1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IP multicast routing support for mrouted 3.6/3.8
5 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6 * Linux Consultancy and Custom Driver Development
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
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.
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>
30 #include <linux/kernel.h>
31 #include <linux/fcntl.h>
32 #include <linux/stat.h>
33 #include <linux/socket.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>
47 #include <net/protocol.h>
48 #include <linux/skbuff.h>
49 #include <net/route.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>
66 #include <linux/nospec.h>
69 struct fib_rule common;
76 /* Big lock, protecting vif table, mrt cache and mroute socket state.
77 * Note that the changes are semaphored via rtnl_lock.
80 static DEFINE_RWLOCK(mrt_lock);
82 /* Multicast router control variables */
84 /* Special spinlock for queue of unresolved entries */
85 static DEFINE_SPINLOCK(mfc_unres_lock);
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.
92 * In this case data path is free of exclusive locks at all.
95 static struct kmem_cache *mrt_cachep __ro_after_init;
97 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
98 static void ipmr_free_table(struct mr_table *mrt);
100 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
101 struct net_device *dev, struct sk_buff *skb,
102 struct mfc_cache *cache, int local);
103 static int ipmr_cache_report(struct mr_table *mrt,
104 struct sk_buff *pkt, vifi_t vifi, int assert);
105 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
107 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
108 static void mroute_clean_tables(struct mr_table *mrt, int flags);
109 static void ipmr_expire_process(struct timer_list *t);
111 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
112 #define ipmr_for_each_table(mrt, net) \
113 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
114 lockdep_rtnl_is_held() || \
115 list_empty(&net->ipv4.mr_tables))
117 static struct mr_table *ipmr_mr_table_iter(struct net *net,
118 struct mr_table *mrt)
120 struct mr_table *ret;
123 ret = list_entry_rcu(net->ipv4.mr_tables.next,
124 struct mr_table, list);
126 ret = list_entry_rcu(mrt->list.next,
127 struct mr_table, list);
129 if (&ret->list == &net->ipv4.mr_tables)
134 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
136 struct mr_table *mrt;
138 ipmr_for_each_table(mrt, net) {
145 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
146 struct mr_table **mrt)
149 struct ipmr_result res;
150 struct fib_lookup_arg arg = {
152 .flags = FIB_LOOKUP_NOREF,
155 /* update flow if oif or iif point to device enslaved to l3mdev */
156 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
158 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
159 flowi4_to_flowi(flp4), 0, &arg);
166 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
167 int flags, struct fib_lookup_arg *arg)
169 struct ipmr_result *res = arg->result;
170 struct mr_table *mrt;
172 switch (rule->action) {
175 case FR_ACT_UNREACHABLE:
177 case FR_ACT_PROHIBIT:
179 case FR_ACT_BLACKHOLE:
184 arg->table = fib_rule_get_table(rule, arg);
186 mrt = ipmr_get_table(rule->fr_net, arg->table);
193 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
198 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
202 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
203 struct fib_rule_hdr *frh, struct nlattr **tb,
204 struct netlink_ext_ack *extack)
209 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
215 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
216 struct fib_rule_hdr *frh)
224 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
225 .family = RTNL_FAMILY_IPMR,
226 .rule_size = sizeof(struct ipmr_rule),
227 .addr_size = sizeof(u32),
228 .action = ipmr_rule_action,
229 .match = ipmr_rule_match,
230 .configure = ipmr_rule_configure,
231 .compare = ipmr_rule_compare,
232 .fill = ipmr_rule_fill,
233 .nlgroup = RTNLGRP_IPV4_RULE,
234 .policy = ipmr_rule_policy,
235 .owner = THIS_MODULE,
238 static int __net_init ipmr_rules_init(struct net *net)
240 struct fib_rules_ops *ops;
241 struct mr_table *mrt;
244 ops = fib_rules_register(&ipmr_rules_ops_template, net);
248 INIT_LIST_HEAD(&net->ipv4.mr_tables);
250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
260 net->ipv4.mr_rules_ops = ops;
265 ipmr_free_table(mrt);
268 fib_rules_unregister(ops);
272 static void __net_exit ipmr_rules_exit(struct net *net)
274 struct mr_table *mrt, *next;
277 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
278 list_del(&mrt->list);
279 ipmr_free_table(mrt);
281 fib_rules_unregister(net->ipv4.mr_rules_ops);
285 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
286 struct netlink_ext_ack *extack)
288 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
291 static unsigned int ipmr_rules_seq_read(struct net *net)
293 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
296 bool ipmr_rule_default(const struct fib_rule *rule)
298 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
300 EXPORT_SYMBOL(ipmr_rule_default);
302 #define ipmr_for_each_table(mrt, net) \
303 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
305 static struct mr_table *ipmr_mr_table_iter(struct net *net,
306 struct mr_table *mrt)
309 return net->ipv4.mrt;
313 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
315 return net->ipv4.mrt;
318 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
319 struct mr_table **mrt)
321 *mrt = net->ipv4.mrt;
325 static int __net_init ipmr_rules_init(struct net *net)
327 struct mr_table *mrt;
329 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
336 static void __net_exit ipmr_rules_exit(struct net *net)
339 ipmr_free_table(net->ipv4.mrt);
340 net->ipv4.mrt = NULL;
344 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
345 struct netlink_ext_ack *extack)
350 static unsigned int ipmr_rules_seq_read(struct net *net)
355 bool ipmr_rule_default(const struct fib_rule *rule)
359 EXPORT_SYMBOL(ipmr_rule_default);
362 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
365 const struct mfc_cache_cmp_arg *cmparg = arg->key;
366 struct mfc_cache *c = (struct mfc_cache *)ptr;
368 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
369 cmparg->mfc_origin != c->mfc_origin;
372 static const struct rhashtable_params ipmr_rht_params = {
373 .head_offset = offsetof(struct mr_mfc, mnode),
374 .key_offset = offsetof(struct mfc_cache, cmparg),
375 .key_len = sizeof(struct mfc_cache_cmp_arg),
377 .obj_cmpfn = ipmr_hash_cmp,
378 .automatic_shrinking = true,
381 static void ipmr_new_table_set(struct mr_table *mrt,
384 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
