projects/linux·blob master · refs

3409 lines · 82.4 KiB

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, <[email protected]>
6 * Linux Consultancy and Custom Driver Development
7 *
8 * Fixes:
9 * Michael Chastain : Incorrect size of copying.
10 * Alan Cox : Added the cache manager code
11 * Alan Cox : Fixed the clone/copy bug and device race.
12 * Mike McLagan : Routing by source
13 * Malcolm Beattie : Buffer handling fixes.
14 * Alexey Kuznetsov : Double buffer free and other fixes.
15 * SVR Anand : Fixed several multicast bugs and problems.
16 * Alexey Kuznetsov : Status, optimisations and more.
17 * Brad Parker : Better behaviour on mrouted upcall
18 * overflow.
19 * Carlos Picoto : PIMv1 Support
20 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
21 * Relax this requirement to work with older peers.
22 */
23
24#include <linux/uaccess.h>
25#include <linux/types.h>
26#include <linux/cache.h>
27#include <linux/capability.h>
28#include <linux/errno.h>
29#include <linux/mm.h>
30#include <linux/kernel.h>
31#include <linux/fcntl.h>
32#include <linux/stat.h>
33#include <linux/socket.h>
34#include <linux/in.h>
35#include <linux/inet.h>
36#include <linux/netdevice.h>
37#include <linux/inetdevice.h>
38#include <linux/igmp.h>
39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
41#include <linux/mroute.h>
42#include <linux/init.h>
43#include <linux/if_ether.h>
44#include <linux/slab.h>
45#include <net/flow.h>
46#include <net/net_namespace.h>
47#include <net/ip.h>
48#include <net/protocol.h>
49#include <linux/skbuff.h>
50#include <net/route.h>
51#include <net/icmp.h>
52#include <net/udp.h>
53#include <net/raw.h>
54#include <linux/notifier.h>
55#include <linux/if_arp.h>
56#include <linux/netfilter_ipv4.h>
57#include <linux/compat.h>
58#include <linux/export.h>
59#include <linux/rhashtable.h>
60#include <net/ip_tunnels.h>
61#include <net/checksum.h>
62#include <net/netlink.h>
63#include <net/fib_rules.h>
64#include <linux/netconf.h>
65#include <net/rtnh.h>
66#include <net/inet_dscp.h>
67
68#include <linux/nospec.h>
69
70struct ipmr_rule {
71 struct fib_rule common;
72};
73
74struct ipmr_result {
75 struct mr_table *mrt;
76};
77
78/* Big lock, protecting vif table, mrt cache and mroute socket state.
79 * Note that the changes are semaphored via rtnl_lock.
80 */
81
82static DEFINE_SPINLOCK(mrt_lock);
83
84static struct net_device *vif_dev_read(const struct vif_device *vif)
85{
86 return rcu_dereference(vif->dev);
87}
88
89/* Multicast router control variables */
90
91/* Special spinlock for queue of unresolved entries */
92static DEFINE_SPINLOCK(mfc_unres_lock);
93
94/* We return to original Alan's scheme. Hash table of resolved
95 * entries is changed only in process context and protected
96 * with weak lock mrt_lock. Queue of unresolved entries is protected
97 * with strong spinlock mfc_unres_lock.
98 *
99 * In this case data path is free of exclusive locks at all.
100 */
101
102static struct kmem_cache *mrt_cachep __ro_after_init;
103
104static struct mr_table *ipmr_new_table(struct net *net, u32 id);
105static void ipmr_free_table(struct mr_table *mrt,
106 struct list_head *dev_kill_list);
107
108static void ip_mr_forward(struct net *net, struct mr_table *mrt,
109 struct net_device *dev, struct sk_buff *skb,
110 struct mfc_cache *cache, int local);
111static int ipmr_cache_report(const struct mr_table *mrt,
112 struct sk_buff *pkt, vifi_t vifi, int assert);
113static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
114 int cmd);
115static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
116static void mroute_clean_tables(struct mr_table *mrt, int flags,
117 struct list_head *dev_kill_list);
118static void ipmr_expire_process(struct timer_list *t);
119
120#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
121#define ipmr_for_each_table(mrt, net) \
122 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
123 lockdep_rtnl_is_held() || \
124 list_empty(&net->ipv4.mr_tables))
125
126static struct mr_table *ipmr_mr_table_iter(struct net *net,
127 struct mr_table *mrt)
128{
129 struct mr_table *ret;
130
131 if (!mrt)
132 ret = list_entry_rcu(net->ipv4.mr_tables.next,
133 struct mr_table, list);
134 else
135 ret = list_entry_rcu(mrt->list.next,
136 struct mr_table, list);
137
138 if (&ret->list == &net->ipv4.mr_tables)
139 return NULL;
140 return ret;
141}
142
143static struct mr_table *__ipmr_get_table(struct net *net, u32 id)
144{
145 struct mr_table *mrt;
146
147 ipmr_for_each_table(mrt, net) {
148 if (mrt->id == id)
149 return mrt;
150 }
151 return NULL;
152}
153
154static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
155 struct mr_table **mrt)
156{
157 int err;
158 struct ipmr_result res;
159 struct fib_lookup_arg arg = {
160 .result = &res,
161 .flags = FIB_LOOKUP_NOREF,
162 };
163
164 /* update flow if oif or iif point to device enslaved to l3mdev */
165 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
166
167 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
168 flowi4_to_flowi(flp4), 0, &arg);
169 if (err < 0)
170 return err;
171 *mrt = res.mrt;
172 return 0;
173}
174
175static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
176 int flags, struct fib_lookup_arg *arg)
177{
178 struct ipmr_result *res = arg->result;
179 struct mr_table *mrt;
180
181 switch (rule->action) {
182 case FR_ACT_TO_TBL:
183 break;
184 case FR_ACT_UNREACHABLE:
185 return -ENETUNREACH;
186 case FR_ACT_PROHIBIT:
187 return -EACCES;
188 case FR_ACT_BLACKHOLE:
189 default:
190 return -EINVAL;
191 }
192
193 arg->table = fib_rule_get_table(rule, arg);
194
195 mrt = __ipmr_get_table(rule->fr_net, arg->table);
196 if (!mrt)
197 return -EAGAIN;
198 res->mrt = mrt;
199 return 0;
200}
201
202static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
203{
204 return 1;
205}
206
207static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
208 struct fib_rule_hdr *frh, struct nlattr **tb,
209 struct netlink_ext_ack *extack)
210{
211 return 0;
212}
213
214static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
215 struct nlattr **tb)
216{
217 return 1;
218}
219
220static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
221 struct fib_rule_hdr *frh)
222{
223 frh->dst_len = 0;
224 frh->src_len = 0;
225 frh->tos = 0;
226 return 0;
227}
228
229static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
230 .family = RTNL_FAMILY_IPMR,
231 .rule_size = sizeof(struct ipmr_rule),
232 .addr_size = sizeof(u32),
233 .action = ipmr_rule_action,
234 .match = ipmr_rule_match,
235 .configure = ipmr_rule_configure,
236 .compare = ipmr_rule_compare,
237 .fill = ipmr_rule_fill,
238 .nlgroup = RTNLGRP_IPV4_RULE,
239 .owner = THIS_MODULE,
240};
241
242static int __net_init ipmr_rules_init(struct net *net)
243{
244 struct fib_rules_ops *ops;
245 LIST_HEAD(dev_kill_list);
246 struct mr_table *mrt;
247 int err;
248
249 ops = fib_rules_register(&ipmr_rules_ops_template, net);
250 if (IS_ERR(ops))
251 return PTR_ERR(ops);
252
253 INIT_LIST_HEAD(&net->ipv4.mr_tables);
254
255 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
256 if (IS_ERR(mrt)) {
257 err = PTR_ERR(mrt);
258 goto err1;
259 }
260
261 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT);
262 if (err < 0)
263 goto err2;
264
265 net->ipv4.mr_rules_ops = ops;
266 return 0;
267
268err2:
269 ipmr_free_table(mrt, &dev_kill_list);
270err1:
271 fib_rules_unregister(ops);
272 return err;
273}
274
275static void __net_exit ipmr_rules_exit(struct net *net)
276{
277 fib_rules_unregister(net->ipv4.mr_rules_ops);
278}
279
280static void __net_exit ipmr_rules_exit_rtnl(struct net *net,
281 struct list_head *dev_kill_list)
282{
283 struct mr_table *mrt, *next;
284
285 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
286 list_del_rcu(&mrt->list);
287 ipmr_free_table(mrt, dev_kill_list);
288 }
289}
290
291static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
292 struct netlink_ext_ack *extack)
293{
294 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
295}
296
297static unsigned int ipmr_rules_seq_read(const struct net *net)
298{
299 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
300}
301
302bool ipmr_rule_default(const struct fib_rule *rule)
303{
304 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
305}
306EXPORT_SYMBOL(ipmr_rule_default);
307#else
308static struct mr_table *ipmr_mr_table_iter(struct net *net,
309 struct mr_table *mrt)
310{
311 if (!mrt)
312 return rcu_dereference(net->ipv4.mrt);
313 return NULL;
314}
315
316static struct mr_table *__ipmr_get_table(struct net *net, u32 id)
317{
318 return rcu_dereference_check(net->ipv4.mrt,
319 lockdep_rtnl_is_held() ||
320 !rcu_access_pointer(net->ipv4.mrt));
321}
322
323#define ipmr_for_each_table(mrt, net) \
324 for (mrt = __ipmr_get_table(net, 0); mrt; mrt = NULL)
325
326static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
327 struct mr_table **mrt)
328{
329 *mrt = rcu_dereference(net->ipv4.mrt);
330 if (!*mrt)
331 return -EAGAIN;
332 return 0;
333}
334
335static int __net_init ipmr_rules_init(struct net *net)
336{
337 struct mr_table *mrt;
338
339 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
340 if (IS_ERR(mrt))
341 return PTR_ERR(mrt);
342
343 rcu_assign_pointer(net->ipv4.mrt, mrt);
344 return 0;
345}
346
347static void __net_exit ipmr_rules_exit(struct net *net)
348{
349}
350
351static void __net_exit ipmr_rules_exit_rtnl(struct net *net,
352 struct list_head *dev_kill_list)
353{
354 struct mr_table *mrt = rcu_dereference_protected(net->ipv4.mrt, 1);
355
356 RCU_INIT_POINTER(net->ipv4.mrt, NULL);
357 ipmr_free_table(mrt, dev_kill_list);
358}
359
360static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
361 struct netlink_ext_ack *extack)
362{
363 return 0;
364}
365
366static unsigned int ipmr_rules_seq_read(const struct net *net)
367{
368 return 0;
369}
370
371bool ipmr_rule_default(const struct fib_rule *rule)
372{
373 return true;
374}
375EXPORT_SYMBOL(ipmr_rule_default);
376#endif
377
378static struct mr_table *ipmr_get_table(struct net *net, u32 id)
