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, fl: flowi4_to_flowi(fl4: flp4));
162
163	err = fib_rules_lookup(net->ipv4.mr_rules_ops,
164	flowi4_to_flowi(fl4: flp4), flags: 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(net: rule->fr_net, id: 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(ptr: ops))
246	return PTR_ERR(ptr: ops);
247
248	INIT_LIST_HEAD(list: &net->ipv4.mr_tables);
249
250	mrt = ipmr_new_table(net, id: RT_TABLE_DEFAULT);
251	if (IS_ERR(ptr: mrt)) {
252	err = PTR_ERR(ptr: mrt);
253	goto err1;
254	}
255
256	err = fib_default_rule_add(ops, pref: 0x7fff, table: RT_TABLE_DEFAULT, flags: 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(entry: &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(new: &mrt->list, head: &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(error: -EINVAL);
404
405	mrt = ipmr_get_table(net, id);
406	if (mrt)
407	return mrt;
408
409	return mr_table_alloc(net, id, ops: &ipmr_mr_table_ops,
410	expire_func: ipmr_expire_process, table_set: ipmr_new_table_set);
411	}
412
413	static void ipmr_free_table(struct mr_table *mrt)
414	{
415	timer_shutdown_sync(timer: &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(hlt: &mrt->mfc_hash);
419	kfree(objp: 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(p: 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, name: "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(buf: p.name, fmt: "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, name: p.name);
466	if (!new_dev)
467	goto out;
468
469	new_dev->flags |= IFF_MULTICAST;
470	if (!ipmr_init_vif_indev(dev: new_dev))
471	goto out_unregister;
472	if (dev_open(dev: new_dev, NULL))
473	goto out_unregister;
474	dev_hold(dev: new_dev);
475	err = dev_set_allmulti(dev: new_dev, inc: 1);
476	if (err) {
477	dev_close(dev: new_dev);
478	tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
479	SIOCDELTUNNEL);
480	dev_put(dev: new_dev);
481	new_dev = ERR_PTR(error: err);
482	}
483	return new_dev;
484
485	out_unregister:
486	unregister_netdevice(dev: new_dev);
487	out:
488	return ERR_PTR(error: -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, flp4: &fl4, mrt: &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, pkt: 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(buf: name, fmt: "pimreg");
549	else
550	sprintf(buf: name, fmt: "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(addr: 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(vif: &mrt->vif_table[vif_num]);
601	if (!reg_dev)
602	return 1;
603
604	skb->mac_header = skb->network_header;
605	skb_pull(skb, len: (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, dev: reg_dev, net: dev_net(dev: 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	ipmr_seq: &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: &mfc->_c, tb_id, ipmr_seq: &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(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(lock: &mrt_lock);
667	call_ipmr_vif_entry_notifiers(net, event_type: FIB_EVENT_VIF_DEL, vif: v, vif_dev: dev,
668	vif_index: vifi, tb_id: 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(lock: &mrt_lock);
686
687	dev_set_allmulti(dev, inc: -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(net: dev_net(dev), RTM_NEWNETCONF,
693	type: NETCONFA_MC_FORWARDING,
694	ifindex: dev->ifindex, devconf: &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, tracker: &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(s: mrt_cachep, objp: (struct mfc_cache *)c);
710	}
711
712	static void ipmr_cache_free(struct mfc_cache *c)
713	{
714	call_rcu(head: &c->_c.rcu, func: 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(pnet: &mrt->net);
723	struct sk_buff *skb;
724	struct nlmsgerr *e;
725
726	atomic_dec(v: &mrt->cache_resolve_queue_len);
727
728	while ((skb = skb_dequeue(list: &c->_c.mfc_un.unres.unresolved))) {
729	if (ip_hdr(skb)->version == 0) {
730	struct nlmsghdr *nlh = skb_pull(skb,
731	len: sizeof(struct iphdr));
732	nlh->nlmsg_type = NLMSG_ERROR;
733	nlh->nlmsg_len = nlmsg_msg_size(payload: sizeof(struct nlmsgerr));
734	skb_trim(skb, len: 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(lock: &mfc_unres_lock)) {
757	mod_timer(timer: &mrt->ipmr_expire_timer, expires: jiffies+HZ/10);
758	return;
759	}
760
761	if (list_empty(head: &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(entry: &c->list);
776	mroute_netlink_event(mrt, mfc: (struct mfc_cache *)c, RTM_DELROUTE);
777	ipmr_destroy_unres(mrt, c: (struct mfc_cache *)c);
778	}
779
780	if (!list_empty(head: &mrt->mfc_unres_queue))
781	mod_timer(timer: &mrt->ipmr_expire_timer, expires: jiffies + expires);
782
783	out:
784	spin_unlock(lock: &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, inc: 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, v: vifc);
845	if (IS_ERR(ptr: dev))
846	return PTR_ERR(ptr: 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, ifindex: 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, addr: vifc->vifc_lcl_addr.s_addr);
858	}
859	if (!dev)
860	return -EADDRNOTAVAIL;
861	err = dev_set_allmulti(dev, inc: 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, type: NETCONFA_MC_FORWARDING,
878	ifindex: dev->ifindex, devconf: &in_dev->cnf);
879	ip_rt_multicast_event(in_dev);
880
881	/* Fill in the VIF structures */
882	vif_device_init(v, dev, rate_limit: vifc->vifc_rate_limit,
883	threshold: vifc->vifc_threshold,
884	flags: vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
885	get_iflink_mask: (VIFF_TUNNEL | VIFF_REGISTER));
886
887	err = dev_get_port_parent_id(dev, ppid: &ppid, recurse: 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(lock: &mrt_lock);
900	rcu_assign_pointer(v->dev, dev);
901	netdev_tracker_alloc(dev, tracker: &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(lock: &mrt_lock);
909	call_ipmr_vif_entry_notifiers(net, event_type: FIB_EVENT_VIF_ADD, vif: v, vif_dev: dev,
910	vif_index: vifi, tb_id: 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, hasharg: &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, hasharg: &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, hasharg: &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(k: 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(r: &c->_c.mfc_un.res.refcount, n: 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(k: mrt_cachep, GFP_ATOMIC);
971
972	if (c) {
973	skb_queue_head_init(list: &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(list: &uc->_c.mfc_un.unres.unresolved))) {
988	if (ip_hdr(skb)->version == 0) {
989	struct nlmsghdr *nlh = skb_pull(skb,
990	len: sizeof(struct iphdr));
991
992	if (mr_fill_mroute(mrt, skb, c: &c->_c,
993	rtm: 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(payload: sizeof(struct nlmsgerr));
999	skb_trim(skb, len: 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, dev: skb->dev, skb, cache: c, local: 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(skb: pkt);
1022	struct sock *mroute_sk;
1023	struct igmphdr *igmp;
1024	struct igmpmsg *msg;
1025	struct sk_buff *skb;
1026	int ret;
1027
1028	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1029	skb = skb_realloc_headroom(skb: pkt, headroom: sizeof(struct iphdr));
1030	else
1031	skb = alloc_skb(size: 128, GFP_ATOMIC);
1032
1033	if (!skb)
1034	return -ENOBUFS;
1035
1036	if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1037	/* Ugly, but we have no choice with this interface.
1038	* Duplicate old header, fix ihl, length etc.
1039	* And all this only to mangle msg->im_msgtype and
1040	* to set msg->im_mbz to "mbz" :-)
1041	*/
1042	skb_push(skb, len: sizeof(struct iphdr));
1043	skb_reset_network_header(skb);
1044	skb_reset_transport_header(skb);
1045	msg = (struct igmpmsg *)skb_network_header(skb);
1046	memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1047	msg->im_msgtype = assert;
1048	msg->im_mbz = 0;
1049	if (assert == IGMPMSG_WRVIFWHOLE) {
1050	msg->im_vif = vifi;
1051	msg->im_vif_hi = vifi >> 8;
1052	} else {
1053	/* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1054	int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1055
1056	msg->im_vif = vif_num;
1057	msg->im_vif_hi = vif_num >> 8;
1058	}
1059	ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1060	ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1061	sizeof(struct iphdr));
1062	} else {
1063	/* Copy the IP header */
1064	skb_set_network_header(skb, offset: skb->len);
1065	skb_put(skb, len: ihl);
1066	skb_copy_to_linear_data(skb, from: pkt->data, len: ihl);
1067	/* Flag to the kernel this is a route add */
1068	ip_hdr(skb)->protocol = 0;
1069	msg = (struct igmpmsg *)skb_network_header(skb);
1070	msg->im_vif = vifi;
1071	msg->im_vif_hi = vifi >> 8;
1072	skb_dst_set(skb, dst: dst_clone(dst: skb_dst(skb: pkt)));
1073	/* Add our header */
1074	igmp = skb_put(skb, len: sizeof(struct igmphdr));
1075	igmp->type = assert;
1076	msg->im_msgtype = assert;
1077	igmp->code = 0;
1078	ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1079	skb->transport_header = skb->network_header;
1080	}
1081
1082	mroute_sk = rcu_dereference(mrt->mroute_sk);
1083	if (!mroute_sk) {
1084	kfree_skb(skb);
1085	return -EINVAL;
1086	}
1087
1088	igmpmsg_netlink_event(mrt, pkt: skb);
1089
1090	/* Deliver to mrouted */
1091	ret = sock_queue_rcv_skb(sk: mroute_sk, skb);
1092
1093	if (ret < 0) {
1094	net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1095	kfree_skb(skb);
1096	}
1097
1098	return ret;
1099	}
1100
1101	/* Queue a packet for resolution. It gets locked cache entry! */
1102	/* Called under rcu_read_lock() */
1103	static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1104	struct sk_buff *skb, struct net_device *dev)
1105	{
1106	const struct iphdr *iph = ip_hdr(skb);
1107	struct mfc_cache *c;
1108	bool found = false;
1109	int err;
