?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
dm writecache: fix incorrect flush sequence when doing SSD mode commit
[linux/fpc-iii.git] / net / ipv4 / ipmr.c
blob6e68def66822f47fc08d94eddd32a4bd4f9fdfb0
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
115 static struct mr_table *ipmr_mr_table_iter(struct net *net,
116 struct mr_table *mrt)
117 {
118 struct mr_table *ret;
119
120 if (!mrt)
121 ret = list_entry_rcu(net->ipv4.mr_tables.next,
122 struct mr_table, list);
123 else
124 ret = list_entry_rcu(mrt->list.next,
125 struct mr_table, list);
126
127 if (&ret->list == &net->ipv4.mr_tables)
128 return NULL;
129 return ret;
130 }
131
132 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
133 {
134 struct mr_table *mrt;
135
136 ipmr_for_each_table(mrt, net) {
137 if (mrt->id == id)
138 return mrt;
139 }
140 return NULL;
141 }
142
143 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
144 struct mr_table **mrt)
145 {
146 int err;
147 struct ipmr_result res;
148 struct fib_lookup_arg arg = {
149 .result = &res,
150 .flags = FIB_LOOKUP_NOREF,
151 };
152
153 /* update flow if oif or iif point to device enslaved to l3mdev */
154 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
155
156 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
157 flowi4_to_flowi(flp4), 0, &arg);
158 if (err < 0)
159 return err;
160 *mrt = res.mrt;
161 return 0;
162 }
163
164 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
165 int flags, struct fib_lookup_arg *arg)
166 {
167 struct ipmr_result *res = arg->result;
168 struct mr_table *mrt;
169
170 switch (rule->action) {
171 case FR_ACT_TO_TBL:
172 break;
173 case FR_ACT_UNREACHABLE:
174 return -ENETUNREACH;
175 case FR_ACT_PROHIBIT:
176 return -EACCES;
177 case FR_ACT_BLACKHOLE:
178 default:
179 return -EINVAL;
180 }
181
182 arg->table = fib_rule_get_table(rule, arg);
183
184 mrt = ipmr_get_table(rule->fr_net, arg->table);
185 if (!mrt)
186 return -EAGAIN;
187 res->mrt = mrt;
188 return 0;
189 }
190
191 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
192 {
193 return 1;
194 }
195
196 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
197 FRA_GENERIC_POLICY,
198 };
199
200 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
201 struct fib_rule_hdr *frh, struct nlattr **tb,
202 struct netlink_ext_ack *extack)
203 {
204 return 0;
205 }
206
207 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
208 struct nlattr **tb)
209 {
210 return 1;
211 }
212
213 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
214 struct fib_rule_hdr *frh)
215 {
216 frh->dst_len = 0;
217 frh->src_len = 0;
218 frh->tos = 0;
219 return 0;
220 }
221
222 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
223 .family = RTNL_FAMILY_IPMR,
224 .rule_size = sizeof(struct ipmr_rule),
225 .addr_size = sizeof(u32),
226 .action = ipmr_rule_action,
227 .match = ipmr_rule_match,
228 .configure = ipmr_rule_configure,
229 .compare = ipmr_rule_compare,
230 .fill = ipmr_rule_fill,
231 .nlgroup = RTNLGRP_IPV4_RULE,
232 .policy = ipmr_rule_policy,
233 .owner = THIS_MODULE,
234 };
235
236 static int __net_init ipmr_rules_init(struct net *net)
237 {
238 struct fib_rules_ops *ops;
239 struct mr_table *mrt;
240 int err;
241
242 ops = fib_rules_register(&ipmr_rules_ops_template, net);
243 if (IS_ERR(ops))
244 return PTR_ERR(ops);
245
246 INIT_LIST_HEAD(&net->ipv4.mr_tables);
247
248 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
249 if (IS_ERR(mrt)) {
250 err = PTR_ERR(mrt);
251 goto err1;
252 }
253
254 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
255 if (err < 0)
256 goto err2;
257
258 net->ipv4.mr_rules_ops = ops;
259 return 0;
260
261 err2:
262 ipmr_free_table(mrt);
263 err1:
264 fib_rules_unregister(ops);
265 return err;
266 }
267
268 static void __net_exit ipmr_rules_exit(struct net *net)
269 {
270 struct mr_table *mrt, *next;
271
272 rtnl_lock();
273 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
274 list_del(&mrt->list);
275 ipmr_free_table(mrt);
276 }
277 fib_rules_unregister(net->ipv4.mr_rules_ops);
278 rtnl_unlock();
279 }
280
281 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
282 struct netlink_ext_ack *extack)
283 {
284 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
285 }
286
287 static unsigned int ipmr_rules_seq_read(struct net *net)
288 {
289 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
290 }
291
292 bool ipmr_rule_default(const struct fib_rule *rule)
293 {
294 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
295 }
296 EXPORT_SYMBOL(ipmr_rule_default);
297 #else
298 #define ipmr_for_each_table(mrt, net) \
299 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
300
301 static struct mr_table *ipmr_mr_table_iter(struct net *net,
302 struct mr_table *mrt)
303 {
304 if (!mrt)
305 return net->ipv4.mrt;
306 return NULL;
307 }
308
309 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
310 {
311 return net->ipv4.mrt;
312 }
313
314 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
315 struct mr_table **mrt)
316 {
317 *mrt = net->ipv4.mrt;
318 return 0;
319 }
320
321 static int __net_init ipmr_rules_init(struct net *net)
322 {
323 struct mr_table *mrt;
324
325 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
326 if (IS_ERR(mrt))
327 return PTR_ERR(mrt);
328 net->ipv4.mrt = mrt;
329 return 0;
330 }
331
332 static void __net_exit ipmr_rules_exit(struct net *net)
333 {
334 rtnl_lock();
335 ipmr_free_table(net->ipv4.mrt);
336 net->ipv4.mrt = NULL;
337 rtnl_unlock();
338 }
339
340 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
341 struct netlink_ext_ack *extack)
342 {
343 return 0;
344 }
345
346 static unsigned int ipmr_rules_seq_read(struct net *net)
347 {
348 return 0;
349 }
350
351 bool ipmr_rule_default(const struct fib_rule *rule)
352 {
353 return true;
354 }
355 EXPORT_SYMBOL(ipmr_rule_default);
356 #endif
357
358 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
359 const void *ptr)
360 {
361 const struct mfc_cache_cmp_arg *cmparg = arg->key;
362 struct mfc_cache *c = (struct mfc_cache *)ptr;
363
364 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
365 cmparg->mfc_origin != c->mfc_origin;
366 }
367
368 static const struct rhashtable_params ipmr_rht_params = {
369 .head_offset = offsetof(struct mr_mfc, mnode),
370 .key_offset = offsetof(struct mfc_cache, cmparg),
371 .key_len = sizeof(struct mfc_cache_cmp_arg),
372 .nelem_hint = 3,
373 .obj_cmpfn = ipmr_hash_cmp,
374 .automatic_shrinking = true,
375 };
376
377 static void ipmr_new_table_set(struct mr_table *mrt,
378 struct net *net)
379 {
380 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
381 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
382 #endif
383 }
384
385 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
386 .mfc_mcastgrp = htonl(INADDR_ANY),
387 .mfc_origin = htonl(INADDR_ANY),
388 };
389
390 static struct mr_table_ops ipmr_mr_table_ops = {
391 .rht_params = &ipmr_rht_params,
392 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
393 };
394
395 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
396 {
397 struct mr_table *mrt;
398
399 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
400 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
401 return ERR_PTR(-EINVAL);
402
403 mrt = ipmr_get_table(net, id);
404 if (mrt)
405 return mrt;
406
407 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
408 ipmr_expire_process, ipmr_new_table_set);
409 }
410
411 static void ipmr_free_table(struct mr_table *mrt)
412 {
413 del_timer_sync(&mrt->ipmr_expire_timer);
414 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
415 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
416 rhltable_destroy(&mrt->mfc_hash);
417 kfree(mrt);
418 }
419
420 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
421
422 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
423 {
424 struct net *net = dev_net(dev);
425
426 dev_close(dev);
427
428 dev = __dev_get_by_name(net, "tunl0");
429 if (dev) {
430 const struct net_device_ops *ops = dev->netdev_ops;
431 struct ifreq ifr;
432 struct ip_tunnel_parm p;
433
434 memset(&p, 0, sizeof(p));
435 p.iph.daddr = v->vifc_rmt_addr.s_addr;
436 p.iph.saddr = v->vifc_lcl_addr.s_addr;
437 p.iph.version = 4;
438 p.iph.ihl = 5;
439 p.iph.protocol = IPPROTO_IPIP;
440 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
441 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
442
443 if (ops->ndo_do_ioctl) {
444 mm_segment_t oldfs = get_fs();
445
446 set_fs(KERNEL_DS);
447 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
448 set_fs(oldfs);
449 }
450 }
451 }
452
453 /* Initialize ipmr pimreg/tunnel in_device */
454 static bool ipmr_init_vif_indev(const struct net_device *dev)
455 {
456 struct in_device *in_dev;
457
458 ASSERT_RTNL();
459
460 in_dev = __in_dev_get_rtnl(dev);
461 if (!in_dev)
462 return false;
463 ipv4_devconf_setall(in_dev);
464 neigh_parms_data_state_setall(in_dev->arp_parms);
465 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
466
467 return true;
468 }
469
470 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
471 {
472 struct net_device *dev;
473
474 dev = __dev_get_by_name(net, "tunl0");
475
476 if (dev) {
477 const struct net_device_ops *ops = dev->netdev_ops;
478 int err;
479 struct ifreq ifr;
480 struct ip_tunnel_parm p;
481
482 memset(&p, 0, sizeof(p));
483 p.iph.daddr = v->vifc_rmt_addr.s_addr;
484 p.iph.saddr = v->vifc_lcl_addr.s_addr;
485 p.iph.version = 4;
486 p.iph.ihl = 5;
487 p.iph.protocol = IPPROTO_IPIP;
488 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
489 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
490
491 if (ops->ndo_do_ioctl) {
492 mm_segment_t oldfs = get_fs();
493
494 set_fs(KERNEL_DS);
495 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
496 set_fs(oldfs);
497 } else {
498 err = -EOPNOTSUPP;
499 }
500 dev = NULL;
501
502 if (err == 0 &&
503 (dev = __dev_get_by_name(net, p.name)) != NULL) {
504 dev->flags |= IFF_MULTICAST;
505 if (!ipmr_init_vif_indev(dev))
506 goto failure;
507 if (dev_open(dev, NULL))
508 goto failure;
509 dev_hold(dev);
510 }
511 }
512 return dev;
513
514 failure:
515 unregister_netdevice(dev);
516 return NULL;
517 }
518