379{
380 struct mr_table *mrt;
381
382 rcu_read_lock();
383 mrt = __ipmr_get_table(net, id);
384 rcu_read_unlock();
385
386 return mrt;
387}
388
389static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
390 const void *ptr)
391{
392 const struct mfc_cache_cmp_arg *cmparg = arg->key;
393 const struct mfc_cache *c = ptr;
394
395 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
396 cmparg->mfc_origin != c->mfc_origin;
397}
398
399static const struct rhashtable_params ipmr_rht_params = {
400 .head_offset = offsetof(struct mr_mfc, mnode),
401 .key_offset = offsetof(struct mfc_cache, cmparg),
402 .key_len = sizeof(struct mfc_cache_cmp_arg),
403 .nelem_hint = 3,
404 .obj_cmpfn = ipmr_hash_cmp,
405 .automatic_shrinking = true,
406};
407
408static void ipmr_new_table_set(struct mr_table *mrt,
409 struct net *net)
410{
411#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
412 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
413#endif
414}
415
416static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
417 .mfc_mcastgrp = htonl(INADDR_ANY),
418 .mfc_origin = htonl(INADDR_ANY),
419};
420
421static struct mr_table_ops ipmr_mr_table_ops = {
422 .rht_params = &ipmr_rht_params,
423 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
424};
425
426static struct mr_table *ipmr_new_table(struct net *net, u32 id)
427{
428 struct mr_table *mrt;
429
430 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
431 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
432 return ERR_PTR(-EINVAL);
433
434 mrt = __ipmr_get_table(net, id);
435 if (mrt)
436 return mrt;
437
438 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
439 ipmr_expire_process, ipmr_new_table_set);
440}
441
442static void ipmr_free_table(struct mr_table *mrt, struct list_head *dev_kill_list)
443{
444 struct net *net = read_pnet(&mrt->net);
445 LIST_HEAD(ipmr_dev_kill_list);
446
447 WARN_ON_ONCE(!mr_can_free_table(net));
448
449 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
450 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC,
451 &ipmr_dev_kill_list);
452 timer_shutdown_sync(&mrt->ipmr_expire_timer);
453 mr_table_free(mrt);
454
455 WARN_ON_ONCE(!net_initialized(net) && !list_empty(&ipmr_dev_kill_list));
456 list_splice(&ipmr_dev_kill_list, dev_kill_list);
457}
458
459/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
460
461/* Initialize ipmr pimreg/tunnel in_device */
462static bool ipmr_init_vif_indev(const struct net_device *dev)
463{
464 struct in_device *in_dev;
465
466 ASSERT_RTNL();
467
468 in_dev = __in_dev_get_rtnl(dev);
469 if (!in_dev)
470 return false;
471 ipv4_devconf_setall(in_dev);
472 neigh_parms_data_state_setall(in_dev->arp_parms);
473 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
474
475 return true;
476}
477
478static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
479{
480 struct net_device *tunnel_dev, *new_dev;
481 struct ip_tunnel_parm_kern p = { };
482 int err;
483
484 tunnel_dev = __dev_get_by_name(net, "tunl0");
485 if (!tunnel_dev)
486 goto out;
487
488 p.iph.daddr = v->vifc_rmt_addr.s_addr;
489 p.iph.saddr = v->vifc_lcl_addr.s_addr;
490 p.iph.version = 4;
491 p.iph.ihl = 5;
492 p.iph.protocol = IPPROTO_IPIP;
493 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
494
495 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
496 goto out;
497 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
498 SIOCADDTUNNEL);
499 if (err)
500 goto out;
501
502 new_dev = __dev_get_by_name(net, p.name);
503 if (!new_dev)
504 goto out;
505
506 new_dev->flags |= IFF_MULTICAST;
507 if (!ipmr_init_vif_indev(new_dev))
508 goto out_unregister;
509 if (dev_open(new_dev, NULL))
510 goto out_unregister;
511 dev_hold(new_dev);
512 err = dev_set_allmulti(new_dev, 1);
513 if (err) {
514 dev_close(new_dev);
515 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
516 SIOCDELTUNNEL);
517 dev_put(new_dev);
518 new_dev = ERR_PTR(err);
519 }
520 return new_dev;
521
522out_unregister:
523 unregister_netdevice(new_dev);
524out:
525 return ERR_PTR(-ENOBUFS);
526}
527
528#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
529static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
530{
531 struct net *net = dev_net(dev);
532 struct mr_table *mrt;
533 struct flowi4 fl4 = {
534 .flowi4_oif = dev->ifindex,
535 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
536 .flowi4_mark = skb->mark,
537 };
538 int err;
539
540 rcu_read_lock();
541 err = ipmr_fib_lookup(net, &fl4, &mrt);
542 if (err < 0) {
543 rcu_read_unlock();
544 kfree_skb(skb);
545 return err;
546 }
547
548 DEV_STATS_ADD(dev, tx_bytes, skb->len);
549 DEV_STATS_INC(dev, tx_packets);
550
551 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
552 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
553 IGMPMSG_WHOLEPKT);
554
555 rcu_read_unlock();
556 kfree_skb(skb);
557 return NETDEV_TX_OK;
558}
559
560static int reg_vif_get_iflink(const struct net_device *dev)
561{
562 return 0;
563}
564
565static const struct net_device_ops reg_vif_netdev_ops = {
566 .ndo_start_xmit = reg_vif_xmit,
567 .ndo_get_iflink = reg_vif_get_iflink,
568};
569
570static void reg_vif_setup(struct net_device *dev)
571{
572 dev->type = ARPHRD_PIMREG;
573 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
574 dev->flags = IFF_NOARP;
575 dev->netdev_ops = &reg_vif_netdev_ops;
576 dev->needs_free_netdev = true;
577 dev->netns_immutable = true;
578}
579
580static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
581{
582 struct net_device *dev;
583 char name[IFNAMSIZ];
584
585 if (mrt->id == RT_TABLE_DEFAULT)
586 sprintf(name, "pimreg");
587 else
588 sprintf(name, "pimreg%u", mrt->id);
589
590 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
591
592 if (!dev)
593 return NULL;
594
595 dev_net_set(dev, net);
596
597 if (register_netdevice(dev)) {
598 free_netdev(dev);
599 return NULL;
600 }
601
602 if (!ipmr_init_vif_indev(dev))
603 goto failure;
604 if (dev_open(dev, NULL))
605 goto failure;
606
607 dev_hold(dev);
608
609 return dev;
610
611failure:
612 unregister_netdevice(dev);
613 return NULL;
614}
615
616/* called with rcu_read_lock() */
617static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
618 unsigned int pimlen)
619{
620 struct net_device *reg_dev = NULL;
621 struct iphdr *encap;
622 int vif_num;
623
624 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
625 /* Check that:
626 * a. packet is really sent to a multicast group
627 * b. packet is not a NULL-REGISTER
628 * c. packet is not truncated
629 */
630 if (!ipv4_is_multicast(encap->daddr) ||
631 encap->tot_len == 0 ||
632 ntohs(encap->tot_len) + pimlen > skb->len)
633 return 1;
634
635 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
636 vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
637 if (vif_num >= 0)
638 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
639 if (!reg_dev)
640 return 1;
641
642 skb->mac_header = skb->network_header;
643 skb_pull(skb, (u8 *)encap - skb->data);
644 skb_reset_network_header(skb);
645 skb->protocol = htons(ETH_P_IP);
646 skb->ip_summed = CHECKSUM_NONE;
647
648 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
649
650 netif_rx(skb);
651
652 return NET_RX_SUCCESS;
653}
654#else
655static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
656{
657 return NULL;
658}
659#endif
660
661static int call_ipmr_vif_entry_notifiers(struct net *net,
662 enum fib_event_type event_type,
663 struct vif_device *vif,
664 struct net_device *vif_dev,
665 vifi_t vif_index, u32 tb_id)
666{
667 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
668 vif, vif_dev, vif_index, tb_id,
669 &net->ipv4.ipmr_seq);
670}
671
672static int call_ipmr_mfc_entry_notifiers(struct net *net,
673 enum fib_event_type event_type,
674 struct mfc_cache *mfc, u32 tb_id)
675{
676 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
677 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
678}
679
680/**
681 * vif_delete - Delete a VIF entry
682 * @mrt: Table to delete from
683 * @vifi: VIF identifier to delete
684 * @notify: Set to 1, if the caller is a notifier_call
685 * @head: if unregistering the VIF, place it on this queue
686 */
687static int vif_delete(struct mr_table *mrt, int vifi, int notify,
688 struct list_head *head)
689{
690 struct net *net = read_pnet(&mrt->net);
691 struct vif_device *v;
692 struct net_device *dev;
693 struct in_device *in_dev;
694
695 if (vifi < 0 || vifi >= mrt->maxvif)
696 return -EADDRNOTAVAIL;
697
698 v = &mrt->vif_table[vifi];
699
700 dev = rtnl_dereference(v->dev);
701 if (!dev)
702 return -EADDRNOTAVAIL;
703
704 spin_lock(&mrt_lock);
705 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
706 vifi, mrt->id);
707 RCU_INIT_POINTER(v->dev, NULL);
708
709 if (vifi == mrt->mroute_reg_vif_num) {
710 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
711 WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
712 }
713 if (vifi + 1 == mrt->maxvif) {
714 int tmp;
715
716 for (tmp = vifi - 1; tmp >= 0; tmp--) {
717 if (VIF_EXISTS(mrt, tmp))
718 break;
719 }
720 WRITE_ONCE(mrt->maxvif, tmp + 1);
721 }
722
723 spin_unlock(&mrt_lock);
724
725 dev_set_allmulti(dev, -1);
726
727 in_dev = __in_dev_get_rtnl(dev);
728 if (in_dev) {
729 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
730 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
731 NETCONFA_MC_FORWARDING,
732 dev->ifindex, &in_dev->cnf);
733 ip_rt_multicast_event(in_dev);
734 }
735
736 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
737 unregister_netdevice_queue(dev, head);
738
739 netdev_put(dev, &v->dev_tracker);
740 return 0;
741}
742
743static void ipmr_cache_free_rcu(struct rcu_head *head)
744{
745 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
746
747 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
748}
749
750static void ipmr_cache_free(struct mfc_cache *c)
751{
752 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
753}
754
755/* Destroy an unresolved cache entry, killing queued skbs
756 * and reporting error to netlink readers.
