2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
19 #include <asm/uaccess.h>
20 #include <linux/types.h>
21 #include <linux/sched.h>
22 #include <linux/errno.h>
23 #include <linux/timer.h>
25 #include <linux/kernel.h>
26 #include <linux/fcntl.h>
27 #include <linux/stat.h>
28 #include <linux/socket.h>
29 #include <linux/inet.h>
30 #include <linux/netdevice.h>
31 #include <linux/inetdevice.h>
32 #include <linux/proc_fs.h>
33 #include <linux/seq_file.h>
34 #include <linux/init.h>
35 #include <linux/slab.h>
36 #include <linux/compat.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
41 #include <linux/notifier.h>
42 #include <linux/if_arp.h>
43 #include <net/checksum.h>
44 #include <net/netlink.h>
45 #include <net/fib_rules.h>
48 #include <net/ip6_route.h>
49 #include <linux/mroute6.h>
50 #include <linux/pim.h>
51 #include <net/addrconf.h>
52 #include <linux/netfilter_ipv6.h>
53 #include <linux/export.h>
54 #include <net/ip6_checksum.h>
57 struct list_head list
;
62 struct sock
*mroute6_sk
;
63 struct timer_list ipmr_expire_timer
;
64 struct list_head mfc6_unres_queue
;
65 struct list_head mfc6_cache_array
[MFC6_LINES
];
66 struct mif_device vif6_table
[MAXMIFS
];
68 atomic_t cache_resolve_queue_len
;
69 bool mroute_do_assert
;
71 #ifdef CONFIG_IPV6_PIMSM_V2
72 int mroute_reg_vif_num
;
77 struct fib_rule common
;
81 struct mr6_table
*mrt
;
84 /* Big lock, protecting vif table, mrt cache and mroute socket state.
85 Note that the changes are semaphored via rtnl_lock.
88 static DEFINE_RWLOCK(mrt_lock
);
91 * Multicast router control variables
94 #define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
96 /* Special spinlock for queue of unresolved entries */
97 static DEFINE_SPINLOCK(mfc_unres_lock
);
99 /* We return to original Alan's scheme. Hash table of resolved
100 entries is changed only in process context and protected
101 with weak lock mrt_lock. Queue of unresolved entries is protected
102 with strong spinlock mfc_unres_lock.
104 In this case data path is free of exclusive locks at all.
107 static struct kmem_cache
*mrt_cachep __read_mostly
;
109 static struct mr6_table
*ip6mr_new_table(struct net
*net
, u32 id
);
110 static void ip6mr_free_table(struct mr6_table
*mrt
);
112 static int ip6_mr_forward(struct net
*net
, struct mr6_table
*mrt
,
113 struct sk_buff
*skb
, struct mfc6_cache
*cache
);
114 static int ip6mr_cache_report(struct mr6_table
*mrt
, struct sk_buff
*pkt
,
115 mifi_t mifi
, int assert);
116 static int __ip6mr_fill_mroute(struct mr6_table
*mrt
, struct sk_buff
*skb
,
117 struct mfc6_cache
*c
, struct rtmsg
*rtm
);
118 static int ip6mr_rtm_dumproute(struct sk_buff
*skb
,
119 struct netlink_callback
*cb
);
120 static void mroute_clean_tables(struct mr6_table
*mrt
);
121 static void ipmr_expire_process(unsigned long arg
);
123 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
124 #define ip6mr_for_each_table(mrt, net) \
125 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
127 static struct mr6_table
*ip6mr_get_table(struct net
*net
, u32 id
)
129 struct mr6_table
*mrt
;
131 ip6mr_for_each_table(mrt
, net
) {
138 static int ip6mr_fib_lookup(struct net
*net
, struct flowi6
*flp6
,
139 struct mr6_table
**mrt
)
141 struct ip6mr_result res
;
142 struct fib_lookup_arg arg
= { .result
= &res
, };
145 err
= fib_rules_lookup(net
->ipv6
.mr6_rules_ops
,
146 flowi6_to_flowi(flp6
), 0, &arg
);
153 static int ip6mr_rule_action(struct fib_rule
*rule
, struct flowi
*flp
,
154 int flags
, struct fib_lookup_arg
*arg
)
156 struct ip6mr_result
*res
= arg
->result
;
157 struct mr6_table
*mrt
;
159 switch (rule
->action
) {
162 case FR_ACT_UNREACHABLE
:
164 case FR_ACT_PROHIBIT
:
166 case FR_ACT_BLACKHOLE
:
171 mrt
= ip6mr_get_table(rule
->fr_net
, rule
->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
)
193 static int ip6mr_rule_compare(struct fib_rule
*rule
, struct fib_rule_hdr
*frh
,
199 static int ip6mr_rule_fill(struct fib_rule
*rule
, struct sk_buff
*skb
,
200 struct fib_rule_hdr
*frh
)
208 static const struct fib_rules_ops __net_initconst ip6mr_rules_ops_template
= {
209 .family
= RTNL_FAMILY_IP6MR
,
210 .rule_size
= sizeof(struct ip6mr_rule
),
211 .addr_size
= sizeof(struct in6_addr
),
212 .action
= ip6mr_rule_action
,
213 .match
= ip6mr_rule_match
,
214 .configure
= ip6mr_rule_configure
,
215 .compare
= ip6mr_rule_compare
,
216 .default_pref
= fib_default_rule_pref
,
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 mr6_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
;
251 fib_rules_unregister(ops
);
255 static void __net_exit
ip6mr_rules_exit(struct net
*net
)
257 struct mr6_table
*mrt
, *next
;
259 list_for_each_entry_safe(mrt
, next
, &net
->ipv6
.mr6_tables
, list
) {
260 list_del(&mrt
->list
);
261 ip6mr_free_table(mrt
);
263 fib_rules_unregister(net
->ipv6
.mr6_rules_ops
);
266 #define ip6mr_for_each_table(mrt, net) \
267 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
269 static struct mr6_table
*ip6mr_get_table(struct net
*net
, u32 id
)
271 return net
->ipv6
.mrt6
;
274 static int ip6mr_fib_lookup(struct net
*net
, struct flowi6
*flp6
,
275 struct mr6_table
**mrt
)
277 *mrt
= net
->ipv6
.mrt6
;
281 static int __net_init
ip6mr_rules_init(struct net
*net
)
283 net
->ipv6
.mrt6
= ip6mr_new_table(net
, RT6_TABLE_DFLT
);
284 return net
->ipv6
.mrt6
? 0 : -ENOMEM
;
287 static void __net_exit
ip6mr_rules_exit(struct net
*net
)
289 ip6mr_free_table(net
->ipv6
.mrt6
);
293 static struct mr6_table
*ip6mr_new_table(struct net
