Loading...
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
68struct ipmr_rule {
69 struct fib_rule common;
70};
71
72struct 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
80static DEFINE_RWLOCK(mrt_lock);
81
82/* Multicast router control variables */
83
84/* Special spinlock for queue of unresolved entries */
85static DEFINE_SPINLOCK(mfc_unres_lock);
86
87/* We return to original Alan's scheme. Hash table of resolved
88 * entries is changed only in process context and protected
89 * with weak lock mrt_lock. Queue of unresolved entries is protected
90 * with strong spinlock mfc_unres_lock.
91 *
92 * In this case data path is free of exclusive locks at all.
93 */
94
95static struct kmem_cache *mrt_cachep __ro_after_init;
96
97static struct mr_table *ipmr_new_table(struct net *net, u32 id);
98static void ipmr_free_table(struct mr_table *mrt);
99
100static void ip_mr_forward(struct net *net, struct mr_table *mrt,
101 struct net_device *dev, struct sk_buff *skb,
102 struct mfc_cache *cache, int local);
103static int ipmr_cache_report(struct mr_table *mrt,
104 struct sk_buff *pkt, vifi_t vifi, int assert);
105static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
106 int cmd);
107static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
108static void mroute_clean_tables(struct mr_table *mrt, int flags);
109static void ipmr_expire_process(struct timer_list *t);
110
111#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
112#define ipmr_for_each_table(mrt, net) \
113 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
114 lockdep_rtnl_is_held() || \
115 list_empty(&net->ipv4.mr_tables))
116
117static struct mr_table *ipmr_mr_table_iter(struct net *net,
118 struct mr_table *mrt)
119{
120 struct mr_table *ret;
121
122 if (!mrt)
123 ret = list_entry_rcu(net->ipv4.mr_tables.next,
124 struct mr_table, list);
125 else
126 ret = list_entry_rcu(mrt->list.next,
127 struct mr_table, list);
128
129 if (&ret->list == &net->ipv4.mr_tables)
130 return NULL;
131 return ret;
132}
133
134static struct mr_table *ipmr_get_table(struct net *net, u32 id)
135{
136 struct mr_table *mrt;
137
138 ipmr_for_each_table(mrt, net) {
139 if (mrt->id == id)
140 return mrt;
141 }
142 return NULL;
143}
144
145static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
146 struct mr_table **mrt)
147{
148 int err;
149 struct ipmr_result res;
150 struct fib_lookup_arg arg = {
151 .result = &res,
152 .flags = FIB_LOOKUP_NOREF,
153 };
154
155 /* update flow if oif or iif point to device enslaved to l3mdev */
156 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
157
158 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
159 flowi4_to_flowi(flp4), 0, &arg);
160 if (err < 0)
161 return err;
162 *mrt = res.mrt;
163 return 0;
164}
165
166static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
167 int flags, struct fib_lookup_arg *arg)
168{
169 struct ipmr_result *res = arg->result;
170 struct mr_table *mrt;
171
172 switch (rule->action) {
173 case FR_ACT_TO_TBL:
174 break;
175 case FR_ACT_UNREACHABLE:
176 return -ENETUNREACH;
177 case FR_ACT_PROHIBIT:
178 return -EACCES;
179 case FR_ACT_BLACKHOLE:
180 default:
181 return -EINVAL;
182 }
183
184 arg->table = fib_rule_get_table(rule, arg);
185
186 mrt = ipmr_get_table(rule->fr_net, arg->table);
187 if (!mrt)
188 return -EAGAIN;
189 res->mrt = mrt;
190 return 0;
191}
192
193static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
194{
195 return 1;
196}
197
198static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
199 FRA_GENERIC_POLICY,
200};
201
202static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
203 struct fib_rule_hdr *frh, struct nlattr **tb,
204 struct netlink_ext_ack *extack)
205{
206 return 0;
207}
208
209static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
210 struct nlattr **tb)
211{
212 return 1;
213}
214
215static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
216 struct fib_rule_hdr *frh)
217{
218 frh->dst_len = 0;
219 frh->src_len = 0;
220 frh->tos = 0;
221 return 0;
222}
223
224static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
225 .family = RTNL_FAMILY_IPMR,
226 .rule_size = sizeof(struct ipmr_rule),
227 .addr_size = sizeof(u32),
228 .action = ipmr_rule_action,
229 .match = ipmr_rule_match,
230 .configure = ipmr_rule_configure,
231 .compare = ipmr_rule_compare,
232 .fill = ipmr_rule_fill,
233 .nlgroup = RTNLGRP_IPV4_RULE,
234 .policy = ipmr_rule_policy,
235 .owner = THIS_MODULE,
236};
237
238static int __net_init ipmr_rules_init(struct net *net)
239{
240 struct fib_rules_ops *ops;
241 struct mr_table *mrt;
242 int err;
243
244 ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 if (IS_ERR(ops))
246 return PTR_ERR(ops);
247
248 INIT_LIST_HEAD(&net->ipv4.mr_tables);
249
250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 if (IS_ERR(mrt)) {
252 err = PTR_ERR(mrt);
253 goto err1;
254 }
255
256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
257 if (err < 0)
258 goto err2;
259
260 net->ipv4.mr_rules_ops = ops;
261 return 0;
262
263err2:
264 ipmr_free_table(mrt);
265err1:
266 fib_rules_unregister(ops);
267 return err;
268}
269
270static void __net_exit ipmr_rules_exit(struct net *net)
271{
272 struct mr_table *mrt, *next;
273
274 rtnl_lock();
275 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
276 list_del(&mrt->list);
277 ipmr_free_table(mrt);
278 }
279 fib_rules_unregister(net->ipv4.mr_rules_ops);
280 rtnl_unlock();
281}
282
283static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
284 struct netlink_ext_ack *extack)
285{
286 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
287}
288
289static unsigned int ipmr_rules_seq_read(struct net *net)
290{
291 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
292}
293
294bool ipmr_rule_default(const struct fib_rule *rule)
295{
296 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
297}
298EXPORT_SYMBOL(ipmr_rule_default);
299#else
300#define ipmr_for_each_table(mrt, net) \
301 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
302
303static struct mr_table *ipmr_mr_table_iter(struct net *net,
304 struct mr_table *mrt)
305{
306 if (!mrt)
307 return net->ipv4.mrt;
308 return NULL;
309}
310
311static struct mr_table *ipmr_get_table(struct net *net, u32 id)
312{
313 return net->ipv4.mrt;
314}
315
316static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
317 struct mr_table **mrt)
318{
319 *mrt = net->ipv4.mrt;
320 return 0;
321}
322
323static int __net_init ipmr_rules_init(struct net *net)
324{
325 struct mr_table *mrt;
326
327 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
328 if (IS_ERR(mrt))
329 return PTR_ERR(mrt);
330 net->ipv4.mrt = mrt;
331 return 0;
332}
333
334static void __net_exit ipmr_rules_exit(struct net *net)
335{
336 rtnl_lock();
337 ipmr_free_table(net->ipv4.mrt);
338 net->ipv4.mrt = NULL;
339 rtnl_unlock();
340}
341
342static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
343 struct netlink_ext_ack *extack)
344{
345 return 0;
346}
347
348static unsigned int ipmr_rules_seq_read(struct net *net)
349{
350 return 0;
351}
352
353bool ipmr_rule_default(const struct fib_rule *rule)
354{
355 return true;
356}
357EXPORT_SYMBOL(ipmr_rule_default);
358#endif
359
360static 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 struct mfc_cache *c = (struct mfc_cache *)ptr;
365
366 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
367 cmparg->mfc_origin != c->mfc_origin;
368}
369
370static 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
379static void ipmr_new_table_set(struct mr_table *mrt,
380 struct net *net)
381{
382#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
383 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384#endif
385}
386
387static 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
392static 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
397static struct mr_table *ipmr_new_table(struct net *net, u32 id)
398{
399 struct mr_table *mrt;
400
401 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
402 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
403 return ERR_PTR(-EINVAL);
404
405 mrt = ipmr_get_table(net, id);
406 if (mrt)
407 return mrt;
408
409 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
410 ipmr_expire_process, ipmr_new_table_set);
411}
412
413static void ipmr_free_table(struct mr_table *mrt)
414{
415 del_timer_sync(&mrt->ipmr_expire_timer);
416 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
417 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
418 rhltable_destroy(&mrt->mfc_hash);
419 kfree(mrt);
420}
421
422/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
423
424/* Initialize ipmr pimreg/tunnel in_device */
425static bool ipmr_init_vif_indev(const struct net_device *dev)
426{
427 struct in_device *in_dev;
428
429 ASSERT_RTNL();
430
431 in_dev = __in_dev_get_rtnl(dev);
432 if (!in_dev)
433 return false;
434 ipv4_devconf_setall(in_dev);
435 neigh_parms_data_state_setall(in_dev->arp_parms);
436 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
437
438 return true;
439}
440
441static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
442{
443 struct net_device *tunnel_dev, *new_dev;
444 struct ip_tunnel_parm p = { };
445 int err;
446
447 tunnel_dev = __dev_get_by_name(net, "tunl0");
448 if (!tunnel_dev)
449 goto out;
450
451 p.iph.daddr = v->vifc_rmt_addr.s_addr;
452 p.iph.saddr = v->vifc_lcl_addr.s_addr;
453 p.iph.version = 4;
454 p.iph.ihl = 5;
455 p.iph.protocol = IPPROTO_IPIP;
456 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
457
458 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
459 goto out;
460 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
461 SIOCADDTUNNEL);
462 if (err)
463 goto out;
464
465 new_dev = __dev_get_by_name(net, p.name);
466 if (!new_dev)
467 goto out;
468
469 new_dev->flags |= IFF_MULTICAST;
470 if (!ipmr_init_vif_indev(new_dev))
471 goto out_unregister;
472 if (dev_open(new_dev, NULL))
473 goto out_unregister;
474 dev_hold(new_dev);
475 err = dev_set_allmulti(new_dev, 1);
476 if (err) {
477 dev_close(new_dev);
478 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
479 SIOCDELTUNNEL);
480 dev_put(new_dev);
481 new_dev = ERR_PTR(err);
482 }
483 return new_dev;
484
485out_unregister:
486 unregister_netdevice(new_dev);
487out:
488 return ERR_PTR(-ENOBUFS);
489}
490
491#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
492static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
493{
494 struct net *net = dev_net(dev);
495 struct mr_table *mrt;
496 struct flowi4 fl4 = {
497 .flowi4_oif = dev->ifindex,
498 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
499 .flowi4_mark = skb->mark,
500 };
501 int err;
502
503 err = ipmr_fib_lookup(net, &fl4, &mrt);
504 if (err < 0) {
505 kfree_skb(skb);
506 return err;
507 }
508
509 read_lock(&mrt_lock);
510 dev->stats.tx_bytes += skb->len;
511 dev->stats.tx_packets++;
512 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
513 read_unlock(&mrt_lock);
514 kfree_skb(skb);
515 return NETDEV_TX_OK;
516}
517
518static int reg_vif_get_iflink(const struct net_device *dev)
519{
520 return 0;
521}
522
523static const struct net_device_ops reg_vif_netdev_ops = {
524 .ndo_start_xmit = reg_vif_xmit,
525 .ndo_get_iflink = reg_vif_get_iflink,
526};
527
528static void reg_vif_setup(struct net_device *dev)
529{
530 dev->type = ARPHRD_PIMREG;
531 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
532 dev->flags = IFF_NOARP;
533 dev->netdev_ops = ®_vif_netdev_ops;
534 dev->needs_free_netdev = true;
535 dev->features |= NETIF_F_NETNS_LOCAL;
536}
537
538static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
539{
540 struct net_device *dev;
541 char name[IFNAMSIZ];
542
543 if (mrt->id == RT_TABLE_DEFAULT)
544 sprintf(name, "pimreg");
545 else
546 sprintf(name, "pimreg%u", mrt->id);
547
548 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
549
550 if (!dev)
551 return NULL;
552
553 dev_net_set(dev, net);
554
555 if (register_netdevice(dev)) {
556 free_netdev(dev);
557 return NULL;
558 }
559
560 if (!ipmr_init_vif_indev(dev))
561 goto failure;
562 if (dev_open(dev, NULL))
563 goto failure;
564
565 dev_hold(dev);
566
567 return dev;
568
569failure:
570 unregister_netdevice(dev);
571 return NULL;
572}
573
574/* called with rcu_read_lock() */
575static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
576 unsigned int pimlen)
577{
578 struct net_device *reg_dev = NULL;
579 struct iphdr *encap;
580
581 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
582 /* Check that:
583 * a. packet is really sent to a multicast group
584 * b. packet is not a NULL-REGISTER
585 * c. packet is not truncated
586 */
587 if (!ipv4_is_multicast(encap->daddr) ||
588 encap->tot_len == 0 ||
589 ntohs(encap->tot_len) + pimlen > skb->len)
590 return 1;
591
592 read_lock(&mrt_lock);
593 if (mrt->mroute_reg_vif_num >= 0)
594 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
595 read_unlock(&mrt_lock);
596
597 if (!reg_dev)
598 return 1;
599
600 skb->mac_header = skb->network_header;
601 skb_pull(skb, (u8 *)encap - skb->data);
602 skb_reset_network_header(skb);
603 skb->protocol = htons(ETH_P_IP);
604 skb->ip_summed = CHECKSUM_NONE;
605
606 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
607
608 netif_rx(skb);
609
610 return NET_RX_SUCCESS;
611}
612#else
613static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
614{
615 return NULL;
616}
617#endif
618
619static int call_ipmr_vif_entry_notifiers(struct net *net,
620 enum fib_event_type event_type,
621 struct vif_device *vif,
622 vifi_t vif_index, u32 tb_id)
623{
624 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
625 vif, vif_index, tb_id,
626 &net->ipv4.ipmr_seq);
627}
628
629static int call_ipmr_mfc_entry_notifiers(struct net *net,
630 enum fib_event_type event_type,
631 struct mfc_cache *mfc, u32 tb_id)
632{
633 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
634 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
635}
636
637/**
638 * vif_delete - Delete a VIF entry
639 * @mrt: Table to delete from
640 * @vifi: VIF identifier to delete
641 * @notify: Set to 1, if the caller is a notifier_call
642 * @head: if unregistering the VIF, place it on this queue
643 */
644static int vif_delete(struct mr_table *mrt, int vifi, int notify,
645 struct list_head *head)
646{
647 struct net *net = read_pnet(&mrt->net);
648 struct vif_device *v;
649 struct net_device *dev;
650 struct in_device *in_dev;
651
652 if (vifi < 0 || vifi >= mrt->maxvif)
653 return -EADDRNOTAVAIL;
654
655 v = &mrt->vif_table[vifi];
656
657 if (VIF_EXISTS(mrt, vifi))
658 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
659 mrt->id);
660
661 write_lock_bh(&mrt_lock);
662 dev = v->dev;
663 v->dev = NULL;
664
665 if (!dev) {
666 write_unlock_bh(&mrt_lock);
667 return -EADDRNOTAVAIL;
668 }
669
670 if (vifi == mrt->mroute_reg_vif_num)
671 mrt->mroute_reg_vif_num = -1;
672
673 if (vifi + 1 == mrt->maxvif) {
674 int tmp;
675
676 for (tmp = vifi - 1; tmp >= 0; tmp--) {
677 if (VIF_EXISTS(mrt, tmp))
678 break;
679 }
680 mrt->maxvif = tmp+1;
681 }
682
683 write_unlock_bh(&mrt_lock);
684
685 dev_set_allmulti(dev, -1);
686
687 in_dev = __in_dev_get_rtnl(dev);
688 if (in_dev) {
689 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
690 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
691 NETCONFA_MC_FORWARDING,
692 dev->ifindex, &in_dev->cnf);
693 ip_rt_multicast_event(in_dev);
694 }
695
696 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
697 unregister_netdevice_queue(dev, head);
698
699 dev_put(dev);
700 return 0;
701}
702
703static void ipmr_cache_free_rcu(struct rcu_head *head)
704{
705 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
706
707 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
708}
709
710static void ipmr_cache_free(struct mfc_cache *c)
711{
712 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
713}
714
715/* Destroy an unresolved cache entry, killing queued skbs
716 * and reporting error to netlink readers.
717 */
718static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
719{
720 struct net *net = read_pnet(&mrt->net);
721 struct sk_buff *skb;
722 struct nlmsgerr *e;
723
724 atomic_dec(&mrt->cache_resolve_queue_len);
725
726 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
727 if (ip_hdr(skb)->version == 0) {
728 struct nlmsghdr *nlh = skb_pull(skb,
729 sizeof(struct iphdr));
730 nlh->nlmsg_type = NLMSG_ERROR;
731 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
732 skb_trim(skb, nlh->nlmsg_len);
733 e = nlmsg_data(nlh);
734 e->error = -ETIMEDOUT;
735 memset(&e->msg, 0, sizeof(e->msg));
736
737 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
738 } else {
739 kfree_skb(skb);
740 }
741 }
742
743 ipmr_cache_free(c);
744}
745
746/* Timer process for the unresolved queue. */
747static void ipmr_expire_process(struct timer_list *t)
748{
749 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
750 struct mr_mfc *c, *next;
751 unsigned long expires;
752 unsigned long now;
753
754 if (!spin_trylock(&mfc_unres_lock)) {
755 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
756 return;
757 }
758
759 if (list_empty(&mrt->mfc_unres_queue))
760 goto out;
761
762 now = jiffies;
763 expires = 10*HZ;
764
765 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
766 if (time_after(c->mfc_un.unres.expires, now)) {
767 unsigned long interval = c->mfc_un.unres.expires - now;
768 if (interval < expires)
769 expires = interval;
770 continue;
771 }
772
773 list_del(&c->list);
774 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
775 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
776 }
777
778 if (!list_empty(&mrt->mfc_unres_queue))
779 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
780
781out:
782 spin_unlock(&mfc_unres_lock);
783}
784
785/* Fill oifs list. It is called under write locked mrt_lock. */
786static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
787 unsigned char *ttls)
788{
789 int vifi;
790
791 cache->mfc_un.res.minvif = MAXVIFS;
792 cache->mfc_un.res.maxvif = 0;
793 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
794
795 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
796 if (VIF_EXISTS(mrt, vifi) &&
797 ttls[vifi] && ttls[vifi] < 255) {
798 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
799 if (cache->mfc_un.res.minvif > vifi)
800 cache->mfc_un.res.minvif = vifi;
801 if (cache->mfc_un.res.maxvif <= vifi)
802 cache->mfc_un.res.maxvif = vifi + 1;
803 }
804 }
805 cache->mfc_un.res.lastuse = jiffies;
806}
807
808static int vif_add(struct net *net, struct mr_table *mrt,
809 struct vifctl *vifc, int mrtsock)
810{
811 struct netdev_phys_item_id ppid = { };
812 int vifi = vifc->vifc_vifi;
813 struct vif_device *v = &mrt->vif_table[vifi];
814 struct net_device *dev;
815 struct in_device *in_dev;
816 int err;
817
818 /* Is vif busy ? */
819 if (VIF_EXISTS(mrt, vifi))
820 return -EADDRINUSE;
821
822 switch (vifc->vifc_flags) {
823 case VIFF_REGISTER:
824 if (!ipmr_pimsm_enabled())
825 return -EINVAL;
826 /* Special Purpose VIF in PIM
827 * All the packets will be sent to the daemon
828 */
829 if (mrt->mroute_reg_vif_num >= 0)
830 return -EADDRINUSE;
831 dev = ipmr_reg_vif(net, mrt);
832 if (!dev)
833 return -ENOBUFS;
834 err = dev_set_allmulti(dev, 1);
835 if (err) {
836 unregister_netdevice(dev);
837 dev_put(dev);
838 return err;
839 }
840 break;
841 case VIFF_TUNNEL:
842 dev = ipmr_new_tunnel(net, vifc);
843 if (IS_ERR(dev))
844 return PTR_ERR(dev);
845 break;
846 case VIFF_USE_IFINDEX:
847 case 0:
848 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
849 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
850 if (dev && !__in_dev_get_rtnl(dev)) {
851 dev_put(dev);
852 return -EADDRNOTAVAIL;
853 }
854 } else {
855 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
856 }
857 if (!dev)
858 return -EADDRNOTAVAIL;
859 err = dev_set_allmulti(dev, 1);
860 if (err) {
861 dev_put(dev);
862 return err;
863 }
864 break;
865 default:
866 return -EINVAL;
867 }
868
869 in_dev = __in_dev_get_rtnl(dev);
870 if (!in_dev) {
871 dev_put(dev);
872 return -EADDRNOTAVAIL;
873 }
874 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
875 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
876 dev->ifindex, &in_dev->cnf);
877 ip_rt_multicast_event(in_dev);
878
879 /* Fill in the VIF structures */
880 vif_device_init(v, dev, vifc->vifc_rate_limit,
881 vifc->vifc_threshold,
882 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
883 (VIFF_TUNNEL | VIFF_REGISTER));
884
885 err = dev_get_port_parent_id(dev, &ppid, true);
886 if (err == 0) {
887 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
888 v->dev_parent_id.id_len = ppid.id_len;
889 } else {
890 v->dev_parent_id.id_len = 0;
891 }
892
893 v->local = vifc->vifc_lcl_addr.s_addr;
894 v->remote = vifc->vifc_rmt_addr.s_addr;
895
896 /* And finish update writing critical data */
897 write_lock_bh(&mrt_lock);
898 v->dev = dev;
899 if (v->flags & VIFF_REGISTER)
900 mrt->mroute_reg_vif_num = vifi;
901 if (vifi+1 > mrt->maxvif)
902 mrt->maxvif = vifi+1;
903 write_unlock_bh(&mrt_lock);
904 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
905 return 0;
906}
907
908/* called with rcu_read_lock() */
909static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
910 __be32 origin,
911 __be32 mcastgrp)
912{
913 struct mfc_cache_cmp_arg arg = {
914 .mfc_mcastgrp = mcastgrp,
915 .mfc_origin = origin
916 };
917
918 return mr_mfc_find(mrt, &arg);
919}
920
921/* Look for a (*,G) entry */
922static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
923 __be32 mcastgrp, int vifi)
924{
925 struct mfc_cache_cmp_arg arg = {
926 .mfc_mcastgrp = mcastgrp,
927 .mfc_origin = htonl(INADDR_ANY)
928 };
929
930 if (mcastgrp == htonl(INADDR_ANY))
931 return mr_mfc_find_any_parent(mrt, vifi);
932 return mr_mfc_find_any(mrt, vifi, &arg);
933}
934
935/* Look for a (S,G,iif) entry if parent != -1 */
936static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
937 __be32 origin, __be32 mcastgrp,
938 int parent)
939{
940 struct mfc_cache_cmp_arg arg = {
941 .mfc_mcastgrp = mcastgrp,
942 .mfc_origin = origin,
943 };
944
945 return mr_mfc_find_parent(mrt, &arg, parent);
946}
947
948/* Allocate a multicast cache entry */
949static struct mfc_cache *ipmr_cache_alloc(void)
950{
951 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
952
953 if (c) {
954 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
955 c->_c.mfc_un.res.minvif = MAXVIFS;
956 c->_c.free = ipmr_cache_free_rcu;
957 refcount_set(&c->_c.mfc_un.res.refcount, 1);
958 }
959 return c;
960}
961
962static struct mfc_cache *ipmr_cache_alloc_unres(void)
963{
964 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
965
966 if (c) {
967 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
968 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
969 }
970 return c;
971}
972
973/* A cache entry has gone into a resolved state from queued */
974static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
975 struct mfc_cache *uc, struct mfc_cache *c)
976{
977 struct sk_buff *skb;
978 struct nlmsgerr *e;
979
980 /* Play the pending entries through our router */
981 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
982 if (ip_hdr(skb)->version == 0) {
983 struct nlmsghdr *nlh = skb_pull(skb,
984 sizeof(struct iphdr));
985
986 if (mr_fill_mroute(mrt, skb, &c->_c,
987 nlmsg_data(nlh)) > 0) {
988 nlh->nlmsg_len = skb_tail_pointer(skb) -
989 (u8 *)nlh;
990 } else {
991 nlh->nlmsg_type = NLMSG_ERROR;
992 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
993 skb_trim(skb, nlh->nlmsg_len);
994 e = nlmsg_data(nlh);
995 e->error = -EMSGSIZE;
996 memset(&e->msg, 0, sizeof(e->msg));
997 }
998
999 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1000 } else {
1001 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1002 }
1003 }
1004}
1005
1006/* Bounce a cache query up to mrouted and netlink.