757 */
758static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
759{
760 struct net *net = read_pnet(&mrt->net);
761 struct sk_buff *skb;
762 struct nlmsgerr *e;
763
764 WRITE_ONCE(mrt->cache_resolve_queue_len,
765 mrt->cache_resolve_queue_len - 1);
766
767 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
768 if (ip_hdr(skb)->version == 0) {
769 struct nlmsghdr *nlh = skb_pull(skb,
770 sizeof(struct iphdr));
771 nlh->nlmsg_type = NLMSG_ERROR;
772 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
773 skb_trim(skb, nlh->nlmsg_len);
774 e = nlmsg_data(nlh);
775 e->error = -ETIMEDOUT;
776 memset(&e->msg, 0, sizeof(e->msg));
777
778 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
779 } else {
780 kfree_skb(skb);
781 }
782 }
783
784 ipmr_cache_free(c);
785}
786
787/* Timer process for the unresolved queue. */
788static void ipmr_expire_process(struct timer_list *t)
789{
790 struct mr_table *mrt = timer_container_of(mrt, t, ipmr_expire_timer);
791 struct mr_mfc *c, *next;
792 unsigned long expires;
793 unsigned long now;
794
795 if (!spin_trylock(&mfc_unres_lock)) {
796 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
797 return;
798 }
799
800 if (list_empty(&mrt->mfc_unres_queue))
801 goto out;
802
803 now = jiffies;
804 expires = 10*HZ;
805
806 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
807 if (time_after(c->mfc_un.unres.expires, now)) {
808 unsigned long interval = c->mfc_un.unres.expires - now;
809 if (interval < expires)
810 expires = interval;
811 continue;
812 }
813
814 list_del(&c->list);
815 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
816 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
817 }
818
819 if (!list_empty(&mrt->mfc_unres_queue))
820 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
821
822out:
823 spin_unlock(&mfc_unres_lock);
824}
825
826/* Fill oifs list. It is called under locked mrt_lock. */
827static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
828 unsigned char *ttls)
829{
830 int vifi;
831
832 cache->mfc_un.res.minvif = MAXVIFS;
833 cache->mfc_un.res.maxvif = 0;
834 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
835
836 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
837 if (VIF_EXISTS(mrt, vifi) &&
838 ttls[vifi] && ttls[vifi] < 255) {
839 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
840 if (cache->mfc_un.res.minvif > vifi)
841 cache->mfc_un.res.minvif = vifi;
842 if (cache->mfc_un.res.maxvif <= vifi)
843 cache->mfc_un.res.maxvif = vifi + 1;
844 }
845 }
846 WRITE_ONCE(cache->mfc_un.res.lastuse, jiffies);
847}
848
849static int vif_add(struct net *net, struct mr_table *mrt,
850 struct vifctl *vifc, int mrtsock)
851{
852 struct netdev_phys_item_id ppid = { };
853 int vifi = vifc->vifc_vifi;
854 struct vif_device *v = &mrt->vif_table[vifi];
855 struct net_device *dev;
856 struct in_device *in_dev;
857 int err;
858
859 /* Is vif busy ? */
860 if (VIF_EXISTS(mrt, vifi))
861 return -EADDRINUSE;
862
863 switch (vifc->vifc_flags) {
864 case VIFF_REGISTER:
865 if (!ipmr_pimsm_enabled())
866 return -EINVAL;
867 /* Special Purpose VIF in PIM
868 * All the packets will be sent to the daemon
869 */
870 if (mrt->mroute_reg_vif_num >= 0)
871 return -EADDRINUSE;
872 dev = ipmr_reg_vif(net, mrt);
873 if (!dev)
874 return -ENOBUFS;
875 err = dev_set_allmulti(dev, 1);
876 if (err) {
877 unregister_netdevice(dev);
878 dev_put(dev);
879 return err;
880 }
881 break;
882 case VIFF_TUNNEL:
883 dev = ipmr_new_tunnel(net, vifc);
884 if (IS_ERR(dev))
885 return PTR_ERR(dev);
886 break;
887 case VIFF_USE_IFINDEX:
888 case 0:
889 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
890 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
891 if (dev && !__in_dev_get_rtnl(dev)) {
892 dev_put(dev);
893 return -EADDRNOTAVAIL;
894 }
895 } else {
896 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
897 }
898 if (!dev)
899 return -EADDRNOTAVAIL;
900 err = dev_set_allmulti(dev, 1);
901 if (err) {
902 dev_put(dev);
903 return err;
904 }
905 break;
906 default:
907 return -EINVAL;
908 }
909
910 in_dev = __in_dev_get_rtnl(dev);
911 if (!in_dev) {
912 dev_put(dev);
913 return -EADDRNOTAVAIL;
914 }
915 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
916 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
917 dev->ifindex, &in_dev->cnf);
918 ip_rt_multicast_event(in_dev);
919
920 /* Fill in the VIF structures */
921 vif_device_init(v, dev, vifc->vifc_rate_limit,
922 vifc->vifc_threshold,
923 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
924 (VIFF_TUNNEL | VIFF_REGISTER));
925
926 err = netif_get_port_parent_id(dev, &ppid, true);
927 if (err == 0) {
928 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
929 v->dev_parent_id.id_len = ppid.id_len;
930 } else {
931 v->dev_parent_id.id_len = 0;
932 }
933
934 v->local = vifc->vifc_lcl_addr.s_addr;
935 v->remote = vifc->vifc_rmt_addr.s_addr;
936
937 /* And finish update writing critical data */
938 spin_lock(&mrt_lock);
939 rcu_assign_pointer(v->dev, dev);
940 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
941 if (v->flags & VIFF_REGISTER) {
942 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
943 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
944 }
945 if (vifi+1 > mrt->maxvif)
946 WRITE_ONCE(mrt->maxvif, vifi + 1);
947 spin_unlock(&mrt_lock);
948 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
949 vifi, mrt->id);
950 return 0;
951}
952
953/* called with rcu_read_lock() */
954static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
955 __be32 origin,
956 __be32 mcastgrp)
957{
958 struct mfc_cache_cmp_arg arg = {
959 .mfc_mcastgrp = mcastgrp,
960 .mfc_origin = origin
961 };
962
963 return mr_mfc_find(mrt, &arg);
964}
965
966/* Look for a (*,G) entry */
967static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
968 __be32 mcastgrp, int vifi)
969{
970 struct mfc_cache_cmp_arg arg = {
971 .mfc_mcastgrp = mcastgrp,
972 .mfc_origin = htonl(INADDR_ANY)
973 };
974
975 if (mcastgrp == htonl(INADDR_ANY))
976 return mr_mfc_find_any_parent(mrt, vifi);
977 return mr_mfc_find_any(mrt, vifi, &arg);
978}
979
980/* Look for a (S,G,iif) entry if parent != -1 */
981static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
982 __be32 origin, __be32 mcastgrp,
983 int parent)
984{
985 struct mfc_cache_cmp_arg arg = {
986 .mfc_mcastgrp = mcastgrp,
987 .mfc_origin = origin,
988 };
989
990 return mr_mfc_find_parent(mrt, &arg, parent);
991}
992
993/* Allocate a multicast cache entry */
994static struct mfc_cache *ipmr_cache_alloc(void)
995{
996 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
997
998 if (c) {
999 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1000 c->_c.mfc_un.res.minvif = MAXVIFS;
1001 c->_c.free = ipmr_cache_free_rcu;
1002 refcount_set(&c->_c.mfc_un.res.refcount, 1);
1003 }
1004 return c;
1005}
1006
1007static struct mfc_cache *ipmr_cache_alloc_unres(void)
1008{
1009 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1010
1011 if (c) {
1012 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1013 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1014 }
1015 return c;
1016}
1017
1018/* A cache entry has gone into a resolved state from queued */
1019static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1020 struct mfc_cache *uc, struct mfc_cache *c)
1021{
1022 struct sk_buff *skb;
1023 struct nlmsgerr *e;
1024
1025 /* Play the pending entries through our router */
1026 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1027 if (ip_hdr(skb)->version == 0) {
1028 struct nlmsghdr *nlh = skb_pull(skb,
1029 sizeof(struct iphdr));
1030
1031 if (mr_fill_mroute(mrt, skb, &c->_c,
1032 nlmsg_data(nlh)) > 0) {
1033 nlh->nlmsg_len = skb_tail_pointer(skb) -
1034 (u8 *)nlh;
1035 } else {
1036 nlh->nlmsg_type = NLMSG_ERROR;
1037 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1038 skb_trim(skb, nlh->nlmsg_len);
1039 e = nlmsg_data(nlh);
1040 e->error = -EMSGSIZE;
1041 memset(&e->msg, 0, sizeof(e->msg));
1042 }
1043
1044 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1045 } else {
1046 rcu_read_lock();
1047 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1048 rcu_read_unlock();
1049 }
1050 }
1051}
1052
1053/* Bounce a cache query up to mrouted and netlink.
1054 *
1055 * Called under rcu_read_lock().
1056 */
1057static int ipmr_cache_report(const struct mr_table *mrt,
1058 struct sk_buff *pkt, vifi_t vifi, int assert)
1059{
1060 const int ihl = ip_hdrlen(pkt);
1061 struct sock *mroute_sk;
1062 struct igmphdr *igmp;
1063 struct igmpmsg *msg;
1064 struct sk_buff *skb;
1065 int ret;
1066
1067 mroute_sk = rcu_dereference(mrt->mroute_sk);
1068 if (!mroute_sk)
1069 return -EINVAL;
1070
1071 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1072 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1073 else
1074 skb = alloc_skb(128, GFP_ATOMIC);
1075
1076 if (!skb)
1077 return -ENOBUFS;
1078
1079 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1080 /* Ugly, but we have no choice with this interface.
1081 * Duplicate old header, fix ihl, length etc.
1082 * And all this only to mangle msg->im_msgtype and
1083 * to set msg->im_mbz to "mbz" :-)
1084 */
1085 skb_push(skb, sizeof(struct iphdr));
1086 skb_reset_network_header(skb);
1087 skb_reset_transport_header(skb);
1088 msg = (struct igmpmsg *)skb_network_header(skb);
1089 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1090 msg->im_msgtype = assert;
1091 msg->im_mbz = 0;
1092 if (assert == IGMPMSG_WRVIFWHOLE) {
1093 msg->im_vif = vifi;
1094 msg->im_vif_hi = vifi >> 8;
1095 } else {
1096 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1097 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1098
1099 msg->im_vif = vif_num;
1100 msg->im_vif_hi = vif_num >> 8;
1101 }
1102 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1103 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1104 sizeof(struct iphdr));
1105 } else {
1106 /* Copy the IP header */
1107 skb_set_network_header(skb, skb->len);
1108 skb_put(skb, ihl);
1109 skb_copy_to_linear_data(skb, pkt->data, ihl);
1110 /* Flag to the kernel this is a route add */
1111 ip_hdr(skb)->protocol = 0;
1112 msg = (struct igmpmsg *)skb_network_header(skb);
1113 msg->im_vif = vifi;
1114 msg->im_vif_hi = vifi >> 8;
1115 ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1116 memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1117 /* Add our header.
1118 * Note that code, csum and group fields are cleared.
1119 */
1120 igmp = skb_put_zero(skb, sizeof(struct igmphdr));
1121 igmp->type = assert;
1122 msg->im_msgtype = assert;
1123 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1124 skb->transport_header = skb->network_header;
1125 }
1126
1127 igmpmsg_netlink_event(mrt, skb);
1128
1129 /* Deliver to mrouted */
1130 ret = sock_queue_rcv_skb(mroute_sk, skb);
1131
1132 if (ret < 0) {
1133 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1134 kfree_skb(skb);
1135 }
1136
1137 return ret;
1138}
1139
1140/* Queue a packet for resolution. It gets locked cache entry! */
1141/* Called under rcu_read_lock() */
1142static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1143 struct sk_buff *skb, struct net_device *dev)
1144{
1145 struct net *net = read_pnet(&mrt->net);
1146 const struct iphdr *iph = ip_hdr(skb);
1147 struct mfc_cache *c = NULL;
1148 bool found = false;
1149 int err;
1150
1151 spin_lock_bh(&mfc_unres_lock);
1152
1153 if (!check_net(net)) {
1154 err = -EINVAL;
1155 goto err;
1156 }
1157
1158 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1159 if (c->mfc_mcastgrp == iph->daddr &&
1160 c->mfc_origin == iph->saddr) {
1161 found = true;
1162 break;
1163 }
1164 }
1165
1166 if (!found) {
1167 /* Create a new entry if allowable */
1168 c = ipmr_cache_alloc_unres();
1169 if (!c) {
1170 err = -ENOBUFS;
1171 goto err;
1172 }
1173
1174 /* Fill in the new cache entry */
1175 c->_c.mfc_parent = -1;
1176 c->mfc_origin = iph->saddr;
1177 c->mfc_mcastgrp = iph->daddr;
1178
1179 /* Reflect first query at mrouted. */
1180 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1181 if (err < 0)
1182 goto err;
1183
1184 WRITE_ONCE(mrt->cache_resolve_queue_len,
1185 mrt->cache_resolve_queue_len + 1);
1186 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1187 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1188
1189 if (mrt->cache_resolve_queue_len == 1)
1190 mod_timer(&mrt->ipmr_expire_timer,
1191 c->_c.mfc_un.unres.expires);
1192 }
1193
1194 /* See if we can append the packet */
1195 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1196 c = NULL;
1197 err = -ENOBUFS;
1198 goto err;
1199 }
1200
1201 if (dev) {
1202 skb->dev = dev;
1203 skb->skb_iif = dev->ifindex;
1204 }
1205
1206 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1207
1208 spin_unlock_bh(&mfc_unres_lock);
1209 return 0;
1210
1211err:
1212 spin_unlock_bh(&mfc_unres_lock);
1213 if (c)
1214 ipmr_cache_free(c);
1215 kfree_skb(skb);
1216 return err;
1217}
1218
1219/* MFC cache manipulation by user space mroute daemon */
1220
1221static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1222{
1223 struct net *net = read_pnet(&mrt->net);
1224 struct mfc_cache *c;
1225
1226 rcu_read_lock();
1227 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1228 mfc->mfcc_mcastgrp.s_addr, parent);
1229 rcu_read_unlock();
1230 if (!c)
1231 return -ENOENT;
1232 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1233 list_del_rcu(&c->_c.list);
1234 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1235 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1236 mr_cache_put(&c->_c);
1237
1238 return 0;
1239}
1240
1241static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1242 struct mfcctl *mfc, int mrtsock, int parent)
1243{
1244 struct mfc_cache *uc, *c;
1245 struct mr_mfc *_uc;
1246 bool found;
1247 int ret;
1248
1249 if (mfc->mfcc_parent >= MAXVIFS)
1250 return -ENFILE;
1251
1252 rcu_read_lock();
1253 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1254 mfc->mfcc_mcastgrp.s_addr, parent);
1255 rcu_read_unlock();
1256 if (c) {
1257 spin_lock(&mrt_lock);
1258 c->_c.mfc_parent = mfc->mfcc_parent;
1259 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1260 if (!mrtsock)
1261 c->_c.mfc_flags |= MFC_STATIC;
1262 spin_unlock(&mrt_lock);
1263 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1264 mrt->id);
1265 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1266 return 0;
1267 }
1268
1269 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1270 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1271 return -EINVAL;
1272
1273 c = ipmr_cache_alloc();
1274 if (!c)
1275 return -ENOMEM;
1276
1277 c->mfc_origin = mfc->mfcc_origin.s_addr;
1278 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1279 c->_c.mfc_parent = mfc->mfcc_parent;
1280 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1281 if (!mrtsock)
1282 c->_c.mfc_flags |= MFC_STATIC;
1283
1284 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1285 ipmr_rht_params);
1286 if (ret) {
1287 pr_err("ipmr: rhtable insert error %d\n", ret);
1288 ipmr_cache_free(c);
1289 return ret;
1290 }
1291 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1292 /* Check to see if we resolved a queued list. If so we
1293 * need to send on the frames and tidy up.