*net
, u32 id
)
295 struct mr6_table
*mrt
;
298 mrt
= ip6mr_get_table(net
, id
);
302 mrt
= kzalloc(sizeof(*mrt
), GFP_KERNEL
);
306 write_pnet(&mrt
->net
, net
);
308 /* Forwarding cache */
309 for (i
= 0; i
< MFC6_LINES
; i
++)
310 INIT_LIST_HEAD(&mrt
->mfc6_cache_array
[i
]);
312 INIT_LIST_HEAD(&mrt
->mfc6_unres_queue
);
314 setup_timer(&mrt
->ipmr_expire_timer
, ipmr_expire_process
,
317 #ifdef CONFIG_IPV6_PIMSM_V2
318 mrt
->mroute_reg_vif_num
= -1;
320 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
321 list_add_tail_rcu(&mrt
->list
, &net
->ipv6
.mr6_tables
);
326 static void ip6mr_free_table(struct mr6_table
*mrt
)
328 del_timer(&mrt
->ipmr_expire_timer
);
329 mroute_clean_tables(mrt
);
333 #ifdef CONFIG_PROC_FS
335 struct ipmr_mfc_iter
{
336 struct seq_net_private p
;
337 struct mr6_table
*mrt
;
338 struct list_head
*cache
;
343 static struct mfc6_cache
*ipmr_mfc_seq_idx(struct net
*net
,
344 struct ipmr_mfc_iter
*it
, loff_t pos
)
346 struct mr6_table
*mrt
= it
->mrt
;
347 struct mfc6_cache
*mfc
;
349 read_lock(&mrt_lock
);
350 for (it
->ct
= 0; it
->ct
< MFC6_LINES
; it
->ct
++) {
351 it
->cache
= &mrt
->mfc6_cache_array
[it
->ct
];
352 list_for_each_entry(mfc
, it
->cache
, list
)
356 read_unlock(&mrt_lock
);
358 spin_lock_bh(&mfc_unres_lock
);
359 it
->cache
= &mrt
->mfc6_unres_queue
;
360 list_for_each_entry(mfc
, it
->cache
, list
)
363 spin_unlock_bh(&mfc_unres_lock
);
370 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
373 struct ipmr_vif_iter
{
374 struct seq_net_private p
;
375 struct mr6_table
*mrt
;
379 static struct mif_device
*ip6mr_vif_seq_idx(struct net
*net
,
380 struct ipmr_vif_iter
*iter
,
383 struct mr6_table
*mrt
= iter
->mrt
;
385 for (iter
->ct
= 0; iter
->ct
< mrt
->maxvif
; ++iter
->ct
) {
386 if (!MIF_EXISTS(mrt
, iter
->ct
))
389 return &mrt
->vif6_table
[iter
->ct
];
394 static void *ip6mr_vif_seq_start(struct seq_file
*seq
, loff_t
*pos
)
397 struct ipmr_vif_iter
*iter
= seq
->private;
398 struct net
*net
= seq_file_net(seq
);
399 struct mr6_table
*mrt
;
401 mrt
= ip6mr_get_table(net
, RT6_TABLE_DFLT
);
403 return ERR_PTR(-ENOENT
);
407 read_lock(&mrt_lock
);
408 return *pos
? ip6mr_vif_seq_idx(net
, seq
->private, *pos
- 1)
412 static void *ip6mr_vif_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
414 struct ipmr_vif_iter
*iter
= seq
->private;
415 struct net
*net
= seq_file_net(seq
);
416 struct mr6_table
*mrt
= iter
->mrt
;
419 if (v
== SEQ_START_TOKEN
)
420 return ip6mr_vif_seq_idx(net
, iter
, 0);
422 while (++iter
->ct
< mrt
->maxvif
) {
423 if (!MIF_EXISTS(mrt
, iter
->ct
))
425 return &mrt
->vif6_table
[iter
->ct
];
430 static void ip6mr_vif_seq_stop(struct seq_file
*seq
, void *v
)
433 read_unlock(&mrt_lock
);
436 static int ip6mr_vif_seq_show(struct seq_file
*seq
, void *v
)
438 struct ipmr_vif_iter
*iter
= seq
->private;
439 struct mr6_table
*mrt
= iter
->mrt
;
441 if (v
== SEQ_START_TOKEN
) {
443 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
445 const struct mif_device
*vif
= v
;
446 const char *name
= vif
->dev
? vif
->dev
->name
: "none";
449 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
450 vif
- mrt
->vif6_table
,
451 name
, vif
->bytes_in
, vif
->pkt_in
,
452 vif
->bytes_out
, vif
->pkt_out
,
458 static const struct seq_operations ip6mr_vif_seq_ops
= {
459 .start
= ip6mr_vif_seq_start
,
460 .next
= ip6mr_vif_seq_next
,
461 .stop
= ip6mr_vif_seq_stop
,
462 .show
= ip6mr_vif_seq_show
,
465 static int ip6mr_vif_open(struct inode
*inode
, struct file
*file
)
467 return seq_open_net(inode
, file
, &ip6mr_vif_seq_ops
,
468 sizeof(struct ipmr_vif_iter
));
471 static const struct file_operations ip6mr_vif_fops
= {
472 .owner
= THIS_MODULE
,
473 .open
= ip6mr_vif_open
,
476 .release
= seq_release_net
,
479 static void *ipmr_mfc_seq_start(struct seq_file
*seq
, loff_t
*pos
)
481 struct ipmr_mfc_iter
*it
= seq
->private;
482 struct net
*net
= seq_file_net(seq
);
483 struct mr6_table
*mrt
;
485 mrt
= ip6mr_get_table(net
, RT6_TABLE_DFLT
);
487 return ERR_PTR(-ENOENT
);
490 return *pos
? ipmr_mfc_seq_idx(net
, seq
->private, *pos
- 1)
494 static void *ipmr_mfc_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
496 struct mfc6_cache
*mfc
= v
;
497 struct ipmr_mfc_iter
*it
= seq
->private;
498 struct net
*net
= seq_file_net(seq
);
499 struct mr6_table
*mrt
= it
->mrt
;
503 if (v
== SEQ_START_TOKEN
)
504 return ipmr_mfc_seq_idx(net
, seq
->private, 0);
506 if (mfc
->list
.next
!= it
->cache
)
507 return list_entry(mfc
->list
.next
, struct mfc6_cache
, list
);
509 if (it
->cache
== &mrt
->mfc6_unres_queue
)
512 BUG_ON(it
->cache
!= &mrt
->mfc6_cache_array
[it
->ct
]);
514 while (++it
->ct
< MFC6_LINES
) {
515 it
->cache
= &mrt
->mfc6_cache_array
[it
->ct
];
516 if (list_empty(it
->cache
))
518 return list_first_entry(it
->cache
, struct mfc6_cache
, list
);
521 /* exhausted cache_array, show unresolved */
522 read_unlock(&mrt_lock
);
523 it
->cache
= &mrt
->mfc6_unres_queue
;
526 spin_lock_bh(&mfc_unres_lock
);
527 if (!list_empty(it
->cache
))
528 return list_first_entry(it
->cache
, struct mfc6_cache
, list
);
531 spin_unlock_bh(&mfc_unres_lock
);
537 static void ipmr_mfc_seq_stop(struct seq_file
*seq
, void *v
)
539 struct ipmr_mfc_iter
*it
= seq
->private;
540 struct mr6_table
*mrt
= it
->mrt
;
542 if (it
->cache
== &mrt
->mfc6_unres_queue
)
543 spin_unlock_bh(&mfc_unres_lock
);
544 else if (it
->cache
== mrt
->mfc6_cache_array
)
545 read_unlock(&mrt_lock
);