1007 *
1008 * Called under mrt_lock.
1009 */
1010static int ipmr_cache_report(struct mr_table *mrt,
1011 struct sk_buff *pkt, vifi_t vifi, int assert)
1012{
1013 const int ihl = ip_hdrlen(pkt);
1014 struct sock *mroute_sk;
1015 struct igmphdr *igmp;
1016 struct igmpmsg *msg;
1017 struct sk_buff *skb;
1018 int ret;
1019
1020 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1021 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1022 else
1023 skb = alloc_skb(128, GFP_ATOMIC);
1024
1025 if (!skb)
1026 return -ENOBUFS;
1027
1028 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1029 /* Ugly, but we have no choice with this interface.
1030 * Duplicate old header, fix ihl, length etc.
1031 * And all this only to mangle msg->im_msgtype and
1032 * to set msg->im_mbz to "mbz" :-)
1033 */
1034 skb_push(skb, sizeof(struct iphdr));
1035 skb_reset_network_header(skb);
1036 skb_reset_transport_header(skb);
1037 msg = (struct igmpmsg *)skb_network_header(skb);
1038 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1039 msg->im_msgtype = assert;
1040 msg->im_mbz = 0;
1041 if (assert == IGMPMSG_WRVIFWHOLE)
1042 msg->im_vif = vifi;
1043 else
1044 msg->im_vif = mrt->mroute_reg_vif_num;
1045 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1046 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1047 sizeof(struct iphdr));
1048 } else {
1049 /* Copy the IP header */
1050 skb_set_network_header(skb, skb->len);
1051 skb_put(skb, ihl);
1052 skb_copy_to_linear_data(skb, pkt->data, ihl);
1053 /* Flag to the kernel this is a route add */
1054 ip_hdr(skb)->protocol = 0;
1055 msg = (struct igmpmsg *)skb_network_header(skb);
1056 msg->im_vif = vifi;
1057 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1058 /* Add our header */
1059 igmp = skb_put(skb, sizeof(struct igmphdr));
1060 igmp->type = assert;
1061 msg->im_msgtype = assert;
1062 igmp->code = 0;
1063 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1064 skb->transport_header = skb->network_header;
1065 }
1066
1067 rcu_read_lock();
1068 mroute_sk = rcu_dereference(mrt->mroute_sk);
1069 if (!mroute_sk) {
1070 rcu_read_unlock();
1071 kfree_skb(skb);
1072 return -EINVAL;
1073 }
1074
1075 igmpmsg_netlink_event(mrt, skb);
1076
1077 /* Deliver to mrouted */
1078 ret = sock_queue_rcv_skb(mroute_sk, skb);
1079 rcu_read_unlock();
1080 if (ret < 0) {
1081 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1082 kfree_skb(skb);
1083 }
1084
1085 return ret;
1086}
1087
1088/* Queue a packet for resolution. It gets locked cache entry! */
1089static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1090 struct sk_buff *skb, struct net_device *dev)
1091{
1092 const struct iphdr *iph = ip_hdr(skb);
1093 struct mfc_cache *c;
1094 bool found = false;
1095 int err;
1096
1097 spin_lock_bh(&mfc_unres_lock);
1098 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1099 if (c->mfc_mcastgrp == iph->daddr &&
1100 c->mfc_origin == iph->saddr) {
1101 found = true;
1102 break;
1103 }
1104 }
1105
1106 if (!found) {
1107 /* Create a new entry if allowable */
1108 c = ipmr_cache_alloc_unres();
1109 if (!c) {
1110 spin_unlock_bh(&mfc_unres_lock);
1111
1112 kfree_skb(skb);
1113 return -ENOBUFS;
1114 }
1115
1116 /* Fill in the new cache entry */
1117 c->_c.mfc_parent = -1;
1118 c->mfc_origin = iph->saddr;
1119 c->mfc_mcastgrp = iph->daddr;
1120
1121 /* Reflect first query at mrouted. */
1122 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1123
1124 if (err < 0) {
1125 /* If the report failed throw the cache entry
1126 out - Brad Parker
1127 */
1128 spin_unlock_bh(&mfc_unres_lock);
1129
1130 ipmr_cache_free(c);
1131 kfree_skb(skb);
1132 return err;
1133 }
1134
1135 atomic_inc(&mrt->cache_resolve_queue_len);
1136 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1137 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1138
1139 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1140 mod_timer(&mrt->ipmr_expire_timer,
1141 c->_c.mfc_un.unres.expires);
1142 }
1143
1144 /* See if we can append the packet */
1145 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1146 kfree_skb(skb);
1147 err = -ENOBUFS;
1148 } else {
1149 if (dev) {
1150 skb->dev = dev;
1151 skb->skb_iif = dev->ifindex;
1152 }
1153 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1154 err = 0;
1155 }
1156
1157 spin_unlock_bh(&mfc_unres_lock);
1158 return err;
1159}
1160
1161/* MFC cache manipulation by user space mroute daemon */
1162
1163static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1164{
1165 struct net *net = read_pnet(&mrt->net);
1166 struct mfc_cache *c;
1167
1168 /* The entries are added/deleted only under RTNL */
1169 rcu_read_lock();
1170 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1171 mfc->mfcc_mcastgrp.s_addr, parent);
1172 rcu_read_unlock();
1173 if (!c)
1174 return -ENOENT;
1175 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1176 list_del_rcu(&c->_c.list);
1177 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1178 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1179 mr_cache_put(&c->_c);
1180
1181 return 0;
1182}
1183
1184static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1185 struct mfcctl *mfc, int mrtsock, int parent)
1186{
1187 struct mfc_cache *uc, *c;
1188 struct mr_mfc *_uc;
1189 bool found;
1190 int ret;
1191
1192 if (mfc->mfcc_parent >= MAXVIFS)
1193 return -ENFILE;
1194
1195 /* The entries are added/deleted only under RTNL */
1196 rcu_read_lock();
1197 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1198 mfc->mfcc_mcastgrp.s_addr, parent);
1199 rcu_read_unlock();
1200 if (c) {
1201 write_lock_bh(&mrt_lock);
1202 c->_c.mfc_parent = mfc->mfcc_parent;
1203 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1204 if (!mrtsock)
1205 c->_c.mfc_flags |= MFC_STATIC;
1206 write_unlock_bh(&mrt_lock);
1207 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1208 mrt->id);
1209 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1210 return 0;
1211 }
1212
1213 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1214 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1215 return -EINVAL;
1216
1217 c = ipmr_cache_alloc();
1218 if (!c)
1219 return -ENOMEM;
1220
1221 c->mfc_origin = mfc->mfcc_origin.s_addr;
1222 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1223 c->_c.mfc_parent = mfc->mfcc_parent;
1224 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1225 if (!mrtsock)
1226 c->_c.mfc_flags |= MFC_STATIC;
1227
1228 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1229 ipmr_rht_params);
1230 if (ret) {
1231 pr_err("ipmr: rhtable insert error %d\n", ret);
1232 ipmr_cache_free(c);
1233 return ret;
1234 }
1235 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1236 /* Check to see if we resolved a queued list. If so we
1237 * need to send on the frames and tidy up.
1238 */
1239 found = false;
1240 spin_lock_bh(&mfc_unres_lock);
1241 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1242 uc = (struct mfc_cache *)_uc;
1243 if (uc->mfc_origin == c->mfc_origin &&
1244 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1245 list_del(&_uc->list);
1246 atomic_dec(&mrt->cache_resolve_queue_len);
1247 found = true;
1248 break;
1249 }
1250 }
1251 if (list_empty(&mrt->mfc_unres_queue))
1252 del_timer(&mrt->ipmr_expire_timer);
1253 spin_unlock_bh(&mfc_unres_lock);
1254
1255 if (found) {
1256 ipmr_cache_resolve(net, mrt, uc, c);
1257 ipmr_cache_free(uc);
1258 }
1259 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1260 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1261 return 0;
1262}
1263
1264/* Close the multicast socket, and clear the vif tables etc */
1265static void mroute_clean_tables(struct mr_table *mrt, int flags)
1266{
1267 struct net *net = read_pnet(&mrt->net);
1268 struct mr_mfc *c, *tmp;
1269 struct mfc_cache *cache;
1270 LIST_HEAD(list);
1271 int i;
1272
1273 /* Shut down all active vif entries */
1274 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1275 for (i = 0; i < mrt->maxvif; i++) {
1276 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1277 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1278 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1279 continue;
1280 vif_delete(mrt, i, 0, &list);
1281 }
1282 unregister_netdevice_many(&list);
1283 }
1284
1285 /* Wipe the cache */
1286 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1287 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1288 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1289 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1290 continue;
1291 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1292 list_del_rcu(&c->list);
1293 cache = (struct mfc_cache *)c;
1294 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1295 mrt->id);
1296 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1297 mr_cache_put(c);
1298 }
1299 }
1300
1301 if (flags & MRT_FLUSH_MFC) {
1302 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1303 spin_lock_bh(&mfc_unres_lock);
1304 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1305 list_del(&c->list);
1306 cache = (struct mfc_cache *)c;
1307 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1308 ipmr_destroy_unres(mrt, cache);
1309 }
1310 spin_unlock_bh(&mfc_unres_lock);
1311 }
1312 }
1313}
1314
1315/* called from ip_ra_control(), before an RCU grace period,
1316 * we dont need to call synchronize_rcu() here
1317 */
1318static void mrtsock_destruct(struct sock *sk)
1319{
1320 struct net *net = sock_net(sk);
1321 struct mr_table *mrt;
1322
1323 rtnl_lock();
1324 ipmr_for_each_table(mrt, net) {
1325 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1326 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1327 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1328 NETCONFA_MC_FORWARDING,
1329 NETCONFA_IFINDEX_ALL,
1330 net->ipv4.devconf_all);
1331 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1332 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1333 }
1334 }
1335 rtnl_unlock();
1336}
1337
1338/* Socket options and virtual interface manipulation. The whole
1339 * virtual interface system is a complete heap, but unfortunately
1340 * that's how BSD mrouted happens to think. Maybe one day with a proper
1341 * MOSPF/PIM router set up we can clean this up.
1342 */
1343
1344int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1345 unsigned int optlen)
1346{
1347 struct net *net = sock_net(sk);
1348 int val, ret = 0, parent = 0;
1349 struct mr_table *mrt;
1350 struct vifctl vif;
1351 struct mfcctl mfc;
1352 bool do_wrvifwhole;
1353 u32 uval;
1354
1355 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1356 rtnl_lock();
1357 if (sk->sk_type != SOCK_RAW ||
1358 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1359 ret = -EOPNOTSUPP;
1360 goto out_unlock;
1361 }
1362
1363 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1364 if (!mrt) {
1365 ret = -ENOENT;
1366 goto out_unlock;
1367 }
1368 if (optname != MRT_INIT) {
1369 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1370 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1371 ret = -EACCES;
1372 goto out_unlock;
1373 }
1374 }
1375
1376 switch (optname) {
1377 case MRT_INIT:
1378 if (optlen != sizeof(int)) {
1379 ret = -EINVAL;
1380 break;
1381 }
1382 if (rtnl_dereference(mrt->mroute_sk)) {
1383 ret = -EADDRINUSE;
1384 break;
1385 }
1386
1387 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1388 if (ret == 0) {
1389 rcu_assign_pointer(mrt->mroute_sk, sk);
1390 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1391 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1392 NETCONFA_MC_FORWARDING,
1393 NETCONFA_IFINDEX_ALL,
1394 net->ipv4.devconf_all);
1395 }
1396 break;
1397 case MRT_DONE:
1398 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1399 ret = -EACCES;
1400 } else {
1401 /* We need to unlock here because mrtsock_destruct takes
1402 * care of rtnl itself and we can't change that due to
1403 * the IP_ROUTER_ALERT setsockopt which runs without it.
1404 */
1405 rtnl_unlock();
1406 ret = ip_ra_control(sk, 0, NULL);
1407 goto out;
1408 }
1409 break;
1410 case MRT_ADD_VIF:
1411 case MRT_DEL_VIF:
1412 if (optlen != sizeof(vif)) {
1413 ret = -EINVAL;
1414 break;
1415 }
1416 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1417 ret = -EFAULT;
1418 break;
1419 }
1420 if (vif.vifc_vifi >= MAXVIFS) {
1421 ret = -ENFILE;
1422 break;
1423 }
1424 if (optname == MRT_ADD_VIF) {
1425 ret = vif_add(net, mrt, &vif,
1426 sk == rtnl_dereference(mrt->mroute_sk));
1427 } else {
1428 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1429 }
1430 break;
1431 /* Manipulate the forwarding caches. These live
1432 * in a sort of kernel/user symbiosis.
1433 */
1434 case MRT_ADD_MFC:
1435 case MRT_DEL_MFC:
1436 parent = -1;
1437 fallthrough;
1438 case MRT_ADD_MFC_PROXY:
1439 case MRT_DEL_MFC_PROXY:
1440 if (optlen != sizeof(mfc)) {
1441 ret = -EINVAL;
1442 break;
1443 }
1444 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1445 ret = -EFAULT;
1446 break;
1447 }
1448 if (parent == 0)
1449 parent = mfc.mfcc_parent;
1450 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1451 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1452 else
1453 ret = ipmr_mfc_add(net, mrt, &mfc,
1454 sk == rtnl_dereference(mrt->mroute_sk),
1455 parent);
1456 break;
1457 case MRT_FLUSH:
1458 if (optlen != sizeof(val)) {
1459 ret = -EINVAL;
1460 break;
1461 }
1462 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1463 ret = -EFAULT;
1464 break;
1465 }
1466 mroute_clean_tables(mrt, val);
1467 break;
1468 /* Control PIM assert. */
1469 case MRT_ASSERT:
1470 if (optlen != sizeof(val)) {
1471 ret = -EINVAL;
1472 break;
1473 }
1474 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1475 ret = -EFAULT;
1476 break;
1477 }
1478 mrt->mroute_do_assert = val;
1479 break;
1480 case MRT_PIM:
1481 if (!ipmr_pimsm_enabled()) {
1482 ret = -ENOPROTOOPT;
1483 break;
1484 }
1485 if (optlen != sizeof(val)) {
1486 ret = -EINVAL;
1487 break;
1488 }
1489 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1490 ret = -EFAULT;
1491 break;
1492 }
1493
1494 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1495 val = !!val;
1496 if (val != mrt->mroute_do_pim) {
1497 mrt->mroute_do_pim = val;
1498 mrt->mroute_do_assert = val;
1499 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1500 }
1501 break;
1502 case MRT_TABLE:
1503 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1504 ret = -ENOPROTOOPT;
1505 break;
1506 }
1507 if (optlen != sizeof(uval)) {
1508 ret = -EINVAL;
1509 break;
1510 }
1511 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1512 ret = -EFAULT;
1513 break;
1514 }
1515
1516 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1517 ret = -EBUSY;
1518 } else {
1519 mrt = ipmr_new_table(net, uval);
1520 if (IS_ERR(mrt))
1521 ret = PTR_ERR(mrt);
1522 else
1523 raw_sk(sk)->ipmr_table = uval;
1524 }
1525 break;
1526 /* Spurious command, or MRT_VERSION which you cannot set. */
1527 default:
1528 ret = -ENOPROTOOPT;
1529 }
1530out_unlock:
1531 rtnl_unlock();
1532out:
1533 return ret;
1534}
1535
1536/* Getsock opt support for the multicast routing system. */
1537int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1538{
1539 int olr;
1540 int val;
1541 struct net *net = sock_net(sk);
1542 struct mr_table *mrt;
1543
1544 if (sk->sk_type != SOCK_RAW ||
1545 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1546 return -EOPNOTSUPP;
1547
1548 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1549 if (!mrt)
1550 return -ENOENT;
1551
1552 switch (optname) {
1553 case MRT_VERSION:
1554 val = 0x0305;
1555 break;
1556 case MRT_PIM:
1557 if (!ipmr_pimsm_enabled())
1558 return -ENOPROTOOPT;
1559 val = mrt->mroute_do_pim;
1560 break;
1561 case MRT_ASSERT:
1562 val = mrt->mroute_do_assert;
1563 break;
1564 default:
1565 return -ENOPROTOOPT;
1566 }
1567
1568 if (get_user(olr, optlen))
1569 return -EFAULT;
1570 olr = min_t(unsigned int, olr, sizeof(int));
1571 if (olr < 0)
1572 return -EINVAL;
1573 if (put_user(olr, optlen))
1574 return -EFAULT;
1575 if (copy_to_user(optval, &val, olr))
1576 return -EFAULT;
1577 return 0;
1578}
1579
1580/* The IP multicast ioctl support routines. */
1581int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1582{
1583 struct sioc_sg_req sr;
1584 struct sioc_vif_req vr;
1585 struct vif_device *vif;
1586 struct mfc_cache *c;
1587 struct net *net = sock_net(sk);
1588 struct mr_table *mrt;
1589
1590 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1591 if (!mrt)
1592 return -ENOENT;
1593
1594 switch (cmd) {
1595 case SIOCGETVIFCNT:
1596 if (copy_from_user(&vr, arg, sizeof(vr)))
1597 return -EFAULT;
1598 if (vr.vifi >= mrt->maxvif)
1599 return -EINVAL;
1600 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1601 read_lock(&mrt_lock);
1602 vif = &mrt->vif_table[vr.vifi];
1603 if (VIF_EXISTS(mrt, vr.vifi)) {
1604 vr.icount = vif->pkt_in;
1605 vr.ocount = vif->pkt_out;
1606 vr.ibytes = vif->bytes_in;
1607 vr.obytes = vif->bytes_out;
1608 read_unlock(&mrt_lock);
1609
1610 if (copy_to_user(arg, &vr, sizeof(vr)))
1611 return -EFAULT;
1612 return 0;
1613 }
1614 read_unlock(&mrt_lock);
1615 return -EADDRNOTAVAIL;
1616 case SIOCGETSGCNT:
1617 if (copy_from_user(&sr, arg, sizeof(sr)))
1618 return -EFAULT;
1619
1620 rcu_read_lock();
1621 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1622 if (c) {
1623 sr.pktcnt = c->_c.mfc_un.res.pkt;
1624 sr.bytecnt = c->_c.mfc_un.res.bytes;
1625 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1626 rcu_read_unlock();
1627
1628 if (copy_to_user(arg, &sr, sizeof(sr)))
1629 return -EFAULT;
1630 return 0;
1631 }
1632 rcu_read_unlock();
1633 return -EADDRNOTAVAIL;
1634 default:
1635 return -ENOIOCTLCMD;
1636 }
1637}
1638
1639#ifdef CONFIG_COMPAT
1640struct compat_sioc_sg_req {
1641 struct in_addr src;
1642 struct in_addr grp;
1643 compat_ulong_t pktcnt;
1644 compat_ulong_t bytecnt;
1645 compat_ulong_t wrong_if;
1646};
1647
1648struct compat_sioc_vif_req {
1649 vifi_t vifi; /* Which iface */
1650 compat_ulong_t icount;
1651 compat_ulong_t ocount;
1652 compat_ulong_t ibytes;
1653 compat_ulong_t obytes;
1654};
1655
1656int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1657{
1658 struct compat_sioc_sg_req sr;
1659 struct compat_sioc_vif_req vr;
1660 struct vif_device *vif;
1661 struct mfc_cache *c;
1662 struct net *net = sock_net(sk);
1663 struct mr_table *mrt;
1664
1665 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1666 if (!mrt)
1667 return -ENOENT;
1668
1669 switch (cmd) {
1670 case SIOCGETVIFCNT:
1671 if (copy_from_user(&vr, arg, sizeof(vr)))
1672 return -EFAULT;
1673 if (vr.vifi >= mrt->maxvif)
1674 return -EINVAL;
1675 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1676 read_lock(&mrt_lock);
1677 vif = &mrt->vif_table[vr.vifi];
1678 if (VIF_EXISTS(mrt, vr.vifi)) {
1679 vr.icount = vif->pkt_in;
1680 vr.ocount = vif->pkt_out;
1681 vr.ibytes = vif->bytes_in;
1682 vr.obytes = vif->bytes_out;
1683 read_unlock(&mrt_lock);
1684
1685 if (copy_to_user(arg, &vr, sizeof(vr)))
1686 return -EFAULT;
1687 return 0;
1688 }
1689 read_unlock(&mrt_lock);
1690 return -EADDRNOTAVAIL;
1691 case SIOCGETSGCNT:
1692 if (copy_from_user(&sr, arg, sizeof(sr)))
1693 return -EFAULT;
1694
1695 rcu_read_lock();
1696 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1697 if (c) {
1698 sr.pktcnt = c->_c.mfc_un.res.pkt;
1699 sr.bytecnt = c->_c.mfc_un.res.bytes;
1700 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1701 rcu_read_unlock();
1702
1703 if (copy_to_user(arg, &sr, sizeof(sr)))
1704 return -EFAULT;
1705 return 0;
1706 }
1707 rcu_read_unlock();
1708 return -EADDRNOTAVAIL;
1709 default:
1710 return -ENOIOCTLCMD;
1711 }
1712}
1713#endif
1714
1715static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1716{
1717 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1718 struct net *net = dev_net(dev);
1719 struct mr_table *mrt;
1720 struct vif_device *v;
1721 int ct;
1722
1723 if (event != NETDEV_UNREGISTER)
1724 return NOTIFY_DONE;
1725
1726 ipmr_for_each_table(mrt, net) {
1727 v = &mrt->vif_table[0];
1728 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1729 if (v->dev == dev)
1730 vif_delete(mrt, ct, 1, NULL);
1731 }
1732 }
1733 return NOTIFY_DONE;
1734}
1735
1736static struct notifier_block ip_mr_notifier = {
1737 .notifier_call = ipmr_device_event,
1738};
1739
1740/* Encapsulate a packet by attaching a valid IPIP header to it.
