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/capability.h>
32 #include <linux/errno.h>
33 #include <linux/timer.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>
59 #include <linux/notifier.h>
60 #include <linux/if_arp.h>
61 #include <linux/netfilter_ipv4.h>
62 #include <linux/compat.h>
63 #include <linux/export.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>
72 struct fib_rule common
;
79 /* Big lock, protecting vif table, mrt cache and mroute socket state.
80 * Note that the changes are semaphored via rtnl_lock.
83 static DEFINE_RWLOCK(mrt_lock
);
85 /* Multicast router control variables */
87 /* Special spinlock for queue of unresolved entries */
88 static DEFINE_SPINLOCK(mfc_unres_lock
);
90 /* We return to original Alan's scheme. Hash table of resolved
91 * entries is changed only in process context and protected
92 * with weak lock mrt_lock. Queue of unresolved entries is protected
93 * with strong spinlock mfc_unres_lock.
95 * In this case data path is free of exclusive locks at all.
98 static struct kmem_cache
*mrt_cachep __read_mostly
;
100 static struct mr_table
*ipmr_new_table(struct net
*net
, u32 id
);
101 static void ipmr_free_table(struct mr_table
*mrt
);
103 static void ip_mr_forward(struct net
*net
, struct mr_table
*mrt
,
104 struct net_device
*dev
, struct sk_buff
*skb
,
105 struct mfc_cache
*cache
, int local
);
106 static int ipmr_cache_report(struct mr_table
*mrt
,
107 struct sk_buff
*pkt
, vifi_t vifi
, int assert);
108 static int __ipmr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
109 struct mfc_cache
*c
, struct rtmsg
*rtm
);
110 static void mroute_netlink_event(struct mr_table
*mrt
, struct mfc_cache
*mfc
,
112 static void igmpmsg_netlink_event(struct mr_table
*mrt
, struct sk_buff
*pkt
);
113 static void mroute_clean_tables(struct mr_table
*mrt
, bool all
);
114 static void ipmr_expire_process(unsigned long arg
);
116 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
117 #define ipmr_for_each_table(mrt, net) \
118 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
120 static struct mr_table
*ipmr_get_table(struct net
*net
, u32 id
)
122 struct mr_table
*mrt
;
124 ipmr_for_each_table(mrt
, net
) {
131 static int ipmr_fib_lookup(struct net
*net
, struct flowi4
*flp4
,
132 struct mr_table
**mrt
)
135 struct ipmr_result res
;
136 struct fib_lookup_arg arg
= {
138 .flags
= FIB_LOOKUP_NOREF
,
141 /* update flow if oif or iif point to device enslaved to l3mdev */
142 l3mdev_update_flow(net
, flowi4_to_flowi(flp4
));
144 err
= fib_rules_lookup(net
->ipv4
.mr_rules_ops
,
145 flowi4_to_flowi(flp4
), 0, &arg
);
152 static int ipmr_rule_action(struct fib_rule
*rule
, struct flowi
*flp
,
153 int flags
, struct fib_lookup_arg
*arg
)
155 struct ipmr_result
*res
= arg
->result
;
156 struct mr_table
*mrt
;
158 switch (rule
->action
) {
161 case FR_ACT_UNREACHABLE
:
163 case FR_ACT_PROHIBIT
:
165 case FR_ACT_BLACKHOLE
:
170 arg
->table
= fib_rule_get_table(rule
, arg
);
172 mrt
= ipmr_get_table(rule
->fr_net
, arg
->table
);
179 static int ipmr_rule_match(struct fib_rule
*rule
, struct flowi
*fl
, int flags
)
184 static const struct nla_policy ipmr_rule_policy
[FRA_MAX
+ 1] = {
188 static int ipmr_rule_configure(struct fib_rule
*rule
, struct sk_buff
*skb
,
189 struct fib_rule_hdr
*frh
, struct nlattr
**tb
)
194 static int ipmr_rule_compare(struct fib_rule
*rule
, struct fib_rule_hdr
*frh
,
200 static int ipmr_rule_fill(struct fib_rule
*rule
, struct sk_buff
*skb
,
201 struct fib_rule_hdr
*frh
)
209 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template
= {
210 .family
= RTNL_FAMILY_IPMR
,
211 .rule_size
= sizeof(struct ipmr_rule
),
212 .addr_size
= sizeof(u32
),
213 .action
= ipmr_rule_action
,
214 .match
= ipmr_rule_match
,
215 .configure
= ipmr_rule_configure
,
216 .compare
= ipmr_rule_compare
,
217 .fill
= ipmr_rule_fill
,
218 .nlgroup
= RTNLGRP_IPV4_RULE
,
219 .policy
= ipmr_rule_policy
,
220 .owner
= THIS_MODULE
,
223 static int __net_init
ipmr_rules_init(struct net
*net
)
225 struct fib_rules_ops
*ops
;
226 struct mr_table
*mrt
;
229 ops
= fib_rules_register(&ipmr_rules_ops_template
, net
);
233 INIT_LIST_HEAD(&net
->ipv4
.mr_tables
);
235 mrt
= ipmr_new_table(net
, RT_TABLE_DEFAULT
);
241 err
= fib_default_rule_add(ops
, 0x7fff, RT_TABLE_DEFAULT
, 0);
245 net
->ipv4
.mr_rules_ops
= ops
;
249 ipmr_free_table(mrt
);
251 fib_rules_unregister(ops
);
255 static void __net_exit
ipmr_rules_exit(struct net
*net
)
257 struct mr_table
*mrt
, *next
;
260 list_for_each_entry_safe(mrt
, next
, &net
->ipv4
.mr_tables
, list
) {
261 list_del(&mrt
->list
);
262 ipmr_free_table(mrt
);
264 fib_rules_unregister(net
->ipv4
.mr_rules_ops
);
268 #define ipmr_for_each_table(mrt, net) \
269 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
271 static struct mr_table
*ipmr_get_table(struct net
*net
, u32 id
)
273 return net
->ipv4
.mrt
;
276 static int ipmr_fib_lookup(struct net
*net
, struct flowi4
*flp4
,
277 struct mr_table
**mrt
)
279 *mrt
= net
->ipv4
.mrt
;
283 static int __net_init
ipmr_rules_init(struct net
*net
)
285 struct mr_table
*mrt
;
287 mrt
= ipmr_new_table(net
, RT_TABLE_DEFAULT
);
294 static void __net_exit
ipmr_rules_exit(struct net
*net
)
297 ipmr_free_table(net
->ipv4
.mrt
);
298 net
->ipv4
.mrt
= NULL
;
303 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg
*arg
,
306 const struct mfc_cache_cmp_arg
*cmparg
= arg
->key
;
307 struct mfc_cache
*c
= (struct mfc_cache
*)ptr
;
309 return cmparg
->mfc_mcastgrp
!= c
->mfc_mcastgrp
||
310 cmparg
->mfc_origin
!= c
->mfc_origin
;
313 static const struct rhashtable_params ipmr_rht_params
= {
314 .head_offset
= offsetof(struct mfc_cache
, mnode
),
315 .key_offset
= offsetof(struct mfc_cache
, cmparg
),
316 .key_len
= sizeof(struct mfc_cache_cmp_arg
),
319 .obj_cmpfn
= ipmr_hash_cmp
,
320 .automatic_shrinking
= true,
323 static struct mr_table
*ipmr_new_table(struct net
*net
, u32 id
)
325 struct mr_table
*mrt
;
327 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
328 if (id
!= RT_TABLE_DEFAULT
&& id
>= 1000000000)
329 return ERR_PTR(-EINVAL
);
331 mrt
= ipmr_get_table(net
, id
);
335 mrt
= kzalloc(sizeof(*mrt
), GFP_KERNEL
);
337 return ERR_PTR(-ENOMEM
);
338 write_pnet(&mrt
->net
, net
);
341 rhltable_init(&mrt
->mfc_hash
, &ipmr_rht_params
);
342 INIT_LIST_HEAD(&mrt
->mfc_cache_list
);
343 INIT_LIST_HEAD(&mrt
->mfc_unres_queue
);
345 setup_timer(&mrt
->ipmr_expire_timer
, ipmr_expire_process
,
348 mrt
->mroute_reg_vif_num
= -1;
349 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
350 list_add_tail_rcu(&mrt
->list
, &net
->ipv4
.mr_tables
);
355 static void ipmr_free_table(struct mr_table
*mrt
)
357 del_timer_sync(&mrt
->ipmr_expire_timer
);
358 mroute_clean_tables(mrt
, true);
359 rhltable_destroy(&mrt
->mfc_hash
);
363 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
365 static void ipmr_del_tunnel(struct net_device
*dev
, struct vifctl
*v
)
367 struct net
*net
= dev_net(dev
);
371 dev
= __dev_get_by_name(net
, "tunl0");
373 const struct net_device_ops
*ops
= dev
->netdev_ops
;
375 struct ip_tunnel_parm p
;
377 memset(&p
, 0, sizeof(p
));
378 p
.iph
.daddr
= v
->vifc_rmt_addr
.s_addr
;
379 p
.iph
.saddr
= v
->vifc_lcl_addr
.s_addr
;
382 p
.iph
.protocol
= IPPROTO_IPIP
;
383 sprintf(p
.name
, "dvmrp%d", v
->vifc_vifi
);
384 ifr
.ifr_ifru
.ifru_data
= (__force
void __user
*)&p
;
386 if (ops
->ndo_do_ioctl
) {
387 mm_segment_t oldfs
= get_fs();
390 ops
->ndo_do_ioctl(dev
, &ifr
, SIOCDELTUNNEL
);
396 /* Initialize ipmr pimreg/tunnel in_device */
397 static bool ipmr_init_vif_indev(const struct net_device
*dev
)
399 struct in_device
*in_dev
;
403 in_dev
= __in_dev_get_rtnl(dev
);
406 ipv4_devconf_setall(in_dev
);
407 neigh_parms_data_state_setall(in_dev
->arp_parms
);
408 IPV4_DEVCONF(in_dev
->cnf
, RP_FILTER
) = 0;
413 static struct net_device
*ipmr_new_tunnel(struct net
*net
, struct vifctl
*v
)
415 struct net_device
*dev
;
417 dev
= __dev_get_by_name(net
, "tunl0");
420 const struct net_device_ops
*ops
= dev
->netdev_ops
;
423 struct ip_tunnel_parm p
;
425 memset(&p
, 0, sizeof(p
));
426 p
.iph
.daddr
= v
->vifc_rmt_addr
.s_addr
;
427 p
.iph
.saddr
= v
->vifc_lcl_addr
.s_addr
;
430 p
.iph
.protocol
= IPPROTO_IPIP
;
431 sprintf(p
.name
, "dvmrp%d", v
->vifc_vifi
);
432 ifr
.ifr_ifru
.ifru_data
= (__force
void __user
*)&p
;
434 if (ops
->ndo_do_ioctl
) {
435 mm_segment_t oldfs
= get_fs();
438 err
= ops
->ndo_do_ioctl(dev
, &ifr
, SIOCADDTUNNEL
);
446 (dev
= __dev_get_by_name(net
, p
.name