548 static int ipmr_mfc_seq_show(struct seq_file
*seq
, void *v
)
552 if (v
== SEQ_START_TOKEN
) {
556 "Iif Pkts Bytes Wrong Oifs\n");
558 const struct mfc6_cache
*mfc
= v
;
559 const struct ipmr_mfc_iter
*it
= seq
->private;
560 struct mr6_table
*mrt
= it
->mrt
;
562 seq_printf(seq
, "%pI6 %pI6 %-3hd",
563 &mfc
->mf6c_mcastgrp
, &mfc
->mf6c_origin
,
566 if (it
->cache
!= &mrt
->mfc6_unres_queue
) {
567 seq_printf(seq
, " %8lu %8lu %8lu",
569 mfc
->mfc_un
.res
.bytes
,
570 mfc
->mfc_un
.res
.wrong_if
);
571 for (n
= mfc
->mfc_un
.res
.minvif
;
572 n
< mfc
->mfc_un
.res
.maxvif
; n
++) {
573 if (MIF_EXISTS(mrt
, n
) &&
574 mfc
->mfc_un
.res
.ttls
[n
] < 255)
577 n
, mfc
->mfc_un
.res
.ttls
[n
]);
580 /* unresolved mfc_caches don't contain
581 * pkt, bytes and wrong_if values
583 seq_printf(seq
, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
590 static const struct seq_operations ipmr_mfc_seq_ops
= {
591 .start
= ipmr_mfc_seq_start
,
592 .next
= ipmr_mfc_seq_next
,
593 .stop
= ipmr_mfc_seq_stop
,
594 .show
= ipmr_mfc_seq_show
,
597 static int ipmr_mfc_open(struct inode
*inode
, struct file
*file
)
599 return seq_open_net(inode
, file
, &ipmr_mfc_seq_ops
,
600 sizeof(struct ipmr_mfc_iter
));
603 static const struct file_operations ip6mr_mfc_fops
= {
604 .owner
= THIS_MODULE
,
605 .open
= ipmr_mfc_open
,
608 .release
= seq_release_net
,
612 #ifdef CONFIG_IPV6_PIMSM_V2
614 static int pim6_rcv(struct sk_buff
*skb
)
616 struct pimreghdr
*pim
;
617 struct ipv6hdr
*encap
;
618 struct net_device
*reg_dev
= NULL
;
619 struct net
*net
= dev_net(skb
->dev
);
620 struct mr6_table
*mrt
;
621 struct flowi6 fl6
= {
622 .flowi6_iif
= skb
->dev
->ifindex
,
623 .flowi6_mark
= skb
->mark
,
627 if (!pskb_may_pull(skb
, sizeof(*pim
) + sizeof(*encap
)))
630 pim
= (struct pimreghdr
*)skb_transport_header(skb
);
631 if (pim
->type
!= ((PIM_VERSION
<< 4) | PIM_REGISTER
) ||
632 (pim
->flags
& PIM_NULL_REGISTER
) ||
633 (csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
, &ipv6_hdr(skb
)->daddr
,
634 sizeof(*pim
), IPPROTO_PIM
,
635 csum_partial((void *)pim
, sizeof(*pim
), 0)) &&
636 csum_fold(skb_checksum(skb
, 0, skb
->len
, 0))))
639 /* check if the inner packet is destined to mcast group */
640 encap
= (struct ipv6hdr
*)(skb_transport_header(skb
) +
643 if (!ipv6_addr_is_multicast(&encap
->daddr
) ||
644 encap
->payload_len
== 0 ||
645 ntohs(encap
->payload_len
) + sizeof(*pim
) > skb
->len
)
648 if (ip6mr_fib_lookup(net
, &fl6
, &mrt
) < 0)
650 reg_vif_num
= mrt
->mroute_reg_vif_num
;
652 read_lock(&mrt_lock
);
653 if (reg_vif_num
>= 0)
654 reg_dev
= mrt
->vif6_table
[reg_vif_num
].dev
;
657 read_unlock(&mrt_lock
);
662 skb
->mac_header
= skb
->network_header
;
663 skb_pull(skb
, (u8
*)encap
- skb
->data
);
664 skb_reset_network_header(skb
);
665 skb
->protocol
= htons(ETH_P_IPV6
);
666 skb
->ip_summed
= CHECKSUM_NONE
;
667 skb
->pkt_type
= PACKET_HOST
;
669 skb_tunnel_rx(skb
, reg_dev
);
680 static const struct inet6_protocol pim6_protocol
= {
684 /* Service routines creating virtual interfaces: PIMREG */
686 static netdev_tx_t
reg_vif_xmit(struct sk_buff
*skb
,
687 struct net_device
*dev
)
689 struct net
*net
= dev_net(dev
);
690 struct mr6_table
*mrt
;
691 struct flowi6 fl6
= {
692 .flowi6_oif
= dev
->ifindex
,
693 .flowi6_iif
= skb
->skb_iif
,
694 .flowi6_mark
= skb
->mark
,
698 err
= ip6mr_fib_lookup(net
, &fl6
, &mrt
);
704 read_lock(&mrt_lock
);
705 dev
->stats
.tx_bytes
+= skb
->len
;
706 dev
->stats
.tx_packets
++;
707 ip6mr_cache_report(mrt
, skb
, mrt
->mroute_reg_vif_num
, MRT6MSG_WHOLEPKT
);
708 read_unlock(&mrt_lock
);
713 static const struct net_device_ops reg_vif_netdev_ops
= {
714 .ndo_start_xmit
= reg_vif_xmit
,
717 static void reg_vif_setup(struct net_device
*dev
)
719 dev
->type
= ARPHRD_PIMREG
;
720 dev
->mtu
= 1500 - sizeof(struct ipv6hdr
) - 8;
721 dev
->flags
= IFF_NOARP
;
722 dev
->netdev_ops
= ®_vif_netdev_ops
;
723 dev
->destructor
= free_netdev
;
724 dev
->features
|= NETIF_F_NETNS_LOCAL
;
727 static struct net_device
*ip6mr_reg_vif(struct net
*net
, struct mr6_table
*mrt
)
729 struct net_device
*dev
;
732 if (mrt
->id
== RT6_TABLE_DFLT
)
733 sprintf(name
, "pim6reg");
735 sprintf(name
, "pim6reg%u", mrt
->id
);
737 dev
= alloc_netdev(0, name
, reg_vif_setup
);
741 dev_net_set(dev
, net
);
743 if (register_netdevice(dev
)) {
756 /* allow the register to be completed before unregistering. */
760 unregister_netdevice(dev
);
769 static int mif6_delete(struct mr6_table
*mrt
, int vifi
, struct list_head
*head
)
771 struct mif_device
*v
;
772 struct net_device
*dev
;
773 struct inet6_dev
*in6_dev
;
775 if (vifi
< 0 || vifi
>= mrt
->maxvif
)
776 return -EADDRNOTAVAIL
;
778 v
= &mrt
->vif6_table
[vifi
];
780 write_lock_bh(&mrt_lock
);
785 write_unlock_bh(&mrt_lock
);
786 return -EADDRNOTAVAIL
;
789 #ifdef CONFIG_IPV6_PIMSM_V2
790 if (vifi
== mrt
->mroute_reg_vif_num
)
791 mrt
->mroute_reg_vif_num
= -1;
794 if (vifi
+ 1 == mrt
->maxvif
) {
796 for (tmp
= vifi
- 1; tmp
>= 0; tmp
--) {
797 if (MIF_EXISTS(mrt
, tmp
))
800 mrt
->maxvif
= tmp
+ 1;
803 write_unlock_bh(&mrt_lock
);
805 dev_set_allmulti(dev
, -1);
807 in6_dev
= __in6_dev_get(dev
);
809 in6_dev
->cnf
.mc_forwarding
--;
811 if (v
->flags
& MIFF_REGISTER
)
812 unregister_netdevice_queue(dev
, head
);
818 static inline void ip6mr_cache_free(struct mfc6_cache
*c
)
820 kmem_cache_free(mrt_cachep
, c
);
823 /* Destroy an unresolved cache entry, killing queued skbs
824 and reporting error to netlink readers.