519 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
520 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
521 {
522 struct net *net = dev_net(dev);
523 struct mr_table *mrt;
524 struct flowi4 fl4 = {
525 .flowi4_oif = dev->ifindex,
526 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
527 .flowi4_mark = skb->mark,
528 };
529 int err;
530
531 err = ipmr_fib_lookup(net, &fl4, &mrt);
532 if (err < 0) {
533 kfree_skb(skb);
534 return err;
535 }
536
537 read_lock(&mrt_lock);
538 dev->stats.tx_bytes += skb->len;
539 dev->stats.tx_packets++;
540 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
541 read_unlock(&mrt_lock);
542 kfree_skb(skb);
543 return NETDEV_TX_OK;
544 }
545
546 static int reg_vif_get_iflink(const struct net_device *dev)
547 {
548 return 0;
549 }
550
551 static const struct net_device_ops reg_vif_netdev_ops = {
552 .ndo_start_xmit = reg_vif_xmit,
553 .ndo_get_iflink = reg_vif_get_iflink,
554 };
555
556 static void reg_vif_setup(struct net_device *dev)
557 {
558 dev->type = ARPHRD_PIMREG;
559 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
560 dev->flags = IFF_NOARP;
561 dev->netdev_ops = &reg_vif_netdev_ops;
562 dev->needs_free_netdev = true;
563 dev->features |= NETIF_F_NETNS_LOCAL;
564 }
565
566 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
567 {
568 struct net_device *dev;
569 char name[IFNAMSIZ];
570
571 if (mrt->id == RT_TABLE_DEFAULT)
572 sprintf(name, "pimreg");
573 else
574 sprintf(name, "pimreg%u", mrt->id);
575
576 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
577
578 if (!dev)
579 return NULL;
580
581 dev_net_set(dev, net);
582
583 if (register_netdevice(dev)) {
584 free_netdev(dev);
585 return NULL;
586 }
587
588 if (!ipmr_init_vif_indev(dev))
589 goto failure;
590 if (dev_open(dev, NULL))
591 goto failure;
592
593 dev_hold(dev);
594
595 return dev;
596
597 failure:
598 unregister_netdevice(dev);
599 return NULL;
600 }
601
602 /* called with rcu_read_lock() */
603 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
604 unsigned int pimlen)
605 {
606 struct net_device *reg_dev = NULL;
607 struct iphdr *encap;
608
609 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
610 /* Check that:
611 * a. packet is really sent to a multicast group
612 * b. packet is not a NULL-REGISTER
613 * c. packet is not truncated
614 */
615 if (!ipv4_is_multicast(encap->daddr) ||
616 encap->tot_len == 0 ||
617 ntohs(encap->tot_len) + pimlen > skb->len)
618 return 1;
619
620 read_lock(&mrt_lock);
621 if (mrt->mroute_reg_vif_num >= 0)
622 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
623 read_unlock(&mrt_lock);
624
625 if (!reg_dev)
626 return 1;
627
628 skb->mac_header = skb->network_header;
629 skb_pull(skb, (u8 *)encap - skb->data);
630 skb_reset_network_header(skb);
631 skb->protocol = htons(ETH_P_IP);
632 skb->ip_summed = CHECKSUM_NONE;
633
634 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
635
636 netif_rx(skb);
637
638 return NET_RX_SUCCESS;
639 }
640 #else
641 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
642 {
643 return NULL;
644 }
645 #endif
646
647 static int call_ipmr_vif_entry_notifiers(struct net *net,
648 enum fib_event_type event_type,
649 struct vif_device *vif,
650 vifi_t vif_index, u32 tb_id)
651 {
652 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
653 vif, vif_index, tb_id,
654 &net->ipv4.ipmr_seq);
655 }
656
657 static int call_ipmr_mfc_entry_notifiers(struct net *net,
658 enum fib_event_type event_type,
659 struct mfc_cache *mfc, u32 tb_id)
660 {
661 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
662 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
663 }
664
665 /**
666 * vif_delete - Delete a VIF entry
667 * @notify: Set to 1, if the caller is a notifier_call
668 */
669 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
670 struct list_head *head)
671 {
672 struct net *net = read_pnet(&mrt->net);
673 struct vif_device *v;
674 struct net_device *dev;
675 struct in_device *in_dev;
676
677 if (vifi < 0 || vifi >= mrt->maxvif)
678 return -EADDRNOTAVAIL;
679
680 v = &mrt->vif_table[vifi];
681
682 if (VIF_EXISTS(mrt, vifi))
683 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
684 mrt->id);
685
686 write_lock_bh(&mrt_lock);
687 dev = v->dev;
688 v->dev = NULL;
689
690 if (!dev) {
691 write_unlock_bh(&mrt_lock);
692 return -EADDRNOTAVAIL;
693 }
694
695 if (vifi == mrt->mroute_reg_vif_num)
696 mrt->mroute_reg_vif_num = -1;
697
698 if (vifi + 1 == mrt->maxvif) {
699 int tmp;
700
701 for (tmp = vifi - 1; tmp >= 0; tmp--) {
702 if (VIF_EXISTS(mrt, tmp))
703 break;
704 }
705 mrt->maxvif = tmp+1;
706 }
707
708 write_unlock_bh(&mrt_lock);
709
710 dev_set_allmulti(dev, -1);
711
712 in_dev = __in_dev_get_rtnl(dev);
713 if (in_dev) {
714 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
715 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
716 NETCONFA_MC_FORWARDING,
717 dev->ifindex, &in_dev->cnf);
718 ip_rt_multicast_event(in_dev);
719 }
720
721 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
722 unregister_netdevice_queue(dev, head);
723
724 dev_put(dev);
725 return 0;
726 }
727
728 static void ipmr_cache_free_rcu(struct rcu_head *head)
729 {
730 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
731
732 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
733 }
734
735 static void ipmr_cache_free(struct mfc_cache *c)
736 {
737 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
738 }
739
740 /* Destroy an unresolved cache entry, killing queued skbs
741 * and reporting error to netlink readers.
742 */
743 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
744 {
745 struct net *net = read_pnet(&mrt->net);
746 struct sk_buff *skb;
747 struct nlmsgerr *e;
748
749 atomic_dec(&mrt->cache_resolve_queue_len);
750
751 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
752 if (ip_hdr(skb)->version == 0) {
753 struct nlmsghdr *nlh = skb_pull(skb,
754 sizeof(struct iphdr));
755 nlh->nlmsg_type = NLMSG_ERROR;
756 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
757 skb_trim(skb, nlh->nlmsg_len);
758 e = nlmsg_data(nlh);
759 e->error = -ETIMEDOUT;
760 memset(&e->msg, 0, sizeof(e->msg));
761
762 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
763 } else {
764 kfree_skb(skb);
765 }
766 }
767
768 ipmr_cache_free(c);
769 }
770
771 /* Timer process for the unresolved queue. */
772 static void ipmr_expire_process(struct timer_list *t)
773 {
774 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
775 struct mr_mfc *c, *next;
776 unsigned long expires;
777 unsigned long now;
778
779 if (!spin_trylock(&mfc_unres_lock)) {
780 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
781 return;
782 }
783
784 if (list_empty(&mrt->mfc_unres_queue))
785 goto out;
786
787 now = jiffies;
788 expires = 10*HZ;
789
790 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
791 if (time_after(c->mfc_un.unres.expires, now)) {
792 unsigned long interval = c->mfc_un.unres.expires - now;
793 if (interval < expires)
794 expires = interval;
795 continue;
796 }
797
798 list_del(&c->list);
799 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
800 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
801 }
802
803 if (!list_empty(&mrt->mfc_unres_queue))
804 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
805
806 out:
807 spin_unlock(&mfc_unres_lock);
808 }
809
810 /* Fill oifs list. It is called under write locked mrt_lock. */
811 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
812 unsigned char *ttls)
813 {
814 int vifi;
815
816 cache->mfc_un.res.minvif = MAXVIFS;
817 cache->mfc_un.res.maxvif = 0;
818 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
819
820 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
821 if (VIF_EXISTS(mrt, vifi) &&
822 ttls[vifi] && ttls[vifi] < 255) {
823 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
824 if (cache->mfc_un.res.minvif > vifi)
825 cache->mfc_un.res.minvif = vifi;
826 if (cache->mfc_un.res.maxvif <= vifi)
827 cache->mfc_un.res.maxvif = vifi + 1;
828 }
829 }
830 cache->mfc_un.res.lastuse = jiffies;
831 }
832
833 static int vif_add(struct net *net, struct mr_table *mrt,
834 struct vifctl *vifc, int mrtsock)
835 {
836 struct netdev_phys_item_id ppid = { };
837 int vifi = vifc->vifc_vifi;
838 struct vif_device *v = &mrt->vif_table[vifi];
839 struct net_device *dev;
840 struct in_device *in_dev;
841 int err;
842
843 /* Is vif busy ? */
844 if (VIF_EXISTS(mrt, vifi))
845 return -EADDRINUSE;
846
847 switch (vifc->vifc_flags) {
848 case VIFF_REGISTER:
849 if (!ipmr_pimsm_enabled())
850 return -EINVAL;
851 /* Special Purpose VIF in PIM
852 * All the packets will be sent to the daemon
853 */
854 if (mrt->mroute_reg_vif_num >= 0)
855 return -EADDRINUSE;
856 dev = ipmr_reg_vif(net, mrt);
857 if (!dev)
858 return -ENOBUFS;
859 err = dev_set_allmulti(dev, 1);
860 if (err) {
861 unregister_netdevice(dev);
862 dev_put(dev);
863 return err;
864 }
865 break;
866 case VIFF_TUNNEL:
867 dev = ipmr_new_tunnel(net, vifc);
868 if (!dev)
869 return -ENOBUFS;
870 err = dev_set_allmulti(dev, 1);
871 if (err) {
872 ipmr_del_tunnel(dev, vifc);
873 dev_put(dev);
874 return err;
875 }
876 break;
877 case VIFF_USE_IFINDEX:
878 case 0:
879 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
880 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
881 if (dev && !__in_dev_get_rtnl(dev)) {
882 dev_put(dev);
883 return -EADDRNOTAVAIL;
884 }
885 } else {
886 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
887 }
888 if (!dev)
889 return -EADDRNOTAVAIL;
890 err = dev_set_allmulti(dev, 1);
891 if (err) {
892 dev_put(dev);
893 return err;
894 }
895 break;
896 default:
897 return -EINVAL;
898 }
899
900 in_dev = __in_dev_get_rtnl(dev);
901 if (!in_dev) {
902 dev_put(dev);
903 return -EADDRNOTAVAIL;
904 }
905 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
906 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
907 dev->ifindex, &in_dev->cnf);
908 ip_rt_multicast_event(in_dev);
909
910 /* Fill in the VIF structures */
911 vif_device_init(v, dev, vifc->vifc_rate_limit,
912 vifc->vifc_threshold,
913 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
914 (VIFF_TUNNEL | VIFF_REGISTER));
915
916 err = dev_get_port_parent_id(dev, &ppid, true);
