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 <asm/system.h>
30 #include <asm/uaccess.h>
31 #include <linux/types.h>
32 #include <linux/capability.h>
33 #include <linux/errno.h>
34 #include <linux/timer.h>
36 #include <linux/kernel.h>
37 #include <linux/fcntl.h>
38 #include <linux/stat.h>
39 #include <linux/socket.h>
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/inetdevice.h>
44 #include <linux/igmp.h>
45 #include <linux/proc_fs.h>
46 #include <linux/seq_file.h>
47 #include <linux/mroute.h>
48 #include <linux/init.h>
49 #include <linux/if_ether.h>
50 #include <linux/slab.h>
51 #include <net/net_namespace.h>
53 #include <net/protocol.h>
54 #include <linux/skbuff.h>
55 #include <net/route.h>
60 #include <linux/notifier.h>
61 #include <linux/if_arp.h>
62 #include <linux/netfilter_ipv4.h>
63 #include <linux/compat.h>
64 #include <linux/export.h>
66 #include <net/checksum.h>
67 #include <net/netlink.h>
68 #include <net/fib_rules.h>
70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
71 #define CONFIG_IP_PIMSM 1
75 struct list_head list
;
80 struct sock __rcu
*mroute_sk
;
81 struct timer_list ipmr_expire_timer
;
82 struct list_head mfc_unres_queue
;
83 struct list_head mfc_cache_array
[MFC_LINES
];
84 struct vif_device vif_table
[MAXVIFS
];
86 atomic_t cache_resolve_queue_len
;
89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
90 int mroute_reg_vif_num
;
95 struct fib_rule common
;
102 /* Big lock, protecting vif table, mrt cache and mroute socket state.
103 * Note that the changes are semaphored via rtnl_lock.
106 static DEFINE_RWLOCK(mrt_lock
);
109 * Multicast router control variables
112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
114 /* Special spinlock for queue of unresolved entries */
115 static DEFINE_SPINLOCK(mfc_unres_lock
);
117 /* We return to original Alan's scheme. Hash table of resolved
118 * entries is changed only in process context and protected
119 * with weak lock mrt_lock. Queue of unresolved entries is protected
120 * with strong spinlock mfc_unres_lock.
122 * In this case data path is free of exclusive locks at all.
125 static struct kmem_cache
*mrt_cachep __read_mostly
;
127 static struct mr_table
*ipmr_new_table(struct net
*net
, u32 id
);
128 static int ip_mr_forward(struct net
*net
, struct mr_table
*mrt
,
129 struct sk_buff
*skb
, struct mfc_cache
*cache
,
131 static int ipmr_cache_report(struct mr_table
*mrt
,
132 struct sk_buff
*pkt
, vifi_t vifi
, int assert);
133 static int __ipmr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
134 struct mfc_cache
*c
, struct rtmsg
*rtm
);
135 static void ipmr_expire_process(unsigned long arg
);
137 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
138 #define ipmr_for_each_table(mrt, net) \
139 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
141 static struct mr_table
*ipmr_get_table(struct net
*net
, u32 id
)
143 struct mr_table
*mrt
;
145 ipmr_for_each_table(mrt
, net
) {
152 static int ipmr_fib_lookup(struct net
*net
, struct flowi4
*flp4
,
153 struct mr_table
**mrt
)
155 struct ipmr_result res
;
156 struct fib_lookup_arg arg
= { .result
= &res
, };
159 err
= fib_rules_lookup(net
->ipv4
.mr_rules_ops
,
160 flowi4_to_flowi(flp4
), 0, &arg
);
167 static int ipmr_rule_action(struct fib_rule
*rule
, struct flowi
*flp
,
168 int flags
, struct fib_lookup_arg
*arg
)
170 struct ipmr_result
*res
= arg
->result
;
171 struct mr_table
*mrt
;
173 switch (rule
->action
) {
176 case FR_ACT_UNREACHABLE
:
178 case FR_ACT_PROHIBIT
:
180 case FR_ACT_BLACKHOLE
:
185 mrt
= ipmr_get_table(rule
->fr_net
, rule
->table
);
192 static int ipmr_rule_match(struct fib_rule
*rule
, struct flowi
*fl
, int flags
)
197 static const struct nla_policy ipmr_rule_policy
[FRA_MAX
+ 1] = {
201 static int ipmr_rule_configure(struct fib_rule
*rule
, struct sk_buff
*skb
,
202 struct fib_rule_hdr
*frh
, struct nlattr
**tb
)
207 static int ipmr_rule_compare(struct fib_rule
*rule
, struct fib_rule_hdr
*frh
,
213 static int ipmr_rule_fill(struct fib_rule
*rule
, struct sk_buff
*skb
,
214 struct fib_rule_hdr
*frh
)
222 static const struct fib_rules_ops __net_initdata ipmr_rules_ops_template
= {
223 .family
= RTNL_FAMILY_IPMR
,
224 .rule_size
= sizeof(struct ipmr_rule
),
225 .addr_size
= sizeof(u32
),
226 .action
= ipmr_rule_action
,
227 .match
= ipmr_rule_match
,
228 .configure
= ipmr_rule_configure
,
229 .compare
= ipmr_rule_compare
,
230 .default_pref
= fib_default_rule_pref
,
231 .fill
= ipmr_rule_fill
,
232 .nlgroup
= RTNLGRP_IPV4_RULE
,
233 .policy
= ipmr_rule_policy
,
234 .owner
= THIS_MODULE
,
237 static int __net_init
ipmr_rules_init(struct net
*net
)
239 struct fib_rules_ops
*ops
;
240 struct mr_table
*mrt
;
243 ops
= fib_rules_register(&ipmr_rules_ops_template
, net
);
247 INIT_LIST_HEAD(&net
->ipv4
.mr_tables
);
249 mrt
= ipmr_new_table(net
, RT_TABLE_DEFAULT
);
255 err
= fib_default_rule_add(ops
, 0x7fff, RT_TABLE_DEFAULT
, 0);
259 net
->ipv4
.mr_rules_ops
= ops
;
265 fib_rules_unregister(ops
);
269 static void __net_exit
ipmr_rules_exit(struct net
*net
)
271 struct mr_table
*mrt
, *next
;
273 list_for_each_entry_safe(mrt
, next
, &net
->ipv4
.mr_tables
, list
) {
274 list_del(&mrt
->list
);
277 fib_rules_unregister(net
->ipv4
.mr_rules_ops
);
280 #define ipmr_for_each_table(mrt, net) \
281 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
283 static struct mr_table
*ipmr_get_table(struct net
*net
, u32 id
)
285 return net
->ipv4
.mrt
;
288 static int ipmr_fib_lookup(struct net
*net
, struct flowi4
*flp4
,
289 struct mr_table
**mrt
)
291 *mrt
= net
->ipv4
.mrt
;
295 static int __net_init
ipmr_rules_init(struct net
*net
)
297 net
->ipv4
.mrt
= ipmr_new_table(net
, RT_TABLE_DEFAULT
);
298 return net
->ipv4
.mrt
? 0 : -ENOMEM
;
301 static void __net_exit
ipmr_rules_exit(struct net
*net
)
303 kfree(net
->ipv4
.mrt
);
307 static struct mr_table
*ipmr_new_table(struct net
*net
, u32 id
)
309 struct mr_table
*mrt
;
312 mrt
= ipmr_get_table(net
, id
);
316 mrt
= kzalloc(sizeof(*mrt
), GFP_KERNEL
);
319 write_pnet(&mrt
->net
, net
);
322 /* Forwarding cache */
323 for (i
= 0; i
< MFC_LINES
; i
++)
324 INIT_LIST_HEAD(&mrt
->mfc_cache_array
[i
]);
326 INIT_LIST_HEAD(&mrt
->mfc_unres_queue
);
328 setup_timer(&mrt