827 static void ip6mr_destroy_unres(struct mr6_table
*mrt
, struct mfc6_cache
*c
)
829 struct net
*net
= read_pnet(&mrt
->net
);
832 atomic_dec(&mrt
->cache_resolve_queue_len
);
834 while((skb
= skb_dequeue(&c
->mfc_un
.unres
.unresolved
)) != NULL
) {
835 if (ipv6_hdr(skb
)->version
== 0) {
836 struct nlmsghdr
*nlh
= (struct nlmsghdr
*)skb_pull(skb
, sizeof(struct ipv6hdr
));
837 nlh
->nlmsg_type
= NLMSG_ERROR
;
838 nlh
->nlmsg_len
= NLMSG_LENGTH(sizeof(struct nlmsgerr
));
839 skb_trim(skb
, nlh
->nlmsg_len
);
840 ((struct nlmsgerr
*)NLMSG_DATA(nlh
))->error
= -ETIMEDOUT
;
841 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).portid
);
850 /* Timer process for all the unresolved queue. */
852 static void ipmr_do_expire_process(struct mr6_table
*mrt
)
854 unsigned long now
= jiffies
;
855 unsigned long expires
= 10 * HZ
;
856 struct mfc6_cache
*c
, *next
;
858 list_for_each_entry_safe(c
, next
, &mrt
->mfc6_unres_queue
, list
) {
859 if (time_after(c
->mfc_un
.unres
.expires
, now
)) {
861 unsigned long interval
= c
->mfc_un
.unres
.expires
- now
;
862 if (interval
< expires
)
868 ip6mr_destroy_unres(mrt
, c
);
871 if (!list_empty(&mrt
->mfc6_unres_queue
))
872 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+ expires
);
875 static void ipmr_expire_process(unsigned long arg
)
877 struct mr6_table
*mrt
= (struct mr6_table
*)arg
;
879 if (!spin_trylock(&mfc_unres_lock
)) {
880 mod_timer(&mrt
->ipmr_expire_timer
, jiffies
+ 1);
884 if (!list_empty(&mrt
->mfc6_unres_queue
))
885 ipmr_do_expire_process(mrt
);
887 spin_unlock(&mfc_unres_lock
);
890 /* Fill oifs list. It is called under write locked mrt_lock. */
892 static void ip6mr_update_thresholds(struct mr6_table
*mrt
, struct mfc6_cache
*cache
,
897 cache
->mfc_un
.res
.minvif
= MAXMIFS
;
898 cache
->mfc_un
.res
.maxvif
= 0;
899 memset(cache
->mfc_un
.res
.ttls
, 255, MAXMIFS
);
901 for (vifi
= 0; vifi
< mrt
->maxvif
; vifi
++) {
902 if (MIF_EXISTS(mrt
, vifi
) &&
903 ttls
[vifi
] && ttls
[vifi
] < 255) {
904 cache
->mfc_un
.res
.ttls
[vifi
] = ttls
[vifi
];
905 if (cache
->mfc_un
.res
.minvif
> vifi
)
906 cache
->mfc_un
.res
.minvif
= vifi
;
907 if (cache
->mfc_un
.res
.maxvif
<= vifi
)
908 cache
->mfc_un
.res
.maxvif
= vifi
+ 1;
913 static int mif6_add(struct net
*net
, struct mr6_table
*mrt
,
914 struct mif6ctl
*vifc
, int mrtsock
)
916 int vifi
= vifc
->mif6c_mifi
;
917 struct mif_device
*v
= &mrt
->vif6_table
[vifi
];
918 struct net_device
*dev
;
919 struct inet6_dev
*in6_dev
;
923 if (MIF_EXISTS(mrt
, vifi
))
926 switch (vifc
->mif6c_flags
) {
927 #ifdef CONFIG_IPV6_PIMSM_V2
930 * Special Purpose VIF in PIM
931 * All the packets will be sent to the daemon
933 if (mrt
->mroute_reg_vif_num
>= 0)
935 dev
= ip6mr_reg_vif(net
, mrt
);
938 err
= dev_set_allmulti(dev
, 1);
940 unregister_netdevice(dev
);
947 dev
= dev_get_by_index(net
, vifc
->mif6c_pifi
);
949 return -EADDRNOTAVAIL
;
950 err
= dev_set_allmulti(dev
, 1);
960 in6_dev
= __in6_dev_get(dev
);
962 in6_dev
->cnf
.mc_forwarding
++;
965 * Fill in the VIF structures
967 v
->rate_limit
= vifc
->vifc_rate_limit
;
968 v
->flags
= vifc
->mif6c_flags
;
970 v
->flags
|= VIFF_STATIC
;
971 v
->threshold
= vifc
->vifc_threshold
;
976 v
->link
= dev
->ifindex
;
977 if (v
->flags
& MIFF_REGISTER
)
978 v
->link
= dev
->iflink
;
980 /* And finish update writing critical data */
981 write_lock_bh(&mrt_lock
);
983 #ifdef CONFIG_IPV6_PIMSM_V2
984 if (v
->flags
& MIFF_REGISTER
)
985 mrt
->mroute_reg_vif_num
= vifi
;
987 if (vifi
+ 1 > mrt
->maxvif
)
988 mrt
->maxvif
= vifi
+ 1;
989 write_unlock_bh(&mrt_lock
);
993 static struct mfc6_cache
*ip6mr_cache_find(struct mr6_table
*mrt
,
994 const struct in6_addr
*origin
,
995 const struct in6_addr
*mcastgrp
)
997 int line
= MFC6_HASH(mcastgrp
, origin
);
998 struct mfc6_cache
*c
;
1000 list_for_each_entry(c
, &mrt
->mfc6_cache_array
[line
], list
) {
1001 if (ipv6_addr_equal(&c
->mf6c_origin
, origin
) &&
1002 ipv6_addr_equal(&c
->mf6c_mcastgrp
, mcastgrp
))
1009 * Allocate a multicast cache entry
1011 static struct mfc6_cache
*ip6mr_cache_alloc(void)
1013 struct mfc6_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_KERNEL
);
1016 c
->mfc_un
.res
.minvif
= MAXMIFS
;
1020 static struct mfc6_cache
*ip6mr_cache_alloc_unres(void)
1022 struct mfc6_cache
*c
= kmem_cache_zalloc(mrt_cachep
, GFP_ATOMIC
);
1025 skb_queue_head_init(&c
->mfc_un
.unres
.unresolved
);
1026 c
->mfc_un
.unres
.expires
= jiffies
+ 10 * HZ
;
1031 * A cache entry has gone into a resolved state from queued
1034 static void ip6mr_cache_resolve(struct net
*net
, struct mr6_table
*mrt
,
1035 struct mfc6_cache
*uc
, struct mfc6_cache
*c
)
1037 struct sk_buff
*skb
;
1040 * Play the pending entries through our router
1043 while((skb
= __skb_dequeue(&uc
->mfc_un
.unres
.unresolved
))) {
1044 if (ipv6_hdr(skb
)->version
== 0) {
1045 struct nlmsghdr
*nlh
= (struct nlmsghdr
*)skb_pull(skb
, sizeof(struct ipv6hdr
));
1047 if (__ip6mr_fill_mroute(mrt
, skb
, c
, NLMSG_DATA(nlh
)) > 0) {
1048 nlh
->nlmsg_len
= skb_tail_pointer(skb
) - (u8
*)nlh
;
1050 nlh
->nlmsg_type
= NLMSG_ERROR
;
1051 nlh
->nlmsg_len
= NLMSG_LENGTH(sizeof(struct nlmsgerr
));
1052 skb_trim(skb
, nlh
->nlmsg_len
);
1053 ((struct nlmsgerr
*)NLMSG_DATA(nlh
))->error
= -EMSGSIZE
;
1055 rtnl_unicast(skb
, net
, NETLINK_CB(skb
).portid
);
1057 ip6_mr_forward(net
, mrt
, skb
, c
);
1062 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1063 * expects the following bizarre scheme.
1065 * Called under mrt_lock.
1068 static int ip6mr_cache_report(struct mr6_table
*mrt
, struct sk_buff
*pkt
,
1069 mifi_t mifi
, int assert)
1071 struct sk_buff
*skb
;
1072 struct mrt6msg
*msg
;
1075 #ifdef CONFIG_IPV6_PIMSM_V2
1076 if (assert == MRT6MSG_WHOLEPKT
)
1077 skb
= skb_realloc_headroom(pkt
, -skb_network_offset(pkt
)
1081 skb
= alloc_skb(sizeof(struct ipv6hdr
) + sizeof(*msg
), GFP_ATOMIC
);
1086 /* I suppose that internal messages
1087 * do not require checksums */
1089 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1091 #ifdef CONFIG_IPV6_PIMSM_V2
1092 if (assert == MRT6MSG_WHOLEPKT
) {
1093 /* Ugly, but we have no choice with this interface.
1094 Duplicate old header, fix length etc.
1095 And all this only to mangle msg->im6_msgtype and
1096 to set msg->im6_mbz to "mbz" :-)
1098 skb_push(skb
, -skb_network_offset(pkt
));
1100 skb_push(skb
, sizeof(*msg
));
1101 skb_reset_transport_header(skb
);
1102 msg
= (struct mrt6msg
*)skb_transport_header(skb
);
1104 msg
->im6_msgtype
= MRT6MSG_WHOLEPKT
;
1105 msg
->im6_mif
= mrt
->mroute_reg_vif_num
;
1107 msg
->im6_src
= ipv6_hdr(pkt
)->saddr
;
1108 msg
->im6_dst
= ipv6_hdr(pkt
)->daddr
;
1110 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1115 * Copy the IP header
1118 skb_put(skb
, sizeof(struct ipv6hdr
));
1119 skb_reset_network_header(skb
);
1120 skb_copy_to_linear_data(skb
, ipv6_hdr(pkt
), sizeof(struct ipv6hdr
));
1125 skb_put(skb
, sizeof(*msg
));
1126 skb_reset_transport_header(skb
);
1127 msg
= (struct mrt6msg
*)skb_transport_header(skb
);
1130 msg
->im6_msgtype
= assert;
1131 msg
->im6_mif
= mifi
;
1133 msg
->im6_src
= ipv6_hdr(pkt
)->saddr
;
1134 msg
->im6_dst
= ipv6_hdr(pkt
)->daddr
;
1136 skb_dst_set(skb
, dst_clone(skb_dst(pkt
)));
1137 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1140 if (mrt
->mroute6_sk
== NULL
) {
1146 * Deliver to user space multicast routing algorithms
1148 ret
= sock_queue_rcv_skb(mrt
->mroute6_sk
, skb
);
1150 net_warn_ratelimited("mroute6: pending queue full, dropping entries\n");
1158 * Queue a packet for resolution. It gets locked cache entry!