1294 */
1295 found = false;
1296 spin_lock_bh(&mfc_unres_lock);
1297 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1298 uc = (struct mfc_cache *)_uc;
1299 if (uc->mfc_origin == c->mfc_origin &&
1300 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1301 list_del(&_uc->list);
1302 WRITE_ONCE(mrt->cache_resolve_queue_len,
1303 mrt->cache_resolve_queue_len - 1);
1304 found = true;
1305 break;
1306 }
1307 }
1308 if (list_empty(&mrt->mfc_unres_queue))
1309 timer_delete(&mrt->ipmr_expire_timer);
1310 spin_unlock_bh(&mfc_unres_lock);
1311
1312 if (found) {
1313 ipmr_cache_resolve(net, mrt, uc, c);
1314 ipmr_cache_free(uc);
1315 }
1316 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1317 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1318 return 0;
1319}
1320
1321/* Close the multicast socket, and clear the vif tables etc */
1322static void mroute_clean_tables(struct mr_table *mrt, int flags,
1323 struct list_head *dev_kill_list)
1324{
1325 struct net *net = read_pnet(&mrt->net);
1326 struct mfc_cache *cache;
1327 struct mr_mfc *c, *tmp;
1328 int i;
1329
1330 /* Shut down all active vif entries */
1331 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1332 for (i = 0; i < mrt->maxvif; i++) {
1333 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1334 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1335 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1336 continue;
1337 vif_delete(mrt, i, 0, dev_kill_list);
1338 }
1339 }
1340
1341 /* Wipe the cache */
1342 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1343 mutex_lock(&net->ipv4.mfc_mutex);
1344
1345 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1346 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1347 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1348 continue;
1349 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1350 list_del_rcu(&c->list);
1351 cache = (struct mfc_cache *)c;
1352 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1353 mrt->id);
1354 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1355 mr_cache_put(c);
1356 }
1357
1358 mutex_unlock(&net->ipv4.mfc_mutex);
1359 }
1360
1361 if (flags & MRT_FLUSH_MFC) {
1362 if (READ_ONCE(mrt->cache_resolve_queue_len) || !check_net(net)) {
1363 spin_lock_bh(&mfc_unres_lock);
1364 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1365 list_del(&c->list);
1366 cache = (struct mfc_cache *)c;
1367 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1368 ipmr_destroy_unres(mrt, cache);
1369 }
1370 spin_unlock_bh(&mfc_unres_lock);
1371 }
1372 }
1373}
1374
1375/* called from ip_ra_control(), before an RCU grace period,
1376 * we don't need to call synchronize_rcu() here
1377 */
1378static void mrtsock_destruct(struct sock *sk)
1379{
1380 struct net *net = sock_net(sk);
1381 LIST_HEAD(dev_kill_list);
1382 struct mr_table *mrt;
1383
1384 rtnl_lock();
1385
1386 ipmr_for_each_table(mrt, net) {
1387 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1388 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1389 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1390 NETCONFA_MC_FORWARDING,
1391 NETCONFA_IFINDEX_ALL,
1392 net->ipv4.devconf_all);
1393 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1394 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC,
1395 &dev_kill_list);
1396 }
1397 }
1398
1399 unregister_netdevice_many(&dev_kill_list);
1400
1401 rtnl_unlock();
1402}
1403
1404/* Socket options and virtual interface manipulation. The whole
1405 * virtual interface system is a complete heap, but unfortunately
1406 * that's how BSD mrouted happens to think. Maybe one day with a proper
1407 * MOSPF/PIM router set up we can clean this up.
1408 */
1409
1410int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1411 unsigned int optlen)
1412{
1413 struct net *net = sock_net(sk);
1414 int val, ret = 0, parent = 0;
1415 struct mr_table *mrt;
1416 struct vifctl vif;
1417 struct mfcctl mfc;
1418 bool do_wrvifwhole;
1419 u32 uval;
1420
1421 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1422 rtnl_lock();
1423 if (sk->sk_type != SOCK_RAW ||
1424 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1425 ret = -EOPNOTSUPP;
1426 goto out_unlock;
1427 }
1428
1429 mrt = __ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1430 if (!mrt) {
1431 ret = -ENOENT;
1432 goto out_unlock;
1433 }
1434 if (optname != MRT_INIT) {
1435 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1436 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1437 ret = -EACCES;
1438 goto out_unlock;
1439 }
1440 }
1441
1442 switch (optname) {
1443 case MRT_INIT:
1444 if (optlen != sizeof(int)) {
1445 ret = -EINVAL;
1446 break;
1447 }
1448 if (rtnl_dereference(mrt->mroute_sk)) {
1449 ret = -EADDRINUSE;
1450 break;
1451 }
1452
1453 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1454 if (ret == 0) {
1455 rcu_assign_pointer(mrt->mroute_sk, sk);
1456 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1457 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1458 NETCONFA_MC_FORWARDING,
1459 NETCONFA_IFINDEX_ALL,
1460 net->ipv4.devconf_all);
1461 }
1462 break;
1463 case MRT_DONE:
1464 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1465 ret = -EACCES;
1466 } else {
1467 /* We need to unlock here because mrtsock_destruct takes
1468 * care of rtnl itself and we can't change that due to
1469 * the IP_ROUTER_ALERT setsockopt which runs without it.
1470 */
1471 rtnl_unlock();
1472 ret = ip_ra_control(sk, 0, NULL);
1473 goto out;
1474 }
1475 break;
1476 case MRT_ADD_VIF:
1477 case MRT_DEL_VIF:
1478 if (optlen != sizeof(vif)) {
1479 ret = -EINVAL;
1480 break;
1481 }
1482 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1483 ret = -EFAULT;
1484 break;
1485 }
1486 if (vif.vifc_vifi >= MAXVIFS) {
1487 ret = -ENFILE;
1488 break;
1489 }
1490 if (optname == MRT_ADD_VIF) {
1491 ret = vif_add(net, mrt, &vif,
1492 sk == rtnl_dereference(mrt->mroute_sk));
1493 } else {
1494 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1495 }
1496 break;
1497 /* Manipulate the forwarding caches. These live
1498 * in a sort of kernel/user symbiosis.
1499 */
1500 case MRT_ADD_MFC:
1501 case MRT_DEL_MFC:
1502 parent = -1;
1503 fallthrough;
1504 case MRT_ADD_MFC_PROXY:
1505 case MRT_DEL_MFC_PROXY:
1506 if (optlen != sizeof(mfc)) {
1507 ret = -EINVAL;
1508 break;
1509 }
1510 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1511 ret = -EFAULT;
1512 break;
1513 }
1514 if (parent == 0)
1515 parent = mfc.mfcc_parent;
1516
1517 mutex_lock(&net->ipv4.mfc_mutex);
1518
1519 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1520 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1521 else
1522 ret = ipmr_mfc_add(net, mrt, &mfc,
1523 sk == rtnl_dereference(mrt->mroute_sk),
1524 parent);
1525
1526 mutex_unlock(&net->ipv4.mfc_mutex);
1527 break;
1528 case MRT_FLUSH: {
1529 LIST_HEAD(dev_kill_list);
1530
1531 if (optlen != sizeof(val)) {
1532 ret = -EINVAL;
1533 break;
1534 }
1535 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1536 ret = -EFAULT;
1537 break;
1538 }
1539
1540 mroute_clean_tables(mrt, val, &dev_kill_list);
1541 unregister_netdevice_many(&dev_kill_list);
1542 break;
1543 }
1544 /* Control PIM assert. */
1545 case MRT_ASSERT:
1546 if (optlen != sizeof(val)) {
1547 ret = -EINVAL;
1548 break;
1549 }
1550 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1551 ret = -EFAULT;
1552 break;
1553 }
1554 WRITE_ONCE(mrt->mroute_do_assert, val);
1555 break;
1556 case MRT_PIM:
1557 if (!ipmr_pimsm_enabled()) {
1558 ret = -ENOPROTOOPT;
1559 break;
1560 }
1561 if (optlen != sizeof(val)) {
1562 ret = -EINVAL;
1563 break;
1564 }
1565 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1566 ret = -EFAULT;
1567 break;
1568 }
1569
1570 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1571 val = !!val;
1572 if (val != mrt->mroute_do_pim) {
1573 WRITE_ONCE(mrt->mroute_do_pim, val);
1574 WRITE_ONCE(mrt->mroute_do_assert, val);
1575 WRITE_ONCE(mrt->mroute_do_wrvifwhole, do_wrvifwhole);
1576 }
1577 break;
1578 case MRT_TABLE:
1579 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1580 ret = -ENOPROTOOPT;
1581 break;
1582 }
1583 if (optlen != sizeof(uval)) {
1584 ret = -EINVAL;
1585 break;
1586 }
1587 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1588 ret = -EFAULT;
1589 break;
1590 }
1591
1592 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1593 ret = -EBUSY;
1594 } else {
1595 mrt = ipmr_new_table(net, uval);
1596 if (IS_ERR(mrt))
1597 ret = PTR_ERR(mrt);
1598 else
1599 raw_sk(sk)->ipmr_table = uval;
1600 }
1601 break;
1602 /* Spurious command, or MRT_VERSION which you cannot set. */
1603 default:
1604 ret = -ENOPROTOOPT;
1605 }
1606out_unlock:
1607 rtnl_unlock();
1608out:
1609 return ret;
1610}
1611
1612/* Execute if this ioctl is a special mroute ioctl */
1613int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1614{
1615 switch (cmd) {
1616 /* These userspace buffers will be consumed by ipmr_ioctl() */
1617 case SIOCGETVIFCNT: {
1618 struct sioc_vif_req buffer;
1619
1620 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1621 sizeof(buffer));
1622 }
1623 case SIOCGETSGCNT: {
1624 struct sioc_sg_req buffer;
1625
1626 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1627 sizeof(buffer));
1628 }
1629 }
1630 /* return code > 0 means that the ioctl was not executed */
1631 return 1;
1632}
1633
1634/* Getsock opt support for the multicast routing system. */
1635int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1636 sockptr_t optlen)
1637{
1638 int olr;
1639 int val;
1640 struct net *net = sock_net(sk);
1641 struct mr_table *mrt;
1642
1643 if (sk->sk_type != SOCK_RAW ||
1644 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1645 return -EOPNOTSUPP;
1646
1647 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1648 if (!mrt)
1649 return -ENOENT;
1650
1651 switch (optname) {
1652 case MRT_VERSION:
1653 val = 0x0305;
1654 break;
1655 case MRT_PIM:
1656 if (!ipmr_pimsm_enabled())
1657 return -ENOPROTOOPT;
1658 val = READ_ONCE(mrt->mroute_do_pim);
1659 break;
1660 case MRT_ASSERT:
1661 val = READ_ONCE(mrt->mroute_do_assert);
1662 break;
1663 default:
1664 return -ENOPROTOOPT;
1665 }
1666
1667 if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1668 return -EFAULT;
1669 if (olr < 0)
1670 return -EINVAL;
1671
1672 olr = min_t(unsigned int, olr, sizeof(int));
1673
1674 if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1675 return -EFAULT;
1676 if (copy_to_sockptr(optval, &val, olr))
1677 return -EFAULT;
1678 return 0;
1679}
1680
1681/* The IP multicast ioctl support routines. */
1682int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1683{
1684 struct vif_device *vif;
1685 struct mfc_cache *c;
1686 struct net *net = sock_net(sk);
1687 struct sioc_vif_req *vr;
1688 struct sioc_sg_req *sr;
1689 struct mr_table *mrt;
1690
1691 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1692 if (!mrt)
1693 return -ENOENT;
1694
1695 switch (cmd) {
1696 case SIOCGETVIFCNT:
1697 vr = (struct sioc_vif_req *)arg;
1698 if (vr->vifi >= mrt->maxvif)
1699 return -EINVAL;
1700 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1701 rcu_read_lock();
1702 vif = &mrt->vif_table[vr->vifi];
1703 if (VIF_EXISTS(mrt, vr->vifi)) {
1704 vr->icount = READ_ONCE(vif->pkt_in);
1705 vr->ocount = READ_ONCE(vif->pkt_out);
1706 vr->ibytes = READ_ONCE(vif->bytes_in);
1707 vr->obytes = READ_ONCE(vif->bytes_out);
1708 rcu_read_unlock();
1709
1710 return 0;
1711 }
1712 rcu_read_unlock();
1713 return -EADDRNOTAVAIL;
1714 case SIOCGETSGCNT:
1715 sr = (struct sioc_sg_req *)arg;
1716
1717 rcu_read_lock();
1718 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1719 if (c) {
1720 sr->pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1721 sr->bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1722 sr->wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1723 rcu_read_unlock();
1724 return 0;
1725 }
1726 rcu_read_unlock();
1727 return -EADDRNOTAVAIL;
1728 default:
1729 return -ENOIOCTLCMD;
1730 }
1731}
1732
1733#ifdef CONFIG_COMPAT
1734struct compat_sioc_sg_req {
1735 struct in_addr src;
1736 struct in_addr grp;
1737 compat_ulong_t pktcnt;
1738 compat_ulong_t bytecnt;
1739 compat_ulong_t wrong_if;
1740};
1741
1742struct compat_sioc_vif_req {
1743 vifi_t vifi; /* Which iface */
1744 compat_ulong_t icount;
1745 compat_ulong_t ocount;
1746 compat_ulong_t ibytes;
1747 compat_ulong_t obytes;
1748};
1749
1750int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1751{
1752 struct compat_sioc_sg_req sr;
1753 struct compat_sioc_vif_req vr;
1754 struct vif_device *vif;
1755 struct mfc_cache *c;
1756 struct net *net = sock_net(sk);
1757 struct mr_table *mrt;
1758
1759 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1760 if (!mrt)
1761 return -ENOENT;
1762
1763 switch (cmd) {
1764 case SIOCGETVIFCNT:
1765 if (copy_from_user(&vr, arg, sizeof(vr)))
1766 return -EFAULT;
1767 if (vr.vifi >= mrt->maxvif)
1768 return -EINVAL;
1769 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1770 rcu_read_lock();
1771 vif = &mrt->vif_table[vr.vifi];
1772 if (VIF_EXISTS(mrt, vr.vifi)) {
1773 vr.icount = READ_ONCE(vif->pkt_in);
1774 vr.ocount = READ_ONCE(vif->pkt_out);
1775 vr.ibytes = READ_ONCE(vif->bytes_in);
1776 vr.obytes = READ_ONCE(vif->bytes_out);
1777 rcu_read_unlock();
1778
1779 if (copy_to_user(arg, &vr, sizeof(vr)))
1780 return -EFAULT;
1781 return 0;
1782 }
1783 rcu_read_unlock();
1784 return -EADDRNOTAVAIL;
1785 case SIOCGETSGCNT:
1786 if (copy_from_user(&sr, arg, sizeof(sr)))
1787 return -EFAULT;
1788
1789 rcu_read_lock();
1790 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1791 if (c) {
1792 sr.pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1793 sr.bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1794 sr.wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1795 rcu_read_unlock();
1796
1797 if (copy_to_user(arg, &sr, sizeof(sr)))
1798 return -EFAULT;
1799 return 0;
1800 }
1801 rcu_read_unlock();
1802 return -EADDRNOTAVAIL;
1803 default:
1804 return -ENOIOCTLCMD;
1805 }
1806}
1807#endif
1808
1809static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1810{
1811 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1812 struct net *net = dev_net(dev);
1813 struct mr_table *mrt;
1814 struct vif_device *v;
1815 int ct;
1816
1817 if (event != NETDEV_UNREGISTER)
1818 return NOTIFY_DONE;
1819
1820 ipmr_for_each_table(mrt, net) {
1821 v = &mrt->vif_table[0];
1822 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1823 if (rcu_access_pointer(v->dev) == dev)
1824 vif_delete(mrt, ct, 1, NULL);
1825 }
1826 }
1827 return NOTIFY_DONE;
1828}
1829
1830static struct notifier_block ip_mr_notifier = {
1831 .notifier_call = ipmr_device_event,
1832};
1833
1834/* Encapsulate a packet by attaching a valid IPIP header to it.