->ipmr_expire_timer
, ipmr_expire_process
,
331 #ifdef CONFIG_IP_PIMSM
332 mrt
->mroute_reg_vif_num
= -1;
334 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
335 list_add_tail_rcu(&mrt
->list
, &net
->ipv4
.mr_tables
);
340 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
342 static void ipmr_del_tunnel(struct net_device
*dev
, struct vifctl
*v
)
344 struct net
*net
= dev_net(dev
);
348 dev
= __dev_get_by_name(net
, "tunl0");
350 const struct net_device_ops
*ops
= dev
->netdev_ops
;
352 struct ip_tunnel_parm p
;
354 memset(&p
, 0, sizeof(p
));
355 p
.iph
.daddr
= v
->vifc_rmt_addr
.s_addr
;
356 p
.iph
.saddr
= v
->vifc_lcl_addr
.s_addr
;
359 p
.iph
.protocol
= IPPROTO_IPIP
;
360 sprintf(p
.name
, "dvmrp%d", v
->vifc_vifi
);
361 ifr
.ifr_ifru
.ifru_data
= (__force
void __user
*)&p
;
363 if (ops
->ndo_do_ioctl
) {
364 mm_segment_t oldfs
= get_fs();
367 ops
->ndo_do_ioctl(dev
, &ifr
, SIOCDELTUNNEL
);
374 struct net_device
*ipmr_new_tunnel(struct net
*net
, struct vifctl
*v
)
376 struct net_device
*dev
;
378 dev
= __dev_get_by_name(net
, "tunl0");
381 const struct net_device_ops
*ops
= dev
->netdev_ops
;
384 struct ip_tunnel_parm p
;
385 struct in_device
*in_dev
;
387 memset(&p
, 0, sizeof(p
));
388 p
.iph
.daddr
= v
->vifc_rmt_addr
.s_addr
;
389 p
.iph
.saddr
= v
->vifc_lcl_addr
.s_addr
;
392 p
.iph
.protocol
= IPPROTO_IPIP
;
393 sprintf(p
.name
, "dvmrp%d", v
->vifc_vifi
);
394 ifr
.ifr_ifru
.ifru_data
= (__force
void __user
*)&p
;
396 if (ops
->ndo_do_ioctl
) {
397 mm_segment_t oldfs
= get_fs();
400 err
= ops
->ndo_do_ioctl(dev
, &ifr
, SIOCADDTUNNEL
);
408 (dev
= __dev_get_by_name(net
, p
.name
)) != NULL
) {
409 dev
->flags
|= IFF_MULTICAST
;
411 in_dev
= __in_dev_get_rtnl(dev
);
415 ipv4_devconf_setall(in_dev
);
416 IPV4_DEVCONF(in_dev
->cnf
, RP_FILTER
) = 0;
426 /* allow the register to be completed before unregistering. */
430 unregister_netdevice(dev
);
434 #ifdef CONFIG_IP_PIMSM
436 static netdev_tx_t
reg_vif_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
438 struct net
*net
= dev_net(dev
);
439 struct mr_table
*mrt
;
440 struct flowi4 fl4
= {
441 .flowi4_oif
= dev
->ifindex
,
442 .flowi4_iif
= skb
->skb_iif
,
443 .flowi4_mark
= skb
->mark
,
447 err
= ipmr_fib_lookup(net
, &fl4
, &mrt
);
453 read_lock(&mrt_lock
);
454 dev
->stats
.tx_bytes
+= skb
->len
;
455 dev
->stats
.tx_packets
++;
456 ipmr_cache_report(mrt
, skb
, mrt
->mroute_reg_vif_num
, IGMPMSG_WHOLEPKT
);
457 read_unlock(&mrt_lock
);
462 static const struct net_device_ops reg_vif_netdev_ops
= {
463 .ndo_start_xmit
= reg_vif_xmit
,
466 static void reg_vif_setup(struct net_device
*dev
)
468 dev
->type
= ARPHRD_PIMREG
;
469 dev
->mtu
= ETH_DATA_LEN
- sizeof(struct iphdr
) - 8;
470 dev
->flags
= IFF_NOARP
;
471 dev
->netdev_ops
= ®_vif_netdev_ops
,
472 dev
->destructor
= free_netdev
;
473 dev
->features
|= NETIF_F_NETNS_LOCAL
;
476 static struct net_device
*ipmr_reg_vif(struct net
*net
, struct mr_table
*mrt
)
478 struct net_device
*dev
;
479 struct in_device
*in_dev
;
482 if (mrt
->id
== RT_TABLE_DEFAULT
)
483 sprintf(name
, "pimreg");
485 sprintf(name
, "pimreg%u", mrt
->id
);
487 dev
= alloc_netdev(0, name
, reg_vif_setup
);
492 dev_net_set(dev
, net
);
494 if (register_netdevice(dev
)) {
501 in_dev
= __in_dev_get_rcu(dev
);
507 ipv4_devconf_setall(in_dev
);
508 IPV4_DEVCONF(in_dev
->cnf
, RP_FILTER
) = 0;
519 /* allow the register to be completed before unregistering. */
523 unregister_netdevice(dev
);
530 * @notify: Set to 1, if the caller is a notifier_call
533 static int vif_delete(struct mr_table
*mrt
, int vifi
, int notify
,
534 struct list_head
*head
)
536 struct vif_device
*v
;
537 struct net_device
*dev
;
538 struct in_device
*in_dev
;
540 if (vifi
< 0 || vifi
>= mrt
->maxvif
)
541 return -EADDRNOTAVAIL
;
543 v
= &mrt
->vif_table
[vifi
];
545 write_lock_bh(&mrt_lock
);
550 write_unlock_bh(&mrt_lock
);
551 return -EADDRNOTAVAIL
;
554 #ifdef CONFIG_IP_PIMSM
555 if (vifi
== mrt
->mroute_reg_vif_num
)
556 mrt
->mroute_reg_vif_num
= -1;
559 if (vifi
+ 1 == mrt
->maxvif
) {
562 for (tmp
= vifi
- 1; tmp
>= 0; tmp
--) {
563 if (VIF_EXISTS(mrt
, tmp
))
569 write_unlock_bh(&mrt_lock
);
571 dev_set_allmulti(dev
, -1);
573 in_dev
= __in_dev_get_rtnl(dev
);
575 IPV4_DEVCONF(in_dev
->cnf
, MC_FORWARDING
)--;
576 ip_rt_multicast_event(in_dev
);
579 if (v
->flags
& (VIFF_TUNNEL
| VIFF_REGISTER
) && !notify
)
580 unregister_netdevice_queue(dev
, head
);
586 static void ipmr_cache_free_rcu(struct rcu_head
*head
)
588 struct mfc_cache
*c
= container_of(head
, struct mfc_cache
, rcu
);
590 kmem_cache_free(mrt_cachep
, c
);
593 static inline void ipmr_cache_free(struct mfc_cache
*c
)
595 call_rcu(&c
->rcu
, ipmr_cache_free_rcu
);
598 /* Destroy an unresolved cache entry, killing queued skbs
599 * and reporting error to netlink readers.
602 static void ipmr_destroy_unres(struct mr_table
*mrt
, struct mfc_cache
*c
)
604 struct net
*net
= read_pnet(&mrt
->net
);
608 atomic_dec(&mrt
->cache_resolve_queue_len
);
610 while ((skb
= skb_dequeue(&c
->mfc_un
.unres
.unresolved
))) {
611 if (ip_hdr(skb
)->version
== 0) {
612 struct nlmsghdr
*nlh
= (struct nlmsghdr
*)skb_pull(skb
, sizeof(struct iphdr
));
613 nlh
->nlmsg_type
= NLMSG_ERROR
;
614 nlh
->nlmsg_len
= NLMSG_LENGTH(sizeof(struct nlmsgerr
));
615 skb_trim(skb
, nlh
->nlmsg_len
);
617 e
->error
= -ETIMEDOUT
;
618 memset(&e
->msg
, 0, sizeof(e
->msg
));
620 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).pid
);
630 /* Timer process for the unresolved queue. */
632 static void ipmr_expire_process(unsigned long arg
)
634 struct mr_table
*mrt
= (struct mr_table
*)arg
;
636 unsigned long expires
;
637 struct mfc_cache
*c
, *next
;
639 if (!spin_trylock(&mfc_unres_lock
)) {
640 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+HZ
/10);
644 if (list_empty(&mrt
->mfc_unres_queue
))
650 list_for_each_entry_safe(c
, next
, &mrt
->mfc_unres_queue
, list
) {
651 if (time_after(c
->mfc_un
.unres
.expires
, now
)) {
652 unsigned long interval
= c
->mfc_un
.unres
.expires
- now
;
653 if (interval
< expires
)
659 ipmr_destroy_unres(mrt
, c
);
662 if (!list_empty(&mrt
->mfc_unres_queue
))
663 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+ expires
);
