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 int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
199 struct fib_rule_hdr *frh, struct nlattr **tb,
200 struct netlink_ext_ack *extack)
205 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
211 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
212 struct fib_rule_hdr *frh)
220 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
221 .family = RTNL_FAMILY_IPMR,
222 .rule_size = sizeof(struct ipmr_rule),
223 .addr_size = sizeof(u32),
224 .action = ipmr_rule_action,
225 .match = ipmr_rule_match,
226 .configure = ipmr_rule_configure,
227 .compare = ipmr_rule_compare,
228 .fill = ipmr_rule_fill,
229 .nlgroup = RTNLGRP_IPV4_RULE,
230 .owner = THIS_MODULE,
233 static int __net_init ipmr_rules_init(struct net *net)
235 struct fib_rules_ops *ops;
236 struct mr_table *mrt;
239 ops = fib_rules_register(&ipmr_rules_ops_template, net);
243 INIT_LIST_HEAD(&net->ipv4.mr_tables);
245 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
255 net->ipv4.mr_rules_ops = ops;
260 ipmr_free_table(mrt);
263 fib_rules_unregister(ops);
267 static void __net_exit ipmr_rules_exit(struct net *net)
269 struct mr_table *mrt, *next;
272 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
273 list_del(&mrt->list);
274 ipmr_free_table(mrt);
276 fib_rules_unregister(net->ipv4.mr_rules_ops);
280 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
281 struct netlink_ext_ack *extack)
283 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
286 static unsigned int ipmr_rules_seq_read(struct net *net)
288 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
291 bool ipmr_rule_default(const struct fib_rule *rule)
293 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
295 EXPORT_SYMBOL(ipmr_rule_default);
297 #define ipmr_for_each_table(mrt, net) \
298 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
300 static struct mr_table *ipmr_mr_table_iter(struct net *net,
301 struct mr_table *mrt)
304 return net->ipv4.mrt;
308 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
310 return net->ipv4.mrt;
313 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
314 struct mr_table **mrt)
316 *mrt = net->ipv4.mrt;
320 static int __net_init ipmr_rules_init(struct net *net)
322 struct mr_table *mrt;
324 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
331 static void __net_exit ipmr_rules_exit(struct net *net)
334 ipmr_free_table(net->ipv4.mrt);
335 net->ipv4.mrt = NULL;
339 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
340 struct netlink_ext_ack *extack)
345 static unsigned int ipmr_rules_seq_read(struct net *net)
350 bool ipmr_rule_default(const struct fib_rule *rule)
354 EXPORT_SYMBOL(ipmr_rule_default);
357 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
360 const struct mfc_cache_cmp_arg *cmparg = arg->key;
361 struct mfc_cache *c = (struct mfc_cache *)ptr;
363 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
364 cmparg->mfc_origin != c->mfc_origin;
367 static const struct rhashtable_params ipmr_rht_params = {
368 .head_offset = offsetof(struct mr_mfc, mnode),
369 .key_offset = offsetof(struct mfc_cache, cmparg),
370 .key_len = sizeof(struct mfc_cache_cmp_arg),
372 .obj_cmpfn = ipmr_hash_cmp,
373 .automatic_shrinking = true,
376 static void ipmr_new_table_set(struct mr_table *mrt,
379 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
380 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
385 .mfc_mcastgrp = htonl(INADDR_ANY),
386 .mfc_origin = htonl(INADDR_ANY),
389 static struct mr_table_ops ipmr_mr_table_ops = {
390 .rht_params = &ipmr_rht_params,
391 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
394 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
396 struct mr_table *mrt;
398 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
399 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
400 return ERR_PTR(-EINVAL);
402 mrt = ipmr_get_table(net, id);
406 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
407 ipmr_expire_process, ipmr_new_table_set);
410 static void ipmr_free_table(struct mr_table *mrt)
412 del_timer_sync(&mrt->ipmr_expire_timer);
413 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
414 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
415 rhltable_destroy(&mrt->mfc_hash);
419 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
421 /* Initialize ipmr pimreg/tunnel in_device */
422 static bool ipmr_init_vif_indev(const struct net_device *dev)
424 struct in_device *in_dev;
428 in_dev = __in_dev_get_rtnl(dev);
431 ipv4_devconf_setall(in_dev);
432 neigh_parms_data_state_setall(in_dev->arp_parms);
433 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
438 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
440 struct net_device *tunnel_dev, *new_dev;
441 struct ip_tunnel_parm p = { };
444 tunnel_dev = __dev_get_by_name(net, "tunl0");
448 p.iph.daddr = v->vifc_rmt_addr.s_addr;
449 p.iph.saddr = v->vifc_lcl_addr.s_addr;
452 p.iph.protocol = IPPROTO_IPIP;
453 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
455 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
457 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
462 new_dev = __dev_get_by_name(net, p.name);
466 new_dev->flags |= IFF_MULTICAST;
467 if (!ipmr_init_vif_indev(new_dev))
469 if (dev_open(new_dev, NULL))
472 err = dev_set_allmulti(new_dev, 1);
475 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
478 new_dev = ERR_PTR(err);
483 unregister_netdevice(new_dev);
485 return ERR_PTR(-ENOBUFS);
488 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
489 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
491 struct net *net = dev_net(dev);
492 struct mr_table *mrt;
493 struct flowi4 fl4 = {
494 .flowi4_oif = dev->ifindex,
495 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
496 .flowi4_mark = skb->mark,
500 err = ipmr_fib_lookup(net, &fl4, &mrt);
506 read_lock(&mrt_lock);
507 dev->stats.tx_bytes += skb->len;
508 dev->stats.tx_packets++;
509 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
510 read_unlock(&mrt_lock);
515 static int reg_vif_get_iflink(const struct net_device *dev)
520 static const struct net_device_ops reg_vif_netdev_ops = {
521 .ndo_start_xmit = reg_vif_xmit,
522 .ndo_get_iflink = reg_vif_get_iflink,
525 static void reg_vif_setup(struct net_device *dev)
527 dev->type = ARPHRD_PIMREG;
528 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
529 dev->flags = IFF_NOARP;
530 dev->netdev_ops = ®_vif_netdev_ops;
531 dev->needs_free_netdev = true;
532 dev->features |= NETIF_F_NETNS_LOCAL;
535 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
537 struct net_device *dev;
540 if (mrt->id == RT_TABLE_DEFAULT)
541 sprintf(name, "pimreg");
543 sprintf(name, "pimreg%u", mrt->id);
545 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
550 dev_net_set(dev, net);
552 if (register_netdevice(dev)) {
557 if (!ipmr_init_vif_indev(dev))
559 if (dev_open(dev, NULL))
567 unregister_netdevice(dev);
571 /* called with rcu_read_lock() */
572 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
575 struct net_device *reg_dev = NULL;
578 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
580 * a. packet is really sent to a multicast group
581 * b. packet is not a NULL-REGISTER
582 * c. packet is not truncated
584 if (!ipv4_is_multicast(encap->daddr) ||
585 encap->tot_len == 0 ||
586 ntohs(encap->tot_len) + pimlen > skb->len)
589 read_lock(&mrt_lock);
590 if (mrt->mroute_reg_vif_num >= 0)
591 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
592 read_unlock(&mrt_lock);
597 skb->mac_header = skb->network_header;
598 skb_pull(skb, (u8 *)encap - skb->data);
599 skb_reset_network_header(skb);
600 skb->protocol = htons(ETH_P_IP);
601 skb->ip_summed = CHECKSUM_NONE;
603 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
607 return NET_RX_SUCCESS;
610 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
616 static int call_ipmr_vif_entry_notifiers(struct net *net,
617 enum fib_event_type event_type,
618 struct vif_device *vif,
619 vifi_t vif_index, u32 tb_id)
621 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
622 vif, vif_index, tb_id,
623 &net->ipv4.ipmr_seq);
626 static int call_ipmr_mfc_entry_notifiers(struct net *net,
627 enum fib_event_type event_type,
628 struct mfc_cache *mfc, u32 tb_id)
