1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux IPv6 multicast routing support for BSD pim6sd
4 * Based on net/ipv4/ipmr.c.
6 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
7 * LSIIT Laboratory, Strasbourg, France
8 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
10 * Copyright (C)2007,2008 USAGI/WIDE Project
11 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
14 #include <linux/uaccess.h>
15 #include <linux/types.h>
16 #include <linux/sched.h>
17 #include <linux/errno.h>
19 #include <linux/kernel.h>
20 #include <linux/fcntl.h>
21 #include <linux/stat.h>
22 #include <linux/socket.h>
23 #include <linux/inet.h>
24 #include <linux/netdevice.h>
25 #include <linux/inetdevice.h>
26 #include <linux/proc_fs.h>
27 #include <linux/seq_file.h>
28 #include <linux/init.h>
29 #include <linux/compat.h>
30 #include <linux/rhashtable.h>
31 #include <net/protocol.h>
32 #include <linux/skbuff.h>
34 #include <linux/notifier.h>
35 #include <linux/if_arp.h>
36 #include <net/checksum.h>
37 #include <net/netlink.h>
38 #include <net/fib_rules.h>
41 #include <net/ip6_route.h>
42 #include <linux/mroute6.h>
43 #include <linux/pim.h>
44 #include <net/addrconf.h>
45 #include <linux/netfilter_ipv6.h>
46 #include <linux/export.h>
47 #include <net/ip6_checksum.h>
48 #include <linux/netconf.h>
49 #include <net/ip_tunnels.h>
51 #include <linux/nospec.h>
54 struct fib_rule common
;
61 /* Big lock, protecting vif table, mrt cache and mroute socket state.
62 Note that the changes are semaphored via rtnl_lock.
65 static DEFINE_RWLOCK(mrt_lock
);
67 /* Multicast router control variables */
69 /* Special spinlock for queue of unresolved entries */
70 static DEFINE_SPINLOCK(mfc_unres_lock
);
72 /* We return to original Alan's scheme. Hash table of resolved
73 entries is changed only in process context and protected
74 with weak lock mrt_lock. Queue of unresolved entries is protected
75 with strong spinlock mfc_unres_lock.
77 In this case data path is free of exclusive locks at all.
80 static struct kmem_cache
*mrt_cachep __read_mostly
;
82 static struct mr_table
*ip6mr_new_table(struct net
*net
, u32 id
);
83 static void ip6mr_free_table(struct mr_table
*mrt
);
85 static void ip6_mr_forward(struct net
*net
, struct mr_table
*mrt
,
86 struct net_device
*dev
, struct sk_buff
*skb
,
87 struct mfc6_cache
*cache
);
88 static int ip6mr_cache_report(struct mr_table
*mrt
, struct sk_buff
*pkt
,
89 mifi_t mifi
, int assert);
90 static void mr6_netlink_event(struct mr_table
*mrt
, struct mfc6_cache
*mfc
,
92 static void mrt6msg_netlink_event(struct mr_table
*mrt
, struct sk_buff
*pkt
);
93 static int ip6mr_rtm_dumproute(struct sk_buff
*skb
,
94 struct netlink_callback
*cb
);
95 static void mroute_clean_tables(struct mr_table
*mrt
, int flags
);
96 static void ipmr_expire_process(struct timer_list
*t
);
98 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
99 #define ip6mr_for_each_table(mrt, net) \
100 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
102 static struct mr_table
*ip6mr_mr_table_iter(struct net
*net
,
103 struct mr_table
*mrt
)
105 struct mr_table
*ret
;
108 ret
= list_entry_rcu(net
->ipv6
.mr6_tables
.next
,
109 struct mr_table
, list
);
111 ret
= list_entry_rcu(mrt
->list
.next
,
112 struct mr_table
, list
);
114 if (&ret
->list
== &net
->ipv6
.mr6_tables
)
119 static struct mr_table
*ip6mr_get_table(struct net
*net
, u32 id
)
121 struct mr_table
*mrt
;
123 ip6mr_for_each_table(mrt
, net
) {
130 static int ip6mr_fib_lookup(struct net
*net
, struct flowi6
*flp6
,
131 struct mr_table
**mrt
)
134 struct ip6mr_result res
;
135 struct fib_lookup_arg arg
= {
137 .flags
= FIB_LOOKUP_NOREF
,
140 /* update flow if oif or iif point to device enslaved to l3mdev */
141 l3mdev_update_flow(net
, flowi6_to_flowi(flp6
));
143 err
= fib_rules_lookup(net
->ipv6
.mr6_rules_ops
,
144 flowi6_to_flowi(flp6
), 0, &arg
);
151 static int ip6mr_rule_action(struct fib_rule
*rule
, struct flowi
*flp
,
152 int flags
, struct fib_lookup_arg
*arg
)
154 struct ip6mr_result
*res
= arg
->result
;
155 struct mr_table
*mrt
;
157 switch (rule
->action
) {
160 case FR_ACT_UNREACHABLE
:
162 case FR_ACT_PROHIBIT
:
164 case FR_ACT_BLACKHOLE
:
169 arg
->table
= fib_rule_get_table(rule
, arg
);
171 mrt
= ip6mr_get_table(rule
->fr_net
, arg
->table
);
178 static int ip6mr_rule_match(struct fib_rule
*rule
, struct flowi
*flp
, int flags
)
183 static const struct nla_policy ip6mr_rule_policy
[FRA_MAX
+ 1] = {
187 static int ip6mr_rule_configure(struct fib_rule
*rule
, struct sk_buff
*skb
,
188 struct fib_rule_hdr
*frh
, struct nlattr
**tb
,
189 struct netlink_ext_ack
*extack
)
194 static int ip6mr_rule_compare(struct fib_rule
*rule
, struct fib_rule_hdr
*frh
,
200 static int ip6mr_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 ip6mr_rules_ops_template
= {
210 .family
= RTNL_FAMILY_IP6MR
,
211 .rule_size
= sizeof(struct ip6mr_rule
),
212 .addr_size
= sizeof(struct in6_addr
),
213 .action
= ip6mr_rule_action
,
214 .match
= ip6mr_rule_match
,
215 .configure
= ip6mr_rule_configure
,
216 .compare
= ip6mr_rule_compare
,
217 .fill
= ip6mr_rule_fill
,
218 .nlgroup
= RTNLGRP_IPV6_RULE
,
219 .policy
= ip6mr_rule_policy
,
220 .owner
= THIS_MODULE
,
223 static int __net_init
ip6mr_rules_init(struct net
*net
)
225 struct fib_rules_ops
*ops
;
226 struct mr_table
*mrt
;
229 ops
= fib_rules_register(&ip6mr_rules_ops_template
, net
);
233 INIT_LIST_HEAD(&net
->ipv6
.mr6_tables
);
235 mrt
= ip6mr_new_table(net
, RT6_TABLE_DFLT
);
241 err
= fib_default_rule_add(ops
, 0x7fff, RT6_TABLE_DFLT
, 0);
245 net
->ipv6
.mr6_rules_ops
= ops
;
249 ip6mr_free_table(mrt
);
251 fib_rules_unregister(ops
);
255 static void __net_exit
ip6mr_rules_exit(struct net
*net
)
257 struct mr_table
*mrt
, *next
;
260 list_for_each_entry_safe(mrt
, next
, &net
->ipv6
.mr6_tables
, list
) {
261 list_del(&mrt
->list
);
262 ip6mr_free_table(mrt
);
264 fib_rules_unregister(net
->ipv6
.mr6_rules_ops
);
268 static int ip6mr_rules_dump(struct net
*net
, struct notifier_block
*nb
)
270 return fib_rules_dump(net
, nb
, RTNL_FAMILY_IP6MR
);
273 static unsigned int ip6mr_rules_seq_read(struct net
*net
)
275 return fib_rules_seq_read(net
, RTNL_FAMILY_IP6MR
);
278 bool ip6mr_rule_default(const struct fib_rule
*rule
)
280 return fib_rule_matchall(rule
) && rule
->action
== FR_ACT_TO_TBL
&&
281 rule
->table
== RT6_TABLE_DFLT
&& !rule
->l3mdev
;
283 EXPORT_SYMBOL(ip6mr_rule_default
);
285 #define ip6mr_for_each_table(mrt, net) \
286 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
288 static struct mr_table
*ip6mr_mr_table_iter(struct net