666 spin_unlock(&mfc_unres_lock
);
669 /* Fill oifs list. It is called under write locked mrt_lock. */
671 static void ipmr_update_thresholds(struct mr_table
*mrt
, struct mfc_cache
*cache
,
676 cache
->mfc_un
.res
.minvif
= MAXVIFS
;
677 cache
->mfc_un
.res
.maxvif
= 0;
678 memset(cache
->mfc_un
.res
.ttls
, 255, MAXVIFS
);
680 for (vifi
= 0; vifi
< mrt
->maxvif
; vifi
++) {
681 if (VIF_EXISTS(mrt
, vifi
) &&
682 ttls
[vifi
] && ttls
[vifi
] < 255) {
683 cache
->mfc_un
.res
.ttls
[vifi
] = ttls
[vifi
];
684 if (cache
->mfc_un
.res
.minvif
> vifi
)
685 cache
->mfc_un
.res
.minvif
= vifi
;
686 if (cache
->mfc_un
.res
.maxvif
<= vifi
)
687 cache
->mfc_un
.res
.maxvif
= vifi
+ 1;
692 static int vif_add(struct net
*net
, struct mr_table
*mrt
,
693 struct vifctl
*vifc
, int mrtsock
)
695 int vifi
= vifc
->vifc_vifi
;
696 struct vif_device
*v
= &mrt
->vif_table
[vifi
];
697 struct net_device
*dev
;
698 struct in_device
*in_dev
;
702 if (VIF_EXISTS(mrt
, vifi
))
705 switch (vifc
->vifc_flags
) {
706 #ifdef CONFIG_IP_PIMSM
709 * Special Purpose VIF in PIM
710 * All the packets will be sent to the daemon
712 if (mrt
->mroute_reg_vif_num
>= 0)
714 dev
= ipmr_reg_vif(net
, mrt
);
717 err
= dev_set_allmulti(dev
, 1);
719 unregister_netdevice(dev
);
726 dev
= ipmr_new_tunnel(net
, vifc
);
729 err
= dev_set_allmulti(dev
, 1);
731 ipmr_del_tunnel(dev
, vifc
);
737 case VIFF_USE_IFINDEX
:
739 if (vifc
->vifc_flags
== VIFF_USE_IFINDEX
) {
740 dev
= dev_get_by_index(net
, vifc
->vifc_lcl_ifindex
);
741 if (dev
&& __in_dev_get_rtnl(dev
) == NULL
) {
743 return -EADDRNOTAVAIL
;
746 dev
= ip_dev_find(net
, vifc
->vifc_lcl_addr
.s_addr
);
749 return -EADDRNOTAVAIL
;
750 err
= dev_set_allmulti(dev
, 1);
760 in_dev
= __in_dev_get_rtnl(dev
);
763 return -EADDRNOTAVAIL
;
765 IPV4_DEVCONF(in_dev
->cnf
, MC_FORWARDING
)++;
766 ip_rt_multicast_event(in_dev
);
768 /* Fill in the VIF structures */
770 v
->rate_limit
= vifc
->vifc_rate_limit
;
771 v
->local
= vifc
->vifc_lcl_addr
.s_addr
;
772 v
->remote
= vifc
->vifc_rmt_addr
.s_addr
;
773 v
->flags
= vifc
->vifc_flags
;
775 v
->flags
|= VIFF_STATIC
;
776 v
->threshold
= vifc
->vifc_threshold
;
781 v
->link
= dev
->ifindex
;
782 if (v
->flags
& (VIFF_TUNNEL
| VIFF_REGISTER
))
783 v
->link
= dev
->iflink
;
785 /* And finish update writing critical data */
786 write_lock_bh(&mrt_lock
);
788 #ifdef CONFIG_IP_PIMSM
789 if (v
->flags
& VIFF_REGISTER
)
790 mrt
->mroute_reg_vif_num
= vifi
;
792 if (vifi
+1 > mrt
->maxvif
)
793 mrt
->maxvif
= vifi
+1;
794 write_unlock_bh(&mrt_lock
);
798 /* called with rcu_read_lock() */
799 static struct mfc_cache
*ipmr_cache_find(struct mr_table
*mrt
,
803 int line
= MFC_HASH(mcastgrp
, origin
);
806 list_for_each_entry_rcu(c
, &mrt
->mfc_cache_array
[line
], list
) {
807 if (c
->mfc_origin
== origin
&& c
->mfc_mcastgrp
== mcastgrp
)
814 * Allocate a multicast cache entry
816 static struct mfc_cache
*ipmr_cache_alloc(void)
818 struct mfc_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_KERNEL
);
821 c
->mfc_un
.res
.minvif
= MAXVIFS
;
825 static struct mfc_cache
*ipmr_cache_alloc_unres(void)
827 struct mfc_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_ATOMIC
);
830 skb_queue_head_init(&c
->mfc_un
.unres
.unresolved
);
831 c
->mfc_un
.unres
.expires
= jiffies
+ 10*HZ
;
837 * A cache entry has gone into a resolved state from queued
840 static void ipmr_cache_resolve(struct net
*net
, struct mr_table
*mrt
,
841 struct mfc_cache
*uc
, struct mfc_cache
*c
)
846 /* Play the pending entries through our router */
848 while ((skb
= __skb_dequeue(&uc
->mfc_un
.unres
.unresolved
))) {
849 if (ip_hdr(skb
)->version
== 0) {
850 struct nlmsghdr
*nlh
= (struct nlmsghdr
*)skb_pull(skb
, sizeof(struct iphdr
));
852 if (__ipmr_fill_mroute(mrt
, skb
, c
, NLMSG_DATA(nlh
)) > 0) {
853 nlh
->nlmsg_len
= skb_tail_pointer(skb
) -
856 nlh
->nlmsg_type
= NLMSG_ERROR
;
857 nlh
->nlmsg_len
= NLMSG_LENGTH(sizeof(struct nlmsgerr
));
858 skb_trim(skb
, nlh
->nlmsg_len
);
860 e
->error
= -EMSGSIZE
;
861 memset(&e
->msg
, 0, sizeof(e
->msg
));
864 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).pid
);
866 ip_mr_forward(net
, mrt
, skb
, c
, 0);
872 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
873 * expects the following bizarre scheme.
875 * Called under mrt_lock.
878 static int ipmr_cache_report(struct mr_table
*mrt
,
879 struct sk_buff
*pkt
, vifi_t vifi
, int assert)
882 const int ihl
= ip_hdrlen(pkt
);
883 struct igmphdr
*igmp
;
885 struct sock
*mroute_sk
;
888 #ifdef CONFIG_IP_PIMSM
889 if (assert == IGMPMSG_WHOLEPKT
)
890 skb
= skb_realloc_headroom(pkt
, sizeof(struct iphdr
));
893 skb
= alloc_skb(128, GFP_ATOMIC
);
898 #ifdef CONFIG_IP_PIMSM
899 if (assert == IGMPMSG_WHOLEPKT
) {
900 /* Ugly, but we have no choice with this interface.
901 * Duplicate old header, fix ihl, length etc.
902 * And all this only to mangle msg->im_msgtype and
903 * to set msg->im_mbz to "mbz" :-)
905 skb_push(skb
, sizeof(struct iphdr
));
906 skb_reset_network_header(skb
);
907 skb_reset_transport_header(skb
);
908 msg
= (struct igmpmsg
*)skb_network_header(skb
);
909 memcpy(msg
, skb_network_header(pkt
), sizeof(struct iphdr
));
910 msg
->im_msgtype
= IGMPMSG_WHOLEPKT
;
912 msg
->im_vif
= mrt
->mroute_reg_vif_num
;
913 ip_hdr(skb
)->ihl
= sizeof(struct iphdr
) >> 2;
914 ip_hdr(skb
)->tot_len
= htons(ntohs(ip_hdr(pkt
)->tot_len
) +
915 sizeof(struct iphdr
));
920 /* Copy the IP header */
922 skb
->network_header
= skb
->tail
;
924 skb_copy_to_linear_data(skb
, pkt
->data
, ihl
);
925 ip_hdr(skb
)->protocol
= 0; /* Flag to the kernel this is a route add */
926 msg
= (struct igmpmsg
*)skb_network_header(skb
);
928 skb_dst_set(skb
, dst_clone(skb_dst(pkt
)));
932 igmp
= (struct igmphdr
*)skb_put(skb
, sizeof(struct igmphdr
));
934 msg
->im_msgtype
= assert;
936 ip_hdr(skb
)->tot_len
= htons(skb
->len
); /* Fix the length */
937 skb
->transport_header
= skb
->network_header
;
941 mroute_sk
= rcu_dereference(mrt
->mroute_sk
);
942 if (mroute_sk
== NULL
) {
948 /* Deliver to mrouted */
950 ret
= sock_queue_rcv_skb(mroute_sk
, skb
);
954 printk(KERN_WARNING
"mroute: pending queue full, dropping entries.\n");
962 * Queue a packet for resolution. It gets locked cache entry!
