2 * IP multicast routing support for mrouted 3.6/3.8
4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5 * Linux Consultancy and Custom Driver Development
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
13 * Michael Chastain : Incorrect size of copying.
14 * Alan Cox : Added the cache manager code
15 * Alan Cox : Fixed the clone/copy bug and device race.
16 * Mike McLagan : Routing by source
17 * Malcolm Beattie : Buffer handling fixes.
18 * Alexey Kuznetsov : Double buffer free and other fixes.
19 * SVR Anand : Fixed several multicast bugs and problems.
20 * Alexey Kuznetsov : Status, optimisations and more.
21 * Brad Parker : Better behaviour on mrouted upcall
23 * Carlos Picoto : PIMv1 Support
24 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
25 * Relax this requirement to work with older peers.
29 #include <linux/uaccess.h>
30 #include <linux/types.h>
31 #include <linux/cache.h>
32 #include <linux/capability.h>
33 #include <linux/errno.h>
35 #include <linux/kernel.h>
36 #include <linux/fcntl.h>
37 #include <linux/stat.h>
38 #include <linux/socket.h>
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/inetdevice.h>
43 #include <linux/igmp.h>
44 #include <linux/proc_fs.h>
45 #include <linux/seq_file.h>
46 #include <linux/mroute.h>
47 #include <linux/init.h>
48 #include <linux/if_ether.h>
49 #include <linux/slab.h>
50 #include <net/net_namespace.h>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.h>
58 #include <linux/notifier.h>
59 #include <linux/if_arp.h>
60 #include <linux/netfilter_ipv4.h>
61 #include <linux/compat.h>
62 #include <linux/export.h>
63 #include <linux/rhashtable.h>
64 #include <net/ip_tunnels.h>
65 #include <net/checksum.h>
66 #include <net/netlink.h>
67 #include <net/fib_rules.h>
68 #include <linux/netconf.h>
69 #include <net/nexthop.h>
70 #include <net/switchdev.h>
72 #include <linux/nospec.h>
75 struct fib_rule common;
82 /* Big lock, protecting vif table, mrt cache and mroute socket state.
83 * Note that the changes are semaphored via rtnl_lock.
86 static DEFINE_RWLOCK(mrt_lock);
88 /* Multicast router control variables */
90 /* Special spinlock for queue of unresolved entries */
91 static DEFINE_SPINLOCK(mfc_unres_lock);
93 /* We return to original Alan's scheme. Hash table of resolved
94 * entries is changed only in process context and protected
95 * with weak lock mrt_lock. Queue of unresolved entries is protected
96 * with strong spinlock mfc_unres_lock.
98 * In this case data path is free of exclusive locks at all.
101 static struct kmem_cache *mrt_cachep __ro_after_init;
103 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
104 static void ipmr_free_table(struct mr_table *mrt);
106 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
107 struct net_device *dev, struct sk_buff *skb,
108 struct mfc_cache *cache, int local);
109 static int ipmr_cache_report(struct mr_table *mrt,
110 struct sk_buff *pkt, vifi_t vifi, int assert);
111 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
113 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
114 static void mroute_clean_tables(struct mr_table *mrt, bool all);
115 static void ipmr_expire_process(struct timer_list *t);
117 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
118 #define ipmr_for_each_table(mrt, net) \
119 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
121 static struct mr_table *ipmr_mr_table_iter(struct net *net,
122 struct mr_table *mrt)
124 struct mr_table *ret;
127 ret = list_entry_rcu(net->ipv4.mr_tables.next,
128 struct mr_table, list);
130 ret = list_entry_rcu(mrt->list.next,
131 struct mr_table, list);
133 if (&ret->list == &net->ipv4.mr_tables)
138 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
140 struct mr_table *mrt;
142 ipmr_for_each_table(mrt, net) {
149 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
150 struct mr_table **mrt)
153 struct ipmr_result res;
154 struct fib_lookup_arg arg = {
156 .flags = FIB_LOOKUP_NOREF,
159 /* update flow if oif or iif point to device enslaved to l3mdev */
160 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
162 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
163 flowi4_to_flowi(flp4), 0, &arg);
170 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
171 int flags, struct fib_lookup_arg *arg)
173 struct ipmr_result *res = arg->result;
174 struct mr_table *mrt;
176 switch (rule->action) {
179 case FR_ACT_UNREACHABLE:
181 case FR_ACT_PROHIBIT:
183 case FR_ACT_BLACKHOLE:
188 arg->table = fib_rule_get_table(rule, arg);
190 mrt = ipmr_get_table(rule->fr_net, arg->table);
197 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
202 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
206 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
207 struct fib_rule_hdr *frh, struct nlattr **tb,
208 struct netlink_ext_ack *extack)
213 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
219 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
220 struct fib_rule_hdr *frh)
228 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
229 .family = RTNL_FAMILY_IPMR,
230 .rule_size = sizeof(struct ipmr_rule),
231 .addr_size = sizeof(u32),
232 .action = ipmr_rule_action,
233 .match = ipmr_rule_match,
234 .configure = ipmr_rule_configure,
235 .compare = ipmr_rule_compare,
236 .fill = ipmr_rule_fill,
237 .nlgroup = RTNLGRP_IPV4_RULE,
238 .policy = ipmr_rule_policy,
239 .owner = THIS_MODULE,
242 static int __net_init ipmr_rules_init(struct net *net)
244 struct fib_rules_ops *ops;
245 struct mr_table *mrt;
248 ops = fib_rules_register(&ipmr_rules_ops_template, net);
252 INIT_LIST_HEAD(&net->ipv4.mr_tables);
254 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
260 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
264 net->ipv4.mr_rules_ops = ops;
269 ipmr_free_table(mrt);
272 fib_rules_unregister(ops);
276 static void __net_exit ipmr_rules_exit(struct net *net)
278 struct mr_table *mrt, *next;
281 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
282 list_del(&mrt->list);
283 ipmr_free_table(mrt);
285 fib_rules_unregister(net->ipv4.mr_rules_ops);
289 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
291 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR);
294 static unsigned int ipmr_rules_seq_read(struct net *net)
296 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
299 bool ipmr_rule_default(const struct fib_rule *rule)
301 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
303 EXPORT_SYMBOL(ipmr_rule_default);
305 #define ipmr_for_each_table(mrt, net) \
306 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
308 static struct mr_table *ipmr_mr_table_iter(struct net *net,
309 struct mr_table *mrt)
312 return net->ipv4.mrt;
316 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
318 return net->ipv4.mrt;
321 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
322 struct mr_table **mrt)
324 *mrt = net->ipv4.mrt;
328 static int __net_init ipmr_rules_init(struct net *net)
330 struct mr_table *mrt;
332 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
339 static void __net_exit ipmr_rules_exit(struct net *net)
342 ipmr_free_table(net->ipv4.mrt);
343 net->ipv4.mrt = NULL;
347 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
352 static unsigned int ipmr_rules_seq_read(struct net *net)
357 bool ipmr_rule_default(const struct fib_rule *rule)
361 EXPORT_SYMBOL(ipmr_rule_default);
364 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
367 const struct mfc_cache_cmp_arg *cmparg = arg->key;
368 struct mfc_cache *c = (struct mfc_cache *)ptr;
370 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
