2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * IPv4 Forwarding Information Base: FIB frontend.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <linux/module.h>
17 #include <asm/uaccess.h>
18 #include <asm/system.h>
19 #include <linux/bitops.h>
20 #include <linux/capability.h>
21 #include <linux/types.h>
22 #include <linux/kernel.h>
24 #include <linux/string.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/errno.h>
29 #include <linux/inet.h>
30 #include <linux/inetdevice.h>
31 #include <linux/netdevice.h>
32 #include <linux/if_addr.h>
33 #include <linux/if_arp.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/list.h>
37 #include <linux/slab.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
45 #include <net/ip_fib.h>
46 #include <net/rtnetlink.h>
48 #ifndef CONFIG_IP_MULTIPLE_TABLES
50 static int __net_init
fib4_rules_init(struct net
*net
)
52 struct fib_table
*local_table
, *main_table
;
54 local_table
= fib_trie_table(RT_TABLE_LOCAL
);
55 if (local_table
== NULL
)
58 main_table
= fib_trie_table(RT_TABLE_MAIN
);
59 if (main_table
== NULL
)
62 hlist_add_head_rcu(&local_table
->tb_hlist
,
63 &net
->ipv4
.fib_table_hash
[TABLE_LOCAL_INDEX
]);
64 hlist_add_head_rcu(&main_table
->tb_hlist
,
65 &net
->ipv4
.fib_table_hash
[TABLE_MAIN_INDEX
]);
74 struct fib_table
*fib_new_table(struct net
*net
, u32 id
)
81 tb
= fib_get_table(net
, id
);
85 tb
= fib_trie_table(id
);
88 h
= id
& (FIB_TABLE_HASHSZ
- 1);
89 hlist_add_head_rcu(&tb
->tb_hlist
, &net
->ipv4
.fib_table_hash
[h
]);
93 struct fib_table
*fib_get_table(struct net
*net
, u32 id
)
96 struct hlist_node
*node
;
97 struct hlist_head
*head
;
102 h
= id
& (FIB_TABLE_HASHSZ
- 1);
105 head
= &net
->ipv4
.fib_table_hash
[h
];
106 hlist_for_each_entry_rcu(tb
, node
, head
, tb_hlist
) {
107 if (tb
->tb_id
== id
) {
115 #endif /* CONFIG_IP_MULTIPLE_TABLES */
117 static void fib_flush(struct net
*net
)
120 struct fib_table
*tb
;
121 struct hlist_node
*node
;
122 struct hlist_head
*head
;
125 for (h
= 0; h
< FIB_TABLE_HASHSZ
; h
++) {
126 head
= &net
->ipv4
.fib_table_hash
[h
];
127 hlist_for_each_entry(tb
, node
, head
, tb_hlist
)
128 flushed
+= fib_table_flush(tb
);
132 rt_cache_flush(net
, -1);
136 * Find address type as if only "dev" was present in the system. If
137 * on_dev is NULL then all interfaces are taken into consideration.
139 static inline unsigned __inet_dev_addr_type(struct net
*net
,
140 const struct net_device
*dev
,
143 struct flowi fl
= { .fl4_dst
= addr
};
144 struct fib_result res
;
145 unsigned ret
= RTN_BROADCAST
;
146 struct fib_table
*local_table
;
148 if (ipv4_is_zeronet(addr
) || ipv4_is_lbcast(addr
))
149 return RTN_BROADCAST
;
150 if (ipv4_is_multicast(addr
))
151 return RTN_MULTICAST
;
153 #ifdef CONFIG_IP_MULTIPLE_TABLES
157 local_table
= fib_get_table(net
, RT_TABLE_LOCAL
);
161 if (!fib_table_lookup(local_table
, &fl
, &res
, FIB_LOOKUP_NOREF
)) {
162 if (!dev
|| dev
== res
.fi
->fib_dev
)
170 unsigned int inet_addr_type(struct net
*net
, __be32 addr
)
172 return __inet_dev_addr_type(net
, NULL
, addr
);
174 EXPORT_SYMBOL(inet_addr_type
);
176 unsigned int inet_dev_addr_type(struct net
*net
, const struct net_device
*dev
,
179 return __inet_dev_addr_type(net
, dev
, addr
);
181 EXPORT_SYMBOL(inet_dev_addr_type
);
183 /* Given (packet source, input interface) and optional (dst, oif, tos):
184 * - (main) check, that source is valid i.e. not broadcast or our local
186 * - figure out what "logical" interface this packet arrived
187 * and calculate "specific destination" address.