630 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
631 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
635 * vif_delete - Delete a VIF entry
636 * @mrt: Table to delete from
637 * @vifi: VIF identifier to delete
638 * @notify: Set to 1, if the caller is a notifier_call
639 * @head: if unregistering the VIF, place it on this queue
641 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
642 struct list_head *head)
644 struct net *net = read_pnet(&mrt->net);
645 struct vif_device *v;
646 struct net_device *dev;
647 struct in_device *in_dev;
649 if (vifi < 0 || vifi >= mrt->maxvif)
650 return -EADDRNOTAVAIL;
652 v = &mrt->vif_table[vifi];
654 if (VIF_EXISTS(mrt, vifi))
655 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
658 write_lock_bh(&mrt_lock);
663 write_unlock_bh(&mrt_lock);
664 return -EADDRNOTAVAIL;
667 if (vifi == mrt->mroute_reg_vif_num)
668 mrt->mroute_reg_vif_num = -1;
670 if (vifi + 1 == mrt->maxvif) {
673 for (tmp = vifi - 1; tmp >= 0; tmp--) {
674 if (VIF_EXISTS(mrt, tmp))
680 write_unlock_bh(&mrt_lock);
682 dev_set_allmulti(dev, -1);
684 in_dev = __in_dev_get_rtnl(dev);
686 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
687 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
688 NETCONFA_MC_FORWARDING,
689 dev->ifindex, &in_dev->cnf);
690 ip_rt_multicast_event(in_dev);
693 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
694 unregister_netdevice_queue(dev, head);
696 dev_put_track(dev, &v->dev_tracker);
700 static void ipmr_cache_free_rcu(struct rcu_head *head)
702 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
704 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
707 static void ipmr_cache_free(struct mfc_cache *c)
709 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
712 /* Destroy an unresolved cache entry, killing queued skbs
713 * and reporting error to netlink readers.
715 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
717 struct net *net = read_pnet(&mrt->net);
721 atomic_dec(&mrt->cache_resolve_queue_len);
723 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
724 if (ip_hdr(skb)->version == 0) {
725 struct nlmsghdr *nlh = skb_pull(skb,
726 sizeof(struct iphdr));
727 nlh->nlmsg_type = NLMSG_ERROR;
728 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
729 skb_trim(skb, nlh->nlmsg_len);
731 e->error = -ETIMEDOUT;
732 memset(&e->msg, 0, sizeof(e->msg));
734 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
743 /* Timer process for the unresolved queue. */
744 static void ipmr_expire_process(struct timer_list *t)
746 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
747 struct mr_mfc *c, *next;
748 unsigned long expires;
751 if (!spin_trylock(&mfc_unres_lock)) {
752 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
756 if (list_empty(&mrt->mfc_unres_queue))
762 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
763 if (time_after(c->mfc_un.unres.expires, now)) {
764 unsigned long interval = c->mfc_un.unres.expires - now;
765 if (interval < expires)
771 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
772 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
775 if (!list_empty(&mrt->mfc_unres_queue))
776 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
779 spin_unlock(&mfc_unres_lock);
782 /* Fill oifs list. It is called under write locked mrt_lock. */
783 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
788 cache->mfc_un.res.minvif = MAXVIFS;
789 cache->mfc_un.res.maxvif = 0;
790 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
792 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
793 if (VIF_EXISTS(mrt, vifi) &&
794 ttls[vifi] && ttls[vifi] < 255) {
795 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
796 if (cache->mfc_un.res.minvif > vifi)
797 cache->mfc_un.res.minvif = vifi;
798 if (cache->mfc_un.res.maxvif <= vifi)
799 cache->mfc_un.res.maxvif = vifi + 1;
802 cache->mfc_un.res.lastuse = jiffies;
805 static int vif_add(struct net *net, struct mr_table *mrt,
806 struct vifctl *vifc, int mrtsock)
808 struct netdev_phys_item_id ppid = { };
809 int vifi = vifc->vifc_vifi;
810 struct vif_device *v = &mrt->vif_table[vifi];
811 struct net_device *dev;
812 struct in_device *in_dev;
816 if (VIF_EXISTS(mrt, vifi))
819 switch (vifc->vifc_flags) {
821 if (!ipmr_pimsm_enabled())
823 /* Special Purpose VIF in PIM
824 * All the packets will be sent to the daemon
826 if (mrt->mroute_reg_vif_num >= 0)
828 dev = ipmr_reg_vif(net, mrt);
831 err = dev_set_allmulti(dev, 1);
833 unregister_netdevice(dev);
839 dev = ipmr_new_tunnel(net, vifc);
843 case VIFF_USE_IFINDEX:
845 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
846 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
847 if (dev && !__in_dev_get_rtnl(dev)) {
849 return -EADDRNOTAVAIL;
852 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
855 return -EADDRNOTAVAIL;
856 err = dev_set_allmulti(dev, 1);
866 in_dev = __in_dev_get_rtnl(dev);
869 return -EADDRNOTAVAIL;
871 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
872 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
873 dev->ifindex, &in_dev->cnf);
874 ip_rt_multicast_event(in_dev);
876 /* Fill in the VIF structures */
877 vif_device_init(v, dev, vifc->vifc_rate_limit,
878 vifc->vifc_threshold,
879 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
880 (VIFF_TUNNEL | VIFF_REGISTER));
882 err = dev_get_port_parent_id(dev, &ppid, true);
884 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
885 v->dev_parent_id.id_len = ppid.id_len;
887 v->dev_parent_id.id_len = 0;
890 v->local = vifc->vifc_lcl_addr.s_addr;
891 v->remote = vifc->vifc_rmt_addr.s_addr;
893 /* And finish update writing critical data */
894 write_lock_bh(&mrt_lock);
896 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
897 if (v->flags & VIFF_REGISTER)
898 mrt->mroute_reg_vif_num = vifi;
899 if (vifi+1 > mrt->maxvif)
900 mrt->maxvif = vifi+1;
901 write_unlock_bh(&mrt_lock);
902 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
906 /* called with rcu_read_lock() */
907 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
911 struct mfc_cache_cmp_arg arg = {
912 .mfc_mcastgrp = mcastgrp,
916 return mr_mfc_find(mrt, &arg);
919 /* Look for a (*,G) entry */
920 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
921 __be32 mcastgrp, int vifi)
923 struct mfc_cache_cmp_arg arg = {
924 .mfc_mcastgrp = mcastgrp,
925 .mfc_origin = htonl(INADDR_ANY)
928 if (mcastgrp == htonl(INADDR_ANY))
929 return mr_mfc_find_any_parent(mrt, vifi);
930 return mr_mfc_find_any(mrt, vifi, &arg);
933 /* Look for a (S,G,iif) entry if parent != -1 */
934 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
935 __be32 origin, __be32 mcastgrp,
938 struct mfc_cache_cmp_arg arg = {
939 .mfc_mcastgrp = mcastgrp,
940 .mfc_origin = origin,
943 return mr_mfc_find_parent(mrt, &arg, parent);
946 /* Allocate a multicast cache entry */
947 static struct mfc_cache *ipmr_cache_alloc(void)
949 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
952 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
953 c->_c.mfc_un.res.minvif = MAXVIFS;
954 c->_c.free = ipmr_cache_free_rcu;
955 refcount_set(&c->_c.mfc_un.res.refcount, 1);
960 static struct mfc_cache *ipmr_cache_alloc_unres(void)
962 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
965 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
966 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
971 /* A cache entry has gone into a resolved state from queued */
972 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
973 struct mfc_cache *uc, struct mfc_cache *c)
978 /* Play the pending entries through our router */
979 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
980 if (ip_hdr(skb)->version == 0) {
981 struct nlmsghdr *nlh = skb_pull(skb,
982 sizeof(struct iphdr));
984 if (mr_fill_mroute(mrt, skb, &c->_c,
985 nlmsg_data(nlh)) > 0) {
986 nlh->nlmsg_len = skb_tail_pointer(skb) -
989 nlh->nlmsg_type = NLMSG_ERROR;
990 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
991 skb_trim(skb, nlh->nlmsg_len);
993 e->error = -EMSGSIZE;
994 memset(&e->msg, 0, sizeof(e->msg));
997 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
999 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1004 /* Bounce a cache query up to mrouted and netlink.