1741 * This avoids tunnel drivers and other mess and gives us the speed so
1742 * important for multicast video.
1743 */
1744static void ip_encap(struct net *net, struct sk_buff *skb,
1745 __be32 saddr, __be32 daddr)
1746{
1747 struct iphdr *iph;
1748 const struct iphdr *old_iph = ip_hdr(skb);
1749
1750 skb_push(skb, sizeof(struct iphdr));
1751 skb->transport_header = skb->network_header;
1752 skb_reset_network_header(skb);
1753 iph = ip_hdr(skb);
1754
1755 iph->version = 4;
1756 iph->tos = old_iph->tos;
1757 iph->ttl = old_iph->ttl;
1758 iph->frag_off = 0;
1759 iph->daddr = daddr;
1760 iph->saddr = saddr;
1761 iph->protocol = IPPROTO_IPIP;
1762 iph->ihl = 5;
1763 iph->tot_len = htons(skb->len);
1764 ip_select_ident(net, skb, NULL);
1765 ip_send_check(iph);
1766
1767 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1768 nf_reset_ct(skb);
1769}
1770
1771static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1772 struct sk_buff *skb)
1773{
1774 struct ip_options *opt = &(IPCB(skb)->opt);
1775
1776 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1777 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1778
1779 if (unlikely(opt->optlen))
1780 ip_forward_options(skb);
1781
1782 return dst_output(net, sk, skb);
1783}
1784
1785#ifdef CONFIG_NET_SWITCHDEV
1786static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1787 int in_vifi, int out_vifi)
1788{
1789 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1790 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1791
1792 if (!skb->offload_l3_fwd_mark)
1793 return false;
1794 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1795 return false;
1796 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1797 &in_vif->dev_parent_id);
1798}
1799#else
1800static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1801 int in_vifi, int out_vifi)
1802{
1803 return false;
1804}
1805#endif
1806
1807/* Processing handlers for ipmr_forward */
1808
1809static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1810 int in_vifi, struct sk_buff *skb, int vifi)
1811{
1812 const struct iphdr *iph = ip_hdr(skb);
1813 struct vif_device *vif = &mrt->vif_table[vifi];
1814 struct net_device *dev;
1815 struct rtable *rt;
1816 struct flowi4 fl4;
1817 int encap = 0;
1818
1819 if (!vif->dev)
1820 goto out_free;
1821
1822 if (vif->flags & VIFF_REGISTER) {
1823 vif->pkt_out++;
1824 vif->bytes_out += skb->len;
1825 vif->dev->stats.tx_bytes += skb->len;
1826 vif->dev->stats.tx_packets++;
1827 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1828 goto out_free;
1829 }
1830
1831 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1832 goto out_free;
1833
1834 if (vif->flags & VIFF_TUNNEL) {
1835 rt = ip_route_output_ports(net, &fl4, NULL,
1836 vif->remote, vif->local,
1837 0, 0,
1838 IPPROTO_IPIP,
1839 RT_TOS(iph->tos), vif->link);
1840 if (IS_ERR(rt))
1841 goto out_free;
1842 encap = sizeof(struct iphdr);
1843 } else {
1844 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1845 0, 0,
1846 IPPROTO_IPIP,
1847 RT_TOS(iph->tos), vif->link);
1848 if (IS_ERR(rt))
1849 goto out_free;
1850 }
1851
1852 dev = rt->dst.dev;
1853
1854 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1855 /* Do not fragment multicasts. Alas, IPv4 does not
1856 * allow to send ICMP, so that packets will disappear
1857 * to blackhole.
1858 */
1859 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1860 ip_rt_put(rt);
1861 goto out_free;
1862 }
1863
1864 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1865
1866 if (skb_cow(skb, encap)) {
1867 ip_rt_put(rt);
1868 goto out_free;
1869 }
1870
1871 vif->pkt_out++;
1872 vif->bytes_out += skb->len;
1873
1874 skb_dst_drop(skb);
1875 skb_dst_set(skb, &rt->dst);
1876 ip_decrease_ttl(ip_hdr(skb));
1877
1878 /* FIXME: forward and output firewalls used to be called here.
1879 * What do we do with netfilter? -- RR
1880 */
1881 if (vif->flags & VIFF_TUNNEL) {
1882 ip_encap(net, skb, vif->local, vif->remote);
1883 /* FIXME: extra output firewall step used to be here. --RR */
1884 vif->dev->stats.tx_packets++;
1885 vif->dev->stats.tx_bytes += skb->len;
1886 }
1887
1888 IPCB(skb)->flags |= IPSKB_FORWARDED;
1889
1890 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1891 * not only before forwarding, but after forwarding on all output
1892 * interfaces. It is clear, if mrouter runs a multicasting
1893 * program, it should receive packets not depending to what interface
1894 * program is joined.
1895 * If we will not make it, the program will have to join on all
1896 * interfaces. On the other hand, multihoming host (or router, but
1897 * not mrouter) cannot join to more than one interface - it will
1898 * result in receiving multiple packets.
1899 */
1900 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1901 net, NULL, skb, skb->dev, dev,
1902 ipmr_forward_finish);
1903 return;
1904
1905out_free:
1906 kfree_skb(skb);
1907}
1908
1909static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1910{
1911 int ct;
1912
1913 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1914 if (mrt->vif_table[ct].dev == dev)
1915 break;
1916 }
1917 return ct;
1918}
1919
1920/* "local" means that we should preserve one skb (for local delivery) */
1921static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1922 struct net_device *dev, struct sk_buff *skb,
1923 struct mfc_cache *c, int local)
1924{
1925 int true_vifi = ipmr_find_vif(mrt, dev);
1926 int psend = -1;
1927 int vif, ct;
1928
1929 vif = c->_c.mfc_parent;
1930 c->_c.mfc_un.res.pkt++;
1931 c->_c.mfc_un.res.bytes += skb->len;
1932 c->_c.mfc_un.res.lastuse = jiffies;
1933
1934 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1935 struct mfc_cache *cache_proxy;
1936
1937 /* For an (*,G) entry, we only check that the incomming
1938 * interface is part of the static tree.
1939 */
1940 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1941 if (cache_proxy &&
1942 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1943 goto forward;
1944 }
1945
1946 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1947 if (mrt->vif_table[vif].dev != dev) {
1948 if (rt_is_output_route(skb_rtable(skb))) {
1949 /* It is our own packet, looped back.
1950 * Very complicated situation...
1951 *
1952 * The best workaround until routing daemons will be
1953 * fixed is not to redistribute packet, if it was
1954 * send through wrong interface. It means, that
1955 * multicast applications WILL NOT work for
1956 * (S,G), which have default multicast route pointing
1957 * to wrong oif. In any case, it is not a good
1958 * idea to use multicasting applications on router.
1959 */
1960 goto dont_forward;
1961 }
1962
1963 c->_c.mfc_un.res.wrong_if++;
1964
1965 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1966 /* pimsm uses asserts, when switching from RPT to SPT,
1967 * so that we cannot check that packet arrived on an oif.
1968 * It is bad, but otherwise we would need to move pretty
1969 * large chunk of pimd to kernel. Ough... --ANK
1970 */
1971 (mrt->mroute_do_pim ||
1972 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1973 time_after(jiffies,
1974 c->_c.mfc_un.res.last_assert +
1975 MFC_ASSERT_THRESH)) {
1976 c->_c.mfc_un.res.last_assert = jiffies;
1977 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1978 if (mrt->mroute_do_wrvifwhole)
1979 ipmr_cache_report(mrt, skb, true_vifi,
1980 IGMPMSG_WRVIFWHOLE);
1981 }
1982 goto dont_forward;
1983 }
1984
1985forward:
1986 mrt->vif_table[vif].pkt_in++;
1987 mrt->vif_table[vif].bytes_in += skb->len;
1988
1989 /* Forward the frame */
1990 if (c->mfc_origin == htonl(INADDR_ANY) &&
1991 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
1992 if (true_vifi >= 0 &&
1993 true_vifi != c->_c.mfc_parent &&
1994 ip_hdr(skb)->ttl >
1995 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
1996 /* It's an (*,*) entry and the packet is not coming from
1997 * the upstream: forward the packet to the upstream
1998 * only.
1999 */
2000 psend = c->_c.mfc_parent;
2001 goto last_forward;
2002 }
2003 goto dont_forward;
2004 }
2005 for (ct = c->_c.mfc_un.res.maxvif - 1;
2006 ct >= c->_c.mfc_un.res.minvif; ct--) {
2007 /* For (*,G) entry, don't forward to the incoming interface */
2008 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2009 ct != true_vifi) &&
2010 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2011 if (psend != -1) {
2012 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2013
2014 if (skb2)
2015 ipmr_queue_xmit(net, mrt, true_vifi,
2016 skb2, psend);
2017 }
2018 psend = ct;
2019 }
2020 }
2021last_forward:
2022 if (psend != -1) {
2023 if (local) {
2024 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2025
2026 if (skb2)
2027 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2028 psend);
2029 } else {
2030 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2031 return;
2032 }
2033 }
2034
2035dont_forward:
2036 if (!local)
2037 kfree_skb(skb);
2038}
2039
2040static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2041{
2042 struct rtable *rt = skb_rtable(skb);
2043 struct iphdr *iph = ip_hdr(skb);
2044 struct flowi4 fl4 = {
2045 .daddr = iph->daddr,
2046 .saddr = iph->saddr,
2047 .flowi4_tos = RT_TOS(iph->tos),
2048 .flowi4_oif = (rt_is_output_route(rt) ?
2049 skb->dev->ifindex : 0),
2050 .flowi4_iif = (rt_is_output_route(rt) ?
2051 LOOPBACK_IFINDEX :
2052 skb->dev->ifindex),
2053 .flowi4_mark = skb->mark,
2054 };
2055 struct mr_table *mrt;
2056 int err;
2057
2058 err = ipmr_fib_lookup(net, &fl4, &mrt);
2059 if (err)
2060 return ERR_PTR(err);
2061 return mrt;
2062}
2063
2064/* Multicast packets for forwarding arrive here
2065 * Called with rcu_read_lock();
2066 */
2067int ip_mr_input(struct sk_buff *skb)
2068{
2069 struct mfc_cache *cache;
2070 struct net *net = dev_net(skb->dev);
2071 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2072 struct mr_table *mrt;
2073 struct net_device *dev;
2074
2075 /* skb->dev passed in is the loX master dev for vrfs.
2076 * As there are no vifs associated with loopback devices,
2077 * get the proper interface that does have a vif associated with it.
2078 */
2079 dev = skb->dev;
2080 if (netif_is_l3_master(skb->dev)) {
2081 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2082 if (!dev) {
2083 kfree_skb(skb);
2084 return -ENODEV;
2085 }
2086 }
2087
2088 /* Packet is looped back after forward, it should not be
2089 * forwarded second time, but still can be delivered locally.
2090 */
2091 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2092 goto dont_forward;
2093
2094 mrt = ipmr_rt_fib_lookup(net, skb);
2095 if (IS_ERR(mrt)) {
2096 kfree_skb(skb);
2097 return PTR_ERR(mrt);
2098 }
2099 if (!local) {
2100 if (IPCB(skb)->opt.router_alert) {
2101 if (ip_call_ra_chain(skb))
2102 return 0;
2103 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2104 /* IGMPv1 (and broken IGMPv2 implementations sort of
2105 * Cisco IOS <= 11.2(8)) do not put router alert
2106 * option to IGMP packets destined to routable
2107 * groups. It is very bad, because it means
2108 * that we can forward NO IGMP messages.
2109 */
2110 struct sock *mroute_sk;
2111
2112 mroute_sk = rcu_dereference(mrt->mroute_sk);
2113 if (mroute_sk) {
2114 nf_reset_ct(skb);
2115 raw_rcv(mroute_sk, skb);
2116 return 0;
2117 }
2118 }
2119 }
2120
2121 /* already under rcu_read_lock() */
2122 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2123 if (!cache) {
2124 int vif = ipmr_find_vif(mrt, dev);
2125
2126 if (vif >= 0)
2127 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2128 vif);
2129 }
2130
2131 /* No usable cache entry */
2132 if (!cache) {
2133 int vif;
2134
2135 if (local) {
2136 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2137 ip_local_deliver(skb);
2138 if (!skb2)
2139 return -ENOBUFS;
2140 skb = skb2;
2141 }
2142
2143 read_lock(&mrt_lock);
2144 vif = ipmr_find_vif(mrt, dev);
2145 if (vif >= 0) {
2146 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2147 read_unlock(&mrt_lock);
2148
2149 return err2;
2150 }
2151 read_unlock(&mrt_lock);
2152 kfree_skb(skb);
2153 return -ENODEV;
2154 }
2155
2156 read_lock(&mrt_lock);
2157 ip_mr_forward(net, mrt, dev, skb, cache, local);
2158 read_unlock(&mrt_lock);
2159
2160 if (local)
2161 return ip_local_deliver(skb);
2162
2163 return 0;
2164
2165dont_forward:
2166 if (local)
2167 return ip_local_deliver(skb);
2168 kfree_skb(skb);
2169 return 0;
2170}
2171
2172#ifdef CONFIG_IP_PIMSM_V1
2173/* Handle IGMP messages of PIMv1 */
2174int pim_rcv_v1(struct sk_buff *skb)
2175{
2176 struct igmphdr *pim;
2177 struct net *net = dev_net(skb->dev);
2178 struct mr_table *mrt;
2179
2180 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2181 goto drop;
2182
2183 pim = igmp_hdr(skb);
2184
2185 mrt = ipmr_rt_fib_lookup(net, skb);
2186 if (IS_ERR(mrt))
2187 goto drop;
2188 if (!mrt->mroute_do_pim ||
2189 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2190 goto drop;
2191
2192 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2193drop:
2194 kfree_skb(skb);
2195 }
2196 return 0;
2197}
2198#endif
2199
2200#ifdef CONFIG_IP_PIMSM_V2
2201static int pim_rcv(struct sk_buff *skb)
2202{
2203 struct pimreghdr *pim;
2204 struct net *net = dev_net(skb->dev);
2205 struct mr_table *mrt;
2206
2207 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2208 goto drop;
2209
2210 pim = (struct pimreghdr *)skb_transport_header(skb);
2211 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2212 (pim->flags & PIM_NULL_REGISTER) ||
2213 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2214 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2215 goto drop;
2216
2217 mrt = ipmr_rt_fib_lookup(net, skb);
2218 if (IS_ERR(mrt))
2219 goto drop;
2220 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2221drop:
2222 kfree_skb(skb);
2223 }
2224 return 0;
2225}
2226#endif
2227
2228int ipmr_get_route(struct net *net, struct sk_buff *skb,
2229 __be32 saddr, __be32 daddr,
2230 struct rtmsg *rtm, u32 portid)
2231{
2232 struct mfc_cache *cache;
2233 struct mr_table *mrt;
2234 int err;
2235
2236 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2237 if (!mrt)
2238 return -ENOENT;
2239
2240 rcu_read_lock();
2241 cache = ipmr_cache_find(mrt, saddr, daddr);
2242 if (!cache && skb->dev) {
2243 int vif = ipmr_find_vif(mrt, skb->dev);
2244
2245 if (vif >= 0)
2246 cache = ipmr_cache_find_any(mrt, daddr, vif);
2247 }
2248 if (!cache) {
2249 struct sk_buff *skb2;
2250 struct iphdr *iph;
2251 struct net_device *dev;
2252 int vif = -1;
2253
2254 dev = skb->dev;
2255 read_lock(&mrt_lock);
2256 if (dev)
2257 vif = ipmr_find_vif(mrt, dev);
2258 if (vif < 0) {
2259 read_unlock(&mrt_lock);
2260 rcu_read_unlock();
2261 return -ENODEV;
2262 }
2263
2264 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2265 if (!skb2) {
2266 read_unlock(&mrt_lock);
2267 rcu_read_unlock();
2268 return -ENOMEM;
2269 }
2270
2271 NETLINK_CB(skb2).portid = portid;
2272 skb_push(skb2, sizeof(struct iphdr));
2273 skb_reset_network_header(skb2);
2274 iph = ip_hdr(skb2);
2275 iph->ihl = sizeof(struct iphdr) >> 2;
2276 iph->saddr = saddr;
2277 iph->daddr = daddr;
2278 iph->version = 0;
2279 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2280 read_unlock(&mrt_lock);
2281 rcu_read_unlock();
2282 return err;
2283 }
2284
2285 read_lock(&mrt_lock);
2286 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2287 read_unlock(&mrt_lock);
2288 rcu_read_unlock();
2289 return err;
2290}
2291
2292static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2293 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2294 int flags)
2295{
2296 struct nlmsghdr *nlh;
2297 struct rtmsg *rtm;
2298 int err;
2299
2300 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2301 if (!nlh)
2302 return -EMSGSIZE;
2303
2304 rtm = nlmsg_data(nlh);
2305 rtm->rtm_family = RTNL_FAMILY_IPMR;
2306 rtm->rtm_dst_len = 32;
2307 rtm->rtm_src_len = 32;
2308 rtm->rtm_tos = 0;
2309 rtm->rtm_table = mrt->id;
2310 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2311 goto nla_put_failure;
2312 rtm->rtm_type = RTN_MULTICAST;
2313 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2314 if (c->_c.mfc_flags & MFC_STATIC)
2315 rtm->rtm_protocol = RTPROT_STATIC;
2316 else
2317 rtm->rtm_protocol = RTPROT_MROUTED;
2318 rtm->rtm_flags = 0;
2319
2320 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2321 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2322 goto nla_put_failure;
2323 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2324 /* do not break the dump if cache is unresolved */
2325 if (err < 0 && err != -ENOENT)
2326 goto nla_put_failure;
2327
2328 nlmsg_end(skb, nlh);
2329 return 0;
2330
2331nla_put_failure:
2332 nlmsg_cancel(skb, nlh);
2333 return -EMSGSIZE;
2334}
2335
2336static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2337 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2338 int flags)
2339{
2340 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2341 cmd, flags);
2342}
2343
2344static size_t mroute_msgsize(bool unresolved, int maxvif)
2345{
2346 size_t len =
2347 NLMSG_ALIGN(sizeof(struct rtmsg))
2348 + nla_total_size(4) /* RTA_TABLE */
2349 + nla_total_size(4) /* RTA_SRC */
2350 + nla_total_size(4) /* RTA_DST */
2351 ;
2352
2353 if (!unresolved)
2354 len = len
2355 + nla_total_size(4) /* RTA_IIF */
2356 + nla_total_size(0) /* RTA_MULTIPATH */
2357 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2358 /* RTA_MFC_STATS */
2359 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2360 ;
2361
2362 return len;
2363}
2364
2365static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2366 int cmd)
2367{
2368 struct net *net = read_pnet(&mrt->net);
2369 struct sk_buff *skb;
2370 int err = -ENOBUFS;
2371
2372 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2373 mrt->maxvif),
2374 GFP_ATOMIC);
2375 if (!skb)
2376 goto errout;
2377
2378 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2379 if (err < 0)
2380 goto errout;
2381
2382 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2383 return;
2384
2385errout:
2386 kfree_skb(skb);
2387 if (err < 0)
2388 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2389}
2390
2391static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2392{
2393 size_t len =
2394 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2395 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2396 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2397 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2398 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2399 /* IPMRA_CREPORT_PKT */
2400 + nla_total_size(payloadlen)
2401 ;
2402
2403 return len;
2404}
2405
2406static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2407{
2408 struct net *net = read_pnet(&mrt->net);
2409 struct nlmsghdr *nlh;
2410 struct rtgenmsg *rtgenm;
2411 struct igmpmsg *msg;
2412 struct sk_buff *skb;
2413 struct nlattr *nla;
2414 int payloadlen;
2415
2416 payloadlen = pkt->len - sizeof(struct igmpmsg);
2417 msg = (struct igmpmsg *)skb_network_header(pkt);
2418
2419 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2420 if (!skb)
2421 goto errout;
2422
2423 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2424 sizeof(struct rtgenmsg), 0);
2425 if (!nlh)
2426 goto errout;
2427 rtgenm = nlmsg_data(nlh);
2428 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2429 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2430 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2431 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2432 msg->im_src.s_addr) ||
2433 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2434 msg->im_dst.s_addr))
2435 goto nla_put_failure;
2436
2437 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2438 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2439 nla_data(nla), payloadlen))
2440 goto nla_put_failure;
2441
2442 nlmsg_end(skb, nlh);
2443
2444 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2445 return;
2446
2447nla_put_failure:
2448 nlmsg_cancel(skb, nlh);
2449errout:
2450 kfree_skb(skb);
2451 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2452}
2453
2454static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2455 const struct nlmsghdr *nlh,
2456 struct nlattr **tb,
2457 struct netlink_ext_ack *extack)
2458{
2459 struct rtmsg *rtm;
2460 int i, err;
2461
2462 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2463 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2464 return -EINVAL;