)) != NULL
) {
447 dev
->flags
|= IFF_MULTICAST
;
448 if (!ipmr_init_vif_indev(dev
))
458 unregister_netdevice(dev
);
462 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
463 static netdev_tx_t
reg_vif_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
465 struct net
*net
= dev_net(dev
);
466 struct mr_table
*mrt
;
467 struct flowi4 fl4
= {
468 .flowi4_oif
= dev
->ifindex
,
469 .flowi4_iif
= skb
->skb_iif
? : LOOPBACK_IFINDEX
,
470 .flowi4_mark
= skb
->mark
,
474 err
= ipmr_fib_lookup(net
, &fl4
, &mrt
);
480 read_lock(&mrt_lock
);
481 dev
->stats
.tx_bytes
+= skb
->len
;
482 dev
->stats
.tx_packets
++;
483 ipmr_cache_report(mrt
, skb
, mrt
->mroute_reg_vif_num
, IGMPMSG_WHOLEPKT
);
484 read_unlock(&mrt_lock
);
489 static int reg_vif_get_iflink(const struct net_device
*dev
)
494 static const struct net_device_ops reg_vif_netdev_ops
= {
495 .ndo_start_xmit
= reg_vif_xmit
,
496 .ndo_get_iflink
= reg_vif_get_iflink
,
499 static void reg_vif_setup(struct net_device
*dev
)
501 dev
->type
= ARPHRD_PIMREG
;
502 dev
->mtu
= ETH_DATA_LEN
- sizeof(struct iphdr
) - 8;
503 dev
->flags
= IFF_NOARP
;
504 dev
->netdev_ops
= ®_vif_netdev_ops
;
505 dev
->needs_free_netdev
= true;
506 dev
->features
|= NETIF_F_NETNS_LOCAL
;
509 static struct net_device
*ipmr_reg_vif(struct net
*net
, struct mr_table
*mrt
)
511 struct net_device
*dev
;
514 if (mrt
->id
== RT_TABLE_DEFAULT
)
515 sprintf(name
, "pimreg");
517 sprintf(name
, "pimreg%u", mrt
->id
);
519 dev
= alloc_netdev(0, name
, NET_NAME_UNKNOWN
, reg_vif_setup
);
524 dev_net_set(dev
, net
);
526 if (register_netdevice(dev
)) {
531 if (!ipmr_init_vif_indev(dev
))
541 unregister_netdevice(dev
);
545 /* called with rcu_read_lock() */
546 static int __pim_rcv(struct mr_table
*mrt
, struct sk_buff
*skb
,
549 struct net_device
*reg_dev
= NULL
;
552 encap
= (struct iphdr
*)(skb_transport_header(skb
) + pimlen
);
554 * a. packet is really sent to a multicast group
555 * b. packet is not a NULL-REGISTER
556 * c. packet is not truncated
558 if (!ipv4_is_multicast(encap
->daddr
) ||
559 encap
->tot_len
== 0 ||
560 ntohs(encap
->tot_len
) + pimlen
> skb
->len
)
563 read_lock(&mrt_lock
);
564 if (mrt
->mroute_reg_vif_num
>= 0)
565 reg_dev
= mrt
->vif_table
[mrt
->mroute_reg_vif_num
].dev
;
566 read_unlock(&mrt_lock
);
571 skb
->mac_header
= skb
->network_header
;
572 skb_pull(skb
, (u8
*)encap
- skb
->data
);
573 skb_reset_network_header(skb
);
574 skb
->protocol
= htons(ETH_P_IP
);
575 skb
->ip_summed
= CHECKSUM_NONE
;
577 skb_tunnel_rx(skb
, reg_dev
, dev_net(reg_dev
));
581 return NET_RX_SUCCESS
;
584 static struct net_device
*ipmr_reg_vif(struct net
*net
, struct mr_table
*mrt
)
591 * vif_delete - Delete a VIF entry
592 * @notify: Set to 1, if the caller is a notifier_call
594 static int vif_delete(struct mr_table
*mrt
, int vifi
, int notify
,
595 struct list_head
*head
)
597 struct vif_device
*v
;
598 struct net_device
*dev
;
599 struct in_device
*in_dev
;
601 if (vifi
< 0 || vifi
>= mrt
->maxvif
)
602 return -EADDRNOTAVAIL
;
604 v
= &mrt
->vif_table
[vifi
];
606 write_lock_bh(&mrt_lock
);
611 write_unlock_bh(&mrt_lock
);
612 return -EADDRNOTAVAIL
;
615 if (vifi
== mrt
->mroute_reg_vif_num
)
616 mrt
->mroute_reg_vif_num
= -1;
618 if (vifi
+ 1 == mrt
->maxvif
) {
621 for (tmp
= vifi
- 1; tmp
>= 0; tmp
--) {
622 if (VIF_EXISTS(mrt
, tmp
))
628 write_unlock_bh(&mrt_lock
);
630 dev_set_allmulti(dev
, -1);
632 in_dev
= __in_dev_get_rtnl(dev
);
634 IPV4_DEVCONF(in_dev
->cnf
, MC_FORWARDING
)--;
635 inet_netconf_notify_devconf(dev_net(dev
), RTM_NEWNETCONF
,
636 NETCONFA_MC_FORWARDING
,
637 dev
->ifindex
, &in_dev
->cnf
);
638 ip_rt_multicast_event(in_dev
);
641 if (v
->flags
& (VIFF_TUNNEL
| VIFF_REGISTER
) && !notify
)
642 unregister_netdevice_queue(dev
, head
);
648 static void ipmr_cache_free_rcu(struct rcu_head
*head
)
650 struct mfc_cache
*c
= container_of(head
, struct mfc_cache
, rcu
);
652 kmem_cache_free(mrt_cachep
, c
);
655 static inline void ipmr_cache_free(struct mfc_cache
*c
)
657 call_rcu(&c
->rcu
, ipmr_cache_free_rcu
);
660 /* Destroy an unresolved cache entry, killing queued skbs
661 * and reporting error to netlink readers.
663 static void ipmr_destroy_unres(struct mr_table
*mrt
, struct mfc_cache
*c
)
665 struct net
*net
= read_pnet(&mrt
->net
);
669 atomic_dec(&mrt
->cache_resolve_queue_len
);
671 while ((skb
= skb_dequeue(&c
->mfc_un
.unres
.unresolved
))) {
672 if (ip_hdr(skb
)->version
== 0) {
673 struct nlmsghdr
*nlh
= skb_pull(skb
,
674 sizeof(struct iphdr
));
675 nlh
->nlmsg_type
= NLMSG_ERROR
;
676 nlh
->nlmsg_len
= nlmsg_msg_size(sizeof(struct nlmsgerr
));
677 skb_trim(skb
, nlh
->nlmsg_len
);
679 e
->error
= -ETIMEDOUT
;
680 memset(&e
->msg
, 0, sizeof(e
->msg
));
682 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).portid
);
691 /* Timer process for the unresolved queue. */
692 static void ipmr_expire_process(unsigned long arg
)
694 struct mr_table
*mrt
= (struct mr_table
*)arg
;
696 unsigned long expires
;
697 struct mfc_cache
*c
, *next
;
699 if (!spin_trylock(&mfc_unres_lock
)) {
700 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+HZ
/10);
704 if (list_empty(&mrt
->mfc_unres_queue
))
710 list_for_each_entry_safe(c
, next
, &mrt
->mfc_unres_queue
, list
) {
711 if (time_after(c
->mfc_un
.unres
.expires
, now
)) {
712 unsigned long interval
= c
->mfc_un
.unres
.expires
- now
;
713 if (interval
< expires
)
719 mroute_netlink_event(mrt
, c
, RTM_DELROUTE
);
720 ipmr_destroy_unres(mrt
, c
);
723 if (!list_empty(&mrt
->mfc_unres_queue
))
724 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+ expires
);
727 spin_unlock(&mfc_unres_lock
);
730 /* Fill oifs list. It is called under write locked mrt_lock. */
731 static void ipmr_update_thresholds(struct mr_table
*mrt
, struct mfc_cache
*cache
,
736 cache
->mfc_un
.res
.minvif
= MAXVIFS
;
737 cache
->mfc_un
.res
.maxvif
= 0;
738 memset(cache
->mfc_un
.res
.ttls
, 255, MAXVIFS
);
740 for (vifi
= 0; vifi
< mrt
->maxvif
; vifi
++) {
741 if (VIF_EXISTS(mrt
, vifi
) &&
742 ttls
[vifi
] && ttls
[vifi
] < 255) {
743 cache
->mfc_un
.res
.ttls
[vifi
] = ttls
[vifi
];
744 if (cache
->mfc_un
.res
.minvif
> vifi
)
745 cache
->mfc_un
.res
.minvif
= vifi
;
746 if (cache
->mfc_un
.res
.maxvif
<= vifi
)
747 cache
->mfc_un
.res
.maxvif
= vifi
+ 1;
750 cache
->mfc_un
.res
.lastuse
= jiffies
;
753 static int vif_add(struct net
*net
, struct mr_table
*mrt
,
754 struct vifctl
*vifc
, int mrtsock
)
756 int vifi
= vifc
->vifc_vifi
;
757 struct vif_device
*v
= &mrt
->vif_table
[vifi
];
758 struct net_device
*dev
;
759 struct in_device
*in_dev
;
763 if (VIF_EXISTS(mrt
, vifi
))
766 switch (vifc
->vifc_flags
) {
768 if (!ipmr_pimsm_enabled())
770 /* Special Purpose VIF in PIM
771 * All the packets will be sent to the daemon
773 if (mrt
->mroute_reg_vif_num
>= 0)
775 dev
= ipmr_reg_vif(net
, mrt
);
778 err
= dev_set_allmulti(dev
, 1);
780 unregister_netdevice(dev
);
786 dev
= ipmr_new_tunnel(net
, vifc
);
789 err
= dev_set_allmulti(dev
, 1);
791 ipmr_del_tunnel(dev
, vifc
);
796 case VIFF_USE_IFINDEX
:
798 if (vifc
->vifc_flags
== VIFF_USE_IFINDEX
) {
799 dev
= dev_get_by_index(net
, vifc
->vifc_lcl_ifindex
);
800 if (dev
&& !__in_dev_get_rtnl(dev
)) {
802 return -EADDRNOTAVAIL
;
805 dev
= ip_dev_find(net
, vifc
->vifc_lcl_addr
.s_addr
);
808 return -EADDRNOTAVAIL
;
809 err
= dev_set_allmulti(dev
, 1);
819 in_dev
= __in_dev_get_rtnl(dev
);
822 return -EADDRNOTAVAIL
;
824 IPV4_DEVCONF(in_dev
->cnf
, MC_FORWARDING
)++;
825 inet_netconf_notify_devconf(net
, RTM_NEWNETCONF
, NETCONFA_MC_FORWARDING
,
826 dev
->ifindex
, &in_dev
->cnf
);
827 ip_rt_multicast_event(in_dev
);
829 /* Fill in the VIF structures */
831 v
->rate_limit
= vifc
->vifc_rate_limit
;
832 v
->local
= vifc
->vifc_lcl_addr
.s_addr
;
833 v
->remote
= vifc
->vifc_rmt_addr
.s_addr
;
834 v
->flags
= vifc
->vifc_flags
;
836 v
->flags
|= VIFF_STATIC
;
837 v
->threshold
= vifc
->vifc_threshold
;
842 v
->link
= dev
->ifindex
;
843 if (v
->flags
& (VIFF_TUNNEL
| VIFF_REGISTER
))
844 v
->link
= dev_get_iflink(dev
);
846 /* And finish update writing critical data */
847 write_lock_bh(&mrt_lock
);
849 if (v
->flags
& VIFF_REGISTER
)
850 mrt
->mroute_reg_vif_num
= vifi
;
851 if (vifi
+1 > mrt
->maxvif
)
852 mrt
->maxvif
= vifi
+1;
853 write_unlock_bh(&mrt_lock
);
857 /* called with rcu_read_lock() */
858 static struct mfc_cache
*ipmr_cache_find(struct mr_table
*mrt
,
862 struct mfc_cache_cmp_arg arg
= {
863 .mfc_mcastgrp
= mcastgrp
,
866 struct rhlist_head
*tmp
, *list
;
869 list