371 cmparg->mfc_origin != c->mfc_origin;
374 static const struct rhashtable_params ipmr_rht_params = {
375 .head_offset = offsetof(struct mr_mfc, mnode),
376 .key_offset = offsetof(struct mfc_cache, cmparg),
377 .key_len = sizeof(struct mfc_cache_cmp_arg),
380 .obj_cmpfn = ipmr_hash_cmp,
381 .automatic_shrinking = true,
384 static void ipmr_new_table_set(struct mr_table *mrt,
387 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
388 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
392 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
393 .mfc_mcastgrp = htonl(INADDR_ANY),
394 .mfc_origin = htonl(INADDR_ANY),
397 static struct mr_table_ops ipmr_mr_table_ops = {
398 .rht_params = &ipmr_rht_params,
399 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
402 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
404 struct mr_table *mrt;
406 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
407 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
408 return ERR_PTR(-EINVAL);
410 mrt = ipmr_get_table(net, id);
414 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
415 ipmr_expire_process, ipmr_new_table_set);
418 static void ipmr_free_table(struct mr_table *mrt)
420 del_timer_sync(&mrt->ipmr_expire_timer);
421 mroute_clean_tables(mrt, true);
422 rhltable_destroy(&mrt->mfc_hash);
426 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
428 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
430 struct net *net = dev_net(dev);
434 dev = __dev_get_by_name(net, "tunl0");
436 const struct net_device_ops *ops = dev->netdev_ops;
438 struct ip_tunnel_parm p;
440 memset(&p, 0, sizeof(p));
441 p.iph.daddr = v->vifc_rmt_addr.s_addr;
442 p.iph.saddr = v->vifc_lcl_addr.s_addr;
445 p.iph.protocol = IPPROTO_IPIP;
446 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
447 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
449 if (ops->ndo_do_ioctl) {
450 mm_segment_t oldfs = get_fs();
453 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
459 /* Initialize ipmr pimreg/tunnel in_device */
460 static bool ipmr_init_vif_indev(const struct net_device *dev)
462 struct in_device *in_dev;
466 in_dev = __in_dev_get_rtnl(dev);
469 ipv4_devconf_setall(in_dev);
470 neigh_parms_data_state_setall(in_dev->arp_parms);
471 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
476 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
478 struct net_device *dev;
480 dev = __dev_get_by_name(net, "tunl0");
483 const struct net_device_ops *ops = dev->netdev_ops;
486 struct ip_tunnel_parm p;
488 memset(&p, 0, sizeof(p));
489 p.iph.daddr = v->vifc_rmt_addr.s_addr;
490 p.iph.saddr = v->vifc_lcl_addr.s_addr;
493 p.iph.protocol = IPPROTO_IPIP;
494 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
495 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
497 if (ops->ndo_do_ioctl) {
498 mm_segment_t oldfs = get_fs();
501 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
509 (dev = __dev_get_by_name(net, p.name)) != NULL) {
510 dev->flags |= IFF_MULTICAST;
511 if (!ipmr_init_vif_indev(dev))
521 unregister_netdevice(dev);
525 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
526 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
528 struct net *net = dev_net(dev);
529 struct mr_table *mrt;
530 struct flowi4 fl4 = {
531 .flowi4_oif = dev->ifindex,
532 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
533 .flowi4_mark = skb->mark,
537 err = ipmr_fib_lookup(net, &fl4, &mrt);
543 read_lock(&mrt_lock);
544 dev->stats.tx_bytes += skb->len;
545 dev->stats.tx_packets++;
546 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
547 read_unlock(&mrt_lock);
552 static int reg_vif_get_iflink(const struct net_device *dev)
557 static const struct net_device_ops reg_vif_netdev_ops = {
558 .ndo_start_xmit = reg_vif_xmit,
559 .ndo_get_iflink = reg_vif_get_iflink,
562 static void reg_vif_setup(struct net_device *dev)
564 dev->type = ARPHRD_PIMREG;
565 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
566 dev->flags = IFF_NOARP;
567 dev->netdev_ops = ®_vif_netdev_ops;
568 dev->needs_free_netdev = true;
569 dev->features |= NETIF_F_NETNS_LOCAL;
572 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
574 struct net_device *dev;
577 if (mrt->id == RT_TABLE_DEFAULT)
578 sprintf(name, "pimreg");
580 sprintf(name, "pimreg%u", mrt->id);
582 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
587 dev_net_set(dev, net);
589 if (register_netdevice(dev)) {
594 if (!ipmr_init_vif_indev(dev))
604 unregister_netdevice(dev);
608 /* called with rcu_read_lock() */
609 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
612 struct net_device *reg_dev = NULL;
615 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
617 * a. packet is really sent to a multicast group
618 * b. packet is not a NULL-REGISTER
619 * c. packet is not truncated
621 if (!ipv4_is_multicast(encap->daddr) ||
622 encap->tot_len == 0 ||
623 ntohs(encap->tot_len) + pimlen > skb->len)
626 read_lock(&mrt_lock);
627 if (mrt->mroute_reg_vif_num >= 0)
628 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
629 read_unlock(&mrt_lock);
634 skb->mac_header = skb->network_header;
635 skb_pull(skb, (u8 *)encap - skb->data);
636 skb_reset_network_header(skb);
637 skb->protocol = htons(ETH_P_IP);
638 skb->ip_summed = CHECKSUM_NONE;
640 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
644 return NET_RX_SUCCESS;
647 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
653 static int call_ipmr_vif_entry_notifiers(struct net *net,
654 enum fib_event_type event_type,
655 struct vif_device *vif,
656 vifi_t vif_index, u32 tb_id)
658 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
659 vif, vif_index, tb_id,
660 &net->ipv4.ipmr_seq);
663 static int call_ipmr_mfc_entry_notifiers(struct net *net,
664 enum fib_event_type event_type,
665 struct mfc_cache *mfc, u32 tb_id)
667 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
668 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
672 * vif_delete - Delete a VIF entry
673 * @notify: Set to 1, if the caller is a notifier_call
675 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
676 struct list_head *head)
678 struct net *net = read_pnet(&mrt->net);
679 struct vif_device *v;
680 struct net_device *dev;
681 struct in_device *in_dev;
683 if (vifi < 0 || vifi >= mrt->maxvif)
684 return -EADDRNOTAVAIL;
686 v = &mrt->vif_table[vifi];
688 if (VIF_EXISTS(mrt, vifi))
689 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
692 write_lock_bh(&mrt_lock);
697 write_unlock_bh(&mrt_lock);
698 return -EADDRNOTAVAIL;
701 if (vifi == mrt->mroute_reg_vif_num)
702 mrt->mroute_reg_vif_num = -1;
704 if (vifi + 1 == mrt->maxvif) {
707 for (tmp = vifi - 1; tmp >= 0; tmp--) {
708 if (VIF_EXISTS(mrt, tmp))
714 write_unlock_bh(&mrt_lock);
716 dev_set_allmulti(dev, -1);
718 in_dev = __in_dev_get_rtnl(dev);
720 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
721 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
722 NETCONFA_MC_FORWARDING,
723 dev->ifindex, &in_dev->cnf);
724 ip_rt_multicast_event(in_dev);
727 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
728 unregister_netdevice_queue(dev, head);
734 static void ipmr_cache_free_rcu(struct rcu_head *head)
736 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
738 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
741 static void ipmr_cache_free(struct mfc_cache *c)
743 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
746 /* Destroy an unresolved cache entry, killing queued skbs
747 * and reporting error to netlink readers.