1006 * Called under mrt_lock.
1008 static int ipmr_cache_report(struct mr_table *mrt,
1009 struct sk_buff *pkt, vifi_t vifi, int assert)
1011 const int ihl = ip_hdrlen(pkt);
1012 struct sock *mroute_sk;
1013 struct igmphdr *igmp;
1014 struct igmpmsg *msg;
1015 struct sk_buff *skb;
1018 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1019 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1021 skb = alloc_skb(128, GFP_ATOMIC);
1026 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1027 /* Ugly, but we have no choice with this interface.
1028 * Duplicate old header, fix ihl, length etc.
1029 * And all this only to mangle msg->im_msgtype and
1030 * to set msg->im_mbz to "mbz" :-)
1032 skb_push(skb, sizeof(struct iphdr));
1033 skb_reset_network_header(skb);
1034 skb_reset_transport_header(skb);
1035 msg = (struct igmpmsg *)skb_network_header(skb);
1036 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1037 msg->im_msgtype = assert;
1039 if (assert == IGMPMSG_WRVIFWHOLE) {
1041 msg->im_vif_hi = vifi >> 8;
1043 msg->im_vif = mrt->mroute_reg_vif_num;
1044 msg->im_vif_hi = mrt->mroute_reg_vif_num >> 8;
1046 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1047 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1048 sizeof(struct iphdr));
1050 /* Copy the IP header */
1051 skb_set_network_header(skb, skb->len);
1053 skb_copy_to_linear_data(skb, pkt->data, ihl);
1054 /* Flag to the kernel this is a route add */
1055 ip_hdr(skb)->protocol = 0;
1056 msg = (struct igmpmsg *)skb_network_header(skb);
1058 msg->im_vif_hi = vifi >> 8;
1059 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1060 /* Add our header */
1061 igmp = skb_put(skb, sizeof(struct igmphdr));
1062 igmp->type = assert;
1063 msg->im_msgtype = assert;
1065 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1066 skb->transport_header = skb->network_header;
1070 mroute_sk = rcu_dereference(mrt->mroute_sk);
1077 igmpmsg_netlink_event(mrt, skb);
1079 /* Deliver to mrouted */
1080 ret = sock_queue_rcv_skb(mroute_sk, skb);
1083 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1090 /* Queue a packet for resolution. It gets locked cache entry! */
1091 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1092 struct sk_buff *skb, struct net_device *dev)
1094 const struct iphdr *iph = ip_hdr(skb);
1095 struct mfc_cache *c;
1099 spin_lock_bh(&mfc_unres_lock);
1100 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1101 if (c->mfc_mcastgrp == iph->daddr &&
1102 c->mfc_origin == iph->saddr) {
1109 /* Create a new entry if allowable */
1110 c = ipmr_cache_alloc_unres();
1112 spin_unlock_bh(&mfc_unres_lock);
1118 /* Fill in the new cache entry */
1119 c->_c.mfc_parent = -1;
1120 c->mfc_origin = iph->saddr;
1121 c->mfc_mcastgrp = iph->daddr;
1123 /* Reflect first query at mrouted. */
1124 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1127 /* If the report failed throw the cache entry
1130 spin_unlock_bh(&mfc_unres_lock);
1137 atomic_inc(&mrt->cache_resolve_queue_len);
1138 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1139 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1141 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1142 mod_timer(&mrt->ipmr_expire_timer,
1143 c->_c.mfc_un.unres.expires);
1146 /* See if we can append the packet */
1147 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1153 skb->skb_iif = dev->ifindex;
1155 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1159 spin_unlock_bh(&mfc_unres_lock);
1163 /* MFC cache manipulation by user space mroute daemon */
1165 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1167 struct net *net = read_pnet(&mrt->net);
1168 struct mfc_cache *c;
1170 /* The entries are added/deleted only under RTNL */
1172 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1173 mfc->mfcc_mcastgrp.s_addr, parent);
1177 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1178 list_del_rcu(&c->_c.list);
1179 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1180 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1181 mr_cache_put(&c->_c);
1186 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1187 struct mfcctl *mfc, int mrtsock, int parent)
1189 struct mfc_cache *uc, *c;
1194 if (mfc->mfcc_parent >= MAXVIFS)
1197 /* The entries are added/deleted only under RTNL */
1199 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1200 mfc->mfcc_mcastgrp.s_addr, parent);
1203 write_lock_bh(&mrt_lock);
1204 c->_c.mfc_parent = mfc->mfcc_parent;
1205 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1207 c->_c.mfc_flags |= MFC_STATIC;
1208 write_unlock_bh(&mrt_lock);
1209 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1211 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1215 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1216 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1219 c = ipmr_cache_alloc();
1223 c->mfc_origin = mfc->mfcc_origin.s_addr;
1224 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1225 c->_c.mfc_parent = mfc->mfcc_parent;
1226 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1228 c->_c.mfc_flags |= MFC_STATIC;
1230 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1233 pr_err("ipmr: rhtable insert error %d\n", ret);
1237 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1238 /* Check to see if we resolved a queued list. If so we
1239 * need to send on the frames and tidy up.
