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 <linux/uaccess.h>
18 #include <linux/bitops.h>
19 #include <linux/capability.h>
20 #include <linux/types.h>
21 #include <linux/kernel.h>
23 #include <linux/string.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/errno.h>
28 #include <linux/inet.h>
29 #include <linux/inetdevice.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_addr.h>
32 #include <linux/if_arp.h>
33 #include <linux/skbuff.h>
34 #include <linux/cache.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 #include <net/l3mdev.h>
49 #include <net/lwtunnel.h>
50 #include <trace/events/fib.h>
52 #ifndef CONFIG_IP_MULTIPLE_TABLES
54 static int __net_init
fib4_rules_init(struct net
*net
)
56 struct fib_table
*local_table
, *main_table
;
58 main_table
= fib_trie_table(RT_TABLE_MAIN
, NULL
);
62 local_table
= fib_trie_table(RT_TABLE_LOCAL
, main_table
);
66 hlist_add_head_rcu(&local_table
->tb_hlist
,
67 &net
->ipv4
.fib_table_hash
[TABLE_LOCAL_INDEX
]);
68 hlist_add_head_rcu(&main_table
->tb_hlist
,
69 &net
->ipv4
.fib_table_hash
[TABLE_MAIN_INDEX
]);
73 fib_free_table(main_table
);
78 struct fib_table
*fib_new_table(struct net
*net
, u32 id
)
80 struct fib_table
*tb
, *alias
= NULL
;
85 tb
= fib_get_table(net
, id
);
89 if (id
== RT_TABLE_LOCAL
&& !net
->ipv4
.fib_has_custom_rules
)
90 alias
= fib_new_table(net
, RT_TABLE_MAIN
);
92 tb
= fib_trie_table(id
, alias
);
98 rcu_assign_pointer(net
->ipv4
.fib_main
, tb
);
100 case RT_TABLE_DEFAULT
:
101 rcu_assign_pointer(net
->ipv4
.fib_default
, tb
);
107 h
= id
& (FIB_TABLE_HASHSZ
- 1);
108 hlist_add_head_rcu(&tb
->tb_hlist
, &net
->ipv4
.fib_table_hash
[h
]);
111 EXPORT_SYMBOL_GPL(fib_new_table
);
113 /* caller must hold either rtnl or rcu read lock */
114 struct fib_table
*fib_get_table(struct net
*net
, u32 id
)
116 struct fib_table
*tb
;
117 struct hlist_head
*head
;
122 h
= id
& (FIB_TABLE_HASHSZ
- 1);
124 head
= &net
->ipv4
.fib_table_hash
[h
];
125 hlist_for_each_entry_rcu(tb
, head
, tb_hlist
) {
131 #endif /* CONFIG_IP_MULTIPLE_TABLES */
133 static void fib_replace_table(struct net
*net
, struct fib_table
*old
,
134 struct fib_table
*new)
136 #ifdef CONFIG_IP_MULTIPLE_TABLES
137 switch (new->tb_id
) {
139 rcu_assign_pointer(net
->ipv4
.fib_main
, new);
141 case RT_TABLE_DEFAULT
:
142 rcu_assign_pointer(net
->ipv4
.fib_default
, new);
149 /* replace the old table in the hlist */
150 hlist_replace_rcu(&old
->tb_hlist
, &new->tb_hlist
);
153 int fib_unmerge(struct net
*net
)
155 struct fib_table
*old
, *new, *main_table
;
157 /* attempt to fetch local table if it has been allocated */
158 old
= fib_get_table(net
, RT_TABLE_LOCAL
);
162 new = fib_trie_unmerge(old
);
166 /* table is already unmerged */
170 /* replace merged table with clean table */
171 fib_replace_table(net
, old
, new);
174 /* attempt to fetch main table if it has been allocated */
175 main_table
= fib_get_table(net
, RT_TABLE_MAIN
);
179 /* flush local entries from main table */
180 fib_table_flush_external(main_table
);
185 static void fib_flush(struct net
*net
)
190 for (h
= 0; h
< FIB_TABLE_HASHSZ
; h
++) {
191 struct hlist_head
*head
= &net
->ipv4
.fib_table_hash
[h
];
192 struct hlist_node
*tmp
;
193 struct fib_table
*tb
;
195 hlist_for_each_entry_safe(tb
, tmp
, head
, tb_hlist
)
196 flushed
+= fib_table_flush(net
, tb
);
204 * Find address type as if only "dev" was present in the system. If
205 * on_dev is NULL then all interfaces are taken into consideration.