188 * - check, that packet arrived from expected physical interface.
189 * called with rcu_read_lock()
191 int fib_validate_source(__be32 src
, __be32 dst
, u8 tos
, int oif
,
192 struct net_device
*dev
, __be32
*spec_dst
,
195 struct in_device
*in_dev
;
203 struct fib_result res
;
204 int no_addr
, rpf
, accept_local
;
209 no_addr
= rpf
= accept_local
= 0;
210 in_dev
= __in_dev_get_rcu(dev
);
212 no_addr
= in_dev
->ifa_list
== NULL
;
213 rpf
= IN_DEV_RPFILTER(in_dev
);
214 accept_local
= IN_DEV_ACCEPT_LOCAL(in_dev
);
215 if (mark
&& !IN_DEV_SRC_VMARK(in_dev
))
223 if (fib_lookup(net
, &fl
, &res
))
225 if (res
.type
!= RTN_UNICAST
) {
226 if (res
.type
!= RTN_LOCAL
|| !accept_local
)
229 *spec_dst
= FIB_RES_PREFSRC(res
);
230 fib_combine_itag(itag
, &res
);
233 #ifdef CONFIG_IP_ROUTE_MULTIPATH
234 for (ret
= 0; ret
< res
.fi
->fib_nhs
; ret
++) {
235 struct fib_nh
*nh
= &res
.fi
->fib_nh
[ret
];
237 if (nh
->nh_dev
== dev
) {
243 if (FIB_RES_DEV(res
) == dev
)
247 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
254 fl
.oif
= dev
->ifindex
;
257 if (fib_lookup(net
, &fl
, &res
) == 0) {
258 if (res
.type
== RTN_UNICAST
) {
259 *spec_dst
= FIB_RES_PREFSRC(res
);
260 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
268 *spec_dst
= inet_select_addr(dev
, 0, RT_SCOPE_UNIVERSE
);
278 static inline __be32
sk_extract_addr(struct sockaddr
*addr
)
280 return ((struct sockaddr_in
*) addr
)->sin_addr
.s_addr
;
283 static int put_rtax(struct nlattr
*mx
, int len
, int type
, u32 value
)
287 nla
= (struct nlattr
*) ((char *) mx
+ len
);
288 nla
->nla_type
= type
;
289 nla
->nla_len
= nla_attr_size(4);
290 *(u32
*) nla_data(nla
) = value
;
292 return len
+ nla_total_size(4);
295 static int rtentry_to_fib_config(struct net
*net
, int cmd
, struct rtentry
*rt
,
296 struct fib_config
*cfg
)
301 memset(cfg
, 0, sizeof(*cfg
));
302 cfg
->fc_nlinfo
.nl_net
= net
;
304 if (rt
->rt_dst
.sa_family
!= AF_INET
)
305 return -EAFNOSUPPORT
;
308 * Check mask for validity:
309 * a) it must be contiguous.
310 * b) destination must have all host bits clear.
311 * c) if application forgot to set correct family (AF_INET),
312 * reject request unless it is absolutely clear i.e.
313 * both family and mask are zero.