1242 spin_lock_bh(&mfc_unres_lock);
1243 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1244 uc = (struct mfc_cache *)_uc;
1245 if (uc->mfc_origin == c->mfc_origin &&
1246 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1247 list_del(&_uc->list);
1248 atomic_dec(&mrt->cache_resolve_queue_len);
1253 if (list_empty(&mrt->mfc_unres_queue))
1254 del_timer(&mrt->ipmr_expire_timer);
1255 spin_unlock_bh(&mfc_unres_lock);
1258 ipmr_cache_resolve(net, mrt, uc, c);
1259 ipmr_cache_free(uc);
1261 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1262 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1266 /* Close the multicast socket, and clear the vif tables etc */
1267 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1269 struct net *net = read_pnet(&mrt->net);
1270 struct mr_mfc *c, *tmp;
1271 struct mfc_cache *cache;
1275 /* Shut down all active vif entries */
1276 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1277 for (i = 0; i < mrt->maxvif; i++) {
1278 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1279 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1280 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1282 vif_delete(mrt, i, 0, &list);
1284 unregister_netdevice_many(&list);
1287 /* Wipe the cache */
1288 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1289 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1290 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1291 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1293 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1294 list_del_rcu(&c->list);
1295 cache = (struct mfc_cache *)c;
1296 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1298 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1303 if (flags & MRT_FLUSH_MFC) {
1304 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1305 spin_lock_bh(&mfc_unres_lock);
1306 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1308 cache = (struct mfc_cache *)c;
1309 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1310 ipmr_destroy_unres(mrt, cache);
1312 spin_unlock_bh(&mfc_unres_lock);
1317 /* called from ip_ra_control(), before an RCU grace period,
1318 * we don't need to call synchronize_rcu() here
1320 static void mrtsock_destruct(struct sock *sk)
1322 struct net *net = sock_net(sk);
1323 struct mr_table *mrt;
1326 ipmr_for_each_table(mrt, net) {
1327 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1328 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1329 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1330 NETCONFA_MC_FORWARDING,
1331 NETCONFA_IFINDEX_ALL,
1332 net->ipv4.devconf_all);
1333 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1334 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1340 /* Socket options and virtual interface manipulation. The whole
1341 * virtual interface system is a complete heap, but unfortunately
1342 * that's how BSD mrouted happens to think. Maybe one day with a proper
1343 * MOSPF/PIM router set up we can clean this up.
1346 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1347 unsigned int optlen)
1349 struct net *net = sock_net(sk);
1350 int val, ret = 0, parent = 0;
1351 struct mr_table *mrt;
1357 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1359 if (sk->sk_type != SOCK_RAW ||
1360 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1365 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1370 if (optname != MRT_INIT) {
1371 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1372 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1380 if (optlen != sizeof(int)) {
1384 if (rtnl_dereference(mrt->mroute_sk)) {
1389 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1391 rcu_assign_pointer(mrt->mroute_sk, sk);
1392 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1393 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1394 NETCONFA_MC_FORWARDING,
1395 NETCONFA_IFINDEX_ALL,
1396 net->ipv4.devconf_all);
1400 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1403 /* We need to unlock here because mrtsock_destruct takes
1404 * care of rtnl itself and we can't change that due to
1405 * the IP_ROUTER_ALERT setsockopt which runs without it.
1408 ret = ip_ra_control(sk, 0, NULL);
1414 if (optlen != sizeof(vif)) {
1418 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1422 if (vif.vifc_vifi >= MAXVIFS) {
1426 if (optname == MRT_ADD_VIF) {
1427 ret = vif_add(net, mrt, &vif,
1428 sk == rtnl_dereference(mrt->mroute_sk));
1430 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1433 /* Manipulate the forwarding caches. These live
1434 * in a sort of kernel/user symbiosis.
1440 case MRT_ADD_MFC_PROXY:
1441 case MRT_DEL_MFC_PROXY:
1442 if (optlen != sizeof(mfc)) {
1446 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1451 parent = mfc.mfcc_parent;
1452 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1453 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1455 ret = ipmr_mfc_add(net, mrt, &mfc,
1456 sk == rtnl_dereference(mrt->mroute_sk),
1460 if (optlen != sizeof(val)) {
1464 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1468 mroute_clean_tables(mrt, val);
1470 /* Control PIM assert. */
1472 if (optlen != sizeof(val)) {
1476 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1480 mrt->mroute_do_assert = val;
1483 if (!ipmr_pimsm_enabled()) {
1487 if (optlen != sizeof(val)) {
1491 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1496 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1498 if (val != mrt->mroute_do_pim) {
1499 mrt->mroute_do_pim = val;
1500 mrt->mroute_do_assert = val;
1501 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1505 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1509 if (optlen != sizeof(uval)) {
1513 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1518 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1521 mrt = ipmr_new_table(net, uval);
1525 raw_sk(sk)->ipmr_table = uval;
1528 /* Spurious command, or MRT_VERSION which you cannot set. */
1538 /* Getsock opt support for the multicast routing system. */
1539 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1543 struct net *net = sock_net(sk);
1544 struct mr_table *mrt;
1546 if (sk->sk_type != SOCK_RAW ||
1547 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1550 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1559 if (!ipmr_pimsm_enabled())
1560 return -ENOPROTOOPT;
1561 val = mrt->mroute_do_pim;
1564 val = mrt->mroute_do_assert;
1567 return -ENOPROTOOPT;
1570 if (get_user(olr, optlen))
1572 olr = min_t(unsigned int, olr, sizeof(int));
1575 if (put_user(olr, optlen))
1577 if (copy_to_user(optval, &val, olr))
1582 /* The IP multicast ioctl support routines. */
1583 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1585 struct sioc_sg_req sr;
1586 struct sioc_vif_req vr;
1587 struct vif_device *vif;
1588 struct mfc_cache *c;
1589 struct net *net = sock_net(sk);
1590 struct mr_table *mrt;
1592 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1598 if (copy_from_user(&vr, arg, sizeof(vr)))
1600 if (vr.vifi >= mrt->maxvif)
1602 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1603 read_lock(&mrt_lock);
1604 vif = &mrt->vif_table[vr.vifi];
1605 if (VIF_EXISTS(mrt, vr.vifi)) {
1606 vr.icount = vif->pkt_in;
1607 vr.ocount = vif->pkt_out;
1608 vr.ibytes = vif->bytes_in;
1609 vr.obytes = vif->bytes_out;
1610 read_unlock(&mrt_lock);
1612 if (copy_to_user(arg, &vr, sizeof(vr)))
1616 read_unlock(&mrt_lock);
1617 return -EADDRNOTAVAIL;
1619 if (copy_from_user(&sr, arg, sizeof(sr)))
1623 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1625 sr.pktcnt = c->_c.mfc_un.res.pkt;