385 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
389 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
390 .mfc_mcastgrp = htonl(INADDR_ANY),
391 .mfc_origin = htonl(INADDR_ANY),
394 static struct mr_table_ops ipmr_mr_table_ops = {
395 .rht_params = &ipmr_rht_params,
396 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
399 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
401 struct mr_table *mrt;
403 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
404 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
405 return ERR_PTR(-EINVAL);
407 mrt = ipmr_get_table(net, id);
411 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
412 ipmr_expire_process, ipmr_new_table_set);
415 static void ipmr_free_table(struct mr_table *mrt)
417 del_timer_sync(&mrt->ipmr_expire_timer);
418 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
419 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
420 rhltable_destroy(&mrt->mfc_hash);
424 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
426 /* Initialize ipmr pimreg/tunnel in_device */
427 static bool ipmr_init_vif_indev(const struct net_device *dev)
429 struct in_device *in_dev;
433 in_dev = __in_dev_get_rtnl(dev);
436 ipv4_devconf_setall(in_dev);
437 neigh_parms_data_state_setall(in_dev->arp_parms);
438 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
443 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
445 struct net_device *tunnel_dev, *new_dev;
446 struct ip_tunnel_parm p = { };
449 tunnel_dev = __dev_get_by_name(net, "tunl0");
453 p.iph.daddr = v->vifc_rmt_addr.s_addr;
454 p.iph.saddr = v->vifc_lcl_addr.s_addr;
457 p.iph.protocol = IPPROTO_IPIP;
458 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
460 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
462 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
467 new_dev = __dev_get_by_name(net, p.name);
471 new_dev->flags |= IFF_MULTICAST;
472 if (!ipmr_init_vif_indev(new_dev))
474 if (dev_open(new_dev, NULL))
477 err = dev_set_allmulti(new_dev, 1);
480 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
483 new_dev = ERR_PTR(err);
488 unregister_netdevice(new_dev);
490 return ERR_PTR(-ENOBUFS);
493 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
494 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
496 struct net *net = dev_net(dev);
497 struct mr_table *mrt;
498 struct flowi4 fl4 = {
499 .flowi4_oif = dev->ifindex,
500 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
501 .flowi4_mark = skb->mark,
505 err = ipmr_fib_lookup(net, &fl4, &mrt);
511 read_lock(&mrt_lock);
512 dev->stats.tx_bytes += skb->len;
513 dev->stats.tx_packets++;
514 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
515 read_unlock(&mrt_lock);
520 static int reg_vif_get_iflink(const struct net_device *dev)
525 static const struct net_device_ops reg_vif_netdev_ops = {
526 .ndo_start_xmit = reg_vif_xmit,
527 .ndo_get_iflink = reg_vif_get_iflink,
530 static void reg_vif_setup(struct net_device *dev)
532 dev->type = ARPHRD_PIMREG;
533 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
534 dev->flags = IFF_NOARP;
535 dev->netdev_ops = ®_vif_netdev_ops;
536 dev->needs_free_netdev = true;
537 dev->features |= NETIF_F_NETNS_LOCAL;
540 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
542 struct net_device *dev;
545 if (mrt->id == RT_TABLE_DEFAULT)
546 sprintf(name, "pimreg");
548 sprintf(name, "pimreg%u", mrt->id);
550 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
555 dev_net_set(dev, net);
557 if (register_netdevice(dev)) {
562 if (!ipmr_init_vif_indev(dev))
564 if (dev_open(dev, NULL))
572 unregister_netdevice(dev);
576 /* called with rcu_read_lock() */
577 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
580 struct net_device *reg_dev = NULL;
583 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
585 * a. packet is really sent to a multicast group
586 * b. packet is not a NULL-REGISTER
587 * c. packet is not truncated
589 if (!ipv4_is_multicast(encap->daddr) ||
590 encap->tot_len == 0 ||
591 ntohs(encap->tot_len) + pimlen > skb->len)
594 read_lock(&mrt_lock);
595 if (mrt->mroute_reg_vif_num >= 0)
596 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
597 read_unlock(&mrt_lock);
602 skb->mac_header = skb->network_header;
603 skb_pull(skb, (u8 *)encap - skb->data);
604 skb_reset_network_header(skb);
605 skb->protocol = htons(ETH_P_IP);
606 skb->ip_summed = CHECKSUM_NONE;
608 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
612 return NET_RX_SUCCESS;
615 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
621 static int call_ipmr_vif_entry_notifiers(struct net *net,
622 enum fib_event_type event_type,
623 struct vif_device *vif,
624 vifi_t vif_index, u32 tb_id)
626 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
627 vif, vif_index, tb_id,
628 &net->ipv4.ipmr_seq);
631 static int call_ipmr_mfc_entry_notifiers(struct net *net,
632 enum fib_event_type event_type,
633 struct mfc_cache *mfc, u32 tb_id)
635 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
636 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
640 * vif_delete - Delete a VIF entry
641 * @mrt: Table to delete from
642 * @vifi: VIF identifier to delete
643 * @notify: Set to 1, if the caller is a notifier_call
644 * @head: if unregistering the VIF, place it on this queue
646 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
647 struct list_head *head)
649 struct net *net = read_pnet(&mrt->net);
650 struct vif_device *v;
651 struct net_device *dev;
652 struct in_device *in_dev;
654 if (vifi < 0 || vifi >= mrt->maxvif)
655 return -EADDRNOTAVAIL;
657 v = &mrt->vif_table[vifi];
659 if (VIF_EXISTS(mrt, vifi))
660 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
663 write_lock_bh(&mrt_lock);
668 write_unlock_bh(&mrt_lock);
669 return -EADDRNOTAVAIL;
672 if (vifi == mrt->mroute_reg_vif_num)
673 mrt->mroute_reg_vif_num = -1;
675 if (vifi + 1 == mrt->maxvif) {
678 for (tmp = vifi - 1; tmp >= 0; tmp--) {
679 if (VIF_EXISTS(mrt, tmp))
685 write_unlock_bh(&mrt_lock);
687 dev_set_allmulti(dev, -1);
689 in_dev = __in_dev_get_rtnl(dev);
691 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
692 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
693 NETCONFA_MC_FORWARDING,
694 dev->ifindex, &in_dev->cnf);
695 ip_rt_multicast_event(in_dev);
698 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
699 unregister_netdevice_queue(dev, head);
705 static void ipmr_cache_free_rcu(struct rcu_head *head)
707 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
709 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
712 static void ipmr_cache_free(struct mfc_cache *c)
714 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
717 /* Destroy an unresolved cache entry, killing queued skbs
718 * and reporting error to netlink readers.