= rhltable_lookup(&mrt
->mfc_hash
, &arg
, ipmr_rht_params
);
870 rhl_for_each_entry_rcu(c
, tmp
, list
, mnode
)
876 /* Look for a (*,*,oif) entry */
877 static struct mfc_cache
*ipmr_cache_find_any_parent(struct mr_table
*mrt
,
880 struct mfc_cache_cmp_arg arg
= {
881 .mfc_mcastgrp
= htonl(INADDR_ANY
),
882 .mfc_origin
= htonl(INADDR_ANY
)
884 struct rhlist_head
*tmp
, *list
;
887 list
= rhltable_lookup(&mrt
->mfc_hash
, &arg
, ipmr_rht_params
);
888 rhl_for_each_entry_rcu(c
, tmp
, list
, mnode
)
889 if (c
->mfc_un
.res
.ttls
[vifi
] < 255)
895 /* Look for a (*,G) entry */
896 static struct mfc_cache
*ipmr_cache_find_any(struct mr_table
*mrt
,
897 __be32 mcastgrp
, int vifi
)
899 struct mfc_cache_cmp_arg arg
= {
900 .mfc_mcastgrp
= mcastgrp
,
901 .mfc_origin
= htonl(INADDR_ANY
)
903 struct rhlist_head
*tmp
, *list
;
904 struct mfc_cache
*c
, *proxy
;
906 if (mcastgrp
== htonl(INADDR_ANY
))
909 list
= rhltable_lookup(&mrt
->mfc_hash
, &arg
, ipmr_rht_params
);
910 rhl_for_each_entry_rcu(c
, tmp
, list
, mnode
) {
911 if (c
->mfc_un
.res
.ttls
[vifi
] < 255)
914 /* It's ok if the vifi is part of the static tree */
915 proxy
= ipmr_cache_find_any_parent(mrt
, c
->mfc_parent
);
916 if (proxy
&& proxy
->mfc_un
.res
.ttls
[vifi
] < 255)
921 return ipmr_cache_find_any_parent(mrt
, vifi
);
924 /* Look for a (S,G,iif) entry if parent != -1 */
925 static struct mfc_cache
*ipmr_cache_find_parent(struct mr_table
*mrt
,
926 __be32 origin
, __be32 mcastgrp
,
929 struct mfc_cache_cmp_arg arg
= {
930 .mfc_mcastgrp
= mcastgrp
,
931 .mfc_origin
= origin
,
933 struct rhlist_head
*tmp
, *list
;
936 list
= rhltable_lookup(&mrt
->mfc_hash
, &arg
, ipmr_rht_params
);
937 rhl_for_each_entry_rcu(c
, tmp
, list
, mnode
)
938 if (parent
== -1 || parent
== c
->mfc_parent
)
944 /* Allocate a multicast cache entry */
945 static struct mfc_cache
*ipmr_cache_alloc(void)
947 struct mfc_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_KERNEL
);
950 c
->mfc_un
.res
.last_assert
= jiffies
- MFC_ASSERT_THRESH
- 1;
951 c
->mfc_un
.res
.minvif
= MAXVIFS
;
956 static struct mfc_cache
*ipmr_cache_alloc_unres(void)
958 struct mfc_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_ATOMIC
);
961 skb_queue_head_init(&c
->mfc_un
.unres
.unresolved
);
962 c
->mfc_un
.unres
.expires
= jiffies
+ 10*HZ
;
967 /* A cache entry has gone into a resolved state from queued */
968 static void ipmr_cache_resolve(struct net
*net
, struct mr_table
*mrt
,
969 struct mfc_cache
*uc
, struct mfc_cache
*c
)
974 /* Play the pending entries through our router */
975 while ((skb
= __skb_dequeue(&uc
->mfc_un
.unres
.unresolved
))) {
976 if (ip_hdr(skb
)->version
== 0) {
977 struct nlmsghdr
*nlh
= skb_pull(skb
,
978 sizeof(struct iphdr
));
980 if (__ipmr_fill_mroute(mrt
, skb
, c
, nlmsg_data(nlh
)) > 0) {
981 nlh
->nlmsg_len
= skb_tail_pointer(skb
) -
984 nlh
->nlmsg_type
= NLMSG_ERROR
;
985 nlh
->nlmsg_len
= nlmsg_msg_size(sizeof(struct nlmsgerr
));
986 skb_trim(skb
, nlh
->nlmsg_len
);
988 e
->error
= -EMSGSIZE
;
989 memset(&e
->msg
, 0, sizeof(e
->msg
));
992 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).portid
);
994 ip_mr_forward(net
, mrt
, skb
->dev
, skb
, c
, 0);
999 /* Bounce a cache query up to mrouted and netlink.
1001 * Called under mrt_lock.
1003 static int ipmr_cache_report(struct mr_table
*mrt
,
1004 struct sk_buff
*pkt
, vifi_t vifi
, int assert)
1006 const int ihl
= ip_hdrlen(pkt
);
1007 struct sock
*mroute_sk
;
1008 struct igmphdr
*igmp
;
1009 struct igmpmsg
*msg
;
1010 struct sk_buff
*skb
;
1013 if (assert == IGMPMSG_WHOLEPKT
)
1014 skb
= skb_realloc_headroom(pkt
, sizeof(struct iphdr
));
1016 skb
= alloc_skb(128, GFP_ATOMIC
);
1021 if (assert == IGMPMSG_WHOLEPKT
) {
1022 /* Ugly, but we have no choice with this interface.
1023 * Duplicate old header, fix ihl, length etc.
1024 * And all this only to mangle msg->im_msgtype and
1025 * to set msg->im_mbz to "mbz" :-)
1027 skb_push(skb
, sizeof(struct iphdr
));
1028 skb_reset_network_header(skb
);
1029 skb_reset_transport_header(skb
);
1030 msg
= (struct igmpmsg
*)skb_network_header(skb
);
1031 memcpy(msg
, skb_network_header(pkt
), sizeof(struct iphdr
));
1032 msg
->im_msgtype
= IGMPMSG_WHOLEPKT
;
1034 msg
->im_vif
= mrt
->mroute_reg_vif_num
;
1035 ip_hdr(skb
)->ihl
= sizeof(struct iphdr
) >> 2;
1036 ip_hdr(skb
)->tot_len
= htons(ntohs(ip_hdr(pkt
)->tot_len
) +
1037 sizeof(struct iphdr
));
1039 /* Copy the IP header */
1040 skb_set_network_header(skb
, skb
->len
);
1042 skb_copy_to_linear_data(skb
, pkt
->data
, ihl
);
1043 /* Flag to the kernel this is a route add */
1044 ip_hdr(skb
)->protocol
= 0;
1045 msg
= (struct igmpmsg
*)skb_network_header(skb
);
1047 skb_dst_set(skb
, dst_clone(skb_dst(pkt
)));
1048 /* Add our header */
1049 igmp
= skb_put(skb
, sizeof(struct igmphdr
));
1050 igmp
->type
= assert;
1051 msg
->im_msgtype
= assert;
1053 ip_hdr(skb
)->tot_len
= htons(skb
->len
); /* Fix the length */
1054 skb
->transport_header
= skb
->network_header
;
1058 mroute_sk
= rcu_dereference(mrt
->mroute_sk
);
1065 igmpmsg_netlink_event(mrt
, skb
);
1067 /* Deliver to mrouted */
1068 ret
= sock_queue_rcv_skb(mroute_sk
, skb
);
1071 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1078 /* Queue a packet for resolution. It gets locked cache entry! */
1079 static int ipmr_cache_unresolved(struct mr_table
*mrt
, vifi_t vifi
,
1080 struct sk_buff
*skb
, struct net_device
*dev
)
1082 const struct iphdr
*iph
= ip_hdr(skb
);
1083 struct mfc_cache
*c
;
1087 spin_lock_bh(&mfc_unres_lock
);
1088 list_for_each_entry(c
, &mrt
->mfc_unres_queue
, list
) {
1089 if (c
->mfc_mcastgrp
== iph
->daddr
&&
1090 c
->mfc_origin
== iph
->saddr
) {
1097 /* Create a new entry if allowable */
1098 if (atomic_read(&mrt
->cache_resolve_queue_len
) >= 10 ||
1099 (c
= ipmr_cache_alloc_unres()) == NULL
) {
1100 spin_unlock_bh(&mfc_unres_lock
);
1106 /* Fill in the new cache entry */
1108 c
->mfc_origin
= iph
->saddr
;
1109 c
->mfc_mcastgrp
= iph
->daddr
;
1111 /* Reflect first query at mrouted. */
1112 err
= ipmr_cache_report(mrt
, skb
, vifi
, IGMPMSG_NOCACHE
);
1114 /* If the report failed throw the cache entry
1117 spin_unlock_bh(&mfc_unres_lock
);
1124 atomic_inc(&mrt
->cache_resolve_queue_len
);
1125 list_add(&c
->list
, &mrt
->mfc_unres_queue
);
1126 mroute_netlink_event(mrt
, c
, RTM_NEWROUTE
);
1128 if (atomic_read(&mrt
->cache_resolve_queue_len
) == 1)
1129 mod_timer(&mrt
->ipmr_expire_timer
, c
->mfc_un
.unres
.expires
);
1132 /* See if we can append the packet */
1133 if (c
->mfc_un
.unres
.unresolved
.qlen
> 3) {
1139 skb
->skb_iif
= dev
->ifindex
;
1141 skb_queue_tail(&c
->mfc_un
.unres
.unresolved
, skb
);
1145 spin_unlock_bh(&mfc_unres_lock
);
1149 /* MFC cache manipulation by user space mroute daemon */
1151 static int ipmr_mfc_delete(struct mr_table
*mrt
, struct mfcctl
*mfc
, int parent
)
1153 struct mfc_cache
*c
;
1155 /* The entries are added/deleted only under RTNL */
1157 c
= ipmr_cache_find_parent(mrt
, mfc
->mfcc_origin
.s_addr
,
1158 mfc
->mfcc_mcastgrp
.s_addr
, parent
);
1162 rhltable_remove(&mrt
->mfc_hash
, &c
->mnode
, ipmr_rht_params
);
1163 list_del_rcu(&c
->list
);
1164 mroute_netlink_event(mrt
, c
, RTM_DELROUTE
);
1170 static int ipmr_mfc_add(struct net
*net
, struct mr_table
*mrt
,
1171 struct mfcctl
*mfc
, int mrtsock
, int parent
)
1173 struct mfc_cache
*uc
, *c
;
1177 if (mfc
->mfcc_parent
>= MAXVIFS
)
1180 /* The entries are added/deleted only under RTNL */
1182 c
= ipmr_cache_find_parent(mrt
, mfc
->mfcc_origin
.s_addr
,
1183 mfc
->mfcc_mcastgrp
.s_addr
, parent
);
1186 write_lock_bh(&mrt_lock
);
1187 c
->mfc_parent
= mfc
->mfcc_parent
;
1188 ipmr_update_thresholds(mrt
, c
, mfc
->mfcc_ttls
);
1190 c
->mfc_flags
|= MFC_STATIC
;
1191 write_unlock_bh(&mrt_lock
);
1192 mroute_netlink_event(mrt
, c
, RTM_NEWROUTE
);
1196 if (mfc
->mfcc_mcastgrp
.s_addr
!= htonl(INADDR_ANY
) &&
1197 !ipv4_is_multicast(mfc
->mfcc_mcastgrp
.s_addr
))
1200 c
= ipmr_cache_alloc();
1204 c
->mfc_origin
= mfc
->mfcc_origin
.s_addr
;
1205 c
->mfc_mcastgrp
= mfc
->mfcc_mcastgrp
.s_addr
;
1206 c
->mfc_parent
= mfc
->mfcc_parent
;
1207 ipmr_update_thresholds(mrt
, c
, mfc
->mfcc_ttls
);
1209 c
->mfc_flags
|= MFC_STATIC
;
1211 ret
= rhltable_insert_key(&mrt
->mfc_hash
, &c
->cmparg
, &c
->mnode
,
1214 pr_err("ipmr: rhtable insert error %d\n", ret
);
1218 list_add_tail_rcu(&c
->list
, &mrt
->mfc_cache_list
);
1219 /* Check to see if we resolved a queued list. If so we
1220 * need to send on the frames and tidy up.