*net
,
289 struct mr_table
*mrt
)
292 return net
->ipv6
.mrt6
;
296 static struct mr_table
*ip6mr_get_table(struct net
*net
, u32 id
)
298 return net
->ipv6
.mrt6
;
301 static int ip6mr_fib_lookup(struct net
*net
, struct flowi6
*flp6
,
302 struct mr_table
**mrt
)
304 *mrt
= net
->ipv6
.mrt6
;
308 static int __net_init
ip6mr_rules_init(struct net
*net
)
310 struct mr_table
*mrt
;
312 mrt
= ip6mr_new_table(net
, RT6_TABLE_DFLT
);
315 net
->ipv6
.mrt6
= mrt
;
319 static void __net_exit
ip6mr_rules_exit(struct net
*net
)
322 ip6mr_free_table(net
->ipv6
.mrt6
);
323 net
->ipv6
.mrt6
= NULL
;
327 static int ip6mr_rules_dump(struct net
*net
, struct notifier_block
*nb
)
332 static unsigned int ip6mr_rules_seq_read(struct net
*net
)
338 static int ip6mr_hash_cmp(struct rhashtable_compare_arg
*arg
,
341 const struct mfc6_cache_cmp_arg
*cmparg
= arg
->key
;
342 struct mfc6_cache
*c
= (struct mfc6_cache
*)ptr
;
344 return !ipv6_addr_equal(&c
->mf6c_mcastgrp
, &cmparg
->mf6c_mcastgrp
) ||
345 !ipv6_addr_equal(&c
->mf6c_origin
, &cmparg
->mf6c_origin
);
348 static const struct rhashtable_params ip6mr_rht_params
= {
349 .head_offset
= offsetof(struct mr_mfc
, mnode
),
350 .key_offset
= offsetof(struct mfc6_cache
, cmparg
),
351 .key_len
= sizeof(struct mfc6_cache_cmp_arg
),
353 .obj_cmpfn
= ip6mr_hash_cmp
,
354 .automatic_shrinking
= true,
357 static void ip6mr_new_table_set(struct mr_table
*mrt
,
360 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
361 list_add_tail_rcu(&mrt
->list
, &net
->ipv6
.mr6_tables
);
365 static struct mfc6_cache_cmp_arg ip6mr_mr_table_ops_cmparg_any
= {
366 .mf6c_origin
= IN6ADDR_ANY_INIT
,
367 .mf6c_mcastgrp
= IN6ADDR_ANY_INIT
,
370 static struct mr_table_ops ip6mr_mr_table_ops
= {
371 .rht_params
= &ip6mr_rht_params
,
372 .cmparg_any
= &ip6mr_mr_table_ops_cmparg_any
,
375 static struct mr_table
*ip6mr_new_table(struct net
*net
, u32 id
)
377 struct mr_table
*mrt
;
379 mrt
= ip6mr_get_table(net
, id
);
383 return mr_table_alloc(net
, id
, &ip6mr_mr_table_ops
,
384 ipmr_expire_process
, ip6mr_new_table_set
);
387 static void ip6mr_free_table(struct mr_table
*mrt
)
389 del_timer_sync(&mrt
->ipmr_expire_timer
);
390 mroute_clean_tables(mrt
, MRT6_FLUSH_MIFS
| MRT6_FLUSH_MIFS_STATIC
|
391 MRT6_FLUSH_MFC
| MRT6_FLUSH_MFC_STATIC
);
392 rhltable_destroy(&mrt
->mfc_hash
);
396 #ifdef CONFIG_PROC_FS
397 /* The /proc interfaces to multicast routing
398 * /proc/ip6_mr_cache /proc/ip6_mr_vif
401 static void *ip6mr_vif_seq_start(struct seq_file
*seq
, loff_t
*pos
)
404 struct mr_vif_iter
*iter
= seq
->private;
405 struct net
*net
= seq_file_net(seq
);
406 struct mr_table
*mrt
;
408 mrt
= ip6mr_get_table(net
, RT6_TABLE_DFLT
);
410 return ERR_PTR(-ENOENT
);
414 read_lock(&mrt_lock
);
415 return mr_vif_seq_start(seq
, pos
);
418 static void ip6mr_vif_seq_stop(struct seq_file
*seq
, void *v
)
421 read_unlock(&mrt_lock
);
424 static int ip6mr_vif_seq_show(struct seq_file
*seq
, void *v
)
426 struct mr_vif_iter
*iter
= seq
->private;
427 struct mr_table
*mrt
= iter
->mrt
;
429 if (v
== SEQ_START_TOKEN
) {
431 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
433 const struct vif_device
*vif
= v
;
434 const char *name
= vif
->dev
? vif
->dev
->name
: "none";
437 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
438 vif
- mrt
->vif_table
,
439 name
, vif
->bytes_in
, vif
->pkt_in
,
440 vif
->bytes_out
, vif
->pkt_out
,
446 static const struct seq_operations ip6mr_vif_seq_ops
= {
447 .start
= ip6mr_vif_seq_start
,
448 .next
= mr_vif_seq_next
,
449 .stop
= ip6mr_vif_seq_stop
,
450 .show
= ip6mr_vif_seq_show
,
453 static void *ipmr_mfc_seq_start(struct seq_file
*seq
, loff_t
*pos
)
455 struct net
*net
= seq_file_net(seq
);
456 struct mr_table
*mrt
;
458 mrt
= ip6mr_get_table(net
, RT6_TABLE_DFLT
);
460 return ERR_PTR(-ENOENT
);
462 return mr_mfc_seq_start(seq
, pos
, mrt
, &mfc_unres_lock
);
465 static int ipmr_mfc_seq_show(struct seq_file
*seq
, void *v
)
469 if (v
== SEQ_START_TOKEN
) {
473 "Iif Pkts Bytes Wrong Oifs\n");
475 const struct mfc6_cache
*mfc
= v
;
476 const struct mr_mfc_iter
*it
= seq
->private;
477 struct mr_table
*mrt
= it
->mrt
;
479 seq_printf(seq
, "%pI6 %pI6 %-3hd",
480 &mfc
->mf6c_mcastgrp
, &mfc
->mf6c_origin
,
483 if (it
->cache
!= &mrt
->mfc_unres_queue
) {
484 seq_printf(seq
, " %8lu %8lu %8lu",
485 mfc
->_c
.mfc_un
.res
.pkt
,
486 mfc
->_c
.mfc_un
.res
.bytes
,
487 mfc
->_c
.mfc_un
.res
.wrong_if
);
488 for (n
= mfc
->_c
.mfc_un
.res
.minvif
;
489 n
< mfc
->_c
.mfc_un
.res
.maxvif
; n
++) {
490 if (VIF_EXISTS(mrt
, n
) &&
491 mfc
->_c
.mfc_un
.res
.ttls
[n
] < 255)
494 mfc
->_c
.mfc_un
.res
.ttls
[n
]);
497 /* unresolved mfc_caches don't contain
498 * pkt, bytes and wrong_if values
500 seq_printf(seq
, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
507 static const struct seq_operations ipmr_mfc_seq_ops
= {
508 .start
= ipmr_mfc_seq_start
,
509 .next
= mr_mfc_seq_next
,
510 .stop
= mr_mfc_seq_stop
,
511 .show
= ipmr_mfc_seq_show
,
515 #ifdef CONFIG_IPV6_PIMSM_V2
517 static int pim6_rcv(struct sk_buff
*skb
)
519 struct pimreghdr
*pim
;
520 struct ipv6hdr
*encap
;
521 struct net_device
*reg_dev
= NULL
;
522 struct net
*net
= dev_net(skb
->dev
);
523 struct mr_table
*mrt
;
524 struct flowi6 fl6
= {
525 .flowi6_iif
= skb
->dev
->ifindex
,
526 .flowi6_mark
= skb
->mark
,
530 if (!pskb_may_pull(skb
, sizeof(*pim
) + sizeof(*encap
)))
533 pim
= (struct pimreghdr
*)skb_transport_header(skb
);
534 if (pim
->type
!= ((PIM_VERSION
<< 4) | PIM_TYPE_REGISTER
) ||
535 (pim
->flags
& PIM_NULL_REGISTER
) ||
536 (csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
, &ipv6_hdr(skb
)->daddr
,
537 sizeof(*pim
), IPPROTO_PIM
,
538 csum_partial((void *)pim
, sizeof(*pim
), 0)) &&
539 csum_fold(skb_checksum(skb
, 0, skb
->len
, 0))))
542 /* check if the inner packet is destined to mcast group */
543 encap
= (struct ipv6hdr
*)(skb_transport_header(skb
) +
546 if (!ipv6_addr_is_multicast(&encap
->daddr
) ||
547 encap
->payload_len
== 0 ||
548 ntohs(encap
->payload_len
) + sizeof(*pim
) > skb
->len
)
551 if (ip6mr_fib_lookup(net
, &fl6
, &mrt
) < 0)
553 reg_vif_num
= mrt
->mroute_reg_vif_num
;
555 read_lock(&mrt_lock
);
556 if (reg_vif_num
>= 0)
557 reg_dev
= mrt
->vif_table
[reg_vif_num
].dev
;
560 read_unlock(&mrt_lock
);
565 skb
->mac_header
= skb
->network_header
;
566 skb_pull(skb
, (u8
*)encap
- skb
->data
);
567 skb_reset_network_header(skb
);
568 skb
->protocol
= htons(ETH_P_IPV6
);
569 skb
->ip_summed
= CHECKSUM_NONE
;
571 skb_tunnel_rx(skb
, reg_dev
, dev_net(reg_dev
));
582 static const struct inet6_protocol pim6_protocol