720 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
722 struct net *net = read_pnet(&mrt->net);
726 atomic_dec(&mrt->cache_resolve_queue_len);
728 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
729 if (ip_hdr(skb)->version == 0) {
730 struct nlmsghdr *nlh = skb_pull(skb,
731 sizeof(struct iphdr));
732 nlh->nlmsg_type = NLMSG_ERROR;
733 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
734 skb_trim(skb, nlh->nlmsg_len);
736 e->error = -ETIMEDOUT;
737 memset(&e->msg, 0, sizeof(e->msg));
739 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
748 /* Timer process for the unresolved queue. */
749 static void ipmr_expire_process(struct timer_list *t)
751 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
752 struct mr_mfc *c, *next;
753 unsigned long expires;
756 if (!spin_trylock(&mfc_unres_lock)) {
757 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
761 if (list_empty(&mrt->mfc_unres_queue))
767 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
768 if (time_after(c->mfc_un.unres.expires, now)) {
769 unsigned long interval = c->mfc_un.unres.expires - now;
770 if (interval < expires)
776 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
777 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
780 if (!list_empty(&mrt->mfc_unres_queue))
781 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
784 spin_unlock(&mfc_unres_lock);
787 /* Fill oifs list. It is called under write locked mrt_lock. */
788 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
793 cache->mfc_un.res.minvif = MAXVIFS;
794 cache->mfc_un.res.maxvif = 0;
795 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
797 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
798 if (VIF_EXISTS(mrt, vifi) &&
799 ttls[vifi] && ttls[vifi] < 255) {
800 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
801 if (cache->mfc_un.res.minvif > vifi)
802 cache->mfc_un.res.minvif = vifi;
803 if (cache->mfc_un.res.maxvif <= vifi)
804 cache->mfc_un.res.maxvif = vifi + 1;
807 cache->mfc_un.res.lastuse = jiffies;
810 static int vif_add(struct net *net, struct mr_table *mrt,
811 struct vifctl *vifc, int mrtsock)
813 struct netdev_phys_item_id ppid = { };
814 int vifi = vifc->vifc_vifi;
815 struct vif_device *v = &mrt->vif_table[vifi];
816 struct net_device *dev;
817 struct in_device *in_dev;
821 if (VIF_EXISTS(mrt, vifi))
824 switch (vifc->vifc_flags) {
826 if (!ipmr_pimsm_enabled())
828 /* Special Purpose VIF in PIM
829 * All the packets will be sent to the daemon
831 if (mrt->mroute_reg_vif_num >= 0)
833 dev = ipmr_reg_vif(net, mrt);
836 err = dev_set_allmulti(dev, 1);
838 unregister_netdevice(dev);
844 dev = ipmr_new_tunnel(net, vifc);
848 case VIFF_USE_IFINDEX:
850 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
851 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
852 if (dev && !__in_dev_get_rtnl(dev)) {
854 return -EADDRNOTAVAIL;
857 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
860 return -EADDRNOTAVAIL;
861 err = dev_set_allmulti(dev, 1);
871 in_dev = __in_dev_get_rtnl(dev);
874 return -EADDRNOTAVAIL;
876 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
877 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
878 dev->ifindex, &in_dev->cnf);
879 ip_rt_multicast_event(in_dev);
881 /* Fill in the VIF structures */
882 vif_device_init(v, dev, vifc->vifc_rate_limit,
883 vifc->vifc_threshold,
884 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
885 (VIFF_TUNNEL | VIFF_REGISTER));
887 err = dev_get_port_parent_id(dev, &ppid, true);
889 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
890 v->dev_parent_id.id_len = ppid.id_len;
892 v->dev_parent_id.id_len = 0;
895 v->local = vifc->vifc_lcl_addr.s_addr;
896 v->remote = vifc->vifc_rmt_addr.s_addr;
898 /* And finish update writing critical data */
899 write_lock_bh(&mrt_lock);
901 if (v->flags & VIFF_REGISTER)
902 mrt->mroute_reg_vif_num = vifi;
903 if (vifi+1 > mrt->maxvif)
904 mrt->maxvif = vifi+1;
905 write_unlock_bh(&mrt_lock);
906 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
910 /* called with rcu_read_lock() */
911 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
915 struct mfc_cache_cmp_arg arg = {
916 .mfc_mcastgrp = mcastgrp,
920 return mr_mfc_find(mrt, &arg);
923 /* Look for a (*,G) entry */
924 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
925 __be32 mcastgrp, int vifi)
927 struct mfc_cache_cmp_arg arg = {
928 .mfc_mcastgrp = mcastgrp,
929 .mfc_origin = htonl(INADDR_ANY)
932 if (mcastgrp == htonl(INADDR_ANY))
933 return mr_mfc_find_any_parent(mrt, vifi);
934 return mr_mfc_find_any(mrt, vifi, &arg);
937 /* Look for a (S,G,iif) entry if parent != -1 */
938 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
939 __be32 origin, __be32 mcastgrp,
942 struct mfc_cache_cmp_arg arg = {
943 .mfc_mcastgrp = mcastgrp,
944 .mfc_origin = origin,
947 return mr_mfc_find_parent(mrt, &arg, parent);
950 /* Allocate a multicast cache entry */
951 static struct mfc_cache *ipmr_cache_alloc(void)
953 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
956 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
957 c->_c.mfc_un.res.minvif = MAXVIFS;
958 c->_c.free = ipmr_cache_free_rcu;
959 refcount_set(&c->_c.mfc_un.res.refcount, 1);
964 static struct mfc_cache *ipmr_cache_alloc_unres(void)
966 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
969 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
970 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
975 /* A cache entry has gone into a resolved state from queued */
976 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
977 struct mfc_cache *uc, struct mfc_cache *c)
982 /* Play the pending entries through our router */
983 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
984 if (ip_hdr(skb)->version == 0) {
985 struct nlmsghdr *nlh = skb_pull(skb,
986 sizeof(struct iphdr));
988 if (mr_fill_mroute(mrt, skb, &c->_c,
989 nlmsg_data(nlh)) > 0) {
990 nlh->nlmsg_len = skb_tail_pointer(skb) -
993 nlh->nlmsg_type = NLMSG_ERROR;
994 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
995 skb_trim(skb, nlh->nlmsg_len);
997 e->error = -EMSGSIZE;
998 memset(&e->msg, 0, sizeof(e->msg));
1001 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1003 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1008 /* Bounce a cache query up to mrouted and netlink.
1010 * Called under mrt_lock.
1012 static int ipmr_cache_report(struct mr_table *mrt,
1013 struct sk_buff *pkt, vifi_t vifi, int assert)
1015 const int ihl = ip_hdrlen(pkt);
1016 struct sock *mroute_sk;
1017 struct igmphdr *igmp;
1018 struct igmpmsg *msg;
1019 struct sk_buff *skb;
1022 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1023 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1025 skb = alloc_skb(128, GFP_ATOMIC);
1030 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1031 /* Ugly, but we have no choice with this interface.
1032 * Duplicate old header, fix ihl, length etc.
1033 * And all this only to mangle msg->im_msgtype and
1034 * to set msg->im_mbz to "mbz" :-)
1036 skb_push(skb, sizeof(struct iphdr));
1037 skb_reset_network_header(skb);
1038 skb_reset_transport_header(skb);
1039 msg = (struct igmpmsg *)skb_network_header(skb);
1040 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1041 msg->im_msgtype = assert;
1043 if (assert == IGMPMSG_WRVIFWHOLE) {
1045 msg->im_vif_hi = vifi >> 8;
1047 msg->im_vif = mrt->mroute_reg_vif_num;
1048 msg->im_vif_hi = mrt->mroute_reg_vif_num >> 8;
1050 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1051 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1052 sizeof(struct iphdr));
1054 /* Copy the IP header */
1055 skb_set_network_header(skb, skb->len);
1057 skb_copy_to_linear_data(skb, pkt->data, ihl);
1058 /* Flag to the kernel this is a route add */
1059 ip_hdr(skb)->protocol = 0;
1060 msg = (struct igmpmsg *)skb_network_header(skb);
1062 msg->im_vif_hi = vifi >> 8;
1063 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1064 /* Add our header */
1065 igmp = skb_put(skb, sizeof(struct igmphdr));
1066 igmp->type = assert;
1067 msg->im_msgtype = assert;
1069 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1070 skb->transport_header = skb->network_header;
1074 mroute_sk = rcu_dereference(mrt->mroute_sk);
1081 igmpmsg_netlink_event(mrt, skb);
1083 /* Deliver to mrouted */
1084 ret = sock_queue_rcv_skb(mroute_sk, skb);
1087 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1094 /* Queue a packet for resolution. It gets locked cache entry! */
1095 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1096 struct sk_buff *skb, struct net_device *dev)
1098 const struct iphdr *iph = ip_hdr(skb);
1099 struct mfc_cache *c;
1103 spin_lock_bh(&mfc_unres_lock);
1104 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1105 if (c->mfc_mcastgrp == iph->daddr &&
1106 c->mfc_origin == iph->saddr) {
1113 /* Create a new entry if allowable */
1114 c = ipmr_cache_alloc_unres();
1116 spin_unlock_bh(&mfc_unres_lock);
1122 /* Fill in the new cache entry */
1123 c->_c.mfc_parent = -1;
1124 c->mfc_origin = iph->saddr;
1125 c->mfc_mcastgrp = iph->daddr;
1127 /* Reflect first query at mrouted. */
1128 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1131 /* If the report failed throw the cache entry
1134 spin_unlock_bh(&mfc_unres_lock);
1141 atomic_inc(&mrt->cache_resolve_queue_len);
1142 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1143 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1145 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1146 mod_timer(&mrt->ipmr_expire_timer,
1147 c->_c.mfc_un.unres.expires);
1150 /* See if we can append the packet */
1151 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1157 skb->skb_iif = dev->ifindex;
1159 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1163 spin_unlock_bh(&mfc_unres_lock);
1167 /* MFC cache manipulation by user space mroute daemon */
1169 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1171 struct net *net = read_pnet(&mrt->net);
1172 struct mfc_cache *c;
1174 /* The entries are added/deleted only under RTNL */
1176 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1177 mfc->mfcc_mcastgrp.s_addr, parent);
1181 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1182 list_del_rcu(&c->_c.list);
1183 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1184 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1185 mr_cache_put(&c->_c);
1190 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1191 struct mfcctl *mfc, int mrtsock, int parent)
1193 struct mfc_cache *uc, *c;
1198 if (mfc->mfcc_parent >= MAXVIFS)
1201 /* The entries are added/deleted only under RTNL */
1203 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1204 mfc->mfcc_mcastgrp.s_addr, parent);
1207 write_lock_bh(&mrt_lock);
1208 c->_c.mfc_parent = mfc->mfcc_parent;
1209 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1211 c->_c.mfc_flags |= MFC_STATIC;
1212 write_unlock_bh(&mrt_lock);
1213 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1215 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1219 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1220 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1223 c = ipmr_cache_alloc();
1227 c->mfc_origin = mfc->mfcc_origin.s_addr;
1228 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1229 c->_c.mfc_parent = mfc->mfcc_parent;
1230 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1232 c->_c.mfc_flags |= MFC_STATIC;
1234 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1237 pr_err("ipmr: rhtable insert error %d\n", ret);
1241 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1242 /* Check to see if we resolved a queued list. If so we
1243 * need to send on the frames and tidy up.
