]> git.ipfire.org Git - thirdparty/linux.git/blob - net/ipv4/ipmr.c
Merge tag 'io_uring-5.7-2020-05-22' of git://git.kernel.dk/linux-block
[thirdparty/linux.git] / net / ipv4 / ipmr.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IP multicast routing support for mrouted 3.6/3.8
4 *
5 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6 * Linux Consultancy and Custom Driver Development
7 *
8 * Fixes:
9 * Michael Chastain : Incorrect size of copying.
10 * Alan Cox : Added the cache manager code
11 * Alan Cox : Fixed the clone/copy bug and device race.
12 * Mike McLagan : Routing by source
13 * Malcolm Beattie : Buffer handling fixes.
14 * Alexey Kuznetsov : Double buffer free and other fixes.
15 * SVR Anand : Fixed several multicast bugs and problems.
16 * Alexey Kuznetsov : Status, optimisations and more.
17 * Brad Parker : Better behaviour on mrouted upcall
18 * overflow.
19 * Carlos Picoto : PIMv1 Support
20 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
21 * Relax this requirement to work with older peers.
22 */
23
24 #include <linux/uaccess.h>
25 #include <linux/types.h>
26 #include <linux/cache.h>
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/mm.h>
30 #include <linux/kernel.h>
31 #include <linux/fcntl.h>
32 #include <linux/stat.h>
33 #include <linux/socket.h>
34 #include <linux/in.h>
35 #include <linux/inet.h>
36 #include <linux/netdevice.h>
37 #include <linux/inetdevice.h>
38 #include <linux/igmp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/mroute.h>
42 #include <linux/init.h>
43 #include <linux/if_ether.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <linux/skbuff.h>
49 #include <net/route.h>
50 #include <net/icmp.h>
51 #include <net/udp.h>
52 #include <net/raw.h>
53 #include <linux/notifier.h>
54 #include <linux/if_arp.h>
55 #include <linux/netfilter_ipv4.h>
56 #include <linux/compat.h>
57 #include <linux/export.h>
58 #include <linux/rhashtable.h>
59 #include <net/ip_tunnels.h>
60 #include <net/checksum.h>
61 #include <net/netlink.h>
62 #include <net/fib_rules.h>
63 #include <linux/netconf.h>
64 #include <net/rtnh.h>
65
66 #include <linux/nospec.h>
67
68 struct ipmr_rule {
69 struct fib_rule common;
70 };
71
72 struct ipmr_result {
73 struct mr_table *mrt;
74 };
75
76 /* Big lock, protecting vif table, mrt cache and mroute socket state.
77 * Note that the changes are semaphored via rtnl_lock.
78 */
79
80 static DEFINE_RWLOCK(mrt_lock);
81
82 /* Multicast router control variables */
83
84 /* Special spinlock for queue of unresolved entries */
85 static DEFINE_SPINLOCK(mfc_unres_lock);
86
87 /* We return to original Alan's scheme. Hash table of resolved
88 * entries is changed only in process context and protected
89 * with weak lock mrt_lock. Queue of unresolved entries is protected
90 * with strong spinlock mfc_unres_lock.
91 *
92 * In this case data path is free of exclusive locks at all.
93 */
94
95 static struct kmem_cache *mrt_cachep __ro_after_init;
96
97 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
98 static void ipmr_free_table(struct mr_table *mrt);
99
100 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
101 struct net_device *dev, struct sk_buff *skb,
102 struct mfc_cache *cache, int local);
103 static int ipmr_cache_report(struct mr_table *mrt,
104 struct sk_buff *pkt, vifi_t vifi, int assert);
105 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
106 int cmd);
107 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
108 static void mroute_clean_tables(struct mr_table *mrt, int flags);
109 static void ipmr_expire_process(struct timer_list *t);
110
111 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
112 #define ipmr_for_each_table(mrt, net) \
113 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
114 lockdep_rtnl_is_held() || \
115 list_empty(&net->ipv4.mr_tables))
116
117 static struct mr_table *ipmr_mr_table_iter(struct net *net,
118 struct mr_table *mrt)
119 {
120 struct mr_table *ret;
121
122 if (!mrt)
123 ret = list_entry_rcu(net->ipv4.mr_tables.next,
124 struct mr_table, list);
125 else
126 ret = list_entry_rcu(mrt->list.next,
127 struct mr_table, list);
128
129 if (&ret->list == &net->ipv4.mr_tables)
130 return NULL;
131 return ret;
132 }
133
134 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
135 {
136 struct mr_table *mrt;
137
138 ipmr_for_each_table(mrt, net) {
139 if (mrt->id == id)
140 return mrt;
141 }
142 return NULL;
143 }
144
145 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
146 struct mr_table **mrt)
147 {
148 int err;
149 struct ipmr_result res;
150 struct fib_lookup_arg arg = {
151 .result = &res,
152 .flags = FIB_LOOKUP_NOREF,
153 };
154
155 /* update flow if oif or iif point to device enslaved to l3mdev */
156 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
157
158 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
159 flowi4_to_flowi(flp4), 0, &arg);
160 if (err < 0)
161 return err;
162 *mrt = res.mrt;
163 return 0;
164 }
165
166 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
167 int flags, struct fib_lookup_arg *arg)
168 {
169 struct ipmr_result *res = arg->result;
170 struct mr_table *mrt;
171
172 switch (rule->action) {
173 case FR_ACT_TO_TBL:
174 break;
175 case FR_ACT_UNREACHABLE:
176 return -ENETUNREACH;
177 case FR_ACT_PROHIBIT:
178 return -EACCES;
179 case FR_ACT_BLACKHOLE:
180 default:
181 return -EINVAL;
182 }
183
184 arg->table = fib_rule_get_table(rule, arg);
185
186 mrt = ipmr_get_table(rule->fr_net, arg->table);
187 if (!mrt)
188 return -EAGAIN;
189 res->mrt = mrt;
190 return 0;
191 }
192
193 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
194 {
195 return 1;
196 }
197
198 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
199 FRA_GENERIC_POLICY,
200 };
201
202 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
203 struct fib_rule_hdr *frh, struct nlattr **tb,
204 struct netlink_ext_ack *extack)
205 {
206 return 0;
207 }
208
209 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
210 struct nlattr **tb)
211 {
212 return 1;
213 }
214
215 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
216 struct fib_rule_hdr *frh)
217 {
218 frh->dst_len = 0;
219 frh->src_len = 0;
220 frh->tos = 0;
221 return 0;
222 }
223
224 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
225 .family = RTNL_FAMILY_IPMR,
226 .rule_size = sizeof(struct ipmr_rule),
227 .addr_size = sizeof(u32),
228 .action = ipmr_rule_action,
229 .match = ipmr_rule_match,
230 .configure = ipmr_rule_configure,
231 .compare = ipmr_rule_compare,
232 .fill = ipmr_rule_fill,
233 .nlgroup = RTNLGRP_IPV4_RULE,
234 .policy = ipmr_rule_policy,
235 .owner = THIS_MODULE,
236 };
237
238 static int __net_init ipmr_rules_init(struct net *net)
239 {
240 struct fib_rules_ops *ops;
241 struct mr_table *mrt;
242 int err;
243
244 ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 if (IS_ERR(ops))
246 return PTR_ERR(ops);
247
248 INIT_LIST_HEAD(&net->ipv4.mr_tables);
249
250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 if (IS_ERR(mrt)) {
252 err = PTR_ERR(mrt);
253 goto err1;
254 }
255
256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
257 if (err < 0)
258 goto err2;
259
260 net->ipv4.mr_rules_ops = ops;
261 return 0;
262
263 err2:
264 ipmr_free_table(mrt);
265 err1:
266 fib_rules_unregister(ops);
267 return err;
268 }
269
270 static void __net_exit ipmr_rules_exit(struct net *net)
271 {
272 struct mr_table *mrt, *next;
273
274 rtnl_lock();
275 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
276 list_del(&mrt->list);
277 ipmr_free_table(mrt);
278 }
279 fib_rules_unregister(net->ipv4.mr_rules_ops);
280 rtnl_unlock();
281 }
282
283 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
284 struct netlink_ext_ack *extack)
285 {
286 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
287 }
288
289 static unsigned int ipmr_rules_seq_read(struct net *net)
290 {
291 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
292 }
293
294 bool ipmr_rule_default(const struct fib_rule *rule)
295 {
296 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
297 }
298 EXPORT_SYMBOL(ipmr_rule_default);
299 #else
300 #define ipmr_for_each_table(mrt, net) \
301 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
302
303 static struct mr_table *ipmr_mr_table_iter(struct net *net,
304 struct mr_table *mrt)
305 {
306 if (!mrt)
307 return net->ipv4.mrt;
308 return NULL;
309 }
310
311 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
312 {
313 return net->ipv4.mrt;
314 }
315
316 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
317 struct mr_table **mrt)
318 {
319 *mrt = net->ipv4.mrt;
320 return 0;
321 }
322
323 static int __net_init ipmr_rules_init(struct net *net)
324 {
325 struct mr_table *mrt;
326
327 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
328 if (IS_ERR(mrt))
329 return PTR_ERR(mrt);
330 net->ipv4.mrt = mrt;
331 return 0;
332 }
333
334 static void __net_exit ipmr_rules_exit(struct net *net)
335 {
336 rtnl_lock();
337 ipmr_free_table(net->ipv4.mrt);
338 net->ipv4.mrt = NULL;
339 rtnl_unlock();
340 }
341
342 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
343 struct netlink_ext_ack *extack)
344 {
345 return 0;
346 }
347
348 static unsigned int ipmr_rules_seq_read(struct net *net)
349 {
350 return 0;
351 }
352
353 bool ipmr_rule_default(const struct fib_rule *rule)
354 {
355 return true;
356 }
357 EXPORT_SYMBOL(ipmr_rule_default);
358 #endif
359
360 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
361 const void *ptr)
362 {
363 const struct mfc_cache_cmp_arg *cmparg = arg->key;
364 struct mfc_cache *c = (struct mfc_cache *)ptr;
365
366 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