= {
586 /* Service routines creating virtual interfaces: PIMREG */
588 static netdev_tx_t
reg_vif_xmit(struct sk_buff
*skb
,
589 struct net_device
*dev
)
591 struct net
*net
= dev_net(dev
);
592 struct mr_table
*mrt
;
593 struct flowi6 fl6
= {
594 .flowi6_oif
= dev
->ifindex
,
595 .flowi6_iif
= skb
->skb_iif
? : LOOPBACK_IFINDEX
,
596 .flowi6_mark
= skb
->mark
,
599 if (!pskb_inet_may_pull(skb
))
602 if (ip6mr_fib_lookup(net
, &fl6
, &mrt
) < 0)
605 read_lock(&mrt_lock
);
606 dev
->stats
.tx_bytes
+= skb
->len
;
607 dev
->stats
.tx_packets
++;
608 ip6mr_cache_report(mrt
, skb
, mrt
->mroute_reg_vif_num
, MRT6MSG_WHOLEPKT
);
609 read_unlock(&mrt_lock
);
614 dev
->stats
.tx_errors
++;
619 static int reg_vif_get_iflink(const struct net_device
*dev
)
624 static const struct net_device_ops reg_vif_netdev_ops
= {
625 .ndo_start_xmit
= reg_vif_xmit
,
626 .ndo_get_iflink
= reg_vif_get_iflink
,
629 static void reg_vif_setup(struct net_device
*dev
)
631 dev
->type
= ARPHRD_PIMREG
;
632 dev
->mtu
= 1500 - sizeof(struct ipv6hdr
) - 8;
633 dev
->flags
= IFF_NOARP
;
634 dev
->netdev_ops
= ®_vif_netdev_ops
;
635 dev
->needs_free_netdev
= true;
636 dev
->features
|= NETIF_F_NETNS_LOCAL
;
639 static struct net_device
*ip6mr_reg_vif(struct net
*net
, struct mr_table
*mrt
)
641 struct net_device
*dev
;
644 if (mrt
->id
== RT6_TABLE_DFLT
)
645 sprintf(name
, "pim6reg");
647 sprintf(name
, "pim6reg%u", mrt
->id
);
649 dev
= alloc_netdev(0, name
, NET_NAME_UNKNOWN
, reg_vif_setup
);
653 dev_net_set(dev
, net
);
655 if (register_netdevice(dev
)) {
660 if (dev_open(dev
, NULL
))
667 unregister_netdevice(dev
);
672 static int call_ip6mr_vif_entry_notifiers(struct net
*net
,
673 enum fib_event_type event_type
,
674 struct vif_device
*vif
,
675 mifi_t vif_index
, u32 tb_id
)
677 return mr_call_vif_notifiers(net
, RTNL_FAMILY_IP6MR
, event_type
,
678 vif
, vif_index
, tb_id
,
679 &net
->ipv6
.ipmr_seq
);
682 static int call_ip6mr_mfc_entry_notifiers(struct net
*net
,
683 enum fib_event_type event_type
,
684 struct mfc6_cache
*mfc
, u32 tb_id
)
686 return mr_call_mfc_notifiers(net
, RTNL_FAMILY_IP6MR
, event_type
,
687 &mfc
->_c
, tb_id
, &net
->ipv6
.ipmr_seq
);
690 /* Delete a VIF entry */
691 static int mif6_delete(struct mr_table
*mrt
, int vifi
, int notify
,
692 struct list_head
*head
)
694 struct vif_device
*v
;
695 struct net_device
*dev
;
696 struct inet6_dev
*in6_dev
;
698 if (vifi
< 0 || vifi
>= mrt
->maxvif
)
699 return -EADDRNOTAVAIL
;
701 v
= &mrt
->vif_table
[vifi
];
703 if (VIF_EXISTS(mrt
, vifi
))
704 call_ip6mr_vif_entry_notifiers(read_pnet(&mrt
->net
),
705 FIB_EVENT_VIF_DEL
, v
, vifi
,
708 write_lock_bh(&mrt_lock
);
713 write_unlock_bh(&mrt_lock
);
714 return -EADDRNOTAVAIL
;
717 #ifdef CONFIG_IPV6_PIMSM_V2
718 if (vifi
== mrt
->mroute_reg_vif_num
)
719 mrt
->mroute_reg_vif_num
= -1;
722 if (vifi
+ 1 == mrt
->maxvif
) {
724 for (tmp
= vifi
- 1; tmp
>= 0; tmp
--) {
725 if (VIF_EXISTS(mrt
, tmp
))
728 mrt
->maxvif
= tmp
+ 1;
731 write_unlock_bh(&mrt_lock
);
733 dev_set_allmulti(dev
, -1);
735 in6_dev
= __in6_dev_get(dev
);
737 in6_dev
->cnf
.mc_forwarding
--;
738 inet6_netconf_notify_devconf(dev_net(dev
), RTM_NEWNETCONF
,
739 NETCONFA_MC_FORWARDING
,
740 dev
->ifindex
, &in6_dev
->cnf
);
743 if ((v
->flags
& MIFF_REGISTER
) && !notify
)
744 unregister_netdevice_queue(dev
, head
);
750 static inline void ip6mr_cache_free_rcu(struct rcu_head
*head
)
752 struct mr_mfc
*c
= container_of(head
, struct mr_mfc
, rcu
);
754 kmem_cache_free(mrt_cachep
, (struct mfc6_cache
*)c
);
757 static inline void ip6mr_cache_free(struct mfc6_cache
*c
)
759 call_rcu(&c
->_c
.rcu
, ip6mr_cache_free_rcu
);
762 /* Destroy an unresolved cache entry, killing queued skbs
763 and reporting error to netlink readers.
766 static void ip6mr_destroy_unres(struct mr_table
*mrt
, struct mfc6_cache
*c
)
768 struct net
*net
= read_pnet(&mrt
->net
);
771 atomic_dec(&mrt
->cache_resolve_queue_len
);
773 while ((skb
= skb_dequeue(&c
->_c
.mfc_un
.unres
.unresolved
)) != NULL
) {
774 if (ipv6_hdr(skb
)->version
== 0) {
775 struct nlmsghdr
*nlh
= skb_pull(skb
,
776 sizeof(struct ipv6hdr
));
777 nlh
->nlmsg_type
= NLMSG_ERROR
;
778 nlh
->nlmsg_len
= nlmsg_msg_size(sizeof(struct nlmsgerr
));
779 skb_trim(skb
, nlh
->nlmsg_len
);
780 ((struct nlmsgerr
*)nlmsg_data(nlh
))->error
= -ETIMEDOUT
;
781 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).portid
);
790 /* Timer process for all the unresolved queue. */
792 static void ipmr_do_expire_process(struct mr_table
*mrt
)
794 unsigned long now
= jiffies
;
795 unsigned long expires
= 10 * HZ
;
796 struct mr_mfc
*c
, *next
;
798 list_for_each_entry_safe(c
, next
, &mrt
->mfc_unres_queue
, list
) {
799 if (time_after(c
->mfc_un
.unres
.expires
, now
)) {
801 unsigned long interval
= c
->mfc_un
.unres
.expires
- now
;
802 if (interval
< expires
)
808 mr6_netlink_event(mrt
, (struct mfc6_cache
*)c
, RTM_DELROUTE
);
809 ip6mr_destroy_unres(mrt
, (struct mfc6_cache
*)c
);
812 if (!list_empty(&mrt
->mfc_unres_queue
))
813 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+ expires
);
816 static void ipmr_expire_process(struct timer_list
*t
)
818 struct mr_table
*mrt
= from_timer(mrt
, t
, ipmr_expire_timer
);
820 if (!spin_trylock(&mfc_unres_lock
)) {
821 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+ 1);
825 if (!list_empty(&mrt
->mfc_unres_queue
))
826 ipmr_do_expire_process(mrt
);
828 spin_unlock(&mfc_unres_lock
);
831 /* Fill oifs list. It is called under write locked mrt_lock. */
833 static void ip6mr_update_thresholds(struct mr_table
*mrt
,
834 struct mr_mfc
*cache
,
839 cache
->mfc_un
.res
.minvif
= MAXMIFS
;
840 cache
->mfc_un
.res
.maxvif
= 0;
841 memset(cache
->mfc_un
.res
.ttls
, 255, MAXMIFS
);
843 for (vifi
= 0; vifi
< mrt
->maxvif
; vifi
++) {
844 if (VIF_EXISTS(mrt
, vifi
) &&
845 ttls
[vifi
] && ttls
[vifi
] < 255) {
846 cache
->mfc_un
.res
.ttls
[vifi
] = ttls
[vifi
];
847 if (cache
->mfc_un
.res
.minvif
> vifi
)
848 cache
->mfc_un
.res
.minvif
= vifi
;
849 if (cache
->mfc_un
.res
.maxvif
<= vifi
)
850 cache
->mfc_un
.res
.maxvif
= vifi
+ 1;
853 cache
->mfc_un
.res
.lastuse
= jiffies
;
856 static int mif6_add(struct net
*net
, struct mr_table
*mrt
,
857 struct mif6ctl
*vifc
, int mrtsock
)
859 int vifi
= vifc
->mif6c_mifi
;
860 struct vif_device
*v
= &mrt
->vif_table
[vifi
];
861 struct net_device
*dev
;
862 struct inet6_dev
*in6_dev
;
866 if (VIF_EXISTS(mrt
, vifi
))
869 switch (vifc
->mif6c_flags
) {
870 #ifdef CONFIG_IPV6_PIMSM_V2
873 * Special Purpose VIF in PIM
874 * All the packets will be sent to the daemon