1223 spin_lock_bh(&mfc_unres_lock
);
1224 list_for_each_entry(uc
, &mrt
->mfc_unres_queue
, list
) {
1225 if (uc
->mfc_origin
== c
->mfc_origin
&&
1226 uc
->mfc_mcastgrp
== c
->mfc_mcastgrp
) {
1227 list_del(&uc
->list
);
1228 atomic_dec(&mrt
->cache_resolve_queue_len
);
1233 if (list_empty(&mrt
->mfc_unres_queue
))
1234 del_timer(&mrt
->ipmr_expire_timer
);
1235 spin_unlock_bh(&mfc_unres_lock
);
1238 ipmr_cache_resolve(net
, mrt
, uc
, c
);
1239 ipmr_cache_free(uc
);
1241 mroute_netlink_event(mrt
, c
, RTM_NEWROUTE
);
1245 /* Close the multicast socket, and clear the vif tables etc */
1246 static void mroute_clean_tables(struct mr_table
*mrt
, bool all
)
1248 struct mfc_cache
*c
, *tmp
;
1252 /* Shut down all active vif entries */
1253 for (i
= 0; i
< mrt
->maxvif
; i
++) {
1254 if (!all
&& (mrt
->vif_table
[i
].flags
& VIFF_STATIC
))
1256 vif_delete(mrt
, i
, 0, &list
);
1258 unregister_netdevice_many(&list
);
1260 /* Wipe the cache */
1261 list_for_each_entry_safe(c
, tmp
, &mrt
->mfc_cache_list
, list
) {
1262 if (!all
&& (c
->mfc_flags
& MFC_STATIC
))
1264 rhltable_remove(&mrt
->mfc_hash
, &c
->mnode
, ipmr_rht_params
);
1265 list_del_rcu(&c
->list
);
1266 mroute_netlink_event(mrt
, c
, RTM_DELROUTE
);
1270 if (atomic_read(&mrt
->cache_resolve_queue_len
) != 0) {
1271 spin_lock_bh(&mfc_unres_lock
);
1272 list_for_each_entry_safe(c
, tmp
, &mrt
->mfc_unres_queue
, list
) {
1274 mroute_netlink_event(mrt
, c
, RTM_DELROUTE
);
1275 ipmr_destroy_unres(mrt
, c
);
1277 spin_unlock_bh(&mfc_unres_lock
);
1281 /* called from ip_ra_control(), before an RCU grace period,
1282 * we dont need to call synchronize_rcu() here
1284 static void mrtsock_destruct(struct sock
*sk
)
1286 struct net
*net
= sock_net(sk
);
1287 struct mr_table
*mrt
;
1290 ipmr_for_each_table(mrt
, net
) {
1291 if (sk
== rtnl_dereference(mrt
->mroute_sk
)) {
1292 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)--;
1293 inet_netconf_notify_devconf(net
, RTM_NEWNETCONF
,
1294 NETCONFA_MC_FORWARDING
,
1295 NETCONFA_IFINDEX_ALL
,
1296 net
->ipv4
.devconf_all
);
1297 RCU_INIT_POINTER(mrt
->mroute_sk
, NULL
);
1298 mroute_clean_tables(mrt
, false);
1303 /* Socket options and virtual interface manipulation. The whole
1304 * virtual interface system is a complete heap, but unfortunately
1305 * that's how BSD mrouted happens to think. Maybe one day with a proper
1306 * MOSPF/PIM router set up we can clean this up.
1309 int ip_mroute_setsockopt(struct sock
*sk
, int optname
, char __user
*optval
,
1310 unsigned int optlen
)
1312 struct net
*net
= sock_net(sk
);
1313 int val
, ret
= 0, parent
= 0;
1314 struct mr_table
*mrt
;
1319 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1321 if (sk
->sk_type
!= SOCK_RAW
||
1322 inet_sk(sk
)->inet_num
!= IPPROTO_IGMP
) {
1327 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1332 if (optname
!= MRT_INIT
) {
1333 if (sk
!= rcu_access_pointer(mrt
->mroute_sk
) &&
1334 !ns_capable(net
->user_ns
, CAP_NET_ADMIN
)) {
1342 if (optlen
!= sizeof(int)) {
1346 if (rtnl_dereference(mrt
->mroute_sk
)) {
1351 ret
= ip_ra_control(sk
, 1, mrtsock_destruct
);
1353 rcu_assign_pointer(mrt
->mroute_sk
, sk
);
1354 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)++;
1355 inet_netconf_notify_devconf(net
, RTM_NEWNETCONF
,
1356 NETCONFA_MC_FORWARDING
,
1357 NETCONFA_IFINDEX_ALL
,
1358 net
->ipv4
.devconf_all
);
1362 if (sk
!= rcu_access_pointer(mrt
->mroute_sk
)) {
1365 ret
= ip_ra_control(sk
, 0, NULL
);
1371 if (optlen
!= sizeof(vif
)) {
1375 if (copy_from_user(&vif
, optval
, sizeof(vif
))) {
1379 if (vif
.vifc_vifi
>= MAXVIFS
) {
1383 if (optname
== MRT_ADD_VIF
) {
1384 ret
= vif_add(net
, mrt
, &vif
,
1385 sk
== rtnl_dereference(mrt
->mroute_sk
));
1387 ret
= vif_delete(mrt
, vif
.vifc_vifi
, 0, NULL
);
1390 /* Manipulate the forwarding caches. These live
1391 * in a sort of kernel/user symbiosis.
1396 case MRT_ADD_MFC_PROXY
:
1397 case MRT_DEL_MFC_PROXY
:
1398 if (optlen
!= sizeof(mfc
)) {
1402 if (copy_from_user(&mfc
, optval
, sizeof(mfc
))) {
1407 parent
= mfc
.mfcc_parent
;
1408 if (optname
== MRT_DEL_MFC
|| optname
== MRT_DEL_MFC_PROXY
)
1409 ret
= ipmr_mfc_delete(mrt
, &mfc
, parent
);
1411 ret
= ipmr_mfc_add(net
, mrt
, &mfc
,
1412 sk
== rtnl_dereference(mrt
->mroute_sk
),
1415 /* Control PIM assert. */
1417 if (optlen
!= sizeof(val
)) {
1421 if (get_user(val
, (int __user
*)optval
)) {
1425 mrt
->mroute_do_assert
= val
;
1428 if (!ipmr_pimsm_enabled()) {
1432 if (optlen
!= sizeof(val
)) {
1436 if (get_user(val
, (int __user
*)optval
)) {
1442 if (val
!= mrt
->mroute_do_pim
) {
1443 mrt
->mroute_do_pim
= val
;
1444 mrt
->mroute_do_assert
= val
;
1448 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES
)) {
1452 if (optlen
!= sizeof(uval
)) {
1456 if (get_user(uval
, (u32 __user
*)optval
)) {
1461 if (sk
== rtnl_dereference(mrt
->mroute_sk
)) {
1464 mrt
= ipmr_new_table(net
, uval
);
1468 raw_sk(sk
)->ipmr_table
= uval
;
1471 /* Spurious command, or MRT_VERSION which you cannot set. */
1480 /* Getsock opt support for the multicast routing system. */
1481 int ip_mroute_getsockopt(struct sock
*sk
, int optname
, char __user
*optval
, int __user
*optlen
)
1485 struct net
*net
= sock_net(sk
);
1486 struct mr_table
*mrt
;
1488 if (sk
->sk_type
!= SOCK_RAW
||
1489 inet_sk(sk
)->inet_num
!= IPPROTO_IGMP
)
1492 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1501 if (!ipmr_pimsm_enabled())
1502 return -ENOPROTOOPT
;
1503 val
= mrt
->mroute_do_pim
;
1506 val
= mrt
->mroute_do_assert
;
1509 return -ENOPROTOOPT
;
1512 if (get_user(olr
, optlen
))
1514 olr
= min_t(unsigned int, olr
, sizeof(int));
1517 if (put_user(olr
, optlen
))
1519 if (copy_to_user(optval
, &val
, olr
))
1524 /* The IP multicast ioctl support routines. */
1525 int ipmr_ioctl(struct sock
*sk
, int cmd
, void __user
*arg
)
1527 struct sioc_sg_req sr
;
1528 struct sioc_vif_req vr
;
1529 struct vif_device
*vif
;
1530 struct mfc_cache
*c
;
1531 struct net
*net
= sock_net(sk
);
1532 struct mr_table
*mrt
;
1534 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1540 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1542 if (vr
.vifi
>= mrt
->maxvif
)
1544 read_lock(&mrt_lock
);
1545 vif
= &mrt
->vif_table
[vr
.vifi
];
1546 if (VIF_EXISTS(mrt
, vr
.vifi
)) {
1547 vr
.icount
= vif
->pkt_in
;
1548 vr
.ocount
= vif
->pkt_out
;
1549 vr
.ibytes
= vif
->bytes_in
;
1550 vr
.obytes
= vif
->bytes_out
;
1551 read_unlock(&mrt_lock
);
1553 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1557 read_unlock(&mrt_lock
);
1558 return -EADDRNOTAVAIL
;
1560 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1564 c
= ipmr_cache_find(mrt
, sr
.src
.s_addr
, sr
.grp
.s_addr
);
1566 sr
.pktcnt
= c
->mfc_un
.res
.pkt
;
1567 sr
.bytecnt
= c
->mfc_un
.res
.bytes
;
1568 sr
.wrong_if
= c
->mfc_un
.res
.wrong_if
;
1571 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1576 return -EADDRNOTAVAIL
;
1578 return -ENOIOCTLCMD
;
1582 #ifdef CONFIG_COMPAT
1583 struct compat_sioc_sg_req
{
1586 compat_ulong_t pktcnt
;
1587 compat_ulong_t bytecnt
;
1588 compat_ulong_t wrong_if
;
1591 struct compat_sioc_vif_req
{
1592 vifi_t vifi
; /* Which iface */
1593 compat_ulong_t icount
;
1594 compat_ulong_t ocount
;
1595 compat_ulong_t ibytes
;
1596 compat_ulong_t obytes
;
1599 int ipmr_compat_ioctl(struct sock
*sk
, unsigned int cmd
, void __user
*arg
)
1601 struct compat_sioc_sg_req sr
;
1602 struct compat_sioc_vif_req vr
;
1603 struct vif_device
*vif
;
1604 struct mfc_cache
*c
;
1605 struct net
*net
= sock_net(sk
);
1606 struct mr_table
*mrt
;
1608 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1614 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1616 if (vr
.vifi
>= mrt
->maxvif
)
1618 read_lock(&mrt_lock
);
1619 vif
= &mrt
->vif_table
[vr
.vifi
];
1620 if (VIF_EXISTS(mrt
, vr
.vifi
)) {
1621 vr
.icount
= vif
->pkt_in
;
1622 vr
.ocount
= vif
->pkt_out
;
1623 vr
.ibytes
= vif
->bytes_in
;
1624 vr
.obytes
= vif
->bytes_out
;
1625 read_unlock(&mrt_lock
);
1627 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1631 read_unlock(&mrt_lock
);
1632 return -EADDRNOTAVAIL
;
1634 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1638 c
= ipmr_cache_find(mrt
, sr
.src
.s_addr
, sr
.grp
.s_addr
);
1640 sr
.pktcnt
= c
->mfc_un
.res
.pkt
;
1641 sr
.bytecnt
= c
->mfc_un
.res
.bytes
;
1642 sr
.wrong_if
= c
->mfc_un
.res
.wrong_if
;
1645 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1650 return -EADDRNOTAVAIL
;
1652 return -ENOIOCTLCMD
;
1657 static int ipmr_device_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1659 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1660 struct net
*net
= dev_net(dev
);
1661 struct mr_table
*mrt
;
1662 struct vif_device
*v
;
1665 if (event
!= NETDEV_UNREGISTER
)
1668 ipmr_for_each_table(mrt
, net
) {
1669 v
= &mrt
->vif_table
[0];
1670 for (ct
= 0; ct
< mrt
->maxvif
; ct
++, v
++) {
1672 vif_delete(mrt
, ct
, 1, NULL
);
1678 static struct notifier_block ip_mr_notifier
= {
1679 .notifier_call
= ipmr_device_event
,
1682 /* Encapsulate a packet by attaching a valid IPIP header to it.