207 static inline unsigned int __inet_dev_addr_type(struct net
*net
,
208 const struct net_device
*dev
,
209 __be32 addr
, u32 tb_id
)
211 struct flowi4 fl4
= { .daddr
= addr
};
212 struct fib_result res
;
213 unsigned int ret
= RTN_BROADCAST
;
214 struct fib_table
*table
;
216 if (ipv4_is_zeronet(addr
) || ipv4_is_lbcast(addr
))
217 return RTN_BROADCAST
;
218 if (ipv4_is_multicast(addr
))
219 return RTN_MULTICAST
;
223 table
= fib_get_table(net
, tb_id
);
226 if (!fib_table_lookup(table
, &fl4
, &res
, FIB_LOOKUP_NOREF
)) {
227 if (!dev
|| dev
== res
.fi
->fib_dev
)
236 unsigned int inet_addr_type_table(struct net
*net
, __be32 addr
, u32 tb_id
)
238 return __inet_dev_addr_type(net
, NULL
, addr
, tb_id
);
240 EXPORT_SYMBOL(inet_addr_type_table
);
242 unsigned int inet_addr_type(struct net
*net
, __be32 addr
)
244 return __inet_dev_addr_type(net
, NULL
, addr
, RT_TABLE_LOCAL
);
246 EXPORT_SYMBOL(inet_addr_type
);
248 unsigned int inet_dev_addr_type(struct net
*net
, const struct net_device
*dev
,
251 u32 rt_table
= l3mdev_fib_table(dev
) ? : RT_TABLE_LOCAL
;
253 return __inet_dev_addr_type(net
, dev
, addr
, rt_table
);
255 EXPORT_SYMBOL(inet_dev_addr_type
);
257 /* inet_addr_type with dev == NULL but using the table from a dev
258 * if one is associated
260 unsigned int inet_addr_type_dev_table(struct net
*net
,
261 const struct net_device
*dev
,
264 u32 rt_table
= l3mdev_fib_table(dev
) ? : RT_TABLE_LOCAL
;
266 return __inet_dev_addr_type(net
, NULL
, addr
, rt_table
);
268 EXPORT_SYMBOL(inet_addr_type_dev_table
);
270 __be32
fib_compute_spec_dst(struct sk_buff
*skb
)
272 struct net_device
*dev
= skb
->dev
;
273 struct in_device
*in_dev
;
274 struct fib_result res
;
279 rt
= skb_rtable(skb
);
280 if ((rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
| RTCF_LOCAL
)) ==
282 return ip_hdr(skb
)->daddr
;
284 in_dev
= __in_dev_get_rcu(dev
);
289 scope
= RT_SCOPE_UNIVERSE
;
290 if (!ipv4_is_zeronet(ip_hdr(skb
)->saddr
)) {
291 struct flowi4 fl4
= {
292 .flowi4_iif
= LOOPBACK_IFINDEX
,
293 .daddr
= ip_hdr(skb
)->saddr
,
294 .flowi4_tos
= RT_TOS(ip_hdr(skb
)->tos
),
295 .flowi4_scope
= scope
,
296 .flowi4_mark
= IN_DEV_SRC_VMARK(in_dev
) ? skb
->mark
: 0,
298 if (!fib_lookup(net
, &fl4
, &res
, 0))
299 return FIB_RES_PREFSRC(net
, res
);
301 scope
= RT_SCOPE_LINK
;
304 return inet_select_addr(dev
, ip_hdr(skb
)->saddr
, scope
);
307 /* Given (packet source, input interface) and optional (dst, oif, tos):
308 * - (main) check, that source is valid i.e. not broadcast or our local
310 * - figure out what "logical" interface this packet arrived
311 * and calculate "specific destination" address.
312 * - check, that packet arrived from expected physical interface.