1110
1111	spin_lock_bh(lock: &mfc_unres_lock);
1112	list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1113	if (c->mfc_mcastgrp == iph->daddr &&
1114	c->mfc_origin == iph->saddr) {
1115	found = true;
1116	break;
1117	}
1118	}
1119
1120	if (!found) {
1121	/* Create a new entry if allowable */
1122	c = ipmr_cache_alloc_unres();
1123	if (!c) {
1124	spin_unlock_bh(lock: &mfc_unres_lock);
1125
1126	kfree_skb(skb);
1127	return -ENOBUFS;
1128	}
1129
1130	/* Fill in the new cache entry */
1131	c->_c.mfc_parent = -1;
1132	c->mfc_origin = iph->saddr;
1133	c->mfc_mcastgrp = iph->daddr;
1134
1135	/* Reflect first query at mrouted. */
1136	err = ipmr_cache_report(mrt, pkt: skb, vifi, IGMPMSG_NOCACHE);
1137
1138	if (err < 0) {
1139	/* If the report failed throw the cache entry
1140	out - Brad Parker
1141	*/
1142	spin_unlock_bh(lock: &mfc_unres_lock);
1143
1144	ipmr_cache_free(c);
1145	kfree_skb(skb);
1146	return err;
1147	}
1148
1149	atomic_inc(v: &mrt->cache_resolve_queue_len);
1150	list_add(new: &c->_c.list, head: &mrt->mfc_unres_queue);
1151	mroute_netlink_event(mrt, mfc: c, RTM_NEWROUTE);
1152
1153	if (atomic_read(v: &mrt->cache_resolve_queue_len) == 1)
1154	mod_timer(timer: &mrt->ipmr_expire_timer,
1155	expires: c->_c.mfc_un.unres.expires);
1156	}
1157
1158	/* See if we can append the packet */
1159	if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1160	kfree_skb(skb);
1161	err = -ENOBUFS;
1162	} else {
1163	if (dev) {
1164	skb->dev = dev;
1165	skb->skb_iif = dev->ifindex;
1166	}
1167	skb_queue_tail(list: &c->_c.mfc_un.unres.unresolved, newsk: skb);
1168	err = 0;
1169	}
1170
1171	spin_unlock_bh(lock: &mfc_unres_lock);
1172	return err;
1173	}
1174
1175	/* MFC cache manipulation by user space mroute daemon */
1176
1177	static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1178	{
1179	struct net *net = read_pnet(pnet: &mrt->net);
1180	struct mfc_cache *c;
1181
1182	/* The entries are added/deleted only under RTNL */
1183	rcu_read_lock();
1184	c = ipmr_cache_find_parent(mrt, origin: mfc->mfcc_origin.s_addr,
1185	mcastgrp: mfc->mfcc_mcastgrp.s_addr, parent);
1186	rcu_read_unlock();
1187	if (!c)
1188	return -ENOENT;
1189	rhltable_remove(hlt: &mrt->mfc_hash, list: &c->_c.mnode, params: ipmr_rht_params);
1190	list_del_rcu(entry: &c->_c.list);
1191	call_ipmr_mfc_entry_notifiers(net, event_type: FIB_EVENT_ENTRY_DEL, mfc: c, tb_id: mrt->id);
1192	mroute_netlink_event(mrt, mfc: c, RTM_DELROUTE);
1193	mr_cache_put(c: &c->_c);
1194
1195	return 0;
1196	}
1197
1198	static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1199	struct mfcctl *mfc, int mrtsock, int parent)
1200	{
1201	struct mfc_cache *uc, *c;
1202	struct mr_mfc *_uc;
1203	bool found;
1204	int ret;
1205
1206	if (mfc->mfcc_parent >= MAXVIFS)
1207	return -ENFILE;
1208
1209	/* The entries are added/deleted only under RTNL */
1210	rcu_read_lock();
1211	c = ipmr_cache_find_parent(mrt, origin: mfc->mfcc_origin.s_addr,
1212	mcastgrp: mfc->mfcc_mcastgrp.s_addr, parent);
1213	rcu_read_unlock();
1214	if (c) {
1215	spin_lock(lock: &mrt_lock);
1216	c->_c.mfc_parent = mfc->mfcc_parent;
1217	ipmr_update_thresholds(mrt, cache: &c->_c, ttls: mfc->mfcc_ttls);
1218	if (!mrtsock)
1219	c->_c.mfc_flags |= MFC_STATIC;
1220	spin_unlock(lock: &mrt_lock);
1221	call_ipmr_mfc_entry_notifiers(net, event_type: FIB_EVENT_ENTRY_REPLACE, mfc: c,
1222	tb_id: mrt->id);
1223	mroute_netlink_event(mrt, mfc: c, RTM_NEWROUTE);
1224	return 0;
1225	}
1226
1227	if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1228	!ipv4_is_multicast(addr: mfc->mfcc_mcastgrp.s_addr))
1229	return -EINVAL;
1230
1231	c = ipmr_cache_alloc();
1232	if (!c)
1233	return -ENOMEM;
1234
1235	c->mfc_origin = mfc->mfcc_origin.s_addr;
1236	c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1237	c->_c.mfc_parent = mfc->mfcc_parent;
1238	ipmr_update_thresholds(mrt, cache: &c->_c, ttls: mfc->mfcc_ttls);
1239	if (!mrtsock)
1240	c->_c.mfc_flags |= MFC_STATIC;
1241
1242	ret = rhltable_insert_key(hlt: &mrt->mfc_hash, key: &c->cmparg, list: &c->_c.mnode,
1243	params: ipmr_rht_params);
1244	if (ret) {
1245	pr_err("ipmr: rhtable insert error %d\n", ret);
1246	ipmr_cache_free(c);
1247	return ret;
1248	}
1249	list_add_tail_rcu(new: &c->_c.list, head: &mrt->mfc_cache_list);
1250	/* Check to see if we resolved a queued list. If so we
1251	* need to send on the frames and tidy up.
1252	*/
1253	found = false;
1254	spin_lock_bh(lock: &mfc_unres_lock);
1255	list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1256	uc = (struct mfc_cache *)_uc;
1257	if (uc->mfc_origin == c->mfc_origin &&
1258	uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1259	list_del(entry: &_uc->list);
1260	atomic_dec(v: &mrt->cache_resolve_queue_len);
1261	found = true;
1262	break;
1263	}
1264	}
1265	if (list_empty(head: &mrt->mfc_unres_queue))
1266	del_timer(timer: &mrt->ipmr_expire_timer);
1267	spin_unlock_bh(lock: &mfc_unres_lock);
1268
1269	if (found) {
1270	ipmr_cache_resolve(net, mrt, uc, c);
1271	ipmr_cache_free(c: uc);
1272	}
1273	call_ipmr_mfc_entry_notifiers(net, event_type: FIB_EVENT_ENTRY_ADD, mfc: c, tb_id: mrt->id);
1274	mroute_netlink_event(mrt, mfc: c, RTM_NEWROUTE);
1275	return 0;
1276	}
1277
1278	/* Close the multicast socket, and clear the vif tables etc */
1279	static void mroute_clean_tables(struct mr_table *mrt, int flags)
1280	{
1281	struct net *net = read_pnet(pnet: &mrt->net);
1282	struct mr_mfc *c, *tmp;
1283	struct mfc_cache *cache;
1284	LIST_HEAD(list);
1285	int i;
1286
1287	/* Shut down all active vif entries */
1288	if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1289	for (i = 0; i < mrt->maxvif; i++) {
1290	if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1291	!(flags & MRT_FLUSH_VIFS_STATIC)) ||
1292	(!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1293	continue;
1294	vif_delete(mrt, vifi: i, notify: 0, head: &list);
1295	}
1296	unregister_netdevice_many(head: &list);
1297	}
1298
1299	/* Wipe the cache */
1300	if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1301	list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1302	if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1303	(!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1304	continue;
1305	rhltable_remove(hlt: &mrt->mfc_hash, list: &c->mnode, params: ipmr_rht_params);
1306	list_del_rcu(entry: &c->list);
1307	cache = (struct mfc_cache *)c;
1308	call_ipmr_mfc_entry_notifiers(net, event_type: FIB_EVENT_ENTRY_DEL, mfc: cache,
1309	tb_id: mrt->id);
1310	mroute_netlink_event(mrt, mfc: cache, RTM_DELROUTE);
1311	mr_cache_put(c);
1312	}
1313	}
1314
1315	if (flags & MRT_FLUSH_MFC) {
1316	if (atomic_read(v: &mrt->cache_resolve_queue_len) != 0) {
1317	spin_lock_bh(lock: &mfc_unres_lock);
1318	list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1319	list_del(entry: &c->list);
1320	cache = (struct mfc_cache *)c;
1321	mroute_netlink_event(mrt, mfc: cache, RTM_DELROUTE);
1322	ipmr_destroy_unres(mrt, c: cache);
1323	}
1324	spin_unlock_bh(lock: &mfc_unres_lock);
1325	}
1326	}
1327	}
1328
1329	/* called from ip_ra_control(), before an RCU grace period,
1330	* we don't need to call synchronize_rcu() here
1331	*/
1332	static void mrtsock_destruct(struct sock *sk)
1333	{
1334	struct net *net = sock_net(sk);
1335	struct mr_table *mrt;
1336
1337	rtnl_lock();
1338	ipmr_for_each_table(mrt, net) {
1339	if (sk == rtnl_dereference(mrt->mroute_sk)) {
1340	IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1341	inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1342	type: NETCONFA_MC_FORWARDING,
1343	NETCONFA_IFINDEX_ALL,
1344	devconf: net->ipv4.devconf_all);
1345	RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1346	mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1347	}
1348	}
1349	rtnl_unlock();
1350	}
1351
1352	/* Socket options and virtual interface manipulation. The whole
1353	* virtual interface system is a complete heap, but unfortunately
1354	* that's how BSD mrouted happens to think. Maybe one day with a proper
1355	* MOSPF/PIM router set up we can clean this up.
1356	*/
1357
1358	int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1359	unsigned int optlen)
1360	{
1361	struct net *net = sock_net(sk);
1362	int val, ret = 0, parent = 0;
1363	struct mr_table *mrt;
1364	struct vifctl vif;
1365	struct mfcctl mfc;
1366	bool do_wrvifwhole;
1367	u32 uval;
1368
1369	/* There's one exception to the lock - MRT_DONE which needs to unlock */
1370	rtnl_lock();
1371	if (sk->sk_type != SOCK_RAW ||
1372	inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1373	ret = -EOPNOTSUPP;
1374	goto out_unlock;
1375	}
1376
1377	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1378	if (!mrt) {
1379	ret = -ENOENT;
1380	goto out_unlock;
1381	}
1382	if (optname != MRT_INIT) {
1383	if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1384	!ns_capable(ns: net->user_ns, CAP_NET_ADMIN)) {
1385	ret = -EACCES;
1386	goto out_unlock;
1387	}
1388	}
1389
1390	switch (optname) {
1391	case MRT_INIT:
1392	if (optlen != sizeof(int)) {
1393	ret = -EINVAL;
1394	break;
1395	}
1396	if (rtnl_dereference(mrt->mroute_sk)) {
1397	ret = -EADDRINUSE;
1398	break;
1399	}
1400
1401	ret = ip_ra_control(sk, on: 1, destructor: mrtsock_destruct);
1402	if (ret == 0) {
1403	rcu_assign_pointer(mrt->mroute_sk, sk);
1404	IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1405	inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1406	type: NETCONFA_MC_FORWARDING,
1407	NETCONFA_IFINDEX_ALL,
1408	devconf: net->ipv4.devconf_all);
1409	}
1410	break;
1411	case MRT_DONE:
1412	if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1413	ret = -EACCES;
1414	} else {
1415	/* We need to unlock here because mrtsock_destruct takes
1416	* care of rtnl itself and we can't change that due to
1417	* the IP_ROUTER_ALERT setsockopt which runs without it.