1246 spin_lock_bh(&mfc_unres_lock);
1247 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1248 uc = (struct mfc_cache *)_uc;
1249 if (uc->mfc_origin == c->mfc_origin &&
1250 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1251 list_del(&_uc->list);
1252 atomic_dec(&mrt->cache_resolve_queue_len);
1257 if (list_empty(&mrt->mfc_unres_queue))
1258 del_timer(&mrt->ipmr_expire_timer);
1259 spin_unlock_bh(&mfc_unres_lock);
1262 ipmr_cache_resolve(net, mrt, uc, c);
1263 ipmr_cache_free(uc);
1265 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1266 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1270 /* Close the multicast socket, and clear the vif tables etc */
1271 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1273 struct net *net = read_pnet(&mrt->net);
1274 struct mr_mfc *c, *tmp;
1275 struct mfc_cache *cache;
1279 /* Shut down all active vif entries */
1280 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1281 for (i = 0; i < mrt->maxvif; i++) {
1282 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1283 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1284 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1286 vif_delete(mrt, i, 0, &list);
1288 unregister_netdevice_many(&list);
1291 /* Wipe the cache */
1292 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1293 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1294 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1295 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1297 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1298 list_del_rcu(&c->list);
1299 cache = (struct mfc_cache *)c;
1300 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1302 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1307 if (flags & MRT_FLUSH_MFC) {
1308 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1309 spin_lock_bh(&mfc_unres_lock);
1310 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1312 cache = (struct mfc_cache *)c;
1313 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1314 ipmr_destroy_unres(mrt, cache);
1316 spin_unlock_bh(&mfc_unres_lock);
1321 /* called from ip_ra_control(), before an RCU grace period,
1322 * we don't need to call synchronize_rcu() here
1324 static void mrtsock_destruct(struct sock *sk)
1326 struct net *net = sock_net(sk);
1327 struct mr_table *mrt;
1330 ipmr_for_each_table(mrt, net) {
1331 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1332 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1333 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1334 NETCONFA_MC_FORWARDING,
1335 NETCONFA_IFINDEX_ALL,
1336 net->ipv4.devconf_all);
1337 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1338 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1344 /* Socket options and virtual interface manipulation. The whole
1345 * virtual interface system is a complete heap, but unfortunately
1346 * that's how BSD mrouted happens to think. Maybe one day with a proper
1347 * MOSPF/PIM router set up we can clean this up.
1350 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1351 unsigned int optlen)
1353 struct net *net = sock_net(sk);
1354 int val, ret = 0, parent = 0;
1355 struct mr_table *mrt;
1361 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1363 if (sk->sk_type != SOCK_RAW ||
1364 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1369 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1374 if (optname != MRT_INIT) {
1375 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1376 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1384 if (optlen != sizeof(int)) {
1388 if (rtnl_dereference(mrt->mroute_sk)) {
1393 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1395 rcu_assign_pointer(mrt->mroute_sk, sk);
1396 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1397 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1398 NETCONFA_MC_FORWARDING,
1399 NETCONFA_IFINDEX_ALL,
1400 net->ipv4.devconf_all);
1404 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1407 /* We need to unlock here because mrtsock_destruct takes
1408 * care of rtnl itself and we can't change that due to
1409 * the IP_ROUTER_ALERT setsockopt which runs without it.
1412 ret = ip_ra_control(sk, 0, NULL);
1418 if (optlen != sizeof(vif)) {
1422 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1426 if (vif.vifc_vifi >= MAXVIFS) {
1430 if (optname == MRT_ADD_VIF) {
1431 ret = vif_add(net, mrt, &vif,
1432 sk == rtnl_dereference(mrt->mroute_sk));
1434 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1437 /* Manipulate the forwarding caches. These live
1438 * in a sort of kernel/user symbiosis.