966 ipmr_cache_unresolved(struct mr_table
*mrt
, vifi_t vifi
, struct sk_buff
*skb
)
971 const struct iphdr
*iph
= ip_hdr(skb
);
973 spin_lock_bh(&mfc_unres_lock
);
974 list_for_each_entry(c
, &mrt
->mfc_unres_queue
, list
) {
975 if (c
->mfc_mcastgrp
== iph
->daddr
&&
976 c
->mfc_origin
== iph
->saddr
) {
983 /* Create a new entry if allowable */
985 if (atomic_read(&mrt
->cache_resolve_queue_len
) >= 10 ||
986 (c
= ipmr_cache_alloc_unres()) == NULL
) {
987 spin_unlock_bh(&mfc_unres_lock
);
993 /* Fill in the new cache entry */
996 c
->mfc_origin
= iph
->saddr
;
997 c
->mfc_mcastgrp
= iph
->daddr
;
999 /* Reflect first query at mrouted. */
1001 err
= ipmr_cache_report(mrt
, skb
, vifi
, IGMPMSG_NOCACHE
);
1003 /* If the report failed throw the cache entry
1006 spin_unlock_bh(&mfc_unres_lock
);
1013 atomic_inc(&mrt
->cache_resolve_queue_len
);
1014 list_add(&c
->list
, &mrt
->mfc_unres_queue
);
1016 if (atomic_read(&mrt
->cache_resolve_queue_len
) == 1)
1017 mod_timer(&mrt
->ipmr_expire_timer
, c
->mfc_un
.unres
.expires
);
1020 /* See if we can append the packet */
1022 if (c
->mfc_un
.unres
.unresolved
.qlen
> 3) {
1026 skb_queue_tail(&c
->mfc_un
.unres
.unresolved
, skb
);
1030 spin_unlock_bh(&mfc_unres_lock
);
1035 * MFC cache manipulation by user space mroute daemon
1038 static int ipmr_mfc_delete(struct mr_table
*mrt
, struct mfcctl
*mfc
)
1041 struct mfc_cache
*c
, *next
;
1043 line
= MFC_HASH(mfc
->mfcc_mcastgrp
.s_addr
, mfc
->mfcc_origin
.s_addr
);
1045 list_for_each_entry_safe(c
, next
, &mrt
->mfc_cache_array
[line
], list
) {
1046 if (c
->mfc_origin
== mfc
->mfcc_origin
.s_addr
&&
1047 c
->mfc_mcastgrp
== mfc
->mfcc_mcastgrp
.s_addr
) {
1048 list_del_rcu(&c
->list
);
1057 static int ipmr_mfc_add(struct net
*net
, struct mr_table
*mrt
,
1058 struct mfcctl
*mfc
, int mrtsock
)
1062 struct mfc_cache
*uc
, *c
;
1064 if (mfc
->mfcc_parent
>= MAXVIFS
)
1067 line
= MFC_HASH(mfc
->mfcc_mcastgrp
.s_addr
, mfc
->mfcc_origin
.s_addr
);
1069 list_for_each_entry(c
, &mrt
->mfc_cache_array
[line
], list
) {
1070 if (c
->mfc_origin
== mfc
->mfcc_origin
.s_addr
&&
1071 c
->mfc_mcastgrp
== mfc
->mfcc_mcastgrp
.s_addr
) {
1078 write_lock_bh(&mrt_lock
);
1079 c
->mfc_parent
= mfc
->mfcc_parent
;
1080 ipmr_update_thresholds(mrt
, c
, mfc
->mfcc_ttls
);
1082 c
->mfc_flags
|= MFC_STATIC
;
1083 write_unlock_bh(&mrt_lock
);
1087 if (!ipv4_is_multicast(mfc
->mfcc_mcastgrp
.s_addr
))
1090 c
= ipmr_cache_alloc();
1094 c
->mfc_origin
= mfc
->mfcc_origin
.s_addr
;
1095 c
->mfc_mcastgrp
= mfc
->mfcc_mcastgrp
.s_addr
;
1096 c
->mfc_parent
= mfc
->mfcc_parent
;
1097 ipmr_update_thresholds(mrt
, c
, mfc
->mfcc_ttls
);
1099 c
->mfc_flags
|= MFC_STATIC
;
1101 list_add_rcu(&c
->list
, &mrt
->mfc_cache_array
[line
]);
1104 * Check to see if we resolved a queued list. If so we
1105 * need to send on the frames and tidy up.
1108 spin_lock_bh(&mfc_unres_lock
);
1109 list_for_each_entry(uc
, &mrt
->mfc_unres_queue
, list
) {
1110 if (uc
->mfc_origin
== c
->mfc_origin
&&
1111 uc
->mfc_mcastgrp
== c
->mfc_mcastgrp
) {
1112 list_del(&uc
->list
);
1113 atomic_dec(&mrt
->cache_resolve_queue_len
);
1118 if (list_empty(&mrt
->mfc_unres_queue
))
1119 del_timer(&mrt
->ipmr_expire_timer
);
1120 spin_unlock_bh(&mfc_unres_lock
);
1123 ipmr_cache_resolve(net
, mrt
, uc
, c
);
1124 ipmr_cache_free(uc
);
1130 * Close the multicast socket, and clear the vif tables etc
1133 static void mroute_clean_tables(struct mr_table
*mrt
)
1137 struct mfc_cache
*c
, *next
;
1139 /* Shut down all active vif entries */
1141 for (i
= 0; i
< mrt
->maxvif
; i
++) {
1142 if (!(mrt
->vif_table
[i
].flags
& VIFF_STATIC
))
1143 vif_delete(mrt
, i
, 0, &list
);
1145 unregister_netdevice_many(&list
);
1147 /* Wipe the cache */
1149 for (i
= 0; i
< MFC_LINES
; i
++) {
1150 list_for_each_entry_safe(c
, next
, &mrt
->mfc_cache_array
[i
], list
) {
1151 if (c
->mfc_flags
& MFC_STATIC
)
1153 list_del_rcu(&c
->list
);
1158 if (atomic_read(&mrt
->cache_resolve_queue_len
) != 0) {
1159 spin_lock_bh(&mfc_unres_lock
);
1160 list_for_each_entry_safe(c
, next
, &mrt
->mfc_unres_queue
, list
) {
1162 ipmr_destroy_unres(mrt
, c
);
1164 spin_unlock_bh(&mfc_unres_lock
);
1168 /* called from ip_ra_control(), before an RCU grace period,
1169 * we dont need to call synchronize_rcu() here
1171 static void mrtsock_destruct(struct sock
*sk
)
1173 struct net
*net
= sock_net(sk
);
1174 struct mr_table
*mrt
;
1177 ipmr_for_each_table(mrt
, net
) {
1178 if (sk
== rtnl_dereference(mrt
->mroute_sk
)) {
1179 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)--;
1180 RCU_INIT_POINTER(mrt
->mroute_sk
, NULL
);
1181 mroute_clean_tables(mrt
);
1188 * Socket options and virtual interface manipulation. The whole
1189 * virtual interface system is a complete heap, but unfortunately
1190 * that's how BSD mrouted happens to think. Maybe one day with a proper
1191 * MOSPF/PIM router set up we can clean this up.
1194 int ip_mroute_setsockopt(struct sock
*sk
, int optname
, char __user
*optval
, unsigned int optlen
)
1199 struct net
*net
= sock_net(sk
);
1200 struct mr_table
*mrt
;
1202 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1206 if (optname
!= MRT_INIT
) {
1207 if (sk
!= rcu_access_pointer(mrt
->mroute_sk
) &&
1208 !capable(CAP_NET_ADMIN
))
1214 if (sk
->sk_type
!= SOCK_RAW
||
1215 inet_sk(sk
)->inet_num
!= IPPROTO_IGMP
)
1217 if (optlen
!= sizeof(int))
1218 return -ENOPROTOOPT
;
1221 if (rtnl_dereference(mrt
->mroute_sk
)) {
1226 ret
= ip_ra_control(sk
, 1, mrtsock_destruct
);
1228 RCU_INIT_POINTER(mrt
->mroute_sk
, sk
);
1229 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)++;
1234 if (sk
!= rcu_access_pointer(mrt
->mroute_sk
))
1236 return ip_ra_control(sk
, 0, NULL
);
1239 if (optlen
!= sizeof(vif
))
1241 if (copy_from_user(&vif
, optval
, sizeof(vif
)))
1243 if (vif
.vifc_vifi
>= MAXVIFS
)
1246 if (optname
== MRT_ADD_VIF
) {
1247 ret
= vif_add(net
, mrt
, &vif
,
1248 sk
== rtnl_dereference(mrt
->mroute_sk
));
1250 ret
= vif_delete(mrt
, vif
.vifc_vifi
, 0, NULL
);
1256 * Manipulate the forwarding caches. These live
1257 * in a sort of kernel/user symbiosis.
1261 if (optlen
!= sizeof(mfc
))
1263 if (copy_from_user(&mfc
, optval
, sizeof(mfc
)))
1266 if (optname
== MRT_DEL_MFC
)
1267 ret
= ipmr_mfc_delete(mrt
, &mfc
);
1269 ret
= ipmr_mfc_add(net
, mrt
, &mfc
,
1270 sk
== rtnl_dereference(mrt
->mroute_sk
));
1274 * Control PIM assert.
1279 if (get_user(v
, (int __user
*)optval
))
1281 mrt
->mroute_do_assert
= (v
) ? 1 : 0;
1284 #ifdef CONFIG_IP_PIMSM
1289 if (get_user(v
, (int __user
*)optval
))
1295 if (v
!= mrt
->mroute_do_pim
) {
1296 mrt
->mroute_do_pim
= v
;
1297 mrt
->mroute_do_assert
= v
;
1303 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1308 if (optlen
!= sizeof(u32
))
1310 if (get_user(v
, (u32 __user
*)optval
))
1315 if (sk
== rtnl_dereference(mrt
->mroute_sk
)) {
1318 if (!ipmr_new_table(net
, v
))
1320 raw_sk(sk
)->ipmr_table
= v
;
1327 * Spurious command, or MRT_VERSION which you cannot
1331 return -ENOPROTOOPT
;
1336 * Getsock opt support for the multicast routing system.