367 cmparg->mfc_origin != c->mfc_origin;
368 }
369
370 static const struct rhashtable_params ipmr_rht_params = {
371 .head_offset = offsetof(struct mr_mfc, mnode),
372 .key_offset = offsetof(struct mfc_cache, cmparg),
373 .key_len = sizeof(struct mfc_cache_cmp_arg),
374 .nelem_hint = 3,
375 .obj_cmpfn = ipmr_hash_cmp,
376 .automatic_shrinking = true,
377 };
378
379 static void ipmr_new_table_set(struct mr_table *mrt,
380 struct net *net)
381 {
382 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
383 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384 #endif
385 }
386
387 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
388 .mfc_mcastgrp = htonl(INADDR_ANY),
389 .mfc_origin = htonl(INADDR_ANY),
390 };
391
392 static struct mr_table_ops ipmr_mr_table_ops = {
393 .rht_params = &ipmr_rht_params,
394 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
395 };
396
397 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
398 {
399 struct mr_table *mrt;
400
401 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
402 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
403 return ERR_PTR(-EINVAL);
404
405 mrt = ipmr_get_table(net, id);
406 if (mrt)
407 return mrt;
408
409 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
410 ipmr_expire_process, ipmr_new_table_set);
411 }
412
413 static void ipmr_free_table(struct mr_table *mrt)
414 {
415 del_timer_sync(&mrt->ipmr_expire_timer);
416 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
417 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
418 rhltable_destroy(&mrt->mfc_hash);
419 kfree(mrt);
420 }
421
422 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
423
424 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
425 {
426 struct net *net = dev_net(dev);
427
428 dev_close(dev);
429
430 dev = __dev_get_by_name(net, "tunl0");
431 if (dev) {
432 const struct net_device_ops *ops = dev->netdev_ops;
433 struct ifreq ifr;
434 struct ip_tunnel_parm p;
435
436 memset(&p, 0, sizeof(p));
437 p.iph.daddr = v->vifc_rmt_addr.s_addr;
438 p.iph.saddr = v->vifc_lcl_addr.s_addr;
439 p.iph.version = 4;
440 p.iph.ihl = 5;
441 p.iph.protocol = IPPROTO_IPIP;
442 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
443 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
444
445 if (ops->ndo_do_ioctl) {
446 mm_segment_t oldfs = get_fs();
447
448 set_fs(KERNEL_DS);
449 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
450 set_fs(oldfs);
451 }
452 }
453 }
454
455 /* Initialize ipmr pimreg/tunnel in_device */
456 static bool ipmr_init_vif_indev(const struct net_device *dev)
457 {
458 struct in_device *in_dev;
459
460 ASSERT_RTNL();
461
462 in_dev = __in_dev_get_rtnl(dev);
463 if (!in_dev)
464 return false;
465 ipv4_devconf_setall(in_dev);
466 neigh_parms_data_state_setall(in_dev->arp_parms);
467 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
468
469 return true;
470 }
471
472 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
473 {
474 struct net_device *dev;
475
476 dev = __dev_get_by_name(net, "tunl0");
477
478 if (dev) {
479 const struct net_device_ops *ops = dev->netdev_ops;
480 int err;
481 struct ifreq ifr;
482 struct ip_tunnel_parm p;
483
484 memset(&p, 0, sizeof(p));
485 p.iph.daddr = v->vifc_rmt_addr.s_addr;
486 p.iph.saddr = v->vifc_lcl_addr.s_addr;
487 p.iph.version = 4;
488 p.iph.ihl = 5;
489 p.iph.protocol = IPPROTO_IPIP;
490 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
491 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
492
493 if (ops->ndo_do_ioctl) {
494 mm_segment_t oldfs = get_fs();
495
496 set_fs(KERNEL_DS);
497 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
498 set_fs(oldfs);
499 } else {
500 err = -EOPNOTSUPP;
501 }
502 dev = NULL;
503
504 if (err == 0 &&
505 (dev = __dev_get_by_name(net, p.name)) != NULL) {
506 dev->flags |= IFF_MULTICAST;
507 if (!ipmr_init_vif_indev(dev))
508 goto failure;
509 if (dev_open(dev, NULL))
510 goto failure;
511 dev_hold(dev);
512 }
513 }
514 return dev;
515
516 failure:
517 unregister_netdevice(dev);
518 return NULL;
519 }
520
521 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
522 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
523 {
524 struct net *net = dev_net(dev);
525 struct mr_table *mrt;
526 struct flowi4 fl4 = {
527 .flowi4_oif = dev->ifindex,
528 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
529 .flowi4_mark = skb->mark,
530 };
531 int err;
532
533 err = ipmr_fib_lookup(net, &fl4, &mrt);
534 if (err < 0) {
535 kfree_skb(skb);
536 return err;
537 }
538
539 read_lock(&mrt_lock);
540 dev->stats.tx_bytes += skb->len;
541 dev->stats.tx_packets++;
542 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
543 read_unlock(&mrt_lock);
544 kfree_skb(skb);
545 return NETDEV_TX_OK;
546 }
547
548 static int reg_vif_get_iflink(const struct net_device *dev)
549 {
550 return 0;
551 }
552
553 static const struct net_device_ops reg_vif_netdev_ops = {
554 .ndo_start_xmit = reg_vif_xmit,
555 .ndo_get_iflink = reg_vif_get_iflink,
556 };
557
558 static void reg_vif_setup(struct net_device *dev)
559 {
560 dev->type = ARPHRD_PIMREG;
561 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
562 dev->flags = IFF_NOARP;
563 dev->netdev_ops = &reg_vif_netdev_ops;
564 dev->needs_free_netdev = true;
565 dev->features |= NETIF_F_NETNS_LOCAL;
566 }
567
568 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
569 {
570 struct net_device *dev;
571 char name[IFNAMSIZ];
572
573 if (mrt->id == RT_TABLE_DEFAULT)
574 sprintf(name, "pimreg");
575 else
576 sprintf(name, "pimreg%u", mrt->id);
577
578 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
579
580 if (!dev)
581 return NULL;
582
583 dev_net_set(dev, net);
584
585 if (register_netdevice(dev)) {
586 free_netdev(dev);
587 return NULL;
588 }
589
590 if (!ipmr_init_vif_indev(dev))
591 goto failure;
592 if (dev_open(dev, NULL))
593 goto failure;
594
595 dev_hold(dev);
596
597 return dev;
598
599 failure:
600 unregister_netdevice(dev);
601 return NULL;
602 }
603
604 /* called with rcu_read_lock() */
605 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
606 unsigned int pimlen)
607 {
608 struct net_device *reg_dev = NULL;
609 struct iphdr *encap;
610
611 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
612 /* Check that:
613 * a. packet is really sent to a multicast group
614 * b. packet is not a NULL-REGISTER
615 * c. packet is not truncated
616 */
617 if (!ipv4_is_multicast(encap->daddr) ||
618 encap->tot_len == 0 ||
619 ntohs(encap->tot_len) + pimlen > skb->len)
620 return 1;
621
622 read_lock(&mrt_lock);
623 if (mrt->mroute_reg_vif_num >= 0)
624 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
625 read_unlock(&mrt_lock);
626
627 if (!reg_dev)
628 return 1;
629
630 skb->mac_header = skb->network_header;
631 skb_pull(skb, (u8 *)encap - skb->data);
632 skb_reset_network_header(skb);
633 skb->protocol = htons(ETH_P_IP);
634 skb->ip_summed = CHECKSUM_NONE;
635
636 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
637
638 netif_rx(skb);
639
640 return NET_RX_SUCCESS;
641 }
642 #else
643 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
644 {
645 return NULL;
646 }
647 #endif
648
649 static int call_ipmr_vif_entry_notifiers(struct net *net,
650 enum fib_event_type event_type,
651 struct vif_device *vif,
652 vifi_t vif_index, u32 tb_id)
653 {
654 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
655 vif, vif_index, tb_id,
656 &net->ipv4.ipmr_seq);
657 }
658
659 static int call_ipmr_mfc_entry_notifiers(struct net *net,
660 enum fib_event_type event_type,
661 struct mfc_cache *mfc, u32 tb_id)
662 {
663 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
664 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
665 }
666
667 /**
668 * vif_delete - Delete a VIF entry
669 * @notify: Set to 1, if the caller is a notifier_call
670 */
671 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
672 struct list_head *head)
673 {
674 struct net *net = read_pnet(&mrt->net);
675 struct vif_device *v;
676 struct net_device *dev;
677 struct in_device *in_dev;
678
679 if (vifi < 0 || vifi >= mrt->maxvif)
680 return -EADDRNOTAVAIL;
681
682 v = &mrt->vif_table[vifi];
683
684 if (VIF_EXISTS(mrt, vifi))
685 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
686 mrt->id);
687
688 write_lock_bh(&mrt_lock);
689 dev = v->dev;
690 v->dev = NULL;
691
692 if (!dev) {
693 write_unlock_bh(&mrt_lock);
694 return -EADDRNOTAVAIL;
695 }
696
697 if (vifi == mrt->mroute_reg_vif_num)
698 mrt->mroute_reg_vif_num = -1;
699
700 if (vifi + 1 == mrt->maxvif) {
701 int tmp;
702
703 for (tmp = vifi - 1; tmp >= 0; tmp--) {
704 if (VIF_EXISTS(mrt, tmp))
705 break;
706 }
707 mrt->maxvif = tmp+1;
708 }
709
710 write_unlock_bh(&mrt_lock);
711
712 dev_set_allmulti(dev, -1);
713
714 in_dev = __in_dev_get_rtnl(dev);
715 if (in_dev) {
716 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
717 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
718 NETCONFA_MC_FORWARDING,
719 dev->ifindex, &in_dev->cnf);
720 ip_rt_multicast_event(in_dev);
721 }
722
723 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
724 unregister_netdevice_queue(dev, head);
725
726 dev_put(dev);
727 return 0;
728 }
729
730 static void ipmr_cache_free_rcu(struct rcu_head *head)
731 {
732 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
733
734 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
735 }
736
737 static void ipmr_cache_free(struct mfc_cache *c)
738 {
739 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
740 }
741
742 /* Destroy an unresolved cache entry, killing queued skbs
743 * and reporting error to netlink readers.