1418	*/
1419	rtnl_unlock();
1420	ret = ip_ra_control(sk, on: 0, NULL);
1421	goto out;
1422	}
1423	break;
1424	case MRT_ADD_VIF:
1425	case MRT_DEL_VIF:
1426	if (optlen != sizeof(vif)) {
1427	ret = -EINVAL;
1428	break;
1429	}
1430	if (copy_from_sockptr(dst: &vif, src: optval, size: sizeof(vif))) {
1431	ret = -EFAULT;
1432	break;
1433	}
1434	if (vif.vifc_vifi >= MAXVIFS) {
1435	ret = -ENFILE;
1436	break;
1437	}
1438	if (optname == MRT_ADD_VIF) {
1439	ret = vif_add(net, mrt, vifc: &vif,
1440	mrtsock: sk == rtnl_dereference(mrt->mroute_sk));
1441	} else {
1442	ret = vif_delete(mrt, vifi: vif.vifc_vifi, notify: 0, NULL);
1443	}
1444	break;
1445	/* Manipulate the forwarding caches. These live
1446	* in a sort of kernel/user symbiosis.
1447	*/
1448	case MRT_ADD_MFC:
1449	case MRT_DEL_MFC:
1450	parent = -1;
1451	fallthrough;
1452	case MRT_ADD_MFC_PROXY:
1453	case MRT_DEL_MFC_PROXY:
1454	if (optlen != sizeof(mfc)) {
1455	ret = -EINVAL;
1456	break;
1457	}
1458	if (copy_from_sockptr(dst: &mfc, src: optval, size: sizeof(mfc))) {
1459	ret = -EFAULT;
1460	break;
1461	}
1462	if (parent == 0)
1463	parent = mfc.mfcc_parent;
1464	if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1465	ret = ipmr_mfc_delete(mrt, mfc: &mfc, parent);
1466	else
1467	ret = ipmr_mfc_add(net, mrt, mfc: &mfc,
1468	mrtsock: sk == rtnl_dereference(mrt->mroute_sk),
1469	parent);
1470	break;
1471	case MRT_FLUSH:
1472	if (optlen != sizeof(val)) {
1473	ret = -EINVAL;
1474	break;
1475	}
1476	if (copy_from_sockptr(dst: &val, src: optval, size: sizeof(val))) {
1477	ret = -EFAULT;
1478	break;
1479	}
1480	mroute_clean_tables(mrt, flags: val);
1481	break;
1482	/* Control PIM assert. */
1483	case MRT_ASSERT:
1484	if (optlen != sizeof(val)) {
1485	ret = -EINVAL;
1486	break;
1487	}
1488	if (copy_from_sockptr(dst: &val, src: optval, size: sizeof(val))) {
1489	ret = -EFAULT;
1490	break;
1491	}
1492	mrt->mroute_do_assert = val;
1493	break;
1494	case MRT_PIM:
1495	if (!ipmr_pimsm_enabled()) {
1496	ret = -ENOPROTOOPT;
1497	break;
1498	}
1499	if (optlen != sizeof(val)) {
1500	ret = -EINVAL;
1501	break;
1502	}
1503	if (copy_from_sockptr(dst: &val, src: optval, size: sizeof(val))) {
1504	ret = -EFAULT;
1505	break;
1506	}
1507
1508	do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1509	val = !!val;
1510	if (val != mrt->mroute_do_pim) {
1511	mrt->mroute_do_pim = val;
1512	mrt->mroute_do_assert = val;
1513	mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1514	}
1515	break;
1516	case MRT_TABLE:
1517	if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1518	ret = -ENOPROTOOPT;
1519	break;
1520	}
1521	if (optlen != sizeof(uval)) {
1522	ret = -EINVAL;
1523	break;
1524	}
1525	if (copy_from_sockptr(dst: &uval, src: optval, size: sizeof(uval))) {
1526	ret = -EFAULT;
1527	break;
1528	}
1529
1530	if (sk == rtnl_dereference(mrt->mroute_sk)) {
1531	ret = -EBUSY;
1532	} else {
1533	mrt = ipmr_new_table(net, id: uval);
1534	if (IS_ERR(ptr: mrt))
1535	ret = PTR_ERR(ptr: mrt);
1536	else
1537	raw_sk(sk)->ipmr_table = uval;
1538	}
1539	break;
1540	/* Spurious command, or MRT_VERSION which you cannot set. */
1541	default:
1542	ret = -ENOPROTOOPT;
1543	}
1544	out_unlock:
1545	rtnl_unlock();
1546	out:
1547	return ret;
1548	}
1549
1550	/* Execute if this ioctl is a special mroute ioctl */
1551	int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1552	{
1553	switch (cmd) {
1554	/* These userspace buffers will be consumed by ipmr_ioctl() */
1555	case SIOCGETVIFCNT: {
1556	struct sioc_vif_req buffer;
1557
1558	return sock_ioctl_inout(sk, cmd, arg, karg: &buffer,
1559	size: sizeof(buffer));
1560	}
1561	case SIOCGETSGCNT: {
1562	struct sioc_sg_req buffer;
1563
1564	return sock_ioctl_inout(sk, cmd, arg, karg: &buffer,
1565	size: sizeof(buffer));
1566	}
1567	}
1568	/* return code > 0 means that the ioctl was not executed */
1569	return 1;
1570	}
1571
1572	/* Getsock opt support for the multicast routing system. */
1573	int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1574	sockptr_t optlen)
1575	{
1576	int olr;
1577	int val;
1578	struct net *net = sock_net(sk);
1579	struct mr_table *mrt;
1580
1581	if (sk->sk_type != SOCK_RAW ||
1582	inet_sk(sk)->inet_num != IPPROTO_IGMP)
1583	return -EOPNOTSUPP;
1584
1585	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1586	if (!mrt)
1587	return -ENOENT;
1588
1589	switch (optname) {
1590	case MRT_VERSION:
1591	val = 0x0305;
1592	break;
1593	case MRT_PIM:
1594	if (!ipmr_pimsm_enabled())
1595	return -ENOPROTOOPT;
1596	val = mrt->mroute_do_pim;
1597	break;
1598	case MRT_ASSERT:
1599	val = mrt->mroute_do_assert;
1600	break;
1601	default:
1602	return -ENOPROTOOPT;
1603	}
1604
1605	if (copy_from_sockptr(dst: &olr, src: optlen, size: sizeof(int)))
1606	return -EFAULT;
1607	olr = min_t(unsigned int, olr, sizeof(int));
1608	if (olr < 0)
1609	return -EINVAL;
1610	if (copy_to_sockptr(dst: optlen, src: &olr, size: sizeof(int)))
1611	return -EFAULT;
1612	if (copy_to_sockptr(dst: optval, src: &val, size: olr))
1613	return -EFAULT;
1614	return 0;
1615	}
1616
1617	/* The IP multicast ioctl support routines. */
1618	int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1619	{
1620	struct vif_device *vif;
1621	struct mfc_cache *c;
1622	struct net *net = sock_net(sk);
1623	struct sioc_vif_req *vr;
1624	struct sioc_sg_req *sr;
1625	struct mr_table *mrt;
1626
1627	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1628	if (!mrt)
1629	return -ENOENT;
1630
1631	switch (cmd) {
1632	case SIOCGETVIFCNT:
1633	vr = (struct sioc_vif_req *)arg;
1634	if (vr->vifi >= mrt->maxvif)
1635	return -EINVAL;
1636	vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1637	rcu_read_lock();
1638	vif = &mrt->vif_table[vr->vifi];
1639	if (VIF_EXISTS(mrt, vr->vifi)) {
1640	vr->icount = READ_ONCE(vif->pkt_in);
1641	vr->ocount = READ_ONCE(vif->pkt_out);
1642	vr->ibytes = READ_ONCE(vif->bytes_in);
1643	vr->obytes = READ_ONCE(vif->bytes_out);
1644	rcu_read_unlock();
1645
1646	return 0;
1647	}
1648	rcu_read_unlock();
1649	return -EADDRNOTAVAIL;
1650	case SIOCGETSGCNT:
1651	sr = (struct sioc_sg_req *)arg;
1652
1653	rcu_read_lock();
1654	c = ipmr_cache_find(mrt, origin: sr->src.s_addr, mcastgrp: sr->grp.s_addr);
1655	if (c) {
1656	sr->pktcnt = c->_c.mfc_un.res.pkt;
1657	sr->bytecnt = c->_c.mfc_un.res.bytes;
1658	sr->wrong_if = c->_c.mfc_un.res.wrong_if;
1659	rcu_read_unlock();
1660	return 0;
1661	}
1662	rcu_read_unlock();
1663	return -EADDRNOTAVAIL;
1664	default:
1665	return -ENOIOCTLCMD;
1666	}
1667	}
1668
1669	#ifdef CONFIG_COMPAT
1670	struct compat_sioc_sg_req {
1671	struct in_addr src;
1672	struct in_addr grp;
1673	compat_ulong_t pktcnt;
1674	compat_ulong_t bytecnt;
1675	compat_ulong_t wrong_if;
1676	};
1677
1678	struct compat_sioc_vif_req {
1679	vifi_t vifi; /* Which iface */
1680	compat_ulong_t icount;
1681	compat_ulong_t ocount;
1682	compat_ulong_t ibytes;
1683	compat_ulong_t obytes;
1684	};
1685
1686	int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1687	{
1688	struct compat_sioc_sg_req sr;
1689	struct compat_sioc_vif_req vr;
1690	struct vif_device *vif;
1691	struct mfc_cache *c;
1692	struct net *net = sock_net(sk);
1693	struct mr_table *mrt;
1694
1695	mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1696	if (!mrt)
1697	return -ENOENT;
1698
1699	switch (cmd) {
1700	case SIOCGETVIFCNT:
1701	if (copy_from_user(to: &vr, from: arg, n: sizeof(vr)))
1702	return -EFAULT;
1703	if (vr.vifi >= mrt->maxvif)
1704	return -EINVAL;
1705	vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1706	rcu_read_lock();
1707	vif = &mrt->vif_table[vr.vifi];
1708	if (VIF_EXISTS(mrt, vr.vifi)) {
1709	vr.icount = READ_ONCE(vif->pkt_in);
1710	vr.ocount = READ_ONCE(vif->pkt_out);
1711	vr.ibytes = READ_ONCE(vif->bytes_in);
1712	vr.obytes = READ_ONCE(vif->bytes_out);
1713	rcu_read_unlock();
1714
1715	if (copy_to_user(to: arg, from: &vr, n: sizeof(vr)))
1716	return -EFAULT;
1717	return 0;
1718	}
1719	rcu_read_unlock();
1720	return -EADDRNOTAVAIL;
1721	case SIOCGETSGCNT:
1722	if (copy_from_user(to: &sr, from: arg, n: sizeof(sr)))
1723	return -EFAULT;
1724
1725	rcu_read_lock();
1726	c = ipmr_cache_find(mrt, origin: sr.src.s_addr, mcastgrp: sr.grp.s_addr);
1727	if (c) {
1728	sr.pktcnt = c->_c.mfc_un.res.pkt;
1729	sr.bytecnt = c->_c.mfc_un.res.bytes;
1730	sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1731	rcu_read_unlock();
1732
1733	if (copy_to_user(to: arg, from: &sr, n: sizeof(sr)))
1734	return -EFAULT;
1735	return 0;
1736	}
1737	rcu_read_unlock();
1738	return -EADDRNOTAVAIL;
1739	default:
1740	return -ENOIOCTLCMD;
1741	}
1742	}
1743	#endif
1744
1745	static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1746	{
1747	struct net_device *dev = netdev_notifier_info_to_dev(info: ptr);
1748	struct net *net = dev_net(dev);
1749	struct mr_table *mrt;
1750	struct vif_device *v;
1751	int ct;
1752
1753	if (event != NETDEV_UNREGISTER)
1754	return NOTIFY_DONE;
1755
1756	ipmr_for_each_table(mrt, net) {
1757	v = &mrt->vif_table[0];
1758	for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1759	if (rcu_access_pointer(v->dev) == dev)
1760	vif_delete(mrt, vifi: ct, notify: 1, NULL);
1761	}
1762	}
1763	return NOTIFY_DONE;
1764	}
1765
1766	static struct notifier_block ip_mr_notifier = {
1767	.notifier_call = ipmr_device_event,
1768	};
1769
1770	/* Encapsulate a packet by attaching a valid IPIP header to it.