1835 * This avoids tunnel drivers and other mess and gives us the speed so
1836 * important for multicast video.
1837 */
1838static void ip_encap(struct net *net, struct sk_buff *skb,
1839 __be32 saddr, __be32 daddr)
1840{
1841 struct iphdr *iph;
1842 const struct iphdr *old_iph = ip_hdr(skb);
1843
1844 skb_push(skb, sizeof(struct iphdr));
1845 skb->transport_header = skb->network_header;
1846 skb_reset_network_header(skb);
1847 iph = ip_hdr(skb);
1848
1849 iph->version = 4;
1850 iph->tos = old_iph->tos;
1851 iph->ttl = old_iph->ttl;
1852 iph->frag_off = 0;
1853 iph->daddr = daddr;
1854 iph->saddr = saddr;
1855 iph->protocol = IPPROTO_IPIP;
1856 iph->ihl = 5;
1857 iph->tot_len = htons(skb->len);
1858 ip_select_ident(net, skb, NULL);
1859 ip_send_check(iph);
1860
1861 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1862 nf_reset_ct(skb);
1863}
1864
1865static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1866 struct sk_buff *skb)
1867{
1868 struct ip_options *opt = &(IPCB(skb)->opt);
1869
1870 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1871
1872 if (unlikely(opt->optlen))
1873 ip_forward_options(skb);
1874
1875 return dst_output(net, sk, skb);
1876}
1877
1878#ifdef CONFIG_NET_SWITCHDEV
1879static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1880 int in_vifi, int out_vifi)
1881{
1882 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1883 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1884
1885 if (!skb->offload_l3_fwd_mark)
1886 return false;
1887 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1888 return false;
1889 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1890 &in_vif->dev_parent_id);
1891}
1892#else
1893static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1894 int in_vifi, int out_vifi)
1895{
1896 return false;
1897}
1898#endif
1899
1900/* Processing handlers for ipmr_forward, under rcu_read_lock() */
1901
1902static int ipmr_prepare_xmit(struct net *net, struct mr_table *mrt,
1903 struct sk_buff *skb, int vifi)
1904{
1905 const struct iphdr *iph = ip_hdr(skb);
1906 struct vif_device *vif = &mrt->vif_table[vifi];
1907 struct net_device *vif_dev;
1908 struct rtable *rt;
1909 struct flowi4 fl4;
1910 int encap = 0;
1911
1912 vif_dev = vif_dev_read(vif);
1913 if (!vif_dev)
1914 return -1;
1915
1916 if (vif->flags & VIFF_REGISTER) {
1917 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1918 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1919 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1920 DEV_STATS_INC(vif_dev, tx_packets);
1921 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1922 return -1;
1923 }
1924
1925 if (vif->flags & VIFF_TUNNEL) {
1926 rt = ip_route_output_ports(net, &fl4, NULL,
1927 vif->remote, vif->local,
1928 0, 0,
1929 IPPROTO_IPIP,
1930 iph->tos & INET_DSCP_MASK, vif->link);
1931 if (IS_ERR(rt))
1932 return -1;
1933 encap = sizeof(struct iphdr);
1934 } else {
1935 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1936 0, 0,
1937 IPPROTO_IPIP,
1938 iph->tos & INET_DSCP_MASK, vif->link);
1939 if (IS_ERR(rt))
1940 return -1;
1941 }
1942
1943 if (skb->len+encap > dst4_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1944 /* Do not fragment multicasts. Alas, IPv4 does not
1945 * allow to send ICMP, so that packets will disappear
1946 * to blackhole.
1947 */
1948 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1949 ip_rt_put(rt);
1950 return -1;
1951 }
1952
1953 encap += LL_RESERVED_SPACE(dst_dev_rcu(&rt->dst)) + rt->dst.header_len;
1954
1955 if (skb_cow(skb, encap)) {
1956 ip_rt_put(rt);
1957 return -1;
1958 }
1959
1960 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1961 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1962
1963 skb_dst_drop(skb);
1964 skb_dst_set(skb, &rt->dst);
1965 ip_decrease_ttl(ip_hdr(skb));
1966
1967 /* FIXME: forward and output firewalls used to be called here.
1968 * What do we do with netfilter? -- RR
1969 */
1970 if (vif->flags & VIFF_TUNNEL) {
1971 ip_encap(net, skb, vif->local, vif->remote);
1972 /* FIXME: extra output firewall step used to be here. --RR */
1973 DEV_STATS_INC(vif_dev, tx_packets);
1974 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1975 }
1976
1977 return 0;
1978}
1979
1980static void ipmr_queue_fwd_xmit(struct net *net, struct mr_table *mrt,
1981 int in_vifi, struct sk_buff *skb, int vifi)
1982{
1983 struct rtable *rt;
1984
1985 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1986 goto out_free;
1987
1988 if (ipmr_prepare_xmit(net, mrt, skb, vifi))
1989 goto out_free;
1990
1991 rt = skb_rtable(skb);
1992
1993 IPCB(skb)->flags |= IPSKB_FORWARDED;
1994
1995 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1996 * not only before forwarding, but after forwarding on all output
1997 * interfaces. It is clear, if mrouter runs a multicasting
1998 * program, it should receive packets not depending to what interface
1999 * program is joined.
2000 * If we will not make it, the program will have to join on all
2001 * interfaces. On the other hand, multihoming host (or router, but
2002 * not mrouter) cannot join to more than one interface - it will
2003 * result in receiving multiple packets.
2004 */
2005 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
2006 net, NULL, skb, skb->dev, dst_dev_rcu(&rt->dst),
2007 ipmr_forward_finish);
2008 return;
2009
2010out_free:
2011 kfree_skb(skb);
2012}
2013
2014static void ipmr_queue_output_xmit(struct net *net, struct mr_table *mrt,
2015 struct sk_buff *skb, int vifi)
2016{
2017 if (ipmr_prepare_xmit(net, mrt, skb, vifi))
2018 goto out_free;
2019
2020 ip_mc_output(net, NULL, skb);
2021 return;
2022
2023out_free:
2024 kfree_skb(skb);
2025}
2026
2027/* Called with mrt_lock or rcu_read_lock() */
2028static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
2029{
2030 int ct;
2031 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
2032 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
2033 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
2034 break;
2035 }
2036 return ct;
2037}
2038
2039/* "local" means that we should preserve one skb (for local delivery) */
2040/* Called uner rcu_read_lock() */
2041static void ip_mr_forward(struct net *net, struct mr_table *mrt,
2042 struct net_device *dev, struct sk_buff *skb,
2043 struct mfc_cache *c, int local)
2044{
2045 int true_vifi = ipmr_find_vif(mrt, dev);
2046 int psend = -1;
2047 int vif, ct;
2048
2049 vif = c->_c.mfc_parent;
2050 atomic_long_inc(&c->_c.mfc_un.res.pkt);
2051 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes);
2052 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies);
2053
2054 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
2055 struct mfc_cache *cache_proxy;
2056
2057 /* For an (*,G) entry, we only check that the incoming
2058 * interface is part of the static tree.
2059 */
2060 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2061 if (cache_proxy &&
2062 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
2063 goto forward;
2064 }
2065
2066 /* Wrong interface: drop packet and (maybe) send PIM assert. */
2067 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
2068 if (rt_is_output_route(skb_rtable(skb))) {
2069 /* It is our own packet, looped back.
2070 * Very complicated situation...
2071 *
2072 * The best workaround until routing daemons will be
2073 * fixed is not to redistribute packet, if it was
2074 * send through wrong interface. It means, that
2075 * multicast applications WILL NOT work for
2076 * (S,G), which have default multicast route pointing
2077 * to wrong oif. In any case, it is not a good
2078 * idea to use multicasting applications on router.
2079 */
2080 goto dont_forward;
2081 }
2082
2083 atomic_long_inc(&c->_c.mfc_un.res.wrong_if);
2084
2085 if (true_vifi >= 0 && READ_ONCE(mrt->mroute_do_assert) &&
2086 /* pimsm uses asserts, when switching from RPT to SPT,
2087 * so that we cannot check that packet arrived on an oif.
2088 * It is bad, but otherwise we would need to move pretty
2089 * large chunk of pimd to kernel. Ough... --ANK
2090 */
2091 (READ_ONCE(mrt->mroute_do_pim) ||
2092 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2093 time_after(jiffies,
2094 c->_c.mfc_un.res.last_assert +
2095 MFC_ASSERT_THRESH)) {
2096 c->_c.mfc_un.res.last_assert = jiffies;
2097 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2098 if (READ_ONCE(mrt->mroute_do_wrvifwhole))
2099 ipmr_cache_report(mrt, skb, true_vifi,
2100 IGMPMSG_WRVIFWHOLE);
2101 }
2102 goto dont_forward;
2103 }
2104
2105forward:
2106 WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2107 mrt->vif_table[vif].pkt_in + 1);
2108 WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2109 mrt->vif_table[vif].bytes_in + skb->len);
2110
2111 /* Forward the frame */
2112 if (c->mfc_origin == htonl(INADDR_ANY) &&
2113 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2114 if (true_vifi >= 0 &&
2115 true_vifi != c->_c.mfc_parent &&
2116 ip_hdr(skb)->ttl >
2117 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2118 /* It's an (*,*) entry and the packet is not coming from
2119 * the upstream: forward the packet to the upstream
2120 * only.
