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