2465 }
2466
2467 if (!netlink_strict_get_check(skb))
2468 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2469 rtm_ipv4_policy, extack);
2470
2471 rtm = nlmsg_data(nlh);
2472 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2473 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2474 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2475 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2476 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2477 return -EINVAL;
2478 }
2479
2480 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2481 rtm_ipv4_policy, extack);
2482 if (err)
2483 return err;
2484
2485 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2486 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2487 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2488 return -EINVAL;
2489 }
2490
2491 for (i = 0; i <= RTA_MAX; i++) {
2492 if (!tb[i])
2493 continue;
2494
2495 switch (i) {
2496 case RTA_SRC:
2497 case RTA_DST:
2498 case RTA_TABLE:
2499 break;
2500 default:
2501 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2502 return -EINVAL;
2503 }
2504 }
2505
2506 return 0;
2507}
2508
2509static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2510 struct netlink_ext_ack *extack)
2511{
2512 struct net *net = sock_net(in_skb->sk);
2513 struct nlattr *tb[RTA_MAX + 1];
2514 struct sk_buff *skb = NULL;
2515 struct mfc_cache *cache;
2516 struct mr_table *mrt;
2517 __be32 src, grp;
2518 u32 tableid;
2519 int err;
2520
2521 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2522 if (err < 0)
2523 goto errout;
2524
2525 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2526 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2527 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2528
2529 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2530 if (!mrt) {
2531 err = -ENOENT;
2532 goto errout_free;
2533 }
2534
2535 /* entries are added/deleted only under RTNL */
2536 rcu_read_lock();
2537 cache = ipmr_cache_find(mrt, src, grp);
2538 rcu_read_unlock();
2539 if (!cache) {
2540 err = -ENOENT;
2541 goto errout_free;
2542 }
2543
2544 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2545 if (!skb) {
2546 err = -ENOBUFS;
2547 goto errout_free;
2548 }
2549
2550 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2551 nlh->nlmsg_seq, cache,
2552 RTM_NEWROUTE, 0);
2553 if (err < 0)
2554 goto errout_free;
2555
2556 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2557
2558errout:
2559 return err;
2560
2561errout_free:
2562 kfree_skb(skb);
2563 goto errout;
2564}
2565
2566static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2567{
2568 struct fib_dump_filter filter = {};
2569 int err;
2570
2571 if (cb->strict_check) {
2572 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2573 &filter, cb);
2574 if (err < 0)
2575 return err;
2576 }
2577
2578 if (filter.table_id) {
2579 struct mr_table *mrt;
2580
2581 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2582 if (!mrt) {
2583 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2584 return skb->len;
2585
2586 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2587 return -ENOENT;
2588 }
2589 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2590 &mfc_unres_lock, &filter);
2591 return skb->len ? : err;
2592 }
2593
2594 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2595 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2596}
2597
2598static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2599 [RTA_SRC] = { .type = NLA_U32 },
2600 [RTA_DST] = { .type = NLA_U32 },
2601 [RTA_IIF] = { .type = NLA_U32 },
2602 [RTA_TABLE] = { .type = NLA_U32 },
2603 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2604};
2605
2606static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2607{
2608 switch (rtm_protocol) {
2609 case RTPROT_STATIC:
2610 case RTPROT_MROUTED:
2611 return true;
2612 }
2613 return false;
2614}
2615
2616static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2617{
2618 struct rtnexthop *rtnh = nla_data(nla);
2619 int remaining = nla_len(nla), vifi = 0;
2620
2621 while (rtnh_ok(rtnh, remaining)) {
2622 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2623 if (++vifi == MAXVIFS)
2624 break;
2625 rtnh = rtnh_next(rtnh, &remaining);
2626 }
2627
2628 return remaining > 0 ? -EINVAL : vifi;
2629}
2630
2631/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2632static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2633 struct mfcctl *mfcc, int *mrtsock,
2634 struct mr_table **mrtret,
2635 struct netlink_ext_ack *extack)
2636{
2637 struct net_device *dev = NULL;
2638 u32 tblid = RT_TABLE_DEFAULT;
2639 struct mr_table *mrt;
2640 struct nlattr *attr;
2641 struct rtmsg *rtm;
2642 int ret, rem;
2643
2644 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2645 rtm_ipmr_policy, extack);
2646 if (ret < 0)
2647 goto out;
2648 rtm = nlmsg_data(nlh);
2649
2650 ret = -EINVAL;
2651 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2652 rtm->rtm_type != RTN_MULTICAST ||
2653 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2654 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2655 goto out;
2656
2657 memset(mfcc, 0, sizeof(*mfcc));
2658 mfcc->mfcc_parent = -1;
2659 ret = 0;
2660 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2661 switch (nla_type(attr)) {
2662 case RTA_SRC:
2663 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2664 break;
2665 case RTA_DST:
2666 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2667 break;
2668 case RTA_IIF:
2669 dev = __dev_get_by_index(net, nla_get_u32(attr));
2670 if (!dev) {
2671 ret = -ENODEV;
2672 goto out;
2673 }
2674 break;
2675 case RTA_MULTIPATH:
2676 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2677 ret = -EINVAL;
2678 goto out;
2679 }
2680 break;
2681 case RTA_PREFSRC:
2682 ret = 1;
2683 break;
2684 case RTA_TABLE:
2685 tblid = nla_get_u32(attr);
2686 break;
2687 }
2688 }
2689 mrt = ipmr_get_table(net, tblid);
2690 if (!mrt) {
2691 ret = -ENOENT;
2692 goto out;
2693 }
2694 *mrtret = mrt;
2695 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2696 if (dev)
2697 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2698
2699out:
2700 return ret;
2701}
2702
2703/* takes care of both newroute and delroute */
2704static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2705 struct netlink_ext_ack *extack)
2706{
2707 struct net *net = sock_net(skb->sk);
2708 int ret, mrtsock, parent;
2709 struct mr_table *tbl;
2710 struct mfcctl mfcc;
2711
2712 mrtsock = 0;
2713 tbl = NULL;
2714 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2715 if (ret < 0)
2716 return ret;
2717
2718 parent = ret ? mfcc.mfcc_parent : -1;
2719 if (nlh->nlmsg_type == RTM_NEWROUTE)
2720 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2721 else
2722 return ipmr_mfc_delete(tbl, &mfcc, parent);
2723}
2724
2725static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2726{
2727 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2728
2729 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2730 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2731 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2732 mrt->mroute_reg_vif_num) ||
2733 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2734 mrt->mroute_do_assert) ||
2735 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2736 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2737 mrt->mroute_do_wrvifwhole))
2738 return false;
2739
2740 return true;
2741}
2742
2743static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2744{
2745 struct nlattr *vif_nest;
2746 struct vif_device *vif;
2747
2748 /* if the VIF doesn't exist just continue */
2749 if (!VIF_EXISTS(mrt, vifid))
2750 return true;
2751
2752 vif = &mrt->vif_table[vifid];
2753 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2754 if (!vif_nest)
2755 return false;
2756 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2757 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2758 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2759 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2760 IPMRA_VIFA_PAD) ||
2761 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2762 IPMRA_VIFA_PAD) ||
2763 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2764 IPMRA_VIFA_PAD) ||
2765 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2766 IPMRA_VIFA_PAD) ||
2767 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2768 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2769 nla_nest_cancel(skb, vif_nest);
2770 return false;
2771 }
2772 nla_nest_end(skb, vif_nest);
2773
2774 return true;
2775}
2776
2777static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2778 struct netlink_ext_ack *extack)
2779{
2780 struct ifinfomsg *ifm;
2781
2782 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2783 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2784 return -EINVAL;
2785 }
2786
2787 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2788 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2789 return -EINVAL;
2790 }
2791
2792 ifm = nlmsg_data(nlh);
2793 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2794 ifm->ifi_change || ifm->ifi_index) {
2795 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2796 return -EINVAL;
2797 }
2798
2799 return 0;
2800}
2801
2802static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2803{
2804 struct net *net = sock_net(skb->sk);
2805 struct nlmsghdr *nlh = NULL;
2806 unsigned int t = 0, s_t;
2807 unsigned int e = 0, s_e;
2808 struct mr_table *mrt;
2809
2810 if (cb->strict_check) {
2811 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2812
2813 if (err < 0)
2814 return err;
2815 }
2816
2817 s_t = cb->args[0];
2818 s_e = cb->args[1];
2819
2820 ipmr_for_each_table(mrt, net) {
2821 struct nlattr *vifs, *af;
2822 struct ifinfomsg *hdr;
2823 u32 i;
2824
2825 if (t < s_t)
2826 goto skip_table;
2827 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2828 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2829 sizeof(*hdr), NLM_F_MULTI);
2830 if (!nlh)
2831 break;
2832
2833 hdr = nlmsg_data(nlh);
2834 memset(hdr, 0, sizeof(*hdr));
2835 hdr->ifi_family = RTNL_FAMILY_IPMR;
2836
2837 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2838 if (!af) {
2839 nlmsg_cancel(skb, nlh);
2840 goto out;
2841 }
2842
2843 if (!ipmr_fill_table(mrt, skb)) {
2844 nlmsg_cancel(skb, nlh);
2845 goto out;
2846 }
2847
2848 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2849 if (!vifs) {
2850 nla_nest_end(skb, af);
2851 nlmsg_end(skb, nlh);
2852 goto out;
2853 }
2854 for (i = 0; i < mrt->maxvif; i++) {
2855 if (e < s_e)
2856 goto skip_entry;
2857 if (!ipmr_fill_vif(mrt, i, skb)) {
2858 nla_nest_end(skb, vifs);
2859 nla_nest_end(skb, af);
2860 nlmsg_end(skb, nlh);
2861 goto out;
2862 }
2863skip_entry:
2864 e++;
2865 }
2866 s_e = 0;
2867 e = 0;
2868 nla_nest_end(skb, vifs);
2869 nla_nest_end(skb, af);
2870 nlmsg_end(skb, nlh);
2871skip_table:
2872 t++;
2873 }
2874
2875out:
2876 cb->args[1] = e;
2877 cb->args[0] = t;
2878
2879 return skb->len;
2880}
2881
2882#ifdef CONFIG_PROC_FS
2883/* The /proc interfaces to multicast routing :
2884 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2885 */
2886
2887static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2888 __acquires(mrt_lock)
2889{
2890 struct mr_vif_iter *iter = seq->private;
2891 struct net *net = seq_file_net(seq);
2892 struct mr_table *mrt;
2893
2894 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2895 if (!mrt)
2896 return ERR_PTR(-ENOENT);
2897
2898 iter->mrt = mrt;
2899
2900 read_lock(&mrt_lock);
2901 return mr_vif_seq_start(seq, pos);
2902}
2903
2904static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2905 __releases(mrt_lock)
2906{
2907 read_unlock(&mrt_lock);
2908}
2909
2910static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2911{
2912 struct mr_vif_iter *iter = seq->private;
2913 struct mr_table *mrt = iter->mrt;
2914
2915 if (v == SEQ_START_TOKEN) {
2916 seq_puts(seq,
2917 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2918 } else {
2919 const struct vif_device *vif = v;
2920 const char *name = vif->dev ?
2921 vif->dev->name : "none";
2922
2923 seq_printf(seq,
2924 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2925 vif - mrt->vif_table,
2926 name, vif->bytes_in, vif->pkt_in,
2927 vif->bytes_out, vif->pkt_out,
2928 vif->flags, vif->local, vif->remote);
2929 }
2930 return 0;
2931}
2932
2933static const struct seq_operations ipmr_vif_seq_ops = {
2934 .start = ipmr_vif_seq_start,
2935 .next = mr_vif_seq_next,
2936 .stop = ipmr_vif_seq_stop,
2937 .show = ipmr_vif_seq_show,
2938};
2939
2940static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2941{
2942 struct net *net = seq_file_net(seq);
2943 struct mr_table *mrt;
2944
2945 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2946 if (!mrt)
2947 return ERR_PTR(-ENOENT);
2948
2949 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2950}
2951
2952static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2953{
2954 int n;
2955
2956 if (v == SEQ_START_TOKEN) {
2957 seq_puts(seq,
2958 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2959 } else {
2960 const struct mfc_cache *mfc = v;
2961 const struct mr_mfc_iter *it = seq->private;
2962 const struct mr_table *mrt = it->mrt;
2963
2964 seq_printf(seq, "%08X %08X %-3hd",
2965 (__force u32) mfc->mfc_mcastgrp,
2966 (__force u32) mfc->mfc_origin,
2967 mfc->_c.mfc_parent);
2968
2969 if (it->cache != &mrt->mfc_unres_queue) {
2970 seq_printf(seq, " %8lu %8lu %8lu",
2971 mfc->_c.mfc_un.res.pkt,
2972 mfc->_c.mfc_un.res.bytes,
2973 mfc->_c.mfc_un.res.wrong_if);
2974 for (n = mfc->_c.mfc_un.res.minvif;
2975 n < mfc->_c.mfc_un.res.maxvif; n++) {
2976 if (VIF_EXISTS(mrt, n) &&
2977 mfc->_c.mfc_un.res.ttls[n] < 255)
2978 seq_printf(seq,
2979 " %2d:%-3d",
2980 n, mfc->_c.mfc_un.res.ttls[n]);
2981 }
2982 } else {
2983 /* unresolved mfc_caches don't contain
2984 * pkt, bytes and wrong_if values
2985 */
2986 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2987 }
2988 seq_putc(seq, '\n');
2989 }
2990 return 0;
2991}
2992
2993static const struct seq_operations ipmr_mfc_seq_ops = {
2994 .start = ipmr_mfc_seq_start,
2995 .next = mr_mfc_seq_next,
2996 .stop = mr_mfc_seq_stop,
2997 .show = ipmr_mfc_seq_show,
2998};
2999#endif
3000
3001#ifdef CONFIG_IP_PIMSM_V2
3002static const struct net_protocol pim_protocol = {
3003 .handler = pim_rcv,
3004 .netns_ok = 1,
3005};
3006#endif
3007
3008static unsigned int ipmr_seq_read(struct net *net)
3009{
3010 ASSERT_RTNL();
3011
3012 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3013}
3014
3015static int ipmr_dump(struct net *net, struct notifier_block *nb,
3016 struct netlink_ext_ack *extack)
3017{
3018 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3019 ipmr_mr_table_iter, &mrt_lock, extack);
3020}
3021
3022static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3023 .family = RTNL_FAMILY_IPMR,
3024 .fib_seq_read = ipmr_seq_read,
3025 .fib_dump = ipmr_dump,
3026 .owner = THIS_MODULE,
3027};
3028
3029static int __net_init ipmr_notifier_init(struct net *net)
3030{
3031 struct fib_notifier_ops *ops;
3032
3033 net->ipv4.ipmr_seq = 0;
3034
3035 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3036 if (IS_ERR(ops))
3037 return PTR_ERR(ops);
3038 net->ipv4.ipmr_notifier_ops = ops;
3039
3040 return 0;
3041}
3042
3043static void __net_exit ipmr_notifier_exit(struct net *net)
3044{
3045 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3046 net->ipv4.ipmr_notifier_ops = NULL;
3047}
3048
3049/* Setup for IP multicast routing */
3050static int __net_init ipmr_net_init(struct net *net)
3051{
3052 int err;
3053
3054 err = ipmr_notifier_init(net);
3055 if (err)
3056 goto ipmr_notifier_fail;
3057
3058 err = ipmr_rules_init(net);
3059 if (err < 0)
3060 goto ipmr_rules_fail;
3061
3062#ifdef CONFIG_PROC_FS
3063 err = -ENOMEM;
3064 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3065 sizeof(struct mr_vif_iter)))
3066 goto proc_vif_fail;
3067 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3068 sizeof(struct mr_mfc_iter)))
3069 goto proc_cache_fail;
3070#endif
3071 return 0;
3072
3073#ifdef CONFIG_PROC_FS
3074proc_cache_fail:
3075 remove_proc_entry("ip_mr_vif", net->proc_net);
3076proc_vif_fail:
3077 ipmr_rules_exit(net);
3078#endif
3079ipmr_rules_fail:
3080 ipmr_notifier_exit(net);
3081ipmr_notifier_fail:
3082 return err;
3083}
3084
3085static void __net_exit ipmr_net_exit(struct net *net)
3086{
3087#ifdef CONFIG_PROC_FS
3088 remove_proc_entry("ip_mr_cache", net->proc_net);
3089 remove_proc_entry("ip_mr_vif", net->proc_net);
3090#endif
3091 ipmr_notifier_exit(net);
3092 ipmr_rules_exit(net);
3093}
3094
3095static struct pernet_operations ipmr_net_ops = {
3096 .init = ipmr_net_init,
3097 .exit = ipmr_net_exit,
3098};
3099
3100int __init ip_mr_init(void)
3101{
3102 int err;
3103
3104 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3105 sizeof(struct mfc_cache),
3106 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3107 NULL);
3108
3109 err = register_pernet_subsys(&ipmr_net_ops);
3110 if (err)
3111 goto reg_pernet_fail;
3112
3113 err = register_netdevice_notifier(&ip_mr_notifier);
3114 if (err)
3115 goto reg_notif_fail;
3116#ifdef CONFIG_IP_PIMSM_V2
3117 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3118 pr_err("%s: can't add PIM protocol\n", __func__);
3119 err = -EAGAIN;
3120 goto add_proto_fail;
3121 }
3122#endif
3123 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3124 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3125 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3126 ipmr_rtm_route, NULL, 0);
3127 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3128 ipmr_rtm_route, NULL, 0);
3129
3130 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3131 NULL, ipmr_rtm_dumplink, 0);
3132 return 0;
3133
3134#ifdef CONFIG_IP_PIMSM_V2
3135add_proto_fail:
3136 unregister_netdevice_notifier(&ip_mr_notifier);
3137#endif
3138reg_notif_fail:
3139 unregister_pernet_subsys(&ipmr_net_ops);
3140reg_pernet_fail:
3141 kmem_cache_destroy(mrt_cachep);
3142 return err;
3143}
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * IP multicast routing support for mrouted 3.6/3.8
4 *
5 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6 * Linux Consultancy and Custom Driver Development
7 *
8 * Fixes:
9 * Michael Chastain : Incorrect size of copying.
10 * Alan Cox : Added the cache manager code
11 * Alan Cox : Fixed the clone/copy bug and device race.
12 * Mike McLagan : Routing by source
13 * Malcolm Beattie : Buffer handling fixes.
14 * Alexey Kuznetsov : Double buffer free and other fixes.
15 * SVR Anand : Fixed several multicast bugs and problems.
16 * Alexey Kuznetsov : Status, optimisations and more.
17 * Brad Parker : Better behaviour on mrouted upcall
18 * overflow.
19 * Carlos Picoto : PIMv1 Support
20 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
21 * Relax this requirement to work with older peers.
22 */
23
24#include <linux/uaccess.h>
25#include <linux/types.h>
26#include <linux/cache.h>
27#include <linux/capability.h>
28#include <linux/errno.h>
29#include <linux/mm.h>
30#include <linux/kernel.h>
31#include <linux/fcntl.h>
32#include <linux/stat.h>
33#include <linux/socket.h>
34#include <linux/in.h>
35#include <linux/inet.h>
36#include <linux/netdevice.h>
37#include <linux/inetdevice.h>
38#include <linux/igmp.h>
39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
41#include <linux/mroute.h>
42#include <linux/init.h>
43#include <linux/if_ether.h>
44#include <linux/slab.h>
45#include <net/net_namespace.h>
46#include <net/ip.h>
47#include <net/protocol.h>
48#include <linux/skbuff.h>
49#include <net/route.h>
50#include <net/icmp.h>
51#include <net/udp.h>
52#include <net/raw.h>
53#include <linux/notifier.h>
54#include <linux/if_arp.h>
55#include <linux/netfilter_ipv4.h>
56#include <linux/compat.h>
57#include <linux/export.h>
58#include <linux/rhashtable.h>
59#include <net/ip_tunnels.h>
60#include <net/checksum.h>
61#include <net/netlink.h>
62#include <net/fib_rules.h>
63#include <linux/netconf.h>
64#include <net/rtnh.h>
65#include <net/inet_dscp.h>
66
67#include <linux/nospec.h>
68
69struct ipmr_rule {
70 struct fib_rule common;
71};
72
73struct ipmr_result {
74 struct mr_table *mrt;
75};
76
77/* Big lock, protecting vif table, mrt cache and mroute socket state.
78 * Note that the changes are semaphored via rtnl_lock.