917 if (err == 0) {
918 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
919 v->dev_parent_id.id_len = ppid.id_len;
920 } else {
921 v->dev_parent_id.id_len = 0;
922 }
923
924 v->local = vifc->vifc_lcl_addr.s_addr;
925 v->remote = vifc->vifc_rmt_addr.s_addr;
926
927 /* And finish update writing critical data */
928 write_lock_bh(&mrt_lock);
929 v->dev = dev;
930 if (v->flags & VIFF_REGISTER)
931 mrt->mroute_reg_vif_num = vifi;
932 if (vifi+1 > mrt->maxvif)
933 mrt->maxvif = vifi+1;
934 write_unlock_bh(&mrt_lock);
935 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
936 return 0;
937 }
938
939 /* called with rcu_read_lock() */
940 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
941 __be32 origin,
942 __be32 mcastgrp)
943 {
944 struct mfc_cache_cmp_arg arg = {
945 .mfc_mcastgrp = mcastgrp,
946 .mfc_origin = origin
947 };
948
949 return mr_mfc_find(mrt, &arg);
950 }
951
952 /* Look for a (*,G) entry */
953 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
954 __be32 mcastgrp, int vifi)
955 {
956 struct mfc_cache_cmp_arg arg = {
957 .mfc_mcastgrp = mcastgrp,
958 .mfc_origin = htonl(INADDR_ANY)
959 };
960
961 if (mcastgrp == htonl(INADDR_ANY))
962 return mr_mfc_find_any_parent(mrt, vifi);
963 return mr_mfc_find_any(mrt, vifi, &arg);
964 }
965
966 /* Look for a (S,G,iif) entry if parent != -1 */
967 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
968 __be32 origin, __be32 mcastgrp,
969 int parent)
970 {
971 struct mfc_cache_cmp_arg arg = {
972 .mfc_mcastgrp = mcastgrp,
973 .mfc_origin = origin,
974 };
975
976 return mr_mfc_find_parent(mrt, &arg, parent);
977 }
978
979 /* Allocate a multicast cache entry */
980 static struct mfc_cache *ipmr_cache_alloc(void)
981 {
982 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
983
984 if (c) {
985 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
986 c->_c.mfc_un.res.minvif = MAXVIFS;
987 c->_c.free = ipmr_cache_free_rcu;
988 refcount_set(&c->_c.mfc_un.res.refcount, 1);
989 }
990 return c;
991 }
992
993 static struct mfc_cache *ipmr_cache_alloc_unres(void)
994 {
995 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
996
997 if (c) {
998 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
999 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1000 }
1001 return c;
1002 }
1003
1004 /* A cache entry has gone into a resolved state from queued */
1005 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1006 struct mfc_cache *uc, struct mfc_cache *c)
1007 {
1008 struct sk_buff *skb;
1009 struct nlmsgerr *e;
1010
1011 /* Play the pending entries through our router */
1012 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1013 if (ip_hdr(skb)->version == 0) {
1014 struct nlmsghdr *nlh = skb_pull(skb,
1015 sizeof(struct iphdr));
1016
1017 if (mr_fill_mroute(mrt, skb, &c->_c,
1018 nlmsg_data(nlh)) > 0) {
1019 nlh->nlmsg_len = skb_tail_pointer(skb) -
1020 (u8 *)nlh;
1021 } else {
1022 nlh->nlmsg_type = NLMSG_ERROR;
1023 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1024 skb_trim(skb, nlh->nlmsg_len);
1025 e = nlmsg_data(nlh);
1026 e->error = -EMSGSIZE;
1027 memset(&e->msg, 0, sizeof(e->msg));
1028 }
1029
1030 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1031 } else {
1032 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1033 }
1034 }
1035 }
1036
1037 /* Bounce a cache query up to mrouted and netlink.
1038 *
1039 * Called under mrt_lock.
1040 */
1041 static int ipmr_cache_report(struct mr_table *mrt,
1042 struct sk_buff *pkt, vifi_t vifi, int assert)
1043 {
1044 const int ihl = ip_hdrlen(pkt);
1045 struct sock *mroute_sk;
1046 struct igmphdr *igmp;
1047 struct igmpmsg *msg;
1048 struct sk_buff *skb;
1049 int ret;
1050
1051 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1052 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1053 else
1054 skb = alloc_skb(128, GFP_ATOMIC);
1055
1056 if (!skb)
1057 return -ENOBUFS;
1058
1059 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1060 /* Ugly, but we have no choice with this interface.
1061 * Duplicate old header, fix ihl, length etc.
1062 * And all this only to mangle msg->im_msgtype and
1063 * to set msg->im_mbz to "mbz" :-)
1064 */
1065 skb_push(skb, sizeof(struct iphdr));
1066 skb_reset_network_header(skb);
1067 skb_reset_transport_header(skb);
1068 msg = (struct igmpmsg *)skb_network_header(skb);
1069 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1070 msg->im_msgtype = assert;
1071 msg->im_mbz = 0;
1072 if (assert == IGMPMSG_WRVIFWHOLE)
1073 msg->im_vif = vifi;
1074 else
1075 msg->im_vif = mrt->mroute_reg_vif_num;
1076 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1077 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1078 sizeof(struct iphdr));
1079 } else {
1080 /* Copy the IP header */
1081 skb_set_network_header(skb, skb->len);
1082 skb_put(skb, ihl);
1083 skb_copy_to_linear_data(skb, pkt->data, ihl);
1084 /* Flag to the kernel this is a route add */
1085 ip_hdr(skb)->protocol = 0;
1086 msg = (struct igmpmsg *)skb_network_header(skb);
1087 msg->im_vif = vifi;
1088 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1089 /* Add our header */
1090 igmp = skb_put(skb, sizeof(struct igmphdr));
1091 igmp->type = assert;
1092 msg->im_msgtype = assert;
1093 igmp->code = 0;
1094 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1095 skb->transport_header = skb->network_header;
1096 }
1097
1098 rcu_read_lock();
1099 mroute_sk = rcu_dereference(mrt->mroute_sk);
1100 if (!mroute_sk) {
1101 rcu_read_unlock();
1102 kfree_skb(skb);
1103 return -EINVAL;
1104 }
1105
1106 igmpmsg_netlink_event(mrt, skb);
1107
1108 /* Deliver to mrouted */
1109 ret = sock_queue_rcv_skb(mroute_sk, skb);
1110 rcu_read_unlock();
1111 if (ret < 0) {
1112 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1113 kfree_skb(skb);
1114 }
1115
1116 return ret;
1117 }
1118
1119 /* Queue a packet for resolution. It gets locked cache entry! */
1120 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1121 struct sk_buff *skb, struct net_device *dev)
1122 {
1123 const struct iphdr *iph = ip_hdr(skb);
1124 struct mfc_cache *c;
1125 bool found = false;
1126 int err;
1127
1128 spin_lock_bh(&mfc_unres_lock);
1129 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1130 if (c->mfc_mcastgrp == iph->daddr &&
1131 c->mfc_origin == iph->saddr) {
1132 found = true;
1133 break;
1134 }
1135 }
1136
1137 if (!found) {
1138 /* Create a new entry if allowable */
1139 c = ipmr_cache_alloc_unres();
1140 if (!c) {
1141 spin_unlock_bh(&mfc_unres_lock);
1142
1143 kfree_skb(skb);
1144 return -ENOBUFS;
1145 }
1146
1147 /* Fill in the new cache entry */
1148 c->_c.mfc_parent = -1;
1149 c->mfc_origin = iph->saddr;
1150 c->mfc_mcastgrp = iph->daddr;
1151
1152 /* Reflect first query at mrouted. */
1153 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1154
1155 if (err < 0) {
1156 /* If the report failed throw the cache entry
1157 out - Brad Parker
1158 */
1159 spin_unlock_bh(&mfc_unres_lock);
1160
1161 ipmr_cache_free(c);
1162 kfree_skb(skb);
1163 return err;
1164 }
1165
1166 atomic_inc(&mrt->cache_resolve_queue_len);
1167 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1168 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1169
1170 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1171 mod_timer(&mrt->ipmr_expire_timer,
1172 c->_c.mfc_un.unres.expires);
1173 }
1174
1175 /* See if we can append the packet */
1176 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1177 kfree_skb(skb);
1178 err = -ENOBUFS;
1179 } else {
1180 if (dev) {
1181 skb->dev = dev;
1182 skb->skb_iif = dev->ifindex;
1183 }
1184 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1185 err = 0;
1186 }
1187
1188 spin_unlock_bh(&mfc_unres_lock);
1189 return err;
1190 }
1191
1192 /* MFC cache manipulation by user space mroute daemon */
1193
1194 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1195 {
1196 struct net *net = read_pnet(&mrt->net);
1197 struct mfc_cache *c;
1198
1199 /* The entries are added/deleted only under RTNL */
1200 rcu_read_lock();
1201 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1202 mfc->mfcc_mcastgrp.s_addr, parent);
1203 rcu_read_unlock();
1204 if (!c)
1205 return -ENOENT;
1206 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1207 list_del_rcu(&c->_c.list);
1208 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1209 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1210 mr_cache_put(&c->_c);
1211
1212 return 0;
1213 }
1214
1215 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1216 struct mfcctl *mfc, int mrtsock, int parent)
1217 {
1218 struct mfc_cache *uc, *c;
1219 struct mr_mfc *_uc;
1220 bool found;
1221 int ret;
1222
1223 if (mfc->mfcc_parent >= MAXVIFS)
1224 return -ENFILE;
1225
1226 /* The entries are added/deleted only under RTNL */
1227 rcu_read_lock();
1228 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1229 mfc->mfcc_mcastgrp.s_addr, parent);
1230 rcu_read_unlock();
1231 if (c) {
1232 write_lock_bh(&mrt_lock);
1233 c->_c.mfc_parent = mfc->mfcc_parent;
1234 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1235 if (!mrtsock)
1236 c->_c.mfc_flags |= MFC_STATIC;
1237 write_unlock_bh(&mrt_lock);
1238 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1239 mrt->id);
1240 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1241 return 0;
1242 }
1243
1244 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1245 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1246 return -EINVAL;
1247
1248 c = ipmr_cache_alloc();
1249 if (!c)
1250 return -ENOMEM;
1251
1252 c->mfc_origin = mfc->mfcc_origin.s_addr;
1253 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1254 c->_c.mfc_parent = mfc->mfcc_parent;
1255 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1256 if (!mrtsock)
1257 c->_c.mfc_flags |= MFC_STATIC;
1258
1259 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1260 ipmr_rht_params);
1261 if (ret) {
1262 pr_err("ipmr: rhtable insert error %d\n", ret);
1263 ipmr_cache_free(c);
1264 return ret;
1265 }
1266 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1267 /* Check to see if we resolved a queued list. If so we
1268 * need to send on the frames and tidy up.