316 addr
= sk_extract_addr(&rt
->rt_dst
);
317 if (!(rt
->rt_flags
& RTF_HOST
)) {
318 __be32 mask
= sk_extract_addr(&rt
->rt_genmask
);
320 if (rt
->rt_genmask
.sa_family
!= AF_INET
) {
321 if (mask
|| rt
->rt_genmask
.sa_family
)
322 return -EAFNOSUPPORT
;
325 if (bad_mask(mask
, addr
))
328 plen
= inet_mask_len(mask
);
331 cfg
->fc_dst_len
= plen
;
334 if (cmd
!= SIOCDELRT
) {
335 cfg
->fc_nlflags
= NLM_F_CREATE
;
336 cfg
->fc_protocol
= RTPROT_BOOT
;
340 cfg
->fc_priority
= rt
->rt_metric
- 1;
342 if (rt
->rt_flags
& RTF_REJECT
) {
343 cfg
->fc_scope
= RT_SCOPE_HOST
;
344 cfg
->fc_type
= RTN_UNREACHABLE
;
348 cfg
->fc_scope
= RT_SCOPE_NOWHERE
;
349 cfg
->fc_type
= RTN_UNICAST
;
353 struct net_device
*dev
;
354 char devname
[IFNAMSIZ
];
356 if (copy_from_user(devname
, rt
->rt_dev
, IFNAMSIZ
-1))
359 devname
[IFNAMSIZ
-1] = 0;
360 colon
= strchr(devname
, ':');
363 dev
= __dev_get_by_name(net
, devname
);
366 cfg
->fc_oif
= dev
->ifindex
;
368 struct in_ifaddr
*ifa
;
369 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
373 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
)
374 if (strcmp(ifa
->ifa_label
, devname
) == 0)
378 cfg
->fc_prefsrc
= ifa
->ifa_local
;
382 addr
= sk_extract_addr(&rt
->rt_gateway
);
383 if (rt
->rt_gateway
.sa_family
== AF_INET
&& addr
) {
385 if (rt
->rt_flags
& RTF_GATEWAY
&&
386 inet_addr_type(net
, addr
) == RTN_UNICAST
)
387 cfg
->fc_scope
= RT_SCOPE_UNIVERSE
;
390 if (cmd
== SIOCDELRT
)
393 if (rt
->rt_flags
& RTF_GATEWAY
&& !cfg
->fc_gw
)
396 if (cfg
->fc_scope
== RT_SCOPE_NOWHERE
)
397 cfg
->fc_scope
= RT_SCOPE_LINK
;
399 if (rt
->rt_flags
& (RTF_MTU
| RTF_WINDOW
| RTF_IRTT
)) {
403 mx
= kzalloc(3 * nla_total_size(4), GFP_KERNEL
);
407 if (rt
->rt_flags
& RTF_MTU
)
408 len
= put_rtax(mx
, len
, RTAX_ADVMSS
, rt
->rt_mtu
- 40);
410 if (rt
->rt_flags
& RTF_WINDOW
)
411 len
= put_rtax(mx
, len
, RTAX_WINDOW
, rt
->rt_window
);
413 if (rt
->rt_flags
& RTF_IRTT
)
414 len
= put_rtax(mx
, len
, RTAX_RTT
, rt
->rt_irtt
<< 3);
417 cfg
->fc_mx_len
= len
;
424 * Handle IP routing ioctl calls.