1162 ip6mr_cache_unresolved(struct mr6_table
*mrt
, mifi_t mifi
, struct sk_buff
*skb
)
1166 struct mfc6_cache
*c
;
1168 spin_lock_bh(&mfc_unres_lock
);
1169 list_for_each_entry(c
, &mrt
->mfc6_unres_queue
, list
) {
1170 if (ipv6_addr_equal(&c
->mf6c_mcastgrp
, &ipv6_hdr(skb
)->daddr
) &&
1171 ipv6_addr_equal(&c
->mf6c_origin
, &ipv6_hdr(skb
)->saddr
)) {
1179 * Create a new entry if allowable
1182 if (atomic_read(&mrt
->cache_resolve_queue_len
) >= 10 ||
1183 (c
= ip6mr_cache_alloc_unres()) == NULL
) {
1184 spin_unlock_bh(&mfc_unres_lock
);
1191 * Fill in the new cache entry
1193 c
->mf6c_parent
= -1;
1194 c
->mf6c_origin
= ipv6_hdr(skb
)->saddr
;
1195 c
->mf6c_mcastgrp
= ipv6_hdr(skb
)->daddr
;
1198 * Reflect first query at pim6sd
1200 err
= ip6mr_cache_report(mrt
, skb
, mifi
, MRT6MSG_NOCACHE
);
1202 /* If the report failed throw the cache entry
1205 spin_unlock_bh(&mfc_unres_lock
);
1207 ip6mr_cache_free(c
);
1212 atomic_inc(&mrt
->cache_resolve_queue_len
);
1213 list_add(&c
->list
, &mrt
->mfc6_unres_queue
);
1215 ipmr_do_expire_process(mrt
);
1219 * See if we can append the packet
1221 if (c
->mfc_un
.unres
.unresolved
.qlen
> 3) {
1225 skb_queue_tail(&c
->mfc_un
.unres
.unresolved
, skb
);
1229 spin_unlock_bh(&mfc_unres_lock
);
1234 * MFC6 cache manipulation by user space
1237 static int ip6mr_mfc_delete(struct mr6_table
*mrt
, struct mf6cctl
*mfc
)
1240 struct mfc6_cache
*c
, *next
;
1242 line
= MFC6_HASH(&mfc
->mf6cc_mcastgrp
.sin6_addr
, &mfc
->mf6cc_origin
.sin6_addr
);
1244 list_for_each_entry_safe(c
, next
, &mrt
->mfc6_cache_array
[line
], list
) {
1245 if (ipv6_addr_equal(&c
->mf6c_origin
, &mfc
->mf6cc_origin
.sin6_addr
) &&
1246 ipv6_addr_equal(&c
->mf6c_mcastgrp
, &mfc
->mf6cc_mcastgrp
.sin6_addr
)) {
1247 write_lock_bh(&mrt_lock
);
1249 write_unlock_bh(&mrt_lock
);
1251 ip6mr_cache_free(c
);
1258 static int ip6mr_device_event(struct notifier_block
*this,
1259 unsigned long event
, void *ptr
)
1261 struct net_device
*dev
= ptr
;
1262 struct net
*net
= dev_net(dev
);
1263 struct mr6_table
*mrt
;
1264 struct mif_device
*v
;
1268 if (event
!= NETDEV_UNREGISTER
)
1271 ip6mr_for_each_table(mrt
, net
) {
1272 v
= &mrt
->vif6_table
[0];
1273 for (ct
= 0; ct
< mrt
->maxvif
; ct
++, v
++) {
1275 mif6_delete(mrt
, ct
, &list
);
1278 unregister_netdevice_many(&list
);
1283 static struct notifier_block ip6_mr_notifier
= {
1284 .notifier_call
= ip6mr_device_event
1288 * Setup for IP multicast routing
1291 static int __net_init
ip6mr_net_init(struct net
*net
)
1295 err
= ip6mr_rules_init(net
);
1299 #ifdef CONFIG_PROC_FS
1301 if (!proc_net_fops_create(net
, "ip6_mr_vif", 0, &ip6mr_vif_fops
))
1303 if (!proc_net_fops_create(net
, "ip6_mr_cache", 0, &ip6mr_mfc_fops
))
1304 goto proc_cache_fail
;
1309 #ifdef CONFIG_PROC_FS
1311 proc_net_remove(net
, "ip6_mr_vif");
1313 ip6mr_rules_exit(net
);
1319 static void __net_exit
ip6mr_net_exit(struct net
*net
)
1321 #ifdef CONFIG_PROC_FS
1322 proc_net_remove(net
, "ip6_mr_cache");
1323 proc_net_remove(net
, "ip6_mr_vif");
1325 ip6mr_rules_exit(net
);
1328 static struct pernet_operations ip6mr_net_ops
= {
1329 .init
= ip6mr_net_init
,
1330 .exit
= ip6mr_net_exit
,
1333 int __init
ip6_mr_init(void)
1337 mrt_cachep
= kmem_cache_create("ip6_mrt_cache",
1338 sizeof(struct mfc6_cache
),
1339 0, SLAB_HWCACHE_ALIGN
,
1344 err
= register_pernet_subsys(&ip6mr_net_ops
);
1346 goto reg_pernet_fail
;
1348 err
= register_netdevice_notifier(&ip6_mr_notifier
);
1350 goto reg_notif_fail
;
1351 #ifdef CONFIG_IPV6_PIMSM_V2
1352 if (inet6_add_protocol(&pim6_protocol
, IPPROTO_PIM
) < 0) {
1353 pr_err("%s: can't add PIM protocol\n", __func__
);
1355 goto add_proto_fail
;
1358 rtnl_register(RTNL_FAMILY_IP6MR
, RTM_GETROUTE
, NULL
,
1359 ip6mr_rtm_dumproute
, NULL
);
1361 #ifdef CONFIG_IPV6_PIMSM_V2
1363 unregister_netdevice_notifier(&ip6_mr_notifier
);
1366 unregister_pernet_subsys(&ip6mr_net_ops
);
1368 kmem_cache_destroy(mrt_cachep
);
1372 void ip6_mr_cleanup(void)
1374 unregister_netdevice_notifier(&ip6_mr_notifier
);
1375 unregister_pernet_subsys(&ip6mr_net_ops
);
1376 kmem_cache_destroy(mrt_cachep
);
1379 static int ip6mr_mfc_add(struct net
*net
, struct mr6_table
*mrt
,
1380 struct mf6cctl
*mfc
, int mrtsock
)
1384 struct mfc6_cache
*uc
, *c
;
1385 unsigned char ttls
[MAXMIFS
];
1388 if (mfc
->mf6cc_parent
>= MAXMIFS
)
1391 memset(ttls
, 255, MAXMIFS
);
1392 for (i
= 0; i
< MAXMIFS
; i
++) {
1393 if (IF_ISSET(i
, &mfc
->mf6cc_ifset
))
1398 line
= MFC6_HASH(&mfc
->mf6cc_mcastgrp
.sin6_addr
, &mfc
->mf6cc_origin
.sin6_addr
);
1400 list_for_each_entry(c
, &mrt
->mfc6_cache_array
[line
], list
) {
1401 if (ipv6_addr_equal(&c
->mf6c_origin
, &mfc
->mf6cc_origin
.sin6_addr
) &&
1402 ipv6_addr_equal(&c
->mf6c_mcastgrp
, &mfc
->mf6cc_mcastgrp
.sin6_addr
)) {
1409 write_lock_bh(&mrt_lock
);
1410 c
->mf6c_parent
= mfc
->mf6cc_parent
;
1411 ip6mr_update_thresholds(mrt
, c
, ttls
);
1413 c
->mfc_flags
|= MFC_STATIC
;
1414 write_unlock_bh(&mrt_lock
);
1418 if (!ipv6_addr_is_multicast(&mfc
->mf6cc_mcastgrp
.sin6_addr
))
1421 c
= ip6mr_cache_alloc();
1425 c
->mf6c_origin
= mfc
->mf6cc_origin
.sin6_addr
;
1426 c
->mf6c_mcastgrp
= mfc
->mf6cc_mcastgrp
.sin6_addr
;
1427 c
->mf6c_parent
= mfc
->mf6cc_parent
;
1428 ip6mr_update_thresholds(mrt
, c
, ttls
);
1430 c
->mfc_flags
|= MFC_STATIC
;
1432 write_lock_bh(&mrt_lock
);
1433 list_add(&c
->list
, &mrt
->mfc6_cache_array
[line
]);
1434 write_unlock_bh(&mrt_lock
);
1437 * Check to see if we resolved a queued list. If so we
1438 * need to send on the frames and tidy up.