744 */
745 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
746 {
747 struct net *net = read_pnet(&mrt->net);
748 struct sk_buff *skb;
749 struct nlmsgerr *e;
750
751 atomic_dec(&mrt->cache_resolve_queue_len);
752
753 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
754 if (ip_hdr(skb)->version == 0) {
755 struct nlmsghdr *nlh = skb_pull(skb,
756 sizeof(struct iphdr));
757 nlh->nlmsg_type = NLMSG_ERROR;
758 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
759 skb_trim(skb, nlh->nlmsg_len);
760 e = nlmsg_data(nlh);
761 e->error = -ETIMEDOUT;
762 memset(&e->msg, 0, sizeof(e->msg));
763
764 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
765 } else {
766 kfree_skb(skb);
767 }
768 }
769
770 ipmr_cache_free(c);
771 }
772
773 /* Timer process for the unresolved queue. */
774 static void ipmr_expire_process(struct timer_list *t)
775 {
776 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
777 struct mr_mfc *c, *next;
778 unsigned long expires;
779 unsigned long now;
780
781 if (!spin_trylock(&mfc_unres_lock)) {
782 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
783 return;
784 }
785
786 if (list_empty(&mrt->mfc_unres_queue))
787 goto out;
788
789 now = jiffies;
790 expires = 10*HZ;
791
792 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
793 if (time_after(c->mfc_un.unres.expires, now)) {
794 unsigned long interval = c->mfc_un.unres.expires - now;
795 if (interval < expires)
796 expires = interval;
797 continue;
798 }
799
800 list_del(&c->list);
801 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
802 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
803 }
804
805 if (!list_empty(&mrt->mfc_unres_queue))
806 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
807
808 out:
809 spin_unlock(&mfc_unres_lock);
810 }
811
812 /* Fill oifs list. It is called under write locked mrt_lock. */
813 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
814 unsigned char *ttls)
815 {
816 int vifi;
817
818 cache->mfc_un.res.minvif = MAXVIFS;
819 cache->mfc_un.res.maxvif = 0;
820 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
821
822 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
823 if (VIF_EXISTS(mrt, vifi) &&
824 ttls[vifi] && ttls[vifi] < 255) {
825 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
826 if (cache->mfc_un.res.minvif > vifi)
827 cache->mfc_un.res.minvif = vifi;
828 if (cache->mfc_un.res.maxvif <= vifi)
829 cache->mfc_un.res.maxvif = vifi + 1;
830 }
831 }
832 cache->mfc_un.res.lastuse = jiffies;
833 }
834
835 static int vif_add(struct net *net, struct mr_table *mrt,
836 struct vifctl *vifc, int mrtsock)
837 {
838 struct netdev_phys_item_id ppid = { };
839 int vifi = vifc->vifc_vifi;
840 struct vif_device *v = &mrt->vif_table[vifi];
841 struct net_device *dev;
842 struct in_device *in_dev;
843 int err;
844
845 /* Is vif busy ? */
846 if (VIF_EXISTS(mrt, vifi))
847 return -EADDRINUSE;
848
849 switch (vifc->vifc_flags) {
850 case VIFF_REGISTER:
851 if (!ipmr_pimsm_enabled())
852 return -EINVAL;
853 /* Special Purpose VIF in PIM
854 * All the packets will be sent to the daemon
855 */
856 if (mrt->mroute_reg_vif_num >= 0)
857 return -EADDRINUSE;
858 dev = ipmr_reg_vif(net, mrt);
859 if (!dev)
860 return -ENOBUFS;
861 err = dev_set_allmulti(dev, 1);
862 if (err) {
863 unregister_netdevice(dev);
864 dev_put(dev);
865 return err;
866 }
867 break;
868 case VIFF_TUNNEL:
869 dev = ipmr_new_tunnel(net, vifc);
870 if (!dev)
871 return -ENOBUFS;
872 err = dev_set_allmulti(dev, 1);
873 if (err) {
874 ipmr_del_tunnel(dev, vifc);
875 dev_put(dev);
876 return err;
877 }
878 break;
879 case VIFF_USE_IFINDEX:
880 case 0:
881 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
882 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
883 if (dev && !__in_dev_get_rtnl(dev)) {
884 dev_put(dev);
885 return -EADDRNOTAVAIL;
886 }
887 } else {
888 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
889 }
890 if (!dev)
891 return -EADDRNOTAVAIL;
892 err = dev_set_allmulti(dev, 1);
893 if (err) {
894 dev_put(dev);
895 return err;
896 }
897 break;
898 default:
899 return -EINVAL;
900 }
901
902 in_dev = __in_dev_get_rtnl(dev);
903 if (!in_dev) {
904 dev_put(dev);
905 return -EADDRNOTAVAIL;
906 }
907 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
908 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
909 dev->ifindex, &in_dev->cnf);
910 ip_rt_multicast_event(in_dev);
911
912 /* Fill in the VIF structures */
913 vif_device_init(v, dev, vifc->vifc_rate_limit,
914 vifc->vifc_threshold,
915 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
916 (VIFF_TUNNEL | VIFF_REGISTER));
917
918 err = dev_get_port_parent_id(dev, &ppid, true);
919 if (err == 0) {
920 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
921 v->dev_parent_id.id_len = ppid.id_len;
922 } else {
923 v->dev_parent_id.id_len = 0;
924 }
925
926 v->local = vifc->vifc_lcl_addr.s_addr;
927 v->remote = vifc->vifc_rmt_addr.s_addr;
928
929 /* And finish update writing critical data */
930 write_lock_bh(&mrt_lock);
931 v->dev = dev;
932 if (v->flags & VIFF_REGISTER)
933 mrt->mroute_reg_vif_num = vifi;
934 if (vifi+1 > mrt->maxvif)
935 mrt->maxvif = vifi+1;
936 write_unlock_bh(&mrt_lock);
937 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
938 return 0;
939 }
940
941 /* called with rcu_read_lock() */
942 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
943 __be32 origin,
944 __be32 mcastgrp)
945 {
946 struct mfc_cache_cmp_arg arg = {
947 .mfc_mcastgrp = mcastgrp,
948 .mfc_origin = origin
949 };
950
951 return mr_mfc_find(mrt, &arg);
952 }
953
954 /* Look for a (*,G) entry */
955 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
956 __be32 mcastgrp, int vifi)
957 {
958 struct mfc_cache_cmp_arg arg = {
959 .mfc_mcastgrp = mcastgrp,
960 .mfc_origin = htonl(INADDR_ANY)
961 };
962
963 if (mcastgrp == htonl(INADDR_ANY))
964 return mr_mfc_find_any_parent(mrt, vifi);
965 return mr_mfc_find_any(mrt, vifi, &arg);
966 }
967
968 /* Look for a (S,G,iif) entry if parent != -1 */
969 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
970 __be32 origin, __be32 mcastgrp,
971 int parent)
972 {
973 struct mfc_cache_cmp_arg arg = {
974 .mfc_mcastgrp = mcastgrp,
975 .mfc_origin = origin,
976 };
977
978 return mr_mfc_find_parent(mrt, &arg, parent);
979 }
980
981 /* Allocate a multicast cache entry */
982 static struct mfc_cache *ipmr_cache_alloc(void)
983 {
984 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
985
986 if (c) {
987 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
988 c->_c.mfc_un.res.minvif = MAXVIFS;
989 c->_c.free = ipmr_cache_free_rcu;
990 refcount_set(&c->_c.mfc_un.res.refcount, 1);
991 }
992 return c;
993 }
994
995 static struct mfc_cache *ipmr_cache_alloc_unres(void)
996 {
997 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
998
999 if (c) {
1000 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1001 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1002 }
1003 return c;
1004 }
1005
1006 /* A cache entry has gone into a resolved state from queued */
1007 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1008 struct mfc_cache *uc, struct mfc_cache *c)
1009 {
1010 struct sk_buff *skb;
1011 struct nlmsgerr *e;
1012
1013 /* Play the pending entries through our router */
1014 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1015 if (ip_hdr(skb)->version == 0) {
1016 struct nlmsghdr *nlh = skb_pull(skb,
1017 sizeof(struct iphdr));
1018
1019 if (mr_fill_mroute(mrt, skb, &c->_c,
1020 nlmsg_data(nlh)) > 0) {
1021 nlh->nlmsg_len = skb_tail_pointer(skb) -
1022 (u8 *)nlh;
1023 } else {
1024 nlh->nlmsg_type = NLMSG_ERROR;
1025 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1026 skb_trim(skb, nlh->nlmsg_len);
1027 e = nlmsg_data(nlh);
1028 e->error = -EMSGSIZE;
1029 memset(&e->msg, 0, sizeof(e->msg));
1030 }
1031
1032 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1033 } else {
1034 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1035 }
1036 }
1037 }
1038
1039 /* Bounce a cache query up to mrouted and netlink.
1040 *
1041 * Called under mrt_lock.
1042 */
1043 static int ipmr_cache_report(struct mr_table *mrt,
1044 struct sk_buff *pkt, vifi_t vifi, int assert)
1045 {
1046 const int ihl = ip_hdrlen(pkt);
1047 struct sock *mroute_sk;
1048 struct igmphdr *igmp;
1049 struct igmpmsg *msg;
1050 struct sk_buff *skb;
1051 int ret;
1052
1053 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1054 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1055 else
1056 skb = alloc_skb(128, GFP_ATOMIC);
1057
1058 if (!skb)
1059 return -ENOBUFS;
1060
1061 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1062 /* Ugly, but we have no choice with this interface.
1063 * Duplicate old header, fix ihl, length etc.
1064 * And all this only to mangle msg->im_msgtype and
1065 * to set msg->im_mbz to "mbz" :-)
1066 */
1067 skb_push(skb, sizeof(struct iphdr));
1068 skb_reset_network_header(skb);
1069 skb_reset_transport_header(skb);
1070 msg = (struct igmpmsg *)skb_network_header(skb);
1071 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1072 msg->im_msgtype = assert;
1073 msg->im_mbz = 0;
1074 if (assert == IGMPMSG_WRVIFWHOLE)
1075 msg->im_vif = vifi;
1076 else
1077 msg->im_vif = mrt->mroute_reg_vif_num;
1078 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1079 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1080 sizeof(struct iphdr));
1081 } else {
1082 /* Copy the IP header */
1083 skb_set_network_header(skb, skb->len);
1084 skb_put(skb, ihl);
1085 skb_copy_to_linear_data(skb, pkt->data, ihl);
1086 /* Flag to the kernel this is a route add */
1087 ip_hdr(skb)->protocol = 0;
1088 msg = (struct igmpmsg *)skb_network_header(skb);
1089 msg->im_vif = vifi;
1090 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1091 /* Add our header */
1092 igmp = skb_put(skb, sizeof(struct igmphdr));
1093 igmp->type = assert;
1094 msg->im_msgtype = assert;
1095 igmp->code = 0;
1096 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1097 skb->transport_header = skb->network_header;
1098 }
1099
1100 rcu_read_lock();
1101 mroute_sk = rcu_dereference(mrt->mroute_sk);
1102 if (!mroute_sk) {
1103 rcu_read_unlock();
1104 kfree_skb(skb);
1105 return -EINVAL;
1106 }
1107
1108 igmpmsg_netlink_event(mrt, skb);
1109
1110 /* Deliver to mrouted */
1111 ret = sock_queue_rcv_skb(mroute_sk, skb);
1112 rcu_read_unlock();
1113 if (ret < 0) {
1114 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1115 kfree_skb(skb);
1116 }
1117
1118 return ret;
1119 }
1120
1121 /* Queue a packet for resolution. It gets locked cache entry! */
1122 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1123 struct sk_buff *skb, struct net_device *dev)
1124 {
1125 const struct iphdr *iph = ip_hdr(skb);
1126 struct mfc_cache *c;
1127 bool found = false;
1128 int err;
1129
1130 spin_lock_bh(&mfc_unres_lock);
1131 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1132 if (c->mfc_mcastgrp == iph->daddr &&
1133 c->mfc_origin == iph->saddr) {
1134 found = true;
1135 break;
1136 }
1137 }
1138
1139 if (!found) {
1140 /* Create a new entry if allowable */
1141 c = ipmr_cache_alloc_unres();
1142 if (!c) {
1143 spin_unlock_bh(&mfc_unres_lock);
1144
1145 kfree_skb(skb);
1146 return -ENOBUFS;
1147 }
1148
1149 /* Fill in the new cache entry */
1150 c->_c.mfc_parent = -1;
1151 c->mfc_origin = iph->saddr;
1152 c->mfc_mcastgrp = iph->daddr;
1153
1154 /* Reflect first query at mrouted. */
1155 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1156
1157 if (err < 0) {
1158 /* If the report failed throw the cache entry
1159 out - Brad Parker
1160 */
1161 spin_unlock_bh(&mfc_unres_lock);
1162
1163 ipmr_cache_free(c);
1164 kfree_skb(skb);
1165 return err;
1166 }
1167
1168 atomic_inc(&mrt->cache_resolve_queue_len);
1169 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1170 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1171
1172 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1173 mod_timer(&mrt->ipmr_expire_timer,
1174 c->_c.mfc_un.unres.expires);
1175 }
1176
1177 /* See if we can append the packet */
1178 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1179 kfree_skb(skb);
1180 err = -ENOBUFS;
1181 } else {
1182 if (dev) {
1183 skb->dev = dev;
1184 skb->skb_iif = dev->ifindex;
1185 }
1186 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1187 err = 0;
1188 }
1189
1190 spin_unlock_bh(&mfc_unres_lock);
1191 return err;
1192 }
1193
1194 /* MFC cache manipulation by user space mroute daemon */
1195
1196 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1197 {
1198 struct net *net = read_pnet(&mrt->net);
1199 struct mfc_cache *c;
1200
1201 /* The entries are added/deleted only under RTNL */
1202 rcu_read_lock();
1203 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1204 mfc->mfcc_mcastgrp.s_addr, parent);
1205 rcu_read_unlock();
1206 if (!c)
1207 return -ENOENT;
1208 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1209 list_del_rcu(&c->_c.list);
1210 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1211 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1212 mr_cache_put(&c->_c);
1213
1214 return 0;
1215 }
1216
1217 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1218 struct mfcctl *mfc, int mrtsock, int parent)
1219 {
1220 struct mfc_cache *uc, *c;
1221 struct mr_mfc *_uc;
1222 bool found;
1223 int ret;
1224
1225 if (mfc->mfcc_parent >= MAXVIFS)
1226 return -ENFILE;
1227
1228 /* The entries are added/deleted only under RTNL */
1229 rcu_read_lock();
1230 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1231 mfc->mfcc_mcastgrp.s_addr, parent);
1232 rcu_read_unlock();
1233 if (c) {
1234 write_lock_bh(&mrt_lock);
1235 c->_c.mfc_parent = mfc->mfcc_parent;
1236 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1237 if (!mrtsock)
1238 c->_c.mfc_flags |= MFC_STATIC;
1239 write_unlock_bh(&mrt_lock);
1240 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1241 mrt->id);
1242 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1243 return 0;
1244 }
1245
1246 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1247 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1248 return -EINVAL;
1249
1250 c = ipmr_cache_alloc();
1251 if (!c)
1252 return -ENOMEM;
1253
1254 c->mfc_origin = mfc->mfcc_origin.s_addr;
1255 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1256 c->_c.mfc_parent = mfc->mfcc_parent;
1257 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1258 if (!mrtsock)
1259 c->_c.mfc_flags |= MFC_STATIC;
1260
1261 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1262 ipmr_rht_params);
1263 if (ret) {
1264 pr_err("ipmr: rhtable insert error %d\n", ret);
1265 ipmr_cache_free(c);
1266 return ret;
1267 }
1268 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1269 /* Check to see if we resolved a queued list. If so we
1270 * need to send on the frames and tidy up.