1339 int ip_mroute_getsockopt(struct sock
*sk
, int optname
, char __user
*optval
, int __user
*optlen
)
1343 struct net
*net
= sock_net(sk
);
1344 struct mr_table
*mrt
;
1346 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1350 if (optname
!= MRT_VERSION
&&
1351 #ifdef CONFIG_IP_PIMSM
1352 optname
!= MRT_PIM
&&
1354 optname
!= MRT_ASSERT
)
1355 return -ENOPROTOOPT
;
1357 if (get_user(olr
, optlen
))
1360 olr
= min_t(unsigned int, olr
, sizeof(int));
1364 if (put_user(olr
, optlen
))
1366 if (optname
== MRT_VERSION
)
1368 #ifdef CONFIG_IP_PIMSM
1369 else if (optname
== MRT_PIM
)
1370 val
= mrt
->mroute_do_pim
;
1373 val
= mrt
->mroute_do_assert
;
1374 if (copy_to_user(optval
, &val
, olr
))
1380 * The IP multicast ioctl support routines.
1383 int ipmr_ioctl(struct sock
*sk
, int cmd
, void __user
*arg
)
1385 struct sioc_sg_req sr
;
1386 struct sioc_vif_req vr
;
1387 struct vif_device
*vif
;
1388 struct mfc_cache
*c
;
1389 struct net
*net
= sock_net(sk
);
1390 struct mr_table
*mrt
;
1392 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1398 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1400 if (vr
.vifi
>= mrt
->maxvif
)
1402 read_lock(&mrt_lock
);
1403 vif
= &mrt
->vif_table
[vr
.vifi
];
1404 if (VIF_EXISTS(mrt
, vr
.vifi
)) {
1405 vr
.icount
= vif
->pkt_in
;
1406 vr
.ocount
= vif
->pkt_out
;
1407 vr
.ibytes
= vif
->bytes_in
;
1408 vr
.obytes
= vif
->bytes_out
;
1409 read_unlock(&mrt_lock
);
1411 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1415 read_unlock(&mrt_lock
);
1416 return -EADDRNOTAVAIL
;
1418 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1422 c
= ipmr_cache_find(mrt
, sr
.src
.s_addr
, sr
.grp
.s_addr
);
1424 sr
.pktcnt
= c
->mfc_un
.res
.pkt
;
1425 sr
.bytecnt
= c
->mfc_un
.res
.bytes
;
1426 sr
.wrong_if
= c
->mfc_un
.res
.wrong_if
;
1429 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1434 return -EADDRNOTAVAIL
;
1436 return -ENOIOCTLCMD
;
1440 #ifdef CONFIG_COMPAT
1441 struct compat_sioc_sg_req
{
1444 compat_ulong_t pktcnt
;
1445 compat_ulong_t bytecnt
;
1446 compat_ulong_t wrong_if
;
1449 struct compat_sioc_vif_req
{
1450 vifi_t vifi
; /* Which iface */
1451 compat_ulong_t icount
;
1452 compat_ulong_t ocount
;
1453 compat_ulong_t ibytes
;
1454 compat_ulong_t obytes
;
1457 int ipmr_compat_ioctl(struct sock
*sk
, unsigned int cmd
, void __user
*arg
)
1459 struct compat_sioc_sg_req sr
;
1460 struct compat_sioc_vif_req vr
;
1461 struct vif_device
*vif
;
1462 struct mfc_cache
*c
;
1463 struct net
*net
= sock_net(sk
);
1464 struct mr_table
*mrt
;
1466 mrt
= ipmr_get_table(net
, raw_sk(sk
)->ipmr_table
? : RT_TABLE_DEFAULT
);
1472 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1474 if (vr
.vifi
>= mrt
->maxvif
)
1476 read_lock(&mrt_lock
);
1477 vif
= &mrt
->vif_table
[vr
.vifi
];
1478 if (VIF_EXISTS(mrt
, vr
.vifi
)) {
1479 vr
.icount
= vif
->pkt_in
;
1480 vr
.ocount
= vif
->pkt_out
;
1481 vr
.ibytes
= vif
->bytes_in
;
1482 vr
.obytes
= vif
->bytes_out
;
1483 read_unlock(&mrt_lock
);
1485 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1489 read_unlock(&mrt_lock
);
1490 return -EADDRNOTAVAIL
;
1492 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1496 c
= ipmr_cache_find(mrt
, sr
.src
.s_addr
, sr
.grp
.s_addr
);
1498 sr
.pktcnt
= c
->mfc_un
.res
.pkt
;
1499 sr
.bytecnt
= c
->mfc_un
.res
.bytes
;
1500 sr
.wrong_if
= c
->mfc_un
.res
.wrong_if
;
1503 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1508 return -EADDRNOTAVAIL
;
1510 return -ENOIOCTLCMD
;
1516 static int ipmr_device_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1518 struct net_device
*dev
= ptr
;
1519 struct net
*net
= dev_net(dev
);
1520 struct mr_table
*mrt
;
1521 struct vif_device
*v
;
1525 if (event
!= NETDEV_UNREGISTER
)
1528 ipmr_for_each_table(mrt
, net
) {
1529 v
= &mrt
->vif_table
[0];
1530 for (ct
= 0; ct
< mrt
->maxvif
; ct
++, v
++) {
1532 vif_delete(mrt
, ct
, 1, &list
);
1535 unregister_netdevice_many(&list
);
1540 static struct notifier_block ip_mr_notifier
= {
1541 .notifier_call
= ipmr_device_event
,
1545 * Encapsulate a packet by attaching a valid IPIP header to it.
1546 * This avoids tunnel drivers and other mess and gives us the speed so
1547 * important for multicast video.
1550 static void ip_encap(struct sk_buff
*skb
, __be32 saddr
, __be32 daddr
)
1553 const struct iphdr
*old_iph
= ip_hdr(skb
);
1555 skb_push(skb
, sizeof(struct iphdr
));
1556 skb
->transport_header
= skb
->network_header
;
1557 skb_reset_network_header(skb
);
1561 iph
->tos
= old_iph
->tos
;
1562 iph
->ttl
= old_iph
->ttl
;
1566 iph
->protocol
= IPPROTO_IPIP
;
1568 iph
->tot_len
= htons(skb
->len
);
1569 ip_select_ident(iph
, skb_dst(skb
), NULL
);
1572 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
1576 static inline int ipmr_forward_finish(struct sk_buff
*skb
)
1578 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
1580 IP_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), IPSTATS_MIB_OUTFORWDATAGRAMS
);
1582 if (unlikely(opt
->optlen
))
1583 ip_forward_options(skb
);
1585 return dst_output(skb
);
1589 * Processing handlers for ipmr_forward
1592 static void ipmr_queue_xmit(struct net
*net
, struct mr_table
*mrt
,
1593 struct sk_buff
*skb
, struct mfc_cache
*c
, int vifi
)
1595 const struct iphdr
*iph
= ip_hdr(skb
);
1596 struct vif_device
*vif
= &mrt
->vif_table
[vifi
];
1597 struct net_device
*dev
;
1602 if (vif
->dev
== NULL
)
1605 #ifdef CONFIG_IP_PIMSM
1606 if (vif
->flags
& VIFF_REGISTER
) {
1608 vif
->bytes_out
+= skb
->len
;
1609 vif
->dev
->stats
.tx_bytes
+= skb
->len
;
1610 vif
->dev
->stats
.tx_packets
++;
1611 ipmr_cache_report(mrt
, skb
, vifi
, IGMPMSG_WHOLEPKT
);
1616 if (vif
->flags
& VIFF_TUNNEL
) {
1617 rt
= ip_route_output_ports(net
, &fl4
, NULL
,
1618 vif
->remote
, vif
->local
,
1621 RT_TOS(iph
->tos
), vif
->link
);
1624 encap
= sizeof(struct iphdr
);
1626 rt
= ip_route_output_ports(net
, &fl4
, NULL
, iph
->daddr
, 0,
1629 RT_TOS(iph
->tos
), vif
->link
);
1636 if (skb
->len
+encap
> dst_mtu(&rt
->dst
) && (ntohs(iph
->frag_off
) & IP_DF
)) {
1637 /* Do not fragment multicasts. Alas, IPv4 does not
1638 * allow to send ICMP, so that packets will disappear
1642 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_FRAGFAILS
);
1647 encap
+= LL_RESERVED_SPACE(dev
) + rt
->dst
.header_len
;
1649 if (skb_cow(skb
, encap
)) {
1655 vif
->bytes_out
+= skb
->len
;
1658 skb_dst_set(skb
, &rt
->dst
);
1659 ip_decrease_ttl(ip_hdr(skb
));
1661 /* FIXME: forward and output firewalls used to be called here.