1269 */
1270 found = false;
1271 spin_lock_bh(&mfc_unres_lock);
1272 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1273 uc = (struct mfc_cache *)_uc;
1274 if (uc->mfc_origin == c->mfc_origin &&
1275 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1276 list_del(&_uc->list);
1277 atomic_dec(&mrt->cache_resolve_queue_len);
1278 found = true;
1279 break;
1280 }
1281 }
1282 if (list_empty(&mrt->mfc_unres_queue))
1283 del_timer(&mrt->ipmr_expire_timer);
1284 spin_unlock_bh(&mfc_unres_lock);
1285
1286 if (found) {
1287 ipmr_cache_resolve(net, mrt, uc, c);
1288 ipmr_cache_free(uc);
1289 }
1290 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1291 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1292 return 0;
1293 }
1294
1295 /* Close the multicast socket, and clear the vif tables etc */
1296 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1297 {
1298 struct net *net = read_pnet(&mrt->net);
1299 struct mr_mfc *c, *tmp;
1300 struct mfc_cache *cache;
1301 LIST_HEAD(list);
1302 int i;
1303
1304 /* Shut down all active vif entries */
1305 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1306 for (i = 0; i < mrt->maxvif; i++) {
1307 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1308 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1309 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1310 continue;
1311 vif_delete(mrt, i, 0, &list);
1312 }
1313 unregister_netdevice_many(&list);
1314 }
1315
1316 /* Wipe the cache */
1317 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1318 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1319 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1320 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1321 continue;
1322 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1323 list_del_rcu(&c->list);
1324 cache = (struct mfc_cache *)c;
1325 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1326 mrt->id);
1327 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1328 mr_cache_put(c);
1329 }
1330 }
1331
1332 if (flags & MRT_FLUSH_MFC) {
1333 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1334 spin_lock_bh(&mfc_unres_lock);
1335 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1336 list_del(&c->list);
1337 cache = (struct mfc_cache *)c;
1338 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1339 ipmr_destroy_unres(mrt, cache);
1340 }
1341 spin_unlock_bh(&mfc_unres_lock);
1342 }
1343 }
1344 }
1345
1346 /* called from ip_ra_control(), before an RCU grace period,
1347 * we dont need to call synchronize_rcu() here
1348 */
1349 static void mrtsock_destruct(struct sock *sk)
1350 {
1351 struct net *net = sock_net(sk);
1352 struct mr_table *mrt;
1353
1354 rtnl_lock();
1355 ipmr_for_each_table(mrt, net) {
1356 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1357 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1358 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1359 NETCONFA_MC_FORWARDING,
1360 NETCONFA_IFINDEX_ALL,
1361 net->ipv4.devconf_all);
1362 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1363 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1364 }
1365 }
1366 rtnl_unlock();
1367 }
1368
1369 /* Socket options and virtual interface manipulation. The whole
1370 * virtual interface system is a complete heap, but unfortunately
1371 * that's how BSD mrouted happens to think. Maybe one day with a proper
1372 * MOSPF/PIM router set up we can clean this up.
1373 */
1374
1375 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1376 unsigned int optlen)
1377 {
1378 struct net *net = sock_net(sk);
1379 int val, ret = 0, parent = 0;
1380 struct mr_table *mrt;
1381 struct vifctl vif;
1382 struct mfcctl mfc;
1383 bool do_wrvifwhole;
1384 u32 uval;
1385
1386 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1387 rtnl_lock();
1388 if (sk->sk_type != SOCK_RAW ||
1389 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1390 ret = -EOPNOTSUPP;
1391 goto out_unlock;
1392 }
1393
1394 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1395 if (!mrt) {
1396 ret = -ENOENT;
1397 goto out_unlock;
1398 }
1399 if (optname != MRT_INIT) {
1400 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1401 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1402 ret = -EACCES;
1403 goto out_unlock;
1404 }
1405 }
1406
1407 switch (optname) {
1408 case MRT_INIT:
1409 if (optlen != sizeof(int)) {
1410 ret = -EINVAL;
1411 break;
1412 }
1413 if (rtnl_dereference(mrt->mroute_sk)) {
1414 ret = -EADDRINUSE;
1415 break;
1416 }
1417
1418 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1419 if (ret == 0) {
1420 rcu_assign_pointer(mrt->mroute_sk, sk);
1421 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1422 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1423 NETCONFA_MC_FORWARDING,
1424 NETCONFA_IFINDEX_ALL,
1425 net->ipv4.devconf_all);
1426 }
1427 break;
1428 case MRT_DONE:
1429 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1430 ret = -EACCES;
1431 } else {
1432 /* We need to unlock here because mrtsock_destruct takes
1433 * care of rtnl itself and we can't change that due to
1434 * the IP_ROUTER_ALERT setsockopt which runs without it.
1435 */
1436 rtnl_unlock();
1437 ret = ip_ra_control(sk, 0, NULL);
1438 goto out;
1439 }
1440 break;
1441 case MRT_ADD_VIF:
1442 case MRT_DEL_VIF:
1443 if (optlen != sizeof(vif)) {
1444 ret = -EINVAL;
1445 break;
1446 }
1447 if (copy_from_user(&vif, optval, sizeof(vif))) {
1448 ret = -EFAULT;
1449 break;
1450 }
1451 if (vif.vifc_vifi >= MAXVIFS) {
1452 ret = -ENFILE;
1453 break;
1454 }
1455 if (optname == MRT_ADD_VIF) {
1456 ret = vif_add(net, mrt, &vif,
1457 sk == rtnl_dereference(mrt->mroute_sk));
1458 } else {
1459 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1460 }
1461 break;
1462 /* Manipulate the forwarding caches. These live
1463 * in a sort of kernel/user symbiosis.
1464 */
1465 case MRT_ADD_MFC:
1466 case MRT_DEL_MFC:
1467 parent = -1;
1468 /* fall through */
1469 case MRT_ADD_MFC_PROXY:
1470 case MRT_DEL_MFC_PROXY:
1471 if (optlen != sizeof(mfc)) {
1472 ret = -EINVAL;
1473 break;
1474 }
1475 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1476 ret = -EFAULT;
1477 break;
1478 }
1479 if (parent == 0)
1480 parent = mfc.mfcc_parent;
1481 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1482 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1483 else
1484 ret = ipmr_mfc_add(net, mrt, &mfc,
1485 sk == rtnl_dereference(mrt->mroute_sk),
1486 parent);
1487 break;
1488 case MRT_FLUSH:
1489 if (optlen != sizeof(val)) {
1490 ret = -EINVAL;
1491 break;
1492 }
1493 if (get_user(val, (int __user *)optval)) {
1494 ret = -EFAULT;
1495 break;
1496 }
1497 mroute_clean_tables(mrt, val);
1498 break;
1499 /* Control PIM assert. */
1500 case MRT_ASSERT:
1501 if (optlen != sizeof(val)) {
1502 ret = -EINVAL;
1503 break;
1504 }
1505 if (get_user(val, (int __user *)optval)) {
1506 ret = -EFAULT;
1507 break;
1508 }
1509 mrt->mroute_do_assert = val;
1510 break;
1511 case MRT_PIM:
1512 if (!ipmr_pimsm_enabled()) {
1513 ret = -ENOPROTOOPT;
1514 break;
1515 }
1516 if (optlen != sizeof(val)) {
1517 ret = -EINVAL;
1518 break;
1519 }
1520 if (get_user(val, (int __user *)optval)) {
1521 ret = -EFAULT;
1522 break;
1523 }
1524
1525 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1526 val = !!val;
1527 if (val != mrt->mroute_do_pim) {
1528 mrt->mroute_do_pim = val;
1529 mrt->mroute_do_assert = val;
1530 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1531 }
1532 break;
1533 case MRT_TABLE:
1534 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1535 ret = -ENOPROTOOPT;
1536 break;
1537 }
1538 if (optlen != sizeof(uval)) {
1539 ret = -EINVAL;
1540 break;
1541 }
1542 if (get_user(uval, (u32 __user *)optval)) {
1543 ret = -EFAULT;
1544 break;
1545 }
1546
1547 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1548 ret = -EBUSY;
1549 } else {
1550 mrt = ipmr_new_table(net, uval);
1551 if (IS_ERR(mrt))
1552 ret = PTR_ERR(mrt);
1553 else
1554 raw_sk(sk)->ipmr_table = uval;
1555 }
1556 break;
1557 /* Spurious command, or MRT_VERSION which you cannot set. */
1558 default:
1559 ret = -ENOPROTOOPT;
1560 }
1561 out_unlock:
1562 rtnl_unlock();
1563 out:
1564 return ret;
1565 }
1566
1567 /* Getsock opt support for the multicast routing system. */
1568 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1569 {
1570 int olr;
1571 int val;
1572 struct net *net = sock_net(sk);
1573 struct mr_table *mrt;
1574
1575 if (sk->sk_type != SOCK_RAW ||
1576 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1577 return -EOPNOTSUPP;
1578
1579 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1580 if (!mrt)
1581 return -ENOENT;
1582
1583 switch (optname) {
1584 case MRT_VERSION:
1585 val = 0x0305;
1586 break;
1587 case MRT_PIM:
1588 if (!ipmr_pimsm_enabled())
1589 return -ENOPROTOOPT;
1590 val = mrt->mroute_do_pim;
1591 break;
1592 case MRT_ASSERT:
1593 val = mrt->mroute_do_assert;
1594 break;
1595 default:
1596 return -ENOPROTOOPT;
1597 }
1598
1599 if (get_user(olr, optlen))
1600 return -EFAULT;
1601 olr = min_t(unsigned int, olr, sizeof(int));
1602 if (olr < 0)
1603 return -EINVAL;
1604 if (put_user(olr, optlen))
1605 return -EFAULT;
1606 if (copy_to_user(optval, &val, olr))
1607 return -EFAULT;
1608 return 0;
1609 }
1610
1611 /* The IP multicast ioctl support routines. */
1612 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1613 {
1614 struct sioc_sg_req sr;
1615 struct sioc_vif_req vr;
1616 struct vif_device *vif;
1617 struct mfc_cache *c;
1618 struct net *net = sock_net(sk);
1619 struct mr_table *mrt;
1620
1621 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1622 if (!mrt)
1623 return -ENOENT;
1624
1625 switch (cmd) {
1626 case SIOCGETVIFCNT:
1627 if (copy_from_user(&vr, arg, sizeof(vr)))
1628 return -EFAULT;
1629 if (vr.vifi >= mrt->maxvif)
1630 return -EINVAL;
1631 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1632 read_lock(&mrt_lock);
1633 vif = &mrt->vif_table[vr.vifi];
1634 if (VIF_EXISTS(mrt, vr.vifi)) {
1635 vr.icount = vif->pkt_in;
1636 vr.ocount = vif->pkt_out;
1637 vr.ibytes = vif->bytes_in;