1771	* This avoids tunnel drivers and other mess and gives us the speed so
1772	* important for multicast video.
1773	*/
1774	static void ip_encap(struct net *net, struct sk_buff *skb,
1775	__be32 saddr, __be32 daddr)
1776	{
1777	struct iphdr *iph;
1778	const struct iphdr *old_iph = ip_hdr(skb);
1779
1780	skb_push(skb, len: sizeof(struct iphdr));
1781	skb->transport_header = skb->network_header;
1782	skb_reset_network_header(skb);
1783	iph = ip_hdr(skb);
1784
1785	iph->version = 4;
1786	iph->tos = old_iph->tos;
1787	iph->ttl = old_iph->ttl;
1788	iph->frag_off = 0;
1789	iph->daddr = daddr;
1790	iph->saddr = saddr;
1791	iph->protocol = IPPROTO_IPIP;
1792	iph->ihl = 5;
1793	iph->tot_len = htons(skb->len);
1794	ip_select_ident(net, skb, NULL);
1795	ip_send_check(ip: iph);
1796
1797	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1798	nf_reset_ct(skb);
1799	}
1800
1801	static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1802	struct sk_buff *skb)
1803	{
1804	struct ip_options *opt = &(IPCB(skb)->opt);
1805
1806	IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1807
1808	if (unlikely(opt->optlen))
1809	ip_forward_options(skb);
1810
1811	return dst_output(net, sk, skb);
1812	}
1813
1814	#ifdef CONFIG_NET_SWITCHDEV
1815	static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1816	int in_vifi, int out_vifi)
1817	{
1818	struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1819	struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1820
1821	if (!skb->offload_l3_fwd_mark)
1822	return false;
1823	if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1824	return false;
1825	return netdev_phys_item_id_same(a: &out_vif->dev_parent_id,
1826	b: &in_vif->dev_parent_id);
1827	}
1828	#else
1829	static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1830	int in_vifi, int out_vifi)
1831	{
1832	return false;
1833	}
1834	#endif
1835
1836	/* Processing handlers for ipmr_forward, under rcu_read_lock() */
1837
1838	static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1839	int in_vifi, struct sk_buff *skb, int vifi)
1840	{
1841	const struct iphdr *iph = ip_hdr(skb);
1842	struct vif_device *vif = &mrt->vif_table[vifi];
1843	struct net_device *vif_dev;
1844	struct net_device *dev;
1845	struct rtable *rt;
1846	struct flowi4 fl4;
1847	int encap = 0;
1848
1849	vif_dev = vif_dev_read(vif);
1850	if (!vif_dev)
1851	goto out_free;
1852
1853	if (vif->flags & VIFF_REGISTER) {
1854	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1855	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1856	DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1857	DEV_STATS_INC(vif_dev, tx_packets);
1858	ipmr_cache_report(mrt, pkt: skb, vifi, IGMPMSG_WHOLEPKT);
1859	goto out_free;
1860	}
1861
1862	if (ipmr_forward_offloaded(skb, mrt, in_vifi, out_vifi: vifi))
1863	goto out_free;
1864
1865	if (vif->flags & VIFF_TUNNEL) {
1866	rt = ip_route_output_ports(net, fl4: &fl4, NULL,
1867	daddr: vif->remote, saddr: vif->local,
1868	dport: 0, sport: 0,
1869	IPPROTO_IPIP,
1870	RT_TOS(iph->tos), oif: vif->link);
1871	if (IS_ERR(ptr: rt))
1872	goto out_free;
1873	encap = sizeof(struct iphdr);
1874	} else {
1875	rt = ip_route_output_ports(net, fl4: &fl4, NULL, daddr: iph->daddr, saddr: 0,
1876	dport: 0, sport: 0,
1877	IPPROTO_IPIP,
1878	RT_TOS(iph->tos), oif: vif->link);
1879	if (IS_ERR(ptr: rt))
1880	goto out_free;
1881	}
1882
1883	dev = rt->dst.dev;
1884
1885	if (skb->len+encap > dst_mtu(dst: &rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1886	/* Do not fragment multicasts. Alas, IPv4 does not
1887	* allow to send ICMP, so that packets will disappear
1888	* to blackhole.
1889	*/
1890	IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1891	ip_rt_put(rt);
1892	goto out_free;
1893	}
1894
1895	encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1896
1897	if (skb_cow(skb, headroom: encap)) {
1898	ip_rt_put(rt);
1899	goto out_free;
1900	}
1901
1902	WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1903	WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1904
1905	skb_dst_drop(skb);
1906	skb_dst_set(skb, dst: &rt->dst);
1907	ip_decrease_ttl(iph: ip_hdr(skb));
1908
1909	/* FIXME: forward and output firewalls used to be called here.
1910	* What do we do with netfilter? -- RR
1911	*/
1912	if (vif->flags & VIFF_TUNNEL) {
1913	ip_encap(net, skb, saddr: vif->local, daddr: vif->remote);
1914	/* FIXME: extra output firewall step used to be here. --RR */
1915	DEV_STATS_INC(vif_dev, tx_packets);
1916	DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1917	}
1918
1919	IPCB(skb)->flags |= IPSKB_FORWARDED;
1920
1921	/* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1922	* not only before forwarding, but after forwarding on all output
1923	* interfaces. It is clear, if mrouter runs a multicasting
1924	* program, it should receive packets not depending to what interface
1925	* program is joined.
1926	* If we will not make it, the program will have to join on all
1927	* interfaces. On the other hand, multihoming host (or router, but
1928	* not mrouter) cannot join to more than one interface - it will
1929	* result in receiving multiple packets.
1930	*/
1931	NF_HOOK(pf: NFPROTO_IPV4, hook: NF_INET_FORWARD,
1932	net, NULL, skb, in: skb->dev, out: dev,
1933	okfn: ipmr_forward_finish);
1934	return;
1935
1936	out_free:
1937	kfree_skb(skb);
1938	}
1939
1940	/* Called with mrt_lock or rcu_read_lock() */
1941	static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1942	{
1943	int ct;
1944	/* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1945	for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1946	if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1947	break;
1948	}
1949	return ct;
1950	}
1951
1952	/* "local" means that we should preserve one skb (for local delivery) */
1953	/* Called uner rcu_read_lock() */
1954	static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1955	struct net_device *dev, struct sk_buff *skb,
1956	struct mfc_cache *c, int local)
1957	{
1958	int true_vifi = ipmr_find_vif(mrt, dev);
1959	int psend = -1;
1960	int vif, ct;
1961
1962	vif = c->_c.mfc_parent;
1963	c->_c.mfc_un.res.pkt++;
1964	c->_c.mfc_un.res.bytes += skb->len;
1965	c->_c.mfc_un.res.lastuse = jiffies;
1966
1967	if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1968	struct mfc_cache *cache_proxy;
1969
1970	/* For an (*,G) entry, we only check that the incoming
1971	* interface is part of the static tree.
1972	*/
1973	cache_proxy = mr_mfc_find_any_parent(mrt, vifi: vif);
1974	if (cache_proxy &&
1975	cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1976	goto forward;
1977	}
1978
1979	/* Wrong interface: drop packet and (maybe) send PIM assert. */
1980	if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
1981	if (rt_is_output_route(rt: skb_rtable(skb))) {
1982	/* It is our own packet, looped back.
1983	* Very complicated situation...
1984	*
1985	* The best workaround until routing daemons will be
1986	* fixed is not to redistribute packet, if it was
1987	* send through wrong interface. It means, that
1988	* multicast applications WILL NOT work for
1989	* (S,G), which have default multicast route pointing
1990	* to wrong oif. In any case, it is not a good
1991	* idea to use multicasting applications on router.
1992	*/
1993	goto dont_forward;
1994	}
1995
1996	c->_c.mfc_un.res.wrong_if++;
1997
1998	if (true_vifi >= 0 && mrt->mroute_do_assert &&
1999	/* pimsm uses asserts, when switching from RPT to SPT,
2000	* so that we cannot check that packet arrived on an oif.