1441 spin_lock_bh(&mfc_unres_lock
);
1442 list_for_each_entry(uc
, &mrt
->mfc6_unres_queue
, list
) {
1443 if (ipv6_addr_equal(&uc
->mf6c_origin
, &c
->mf6c_origin
) &&
1444 ipv6_addr_equal(&uc
->mf6c_mcastgrp
, &c
->mf6c_mcastgrp
)) {
1445 list_del(&uc
->list
);
1446 atomic_dec(&mrt
->cache_resolve_queue_len
);
1451 if (list_empty(&mrt
->mfc6_unres_queue
))
1452 del_timer(&mrt
->ipmr_expire_timer
);
1453 spin_unlock_bh(&mfc_unres_lock
);
1456 ip6mr_cache_resolve(net
, mrt
, uc
, c
);
1457 ip6mr_cache_free(uc
);
1463 * Close the multicast socket, and clear the vif tables etc
1466 static void mroute_clean_tables(struct mr6_table
*mrt
)
1470 struct mfc6_cache
*c
, *next
;
1473 * Shut down all active vif entries
1475 for (i
= 0; i
< mrt
->maxvif
; i
++) {
1476 if (!(mrt
->vif6_table
[i
].flags
& VIFF_STATIC
))
1477 mif6_delete(mrt
, i
, &list
);
1479 unregister_netdevice_many(&list
);
1484 for (i
= 0; i
< MFC6_LINES
; i
++) {
1485 list_for_each_entry_safe(c
, next
, &mrt
->mfc6_cache_array
[i
], list
) {
1486 if (c
->mfc_flags
& MFC_STATIC
)
1488 write_lock_bh(&mrt_lock
);
1490 write_unlock_bh(&mrt_lock
);
1492 ip6mr_cache_free(c
);
1496 if (atomic_read(&mrt
->cache_resolve_queue_len
) != 0) {
1497 spin_lock_bh(&mfc_unres_lock
);
1498 list_for_each_entry_safe(c
, next
, &mrt
->mfc6_unres_queue
, list
) {
1500 ip6mr_destroy_unres(mrt
, c
);
1502 spin_unlock_bh(&mfc_unres_lock
);
1506 static int ip6mr_sk_init(struct mr6_table
*mrt
, struct sock
*sk
)
1509 struct net
*net
= sock_net(sk
);
1512 write_lock_bh(&mrt_lock
);
1513 if (likely(mrt
->mroute6_sk
== NULL
)) {
1514 mrt
->mroute6_sk
= sk
;
1515 net
->ipv6
.devconf_all
->mc_forwarding
++;
1519 write_unlock_bh(&mrt_lock
);
1526 int ip6mr_sk_done(struct sock
*sk
)
1529 struct net
*net
= sock_net(sk
);
1530 struct mr6_table
*mrt
;
1533 ip6mr_for_each_table(mrt
, net
) {
1534 if (sk
== mrt
->mroute6_sk
) {
1535 write_lock_bh(&mrt_lock
);
1536 mrt
->mroute6_sk
= NULL
;
1537 net
->ipv6
.devconf_all
->mc_forwarding
--;
1538 write_unlock_bh(&mrt_lock
);
1540 mroute_clean_tables(mrt
);
1550 struct sock
*mroute6_socket(struct net
*net
, struct sk_buff
*skb
)
1552 struct mr6_table
*mrt
;
1553 struct flowi6 fl6
= {
1554 .flowi6_iif
= skb
->skb_iif
,
1555 .flowi6_oif
= skb
->dev
->ifindex
,
1556 .flowi6_mark
= skb
->mark
,
1559 if (ip6mr_fib_lookup(net
, &fl6
, &mrt
) < 0)
1562 return mrt
->mroute6_sk
;
1566 * Socket options and virtual interface manipulation. The whole
1567 * virtual interface system is a complete heap, but unfortunately
1568 * that's how BSD mrouted happens to think. Maybe one day with a proper
1569 * MOSPF/PIM router set up we can clean this up.
1572 int ip6_mroute_setsockopt(struct sock
*sk
, int optname
, char __user
*optval
, unsigned int optlen
)
1578 struct net
*net
= sock_net(sk
);
1579 struct mr6_table
*mrt
;
1581 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1585 if (optname
!= MRT6_INIT
) {
1586 if (sk
!= mrt
->mroute6_sk
&& !ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1592 if (sk
->sk_type
!= SOCK_RAW
||
1593 inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1595 if (optlen
< sizeof(int))
1598 return ip6mr_sk_init(mrt
, sk
);
1601 return ip6mr_sk_done(sk
);
1604 if (optlen
< sizeof(vif
))
1606 if (copy_from_user(&vif
, optval
, sizeof(vif
)))
1608 if (vif
.mif6c_mifi
>= MAXMIFS
)
1611 ret
= mif6_add(net
, mrt
, &vif
, sk
== mrt
->mroute6_sk
);
1616 if (optlen
< sizeof(mifi_t
))
1618 if (copy_from_user(&mifi
, optval
, sizeof(mifi_t
)))
1621 ret
= mif6_delete(mrt
, mifi
, NULL
);
1626 * Manipulate the forwarding caches. These live
1627 * in a sort of kernel/user symbiosis.
1631 if (optlen
< sizeof(mfc
))
1633 if (copy_from_user(&mfc
, optval
, sizeof(mfc
)))
1636 if (optname
== MRT6_DEL_MFC
)
1637 ret
= ip6mr_mfc_delete(mrt
, &mfc
);
1639 ret
= ip6mr_mfc_add(net
, mrt
, &mfc
, sk
== mrt
->mroute6_sk
);
1644 * Control PIM assert (to activate pim will activate assert)
1650 if (optlen
!= sizeof(v
))
1652 if (get_user(v
, (int __user
*)optval
))
1654 mrt
->mroute_do_assert
= v
;
1658 #ifdef CONFIG_IPV6_PIMSM_V2
1663 if (optlen
!= sizeof(v
))
1665 if (get_user(v
, (int __user
*)optval
))
1670 if (v
!= mrt
->mroute_do_pim
) {
1671 mrt
->mroute_do_pim
= v
;
1672 mrt
->mroute_do_assert
= v
;
1679 #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1684 if (optlen
!= sizeof(u32
))
1686 if (get_user(v
, (u32 __user
*)optval
))
1688 if (sk
== mrt
->mroute6_sk
)
1693 if (!ip6mr_new_table(net
, v
))
1695 raw6_sk(sk
)->ip6mr_table
= v
;
1701 * Spurious command, or MRT6_VERSION which you cannot
1705 return -ENOPROTOOPT
;
1710 * Getsock opt support for the multicast routing system.
1713 int ip6_mroute_getsockopt(struct sock
*sk
, int optname
, char __user
*optval
,
1718 struct net
*net
= sock_net(sk
);
1719 struct mr6_table
*mrt
;
1721 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1729 #ifdef CONFIG_IPV6_PIMSM_V2
1731 val
= mrt
->mroute_do_pim
;
1735 val
= mrt
->mroute_do_assert
;
1738 return -ENOPROTOOPT
;
1741 if (get_user(olr
, optlen
))
1744 olr
= min_t(int, olr
, sizeof(int));
1748 if (put_user(olr
, optlen
))
1750 if (copy_to_user(optval
, &val
, olr
))
1756 * The IP multicast ioctl support routines.