749 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
751 struct net *net = read_pnet(&mrt->net);
755 atomic_dec(&mrt->cache_resolve_queue_len);
757 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
758 if (ip_hdr(skb)->version == 0) {
759 struct nlmsghdr *nlh = skb_pull(skb,
760 sizeof(struct iphdr));
761 nlh->nlmsg_type = NLMSG_ERROR;
762 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
763 skb_trim(skb, nlh->nlmsg_len);
765 e->error = -ETIMEDOUT;
766 memset(&e->msg, 0, sizeof(e->msg));
768 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
777 /* Timer process for the unresolved queue. */
778 static void ipmr_expire_process(struct timer_list *t)
780 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
781 struct mr_mfc *c, *next;
782 unsigned long expires;
785 if (!spin_trylock(&mfc_unres_lock)) {
786 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
790 if (list_empty(&mrt->mfc_unres_queue))
796 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
797 if (time_after(c->mfc_un.unres.expires, now)) {
798 unsigned long interval = c->mfc_un.unres.expires - now;
799 if (interval < expires)
805 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
806 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
809 if (!list_empty(&mrt->mfc_unres_queue))
810 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
813 spin_unlock(&mfc_unres_lock);
816 /* Fill oifs list. It is called under write locked mrt_lock. */
817 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
822 cache->mfc_un.res.minvif = MAXVIFS;
823 cache->mfc_un.res.maxvif = 0;
824 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
826 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
827 if (VIF_EXISTS(mrt, vifi) &&
828 ttls[vifi] && ttls[vifi] < 255) {
829 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
830 if (cache->mfc_un.res.minvif > vifi)
831 cache->mfc_un.res.minvif = vifi;
832 if (cache->mfc_un.res.maxvif <= vifi)
833 cache->mfc_un.res.maxvif = vifi + 1;
836 cache->mfc_un.res.lastuse = jiffies;
839 static int vif_add(struct net *net, struct mr_table *mrt,
840 struct vifctl *vifc, int mrtsock)
842 int vifi = vifc->vifc_vifi;
843 struct switchdev_attr attr = {
844 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
846 struct vif_device *v = &mrt->vif_table[vifi];
847 struct net_device *dev;
848 struct in_device *in_dev;
852 if (VIF_EXISTS(mrt, vifi))
855 switch (vifc->vifc_flags) {
857 if (!ipmr_pimsm_enabled())
859 /* Special Purpose VIF in PIM
860 * All the packets will be sent to the daemon
862 if (mrt->mroute_reg_vif_num >= 0)
864 dev = ipmr_reg_vif(net, mrt);
867 err = dev_set_allmulti(dev, 1);
869 unregister_netdevice(dev);
875 dev = ipmr_new_tunnel(net, vifc);
878 err = dev_set_allmulti(dev, 1);
880 ipmr_del_tunnel(dev, vifc);
885 case VIFF_USE_IFINDEX:
887 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
888 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
889 if (dev && !__in_dev_get_rtnl(dev)) {
891 return -EADDRNOTAVAIL;
894 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
897 return -EADDRNOTAVAIL;
898 err = dev_set_allmulti(dev, 1);
908 in_dev = __in_dev_get_rtnl(dev);
911 return -EADDRNOTAVAIL;
913 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
914 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
915 dev->ifindex, &in_dev->cnf);
916 ip_rt_multicast_event(in_dev);
918 /* Fill in the VIF structures */
919 vif_device_init(v, dev, vifc->vifc_rate_limit,
920 vifc->vifc_threshold,
921 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
922 (VIFF_TUNNEL | VIFF_REGISTER));
925 if (!switchdev_port_attr_get(dev, &attr)) {
926 memcpy(v->dev_parent_id.id, attr.u.ppid.id, attr.u.ppid.id_len);
927 v->dev_parent_id.id_len = attr.u.ppid.id_len;
929 v->dev_parent_id.id_len = 0;
932 v->local = vifc->vifc_lcl_addr.s_addr;
933 v->remote = vifc->vifc_rmt_addr.s_addr;
935 /* And finish update writing critical data */
936 write_lock_bh(&mrt_lock);
938 if (v->flags & VIFF_REGISTER)
939 mrt->mroute_reg_vif_num = vifi;
940 if (vifi+1 > mrt->maxvif)
941 mrt->maxvif = vifi+1;
942 write_unlock_bh(&mrt_lock);
943 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
947 /* called with rcu_read_lock() */
948 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
952 struct mfc_cache_cmp_arg arg = {
953 .mfc_mcastgrp = mcastgrp,
957 return mr_mfc_find(mrt, &arg);
960 /* Look for a (*,G) entry */
961 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
962 __be32 mcastgrp, int vifi)
964 struct mfc_cache_cmp_arg arg = {
965 .mfc_mcastgrp = mcastgrp,
966 .mfc_origin = htonl(INADDR_ANY)
969 if (mcastgrp == htonl(INADDR_ANY))
970 return mr_mfc_find_any_parent(mrt, vifi);
971 return mr_mfc_find_any(mrt, vifi, &arg);
974 /* Look for a (S,G,iif) entry if parent != -1 */
975 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
976 __be32 origin, __be32 mcastgrp,
979 struct mfc_cache_cmp_arg arg = {
980 .mfc_mcastgrp = mcastgrp,
981 .mfc_origin = origin,
984 return mr_mfc_find_parent(mrt, &arg, parent);
987 /* Allocate a multicast cache entry */
988 static struct mfc_cache *ipmr_cache_alloc(void)
990 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
993 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
994 c->_c.mfc_un.res.minvif = MAXVIFS;
995 c->_c.free = ipmr_cache_free_rcu;
996 refcount_set(&c->_c.mfc_un.res.refcount, 1);
1001 static struct mfc_cache *ipmr_cache_alloc_unres(void)
1003 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1006 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1007 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1012 /* A cache entry has gone into a resolved state from queued */
1013 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1014 struct mfc_cache *uc, struct mfc_cache *c)
1016 struct sk_buff *skb;
1019 /* Play the pending entries through our router */
1020 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1021 if (ip_hdr(skb)->version == 0) {
1022 struct nlmsghdr *nlh = skb_pull(skb,
1023 sizeof(struct iphdr));
1025 if (mr_fill_mroute(mrt, skb, &c->_c,
1026 nlmsg_data(nlh)) > 0) {
1027 nlh->nlmsg_len = skb_tail_pointer(skb) -
1030 nlh->nlmsg_type = NLMSG_ERROR;
1031 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1032 skb_trim(skb, nlh->nlmsg_len);
1033 e = nlmsg_data(nlh);
1034 e->error = -EMSGSIZE;
1035 memset(&e->msg, 0, sizeof(e->msg));
1038 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1040 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1045 /* Bounce a cache query up to mrouted and netlink.
1047 * Called under mrt_lock.
1049 static int ipmr_cache_report(struct mr_table *mrt,
1050 struct sk_buff *pkt, vifi_t vifi, int assert)
1052 const int ihl = ip_hdrlen(pkt);
1053 struct sock *mroute_sk;
1054 struct igmphdr *igmp;
1055 struct igmpmsg *msg;
1056 struct sk_buff *skb;
1059 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1060 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1062 skb = alloc_skb(128, GFP_ATOMIC);
1067 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1068 /* Ugly, but we have no choice with this interface.
1069 * Duplicate old header, fix ihl, length etc.