1626 sr.bytecnt = c->_c.mfc_un.res.bytes;
1627 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1630 if (copy_to_user(arg, &sr, sizeof(sr)))
1635 return -EADDRNOTAVAIL;
1637 return -ENOIOCTLCMD;
1641 #ifdef CONFIG_COMPAT
1642 struct compat_sioc_sg_req {
1645 compat_ulong_t pktcnt;
1646 compat_ulong_t bytecnt;
1647 compat_ulong_t wrong_if;
1650 struct compat_sioc_vif_req {
1651 vifi_t vifi; /* Which iface */
1652 compat_ulong_t icount;
1653 compat_ulong_t ocount;
1654 compat_ulong_t ibytes;
1655 compat_ulong_t obytes;
1658 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1660 struct compat_sioc_sg_req sr;
1661 struct compat_sioc_vif_req vr;
1662 struct vif_device *vif;
1663 struct mfc_cache *c;
1664 struct net *net = sock_net(sk);
1665 struct mr_table *mrt;
1667 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1673 if (copy_from_user(&vr, arg, sizeof(vr)))
1675 if (vr.vifi >= mrt->maxvif)
1677 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1678 read_lock(&mrt_lock);
1679 vif = &mrt->vif_table[vr.vifi];
1680 if (VIF_EXISTS(mrt, vr.vifi)) {
1681 vr.icount = vif->pkt_in;
1682 vr.ocount = vif->pkt_out;
1683 vr.ibytes = vif->bytes_in;
1684 vr.obytes = vif->bytes_out;
1685 read_unlock(&mrt_lock);
1687 if (copy_to_user(arg, &vr, sizeof(vr)))
1691 read_unlock(&mrt_lock);
1692 return -EADDRNOTAVAIL;
1694 if (copy_from_user(&sr, arg, sizeof(sr)))
1698 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1700 sr.pktcnt = c->_c.mfc_un.res.pkt;
1701 sr.bytecnt = c->_c.mfc_un.res.bytes;
1702 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1705 if (copy_to_user(arg, &sr, sizeof(sr)))
1710 return -EADDRNOTAVAIL;
1712 return -ENOIOCTLCMD;
1717 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1719 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1720 struct net *net = dev_net(dev);
1721 struct mr_table *mrt;
1722 struct vif_device *v;
1725 if (event != NETDEV_UNREGISTER)
1728 ipmr_for_each_table(mrt, net) {
1729 v = &mrt->vif_table[0];
1730 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1732 vif_delete(mrt, ct, 1, NULL);
1738 static struct notifier_block ip_mr_notifier = {
1739 .notifier_call = ipmr_device_event,
1742 /* Encapsulate a packet by attaching a valid IPIP header to it.
1743 * This avoids tunnel drivers and other mess and gives us the speed so
1744 * important for multicast video.
1746 static void ip_encap(struct net *net, struct sk_buff *skb,
1747 __be32 saddr, __be32 daddr)
1750 const struct iphdr *old_iph = ip_hdr(skb);
1752 skb_push(skb, sizeof(struct iphdr));
1753 skb->transport_header = skb->network_header;
1754 skb_reset_network_header(skb);
1758 iph->tos = old_iph->tos;
1759 iph->ttl = old_iph->ttl;
1763 iph->protocol = IPPROTO_IPIP;
1765 iph->tot_len = htons(skb->len);
1766 ip_select_ident(net, skb, NULL);
1769 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1773 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1774 struct sk_buff *skb)
1776 struct ip_options *opt = &(IPCB(skb)->opt);
1778 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1779 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1781 if (unlikely(opt->optlen))
1782 ip_forward_options(skb);
1784 return dst_output(net, sk, skb);
1787 #ifdef CONFIG_NET_SWITCHDEV
1788 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1789 int in_vifi, int out_vifi)
1791 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1792 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1794 if (!skb->offload_l3_fwd_mark)
1796 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1798 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1799 &in_vif->dev_parent_id);
1802 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1803 int in_vifi, int out_vifi)
1809 /* Processing handlers for ipmr_forward */
1811 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1812 int in_vifi, struct sk_buff *skb, int vifi)
1814 const struct iphdr *iph = ip_hdr(skb);
1815 struct vif_device *vif = &mrt->vif_table[vifi];
1816 struct net_device *dev;
1824 if (vif->flags & VIFF_REGISTER) {
1826 vif->bytes_out += skb->len;
1827 vif->dev->stats.tx_bytes += skb->len;
1828 vif->dev->stats.tx_packets++;
1829 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1833 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1836 if (vif->flags & VIFF_TUNNEL) {
1837 rt = ip_route_output_ports(net, &fl4, NULL,
1838 vif->remote, vif->local,
1841 RT_TOS(iph->tos), vif->link);
1844 encap = sizeof(struct iphdr);
1846 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1849 RT_TOS(iph->tos), vif->link);
1856 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1857 /* Do not fragment multicasts. Alas, IPv4 does not
1858 * allow to send ICMP, so that packets will disappear
1861 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1866 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1868 if (skb_cow(skb, encap)) {
1874 vif->bytes_out += skb->len;
1877 skb_dst_set(skb, &rt->dst);
1878 ip_decrease_ttl(ip_hdr(skb));
1880 /* FIXME: forward and output firewalls used to be called here.
1881 * What do we do with netfilter? -- RR
1883 if (vif->flags & VIFF_TUNNEL) {
1884 ip_encap(net, skb, vif->local, vif->remote);
1885 /* FIXME: extra output firewall step used to be here. --RR */
1886 vif->dev->stats.tx_packets++;
1887 vif->dev->stats.tx_bytes += skb->len;
1890 IPCB(skb)->flags |= IPSKB_FORWARDED;
1892 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1893 * not only before forwarding, but after forwarding on all output
1894 * interfaces. It is clear, if mrouter runs a multicasting
1895 * program, it should receive packets not depending to what interface
1896 * program is joined.
1897 * If we will not make it, the program will have to join on all
1898 * interfaces. On the other hand, multihoming host (or router, but
1899 * not mrouter) cannot join to more than one interface - it will
1900 * result in receiving multiple packets.
1902 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1903 net, NULL, skb, skb->dev, dev,
1904 ipmr_forward_finish);
1911 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1915 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1916 if (mrt->vif_table[ct].dev == dev)
1922 /* "local" means that we should preserve one skb (for local delivery) */
1923 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1924 struct net_device *dev, struct sk_buff *skb,
1925 struct mfc_cache *c, int local)
1927 int true_vifi = ipmr_find_vif(mrt, dev);
1931 vif = c->_c.mfc_parent;
1932 c->_c.mfc_un.res.pkt++;
1933 c->_c.mfc_un.res.bytes += skb->len;
1934 c->_c.mfc_un.res.lastuse = jiffies;
1936 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1937 struct mfc_cache *cache_proxy;
1939 /* For an (*,G) entry, we only check that the incoming
1940 * interface is part of the static tree.
1942 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1944 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1948 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1949 if (mrt->vif_table[vif].dev != dev) {
1950 if (rt_is_output_route(skb_rtable(skb))) {
1951 /* It is our own packet, looped back.
1952 * Very complicated situation...
1954 * The best workaround until routing daemons will be
1955 * fixed is not to redistribute packet, if it was
1956 * send through wrong interface. It means, that
1957 * multicast applications WILL NOT work for
1958 * (S,G), which have default multicast route pointing
1959 * to wrong oif. In any case, it is not a good
1960 * idea to use multicasting applications on router.