2121 */
2122 psend = c->_c.mfc_parent;
2123 goto last_forward;
2124 }
2125 goto dont_forward;
2126 }
2127 for (ct = c->_c.mfc_un.res.maxvif - 1;
2128 ct >= c->_c.mfc_un.res.minvif; ct--) {
2129 /* For (*,G) entry, don't forward to the incoming interface */
2130 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2131 ct != true_vifi) &&
2132 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2133 if (psend != -1) {
2134 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2135
2136 if (skb2)
2137 ipmr_queue_fwd_xmit(net, mrt, true_vifi,
2138 skb2, psend);
2139 }
2140 psend = ct;
2141 }
2142 }
2143last_forward:
2144 if (psend != -1) {
2145 if (local) {
2146 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2147
2148 if (skb2)
2149 ipmr_queue_fwd_xmit(net, mrt, true_vifi, skb2,
2150 psend);
2151 } else {
2152 ipmr_queue_fwd_xmit(net, mrt, true_vifi, skb, psend);
2153 return;
2154 }
2155 }
2156
2157dont_forward:
2158 if (!local)
2159 kfree_skb(skb);
2160}
2161
2162static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2163{
2164 struct rtable *rt = skb_rtable(skb);
2165 struct iphdr *iph = ip_hdr(skb);
2166 struct flowi4 fl4 = {
2167 .daddr = iph->daddr,
2168 .saddr = iph->saddr,
2169 .flowi4_dscp = ip4h_dscp(iph),
2170 .flowi4_oif = (rt_is_output_route(rt) ?
2171 skb->dev->ifindex : 0),
2172 .flowi4_iif = (rt_is_output_route(rt) ?
2173 LOOPBACK_IFINDEX :
2174 skb->dev->ifindex),
2175 .flowi4_mark = skb->mark,
2176 };
2177 struct mr_table *mrt;
2178 int err;
2179
2180 err = ipmr_fib_lookup(net, &fl4, &mrt);
2181 if (err)
2182 return ERR_PTR(err);
2183 return mrt;
2184}
2185
2186/* Multicast packets for forwarding arrive here
2187 * Called with rcu_read_lock();
2188 */
2189int ip_mr_input(struct sk_buff *skb)
2190{
2191 struct mfc_cache *cache;
2192 struct net *net = dev_net(skb->dev);
2193 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2194 struct mr_table *mrt;
2195 struct net_device *dev;
2196
2197 /* skb->dev passed in is the loX master dev for vrfs.
2198 * As there are no vifs associated with loopback devices,
2199 * get the proper interface that does have a vif associated with it.
2200 */
2201 dev = skb->dev;
2202 if (netif_is_l3_master(skb->dev)) {
2203 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2204 if (!dev) {
2205 kfree_skb(skb);
2206 return -ENODEV;
2207 }
2208 }
2209
2210 /* Packet is looped back after forward, it should not be
2211 * forwarded second time, but still can be delivered locally.
2212 */
2213 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2214 goto dont_forward;
2215
2216 if (!local)
2217 skb_orphan(skb);
2218
2219 mrt = ipmr_rt_fib_lookup(net, skb);
2220 if (IS_ERR(mrt)) {
2221 kfree_skb(skb);
2222 return PTR_ERR(mrt);
2223 }
2224 if (!local) {
2225 if (IPCB(skb)->opt.router_alert) {
2226 if (ip_call_ra_chain(skb))
2227 return 0;
2228 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2229 /* IGMPv1 (and broken IGMPv2 implementations sort of
2230 * Cisco IOS <= 11.2(8)) do not put router alert
2231 * option to IGMP packets destined to routable
2232 * groups. It is very bad, because it means
2233 * that we can forward NO IGMP messages.
2234 */
2235 struct sock *mroute_sk;
2236
2237 mroute_sk = rcu_dereference(mrt->mroute_sk);
2238 if (mroute_sk) {
2239 nf_reset_ct(skb);
2240 raw_rcv(mroute_sk, skb);
2241 return 0;
2242 }
2243 }
2244 }
2245
2246 /* already under rcu_read_lock() */
2247 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2248 if (!cache) {
2249 int vif = ipmr_find_vif(mrt, dev);
2250
2251 if (vif >= 0)
2252 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2253 vif);
2254 }
2255
2256 /* No usable cache entry */
2257 if (!cache) {
2258 int vif;
2259
2260 if (local) {
2261 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2262 ip_local_deliver(skb);
2263 if (!skb2)
2264 return -ENOBUFS;
2265 skb = skb2;
2266 }
2267
2268 vif = ipmr_find_vif(mrt, dev);
2269 if (vif >= 0)
2270 return ipmr_cache_unresolved(mrt, vif, skb, dev);
2271 kfree_skb(skb);
2272 return -ENODEV;
2273 }
2274
2275 ip_mr_forward(net, mrt, dev, skb, cache, local);
2276
2277 if (local)
2278 return ip_local_deliver(skb);
2279
2280 return 0;
2281
2282dont_forward:
2283 if (local)
2284 return ip_local_deliver(skb);
2285 kfree_skb(skb);
2286 return 0;
2287}
2288
2289static void ip_mr_output_finish(struct net *net, struct mr_table *mrt,
2290 struct net_device *dev, struct sk_buff *skb,
2291 struct mfc_cache *c)
2292{
2293 int psend = -1;
2294 int ct;
2295
2296 atomic_long_inc(&c->_c.mfc_un.res.pkt);
2297 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes);
2298 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies);
2299
2300 /* Forward the frame */
2301 if (c->mfc_origin == htonl(INADDR_ANY) &&
2302 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2303 if (ip_hdr(skb)->ttl >
2304 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2305 /* It's an (*,*) entry and the packet is not coming from
2306 * the upstream: forward the packet to the upstream
2307 * only.
2308 */
2309 psend = c->_c.mfc_parent;
2310 goto last_xmit;
2311 }
2312 goto dont_xmit;
2313 }
2314
2315 for (ct = c->_c.mfc_un.res.maxvif - 1;
2316 ct >= c->_c.mfc_un.res.minvif; ct--) {
2317 if (ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2318 if (psend != -1) {
2319 struct sk_buff *skb2;
2320
2321 skb2 = skb_clone(skb, GFP_ATOMIC);
2322 if (skb2)
2323 ipmr_queue_output_xmit(net, mrt,
2324 skb2, psend);
2325 }
2326 psend = ct;
2327 }
2328 }
2329
2330last_xmit:
2331 if (psend != -1) {
2332 ipmr_queue_output_xmit(net, mrt, skb, psend);
2333 return;
2334 }
2335
2336dont_xmit:
2337 kfree_skb(skb);
2338}
2339
2340/* Multicast packets for forwarding arrive here
2341 * Called with rcu_read_lock();
2342 */
2343int ip_mr_output(struct net *net, struct sock *sk, struct sk_buff *skb)
2344{
2345 struct rtable *rt = skb_rtable(skb);
2346 struct mfc_cache *cache;
2347 struct net_device *dev;
2348 struct mr_table *mrt;
2349 int vif;
2350
2351 guard(rcu)();
2352
2353 dev = dst_dev_rcu(&rt->dst);
2354
2355 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2356 goto mc_output;
2357 if (!(IPCB(skb)->flags & IPSKB_MCROUTE))
2358 goto mc_output;
2359
2360 skb->dev = dev;
2361
2362 mrt = ipmr_rt_fib_lookup(net, skb);
2363 if (IS_ERR(mrt))
2364 goto mc_output;
2365
2366 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2367 if (!cache) {
2368 vif = ipmr_find_vif(mrt, dev);
2369 if (vif >= 0)
2370 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2371 vif);
2372 }
2373
2374 /* No usable cache entry */
2375 if (!cache) {
2376 vif = ipmr_find_vif(mrt, dev);
2377 if (vif >= 0)
2378 return ipmr_cache_unresolved(mrt, vif, skb, dev);
2379 goto mc_output;
2380 }
2381
2382 vif = cache->_c.mfc_parent;
2383 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev)
2384 goto mc_output;
2385
2386 ip_mr_output_finish(net, mrt, dev, skb, cache);
2387 return 0;
2388
2389mc_output:
2390 return ip_mc_output(net, sk, skb);
2391}
2392
2393#ifdef CONFIG_IP_PIMSM_V1
2394/* Handle IGMP messages of PIMv1 */
2395int pim_rcv_v1(struct sk_buff *skb)
2396{
2397 struct igmphdr *pim;
2398 struct net *net = dev_net(skb->dev);
2399 struct mr_table *mrt;
2400
2401 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2402 goto drop;
2403
2404 pim = igmp_hdr(skb);
2405
2406 mrt = ipmr_rt_fib_lookup(net, skb);
2407 if (IS_ERR(mrt))
2408 goto drop;
2409 if (!READ_ONCE(mrt->mroute_do_pim) ||
2410 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2411 goto drop;
2412
2413 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2414drop:
2415 kfree_skb(skb);
2416 }
2417 return 0;
2418}
2419#endif
2420
2421#ifdef CONFIG_IP_PIMSM_V2
2422static int pim_rcv(struct sk_buff *skb)
2423{
2424 struct pimreghdr *pim;
2425 struct net *net = dev_net(skb->dev);
2426 struct mr_table *mrt;
2427
2428 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2429 goto drop;
2430
2431 pim = (struct pimreghdr *)skb_transport_header(skb);
2432 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2433 (pim->flags & PIM_NULL_REGISTER) ||
2434 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2435 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2436 goto drop;
2437
2438 mrt = ipmr_rt_fib_lookup(net, skb);
2439 if (IS_ERR(mrt))
2440 goto drop;
2441 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2442drop:
2443 kfree_skb(skb);
2444 }
2445 return 0;
2446}
2447#endif
2448
2449int ipmr_get_route(struct net *net, struct sk_buff *skb,
2450 __be32 saddr, __be32 daddr,
2451 struct rtmsg *rtm, u32 portid)
2452{
2453 struct mfc_cache *cache;
2454 struct mr_table *mrt;
2455 int err;
2456
2457 rcu_read_lock();
2458 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
2459 if (!mrt) {
2460 rcu_read_unlock();
2461 return -ENOENT;
2462 }
2463
2464 cache = ipmr_cache_find(mrt, saddr, daddr);
2465 if (!cache && skb->dev) {
2466 int vif = ipmr_find_vif(mrt, skb->dev);
2467
2468 if (vif >= 0)
2469 cache = ipmr_cache_find_any(mrt, daddr, vif);
2470 }
2471 if (!cache) {
2472 struct sk_buff *skb2;
2473 struct iphdr *iph;
2474 struct net_device *dev;
2475 int vif = -1;
2476
2477 dev = skb->dev;
2478 if (dev)
2479 vif = ipmr_find_vif(mrt, dev);
2480 if (vif < 0) {
2481 rcu_read_unlock();
2482 return -ENODEV;
2483 }
2484
2485 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2486 if (!skb2) {
2487 rcu_read_unlock();
2488 return -ENOMEM;
2489 }
2490
2491 NETLINK_CB(skb2).portid = portid;
2492 skb_push(skb2, sizeof(struct iphdr));
2493 skb_reset_network_header(skb2);
2494 iph = ip_hdr(skb2);
2495 iph->ihl = sizeof(struct iphdr) >> 2;
2496 iph->saddr = saddr;
2497 iph->daddr = daddr;
2498 iph->version = 0;
2499 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2500 rcu_read_unlock();
2501 return err;
2502 }
2503
2504 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2505 rcu_read_unlock();
2506 return err;
2507}
2508
2509static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2510 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2511 int flags)
2512{
2513 struct nlmsghdr *nlh;
2514 struct rtmsg *rtm;
2515 int err;
2516
2517 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2518 if (!nlh)
2519 return -EMSGSIZE;
2520
2521 rtm = nlmsg_data(nlh);
2522 rtm->rtm_family = RTNL_FAMILY_IPMR;
2523 rtm->rtm_dst_len = 32;
2524 rtm->rtm_src_len = 32;
2525 rtm->rtm_tos = 0;
2526 rtm->rtm_table = mrt->id;
2527 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2528 goto nla_put_failure;
2529 rtm->rtm_type = RTN_MULTICAST;
2530 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2531 if (c->_c.mfc_flags & MFC_STATIC)
2532 rtm->rtm_protocol = RTPROT_STATIC;
2533 else
2534 rtm->rtm_protocol = RTPROT_MROUTED;
2535 rtm->rtm_flags = 0;
2536
2537 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2538 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2539 goto nla_put_failure;
2540 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2541 /* do not break the dump if cache is unresolved */
2542 if (err < 0 && err != -ENOENT)
2543 goto nla_put_failure;
2544
2545 nlmsg_end(skb, nlh);
2546 return 0;
2547
2548nla_put_failure:
2549 nlmsg_cancel(skb, nlh);
2550 return -EMSGSIZE;
2551}
2552
2553static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2554 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2555 int flags)
2556{
2557 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2558 cmd, flags);
2559}
2560
2561static size_t mroute_msgsize(bool unresolved)
2562{