1638 vr.obytes = vif->bytes_out;
1639 read_unlock(&mrt_lock);
1640
1641 if (copy_to_user(arg, &vr, sizeof(vr)))
1642 return -EFAULT;
1643 return 0;
1644 }
1645 read_unlock(&mrt_lock);
1646 return -EADDRNOTAVAIL;
1647 case SIOCGETSGCNT:
1648 if (copy_from_user(&sr, arg, sizeof(sr)))
1649 return -EFAULT;
1650
1651 rcu_read_lock();
1652 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1653 if (c) {
1654 sr.pktcnt = c->_c.mfc_un.res.pkt;
1655 sr.bytecnt = c->_c.mfc_un.res.bytes;
1656 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1657 rcu_read_unlock();
1658
1659 if (copy_to_user(arg, &sr, sizeof(sr)))
1660 return -EFAULT;
1661 return 0;
1662 }
1663 rcu_read_unlock();
1664 return -EADDRNOTAVAIL;
1665 default:
1666 return -ENOIOCTLCMD;
1667 }
1668 }
1669
1670 #ifdef CONFIG_COMPAT
1671 struct compat_sioc_sg_req {
1672 struct in_addr src;
1673 struct in_addr grp;
1674 compat_ulong_t pktcnt;
1675 compat_ulong_t bytecnt;
1676 compat_ulong_t wrong_if;
1677 };
1678
1679 struct compat_sioc_vif_req {
1680 vifi_t vifi; /* Which iface */
1681 compat_ulong_t icount;
1682 compat_ulong_t ocount;
1683 compat_ulong_t ibytes;
1684 compat_ulong_t obytes;
1685 };
1686
1687 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1688 {
1689 struct compat_sioc_sg_req sr;
1690 struct compat_sioc_vif_req vr;
1691 struct vif_device *vif;
1692 struct mfc_cache *c;
1693 struct net *net = sock_net(sk);
1694 struct mr_table *mrt;
1695
1696 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1697 if (!mrt)
1698 return -ENOENT;
1699
1700 switch (cmd) {
1701 case SIOCGETVIFCNT:
1702 if (copy_from_user(&vr, arg, sizeof(vr)))
1703 return -EFAULT;
1704 if (vr.vifi >= mrt->maxvif)
1705 return -EINVAL;
1706 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1707 read_lock(&mrt_lock);
1708 vif = &mrt->vif_table[vr.vifi];
1709 if (VIF_EXISTS(mrt, vr.vifi)) {
1710 vr.icount = vif->pkt_in;
1711 vr.ocount = vif->pkt_out;
1712 vr.ibytes = vif->bytes_in;
1713 vr.obytes = vif->bytes_out;
1714 read_unlock(&mrt_lock);
1715
1716 if (copy_to_user(arg, &vr, sizeof(vr)))
1717 return -EFAULT;
1718 return 0;
1719 }
1720 read_unlock(&mrt_lock);
1721 return -EADDRNOTAVAIL;
1722 case SIOCGETSGCNT:
1723 if (copy_from_user(&sr, arg, sizeof(sr)))
1724 return -EFAULT;
1725
1726 rcu_read_lock();
1727 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1728 if (c) {
1729 sr.pktcnt = c->_c.mfc_un.res.pkt;
1730 sr.bytecnt = c->_c.mfc_un.res.bytes;
1731 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1732 rcu_read_unlock();
1733
1734 if (copy_to_user(arg, &sr, sizeof(sr)))
1735 return -EFAULT;
1736 return 0;
1737 }
1738 rcu_read_unlock();
1739 return -EADDRNOTAVAIL;
1740 default:
1741 return -ENOIOCTLCMD;
1742 }
1743 }
1744 #endif
1745
1746 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1747 {
1748 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1749 struct net *net = dev_net(dev);
1750 struct mr_table *mrt;
1751 struct vif_device *v;
1752 int ct;
1753
1754 if (event != NETDEV_UNREGISTER)
1755 return NOTIFY_DONE;
1756
1757 ipmr_for_each_table(mrt, net) {
1758 v = &mrt->vif_table[0];
1759 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1760 if (v->dev == dev)
1761 vif_delete(mrt, ct, 1, NULL);
1762 }
1763 }
1764 return NOTIFY_DONE;
1765 }
1766
1767 static struct notifier_block ip_mr_notifier = {
1768 .notifier_call = ipmr_device_event,
1769 };
1770
1771 /* Encapsulate a packet by attaching a valid IPIP header to it.
1772 * This avoids tunnel drivers and other mess and gives us the speed so
1773 * important for multicast video.
1774 */
1775 static void ip_encap(struct net *net, struct sk_buff *skb,
1776 __be32 saddr, __be32 daddr)
1777 {
1778 struct iphdr *iph;
1779 const struct iphdr *old_iph = ip_hdr(skb);
1780
1781 skb_push(skb, sizeof(struct iphdr));
1782 skb->transport_header = skb->network_header;
1783 skb_reset_network_header(skb);
1784 iph = ip_hdr(skb);
1785
1786 iph->version = 4;
1787 iph->tos = old_iph->tos;
1788 iph->ttl = old_iph->ttl;
1789 iph->frag_off = 0;
1790 iph->daddr = daddr;
1791 iph->saddr = saddr;
1792 iph->protocol = IPPROTO_IPIP;
1793 iph->ihl = 5;
1794 iph->tot_len = htons(skb->len);
1795 ip_select_ident(net, skb, NULL);
1796 ip_send_check(iph);
1797
1798 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1799 nf_reset_ct(skb);
1800 }
1801
1802 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1803 struct sk_buff *skb)
1804 {
1805 struct ip_options *opt = &(IPCB(skb)->opt);
1806
1807 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1808 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1809
1810 if (unlikely(opt->optlen))
1811 ip_forward_options(skb);
1812
1813 return dst_output(net, sk, skb);
1814 }
1815
1816 #ifdef CONFIG_NET_SWITCHDEV
1817 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1818 int in_vifi, int out_vifi)
1819 {
1820 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1821 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1822
1823 if (!skb->offload_l3_fwd_mark)
1824 return false;
1825 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1826 return false;
1827 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1828 &in_vif->dev_parent_id);
1829 }
1830 #else
1831 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1832 int in_vifi, int out_vifi)
1833 {
1834 return false;
1835 }
1836 #endif
1837
1838 /* Processing handlers for ipmr_forward */
1839
1840 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1841 int in_vifi, struct sk_buff *skb, int vifi)
1842 {
1843 const struct iphdr *iph = ip_hdr(skb);
1844 struct vif_device *vif = &mrt->vif_table[vifi];
1845 struct net_device *dev;
1846 struct rtable *rt;
1847 struct flowi4 fl4;
1848 int encap = 0;
1849
1850 if (!vif->dev)
1851 goto out_free;
1852
1853 if (vif->flags & VIFF_REGISTER) {
1854 vif->pkt_out++;
1855 vif->bytes_out += skb->len;
1856 vif->dev->stats.tx_bytes += skb->len;
1857 vif->dev->stats.tx_packets++;
1858 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1859 goto out_free;
1860 }
1861
1862 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1863 goto out_free;
1864
1865 if (vif->flags & VIFF_TUNNEL) {
1866 rt = ip_route_output_ports(net, &fl4, NULL,
1867 vif->remote, vif->local,
1868 0, 0,
1869 IPPROTO_IPIP,
1870 RT_TOS(iph->tos), vif->link);
1871 if (IS_ERR(rt))
1872 goto out_free;
1873 encap = sizeof(struct iphdr);
1874 } else {
1875 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1876 0, 0,
1877 IPPROTO_IPIP,
1878 RT_TOS(iph->tos), vif->link);
1879 if (IS_ERR(rt))
1880 goto out_free;
1881 }
1882
1883 dev = rt->dst.dev;
1884
1885 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1886 /* Do not fragment multicasts. Alas, IPv4 does not
1887 * allow to send ICMP, so that packets will disappear
1888 * to blackhole.
1889 */
1890 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1891 ip_rt_put(rt);
1892 goto out_free;
1893 }
1894
1895 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1896
1897 if (skb_cow(skb, encap)) {
1898 ip_rt_put(rt);
1899 goto out_free;
1900 }
1901
1902 vif->pkt_out++;
1903 vif->bytes_out += skb->len;
1904
1905 skb_dst_drop(skb);
1906 skb_dst_set(skb, &rt->dst);
1907 ip_decrease_ttl(ip_hdr(skb));
1908
1909 /* FIXME: forward and output firewalls used to be called here.
1910 * What do we do with netfilter? -- RR
1911 */
1912 if (vif->flags & VIFF_TUNNEL) {
1913 ip_encap(net, skb, vif->local, vif->remote);
1914 /* FIXME: extra output firewall step used to be here. --RR */
1915 vif->dev->stats.tx_packets++;
1916 vif->dev->stats.tx_bytes += skb->len;
1917 }
1918
1919 IPCB(skb)->flags |= IPSKB_FORWARDED;
1920
1921 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1922 * not only before forwarding, but after forwarding on all output
1923 * interfaces. It is clear, if mrouter runs a multicasting
1924 * program, it should receive packets not depending to what interface
1925 * program is joined.
1926 * If we will not make it, the program will have to join on all
1927 * interfaces. On the other hand, multihoming host (or router, but
1928 * not mrouter) cannot join to more than one interface - it will
1929 * result in receiving multiple packets.
1930 */
1931 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1932 net, NULL, skb, skb->dev, dev,
1933 ipmr_forward_finish);
1934 return;
1935
1936 out_free:
1937 kfree_skb(skb);
1938 }
1939
1940 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1941 {
1942 int ct;
1943
1944 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1945 if (mrt->vif_table[ct].dev == dev)
1946 break;
1947 }
1948 return ct;
1949 }
1950
1951 /* "local" means that we should preserve one skb (for local delivery) */
1952 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1953 struct net_device *dev, struct sk_buff *skb,
1954 struct mfc_cache *c, int local)
1955 {
1956 int true_vifi = ipmr_find_vif(mrt, dev);
1957 int psend = -1;
1958 int vif, ct;
1959
1960 vif = c->_c.mfc_parent;
1961 c->_c.mfc_un.res.pkt++;
1962 c->_c.mfc_un.res.bytes += skb->len;
1963 c->_c.mfc_un.res.lastuse = jiffies;
1964
1965 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1966 struct mfc_cache *cache_proxy;
1967
1968 /* For an (*,G) entry, we only check that the incomming
1969 * interface is part of the static tree.
1970 */
1971 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1972 if (cache_proxy &&
1973 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1974 goto forward;
1975 }
1976
1977 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1978 if (mrt->vif_table[vif].dev != dev) {
1979 if (rt_is_output_route(skb_rtable(skb))) {
1980 /* It is our own packet, looped back.
1981 * Very complicated situation...
1982 *
1983 * The best workaround until routing daemons will be
1984 * fixed is not to redistribute packet, if it was
1985 * send through wrong interface. It means, that
1986 * multicast applications WILL NOT work for
1987 * (S,G), which have default multicast route pointing
1988 * to wrong oif. In any case, it is not a good
1989 * idea to use multicasting applications on router.