1271 */
1272 found = false;
1273 spin_lock_bh(&mfc_unres_lock);
1274 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1275 uc = (struct mfc_cache *)_uc;
1276 if (uc->mfc_origin == c->mfc_origin &&
1277 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1278 list_del(&_uc->list);
1279 atomic_dec(&mrt->cache_resolve_queue_len);
1280 found = true;
1281 break;
1282 }
1283 }
1284 if (list_empty(&mrt->mfc_unres_queue))
1285 del_timer(&mrt->ipmr_expire_timer);
1286 spin_unlock_bh(&mfc_unres_lock);
1287
1288 if (found) {
1289 ipmr_cache_resolve(net, mrt, uc, c);
1290 ipmr_cache_free(uc);
1291 }
1292 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1293 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1294 return 0;
1295 }
1296
1297 /* Close the multicast socket, and clear the vif tables etc */
1298 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1299 {
1300 struct net *net = read_pnet(&mrt->net);
1301 struct mr_mfc *c, *tmp;
1302 struct mfc_cache *cache;
1303 LIST_HEAD(list);
1304 int i;
1305
1306 /* Shut down all active vif entries */
1307 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1308 for (i = 0; i < mrt->maxvif; i++) {
1309 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1310 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1311 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1312 continue;
1313 vif_delete(mrt, i, 0, &list);
1314 }
1315 unregister_netdevice_many(&list);
1316 }
1317
1318 /* Wipe the cache */
1319 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1320 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1321 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1322 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1323 continue;
1324 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1325 list_del_rcu(&c->list);
1326 cache = (struct mfc_cache *)c;
1327 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1328 mrt->id);
1329 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1330 mr_cache_put(c);
1331 }
1332 }
1333
1334 if (flags & MRT_FLUSH_MFC) {
1335 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1336 spin_lock_bh(&mfc_unres_lock);
1337 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1338 list_del(&c->list);
1339 cache = (struct mfc_cache *)c;
1340 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1341 ipmr_destroy_unres(mrt, cache);
1342 }
1343 spin_unlock_bh(&mfc_unres_lock);
1344 }
1345 }
1346 }
1347
1348 /* called from ip_ra_control(), before an RCU grace period,
1349 * we dont need to call synchronize_rcu() here
1350 */
1351 static void mrtsock_destruct(struct sock *sk)
1352 {
1353 struct net *net = sock_net(sk);
1354 struct mr_table *mrt;
1355
1356 rtnl_lock();
1357 ipmr_for_each_table(mrt, net) {
1358 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1359 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1360 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1361 NETCONFA_MC_FORWARDING,
1362 NETCONFA_IFINDEX_ALL,
1363 net->ipv4.devconf_all);
1364 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1365 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1366 }
1367 }
1368 rtnl_unlock();
1369 }
1370
1371 /* Socket options and virtual interface manipulation. The whole
1372 * virtual interface system is a complete heap, but unfortunately
1373 * that's how BSD mrouted happens to think. Maybe one day with a proper
1374 * MOSPF/PIM router set up we can clean this up.
1375 */
1376
1377 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1378 unsigned int optlen)
1379 {
1380 struct net *net = sock_net(sk);
1381 int val, ret = 0, parent = 0;
1382 struct mr_table *mrt;
1383 struct vifctl vif;
1384 struct mfcctl mfc;
1385 bool do_wrvifwhole;
1386 u32 uval;
1387
1388 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1389 rtnl_lock();
1390 if (sk->sk_type != SOCK_RAW ||
1391 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1392 ret = -EOPNOTSUPP;
1393 goto out_unlock;
1394 }
1395
1396 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1397 if (!mrt) {
1398 ret = -ENOENT;
1399 goto out_unlock;
1400 }
1401 if (optname != MRT_INIT) {
1402 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1403 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1404 ret = -EACCES;
1405 goto out_unlock;
1406 }
1407 }
1408
1409 switch (optname) {
1410 case MRT_INIT:
1411 if (optlen != sizeof(int)) {
1412 ret = -EINVAL;
1413 break;
1414 }
1415 if (rtnl_dereference(mrt->mroute_sk)) {
1416 ret = -EADDRINUSE;
1417 break;
1418 }
1419
1420 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1421 if (ret == 0) {
1422 rcu_assign_pointer(mrt->mroute_sk, sk);
1423 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1424 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1425 NETCONFA_MC_FORWARDING,
1426 NETCONFA_IFINDEX_ALL,
1427 net->ipv4.devconf_all);
1428 }
1429 break;
1430 case MRT_DONE:
1431 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1432 ret = -EACCES;
1433 } else {
1434 /* We need to unlock here because mrtsock_destruct takes
1435 * care of rtnl itself and we can't change that due to
1436 * the IP_ROUTER_ALERT setsockopt which runs without it.
1437 */
1438 rtnl_unlock();
1439 ret = ip_ra_control(sk, 0, NULL);
1440 goto out;
1441 }
1442 break;
1443 case MRT_ADD_VIF:
1444 case MRT_DEL_VIF:
1445 if (optlen != sizeof(vif)) {
1446 ret = -EINVAL;
1447 break;
1448 }
1449 if (copy_from_user(&vif, optval, sizeof(vif))) {
1450 ret = -EFAULT;
1451 break;
1452 }
1453 if (vif.vifc_vifi >= MAXVIFS) {
1454 ret = -ENFILE;
1455 break;
1456 }
1457 if (optname == MRT_ADD_VIF) {
1458 ret = vif_add(net, mrt, &vif,
1459 sk == rtnl_dereference(mrt->mroute_sk));
1460 } else {
1461 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1462 }
1463 break;
1464 /* Manipulate the forwarding caches. These live
1465 * in a sort of kernel/user symbiosis.
1466 */
1467 case MRT_ADD_MFC:
1468 case MRT_DEL_MFC:
1469 parent = -1;
1470 fallthrough;
1471 case MRT_ADD_MFC_PROXY:
1472 case MRT_DEL_MFC_PROXY:
1473 if (optlen != sizeof(mfc)) {
1474 ret = -EINVAL;
1475 break;
1476 }
1477 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1478 ret = -EFAULT;
1479 break;
1480 }
1481 if (parent == 0)
1482 parent = mfc.mfcc_parent;
1483 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1484 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1485 else
1486 ret = ipmr_mfc_add(net, mrt, &mfc,
1487 sk == rtnl_dereference(mrt->mroute_sk),
1488 parent);
1489 break;
1490 case MRT_FLUSH:
1491 if (optlen != sizeof(val)) {
1492 ret = -EINVAL;
1493 break;
1494 }
1495 if (get_user(val, (int __user *)optval)) {
1496 ret = -EFAULT;
1497 break;
1498 }
1499 mroute_clean_tables(mrt, val);
1500 break;
1501 /* Control PIM assert. */
1502 case MRT_ASSERT:
1503 if (optlen != sizeof(val)) {
1504 ret = -EINVAL;
1505 break;
1506 }
1507 if (get_user(val, (int __user *)optval)) {
1508 ret = -EFAULT;
1509 break;
1510 }
1511 mrt->mroute_do_assert = val;
1512 break;
1513 case MRT_PIM:
1514 if (!ipmr_pimsm_enabled()) {
1515 ret = -ENOPROTOOPT;
1516 break;
1517 }
1518 if (optlen != sizeof(val)) {
1519 ret = -EINVAL;
1520 break;
1521 }
1522 if (get_user(val, (int __user *)optval)) {
1523 ret = -EFAULT;
1524 break;
1525 }
1526
1527 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1528 val = !!val;
1529 if (val != mrt->mroute_do_pim) {
1530 mrt->mroute_do_pim = val;
1531 mrt->mroute_do_assert = val;
1532 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1533 }
1534 break;
1535 case MRT_TABLE:
1536 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1537 ret = -ENOPROTOOPT;
1538 break;
1539 }
1540 if (optlen != sizeof(uval)) {
1541 ret = -EINVAL;
1542 break;
1543 }
1544 if (get_user(uval, (u32 __user *)optval)) {
1545 ret = -EFAULT;
1546 break;
1547 }
1548
1549 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1550 ret = -EBUSY;
1551 } else {
1552 mrt = ipmr_new_table(net, uval);
1553 if (IS_ERR(mrt))
1554 ret = PTR_ERR(mrt);
1555 else
1556 raw_sk(sk)->ipmr_table = uval;
1557 }
1558 break;
1559 /* Spurious command, or MRT_VERSION which you cannot set. */
1560 default:
1561 ret = -ENOPROTOOPT;
1562 }
1563 out_unlock:
1564 rtnl_unlock();
1565 out:
1566 return ret;
1567 }
1568
1569 /* Getsock opt support for the multicast routing system. */
1570 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1571 {
1572 int olr;
1573 int val;
1574 struct net *net = sock_net(sk);
1575 struct mr_table *mrt;
1576
1577 if (sk->sk_type != SOCK_RAW ||
1578 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1579 return -EOPNOTSUPP;
1580
1581 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1582 if (!mrt)
1583 return -ENOENT;
1584
1585 switch (optname) {
1586 case MRT_VERSION:
1587 val = 0x0305;
1588 break;
1589 case MRT_PIM:
1590 if (!ipmr_pimsm_enabled())
1591 return -ENOPROTOOPT;
1592 val = mrt->mroute_do_pim;
1593 break;
1594 case MRT_ASSERT:
1595 val = mrt->mroute_do_assert;
1596 break;
1597 default:
1598 return -ENOPROTOOPT;
1599 }
1600
1601 if (get_user(olr, optlen))
1602 return -EFAULT;
1603 olr = min_t(unsigned int, olr, sizeof(int));
1604 if (olr < 0)
1605 return -EINVAL;