876 if (mrt
->mroute_reg_vif_num
>= 0)
878 dev
= ip6mr_reg_vif(net
, mrt
);
881 err
= dev_set_allmulti(dev
, 1);
883 unregister_netdevice(dev
);
890 dev
= dev_get_by_index(net
, vifc
->mif6c_pifi
);
892 return -EADDRNOTAVAIL
;
893 err
= dev_set_allmulti(dev
, 1);
903 in6_dev
= __in6_dev_get(dev
);
905 in6_dev
->cnf
.mc_forwarding
++;
906 inet6_netconf_notify_devconf(dev_net(dev
), RTM_NEWNETCONF
,
907 NETCONFA_MC_FORWARDING
,
908 dev
->ifindex
, &in6_dev
->cnf
);
911 /* Fill in the VIF structures */
912 vif_device_init(v
, dev
, vifc
->vifc_rate_limit
, vifc
->vifc_threshold
,
913 vifc
->mif6c_flags
| (!mrtsock
? VIFF_STATIC
: 0),
916 /* And finish update writing critical data */
917 write_lock_bh(&mrt_lock
);
919 #ifdef CONFIG_IPV6_PIMSM_V2
920 if (v
->flags
& MIFF_REGISTER
)
921 mrt
->mroute_reg_vif_num
= vifi
;
923 if (vifi
+ 1 > mrt
->maxvif
)
924 mrt
->maxvif
= vifi
+ 1;
925 write_unlock_bh(&mrt_lock
);
926 call_ip6mr_vif_entry_notifiers(net
, FIB_EVENT_VIF_ADD
,
931 static struct mfc6_cache
*ip6mr_cache_find(struct mr_table
*mrt
,
932 const struct in6_addr
*origin
,
933 const struct in6_addr
*mcastgrp
)
935 struct mfc6_cache_cmp_arg arg
= {
936 .mf6c_origin
= *origin
,
937 .mf6c_mcastgrp
= *mcastgrp
,
940 return mr_mfc_find(mrt
, &arg
);
943 /* Look for a (*,G) entry */
944 static struct mfc6_cache
*ip6mr_cache_find_any(struct mr_table
*mrt
,
945 struct in6_addr
*mcastgrp
,
948 struct mfc6_cache_cmp_arg arg
= {
949 .mf6c_origin
= in6addr_any
,
950 .mf6c_mcastgrp
= *mcastgrp
,
953 if (ipv6_addr_any(mcastgrp
))
954 return mr_mfc_find_any_parent(mrt
, mifi
);
955 return mr_mfc_find_any(mrt
, mifi
, &arg
);
958 /* Look for a (S,G,iif) entry if parent != -1 */
959 static struct mfc6_cache
*
960 ip6mr_cache_find_parent(struct mr_table
*mrt
,
961 const struct in6_addr
*origin
,
962 const struct in6_addr
*mcastgrp
,
965 struct mfc6_cache_cmp_arg arg
= {
966 .mf6c_origin
= *origin
,
967 .mf6c_mcastgrp
= *mcastgrp
,
970 return mr_mfc_find_parent(mrt
, &arg
, parent
);
973 /* Allocate a multicast cache entry */
974 static struct mfc6_cache
*ip6mr_cache_alloc(void)
976 struct mfc6_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_KERNEL
);
979 c
->_c
.mfc_un
.res
.last_assert
= jiffies
- MFC_ASSERT_THRESH
- 1;
980 c
->_c
.mfc_un
.res
.minvif
= MAXMIFS
;
981 c
->_c
.free
= ip6mr_cache_free_rcu
;
982 refcount_set(&c
->_c
.mfc_un
.res
.refcount
, 1);
986 static struct mfc6_cache
*ip6mr_cache_alloc_unres(void)
988 struct mfc6_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_ATOMIC
);
991 skb_queue_head_init(&c
->_c
.mfc_un
.unres
.unresolved
);
992 c
->_c
.mfc_un
.unres
.expires
= jiffies
+ 10 * HZ
;
997 * A cache entry has gone into a resolved state from queued
1000 static void ip6mr_cache_resolve(struct net
*net
, struct mr_table
*mrt
,
1001 struct mfc6_cache
*uc
, struct mfc6_cache
*c
)
1003 struct sk_buff
*skb
;
1006 * Play the pending entries through our router
1009 while ((skb
= __skb_dequeue(&uc
->_c
.mfc_un
.unres
.unresolved
))) {
1010 if (ipv6_hdr(skb
)->version
== 0) {
1011 struct nlmsghdr
*nlh
= skb_pull(skb
,
1012 sizeof(struct ipv6hdr
));
1014 if (mr_fill_mroute(mrt
, skb
, &c
->_c
,
1015 nlmsg_data(nlh
)) > 0) {
1016 nlh
->nlmsg_len
= skb_tail_pointer(skb
) - (u8
*)nlh
;
1018 nlh
->nlmsg_type
= NLMSG_ERROR
;
1019 nlh
->nlmsg_len
= nlmsg_msg_size(sizeof(struct nlmsgerr
));
1020 skb_trim(skb
, nlh
->nlmsg_len
);
1021 ((struct nlmsgerr
*)nlmsg_data(nlh
))->error
= -EMSGSIZE
;
1023 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).portid
);
1025 ip6_mr_forward(net
, mrt
, skb
->dev
, skb
, c
);
1030 * Bounce a cache query up to pim6sd and netlink.
1032 * Called under mrt_lock.
1035 static int ip6mr_cache_report(struct mr_table
*mrt
, struct sk_buff
*pkt
,
1036 mifi_t mifi
, int assert)
1038 struct sock
*mroute6_sk
;
1039 struct sk_buff
*skb
;
1040 struct mrt6msg
*msg
;
1043 #ifdef CONFIG_IPV6_PIMSM_V2
1044 if (assert == MRT6MSG_WHOLEPKT
)
1045 skb
= skb_realloc_headroom(pkt
, -skb_network_offset(pkt
)
1049 skb
= alloc_skb(sizeof(struct ipv6hdr
) + sizeof(*msg
), GFP_ATOMIC
);
1054 /* I suppose that internal messages
1055 * do not require checksums */
1057 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1059 #ifdef CONFIG_IPV6_PIMSM_V2
1060 if (assert == MRT6MSG_WHOLEPKT
) {
1061 /* Ugly, but we have no choice with this interface.
1062 Duplicate old header, fix length etc.
1063 And all this only to mangle msg->im6_msgtype and
1064 to set msg->im6_mbz to "mbz" :-)
1066 skb_push(skb
, -skb_network_offset(pkt
));
1068 skb_push(skb
, sizeof(*msg
));
1069 skb_reset_transport_header(skb
);
1070 msg
= (struct mrt6msg
*)skb_transport_header(skb
);
1072 msg
->im6_msgtype
= MRT6MSG_WHOLEPKT
;
1073 msg
->im6_mif
= mrt
->mroute_reg_vif_num
;
1075 msg
->im6_src
= ipv6_hdr(pkt
)->saddr
;
1076 msg
->im6_dst
= ipv6_hdr(pkt
)->daddr
;
1078 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1083 * Copy the IP header
1086 skb_put(skb
, sizeof(struct ipv6hdr
));
1087 skb_reset_network_header(skb
);
1088 skb_copy_to_linear_data(skb
, ipv6_hdr(pkt
), sizeof(struct ipv6hdr
));
1093 skb_put(skb
, sizeof(*msg
));
1094 skb_reset_transport_header(skb
);
1095 msg
= (struct mrt6msg
*)skb_transport_header(skb
);
1098 msg
->im6_msgtype
= assert;
1099 msg
->im6_mif
= mifi
;
1101 msg
->im6_src
= ipv6_hdr(pkt
)->saddr
;
1102 msg
->im6_dst
= ipv6_hdr(pkt
)->daddr
;
1104 skb_dst_set(skb
, dst_clone(skb_dst(pkt
)));
1105 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1109 mroute6_sk
= rcu_dereference(mrt
->mroute_sk
);
1116 mrt6msg_netlink_event(mrt
, skb
);
1118 /* Deliver to user space multicast routing algorithms */
1119 ret
= sock_queue_rcv_skb(mroute6_sk
, skb
);
1122 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1129 /* Queue a packet for resolution. It gets locked cache entry! */
1130 static int ip6mr_cache_unresolved(struct mr_table
*mrt
, mifi_t mifi
,
1131 struct sk_buff
*skb
, struct net_device
*dev
)
1133 struct mfc6_cache
*c
;
1137 spin_lock_bh(&mfc_unres_lock
);
1138 list_for_each_entry(c
, &mrt
->mfc_unres_queue
, _c
.list
) {
1139 if (ipv6_addr_equal(&c
->mf6c_mcastgrp
, &ipv6_hdr(skb
)->daddr
) &&
1140 ipv6_addr_equal(&c
->mf6c_origin
, &ipv6_hdr(skb
)->saddr
)) {
1148 * Create a new entry if allowable
1151 if (atomic_read(&mrt
->cache_resolve_queue_len
) >= 10 ||
1152 (c
= ip6mr_cache_alloc_unres()) == NULL
) {
1153 spin_unlock_bh(&mfc_unres_lock
);
1159 /* Fill in the new cache entry */
1160 c
->_c
.mfc_parent
= -1;
1161 c
->mf6c_origin
= ipv6_hdr(skb
)->saddr
;
1162 c
->mf6c_mcastgrp
= ipv6_hdr(skb
)->daddr
;
