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