1070 * And all this only to mangle msg->im_msgtype and
1071 * to set msg->im_mbz to "mbz" :-)
1073 skb_push(skb, sizeof(struct iphdr));
1074 skb_reset_network_header(skb);
1075 skb_reset_transport_header(skb);
1076 msg = (struct igmpmsg *)skb_network_header(skb);
1077 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1078 msg->im_msgtype = assert;
1080 if (assert == IGMPMSG_WRVIFWHOLE)
1083 msg->im_vif = mrt->mroute_reg_vif_num;
1084 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1085 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1086 sizeof(struct iphdr));
1088 /* Copy the IP header */
1089 skb_set_network_header(skb, skb->len);
1091 skb_copy_to_linear_data(skb, pkt->data, ihl);
1092 /* Flag to the kernel this is a route add */
1093 ip_hdr(skb)->protocol = 0;
1094 msg = (struct igmpmsg *)skb_network_header(skb);
1096 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1097 /* Add our header */
1098 igmp = skb_put(skb, sizeof(struct igmphdr));
1099 igmp->type = assert;
1100 msg->im_msgtype = assert;
1102 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1103 skb->transport_header = skb->network_header;
1107 mroute_sk = rcu_dereference(mrt->mroute_sk);
1114 igmpmsg_netlink_event(mrt, skb);
1116 /* Deliver to mrouted */
1117 ret = sock_queue_rcv_skb(mroute_sk, skb);
1120 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1127 /* Queue a packet for resolution. It gets locked cache entry! */
1128 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1129 struct sk_buff *skb, struct net_device *dev)
1131 const struct iphdr *iph = ip_hdr(skb);
1132 struct mfc_cache *c;
1136 spin_lock_bh(&mfc_unres_lock);
1137 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1138 if (c->mfc_mcastgrp == iph->daddr &&
1139 c->mfc_origin == iph->saddr) {
1146 /* Create a new entry if allowable */
1147 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1148 (c = ipmr_cache_alloc_unres()) == NULL) {
1149 spin_unlock_bh(&mfc_unres_lock);
1155 /* Fill in the new cache entry */
1156 c->_c.mfc_parent = -1;
1157 c->mfc_origin = iph->saddr;
1158 c->mfc_mcastgrp = iph->daddr;
1160 /* Reflect first query at mrouted. */
1161 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1164 /* If the report failed throw the cache entry
1167 spin_unlock_bh(&mfc_unres_lock);
1174 atomic_inc(&mrt->cache_resolve_queue_len);
1175 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1176 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1178 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1179 mod_timer(&mrt->ipmr_expire_timer,
1180 c->_c.mfc_un.unres.expires);
1183 /* See if we can append the packet */
1184 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1190 skb->skb_iif = dev->ifindex;
1192 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1196 spin_unlock_bh(&mfc_unres_lock);
1200 /* MFC cache manipulation by user space mroute daemon */
1202 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1204 struct net *net = read_pnet(&mrt->net);
1205 struct mfc_cache *c;
1207 /* The entries are added/deleted only under RTNL */
1209 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1210 mfc->mfcc_mcastgrp.s_addr, parent);
1214 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1215 list_del_rcu(&c->_c.list);
1216 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1217 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1218 mr_cache_put(&c->_c);
1223 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1224 struct mfcctl *mfc, int mrtsock, int parent)
1226 struct mfc_cache *uc, *c;
1231 if (mfc->mfcc_parent >= MAXVIFS)
1234 /* The entries are added/deleted only under RTNL */
1236 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1237 mfc->mfcc_mcastgrp.s_addr, parent);
1240 write_lock_bh(&mrt_lock);
1241 c->_c.mfc_parent = mfc->mfcc_parent;
1242 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1244 c->_c.mfc_flags |= MFC_STATIC;
1245 write_unlock_bh(&mrt_lock);
1246 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1248 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1252 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1253 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1256 c = ipmr_cache_alloc();
1260 c->mfc_origin = mfc->mfcc_origin.s_addr;
1261 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1262 c->_c.mfc_parent = mfc->mfcc_parent;
1263 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1265 c->_c.mfc_flags |= MFC_STATIC;
1267 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1270 pr_err("ipmr: rhtable insert error %d\n", ret);
1274 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1275 /* Check to see if we resolved a queued list. If so we
1276 * need to send on the frames and tidy up.
1279 spin_lock_bh(&mfc_unres_lock);
1280 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1281 uc = (struct mfc_cache *)_uc;
1282 if (uc->mfc_origin == c->mfc_origin &&
1283 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1284 list_del(&_uc->list);
1285 atomic_dec(&mrt->cache_resolve_queue_len);
1290 if (list_empty(&mrt->mfc_unres_queue))
1291 del_timer(&mrt->ipmr_expire_timer);
1292 spin_unlock_bh(&mfc_unres_lock);
1295 ipmr_cache_resolve(net, mrt, uc, c);
1296 ipmr_cache_free(uc);
1298 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1299 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1303 /* Close the multicast socket, and clear the vif tables etc */
1304 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1306 struct net *net = read_pnet(&mrt->net);
1307 struct mr_mfc *c, *tmp;
1308 struct mfc_cache *cache;
1312 /* Shut down all active vif entries */
1313 for (i = 0; i < mrt->maxvif; i++) {
1314 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1316 vif_delete(mrt, i, 0, &list);
1318 unregister_netdevice_many(&list);
1320 /* Wipe the cache */
1321 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1322 if (!all && (c->mfc_flags & MFC_STATIC))
1324 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1325 list_del_rcu(&c->list);
1326 cache = (struct mfc_cache *)c;
1327 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1329 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1333 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1334 spin_lock_bh(&mfc_unres_lock);
1335 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1337 cache = (struct mfc_cache *)c;
1338 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1339 ipmr_destroy_unres(mrt, cache);
1341 spin_unlock_bh(&mfc_unres_lock);
1345 /* called from ip_ra_control(), before an RCU grace period,
1346 * we dont need to call synchronize_rcu() here
1348 static void mrtsock_destruct(struct sock *sk)
1350 struct net *net = sock_net(sk);
1351 struct mr_table *mrt;
1354 ipmr_for_each_table(mrt, net) {
1355 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1356 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1357 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1358 NETCONFA_MC_FORWARDING,
1359 NETCONFA_IFINDEX_ALL,
1360 net->ipv4.devconf_all);
1361 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1362 mroute_clean_tables(mrt, false);
1368 /* Socket options and virtual interface manipulation. The whole
1369 * virtual interface system is a complete heap, but unfortunately
1370 * that's how BSD mrouted happens to think. Maybe one day with a proper
1371 * MOSPF/PIM router set up we can clean this up.
1374 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1375 unsigned int optlen)
1377 struct net *net = sock_net(sk);
1378 int val, ret = 0, parent = 0;
1379 struct mr_table *mrt;
1385 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1387 if (sk->sk_type != SOCK_RAW ||
1388 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1393 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1398 if (optname != MRT_INIT) {
1399 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1400 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1408 if (optlen != sizeof(int)) {
1412 if (rtnl_dereference(mrt->mroute_sk)) {
1417 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1419 rcu_assign_pointer(mrt->mroute_sk, sk);
1420 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1421 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1422 NETCONFA_MC_FORWARDING,
1423 NETCONFA_IFINDEX_ALL,
1424 net->ipv4.devconf_all);
1428 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1431 /* We need to unlock here because mrtsock_destruct takes
1432 * care of rtnl itself and we can't change that due to
1433 * the IP_ROUTER_ALERT setsockopt which runs without it.
1436 ret = ip_ra_control(sk, 0, NULL);
1442 if (optlen != sizeof(vif)) {
1446 if (copy_from_user(&vif, optval, sizeof(vif))) {
1450 if (vif.vifc_vifi >= MAXVIFS) {
1454 if (optname == MRT_ADD_VIF) {
1455 ret = vif_add(net, mrt, &vif,
1456 sk == rtnl_dereference(mrt->mroute_sk));
1458 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1461 /* Manipulate the forwarding caches. These live
1462 * in a sort of kernel/user symbiosis.