1165 * Reflect first query at pim6sd
1167 err
= ip6mr_cache_report(mrt
, skb
, mifi
, MRT6MSG_NOCACHE
);
1169 /* If the report failed throw the cache entry
1172 spin_unlock_bh(&mfc_unres_lock
);
1174 ip6mr_cache_free(c
);
1179 atomic_inc(&mrt
->cache_resolve_queue_len
);
1180 list_add(&c
->_c
.list
, &mrt
->mfc_unres_queue
);
1181 mr6_netlink_event(mrt
, c
, RTM_NEWROUTE
);
1183 ipmr_do_expire_process(mrt
);
1186 /* See if we can append the packet */
1187 if (c
->_c
.mfc_un
.unres
.unresolved
.qlen
> 3) {
1193 skb
->skb_iif
= dev
->ifindex
;
1195 skb_queue_tail(&c
->_c
.mfc_un
.unres
.unresolved
, skb
);
1199 spin_unlock_bh(&mfc_unres_lock
);
1204 * MFC6 cache manipulation by user space
1207 static int ip6mr_mfc_delete(struct mr_table
*mrt
, struct mf6cctl
*mfc
,
1210 struct mfc6_cache
*c
;
1212 /* The entries are added/deleted only under RTNL */
1214 c
= ip6mr_cache_find_parent(mrt
, &mfc
->mf6cc_origin
.sin6_addr
,
1215 &mfc
->mf6cc_mcastgrp
.sin6_addr
, parent
);
1219 rhltable_remove(&mrt
->mfc_hash
, &c
->_c
.mnode
, ip6mr_rht_params
);
1220 list_del_rcu(&c
->_c
.list
);
1222 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt
->net
),
1223 FIB_EVENT_ENTRY_DEL
, c
, mrt
->id
);
1224 mr6_netlink_event(mrt
, c
, RTM_DELROUTE
);
1225 mr_cache_put(&c
->_c
);
1229 static int ip6mr_device_event(struct notifier_block
*this,
1230 unsigned long event
, void *ptr
)
1232 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1233 struct net
*net
= dev_net(dev
);
1234 struct mr_table
*mrt
;
1235 struct vif_device
*v
;
1238 if (event
!= NETDEV_UNREGISTER
)
1241 ip6mr_for_each_table(mrt
, net
) {
1242 v
= &mrt
->vif_table
[0];
1243 for (ct
= 0; ct
< mrt
->maxvif
; ct
++, v
++) {
1245 mif6_delete(mrt
, ct
, 1, NULL
);
1252 static unsigned int ip6mr_seq_read(struct net
*net
)
1256 return net
->ipv6
.ipmr_seq
+ ip6mr_rules_seq_read(net
);
1259 static int ip6mr_dump(struct net
*net
, struct notifier_block
*nb
)
1261 return mr_dump(net
, nb
, RTNL_FAMILY_IP6MR
, ip6mr_rules_dump
,
1262 ip6mr_mr_table_iter
, &mrt_lock
);
1265 static struct notifier_block ip6_mr_notifier
= {
1266 .notifier_call
= ip6mr_device_event
1269 static const struct fib_notifier_ops ip6mr_notifier_ops_template
= {
1270 .family
= RTNL_FAMILY_IP6MR
,
1271 .fib_seq_read
= ip6mr_seq_read
,
1272 .fib_dump
= ip6mr_dump
,
1273 .owner
= THIS_MODULE
,
1276 static int __net_init
ip6mr_notifier_init(struct net
*net
)
1278 struct fib_notifier_ops
*ops
;
1280 net
->ipv6
.ipmr_seq
= 0;
1282 ops
= fib_notifier_ops_register(&ip6mr_notifier_ops_template
, net
);
1284 return PTR_ERR(ops
);
1286 net
->ipv6
.ip6mr_notifier_ops
= ops
;
1291 static void __net_exit
ip6mr_notifier_exit(struct net
*net
)
1293 fib_notifier_ops_unregister(net
->ipv6
.ip6mr_notifier_ops
);
1294 net
->ipv6
.ip6mr_notifier_ops
= NULL
;
1297 /* Setup for IP multicast routing */
1298 static int __net_init
ip6mr_net_init(struct net
*net
)
1302 err
= ip6mr_notifier_init(net
);
1306 err
= ip6mr_rules_init(net
);
1308 goto ip6mr_rules_fail
;
1310 #ifdef CONFIG_PROC_FS
1312 if (!proc_create_net("ip6_mr_vif", 0, net
->proc_net
, &ip6mr_vif_seq_ops
,
1313 sizeof(struct mr_vif_iter
)))
1315 if (!proc_create_net("ip6_mr_cache", 0, net
->proc_net
, &ipmr_mfc_seq_ops
,
1316 sizeof(struct mr_mfc_iter
)))
1317 goto proc_cache_fail
;
1322 #ifdef CONFIG_PROC_FS
1324 remove_proc_entry("ip6_mr_vif", net
->proc_net
);
1326 ip6mr_rules_exit(net
);
1329 ip6mr_notifier_exit(net
);
1333 static void __net_exit
ip6mr_net_exit(struct net
*net
)
1335 #ifdef CONFIG_PROC_FS
1336 remove_proc_entry("ip6_mr_cache", net
->proc_net
);
1337 remove_proc_entry("ip6_mr_vif", net
->proc_net
);
1339 ip6mr_rules_exit(net
);
1340 ip6mr_notifier_exit(net
);
1343 static struct pernet_operations ip6mr_net_ops
= {
1344 .init
= ip6mr_net_init
,
1345 .exit
= ip6mr_net_exit
,
1348 int __init
ip6_mr_init(void)
1352 mrt_cachep
= kmem_cache_create("ip6_mrt_cache",
1353 sizeof(struct mfc6_cache
),
1354 0, SLAB_HWCACHE_ALIGN
,
1359 err
= register_pernet_subsys(&ip6mr_net_ops
);
1361 goto reg_pernet_fail
;
1363 err
= register_netdevice_notifier(&ip6_mr_notifier
);
1365 goto reg_notif_fail
;
1366 #ifdef CONFIG_IPV6_PIMSM_V2
1367 if (inet6_add_protocol(&pim6_protocol
, IPPROTO_PIM
) < 0) {
1368 pr_err("%s: can't add PIM protocol\n", __func__
);
1370 goto add_proto_fail
;
1373 err
= rtnl_register_module(THIS_MODULE
, RTNL_FAMILY_IP6MR
, RTM_GETROUTE
,
1374 NULL
, ip6mr_rtm_dumproute
, 0);
1378 #ifdef CONFIG_IPV6_PIMSM_V2
1379 inet6_del_protocol(&pim6_protocol
, IPPROTO_PIM
);
1381 unregister_netdevice_notifier(&ip6_mr_notifier
);
1384 unregister_pernet_subsys(&ip6mr_net_ops
);
1386 kmem_cache_destroy(mrt_cachep
);
1390 void ip6_mr_cleanup(void)
1392 rtnl_unregister(RTNL_FAMILY_IP6MR
, RTM_GETROUTE
);
1393 #ifdef CONFIG_IPV6_PIMSM_V2
1394 inet6_del_protocol(&pim6_protocol
, IPPROTO_PIM
);
1396 unregister_netdevice_notifier(&ip6_mr_notifier
);
1397 unregister_pernet_subsys(&ip6mr_net_ops
);
1398 kmem_cache_destroy(mrt_cachep
);
1401 static int ip6mr_mfc_add(struct net
*net
, struct mr_table
*mrt
,
1402 struct mf6cctl
*mfc
, int mrtsock
, int parent
)
1404 unsigned char ttls
[MAXMIFS
];
1405 struct mfc6_cache
*uc
, *c
;
1410 if (mfc
->mf6cc_parent
>= MAXMIFS
)
1413 memset(ttls
, 255, MAXMIFS
);
1414 for (i
= 0; i
< MAXMIFS
; i
++) {
1415 if (IF_ISSET(i
, &mfc
->mf6cc_ifset
))
1419 /* The entries are added/deleted only under RTNL */
1421 c
= ip6mr_cache_find_parent(mrt
, &mfc
->mf6cc_origin
.sin6_addr
,
1422 &mfc
->mf6cc_mcastgrp
.sin6_addr
, parent
);
1425 write_lock_bh(&mrt_lock
);
1426 c
->_c
.mfc_parent
= mfc
->mf6cc_parent
;
1427 ip6mr_update_thresholds(mrt
, &c
->_c
, ttls
);
1429 c
->_c
.mfc_flags
|= MFC_STATIC
;
1430 write_unlock_bh(&mrt_lock
);
1431 call_ip6mr_mfc_entry_notifiers(net
, FIB_EVENT_ENTRY_REPLACE
,
1433 mr6_netlink_event(mrt
, c
, RTM_NEWROUTE
);
1437 if (!ipv6_addr_any(&mfc
->mf6cc_mcastgrp
.sin6_addr
) &&
1438 !ipv6_addr_is_multicast(&mfc
->mf6cc_mcastgrp
.sin6_addr
))
1441 c
= ip6mr_cache_alloc();
1445 c
->mf6c_origin
= mfc
->mf6cc_origin
.sin6_addr
;
1446 c
->mf6c_mcastgrp
= mfc
->mf6cc_mcastgrp
.sin6_addr
;
1447 c
->_c
.mfc_parent
= mfc
->mf6cc_parent
;
1448 ip6mr_update_thresholds(mrt
, &c
->_c
, ttls
);
1450 c
->_c
.mfc_flags
|= MFC_STATIC
;
1452 err
= rhltable_insert_key(&mrt
->mfc_hash
, &c
->cmparg
, &c
->_c
.mnode
,
1455 pr_err("ip6mr: rhtable insert error %d\n", err
);
1456 ip6mr_cache_free(c
);
1459 list_add_tail_rcu(&c
->_c
.list
, &mrt
->mfc_cache_list
);
1461 /* Check to see if we resolved a queued list. If so we
1462 * need to send on the frames and tidy up.