425 * These are used to manipulate the routing tables
427 int ip_rt_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
429 struct fib_config cfg
;
434 case SIOCADDRT
: /* Add a route */
435 case SIOCDELRT
: /* Delete a route */
436 if (!capable(CAP_NET_ADMIN
))
439 if (copy_from_user(&rt
, arg
, sizeof(rt
)))
443 err
= rtentry_to_fib_config(net
, cmd
, &rt
, &cfg
);
445 struct fib_table
*tb
;
447 if (cmd
== SIOCDELRT
) {
448 tb
= fib_get_table(net
, cfg
.fc_table
);
450 err
= fib_table_delete(tb
, &cfg
);
454 tb
= fib_new_table(net
, cfg
.fc_table
);
456 err
= fib_table_insert(tb
, &cfg
);
461 /* allocated by rtentry_to_fib_config() */
470 const struct nla_policy rtm_ipv4_policy
[RTA_MAX
+ 1] = {
471 [RTA_DST
] = { .type
= NLA_U32
},
472 [RTA_SRC
] = { .type
= NLA_U32
},
473 [RTA_IIF
] = { .type
= NLA_U32
},
474 [RTA_OIF
] = { .type
= NLA_U32
},
475 [RTA_GATEWAY
] = { .type
= NLA_U32
},
476 [RTA_PRIORITY
] = { .type
= NLA_U32
},
477 [RTA_PREFSRC
] = { .type
= NLA_U32
},
478 [RTA_METRICS
] = { .type
= NLA_NESTED
},
479 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
480 [RTA_FLOW
] = { .type
= NLA_U32
},
483 static int rtm_to_fib_config(struct net
*net
, struct sk_buff
*skb
,
484 struct nlmsghdr
*nlh
, struct fib_config
*cfg
)
490 err
= nlmsg_validate(nlh
, sizeof(*rtm
), RTA_MAX
, rtm_ipv4_policy
);
494 memset(cfg
, 0, sizeof(*cfg
));
496 rtm
= nlmsg_data(nlh
);
497 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
498 cfg
->fc_tos
= rtm
->rtm_tos
;
499 cfg
->fc_table
= rtm
->rtm_table
;
500 cfg
->fc_protocol
= rtm
->rtm_protocol
;
501 cfg
->fc_scope
= rtm
->rtm_scope
;
502 cfg
->fc_type
= rtm
->rtm_type
;
503 cfg
->fc_flags
= rtm
->rtm_flags
;
504 cfg
->fc_nlflags
= nlh
->nlmsg_flags
;
506 cfg
->fc_nlinfo
.pid
= NETLINK_CB(skb
).pid
;
507 cfg
->fc_nlinfo
.nlh
= nlh
;
508 cfg
->fc_nlinfo
.nl_net
= net
;
510 if (cfg
->fc_type
> RTN_MAX
) {
515 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct rtmsg
), remaining
) {
516 switch (nla_type(attr
)) {
518 cfg
->fc_dst
= nla_get_be32(attr
);
521 cfg
->fc_oif
= nla_get_u32(attr
);
524 cfg
->fc_gw
= nla_get_be32(attr
);
527 cfg
->fc_priority
= nla_get_u32(attr
);
530 cfg
->fc_prefsrc
= nla_get_be32(attr
);
533 cfg
->fc_mx
= nla_data(attr
);
534 cfg
->fc_mx_len
= nla_len(attr
);
537 cfg
->fc_mp
= nla_data(attr
);
538 cfg
->fc_mp_len
= nla_len(attr
);
541 cfg
->fc_flow
= nla_get_u32(attr
);
544 cfg
->fc_table
= nla_get_u32(attr
);
554 static int inet_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
556 struct net
*net
= sock_net(skb
->sk
);
557 struct fib_config cfg
;
558 struct fib_table
*tb
;
561 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
);
565 tb
= fib_get_table(net
, cfg
.fc_table
);
571 err
= fib_table_delete(tb
, &cfg
);
576 static int inet_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
578 struct net
*net
= sock_net(skb
->sk
);
579 struct fib_config cfg
;
580 struct fib_table
*tb
;
583 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
);
587 tb
= fib_new_table(net
, cfg
.fc_table
);
593 err
= fib_table_insert(tb
, &cfg
);
598 static int inet_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
600 struct net
*net
= sock_net(skb
->sk
);
602 unsigned int e
= 0, s_e
;
603 struct fib_table
*tb
;
604 struct hlist_node
*node
;
605 struct hlist_head
*head
;
608 if (nlmsg_len(cb
->nlh
) >= sizeof(struct rtmsg
) &&
609 ((struct rtmsg
*) nlmsg_data(cb
->nlh
))->rtm_flags
& RTM_F_CLONED
)
610 return ip_rt_dump(skb
, cb
);
615 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
617 head
= &net
->ipv4
.fib_table_hash
[h
];
618 hlist_for_each_entry(tb
, node
, head
, tb_hlist
) {
622 memset(&cb
->args
[2], 0, sizeof(cb
->args
) -
623 2 * sizeof(cb
->args
[0]));
624 if (fib_table_dump(tb
, skb
, cb
) < 0)
638 /* Prepare and feed intra-kernel routing request.
