsg_seq,
2371 						     mfc, RTM_NEWROUTE) < 0)
2372 					goto done;
2373 next_entry:
2374 				e++;
2375 			}
2376 			e = s_e = 0;
2377 		}
2378 		spin_lock_bh(&mfc_unres_lock);
2379 		list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2380 			if (e < s_e)
2381 				goto next_entry2;
2382 			if (ipmr_fill_mroute(mrt, skb,
2383 					     NETLINK_CB(cb->skb).portid,
2384 					     cb->nlh->nlmsg_seq,
2385 					     mfc, RTM_NEWROUTE) < 0) {
2386 				spin_unlock_bh(&mfc_unres_lock);
2387 				goto done;
2388 			}
2389 next_entry2:
2390 			e++;
2391 		}
2392 		spin_unlock_bh(&mfc_unres_lock);
2393 		e = s_e = 0;
2394 		s_h = 0;
2395 next_table:
2396 		t++;
2397 	}
2398 done:
2399 	rcu_read_unlock();
2400 
2401 	cb->args[2] = e;
2402 	cb->args[1] = h;
2403 	cb->args[0] = t;
2404 
2405 	return skb->len;
2406 }
2407 
2408 #ifdef CONFIG_PROC_FS
2409 /*
2410  *	The /proc interfaces to multicast routing :
2411  *	/proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2412  */
2413 struct ipmr_vif_iter {
2414 	struct seq_net_private p;
2415 	struct mr_table *mrt;
2416 	int ct;
2417 };
2418 
2419 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2420 					   struct ipmr_vif_iter *iter,
2421 					   loff_t pos)
2422 {
2423 	struct mr_table *mrt = iter->mrt;
2424 
2425 	for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2426 		if (!VIF_EXISTS(mrt, iter->ct))
2427 			continue;
2428 		if (pos-- == 0)
2429 			return &mrt->vif_table[iter->ct];
2430 	}
2431 	return NULL;
2432 }
2433 
2434 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2435 	__acquires(mrt_lock)
2436 {
2437 	struct ipmr_vif_iter *iter = seq->private;
2438 	struct net *net = seq_file_net(seq);
2439 	struct mr_table *mrt;
2440 
2441 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2442 	if (mrt == NULL)
2443 		return ERR_PTR(-ENOENT);
2444 
2445 	iter->mrt = mrt;
2446 
2447 	read_lock(&mrt_lock);
2448 	return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2449 		: SEQ_START_TOKEN;
2450 }
2451 
2452 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2453 {
2454 	struct ipmr_vif_iter *iter = seq->private;
2455 	struct net *net = seq_file_net(seq);
2456 	struct mr_table *mrt = iter->mrt;
2457 
2458 	++*pos;
2459 	if (v == SEQ_START_TOKEN)
2460 		return ipmr_vif_seq_idx(net, iter, 0);
2461 
2462 	while (++iter->ct < mrt->maxvif) {
2463 		if (!VIF_EXISTS(mrt, iter->ct))
2464 			continue;
2465 		return &mrt->vif_table[iter->ct];
2466 	}
2467 	return NULL;
2468 }
2469 
2470 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2471 	__releases(mrt_lock)
2472 {
2473 	read_unlock(&mrt_lock);
2474 }
2475 
2476 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2477 {
2478 	struct ipmr_vif_iter *iter = seq->private;
2479 	struct mr_table *mrt = iter->mrt;
2480 
2481 	if (v == SEQ_START_TOKEN) {
2482 		seq_puts(seq,
2483 			 "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2484 	} else {
2485 		const struct vif_device *vif = v;
2486 		const char *name =  vif->dev ? vif->dev->name : "none";
2487 
2488 		seq_printf(seq,
2489 			   "%2Zd %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2490 			   vif - mrt->vif_table,
2491 			   name, vif->bytes_in, vif->pkt_in,
2492 			   vif->bytes_out, vif->pkt_out,
2493 			   vif->flags, vif->local, vif->remote);
2494 	}
2495 	return 0;
2496 }
2497 
2498 static const struct seq_operations ipmr_vif_seq_ops = {
2499 	.start = ipmr_vif_seq_start,
2500 	.next  = ipmr_vif_seq_next,
2501 	.stop  = ipmr_vif_seq_stop,
2502 	.show  = ipmr_vif_seq_show,
2503 };
2504 
2505 static int ipmr_vif_open(struct inode *inode, struct file *file)
2506 {
2507 	return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2508 			    sizeof(struct ipmr_vif_iter));