1683 * This avoids tunnel drivers and other mess and gives us the speed so
1684 * important for multicast video.
1686 static void ip_encap(struct net
*net
, struct sk_buff
*skb
,
1687 __be32 saddr
, __be32 daddr
)
1690 const struct iphdr
*old_iph
= ip_hdr(skb
);
1692 skb_push(skb
, sizeof(struct iphdr
));
1693 skb
->transport_header
= skb
->network_header
;
1694 skb_reset_network_header(skb
);
1698 iph
->tos
= old_iph
->tos
;
1699 iph
->ttl
= old_iph
->ttl
;
1703 iph
->protocol
= IPPROTO_IPIP
;
1705 iph
->tot_len
= htons(skb
->len
);
1706 ip_select_ident(net
, skb
, NULL
);
1709 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1713 static inline int ipmr_forward_finish(struct net
*net
, struct sock
*sk
,
1714 struct sk_buff
*skb
)
1716 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
1718 IP_INC_STATS(net
, IPSTATS_MIB_OUTFORWDATAGRAMS
);
1719 IP_ADD_STATS(net
, IPSTATS_MIB_OUTOCTETS
, skb
->len
);
1721 if (unlikely(opt
->optlen
))
1722 ip_forward_options(skb
);
1724 return dst_output(net
, sk
, skb
);
1727 /* Processing handlers for ipmr_forward */
1729 static void ipmr_queue_xmit(struct net
*net
, struct mr_table
*mrt
,
1730 struct sk_buff
*skb
, struct mfc_cache
*c
, int vifi
)
1732 const struct iphdr
*iph
= ip_hdr(skb
);
1733 struct vif_device
*vif
= &mrt
->vif_table
[vifi
];
1734 struct net_device
*dev
;
1742 if (vif
->flags
& VIFF_REGISTER
) {
1744 vif
->bytes_out
+= skb
->len
;
1745 vif
->dev
->stats
.tx_bytes
+= skb
->len
;
1746 vif
->dev
->stats
.tx_packets
++;
1747 ipmr_cache_report(mrt
, skb
, vifi
, IGMPMSG_WHOLEPKT
);
1751 if (vif
->flags
& VIFF_TUNNEL
) {
1752 rt
= ip_route_output_ports(net
, &fl4
, NULL
,
1753 vif
->remote
, vif
->local
,
1756 RT_TOS(iph
->tos
), vif
->link
);
1759 encap
= sizeof(struct iphdr
);
1761 rt
= ip_route_output_ports(net
, &fl4
, NULL
, iph
->daddr
, 0,
1764 RT_TOS(iph
->tos
), vif
->link
);
1771 if (skb
->len
+encap
> dst_mtu(&rt
->dst
) && (ntohs(iph
->frag_off
) & IP_DF
)) {
1772 /* Do not fragment multicasts. Alas, IPv4 does not
1773 * allow to send ICMP, so that packets will disappear
1776 IP_INC_STATS(net
, IPSTATS_MIB_FRAGFAILS
);
1781 encap
+= LL_RESERVED_SPACE(dev
) + rt
->dst
.header_len
;
1783 if (skb_cow(skb
, encap
)) {
1789 vif
->bytes_out
+= skb
->len
;
1792 skb_dst_set(skb
, &rt
->dst
);
1793 ip_decrease_ttl(ip_hdr(skb
));
1795 /* FIXME: forward and output firewalls used to be called here.
1796 * What do we do with netfilter? -- RR
1798 if (vif
->flags
& VIFF_TUNNEL
) {
1799 ip_encap(net
, skb
, vif
->local
, vif
->remote
);
1800 /* FIXME: extra output firewall step used to be here. --RR */
1801 vif
->dev
->stats
.tx_packets
++;
1802 vif
->dev
->stats
.tx_bytes
+= skb
->len
;
1805 IPCB(skb
)->flags
|= IPSKB_FORWARDED
;
1807 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1808 * not only before forwarding, but after forwarding on all output
1809 * interfaces. It is clear, if mrouter runs a multicasting
1810 * program, it should receive packets not depending to what interface
1811 * program is joined.
1812 * If we will not make it, the program will have to join on all
1813 * interfaces. On the other hand, multihoming host (or router, but
1814 * not mrouter) cannot join to more than one interface - it will
1815 * result in receiving multiple packets.
1817 NF_HOOK(NFPROTO_IPV4
, NF_INET_FORWARD
,
1818 net
, NULL
, skb
, skb
->dev
, dev
,
1819 ipmr_forward_finish
);
1826 static int ipmr_find_vif(struct mr_table
*mrt
, struct net_device
*dev
)
1830 for (ct
= mrt
->maxvif
-1; ct
>= 0; ct
--) {
1831 if (mrt
->vif_table
[ct
].dev
== dev
)
1837 /* "local" means that we should preserve one skb (for local delivery) */
1838 static void ip_mr_forward(struct net
*net
, struct mr_table
*mrt
,
1839 struct net_device
*dev
, struct sk_buff
*skb
,
1840 struct mfc_cache
*cache
, int local
)
1842 int true_vifi
= ipmr_find_vif(mrt
, dev
);
1846 vif
= cache
->mfc_parent
;
1847 cache
->mfc_un
.res
.pkt
++;
1848 cache
->mfc_un
.res
.bytes
+= skb
->len
;
1849 cache
->mfc_un
.res
.lastuse
= jiffies
;
1851 if (cache
->mfc_origin
== htonl(INADDR_ANY
) && true_vifi
>= 0) {
1852 struct mfc_cache
*cache_proxy
;
1854 /* For an (*,G) entry, we only check that the incomming
1855 * interface is part of the static tree.
1857 cache_proxy
= ipmr_cache_find_any_parent(mrt
, vif
);
1859 cache_proxy
->mfc_un
.res
.ttls
[true_vifi
] < 255)
1863 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1864 if (mrt
->vif_table
[vif
].dev
!= dev
) {
1865 if (rt_is_output_route(skb_rtable(skb
))) {
1866 /* It is our own packet, looped back.
1867 * Very complicated situation...
1869 * The best workaround until routing daemons will be
1870 * fixed is not to redistribute packet, if it was
1871 * send through wrong interface. It means, that
1872 * multicast applications WILL NOT work for
1873 * (S,G), which have default multicast route pointing
1874 * to wrong oif. In any case, it is not a good
1875 * idea to use multicasting applications on router.
1880 cache
->mfc_un
.res
.wrong_if
++;
1882 if (true_vifi
>= 0 && mrt
->mroute_do_assert
&&
1883 /* pimsm uses asserts, when switching from RPT to SPT,
1884 * so that we cannot check that packet arrived on an oif.
1885 * It is bad, but otherwise we would need to move pretty
1886 * large chunk of pimd to kernel. Ough... --ANK
1888 (mrt
->mroute_do_pim
||
1889 cache
->mfc_un
.res
.ttls
[true_vifi
] < 255) &&
1891 cache
->mfc_un
.res
.last_assert
+ MFC_ASSERT_THRESH
)) {
1892 cache
->mfc_un
.res
.last_assert
= jiffies
;
1893 ipmr_cache_report(mrt
, skb
, true_vifi
, IGMPMSG_WRONGVIF
);
1899 mrt
->vif_table
[vif
].pkt_in
++;
1900 mrt
->vif_table
[vif
].bytes_in
+= skb
->len
;
1902 /* Forward the frame */
1903 if (cache
->mfc_origin
== htonl(INADDR_ANY
) &&
1904 cache
->mfc_mcastgrp
== htonl(INADDR_ANY
)) {
1905 if (true_vifi
>= 0 &&
1906 true_vifi
!= cache
->mfc_parent
&&
1908 cache
->mfc_un
.res
.ttls
[cache
->mfc_parent
]) {
1909 /* It's an (*,*) entry and the packet is not coming from
1910 * the upstream: forward the packet to the upstream
1913 psend
= cache
->mfc_parent
;
1918 for (ct
= cache
->mfc_un
.res
.maxvif
- 1;
1919 ct
>= cache
->mfc_un
.res
.minvif
; ct
--) {
1920 /* For (*,G) entry, don't forward to the incoming interface */
1921 if ((cache
->mfc_origin
!= htonl(INADDR_ANY
) ||
1923 ip_hdr(skb
)->ttl
> cache
->mfc_un
.res
.ttls
[ct
]) {
1925 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
1928 ipmr_queue_xmit(net
, mrt
, skb2
, cache
,
1937 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
1940 ipmr_queue_xmit(net
, mrt
, skb2
, cache
, psend
);
1942 ipmr_queue_xmit(net
, mrt
, skb
, cache
, psend
);
1952 static struct mr_table
*ipmr_rt_fib_lookup(struct net
*net
, struct sk_buff
*skb
)
1954 struct rtable
*rt
= skb_rtable(skb
);
1955 struct iphdr
*iph
= ip_hdr(skb
);
1956 struct flowi4 fl4
= {
1957 .daddr
= iph
->daddr
,
1958 .saddr
= iph
->saddr
,
1959 .flowi4_tos
= RT_TOS(iph
->tos
),
1960 .flowi4_oif
= (rt_is_output_route(rt
) ?
1961 skb
->dev
->ifindex
: 0),
1962 .flowi4_iif
= (rt_is_output_route(rt
) ?
1965 .flowi4_mark
= skb
->mark
,
1967 struct mr_table
*mrt
;
1970 err
= ipmr_fib_lookup(net
, &fl4
, &mrt
);
1972 return ERR_PTR(err
);
1976 /* Multicast packets for forwarding arrive here
1977 * Called with rcu_read_lock();
1979 int ip_mr_input(struct sk_buff
*skb
)
1981 struct mfc_cache
*cache
;
1982 struct net
*net
= dev_net(skb
->dev
);
1983 int local
= skb_rtable(skb
)->rt_flags
& RTCF_LOCAL
;
1984 struct mr_table
*mrt
;
1985 struct net_device
*dev
;
1987 /* skb->dev passed in is the loX master dev for vrfs.
1988 * As there are no vifs associated with loopback devices,
1989 * get the proper interface that does have a vif associated with it.
1992 if (netif_is_l3_master(skb
->dev
)) {
1993 dev
= dev_get_by_index_rcu(net
, IPCB(skb
)->iif
);
2000 /* Packet is looped back after forward, it should not be
2001 * forwarded second time, but still can be delivered locally.