639 * Really, it should be netlink message, but :-( netlink
640 * can be not configured, so that we feed it directly
641 * to fib engine. It is legal, because all events occur
642 * only when netlink is already locked.
644 static void fib_magic(int cmd
, int type
, __be32 dst
, int dst_len
, struct in_ifaddr
*ifa
)
646 struct net
*net
= dev_net(ifa
->ifa_dev
->dev
);
647 struct fib_table
*tb
;
648 struct fib_config cfg
= {
649 .fc_protocol
= RTPROT_KERNEL
,
652 .fc_dst_len
= dst_len
,
653 .fc_prefsrc
= ifa
->ifa_local
,
654 .fc_oif
= ifa
->ifa_dev
->dev
->ifindex
,
655 .fc_nlflags
= NLM_F_CREATE
| NLM_F_APPEND
,
661 if (type
== RTN_UNICAST
)
662 tb
= fib_new_table(net
, RT_TABLE_MAIN
);
664 tb
= fib_new_table(net
, RT_TABLE_LOCAL
);
669 cfg
.fc_table
= tb
->tb_id
;
671 if (type
!= RTN_LOCAL
)
672 cfg
.fc_scope
= RT_SCOPE_LINK
;
674 cfg
.fc_scope
= RT_SCOPE_HOST
;
676 if (cmd
== RTM_NEWROUTE
)
677 fib_table_insert(tb
, &cfg
);
679 fib_table_delete(tb
, &cfg
);
682 void fib_add_ifaddr(struct in_ifaddr
*ifa
)
684 struct in_device
*in_dev
= ifa
->ifa_dev
;
685 struct net_device
*dev
= in_dev
->dev
;
686 struct in_ifaddr
*prim
= ifa
;
687 __be32 mask
= ifa
->ifa_mask
;
688 __be32 addr
= ifa
->ifa_local
;
689 __be32 prefix
= ifa
->ifa_address
& mask
;
691 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
692 prim
= inet_ifa_byprefix(in_dev
, prefix
, mask
);
694 printk(KERN_WARNING
"fib_add_ifaddr: bug: prim == NULL\n");
699 fib_magic(RTM_NEWROUTE
, RTN_LOCAL
, addr
, 32, prim
);
701 if (!(dev
->flags
& IFF_UP
))
704 /* Add broadcast address, if it is explicitly assigned. */
705 if (ifa
->ifa_broadcast
&& ifa
->ifa_broadcast
!= htonl(0xFFFFFFFF))
706 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
708 if (!ipv4_is_zeronet(prefix
) && !(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
709 (prefix
!= addr
|| ifa
->ifa_prefixlen
< 32)) {
710 fib_magic(RTM_NEWROUTE
,
711 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
712 prefix
, ifa
->ifa_prefixlen
, prim
);
714 /* Add network specific broadcasts, when it takes a sense */
715 if (ifa
->ifa_prefixlen
< 31) {
716 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
, 32, prim
);
717 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
| ~mask
,
723 static void fib_del_ifaddr(struct in_ifaddr
*ifa
)
725 struct in_device
*in_dev
= ifa
->ifa_dev
;
726 struct net_device
*dev
= in_dev
->dev
;
727 struct in_ifaddr
*ifa1
;
728 struct in_ifaddr
*prim
= ifa
;
729 __be32 brd
= ifa
->ifa_address
| ~ifa
->ifa_mask
;
730 __be32 any
= ifa
->ifa_address
& ifa
->ifa_mask
;
737 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
))
738 fib_magic(RTM_DELROUTE
,
739 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
740 any
, ifa
->ifa_prefixlen
, prim
);
742 prim
= inet_ifa_byprefix(in_dev
, any
, ifa
->ifa_mask
);
744 printk(KERN_WARNING
"fib_del_ifaddr: bug: prim == NULL\n");
749 /* Deletion is more complicated than add.