2509 }
2510 
2511 static const struct file_operations ipmr_vif_fops = {
2512 	.owner	 = THIS_MODULE,
2513 	.open    = ipmr_vif_open,
2514 	.read    = seq_read,
2515 	.llseek  = seq_lseek,
2516 	.release = seq_release_net,
2517 };
2518 
2519 struct ipmr_mfc_iter {
2520 	struct seq_net_private p;
2521 	struct mr_table *mrt;
2522 	struct list_head *cache;
2523 	int ct;
2524 };
2525 
2526 
2527 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2528 					  struct ipmr_mfc_iter *it, loff_t pos)
2529 {
2530 	struct mr_table *mrt = it->mrt;
2531 	struct mfc_cache *mfc;
2532 
2533 	rcu_read_lock();
2534 	for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2535 		it->cache = &mrt->mfc_cache_array[it->ct];
2536 		list_for_each_entry_rcu(mfc, it->cache, list)
2537 			if (pos-- == 0)
2538 				return mfc;
2539 	}
2540 	rcu_read_unlock();
2541 
2542 	spin_lock_bh(&mfc_unres_lock);
2543 	it->cache = &mrt->mfc_unres_queue;
2544 	list_for_each_entry(mfc, it->cache, list)
2545 		if (pos-- == 0)
2546 			return mfc;
2547 	spin_unlock_bh(&mfc_unres_lock);
2548 
2549 	it->cache = NULL;
2550 	return NULL;
2551 }
2552 
2553 
2554 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2555 {
2556 	struct ipmr_mfc_iter *it = seq->private;
2557 	struct net *net = seq_file_net(seq);
2558 	struct mr_table *mrt;
2559 
2560 	mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2561 	if (mrt == NULL)
2562 		return ERR_PTR(-ENOENT);
2563 
2564 	it->mrt = mrt;
2565 	it->cache = NULL;
2566 	it->ct = 0;
2567 	return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2568 		: SEQ_START_TOKEN;
2569 }
2570 
2571 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2572 {
2573 	struct mfc_cache *mfc = v;
2574 	struct ipmr_mfc_iter *it = seq->private;
2575 	struct net *net = seq_file_net(seq);
2576 	struct mr_table *mrt = it->mrt;
2577 
2578 	++*pos;
2579 
2580 	if (v == SEQ_START_TOKEN)
2581 		return ipmr_mfc_seq_idx(net, seq->private, 0);
2582 
2583 	if (mfc->list.next != it->cache)
2584 		return list_entry(mfc->list.next, struct mfc_cache, list);
2585 
2586 	if (it->cache == &mrt->mfc_unres_queue)
2587 		goto end_of_list;
2588 
2589 	BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2590 
2591 	while (++it->ct < MFC_LINES) {
2592 		it->cache = &mrt->mfc_cache_array[it->ct];
2593 		if (list_empty(it->cache))
2594 			continue;
2595 		return list_first_entry(it->cache, struct mfc_cache, list);
2596 	}
2597 
2598 	/* exhausted cache_array, show unresolved */
2599 	rcu_read_unlock();
2600 	it->cache = &mrt->mfc_unres_queue;
2601 	it->ct = 0;
2602 
2603 	spin_lock_bh(&mfc_unres_lock);
2604 	if (!list_empty(it->cache))
2605 		return list_first_entry(it->cache, struct mfc_cache, list);
2606 
2607 end_of_list:
2608 	spin_unlock_bh(&mfc_unres_lock);
2609 	it->cache = NULL;
2610 
2611 	return NULL;
2612 }
2613 
2614 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2615 {
2616 	struct ipmr_mfc_iter *it = seq->private;
2617 	struct mr_table *mrt = it->mrt;
2618 
2619 	if (it->cache == &mrt->mfc_unres_queue)
2620 		spin_unlock_bh(&mfc_unres_lock);
2621 	else if (it->cache == &mrt->mfc_cache_array[it->ct])
2622 		rcu_read_unlock();
2623 }
2624 
2625 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2626 {
2627 	int n;
2628 
2629 	if (v == SEQ_START_TOKEN) {
2630 		seq_puts(seq,
2631 		 "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2632 	} else {
2633 		const struct mfc_cache *mfc = v;
2634 		const struct ipmr_mfc_iter *it = seq->private;
2635 		const struct mr_table *mrt = it->mrt;
2636 
2637 		seq_printf(seq, "%08X %08X %-3hd",
2638 			   (__force u32) mfc->mfc_mcastgrp,
2639 			   (__force u32) mfc->mfc_origin,
2640 			   mfc->mfc_parent);
2641 