79 */
80
81static DEFINE_SPINLOCK(mrt_lock);
82
83static struct net_device *vif_dev_read(const struct vif_device *vif)
84{
85 return rcu_dereference(vif->dev);
86}
87
88/* Multicast router control variables */
89
90/* Special spinlock for queue of unresolved entries */
91static DEFINE_SPINLOCK(mfc_unres_lock);
92
93/* We return to original Alan's scheme. Hash table of resolved
94 * entries is changed only in process context and protected
95 * with weak lock mrt_lock. Queue of unresolved entries is protected
96 * with strong spinlock mfc_unres_lock.
97 *
98 * In this case data path is free of exclusive locks at all.
99 */
100
101static struct kmem_cache *mrt_cachep __ro_after_init;
102
103static struct mr_table *ipmr_new_table(struct net *net, u32 id);
104static void ipmr_free_table(struct mr_table *mrt);
105
106static void ip_mr_forward(struct net *net, struct mr_table *mrt,
107 struct net_device *dev, struct sk_buff *skb,
108 struct mfc_cache *cache, int local);
109static int ipmr_cache_report(const struct mr_table *mrt,
110 struct sk_buff *pkt, vifi_t vifi, int assert);
111static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
112 int cmd);
113static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
114static void mroute_clean_tables(struct mr_table *mrt, int flags);
115static void ipmr_expire_process(struct timer_list *t);
116
117#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
118#define ipmr_for_each_table(mrt, net) \
119 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
120 lockdep_rtnl_is_held() || \
121 list_empty(&net->ipv4.mr_tables))
122
123static bool ipmr_can_free_table(struct net *net)
124{
125 return !check_net(net) || !net_initialized(net);
126}
127
128static struct mr_table *ipmr_mr_table_iter(struct net *net,
129 struct mr_table *mrt)
130{
131 struct mr_table *ret;
132
133 if (!mrt)
134 ret = list_entry_rcu(net->ipv4.mr_tables.next,
135 struct mr_table, list);
136 else
137 ret = list_entry_rcu(mrt->list.next,
138 struct mr_table, list);
139
140 if (&ret->list == &net->ipv4.mr_tables)
141 return NULL;
142 return ret;
143}
144
145static struct mr_table *__ipmr_get_table(struct net *net, u32 id)
146{
147 struct mr_table *mrt;
148
149 ipmr_for_each_table(mrt, net) {
150 if (mrt->id == id)
151 return mrt;
152 }
153 return NULL;
154}
155
156static struct mr_table *ipmr_get_table(struct net *net, u32 id)
157{
158 struct mr_table *mrt;
159
160 rcu_read_lock();
161 mrt = __ipmr_get_table(net, id);
162 rcu_read_unlock();
163 return mrt;
164}
165
166static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
167 struct mr_table **mrt)
168{
169 int err;
170 struct ipmr_result res;
171 struct fib_lookup_arg arg = {
172 .result = &res,
173 .flags = FIB_LOOKUP_NOREF,
174 };
175
176 /* update flow if oif or iif point to device enslaved to l3mdev */
177 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
178
179 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
180 flowi4_to_flowi(flp4), 0, &arg);
181 if (err < 0)
182 return err;
183 *mrt = res.mrt;
184 return 0;
185}
186
187static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
188 int flags, struct fib_lookup_arg *arg)
189{
190 struct ipmr_result *res = arg->result;
191 struct mr_table *mrt;
192
193 switch (rule->action) {
194 case FR_ACT_TO_TBL:
195 break;
196 case FR_ACT_UNREACHABLE:
197 return -ENETUNREACH;
198 case FR_ACT_PROHIBIT:
199 return -EACCES;
200 case FR_ACT_BLACKHOLE:
201 default:
202 return -EINVAL;
203 }
204
205 arg->table = fib_rule_get_table(rule, arg);
206
207 mrt = __ipmr_get_table(rule->fr_net, arg->table);
208 if (!mrt)
209 return -EAGAIN;
210 res->mrt = mrt;
211 return 0;
212}
213
214static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
215{
216 return 1;
217}
218
219static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
220 struct fib_rule_hdr *frh, struct nlattr **tb,
221 struct netlink_ext_ack *extack)
222{
223 return 0;
224}
225
226static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
227 struct nlattr **tb)
228{
229 return 1;
230}
231
232static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
233 struct fib_rule_hdr *frh)
234{
235 frh->dst_len = 0;
236 frh->src_len = 0;
237 frh->tos = 0;
238 return 0;
239}
240
241static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
242 .family = RTNL_FAMILY_IPMR,
243 .rule_size = sizeof(struct ipmr_rule),
244 .addr_size = sizeof(u32),
245 .action = ipmr_rule_action,
246 .match = ipmr_rule_match,
247 .configure = ipmr_rule_configure,
248 .compare = ipmr_rule_compare,
249 .fill = ipmr_rule_fill,
250 .nlgroup = RTNLGRP_IPV4_RULE,
251 .owner = THIS_MODULE,
252};
253
254static int __net_init ipmr_rules_init(struct net *net)
255{
256 struct fib_rules_ops *ops;
257 struct mr_table *mrt;
258 int err;
259
260 ops = fib_rules_register(&ipmr_rules_ops_template, net);
261 if (IS_ERR(ops))
262 return PTR_ERR(ops);
263
264 INIT_LIST_HEAD(&net->ipv4.mr_tables);
265
266 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
267 if (IS_ERR(mrt)) {
268 err = PTR_ERR(mrt);
269 goto err1;
270 }
271
272 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT);
273 if (err < 0)
274 goto err2;
275
276 net->ipv4.mr_rules_ops = ops;
277 return 0;
278
279err2:
280 rtnl_lock();
281 ipmr_free_table(mrt);
282 rtnl_unlock();
283err1:
284 fib_rules_unregister(ops);
285 return err;
286}
287
288static void __net_exit ipmr_rules_exit(struct net *net)
289{
290 struct mr_table *mrt, *next;
291
292 ASSERT_RTNL();
293 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
294 list_del(&mrt->list);
295 ipmr_free_table(mrt);
296 }
297 fib_rules_unregister(net->ipv4.mr_rules_ops);
298}
299
300static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
301 struct netlink_ext_ack *extack)
302{
303 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
304}
305
306static unsigned int ipmr_rules_seq_read(const struct net *net)
307{
308 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
309}
310
311bool ipmr_rule_default(const struct fib_rule *rule)
312{
313 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
314}
315EXPORT_SYMBOL(ipmr_rule_default);
316#else
317#define ipmr_for_each_table(mrt, net) \
318 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
319
320static bool ipmr_can_free_table(struct net *net)
321{
322 return !check_net(net);
323}
324
325static struct mr_table *ipmr_mr_table_iter(struct net *net,
326 struct mr_table *mrt)
327{
328 if (!mrt)
329 return net->ipv4.mrt;
330 return NULL;
331}
332
333static struct mr_table *ipmr_get_table(struct net *net, u32 id)
334{
335 return net->ipv4.mrt;
336}
337
338#define __ipmr_get_table ipmr_get_table
339
340static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
341 struct mr_table **mrt)
342{
343 *mrt = net->ipv4.mrt;
344 return 0;
345}
346
347static int __net_init ipmr_rules_init(struct net *net)
348{
349 struct mr_table *mrt;
350
351 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
352 if (IS_ERR(mrt))
353 return PTR_ERR(mrt);
354 net->ipv4.mrt = mrt;
355 return 0;
356}
357
358static void __net_exit ipmr_rules_exit(struct net *net)
359{
360 ASSERT_RTNL();
361 ipmr_free_table(net->ipv4.mrt);
362 net->ipv4.mrt = NULL;
363}
364
365static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
366 struct netlink_ext_ack *extack)
367{
368 return 0;
369}
370
371static unsigned int ipmr_rules_seq_read(const struct net *net)
372{
373 return 0;
374}
375
376bool ipmr_rule_default(const struct fib_rule *rule)
377{
378 return true;
379}
380EXPORT_SYMBOL(ipmr_rule_default);
381#endif
382
383static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
384 const void *ptr)
385{
386 const struct mfc_cache_cmp_arg *cmparg = arg->key;
387 const struct mfc_cache *c = ptr;
388
389 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
390 cmparg->mfc_origin != c->mfc_origin;
391}
392
393static const struct rhashtable_params ipmr_rht_params = {
394 .head_offset = offsetof(struct mr_mfc, mnode),
395 .key_offset = offsetof(struct mfc_cache, cmparg),
396 .key_len = sizeof(struct mfc_cache_cmp_arg),
397 .nelem_hint = 3,
398 .obj_cmpfn = ipmr_hash_cmp,
399 .automatic_shrinking = true,
400};
401
402static void ipmr_new_table_set(struct mr_table *mrt,
403 struct net *net)
404{
405#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
406 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
407#endif
408}
409
410static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
411 .mfc_mcastgrp = htonl(INADDR_ANY),
412 .mfc_origin = htonl(INADDR_ANY),
413};
414
415static struct mr_table_ops ipmr_mr_table_ops = {
416 .rht_params = &ipmr_rht_params,
417 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
418};
419
420static struct mr_table *ipmr_new_table(struct net *net, u32 id)
421{
422 struct mr_table *mrt;
423
424 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
425 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
426 return ERR_PTR(-EINVAL);
427
428 mrt = __ipmr_get_table(net, id);
429 if (mrt)
430 return mrt;
431
432 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
433 ipmr_expire_process, ipmr_new_table_set);
434}
435
436static void ipmr_free_table(struct mr_table *mrt)
437{
438 struct net *net = read_pnet(&mrt->net);
439
440 WARN_ON_ONCE(!ipmr_can_free_table(net));
441
442 timer_shutdown_sync(&mrt->ipmr_expire_timer);
443 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
444 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
445 rhltable_destroy(&mrt->mfc_hash);
446 kfree(mrt);
447}
448
449/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
450
451/* Initialize ipmr pimreg/tunnel in_device */
452static bool ipmr_init_vif_indev(const struct net_device *dev)
453{
454 struct in_device *in_dev;
455
456 ASSERT_RTNL();
457
458 in_dev = __in_dev_get_rtnl(dev);
459 if (!in_dev)
460 return false;
461 ipv4_devconf_setall(in_dev);
462 neigh_parms_data_state_setall(in_dev->arp_parms);
463 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
464
465 return true;
466}
467
468static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
469{
470 struct net_device *tunnel_dev, *new_dev;
471 struct ip_tunnel_parm_kern p = { };
472 int err;
473
474 tunnel_dev = __dev_get_by_name(net, "tunl0");
475 if (!tunnel_dev)
476 goto out;
477
478 p.iph.daddr = v->vifc_rmt_addr.s_addr;
479 p.iph.saddr = v->vifc_lcl_addr.s_addr;
480 p.iph.version = 4;
481 p.iph.ihl = 5;
482 p.iph.protocol = IPPROTO_IPIP;
483 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
484
485 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
486 goto out;
487 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
488 SIOCADDTUNNEL);
489 if (err)
490 goto out;
491
492 new_dev = __dev_get_by_name(net, p.name);
493 if (!new_dev)
494 goto out;
495
496 new_dev->flags |= IFF_MULTICAST;
497 if (!ipmr_init_vif_indev(new_dev))
498 goto out_unregister;
499 if (dev_open(new_dev, NULL))
500 goto out_unregister;
501 dev_hold(new_dev);
502 err = dev_set_allmulti(new_dev, 1);
503 if (err) {
504 dev_close(new_dev);
505 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
506 SIOCDELTUNNEL);
507 dev_put(new_dev);
508 new_dev = ERR_PTR(err);
509 }
510 return new_dev;
511
512out_unregister:
513 unregister_netdevice(new_dev);
514out:
515 return ERR_PTR(-ENOBUFS);
516}
517
518#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
519static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
520{
521 struct net *net = dev_net(dev);
522 struct mr_table *mrt;
523 struct flowi4 fl4 = {
524 .flowi4_oif = dev->ifindex,
525 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
526 .flowi4_mark = skb->mark,
527 };
528 int err;
529
530 err = ipmr_fib_lookup(net, &fl4, &mrt);
531 if (err < 0) {
532 kfree_skb(skb);
533 return err;
534 }
535
536 DEV_STATS_ADD(dev, tx_bytes, skb->len);
537 DEV_STATS_INC(dev, tx_packets);
538 rcu_read_lock();
539
540 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
541 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
542 IGMPMSG_WHOLEPKT);
543
544 rcu_read_unlock();
545 kfree_skb(skb);
546 return NETDEV_TX_OK;
547}
548
549static int reg_vif_get_iflink(const struct net_device *dev)
550{
551 return 0;
552}
553
554static const struct net_device_ops reg_vif_netdev_ops = {
555 .ndo_start_xmit = reg_vif_xmit,
556 .ndo_get_iflink = reg_vif_get_iflink,
557};
558
559static void reg_vif_setup(struct net_device *dev)
560{
561 dev->type = ARPHRD_PIMREG;
562 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
563 dev->flags = IFF_NOARP;
564 dev->netdev_ops = ®_vif_netdev_ops;
565 dev->needs_free_netdev = true;
566 dev->netns_local = true;
567}
568
569static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
570{
571 struct net_device *dev;
572 char name[IFNAMSIZ];
573
574 if (mrt->id == RT_TABLE_DEFAULT)
575 sprintf(name, "pimreg");
576 else
577 sprintf(name, "pimreg%u", mrt->id);
578
579 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
580
581 if (!dev)
582 return NULL;
583
584 dev_net_set(dev, net);
585
586 if (register_netdevice(dev)) {
587 free_netdev(dev);
588 return NULL;
589 }
590
591 if (!ipmr_init_vif_indev(dev))
592 goto failure;
593 if (dev_open(dev, NULL))
594 goto failure;
595
596 dev_hold(dev);
597
598 return dev;
599
600failure:
601 unregister_netdevice(dev);
602 return NULL;
603}
604
605/* called with rcu_read_lock() */
606static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
607 unsigned int pimlen)
608{
609 struct net_device *reg_dev = NULL;
610 struct iphdr *encap;
611 int vif_num;
612
613 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
614 /* Check that:
615 * a. packet is really sent to a multicast group
616 * b. packet is not a NULL-REGISTER
617 * c. packet is not truncated
618 */
619 if (!ipv4_is_multicast(encap->daddr) ||
620 encap->tot_len == 0 ||
621 ntohs(encap->tot_len) + pimlen > skb->len)
622 return 1;
623
624 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
625 vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
626 if (vif_num >= 0)
627 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
628 if (!reg_dev)
629 return 1;
630
631 skb->mac_header = skb->network_header;
632 skb_pull(skb, (u8 *)encap - skb->data);
633 skb_reset_network_header(skb);
634 skb->protocol = htons(ETH_P_IP);
635 skb->ip_summed = CHECKSUM_NONE;
636
637 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
638
639 netif_rx(skb);
640
641 return NET_RX_SUCCESS;
642}
643#else
644static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
645{
646 return NULL;
647}
648#endif
649
650static int call_ipmr_vif_entry_notifiers(struct net *net,
651 enum fib_event_type event_type,
652 struct vif_device *vif,
653 struct net_device *vif_dev,
654 vifi_t vif_index, u32 tb_id)
655{
656 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
657 vif, vif_dev, vif_index, tb_id,
658 &net->ipv4.ipmr_seq);
659}
660
661static int call_ipmr_mfc_entry_notifiers(struct net *net,
662 enum fib_event_type event_type,
663 struct mfc_cache *mfc, u32 tb_id)
664{
665 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
666 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
667}
668
669/**
670 * vif_delete - Delete a VIF entry
671 * @mrt: Table to delete from
672 * @vifi: VIF identifier to delete
673 * @notify: Set to 1, if the caller is a notifier_call
674 * @head: if unregistering the VIF, place it on this queue
675 */
676static int vif_delete(struct mr_table *mrt, int vifi, int notify,
677 struct list_head *head)
678{
679 struct net *net = read_pnet(&mrt->net);
680 struct vif_device *v;
681 struct net_device *dev;
682 struct in_device *in_dev;
683
684 if (vifi < 0 || vifi >= mrt->maxvif)
685 return -EADDRNOTAVAIL;
686
687 v = &mrt->vif_table[vifi];
688
689 dev = rtnl_dereference(v->dev);
690 if (!dev)
691 return -EADDRNOTAVAIL;
692
693 spin_lock(&mrt_lock);
694 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
695 vifi, mrt->id);
696 RCU_INIT_POINTER(v->dev, NULL);
697
698 if (vifi == mrt->mroute_reg_vif_num) {
699 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
700 WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
701 }
702 if (vifi + 1 == mrt->maxvif) {
703 int tmp;
704
705 for (tmp = vifi - 1; tmp >= 0; tmp--) {
706 if (VIF_EXISTS(mrt, tmp))
707 break;
708 }
709 WRITE_ONCE(mrt->maxvif, tmp + 1);
710 }
711
712 spin_unlock(&mrt_lock);
713
714 dev_set_allmulti(dev, -1);
715
716 in_dev = __in_dev_get_rtnl(dev);
717 if (in_dev) {
718 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
719 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
720 NETCONFA_MC_FORWARDING,
721 dev->ifindex, &in_dev->cnf);
722 ip_rt_multicast_event(in_dev);
723 }
724
725 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
726 unregister_netdevice_queue(dev, head);
727
728 netdev_put(dev, &v->dev_tracker);
729 return 0;
730}
731
732static void ipmr_cache_free_rcu(struct rcu_head *head)
733{
734 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
735
736 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
737}
738
739static void ipmr_cache_free(struct mfc_cache *c)
740{
741 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
742}
743
744/* Destroy an unresolved cache entry, killing queued skbs
745 * and reporting error to netlink readers.
746 */
747static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
748{
749 struct net *net = read_pnet(&mrt->net);
750 struct sk_buff *skb;
751 struct nlmsgerr *e;
752
753 atomic_dec(&mrt->cache_resolve_queue_len);
754
755 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
756 if (ip_hdr(skb)->version == 0) {
757 struct nlmsghdr *nlh = skb_pull(skb,
758 sizeof(struct iphdr));
759 nlh->nlmsg_type = NLMSG_ERROR;
760 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
761 skb_trim(skb, nlh->nlmsg_len);
762 e = nlmsg_data(nlh);
763 e->error = -ETIMEDOUT;
764 memset(&e->msg, 0, sizeof(e->msg));
765
766 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
767 } else {
768 kfree_skb(skb);
769 }
770 }
771
772 ipmr_cache_free(c);
773}
774
775/* Timer process for the unresolved queue. */
776static void ipmr_expire_process(struct timer_list *t)
777{
778 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
779 struct mr_mfc *c, *next;
780 unsigned long expires;
781 unsigned long now;
782
783 if (!spin_trylock(&mfc_unres_lock)) {
784 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
785 return;
786 }
787
788 if (list_empty(&mrt->mfc_unres_queue))
789 goto out;
790
791 now = jiffies;
792 expires = 10*HZ;
793
794 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
795 if (time_after(c->mfc_un.unres.expires, now)) {
796 unsigned long interval = c->mfc_un.unres.expires - now;
797 if (interval < expires)
798 expires = interval;
799 continue;
800 }
801
802 list_del(&c->list);
803 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
804 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
805 }
806
807 if (!list_empty(&mrt->mfc_unres_queue))
808 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
809
810out:
811 spin_unlock(&mfc_unres_lock);
812}
813
814/* Fill oifs list. It is called under locked mrt_lock. */
815static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
816 unsigned char *ttls)
817{
818 int vifi;
819
820 cache->mfc_un.res.minvif = MAXVIFS;
821 cache->mfc_un.res.maxvif = 0;
822 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
823
824 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
825 if (VIF_EXISTS(mrt, vifi) &&
826 ttls[vifi] && ttls[vifi] < 255) {
827 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
828 if (cache->mfc_un.res.minvif > vifi)
829 cache->mfc_un.res.minvif = vifi;
830 if (cache->mfc_un.res.maxvif <= vifi)
831 cache->mfc_un.res.maxvif = vifi + 1;
832 }
833 }
834 WRITE_ONCE(cache->mfc_un.res.lastuse, jiffies);
835}
836
837static int vif_add(struct net *net, struct mr_table *mrt,
838 struct vifctl *vifc, int mrtsock)
839{
840 struct netdev_phys_item_id ppid = { };
841 int vifi = vifc->vifc_vifi;
842 struct vif_device *v = &mrt->vif_table[vifi];
843 struct net_device *dev;
844 struct in_device *in_dev;
845 int err;
846
847 /* Is vif busy ? */
848 if (VIF_EXISTS(mrt, vifi))
849 return -EADDRINUSE;
850
851 switch (vifc->vifc_flags) {
852 case VIFF_REGISTER:
853 if (!ipmr_pimsm_enabled())
854 return -EINVAL;
855 /* Special Purpose VIF in PIM
856 * All the packets will be sent to the daemon
857 */
858 if (mrt->mroute_reg_vif_num >= 0)
859 return -EADDRINUSE;
860 dev = ipmr_reg_vif(net, mrt);
861 if (!dev)
862 return -ENOBUFS;
863 err = dev_set_allmulti(dev, 1);
864 if (err) {
865 unregister_netdevice(dev);
866 dev_put(dev);
867 return err;
868 }
869 break;
870 case VIFF_TUNNEL:
871 dev = ipmr_new_tunnel(net, vifc);
872 if (IS_ERR(dev))
873 return PTR_ERR(dev);
874 break;
875 case VIFF_USE_IFINDEX:
876 case 0:
877 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
878 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
879 if (dev && !__in_dev_get_rtnl(dev)) {
880 dev_put(dev);
881 return -EADDRNOTAVAIL;
882 }
883 } else {
884 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
885 }
886 if (!dev)
887 return -EADDRNOTAVAIL;
888 err = dev_set_allmulti(dev, 1);
889 if (err) {
890 dev_put(dev);
891 return err;
892 }
893 break;
894 default:
895 return -EINVAL;
896 }
897
898 in_dev = __in_dev_get_rtnl(dev);
899 if (!in_dev) {
900 dev_put(dev);
901 return -EADDRNOTAVAIL;
902 }
903 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
904 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
905 dev->ifindex, &in_dev->cnf);
906 ip_rt_multicast_event(in_dev);
907
908 /* Fill in the VIF structures */
909 vif_device_init(v, dev, vifc->vifc_rate_limit,
910 vifc->vifc_threshold,
911 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
912 (VIFF_TUNNEL | VIFF_REGISTER));
913
914 err = dev_get_port_parent_id(dev, &ppid, true);
915 if (err == 0) {
916 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
917 v->dev_parent_id.id_len = ppid.id_len;
918 } else {
919 v->dev_parent_id.id_len = 0;
920 }
921
922 v->local = vifc->vifc_lcl_addr.s_addr;
923 v->remote = vifc->vifc_rmt_addr.s_addr;
924
925 /* And finish update writing critical data */
926 spin_lock(&mrt_lock);
927 rcu_assign_pointer(v->dev, dev);
928 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
929 if (v->flags & VIFF_REGISTER) {
930 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
931 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
932 }
933 if (vifi+1 > mrt->maxvif)
934 WRITE_ONCE(mrt->maxvif, vifi + 1);
935 spin_unlock(&mrt_lock);
936 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
937 vifi, mrt->id);
938 return 0;
939}
940
941/* called with rcu_read_lock() */
942static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
943 __be32 origin,
944 __be32 mcastgrp)
945{
946 struct mfc_cache_cmp_arg arg = {
947 .mfc_mcastgrp = mcastgrp,
948 .mfc_origin = origin
949 };
950
951 return mr_mfc_find(mrt, &arg);
952}
953
954/* Look for a (*,G) entry */
955static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
956 __be32 mcastgrp, int vifi)
957{
958 struct mfc_cache_cmp_arg arg = {
959 .mfc_mcastgrp = mcastgrp,
960 .mfc_origin = htonl(INADDR_ANY)
961 };
962
963 if (mcastgrp == htonl(INADDR_ANY))
964 return mr_mfc_find_any_parent(mrt, vifi);
965 return mr_mfc_find_any(mrt, vifi, &arg);
966}
967
968/* Look for a (S,G,iif) entry if parent != -1 */
969static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
970 __be32 origin, __be32 mcastgrp,
971 int parent)
972{
973 struct mfc_cache_cmp_arg arg = {
974 .mfc_mcastgrp = mcastgrp,
975 .mfc_origin = origin,
976 };
977
978 return mr_mfc_find_parent(mrt, &arg, parent);
979}
980
981/* Allocate a multicast cache entry */
982static struct mfc_cache *ipmr_cache_alloc(void)
983{
984 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
985
986 if (c) {
987 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
988 c->_c.mfc_un.res.minvif = MAXVIFS;
989 c->_c.free = ipmr_cache_free_rcu;
990 refcount_set(&c->_c.mfc_un.res.refcount, 1);
991 }
992 return c;
993}
994
995static struct mfc_cache *ipmr_cache_alloc_unres(void)
996{
997 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
998
999 if (c) {
1000 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1001 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1002 }
1003 return c;
1004}
1005
1006/* A cache entry has gone into a resolved state from queued */
1007static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1008 struct mfc_cache *uc, struct mfc_cache *c)
1009{
1010 struct sk_buff *skb;
1011 struct nlmsgerr *e;
1012
1013 /* Play the pending entries through our router */
1014 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1015 if (ip_hdr(skb)->version == 0) {
1016 struct nlmsghdr *nlh = skb_pull(skb,
1017 sizeof(struct iphdr));
1018
1019 if (mr_fill_mroute(mrt, skb, &c->_c,
1020 nlmsg_data(nlh)) > 0) {
1021 nlh->nlmsg_len = skb_tail_pointer(skb) -
1022 (u8 *)nlh;
1023 } else {
1024 nlh->nlmsg_type = NLMSG_ERROR;
1025 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1026 skb_trim(skb, nlh->nlmsg_len);
1027 e = nlmsg_data(nlh);
1028 e->error = -EMSGSIZE;
1029 memset(&e->msg, 0, sizeof(e->msg));
1030 }
1031
1032 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1033 } else {
1034 rcu_read_lock();
1035 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1036 rcu_read_unlock();
1037 }
1038 }
1039}
1040
1041/* Bounce a cache query up to mrouted and netlink.