2003 if (IPCB(skb
)->flags
& IPSKB_FORWARDED
)
2006 mrt
= ipmr_rt_fib_lookup(net
, skb
);
2009 return PTR_ERR(mrt
);
2012 if (IPCB(skb
)->opt
.router_alert
) {
2013 if (ip_call_ra_chain(skb
))
2015 } else if (ip_hdr(skb
)->protocol
== IPPROTO_IGMP
) {
2016 /* IGMPv1 (and broken IGMPv2 implementations sort of
2017 * Cisco IOS <= 11.2(8)) do not put router alert
2018 * option to IGMP packets destined to routable
2019 * groups. It is very bad, because it means
2020 * that we can forward NO IGMP messages.
2022 struct sock
*mroute_sk
;
2024 mroute_sk
= rcu_dereference(mrt
->mroute_sk
);
2027 raw_rcv(mroute_sk
, skb
);
2033 /* already under rcu_read_lock() */
2034 cache
= ipmr_cache_find(mrt
, ip_hdr(skb
)->saddr
, ip_hdr(skb
)->daddr
);
2036 int vif
= ipmr_find_vif(mrt
, dev
);
2039 cache
= ipmr_cache_find_any(mrt
, ip_hdr(skb
)->daddr
,
2043 /* No usable cache entry */
2048 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
2049 ip_local_deliver(skb
);
2055 read_lock(&mrt_lock
);
2056 vif
= ipmr_find_vif(mrt
, dev
);
2058 int err2
= ipmr_cache_unresolved(mrt
, vif
, skb
, dev
);
2059 read_unlock(&mrt_lock
);
2063 read_unlock(&mrt_lock
);
2068 read_lock(&mrt_lock
);
2069 ip_mr_forward(net
, mrt
, dev
, skb
, cache
, local
);
2070 read_unlock(&mrt_lock
);
2073 return ip_local_deliver(skb
);
2079 return ip_local_deliver(skb
);
2084 #ifdef CONFIG_IP_PIMSM_V1
2085 /* Handle IGMP messages of PIMv1 */
2086 int pim_rcv_v1(struct sk_buff
*skb
)
2088 struct igmphdr
*pim
;
2089 struct net
*net
= dev_net(skb
->dev
);
2090 struct mr_table
*mrt
;
2092 if (!pskb_may_pull(skb
, sizeof(*pim
) + sizeof(struct iphdr
)))
2095 pim
= igmp_hdr(skb
);
2097 mrt
= ipmr_rt_fib_lookup(net
, skb
);
2100 if (!mrt
->mroute_do_pim
||
2101 pim
->group
!= PIM_V1_VERSION
|| pim
->code
!= PIM_V1_REGISTER
)
2104 if (__pim_rcv(mrt
, skb
, sizeof(*pim
))) {
2112 #ifdef CONFIG_IP_PIMSM_V2
2113 static int pim_rcv(struct sk_buff
*skb
)
2115 struct pimreghdr
*pim
;
2116 struct net
*net
= dev_net(skb
->dev
);
2117 struct mr_table
*mrt
;
2119 if (!pskb_may_pull(skb
, sizeof(*pim
) + sizeof(struct iphdr
)))
2122 pim
= (struct pimreghdr
*)skb_transport_header(skb
);
2123 if (pim
->type
!= ((PIM_VERSION
<< 4) | (PIM_TYPE_REGISTER
)) ||
2124 (pim
->flags
& PIM_NULL_REGISTER
) ||
2125 (ip_compute_csum((void *)pim
, sizeof(*pim
)) != 0 &&
2126 csum_fold(skb_checksum(skb
, 0, skb
->len
, 0))))
2129 mrt
= ipmr_rt_fib_lookup(net
, skb
);
2132 if (__pim_rcv(mrt
, skb
, sizeof(*pim
))) {
2140 static int __ipmr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
2141 struct mfc_cache
*c
, struct rtmsg
*rtm
)
2143 struct rta_mfc_stats mfcs
;
2144 struct nlattr
*mp_attr
;
2145 struct rtnexthop
*nhp
;
2146 unsigned long lastuse
;
2149 /* If cache is unresolved, don't try to parse IIF and OIF */
2150 if (c
->mfc_parent
>= MAXVIFS
) {
2151 rtm
->rtm_flags
|= RTNH_F_UNRESOLVED
;
2155 if (VIF_EXISTS(mrt
, c
->mfc_parent
) &&
2156 nla_put_u32(skb
, RTA_IIF
, mrt
->vif_table
[c
->mfc_parent
].dev
->ifindex
) < 0)
2159 if (!(mp_attr
= nla_nest_start(skb
, RTA_MULTIPATH
)))
2162 for (ct
= c
->mfc_un
.res
.minvif
; ct
< c
->mfc_un
.res
.maxvif
; ct
++) {
2163 if (VIF_EXISTS(mrt
, ct
) && c
->mfc_un
.res
.ttls
[ct
] < 255) {
2164 if (!(nhp
= nla_reserve_nohdr(skb
, sizeof(*nhp
)))) {
2165 nla_nest_cancel(skb
, mp_attr
);
2169 nhp
->rtnh_flags
= 0;
2170 nhp
->rtnh_hops
= c
->mfc_un
.res
.ttls
[ct
];
2171 nhp
->rtnh_ifindex
= mrt
->vif_table
[ct
].dev
->ifindex
;
2172 nhp
->rtnh_len
= sizeof(*nhp
);
2176 nla_nest_end(skb
, mp_attr
);
2178 lastuse
= READ_ONCE(c
->mfc_un
.res
.lastuse
);
2179 lastuse
= time_after_eq(jiffies
, lastuse
) ? jiffies
- lastuse
: 0;
2181 mfcs
.mfcs_packets
= c
->mfc_un
.res
.pkt
;
2182 mfcs
.mfcs_bytes
= c
->mfc_un
.res
.bytes
;
2183 mfcs
.mfcs_wrong_if
= c
->mfc_un
.res
.wrong_if
;
2184 if (nla_put_64bit(skb
, RTA_MFC_STATS
, sizeof(mfcs
), &mfcs
, RTA_PAD
) ||
2185 nla_put_u64_64bit(skb
, RTA_EXPIRES
, jiffies_to_clock_t(lastuse
),
2189 rtm
->rtm_type
= RTN_MULTICAST
;
2193 int ipmr_get_route(struct net
*net
, struct sk_buff
*skb
,
2194 __be32 saddr
, __be32 daddr
,
2195 struct rtmsg
*rtm
, u32 portid
)
2197 struct mfc_cache
*cache
;
2198 struct mr_table
*mrt
;
2201 mrt
= ipmr_get_table(net
, RT_TABLE_DEFAULT
);
2206 cache
= ipmr_cache_find(mrt
, saddr
, daddr
);
2207 if (!cache
&& skb
->dev
) {
2208 int vif
= ipmr_find_vif(mrt
, skb
->dev
);
2211 cache
= ipmr_cache_find_any(mrt
, daddr
, vif
);
2214 struct sk_buff
*skb2
;
2216 struct net_device
*dev
;
2220 read_lock(&mrt_lock
);
2222 vif
= ipmr_find_vif(mrt
, dev
);
2224 read_unlock(&mrt_lock
);
2228 skb2
= skb_clone(skb
, GFP_ATOMIC
);
2230 read_unlock(&mrt_lock
);
2235 NETLINK_CB(skb2
).portid
= portid
;
2236 skb_push(skb2
, sizeof(struct iphdr
));
2237 skb_reset_network_header(skb2
);
2239 iph
->ihl
= sizeof(struct iphdr
) >> 2;
2243 err
= ipmr_cache_unresolved(mrt
, vif
, skb2
, dev
);
2244 read_unlock(&mrt_lock
);
2249 read_lock(&mrt_lock
);
2250 err
= __ipmr_fill_mroute(mrt
, skb
, cache
, rtm
);
2251 read_unlock(&mrt_lock
);
2256 static int ipmr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
2257 u32 portid
, u32 seq
, struct mfc_cache
*c
, int cmd
,
2260 struct nlmsghdr
*nlh
;
2264 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rtm
), flags
);
2268 rtm
= nlmsg_data(nlh
);
2269 rtm
->rtm_family
= RTNL_FAMILY_IPMR
;
2270 rtm
->rtm_dst_len
= 32;
2271 rtm
->rtm_src_len
= 32;
2273 rtm
->rtm_table
= mrt
->id
;
2274 if (nla_put_u32(skb
, RTA_TABLE
, mrt
->id
))
2275 goto nla_put_failure
;
2276 rtm
->rtm_type
= RTN_MULTICAST
;
2277 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2278 if (c
->mfc_flags
& MFC_STATIC
)
2279 rtm
->rtm_protocol
= RTPROT_STATIC
;
2281 rtm
->rtm_protocol
= RTPROT_MROUTED
;
2284 if (nla_put_in_addr(skb
, RTA_SRC
, c
->mfc_origin
) ||
2285 nla_put_in_addr(skb
, RTA_DST
, c
->mfc_mcastgrp
))
2286 goto nla_put_failure
;
2287 err
= __ipmr_fill_mroute(mrt
, skb
, c
, rtm
);
2288 /* do not break the dump if cache is unresolved */
2289 if (err
< 0 && err
!= -ENOENT
)
2290 goto nla_put_failure
;
2292 nlmsg_end(skb
, nlh
);
2296 nlmsg_cancel(skb
, nlh
);
2300 static size_t mroute_msgsize(bool unresolved
, int maxvif
)
2303 NLMSG_ALIGN(sizeof(struct rtmsg
))
2304 + nla_total_size(4) /* RTA_TABLE */
2305 + nla_total_size(4) /* RTA_SRC */
2306 + nla_total_size(4) /* RTA_DST */
2311 + nla_total_size(4) /* RTA_IIF */
2312 + nla_total_size(0) /* RTA_MULTIPATH */
2313 + maxvif
* NLA_ALIGN(sizeof(struct rtnexthop
))
2315 + nla_total_size_64bit(sizeof(struct rta_mfc_stats
))
2321 static void mroute_netlink_event(struct mr_table
*mrt
, struct mfc_cache
*mfc
,
2324 struct net
*net
= read_pnet(&mrt
->net
);
2325 struct sk_buff
*skb
;
2328 skb
= nlmsg_new(mroute_msgsize(mfc
->mfc_parent
>= MAXVIFS
, mrt
->maxvif
),
2333 err
= ipmr_fill_mroute(mrt
, skb
, 0, 0, mfc
, cmd
, 0);
2337 rtnl_notify(skb
, net
, 0, RTNLGRP_IPV4_MROUTE
, NULL
, GFP_ATOMIC
);
2343 rtnl_set_sk_err(net
, RTNLGRP_IPV4_MROUTE
, err
);
2346 static size_t igmpmsg_netlink_msgsize(size_t payloadlen
)
2349 NLMSG_ALIGN(sizeof(struct rtgenmsg
))
2350 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2351 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2352 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2353 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2354 /* IPMRA_CREPORT_PKT */
2355 + nla_total_size(payloadlen
)
2361 static void igmpmsg_netlink_event(struct mr_table
*mrt
, struct sk_buff
*pkt
)
2363 struct net
*net
= read_pnet(&mrt
->net
);
2364 struct nlmsghdr
*nlh
;
2365 struct rtgenmsg
*rtgenm
;
2366 struct igmpmsg
*msg
;
2367 struct sk_buff
*skb
;
2371 payloadlen
= pkt
->len
- sizeof(struct igmpmsg
);
2372 msg
= (struct igmpmsg
*)skb_network_header(pkt
);
2374 skb
= nlmsg_new(igmpmsg_netlink_msgsize(payloadlen
), GFP_ATOMIC
);
2378 nlh
= nlmsg_put(skb
, 0, 0, RTM_NEWCACHEREPORT
,
2379 sizeof(struct rtgenmsg
), 0);