1606 if (put_user(olr, optlen))
1607 return -EFAULT;
1608 if (copy_to_user(optval, &val, olr))
1609 return -EFAULT;
1610 return 0;
1611 }
1612
1613 /* The IP multicast ioctl support routines. */
1614 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1615 {
1616 struct sioc_sg_req sr;
1617 struct sioc_vif_req vr;
1618 struct vif_device *vif;
1619 struct mfc_cache *c;
1620 struct net *net = sock_net(sk);
1621 struct mr_table *mrt;
1622
1623 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1624 if (!mrt)
1625 return -ENOENT;
1626
1627 switch (cmd) {
1628 case SIOCGETVIFCNT:
1629 if (copy_from_user(&vr, arg, sizeof(vr)))
1630 return -EFAULT;
1631 if (vr.vifi >= mrt->maxvif)
1632 return -EINVAL;
1633 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1634 read_lock(&mrt_lock);
1635 vif = &mrt->vif_table[vr.vifi];
1636 if (VIF_EXISTS(mrt, vr.vifi)) {
1637 vr.icount = vif->pkt_in;
1638 vr.ocount = vif->pkt_out;
1639 vr.ibytes = vif->bytes_in;
1640 vr.obytes = vif->bytes_out;
1641 read_unlock(&mrt_lock);
1642
1643 if (copy_to_user(arg, &vr, sizeof(vr)))
1644 return -EFAULT;
1645 return 0;
1646 }
1647 read_unlock(&mrt_lock);
1648 return -EADDRNOTAVAIL;
1649 case SIOCGETSGCNT:
1650 if (copy_from_user(&sr, arg, sizeof(sr)))
1651 return -EFAULT;
1652
1653 rcu_read_lock();
1654 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1655 if (c) {
1656 sr.pktcnt = c->_c.mfc_un.res.pkt;
1657 sr.bytecnt = c->_c.mfc_un.res.bytes;
1658 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1659 rcu_read_unlock();
1660
1661 if (copy_to_user(arg, &sr, sizeof(sr)))
1662 return -EFAULT;
1663 return 0;
1664 }
1665 rcu_read_unlock();
1666 return -EADDRNOTAVAIL;
1667 default:
1668 return -ENOIOCTLCMD;
1669 }
1670 }
1671
1672 #ifdef CONFIG_COMPAT
1673 struct compat_sioc_sg_req {
1674 struct in_addr src;
1675 struct in_addr grp;
1676 compat_ulong_t pktcnt;
1677 compat_ulong_t bytecnt;
1678 compat_ulong_t wrong_if;
1679 };
1680
1681 struct compat_sioc_vif_req {
1682 vifi_t vifi; /* Which iface */
1683 compat_ulong_t icount;
1684 compat_ulong_t ocount;
1685 compat_ulong_t ibytes;
1686 compat_ulong_t obytes;
1687 };
1688
1689 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1690 {
1691 struct compat_sioc_sg_req sr;
1692 struct compat_sioc_vif_req vr;
1693 struct vif_device *vif;
1694 struct mfc_cache *c;
1695 struct net *net = sock_net(sk);
1696 struct mr_table *mrt;
1697
1698 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1699 if (!mrt)
1700 return -ENOENT;
1701
1702 switch (cmd) {
1703 case SIOCGETVIFCNT:
1704 if (copy_from_user(&vr, arg, sizeof(vr)))
1705 return -EFAULT;
1706 if (vr.vifi >= mrt->maxvif)
1707 return -EINVAL;
1708 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1709 read_lock(&mrt_lock);
1710 vif = &mrt->vif_table[vr.vifi];
1711 if (VIF_EXISTS(mrt, vr.vifi)) {
1712 vr.icount = vif->pkt_in;
1713 vr.ocount = vif->pkt_out;
1714 vr.ibytes = vif->bytes_in;
1715 vr.obytes = vif->bytes_out;
1716 read_unlock(&mrt_lock);
1717
1718 if (copy_to_user(arg, &vr, sizeof(vr)))
1719 return -EFAULT;
1720 return 0;
1721 }
1722 read_unlock(&mrt_lock);
1723 return -EADDRNOTAVAIL;
1724 case SIOCGETSGCNT:
1725 if (copy_from_user(&sr, arg, sizeof(sr)))
1726 return -EFAULT;
1727
1728 rcu_read_lock();
1729 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1730 if (c) {
1731 sr.pktcnt = c->_c.mfc_un.res.pkt;
1732 sr.bytecnt = c->_c.mfc_un.res.bytes;
1733 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1734 rcu_read_unlock();
1735
1736 if (copy_to_user(arg, &sr, sizeof(sr)))
1737 return -EFAULT;
1738 return 0;
1739 }
1740 rcu_read_unlock();
1741 return -EADDRNOTAVAIL;
1742 default:
1743 return -ENOIOCTLCMD;
1744 }
1745 }
1746 #endif
1747
1748 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1749 {
1750 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1751 struct net *net = dev_net(dev);
1752 struct mr_table *mrt;
1753 struct vif_device *v;
1754 int ct;
1755
1756 if (event != NETDEV_UNREGISTER)
1757 return NOTIFY_DONE;
1758
1759 ipmr_for_each_table(mrt, net) {
1760 v = &mrt->vif_table[0];
1761 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1762 if (v->dev == dev)
1763 vif_delete(mrt, ct, 1, NULL);
1764 }
1765 }
1766 return NOTIFY_DONE;
1767 }
1768
1769 static struct notifier_block ip_mr_notifier = {
1770 .notifier_call = ipmr_device_event,
1771 };
1772
1773 /* Encapsulate a packet by attaching a valid IPIP header to it.
1774 * This avoids tunnel drivers and other mess and gives us the speed so
1775 * important for multicast video.
1776 */
1777 static void ip_encap(struct net *net, struct sk_buff *skb,
1778 __be32 saddr, __be32 daddr)
1779 {
1780 struct iphdr *iph;
1781 const struct iphdr *old_iph = ip_hdr(skb);
1782
1783 skb_push(skb, sizeof(struct iphdr));
1784 skb->transport_header = skb->network_header;
1785 skb_reset_network_header(skb);
1786 iph = ip_hdr(skb);
1787
1788 iph->version = 4;
1789 iph->tos = old_iph->tos;
1790 iph->ttl = old_iph->ttl;
1791 iph->frag_off = 0;
1792 iph->daddr = daddr;
1793 iph->saddr = saddr;
1794 iph->protocol = IPPROTO_IPIP;
1795 iph->ihl = 5;
1796 iph->tot_len = htons(skb->len);
1797 ip_select_ident(net, skb, NULL);
1798 ip_send_check(iph);
1799
1800 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1801 nf_reset_ct(skb);
1802 }
1803
1804 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1805 struct sk_buff *skb)
1806 {
1807 struct ip_options *opt = &(IPCB(skb)->opt);
1808
1809 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1810 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1811
1812 if (unlikely(opt->optlen))
1813 ip_forward_options(skb);
1814
1815 return dst_output(net, sk, skb);
1816 }
1817
1818 #ifdef CONFIG_NET_SWITCHDEV
1819 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1820 int in_vifi, int out_vifi)
1821 {
1822 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1823 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1824
1825 if (!skb->offload_l3_fwd_mark)
1826 return false;
1827 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1828 return false;
1829 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1830 &in_vif->dev_parent_id);
1831 }
1832 #else
1833 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1834 int in_vifi, int out_vifi)
1835 {
1836 return false;
1837 }
1838 #endif
1839
1840 /* Processing handlers for ipmr_forward */
1841
1842 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1843 int in_vifi, struct sk_buff *skb, int vifi)
1844 {
1845 const struct iphdr *iph = ip_hdr(skb);
1846 struct vif_device *vif = &mrt->vif_table[vifi];
1847 struct net_device *dev;
1848 struct rtable *rt;
1849 struct flowi4 fl4;
1850 int encap = 0;
1851
1852 if (!vif->dev)
1853 goto out_free;
1854
1855 if (vif->flags & VIFF_REGISTER) {
1856 vif->pkt_out++;
1857 vif->bytes_out += skb->len;
1858 vif->dev->stats.tx_bytes += skb->len;
1859 vif->dev->stats.tx_packets++;
1860 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1861 goto out_free;
1862 }
1863
1864 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1865 goto out_free;
1866
1867 if (vif->flags & VIFF_TUNNEL) {
1868 rt = ip_route_output_ports(net, &fl4, NULL,
1869 vif->remote, vif->local,
1870 0, 0,
1871 IPPROTO_IPIP,
1872 RT_TOS(iph->tos), vif->link);
1873 if (IS_ERR(rt))
1874 goto out_free;
1875 encap = sizeof(struct iphdr);
1876 } else {
1877 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1878 0, 0,
1879 IPPROTO_IPIP,
1880 RT_TOS(iph->tos), vif->link);
1881 if (IS_ERR(rt))
1882 goto out_free;
1883 }
1884
1885 dev = rt->dst.dev;
1886
1887 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1888 /* Do not fragment multicasts. Alas, IPv4 does not
1889 * allow to send ICMP, so that packets will disappear
1890 * to blackhole.
1891 */
1892 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1893 ip_rt_put(rt);
1894 goto out_free;
1895 }
1896
1897 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1898
1899 if (skb_cow(skb, encap)) {
1900 ip_rt_put(rt);
1901 goto out_free;
1902 }
1903
1904 vif->pkt_out++;
1905 vif->bytes_out += skb->len;
1906
1907 skb_dst_drop(skb);
1908 skb_dst_set(skb, &rt->dst);
1909 ip_decrease_ttl(ip_hdr(skb));
1910
1911 /* FIXME: forward and output firewalls used to be called here.
1912 * What do we do with netfilter? -- RR
1913 */
1914 if (vif->flags & VIFF_TUNNEL) {
1915 ip_encap(net, skb, vif->local, vif->remote);
1916 /* FIXME: extra output firewall step used to be here. --RR */
1917 vif->dev->stats.tx_packets++;
1918 vif->dev->stats.tx_bytes += skb->len;
1919 }
1920
1921 IPCB(skb)->flags |= IPSKB_FORWARDED;
1922
1923 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1924 * not only before forwarding, but after forwarding on all output
1925 * interfaces. It is clear, if mrouter runs a multicasting
1926 * program, it should receive packets not depending to what interface
1927 * program is joined.
1928 * If we will not make it, the program will have to join on all
1929 * interfaces. On the other hand, multihoming host (or router, but
1930 * not mrouter) cannot join to more than one interface - it will
1931 * result in receiving multiple packets.