1042 *
1043 * Called under rcu_read_lock().
1044 */
1045static int ipmr_cache_report(const struct mr_table *mrt,
1046 struct sk_buff *pkt, vifi_t vifi, int assert)
1047{
1048 const int ihl = ip_hdrlen(pkt);
1049 struct sock *mroute_sk;
1050 struct igmphdr *igmp;
1051 struct igmpmsg *msg;
1052 struct sk_buff *skb;
1053 int ret;
1054
1055 mroute_sk = rcu_dereference(mrt->mroute_sk);
1056 if (!mroute_sk)
1057 return -EINVAL;
1058
1059 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1060 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1061 else
1062 skb = alloc_skb(128, GFP_ATOMIC);
1063
1064 if (!skb)
1065 return -ENOBUFS;
1066
1067 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1068 /* Ugly, but we have no choice with this interface.
1069 * Duplicate old header, fix ihl, length etc.
1070 * And all this only to mangle msg->im_msgtype and
1071 * to set msg->im_mbz to "mbz" :-)
1072 */
1073 skb_push(skb, sizeof(struct iphdr));
1074 skb_reset_network_header(skb);
1075 skb_reset_transport_header(skb);
1076 msg = (struct igmpmsg *)skb_network_header(skb);
1077 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1078 msg->im_msgtype = assert;
1079 msg->im_mbz = 0;
1080 if (assert == IGMPMSG_WRVIFWHOLE) {
1081 msg->im_vif = vifi;
1082 msg->im_vif_hi = vifi >> 8;
1083 } else {
1084 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1085 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1086
1087 msg->im_vif = vif_num;
1088 msg->im_vif_hi = vif_num >> 8;
1089 }
1090 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1091 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1092 sizeof(struct iphdr));
1093 } else {
1094 /* Copy the IP header */
1095 skb_set_network_header(skb, skb->len);
1096 skb_put(skb, ihl);
1097 skb_copy_to_linear_data(skb, pkt->data, ihl);
1098 /* Flag to the kernel this is a route add */
1099 ip_hdr(skb)->protocol = 0;
1100 msg = (struct igmpmsg *)skb_network_header(skb);
1101 msg->im_vif = vifi;
1102 msg->im_vif_hi = vifi >> 8;
1103 ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1104 memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1105 /* Add our header */
1106 igmp = skb_put(skb, sizeof(struct igmphdr));
1107 igmp->type = assert;
1108 msg->im_msgtype = assert;
1109 igmp->code = 0;
1110 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1111 skb->transport_header = skb->network_header;
1112 }
1113
1114 igmpmsg_netlink_event(mrt, skb);
1115
1116 /* Deliver to mrouted */
1117 ret = sock_queue_rcv_skb(mroute_sk, skb);
1118
1119 if (ret < 0) {
1120 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1121 kfree_skb(skb);
1122 }
1123
1124 return ret;
1125}
1126
1127/* Queue a packet for resolution. It gets locked cache entry! */
1128/* Called under rcu_read_lock() */
1129static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1130 struct sk_buff *skb, struct net_device *dev)
1131{
1132 const struct iphdr *iph = ip_hdr(skb);
1133 struct mfc_cache *c;
1134 bool found = false;
1135 int err;
1136
1137 spin_lock_bh(&mfc_unres_lock);
1138 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1139 if (c->mfc_mcastgrp == iph->daddr &&
1140 c->mfc_origin == iph->saddr) {
1141 found = true;
1142 break;
1143 }
1144 }
1145
1146 if (!found) {
1147 /* Create a new entry if allowable */
1148 c = ipmr_cache_alloc_unres();
1149 if (!c) {
1150 spin_unlock_bh(&mfc_unres_lock);
1151
1152 kfree_skb(skb);
1153 return -ENOBUFS;
1154 }
1155
1156 /* Fill in the new cache entry */
1157 c->_c.mfc_parent = -1;
1158 c->mfc_origin = iph->saddr;
1159 c->mfc_mcastgrp = iph->daddr;
1160
1161 /* Reflect first query at mrouted. */
1162 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1163
1164 if (err < 0) {
1165 /* If the report failed throw the cache entry
1166 out - Brad Parker
1167 */
1168 spin_unlock_bh(&mfc_unres_lock);
1169
1170 ipmr_cache_free(c);
1171 kfree_skb(skb);
1172 return err;
1173 }
1174
1175 atomic_inc(&mrt->cache_resolve_queue_len);
1176 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1177 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1178
1179 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1180 mod_timer(&mrt->ipmr_expire_timer,
1181 c->_c.mfc_un.unres.expires);
1182 }
1183
1184 /* See if we can append the packet */
1185 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1186 kfree_skb(skb);
1187 err = -ENOBUFS;
1188 } else {
1189 if (dev) {
1190 skb->dev = dev;
1191 skb->skb_iif = dev->ifindex;
1192 }
1193 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1194 err = 0;
1195 }
1196
1197 spin_unlock_bh(&mfc_unres_lock);
1198 return err;
1199}
1200
1201/* MFC cache manipulation by user space mroute daemon */
1202
1203static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1204{
1205 struct net *net = read_pnet(&mrt->net);
1206 struct mfc_cache *c;
1207
1208 /* The entries are added/deleted only under RTNL */
1209 rcu_read_lock();
1210 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1211 mfc->mfcc_mcastgrp.s_addr, parent);
1212 rcu_read_unlock();
1213 if (!c)
1214 return -ENOENT;
1215 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1216 list_del_rcu(&c->_c.list);
1217 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1218 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1219 mr_cache_put(&c->_c);
1220
1221 return 0;
1222}
1223
1224static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1225 struct mfcctl *mfc, int mrtsock, int parent)
1226{
1227 struct mfc_cache *uc, *c;
1228 struct mr_mfc *_uc;
1229 bool found;
1230 int ret;
1231
1232 if (mfc->mfcc_parent >= MAXVIFS)
1233 return -ENFILE;
1234
1235 /* The entries are added/deleted only under RTNL */
1236 rcu_read_lock();
1237 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1238 mfc->mfcc_mcastgrp.s_addr, parent);
1239 rcu_read_unlock();
1240 if (c) {
1241 spin_lock(&mrt_lock);
1242 c->_c.mfc_parent = mfc->mfcc_parent;
1243 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1244 if (!mrtsock)
1245 c->_c.mfc_flags |= MFC_STATIC;
1246 spin_unlock(&mrt_lock);
1247 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1248 mrt->id);
1249 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1250 return 0;
1251 }
1252
1253 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1254 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1255 return -EINVAL;
1256
1257 c = ipmr_cache_alloc();
1258 if (!c)
1259 return -ENOMEM;
1260
1261 c->mfc_origin = mfc->mfcc_origin.s_addr;
1262 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1263 c->_c.mfc_parent = mfc->mfcc_parent;
1264 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1265 if (!mrtsock)
1266 c->_c.mfc_flags |= MFC_STATIC;
1267
1268 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1269 ipmr_rht_params);
1270 if (ret) {
1271 pr_err("ipmr: rhtable insert error %d\n", ret);
1272 ipmr_cache_free(c);
1273 return ret;
1274 }
1275 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1276 /* Check to see if we resolved a queued list. If so we
1277 * need to send on the frames and tidy up.
1278 */
1279 found = false;
1280 spin_lock_bh(&mfc_unres_lock);
1281 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1282 uc = (struct mfc_cache *)_uc;
1283 if (uc->mfc_origin == c->mfc_origin &&
1284 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1285 list_del(&_uc->list);
1286 atomic_dec(&mrt->cache_resolve_queue_len);
1287 found = true;
1288 break;
1289 }
1290 }
1291 if (list_empty(&mrt->mfc_unres_queue))
1292 del_timer(&mrt->ipmr_expire_timer);
1293 spin_unlock_bh(&mfc_unres_lock);
1294
1295 if (found) {
1296 ipmr_cache_resolve(net, mrt, uc, c);
1297 ipmr_cache_free(uc);
1298 }
1299 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1300 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1301 return 0;
1302}
1303
1304/* Close the multicast socket, and clear the vif tables etc */
1305static void mroute_clean_tables(struct mr_table *mrt, int flags)
1306{
1307 struct net *net = read_pnet(&mrt->net);
1308 struct mr_mfc *c, *tmp;
1309 struct mfc_cache *cache;
1310 LIST_HEAD(list);
1311 int i;
1312
1313 /* Shut down all active vif entries */
1314 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1315 for (i = 0; i < mrt->maxvif; i++) {
1316 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1317 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1318 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1319 continue;
1320 vif_delete(mrt, i, 0, &list);
1321 }
1322 unregister_netdevice_many(&list);
1323 }
1324
1325 /* Wipe the cache */
1326 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1327 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1328 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1329 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1330 continue;
1331 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1332 list_del_rcu(&c->list);
1333 cache = (struct mfc_cache *)c;
1334 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1335 mrt->id);
1336 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1337 mr_cache_put(c);
1338 }
1339 }
1340
1341 if (flags & MRT_FLUSH_MFC) {
1342 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1343 spin_lock_bh(&mfc_unres_lock);
1344 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1345 list_del(&c->list);
1346 cache = (struct mfc_cache *)c;
1347 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1348 ipmr_destroy_unres(mrt, cache);
1349 }
1350 spin_unlock_bh(&mfc_unres_lock);
1351 }
1352 }
1353}
1354
1355/* called from ip_ra_control(), before an RCU grace period,
1356 * we don't need to call synchronize_rcu() here
1357 */
1358static void mrtsock_destruct(struct sock *sk)
1359{
1360 struct net *net = sock_net(sk);
1361 struct mr_table *mrt;
1362
1363 rtnl_lock();
1364 ipmr_for_each_table(mrt, net) {
1365 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1366 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1367 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1368 NETCONFA_MC_FORWARDING,
1369 NETCONFA_IFINDEX_ALL,
1370 net->ipv4.devconf_all);
1371 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1372 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1373 }
1374 }
1375 rtnl_unlock();
1376}
1377
1378/* Socket options and virtual interface manipulation. The whole
1379 * virtual interface system is a complete heap, but unfortunately
1380 * that's how BSD mrouted happens to think. Maybe one day with a proper
1381 * MOSPF/PIM router set up we can clean this up.
1382 */
1383
1384int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1385 unsigned int optlen)
1386{
1387 struct net *net = sock_net(sk);
1388 int val, ret = 0, parent = 0;
1389 struct mr_table *mrt;
1390 struct vifctl vif;
1391 struct mfcctl mfc;
1392 bool do_wrvifwhole;
1393 u32 uval;
1394
1395 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1396 rtnl_lock();
1397 if (sk->sk_type != SOCK_RAW ||
1398 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1399 ret = -EOPNOTSUPP;
1400 goto out_unlock;
1401 }
1402
1403 mrt = __ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1404 if (!mrt) {
1405 ret = -ENOENT;
1406 goto out_unlock;
1407 }
1408 if (optname != MRT_INIT) {
1409 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1410 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1411 ret = -EACCES;
1412 goto out_unlock;
1413 }
1414 }
1415
1416 switch (optname) {
1417 case MRT_INIT:
1418 if (optlen != sizeof(int)) {
1419 ret = -EINVAL;
1420 break;
1421 }
1422 if (rtnl_dereference(mrt->mroute_sk)) {
1423 ret = -EADDRINUSE;
1424 break;
1425 }
1426
1427 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1428 if (ret == 0) {
1429 rcu_assign_pointer(mrt->mroute_sk, sk);
1430 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1431 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1432 NETCONFA_MC_FORWARDING,
1433 NETCONFA_IFINDEX_ALL,
1434 net->ipv4.devconf_all);
1435 }
1436 break;
1437 case MRT_DONE:
1438 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1439 ret = -EACCES;
1440 } else {
1441 /* We need to unlock here because mrtsock_destruct takes
1442 * care of rtnl itself and we can't change that due to
1443 * the IP_ROUTER_ALERT setsockopt which runs without it.
1444 */
1445 rtnl_unlock();
1446 ret = ip_ra_control(sk, 0, NULL);
1447 goto out;
1448 }
1449 break;
1450 case MRT_ADD_VIF:
1451 case MRT_DEL_VIF:
1452 if (optlen != sizeof(vif)) {
1453 ret = -EINVAL;
1454 break;
1455 }
1456 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1457 ret = -EFAULT;
1458 break;
1459 }
1460 if (vif.vifc_vifi >= MAXVIFS) {
1461 ret = -ENFILE;
1462 break;
1463 }
1464 if (optname == MRT_ADD_VIF) {
1465 ret = vif_add(net, mrt, &vif,
1466 sk == rtnl_dereference(mrt->mroute_sk));
1467 } else {
1468 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1469 }
1470 break;
1471 /* Manipulate the forwarding caches. These live
1472 * in a sort of kernel/user symbiosis.
1473 */
1474 case MRT_ADD_MFC:
1475 case MRT_DEL_MFC:
1476 parent = -1;
1477 fallthrough;
1478 case MRT_ADD_MFC_PROXY:
1479 case MRT_DEL_MFC_PROXY:
1480 if (optlen != sizeof(mfc)) {
1481 ret = -EINVAL;
1482 break;
1483 }
1484 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1485 ret = -EFAULT;
1486 break;
1487 }
1488 if (parent == 0)
1489 parent = mfc.mfcc_parent;
1490 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1491 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1492 else
1493 ret = ipmr_mfc_add(net, mrt, &mfc,
1494 sk == rtnl_dereference(mrt->mroute_sk),
1495 parent);
1496 break;
1497 case MRT_FLUSH:
1498 if (optlen != sizeof(val)) {
1499 ret = -EINVAL;
1500 break;
1501 }
1502 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1503 ret = -EFAULT;
1504 break;
1505 }
1506 mroute_clean_tables(mrt, val);
1507 break;
1508 /* Control PIM assert. */
1509 case MRT_ASSERT:
1510 if (optlen != sizeof(val)) {
1511 ret = -EINVAL;
1512 break;
1513 }
1514 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1515 ret = -EFAULT;
1516 break;
1517 }
1518 mrt->mroute_do_assert = val;
1519 break;
1520 case MRT_PIM:
1521 if (!ipmr_pimsm_enabled()) {
1522 ret = -ENOPROTOOPT;
1523 break;
1524 }
1525 if (optlen != sizeof(val)) {
1526 ret = -EINVAL;
1527 break;
1528 }
1529 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1530 ret = -EFAULT;
1531 break;
1532 }
1533
1534 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1535 val = !!val;
1536 if (val != mrt->mroute_do_pim) {
1537 mrt->mroute_do_pim = val;
1538 mrt->mroute_do_assert = val;
1539 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1540 }
1541 break;
1542 case MRT_TABLE:
1543 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1544 ret = -ENOPROTOOPT;
1545 break;
1546 }
1547 if (optlen != sizeof(uval)) {
1548 ret = -EINVAL;
1549 break;
1550 }
1551 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1552 ret = -EFAULT;
1553 break;
1554 }
1555
1556 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1557 ret = -EBUSY;
1558 } else {
1559 mrt = ipmr_new_table(net, uval);
1560 if (IS_ERR(mrt))
1561 ret = PTR_ERR(mrt);
1562 else
1563 raw_sk(sk)->ipmr_table = uval;
1564 }
1565 break;
1566 /* Spurious command, or MRT_VERSION which you cannot set. */
1567 default:
1568 ret = -ENOPROTOOPT;
1569 }
1570out_unlock:
1571 rtnl_unlock();
1572out:
1573 return ret;
1574}
1575
1576/* Execute if this ioctl is a special mroute ioctl */
1577int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1578{
1579 switch (cmd) {
1580 /* These userspace buffers will be consumed by ipmr_ioctl() */
1581 case SIOCGETVIFCNT: {
1582 struct sioc_vif_req buffer;
1583
1584 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1585 sizeof(buffer));
1586 }
1587 case SIOCGETSGCNT: {
1588 struct sioc_sg_req buffer;
1589
1590 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1591 sizeof(buffer));
1592 }
1593 }
1594 /* return code > 0 means that the ioctl was not executed */
1595 return 1;
1596}
1597
1598/* Getsock opt support for the multicast routing system. */
1599int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1600 sockptr_t optlen)
1601{
1602 int olr;
1603 int val;
1604 struct net *net = sock_net(sk);
1605 struct mr_table *mrt;
1606
1607 if (sk->sk_type != SOCK_RAW ||
1608 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1609 return -EOPNOTSUPP;
1610
1611 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1612 if (!mrt)
1613 return -ENOENT;
1614
1615 switch (optname) {
1616 case MRT_VERSION:
1617 val = 0x0305;
1618 break;
1619 case MRT_PIM:
1620 if (!ipmr_pimsm_enabled())
1621 return -ENOPROTOOPT;
1622 val = mrt->mroute_do_pim;
1623 break;
1624 case MRT_ASSERT:
1625 val = mrt->mroute_do_assert;
1626 break;
1627 default:
1628 return -ENOPROTOOPT;
1629 }
1630
1631 if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1632 return -EFAULT;
1633 if (olr < 0)
1634 return -EINVAL;
1635
1636 olr = min_t(unsigned int, olr, sizeof(int));
1637
1638 if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1639 return -EFAULT;
1640 if (copy_to_sockptr(optval, &val, olr))
1641 return -EFAULT;
1642 return 0;
1643}
1644
1645/* The IP multicast ioctl support routines. */
1646int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1647{
1648 struct vif_device *vif;
1649 struct mfc_cache *c;
1650 struct net *net = sock_net(sk);
1651 struct sioc_vif_req *vr;
1652 struct sioc_sg_req *sr;
1653 struct mr_table *mrt;
1654
1655 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1656 if (!mrt)
1657 return -ENOENT;
1658
1659 switch (cmd) {
1660 case SIOCGETVIFCNT:
1661 vr = (struct sioc_vif_req *)arg;
1662 if (vr->vifi >= mrt->maxvif)
1663 return -EINVAL;
1664 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1665 rcu_read_lock();
1666 vif = &mrt->vif_table[vr->vifi];
1667 if (VIF_EXISTS(mrt, vr->vifi)) {
1668 vr->icount = READ_ONCE(vif->pkt_in);
1669 vr->ocount = READ_ONCE(vif->pkt_out);
1670 vr->ibytes = READ_ONCE(vif->bytes_in);
1671 vr->obytes = READ_ONCE(vif->bytes_out);
1672 rcu_read_unlock();
1673
1674 return 0;
1675 }
1676 rcu_read_unlock();
1677 return -EADDRNOTAVAIL;
1678 case SIOCGETSGCNT:
1679 sr = (struct sioc_sg_req *)arg;
1680
1681 rcu_read_lock();
1682 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1683 if (c) {
1684 sr->pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1685 sr->bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1686 sr->wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1687 rcu_read_unlock();
1688 return 0;
1689 }
1690 rcu_read_unlock();
1691 return -EADDRNOTAVAIL;
1692 default:
1693 return -ENOIOCTLCMD;
1694 }
1695}
1696
1697#ifdef CONFIG_COMPAT
1698struct compat_sioc_sg_req {
1699 struct in_addr src;
1700 struct in_addr grp;
1701 compat_ulong_t pktcnt;
1702 compat_ulong_t bytecnt;
1703 compat_ulong_t wrong_if;
1704};
1705
1706struct compat_sioc_vif_req {
1707 vifi_t vifi; /* Which iface */
1708 compat_ulong_t icount;
1709 compat_ulong_t ocount;
1710 compat_ulong_t ibytes;
1711 compat_ulong_t obytes;
1712};
1713
1714int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1715{
1716 struct compat_sioc_sg_req sr;
1717 struct compat_sioc_vif_req vr;
1718 struct vif_device *vif;
1719 struct mfc_cache *c;
1720 struct net *net = sock_net(sk);
1721 struct mr_table *mrt;
1722
1723 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1724 if (!mrt)
1725 return -ENOENT;
1726
1727 switch (cmd) {
1728 case SIOCGETVIFCNT:
1729 if (copy_from_user(&vr, arg, sizeof(vr)))
1730 return -EFAULT;
1731 if (vr.vifi >= mrt->maxvif)
1732 return -EINVAL;
1733 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1734 rcu_read_lock();
1735 vif = &mrt->vif_table[vr.vifi];
1736 if (VIF_EXISTS(mrt, vr.vifi)) {
1737 vr.icount = READ_ONCE(vif->pkt_in);
1738 vr.ocount = READ_ONCE(vif->pkt_out);
1739 vr.ibytes = READ_ONCE(vif->bytes_in);
1740 vr.obytes = READ_ONCE(vif->bytes_out);
1741 rcu_read_unlock();
1742
1743 if (copy_to_user(arg, &vr, sizeof(vr)))
1744 return -EFAULT;
1745 return 0;
1746 }
1747 rcu_read_unlock();
1748 return -EADDRNOTAVAIL;
1749 case SIOCGETSGCNT:
1750 if (copy_from_user(&sr, arg, sizeof(sr)))
1751 return -EFAULT;
1752
1753 rcu_read_lock();
1754 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1755 if (c) {
1756 sr.pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1757 sr.bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1758 sr.wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1759 rcu_read_unlock();
1760
1761 if (copy_to_user(arg, &sr, sizeof(sr)))
1762 return -EFAULT;
1763 return 0;
1764 }
1765 rcu_read_unlock();
1766 return -EADDRNOTAVAIL;
1767 default:
1768 return -ENOIOCTLCMD;
1769 }
1770}
1771#endif
1772
1773static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1774{
1775 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1776 struct net *net = dev_net(dev);
1777 struct mr_table *mrt;
1778 struct vif_device *v;
1779 int ct;
1780
1781 if (event != NETDEV_UNREGISTER)
1782 return NOTIFY_DONE;
1783
1784 ipmr_for_each_table(mrt, net) {
1785 v = &mrt->vif_table[0];
1786 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1787 if (rcu_access_pointer(v->dev) == dev)
1788 vif_delete(mrt, ct, 1, NULL);
1789 }
1790 }
1791 return NOTIFY_DONE;
1792}
1793
1794static struct notifier_block ip_mr_notifier = {
1795 .notifier_call = ipmr_device_event,
1796};
1797
1798/* Encapsulate a packet by attaching a valid IPIP header to it.