2382 rtgenm
= nlmsg_data(nlh
);
2383 rtgenm
->rtgen_family
= RTNL_FAMILY_IPMR
;
2384 if (nla_put_u8(skb
, IPMRA_CREPORT_MSGTYPE
, msg
->im_msgtype
) ||
2385 nla_put_u32(skb
, IPMRA_CREPORT_VIF_ID
, msg
->im_vif
) ||
2386 nla_put_in_addr(skb
, IPMRA_CREPORT_SRC_ADDR
,
2387 msg
->im_src
.s_addr
) ||
2388 nla_put_in_addr(skb
, IPMRA_CREPORT_DST_ADDR
,
2389 msg
->im_dst
.s_addr
))
2390 goto nla_put_failure
;
2392 nla
= nla_reserve(skb
, IPMRA_CREPORT_PKT
, payloadlen
);
2393 if (!nla
|| skb_copy_bits(pkt
, sizeof(struct igmpmsg
),
2394 nla_data(nla
), payloadlen
))
2395 goto nla_put_failure
;
2397 nlmsg_end(skb
, nlh
);
2399 rtnl_notify(skb
, net
, 0, RTNLGRP_IPV4_MROUTE_R
, NULL
, GFP_ATOMIC
);
2403 nlmsg_cancel(skb
, nlh
);
2406 rtnl_set_sk_err(net
, RTNLGRP_IPV4_MROUTE_R
, -ENOBUFS
);
2409 static int ipmr_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
,
2410 struct netlink_ext_ack
*extack
)
2412 struct net
*net
= sock_net(in_skb
->sk
);
2413 struct nlattr
*tb
[RTA_MAX
+ 1];
2414 struct sk_buff
*skb
= NULL
;
2415 struct mfc_cache
*cache
;
2416 struct mr_table
*mrt
;
2422 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
,
2423 rtm_ipv4_policy
, extack
);
2427 rtm
= nlmsg_data(nlh
);
2429 src
= tb
[RTA_SRC
] ? nla_get_in_addr(tb
[RTA_SRC
]) : 0;
2430 grp
= tb
[RTA_DST
] ? nla_get_in_addr(tb
[RTA_DST
]) : 0;
2431 tableid
= tb
[RTA_TABLE
] ? nla_get_u32(tb
[RTA_TABLE
]) : 0;
2433 mrt
= ipmr_get_table(net
, tableid
? tableid
: RT_TABLE_DEFAULT
);
2439 /* entries are added/deleted only under RTNL */
2441 cache
= ipmr_cache_find(mrt
, src
, grp
);
2448 skb
= nlmsg_new(mroute_msgsize(false, mrt
->maxvif
), GFP_KERNEL
);
2454 err
= ipmr_fill_mroute(mrt
, skb
, NETLINK_CB(in_skb
).portid
,
2455 nlh
->nlmsg_seq
, cache
,
2460 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2470 static int ipmr_rtm_dumproute(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2472 struct net
*net
= sock_net(skb
->sk
);
2473 struct mr_table
*mrt
;
2474 struct mfc_cache
*mfc
;
2475 unsigned int t
= 0, s_t
;
2476 unsigned int e
= 0, s_e
;
2482 ipmr_for_each_table(mrt
, net
) {
2485 list_for_each_entry_rcu(mfc
, &mrt
->mfc_cache_list
, list
) {
2488 if (ipmr_fill_mroute(mrt
, skb
,
2489 NETLINK_CB(cb
->skb
).portid
,
2500 spin_lock_bh(&mfc_unres_lock
);
2501 list_for_each_entry(mfc
, &mrt
->mfc_unres_queue
, list
) {
2504 if (ipmr_fill_mroute(mrt
, skb
,
2505 NETLINK_CB(cb
->skb
).portid
,
2509 spin_unlock_bh(&mfc_unres_lock
);
2515 spin_unlock_bh(&mfc_unres_lock
);
2530 static const struct nla_policy rtm_ipmr_policy
[RTA_MAX
+ 1] = {
2531 [RTA_SRC
] = { .type
= NLA_U32
},
2532 [RTA_DST
] = { .type
= NLA_U32
},
2533 [RTA_IIF
] = { .type
= NLA_U32
},
2534 [RTA_TABLE
] = { .type
= NLA_U32
},
2535 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2538 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol
)
2540 switch (rtm_protocol
) {
2542 case RTPROT_MROUTED
:
2548 static int ipmr_nla_get_ttls(const struct nlattr
*nla
, struct mfcctl
*mfcc
)
2550 struct rtnexthop
*rtnh
= nla_data(nla
);
2551 int remaining
= nla_len(nla
), vifi
= 0;
2553 while (rtnh_ok(rtnh
, remaining
)) {
2554 mfcc
->mfcc_ttls
[vifi
] = rtnh
->rtnh_hops
;
2555 if (++vifi
== MAXVIFS
)
2557 rtnh
= rtnh_next(rtnh
, &remaining
);
2560 return remaining
> 0 ? -EINVAL
: vifi
;
2563 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2564 static int rtm_to_ipmr_mfcc(struct net
*net
, struct nlmsghdr
*nlh
,
2565 struct mfcctl
*mfcc
, int *mrtsock
,
2566 struct mr_table
**mrtret
,
2567 struct netlink_ext_ack
*extack
)
2569 struct net_device
*dev
= NULL
;
2570 u32 tblid
= RT_TABLE_DEFAULT
;
2571 struct mr_table
*mrt
;
2572 struct nlattr
*attr
;
2576 ret
= nlmsg_validate(nlh
, sizeof(*rtm
), RTA_MAX
, rtm_ipmr_policy
,
2580 rtm
= nlmsg_data(nlh
);
2583 if (rtm
->rtm_family
!= RTNL_FAMILY_IPMR
|| rtm
->rtm_dst_len
!= 32 ||
2584 rtm
->rtm_type
!= RTN_MULTICAST
||
2585 rtm
->rtm_scope
!= RT_SCOPE_UNIVERSE
||
2586 !ipmr_rtm_validate_proto(rtm
->rtm_protocol
))
2589 memset(mfcc
, 0, sizeof(*mfcc
));
2590 mfcc
->mfcc_parent
= -1;
2592 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct rtmsg
), rem
) {
2593 switch (nla_type(attr
)) {
2595 mfcc
->mfcc_origin
.s_addr
= nla_get_be32(attr
);
2598 mfcc
->mfcc_mcastgrp
.s_addr
= nla_get_be32(attr
);
2601 dev
= __dev_get_by_index(net
, nla_get_u32(attr
));
2608 if (ipmr_nla_get_ttls(attr
, mfcc
) < 0) {
2617 tblid
= nla_get_u32(attr
);
2621 mrt
= ipmr_get_table(net
, tblid
);
2627 *mrtsock
= rtm
->rtm_protocol
== RTPROT_MROUTED
? 1 : 0;
2629 mfcc
->mfcc_parent
= ipmr_find_vif(mrt
, dev
);
2635 /* takes care of both newroute and delroute */
2636 static int ipmr_rtm_route(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2637 struct netlink_ext_ack
*extack
)
2639 struct net
*net
= sock_net(skb
->sk
);
2640 int ret
, mrtsock
, parent
;
2641 struct mr_table
*tbl
;
2646 ret
= rtm_to_ipmr_mfcc(net
, nlh
, &mfcc
, &mrtsock
, &tbl
, extack
);
2650 parent
= ret
? mfcc
.mfcc_parent
: -1;
2651 if (nlh
->nlmsg_type
== RTM_NEWROUTE
)
2652 return ipmr_mfc_add(net
, tbl
, &mfcc
, mrtsock
, parent
);
2654 return ipmr_mfc_delete(tbl
, &mfcc
, parent
);
2657 static bool ipmr_fill_table(struct mr_table
*mrt
, struct sk_buff
*skb
)
2659 u32 queue_len
= atomic_read(&mrt
->cache_resolve_queue_len
);
2661 if (nla_put_u32(skb
, IPMRA_TABLE_ID
, mrt
->id
) ||
2662 nla_put_u32(skb
, IPMRA_TABLE_CACHE_RES_QUEUE_LEN
, queue_len
) ||
2663 nla_put_s32(skb
, IPMRA_TABLE_MROUTE_REG_VIF_NUM
,
2664 mrt
->mroute_reg_vif_num
) ||
2665 nla_put_u8(skb
, IPMRA_TABLE_MROUTE_DO_ASSERT
,
2666 mrt
->mroute_do_assert
) ||
2667 nla_put_u8(skb
, IPMRA_TABLE_MROUTE_DO_PIM
, mrt
->mroute_do_pim
))
2673 static bool ipmr_fill_vif(struct mr_table
*mrt
, u32 vifid
, struct sk_buff
*skb
)
2675 struct nlattr
*vif_nest
;
2676 struct vif_device
*vif
;
2678 /* if the VIF doesn't exist just continue */
2679 if (!VIF_EXISTS(mrt
, vifid
))
2682 vif
= &mrt
->vif_table
[vifid
];
2683 vif_nest
= nla_nest_start(skb
, IPMRA_VIF
);
2686 if (nla_put_u32(skb
, IPMRA_VIFA_IFINDEX
, vif
->dev
->ifindex
) ||
2687 nla_put_u32(skb
, IPMRA_VIFA_VIF_ID
, vifid
) ||
2688 nla_put_u16(skb
, IPMRA_VIFA_FLAGS
, vif
->flags
) ||
2689 nla_put_u64_64bit(skb
, IPMRA_VIFA_BYTES_IN
, vif
->bytes_in
,
2691 nla_put_u64_64bit(skb
, IPMRA_VIFA_BYTES_OUT
, vif
->bytes_out
,
2693 nla_put_u64_64bit(skb
, IPMRA_VIFA_PACKETS_IN
, vif
->pkt_in
,
2695 nla_put_u64_64bit(skb
, IPMRA_VIFA_PACKETS_OUT
, vif
->pkt_out
,
2697 nla_put_be32(skb
, IPMRA_VIFA_LOCAL_ADDR
, vif
->local
) ||
2698 nla_put_be32(skb
, IPMRA_VIFA_REMOTE_ADDR
, vif
->remote
)) {
2699 nla_nest_cancel(skb
, vif_nest
);
2702 nla_nest_end(skb
, vif_nest
);
2707 static int ipmr_rtm_dumplink(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2709 struct net
*net
= sock_net(skb
->sk
);
2710 struct nlmsghdr
*nlh
= NULL
;
2711 unsigned int t
= 0, s_t
;
2712 unsigned int e
= 0, s_e
;
2713 struct mr_table
*mrt
;
2718 ipmr_for_each_table(mrt
, net
) {
2719 struct nlattr
*vifs
, *af
;
2720 struct ifinfomsg
*hdr
;
2725 nlh
= nlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
,
2726 cb
->nlh
->nlmsg_seq
, RTM_NEWLINK
,
2727 sizeof(*hdr
), NLM_F_MULTI
);
2731 hdr
= nlmsg_data(nlh
);
2732 memset(hdr
, 0, sizeof(*hdr
));
2733 hdr
->ifi_family
= RTNL_FAMILY_IPMR
;
2735 af
= nla_nest_start(skb
, IFLA_AF_SPEC
);
2737 nlmsg_cancel(skb
, nlh
);
2741 if (!ipmr_fill_table(mrt
, skb
)) {
2742 nlmsg_cancel(skb
, nlh
);
2746 vifs
= nla_nest_start(skb
, IPMRA_TABLE_VIFS
);
2748 nla_nest_end(skb
, af
);
2749 nlmsg_end(skb
, nlh
);
2752 for (i
= 0; i
< mrt
->maxvif
; i
++) {
2755 if (!ipmr_fill_vif(mrt
, i
, skb
)) {
2756 nla_nest_end(skb
, vifs
);
2757 nla_nest_end(skb
, af
);
2758 nlmsg_end(skb
, nlh
);
2766 nla_nest_end(skb
, vifs
);
2767 nla_nest_end(skb
, af
);
2768 nlmsg_end(skb
, nlh
);
2780 #ifdef CONFIG_PROC_FS
2781 /* The /proc interfaces to multicast routing :