1932 */
1933 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1934 net, NULL, skb, skb->dev, dev,
1935 ipmr_forward_finish);
1936 return;
1937
1938 out_free:
1939 kfree_skb(skb);
1940 }
1941
1942 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1943 {
1944 int ct;
1945
1946 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1947 if (mrt->vif_table[ct].dev == dev)
1948 break;
1949 }
1950 return ct;
1951 }
1952
1953 /* "local" means that we should preserve one skb (for local delivery) */
1954 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1955 struct net_device *dev, struct sk_buff *skb,
1956 struct mfc_cache *c, int local)
1957 {
1958 int true_vifi = ipmr_find_vif(mrt, dev);
1959 int psend = -1;
1960 int vif, ct;
1961
1962 vif = c->_c.mfc_parent;
1963 c->_c.mfc_un.res.pkt++;
1964 c->_c.mfc_un.res.bytes += skb->len;
1965 c->_c.mfc_un.res.lastuse = jiffies;
1966
1967 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1968 struct mfc_cache *cache_proxy;
1969
1970 /* For an (*,G) entry, we only check that the incomming
1971 * interface is part of the static tree.
1972 */
1973 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1974 if (cache_proxy &&
1975 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1976 goto forward;
1977 }
1978
1979 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1980 if (mrt->vif_table[vif].dev != dev) {
1981 if (rt_is_output_route(skb_rtable(skb))) {
1982 /* It is our own packet, looped back.
1983 * Very complicated situation...
1984 *
1985 * The best workaround until routing daemons will be
1986 * fixed is not to redistribute packet, if it was
1987 * send through wrong interface. It means, that
1988 * multicast applications WILL NOT work for
1989 * (S,G), which have default multicast route pointing
1990 * to wrong oif. In any case, it is not a good
1991 * idea to use multicasting applications on router.
1992 */
1993 goto dont_forward;
1994 }
1995
1996 c->_c.mfc_un.res.wrong_if++;
1997
1998 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1999 /* pimsm uses asserts, when switching from RPT to SPT,
2000 * so that we cannot check that packet arrived on an oif.
2001 * It is bad, but otherwise we would need to move pretty
2002 * large chunk of pimd to kernel. Ough... --ANK
2003 */
2004 (mrt->mroute_do_pim ||
2005 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2006 time_after(jiffies,
2007 c->_c.mfc_un.res.last_assert +
2008 MFC_ASSERT_THRESH)) {
2009 c->_c.mfc_un.res.last_assert = jiffies;
2010 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2011 if (mrt->mroute_do_wrvifwhole)
2012 ipmr_cache_report(mrt, skb, true_vifi,
2013 IGMPMSG_WRVIFWHOLE);
2014 }
2015 goto dont_forward;
2016 }
2017
2018 forward:
2019 mrt->vif_table[vif].pkt_in++;
2020 mrt->vif_table[vif].bytes_in += skb->len;
2021
2022 /* Forward the frame */
2023 if (c->mfc_origin == htonl(INADDR_ANY) &&
2024 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2025 if (true_vifi >= 0 &&
2026 true_vifi != c->_c.mfc_parent &&
2027 ip_hdr(skb)->ttl >
2028 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2029 /* It's an (*,*) entry and the packet is not coming from
2030 * the upstream: forward the packet to the upstream
2031 * only.
2032 */
2033 psend = c->_c.mfc_parent;
2034 goto last_forward;
2035 }
2036 goto dont_forward;
2037 }
2038 for (ct = c->_c.mfc_un.res.maxvif - 1;
2039 ct >= c->_c.mfc_un.res.minvif; ct--) {
2040 /* For (*,G) entry, don't forward to the incoming interface */
2041 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2042 ct != true_vifi) &&
2043 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2044 if (psend != -1) {
2045 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2046
2047 if (skb2)
2048 ipmr_queue_xmit(net, mrt, true_vifi,
2049 skb2, psend);
2050 }
2051 psend = ct;
2052 }
2053 }
2054 last_forward:
2055 if (psend != -1) {
2056 if (local) {
2057 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2058
2059 if (skb2)
2060 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2061 psend);
2062 } else {
2063 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2064 return;
2065 }
2066 }
2067
2068 dont_forward:
2069 if (!local)
2070 kfree_skb(skb);
2071 }
2072
2073 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2074 {
2075 struct rtable *rt = skb_rtable(skb);
2076 struct iphdr *iph = ip_hdr(skb);
2077 struct flowi4 fl4 = {
2078 .daddr = iph->daddr,
2079 .saddr = iph->saddr,
2080 .flowi4_tos = RT_TOS(iph->tos),
2081 .flowi4_oif = (rt_is_output_route(rt) ?
2082 skb->dev->ifindex : 0),
2083 .flowi4_iif = (rt_is_output_route(rt) ?
2084 LOOPBACK_IFINDEX :
2085 skb->dev->ifindex),
2086 .flowi4_mark = skb->mark,
2087 };
2088 struct mr_table *mrt;
2089 int err;
2090
2091 err = ipmr_fib_lookup(net, &fl4, &mrt);
2092 if (err)
2093 return ERR_PTR(err);
2094 return mrt;
2095 }
2096
2097 /* Multicast packets for forwarding arrive here
2098 * Called with rcu_read_lock();
2099 */
2100 int ip_mr_input(struct sk_buff *skb)
2101 {
2102 struct mfc_cache *cache;
2103 struct net *net = dev_net(skb->dev);
2104 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2105 struct mr_table *mrt;
2106 struct net_device *dev;
2107
2108 /* skb->dev passed in is the loX master dev for vrfs.
2109 * As there are no vifs associated with loopback devices,
2110 * get the proper interface that does have a vif associated with it.
2111 */
2112 dev = skb->dev;
2113 if (netif_is_l3_master(skb->dev)) {
2114 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2115 if (!dev) {
2116 kfree_skb(skb);
2117 return -ENODEV;
2118 }
2119 }
2120
2121 /* Packet is looped back after forward, it should not be
2122 * forwarded second time, but still can be delivered locally.
2123 */
2124 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2125 goto dont_forward;
2126
2127 mrt = ipmr_rt_fib_lookup(net, skb);
2128 if (IS_ERR(mrt)) {
2129 kfree_skb(skb);
2130 return PTR_ERR(mrt);
2131 }
2132 if (!local) {
2133 if (IPCB(skb)->opt.router_alert) {
2134 if (ip_call_ra_chain(skb))
2135 return 0;
2136 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2137 /* IGMPv1 (and broken IGMPv2 implementations sort of
2138 * Cisco IOS <= 11.2(8)) do not put router alert
2139 * option to IGMP packets destined to routable
2140 * groups. It is very bad, because it means
2141 * that we can forward NO IGMP messages.
2142 */
2143 struct sock *mroute_sk;
2144
2145 mroute_sk = rcu_dereference(mrt->mroute_sk);
2146 if (mroute_sk) {
2147 nf_reset_ct(skb);
2148 raw_rcv(mroute_sk, skb);
2149 return 0;
2150 }
2151 }
2152 }
2153
2154 /* already under rcu_read_lock() */
2155 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2156 if (!cache) {
2157 int vif = ipmr_find_vif(mrt, dev);
2158
2159 if (vif >= 0)
2160 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2161 vif);
2162 }
2163
2164 /* No usable cache entry */
2165 if (!cache) {
2166 int vif;
2167
2168 if (local) {
2169 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2170 ip_local_deliver(skb);
2171 if (!skb2)
2172 return -ENOBUFS;
2173 skb = skb2;
2174 }
2175
2176 read_lock(&mrt_lock);
2177 vif = ipmr_find_vif(mrt, dev);
2178 if (vif >= 0) {
2179 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2180 read_unlock(&mrt_lock);
2181
2182 return err2;
2183 }
2184 read_unlock(&mrt_lock);
2185 kfree_skb(skb);
2186 return -ENODEV;
2187 }
2188
2189 read_lock(&mrt_lock);
2190 ip_mr_forward(net, mrt, dev, skb, cache, local);
2191 read_unlock(&mrt_lock);
2192
2193 if (local)
2194 return ip_local_deliver(skb);
2195
2196 return 0;
2197
2198 dont_forward:
2199 if (local)
2200 return ip_local_deliver(skb);
2201 kfree_skb(skb);
2202 return 0;
2203 }
2204
2205 #ifdef CONFIG_IP_PIMSM_V1
2206 /* Handle IGMP messages of PIMv1 */
2207 int pim_rcv_v1(struct sk_buff *skb)
2208 {
2209 struct igmphdr *pim;
2210 struct net *net = dev_net(skb->dev);
2211 struct mr_table *mrt;
2212
2213 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2214 goto drop;
2215
2216 pim = igmp_hdr(skb);
2217
2218 mrt = ipmr_rt_fib_lookup(net, skb);
2219 if (IS_ERR(mrt))
2220 goto drop;
2221 if (!mrt->mroute_do_pim ||
2222 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2223 goto drop;
2224
2225 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2226 drop:
2227 kfree_skb(skb);
2228 }
2229 return 0;
2230 }
2231 #endif
2232
2233 #ifdef CONFIG_IP_PIMSM_V2
2234 static int pim_rcv(struct sk_buff *skb)
2235 {
2236 struct pimreghdr *pim;
2237 struct net *net = dev_net(skb->dev);
2238 struct mr_table *mrt;
2239
2240 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2241 goto drop;
2242
2243 pim = (struct pimreghdr *)skb_transport_header(skb);
2244 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2245 (pim->flags & PIM_NULL_REGISTER) ||
2246 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2247 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2248 goto drop;
2249
2250 mrt = ipmr_rt_fib_lookup(net, skb);
2251 if (IS_ERR(mrt))
2252 goto drop;
2253 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2254 drop:
2255 kfree_skb(skb);
2256 }
2257 return 0;
2258 }
2259 #endif
2260
2261 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2262 __be32 saddr, __be32 daddr,
2263 struct rtmsg *rtm, u32 portid)
2264 {
2265 struct mfc_cache *cache;
2266 struct mr_table *mrt;
2267 int err;
2268
2269 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2270 if (!mrt)
2271 return -ENOENT;
2272
2273 rcu_read_lock();
2274 cache = ipmr_cache_find(mrt, saddr, daddr);
2275 if (!cache && skb->dev) {
2276 int vif = ipmr_find_vif(mrt, skb->dev);
2277
2278 if (vif >= 0)
2279 cache = ipmr_cache_find_any(mrt, daddr, vif);
2280 }
2281 if (!cache) {
2282 struct sk_buff *skb2;
2283 struct iphdr *iph;
2284 struct net_device *dev;
2285 int vif = -1;
2286
2287 dev = skb->dev;
2288 read_lock(&mrt_lock);
2289 if (dev)
2290 vif = ipmr_find_vif(mrt, dev);
2291 if (vif < 0) {
2292 read_unlock(&mrt_lock);
2293 rcu_read_unlock();
2294 return -ENODEV;
2295 }
2296
2297 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2298 if (!skb2) {
2299 read_unlock(&mrt_lock);
2300 rcu_read_unlock();
2301 return -ENOMEM;
2302 }
2303
2304 NETLINK_CB(skb2).portid = portid;
2305 skb_push(skb2, sizeof(struct iphdr));
2306 skb_reset_network_header(skb2);
2307 iph = ip_hdr(skb2);
2308 iph->ihl = sizeof(struct iphdr) >> 2;
2309 iph->saddr = saddr;
2310 iph->daddr = daddr;
2311 iph->version = 0;
2312 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2313 read_unlock(&mrt_lock);
2314 rcu_read_unlock();
2315 return err;
2316 }
2317
2318 read_lock(&mrt_lock);
2319 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2320 read_unlock(&mrt_lock);
2321 rcu_read_unlock();
2322 return err;
2323 }
2324
2325 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2326 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2327 int flags)
2328 {
2329 struct nlmsghdr *nlh;
2330 struct rtmsg *rtm;
2331 int err;
2332
2333 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2334 if (!nlh)
2335 return -EMSGSIZE;
2336
2337 rtm = nlmsg_data(nlh);
2338 rtm->rtm_family = RTNL_FAMILY_IPMR;
2339 rtm->rtm_dst_len = 32;
2340 rtm->rtm_src_len = 32;
2341 rtm->rtm_tos = 0;
2342 rtm->rtm_table = mrt->id;
2343 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2344 goto nla_put_failure;