1965 c->_c.mfc_un.res.wrong_if++;
1967 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1968 /* pimsm uses asserts, when switching from RPT to SPT,
1969 * so that we cannot check that packet arrived on an oif.
1970 * It is bad, but otherwise we would need to move pretty
1971 * large chunk of pimd to kernel. Ough... --ANK
1973 (mrt->mroute_do_pim ||
1974 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1976 c->_c.mfc_un.res.last_assert +
1977 MFC_ASSERT_THRESH)) {
1978 c->_c.mfc_un.res.last_assert = jiffies;
1979 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1980 if (mrt->mroute_do_wrvifwhole)
1981 ipmr_cache_report(mrt, skb, true_vifi,
1982 IGMPMSG_WRVIFWHOLE);
1988 mrt->vif_table[vif].pkt_in++;
1989 mrt->vif_table[vif].bytes_in += skb->len;
1991 /* Forward the frame */
1992 if (c->mfc_origin == htonl(INADDR_ANY) &&
1993 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
1994 if (true_vifi >= 0 &&
1995 true_vifi != c->_c.mfc_parent &&
1997 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
1998 /* It's an (*,*) entry and the packet is not coming from
1999 * the upstream: forward the packet to the upstream
2002 psend = c->_c.mfc_parent;
2007 for (ct = c->_c.mfc_un.res.maxvif - 1;
2008 ct >= c->_c.mfc_un.res.minvif; ct--) {
2009 /* For (*,G) entry, don't forward to the incoming interface */
2010 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2012 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2014 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2017 ipmr_queue_xmit(net, mrt, true_vifi,
2026 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2029 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2032 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2042 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2044 struct rtable *rt = skb_rtable(skb);
2045 struct iphdr *iph = ip_hdr(skb);
2046 struct flowi4 fl4 = {
2047 .daddr = iph->daddr,
2048 .saddr = iph->saddr,
2049 .flowi4_tos = RT_TOS(iph->tos),
2050 .flowi4_oif = (rt_is_output_route(rt) ?
2051 skb->dev->ifindex : 0),
2052 .flowi4_iif = (rt_is_output_route(rt) ?
2055 .flowi4_mark = skb->mark,
2057 struct mr_table *mrt;
2060 err = ipmr_fib_lookup(net, &fl4, &mrt);
2062 return ERR_PTR(err);
2066 /* Multicast packets for forwarding arrive here
2067 * Called with rcu_read_lock();
2069 int ip_mr_input(struct sk_buff *skb)
2071 struct mfc_cache *cache;
2072 struct net *net = dev_net(skb->dev);
2073 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2074 struct mr_table *mrt;
2075 struct net_device *dev;
2077 /* skb->dev passed in is the loX master dev for vrfs.
2078 * As there are no vifs associated with loopback devices,
2079 * get the proper interface that does have a vif associated with it.
2082 if (netif_is_l3_master(skb->dev)) {
2083 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2090 /* Packet is looped back after forward, it should not be
2091 * forwarded second time, but still can be delivered locally.
2093 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2096 mrt = ipmr_rt_fib_lookup(net, skb);
2099 return PTR_ERR(mrt);
2102 if (IPCB(skb)->opt.router_alert) {
2103 if (ip_call_ra_chain(skb))
2105 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2106 /* IGMPv1 (and broken IGMPv2 implementations sort of
2107 * Cisco IOS <= 11.2(8)) do not put router alert
2108 * option to IGMP packets destined to routable
2109 * groups. It is very bad, because it means
2110 * that we can forward NO IGMP messages.
2112 struct sock *mroute_sk;
2114 mroute_sk = rcu_dereference(mrt->mroute_sk);
2117 raw_rcv(mroute_sk, skb);
2123 /* already under rcu_read_lock() */
2124 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2126 int vif = ipmr_find_vif(mrt, dev);
2129 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2133 /* No usable cache entry */
2138 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2139 ip_local_deliver(skb);
2145 read_lock(&mrt_lock);
2146 vif = ipmr_find_vif(mrt, dev);
2148 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2149 read_unlock(&mrt_lock);
2153 read_unlock(&mrt_lock);
2158 read_lock(&mrt_lock);
2159 ip_mr_forward(net, mrt, dev, skb, cache, local);
2160 read_unlock(&mrt_lock);
2163 return ip_local_deliver(skb);
2169 return ip_local_deliver(skb);
2174 #ifdef CONFIG_IP_PIMSM_V1
2175 /* Handle IGMP messages of PIMv1 */
2176 int pim_rcv_v1(struct sk_buff *skb)
2178 struct igmphdr *pim;
2179 struct net *net = dev_net(skb->dev);
2180 struct mr_table *mrt;
2182 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2185 pim = igmp_hdr(skb);
2187 mrt = ipmr_rt_fib_lookup(net, skb);
2190 if (!mrt->mroute_do_pim ||
2191 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2194 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2202 #ifdef CONFIG_IP_PIMSM_V2
2203 static int pim_rcv(struct sk_buff *skb)
2205 struct pimreghdr *pim;
2206 struct net *net = dev_net(skb->dev);
2207 struct mr_table *mrt;
2209 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2212 pim = (struct pimreghdr *)skb_transport_header(skb);
2213 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2214 (pim->flags & PIM_NULL_REGISTER) ||
2215 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2216 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2219 mrt = ipmr_rt_fib_lookup(net, skb);
2222 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2230 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2231 __be32 saddr, __be32 daddr,
2232 struct rtmsg *rtm, u32 portid)
2234 struct mfc_cache *cache;
2235 struct mr_table *mrt;
2238 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2243 cache = ipmr_cache_find(mrt, saddr, daddr);
2244 if (!cache && skb->dev) {
2245 int vif = ipmr_find_vif(mrt, skb->dev);
2248 cache = ipmr_cache_find_any(mrt, daddr, vif);
2251 struct sk_buff *skb2;
2253 struct net_device *dev;
2257 read_lock(&mrt_lock);
2259 vif = ipmr_find_vif(mrt, dev);
2261 read_unlock(&mrt_lock);
2266 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2268 read_unlock(&mrt_lock);
2273 NETLINK_CB(skb2).portid = portid;
2274 skb_push(skb2, sizeof(struct iphdr));
2275 skb_reset_network_header(skb2);
2277 iph->ihl = sizeof(struct iphdr) >> 2;
2281 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2282 read_unlock(&mrt_lock);
2287 read_lock(&mrt_lock);
2288 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2289 read_unlock(&mrt_lock);
2294 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2295 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2298 struct nlmsghdr *nlh;
2302 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2306 rtm = nlmsg_data(nlh);
2307 rtm->rtm_family = RTNL_FAMILY_IPMR;
2308 rtm->rtm_dst_len = 32;
2309 rtm->rtm_src_len = 32;
2311 rtm->rtm_table = mrt->id;
2312 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2313 goto nla_put_failure;
2314 rtm->rtm_type = RTN_MULTICAST;
2315 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2316 if (c->_c.mfc_flags & MFC_STATIC)
2317 rtm->rtm_protocol = RTPROT_STATIC;
2319 rtm->rtm_protocol = RTPROT_MROUTED;
2322 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2323 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2324 goto nla_put_failure;
2325 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2326 /* do not break the dump if cache is unresolved */
2327 if (err < 0 && err != -ENOENT)
2328 goto nla_put_failure;
2330 nlmsg_end(skb, nlh);
2334 nlmsg_cancel(skb, nlh);
2338 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2339 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2342 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2346 static size_t mroute_msgsize(bool unresolved, int maxvif)
2349 NLMSG_ALIGN(sizeof(struct rtmsg))
2350 + nla_total_size(4) /* RTA_TABLE */
2351 + nla_total_size(4) /* RTA_SRC */
2352 + nla_total_size(4) /* RTA_DST */
2357 + nla_total_size(4) /* RTA_IIF */
2358 + nla_total_size(0) /* RTA_MULTIPATH */
2359 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2361 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2367 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2370 struct net *net = read_pnet(&mrt->net);
2371 struct sk_buff *skb;
2374 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2380 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2384 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2390 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2393 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2396 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2397 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2398 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2399 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2400 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2401 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */
2402 /* IPMRA_CREPORT_PKT */
2403 + nla_total_size(payloadlen)
2409 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2411 struct net *net = read_pnet(&mrt->net);
2412 struct nlmsghdr *nlh;
2413 struct rtgenmsg *rtgenm;
2414 struct igmpmsg *msg;
2415 struct sk_buff *skb;
2419 payloadlen = pkt->len - sizeof(struct igmpmsg);
2420 msg = (struct igmpmsg *)skb_network_header(pkt);
2422 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2426 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2427 sizeof(struct rtgenmsg), 0);
2430 rtgenm = nlmsg_data(nlh);
2431 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2432 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2433 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2434 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2435 msg->im_src.s_addr) ||