1468 case MRT_ADD_MFC_PROXY:
1469 case MRT_DEL_MFC_PROXY:
1470 if (optlen != sizeof(mfc)) {
1474 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1479 parent = mfc.mfcc_parent;
1480 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1481 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1483 ret = ipmr_mfc_add(net, mrt, &mfc,
1484 sk == rtnl_dereference(mrt->mroute_sk),
1487 /* Control PIM assert. */
1489 if (optlen != sizeof(val)) {
1493 if (get_user(val, (int __user *)optval)) {
1497 mrt->mroute_do_assert = val;
1500 if (!ipmr_pimsm_enabled()) {
1504 if (optlen != sizeof(val)) {
1508 if (get_user(val, (int __user *)optval)) {
1513 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1515 if (val != mrt->mroute_do_pim) {
1516 mrt->mroute_do_pim = val;
1517 mrt->mroute_do_assert = val;
1518 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1522 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1526 if (optlen != sizeof(uval)) {
1530 if (get_user(uval, (u32 __user *)optval)) {
1535 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1538 mrt = ipmr_new_table(net, uval);
1542 raw_sk(sk)->ipmr_table = uval;
1545 /* Spurious command, or MRT_VERSION which you cannot set. */
1555 /* Getsock opt support for the multicast routing system. */
1556 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1560 struct net *net = sock_net(sk);
1561 struct mr_table *mrt;
1563 if (sk->sk_type != SOCK_RAW ||
1564 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1567 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1576 if (!ipmr_pimsm_enabled())
1577 return -ENOPROTOOPT;
1578 val = mrt->mroute_do_pim;
1581 val = mrt->mroute_do_assert;
1584 return -ENOPROTOOPT;
1587 if (get_user(olr, optlen))
1589 olr = min_t(unsigned int, olr, sizeof(int));
1592 if (put_user(olr, optlen))
1594 if (copy_to_user(optval, &val, olr))
1599 /* The IP multicast ioctl support routines. */
1600 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1602 struct sioc_sg_req sr;
1603 struct sioc_vif_req vr;
1604 struct vif_device *vif;
1605 struct mfc_cache *c;
1606 struct net *net = sock_net(sk);
1607 struct mr_table *mrt;
1609 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1615 if (copy_from_user(&vr, arg, sizeof(vr)))
1617 if (vr.vifi >= mrt->maxvif)
1619 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1620 read_lock(&mrt_lock);
1621 vif = &mrt->vif_table[vr.vifi];
1622 if (VIF_EXISTS(mrt, vr.vifi)) {
1623 vr.icount = vif->pkt_in;
1624 vr.ocount = vif->pkt_out;
1625 vr.ibytes = vif->bytes_in;
1626 vr.obytes = vif->bytes_out;
1627 read_unlock(&mrt_lock);
1629 if (copy_to_user(arg, &vr, sizeof(vr)))
1633 read_unlock(&mrt_lock);
1634 return -EADDRNOTAVAIL;
1636 if (copy_from_user(&sr, arg, sizeof(sr)))
1640 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1642 sr.pktcnt = c->_c.mfc_un.res.pkt;
1643 sr.bytecnt = c->_c.mfc_un.res.bytes;
1644 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1647 if (copy_to_user(arg, &sr, sizeof(sr)))
1652 return -EADDRNOTAVAIL;
1654 return -ENOIOCTLCMD;
1658 #ifdef CONFIG_COMPAT
1659 struct compat_sioc_sg_req {
1662 compat_ulong_t pktcnt;
1663 compat_ulong_t bytecnt;
1664 compat_ulong_t wrong_if;
1667 struct compat_sioc_vif_req {
1668 vifi_t vifi; /* Which iface */
1669 compat_ulong_t icount;
1670 compat_ulong_t ocount;
1671 compat_ulong_t ibytes;
1672 compat_ulong_t obytes;
1675 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1677 struct compat_sioc_sg_req sr;
1678 struct compat_sioc_vif_req vr;
1679 struct vif_device *vif;
1680 struct mfc_cache *c;
1681 struct net *net = sock_net(sk);
1682 struct mr_table *mrt;
1684 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1690 if (copy_from_user(&vr, arg, sizeof(vr)))
1692 if (vr.vifi >= mrt->maxvif)
1694 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1695 read_lock(&mrt_lock);
1696 vif = &mrt->vif_table[vr.vifi];
1697 if (VIF_EXISTS(mrt, vr.vifi)) {
1698 vr.icount = vif->pkt_in;
1699 vr.ocount = vif->pkt_out;
1700 vr.ibytes = vif->bytes_in;
1701 vr.obytes = vif->bytes_out;
1702 read_unlock(&mrt_lock);
1704 if (copy_to_user(arg, &vr, sizeof(vr)))
1708 read_unlock(&mrt_lock);
1709 return -EADDRNOTAVAIL;
1711 if (copy_from_user(&sr, arg, sizeof(sr)))
1715 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1717 sr.pktcnt = c->_c.mfc_un.res.pkt;
1718 sr.bytecnt = c->_c.mfc_un.res.bytes;
1719 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1722 if (copy_to_user(arg, &sr, sizeof(sr)))
1727 return -EADDRNOTAVAIL;
1729 return -ENOIOCTLCMD;
1734 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1736 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1737 struct net *net = dev_net(dev);
1738 struct mr_table *mrt;
1739 struct vif_device *v;
1742 if (event != NETDEV_UNREGISTER)
1745 ipmr_for_each_table(mrt, net) {
1746 v = &mrt->vif_table[0];
1747 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1749 vif_delete(mrt, ct, 1, NULL);
1755 static struct notifier_block ip_mr_notifier = {
1756 .notifier_call = ipmr_device_event,
1759 /* Encapsulate a packet by attaching a valid IPIP header to it.
1760 * This avoids tunnel drivers and other mess and gives us the speed so
1761 * important for multicast video.
1763 static void ip_encap(struct net *net, struct sk_buff *skb,
1764 __be32 saddr, __be32 daddr)
1767 const struct iphdr *old_iph = ip_hdr(skb);
1769 skb_push(skb, sizeof(struct iphdr));
1770 skb->transport_header = skb->network_header;
1771 skb_reset_network_header(skb);
1775 iph->tos = old_iph->tos;
1776 iph->ttl = old_iph->ttl;
1780 iph->protocol = IPPROTO_IPIP;
1782 iph->tot_len = htons(skb->len);
1783 ip_select_ident(net, skb, NULL);
1786 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1790 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1791 struct sk_buff *skb)
1793 struct ip_options *opt = &(IPCB(skb)->opt);
1795 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1796 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1798 if (unlikely(opt->optlen))
1799 ip_forward_options(skb);
1801 return dst_output(net, sk, skb);
1804 #ifdef CONFIG_NET_SWITCHDEV
1805 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1806 int in_vifi, int out_vifi)
1808 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1809 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1811 if (!skb->offload_mr_fwd_mark)
1813 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1815 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1816 &in_vif->dev_parent_id);
1819 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1820 int in_vifi, int out_vifi)
1826 /* Processing handlers for ipmr_forward */
1828 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1829 int in_vifi, struct sk_buff *skb,
1830 struct mfc_cache *c, int vifi)
1832 const struct iphdr *iph = ip_hdr(skb);
1833 struct vif_device *vif = &mrt->vif_table[vifi];
1834 struct net_device *dev;
1842 if (vif->flags & VIFF_REGISTER) {
1844 vif->bytes_out += skb->len;
1845 vif->dev->stats.tx_bytes += skb->len;
1846 vif->dev->stats.tx_packets++;
1847 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1851 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1854 if (vif->flags & VIFF_TUNNEL) {
1855 rt = ip_route_output_ports(net, &fl4, NULL,
1856 vif->remote, vif->local,
1859 RT_TOS(iph->tos), vif->link);
1862 encap = sizeof(struct iphdr);
1864 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1867 RT_TOS(iph->tos), vif->link);
1874 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1875 /* Do not fragment multicasts. Alas, IPv4 does not
1876 * allow to send ICMP, so that packets will disappear
1879 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1884 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1886 if (skb_cow(skb, encap)) {
1892 vif->bytes_out += skb->len;
1895 skb_dst_set(skb, &rt->dst);
1896 ip_decrease_ttl(ip_hdr(skb));
1898 /* FIXME: forward and output firewalls used to be called here.