1444 case MRT_ADD_MFC_PROXY:
1445 case MRT_DEL_MFC_PROXY:
1446 if (optlen != sizeof(mfc)) {
1450 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1455 parent = mfc.mfcc_parent;
1456 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1457 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1459 ret = ipmr_mfc_add(net, mrt, &mfc,
1460 sk == rtnl_dereference(mrt->mroute_sk),
1464 if (optlen != sizeof(val)) {
1468 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1472 mroute_clean_tables(mrt, val);
1474 /* Control PIM assert. */
1476 if (optlen != sizeof(val)) {
1480 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1484 mrt->mroute_do_assert = val;
1487 if (!ipmr_pimsm_enabled()) {
1491 if (optlen != sizeof(val)) {
1495 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1500 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1502 if (val != mrt->mroute_do_pim) {
1503 mrt->mroute_do_pim = val;
1504 mrt->mroute_do_assert = val;
1505 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1509 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1513 if (optlen != sizeof(uval)) {
1517 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1522 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1525 mrt = ipmr_new_table(net, uval);
1529 raw_sk(sk)->ipmr_table = uval;
1532 /* Spurious command, or MRT_VERSION which you cannot set. */
1542 /* Getsock opt support for the multicast routing system. */
1543 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1547 struct net *net = sock_net(sk);
1548 struct mr_table *mrt;
1550 if (sk->sk_type != SOCK_RAW ||
1551 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1554 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1563 if (!ipmr_pimsm_enabled())
1564 return -ENOPROTOOPT;
1565 val = mrt->mroute_do_pim;
1568 val = mrt->mroute_do_assert;
1571 return -ENOPROTOOPT;
1574 if (get_user(olr, optlen))
1576 olr = min_t(unsigned int, olr, sizeof(int));
1579 if (put_user(olr, optlen))
1581 if (copy_to_user(optval, &val, olr))
1586 /* The IP multicast ioctl support routines. */
1587 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1589 struct sioc_sg_req sr;
1590 struct sioc_vif_req vr;
1591 struct vif_device *vif;
1592 struct mfc_cache *c;
1593 struct net *net = sock_net(sk);
1594 struct mr_table *mrt;
1596 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1602 if (copy_from_user(&vr, arg, sizeof(vr)))
1604 if (vr.vifi >= mrt->maxvif)
1606 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1607 read_lock(&mrt_lock);
1608 vif = &mrt->vif_table[vr.vifi];
1609 if (VIF_EXISTS(mrt, vr.vifi)) {
1610 vr.icount = vif->pkt_in;
1611 vr.ocount = vif->pkt_out;
1612 vr.ibytes = vif->bytes_in;
1613 vr.obytes = vif->bytes_out;
1614 read_unlock(&mrt_lock);
1616 if (copy_to_user(arg, &vr, sizeof(vr)))
1620 read_unlock(&mrt_lock);
1621 return -EADDRNOTAVAIL;
1623 if (copy_from_user(&sr, arg, sizeof(sr)))
1627 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1629 sr.pktcnt = c->_c.mfc_un.res.pkt;
1630 sr.bytecnt = c->_c.mfc_un.res.bytes;
1631 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1634 if (copy_to_user(arg, &sr, sizeof(sr)))
1639 return -EADDRNOTAVAIL;
1641 return -ENOIOCTLCMD;
1645 #ifdef CONFIG_COMPAT
1646 struct compat_sioc_sg_req {
1649 compat_ulong_t pktcnt;
1650 compat_ulong_t bytecnt;
1651 compat_ulong_t wrong_if;
1654 struct compat_sioc_vif_req {
1655 vifi_t vifi; /* Which iface */
1656 compat_ulong_t icount;
1657 compat_ulong_t ocount;
1658 compat_ulong_t ibytes;
1659 compat_ulong_t obytes;
1662 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1664 struct compat_sioc_sg_req sr;
1665 struct compat_sioc_vif_req vr;
1666 struct vif_device *vif;
1667 struct mfc_cache *c;
1668 struct net *net = sock_net(sk);
1669 struct mr_table *mrt;
1671 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1677 if (copy_from_user(&vr, arg, sizeof(vr)))
1679 if (vr.vifi >= mrt->maxvif)
1681 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1682 read_lock(&mrt_lock);
1683 vif = &mrt->vif_table[vr.vifi];
1684 if (VIF_EXISTS(mrt, vr.vifi)) {
1685 vr.icount = vif->pkt_in;
1686 vr.ocount = vif->pkt_out;
1687 vr.ibytes = vif->bytes_in;
1688 vr.obytes = vif->bytes_out;
1689 read_unlock(&mrt_lock);
1691 if (copy_to_user(arg, &vr, sizeof(vr)))
1695 read_unlock(&mrt_lock);
1696 return -EADDRNOTAVAIL;
1698 if (copy_from_user(&sr, arg, sizeof(sr)))
1702 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1704 sr.pktcnt = c->_c.mfc_un.res.pkt;
1705 sr.bytecnt = c->_c.mfc_un.res.bytes;
1706 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1709 if (copy_to_user(arg, &sr, sizeof(sr)))
1714 return -EADDRNOTAVAIL;
1716 return -ENOIOCTLCMD;
1721 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1723 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1724 struct net *net = dev_net(dev);
1725 struct mr_table *mrt;
1726 struct vif_device *v;
1729 if (event != NETDEV_UNREGISTER)
1732 ipmr_for_each_table(mrt, net) {
1733 v = &mrt->vif_table[0];
1734 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1736 vif_delete(mrt, ct, 1, NULL);
1742 static struct notifier_block ip_mr_notifier = {
1743 .notifier_call = ipmr_device_event,
1746 /* Encapsulate a packet by attaching a valid IPIP header to it.
1747 * This avoids tunnel drivers and other mess and gives us the speed so
1748 * important for multicast video.
1750 static void ip_encap(struct net *net, struct sk_buff *skb,
1751 __be32 saddr, __be32 daddr)
1754 const struct iphdr *old_iph = ip_hdr(skb);
1756 skb_push(skb, sizeof(struct iphdr));
1757 skb->transport_header = skb->network_header;
1758 skb_reset_network_header(skb);
1762 iph->tos = old_iph->tos;
1763 iph->ttl = old_iph->ttl;
1767 iph->protocol = IPPROTO_IPIP;
1769 iph->tot_len = htons(skb->len);
1770 ip_select_ident(net, skb, NULL);
1773 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1777 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1778 struct sk_buff *skb)
1780 struct ip_options *opt = &(IPCB(skb)->opt);
1782 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1783 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1785 if (unlikely(opt->optlen))
1786 ip_forward_options(skb);
1788 return dst_output(net, sk, skb);
1791 #ifdef CONFIG_NET_SWITCHDEV
1792 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1793 int in_vifi, int out_vifi)
1795 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1796 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1798 if (!skb->offload_l3_fwd_mark)
1800 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1802 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1803 &in_vif->dev_parent_id);
1806 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1807 int in_vifi, int out_vifi)
1813 /* Processing handlers for ipmr_forward */
1815 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1816 int in_vifi, struct sk_buff *skb, int vifi)
1818 const struct iphdr *iph = ip_hdr(skb);
1819 struct vif_device *vif = &mrt->vif_table[vifi];
1820 struct net_device *dev;
1828 if (vif->flags & VIFF_REGISTER) {
1830 vif->bytes_out += skb->len;
1831 vif->dev->stats.tx_bytes += skb->len;
1832 vif->dev->stats.tx_packets++;
1833 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1837 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1840 if (vif->flags & VIFF_TUNNEL) {
1841 rt = ip_route_output_ports(net, &fl4, NULL,
1842 vif->remote, vif->local,
1845 RT_TOS(iph->tos), vif->link);
1848 encap = sizeof(struct iphdr);
1850 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1853 RT_TOS(iph->tos), vif->link);
1860 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1861 /* Do not fragment multicasts. Alas, IPv4 does not
1862 * allow to send ICMP, so that packets will disappear
1865 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1870 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1872 if (skb_cow(skb, encap)) {
1878 vif->bytes_out += skb->len;
1881 skb_dst_set(skb, &rt->dst);
1882 ip_decrease_ttl(ip_hdr(skb));
1884 /* FIXME: forward and output firewalls used to be called here.
1885 * What do we do with netfilter? -- RR
1887 if (vif->flags & VIFF_TUNNEL) {
1888 ip_encap(net, skb, vif->local, vif->remote);
1889 /* FIXME: extra output firewall step used to be here. --RR */
1890 vif->dev->stats.tx_packets++;
1891 vif->dev->stats.tx_bytes += skb->len;
1894 IPCB(skb)->flags |= IPSKB_FORWARDED;
1896 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1897 * not only before forwarding, but after forwarding on all output
1898 * interfaces. It is clear, if mrouter runs a multicasting
1899 * program, it should receive packets not depending to what interface
1900 * program is joined.
1901 * If we will not make it, the program will have to join on all
1902 * interfaces. On the other hand, multihoming host (or router, but
1903 * not mrouter) cannot join to more than one interface - it will
1904 * result in receiving multiple packets.