1990 */
1991 goto dont_forward;
1992 }
1993
1994 c->_c.mfc_un.res.wrong_if++;
1995
1996 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1997 /* pimsm uses asserts, when switching from RPT to SPT,
1998 * so that we cannot check that packet arrived on an oif.
1999 * It is bad, but otherwise we would need to move pretty
2000 * large chunk of pimd to kernel. Ough... --ANK
2001 */
2002 (mrt->mroute_do_pim ||
2003 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2004 time_after(jiffies,
2005 c->_c.mfc_un.res.last_assert +
2006 MFC_ASSERT_THRESH)) {
2007 c->_c.mfc_un.res.last_assert = jiffies;
2008 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2009 if (mrt->mroute_do_wrvifwhole)
2010 ipmr_cache_report(mrt, skb, true_vifi,
2011 IGMPMSG_WRVIFWHOLE);
2012 }
2013 goto dont_forward;
2014 }
2015
2016 forward:
2017 mrt->vif_table[vif].pkt_in++;
2018 mrt->vif_table[vif].bytes_in += skb->len;
2019
2020 /* Forward the frame */
2021 if (c->mfc_origin == htonl(INADDR_ANY) &&
2022 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2023 if (true_vifi >= 0 &&
2024 true_vifi != c->_c.mfc_parent &&
2025 ip_hdr(skb)->ttl >
2026 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2027 /* It's an (*,*) entry and the packet is not coming from
2028 * the upstream: forward the packet to the upstream
2029 * only.
2030 */
2031 psend = c->_c.mfc_parent;
2032 goto last_forward;
2033 }
2034 goto dont_forward;
2035 }
2036 for (ct = c->_c.mfc_un.res.maxvif - 1;
2037 ct >= c->_c.mfc_un.res.minvif; ct--) {
2038 /* For (*,G) entry, don't forward to the incoming interface */
2039 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2040 ct != true_vifi) &&
2041 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2042 if (psend != -1) {
2043 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2044
2045 if (skb2)
2046 ipmr_queue_xmit(net, mrt, true_vifi,
2047 skb2, psend);
2048 }
2049 psend = ct;
2050 }
2051 }
2052 last_forward:
2053 if (psend != -1) {
2054 if (local) {
2055 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2056
2057 if (skb2)
2058 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2059 psend);
2060 } else {
2061 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2062 return;
2063 }
2064 }
2065
2066 dont_forward:
2067 if (!local)
2068 kfree_skb(skb);
2069 }
2070
2071 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2072 {
2073 struct rtable *rt = skb_rtable(skb);
2074 struct iphdr *iph = ip_hdr(skb);
2075 struct flowi4 fl4 = {
2076 .daddr = iph->daddr,
2077 .saddr = iph->saddr,
2078 .flowi4_tos = RT_TOS(iph->tos),
2079 .flowi4_oif = (rt_is_output_route(rt) ?
2080 skb->dev->ifindex : 0),
2081 .flowi4_iif = (rt_is_output_route(rt) ?
2082 LOOPBACK_IFINDEX :
2083 skb->dev->ifindex),
2084 .flowi4_mark = skb->mark,
2085 };
2086 struct mr_table *mrt;
2087 int err;
2088
2089 err = ipmr_fib_lookup(net, &fl4, &mrt);
2090 if (err)
2091 return ERR_PTR(err);
2092 return mrt;
2093 }
2094
2095 /* Multicast packets for forwarding arrive here
2096 * Called with rcu_read_lock();
2097 */
2098 int ip_mr_input(struct sk_buff *skb)
2099 {
2100 struct mfc_cache *cache;
2101 struct net *net = dev_net(skb->dev);
2102 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2103 struct mr_table *mrt;
2104 struct net_device *dev;
2105
2106 /* skb->dev passed in is the loX master dev for vrfs.
2107 * As there are no vifs associated with loopback devices,
2108 * get the proper interface that does have a vif associated with it.
2109 */
2110 dev = skb->dev;
2111 if (netif_is_l3_master(skb->dev)) {
2112 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2113 if (!dev) {
2114 kfree_skb(skb);
2115 return -ENODEV;
2116 }
2117 }
2118
2119 /* Packet is looped back after forward, it should not be
2120 * forwarded second time, but still can be delivered locally.
2121 */
2122 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2123 goto dont_forward;
2124
2125 mrt = ipmr_rt_fib_lookup(net, skb);
2126 if (IS_ERR(mrt)) {
2127 kfree_skb(skb);
2128 return PTR_ERR(mrt);
2129 }
2130 if (!local) {
2131 if (IPCB(skb)->opt.router_alert) {
2132 if (ip_call_ra_chain(skb))
2133 return 0;
2134 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2135 /* IGMPv1 (and broken IGMPv2 implementations sort of
2136 * Cisco IOS <= 11.2(8)) do not put router alert
2137 * option to IGMP packets destined to routable
2138 * groups. It is very bad, because it means
2139 * that we can forward NO IGMP messages.
2140 */
2141 struct sock *mroute_sk;
2142
2143 mroute_sk = rcu_dereference(mrt->mroute_sk);
2144 if (mroute_sk) {
2145 nf_reset_ct(skb);
2146 raw_rcv(mroute_sk, skb);
2147 return 0;
2148 }
2149 }
2150 }
2151
2152 /* already under rcu_read_lock() */
2153 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2154 if (!cache) {
2155 int vif = ipmr_find_vif(mrt, dev);
2156
2157 if (vif >= 0)
2158 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2159 vif);
2160 }
2161
2162 /* No usable cache entry */
2163 if (!cache) {
2164 int vif;
2165
2166 if (local) {
2167 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2168 ip_local_deliver(skb);
2169 if (!skb2)
2170 return -ENOBUFS;
2171 skb = skb2;
2172 }
2173
2174 read_lock(&mrt_lock);
2175 vif = ipmr_find_vif(mrt, dev);
2176 if (vif >= 0) {
2177 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2178 read_unlock(&mrt_lock);
2179
2180 return err2;
2181 }
2182 read_unlock(&mrt_lock);
2183 kfree_skb(skb);
2184 return -ENODEV;
2185 }
2186
2187 read_lock(&mrt_lock);
2188 ip_mr_forward(net, mrt, dev, skb, cache, local);
2189 read_unlock(&mrt_lock);
2190
2191 if (local)
2192 return ip_local_deliver(skb);
2193
2194 return 0;
2195
2196 dont_forward:
2197 if (local)
2198 return ip_local_deliver(skb);
2199 kfree_skb(skb);
2200 return 0;
2201 }
2202
2203 #ifdef CONFIG_IP_PIMSM_V1
2204 /* Handle IGMP messages of PIMv1 */
2205 int pim_rcv_v1(struct sk_buff *skb)
2206 {
2207 struct igmphdr *pim;
2208 struct net *net = dev_net(skb->dev);
2209 struct mr_table *mrt;
2210
2211 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2212 goto drop;
2213
2214 pim = igmp_hdr(skb);
2215
2216 mrt = ipmr_rt_fib_lookup(net, skb);
2217 if (IS_ERR(mrt))
2218 goto drop;
2219 if (!mrt->mroute_do_pim ||
2220 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2221 goto drop;
2222
2223 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2224 drop:
2225 kfree_skb(skb);
2226 }
2227 return 0;
2228 }
2229 #endif
2230
2231 #ifdef CONFIG_IP_PIMSM_V2
2232 static int pim_rcv(struct sk_buff *skb)
2233 {
2234 struct pimreghdr *pim;
2235 struct net *net = dev_net(skb->dev);
2236 struct mr_table *mrt;
2237
2238 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2239 goto drop;
2240
2241 pim = (struct pimreghdr *)skb_transport_header(skb);
2242 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2243 (pim->flags & PIM_NULL_REGISTER) ||
2244 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2245 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2246 goto drop;
2247
2248 mrt = ipmr_rt_fib_lookup(net, skb);
2249 if (IS_ERR(mrt))
2250 goto drop;
2251 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2252 drop:
2253 kfree_skb(skb);
2254 }
2255 return 0;
2256 }
2257 #endif
2258
2259 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2260 __be32 saddr, __be32 daddr,
2261 struct rtmsg *rtm, u32 portid)
2262 {
2263 struct mfc_cache *cache;
2264 struct mr_table *mrt;
2265 int err;
2266
2267 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2268 if (!mrt)
2269 return -ENOENT;
2270
2271 rcu_read_lock();
2272 cache = ipmr_cache_find(mrt, saddr, daddr);
2273 if (!cache && skb->dev) {
2274 int vif = ipmr_find_vif(mrt, skb->dev);
2275
2276 if (vif >= 0)
2277 cache = ipmr_cache_find_any(mrt, daddr, vif);
2278 }
2279 if (!cache) {
2280 struct sk_buff *skb2;
2281 struct iphdr *iph;
2282 struct net_device *dev;
2283 int vif = -1;
2284
2285 dev = skb->dev;
2286 read_lock(&mrt_lock);
2287 if (dev)
2288 vif = ipmr_find_vif(mrt, dev);
2289 if (vif < 0) {
2290 read_unlock(&mrt_lock);
2291 rcu_read_unlock();
2292 return -ENODEV;
2293 }
2294
2295 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2296 if (!skb2) {
2297 read_unlock(&mrt_lock);
2298 rcu_read_unlock();
2299 return -ENOMEM;
2300 }
2301
2302 NETLINK_CB(skb2).portid = portid;
2303 skb_push(skb2, sizeof(struct iphdr));
2304 skb_reset_network_header(skb2);
2305 iph = ip_hdr(skb2);
2306 iph->ihl = sizeof(struct iphdr) >> 2;
2307 iph->saddr = saddr;
2308 iph->daddr = daddr;
2309 iph->version = 0;
2310 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2311 read_unlock(&mrt_lock);
2312 rcu_read_unlock();
2313 return err;
2314 }
2315
2316 read_lock(&mrt_lock);
2317 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2318 read_unlock(&mrt_lock);
2319 rcu_read_unlock();
2320 return err;
2321 }
2322
2323 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2324 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2325 int flags)
2326 {
2327 struct nlmsghdr *nlh;
2328 struct rtmsg *rtm;
2329 int err;
2330
2331 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2332 if (!nlh)
2333 return -EMSGSIZE;
2334
2335 rtm = nlmsg_data(nlh);
2336 rtm->rtm_family = RTNL_FAMILY_IPMR;
2337 rtm->rtm_dst_len = 32;
2338 rtm->rtm_src_len = 32;
2339 rtm->rtm_tos = 0;
2340 rtm->rtm_table = mrt->id;
2341 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2342 goto nla_put_failure;
2343 rtm->rtm_type = RTN_MULTICAST;
2344 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2345 if (c->_c.mfc_flags & MFC_STATIC)
2346 rtm->rtm_protocol = RTPROT_STATIC;
2347 else
2348 rtm->rtm_protocol = RTPROT_MROUTED;
2349 rtm->rtm_flags = 0;
2350
2351 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2352 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2353 goto nla_put_failure;
2354 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2355 /* do not break the dump if cache is unresolved */
2356 if (err < 0 && err != -ENOENT)
2357 goto nla_put_failure;
2358
2359 nlmsg_end(skb, nlh);
2360 return 0;
2361
2362 nla_put_failure:
2363 nlmsg_cancel(skb, nlh);
2364 return -EMSGSIZE;
2365 }
2366
2367 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2368 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2369 int flags)
2370 {
2371 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2372 cmd, flags);
2373 }
2374
2375 static size_t mroute_msgsize(bool unresolved, int maxvif)
2376 {
2377 size_t len =
2378 NLMSG_ALIGN(sizeof(struct rtmsg))
2379 + nla_total_size(4) /* RTA_TABLE */
2380 + nla_total_size(4) /* RTA_SRC */
2381 + nla_total_size(4) /* RTA_DST */
2382 ;
2383
2384 if (!unresolved)
2385 len = len