2782 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2784 struct ipmr_vif_iter
{
2785 struct seq_net_private p
;
2786 struct mr_table
*mrt
;
2790 static struct vif_device
*ipmr_vif_seq_idx(struct net
*net
,
2791 struct ipmr_vif_iter
*iter
,
2794 struct mr_table
*mrt
= iter
->mrt
;
2796 for (iter
->ct
= 0; iter
->ct
< mrt
->maxvif
; ++iter
->ct
) {
2797 if (!VIF_EXISTS(mrt
, iter
->ct
))
2800 return &mrt
->vif_table
[iter
->ct
];
2805 static void *ipmr_vif_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2806 __acquires(mrt_lock
)
2808 struct ipmr_vif_iter
*iter
= seq
->private;
2809 struct net
*net
= seq_file_net(seq
);
2810 struct mr_table
*mrt
;
2812 mrt
= ipmr_get_table(net
, RT_TABLE_DEFAULT
);
2814 return ERR_PTR(-ENOENT
);
2818 read_lock(&mrt_lock
);
2819 return *pos
? ipmr_vif_seq_idx(net
, seq
->private, *pos
- 1)
2823 static void *ipmr_vif_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2825 struct ipmr_vif_iter
*iter
= seq
->private;
2826 struct net
*net
= seq_file_net(seq
);
2827 struct mr_table
*mrt
= iter
->mrt
;
2830 if (v
== SEQ_START_TOKEN
)
2831 return ipmr_vif_seq_idx(net
, iter
, 0);
2833 while (++iter
->ct
< mrt
->maxvif
) {
2834 if (!VIF_EXISTS(mrt
, iter
->ct
))
2836 return &mrt
->vif_table
[iter
->ct
];
2841 static void ipmr_vif_seq_stop(struct seq_file
*seq
, void *v
)
2842 __releases(mrt_lock
)
2844 read_unlock(&mrt_lock
);
2847 static int ipmr_vif_seq_show(struct seq_file
*seq
, void *v
)
2849 struct ipmr_vif_iter
*iter
= seq
->private;
2850 struct mr_table
*mrt
= iter
->mrt
;
2852 if (v
== SEQ_START_TOKEN
) {
2854 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2856 const struct vif_device
*vif
= v
;
2857 const char *name
= vif
->dev
? vif
->dev
->name
: "none";
2860 "%2zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2861 vif
- mrt
->vif_table
,
2862 name
, vif
->bytes_in
, vif
->pkt_in
,
2863 vif
->bytes_out
, vif
->pkt_out
,
2864 vif
->flags
, vif
->local
, vif
->remote
);
2869 static const struct seq_operations ipmr_vif_seq_ops
= {
2870 .start
= ipmr_vif_seq_start
,
2871 .next
= ipmr_vif_seq_next
,
2872 .stop
= ipmr_vif_seq_stop
,
2873 .show
= ipmr_vif_seq_show
,
2876 static int ipmr_vif_open(struct inode
*inode
, struct file
*file
)
2878 return seq_open_net(inode
, file
, &ipmr_vif_seq_ops
,
2879 sizeof(struct ipmr_vif_iter
));
2882 static const struct file_operations ipmr_vif_fops
= {
2883 .owner
= THIS_MODULE
,
2884 .open
= ipmr_vif_open
,
2886 .llseek
= seq_lseek
,
2887 .release
= seq_release_net
,
2890 struct ipmr_mfc_iter
{
2891 struct seq_net_private p
;
2892 struct mr_table
*mrt
;
2893 struct list_head
*cache
;
2896 static struct mfc_cache
*ipmr_mfc_seq_idx(struct net
*net
,
2897 struct ipmr_mfc_iter
*it
, loff_t pos
)
2899 struct mr_table
*mrt
= it
->mrt
;
2900 struct mfc_cache
*mfc
;
2903 it
->cache
= &mrt
->mfc_cache_list
;
2904 list_for_each_entry_rcu(mfc
, &mrt
->mfc_cache_list
, list
)
2909 spin_lock_bh(&mfc_unres_lock
);
2910 it
->cache
= &mrt
->mfc_unres_queue
;
2911 list_for_each_entry(mfc
, it
->cache
, list
)
2914 spin_unlock_bh(&mfc_unres_lock
);
2921 static void *ipmr_mfc_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2923 struct ipmr_mfc_iter
*it
= seq
->private;
2924 struct net
*net
= seq_file_net(seq
);
2925 struct mr_table
*mrt
;
2927 mrt
= ipmr_get_table(net
, RT_TABLE_DEFAULT
);
2929 return ERR_PTR(-ENOENT
);
2933 return *pos
? ipmr_mfc_seq_idx(net
, seq
->private, *pos
- 1)
2937 static void *ipmr_mfc_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2939 struct ipmr_mfc_iter
*it
= seq
->private;
2940 struct net
*net
= seq_file_net(seq
);
2941 struct mr_table
*mrt
= it
->mrt
;
2942 struct mfc_cache
*mfc
= v
;
2946 if (v
== SEQ_START_TOKEN
)
2947 return ipmr_mfc_seq_idx(net
, seq
->private, 0);
2949 if (mfc
->list
.next
!= it
->cache
)
2950 return list_entry(mfc
->list
.next
, struct mfc_cache
, list
);
2952 if (it
->cache
== &mrt
->mfc_unres_queue
)
2955 /* exhausted cache_array, show unresolved */
2957 it
->cache
= &mrt
->mfc_unres_queue
;
2959 spin_lock_bh(&mfc_unres_lock
);
2960 if (!list_empty(it
->cache
))
2961 return list_first_entry(it
->cache
, struct mfc_cache
, list
);
2964 spin_unlock_bh(&mfc_unres_lock
);
2970 static void ipmr_mfc_seq_stop(struct seq_file
*seq
, void *v
)
2972 struct ipmr_mfc_iter
*it
= seq
->private;
2973 struct mr_table
*mrt
= it
->mrt
;
2975 if (it
->cache
== &mrt
->mfc_unres_queue
)
2976 spin_unlock_bh(&mfc_unres_lock
);
2977 else if (it
->cache
== &mrt
->mfc_cache_list
)
2981 static int ipmr_mfc_seq_show(struct seq_file
*seq
, void *v
)
2985 if (v
== SEQ_START_TOKEN
) {
2987 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2989 const struct mfc_cache
*mfc
= v
;
2990 const struct ipmr_mfc_iter
*it
= seq
->private;
2991 const struct mr_table
*mrt
= it
->mrt
;
2993 seq_printf(seq
, "%08X %08X %-3hd",
2994 (__force u32
) mfc
->mfc_mcastgrp
,
2995 (__force u32
) mfc
->mfc_origin
,
2998 if (it
->cache
!= &mrt
->mfc_unres_queue
) {
2999 seq_printf(seq
, " %8lu %8lu %8lu",
3000 mfc
->mfc_un
.res
.pkt
,
3001 mfc
->mfc_un
.res
.bytes
,
3002 mfc
->mfc_un
.res
.wrong_if
);
3003 for (n
= mfc
->mfc_un
.res
.minvif
;
3004 n
< mfc
->mfc_un
.res
.maxvif
; n
++) {
3005 if (VIF_EXISTS(mrt
, n
) &&
3006 mfc
->mfc_un
.res
.ttls
[n
] < 255)
3009 n
, mfc
->mfc_un
.res
.ttls
[n
]);
3012 /* unresolved mfc_caches don't contain
3013 * pkt, bytes and wrong_if values
3015 seq_printf(seq
, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3017 seq_putc(seq
, '\n');
3022 static const struct seq_operations ipmr_mfc_seq_ops
= {
3023 .start
= ipmr_mfc_seq_start
,
3024 .next
= ipmr_mfc_seq_next
,
3025 .stop
= ipmr_mfc_seq_stop
,
3026 .show
= ipmr_mfc_seq_show
,
3029 static int ipmr_mfc_open(struct inode
*inode
, struct file
*file
)
3031 return seq_open_net(inode
, file
, &ipmr_mfc_seq_ops
,
3032 sizeof(struct ipmr_mfc_iter
));
3035 static const struct file_operations ipmr_mfc_fops
= {
3036 .owner
= THIS_MODULE
,
3037 .open
= ipmr_mfc_open
,
3039 .llseek
= seq_lseek
,
3040 .release
= seq_release_net
,
3044 #ifdef CONFIG_IP_PIMSM_V2
3045 static const struct net_protocol pim_protocol
= {
3051 /* Setup for IP multicast routing */
3052 static int __net_init
ipmr_net_init(struct net
*net
)
3056 err
= ipmr_rules_init(net
);
3060 #ifdef CONFIG_PROC_FS
3062 if (!proc_create("ip_mr_vif", 0, net
->proc_net
, &ipmr_vif_fops
))
3064 if (!proc_create("ip_mr_cache", 0, net
->proc_net
, &ipmr_mfc_fops
))
3065 goto proc_cache_fail
;
3069 #ifdef CONFIG_PROC_FS
3071 remove_proc_entry("ip_mr_vif", net
->proc_net
);
3073 ipmr_rules_exit(net
);
3079 static void __net_exit
ipmr_net_exit(struct net
*net
)
3081 #ifdef CONFIG_PROC_FS
3082 remove_proc_entry("ip_mr_cache", net
->proc_net
);
3083 remove_proc_entry("ip_mr_vif", net
->proc_net
);
3085 ipmr_rules_exit(net
);
3088 static struct pernet_operations ipmr_net_ops
= {
3089 .init
= ipmr_net_init
,
3090 .exit
= ipmr_net_exit
,
3093 int __init
ip_mr_init(void)
3097 mrt_cachep
= kmem_cache_create("ip_mrt_cache",
3098 sizeof(struct mfc_cache
),
3099 0, SLAB_HWCACHE_ALIGN
| SLAB_PANIC
,
3102 err
= register_pernet_subsys(&ipmr_net_ops
);
3104 goto reg_pernet_fail
;
3106 err
= register_netdevice_notifier(&ip_mr_notifier
);
3108 goto reg_notif_fail
;
3109 #ifdef CONFIG_IP_PIMSM_V2
3110 if (inet_add_protocol(&pim_protocol
, IPPROTO_PIM
) < 0) {
3111 pr_err("%s: can't add PIM protocol\n", __func__
);
3113 goto add_proto_fail
;
3116 rtnl_register(RTNL_FAMILY_IPMR
, RTM_GETROUTE
,
3117 ipmr_rtm_getroute
, ipmr_rtm_dumproute
, NULL
);
3118 rtnl_register(RTNL_FAMILY_IPMR
, RTM_NEWROUTE
,
3119 ipmr_rtm_route
, NULL
, NULL
);
3120 rtnl_register(RTNL_FAMILY_IPMR
, RTM_DELROUTE
,
3121 ipmr_rtm_route
, NULL
, NULL
);
3123 rtnl_register(RTNL_FAMILY_IPMR
, RTM_GETLINK
,
3124 NULL
, ipmr_rtm_dumplink
, NULL
);
3127 #ifdef CONFIG_IP_PIMSM_V2
3129 unregister_netdevice_notifier(&ip_mr_notifier
);
3132 unregister_pernet_subsys(&ipmr_net_ops
);
3134 kmem_cache_destroy(mrt_cachep
);