1899 * What do we do with netfilter? -- RR
1901 if (vif->flags & VIFF_TUNNEL) {
1902 ip_encap(net, skb, vif->local, vif->remote);
1903 /* FIXME: extra output firewall step used to be here. --RR */
1904 vif->dev->stats.tx_packets++;
1905 vif->dev->stats.tx_bytes += skb->len;
1908 IPCB(skb)->flags |= IPSKB_FORWARDED;
1910 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1911 * not only before forwarding, but after forwarding on all output
1912 * interfaces. It is clear, if mrouter runs a multicasting
1913 * program, it should receive packets not depending to what interface
1914 * program is joined.
1915 * If we will not make it, the program will have to join on all
1916 * interfaces. On the other hand, multihoming host (or router, but
1917 * not mrouter) cannot join to more than one interface - it will
1918 * result in receiving multiple packets.
1920 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1921 net, NULL, skb, skb->dev, dev,
1922 ipmr_forward_finish);
1929 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1933 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1934 if (mrt->vif_table[ct].dev == dev)
1940 /* "local" means that we should preserve one skb (for local delivery) */
1941 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1942 struct net_device *dev, struct sk_buff *skb,
1943 struct mfc_cache *c, int local)
1945 int true_vifi = ipmr_find_vif(mrt, dev);
1949 vif = c->_c.mfc_parent;
1950 c->_c.mfc_un.res.pkt++;
1951 c->_c.mfc_un.res.bytes += skb->len;
1952 c->_c.mfc_un.res.lastuse = jiffies;
1954 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1955 struct mfc_cache *cache_proxy;
1957 /* For an (*,G) entry, we only check that the incomming
1958 * interface is part of the static tree.
1960 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1962 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1966 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1967 if (mrt->vif_table[vif].dev != dev) {
1968 if (rt_is_output_route(skb_rtable(skb))) {
1969 /* It is our own packet, looped back.
1970 * Very complicated situation...
1972 * The best workaround until routing daemons will be
1973 * fixed is not to redistribute packet, if it was
1974 * send through wrong interface. It means, that
1975 * multicast applications WILL NOT work for
1976 * (S,G), which have default multicast route pointing
1977 * to wrong oif. In any case, it is not a good
1978 * idea to use multicasting applications on router.
1983 c->_c.mfc_un.res.wrong_if++;
1985 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1986 /* pimsm uses asserts, when switching from RPT to SPT,
1987 * so that we cannot check that packet arrived on an oif.
1988 * It is bad, but otherwise we would need to move pretty
1989 * large chunk of pimd to kernel. Ough... --ANK
1991 (mrt->mroute_do_pim ||
1992 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1994 c->_c.mfc_un.res.last_assert +
1995 MFC_ASSERT_THRESH)) {
1996 c->_c.mfc_un.res.last_assert = jiffies;
1997 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1998 if (mrt->mroute_do_wrvifwhole)
1999 ipmr_cache_report(mrt, skb, true_vifi,
2000 IGMPMSG_WRVIFWHOLE);
2006 mrt->vif_table[vif].pkt_in++;
2007 mrt->vif_table[vif].bytes_in += skb->len;
2009 /* Forward the frame */
2010 if (c->mfc_origin == htonl(INADDR_ANY) &&
2011 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2012 if (true_vifi >= 0 &&
2013 true_vifi != c->_c.mfc_parent &&
2015 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2016 /* It's an (*,*) entry and the packet is not coming from
2017 * the upstream: forward the packet to the upstream
2020 psend = c->_c.mfc_parent;
2025 for (ct = c->_c.mfc_un.res.maxvif - 1;
2026 ct >= c->_c.mfc_un.res.minvif; ct--) {
2027 /* For (*,G) entry, don't forward to the incoming interface */
2028 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2030 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2032 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2035 ipmr_queue_xmit(net, mrt, true_vifi,
2044 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2047 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2050 ipmr_queue_xmit(net, mrt, true_vifi, skb, c, psend);
2060 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2062 struct rtable *rt = skb_rtable(skb);
2063 struct iphdr *iph = ip_hdr(skb);
2064 struct flowi4 fl4 = {
2065 .daddr = iph->daddr,
2066 .saddr = iph->saddr,
2067 .flowi4_tos = RT_TOS(iph->tos),
2068 .flowi4_oif = (rt_is_output_route(rt) ?
2069 skb->dev->ifindex : 0),
2070 .flowi4_iif = (rt_is_output_route(rt) ?
2073 .flowi4_mark = skb->mark,
2075 struct mr_table *mrt;
2078 err = ipmr_fib_lookup(net, &fl4, &mrt);
2080 return ERR_PTR(err);
2084 /* Multicast packets for forwarding arrive here
2085 * Called with rcu_read_lock();
2087 int ip_mr_input(struct sk_buff *skb)
2089 struct mfc_cache *cache;
2090 struct net *net = dev_net(skb->dev);
2091 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2092 struct mr_table *mrt;
2093 struct net_device *dev;
2095 /* skb->dev passed in is the loX master dev for vrfs.
2096 * As there are no vifs associated with loopback devices,
2097 * get the proper interface that does have a vif associated with it.
2100 if (netif_is_l3_master(skb->dev)) {
2101 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2108 /* Packet is looped back after forward, it should not be
2109 * forwarded second time, but still can be delivered locally.
2111 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2114 mrt = ipmr_rt_fib_lookup(net, skb);
2117 return PTR_ERR(mrt);
2120 if (IPCB(skb)->opt.router_alert) {
2121 if (ip_call_ra_chain(skb))
2123 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2124 /* IGMPv1 (and broken IGMPv2 implementations sort of
2125 * Cisco IOS <= 11.2(8)) do not put router alert
2126 * option to IGMP packets destined to routable
2127 * groups. It is very bad, because it means
2128 * that we can forward NO IGMP messages.