2386 + nla_total_size(4) /* RTA_IIF */
2387 + nla_total_size(0) /* RTA_MULTIPATH */
2388 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2389 /* RTA_MFC_STATS */
2390 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2391 ;
2392
2393 return len;
2394 }
2395
2396 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2397 int cmd)
2398 {
2399 struct net *net = read_pnet(&mrt->net);
2400 struct sk_buff *skb;
2401 int err = -ENOBUFS;
2402
2403 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2404 mrt->maxvif),
2405 GFP_ATOMIC);
2406 if (!skb)
2407 goto errout;
2408
2409 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2410 if (err < 0)
2411 goto errout;
2412
2413 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2414 return;
2415
2416 errout:
2417 kfree_skb(skb);
2418 if (err < 0)
2419 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2420 }
2421
2422 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2423 {
2424 size_t len =
2425 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2426 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2427 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2428 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2429 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2430 /* IPMRA_CREPORT_PKT */
2431 + nla_total_size(payloadlen)
2432 ;
2433
2434 return len;
2435 }
2436
2437 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2438 {
2439 struct net *net = read_pnet(&mrt->net);
2440 struct nlmsghdr *nlh;
2441 struct rtgenmsg *rtgenm;
2442 struct igmpmsg *msg;
2443 struct sk_buff *skb;
2444 struct nlattr *nla;
2445 int payloadlen;
2446
2447 payloadlen = pkt->len - sizeof(struct igmpmsg);
2448 msg = (struct igmpmsg *)skb_network_header(pkt);
2449
2450 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2451 if (!skb)
2452 goto errout;
2453
2454 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2455 sizeof(struct rtgenmsg), 0);
2456 if (!nlh)
2457 goto errout;
2458 rtgenm = nlmsg_data(nlh);
2459 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2460 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2461 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2462 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2463 msg->im_src.s_addr) ||
2464 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2465 msg->im_dst.s_addr))
2466 goto nla_put_failure;
2467
2468 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2469 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2470 nla_data(nla), payloadlen))
2471 goto nla_put_failure;
2472
2473 nlmsg_end(skb, nlh);
2474
2475 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2476 return;
2477
2478 nla_put_failure:
2479 nlmsg_cancel(skb, nlh);
2480 errout:
2481 kfree_skb(skb);
2482 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2483 }
2484
2485 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2486 const struct nlmsghdr *nlh,
2487 struct nlattr **tb,
2488 struct netlink_ext_ack *extack)
2489 {
2490 struct rtmsg *rtm;
2491 int i, err;
2492
2493 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2494 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2495 return -EINVAL;
2496 }
2497
2498 if (!netlink_strict_get_check(skb))
2499 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2500 rtm_ipv4_policy, extack);
2501
2502 rtm = nlmsg_data(nlh);
2503 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2504 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2505 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2506 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2507 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2508 return -EINVAL;
2509 }
2510
2511 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2512 rtm_ipv4_policy, extack);
2513 if (err)
2514 return err;
2515
2516 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2517 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2518 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2519 return -EINVAL;
2520 }
2521
2522 for (i = 0; i <= RTA_MAX; i++) {
2523 if (!tb[i])
2524 continue;
2525
2526 switch (i) {
2527 case RTA_SRC:
2528 case RTA_DST:
2529 case RTA_TABLE:
2530 break;
2531 default:
2532 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2533 return -EINVAL;
2534 }
2535 }
2536
2537 return 0;
2538 }
2539
2540 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2541 struct netlink_ext_ack *extack)
2542 {
2543 struct net *net = sock_net(in_skb->sk);
2544 struct nlattr *tb[RTA_MAX + 1];
2545 struct sk_buff *skb = NULL;
2546 struct mfc_cache *cache;
2547 struct mr_table *mrt;
2548 __be32 src, grp;
2549 u32 tableid;
2550 int err;
2551
2552 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2553 if (err < 0)
2554 goto errout;
2555
2556 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2557 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2558 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2559
2560 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2561 if (!mrt) {
2562 err = -ENOENT;
2563 goto errout_free;
2564 }
2565
2566 /* entries are added/deleted only under RTNL */
2567 rcu_read_lock();
2568 cache = ipmr_cache_find(mrt, src, grp);
2569 rcu_read_unlock();
2570 if (!cache) {
2571 err = -ENOENT;
2572 goto errout_free;
2573 }
2574
2575 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2576 if (!skb) {
2577 err = -ENOBUFS;
2578 goto errout_free;
2579 }
2580
2581 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2582 nlh->nlmsg_seq, cache,
2583 RTM_NEWROUTE, 0);
2584 if (err < 0)
2585 goto errout_free;
2586
2587 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2588
2589 errout:
2590 return err;
2591
2592 errout_free:
2593 kfree_skb(skb);
2594 goto errout;
2595 }
2596
2597 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2598 {
2599 struct fib_dump_filter filter = {};
2600 int err;
2601
2602 if (cb->strict_check) {
2603 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2604 &filter, cb);
2605 if (err < 0)
2606 return err;
2607 }
2608
2609 if (filter.table_id) {
2610 struct mr_table *mrt;
2611
2612 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2613 if (!mrt) {
2614 if (filter.dump_all_families)
2615 return skb->len;
2616
2617 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2618 return -ENOENT;
2619 }
2620 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2621 &mfc_unres_lock, &filter);
2622 return skb->len ? : err;
2623 }
2624
2625 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2626 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2627 }
2628
2629 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2630 [RTA_SRC] = { .type = NLA_U32 },
2631 [RTA_DST] = { .type = NLA_U32 },
2632 [RTA_IIF] = { .type = NLA_U32 },
2633 [RTA_TABLE] = { .type = NLA_U32 },
2634 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2635 };
2636
2637 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2638 {
2639 switch (rtm_protocol) {
2640 case RTPROT_STATIC:
2641 case RTPROT_MROUTED:
2642 return true;
2643 }
2644 return false;
2645 }
2646
2647 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2648 {
2649 struct rtnexthop *rtnh = nla_data(nla);
2650 int remaining = nla_len(nla), vifi = 0;
2651
2652 while (rtnh_ok(rtnh, remaining)) {
2653 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2654 if (++vifi == MAXVIFS)
2655 break;
2656 rtnh = rtnh_next(rtnh, &remaining);
2657 }
2658
2659 return remaining > 0 ? -EINVAL : vifi;
2660 }
2661
2662 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2663 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2664 struct mfcctl *mfcc, int *mrtsock,
2665 struct mr_table **mrtret,
2666 struct netlink_ext_ack *extack)
2667 {
2668 struct net_device *dev = NULL;
2669 u32 tblid = RT_TABLE_DEFAULT;
2670 struct mr_table *mrt;
2671 struct nlattr *attr;
2672 struct rtmsg *rtm;
2673 int ret, rem;
2674
2675 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2676 rtm_ipmr_policy, extack);
2677 if (ret < 0)
2678 goto out;
2679 rtm = nlmsg_data(nlh);
2680
2681 ret = -EINVAL;
2682 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2683 rtm->rtm_type != RTN_MULTICAST ||
2684 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2685 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2686 goto out;
2687
2688 memset(mfcc, 0, sizeof(*mfcc));
2689 mfcc->mfcc_parent = -1;
2690 ret = 0;
2691 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2692 switch (nla_type(attr)) {
2693 case RTA_SRC:
2694 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2695 break;
2696 case RTA_DST:
2697 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2698 break;
2699 case RTA_IIF:
2700 dev = __dev_get_by_index(net, nla_get_u32(attr));
2701 if (!dev) {
2702 ret = -ENODEV;
2703 goto out;
2704 }
2705 break;
2706 case RTA_MULTIPATH:
2707 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2708 ret = -EINVAL;
2709 goto out;
2710 }
2711 break;
2712 case RTA_PREFSRC:
2713 ret = 1;
2714 break;
2715 case RTA_TABLE:
2716 tblid = nla_get_u32(attr);
2717 break;
2718 }
2719 }
2720 mrt = ipmr_get_table(net, tblid);
2721 if (!mrt) {
2722 ret = -ENOENT;
2723 goto out;
2724 }
2725 *mrtret = mrt;
2726 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2727 if (dev)
2728 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2729
2730 out:
2731 return ret;
2732 }
2733
2734 /* takes care of both newroute and delroute */
2735 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2736 struct netlink_ext_ack *extack)
2737 {
2738 struct net *net = sock_net(skb->sk);
2739 int ret, mrtsock, parent;
2740 struct mr_table *tbl;
2741 struct mfcctl mfcc;
2742
2743 mrtsock = 0;
2744 tbl = NULL;
2745 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2746 if (ret < 0)
2747 return ret;
2748
2749 parent = ret ? mfcc.mfcc_parent : -1;
2750 if (nlh->nlmsg_type == RTM_NEWROUTE)
2751 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2752 else
2753 return ipmr_mfc_delete(tbl, &mfcc, parent);
2754 }
2755
2756 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2757 {
2758 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2759
2760 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2761 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2762 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2763 mrt->mroute_reg_vif_num) ||
2764 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2765 mrt->mroute_do_assert) ||
2766 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2767 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2768 mrt->mroute_do_wrvifwhole))
2769 return false;
2770
2771 return true;
2772 }
2773
2774 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2775 {