2642 		if (it->cache != &mrt->mfc_unres_queue) {
2643 			seq_printf(seq, " %8lu %8lu %8lu",
2644 				   mfc->mfc_un.res.pkt,
2645 				   mfc->mfc_un.res.bytes,
2646 				   mfc->mfc_un.res.wrong_if);
2647 			for (n = mfc->mfc_un.res.minvif;
2648 			     n < mfc->mfc_un.res.maxvif; n++) {
2649 				if (VIF_EXISTS(mrt, n) &&
2650 				    mfc->mfc_un.res.ttls[n] < 255)
2651 					seq_printf(seq,
2652 					   " %2d:%-3d",
2653 					   n, mfc->mfc_un.res.ttls[n]);
2654 			}
2655 		} else {
2656 			/* unresolved mfc_caches don't contain
2657 			 * pkt, bytes and wrong_if values
2658 			 */
2659 			seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2660 		}
2661 		seq_putc(seq, '\n');
2662 	}
2663 	return 0;
2664 }
2665 
2666 static const struct seq_operations ipmr_mfc_seq_ops = {
2667 	.start = ipmr_mfc_seq_start,
2668 	.next  = ipmr_mfc_seq_next,
2669 	.stop  = ipmr_mfc_seq_stop,
2670 	.show  = ipmr_mfc_seq_show,
2671 };
2672 
2673 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2674 {
2675 	return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2676 			    sizeof(struct ipmr_mfc_iter));
2677 }
2678 
2679 static const struct file_operations ipmr_mfc_fops = {
2680 	.owner	 = THIS_MODULE,
2681 	.open    = ipmr_mfc_open,
2682 	.read    = seq_read,
2683 	.llseek  = seq_lseek,
2684 	.release = seq_release_net,
2685 };
2686 #endif
2687 
2688 #ifdef CONFIG_IP_PIMSM_V2
2689 static const struct net_protocol pim_protocol = {
2690 	.handler	=	pim_rcv,
2691 	.netns_ok	=	1,
2692 };
2693 #endif
2694 
2695 
2696 /*
2697  *	Setup for IP multicast routing
2698  */
2699 static int __net_init ipmr_net_init(struct net *net)
2700 {
2701 	int err;
2702 
2703 	err = ipmr_rules_init(net);
2704 	if (err < 0)
2705 		goto fail;
2706 
2707 #ifdef CONFIG_PROC_FS
2708 	err = -ENOMEM;
2709 	if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2710 		goto proc_vif_fail;
2711 	if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2712 		goto proc_cache_fail;
2713 #endif
2714 	return 0;
2715 
2716 #ifdef CONFIG_PROC_FS
2717 proc_cache_fail:
2718 	remove_proc_entry("ip_mr_vif", net->proc_net);
2719 proc_vif_fail:
2720 	ipmr_rules_exit(net);
2721 #endif
2722 fail:
2723 	return err;
2724 }
2725 
2726 static void __net_exit ipmr_net_exit(struct net *net)
2727 {
2728 #ifdef CONFIG_PROC_FS
2729 	remove_proc_entry("ip_mr_cache", net->proc_net);
2730 	remove_proc_entry("ip_mr_vif", net->proc_net);
2731 #endif
2732 	ipmr_rules_exit(net);
2733 }
2734 
2735 static struct pernet_operations ipmr_net_ops = {
2736 	.init = ipmr_net_init,
2737 	.exit = ipmr_net_exit,
2738 };
2739 
2740 int __init ip_mr_init(void)
2741 {
2742 	int err;
2743 
2744 	mrt_cachep = kmem_cache_create("ip_mrt_cache",
2745 				       sizeof(struct mfc_cache),
2746 				       0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2747 				       NULL);
2748 	if (!mrt_cachep)
2749 		return -ENOMEM;
2750 
2751 	err = register_pernet_subsys(&ipmr_net_ops);
2752 	if (err)
2753 		goto reg_pernet_fail;
2754 
2755 	err = register_netdevice_notifier(&ip_mr_notifier);
2756 	if (err)
2757 		goto reg_notif_fail;
2758 #ifdef CONFIG_IP_PIMSM_V2
2759 	if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2760 		pr_err("%s: can't add PIM protocol\n", __func__);
2761 		err = -EAGAIN;
2762 		goto add_proto_fail;
2763 	}
2764 #endif
2765 	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2766 		      NULL, ipmr_rtm_dumproute, NULL);
2767 	return 0;
2768 
2769 #ifdef CONFIG_IP_PIMSM_V2
2770 add_proto_fail:
2771 	unregister_netdevice_notifier(&ip_mr_notifier);
2772 #endif
2773 reg_notif_fail:
2774 	unregister_pernet_subsys(&ipmr_net_ops);
2775 reg_pernet_fail:
2776 	kmem_cache_destroy(mrt_cachep);
2777 	return err;
2778 }
2779