2436 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2437 msg->im_dst.s_addr) ||
2438 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2439 goto nla_put_failure;
2441 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2442 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2443 nla_data(nla), payloadlen))
2444 goto nla_put_failure;
2446 nlmsg_end(skb, nlh);
2448 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2452 nlmsg_cancel(skb, nlh);
2455 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2458 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2459 const struct nlmsghdr *nlh,
2461 struct netlink_ext_ack *extack)
2466 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2467 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2471 if (!netlink_strict_get_check(skb))
2472 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2473 rtm_ipv4_policy, extack);
2475 rtm = nlmsg_data(nlh);
2476 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2477 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2478 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2479 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2480 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2484 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2485 rtm_ipv4_policy, extack);
2489 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2490 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2491 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2495 for (i = 0; i <= RTA_MAX; i++) {
2505 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2513 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2514 struct netlink_ext_ack *extack)
2516 struct net *net = sock_net(in_skb->sk);
2517 struct nlattr *tb[RTA_MAX + 1];
2518 struct sk_buff *skb = NULL;
2519 struct mfc_cache *cache;
2520 struct mr_table *mrt;
2525 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2529 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2530 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2531 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2533 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2539 /* entries are added/deleted only under RTNL */
2541 cache = ipmr_cache_find(mrt, src, grp);
2548 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2554 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2555 nlh->nlmsg_seq, cache,
2560 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2570 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2572 struct fib_dump_filter filter = {};
2575 if (cb->strict_check) {
2576 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2582 if (filter.table_id) {
2583 struct mr_table *mrt;
2585 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2587 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2590 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2593 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2594 &mfc_unres_lock, &filter);
2595 return skb->len ? : err;
2598 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2599 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2602 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2603 [RTA_SRC] = { .type = NLA_U32 },
2604 [RTA_DST] = { .type = NLA_U32 },
2605 [RTA_IIF] = { .type = NLA_U32 },
2606 [RTA_TABLE] = { .type = NLA_U32 },
2607 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2610 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2612 switch (rtm_protocol) {
2614 case RTPROT_MROUTED:
2620 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2622 struct rtnexthop *rtnh = nla_data(nla);
2623 int remaining = nla_len(nla), vifi = 0;
2625 while (rtnh_ok(rtnh, remaining)) {
2626 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2627 if (++vifi == MAXVIFS)
2629 rtnh = rtnh_next(rtnh, &remaining);
2632 return remaining > 0 ? -EINVAL : vifi;
2635 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2636 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2637 struct mfcctl *mfcc, int *mrtsock,
2638 struct mr_table **mrtret,
2639 struct netlink_ext_ack *extack)
2641 struct net_device *dev = NULL;
2642 u32 tblid = RT_TABLE_DEFAULT;
2643 struct mr_table *mrt;
2644 struct nlattr *attr;
2648 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2649 rtm_ipmr_policy, extack);
2652 rtm = nlmsg_data(nlh);
2655 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2656 rtm->rtm_type != RTN_MULTICAST ||
2657 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2658 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2661 memset(mfcc, 0, sizeof(*mfcc));
2662 mfcc->mfcc_parent = -1;
2664 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2665 switch (nla_type(attr)) {
2667 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2670 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2673 dev = __dev_get_by_index(net, nla_get_u32(attr));
2680 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2689 tblid = nla_get_u32(attr);
2693 mrt = ipmr_get_table(net, tblid);
2699 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2701 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2707 /* takes care of both newroute and delroute */
2708 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2709 struct netlink_ext_ack *extack)
2711 struct net *net = sock_net(skb->sk);
2712 int ret, mrtsock, parent;
2713 struct mr_table *tbl;
2718 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2722 parent = ret ? mfcc.mfcc_parent : -1;
2723 if (nlh->nlmsg_type == RTM_NEWROUTE)
2724 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2726 return ipmr_mfc_delete(tbl, &mfcc, parent);
2729 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2731 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2733 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2734 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2735 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2736 mrt->mroute_reg_vif_num) ||
2737 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2738 mrt->mroute_do_assert) ||
2739 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2740 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2741 mrt->mroute_do_wrvifwhole))
2747 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2749 struct nlattr *vif_nest;
2750 struct vif_device *vif;
2752 /* if the VIF doesn't exist just continue */
2753 if (!VIF_EXISTS(mrt, vifid))
2756 vif = &mrt->vif_table[vifid];
2757 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2760 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2761 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2762 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2763 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2765 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2767 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2769 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2771 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2772 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2773 nla_nest_cancel(skb, vif_nest);
2776 nla_nest_end(skb, vif_nest);
2781 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2782 struct netlink_ext_ack *extack)
2784 struct ifinfomsg *ifm;
2786 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2787 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2791 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2792 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2796 ifm = nlmsg_data(nlh);
2797 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2798 ifm->ifi_change || ifm->ifi_index) {
2799 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2806 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2808 struct net *net = sock_net(skb->sk);
2809 struct nlmsghdr *nlh = NULL;
2810 unsigned int t = 0, s_t;
2811 unsigned int e = 0, s_e;
2812 struct mr_table *mrt;
2814 if (cb->strict_check) {
2815 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2824 ipmr_for_each_table(mrt, net) {
2825 struct nlattr *vifs, *af;
2826 struct ifinfomsg *hdr;
2831 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2832 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2833 sizeof(*hdr), NLM_F_MULTI);
2837 hdr = nlmsg_data(nlh);
2838 memset(hdr, 0, sizeof(*hdr));
2839 hdr->ifi_family = RTNL_FAMILY_IPMR;
2841 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2843 nlmsg_cancel(skb, nlh);
2847 if (!ipmr_fill_table(mrt, skb)) {
2848 nlmsg_cancel(skb, nlh);
2852 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2854 nla_nest_end(skb, af);
2855 nlmsg_end(skb, nlh);
2858 for (i = 0; i < mrt->maxvif; i++) {
2861 if (!ipmr_fill_vif(mrt, i, skb)) {
2862 nla_nest_end(skb, vifs);
2863 nla_nest_end(skb, af);
2864 nlmsg_end(skb, nlh);
2872 nla_nest_end(skb, vifs);
2873 nla_nest_end(skb, af);
2874 nlmsg_end(skb, nlh);
2886 #ifdef CONFIG_PROC_FS
2887 /* The /proc interfaces to multicast routing :
2888 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2891 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2892 __acquires(mrt_lock)
2894 struct mr_vif_iter *iter = seq->private;
2895 struct net *net = seq_file_net(seq);
2896 struct mr_table *mrt;
2898 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2900 return ERR_PTR(-ENOENT);
2904 read_lock(&mrt_lock);
2905 return mr_vif_seq_start(seq, pos);
2908 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2909 __releases(mrt_lock)
2911 read_unlock(&mrt_lock);
2914 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2916 struct mr_vif_iter *iter = seq->private;
2917 struct mr_table *mrt = iter->mrt;
2919 if (v == SEQ_START_TOKEN) {
2921 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2923 const struct vif_device *vif = v;
2924 const char *name = vif->dev ?