2130 struct sock *mroute_sk;
2132 mroute_sk = rcu_dereference(mrt->mroute_sk);
2135 raw_rcv(mroute_sk, skb);
2141 /* already under rcu_read_lock() */
2142 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2144 int vif = ipmr_find_vif(mrt, dev);
2147 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2151 /* No usable cache entry */
2156 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2157 ip_local_deliver(skb);
2163 read_lock(&mrt_lock);
2164 vif = ipmr_find_vif(mrt, dev);
2166 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2167 read_unlock(&mrt_lock);
2171 read_unlock(&mrt_lock);
2176 read_lock(&mrt_lock);
2177 ip_mr_forward(net, mrt, dev, skb, cache, local);
2178 read_unlock(&mrt_lock);
2181 return ip_local_deliver(skb);
2187 return ip_local_deliver(skb);
2192 #ifdef CONFIG_IP_PIMSM_V1
2193 /* Handle IGMP messages of PIMv1 */
2194 int pim_rcv_v1(struct sk_buff *skb)
2196 struct igmphdr *pim;
2197 struct net *net = dev_net(skb->dev);
2198 struct mr_table *mrt;
2200 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2203 pim = igmp_hdr(skb);
2205 mrt = ipmr_rt_fib_lookup(net, skb);
2208 if (!mrt->mroute_do_pim ||
2209 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2212 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2220 #ifdef CONFIG_IP_PIMSM_V2
2221 static int pim_rcv(struct sk_buff *skb)
2223 struct pimreghdr *pim;
2224 struct net *net = dev_net(skb->dev);
2225 struct mr_table *mrt;
2227 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2230 pim = (struct pimreghdr *)skb_transport_header(skb);
2231 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2232 (pim->flags & PIM_NULL_REGISTER) ||
2233 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2234 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2237 mrt = ipmr_rt_fib_lookup(net, skb);
2240 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2248 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2249 __be32 saddr, __be32 daddr,
2250 struct rtmsg *rtm, u32 portid)
2252 struct mfc_cache *cache;
2253 struct mr_table *mrt;
2256 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2261 cache = ipmr_cache_find(mrt, saddr, daddr);
2262 if (!cache && skb->dev) {
2263 int vif = ipmr_find_vif(mrt, skb->dev);
2266 cache = ipmr_cache_find_any(mrt, daddr, vif);
2269 struct sk_buff *skb2;
2271 struct net_device *dev;
2275 read_lock(&mrt_lock);
2277 vif = ipmr_find_vif(mrt, dev);
2279 read_unlock(&mrt_lock);
2284 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2286 read_unlock(&mrt_lock);
2291 NETLINK_CB(skb2).portid = portid;
2292 skb_push(skb2, sizeof(struct iphdr));
2293 skb_reset_network_header(skb2);
2295 iph->ihl = sizeof(struct iphdr) >> 2;
2299 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2300 read_unlock(&mrt_lock);
2305 read_lock(&mrt_lock);
2306 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2307 read_unlock(&mrt_lock);
2312 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2313 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2316 struct nlmsghdr *nlh;
2320 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2324 rtm = nlmsg_data(nlh);
2325 rtm->rtm_family = RTNL_FAMILY_IPMR;
2326 rtm->rtm_dst_len = 32;
2327 rtm->rtm_src_len = 32;
2329 rtm->rtm_table = mrt->id;
2330 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2331 goto nla_put_failure;
2332 rtm->rtm_type = RTN_MULTICAST;
2333 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2334 if (c->_c.mfc_flags & MFC_STATIC)
2335 rtm->rtm_protocol = RTPROT_STATIC;
2337 rtm->rtm_protocol = RTPROT_MROUTED;
2340 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2341 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2342 goto nla_put_failure;
2343 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2344 /* do not break the dump if cache is unresolved */
2345 if (err < 0 && err != -ENOENT)
2346 goto nla_put_failure;
2348 nlmsg_end(skb, nlh);
2352 nlmsg_cancel(skb, nlh);
2356 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2357 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2360 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2364 static size_t mroute_msgsize(bool unresolved, int maxvif)
2367 NLMSG_ALIGN(sizeof(struct rtmsg))
2368 + nla_total_size(4) /* RTA_TABLE */
2369 + nla_total_size(4) /* RTA_SRC */
2370 + nla_total_size(4) /* RTA_DST */
2375 + nla_total_size(4) /* RTA_IIF */
2376 + nla_total_size(0) /* RTA_MULTIPATH */
2377 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2379 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2385 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2388 struct net *net = read_pnet(&mrt->net);
2389 struct sk_buff *skb;
2392 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2398 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2402 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2408 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2411 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2414 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2415 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2416 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2417 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2418 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2419 /* IPMRA_CREPORT_PKT */
2420 + nla_total_size(payloadlen)
2426 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2428 struct net *net = read_pnet(&mrt->net);
2429 struct nlmsghdr *nlh;
2430 struct rtgenmsg *rtgenm;
2431 struct igmpmsg *msg;
2432 struct sk_buff *skb;
2436 payloadlen = pkt->len - sizeof(struct igmpmsg);
2437 msg = (struct igmpmsg *)skb_network_header(pkt);
2439 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2443 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2444 sizeof(struct rtgenmsg), 0);
2447 rtgenm = nlmsg_data(nlh);
2448 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2449 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2450 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2451 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2452 msg->im_src.s_addr) ||
2453 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2454 msg->im_dst.s_addr))
2455 goto nla_put_failure;
2457 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2458 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2459 nla_data(nla), payloadlen))
2460 goto nla_put_failure;
2462 nlmsg_end(skb, nlh);
2464 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2468 nlmsg_cancel(skb, nlh);
2471 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2474 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2475 struct netlink_ext_ack *extack)
2477 struct net *net = sock_net(in_skb->sk);
2478 struct nlattr *tb[RTA_MAX + 1];
2479 struct sk_buff *skb = NULL;
2480 struct mfc_cache *cache;
2481 struct mr_table *mrt;
2487 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX,
2488 rtm_ipv4_policy, extack);
2492 rtm = nlmsg_data(nlh);
2494 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2495 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2496 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2498 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2504 /* entries are added/deleted only under RTNL */
2506 cache = ipmr_cache_find(mrt, src, grp);
2513 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2519 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2520 nlh->nlmsg_seq, cache,
2525 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2535 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2537 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2538 _ipmr_fill_mroute, &mfc_unres_lock);
2541 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2542 [RTA_SRC] = { .type = NLA_U32 },
2543 [RTA_DST] = { .type = NLA_U32 },
2544 [RTA_IIF] = { .type = NLA_U32 },
2545 [RTA_TABLE] = { .type = NLA_U32 },
2546 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2549 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2551 switch (rtm_protocol) {
2553 case RTPROT_MROUTED:
2559 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2561 struct rtnexthop *rtnh = nla_data(nla);
2562 int remaining = nla_len(nla), vifi = 0;
2564 while (rtnh_ok(rtnh, remaining)) {
2565 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2566 if (++vifi == MAXVIFS)
2568 rtnh = rtnh_next(rtnh, &remaining);
2571 return remaining > 0 ? -EINVAL : vifi;
2574 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2575 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2576 struct mfcctl *mfcc, int *mrtsock,
2577 struct mr_table **mrtret,
2578 struct netlink_ext_ack *extack)
2580 struct net_device *dev = NULL;
2581 u32 tblid = RT_TABLE_DEFAULT;
2582 struct mr_table *mrt;
2583 struct nlattr *attr;
2587 ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy,
2591 rtm = nlmsg_data(nlh);
2594 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2595 rtm->rtm_type != RTN_MULTICAST ||
2596 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2597 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2600 memset(mfcc, 0, sizeof(*mfcc));
2601 mfcc->mfcc_parent = -1;
2603 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2604 switch (nla_type(attr)) {
2606 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2609 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2612 dev = __dev_get_by_index(net, nla_get_u32(attr));
2619 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2628 tblid = nla_get_u32(attr);
2632 mrt = ipmr_get_table(net, tblid);
2638 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2640 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2646 /* takes care of both newroute and delroute */
2647 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2648 struct netlink_ext_ack *extack)
2650 struct net *net = sock_net(skb->sk);
2651 int ret, mrtsock, parent;
2652 struct mr_table *tbl;
2657 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2661 parent = ret ? mfcc.mfcc_parent : -1;
2662 if (nlh->nlmsg_type == RTM_NEWROUTE)
2663 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2665 return ipmr_mfc_delete(tbl, &mfcc, parent);
2668 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2670 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2672 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2673 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2674 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2675 mrt->mroute_reg_vif_num) ||
2676 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2677 mrt->mroute_do_assert) ||
2678 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2679 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2680 mrt->mroute_do_wrvifwhole))
2686 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2688 struct nlattr *vif_nest;
2689 struct vif_device *vif;
2691 /* if the VIF doesn't exist just continue */
2692 if (!VIF_EXISTS(mrt, vifid))
2695 vif = &mrt->vif_table[vifid];
2696 vif_nest = nla_nest_start(skb, IPMRA_VIF);
2699 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2700 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2701 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2702 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2704 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2706 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2708 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2710 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2711 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2712 nla_nest_cancel(skb, vif_nest);
2715 nla_nest_end(skb, vif_nest);
2720 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2722 struct net *net = sock_net(skb->sk);
2723 struct nlmsghdr *nlh = NULL;
2724 unsigned int t = 0, s_t;
2725 unsigned int e = 0, s_e;
2726 struct mr_table *mrt;
2731 ipmr_for_each_table(mrt, net) {
2732 struct nlattr *vifs, *af;
2733 struct ifinfomsg *hdr;
2738 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2739 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2740 sizeof(*hdr), NLM_F_MULTI);
2744 hdr = nlmsg_data(nlh);
2745 memset(hdr, 0, sizeof(*hdr));
2746 hdr->ifi_family = RTNL_FAMILY_IPMR;
2748 af = nla_nest_start(skb, IFLA_AF_SPEC);
2750 nlmsg_cancel(skb, nlh);
2754 if (!ipmr_fill_table(mrt, skb)) {
2755 nlmsg_cancel(skb, nlh);
2759 vifs = nla_nest_start(skb, IPMRA_TABLE_VIFS);
2761 nla_nest_end(skb, af);
2762 nlmsg_end(skb, nlh);
2765 for (i = 0; i < mrt->maxvif; i++) {
2768 if (!ipmr_fill_vif(mrt, i, skb)) {
2769 nla_nest_end(skb, vifs);
2770 nla_nest_end(skb, af);
2771 nlmsg_end(skb, nlh);
2779 nla_nest_end(skb, vifs);
2780 nla_nest_end(skb, af);
2781 nlmsg_end(skb, nlh);
2793 #ifdef CONFIG_PROC_FS
2794 /* The /proc interfaces to multicast routing :
2795 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2798 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2799 __acquires(mrt_lock)
2801 struct mr_vif_iter *iter = seq->private;
2802 struct net *net = seq_file_net(seq);
2803 struct mr_table *mrt;
2805 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2807 return ERR_PTR(-ENOENT);
2811 read_lock(&mrt_lock);
2812 return mr_vif_seq_start(seq, pos);
2815 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2816 __releases(mrt_lock)
2818 read_unlock(&mrt_lock);
2821 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2823 struct mr_vif_iter *iter = seq->private;
2824 struct mr_table *mrt = iter->mrt;
2826 if (v == SEQ_START_TOKEN) {
2828 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2830 const struct vif_device *vif = v;
2831 const char *name = vif->dev ?