1465 spin_lock_bh(&mfc_unres_lock
);
1466 list_for_each_entry(_uc
, &mrt
->mfc_unres_queue
, list
) {
1467 uc
= (struct mfc6_cache
*)_uc
;
1468 if (ipv6_addr_equal(&uc
->mf6c_origin
, &c
->mf6c_origin
) &&
1469 ipv6_addr_equal(&uc
->mf6c_mcastgrp
, &c
->mf6c_mcastgrp
)) {
1470 list_del(&_uc
->list
);
1471 atomic_dec(&mrt
->cache_resolve_queue_len
);
1476 if (list_empty(&mrt
->mfc_unres_queue
))
1477 del_timer(&mrt
->ipmr_expire_timer
);
1478 spin_unlock_bh(&mfc_unres_lock
);
1481 ip6mr_cache_resolve(net
, mrt
, uc
, c
);
1482 ip6mr_cache_free(uc
);
1484 call_ip6mr_mfc_entry_notifiers(net
, FIB_EVENT_ENTRY_ADD
,
1486 mr6_netlink_event(mrt
, c
, RTM_NEWROUTE
);
1491 * Close the multicast socket, and clear the vif tables etc
1494 static void mroute_clean_tables(struct mr_table
*mrt
, int flags
)
1496 struct mr_mfc
*c
, *tmp
;
1500 /* Shut down all active vif entries */
1501 if (flags
& (MRT6_FLUSH_MIFS
| MRT6_FLUSH_MIFS_STATIC
)) {
1502 for (i
= 0; i
< mrt
->maxvif
; i
++) {
1503 if (((mrt
->vif_table
[i
].flags
& VIFF_STATIC
) &&
1504 !(flags
& MRT6_FLUSH_MIFS_STATIC
)) ||
1505 (!(mrt
->vif_table
[i
].flags
& VIFF_STATIC
) && !(flags
& MRT6_FLUSH_MIFS
)))
1507 mif6_delete(mrt
, i
, 0, &list
);
1509 unregister_netdevice_many(&list
);
1512 /* Wipe the cache */
1513 if (flags
& (MRT6_FLUSH_MFC
| MRT6_FLUSH_MFC_STATIC
)) {
1514 list_for_each_entry_safe(c
, tmp
, &mrt
->mfc_cache_list
, list
) {
1515 if (((c
->mfc_flags
& MFC_STATIC
) && !(flags
& MRT6_FLUSH_MFC_STATIC
)) ||
1516 (!(c
->mfc_flags
& MFC_STATIC
) && !(flags
& MRT6_FLUSH_MFC
)))
1518 rhltable_remove(&mrt
->mfc_hash
, &c
->mnode
, ip6mr_rht_params
);
1519 list_del_rcu(&c
->list
);
1520 call_ip6mr_mfc_entry_notifiers(read_pnet(&mrt
->net
),
1521 FIB_EVENT_ENTRY_DEL
,
1522 (struct mfc6_cache
*)c
, mrt
->id
);
1523 mr6_netlink_event(mrt
, (struct mfc6_cache
*)c
, RTM_DELROUTE
);
1528 if (flags
& MRT6_FLUSH_MFC
) {
1529 if (atomic_read(&mrt
->cache_resolve_queue_len
) != 0) {
1530 spin_lock_bh(&mfc_unres_lock
);
1531 list_for_each_entry_safe(c
, tmp
, &mrt
->mfc_unres_queue
, list
) {
1533 mr6_netlink_event(mrt
, (struct mfc6_cache
*)c
,
1535 ip6mr_destroy_unres(mrt
, (struct mfc6_cache
*)c
);
1537 spin_unlock_bh(&mfc_unres_lock
);
1542 static int ip6mr_sk_init(struct mr_table
*mrt
, struct sock
*sk
)
1545 struct net
*net
= sock_net(sk
);
1548 write_lock_bh(&mrt_lock
);
1549 if (rtnl_dereference(mrt
->mroute_sk
)) {
1552 rcu_assign_pointer(mrt
->mroute_sk
, sk
);
1553 sock_set_flag(sk
, SOCK_RCU_FREE
);
1554 net
->ipv6
.devconf_all
->mc_forwarding
++;
1556 write_unlock_bh(&mrt_lock
);
1559 inet6_netconf_notify_devconf(net
, RTM_NEWNETCONF
,
1560 NETCONFA_MC_FORWARDING
,
1561 NETCONFA_IFINDEX_ALL
,
1562 net
->ipv6
.devconf_all
);
1568 int ip6mr_sk_done(struct sock
*sk
)
1571 struct net
*net
= sock_net(sk
);
1572 struct mr_table
*mrt
;
1574 if (sk
->sk_type
!= SOCK_RAW
||
1575 inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1579 ip6mr_for_each_table(mrt
, net
) {
1580 if (sk
== rtnl_dereference(mrt
->mroute_sk
)) {
1581 write_lock_bh(&mrt_lock
);
1582 RCU_INIT_POINTER(mrt
->mroute_sk
, NULL
);
1583 /* Note that mroute_sk had SOCK_RCU_FREE set,
1584 * so the RCU grace period before sk freeing
1585 * is guaranteed by sk_destruct()
1587 net
->ipv6
.devconf_all
->mc_forwarding
--;
1588 write_unlock_bh(&mrt_lock
);
1589 inet6_netconf_notify_devconf(net
, RTM_NEWNETCONF
,
1590 NETCONFA_MC_FORWARDING
,
1591 NETCONFA_IFINDEX_ALL
,
1592 net
->ipv6
.devconf_all
);
1594 mroute_clean_tables(mrt
, MRT6_FLUSH_MIFS
| MRT6_FLUSH_MFC
);
1604 bool mroute6_is_socket(struct net
*net
, struct sk_buff
*skb
)
1606 struct mr_table
*mrt
;
1607 struct flowi6 fl6
= {
1608 .flowi6_iif
= skb
->skb_iif
? : LOOPBACK_IFINDEX
,
1609 .flowi6_oif
= skb
->dev
->ifindex
,
1610 .flowi6_mark
= skb
->mark
,
1613 if (ip6mr_fib_lookup(net
, &fl6
, &mrt
) < 0)
1616 return rcu_access_pointer(mrt
->mroute_sk
);
1618 EXPORT_SYMBOL(mroute6_is_socket
);
1621 * Socket options and virtual interface manipulation. The whole
1622 * virtual interface system is a complete heap, but unfortunately
1623 * that's how BSD mrouted happens to think. Maybe one day with a proper
1624 * MOSPF/PIM router set up we can clean this up.
1627 int ip6_mroute_setsockopt(struct sock
*sk
, int optname
, char __user
*optval
, unsigned int optlen
)
1629 int ret
, parent
= 0;
1633 struct net
*net
= sock_net(sk
);
1634 struct mr_table
*mrt
;
1636 if (sk
->sk_type
!= SOCK_RAW
||
1637 inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1640 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1644 if (optname
!= MRT6_INIT
) {
1645 if (sk
!= rcu_access_pointer(mrt
->mroute_sk
) &&
1646 !ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1652 if (optlen
< sizeof(int))
1655 return ip6mr_sk_init(mrt
, sk
);
1658 return ip6mr_sk_done(sk
);
1661 if (optlen
< sizeof(vif
))
1663 if (copy_from_user(&vif
, optval
, sizeof(vif
)))
1665 if (vif
.mif6c_mifi
>= MAXMIFS
)
1668 ret
= mif6_add(net
, mrt
, &vif
,
1669 sk
== rtnl_dereference(mrt
->mroute_sk
));
1674 if (optlen
< sizeof(mifi_t
))
1676 if (copy_from_user(&mifi
, optval
, sizeof(mifi_t
)))
1679 ret
= mif6_delete(mrt
, mifi
, 0, NULL
);
1684 * Manipulate the forwarding caches. These live
1685 * in a sort of kernel/user symbiosis.
1691 case MRT6_ADD_MFC_PROXY
:
1692 case MRT6_DEL_MFC_PROXY
:
1693 if (optlen
< sizeof(mfc
))
1695 if (copy_from_user(&mfc
, optval
, sizeof(mfc
)))
1698 parent
= mfc
.mf6cc_parent
;
1700 if (optname
== MRT6_DEL_MFC
|| optname
== MRT6_DEL_MFC_PROXY
)
1701 ret
= ip6mr_mfc_delete(mrt
, &mfc
, parent
);
1703 ret
= ip6mr_mfc_add(net
, mrt
, &mfc
,
1705 rtnl_dereference(mrt
->mroute_sk
),
1714 if (optlen
!= sizeof(flags
))
1716 if (get_user(flags
, (int __user
*)optval
))
1719 mroute_clean_tables(mrt
, flags
);
1725 * Control PIM assert (to activate pim will activate assert)
1731 if (optlen
!= sizeof(v
))
1733 if (get_user(v
, (int __user
*)optval
))
1735 mrt
->mroute_do_assert
= v
;
1739 #ifdef CONFIG_IPV6_PIMSM_V2
1744 if (optlen
!= sizeof(v
))
1746 if (get_user(v
, (int __user
*)optval
))
1751 if (v
!= mrt
->mroute_do_pim
) {
1752 mrt
->mroute_do_pim
= v
;
1753 mrt
->mroute_do_assert
= v
;
1760 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1765 if (optlen
!= sizeof(u32
))
1767 if (get_user(v
, (u32 __user
*)optval
))
1769 /* "pim6reg%u" should not exceed 16 bytes (IFNAMSIZ) */
1770 if (v
!= RT_TABLE_DEFAULT
&& v
>= 100000000)
1772 if (sk
== rcu_access_pointer(mrt
->mroute_sk
))
1777 mrt
= ip6mr_new_table(net
, v
);
1781 raw6_sk(sk
)->ip6mr_table
= v
;
1787 * Spurious command, or MRT6_VERSION which you cannot
1791 return -ENOPROTOOPT
;
1796 * Getsock opt support for the multicast routing system.
1799 int ip6_mroute_getsockopt(struct sock
*sk
, int optname
, char __user
*optval
,
1804 struct net
*net
= sock_net(sk
);
1805 struct mr_table
*mrt
;
1807 if (sk
->sk_type
!= SOCK_RAW
||
1808 inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1811 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1819 #ifdef CONFIG_IPV6_PIMSM_V2
1821 val
= mrt
->mroute_do_pim
;
1825 val
= mrt
->mroute_do_assert
;
1828 return -ENOPROTOOPT
;
1831 if (get_user(olr
, optlen
))
1834 olr
= min_t(int, olr
, sizeof(int));
1838 if (put_user(olr
, optlen
))
1840 if (copy_to_user(optval
, &val
, olr
))
1846 * The IP multicast ioctl support routines.