2001	* It is bad, but otherwise we would need to move pretty
2002	* large chunk of pimd to kernel. Ough... --ANK
2003	*/
2004	(mrt->mroute_do_pim ||
2005	c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2006	time_after(jiffies,
2007	c->_c.mfc_un.res.last_assert +
2008	MFC_ASSERT_THRESH)) {
2009	c->_c.mfc_un.res.last_assert = jiffies;
2010	ipmr_cache_report(mrt, pkt: skb, vifi: true_vifi, IGMPMSG_WRONGVIF);
2011	if (mrt->mroute_do_wrvifwhole)
2012	ipmr_cache_report(mrt, pkt: skb, vifi: true_vifi,
2013	IGMPMSG_WRVIFWHOLE);
2014	}
2015	goto dont_forward;
2016	}
2017
2018	forward:
2019	WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2020	mrt->vif_table[vif].pkt_in + 1);
2021	WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2022	mrt->vif_table[vif].bytes_in + skb->len);
2023
2024	/* Forward the frame */
2025	if (c->mfc_origin == htonl(INADDR_ANY) &&
2026	c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2027	if (true_vifi >= 0 &&
2028	true_vifi != c->_c.mfc_parent &&
2029	ip_hdr(skb)->ttl >
2030	c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2031	/* It's an (*,*) entry and the packet is not coming from
2032	* the upstream: forward the packet to the upstream
2033	* only.
2034	*/
2035	psend = c->_c.mfc_parent;
2036	goto last_forward;
2037	}
2038	goto dont_forward;
2039	}
2040	for (ct = c->_c.mfc_un.res.maxvif - 1;
2041	ct >= c->_c.mfc_un.res.minvif; ct--) {
2042	/* For (*,G) entry, don't forward to the incoming interface */
2043	if ((c->mfc_origin != htonl(INADDR_ANY) ||
2044	ct != true_vifi) &&
2045	ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2046	if (psend != -1) {
2047	struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2048
2049	if (skb2)
2050	ipmr_queue_xmit(net, mrt, in_vifi: true_vifi,
2051	skb: skb2, vifi: psend);
2052	}
2053	psend = ct;
2054	}
2055	}
2056	last_forward:
2057	if (psend != -1) {
2058	if (local) {
2059	struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2060
2061	if (skb2)
2062	ipmr_queue_xmit(net, mrt, in_vifi: true_vifi, skb: skb2,
2063	vifi: psend);
2064	} else {
2065	ipmr_queue_xmit(net, mrt, in_vifi: true_vifi, skb, vifi: psend);
2066	return;
2067	}
2068	}
2069
2070	dont_forward:
2071	if (!local)
2072	kfree_skb(skb);
2073	}
2074
2075	static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2076	{
2077	struct rtable *rt = skb_rtable(skb);
2078	struct iphdr *iph = ip_hdr(skb);
2079	struct flowi4 fl4 = {
2080	.daddr = iph->daddr,
2081	.saddr = iph->saddr,
2082	.flowi4_tos = RT_TOS(iph->tos),
2083	.flowi4_oif = (rt_is_output_route(rt) ?
2084	skb->dev->ifindex : 0),
2085	.flowi4_iif = (rt_is_output_route(rt) ?
2086	LOOPBACK_IFINDEX :
2087	skb->dev->ifindex),
2088	.flowi4_mark = skb->mark,
2089	};
2090	struct mr_table *mrt;
2091	int err;
2092
2093	err = ipmr_fib_lookup(net, flp4: &fl4, mrt: &mrt);
2094	if (err)
2095	return ERR_PTR(error: err);
2096	return mrt;
2097	}
2098
2099	/* Multicast packets for forwarding arrive here
2100	* Called with rcu_read_lock();
2101	*/
2102	int ip_mr_input(struct sk_buff *skb)
2103	{
2104	struct mfc_cache *cache;
2105	struct net *net = dev_net(dev: skb->dev);
2106	int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2107	struct mr_table *mrt;
2108	struct net_device *dev;
2109
2110	/* skb->dev passed in is the loX master dev for vrfs.
2111	* As there are no vifs associated with loopback devices,
2112	* get the proper interface that does have a vif associated with it.
2113	*/
2114	dev = skb->dev;
2115	if (netif_is_l3_master(dev: skb->dev)) {
2116	dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2117	if (!dev) {
2118	kfree_skb(skb);
2119	return -ENODEV;
2120	}
2121	}
2122
2123	/* Packet is looped back after forward, it should not be
2124	* forwarded second time, but still can be delivered locally.
2125	*/
2126	if (IPCB(skb)->flags & IPSKB_FORWARDED)
2127	goto dont_forward;
2128
2129	mrt = ipmr_rt_fib_lookup(net, skb);
2130	if (IS_ERR(ptr: mrt)) {
2131	kfree_skb(skb);
2132	return PTR_ERR(ptr: mrt);
2133	}
2134	if (!local) {
2135	if (IPCB(skb)->opt.router_alert) {
2136	if (ip_call_ra_chain(skb))
2137	return 0;
2138	} else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2139	/* IGMPv1 (and broken IGMPv2 implementations sort of
2140	* Cisco IOS <= 11.2(8)) do not put router alert
2141	* option to IGMP packets destined to routable
2142	* groups. It is very bad, because it means
2143	* that we can forward NO IGMP messages.
2144	*/
2145	struct sock *mroute_sk;
2146
2147	mroute_sk = rcu_dereference(mrt->mroute_sk);
2148	if (mroute_sk) {
2149	nf_reset_ct(skb);
2150	raw_rcv(mroute_sk, skb);
2151	return 0;
2152	}
2153	}
2154	}
2155
2156	/* already under rcu_read_lock() */
2157	cache = ipmr_cache_find(mrt, origin: ip_hdr(skb)->saddr, mcastgrp: ip_hdr(skb)->daddr);
2158	if (!cache) {
2159	int vif = ipmr_find_vif(mrt, dev);
2160
2161	if (vif >= 0)
2162	cache = ipmr_cache_find_any(mrt, mcastgrp: ip_hdr(skb)->daddr,
2163	vifi: vif);
2164	}
2165
2166	/* No usable cache entry */
2167	if (!cache) {
2168	int vif;
2169
2170	if (local) {
2171	struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2172	ip_local_deliver(skb);
2173	if (!skb2)
2174	return -ENOBUFS;
2175	skb = skb2;
2176	}
2177
2178	vif = ipmr_find_vif(mrt, dev);
2179	if (vif >= 0)
2180	return ipmr_cache_unresolved(mrt, vifi: vif, skb, dev);
2181	kfree_skb(skb);
2182	return -ENODEV;
2183	}
2184
2185	ip_mr_forward(net, mrt, dev, skb, c: cache, local);
2186
2187	if (local)
2188	return ip_local_deliver(skb);
2189
2190	return 0;
2191
2192	dont_forward:
2193	if (local)
2194	return ip_local_deliver(skb);
2195	kfree_skb(skb);
2196	return 0;
2197	}
2198
2199	#ifdef CONFIG_IP_PIMSM_V1
2200	/* Handle IGMP messages of PIMv1 */
2201	int pim_rcv_v1(struct sk_buff *skb)
2202	{
2203	struct igmphdr *pim;
2204	struct net *net = dev_net(dev: skb->dev);
2205	struct mr_table *mrt;
2206
2207	if (!pskb_may_pull(skb, len: sizeof(*pim) + sizeof(struct iphdr)))
2208	goto drop;
2209
2210	pim = igmp_hdr(skb);
2211
2212	mrt = ipmr_rt_fib_lookup(net, skb);
2213	if (IS_ERR(ptr: mrt))
2214	goto drop;
2215	if (!mrt->mroute_do_pim ||
2216	pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2217	goto drop;
2218
2219	if (__pim_rcv(mrt, skb, pimlen: sizeof(*pim))) {
2220	drop:
2221	kfree_skb(skb);
2222	}
2223	return 0;
2224	}
2225	#endif
2226
2227	#ifdef CONFIG_IP_PIMSM_V2
2228	static int pim_rcv(struct sk_buff *skb)
2229	{
2230	struct pimreghdr *pim;
2231	struct net *net = dev_net(dev: skb->dev);
2232	struct mr_table *mrt;
2233
2234	if (!pskb_may_pull(skb, len: sizeof(*pim) + sizeof(struct iphdr)))
2235	goto drop;
2236
2237	pim = (struct pimreghdr *)skb_transport_header(skb);
2238	if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2239	(pim->flags & PIM_NULL_REGISTER) ||
2240	(ip_compute_csum(buff: (void *)pim, len: sizeof(*pim)) != 0 &&
2241	csum_fold(sum: skb_checksum(skb, offset: 0, len: skb->len, csum: 0))))
2242	goto drop;
2243
2244	mrt = ipmr_rt_fib_lookup(net, skb);
2245	if (IS_ERR(ptr: mrt))
2246	goto drop;
2247	if (__pim_rcv(mrt, skb, pimlen: sizeof(*pim))) {
2248	drop:
2249	kfree_skb(skb);
2250	}
2251	return 0;
2252	}
2253	#endif
2254
2255	int ipmr_get_route(struct net *net, struct sk_buff *skb,
2256	__be32 saddr, __be32 daddr,
2257	struct rtmsg *rtm, u32 portid)
2258	{
2259	struct mfc_cache *cache;
2260	struct mr_table *mrt;
2261	int err;
2262
2263	mrt = ipmr_get_table(net, id: RT_TABLE_DEFAULT);
2264	if (!mrt)
2265	return -ENOENT;
2266
2267	rcu_read_lock();
2268	cache = ipmr_cache_find(mrt, origin: saddr, mcastgrp: daddr);
2269	if (!cache && skb->dev) {
2270	int vif = ipmr_find_vif(mrt, dev: skb->dev);
2271
2272	if (vif >= 0)
2273	cache = ipmr_cache_find_any(mrt, mcastgrp: daddr, vifi: vif);
2274	}
2275	if (!cache) {
2276	struct sk_buff *skb2;
2277	struct iphdr *iph;
2278	struct net_device *dev;
2279	int vif = -1;
2280
2281	dev = skb->dev;
2282	if (dev)
2283	vif = ipmr_find_vif(mrt, dev);
2284	if (vif < 0) {
2285	rcu_read_unlock();
2286	return -ENODEV;
2287	}
2288
2289	skb2 = skb_realloc_headroom(skb, headroom: sizeof(struct iphdr));
2290	if (!skb2) {
2291	rcu_read_unlock();
2292	return -ENOMEM;
2293	}
2294
2295	NETLINK_CB(skb2).portid = portid;
2296	skb_push(skb: skb2, len: sizeof(struct iphdr));
2297	skb_reset_network_header(skb: skb2);
2298	iph = ip_hdr(skb: skb2);
2299	iph->ihl = sizeof(struct iphdr) >> 2;
2300	iph->saddr = saddr;
2301	iph->daddr = daddr;
2302	iph->version = 0;
2303	err = ipmr_cache_unresolved(mrt, vifi: vif, skb: skb2, dev);
2304	rcu_read_unlock();