2776 struct nlattr *vif_nest;
2777 struct vif_device *vif;
2778
2779 /* if the VIF doesn't exist just continue */
2780 if (!VIF_EXISTS(mrt, vifid))
2781 return true;
2782
2783 vif = &mrt->vif_table[vifid];
2784 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2785 if (!vif_nest)
2786 return false;
2787 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2788 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2789 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2790 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2791 IPMRA_VIFA_PAD) ||
2792 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2793 IPMRA_VIFA_PAD) ||
2794 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2795 IPMRA_VIFA_PAD) ||
2796 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2797 IPMRA_VIFA_PAD) ||
2798 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2799 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2800 nla_nest_cancel(skb, vif_nest);
2801 return false;
2802 }
2803 nla_nest_end(skb, vif_nest);
2804
2805 return true;
2806 }
2807
2808 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2809 struct netlink_ext_ack *extack)
2810 {
2811 struct ifinfomsg *ifm;
2812
2813 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2814 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2815 return -EINVAL;
2816 }
2817
2818 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2819 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2820 return -EINVAL;
2821 }
2822
2823 ifm = nlmsg_data(nlh);
2824 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2825 ifm->ifi_change || ifm->ifi_index) {
2826 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2827 return -EINVAL;
2828 }
2829
2830 return 0;
2831 }
2832
2833 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2834 {
2835 struct net *net = sock_net(skb->sk);
2836 struct nlmsghdr *nlh = NULL;
2837 unsigned int t = 0, s_t;
2838 unsigned int e = 0, s_e;
2839 struct mr_table *mrt;
2840
2841 if (cb->strict_check) {
2842 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2843
2844 if (err < 0)
2845 return err;
2846 }
2847
2848 s_t = cb->args[0];
2849 s_e = cb->args[1];
2850
2851 ipmr_for_each_table(mrt, net) {
2852 struct nlattr *vifs, *af;
2853 struct ifinfomsg *hdr;
2854 u32 i;
2855
2856 if (t < s_t)
2857 goto skip_table;
2858 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2859 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2860 sizeof(*hdr), NLM_F_MULTI);
2861 if (!nlh)
2862 break;
2863
2864 hdr = nlmsg_data(nlh);
2865 memset(hdr, 0, sizeof(*hdr));
2866 hdr->ifi_family = RTNL_FAMILY_IPMR;
2867
2868 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2869 if (!af) {
2870 nlmsg_cancel(skb, nlh);
2871 goto out;
2872 }
2873
2874 if (!ipmr_fill_table(mrt, skb)) {
2875 nlmsg_cancel(skb, nlh);
2876 goto out;
2877 }
2878
2879 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2880 if (!vifs) {
2881 nla_nest_end(skb, af);
2882 nlmsg_end(skb, nlh);
2883 goto out;
2884 }
2885 for (i = 0; i < mrt->maxvif; i++) {
2886 if (e < s_e)
2887 goto skip_entry;
2888 if (!ipmr_fill_vif(mrt, i, skb)) {
2889 nla_nest_end(skb, vifs);
2890 nla_nest_end(skb, af);
2891 nlmsg_end(skb, nlh);
2892 goto out;
2893 }
2894 skip_entry:
2895 e++;
2896 }
2897 s_e = 0;
2898 e = 0;
2899 nla_nest_end(skb, vifs);
2900 nla_nest_end(skb, af);
2901 nlmsg_end(skb, nlh);
2902 skip_table:
2903 t++;
2904 }
2905
2906 out:
2907 cb->args[1] = e;
2908 cb->args[0] = t;
2909
2910 return skb->len;
2911 }
2912
2913 #ifdef CONFIG_PROC_FS
2914 /* The /proc interfaces to multicast routing :
2915 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2916 */
2917
2918 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2919 __acquires(mrt_lock)
2920 {
2921 struct mr_vif_iter *iter = seq->private;
2922 struct net *net = seq_file_net(seq);
2923 struct mr_table *mrt;
2924
2925 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2926 if (!mrt)
2927 return ERR_PTR(-ENOENT);
2928
2929 iter->mrt = mrt;
2930
2931 read_lock(&mrt_lock);
2932 return mr_vif_seq_start(seq, pos);
2933 }
2934
2935 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2936 __releases(mrt_lock)
2937 {
2938 read_unlock(&mrt_lock);
2939 }
2940
2941 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2942 {
2943 struct mr_vif_iter *iter = seq->private;
2944 struct mr_table *mrt = iter->mrt;
2945
2946 if (v == SEQ_START_TOKEN) {
2947 seq_puts(seq,
2948 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2949 } else {
2950 const struct vif_device *vif = v;
2951 const char *name = vif->dev ?
2952 vif->dev->name : "none";
2953
2954 seq_printf(seq,
2955 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2956 vif - mrt->vif_table,
2957 name, vif->bytes_in, vif->pkt_in,
2958 vif->bytes_out, vif->pkt_out,
2959 vif->flags, vif->local, vif->remote);
2960 }
2961 return 0;
2962 }
2963
2964 static const struct seq_operations ipmr_vif_seq_ops = {
2965 .start = ipmr_vif_seq_start,
2966 .next = mr_vif_seq_next,
2967 .stop = ipmr_vif_seq_stop,
2968 .show = ipmr_vif_seq_show,
2969 };
2970
2971 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2972 {
2973 struct net *net = seq_file_net(seq);
2974 struct mr_table *mrt;
2975
2976 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2977 if (!mrt)
2978 return ERR_PTR(-ENOENT);
2979
2980 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2981 }
2982
2983 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2984 {
2985 int n;
2986
2987 if (v == SEQ_START_TOKEN) {
2988 seq_puts(seq,
2989 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2990 } else {
2991 const struct mfc_cache *mfc = v;
2992 const struct mr_mfc_iter *it = seq->private;
2993 const struct mr_table *mrt = it->mrt;
2994
2995 seq_printf(seq, "%08X %08X %-3hd",
2996 (__force u32) mfc->mfc_mcastgrp,
2997 (__force u32) mfc->mfc_origin,
2998 mfc->_c.mfc_parent);
2999
3000 if (it->cache != &mrt->mfc_unres_queue) {
3001 seq_printf(seq, " %8lu %8lu %8lu",
3002 mfc->_c.mfc_un.res.pkt,
3003 mfc->_c.mfc_un.res.bytes,
3004 mfc->_c.mfc_un.res.wrong_if);
3005 for (n = mfc->_c.mfc_un.res.minvif;
3006 n < mfc->_c.mfc_un.res.maxvif; n++) {
3007 if (VIF_EXISTS(mrt, n) &&
3008 mfc->_c.mfc_un.res.ttls[n] < 255)
3009 seq_printf(seq,
3010 " %2d:%-3d",
3011 n, mfc->_c.mfc_un.res.ttls[n]);
3012 }
3013 } else {
3014 /* unresolved mfc_caches don't contain
3015 * pkt, bytes and wrong_if values
3016 */
3017 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3018 }
3019 seq_putc(seq, '\n');
3020 }
3021 return 0;
3022 }
3023
3024 static const struct seq_operations ipmr_mfc_seq_ops = {
3025 .start = ipmr_mfc_seq_start,
3026 .next = mr_mfc_seq_next,
3027 .stop = mr_mfc_seq_stop,
3028 .show = ipmr_mfc_seq_show,
3029 };
3030 #endif
3031
3032 #ifdef CONFIG_IP_PIMSM_V2
3033 static const struct net_protocol pim_protocol = {
3034 .handler = pim_rcv,
3035 .netns_ok = 1,
3036 };
3037 #endif
3038
3039 static unsigned int ipmr_seq_read(struct net *net)
3040 {
3041 ASSERT_RTNL();
3042
3043 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3044 }
3045
3046 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3047 struct netlink_ext_ack *extack)
3048 {
3049 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3050 ipmr_mr_table_iter, &mrt_lock, extack);
3051 }
3052
3053 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3054 .family = RTNL_FAMILY_IPMR,
3055 .fib_seq_read = ipmr_seq_read,
3056 .fib_dump = ipmr_dump,
3057 .owner = THIS_MODULE,
3058 };
3059
3060 static int __net_init ipmr_notifier_init(struct net *net)
3061 {
3062 struct fib_notifier_ops *ops;
3063
3064 net->ipv4.ipmr_seq = 0;
3065
3066 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3067 if (IS_ERR(ops))
3068 return PTR_ERR(ops);
3069 net->ipv4.ipmr_notifier_ops = ops;
3070
3071 return 0;
3072 }
3073
3074 static void __net_exit ipmr_notifier_exit(struct net *net)
3075 {
3076 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3077 net->ipv4.ipmr_notifier_ops = NULL;
3078 }
3079
3080 /* Setup for IP multicast routing */
3081 static int __net_init ipmr_net_init(struct net *net)
3082 {
3083 int err;
3084
3085 err = ipmr_notifier_init(net);
3086 if (err)
3087 goto ipmr_notifier_fail;
3088
3089 err = ipmr_rules_init(net);
3090 if (err < 0)
3091 goto ipmr_rules_fail;
3092
3093 #ifdef CONFIG_PROC_FS
3094 err = -ENOMEM;
3095 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3096 sizeof(struct mr_vif_iter)))
3097 goto proc_vif_fail;
3098 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3099 sizeof(struct mr_mfc_iter)))
3100 goto proc_cache_fail;
3101 #endif
3102 return 0;
3103
3104 #ifdef CONFIG_PROC_FS
3105 proc_cache_fail:
3106 remove_proc_entry("ip_mr_vif", net->proc_net);
3107 proc_vif_fail:
3108 ipmr_rules_exit(net);
3109 #endif
3110 ipmr_rules_fail:
3111 ipmr_notifier_exit(net);
3112 ipmr_notifier_fail:
3113 return err;
3114 }
3115
3116 static void __net_exit ipmr_net_exit(struct net *net)
3117 {
3118 #ifdef CONFIG_PROC_FS
3119 remove_proc_entry("ip_mr_cache", net->proc_net);
3120 remove_proc_entry("ip_mr_vif", net->proc_net);
3121 #endif
3122 ipmr_notifier_exit(net);
3123 ipmr_rules_exit(net);
3124 }
3125
3126 static struct pernet_operations ipmr_net_ops = {
3127 .init = ipmr_net_init,
3128 .exit = ipmr_net_exit,
3129 };
3130
3131 int __init ip_mr_init(void)
3132 {
3133 int err;
3134
3135 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3136 sizeof(struct mfc_cache),
3137 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3138 NULL);
3139
3140 err = register_pernet_subsys(&ipmr_net_ops);
3141 if (err)
3142 goto reg_pernet_fail;
3143
3144 err = register_netdevice_notifier(&ip_mr_notifier);
3145 if (err)
3146 goto reg_notif_fail;
3147 #ifdef CONFIG_IP_PIMSM_V2
3148 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3149 pr_err("%s: can't add PIM protocol\n", __func__);
3150 err = -EAGAIN;
3151 goto add_proto_fail;
3152 }
3153 #endif
3154 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3155 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3156 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3157 ipmr_rtm_route, NULL, 0);
3158 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3159 ipmr_rtm_route, NULL, 0);
3160
3161 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3162 NULL, ipmr_rtm_dumplink, 0);
3163 return 0;
3164
3165 #ifdef CONFIG_IP_PIMSM_V2
3166 add_proto_fail:
3167 unregister_netdevice_notifier(&ip_mr_notifier);
3168 #endif
3169 reg_notif_fail:
3170 unregister_pernet_subsys(&ipmr_net_ops);
3171 reg_pernet_fail:
3172 kmem_cache_destroy(mrt_cachep);
3173 return err;
3174 }
Linux with added FPC-III support