2345 rtm->rtm_type = RTN_MULTICAST;
2346 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2347 if (c->_c.mfc_flags & MFC_STATIC)
2348 rtm->rtm_protocol = RTPROT_STATIC;
2349 else
2350 rtm->rtm_protocol = RTPROT_MROUTED;
2351 rtm->rtm_flags = 0;
2352
2353 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2354 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2355 goto nla_put_failure;
2356 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2357 /* do not break the dump if cache is unresolved */
2358 if (err < 0 && err != -ENOENT)
2359 goto nla_put_failure;
2360
2361 nlmsg_end(skb, nlh);
2362 return 0;
2363
2364 nla_put_failure:
2365 nlmsg_cancel(skb, nlh);
2366 return -EMSGSIZE;
2367 }
2368
2369 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2370 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2371 int flags)
2372 {
2373 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2374 cmd, flags);
2375 }
2376
2377 static size_t mroute_msgsize(bool unresolved, int maxvif)
2378 {
2379 size_t len =
2380 NLMSG_ALIGN(sizeof(struct rtmsg))
2381 + nla_total_size(4) /* RTA_TABLE */
2382 + nla_total_size(4) /* RTA_SRC */
2383 + nla_total_size(4) /* RTA_DST */
2384 ;
2385
2386 if (!unresolved)
2387 len = len
2388 + nla_total_size(4) /* RTA_IIF */
2389 + nla_total_size(0) /* RTA_MULTIPATH */
2390 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2391 /* RTA_MFC_STATS */
2392 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2393 ;
2394
2395 return len;
2396 }
2397
2398 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2399 int cmd)
2400 {
2401 struct net *net = read_pnet(&mrt->net);
2402 struct sk_buff *skb;
2403 int err = -ENOBUFS;
2404
2405 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2406 mrt->maxvif),
2407 GFP_ATOMIC);
2408 if (!skb)
2409 goto errout;
2410
2411 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2412 if (err < 0)
2413 goto errout;
2414
2415 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2416 return;
2417
2418 errout:
2419 kfree_skb(skb);
2420 if (err < 0)
2421 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2422 }
2423
2424 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2425 {
2426 size_t len =
2427 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2428 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2429 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2430 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2431 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2432 /* IPMRA_CREPORT_PKT */
2433 + nla_total_size(payloadlen)
2434 ;
2435
2436 return len;
2437 }
2438
2439 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2440 {
2441 struct net *net = read_pnet(&mrt->net);
2442 struct nlmsghdr *nlh;
2443 struct rtgenmsg *rtgenm;
2444 struct igmpmsg *msg;
2445 struct sk_buff *skb;
2446 struct nlattr *nla;
2447 int payloadlen;
2448
2449 payloadlen = pkt->len - sizeof(struct igmpmsg);
2450 msg = (struct igmpmsg *)skb_network_header(pkt);
2451
2452 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2453 if (!skb)
2454 goto errout;
2455
2456 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2457 sizeof(struct rtgenmsg), 0);
2458 if (!nlh)
2459 goto errout;
2460 rtgenm = nlmsg_data(nlh);
2461 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2462 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2463 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2464 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2465 msg->im_src.s_addr) ||
2466 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2467 msg->im_dst.s_addr))
2468 goto nla_put_failure;
2469
2470 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2471 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2472 nla_data(nla), payloadlen))
2473 goto nla_put_failure;
2474
2475 nlmsg_end(skb, nlh);
2476
2477 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2478 return;
2479
2480 nla_put_failure:
2481 nlmsg_cancel(skb, nlh);
2482 errout:
2483 kfree_skb(skb);
2484 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2485 }
2486
2487 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2488 const struct nlmsghdr *nlh,
2489 struct nlattr **tb,
2490 struct netlink_ext_ack *extack)
2491 {
2492 struct rtmsg *rtm;
2493 int i, err;
2494
2495 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2496 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2497 return -EINVAL;
2498 }
2499
2500 if (!netlink_strict_get_check(skb))
2501 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2502 rtm_ipv4_policy, extack);
2503
2504 rtm = nlmsg_data(nlh);
2505 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2506 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2507 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2508 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2509 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2510 return -EINVAL;
2511 }
2512
2513 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2514 rtm_ipv4_policy, extack);
2515 if (err)
2516 return err;
2517
2518 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2519 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2520 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2521 return -EINVAL;
2522 }
2523
2524 for (i = 0; i <= RTA_MAX; i++) {
2525 if (!tb[i])
2526 continue;
2527
2528 switch (i) {
2529 case RTA_SRC:
2530 case RTA_DST:
2531 case RTA_TABLE:
2532 break;
2533 default:
2534 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2535 return -EINVAL;
2536 }
2537 }
2538
2539 return 0;
2540 }
2541
2542 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2543 struct netlink_ext_ack *extack)
2544 {
2545 struct net *net = sock_net(in_skb->sk);
2546 struct nlattr *tb[RTA_MAX + 1];
2547 struct sk_buff *skb = NULL;
2548 struct mfc_cache *cache;
2549 struct mr_table *mrt;
2550 __be32 src, grp;
2551 u32 tableid;
2552 int err;
2553
2554 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2555 if (err < 0)
2556 goto errout;
2557
2558 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2559 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2560 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2561
2562 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2563 if (!mrt) {
2564 err = -ENOENT;
2565 goto errout_free;
2566 }
2567
2568 /* entries are added/deleted only under RTNL */
2569 rcu_read_lock();
2570 cache = ipmr_cache_find(mrt, src, grp);
2571 rcu_read_unlock();
2572 if (!cache) {
2573 err = -ENOENT;
2574 goto errout_free;
2575 }
2576
2577 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2578 if (!skb) {
2579 err = -ENOBUFS;
2580 goto errout_free;
2581 }
2582
2583 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2584 nlh->nlmsg_seq, cache,
2585 RTM_NEWROUTE, 0);
2586 if (err < 0)
2587 goto errout_free;
2588
2589 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2590
2591 errout:
2592 return err;
2593
2594 errout_free:
2595 kfree_skb(skb);
2596 goto errout;
2597 }
2598
2599 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2600 {
2601 struct fib_dump_filter filter = {};
2602 int err;
2603
2604 if (cb->strict_check) {
2605 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2606 &filter, cb);
2607 if (err < 0)
2608 return err;
2609 }
2610
2611 if (filter.table_id) {
2612 struct mr_table *mrt;
2613
2614 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2615 if (!mrt) {
2616 if (filter.dump_all_families)
2617 return skb->len;
2618
2619 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2620 return -ENOENT;
2621 }
2622 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2623 &mfc_unres_lock, &filter);
2624 return skb->len ? : err;
2625 }
2626
2627 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2628 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2629 }
2630
2631 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2632 [RTA_SRC] = { .type = NLA_U32 },
2633 [RTA_DST] = { .type = NLA_U32 },
2634 [RTA_IIF] = { .type = NLA_U32 },
2635 [RTA_TABLE] = { .type = NLA_U32 },
2636 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2637 };
2638
2639 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2640 {
2641 switch (rtm_protocol) {
2642 case RTPROT_STATIC:
2643 case RTPROT_MROUTED:
2644 return true;
2645 }
2646 return false;
2647 }
2648
2649 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2650 {
2651 struct rtnexthop *rtnh = nla_data(nla);
2652 int remaining = nla_len(nla), vifi = 0;
2653
2654 while (rtnh_ok(rtnh, remaining)) {
2655 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2656 if (++vifi == MAXVIFS)
2657 break;
2658 rtnh = rtnh_next(rtnh, &remaining);
2659 }
2660
2661 return remaining > 0 ? -EINVAL : vifi;
2662 }
2663
2664 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2665 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2666 struct mfcctl *mfcc, int *mrtsock,
2667 struct mr_table **mrtret,
2668 struct netlink_ext_ack *extack)
2669 {
2670 struct net_device *dev = NULL;
2671 u32 tblid = RT_TABLE_DEFAULT;
2672 struct mr_table *mrt;
2673 struct nlattr *attr;
2674 struct rtmsg *rtm;
2675 int ret, rem;
2676
2677 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2678 rtm_ipmr_policy, extack);
2679 if (ret < 0)
2680 goto out;
2681 rtm = nlmsg_data(nlh);
2682
2683 ret = -EINVAL;
2684 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2685 rtm->rtm_type != RTN_MULTICAST ||
2686 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2687 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2688 goto out;
2689
2690 memset(mfcc, 0, sizeof(*mfcc));
2691 mfcc->mfcc_parent = -1;
2692 ret = 0;
2693 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2694 switch (nla_type(attr)) {
2695 case RTA_SRC:
2696 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2697 break;
2698 case RTA_DST:
2699 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2700 break;
2701 case RTA_IIF:
2702 dev = __dev_get_by_index(net, nla_get_u32(attr));
2703 if (!dev) {
2704 ret = -ENODEV;
2705 goto out;
2706 }
2707 break;
2708 case RTA_MULTIPATH:
2709 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2710 ret = -EINVAL;
2711 goto out;
2712 }
2713 break;
2714 case RTA_PREFSRC:
2715 ret = 1;
2716 break;
2717 case RTA_TABLE:
2718 tblid = nla_get_u32(attr);
2719 break;
2720 }
2721 }
2722 mrt = ipmr_get_table(net, tblid);
2723 if (!mrt) {
2724 ret = -ENOENT;
2725 goto out;
2726 }
2727 *mrtret = mrt;
2728 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2729 if (dev)
2730 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2731
2732 out:
2733 return ret;
2734 }
2735
2736 /* takes care of both newroute and delroute */
2737 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2738 struct netlink_ext_ack *extack)
2739 {
2740 struct net *net = sock_net(skb->sk);
2741 int ret, mrtsock, parent;
2742 struct mr_table *tbl;
2743 struct mfcctl mfcc;
2744
2745 mrtsock = 0;
2746 tbl = NULL;
2747 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2748 if (ret < 0)
2749 return ret;
2750
2751 parent = ret ? mfcc.mfcc_parent : -1;
2752 if (nlh->nlmsg_type == RTM_NEWROUTE)
2753 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2754 else
2755 return ipmr_mfc_delete(tbl, &mfcc, parent);
2756 }
2757