750 * We should take care of not to delete too much :-)
752 * Scan address list to be sure that addresses are really gone.
755 for (ifa1
= in_dev
->ifa_list
; ifa1
; ifa1
= ifa1
->ifa_next
) {
756 if (ifa
->ifa_local
== ifa1
->ifa_local
)
758 if (ifa
->ifa_broadcast
== ifa1
->ifa_broadcast
)
760 if (brd
== ifa1
->ifa_broadcast
)
762 if (any
== ifa1
->ifa_broadcast
)
767 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
769 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, brd
, 32, prim
);
771 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, any
, 32, prim
);
772 if (!(ok
& LOCAL_OK
)) {
773 fib_magic(RTM_DELROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 32, prim
);
775 /* Check, that this local address finally disappeared. */
776 if (inet_addr_type(dev_net(dev
), ifa
->ifa_local
) != RTN_LOCAL
) {
777 /* And the last, but not the least thing.
778 * We must flush stray FIB entries.
780 * First of all, we scan fib_info list searching
781 * for stray nexthop entries, then ignite fib_flush.
783 if (fib_sync_down_addr(dev_net(dev
), ifa
->ifa_local
))
784 fib_flush(dev_net(dev
));
793 static void nl_fib_lookup(struct fib_result_nl
*frn
, struct fib_table
*tb
)
796 struct fib_result res
;
798 .mark
= frn
->fl_mark
,
799 .fl4_dst
= frn
->fl_addr
,
800 .fl4_tos
= frn
->fl_tos
,
801 .fl4_scope
= frn
->fl_scope
,
804 #ifdef CONFIG_IP_MULTIPLE_TABLES
812 frn
->tb_id
= tb
->tb_id
;
814 frn
->err
= fib_table_lookup(tb
, &fl
, &res
, FIB_LOOKUP_NOREF
);
817 frn
->prefixlen
= res
.prefixlen
;
818 frn
->nh_sel
= res
.nh_sel
;
819 frn
->type
= res
.type
;
820 frn
->scope
= res
.scope
;
827 static void nl_fib_input(struct sk_buff
*skb
)
830 struct fib_result_nl
*frn
;
831 struct nlmsghdr
*nlh
;
832 struct fib_table
*tb
;
835 net
= sock_net(skb
->sk
);
836 nlh
= nlmsg_hdr(skb
);
837 if (skb
->len
< NLMSG_SPACE(0) || skb
->len
< nlh
->nlmsg_len
||
838 nlh
->nlmsg_len
< NLMSG_LENGTH(sizeof(*frn
)))
841 skb
= skb_clone(skb
, GFP_KERNEL
);
844 nlh
= nlmsg_hdr(skb
);
846 frn
= (struct fib_result_nl
*) NLMSG_DATA(nlh
);
847 tb
= fib_get_table(net
, frn
->tb_id_in
);
849 nl_fib_lookup(frn
, tb
);
851 pid
= NETLINK_CB(skb
).pid
; /* pid of sending process */
852 NETLINK_CB(skb
).pid
= 0; /* from kernel */
853 NETLINK_CB(skb
).dst_group
= 0; /* unicast */
854 netlink_unicast(net
->ipv4
.fibnl
, skb
, pid
, MSG_DONTWAIT
);
857 static int __net_init
nl_fib_lookup_init(struct net
*net
)
860 sk
= netlink_kernel_create(net
, NETLINK_FIB_LOOKUP
, 0,
861 nl_fib_input
, NULL
, THIS_MODULE
);
863 return -EAFNOSUPPORT
;
864 net
->ipv4
.fibnl
= sk
;
868 static void nl_fib_lookup_exit(struct net
*net
)
870 netlink_kernel_release(net
->ipv4
.fibnl
);
871 net
->ipv4
.fibnl
= NULL
;
874 static void fib_disable_ip(struct net_device
*dev
, int force
, int delay
)
876 if (fib_sync_down_dev(dev
, force
))
877 fib_flush(dev_net(dev
));
878 rt_cache_flush(dev_net(dev
), delay
);
882 static int fib_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
884 struct in_ifaddr
*ifa
= (struct in_ifaddr
*)ptr
;
885 struct net_device
*dev
= ifa
->ifa_dev
->dev
;
890 #ifdef CONFIG_IP_ROUTE_MULTIPATH
893 rt_cache_flush(dev_net(dev
), -1);
897 if (ifa
->ifa_dev
->ifa_list
== NULL
) {
898 /* Last address was deleted from this interface.