1759 int ip6mr_ioctl(struct sock
*sk
, int cmd
, void __user
*arg
)
1761 struct sioc_sg_req6 sr
;
1762 struct sioc_mif_req6 vr
;
1763 struct mif_device
*vif
;
1764 struct mfc6_cache
*c
;
1765 struct net
*net
= sock_net(sk
);
1766 struct mr6_table
*mrt
;
1768 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1773 case SIOCGETMIFCNT_IN6
:
1774 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1776 if (vr
.mifi
>= mrt
->maxvif
)
1778 read_lock(&mrt_lock
);
1779 vif
= &mrt
->vif6_table
[vr
.mifi
];
1780 if (MIF_EXISTS(mrt
, vr
.mifi
)) {
1781 vr
.icount
= vif
->pkt_in
;
1782 vr
.ocount
= vif
->pkt_out
;
1783 vr
.ibytes
= vif
->bytes_in
;
1784 vr
.obytes
= vif
->bytes_out
;
1785 read_unlock(&mrt_lock
);
1787 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1791 read_unlock(&mrt_lock
);
1792 return -EADDRNOTAVAIL
;
1793 case SIOCGETSGCNT_IN6
:
1794 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1797 read_lock(&mrt_lock
);
1798 c
= ip6mr_cache_find(mrt
, &sr
.src
.sin6_addr
, &sr
.grp
.sin6_addr
);
1800 sr
.pktcnt
= c
->mfc_un
.res
.pkt
;
1801 sr
.bytecnt
= c
->mfc_un
.res
.bytes
;
1802 sr
.wrong_if
= c
->mfc_un
.res
.wrong_if
;
1803 read_unlock(&mrt_lock
);
1805 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1809 read_unlock(&mrt_lock
);
1810 return -EADDRNOTAVAIL
;
1812 return -ENOIOCTLCMD
;
1816 #ifdef CONFIG_COMPAT
1817 struct compat_sioc_sg_req6
{
1818 struct sockaddr_in6 src
;
1819 struct sockaddr_in6 grp
;
1820 compat_ulong_t pktcnt
;
1821 compat_ulong_t bytecnt
;
1822 compat_ulong_t wrong_if
;
1825 struct compat_sioc_mif_req6
{
1827 compat_ulong_t icount
;
1828 compat_ulong_t ocount
;
1829 compat_ulong_t ibytes
;
1830 compat_ulong_t obytes
;
1833 int ip6mr_compat_ioctl(struct sock
*sk
, unsigned int cmd
, void __user
*arg
)
1835 struct compat_sioc_sg_req6 sr
;
1836 struct compat_sioc_mif_req6 vr
;
1837 struct mif_device
*vif
;
1838 struct mfc6_cache
*c
;
1839 struct net
*net
= sock_net(sk
);
1840 struct mr6_table
*mrt
;
1842 mrt
= ip6mr_get_table(net
, raw6_sk(sk
)->ip6mr_table
? : RT6_TABLE_DFLT
);
1847 case SIOCGETMIFCNT_IN6
:
1848 if (copy_from_user(&vr
, arg
, sizeof(vr
)))
1850 if (vr
.mifi
>= mrt
->maxvif
)
1852 read_lock(&mrt_lock
);
1853 vif
= &mrt
->vif6_table
[vr
.mifi
];
1854 if (MIF_EXISTS(mrt
, vr
.mifi
)) {
1855 vr
.icount
= vif
->pkt_in
;
1856 vr
.ocount
= vif
->pkt_out
;
1857 vr
.ibytes
= vif
->bytes_in
;
1858 vr
.obytes
= vif
->bytes_out
;
1859 read_unlock(&mrt_lock
);
1861 if (copy_to_user(arg
, &vr
, sizeof(vr
)))
1865 read_unlock(&mrt_lock
);
1866 return -EADDRNOTAVAIL
;
1867 case SIOCGETSGCNT_IN6
:
1868 if (copy_from_user(&sr
, arg
, sizeof(sr
)))
1871 read_lock(&mrt_lock
);
1872 c
= ip6mr_cache_find(mrt
, &sr
.src
.sin6_addr
, &sr
.grp
.sin6_addr
);
1874 sr
.pktcnt
= c
->mfc_un
.res
.pkt
;
1875 sr
.bytecnt
= c
->mfc_un
.res
.bytes
;
1876 sr
.wrong_if
= c
->mfc_un
.res
.wrong_if
;
1877 read_unlock(&mrt_lock
);
1879 if (copy_to_user(arg
, &sr
, sizeof(sr
)))
1883 read_unlock(&mrt_lock
);
1884 return -EADDRNOTAVAIL
;
1886 return -ENOIOCTLCMD
;
1891 static inline int ip6mr_forward2_finish(struct sk_buff
*skb
)
1893 IP6_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), ip6_dst_idev(skb_dst(skb
)),
1894 IPSTATS_MIB_OUTFORWDATAGRAMS
);
1895 IP6_ADD_STATS_BH(dev_net(skb_dst(skb
)->dev
), ip6_dst_idev(skb_dst(skb
)),
1896 IPSTATS_MIB_OUTOCTETS
, skb
->len
);
1897 return dst_output(skb
);
1901 * Processing handlers for ip6mr_forward
1904 static int ip6mr_forward2(struct net
*net
, struct mr6_table
*mrt
,
1905 struct sk_buff
*skb
, struct mfc6_cache
*c
, int vifi
)
1907 struct ipv6hdr
*ipv6h
;
1908 struct mif_device
*vif
= &mrt
->vif6_table
[vifi
];
1909 struct net_device
*dev
;
1910 struct dst_entry
*dst
;
1913 if (vif
->dev
== NULL
)
1916 #ifdef CONFIG_IPV6_PIMSM_V2
1917 if (vif
->flags
& MIFF_REGISTER
) {
1919 vif
->bytes_out
+= skb
->len
;
1920 vif
->dev
->stats
.tx_bytes
+= skb
->len
;
1921 vif
->dev
->stats
.tx_packets
++;
1922 ip6mr_cache_report(mrt
, skb
, vifi
, MRT6MSG_WHOLEPKT
);
1927 ipv6h
= ipv6_hdr(skb
);
1929 fl6
= (struct flowi6
) {
1930 .flowi6_oif
= vif
->link
,
1931 .daddr
= ipv6h
->daddr
,
1934 dst
= ip6_route_output(net
, NULL
, &fl6
);
1941 skb_dst_set(skb
, dst
);
1944 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1945 * not only before forwarding, but after forwarding on all output
1946 * interfaces. It is clear, if mrouter runs a multicasting
1947 * program, it should receive packets not depending to what interface
1948 * program is joined.
1949 * If we will not make it, the program will have to join on all
1950 * interfaces. On the other hand, multihoming host (or router, but
1951 * not mrouter) cannot join to more than one interface - it will
1952 * result in receiving multiple packets.
1957 vif
->bytes_out
+= skb
->len
;
1959 /* We are about to write */
1960 /* XXX: extension headers? */
1961 if (skb_cow(skb
, sizeof(*ipv6h
) + LL_RESERVED_SPACE(dev
)))
1964 ipv6h
= ipv6_hdr(skb
);
1967 IP6CB(skb
)->flags
|= IP6SKB_FORWARDED
;
1969 return NF_HOOK(NFPROTO_IPV6
, NF_INET_FORWARD
, skb
, skb
->dev
, dev
,
1970 ip6mr_forward2_finish
);
1977 static int ip6mr_find_vif(struct mr6_table
*mrt
, struct net_device
*dev
)
1981 for (ct
= mrt
->maxvif
- 1; ct
>= 0; ct
--) {
1982 if (mrt
->vif6_table
[ct
].dev
== dev
)
1988 static int ip6_mr_forward(struct net
*net
, struct mr6_table
*mrt
,
1989 struct sk_buff
*skb
, struct mfc6_cache
*cache
)
1994 vif
= cache
->mf6c_parent
;
1995 cache
->mfc_un
.res
.pkt
++;
1996 cache
->mfc_un
.res
.bytes
+= skb
->len
;
1999 * Wrong interface: drop packet and (maybe) send PIM assert.
2001 if (mrt
->vif6_table
[vif
].dev
!= skb
->dev
) {
2004 cache
->mfc_un
.res
.wrong_if
++;
2005 true_vifi
= ip6mr_find_vif(mrt
, skb
->dev
);
2007 if (true_vifi
>= 0 && mrt
->mroute_do_assert
&&
2008 /* pimsm uses asserts, when switching from RPT to SPT,
2009 so that we cannot check that packet arrived on an oif.