1799 * This avoids tunnel drivers and other mess and gives us the speed so
1800 * important for multicast video.
1801 */
1802static void ip_encap(struct net *net, struct sk_buff *skb,
1803 __be32 saddr, __be32 daddr)
1804{
1805 struct iphdr *iph;
1806 const struct iphdr *old_iph = ip_hdr(skb);
1807
1808 skb_push(skb, sizeof(struct iphdr));
1809 skb->transport_header = skb->network_header;
1810 skb_reset_network_header(skb);
1811 iph = ip_hdr(skb);
1812
1813 iph->version = 4;
1814 iph->tos = old_iph->tos;
1815 iph->ttl = old_iph->ttl;
1816 iph->frag_off = 0;
1817 iph->daddr = daddr;
1818 iph->saddr = saddr;
1819 iph->protocol = IPPROTO_IPIP;
1820 iph->ihl = 5;
1821 iph->tot_len = htons(skb->len);
1822 ip_select_ident(net, skb, NULL);
1823 ip_send_check(iph);
1824
1825 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1826 nf_reset_ct(skb);
1827}
1828
1829static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1830 struct sk_buff *skb)
1831{
1832 struct ip_options *opt = &(IPCB(skb)->opt);
1833
1834 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1835
1836 if (unlikely(opt->optlen))
1837 ip_forward_options(skb);
1838
1839 return dst_output(net, sk, skb);
1840}
1841
1842#ifdef CONFIG_NET_SWITCHDEV
1843static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1844 int in_vifi, int out_vifi)
1845{
1846 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1847 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1848
1849 if (!skb->offload_l3_fwd_mark)
1850 return false;
1851 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1852 return false;
1853 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1854 &in_vif->dev_parent_id);
1855}
1856#else
1857static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1858 int in_vifi, int out_vifi)
1859{
1860 return false;
1861}
1862#endif
1863
1864/* Processing handlers for ipmr_forward, under rcu_read_lock() */
1865
1866static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1867 int in_vifi, struct sk_buff *skb, int vifi)
1868{
1869 const struct iphdr *iph = ip_hdr(skb);
1870 struct vif_device *vif = &mrt->vif_table[vifi];
1871 struct net_device *vif_dev;
1872 struct net_device *dev;
1873 struct rtable *rt;
1874 struct flowi4 fl4;
1875 int encap = 0;
1876
1877 vif_dev = vif_dev_read(vif);
1878 if (!vif_dev)
1879 goto out_free;
1880
1881 if (vif->flags & VIFF_REGISTER) {
1882 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1883 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1884 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1885 DEV_STATS_INC(vif_dev, tx_packets);
1886 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1887 goto out_free;
1888 }
1889
1890 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1891 goto out_free;
1892
1893 if (vif->flags & VIFF_TUNNEL) {
1894 rt = ip_route_output_ports(net, &fl4, NULL,
1895 vif->remote, vif->local,
1896 0, 0,
1897 IPPROTO_IPIP,
1898 iph->tos & INET_DSCP_MASK, vif->link);
1899 if (IS_ERR(rt))
1900 goto out_free;
1901 encap = sizeof(struct iphdr);
1902 } else {
1903 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1904 0, 0,
1905 IPPROTO_IPIP,
1906 iph->tos & INET_DSCP_MASK, vif->link);
1907 if (IS_ERR(rt))
1908 goto out_free;
1909 }
1910
1911 dev = rt->dst.dev;
1912
1913 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1914 /* Do not fragment multicasts. Alas, IPv4 does not
1915 * allow to send ICMP, so that packets will disappear
1916 * to blackhole.
1917 */
1918 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1919 ip_rt_put(rt);
1920 goto out_free;
1921 }
1922
1923 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1924
1925 if (skb_cow(skb, encap)) {
1926 ip_rt_put(rt);
1927 goto out_free;
1928 }
1929
1930 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1931 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1932
1933 skb_dst_drop(skb);
1934 skb_dst_set(skb, &rt->dst);
1935 ip_decrease_ttl(ip_hdr(skb));
1936
1937 /* FIXME: forward and output firewalls used to be called here.
1938 * What do we do with netfilter? -- RR
1939 */
1940 if (vif->flags & VIFF_TUNNEL) {
1941 ip_encap(net, skb, vif->local, vif->remote);
1942 /* FIXME: extra output firewall step used to be here. --RR */
1943 DEV_STATS_INC(vif_dev, tx_packets);
1944 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1945 }
1946
1947 IPCB(skb)->flags |= IPSKB_FORWARDED;
1948
1949 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1950 * not only before forwarding, but after forwarding on all output
1951 * interfaces. It is clear, if mrouter runs a multicasting
1952 * program, it should receive packets not depending to what interface
1953 * program is joined.
1954 * If we will not make it, the program will have to join on all
1955 * interfaces. On the other hand, multihoming host (or router, but
1956 * not mrouter) cannot join to more than one interface - it will
1957 * result in receiving multiple packets.
1958 */
1959 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1960 net, NULL, skb, skb->dev, dev,
1961 ipmr_forward_finish);
1962 return;
1963
1964out_free:
1965 kfree_skb(skb);
1966}
1967
1968/* Called with mrt_lock or rcu_read_lock() */
1969static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1970{
1971 int ct;
1972 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1973 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1974 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1975 break;
1976 }
1977 return ct;
1978}
1979
1980/* "local" means that we should preserve one skb (for local delivery) */
1981/* Called uner rcu_read_lock() */
1982static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1983 struct net_device *dev, struct sk_buff *skb,
1984 struct mfc_cache *c, int local)
1985{
1986 int true_vifi = ipmr_find_vif(mrt, dev);
1987 int psend = -1;
1988 int vif, ct;
1989
1990 vif = c->_c.mfc_parent;
1991 atomic_long_inc(&c->_c.mfc_un.res.pkt);
1992 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes);
1993 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies);
1994
1995 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1996 struct mfc_cache *cache_proxy;
1997
1998 /* For an (*,G) entry, we only check that the incoming
1999 * interface is part of the static tree.
2000 */
2001 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2002 if (cache_proxy &&
2003 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
2004 goto forward;
2005 }
2006
2007 /* Wrong interface: drop packet and (maybe) send PIM assert. */
2008 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
2009 if (rt_is_output_route(skb_rtable(skb))) {
2010 /* It is our own packet, looped back.
2011 * Very complicated situation...
2012 *
2013 * The best workaround until routing daemons will be
2014 * fixed is not to redistribute packet, if it was
2015 * send through wrong interface. It means, that
2016 * multicast applications WILL NOT work for
2017 * (S,G), which have default multicast route pointing
2018 * to wrong oif. In any case, it is not a good
2019 * idea to use multicasting applications on router.
2020 */
2021 goto dont_forward;
2022 }
2023
2024 atomic_long_inc(&c->_c.mfc_un.res.wrong_if);
2025
2026 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2027 /* pimsm uses asserts, when switching from RPT to SPT,
2028 * so that we cannot check that packet arrived on an oif.
2029 * It is bad, but otherwise we would need to move pretty
2030 * large chunk of pimd to kernel. Ough... --ANK
2031 */
2032 (mrt->mroute_do_pim ||
2033 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2034 time_after(jiffies,
2035 c->_c.mfc_un.res.last_assert +
2036 MFC_ASSERT_THRESH)) {
2037 c->_c.mfc_un.res.last_assert = jiffies;
2038 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2039 if (mrt->mroute_do_wrvifwhole)
2040 ipmr_cache_report(mrt, skb, true_vifi,
2041 IGMPMSG_WRVIFWHOLE);
2042 }
2043 goto dont_forward;
2044 }
2045
2046forward:
2047 WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2048 mrt->vif_table[vif].pkt_in + 1);
2049 WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2050 mrt->vif_table[vif].bytes_in + skb->len);
2051
2052 /* Forward the frame */
2053 if (c->mfc_origin == htonl(INADDR_ANY) &&
2054 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2055 if (true_vifi >= 0 &&
2056 true_vifi != c->_c.mfc_parent &&
2057 ip_hdr(skb)->ttl >
2058 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2059 /* It's an (*,*) entry and the packet is not coming from
2060 * the upstream: forward the packet to the upstream
2061 * only.
2062 */
2063 psend = c->_c.mfc_parent;
2064 goto last_forward;
2065 }
2066 goto dont_forward;
2067 }
2068 for (ct = c->_c.mfc_un.res.maxvif - 1;
2069 ct >= c->_c.mfc_un.res.minvif; ct--) {
2070 /* For (*,G) entry, don't forward to the incoming interface */
2071 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2072 ct != true_vifi) &&
2073 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2074 if (psend != -1) {
2075 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2076
2077 if (skb2)
2078 ipmr_queue_xmit(net, mrt, true_vifi,
2079 skb2, psend);
2080 }
2081 psend = ct;
2082 }
2083 }
2084last_forward:
2085 if (psend != -1) {
2086 if (local) {
2087 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2088
2089 if (skb2)
2090 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2091 psend);
2092 } else {
2093 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2094 return;
2095 }
2096 }
2097
2098dont_forward:
2099 if (!local)
2100 kfree_skb(skb);
2101}
2102
2103static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2104{
2105 struct rtable *rt = skb_rtable(skb);
2106 struct iphdr *iph = ip_hdr(skb);
2107 struct flowi4 fl4 = {
2108 .daddr = iph->daddr,
2109 .saddr = iph->saddr,
2110 .flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)),
2111 .flowi4_oif = (rt_is_output_route(rt) ?
2112 skb->dev->ifindex : 0),
2113 .flowi4_iif = (rt_is_output_route(rt) ?
2114 LOOPBACK_IFINDEX :
2115 skb->dev->ifindex),
2116 .flowi4_mark = skb->mark,
2117 };
2118 struct mr_table *mrt;
2119 int err;
2120
2121 err = ipmr_fib_lookup(net, &fl4, &mrt);
2122 if (err)
2123 return ERR_PTR(err);
2124 return mrt;
2125}
2126
2127/* Multicast packets for forwarding arrive here
2128 * Called with rcu_read_lock();
2129 */
2130int ip_mr_input(struct sk_buff *skb)
2131{
2132 struct mfc_cache *cache;
2133 struct net *net = dev_net(skb->dev);
2134 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2135 struct mr_table *mrt;
2136 struct net_device *dev;
2137
2138 /* skb->dev passed in is the loX master dev for vrfs.
2139 * As there are no vifs associated with loopback devices,
2140 * get the proper interface that does have a vif associated with it.
2141 */
2142 dev = skb->dev;
2143 if (netif_is_l3_master(skb->dev)) {
2144 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2145 if (!dev) {
2146 kfree_skb(skb);
2147 return -ENODEV;
2148 }
2149 }
2150
2151 /* Packet is looped back after forward, it should not be
2152 * forwarded second time, but still can be delivered locally.
2153 */
2154 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2155 goto dont_forward;
2156
2157 mrt = ipmr_rt_fib_lookup(net, skb);
2158 if (IS_ERR(mrt)) {
2159 kfree_skb(skb);
2160 return PTR_ERR(mrt);
2161 }
2162 if (!local) {
2163 if (IPCB(skb)->opt.router_alert) {
2164 if (ip_call_ra_chain(skb))
2165 return 0;
2166 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2167 /* IGMPv1 (and broken IGMPv2 implementations sort of
2168 * Cisco IOS <= 11.2(8)) do not put router alert
2169 * option to IGMP packets destined to routable
2170 * groups. It is very bad, because it means
2171 * that we can forward NO IGMP messages.