2832 vif->dev->name : "none";
2835 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2836 vif - mrt->vif_table,
2837 name, vif->bytes_in, vif->pkt_in,
2838 vif->bytes_out, vif->pkt_out,
2839 vif->flags, vif->local, vif->remote);
2844 static const struct seq_operations ipmr_vif_seq_ops = {
2845 .start = ipmr_vif_seq_start,
2846 .next = mr_vif_seq_next,
2847 .stop = ipmr_vif_seq_stop,
2848 .show = ipmr_vif_seq_show,
2851 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2853 struct net *net = seq_file_net(seq);
2854 struct mr_table *mrt;
2856 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2858 return ERR_PTR(-ENOENT);
2860 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2863 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2867 if (v == SEQ_START_TOKEN) {
2869 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2871 const struct mfc_cache *mfc = v;
2872 const struct mr_mfc_iter *it = seq->private;
2873 const struct mr_table *mrt = it->mrt;
2875 seq_printf(seq, "%08X %08X %-3hd",
2876 (__force u32) mfc->mfc_mcastgrp,
2877 (__force u32) mfc->mfc_origin,
2878 mfc->_c.mfc_parent);
2880 if (it->cache != &mrt->mfc_unres_queue) {
2881 seq_printf(seq, " %8lu %8lu %8lu",
2882 mfc->_c.mfc_un.res.pkt,
2883 mfc->_c.mfc_un.res.bytes,
2884 mfc->_c.mfc_un.res.wrong_if);
2885 for (n = mfc->_c.mfc_un.res.minvif;
2886 n < mfc->_c.mfc_un.res.maxvif; n++) {
2887 if (VIF_EXISTS(mrt, n) &&
2888 mfc->_c.mfc_un.res.ttls[n] < 255)
2891 n, mfc->_c.mfc_un.res.ttls[n]);
2894 /* unresolved mfc_caches don't contain
2895 * pkt, bytes and wrong_if values
2897 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2899 seq_putc(seq, '\n');
2904 static const struct seq_operations ipmr_mfc_seq_ops = {
2905 .start = ipmr_mfc_seq_start,
2906 .next = mr_mfc_seq_next,
2907 .stop = mr_mfc_seq_stop,
2908 .show = ipmr_mfc_seq_show,
2912 #ifdef CONFIG_IP_PIMSM_V2
2913 static const struct net_protocol pim_protocol = {
2919 static unsigned int ipmr_seq_read(struct net *net)
2923 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
2926 static int ipmr_dump(struct net *net, struct notifier_block *nb)
2928 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
2929 ipmr_mr_table_iter, &mrt_lock);
2932 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
2933 .family = RTNL_FAMILY_IPMR,
2934 .fib_seq_read = ipmr_seq_read,
2935 .fib_dump = ipmr_dump,
2936 .owner = THIS_MODULE,
2939 static int __net_init ipmr_notifier_init(struct net *net)
2941 struct fib_notifier_ops *ops;
2943 net->ipv4.ipmr_seq = 0;
2945 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
2947 return PTR_ERR(ops);
2948 net->ipv4.ipmr_notifier_ops = ops;
2953 static void __net_exit ipmr_notifier_exit(struct net *net)
2955 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
2956 net->ipv4.ipmr_notifier_ops = NULL;
2959 /* Setup for IP multicast routing */
2960 static int __net_init ipmr_net_init(struct net *net)
2964 err = ipmr_notifier_init(net);
2966 goto ipmr_notifier_fail;
2968 err = ipmr_rules_init(net);
2970 goto ipmr_rules_fail;
2972 #ifdef CONFIG_PROC_FS
2974 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
2975 sizeof(struct mr_vif_iter)))
2977 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
2978 sizeof(struct mr_mfc_iter)))
2979 goto proc_cache_fail;
2983 #ifdef CONFIG_PROC_FS
2985 remove_proc_entry("ip_mr_vif", net->proc_net);
2987 ipmr_rules_exit(net);
2990 ipmr_notifier_exit(net);
2995 static void __net_exit ipmr_net_exit(struct net *net)
2997 #ifdef CONFIG_PROC_FS
2998 remove_proc_entry("ip_mr_cache", net->proc_net);
2999 remove_proc_entry("ip_mr_vif", net->proc_net);
3001 ipmr_notifier_exit(net);
3002 ipmr_rules_exit(net);
3005 static struct pernet_operations ipmr_net_ops = {
3006 .init = ipmr_net_init,
3007 .exit = ipmr_net_exit,
3010 int __init ip_mr_init(void)
3014 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3015 sizeof(struct mfc_cache),
3016 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3019 err = register_pernet_subsys(&ipmr_net_ops);
3021 goto reg_pernet_fail;
3023 err = register_netdevice_notifier(&ip_mr_notifier);
3025 goto reg_notif_fail;
3026 #ifdef CONFIG_IP_PIMSM_V2
3027 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3028 pr_err("%s: can't add PIM protocol\n", __func__);
3030 goto add_proto_fail;
3033 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3034 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3035 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3036 ipmr_rtm_route, NULL, 0);
3037 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3038 ipmr_rtm_route, NULL, 0);
3040 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3041 NULL, ipmr_rtm_dumplink, 0);
3044 #ifdef CONFIG_IP_PIMSM_V2
3046 unregister_netdevice_notifier(&ip_mr_notifier);
3049 unregister_pernet_subsys(&ipmr_net_ops);
3051 kmem_cache_destroy(mrt_cachep);