2305	return err;
2306	}
2307
2308	err = mr_fill_mroute(mrt, skb, c: &cache->_c, rtm);
2309	rcu_read_unlock();
2310	return err;
2311	}
2312
2313	static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2314	u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2315	int flags)
2316	{
2317	struct nlmsghdr *nlh;
2318	struct rtmsg *rtm;
2319	int err;
2320
2321	nlh = nlmsg_put(skb, portid, seq, type: cmd, payload: sizeof(*rtm), flags);
2322	if (!nlh)
2323	return -EMSGSIZE;
2324
2325	rtm = nlmsg_data(nlh);
2326	rtm->rtm_family = RTNL_FAMILY_IPMR;
2327	rtm->rtm_dst_len = 32;
2328	rtm->rtm_src_len = 32;
2329	rtm->rtm_tos = 0;
2330	rtm->rtm_table = mrt->id;
2331	if (nla_put_u32(skb, attrtype: RTA_TABLE, value: mrt->id))
2332	goto nla_put_failure;
2333	rtm->rtm_type = RTN_MULTICAST;
2334	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2335	if (c->_c.mfc_flags & MFC_STATIC)
2336	rtm->rtm_protocol = RTPROT_STATIC;
2337	else
2338	rtm->rtm_protocol = RTPROT_MROUTED;
2339	rtm->rtm_flags = 0;
2340
2341	if (nla_put_in_addr(skb, attrtype: RTA_SRC, addr: c->mfc_origin) ||
2342	nla_put_in_addr(skb, attrtype: RTA_DST, addr: c->mfc_mcastgrp))
2343	goto nla_put_failure;
2344	err = mr_fill_mroute(mrt, skb, c: &c->_c, rtm);
2345	/* do not break the dump if cache is unresolved */
2346	if (err < 0 && err != -ENOENT)
2347	goto nla_put_failure;
2348
2349	nlmsg_end(skb, nlh);
2350	return 0;
2351
2352	nla_put_failure:
2353	nlmsg_cancel(skb, nlh);
2354	return -EMSGSIZE;
2355	}
2356
2357	static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2358	u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2359	int flags)
2360	{
2361	return ipmr_fill_mroute(mrt, skb, portid, seq, c: (struct mfc_cache *)c,
2362	cmd, flags);
2363	}
2364
2365	static size_t mroute_msgsize(bool unresolved, int maxvif)
2366	{
2367	size_t len =
2368	NLMSG_ALIGN(sizeof(struct rtmsg))
2369	+ nla_total_size(payload: 4) /* RTA_TABLE */
2370	+ nla_total_size(payload: 4) /* RTA_SRC */
2371	+ nla_total_size(payload: 4) /* RTA_DST */
2372	;
2373
2374	if (!unresolved)
2375	len = len
2376	+ nla_total_size(payload: 4) /* RTA_IIF */
2377	+ nla_total_size(payload: 0) /* RTA_MULTIPATH */
2378	+ maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2379	/* RTA_MFC_STATS */
2380	+ nla_total_size_64bit(payload: sizeof(struct rta_mfc_stats))
2381	;
2382
2383	return len;
2384	}
2385
2386	static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2387	int cmd)
2388	{
2389	struct net *net = read_pnet(pnet: &mrt->net);
2390	struct sk_buff *skb;
2391	int err = -ENOBUFS;
2392
2393	skb = nlmsg_new(payload: mroute_msgsize(unresolved: mfc->_c.mfc_parent >= MAXVIFS,
2394	maxvif: mrt->maxvif),
2395	GFP_ATOMIC);
2396	if (!skb)
2397	goto errout;
2398
2399	err = ipmr_fill_mroute(mrt, skb, portid: 0, seq: 0, c: mfc, cmd, flags: 0);
2400	if (err < 0)
2401	goto errout;
2402
2403	rtnl_notify(skb, net, pid: 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2404	return;
2405
2406	errout:
2407	kfree_skb(skb);
2408	if (err < 0)
2409	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, error: err);
2410	}
2411
2412	static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2413	{
2414	size_t len =
2415	NLMSG_ALIGN(sizeof(struct rtgenmsg))
2416	+ nla_total_size(payload: 1) /* IPMRA_CREPORT_MSGTYPE */
2417	+ nla_total_size(payload: 4) /* IPMRA_CREPORT_VIF_ID */
2418	+ nla_total_size(payload: 4) /* IPMRA_CREPORT_SRC_ADDR */
2419	+ nla_total_size(payload: 4) /* IPMRA_CREPORT_DST_ADDR */
2420	+ nla_total_size(payload: 4) /* IPMRA_CREPORT_TABLE */
2421	/* IPMRA_CREPORT_PKT */
2422	+ nla_total_size(payload: payloadlen)
2423	;
2424
2425	return len;
2426	}
2427
2428	static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2429	{
2430	struct net *net = read_pnet(pnet: &mrt->net);
2431	struct nlmsghdr *nlh;
2432	struct rtgenmsg *rtgenm;
2433	struct igmpmsg *msg;
2434	struct sk_buff *skb;
2435	struct nlattr *nla;
2436	int payloadlen;
2437
2438	payloadlen = pkt->len - sizeof(struct igmpmsg);
2439	msg = (struct igmpmsg *)skb_network_header(skb: pkt);
2440
2441	skb = nlmsg_new(payload: igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2442	if (!skb)
2443	goto errout;
2444
2445	nlh = nlmsg_put(skb, portid: 0, seq: 0, RTM_NEWCACHEREPORT,
2446	payload: sizeof(struct rtgenmsg), flags: 0);
2447	if (!nlh)
2448	goto errout;
2449	rtgenm = nlmsg_data(nlh);
2450	rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2451	if (nla_put_u8(skb, attrtype: IPMRA_CREPORT_MSGTYPE, value: msg->im_msgtype) ||
2452	nla_put_u32(skb, attrtype: IPMRA_CREPORT_VIF_ID, value: msg->im_vif | (msg->im_vif_hi << 8)) ||
2453	nla_put_in_addr(skb, attrtype: IPMRA_CREPORT_SRC_ADDR,
2454	addr: msg->im_src.s_addr) ||
2455	nla_put_in_addr(skb, attrtype: IPMRA_CREPORT_DST_ADDR,
2456	addr: msg->im_dst.s_addr) ||
2457	nla_put_u32(skb, attrtype: IPMRA_CREPORT_TABLE, value: mrt->id))
2458	goto nla_put_failure;
2459
2460	nla = nla_reserve(skb, attrtype: IPMRA_CREPORT_PKT, attrlen: payloadlen);
2461	if (!nla || skb_copy_bits(skb: pkt, offset: sizeof(struct igmpmsg),
2462	to: nla_data(nla), len: payloadlen))
2463	goto nla_put_failure;
2464
2465	nlmsg_end(skb, nlh);
2466
2467	rtnl_notify(skb, net, pid: 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2468	return;
2469
2470	nla_put_failure:
2471	nlmsg_cancel(skb, nlh);
2472	errout:
2473	kfree_skb(skb);
2474	rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, error: -ENOBUFS);
2475	}
2476
2477	static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2478	const struct nlmsghdr *nlh,
2479	struct nlattr **tb,
2480	struct netlink_ext_ack *extack)
2481	{
2482	struct rtmsg *rtm;
2483	int i, err;
2484
2485	if (nlh->nlmsg_len < nlmsg_msg_size(payload: sizeof(*rtm))) {
2486	NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2487	return -EINVAL;
2488	}
2489
2490	if (!netlink_strict_get_check(skb))
2491	return nlmsg_parse_deprecated(nlh, hdrlen: sizeof(*rtm), tb, RTA_MAX,
2492	policy: rtm_ipv4_policy, extack);
2493
2494	rtm = nlmsg_data(nlh);
2495	if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2496	(rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2497	rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2498	rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2499	NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2500	return -EINVAL;
2501	}
2502
2503	err = nlmsg_parse_deprecated_strict(nlh, hdrlen: sizeof(*rtm), tb, RTA_MAX,
2504	policy: rtm_ipv4_policy, extack);
2505	if (err)
2506	return err;
2507
2508	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2509	(tb[RTA_DST] && !rtm->rtm_dst_len)) {
2510	NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2511	return -EINVAL;
2512	}
2513
2514	for (i = 0; i <= RTA_MAX; i++) {
2515	if (!tb[i])
2516	continue;
2517
2518	switch (i) {
2519	case RTA_SRC:
2520	case RTA_DST:
2521	case RTA_TABLE:
2522	break;
2523	default:
2524	NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2525	return -EINVAL;
2526	}
2527	}
2528
2529	return 0;
2530	}
2531
2532	static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2533	struct netlink_ext_ack *extack)
2534	{
2535	struct net *net = sock_net(sk: in_skb->sk);
2536	struct nlattr *tb[RTA_MAX + 1];
2537	struct sk_buff *skb = NULL;
2538	struct mfc_cache *cache;
2539	struct mr_table *mrt;
2540	__be32 src, grp;
2541	u32 tableid;
2542	int err;
2543
2544	err = ipmr_rtm_valid_getroute_req(skb: in_skb, nlh, tb, extack);
2545	if (err < 0)
2546	goto errout;
2547
2548	src = tb[RTA_SRC] ? nla_get_in_addr(nla: tb[RTA_SRC]) : 0;
2549	grp = tb[RTA_DST] ? nla_get_in_addr(nla: tb[RTA_DST]) : 0;
2550	tableid = tb[RTA_TABLE] ? nla_get_u32(nla: tb[RTA_TABLE]) : 0;
2551
2552	mrt = ipmr_get_table(net, id: tableid ? tableid : RT_TABLE_DEFAULT);
2553	if (!mrt) {
2554	err = -ENOENT;
2555	goto errout_free;
2556	}
2557
2558	/* entries are added/deleted only under RTNL */
2559	rcu_read_lock();
2560	cache = ipmr_cache_find(mrt, origin: src, mcastgrp: grp);
2561	rcu_read_unlock();
2562	if (!cache) {
2563	err = -ENOENT;
2564	goto errout_free;
2565	}
2566
2567	skb = nlmsg_new(payload: mroute_msgsize(unresolved: false, maxvif: mrt->maxvif), GFP_KERNEL);
2568	if (!skb) {
2569	err = -ENOBUFS;
2570	goto errout_free;
2571	}
2572
2573	err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2574	seq: nlh->nlmsg_seq, c: cache,
2575	RTM_NEWROUTE, flags: 0);
2576	if (err < 0)
2577	goto errout_free;
2578
2579	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2580
2581	errout:
2582	return err;
2583
2584	errout_free:
2585	kfree_skb(skb);
2586	goto errout;
2587	}
2588
2589	static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2590	{
2591	struct fib_dump_filter filter = {};
2592	int err;
2593
2594	if (cb->strict_check) {
2595	err = ip_valid_fib_dump_req(net: sock_net(sk: skb->sk), nlh: cb->nlh,
2596	filter: &filter, cb);
2597	if (err < 0)
2598	return err;
2599	}
2600
2601	if (filter.table_id) {
2602	struct mr_table *mrt;
2603
2604	mrt = ipmr_get_table(net: sock_net(sk: skb->sk), id: filter.table_id);
2605	if (!mrt) {
2606	if (rtnl_msg_family(nlh: cb->nlh) != RTNL_FAMILY_IPMR)
2607	return skb->len;
2608
2609	NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2610	return -ENOENT;
2611	}
2612	err = mr_table_dump(mrt, skb, cb, fill: _ipmr_fill_mroute,
2613	lock: &mfc_unres_lock, filter: &filter);
2614	return skb->len ? : err;
2615	}
2616
2617	return mr_rtm_dumproute(skb, cb, iter: ipmr_mr_table_iter,
2618	fill: _ipmr_fill_mroute, lock: &mfc_unres_lock, filter: &filter);
2619	}
2620
2621	static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2622	[RTA_SRC] = { .type = NLA_U32 },
2623	[RTA_DST] = { .type = NLA_U32 },
2624	[RTA_IIF] = { .type = NLA_U32 },
2625	[RTA_TABLE] = { .type = NLA_U32 },
2626	[RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2627	};
2628
2629	static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2630	{
2631	switch (rtm_protocol) {
2632	case RTPROT_STATIC:
2633	case RTPROT_MROUTED:
2634	return true;
2635	}
2636	return false;
2637	}
2638
2639	static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2640	{
2641	struct rtnexthop *rtnh = nla_data(nla);
2642	int remaining = nla_len(nla), vifi = 0;
2643
2644	while (rtnh_ok(rtnh, remaining)) {
2645	mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2646	if (++vifi == MAXVIFS)
2647	break;
2648	rtnh = rtnh_next(rtnh, remaining: &remaining);
2649	}
2650
2651	return remaining > 0 ? -EINVAL : vifi;
2652	}
2653
2654	/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2655	static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2656	struct mfcctl *mfcc, int *mrtsock,
2657	struct mr_table **mrtret,
2658	struct netlink_ext_ack *extack)
2659	{
2660	struct net_device *dev = NULL;
2661	u32 tblid = RT_TABLE_DEFAULT;
2662	struct mr_table *mrt;
2663	struct nlattr *attr;
2664	struct rtmsg *rtm;
2665	int ret, rem;
2666
2667	ret = nlmsg_validate_deprecated(nlh, hdrlen: sizeof(*rtm), RTA_MAX,
2668	policy: rtm_ipmr_policy, extack);
2669	if (ret < 0)
2670	goto out;
2671	rtm = nlmsg_data(nlh);
2672
2673	ret = -EINVAL;
2674	if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2675	rtm->rtm_type != RTN_MULTICAST ||
2676	rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2677	!ipmr_rtm_validate_proto(rtm_protocol: rtm->rtm_protocol))
2678	goto out;
2679
2680	memset(mfcc, 0, sizeof(*mfcc));
2681	mfcc->mfcc_parent = -1;
2682	ret = 0;
2683	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2684	switch (nla_type(nla: attr)) {
2685	case RTA_SRC:
2686	mfcc->mfcc_origin.s_addr = nla_get_be32(nla: attr);
2687	break;
2688	case RTA_DST:
2689	mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(nla: attr);
2690	break;
2691	case RTA_IIF:
2692	dev = __dev_get_by_index(net, ifindex: nla_get_u32(nla: attr));
2693	if (!dev) {
2694	ret = -ENODEV;
2695	goto out;
2696	}
2697	break;
2698	case RTA_MULTIPATH:
2699	if (ipmr_nla_get_ttls(nla: attr, mfcc) < 0) {
2700	ret = -EINVAL;
2701	goto out;
2702	}
2703	break;
2704	case RTA_PREFSRC:
2705	ret = 1;
2706	break;
2707	case RTA_TABLE:
2708	tblid = nla_get_u32(nla: attr);
2709	break;
2710	}
2711	}
2712	mrt = ipmr_get_table(net, id: tblid);
2713	if (!mrt) {
2714	ret = -ENOENT;
2715	goto out;
2716	}
2717	*mrtret = mrt;
2718	*mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2719	if (dev)
2720	mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2721
2722	out:
2723	return ret;
2724	}
2725
2726	/* takes care of both newroute and delroute */
2727	static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2728	struct netlink_ext_ack *extack)
2729	{
2730	struct net *net = sock_net(sk: skb->sk);
2731	int ret, mrtsock, parent;
2732	struct mr_table *tbl;
2733	struct mfcctl mfcc;
2734
2735	mrtsock = 0;
2736	tbl = NULL;
2737	ret = rtm_to_ipmr_mfcc(net, nlh, mfcc: &mfcc, mrtsock: &mrtsock, mrtret: &tbl, extack);
2738	if (ret < 0)
2739	return ret;
2740
2741	parent = ret ? mfcc.mfcc_parent : -1;
2742	if (nlh->nlmsg_type == RTM_NEWROUTE)
2743	return ipmr_mfc_add(net, mrt: tbl, mfc: &mfcc, mrtsock, parent);
2744	else
2745	return ipmr_mfc_delete(mrt: tbl, mfc: &mfcc, parent);
2746	}
2747
2748	static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2749	{
2750	u32 queue_len = atomic_read(v: &mrt->cache_resolve_queue_len);
2751
2752	if (nla_put_u32(skb, attrtype: IPMRA_TABLE_ID, value: mrt->id) ||
2753	nla_put_u32(skb, attrtype: IPMRA_TABLE_CACHE_RES_QUEUE_LEN, value: queue_len) ||
2754	nla_put_s32(skb, attrtype: IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2755	value: mrt->mroute_reg_vif_num) ||
2756	nla_put_u8(skb, attrtype: IPMRA_TABLE_MROUTE_DO_ASSERT,
2757	value: mrt->mroute_do_assert) ||
2758	nla_put_u8(skb, attrtype: IPMRA_TABLE_MROUTE_DO_PIM, value: mrt->mroute_do_pim) ||
2759	nla_put_u8(skb, attrtype: IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2760	value: mrt->mroute_do_wrvifwhole))
2761	return false;
2762
2763	return true;
2764	}
2765
2766	static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2767	{
2768	struct net_device *vif_dev;
2769	struct nlattr *vif_nest;
2770	struct vif_device *vif;
2771
2772	vif = &mrt->vif_table[vifid];
2773	vif_dev = rtnl_dereference(vif->dev);
2774	/* if the VIF doesn't exist just continue */
2775	if (!vif_dev)
2776	return true;
2777
2778	vif_nest = nla_nest_start_noflag(skb, attrtype: IPMRA_VIF);
2779	if (!vif_nest)
2780	return false;
2781
2782	if (nla_put_u32(skb, attrtype: IPMRA_VIFA_IFINDEX, value: vif_dev->ifindex) ||
2783	nla_put_u32(skb, attrtype: IPMRA_VIFA_VIF_ID, value: vifid) ||
2784	nla_put_u16(skb, attrtype: IPMRA_VIFA_FLAGS, value: vif->flags) ||
2785	nla_put_u64_64bit(skb, attrtype: IPMRA_VIFA_BYTES_IN, value: vif->bytes_in,
2786	padattr: IPMRA_VIFA_PAD) ||
2787	nla_put_u64_64bit(skb, attrtype: IPMRA_VIFA_BYTES_OUT, value: vif->bytes_out,
2788	padattr: IPMRA_VIFA_PAD) ||
2789	nla_put_u64_64bit(skb, attrtype: IPMRA_VIFA_PACKETS_IN, value: vif->pkt_in,
2790	padattr: IPMRA_VIFA_PAD) ||
2791	nla_put_u64_64bit(skb, attrtype: IPMRA_VIFA_PACKETS_OUT, value: vif->pkt_out,
2792	padattr: IPMRA_VIFA_PAD) ||
2793	nla_put_be32(skb, attrtype: IPMRA_VIFA_LOCAL_ADDR, value: vif->local) ||
2794	nla_put_be32(skb, attrtype: IPMRA_VIFA_REMOTE_ADDR, value: vif->remote)) {
2795	nla_nest_cancel(skb, start: vif_nest);
2796	return false;
2797	}
2798	nla_nest_end(skb, start: vif_nest);
2799
2800	return true;
2801	}
2802
2803	static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2804	struct netlink_ext_ack *extack)
2805	{
2806	struct ifinfomsg *ifm;
2807
2808	if (nlh->nlmsg_len < nlmsg_msg_size(payload: sizeof(*ifm))) {
2809	NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2810	return -EINVAL;
2811	}
2812
2813	if (nlmsg_attrlen(nlh, hdrlen: sizeof(*ifm))) {
2814	NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2815	return -EINVAL;
2816	}
2817
2818	ifm = nlmsg_data(nlh);
2819	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2820	ifm->ifi_change || ifm->ifi_index) {
2821	NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2822	return -EINVAL;
2823	}
2824
2825	return 0;
2826	}
2827
2828	static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2829	{
2830	struct net *net = sock_net(sk: skb->sk);
2831	struct nlmsghdr *nlh = NULL;
2832	unsigned int t = 0, s_t;
2833	unsigned int e = 0, s_e;
2834	struct mr_table *mrt;
2835
2836	if (cb->strict_check) {
2837	int err = ipmr_valid_dumplink(nlh: cb->nlh, extack: cb->extack);
2838
2839	if (err < 0)
2840	return err;
2841	}
2842
2843	s_t = cb->args[0];
2844	s_e = cb->args[1];
2845
2846	ipmr_for_each_table(mrt, net) {
2847	struct nlattr *vifs, *af;
2848	struct ifinfomsg *hdr;
2849	u32 i;
2850
2851	if (t < s_t)
2852	goto skip_table;
2853	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2854	seq: cb->nlh->nlmsg_seq, RTM_NEWLINK,
2855	payload: sizeof(*hdr), NLM_F_MULTI);
2856	if (!nlh)
2857	break;
2858
2859	hdr = nlmsg_data(nlh);
2860	memset(hdr, 0, sizeof(*hdr));
2861	hdr->ifi_family = RTNL_FAMILY_IPMR;
2862
2863	af = nla_nest_start_noflag(skb, attrtype: IFLA_AF_SPEC);
2864	if (!af) {
2865	nlmsg_cancel(skb, nlh);