1849 int ip6mr_ioctl(struct sock
*sk
, int cmd
, void __user
*arg
)
1851 struct sioc_sg_req6 sr
;
1852 struct sioc_mif_req6 vr
;
1853 struct vif_device
*vif
;
1854 struct mfc6_cache
*c
;
1855 struct net
*net
= sock_net(sk
);
1856 struct mr_table
*mrt
;
1858 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1863 case SIOCGETMIFCNT_IN6
:
1864 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1866 if (vr
.mifi
>= mrt
->maxvif
)
1868 vr
.mifi
= array_index_nospec(vr
.mifi
, mrt
->maxvif
);
1869 read_lock(&mrt_lock
);
1870 vif
= &mrt
->vif_table
[vr
.mifi
];
1871 if (VIF_EXISTS(mrt
, vr
.mifi
)) {
1872 vr
.icount
= vif
->pkt_in
;
1873 vr
.ocount
= vif
->pkt_out
;
1874 vr
.ibytes
= vif
->bytes_in
;
1875 vr
.obytes
= vif
->bytes_out
;
1876 read_unlock(&mrt_lock
);
1878 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1882 read_unlock(&mrt_lock
);
1883 return -EADDRNOTAVAIL
;
1884 case SIOCGETSGCNT_IN6
:
1885 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1889 c
= ip6mr_cache_find(mrt
, &sr
.src
.sin6_addr
, &sr
.grp
.sin6_addr
);
1891 sr
.pktcnt
= c
->_c
.mfc_un
.res
.pkt
;
1892 sr
.bytecnt
= c
->_c
.mfc_un
.res
.bytes
;
1893 sr
.wrong_if
= c
->_c
.mfc_un
.res
.wrong_if
;
1896 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1901 return -EADDRNOTAVAIL
;
1903 return -ENOIOCTLCMD
;
1907 #ifdef CONFIG_COMPAT
1908 struct compat_sioc_sg_req6
{
1909 struct sockaddr_in6 src
;
1910 struct sockaddr_in6 grp
;
1911 compat_ulong_t pktcnt
;
1912 compat_ulong_t bytecnt
;
1913 compat_ulong_t wrong_if
;
1916 struct compat_sioc_mif_req6
{
1918 compat_ulong_t icount
;
1919 compat_ulong_t ocount
;
1920 compat_ulong_t ibytes
;
1921 compat_ulong_t obytes
;
1924 int ip6mr_compat_ioctl(struct sock
*sk
, unsigned int cmd
, void __user
*arg
)
1926 struct compat_sioc_sg_req6 sr
;
1927 struct compat_sioc_mif_req6 vr
;
1928 struct vif_device
*vif
;
1929 struct mfc6_cache
*c
;
1930 struct net
*net
= sock_net(sk
);
1931 struct mr_table
*mrt
;
1933 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1938 case SIOCGETMIFCNT_IN6
:
1939 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1941 if (vr
.mifi
>= mrt
->maxvif
)
1943 vr
.mifi
= array_index_nospec(vr
.mifi
, mrt
->maxvif
);
1944 read_lock(&mrt_lock
);
1945 vif
= &mrt
->vif_table
[vr
.mifi
];
1946 if (VIF_EXISTS(mrt
, vr
.mifi
)) {
1947 vr
.icount
= vif
->pkt_in
;
1948 vr
.ocount
= vif
->pkt_out
;
1949 vr
.ibytes
= vif
->bytes_in
;
1950 vr
.obytes
= vif
->bytes_out
;
1951 read_unlock(&mrt_lock
);
1953 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1957 read_unlock(&mrt_lock
);
1958 return -EADDRNOTAVAIL
;
1959 case SIOCGETSGCNT_IN6
:
1960 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1964 c
= ip6mr_cache_find(mrt
, &sr
.src
.sin6_addr
, &sr
.grp
.sin6_addr
);
1966 sr
.pktcnt
= c
->_c
.mfc_un
.res
.pkt
;
1967 sr
.bytecnt
= c
->_c
.mfc_un
.res
.bytes
;
1968 sr
.wrong_if
= c
->_c
.mfc_un
.res
.wrong_if
;
1971 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1976 return -EADDRNOTAVAIL
;
1978 return -ENOIOCTLCMD
;
1983 static inline int ip6mr_forward2_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
1985 IP6_INC_STATS(net
, ip6_dst_idev(skb_dst(skb
)),
1986 IPSTATS_MIB_OUTFORWDATAGRAMS
);
1987 IP6_ADD_STATS(net
, ip6_dst_idev(skb_dst(skb
)),
1988 IPSTATS_MIB_OUTOCTETS
, skb
->len
);
1989 return dst_output(net
, sk
, skb
);
1993 * Processing handlers for ip6mr_forward
1996 static int ip6mr_forward2(struct net
*net
, struct mr_table
*mrt
,
1997 struct sk_buff
*skb
, int vifi
)
1999 struct ipv6hdr
*ipv6h
;
2000 struct vif_device
*vif
= &mrt
->vif_table
[vifi
];
2001 struct net_device
*dev
;
2002 struct dst_entry
*dst
;
2008 #ifdef CONFIG_IPV6_PIMSM_V2
2009 if (vif
->flags
& MIFF_REGISTER
) {
2011 vif
->bytes_out
+= skb
->len
;
2012 vif
->dev
->stats
.tx_bytes
+= skb
->len
;
2013 vif
->dev
->stats
.tx_packets
++;
2014 ip6mr_cache_report(mrt
, skb
, vifi
, MRT6MSG_WHOLEPKT
);
2019 ipv6h
= ipv6_hdr(skb
);
2021 fl6
= (struct flowi6
) {
2022 .flowi6_oif
= vif
->link
,
2023 .daddr
= ipv6h
->daddr
,
2026 dst
= ip6_route_output(net
, NULL
, &fl6
);
2033 skb_dst_set(skb
, dst
);
2036 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
2037 * not only before forwarding, but after forwarding on all output
2038 * interfaces. It is clear, if mrouter runs a multicasting
2039 * program, it should receive packets not depending to what interface
2040 * program is joined.
2041 * If we will not make it, the program will have to join on all
2042 * interfaces. On the other hand, multihoming host (or router, but
2043 * not mrouter) cannot join to more than one interface - it will
2044 * result in receiving multiple packets.
2049 vif
->bytes_out
+= skb
->len
;
2051 /* We are about to write */
2052 /* XXX: extension headers? */
2053 if (skb_cow(skb
, sizeof(*ipv6h
) + LL_RESERVED_SPACE(dev
)))
2056 ipv6h
= ipv6_hdr(skb
);
2059 IP6CB(skb
)->flags
|= IP6SKB_FORWARDED
;
2061 return NF_HOOK(NFPROTO_IPV6
, NF_INET_FORWARD
,
2062 net
, NULL
, skb
, skb
->dev
, dev
,
2063 ip6mr_forward2_finish
);
2070 static int ip6mr_find_vif(struct mr_table
*mrt
, struct net_device
*dev
)
2074 for (ct
= mrt
->maxvif
- 1; ct
>= 0; ct
--) {
2075 if (mrt
->vif_table
[ct
].dev
== dev
)
2081 static void ip6_mr_forward(struct net
*net
, struct mr_table
*mrt
,
2082 struct net_device
*dev
, struct sk_buff
*skb
,
2083 struct mfc6_cache
*c
)
2087 int true_vifi
= ip6mr_find_vif(mrt
, dev
);
2089 vif
= c
->_c
.mfc_parent
;
2090 c
->_c
.mfc_un
.res
.pkt
++;
2091 c
->_c
.mfc_un
.res
.bytes
+= skb
->len
;
2092 c
->_c
.mfc_un
.res
.lastuse
= jiffies
;
2094 if (ipv6_addr_any(&c
->mf6c_origin
) && true_vifi
>= 0) {
2095 struct mfc6_cache
*cache_proxy
;
2097 /* For an (*,G) entry, we only check that the incoming
2098 * interface is part of the static tree.
2101 cache_proxy
= mr_mfc_find_any_parent(mrt
, vif
);
2103 cache_proxy
->_c
.mfc_un
.res
.ttls
[true_vifi
] < 255) {
2111 * Wrong interface: drop packet and (maybe) send PIM assert.
2113 if (mrt
->vif_table
[vif
].dev
!= dev
) {
2114 c
->_c
.mfc_un
.res
.wrong_if
++;
2116 if (true_vifi
>= 0 && mrt
->mroute_do_assert
&&
2117 /* pimsm uses asserts, when switching from RPT to SPT,
2118 so that we cannot check that packet arrived on an oif.
2119 It is bad, but otherwise we would need to move pretty
2120 large chunk of pimd to kernel. Ough... --ANK
2122 (mrt
->mroute_do_pim
||
2123 c
->_c
.mfc_un
.res
.ttls
[true_vifi
] < 255) &&
2125 c
->_c
.mfc_un
.res
.last_assert
+
2126 MFC_ASSERT_THRESH
)) {
2127 c
->_c
.mfc_un
.res
.last_assert
= jiffies
;
2128 ip6mr_cache_report(mrt
, skb
, true_vifi
, MRT6MSG_WRONGMIF
);
2134 mrt
->vif_table
[vif
].pkt_in
++;
2135 mrt
->vif_table
[vif
].bytes_in
+= skb
->len
;
2140 if (ipv6_addr_any(&c
->mf6c_origin
) &&
2141 ipv6_addr_any(&c
->mf6c_mcastgrp
)) {
2142 if (true_vifi
>= 0 &&
2143 true_vifi
!= c
->_c
.mfc_parent
&&
2144 ipv6_hdr(skb
)->hop_limit
>
2145 c
->_c
.mfc_un
.res
.ttls
[c
->_c
.mfc_parent
]) {
2146 /* It's an (*,*) entry and the packet is not coming from
2147 * the upstream: forward the packet to the upstream
2150 psend
= c
->_c
.mfc_parent
;
2155 for (ct
= c
->_c
.mfc_un
.res
.maxvif
- 1;
2156 ct
>= c
->_c
.mfc_un
.res
.minvif
; ct
--) {
2157 /* For (*,G) entry, don't forward to the incoming interface */
2158 if ((!ipv6_addr_any(&c
->mf6c_origin
) || ct
!= true_vifi
) &&
2159 ipv6_hdr(skb
)->hop_limit
> c
->_c
.mfc_un
.res
.ttls
[ct
]) {
2161 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
2163 ip6mr_forward2(net
, mrt
, skb2
, psend
);
2170 ip6mr_forward2(net
, mrt
, skb
, psend
);
2180 * Multicast packets for forwarding arrive here
2183 int ip6_mr_input(struct sk_buff
*skb
)
2185 struct mfc6_cache
*cache
;
2186 struct net
*net
= dev_net(skb
->dev
);
2187 struct mr_table
*mrt
;
2188 struct flowi6 fl6
= {
2189 .flowi6_iif
= skb
->dev
->ifindex
,
2190 .flowi6_mark
= skb
->mark
,
2193 struct net_device
*dev
;
2195 /* skb->dev passed in is the master dev for vrfs.
2196 * Get the proper interface that does have a vif associated with it.