2925 vif->dev->name : "none";
2928 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2929 vif - mrt->vif_table,
2930 name, vif->bytes_in, vif->pkt_in,
2931 vif->bytes_out, vif->pkt_out,
2932 vif->flags, vif->local, vif->remote);
2937 static const struct seq_operations ipmr_vif_seq_ops = {
2938 .start = ipmr_vif_seq_start,
2939 .next = mr_vif_seq_next,
2940 .stop = ipmr_vif_seq_stop,
2941 .show = ipmr_vif_seq_show,
2944 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2946 struct net *net = seq_file_net(seq);
2947 struct mr_table *mrt;
2949 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2951 return ERR_PTR(-ENOENT);
2953 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2956 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2960 if (v == SEQ_START_TOKEN) {
2962 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2964 const struct mfc_cache *mfc = v;
2965 const struct mr_mfc_iter *it = seq->private;
2966 const struct mr_table *mrt = it->mrt;
2968 seq_printf(seq, "%08X %08X %-3hd",
2969 (__force u32) mfc->mfc_mcastgrp,
2970 (__force u32) mfc->mfc_origin,
2971 mfc->_c.mfc_parent);
2973 if (it->cache != &mrt->mfc_unres_queue) {
2974 seq_printf(seq, " %8lu %8lu %8lu",
2975 mfc->_c.mfc_un.res.pkt,
2976 mfc->_c.mfc_un.res.bytes,
2977 mfc->_c.mfc_un.res.wrong_if);
2978 for (n = mfc->_c.mfc_un.res.minvif;
2979 n < mfc->_c.mfc_un.res.maxvif; n++) {
2980 if (VIF_EXISTS(mrt, n) &&
2981 mfc->_c.mfc_un.res.ttls[n] < 255)
2984 n, mfc->_c.mfc_un.res.ttls[n]);
2987 /* unresolved mfc_caches don't contain
2988 * pkt, bytes and wrong_if values
2990 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2992 seq_putc(seq, '\n');
2997 static const struct seq_operations ipmr_mfc_seq_ops = {
2998 .start = ipmr_mfc_seq_start,
2999 .next = mr_mfc_seq_next,
3000 .stop = mr_mfc_seq_stop,
3001 .show = ipmr_mfc_seq_show,
3005 #ifdef CONFIG_IP_PIMSM_V2
3006 static const struct net_protocol pim_protocol = {
3011 static unsigned int ipmr_seq_read(struct net *net)
3015 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3018 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3019 struct netlink_ext_ack *extack)
3021 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3022 ipmr_mr_table_iter, &mrt_lock, extack);
3025 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3026 .family = RTNL_FAMILY_IPMR,
3027 .fib_seq_read = ipmr_seq_read,
3028 .fib_dump = ipmr_dump,
3029 .owner = THIS_MODULE,
3032 static int __net_init ipmr_notifier_init(struct net *net)
3034 struct fib_notifier_ops *ops;
3036 net->ipv4.ipmr_seq = 0;
3038 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3040 return PTR_ERR(ops);
3041 net->ipv4.ipmr_notifier_ops = ops;
3046 static void __net_exit ipmr_notifier_exit(struct net *net)
3048 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3049 net->ipv4.ipmr_notifier_ops = NULL;
3052 /* Setup for IP multicast routing */
3053 static int __net_init ipmr_net_init(struct net *net)
3057 err = ipmr_notifier_init(net);
3059 goto ipmr_notifier_fail;
3061 err = ipmr_rules_init(net);
3063 goto ipmr_rules_fail;
3065 #ifdef CONFIG_PROC_FS
3067 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3068 sizeof(struct mr_vif_iter)))
3070 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3071 sizeof(struct mr_mfc_iter)))
3072 goto proc_cache_fail;
3076 #ifdef CONFIG_PROC_FS
3078 remove_proc_entry("ip_mr_vif", net->proc_net);
3080 ipmr_rules_exit(net);
3083 ipmr_notifier_exit(net);
3088 static void __net_exit ipmr_net_exit(struct net *net)
3090 #ifdef CONFIG_PROC_FS
3091 remove_proc_entry("ip_mr_cache", net->proc_net);
3092 remove_proc_entry("ip_mr_vif", net->proc_net);
3094 ipmr_notifier_exit(net);
3095 ipmr_rules_exit(net);
3098 static struct pernet_operations ipmr_net_ops = {
3099 .init = ipmr_net_init,
3100 .exit = ipmr_net_exit,
3103 int __init ip_mr_init(void)
3107 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3108 sizeof(struct mfc_cache),
3109 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3112 err = register_pernet_subsys(&ipmr_net_ops);
3114 goto reg_pernet_fail;
3116 err = register_netdevice_notifier(&ip_mr_notifier);
3118 goto reg_notif_fail;
3119 #ifdef CONFIG_IP_PIMSM_V2
3120 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3121 pr_err("%s: can't add PIM protocol\n", __func__);
3123 goto add_proto_fail;
3126 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3127 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3128 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3129 ipmr_rtm_route, NULL, 0);
3130 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3131 ipmr_rtm_route, NULL, 0);
3133 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3134 NULL, ipmr_rtm_dumplink, 0);
3137 #ifdef CONFIG_IP_PIMSM_V2
3139 unregister_netdevice_notifier(&ip_mr_notifier);
3142 unregister_pernet_subsys(&ipmr_net_ops);
3144 kmem_cache_destroy(mrt_cachep);