1906 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1907 net, NULL, skb, skb->dev, dev,
1908 ipmr_forward_finish);
1915 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1919 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1920 if (mrt->vif_table[ct].dev == dev)
1926 /* "local" means that we should preserve one skb (for local delivery) */
1927 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1928 struct net_device *dev, struct sk_buff *skb,
1929 struct mfc_cache *c, int local)
1931 int true_vifi = ipmr_find_vif(mrt, dev);
1935 vif = c->_c.mfc_parent;
1936 c->_c.mfc_un.res.pkt++;
1937 c->_c.mfc_un.res.bytes += skb->len;
1938 c->_c.mfc_un.res.lastuse = jiffies;
1940 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1941 struct mfc_cache *cache_proxy;
1943 /* For an (*,G) entry, we only check that the incoming
1944 * interface is part of the static tree.
1946 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1948 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1952 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1953 if (mrt->vif_table[vif].dev != dev) {
1954 if (rt_is_output_route(skb_rtable(skb))) {
1955 /* It is our own packet, looped back.
1956 * Very complicated situation...
1958 * The best workaround until routing daemons will be
1959 * fixed is not to redistribute packet, if it was
1960 * send through wrong interface. It means, that
1961 * multicast applications WILL NOT work for
1962 * (S,G), which have default multicast route pointing
1963 * to wrong oif. In any case, it is not a good
1964 * idea to use multicasting applications on router.
1969 c->_c.mfc_un.res.wrong_if++;
1971 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1972 /* pimsm uses asserts, when switching from RPT to SPT,
1973 * so that we cannot check that packet arrived on an oif.
1974 * It is bad, but otherwise we would need to move pretty
1975 * large chunk of pimd to kernel. Ough... --ANK
1977 (mrt->mroute_do_pim ||
1978 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1980 c->_c.mfc_un.res.last_assert +
1981 MFC_ASSERT_THRESH)) {
1982 c->_c.mfc_un.res.last_assert = jiffies;
1983 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1984 if (mrt->mroute_do_wrvifwhole)
1985 ipmr_cache_report(mrt, skb, true_vifi,
1986 IGMPMSG_WRVIFWHOLE);
1992 mrt->vif_table[vif].pkt_in++;
1993 mrt->vif_table[vif].bytes_in += skb->len;
1995 /* Forward the frame */
1996 if (c->mfc_origin == htonl(INADDR_ANY) &&
1997 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
1998 if (true_vifi >= 0 &&
1999 true_vifi != c->_c.mfc_parent &&
2001 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2002 /* It's an (*,*) entry and the packet is not coming from
2003 * the upstream: forward the packet to the upstream
2006 psend = c->_c.mfc_parent;
2011 for (ct = c->_c.mfc_un.res.maxvif - 1;
2012 ct >= c->_c.mfc_un.res.minvif; ct--) {
2013 /* For (*,G) entry, don't forward to the incoming interface */
2014 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2016 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2018 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2021 ipmr_queue_xmit(net, mrt, true_vifi,
2030 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2033 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2036 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2046 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2048 struct rtable *rt = skb_rtable(skb);
2049 struct iphdr *iph = ip_hdr(skb);
2050 struct flowi4 fl4 = {
2051 .daddr = iph->daddr,
2052 .saddr = iph->saddr,
2053 .flowi4_tos = RT_TOS(iph->tos),
2054 .flowi4_oif = (rt_is_output_route(rt) ?
2055 skb->dev->ifindex : 0),
2056 .flowi4_iif = (rt_is_output_route(rt) ?
2059 .flowi4_mark = skb->mark,
2061 struct mr_table *mrt;
2064 err = ipmr_fib_lookup(net, &fl4, &mrt);
2066 return ERR_PTR(err);
2070 /* Multicast packets for forwarding arrive here
2071 * Called with rcu_read_lock();
2073 int ip_mr_input(struct sk_buff *skb)
2075 struct mfc_cache *cache;
2076 struct net *net = dev_net(skb->dev);
2077 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2078 struct mr_table *mrt;
2079 struct net_device *dev;
2081 /* skb->dev passed in is the loX master dev for vrfs.
2082 * As there are no vifs associated with loopback devices,
2083 * get the proper interface that does have a vif associated with it.
2086 if (netif_is_l3_master(skb->dev)) {
2087 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2094 /* Packet is looped back after forward, it should not be
2095 * forwarded second time, but still can be delivered locally.
2097 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2100 mrt = ipmr_rt_fib_lookup(net, skb);
2103 return PTR_ERR(mrt);
2106 if (IPCB(skb)->opt.router_alert) {
2107 if (ip_call_ra_chain(skb))
2109 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2110 /* IGMPv1 (and broken IGMPv2 implementations sort of
2111 * Cisco IOS <= 11.2(8)) do not put router alert
2112 * option to IGMP packets destined to routable
2113 * groups. It is very bad, because it means
2114 * that we can forward NO IGMP messages.
2116 struct sock *mroute_sk;
2118 mroute_sk = rcu_dereference(mrt->mroute_sk);
2121 raw_rcv(mroute_sk, skb);
2127 /* already under rcu_read_lock() */
2128 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2130 int vif = ipmr_find_vif(mrt, dev);
2133 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2137 /* No usable cache entry */
2142 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2143 ip_local_deliver(skb);
2149 read_lock(&mrt_lock);
2150 vif = ipmr_find_vif(mrt, dev);
2152 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2153 read_unlock(&mrt_lock);
2157 read_unlock(&mrt_lock);
2162 read_lock(&mrt_lock);
2163 ip_mr_forward(net, mrt, dev, skb, cache, local);
2164 read_unlock(&mrt_lock);
2167 return ip_local_deliver(skb);
2173 return ip_local_deliver(skb);
2178 #ifdef CONFIG_IP_PIMSM_V1
2179 /* Handle IGMP messages of PIMv1 */
2180 int pim_rcv_v1(struct sk_buff *skb)
2182 struct igmphdr *pim;
2183 struct net *net = dev_net(skb->dev);
2184 struct mr_table *mrt;
2186 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2189 pim = igmp_hdr(skb);
2191 mrt = ipmr_rt_fib_lookup(net, skb);
2194 if (!mrt->mroute_do_pim ||
2195 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2198 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2206 #ifdef CONFIG_IP_PIMSM_V2
2207 static int pim_rcv(struct sk_buff *skb)
2209 struct pimreghdr *pim;
2210 struct net *net = dev_net(skb->dev);
2211 struct mr_table *mrt;
2213 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2216 pim = (struct pimreghdr *)skb_transport_header(skb);
2217 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2218 (pim->flags & PIM_NULL_REGISTER) ||
2219 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2220 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2223 mrt = ipmr_rt_fib_lookup(net, skb);
2226 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2234 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2235 __be32 saddr, __be32 daddr,
2236 struct rtmsg *rtm, u32 portid)
2238 struct mfc_cache *cache;
2239 struct mr_table *mrt;
2242 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2247 cache = ipmr_cache_find(mrt, saddr, daddr);
2248 if (!cache && skb->dev) {
2249 int vif = ipmr_find_vif(mrt, skb->dev);
2252 cache = ipmr_cache_find_any(mrt, daddr, vif);
2255 struct sk_buff *skb2;
2257 struct net_device *dev;
2261 read_lock(&mrt_lock);
2263 vif = ipmr_find_vif(mrt, dev);
2265 read_unlock(&mrt_lock);
2270 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2272 read_unlock(&mrt_lock);
2277 NETLINK_CB(skb2).portid = portid;
2278 skb_push(skb2, sizeof(struct iphdr));
2279 skb_reset_network_header(skb2);
2281 iph->ihl = sizeof(struct iphdr) >> 2;
2285 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2286 read_unlock(&mrt_lock);
2291 read_lock(&mrt_lock);
2292 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2293 read_unlock(&mrt_lock);
2298 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2299 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2302 struct nlmsghdr *nlh;
2306 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2310 rtm = nlmsg_data(nlh);
2311 rtm->rtm_family = RTNL_FAMILY_IPMR;
2312 rtm->rtm_dst_len = 32;
2313 rtm->rtm_src_len = 32;
2315 rtm->rtm_table = mrt->id;
2316 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2317 goto nla_put_failure;
2318 rtm->rtm_type = RTN_MULTICAST;
2319 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2320 if (c->_c.mfc_flags & MFC_STATIC)
2321 rtm->rtm_protocol = RTPROT_STATIC;
2323 rtm->rtm_protocol = RTPROT_MROUTED;
2326 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2327 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2328 goto nla_put_failure;
2329 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2330 /* do not break the dump if cache is unresolved */
2331 if (err < 0 && err != -ENOENT)
2332 goto nla_put_failure;
2334 nlmsg_end(skb, nlh);
2338 nlmsg_cancel(skb, nlh);
2342 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2343 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2346 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2350 static size_t mroute_msgsize(bool unresolved, int maxvif)
2353 NLMSG_ALIGN(sizeof(struct rtmsg))
2354 + nla_total_size(4) /* RTA_TABLE */
2355 + nla_total_size(4) /* RTA_SRC */
2356 + nla_total_size(4) /* RTA_DST */
2361 + nla_total_size(4) /* RTA_IIF */
2362 + nla_total_size(0) /* RTA_MULTIPATH */
2363 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2365 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2371 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2374 struct net *net = read_pnet(&mrt->net);
2375 struct sk_buff *skb;
2378 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2384 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2388 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2394 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2397 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2400 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2401 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2402 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2403 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2404 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2405 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */
2406 /* IPMRA_CREPORT_PKT */
2407 + nla_total_size(payloadlen)