2758 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2759 {
2760 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2761
2762 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2763 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2764 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2765 mrt->mroute_reg_vif_num) ||
2766 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2767 mrt->mroute_do_assert) ||
2768 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2769 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2770 mrt->mroute_do_wrvifwhole))
2771 return false;
2772
2773 return true;
2774 }
2775
2776 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2777 {
2778 struct nlattr *vif_nest;
2779 struct vif_device *vif;
2780
2781 /* if the VIF doesn't exist just continue */
2782 if (!VIF_EXISTS(mrt, vifid))
2783 return true;
2784
2785 vif = &mrt->vif_table[vifid];
2786 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2787 if (!vif_nest)
2788 return false;
2789 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2790 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2791 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2792 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2793 IPMRA_VIFA_PAD) ||
2794 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2795 IPMRA_VIFA_PAD) ||
2796 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2797 IPMRA_VIFA_PAD) ||
2798 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2799 IPMRA_VIFA_PAD) ||
2800 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2801 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2802 nla_nest_cancel(skb, vif_nest);
2803 return false;
2804 }
2805 nla_nest_end(skb, vif_nest);
2806
2807 return true;
2808 }
2809
2810 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2811 struct netlink_ext_ack *extack)
2812 {
2813 struct ifinfomsg *ifm;
2814
2815 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2816 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2817 return -EINVAL;
2818 }
2819
2820 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2821 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2822 return -EINVAL;
2823 }
2824
2825 ifm = nlmsg_data(nlh);
2826 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2827 ifm->ifi_change || ifm->ifi_index) {
2828 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2829 return -EINVAL;
2830 }
2831
2832 return 0;
2833 }
2834
2835 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2836 {
2837 struct net *net = sock_net(skb->sk);
2838 struct nlmsghdr *nlh = NULL;
2839 unsigned int t = 0, s_t;
2840 unsigned int e = 0, s_e;
2841 struct mr_table *mrt;
2842
2843 if (cb->strict_check) {
2844 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2845
2846 if (err < 0)
2847 return err;
2848 }
2849
2850 s_t = cb->args[0];
2851 s_e = cb->args[1];
2852
2853 ipmr_for_each_table(mrt, net) {
2854 struct nlattr *vifs, *af;
2855 struct ifinfomsg *hdr;
2856 u32 i;
2857
2858 if (t < s_t)
2859 goto skip_table;
2860 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2861 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2862 sizeof(*hdr), NLM_F_MULTI);
2863 if (!nlh)
2864 break;
2865
2866 hdr = nlmsg_data(nlh);
2867 memset(hdr, 0, sizeof(*hdr));
2868 hdr->ifi_family = RTNL_FAMILY_IPMR;
2869
2870 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2871 if (!af) {
2872 nlmsg_cancel(skb, nlh);
2873 goto out;
2874 }
2875
2876 if (!ipmr_fill_table(mrt, skb)) {
2877 nlmsg_cancel(skb, nlh);
2878 goto out;
2879 }
2880
2881 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2882 if (!vifs) {
2883 nla_nest_end(skb, af);
2884 nlmsg_end(skb, nlh);
2885 goto out;
2886 }
2887 for (i = 0; i < mrt->maxvif; i++) {
2888 if (e < s_e)
2889 goto skip_entry;
2890 if (!ipmr_fill_vif(mrt, i, skb)) {
2891 nla_nest_end(skb, vifs);
2892 nla_nest_end(skb, af);
2893 nlmsg_end(skb, nlh);
2894 goto out;
2895 }
2896 skip_entry:
2897 e++;
2898 }
2899 s_e = 0;
2900 e = 0;
2901 nla_nest_end(skb, vifs);
2902 nla_nest_end(skb, af);
2903 nlmsg_end(skb, nlh);
2904 skip_table:
2905 t++;
2906 }
2907
2908 out:
2909 cb->args[1] = e;
2910 cb->args[0] = t;
2911
2912 return skb->len;
2913 }
2914
2915 #ifdef CONFIG_PROC_FS
2916 /* The /proc interfaces to multicast routing :
2917 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2918 */
2919
2920 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2921 __acquires(mrt_lock)
2922 {
2923 struct mr_vif_iter *iter = seq->private;
2924 struct net *net = seq_file_net(seq);
2925 struct mr_table *mrt;
2926
2927 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2928 if (!mrt)
2929 return ERR_PTR(-ENOENT);
2930
2931 iter->mrt = mrt;
2932
2933 read_lock(&mrt_lock);
2934 return mr_vif_seq_start(seq, pos);
2935 }
2936
2937 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2938 __releases(mrt_lock)
2939 {
2940 read_unlock(&mrt_lock);
2941 }
2942
2943 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2944 {
2945 struct mr_vif_iter *iter = seq->private;
2946 struct mr_table *mrt = iter->mrt;
2947
2948 if (v == SEQ_START_TOKEN) {
2949 seq_puts(seq,
2950 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2951 } else {
2952 const struct vif_device *vif = v;
2953 const char *name = vif->dev ?
2954 vif->dev->name : "none";
2955
2956 seq_printf(seq,
2957 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2958 vif - mrt->vif_table,
2959 name, vif->bytes_in, vif->pkt_in,
2960 vif->bytes_out, vif->pkt_out,
2961 vif->flags, vif->local, vif->remote);
2962 }
2963 return 0;
2964 }
2965
2966 static const struct seq_operations ipmr_vif_seq_ops = {
2967 .start = ipmr_vif_seq_start,
2968 .next = mr_vif_seq_next,
2969 .stop = ipmr_vif_seq_stop,
2970 .show = ipmr_vif_seq_show,
2971 };
2972
2973 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2974 {
2975 struct net *net = seq_file_net(seq);
2976 struct mr_table *mrt;
2977
2978 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2979 if (!mrt)
2980 return ERR_PTR(-ENOENT);
2981
2982 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2983 }
2984
2985 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2986 {
2987 int n;
2988
2989 if (v == SEQ_START_TOKEN) {
2990 seq_puts(seq,
2991 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2992 } else {
2993 const struct mfc_cache *mfc = v;
2994 const struct mr_mfc_iter *it = seq->private;
2995 const struct mr_table *mrt = it->mrt;
2996
2997 seq_printf(seq, "%08X %08X %-3hd",
2998 (__force u32) mfc->mfc_mcastgrp,
2999 (__force u32) mfc->mfc_origin,
3000 mfc->_c.mfc_parent);
3001
3002 if (it->cache != &mrt->mfc_unres_queue) {
3003 seq_printf(seq, " %8lu %8lu %8lu",
3004 mfc->_c.mfc_un.res.pkt,
3005 mfc->_c.mfc_un.res.bytes,
3006 mfc->_c.mfc_un.res.wrong_if);
3007 for (n = mfc->_c.mfc_un.res.minvif;
3008 n < mfc->_c.mfc_un.res.maxvif; n++) {
3009 if (VIF_EXISTS(mrt, n) &&
3010 mfc->_c.mfc_un.res.ttls[n] < 255)
3011 seq_printf(seq,
3012 " %2d:%-3d",
3013 n, mfc->_c.mfc_un.res.ttls[n]);
3014 }
3015 } else {
3016 /* unresolved mfc_caches don't contain
3017 * pkt, bytes and wrong_if values
3018 */
3019 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3020 }
3021 seq_putc(seq, '\n');
3022 }
3023 return 0;
3024 }
3025
3026 static const struct seq_operations ipmr_mfc_seq_ops = {
3027 .start = ipmr_mfc_seq_start,
3028 .next = mr_mfc_seq_next,
3029 .stop = mr_mfc_seq_stop,
3030 .show = ipmr_mfc_seq_show,
3031 };
3032 #endif
3033
3034 #ifdef CONFIG_IP_PIMSM_V2
3035 static const struct net_protocol pim_protocol = {
3036 .handler = pim_rcv,
3037 .netns_ok = 1,
3038 };
3039 #endif
3040
3041 static unsigned int ipmr_seq_read(struct net *net)
3042 {
3043 ASSERT_RTNL();
3044
3045 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3046 }
3047
3048 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3049 struct netlink_ext_ack *extack)
3050 {
3051 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3052 ipmr_mr_table_iter, &mrt_lock, extack);
3053 }
3054
3055 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3056 .family = RTNL_FAMILY_IPMR,
3057 .fib_seq_read = ipmr_seq_read,
3058 .fib_dump = ipmr_dump,
3059 .owner = THIS_MODULE,
3060 };
3061
3062 static int __net_init ipmr_notifier_init(struct net *net)
3063 {
3064 struct fib_notifier_ops *ops;
3065
3066 net->ipv4.ipmr_seq = 0;
3067
3068 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3069 if (IS_ERR(ops))
3070 return PTR_ERR(ops);
3071 net->ipv4.ipmr_notifier_ops = ops;
3072
3073 return 0;
3074 }
3075
3076 static void __net_exit ipmr_notifier_exit(struct net *net)
3077 {
3078 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3079 net->ipv4.ipmr_notifier_ops = NULL;
3080 }
3081
3082 /* Setup for IP multicast routing */
3083 static int __net_init ipmr_net_init(struct net *net)
3084 {
3085 int err;
3086
3087 err = ipmr_notifier_init(net);
3088 if (err)
3089 goto ipmr_notifier_fail;
3090
3091 err = ipmr_rules_init(net);
3092 if (err < 0)
3093 goto ipmr_rules_fail;
3094
3095 #ifdef CONFIG_PROC_FS
3096 err = -ENOMEM;
3097 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3098 sizeof(struct mr_vif_iter)))
3099 goto proc_vif_fail;
3100 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3101 sizeof(struct mr_mfc_iter)))
3102 goto proc_cache_fail;
3103 #endif
3104 return 0;
3105
3106 #ifdef CONFIG_PROC_FS
3107 proc_cache_fail:
3108 remove_proc_entry("ip_mr_vif", net->proc_net);
3109 proc_vif_fail:
3110 ipmr_rules_exit(net);
3111 #endif
3112 ipmr_rules_fail:
3113 ipmr_notifier_exit(net);
3114 ipmr_notifier_fail:
3115 return err;
3116 }
3117
3118 static void __net_exit ipmr_net_exit(struct net *net)
3119 {
3120 #ifdef CONFIG_PROC_FS
3121 remove_proc_entry("ip_mr_cache", net->proc_net);
3122 remove_proc_entry("ip_mr_vif", net->proc_net);
3123 #endif
3124 ipmr_notifier_exit(net);
3125 ipmr_rules_exit(net);
3126 }
3127
3128 static struct pernet_operations ipmr_net_ops = {
3129 .init = ipmr_net_init,
3130 .exit = ipmr_net_exit,
3131 };
3132
3133 int __init ip_mr_init(void)
3134 {
3135 int err;
3136
3137 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3138 sizeof(struct mfc_cache),
3139 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3140 NULL);
3141
3142 err = register_pernet_subsys(&ipmr_net_ops);
3143 if (err)
3144 goto reg_pernet_fail;
3145
3146 err = register_netdevice_notifier(&ip_mr_notifier);
3147 if (err)
3148 goto reg_notif_fail;
3149 #ifdef CONFIG_IP_PIMSM_V2
3150 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3151 pr_err("%s: can't add PIM protocol\n", __func__);
3152 err = -EAGAIN;
3153 goto add_proto_fail;
3154 }
3155 #endif
3156 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3157 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3158 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3159 ipmr_rtm_route, NULL, 0);
3160 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3161 ipmr_rtm_route, NULL, 0);
3162
3163 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3164 NULL, ipmr_rtm_dumplink, 0);
3165 return 0;
3166
3167 #ifdef CONFIG_IP_PIMSM_V2
3168 add_proto_fail:
3169 unregister_netdevice_notifier(&ip_mr_notifier);
3170 #endif
3171 reg_notif_fail:
3172 unregister_pernet_subsys(&ipmr_net_ops);
3173 reg_pernet_fail:
3174 kmem_cache_destroy(mrt_cachep);
3175 return err;
3176 }