2563 size_t len =
2564 NLMSG_ALIGN(sizeof(struct rtmsg))
2565 + nla_total_size(4) /* RTA_TABLE */
2566 + nla_total_size(4) /* RTA_SRC */
2567 + nla_total_size(4) /* RTA_DST */
2568 ;
2569
2570 if (!unresolved)
2571 len = len
2572 + nla_total_size(4) /* RTA_IIF */
2573 + nla_total_size(0) /* RTA_MULTIPATH */
2574 + MAXVIFS * NLA_ALIGN(sizeof(struct rtnexthop))
2575 /* RTA_MFC_STATS */
2576 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2577 ;
2578
2579 return len;
2580}
2581
2582static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2583 int cmd)
2584{
2585 struct net *net = read_pnet(&mrt->net);
2586 struct sk_buff *skb;
2587 int err = -ENOBUFS;
2588
2589 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS),
2590 GFP_ATOMIC);
2591 if (!skb)
2592 goto errout;
2593
2594 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2595 if (err < 0)
2596 goto errout;
2597
2598 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2599 return;
2600
2601errout:
2602 kfree_skb(skb);
2603 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2604}
2605
2606static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2607{
2608 size_t len =
2609 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2610 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2611 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2612 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2613 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2614 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */
2615 /* IPMRA_CREPORT_PKT */
2616 + nla_total_size(payloadlen)
2617 ;
2618
2619 return len;
2620}
2621
2622static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2623{
2624 struct net *net = read_pnet(&mrt->net);
2625 struct nlmsghdr *nlh;
2626 struct rtgenmsg *rtgenm;
2627 struct igmpmsg *msg;
2628 struct sk_buff *skb;
2629 struct nlattr *nla;
2630 int payloadlen;
2631
2632 payloadlen = pkt->len - sizeof(struct igmpmsg);
2633 msg = (struct igmpmsg *)skb_network_header(pkt);
2634
2635 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2636 if (!skb)
2637 goto errout;
2638
2639 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2640 sizeof(struct rtgenmsg), 0);
2641 if (!nlh)
2642 goto errout;
2643 rtgenm = nlmsg_data(nlh);
2644 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2645 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2646 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2647 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2648 msg->im_src.s_addr) ||
2649 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2650 msg->im_dst.s_addr) ||
2651 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2652 goto nla_put_failure;
2653
2654 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2655 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2656 nla_data(nla), payloadlen))
2657 goto nla_put_failure;
2658
2659 nlmsg_end(skb, nlh);
2660
2661 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2662 return;
2663
2664nla_put_failure:
2665 nlmsg_cancel(skb, nlh);
2666errout:
2667 kfree_skb(skb);
2668 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2669}
2670
2671static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2672 const struct nlmsghdr *nlh,
2673 struct nlattr **tb,
2674 struct netlink_ext_ack *extack)
2675{
2676 struct rtmsg *rtm;
2677 int i, err;
2678
2679 rtm = nlmsg_payload(nlh, sizeof(*rtm));
2680 if (!rtm) {
2681 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2682 return -EINVAL;
2683 }
2684
2685 if (!netlink_strict_get_check(skb))
2686 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2687 rtm_ipv4_policy, extack);
2688
2689 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2690 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2691 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2692 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2693 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2694 return -EINVAL;
2695 }
2696
2697 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2698 rtm_ipv4_policy, extack);
2699 if (err)
2700 return err;
2701
2702 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2703 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2704 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2705 return -EINVAL;
2706 }
2707
2708 for (i = 0; i <= RTA_MAX; i++) {
2709 if (!tb[i])
2710 continue;
2711
2712 switch (i) {
2713 case RTA_SRC:
2714 case RTA_DST:
2715 case RTA_TABLE:
2716 break;
2717 default:
2718 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2719 return -EINVAL;
2720 }
2721 }
2722
2723 return 0;
2724}
2725
2726static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2727 struct netlink_ext_ack *extack)
2728{
2729 struct net *net = sock_net(in_skb->sk);
2730 struct nlattr *tb[RTA_MAX + 1];
2731 struct mfc_cache *cache;
2732 struct mr_table *mrt;
2733 struct sk_buff *skb;
2734 __be32 src, grp;
2735 u32 tableid;
2736 int err;
2737
2738 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2739 if (err < 0)
2740 goto errout;
2741
2742 src = nla_get_in_addr_default(tb[RTA_SRC], 0);
2743 grp = nla_get_in_addr_default(tb[RTA_DST], 0);
2744 tableid = nla_get_u32_default(tb[RTA_TABLE], 0);
2745
2746 skb = nlmsg_new(mroute_msgsize(false), GFP_KERNEL);
2747 if (!skb) {
2748 err = -ENOBUFS;
2749 goto errout;
2750 }
2751
2752 rcu_read_lock();
2753
2754 mrt = __ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2755 if (!mrt) {
2756 err = -ENOENT;
2757 goto errout_unlock;
2758 }
2759
2760 cache = ipmr_cache_find(mrt, src, grp);
2761 if (!cache) {
2762 err = -ENOENT;
2763 goto errout_unlock;
2764 }
2765
2766 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2767 nlh->nlmsg_seq, cache,
2768 RTM_NEWROUTE, 0);
2769 if (err < 0)
2770 goto errout_unlock;
2771
2772 rcu_read_unlock();
2773
2774 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2775errout:
2776 return err;
2777
2778errout_unlock:
2779 rcu_read_unlock();
2780 kfree_skb(skb);
2781 goto errout;
2782}
2783
2784static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2785{
2786 struct fib_dump_filter filter = {};
2787 int err;
2788
2789 rcu_read_lock();
2790
2791 if (cb->strict_check) {
2792 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2793 &filter, cb);
2794 if (err < 0)
2795 goto out;
2796 }
2797
2798 if (filter.table_id) {
2799 struct mr_table *mrt;
2800
2801 mrt = __ipmr_get_table(sock_net(skb->sk), filter.table_id);
2802 if (!mrt) {
2803 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR) {
2804 err = skb->len;
2805 goto out;
2806 }
2807
2808 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2809 err = -ENOENT;
2810 goto out;
2811 }
2812
2813 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2814 &mfc_unres_lock, &filter);
2815 err = skb->len ? : err;
2816 goto out;
2817 }
2818
2819 err = mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2820 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2821out:
2822 rcu_read_unlock();
2823
2824 return err;
2825}
2826
2827static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2828 [RTA_SRC] = { .type = NLA_U32 },
2829 [RTA_DST] = { .type = NLA_U32 },
2830 [RTA_IIF] = { .type = NLA_U32 },
2831 [RTA_TABLE] = { .type = NLA_U32 },
2832 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2833};
2834
2835static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2836{
2837 switch (rtm_protocol) {
2838 case RTPROT_STATIC:
2839 case RTPROT_MROUTED:
2840 return true;
2841 }
2842 return false;
2843}
2844
2845static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2846{
2847 struct rtnexthop *rtnh = nla_data(nla);
2848 int remaining = nla_len(nla), vifi = 0;
2849
2850 while (rtnh_ok(rtnh, remaining)) {
2851 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2852 if (++vifi == MAXVIFS)
2853 break;
2854 rtnh = rtnh_next(rtnh, &remaining);
2855 }
2856
2857 return remaining > 0 ? -EINVAL : vifi;
2858}
2859
2860/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2861static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2862 struct mfcctl *mfcc, int *mrtsock,
2863 struct mr_table **mrtret,
2864 struct netlink_ext_ack *extack)
2865{
2866 struct net_device *dev = NULL;
2867 u32 tblid = RT_TABLE_DEFAULT;
2868 int ret, rem, iif = 0;
2869 struct mr_table *mrt;
2870 struct nlattr *attr;
2871 struct rtmsg *rtm;
2872
2873 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2874 rtm_ipmr_policy, extack);
2875 if (ret < 0)
2876 goto out;
2877 rtm = nlmsg_data(nlh);
2878
2879 ret = -EINVAL;
2880 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2881 rtm->rtm_type != RTN_MULTICAST ||
2882 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2883 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2884 goto out;
2885
2886 memset(mfcc, 0, sizeof(*mfcc));
2887 mfcc->mfcc_parent = -1;
2888 ret = 0;
2889 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2890 switch (nla_type(attr)) {
2891 case RTA_SRC:
2892 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2893 break;
2894 case RTA_DST:
2895 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2896 break;
2897 case RTA_IIF:
2898 iif = nla_get_u32(attr);
2899 break;
2900 case RTA_MULTIPATH:
2901 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2902 ret = -EINVAL;
2903 goto out;
2904 }
2905 break;
2906 case RTA_PREFSRC:
2907 ret = 1;
2908 break;
2909 case RTA_TABLE:
2910 tblid = nla_get_u32(attr);
2911 break;
2912 }
2913 }
2914
2915 rcu_read_lock();
2916
2917 mrt = __ipmr_get_table(net, tblid);
2918 if (!mrt) {
2919 ret = -ENOENT;
2920 goto unlock;
2921 }
2922
2923 if (iif) {
2924 dev = dev_get_by_index_rcu(net, iif);
2925 if (!dev) {
2926 ret = -ENODEV;
2927 goto unlock;
2928 }
2929
2930 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2931 }
2932
2933 *mrtret = mrt;
2934 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2935
2936unlock:
2937 rcu_read_unlock();
2938out:
2939 return ret;
2940}
2941
2942/* takes care of both newroute and delroute */
2943static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2944 struct netlink_ext_ack *extack)
2945{
2946 struct net *net = sock_net(skb->sk);
2947 int ret, mrtsock = 0, parent;
2948 struct mr_table *tbl = NULL;
2949 struct mfcctl mfcc;
2950
2951 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2952 if (ret < 0)
2953 return ret;
2954
2955 parent = ret ? mfcc.mfcc_parent : -1;
2956
2957 mutex_lock(&net->ipv4.mfc_mutex);
2958
2959 if (nlh->nlmsg_type == RTM_NEWROUTE)
2960 ret = ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2961 else
2962 ret = ipmr_mfc_delete(tbl, &mfcc, parent);
2963
2964 mutex_unlock(&net->ipv4.mfc_mutex);
2965
2966 return ret;
2967}
2968
2969static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2970{
2971 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2972 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN,
2973 READ_ONCE(mrt->cache_resolve_queue_len)) ||
2974 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2975 READ_ONCE(mrt->mroute_reg_vif_num)) ||
2976 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2977 READ_ONCE(mrt->mroute_do_assert)) ||
2978 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM,
2979 READ_ONCE(mrt->mroute_do_pim)) ||
2980 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2981 READ_ONCE(mrt->mroute_do_wrvifwhole)))
2982 return false;
2983
2984 return true;
2985}
2986
2987static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2988{
2989 struct net_device *vif_dev;
2990 struct nlattr *vif_nest;