2172 */
2173 struct sock *mroute_sk;
2174
2175 mroute_sk = rcu_dereference(mrt->mroute_sk);
2176 if (mroute_sk) {
2177 nf_reset_ct(skb);
2178 raw_rcv(mroute_sk, skb);
2179 return 0;
2180 }
2181 }
2182 }
2183
2184 /* already under rcu_read_lock() */
2185 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2186 if (!cache) {
2187 int vif = ipmr_find_vif(mrt, dev);
2188
2189 if (vif >= 0)
2190 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2191 vif);
2192 }
2193
2194 /* No usable cache entry */
2195 if (!cache) {
2196 int vif;
2197
2198 if (local) {
2199 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2200 ip_local_deliver(skb);
2201 if (!skb2)
2202 return -ENOBUFS;
2203 skb = skb2;
2204 }
2205
2206 vif = ipmr_find_vif(mrt, dev);
2207 if (vif >= 0)
2208 return ipmr_cache_unresolved(mrt, vif, skb, dev);
2209 kfree_skb(skb);
2210 return -ENODEV;
2211 }
2212
2213 ip_mr_forward(net, mrt, dev, skb, cache, local);
2214
2215 if (local)
2216 return ip_local_deliver(skb);
2217
2218 return 0;
2219
2220dont_forward:
2221 if (local)
2222 return ip_local_deliver(skb);
2223 kfree_skb(skb);
2224 return 0;
2225}
2226
2227#ifdef CONFIG_IP_PIMSM_V1
2228/* Handle IGMP messages of PIMv1 */
2229int pim_rcv_v1(struct sk_buff *skb)
2230{
2231 struct igmphdr *pim;
2232 struct net *net = dev_net(skb->dev);
2233 struct mr_table *mrt;
2234
2235 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2236 goto drop;
2237
2238 pim = igmp_hdr(skb);
2239
2240 mrt = ipmr_rt_fib_lookup(net, skb);
2241 if (IS_ERR(mrt))
2242 goto drop;
2243 if (!mrt->mroute_do_pim ||
2244 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2245 goto drop;
2246
2247 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2248drop:
2249 kfree_skb(skb);
2250 }
2251 return 0;
2252}
2253#endif
2254
2255#ifdef CONFIG_IP_PIMSM_V2
2256static int pim_rcv(struct sk_buff *skb)
2257{
2258 struct pimreghdr *pim;
2259 struct net *net = dev_net(skb->dev);
2260 struct mr_table *mrt;
2261
2262 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2263 goto drop;
2264
2265 pim = (struct pimreghdr *)skb_transport_header(skb);
2266 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2267 (pim->flags & PIM_NULL_REGISTER) ||
2268 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2269 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2270 goto drop;
2271
2272 mrt = ipmr_rt_fib_lookup(net, skb);
2273 if (IS_ERR(mrt))
2274 goto drop;
2275 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2276drop:
2277 kfree_skb(skb);
2278 }
2279 return 0;
2280}
2281#endif
2282
2283int ipmr_get_route(struct net *net, struct sk_buff *skb,
2284 __be32 saddr, __be32 daddr,
2285 struct rtmsg *rtm, u32 portid)
2286{
2287 struct mfc_cache *cache;
2288 struct mr_table *mrt;
2289 int err;
2290
2291 rcu_read_lock();
2292 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
2293 if (!mrt) {
2294 rcu_read_unlock();
2295 return -ENOENT;
2296 }
2297
2298 cache = ipmr_cache_find(mrt, saddr, daddr);
2299 if (!cache && skb->dev) {
2300 int vif = ipmr_find_vif(mrt, skb->dev);
2301
2302 if (vif >= 0)
2303 cache = ipmr_cache_find_any(mrt, daddr, vif);
2304 }
2305 if (!cache) {
2306 struct sk_buff *skb2;
2307 struct iphdr *iph;
2308 struct net_device *dev;
2309 int vif = -1;
2310
2311 dev = skb->dev;
2312 if (dev)
2313 vif = ipmr_find_vif(mrt, dev);
2314 if (vif < 0) {
2315 rcu_read_unlock();
2316 return -ENODEV;
2317 }
2318
2319 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2320 if (!skb2) {
2321 rcu_read_unlock();
2322 return -ENOMEM;
2323 }
2324
2325 NETLINK_CB(skb2).portid = portid;
2326 skb_push(skb2, sizeof(struct iphdr));
2327 skb_reset_network_header(skb2);
2328 iph = ip_hdr(skb2);
2329 iph->ihl = sizeof(struct iphdr) >> 2;
2330 iph->saddr = saddr;
2331 iph->daddr = daddr;
2332 iph->version = 0;
2333 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2334 rcu_read_unlock();
2335 return err;
2336 }
2337
2338 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2339 rcu_read_unlock();
2340 return err;
2341}
2342
2343static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2344 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2345 int flags)
2346{
2347 struct nlmsghdr *nlh;
2348 struct rtmsg *rtm;
2349 int err;
2350
2351 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2352 if (!nlh)
2353 return -EMSGSIZE;
2354
2355 rtm = nlmsg_data(nlh);
2356 rtm->rtm_family = RTNL_FAMILY_IPMR;
2357 rtm->rtm_dst_len = 32;
2358 rtm->rtm_src_len = 32;
2359 rtm->rtm_tos = 0;
2360 rtm->rtm_table = mrt->id;
2361 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2362 goto nla_put_failure;
2363 rtm->rtm_type = RTN_MULTICAST;
2364 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2365 if (c->_c.mfc_flags & MFC_STATIC)
2366 rtm->rtm_protocol = RTPROT_STATIC;
2367 else
2368 rtm->rtm_protocol = RTPROT_MROUTED;
2369 rtm->rtm_flags = 0;
2370
2371 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2372 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2373 goto nla_put_failure;
2374 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2375 /* do not break the dump if cache is unresolved */
2376 if (err < 0 && err != -ENOENT)
2377 goto nla_put_failure;
2378
2379 nlmsg_end(skb, nlh);
2380 return 0;
2381
2382nla_put_failure:
2383 nlmsg_cancel(skb, nlh);
2384 return -EMSGSIZE;
2385}
2386
2387static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2388 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2389 int flags)
2390{
2391 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2392 cmd, flags);
2393}
2394
2395static size_t mroute_msgsize(bool unresolved, int maxvif)
2396{
2397 size_t len =
2398 NLMSG_ALIGN(sizeof(struct rtmsg))
2399 + nla_total_size(4) /* RTA_TABLE */
2400 + nla_total_size(4) /* RTA_SRC */
2401 + nla_total_size(4) /* RTA_DST */
2402 ;
2403
2404 if (!unresolved)
2405 len = len
2406 + nla_total_size(4) /* RTA_IIF */
2407 + nla_total_size(0) /* RTA_MULTIPATH */
2408 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2409 /* RTA_MFC_STATS */
2410 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2411 ;
2412
2413 return len;
2414}
2415
2416static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2417 int cmd)
2418{
2419 struct net *net = read_pnet(&mrt->net);
2420 struct sk_buff *skb;
2421 int err = -ENOBUFS;
2422
2423 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2424 mrt->maxvif),
2425 GFP_ATOMIC);
2426 if (!skb)
2427 goto errout;
2428
2429 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2430 if (err < 0)
2431 goto errout;
2432
2433 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2434 return;
2435
2436errout:
2437 kfree_skb(skb);
2438 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2439}
2440
2441static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2442{
2443 size_t len =
2444 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2445 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2446 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2447 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2448 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2449 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */
2450 /* IPMRA_CREPORT_PKT */
2451 + nla_total_size(payloadlen)
2452 ;
2453
2454 return len;
2455}
2456
2457static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2458{
2459 struct net *net = read_pnet(&mrt->net);
2460 struct nlmsghdr *nlh;
2461 struct rtgenmsg *rtgenm;
2462 struct igmpmsg *msg;
2463 struct sk_buff *skb;
2464 struct nlattr *nla;
2465 int payloadlen;
2466
2467 payloadlen = pkt->len - sizeof(struct igmpmsg);
2468 msg = (struct igmpmsg *)skb_network_header(pkt);
2469
2470 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2471 if (!skb)
2472 goto errout;
2473
2474 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2475 sizeof(struct rtgenmsg), 0);
2476 if (!nlh)
2477 goto errout;
2478 rtgenm = nlmsg_data(nlh);
2479 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2480 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2481 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2482 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2483 msg->im_src.s_addr) ||
2484 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2485 msg->im_dst.s_addr) ||
2486 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2487 goto nla_put_failure;
2488
2489 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2490 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2491 nla_data(nla), payloadlen))
2492 goto nla_put_failure;
2493
2494 nlmsg_end(skb, nlh);
2495
2496 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2497 return;
2498
2499nla_put_failure:
2500 nlmsg_cancel(skb, nlh);
2501errout:
2502 kfree_skb(skb);
2503 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2504}
2505
2506static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2507 const struct nlmsghdr *nlh,
2508 struct nlattr **tb,
2509 struct netlink_ext_ack *extack)
2510{
2511 struct rtmsg *rtm;
2512 int i, err;
2513
2514 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2515 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2516 return -EINVAL;
2517 }
2518
2519 if (!netlink_strict_get_check(skb))
2520 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2521 rtm_ipv4_policy, extack);
2522
2523 rtm = nlmsg_data(nlh);
2524 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2525 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2526 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2527 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2528 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2529 return -EINVAL;
2530 }
2531
2532 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2533 rtm_ipv4_policy, extack);
2534 if (err)
2535 return err;
2536
2537 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2538 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2539 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2540 return -EINVAL;
2541 }
2542
2543 for (i = 0; i <= RTA_MAX; i++) {
2544 if (!tb[i])
2545 continue;
2546
2547 switch (i) {
2548 case RTA_SRC:
2549 case RTA_DST:
2550 case RTA_TABLE:
2551 break;
2552 default:
2553 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2554 return -EINVAL;
2555 }
2556 }
2557
2558 return 0;
2559}
2560
2561static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2562 struct netlink_ext_ack *extack)
2563{
2564 struct net *net = sock_net(in_skb->sk);
2565 struct nlattr *tb[RTA_MAX + 1];
2566 struct sk_buff *skb = NULL;
2567 struct mfc_cache *cache;
2568 struct mr_table *mrt;
2569 __be32 src, grp;
2570 u32 tableid;
2571 int err;
2572
2573 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2574 if (err < 0)
2575 goto errout;
2576
2577 src = nla_get_in_addr_default(tb[RTA_SRC], 0);
2578 grp = nla_get_in_addr_default(tb[RTA_DST], 0);
2579 tableid = nla_get_u32_default(tb[RTA_TABLE], 0);
2580
2581 mrt = __ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2582 if (!mrt) {
2583 err = -ENOENT;
2584 goto errout_free;
2585 }
2586
2587 /* entries are added/deleted only under RTNL */
2588 rcu_read_lock();
2589 cache = ipmr_cache_find(mrt, src, grp);
2590 rcu_read_unlock();
2591 if (!cache) {
2592 err = -ENOENT;
2593 goto errout_free;
2594 }
2595
2596 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2597 if (!skb) {
2598 err = -ENOBUFS;
2599 goto errout_free;
2600 }
2601
2602 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2603 nlh->nlmsg_seq, cache,
2604 RTM_NEWROUTE, 0);
2605 if (err < 0)
2606 goto errout_free;
2607
2608 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2609
2610errout:
2611 return err;
2612
2613errout_free:
2614 kfree_skb(skb);
2615 goto errout;
2616}
2617
2618static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2619{
2620 struct fib_dump_filter filter = {
2621 .rtnl_held = true,
2622 };
2623 int err;
2624
2625 if (cb->strict_check) {
2626 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2627 &filter, cb);
2628 if (err < 0)
2629 return err;
2630 }
2631
2632 if (filter.table_id) {
2633 struct mr_table *mrt;
2634
2635 mrt = __ipmr_get_table(sock_net(skb->sk), filter.table_id);
2636 if (!mrt) {
2637 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2638 return skb->len;
2639
2640 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2641 return -ENOENT;
2642 }
2643 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2644 &mfc_unres_lock, &filter);
2645 return skb->len ? : err;
2646 }
2647
2648 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2649 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2650}
2651
2652static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2653 [RTA_SRC] = { .type = NLA_U32 },
2654 [RTA_DST] = { .type = NLA_U32 },
2655 [RTA_IIF] = { .type = NLA_U32 },
2656 [RTA_TABLE] = { .type = NLA_U32 },
2657 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2658};
2659
2660static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2661{
2662 switch (rtm_protocol) {
2663 case RTPROT_STATIC:
2664 case RTPROT_MROUTED:
2665 return true;
2666 }
2667 return false;
2668}
2669
2670static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2671{
2672 struct rtnexthop *rtnh = nla_data(nla);
2673 int remaining = nla_len(nla), vifi = 0;
2674
2675 while (rtnh_ok(rtnh, remaining)) {
2676 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2677 if (++vifi == MAXVIFS)
2678 break;
2679 rtnh = rtnh_next(rtnh, &remaining);
2680 }
2681
2682 return remaining > 0 ? -EINVAL : vifi;
2683}
2684
2685/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2686static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2687 struct mfcctl *mfcc, int *mrtsock,
2688 struct mr_table **mrtret,
2689 struct netlink_ext_ack *extack)
2690{
2691 struct net_device *dev = NULL;
2692 u32 tblid = RT_TABLE_DEFAULT;
2693 struct mr_table *mrt;
2694 struct nlattr *attr;
2695 struct rtmsg *rtm;
2696 int ret, rem;
2697
2698 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2699 rtm_ipmr_policy, extack);
2700 if (ret < 0)
2701 goto out;
2702 rtm = nlmsg_data(nlh);
2703
2704 ret = -EINVAL;
2705 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2706 rtm->rtm_type != RTN_MULTICAST ||
2707 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2708 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2709 goto out;
2710
2711 memset(mfcc, 0, sizeof(*mfcc));
2712 mfcc->mfcc_parent = -1;
2713 ret = 0;
2714 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2715 switch (nla_type(attr)) {
2716 case RTA_SRC:
2717 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2718 break;
2719 case RTA_DST:
2720 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2721 break;
2722 case RTA_IIF:
2723 dev = __dev_get_by_index(net, nla_get_u32(attr));
2724 if (!dev) {
2725 ret = -ENODEV;
2726 goto out;
2727 }
2728 break;
2729 case RTA_MULTIPATH:
2730 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2731 ret = -EINVAL;
2732 goto out;
2733 }
2734 break;
2735 case RTA_PREFSRC:
2736 ret = 1;
2737 break;
2738 case RTA_TABLE:
2739 tblid = nla_get_u32(attr);
2740 break;
2741 }
2742 }
2743 mrt = __ipmr_get_table(net, tblid);
2744 if (!mrt) {
2745 ret = -ENOENT;
2746 goto out;
2747 }
2748 *mrtret = mrt;
2749 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2750 if (dev)
2751 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2752
2753out:
2754 return ret;
2755}
2756
2757/* takes care of both newroute and delroute */
2758static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2759 struct netlink_ext_ack *extack)
2760{
2761 struct net *net = sock_net(skb->sk);
2762 int ret, mrtsock, parent;
2763 struct mr_table *tbl;
2764 struct mfcctl mfcc;
2765
2766 mrtsock = 0;
2767 tbl = NULL;
2768 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2769 if (ret < 0)
2770 return ret;
2771
2772 parent = ret ? mfcc.mfcc_parent : -1;
2773 if (nlh->nlmsg_type == RTM_NEWROUTE)
2774 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2775 else
2776 return ipmr_mfc_delete(tbl, &mfcc, parent);
2777}
2778
2779static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2780{
2781 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2782
2783 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2784 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2785 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2786 mrt->mroute_reg_vif_num) ||
2787 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2788 mrt->mroute_do_assert) ||
2789 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2790 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2791 mrt->mroute_do_wrvifwhole))
2792 return false;
2793
2794 return true;
2795}
2796
2797static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2798{
2799 struct net_device *vif_dev;
2800 struct nlattr *vif_nest;
2801 struct vif_device *vif;
2802
2803 vif = &mrt->vif_table[vifid];
2804 vif_dev = rtnl_dereference(vif->dev);
2805 /* if the VIF doesn't exist just continue */
2806 if (!vif_dev)
2807 return true;
2808
2809 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2810 if (!vif_nest)
2811 return false;
2812
2813 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2814 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2815 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2816 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2817 IPMRA_VIFA_PAD) ||
2818 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2819 IPMRA_VIFA_PAD) ||
2820 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2821 IPMRA_VIFA_PAD) ||
2822 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2823 IPMRA_VIFA_PAD) ||
2824 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2825 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2826 nla_nest_cancel(skb, vif_nest);
2827 return false;
2828 }
2829 nla_nest_end(skb, vif_nest);
2830
2831 return true;
2832}
2833
2834static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2835 struct netlink_ext_ack *extack)
2836{
2837 struct ifinfomsg *ifm;
2838
2839 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2840 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2841 return -EINVAL;
2842 }
2843
2844 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2845 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2846 return -EINVAL;
2847 }
2848
2849 ifm = nlmsg_data(nlh);
2850 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2851 ifm->ifi_change || ifm->ifi_index) {
2852 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2853 return -EINVAL;
2854 }
2855
2856 return 0;
2857}
2858
2859static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2860{
2861 struct net *net = sock_net(skb->sk);
2862 struct nlmsghdr *nlh = NULL;
2863 unsigned int t = 0, s_t;
2864 unsigned int e = 0, s_e;
2865 struct mr_table *mrt;
2866
2867 if (cb->strict_check) {
2868 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2869
2870 if (err < 0)
2871 return err;
2872 }
2873
2874 s_t = cb->args[0];
2875 s_e = cb->args[1];
2876
2877 ipmr_for_each_table(mrt, net) {
2878 struct nlattr *vifs, *af;
2879 struct ifinfomsg *hdr;
2880 u32 i;
2881
2882 if (t < s_t)
2883 goto skip_table;
2884 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2885 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2886 sizeof(*hdr), NLM_F_MULTI);
2887 if (!nlh)
2888 break;
2889
2890 hdr = nlmsg_data(nlh);
2891 memset(hdr, 0, sizeof(*hdr));
2892 hdr->ifi_family = RTNL_FAMILY_IPMR;
2893
2894 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2895 if (!af) {
2896 nlmsg_cancel(skb, nlh);
2897 goto out;
2898 }
2899
2900 if (!ipmr_fill_table(mrt, skb)) {
2901 nlmsg_cancel(skb, nlh);
2902 goto out;
2903 }
2904
2905 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2906 if (!vifs) {
2907 nla_nest_end(skb, af);
2908 nlmsg_end(skb, nlh);
2909 goto out;
2910 }
2911 for (i = 0; i < mrt->maxvif; i++) {
2912 if (e < s_e)
2913 goto skip_entry;
2914 if (!ipmr_fill_vif(mrt, i, skb)) {
2915 nla_nest_end(skb, vifs);
2916 nla_nest_end(skb, af);
2917 nlmsg_end(skb, nlh);
2918 goto out;
2919 }
2920skip_entry:
2921 e++;
2922 }
2923 s_e = 0;
2924 e = 0;
2925 nla_nest_end(skb, vifs);
2926 nla_nest_end(skb, af);
2927 nlmsg_end(skb, nlh);
2928skip_table:
2929 t++;
2930 }
2931
2932out:
2933 cb->args[1] = e;
2934 cb->args[0] = t;
2935
2936 return skb->len;
2937}
2938
2939#ifdef CONFIG_PROC_FS
2940/* The /proc interfaces to multicast routing :
2941 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2942 */
2943
2944static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2945 __acquires(RCU)
2946{
2947 struct mr_vif_iter *iter = seq->private;
2948 struct net *net = seq_file_net(seq);
2949 struct mr_table *mrt;
2950
2951 rcu_read_lock();
2952 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
2953 if (!mrt) {
2954 rcu_read_unlock();
2955 return ERR_PTR(-ENOENT);
2956 }
2957
2958 iter->mrt = mrt;
2959
2960 return mr_vif_seq_start(seq, pos);
2961}
2962
2963static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2964 __releases(RCU)
2965{
2966 rcu_read_unlock();
2967}
2968
2969static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2970{
2971 struct mr_vif_iter *iter = seq->private;
2972 struct mr_table *mrt = iter->mrt;
2973
2974 if (v == SEQ_START_TOKEN) {
2975 seq_puts(seq,
2976 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2977 } else {
2978 const struct vif_device *vif = v;
2979 const struct net_device *vif_dev;
2980 const char *name;
2981
2982 vif_dev = vif_dev_read(vif);
2983 name = vif_dev ? vif_dev->name : "none";
2984 seq_printf(seq,
2985 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2986 vif - mrt->vif_table,
2987 name, vif->bytes_in, vif->pkt_in,
2988 vif->bytes_out, vif->pkt_out,
2989 vif->flags, vif->local, vif->remote);
2990 }
2991 return 0;
2992}
2993
2994static const struct seq_operations ipmr_vif_seq_ops = {
2995 .start = ipmr_vif_seq_start,
2996 .next = mr_vif_seq_next,
2997 .stop = ipmr_vif_seq_stop,
2998 .show = ipmr_vif_seq_show,
2999};
3000
3001static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
3002{
3003 struct net *net = seq_file_net(seq);
3004 struct mr_table *mrt;
3005
3006 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
3007 if (!mrt)
3008 return ERR_PTR(-ENOENT);
3009
3010 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
3011}
3012
3013static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
3014{
3015 int n;
3016
3017 if (v == SEQ_START_TOKEN) {
3018 seq_puts(seq,
3019 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
3020 } else {
3021 const struct mfc_cache *mfc = v;
3022 const struct mr_mfc_iter *it = seq->private;
3023 const struct mr_table *mrt = it->mrt;
3024
3025 seq_printf(seq, "%08X %08X %-3hd",
3026 (__force u32) mfc->mfc_mcastgrp,
3027 (__force u32) mfc->mfc_origin,
3028 mfc->_c.mfc_parent);
3029
3030 if (it->cache != &mrt->mfc_unres_queue) {
3031 seq_printf(seq, " %8lu %8lu %8lu",
3032 atomic_long_read(&mfc->_c.mfc_un.res.pkt),
3033 atomic_long_read(&mfc->_c.mfc_un.res.bytes),
3034 atomic_long_read(&mfc->_c.mfc_un.res.wrong_if));
3035 for (n = mfc->_c.mfc_un.res.minvif;
3036 n < mfc->_c.mfc_un.res.maxvif; n++) {
3037 if (VIF_EXISTS(mrt, n) &&
3038 mfc->_c.mfc_un.res.ttls[n] < 255)
3039 seq_printf(seq,
3040 " %2d:%-3d",
3041 n, mfc->_c.mfc_un.res.ttls[n]);
3042 }
3043 } else {
3044 /* unresolved mfc_caches don't contain
3045 * pkt, bytes and wrong_if values
3046 */
3047 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3048 }
3049 seq_putc(seq, '\n');
3050 }
3051 return 0;
3052}
3053
3054static const struct seq_operations ipmr_mfc_seq_ops = {
3055 .start = ipmr_mfc_seq_start,
3056 .next = mr_mfc_seq_next,
3057 .stop = mr_mfc_seq_stop,
3058 .show = ipmr_mfc_seq_show,
3059};
3060#endif
3061
3062#ifdef CONFIG_IP_PIMSM_V2
3063static const struct net_protocol pim_protocol = {
3064 .handler = pim_rcv,
3065};
3066#endif
3067
3068static unsigned int ipmr_seq_read(const struct net *net)
3069{
3070 return READ_ONCE(net->ipv4.ipmr_seq) + ipmr_rules_seq_read(net);
3071}
3072
3073static int ipmr_dump(struct net *net, struct notifier_block *nb,
3074 struct netlink_ext_ack *extack)
3075{
3076 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3077 ipmr_mr_table_iter, extack);
3078}
3079
3080static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3081 .family = RTNL_FAMILY_IPMR,
3082 .fib_seq_read = ipmr_seq_read,
3083 .fib_dump = ipmr_dump,
3084 .owner = THIS_MODULE,
3085};
3086
3087static int __net_init ipmr_notifier_init(struct net *net)
3088{
3089 struct fib_notifier_ops *ops;
3090
3091 net->ipv4.ipmr_seq = 0;
3092
3093 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3094 if (IS_ERR(ops))
3095 return PTR_ERR(ops);
3096 net->ipv4.ipmr_notifier_ops = ops;
3097
3098 return 0;
3099}
3100
3101static void __net_exit ipmr_notifier_exit(struct net *net)
3102{
3103 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3104 net->ipv4.ipmr_notifier_ops = NULL;
3105}
3106
3107/* Setup for IP multicast routing */
3108static int __net_init ipmr_net_init(struct net *net)
3109{
3110 int err;
3111
3112 err = ipmr_notifier_init(net);
3113 if (err)
3114 goto ipmr_notifier_fail;
3115
3116 err = ipmr_rules_init(net);
3117 if (err < 0)
3118 goto ipmr_rules_fail;
3119
3120#ifdef CONFIG_PROC_FS
3121 err = -ENOMEM;
3122 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3123 sizeof(struct mr_vif_iter)))
3124 goto proc_vif_fail;
3125 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3126 sizeof(struct mr_mfc_iter)))
3127 goto proc_cache_fail;
3128#endif
3129 return 0;
3130
3131#ifdef CONFIG_PROC_FS
3132proc_cache_fail:
3133 remove_proc_entry("ip_mr_vif", net->proc_net);
3134proc_vif_fail:
3135 rtnl_lock();
3136 ipmr_rules_exit(net);
3137 rtnl_unlock();
3138#endif
3139ipmr_rules_fail:
3140 ipmr_notifier_exit(net);
3141ipmr_notifier_fail:
3142 return err;
3143}
3144
3145static void __net_exit ipmr_net_exit(struct net *net)
3146{
3147#ifdef CONFIG_PROC_FS
3148 remove_proc_entry("ip_mr_cache", net->proc_net);
3149 remove_proc_entry("ip_mr_vif", net->proc_net);
3150#endif
3151 ipmr_notifier_exit(net);
3152}
3153
3154static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3155{
3156 struct net *net;
3157
3158 rtnl_lock();
3159 list_for_each_entry(net, net_list, exit_list)
3160 ipmr_rules_exit(net);
3161 rtnl_unlock();
3162}
3163
3164static struct pernet_operations ipmr_net_ops = {
3165 .init = ipmr_net_init,
3166 .exit = ipmr_net_exit,
3167 .exit_batch = ipmr_net_exit_batch,
3168};
3169
3170static const struct rtnl_msg_handler ipmr_rtnl_msg_handlers[] __initconst = {
3171 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETLINK,
3172 .dumpit = ipmr_rtm_dumplink},
3173 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_NEWROUTE,
3174 .doit = ipmr_rtm_route},
3175 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_DELROUTE,
3176 .doit = ipmr_rtm_route},
3177 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETROUTE,
3178 .doit = ipmr_rtm_getroute, .dumpit = ipmr_rtm_dumproute},
3179};
3180
3181int __init ip_mr_init(void)
3182{
3183 int err;
3184
3185 mrt_cachep = KMEM_CACHE(mfc_cache, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3186
3187 err = register_pernet_subsys(&ipmr_net_ops);
3188 if (err)
3189 goto reg_pernet_fail;
3190
3191 err = register_netdevice_notifier(&ip_mr_notifier);
3192 if (err)
3193 goto reg_notif_fail;
3194#ifdef CONFIG_IP_PIMSM_V2
3195 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3196 pr_err("%s: can't add PIM protocol\n", __func__);
3197 err = -EAGAIN;
3198 goto add_proto_fail;
3199 }
3200#endif
3201 rtnl_register_many(ipmr_rtnl_msg_handlers);
3202
3203 return 0;
3204
3205#ifdef CONFIG_IP_PIMSM_V2
3206add_proto_fail:
3207 unregister_netdevice_notifier(&ip_mr_notifier);
3208#endif
3209reg_notif_fail:
3210 unregister_pernet_subsys(&ipmr_net_ops);
3211reg_pernet_fail:
3212 kmem_cache_destroy(mrt_cachep);
3213 return err;
3214}