1662 * What do we do with netfilter? -- RR
1664 if (vif
->flags
& VIFF_TUNNEL
) {
1665 ip_encap(skb
, vif
->local
, vif
->remote
);
1666 /* FIXME: extra output firewall step used to be here. --RR */
1667 vif
->dev
->stats
.tx_packets
++;
1668 vif
->dev
->stats
.tx_bytes
+= skb
->len
;
1671 IPCB(skb
)->flags
|= IPSKB_FORWARDED
;
1674 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1675 * not only before forwarding, but after forwarding on all output
1676 * interfaces. It is clear, if mrouter runs a multicasting
1677 * program, it should receive packets not depending to what interface
1678 * program is joined.
1679 * If we will not make it, the program will have to join on all
1680 * interfaces. On the other hand, multihoming host (or router, but
1681 * not mrouter) cannot join to more than one interface - it will
1682 * result in receiving multiple packets.
1684 NF_HOOK(NFPROTO_IPV4
, NF_INET_FORWARD
, skb
, skb
->dev
, dev
,
1685 ipmr_forward_finish
);
1692 static int ipmr_find_vif(struct mr_table
*mrt
, struct net_device
*dev
)
1696 for (ct
= mrt
->maxvif
-1; ct
>= 0; ct
--) {
1697 if (mrt
->vif_table
[ct
].dev
== dev
)
1703 /* "local" means that we should preserve one skb (for local delivery) */
1705 static int ip_mr_forward(struct net
*net
, struct mr_table
*mrt
,
1706 struct sk_buff
*skb
, struct mfc_cache
*cache
,
1712 vif
= cache
->mfc_parent
;
1713 cache
->mfc_un
.res
.pkt
++;
1714 cache
->mfc_un
.res
.bytes
+= skb
->len
;
1717 * Wrong interface: drop packet and (maybe) send PIM assert.
1719 if (mrt
->vif_table
[vif
].dev
!= skb
->dev
) {
1722 if (rt_is_output_route(skb_rtable(skb
))) {
1723 /* It is our own packet, looped back.
1724 * Very complicated situation...
1726 * The best workaround until routing daemons will be
1727 * fixed is not to redistribute packet, if it was
1728 * send through wrong interface. It means, that
1729 * multicast applications WILL NOT work for
1730 * (S,G), which have default multicast route pointing
1731 * to wrong oif. In any case, it is not a good
1732 * idea to use multicasting applications on router.
1737 cache
->mfc_un
.res
.wrong_if
++;
1738 true_vifi
= ipmr_find_vif(mrt
, skb
->dev
);
1740 if (true_vifi
>= 0 && mrt
->mroute_do_assert
&&
1741 /* pimsm uses asserts, when switching from RPT to SPT,
1742 * so that we cannot check that packet arrived on an oif.
1743 * It is bad, but otherwise we would need to move pretty
1744 * large chunk of pimd to kernel. Ough... --ANK
1746 (mrt
->mroute_do_pim
||
1747 cache
->mfc_un
.res
.ttls
[true_vifi
] < 255) &&
1749 cache
->mfc_un
.res
.last_assert
+ MFC_ASSERT_THRESH
)) {
1750 cache
->mfc_un
.res
.last_assert
= jiffies
;
1751 ipmr_cache_report(mrt
, skb
, true_vifi
, IGMPMSG_WRONGVIF
);
1756 mrt
->vif_table
[vif
].pkt_in
++;
1757 mrt
->vif_table
[vif
].bytes_in
+= skb
->len
;
1762 for (ct
= cache
->mfc_un
.res
.maxvif
- 1;
1763 ct
>= cache
->mfc_un
.res
.minvif
; ct
--) {
1764 if (ip_hdr(skb
)->ttl
> cache
->mfc_un
.res
.ttls
[ct
]) {
1766 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
1769 ipmr_queue_xmit(net
, mrt
, skb2
, cache
,
1777 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
1780 ipmr_queue_xmit(net
, mrt
, skb2
, cache
, psend
);
1782 ipmr_queue_xmit(net
, mrt
, skb
, cache
, psend
);
1793 static struct mr_table
*ipmr_rt_fib_lookup(struct net
*net
, struct sk_buff
*skb
)
1795 struct rtable
*rt
= skb_rtable(skb
);
1796 struct iphdr
*iph
= ip_hdr(skb
);
1797 struct flowi4 fl4
= {
1798 .daddr
= iph
->daddr
,
1799 .saddr
= iph
->saddr
,
1800 .flowi4_tos
= RT_TOS(iph
->tos
),
1801 .flowi4_oif
= rt
->rt_oif
,
1802 .flowi4_iif
= rt
->rt_iif
,
1803 .flowi4_mark
= rt
->rt_mark
,
1805 struct mr_table
*mrt
;
1808 err
= ipmr_fib_lookup(net
, &fl4
, &mrt
);
1810 return ERR_PTR(err
);
1815 * Multicast packets for forwarding arrive here
1816 * Called with rcu_read_lock();
1819 int ip_mr_input(struct sk_buff
*skb
)
1821 struct mfc_cache
*cache
;
1822 struct net
*net
= dev_net(skb
->dev
);
1823 int local
= skb_rtable(skb
)->rt_flags
& RTCF_LOCAL
;
1824 struct mr_table
*mrt
;
1826 /* Packet is looped back after forward, it should not be
1827 * forwarded second time, but still can be delivered locally.
1829 if (IPCB(skb
)->flags
& IPSKB_FORWARDED
)
1832 mrt
= ipmr_rt_fib_lookup(net
, skb
);
1835 return PTR_ERR(mrt
);
1838 if (IPCB(skb
)->opt
.router_alert
) {
1839 if (ip_call_ra_chain(skb
))
1841 } else if (ip_hdr(skb
)->protocol
== IPPROTO_IGMP
) {
1842 /* IGMPv1 (and broken IGMPv2 implementations sort of
1843 * Cisco IOS <= 11.2(8)) do not put router alert
1844 * option to IGMP packets destined to routable
1845 * groups. It is very bad, because it means
1846 * that we can forward NO IGMP messages.
1848 struct sock
*mroute_sk
;
1850 mroute_sk
= rcu_dereference(mrt
->mroute_sk
);
1853 raw_rcv(mroute_sk
, skb
);
1859 /* already under rcu_read_lock() */
1860 cache
= ipmr_cache_find(mrt
, ip_hdr(skb
)->saddr
, ip_hdr(skb
)->daddr
);
1863 * No usable cache entry
1865 if (cache
== NULL
) {
1869 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
1870 ip_local_deliver(skb
);
1876 read_lock(&mrt_lock
);
1877 vif
= ipmr_find_vif(mrt
, skb
->dev
);
1879 int err2
= ipmr_cache_unresolved(mrt
, vif
, skb
);
1880 read_unlock(&mrt_lock
);
1884 read_unlock(&mrt_lock
);
1889 read_lock(&mrt_lock
);
1890 ip_mr_forward(net
, mrt
, skb
, cache
, local
);
1891 read_unlock(&mrt_lock
);
1894 return ip_local_deliver(skb
);
1900 return ip_local_deliver(skb
);
1905 #ifdef CONFIG_IP_PIMSM
1906 /* called with rcu_read_lock() */
1907 static int __pim_rcv(struct mr_table
*mrt
, struct sk_buff
*skb
,
1908 unsigned int pimlen
)
1910 struct net_device
*reg_dev
= NULL
;
1911 struct iphdr
*encap
;
1913 encap
= (struct iphdr
*)(skb_transport_header(skb
) + pimlen
);
1916 * a. packet is really sent to a multicast group
1917 * b. packet is not a NULL-REGISTER
1918 * c. packet is not truncated
1920 if (!ipv4_is_multicast(encap
->daddr
) ||
1921 encap
->tot_len
== 0 ||
1922 ntohs(encap
->tot_len
) + pimlen
> skb
->len
)
1925 read_lock(&mrt_lock
);
1926 if (mrt
->mroute_reg_vif_num
>= 0)
1927 reg_dev
= mrt
->vif_table
[mrt
->mroute_reg_vif_num
].dev
;
1928 read_unlock(&mrt_lock
);
1930 if (reg_dev
== NULL
)
1933 skb
->mac_header
= skb
->network_header
;
1934 skb_pull(skb
, (u8
*)encap
- skb
->data
);
1935 skb_reset_network_header(skb
);
1936 skb
->protocol
= htons(ETH_P_IP
);
1937 skb
->ip_summed