2413 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2415 struct net *net = read_pnet(&mrt->net);
2416 struct nlmsghdr *nlh;
2417 struct rtgenmsg *rtgenm;
2418 struct igmpmsg *msg;
2419 struct sk_buff *skb;
2423 payloadlen = pkt->len - sizeof(struct igmpmsg);
2424 msg = (struct igmpmsg *)skb_network_header(pkt);
2426 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2430 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2431 sizeof(struct rtgenmsg), 0);
2434 rtgenm = nlmsg_data(nlh);
2435 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2436 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2437 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2438 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2439 msg->im_src.s_addr) ||
2440 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2441 msg->im_dst.s_addr) ||
2442 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2443 goto nla_put_failure;
2445 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2446 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2447 nla_data(nla), payloadlen))
2448 goto nla_put_failure;
2450 nlmsg_end(skb, nlh);
2452 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2456 nlmsg_cancel(skb, nlh);
2459 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2462 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2463 const struct nlmsghdr *nlh,
2465 struct netlink_ext_ack *extack)
2470 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2471 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2475 if (!netlink_strict_get_check(skb))
2476 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2477 rtm_ipv4_policy, extack);
2479 rtm = nlmsg_data(nlh);
2480 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2481 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2482 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2483 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2484 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2488 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2489 rtm_ipv4_policy, extack);
2493 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2494 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2495 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2499 for (i = 0; i <= RTA_MAX; i++) {
2509 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2517 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2518 struct netlink_ext_ack *extack)
2520 struct net *net = sock_net(in_skb->sk);
2521 struct nlattr *tb[RTA_MAX + 1];
2522 struct sk_buff *skb = NULL;
2523 struct mfc_cache *cache;
2524 struct mr_table *mrt;
2529 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2533 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2534 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2535 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2537 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2543 /* entries are added/deleted only under RTNL */
2545 cache = ipmr_cache_find(mrt, src, grp);
2552 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2558 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2559 nlh->nlmsg_seq, cache,
2564 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2574 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2576 struct fib_dump_filter filter = {};
2579 if (cb->strict_check) {
2580 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2586 if (filter.table_id) {
2587 struct mr_table *mrt;
2589 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2591 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2594 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2597 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2598 &mfc_unres_lock, &filter);
2599 return skb->len ? : err;
2602 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2603 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2606 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2607 [RTA_SRC] = { .type = NLA_U32 },
2608 [RTA_DST] = { .type = NLA_U32 },
2609 [RTA_IIF] = { .type = NLA_U32 },
2610 [RTA_TABLE] = { .type = NLA_U32 },
2611 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2614 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2616 switch (rtm_protocol) {
2618 case RTPROT_MROUTED:
2624 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2626 struct rtnexthop *rtnh = nla_data(nla);
2627 int remaining = nla_len(nla), vifi = 0;
2629 while (rtnh_ok(rtnh, remaining)) {
2630 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2631 if (++vifi == MAXVIFS)
2633 rtnh = rtnh_next(rtnh, &remaining);
2636 return remaining > 0 ? -EINVAL : vifi;
2639 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2640 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2641 struct mfcctl *mfcc, int *mrtsock,
2642 struct mr_table **mrtret,
2643 struct netlink_ext_ack *extack)
2645 struct net_device *dev = NULL;
2646 u32 tblid = RT_TABLE_DEFAULT;
2647 struct mr_table *mrt;
2648 struct nlattr *attr;
2652 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2653 rtm_ipmr_policy, extack);
2656 rtm = nlmsg_data(nlh);
2659 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2660 rtm->rtm_type != RTN_MULTICAST ||
2661 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2662 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2665 memset(mfcc, 0, sizeof(*mfcc));
2666 mfcc->mfcc_parent = -1;
2668 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2669 switch (nla_type(attr)) {
2671 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2674 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2677 dev = __dev_get_by_index(net, nla_get_u32(attr));
2684 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2693 tblid = nla_get_u32(attr);
2697 mrt = ipmr_get_table(net, tblid);
2703 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2705 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2711 /* takes care of both newroute and delroute */
2712 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2713 struct netlink_ext_ack *extack)
2715 struct net *net = sock_net(skb->sk);
2716 int ret, mrtsock, parent;
2717 struct mr_table *tbl;
2722 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2726 parent = ret ? mfcc.mfcc_parent : -1;
2727 if (nlh->nlmsg_type == RTM_NEWROUTE)
2728 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2730 return ipmr_mfc_delete(tbl, &mfcc, parent);
2733 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2735 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2737 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2738 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2739 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2740 mrt->mroute_reg_vif_num) ||
2741 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2742 mrt->mroute_do_assert) ||
2743 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2744 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2745 mrt->mroute_do_wrvifwhole))
2751 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2753 struct nlattr *vif_nest;
2754 struct vif_device *vif;
2756 /* if the VIF doesn't exist just continue */
2757 if (!VIF_EXISTS(mrt, vifid))
2760 vif = &mrt->vif_table[vifid];
2761 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2764 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2765 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2766 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2767 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2769 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2771 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2773 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2775 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2776 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2777 nla_nest_cancel(skb, vif_nest);
2780 nla_nest_end(skb, vif_nest);
2785 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2786 struct netlink_ext_ack *extack)
2788 struct ifinfomsg *ifm;
2790 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2791 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2795 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2796 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2800 ifm = nlmsg_data(nlh);
2801 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2802 ifm->ifi_change || ifm->ifi_index) {
2803 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2810 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2812 struct net *net = sock_net(skb->sk);
2813 struct nlmsghdr *nlh = NULL;
2814 unsigned int t = 0, s_t;
2815 unsigned int e = 0, s_e;
2816 struct mr_table *mrt;
2818 if (cb->strict_check) {
2819 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2828 ipmr_for_each_table(mrt, net) {
2829 struct nlattr *vifs, *af;
2830 struct ifinfomsg *hdr;
2835 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2836 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2837 sizeof(*hdr), NLM_F_MULTI);
2841 hdr = nlmsg_data(nlh);
2842 memset(hdr, 0, sizeof(*hdr));
2843 hdr->ifi_family = RTNL_FAMILY_IPMR;
2845 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2847 nlmsg_cancel(skb, nlh);
2851 if (!ipmr_fill_table(mrt, skb)) {
2852 nlmsg_cancel(skb, nlh);
2856 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2858 nla_nest_end(skb, af);
2859 nlmsg_end(skb, nlh);
2862 for (i = 0; i < mrt->maxvif; i++) {
2865 if (!ipmr_fill_vif(mrt, i, skb)) {
2866 nla_nest_end(skb, vifs);
2867 nla_nest_end(skb, af);
2868 nlmsg_end(skb, nlh);
2876 nla_nest_end(skb, vifs);
2877 nla_nest_end(skb, af);
2878 nlmsg_end(skb, nlh);
2890 #ifdef CONFIG_PROC_FS
2891 /* The /proc interfaces to multicast routing :
2892 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2895 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2896 __acquires(mrt_lock)
2898 struct mr_vif_iter *iter = seq->private;
2899 struct net *net = seq_file_net(seq);
2900 struct mr_table *mrt;
2902 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2904 return ERR_PTR(-ENOENT);
2908 read_lock(&mrt_lock);
2909 return mr_vif_seq_start(seq, pos);
2912 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2913 __releases(mrt_lock)
2915 read_unlock(&mrt_lock);
2918 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2920 struct mr_vif_iter *iter = seq->private;
2921 struct mr_table *mrt = iter->mrt;
2923 if (v == SEQ_START_TOKEN) {
2925 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2927 const struct vif_device *vif = v;
2928 const char *name = vif->dev ?