901 fib_disable_ip(dev
, 1, 0);
903 rt_cache_flush(dev_net(dev
), -1);
910 static int fib_netdev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
912 struct net_device
*dev
= ptr
;
913 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
915 if (event
== NETDEV_UNREGISTER
) {
916 fib_disable_ip(dev
, 2, -1);
927 } endfor_ifa(in_dev
);
928 #ifdef CONFIG_IP_ROUTE_MULTIPATH
931 rt_cache_flush(dev_net(dev
), -1);
934 fib_disable_ip(dev
, 0, 0);
936 case NETDEV_CHANGEMTU
:
938 rt_cache_flush(dev_net(dev
), 0);
940 case NETDEV_UNREGISTER_BATCH
:
941 /* The batch unregister is only called on the first
942 * device in the list of devices being unregistered.
943 * Therefore we should not pass dev_net(dev) in here.
945 rt_cache_flush_batch(NULL
);
951 static struct notifier_block fib_inetaddr_notifier
= {
952 .notifier_call
= fib_inetaddr_event
,
955 static struct notifier_block fib_netdev_notifier
= {
956 .notifier_call
= fib_netdev_event
,
959 static int __net_init
ip_fib_net_init(struct net
*net
)
962 size_t size
= sizeof(struct hlist_head
) * FIB_TABLE_HASHSZ
;
964 /* Avoid false sharing : Use at least a full cache line */
965 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
967 net
->ipv4
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
968 if (net
->ipv4
.fib_table_hash
== NULL
)
971 err
= fib4_rules_init(net
);
977 kfree(net
->ipv4
.fib_table_hash
);
981 static void ip_fib_net_exit(struct net
*net
)
985 #ifdef CONFIG_IP_MULTIPLE_TABLES
986 fib4_rules_exit(net
);
989 for (i
= 0; i
< FIB_TABLE_HASHSZ
; i
++) {
990 struct fib_table
*tb
;
991 struct hlist_head
*head
;
992 struct hlist_node
*node
, *tmp
;
994 head
= &net
->ipv4
.fib_table_hash
[i
];
995 hlist_for_each_entry_safe(tb
, node
, tmp
, head
, tb_hlist
) {
1001 kfree(net
->ipv4
.fib_table_hash
);
1004 static int __net_init
fib_net_init(struct net
*net
)
1008 error
= ip_fib_net_init(net
);
1011 error
= nl_fib_lookup_init(net
);
1014 error
= fib_proc_init(net
);
1021 nl_fib_lookup_exit(net
);
1023 ip_fib_net_exit(net
);
1027 static void __net_exit
fib_net_exit(struct net
*net
)
1030 nl_fib_lookup_exit(net
);
1031 ip_fib_net_exit(net
);
1034 static struct pernet_operations fib_net_ops
= {
1035 .init
= fib_net_init
,
1036 .exit
= fib_net_exit
,
1039 void __init
ip_fib_init(void)
1041 rtnl_register(PF_INET
, RTM_NEWROUTE
, inet_rtm_newroute
, NULL
);
1042 rtnl_register(PF_INET
, RTM_DELROUTE
, inet_rtm_delroute
, NULL
);
1043 rtnl_register(PF_INET
, RTM_GETROUTE
, NULL
, inet_dump_fib
);
1045 register_pernet_subsys(&fib_net_ops
);
1046 register_netdevice_notifier(&fib_netdev_notifier
);
1047 register_inetaddr_notifier(&fib_inetaddr_notifier
);