= CHECKSUM_NONE
;
1938 skb
->pkt_type
= PACKET_HOST
;
1940 skb_tunnel_rx(skb
, reg_dev
);
1944 return NET_RX_SUCCESS
;
1948 #ifdef CONFIG_IP_PIMSM_V1
1950 * Handle IGMP messages of PIMv1
1953 int pim_rcv_v1(struct sk_buff
*skb
)
1955 struct igmphdr
*pim
;
1956 struct net
*net
= dev_net(skb
->dev
);
1957 struct mr_table
*mrt
;
1959 if (!pskb_may_pull(skb
, sizeof(*pim
) + sizeof(struct iphdr
)))
1962 pim
= igmp_hdr(skb
);
1964 mrt
= ipmr_rt_fib_lookup(net
, skb
);
1967 if (!mrt
->mroute_do_pim
||
1968 pim
->group
!= PIM_V1_VERSION
|| pim
->code
!= PIM_V1_REGISTER
)
1971 if (__pim_rcv(mrt
, skb
, sizeof(*pim
))) {
1979 #ifdef CONFIG_IP_PIMSM_V2
1980 static int pim_rcv(struct sk_buff
*skb
)
1982 struct pimreghdr
*pim
;
1983 struct net
*net
= dev_net(skb
->dev
);
1984 struct mr_table
*mrt
;
1986 if (!pskb_may_pull(skb
, sizeof(*pim
) + sizeof(struct iphdr
)))
1989 pim
= (struct pimreghdr
*)skb_transport_header(skb
);
1990 if (pim
->type
!= ((PIM_VERSION
<< 4) | (PIM_REGISTER
)) ||
1991 (pim
->flags
& PIM_NULL_REGISTER
) ||
1992 (ip_compute_csum((void *)pim
, sizeof(*pim
)) != 0 &&
1993 csum_fold(skb_checksum(skb
, 0, skb
->len
, 0))))
1996 mrt
= ipmr_rt_fib_lookup(net
, skb
);
1999 if (__pim_rcv(mrt
, skb
, sizeof(*pim
))) {
2007 static int __ipmr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
2008 struct mfc_cache
*c
, struct rtmsg
*rtm
)
2011 struct rtnexthop
*nhp
;
2012 u8
*b
= skb_tail_pointer(skb
);
2013 struct rtattr
*mp_head
;
2015 /* If cache is unresolved, don't try to parse IIF and OIF */
2016 if (c
->mfc_parent
>= MAXVIFS
)
2019 if (VIF_EXISTS(mrt
, c
->mfc_parent
))
2020 RTA_PUT(skb
, RTA_IIF
, 4, &mrt
->vif_table
[c
->mfc_parent
].dev
->ifindex
);
2022 mp_head
= (struct rtattr
*)skb_put(skb
, RTA_LENGTH(0));
2024 for (ct
= c
->mfc_un
.res
.minvif
; ct
< c
->mfc_un
.res
.maxvif
; ct
++) {
2025 if (VIF_EXISTS(mrt
, ct
) && c
->mfc_un
.res
.ttls
[ct
] < 255) {
2026 if (skb_tailroom(skb
) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp
)) + 4))
2027 goto rtattr_failure
;
2028 nhp
= (struct rtnexthop
*)skb_put(skb
, RTA_ALIGN(sizeof(*nhp
)));
2029 nhp
->rtnh_flags
= 0;
2030 nhp
->rtnh_hops
= c
->mfc_un
.res
.ttls
[ct
];
2031 nhp
->rtnh_ifindex
= mrt
->vif_table
[ct
].dev
->ifindex
;
2032 nhp
->rtnh_len
= sizeof(*nhp
);
2035 mp_head
->rta_type
= RTA_MULTIPATH
;
2036 mp_head
->rta_len
= skb_tail_pointer(skb
) - (u8
*)mp_head
;
2037 rtm
->rtm_type
= RTN_MULTICAST
;
2045 int ipmr_get_route(struct net
*net
, struct sk_buff
*skb
,
2046 __be32 saddr
, __be32 daddr
,
2047 struct rtmsg
*rtm
, int nowait
)
2049 struct mfc_cache
*cache
;
2050 struct mr_table
*mrt
;
2053 mrt
= ipmr_get_table(net
, RT_TABLE_DEFAULT
);
2058 cache
= ipmr_cache_find(mrt
, saddr
, daddr
);
2060 if (cache
== NULL
) {
2061 struct sk_buff
*skb2
;
2063 struct net_device
*dev
;
2072 read_lock(&mrt_lock
);
2074 vif
= ipmr_find_vif(mrt
, dev
);
2076 read_unlock(&mrt_lock
);
2080 skb2
= skb_clone(skb
, GFP_ATOMIC
);
2082 read_unlock(&mrt_lock
);
2087 skb_push(skb2
, sizeof(struct iphdr
));
2088 skb_reset_network_header(skb2
);
2090 iph
->ihl
= sizeof(struct iphdr
) >> 2;
2094 err
= ipmr_cache_unresolved(mrt
, vif
, skb2
);
2095 read_unlock(&mrt_lock
);
2100 read_lock(&mrt_lock
);
2101 if (!nowait
&& (rtm
->rtm_flags
& RTM_F_NOTIFY
))
2102 cache
->mfc_flags
|= MFC_NOTIFY
;
2103 err
= __ipmr_fill_mroute(mrt
, skb
, cache
, rtm
);
2104 read_unlock(&mrt_lock
);
2109 static int ipmr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
2110 u32 pid
, u32 seq
, struct mfc_cache
*c
)
2112 struct nlmsghdr
*nlh
;
2115 nlh
= nlmsg_put(skb
, pid
, seq
, RTM_NEWROUTE
, sizeof(*rtm
), NLM_F_MULTI
);
2119 rtm
= nlmsg_data(nlh
);
2120 rtm
->rtm_family
= RTNL_FAMILY_IPMR
;
2121 rtm
->rtm_dst_len
= 32;
2122 rtm
->rtm_src_len
= 32;
2124 rtm
->rtm_table
= mrt
->id
;
2125 NLA_PUT_U32(skb
, RTA_TABLE
, mrt
->id
);
2126 rtm
->rtm_type
= RTN_MULTICAST
;
2127 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2128 rtm
->rtm_protocol
= RTPROT_UNSPEC
;
2131 NLA_PUT_BE32(skb
, RTA_SRC
, c
->mfc_origin
);
2132 NLA_PUT_BE32(skb
, RTA_DST
, c
->mfc_mcastgrp
);
2134 if (__ipmr_fill_mroute(mrt
, skb
, c
, rtm
) < 0)
2135 goto nla_put_failure
;
2137 return nlmsg_end(skb
, nlh
);
2140 nlmsg_cancel(skb
, nlh
);
2144 static int ipmr_rtm_dumproute(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2146 struct net
*net
= sock_net(skb
->sk
);
2147 struct mr_table
*mrt
;
2148 struct mfc_cache
*mfc
;
2149 unsigned int t
= 0, s_t
;
2150 unsigned int h
= 0, s_h
;
2151 unsigned int e
= 0, s_e
;
2158 ipmr_for_each_table(mrt
, net
) {
2163 for (h
= s_h
; h
< MFC_LINES
; h
++) {
2164 list_for_each_entry_rcu(mfc
, &mrt
->mfc_cache_array
[h
], list
) {
2167 if (ipmr_fill_mroute(mrt
, skb
,
2168 NETLINK_CB(cb
->skb
).pid
,
2191 #ifdef CONFIG_PROC_FS
2193 * The /proc interfaces to multicast routing :
2194 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2196 struct ipmr_vif_iter
{
2197 struct seq_net_private p
;
2198 struct mr_table
*mrt
;
2202 static struct vif_device
*ipmr_vif_seq_idx(struct net
*net
,
2203 struct ipmr_vif_iter
*iter
,
2206 struct mr_table
*mrt
= iter
->mrt
;
2208 for (iter
->ct
= 0; iter
->ct
< mrt
->maxvif
; ++iter
->ct
) {
2209 if (!VIF_EXISTS(mrt
, iter
->ct
))
2212 return &mrt
->vif_table
[iter
->ct
];
2217 static void *ipmr_vif_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2218 __acquires(mrt_lock
)
2220 struct ipmr_vif_iter
*iter
= seq
->private;
2221 struct net
*net
= seq_file_net(seq
);
2222 struct mr_table
*mrt
;
2224 mrt
= ipmr_get_table(net
, RT_TABLE_DEFAULT
);
2226 return ERR_PTR(-ENOENT
);
2230 read_lock(&mrt_lock
);
2231 return *pos
? ipmr_vif_seq_idx(net
, seq
->private, *pos
- 1)
2235 static void *ipmr_vif_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2237 struct ipmr_vif_iter
*iter
= seq
->private;
2238 struct net
*net
= seq_file_net(seq
);
2239 struct mr_table
*mrt
= iter
->mrt
;
2242 if (v
== SEQ_START_TOKEN
)
2243 return ipmr_vif_seq_idx(net
, iter
, 0);
2245 while (++iter
->ct
< mrt
->maxvif
) {
2246 if (!VIF_EXISTS(mrt
, iter
->ct
))
2248 return &mrt
->vif_table
[iter
->ct
];
2253 static void ipmr_vif_seq_stop(struct seq_file
*seq
, void *v
)
2254 __releases(mrt_lock
)
2256 read_unlock(&mrt_lock
);
2259 static int ipmr_vif_seq_show(struct seq_file
*seq
, void *v
)
2261 struct ipmr_vif_iter
*iter
= seq
->private;
2262 struct mr_table
*mrt
= iter
->mrt
;
2264 if (v