2929 vif->dev->name : "none";
2932 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2933 vif - mrt->vif_table,
2934 name, vif->bytes_in, vif->pkt_in,
2935 vif->bytes_out, vif->pkt_out,
2936 vif->flags, vif->local, vif->remote);
2941 static const struct seq_operations ipmr_vif_seq_ops = {
2942 .start = ipmr_vif_seq_start,
2943 .next = mr_vif_seq_next,
2944 .stop = ipmr_vif_seq_stop,
2945 .show = ipmr_vif_seq_show,
2948 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2950 struct net *net = seq_file_net(seq);
2951 struct mr_table *mrt;
2953 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2955 return ERR_PTR(-ENOENT);
2957 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2960 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2964 if (v == SEQ_START_TOKEN) {
2966 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2968 const struct mfc_cache *mfc = v;
2969 const struct mr_mfc_iter *it = seq->private;
2970 const struct mr_table *mrt = it->mrt;
2972 seq_printf(seq, "%08X %08X %-3hd",
2973 (__force u32) mfc->mfc_mcastgrp,
2974 (__force u32) mfc->mfc_origin,
2975 mfc->_c.mfc_parent);
2977 if (it->cache != &mrt->mfc_unres_queue) {
2978 seq_printf(seq, " %8lu %8lu %8lu",
2979 mfc->_c.mfc_un.res.pkt,
2980 mfc->_c.mfc_un.res.bytes,
2981 mfc->_c.mfc_un.res.wrong_if);
2982 for (n = mfc->_c.mfc_un.res.minvif;
2983 n < mfc->_c.mfc_un.res.maxvif; n++) {
2984 if (VIF_EXISTS(mrt, n) &&
2985 mfc->_c.mfc_un.res.ttls[n] < 255)
2988 n, mfc->_c.mfc_un.res.ttls[n]);
2991 /* unresolved mfc_caches don't contain
2992 * pkt, bytes and wrong_if values
2994 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2996 seq_putc(seq, '\n');
3001 static const struct seq_operations ipmr_mfc_seq_ops = {
3002 .start = ipmr_mfc_seq_start,
3003 .next = mr_mfc_seq_next,
3004 .stop = mr_mfc_seq_stop,
3005 .show = ipmr_mfc_seq_show,
3009 #ifdef CONFIG_IP_PIMSM_V2
3010 static const struct net_protocol pim_protocol = {
3015 static unsigned int ipmr_seq_read(struct net *net)
3019 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3022 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3023 struct netlink_ext_ack *extack)
3025 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3026 ipmr_mr_table_iter, &mrt_lock, extack);
3029 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3030 .family = RTNL_FAMILY_IPMR,
3031 .fib_seq_read = ipmr_seq_read,
3032 .fib_dump = ipmr_dump,
3033 .owner = THIS_MODULE,
3036 static int __net_init ipmr_notifier_init(struct net *net)
3038 struct fib_notifier_ops *ops;
3040 net->ipv4.ipmr_seq = 0;
3042 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3044 return PTR_ERR(ops);
3045 net->ipv4.ipmr_notifier_ops = ops;
3050 static void __net_exit ipmr_notifier_exit(struct net *net)
3052 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3053 net->ipv4.ipmr_notifier_ops = NULL;
3056 /* Setup for IP multicast routing */
3057 static int __net_init ipmr_net_init(struct net *net)
3061 err = ipmr_notifier_init(net);
3063 goto ipmr_notifier_fail;
3065 err = ipmr_rules_init(net);
3067 goto ipmr_rules_fail;
3069 #ifdef CONFIG_PROC_FS
3071 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3072 sizeof(struct mr_vif_iter)))
3074 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3075 sizeof(struct mr_mfc_iter)))
3076 goto proc_cache_fail;
3080 #ifdef CONFIG_PROC_FS
3082 remove_proc_entry("ip_mr_vif", net->proc_net);
3084 ipmr_rules_exit(net);
3087 ipmr_notifier_exit(net);
3092 static void __net_exit ipmr_net_exit(struct net *net)
3094 #ifdef CONFIG_PROC_FS
3095 remove_proc_entry("ip_mr_cache", net->proc_net);
3096 remove_proc_entry("ip_mr_vif", net->proc_net);
3098 ipmr_notifier_exit(net);
3099 ipmr_rules_exit(net);
3102 static struct pernet_operations ipmr_net_ops = {
3103 .init = ipmr_net_init,
3104 .exit = ipmr_net_exit,
3107 int __init ip_mr_init(void)
3111 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3112 sizeof(struct mfc_cache),
3113 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3116 err = register_pernet_subsys(&ipmr_net_ops);
3118 goto reg_pernet_fail;
3120 err = register_netdevice_notifier(&ip_mr_notifier);
3122 goto reg_notif_fail;
3123 #ifdef CONFIG_IP_PIMSM_V2
3124 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3125 pr_err("%s: can't add PIM protocol\n", __func__);
3127 goto add_proto_fail;
3130 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3131 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3132 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3133 ipmr_rtm_route, NULL, 0);
3134 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3135 ipmr_rtm_route, NULL, 0);
3137 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3138 NULL, ipmr_rtm_dumplink, 0);
3141 #ifdef CONFIG_IP_PIMSM_V2
3143 unregister_netdevice_notifier(&ip_mr_notifier);
3146 unregister_pernet_subsys(&ipmr_net_ops);
3148 kmem_cache_destroy(mrt_cachep);