2991 struct vif_device *vif;
2992
2993 vif = &mrt->vif_table[vifid];
2994 vif_dev = vif_dev_read(vif);
2995 /* if the VIF doesn't exist just continue */
2996 if (!vif_dev)
2997 return true;
2998
2999 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
3000 if (!vif_nest)
3001 return false;
3002
3003 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, READ_ONCE(vif_dev->ifindex)) ||
3004 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
3005 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
3006 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, READ_ONCE(vif->bytes_in),
3007 IPMRA_VIFA_PAD) ||
3008 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, READ_ONCE(vif->bytes_out),
3009 IPMRA_VIFA_PAD) ||
3010 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, READ_ONCE(vif->pkt_in),
3011 IPMRA_VIFA_PAD) ||
3012 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, READ_ONCE(vif->pkt_out),
3013 IPMRA_VIFA_PAD) ||
3014 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
3015 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
3016 nla_nest_cancel(skb, vif_nest);
3017 return false;
3018 }
3019 nla_nest_end(skb, vif_nest);
3020
3021 return true;
3022}
3023
3024static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
3025 struct netlink_ext_ack *extack)
3026{
3027 struct ifinfomsg *ifm;
3028
3029 ifm = nlmsg_payload(nlh, sizeof(*ifm));
3030 if (!ifm) {
3031 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
3032 return -EINVAL;
3033 }
3034
3035 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
3036 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
3037 return -EINVAL;
3038 }
3039
3040 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3041 ifm->ifi_change || ifm->ifi_index) {
3042 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
3043 return -EINVAL;
3044 }
3045
3046 return 0;
3047}
3048
3049static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
3050{
3051 struct net *net = sock_net(skb->sk);
3052 struct nlmsghdr *nlh = NULL;
3053 unsigned int t = 0, s_t;
3054 unsigned int e = 0, s_e;
3055 struct mr_table *mrt;
3056
3057 if (cb->strict_check) {
3058 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
3059
3060 if (err < 0)
3061 return err;
3062 }
3063
3064 s_t = cb->args[0];
3065 s_e = cb->args[1];
3066
3067 rcu_read_lock();
3068
3069 ipmr_for_each_table(mrt, net) {
3070 struct nlattr *vifs, *af;
3071 struct ifinfomsg *hdr;
3072 u32 i;
3073
3074 if (t < s_t)
3075 goto skip_table;
3076 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
3077 cb->nlh->nlmsg_seq, RTM_NEWLINK,
3078 sizeof(*hdr), NLM_F_MULTI);
3079 if (!nlh)
3080 break;
3081
3082 hdr = nlmsg_data(nlh);
3083 memset(hdr, 0, sizeof(*hdr));
3084 hdr->ifi_family = RTNL_FAMILY_IPMR;
3085
3086 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
3087 if (!af) {
3088 nlmsg_cancel(skb, nlh);
3089 goto out;
3090 }
3091
3092 if (!ipmr_fill_table(mrt, skb)) {
3093 nlmsg_cancel(skb, nlh);
3094 goto out;
3095 }
3096
3097 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
3098 if (!vifs) {
3099 nla_nest_end(skb, af);
3100 nlmsg_end(skb, nlh);
3101 goto out;
3102 }
3103 for (i = 0; i < READ_ONCE(mrt->maxvif); i++) {
3104 if (e < s_e)
3105 goto skip_entry;
3106 if (!ipmr_fill_vif(mrt, i, skb)) {
3107 nla_nest_end(skb, vifs);
3108 nla_nest_end(skb, af);
3109 nlmsg_end(skb, nlh);
3110 goto out;
3111 }
3112skip_entry:
3113 e++;
3114 }
3115 s_e = 0;
3116 e = 0;
3117 nla_nest_end(skb, vifs);
3118 nla_nest_end(skb, af);
3119 nlmsg_end(skb, nlh);
3120skip_table:
3121 t++;
3122 }
3123
3124out:
3125 rcu_read_unlock();
3126
3127 cb->args[1] = e;
3128 cb->args[0] = t;
3129
3130 return skb->len;
3131}
3132
3133#ifdef CONFIG_PROC_FS
3134/* The /proc interfaces to multicast routing :
3135 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
3136 */
3137
3138static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
3139 __acquires(RCU)
3140{
3141 struct mr_vif_iter *iter = seq->private;
3142 struct net *net = seq_file_net(seq);
3143 struct mr_table *mrt;
3144
3145 rcu_read_lock();
3146 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
3147 if (!mrt) {
3148 rcu_read_unlock();
3149 return ERR_PTR(-ENOENT);
3150 }
3151
3152 iter->mrt = mrt;
3153
3154 return mr_vif_seq_start(seq, pos);
3155}
3156
3157static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
3158 __releases(RCU)
3159{
3160 rcu_read_unlock();
3161}
3162
3163static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
3164{
3165 struct mr_vif_iter *iter = seq->private;
3166 struct mr_table *mrt = iter->mrt;
3167
3168 if (v == SEQ_START_TOKEN) {
3169 seq_puts(seq,
3170 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
3171 } else {
3172 const struct vif_device *vif = v;
3173 const struct net_device *vif_dev;
3174 const char *name;
3175
3176 vif_dev = vif_dev_read(vif);
3177 name = vif_dev ? vif_dev->name : "none";
3178 seq_printf(seq,
3179 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
3180 vif - mrt->vif_table,
3181 name, vif->bytes_in, vif->pkt_in,
3182 vif->bytes_out, vif->pkt_out,
3183 vif->flags, vif->local, vif->remote);
3184 }
3185 return 0;
3186}
3187
3188static const struct seq_operations ipmr_vif_seq_ops = {
3189 .start = ipmr_vif_seq_start,
3190 .next = mr_vif_seq_next,
3191 .stop = ipmr_vif_seq_stop,
3192 .show = ipmr_vif_seq_show,
3193};
3194
3195static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
3196{
3197 struct net *net = seq_file_net(seq);
3198 struct mr_table *mrt;
3199
3200 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
3201 if (!mrt)
3202 return ERR_PTR(-ENOENT);
3203
3204 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
3205}
3206
3207static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
3208{
3209 int n;
3210
3211 if (v == SEQ_START_TOKEN) {
3212 seq_puts(seq,
3213 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
3214 } else {
3215 const struct mfc_cache *mfc = v;
3216 const struct mr_mfc_iter *it = seq->private;
3217 const struct mr_table *mrt = it->mrt;
3218
3219 seq_printf(seq, "%08X %08X %-3hd",
3220 (__force u32) mfc->mfc_mcastgrp,
3221 (__force u32) mfc->mfc_origin,
3222 mfc->_c.mfc_parent);
3223
3224 if (it->cache != &mrt->mfc_unres_queue) {
3225 seq_printf(seq, " %8lu %8lu %8lu",
3226 atomic_long_read(&mfc->_c.mfc_un.res.pkt),
3227 atomic_long_read(&mfc->_c.mfc_un.res.bytes),
3228 atomic_long_read(&mfc->_c.mfc_un.res.wrong_if));
3229 for (n = mfc->_c.mfc_un.res.minvif;
3230 n < mfc->_c.mfc_un.res.maxvif; n++) {
3231 if (VIF_EXISTS(mrt, n) &&
3232 mfc->_c.mfc_un.res.ttls[n] < 255)
3233 seq_printf(seq,
3234 " %2d:%-3d",
3235 n, mfc->_c.mfc_un.res.ttls[n]);
3236 }
3237 } else {
3238 /* unresolved mfc_caches don't contain
3239 * pkt, bytes and wrong_if values
3240 */
3241 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3242 }
3243 seq_putc(seq, '\n');
3244 }
3245 return 0;
3246}
3247
3248static const struct seq_operations ipmr_mfc_seq_ops = {
3249 .start = ipmr_mfc_seq_start,
3250 .next = mr_mfc_seq_next,
3251 .stop = mr_mfc_seq_stop,
3252 .show = ipmr_mfc_seq_show,
3253};
3254#endif
3255
3256#ifdef CONFIG_IP_PIMSM_V2
3257static const struct net_protocol pim_protocol = {
3258 .handler = pim_rcv,
3259};
3260#endif
3261
3262static unsigned int ipmr_seq_read(const struct net *net)
3263{
3264 return atomic_read(&net->ipv4.ipmr_seq) + ipmr_rules_seq_read(net);
3265}
3266
3267static int ipmr_dump(struct net *net, struct notifier_block *nb,
3268 struct netlink_ext_ack *extack)
3269{
3270 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3271 ipmr_mr_table_iter, extack);
3272}
3273
3274static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3275 .family = RTNL_FAMILY_IPMR,
3276 .fib_seq_read = ipmr_seq_read,
3277 .fib_dump = ipmr_dump,
3278 .owner = THIS_MODULE,
3279};
3280
3281static int __net_init ipmr_notifier_init(struct net *net)
3282{
3283 struct fib_notifier_ops *ops;
3284
3285 atomic_set(&net->ipv4.ipmr_seq, 0);
3286
3287 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3288 if (IS_ERR(ops))
3289 return PTR_ERR(ops);
3290 net->ipv4.ipmr_notifier_ops = ops;
3291
3292 return 0;
3293}
3294
3295static void __net_exit ipmr_notifier_exit(struct net *net)
3296{
3297 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3298 net->ipv4.ipmr_notifier_ops = NULL;
3299}
3300
3301/* Setup for IP multicast routing */
3302static int __net_init ipmr_net_init(struct net *net)
3303{
3304 LIST_HEAD(dev_kill_list);
3305 int err;
3306
3307 mutex_init(&net->ipv4.mfc_mutex);
3308
3309 err = ipmr_notifier_init(net);
3310 if (err)
3311 goto ipmr_notifier_fail;
3312
3313 err = ipmr_rules_init(net);
3314 if (err < 0)
3315 goto ipmr_rules_fail;
3316
3317#ifdef CONFIG_PROC_FS
3318 err = -ENOMEM;
3319 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3320 sizeof(struct mr_vif_iter)))
3321 goto proc_vif_fail;
3322 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3323 sizeof(struct mr_mfc_iter)))
3324 goto proc_cache_fail;
3325#endif
3326 return 0;
3327
3328#ifdef CONFIG_PROC_FS
3329proc_cache_fail:
3330 remove_proc_entry("ip_mr_vif", net->proc_net);
3331proc_vif_fail:
3332 ipmr_rules_exit_rtnl(net, &dev_kill_list);
3333 ipmr_rules_exit(net);
3334#endif
3335ipmr_rules_fail:
3336 ipmr_notifier_exit(net);
3337ipmr_notifier_fail:
3338 return err;
3339}
3340
3341static void __net_exit ipmr_net_exit(struct net *net)
3342{
3343#ifdef CONFIG_PROC_FS
3344 remove_proc_entry("ip_mr_cache", net->proc_net);
3345 remove_proc_entry("ip_mr_vif", net->proc_net);
3346#endif
3347 ipmr_rules_exit(net);
3348 ipmr_notifier_exit(net);
3349}
3350
3351static void __net_exit ipmr_net_exit_rtnl(struct net *net,
3352 struct list_head *dev_kill_list)
3353{
3354 ipmr_rules_exit_rtnl(net, dev_kill_list);
3355}
3356
3357static struct pernet_operations ipmr_net_ops = {
3358 .init = ipmr_net_init,
3359 .exit = ipmr_net_exit,
3360 .exit_rtnl = ipmr_net_exit_rtnl,
3361};
3362
3363static const struct rtnl_msg_handler ipmr_rtnl_msg_handlers[] __initconst = {
3364 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETLINK,
3365 .dumpit = ipmr_rtm_dumplink, .flags = RTNL_FLAG_DUMP_UNLOCKED},
3366 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_NEWROUTE,
3367 .doit = ipmr_rtm_route, .flags = RTNL_FLAG_DOIT_UNLOCKED},
3368 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_DELROUTE,
3369 .doit = ipmr_rtm_route, .flags = RTNL_FLAG_DOIT_UNLOCKED},
3370 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETROUTE,
3371 .doit = ipmr_rtm_getroute, .dumpit = ipmr_rtm_dumproute,
3372 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
3373};
3374
3375int __init ip_mr_init(void)
3376{
3377 int err;
3378
3379 mrt_cachep = KMEM_CACHE(mfc_cache,
3380 SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT);
3381
3382 err = register_pernet_subsys(&ipmr_net_ops);
3383 if (err)
3384 goto reg_pernet_fail;
3385
3386 err = register_netdevice_notifier(&ip_mr_notifier);
3387 if (err)
3388 goto reg_notif_fail;
3389#ifdef CONFIG_IP_PIMSM_V2
3390 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3391 pr_err("%s: can't add PIM protocol\n", __func__);
3392 err = -EAGAIN;
3393 goto add_proto_fail;
3394 }
3395#endif
3396 rtnl_register_many(ipmr_rtnl_msg_handlers);
3397
3398 return 0;
3399
3400#ifdef CONFIG_IP_PIMSM_V2
3401add_proto_fail:
3402 unregister_netdevice_notifier(&ip_mr_notifier);
3403#endif
3404reg_notif_fail:
3405 unregister_pernet_subsys(&ipmr_net_ops);
3406reg_pernet_fail:
3407 kmem_cache_destroy(mrt_cachep);
3408 return err;
3409}