== SEQ_START_TOKEN
) {
2266 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2268 const struct vif_device
*vif
= v
;
2269 const char *name
= vif
->dev
? vif
->dev
->name
: "none";
2272 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2273 vif
- mrt
->vif_table
,
2274 name
, vif
->bytes_in
, vif
->pkt_in
,
2275 vif
->bytes_out
, vif
->pkt_out
,
2276 vif
->flags
, vif
->local
, vif
->remote
);
2281 static const struct seq_operations ipmr_vif_seq_ops
= {
2282 .start
= ipmr_vif_seq_start
,
2283 .next
= ipmr_vif_seq_next
,
2284 .stop
= ipmr_vif_seq_stop
,
2285 .show
= ipmr_vif_seq_show
,
2288 static int ipmr_vif_open(struct inode
*inode
, struct file
*file
)
2290 return seq_open_net(inode
, file
, &ipmr_vif_seq_ops
,
2291 sizeof(struct ipmr_vif_iter
));
2294 static const struct file_operations ipmr_vif_fops
= {
2295 .owner
= THIS_MODULE
,
2296 .open
= ipmr_vif_open
,
2298 .llseek
= seq_lseek
,
2299 .release
= seq_release_net
,
2302 struct ipmr_mfc_iter
{
2303 struct seq_net_private p
;
2304 struct mr_table
*mrt
;
2305 struct list_head
*cache
;
2310 static struct mfc_cache
*ipmr_mfc_seq_idx(struct net
*net
,
2311 struct ipmr_mfc_iter
*it
, loff_t pos
)
2313 struct mr_table
*mrt
= it
->mrt
;
2314 struct mfc_cache
*mfc
;
2317 for (it
->ct
= 0; it
->ct
< MFC_LINES
; it
->ct
++) {
2318 it
->cache
= &mrt
->mfc_cache_array
[it
->ct
];
2319 list_for_each_entry_rcu(mfc
, it
->cache
, list
)
2325 spin_lock_bh(&mfc_unres_lock
);
2326 it
->cache
= &mrt
->mfc_unres_queue
;
2327 list_for_each_entry(mfc
, it
->cache
, list
)
2330 spin_unlock_bh(&mfc_unres_lock
);
2337 static void *ipmr_mfc_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2339 struct ipmr_mfc_iter
*it
= seq
->private;
2340 struct net
*net
= seq_file_net(seq
);
2341 struct mr_table
*mrt
;
2343 mrt
= ipmr_get_table(net
, RT_TABLE_DEFAULT
);
2345 return ERR_PTR(-ENOENT
);
2350 return *pos
? ipmr_mfc_seq_idx(net
, seq
->private, *pos
- 1)
2354 static void *ipmr_mfc_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2356 struct mfc_cache
*mfc
= v
;
2357 struct ipmr_mfc_iter
*it
= seq
->private;
2358 struct net
*net
= seq_file_net(seq
);
2359 struct mr_table
*mrt
= it
->mrt
;
2363 if (v
== SEQ_START_TOKEN
)
2364 return ipmr_mfc_seq_idx(net
, seq
->private, 0);
2366 if (mfc
->list
.next
!= it
->cache
)
2367 return list_entry(mfc
->list
.next
, struct mfc_cache
, list
);
2369 if (it
->cache
== &mrt
->mfc_unres_queue
)
2372 BUG_ON(it
->cache
!= &mrt
->mfc_cache_array
[it
->ct
]);
2374 while (++it
->ct
< MFC_LINES
) {
2375 it
->cache
= &mrt
->mfc_cache_array
[it
->ct
];
2376 if (list_empty(it
->cache
))
2378 return list_first_entry(it
->cache
, struct mfc_cache
, list
);
2381 /* exhausted cache_array, show unresolved */
2383 it
->cache
= &mrt
->mfc_unres_queue
;
2386 spin_lock_bh(&mfc_unres_lock
);
2387 if (!list_empty(it
->cache
))
2388 return list_first_entry(it
->cache
, struct mfc_cache
, list
);
2391 spin_unlock_bh(&mfc_unres_lock
);
2397 static void ipmr_mfc_seq_stop(struct seq_file
*seq
, void *v
)
2399 struct ipmr_mfc_iter
*it
= seq
->private;
2400 struct mr_table
*mrt
= it
->mrt
;
2402 if (it
->cache
== &mrt
->mfc_unres_queue
)
2403 spin_unlock_bh(&mfc_unres_lock
);
2404 else if (it
->cache
== &mrt
->mfc_cache_array
[it
->ct
])
2408 static int ipmr_mfc_seq_show(struct seq_file
*seq
, void *v
)
2412 if (v
== SEQ_START_TOKEN
) {
2414 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2416 const struct mfc_cache
*mfc
= v
;
2417 const struct ipmr_mfc_iter
*it
= seq
->private;
2418 const struct mr_table
*mrt
= it
->mrt
;
2420 seq_printf(seq
, "%08X %08X %-3hd",
2421 (__force u32
) mfc
->mfc_mcastgrp
,
2422 (__force u32
) mfc
->mfc_origin
,
2425 if (it
->cache
!= &mrt
->mfc_unres_queue
) {
2426 seq_printf(seq
, " %8lu %8lu %8lu",
2427 mfc
->mfc_un
.res
.pkt
,
2428 mfc
->mfc_un
.res
.bytes
,
2429 mfc
->mfc_un
.res
.wrong_if
);
2430 for (n
= mfc
->mfc_un
.res
.minvif
;
2431 n
< mfc
->mfc_un
.res
.maxvif
; n
++) {
2432 if (VIF_EXISTS(mrt
, n
) &&
2433 mfc
->mfc_un
.res
.ttls
[n
] < 255)
2436 n
, mfc
->mfc_un
.res
.ttls
[n
]);
2439 /* unresolved mfc_caches don't contain
2440 * pkt, bytes and wrong_if values
2442 seq_printf(seq
, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2444 seq_putc(seq
, '\n');
2449 static const struct seq_operations ipmr_mfc_seq_ops
= {
2450 .start
= ipmr_mfc_seq_start
,
2451 .next
= ipmr_mfc_seq_next
,
2452 .stop
= ipmr_mfc_seq_stop
,
2453 .show
= ipmr_mfc_seq_show
,
2456 static int ipmr_mfc_open(struct inode
*inode
, struct file
*file
)
2458 return seq_open_net(inode
, file
, &ipmr_mfc_seq_ops
,
2459 sizeof(struct ipmr_mfc_iter
));
2462 static const struct file_operations ipmr_mfc_fops
= {
2463 .owner
= THIS_MODULE
,
2464 .open
= ipmr_mfc_open
,
2466 .llseek
= seq_lseek
,
2467 .release
= seq_release_net
,
2471 #ifdef CONFIG_IP_PIMSM_V2
2472 static const struct net_protocol pim_protocol
= {
2480 * Setup for IP multicast routing
2482 static int __net_init
ipmr_net_init(struct net
*net
)
2486 err
= ipmr_rules_init(net
);
2490 #ifdef CONFIG_PROC_FS
2492 if (!proc_net_fops_create(net
, "ip_mr_vif", 0, &ipmr_vif_fops
))
2494 if (!proc_net_fops_create(net
, "ip_mr_cache", 0, &ipmr_mfc_fops
))
2495 goto proc_cache_fail
;
2499 #ifdef CONFIG_PROC_FS
2501 proc_net_remove(net
, "ip_mr_vif");
2503 ipmr_rules_exit(net
);
2509 static void __net_exit
ipmr_net_exit(struct net
*net
)
2511 #ifdef CONFIG_PROC_FS
2512 proc_net_remove(net
, "ip_mr_cache");
2513 proc_net_remove(net
, "ip_mr_vif");
2515 ipmr_rules_exit(net
);
2518 static struct pernet_operations ipmr_net_ops
= {
2519 .init
= ipmr_net_init
,
2520 .exit
= ipmr_net_exit
,
2523 int __init
ip_mr_init(void)
2527 mrt_cachep
= kmem_cache_create("ip_mrt_cache",
2528 sizeof(struct mfc_cache
),
2529 0, SLAB_HWCACHE_ALIGN
| SLAB_PANIC
,
2534 err
= register_pernet_subsys(&ipmr_net_ops
);
2536 goto reg_pernet_fail
;
2538 err
= register_netdevice_notifier(&ip_mr_notifier
);
2540 goto reg_notif_fail
;
2541 #ifdef CONFIG_IP_PIMSM_V2
2542 if (inet_add_protocol(&pim_protocol
, IPPROTO_PIM
) < 0) {
2543 printk(KERN_ERR
"ip_mr_init: can't add PIM protocol\n");
2545 goto add_proto_fail
;
2548 rtnl_register(RTNL_FAMILY_IPMR
, RTM_GETROUTE
,
2549 NULL
, ipmr_rtm_dumproute
, NULL
);
2552 #ifdef CONFIG_IP_PIMSM_V2
2554 unregister_netdevice_notifier(&ip_mr_notifier
);
2557 unregister_pernet_subsys(&ipmr_net_ops
);
2559 kmem_cache_destroy(mrt_cachep
);