2199 if (netif_is_l3_master(skb
->dev
)) {
2200 dev
= dev_get_by_index_rcu(net
, IPCB(skb
)->iif
);
2207 err
= ip6mr_fib_lookup(net
, &fl6
, &mrt
);
2213 read_lock(&mrt_lock
);
2214 cache
= ip6mr_cache_find(mrt
,
2215 &ipv6_hdr(skb
)->saddr
, &ipv6_hdr(skb
)->daddr
);
2217 int vif
= ip6mr_find_vif(mrt
, dev
);
2220 cache
= ip6mr_cache_find_any(mrt
,
2221 &ipv6_hdr(skb
)->daddr
,
2226 * No usable cache entry
2231 vif
= ip6mr_find_vif(mrt
, dev
);
2233 int err
= ip6mr_cache_unresolved(mrt
, vif
, skb
, dev
);
2234 read_unlock(&mrt_lock
);
2238 read_unlock(&mrt_lock
);
2243 ip6_mr_forward(net
, mrt
, dev
, skb
, cache
);
2245 read_unlock(&mrt_lock
);
2250 int ip6mr_get_route(struct net
*net
, struct sk_buff
*skb
, struct rtmsg
*rtm
,
2254 struct mr_table
*mrt
;
2255 struct mfc6_cache
*cache
;
2256 struct rt6_info
*rt
= (struct rt6_info
*)skb_dst(skb
);
2258 mrt
= ip6mr_get_table(net
, RT6_TABLE_DFLT
);
2262 read_lock(&mrt_lock
);
2263 cache
= ip6mr_cache_find(mrt
, &rt
->rt6i_src
.addr
, &rt
->rt6i_dst
.addr
);
2264 if (!cache
&& skb
->dev
) {
2265 int vif
= ip6mr_find_vif(mrt
, skb
->dev
);
2268 cache
= ip6mr_cache_find_any(mrt
, &rt
->rt6i_dst
.addr
,
2273 struct sk_buff
*skb2
;
2274 struct ipv6hdr
*iph
;
2275 struct net_device
*dev
;
2279 if (!dev
|| (vif
= ip6mr_find_vif(mrt
, dev
)) < 0) {
2280 read_unlock(&mrt_lock
);
2284 /* really correct? */
2285 skb2
= alloc_skb(sizeof(struct ipv6hdr
), GFP_ATOMIC
);
2287 read_unlock(&mrt_lock
);
2291 NETLINK_CB(skb2
).portid
= portid
;
2292 skb_reset_transport_header(skb2
);
2294 skb_put(skb2
, sizeof(struct ipv6hdr
));
2295 skb_reset_network_header(skb2
);
2297 iph
= ipv6_hdr(skb2
);
2300 iph
->flow_lbl
[0] = 0;
2301 iph
->flow_lbl
[1] = 0;
2302 iph
->flow_lbl
[2] = 0;
2303 iph
->payload_len
= 0;
2304 iph
->nexthdr
= IPPROTO_NONE
;
2306 iph
->saddr
= rt
->rt6i_src
.addr
;
2307 iph
->daddr
= rt
->rt6i_dst
.addr
;
2309 err
= ip6mr_cache_unresolved(mrt
, vif
, skb2
, dev
);
2310 read_unlock(&mrt_lock
);
2315 err
= mr_fill_mroute(mrt
, skb
, &cache
->_c
, rtm
);
2316 read_unlock(&mrt_lock
);
2320 static int ip6mr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
2321 u32 portid
, u32 seq
, struct mfc6_cache
*c
, int cmd
,
2324 struct nlmsghdr
*nlh
;
2328 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rtm
), flags
);
2332 rtm
= nlmsg_data(nlh
);
2333 rtm
->rtm_family
= RTNL_FAMILY_IP6MR
;
2334 rtm
->rtm_dst_len
= 128;
2335 rtm
->rtm_src_len
= 128;
2337 rtm
->rtm_table
= mrt
->id
;
2338 if (nla_put_u32(skb
, RTA_TABLE
, mrt
->id
))
2339 goto nla_put_failure
;
2340 rtm
->rtm_type
= RTN_MULTICAST
;
2341 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2342 if (c
->_c
.mfc_flags
& MFC_STATIC
)
2343 rtm
->rtm_protocol
= RTPROT_STATIC
;
2345 rtm
->rtm_protocol
= RTPROT_MROUTED
;
2348 if (nla_put_in6_addr(skb
, RTA_SRC
, &c
->mf6c_origin
) ||
2349 nla_put_in6_addr(skb
, RTA_DST
, &c
->mf6c_mcastgrp
))
2350 goto nla_put_failure
;
2351 err
= mr_fill_mroute(mrt
, skb
, &c
->_c
, rtm
);
2352 /* do not break the dump if cache is unresolved */
2353 if (err
< 0 && err
!= -ENOENT
)
2354 goto nla_put_failure
;
2356 nlmsg_end(skb
, nlh
);
2360 nlmsg_cancel(skb
, nlh
);
2364 static int _ip6mr_fill_mroute(struct mr_table
*mrt
, struct sk_buff
*skb
,
2365 u32 portid
, u32 seq
, struct mr_mfc
*c
,
2368 return ip6mr_fill_mroute(mrt
, skb
, portid
, seq
, (struct mfc6_cache
*)c
,
2372 static int mr6_msgsize(bool unresolved
, int maxvif
)
2375 NLMSG_ALIGN(sizeof(struct rtmsg
))
2376 + nla_total_size(4) /* RTA_TABLE */
2377 + nla_total_size(sizeof(struct in6_addr
)) /* RTA_SRC */
2378 + nla_total_size(sizeof(struct in6_addr
)) /* RTA_DST */
2383 + nla_total_size(4) /* RTA_IIF */
2384 + nla_total_size(0) /* RTA_MULTIPATH */
2385 + maxvif
* NLA_ALIGN(sizeof(struct rtnexthop
))
2387 + nla_total_size_64bit(sizeof(struct rta_mfc_stats
))
2393 static void mr6_netlink_event(struct mr_table
*mrt
, struct mfc6_cache
*mfc
,
2396 struct net
*net
= read_pnet(&mrt
->net
);
2397 struct sk_buff
*skb
;
2400 skb
= nlmsg_new(mr6_msgsize(mfc
->_c
.mfc_parent
>= MAXMIFS
, mrt
->maxvif
),
2405 err
= ip6mr_fill_mroute(mrt
, skb
, 0, 0, mfc
, cmd
, 0);
2409 rtnl_notify(skb
, net
, 0, RTNLGRP_IPV6_MROUTE
, NULL
, GFP_ATOMIC
);
2415 rtnl_set_sk_err(net
, RTNLGRP_IPV6_MROUTE
, err
);
2418 static size_t mrt6msg_netlink_msgsize(size_t payloadlen
)
2421 NLMSG_ALIGN(sizeof(struct rtgenmsg
))
2422 + nla_total_size(1) /* IP6MRA_CREPORT_MSGTYPE */
2423 + nla_total_size(4) /* IP6MRA_CREPORT_MIF_ID */
2424 /* IP6MRA_CREPORT_SRC_ADDR */
2425 + nla_total_size(sizeof(struct in6_addr
))
2426 /* IP6MRA_CREPORT_DST_ADDR */
2427 + nla_total_size(sizeof(struct in6_addr
))
2428 /* IP6MRA_CREPORT_PKT */
2429 + nla_total_size(payloadlen
)
2435 static void mrt6msg_netlink_event(struct mr_table
*mrt
, struct sk_buff
*pkt
)
2437 struct net
*net
= read_pnet(&mrt
->net
);
2438 struct nlmsghdr
*nlh
;
2439 struct rtgenmsg
*rtgenm
;
2440 struct mrt6msg
*msg
;
2441 struct sk_buff
*skb
;
2445 payloadlen
= pkt
->len
- sizeof(struct mrt6msg
);
2446 msg
= (struct mrt6msg
*)skb_transport_header(pkt
);
2448 skb
= nlmsg_new(mrt6msg_netlink_msgsize(payloadlen
), GFP_ATOMIC
);
2452 nlh
= nlmsg_put(skb
, 0, 0, RTM_NEWCACHEREPORT
,
2453 sizeof(struct rtgenmsg
), 0);
2456 rtgenm
= nlmsg_data(nlh
);
2457 rtgenm
->rtgen_family
= RTNL_FAMILY_IP6MR
;
2458 if (nla_put_u8(skb
, IP6MRA_CREPORT_MSGTYPE
, msg
->im6_msgtype
) ||
2459 nla_put_u32(skb
, IP6MRA_CREPORT_MIF_ID
, msg
->im6_mif
) ||
2460 nla_put_in6_addr(skb
, IP6MRA_CREPORT_SRC_ADDR
,
2462 nla_put_in6_addr(skb
, IP6MRA_CREPORT_DST_ADDR
,
2464 goto nla_put_failure
;
2466 nla
= nla_reserve(skb
, IP6MRA_CREPORT_PKT
, payloadlen
);
2467 if (!nla
|| skb_copy_bits(pkt
, sizeof(struct mrt6msg
),
2468 nla_data(nla
), payloadlen
))
2469 goto nla_put_failure
;
2471 nlmsg_end(skb
, nlh
);
2473 rtnl_notify(skb
, net
, 0, RTNLGRP_IPV6_MROUTE_R
, NULL
, GFP_ATOMIC
);
2477 nlmsg_cancel(skb
, nlh
);
2480 rtnl_set_sk_err(net
, RTNLGRP_IPV6_MROUTE_R
, -ENOBUFS
);
2483 static int ip6mr_rtm_dumproute(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2485 const struct nlmsghdr
*nlh
= cb
->nlh
;
2486 struct fib_dump_filter filter
= {};
2489 if (cb
->strict_check
) {
2490 err
= ip_valid_fib_dump_req(sock_net(skb
->sk
), nlh
,
2496 if (filter
.table_id
) {
2497 struct mr_table
*mrt
;
2499 mrt
= ip6mr_get_table(sock_net(skb
->sk
), filter
.table_id
);
2501 if (filter
.dump_all_families
)
2504 NL_SET_ERR_MSG_MOD(cb
->extack
, "MR table does not exist");
2507 err
= mr_table_dump(mrt
, skb
, cb
, _ip6mr_fill_mroute
,
2508 &mfc_unres_lock
, &filter
);
2509 return skb
->len
? : err
;
2512 return mr_rtm_dumproute(skb
, cb
, ip6mr_mr_table_iter
,
2513 _ip6mr_fill_mroute
, &mfc_unres_lock
, &filter
);