2866	goto out;
2867	}
2868
2869	if (!ipmr_fill_table(mrt, skb)) {
2870	nlmsg_cancel(skb, nlh);
2871	goto out;
2872	}
2873
2874	vifs = nla_nest_start_noflag(skb, attrtype: IPMRA_TABLE_VIFS);
2875	if (!vifs) {
2876	nla_nest_end(skb, start: af);
2877	nlmsg_end(skb, nlh);
2878	goto out;
2879	}
2880	for (i = 0; i < mrt->maxvif; i++) {
2881	if (e < s_e)
2882	goto skip_entry;
2883	if (!ipmr_fill_vif(mrt, vifid: i, skb)) {
2884	nla_nest_end(skb, start: vifs);
2885	nla_nest_end(skb, start: af);
2886	nlmsg_end(skb, nlh);
2887	goto out;
2888	}
2889	skip_entry:
2890	e++;
2891	}
2892	s_e = 0;
2893	e = 0;
2894	nla_nest_end(skb, start: vifs);
2895	nla_nest_end(skb, start: af);
2896	nlmsg_end(skb, nlh);
2897	skip_table:
2898	t++;
2899	}
2900
2901	out:
2902	cb->args[1] = e;
2903	cb->args[0] = t;
2904
2905	return skb->len;
2906	}
2907
2908	#ifdef CONFIG_PROC_FS
2909	/* The /proc interfaces to multicast routing :
2910	* /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2911	*/
2912
2913	static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2914	__acquires(RCU)
2915	{
2916	struct mr_vif_iter *iter = seq->private;
2917	struct net *net = seq_file_net(seq);
2918	struct mr_table *mrt;
2919
2920	mrt = ipmr_get_table(net, id: RT_TABLE_DEFAULT);
2921	if (!mrt)
2922	return ERR_PTR(error: -ENOENT);
2923
2924	iter->mrt = mrt;
2925
2926	rcu_read_lock();
2927	return mr_vif_seq_start(seq, pos);
2928	}
2929
2930	static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2931	__releases(RCU)
2932	{
2933	rcu_read_unlock();
2934	}
2935
2936	static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2937	{
2938	struct mr_vif_iter *iter = seq->private;
2939	struct mr_table *mrt = iter->mrt;
2940
2941	if (v == SEQ_START_TOKEN) {
2942	seq_puts(m: seq,
2943	s: "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2944	} else {
2945	const struct vif_device *vif = v;
2946	const struct net_device *vif_dev;
2947	const char *name;
2948
2949	vif_dev = vif_dev_read(vif);
2950	name = vif_dev ? vif_dev->name : "none";
2951	seq_printf(m: seq,
2952	fmt: "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2953	vif - mrt->vif_table,
2954	name, vif->bytes_in, vif->pkt_in,
2955	vif->bytes_out, vif->pkt_out,
2956	vif->flags, vif->local, vif->remote);
2957	}
2958	return 0;
2959	}
2960
2961	static const struct seq_operations ipmr_vif_seq_ops = {
2962	.start = ipmr_vif_seq_start,
2963	.next = mr_vif_seq_next,
2964	.stop = ipmr_vif_seq_stop,
2965	.show = ipmr_vif_seq_show,
2966	};
2967
2968	static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2969	{
2970	struct net *net = seq_file_net(seq);
2971	struct mr_table *mrt;
2972
2973	mrt = ipmr_get_table(net, id: RT_TABLE_DEFAULT);
2974	if (!mrt)
2975	return ERR_PTR(error: -ENOENT);
2976
2977	return mr_mfc_seq_start(seq, pos, mrt, lock: &mfc_unres_lock);
2978	}
2979
2980	static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2981	{
2982	int n;
2983
2984	if (v == SEQ_START_TOKEN) {
2985	seq_puts(m: seq,
2986	s: "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2987	} else {
2988	const struct mfc_cache *mfc = v;
2989	const struct mr_mfc_iter *it = seq->private;
2990	const struct mr_table *mrt = it->mrt;
2991
2992	seq_printf(m: seq, fmt: "%08X %08X %-3hd",
2993	(__force u32) mfc->mfc_mcastgrp,
2994	(__force u32) mfc->mfc_origin,
2995	mfc->_c.mfc_parent);
2996
2997	if (it->cache != &mrt->mfc_unres_queue) {
2998	seq_printf(m: seq, fmt: " %8lu %8lu %8lu",
2999	mfc->_c.mfc_un.res.pkt,
3000	mfc->_c.mfc_un.res.bytes,
3001	mfc->_c.mfc_un.res.wrong_if);
3002	for (n = mfc->_c.mfc_un.res.minvif;
3003	n < mfc->_c.mfc_un.res.maxvif; n++) {
3004	if (VIF_EXISTS(mrt, n) &&
3005	mfc->_c.mfc_un.res.ttls[n] < 255)
3006	seq_printf(m: seq,
3007	fmt: " %2d:%-3d",
3008	n, mfc->_c.mfc_un.res.ttls[n]);
3009	}
3010	} else {
3011	/* unresolved mfc_caches don't contain
3012	* pkt, bytes and wrong_if values
3013	*/
3014	seq_printf(m: seq, fmt: " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3015	}
3016	seq_putc(m: seq, c: '\n');
3017	}
3018	return 0;
3019	}
3020
3021	static const struct seq_operations ipmr_mfc_seq_ops = {
3022	.start = ipmr_mfc_seq_start,
3023	.next = mr_mfc_seq_next,
3024	.stop = mr_mfc_seq_stop,
3025	.show = ipmr_mfc_seq_show,
3026	};
3027	#endif
3028
3029	#ifdef CONFIG_IP_PIMSM_V2
3030	static const struct net_protocol pim_protocol = {
3031	.handler = pim_rcv,
3032	};
3033	#endif
3034
3035	static unsigned int ipmr_seq_read(struct net *net)
3036	{
3037	ASSERT_RTNL();
3038
3039	return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3040	}
3041
3042	static int ipmr_dump(struct net *net, struct notifier_block *nb,
3043	struct netlink_ext_ack *extack)
3044	{
3045	return mr_dump(net, nb, RTNL_FAMILY_IPMR, rules_dump: ipmr_rules_dump,
3046	mr_iter: ipmr_mr_table_iter, extack);
3047	}
3048
3049	static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3050	.family = RTNL_FAMILY_IPMR,
3051	.fib_seq_read = ipmr_seq_read,
3052	.fib_dump = ipmr_dump,
3053	.owner = THIS_MODULE,
3054	};
3055
3056	static int __net_init ipmr_notifier_init(struct net *net)
3057	{
3058	struct fib_notifier_ops *ops;
3059
3060	net->ipv4.ipmr_seq = 0;
3061
3062	ops = fib_notifier_ops_register(tmpl: &ipmr_notifier_ops_template, net);
3063	if (IS_ERR(ptr: ops))
3064	return PTR_ERR(ptr: ops);
3065	net->ipv4.ipmr_notifier_ops = ops;
3066
3067	return 0;
3068	}
3069
3070	static void __net_exit ipmr_notifier_exit(struct net *net)
3071	{
3072	fib_notifier_ops_unregister(ops: net->ipv4.ipmr_notifier_ops);
3073	net->ipv4.ipmr_notifier_ops = NULL;
3074	}
3075
3076	/* Setup for IP multicast routing */
3077	static int __net_init ipmr_net_init(struct net *net)
3078	{
3079	int err;
3080
3081	err = ipmr_notifier_init(net);
3082	if (err)
3083	goto ipmr_notifier_fail;
3084
3085	err = ipmr_rules_init(net);
3086	if (err < 0)
3087	goto ipmr_rules_fail;
3088
3089	#ifdef CONFIG_PROC_FS
3090	err = -ENOMEM;
3091	if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3092	sizeof(struct mr_vif_iter)))
3093	goto proc_vif_fail;
3094	if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3095	sizeof(struct mr_mfc_iter)))
3096	goto proc_cache_fail;
3097	#endif
3098	return 0;
3099
3100	#ifdef CONFIG_PROC_FS
3101	proc_cache_fail:
3102	remove_proc_entry("ip_mr_vif", net->proc_net);
3103	proc_vif_fail:
3104	rtnl_lock();
3105	ipmr_rules_exit(net);
3106	rtnl_unlock();
3107	#endif
3108	ipmr_rules_fail:
3109	ipmr_notifier_exit(net);
3110	ipmr_notifier_fail:
3111	return err;
3112	}
3113
3114	static void __net_exit ipmr_net_exit(struct net *net)
3115	{
3116	#ifdef CONFIG_PROC_FS
3117	remove_proc_entry("ip_mr_cache", net->proc_net);
3118	remove_proc_entry("ip_mr_vif", net->proc_net);
3119	#endif
3120	ipmr_notifier_exit(net);
3121	}
3122
3123	static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3124	{
3125	struct net *net;
3126
3127	rtnl_lock();
3128	list_for_each_entry(net, net_list, exit_list)
3129	ipmr_rules_exit(net);
3130	rtnl_unlock();
3131	}
3132
3133	static struct pernet_operations ipmr_net_ops = {
3134	.init = ipmr_net_init,
3135	.exit = ipmr_net_exit,
3136	.exit_batch = ipmr_net_exit_batch,
3137	};
3138
3139	int __init ip_mr_init(void)
3140	{
3141	int err;
3142
3143	mrt_cachep = kmem_cache_create(name: "ip_mrt_cache",
3144	size: sizeof(struct mfc_cache),
3145	align: 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3146	NULL);
3147
3148	err = register_pernet_subsys(&ipmr_net_ops);
3149	if (err)
3150	goto reg_pernet_fail;
3151
3152	err = register_netdevice_notifier(nb: &ip_mr_notifier);
3153	if (err)
3154	goto reg_notif_fail;
3155	#ifdef CONFIG_IP_PIMSM_V2
3156	if (inet_add_protocol(prot: &pim_protocol, IPPROTO_PIM) < 0) {
3157	pr_err("%s: can't add PIM protocol\n", __func__);
3158	err = -EAGAIN;
3159	goto add_proto_fail;
3160	}
3161	#endif
3162	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3163	ipmr_rtm_getroute, ipmr_rtm_dumproute, flags: 0);
3164	rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3165	ipmr_rtm_route, NULL, flags: 0);
3166	rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3167	ipmr_rtm_route, NULL, flags: 0);
3168
3169	rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3170	NULL, ipmr_rtm_dumplink, flags: 0);
3171	return 0;
3172
3173	#ifdef CONFIG_IP_PIMSM_V2
3174	add_proto_fail:
3175	unregister_netdevice_notifier(nb: &ip_mr_notifier);
3176	#endif
3177	reg_notif_fail:
3178	unregister_pernet_subsys(&ipmr_net_ops);
3179	reg_pernet_fail:
3180	kmem_cache_destroy(s: mrt_cachep);
3181	return err;
3182	}
3183