2010 It is bad, but otherwise we would need to move pretty
2011 large chunk of pimd to kernel. Ough... --ANK
2013 (mrt
->mroute_do_pim
||
2014 cache
->mfc_un
.res
.ttls
[true_vifi
] < 255) &&
2016 cache
->mfc_un
.res
.last_assert
+ MFC_ASSERT_THRESH
)) {
2017 cache
->mfc_un
.res
.last_assert
= jiffies
;
2018 ip6mr_cache_report(mrt
, skb
, true_vifi
, MRT6MSG_WRONGMIF
);
2023 mrt
->vif6_table
[vif
].pkt_in
++;
2024 mrt
->vif6_table
[vif
].bytes_in
+= skb
->len
;
2029 for (ct
= cache
->mfc_un
.res
.maxvif
- 1; ct
>= cache
->mfc_un
.res
.minvif
; ct
--) {
2030 if (ipv6_hdr(skb
)->hop_limit
> cache
->mfc_un
.res
.ttls
[ct
]) {
2032 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
2034 ip6mr_forward2(net
, mrt
, skb2
, cache
, psend
);
2040 ip6mr_forward2(net
, mrt
, skb
, cache
, psend
);
2051 * Multicast packets for forwarding arrive here
2054 int ip6_mr_input(struct sk_buff
*skb
)
2056 struct mfc6_cache
*cache
;
2057 struct net
*net
= dev_net(skb
->dev
);
2058 struct mr6_table
*mrt
;
2059 struct flowi6 fl6
= {
2060 .flowi6_iif
= skb
->dev
->ifindex
,
2061 .flowi6_mark
= skb
->mark
,
2065 err
= ip6mr_fib_lookup(net
, &fl6
, &mrt
);
2071 read_lock(&mrt_lock
);
2072 cache
= ip6mr_cache_find(mrt
,
2073 &ipv6_hdr(skb
)->saddr
, &ipv6_hdr(skb
)->daddr
);
2076 * No usable cache entry
2078 if (cache
== NULL
) {
2081 vif
= ip6mr_find_vif(mrt
, skb
->dev
);
2083 int err
= ip6mr_cache_unresolved(mrt
, vif
, skb
);
2084 read_unlock(&mrt_lock
);
2088 read_unlock(&mrt_lock
);
2093 ip6_mr_forward(net
, mrt
, skb
, cache
);
2095 read_unlock(&mrt_lock
);
2101 static int __ip6mr_fill_mroute(struct mr6_table
*mrt
, struct sk_buff
*skb
,
2102 struct mfc6_cache
*c
, struct rtmsg
*rtm
)
2105 struct rtnexthop
*nhp
;
2106 u8
*b
= skb_tail_pointer(skb
);
2107 struct rtattr
*mp_head
;
2109 /* If cache is unresolved, don't try to parse IIF and OIF */
2110 if (c
->mf6c_parent
>= MAXMIFS
)
2113 if (MIF_EXISTS(mrt
, c
->mf6c_parent
) &&
2114 nla_put_u32(skb
, RTA_IIF
, mrt
->vif6_table
[c
->mf6c_parent
].dev
->ifindex
) < 0)
2117 mp_head
= (struct rtattr
*)skb_put(skb
, RTA_LENGTH(0));
2119 for (ct
= c
->mfc_un
.res
.minvif
; ct
< c
->mfc_un
.res
.maxvif
; ct
++) {
2120 if (MIF_EXISTS(mrt
, ct
) && c
->mfc_un
.res
.ttls
[ct
] < 255) {
2121 if (skb_tailroom(skb
) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp
)) + 4))
2122 goto rtattr_failure
;
2123 nhp
= (struct rtnexthop
*)skb_put(skb
, RTA_ALIGN(sizeof(*nhp
)));
2124 nhp
->rtnh_flags
= 0;
2125 nhp
->rtnh_hops
= c
->mfc_un
.res
.ttls
[ct
];
2126 nhp
->rtnh_ifindex
= mrt
->vif6_table
[ct
].dev
->ifindex
;
2127 nhp
->rtnh_len
= sizeof(*nhp
);
2130 mp_head
->rta_type
= RTA_MULTIPATH
;
2131 mp_head
->rta_len
= skb_tail_pointer(skb
) - (u8
*)mp_head
;
2132 rtm
->rtm_type
= RTN_MULTICAST
;
2140 int ip6mr_get_route(struct net
*net
,
2141 struct sk_buff
*skb
, struct rtmsg
*rtm
, int nowait
)
2144 struct mr6_table
*mrt
;
2145 struct mfc6_cache
*cache
;
2146 struct rt6_info
*rt
= (struct rt6_info
*)skb_dst(skb
);
2148 mrt
= ip6mr_get_table(net
, RT6_TABLE_DFLT
);
2152 read_lock(&mrt_lock
);
2153 cache
= ip6mr_cache_find(mrt
, &rt
->rt6i_src
.addr
, &rt
->rt6i_dst
.addr
);
2156 struct sk_buff
*skb2
;
2157 struct ipv6hdr
*iph
;
2158 struct net_device
*dev
;
2162 read_unlock(&mrt_lock
);
2167 if (dev
== NULL
|| (vif
= ip6mr_find_vif(mrt
, dev
)) < 0) {
2168 read_unlock(&mrt_lock
);
2172 /* really correct? */
2173 skb2
= alloc_skb(sizeof(struct ipv6hdr
), GFP_ATOMIC
);
2175 read_unlock(&mrt_lock
);
2179 skb_reset_transport_header(skb2
);
2181 skb_put(skb2
, sizeof(struct ipv6hdr
));
2182 skb_reset_network_header(skb2
);
2184 iph
= ipv6_hdr(skb2
);
2187 iph
->flow_lbl
[0] = 0;
2188 iph
->flow_lbl
[1] = 0;
2189 iph
->flow_lbl
[2] = 0;
2190 iph
->payload_len
= 0;
2191 iph
->nexthdr
= IPPROTO_NONE
;
2193 iph
->saddr
= rt
->rt6i_src
.addr
;
2194 iph
->daddr
= rt
->rt6i_dst
.addr
;
2196 err
= ip6mr_cache_unresolved(mrt
, vif
, skb2
);
2197 read_unlock(&mrt_lock
);
2202 if (!nowait
&& (rtm
->rtm_flags
&RTM_F_NOTIFY
))
2203 cache
->mfc_flags
|= MFC_NOTIFY
;
2205 err
= __ip6mr_fill_mroute(mrt
, skb
, cache
, rtm
);
2206 read_unlock(&mrt_lock
);
2210 static int ip6mr_fill_mroute(struct mr6_table
*mrt
, struct sk_buff
*skb
,
2211 u32 portid
, u32 seq
, struct mfc6_cache
*c
)
2213 struct nlmsghdr
*nlh
;
2216 nlh
= nlmsg_put(skb
, portid
, seq
, RTM_NEWROUTE
, sizeof(*rtm
), NLM_F_MULTI
);
2220 rtm
= nlmsg_data(nlh
);
2221 rtm
->rtm_family
= RTNL_FAMILY_IPMR
;
2222 rtm
->rtm_dst_len
= 128;
2223 rtm
->rtm_src_len
= 128;
2225 rtm
->rtm_table
= mrt
->id
;
2226 if (nla_put_u32(skb
, RTA_TABLE
, mrt
->id
))
2227 goto nla_put_failure
;
2228 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2229 rtm
->rtm_protocol
= RTPROT_UNSPEC
;
2232 if (nla_put(skb
, RTA_SRC
, 16, &c
->mf6c_origin
) ||
2233 nla_put(skb
, RTA_DST
, 16, &c
->mf6c_mcastgrp
))
2234 goto nla_put_failure
;
2235 if (__ip6mr_fill_mroute(mrt
, skb
, c
, rtm
) < 0)
2236 goto nla_put_failure
;
2238 return nlmsg_end(skb
, nlh
);
2241 nlmsg_cancel(skb
, nlh
);
2245 static int ip6mr_rtm_dumproute(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2247 struct net
*net
= sock_net(skb
->sk
);
2248 struct mr6_table
*mrt
;
2249 struct mfc6_cache
*mfc
;
2250 unsigned int t
= 0, s_t
;
2251 unsigned int h
= 0, s_h
;
2252 unsigned int e
= 0, s_e
;
2258 read_lock(&mrt_lock
);
2259 ip6mr_for_each_table(mrt
, net
) {
2264 for (h
= s_h
; h
< MFC6_LINES
; h
++) {
2265 list_for_each_entry(mfc
, &mrt
->mfc6_cache_array
[h
], list
) {
2268 if (ip6mr_fill_mroute(mrt
, skb
,
2269 NETLINK_CB(cb
->skb
).portid
,
2283 read_unlock(&mrt_lock
);