313 * called with rcu_read_lock()
315 static int __fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
316 u8 tos
, int oif
, struct net_device
*dev
,
317 int rpf
, struct in_device
*idev
, u32
*itag
)
320 struct fib_result res
;
322 struct net
*net
= dev_net(dev
);
326 fl4
.flowi4_iif
= l3mdev_master_ifindex_rcu(dev
);
328 fl4
.flowi4_iif
= oif
? : LOOPBACK_IFINDEX
;
331 fl4
.flowi4_tos
= tos
;
332 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
333 fl4
.flowi4_tun_key
.tun_id
= 0;
334 fl4
.flowi4_flags
= 0;
335 fl4
.flowi4_uid
= sock_net_uid(net
, NULL
);
337 no_addr
= idev
->ifa_list
== NULL
;
339 fl4
.flowi4_mark
= IN_DEV_SRC_VMARK(idev
) ? skb
->mark
: 0;
341 trace_fib_validate_source(dev
, &fl4
);
343 if (fib_lookup(net
, &fl4
, &res
, 0))
345 if (res
.type
!= RTN_UNICAST
&&
346 (res
.type
!= RTN_LOCAL
|| !IN_DEV_ACCEPT_LOCAL(idev
)))
348 if (!rpf
&& !fib_num_tclassid_users(net
) &&
349 (dev
->ifindex
!= oif
|| !IN_DEV_TX_REDIRECTS(idev
)))
351 fib_combine_itag(itag
, &res
);
354 #ifdef CONFIG_IP_ROUTE_MULTIPATH
355 for (ret
= 0; ret
< res
.fi
->fib_nhs
; ret
++) {
356 struct fib_nh
*nh
= &res
.fi
->fib_nh
[ret
];
358 if (nh
->nh_dev
== dev
) {
361 } else if (l3mdev_master_ifindex_rcu(nh
->nh_dev
) == dev
->ifindex
) {
367 if (FIB_RES_DEV(res
) == dev
)
371 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
378 fl4
.flowi4_oif
= dev
->ifindex
;
381 if (fib_lookup(net
, &fl4
, &res
, FIB_LOOKUP_IGNORE_LINKSTATE
) == 0) {
382 if (res
.type
== RTN_UNICAST
)
383 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
399 /* Ignore rp_filter for packets protected by IPsec. */
400 int fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
401 u8 tos
, int oif
, struct net_device
*dev
,
402 struct in_device
*idev
, u32
*itag
)
404 int r
= secpath_exists(skb
) ? 0 : IN_DEV_RPFILTER(idev
);
406 if (!r
&& !fib_num_tclassid_users(dev_net(dev
)) &&
407 IN_DEV_ACCEPT_LOCAL(idev
) &&
408 (dev
->ifindex
!= oif
|| !IN_DEV_TX_REDIRECTS(idev
))) {
412 return __fib_validate_source(skb
, src
, dst
, tos
, oif
, dev
, r
, idev
, itag
);
415 static inline __be32
sk_extract_addr(struct sockaddr
*addr
)
417 return ((struct sockaddr_in
*) addr
)->sin_addr
.s_addr
;
420 static int put_rtax(struct nlattr
*mx
, int len
, int type
, u32 value
)
424 nla
= (struct nlattr
*) ((char *) mx
+ len
);
425 nla
->nla_type
= type
;
426 nla
->nla_len
= nla_attr_size(4);
427 *(u32
*) nla_data(nla
) = value
;
429 return len
+ nla_total_size(4);
432 static int rtentry_to_fib_config(struct net
*net
, int cmd
, struct rtentry
*rt
,
433 struct fib_config
*cfg
)
438 memset(cfg
, 0, sizeof(*cfg
));
439 cfg
->fc_nlinfo
.nl_net
= net
;
441 if (rt
->rt_dst
.sa_family
!= AF_INET
)
442 return -EAFNOSUPPORT
;
445 * Check mask for validity:
446 * a) it must be contiguous.
447 * b) destination must have all host bits clear.
448 * c) if application forgot to set correct family (AF_INET),
449 * reject request unless it is absolutely clear i.e.
450 * both family and mask are zero.
453 addr
= sk_extract_addr(&rt
->rt_dst
);
454 if (!(rt
->rt_flags
& RTF_HOST
)) {
455 __be32 mask
= sk_extract_addr(&rt
->rt_genmask
);
457 if (rt
->rt_genmask
.sa_family
!= AF_INET
) {
458 if (mask
|| rt
->rt_genmask
.sa_family
)
459 return -EAFNOSUPPORT
;
462 if (bad_mask(mask
, addr
))
465 plen
= inet_mask_len(mask
);
468 cfg
->fc_dst_len
= plen
;
471 if (cmd
!= SIOCDELRT
) {
472 cfg
->fc_nlflags
= NLM_F_CREATE
;
473 cfg
->fc_protocol
= RTPROT_BOOT
;
477 cfg
->fc_priority
= rt
->rt_metric
- 1;
479 if (rt
->rt_flags
& RTF_REJECT
) {
480 cfg
->fc_scope
= RT_SCOPE_HOST
;
481 cfg
->fc_type
= RTN_UNREACHABLE
;
485 cfg
->fc_scope
= RT_SCOPE_NOWHERE
;
486 cfg
->fc_type
= RTN_UNICAST
;
490 struct net_device
*dev
;
491 char devname
[IFNAMSIZ
];
493 if (copy_from_user(devname
, rt
->rt_dev
, IFNAMSIZ
-1))
496 devname
[IFNAMSIZ
-1] = 0;
497 colon
= strchr(devname
, ':');
500 dev
= __dev_get_by_name(net
, devname
);
503 cfg
->fc_oif
= dev
->ifindex
;
504 cfg
->fc_table
= l3mdev_fib_table(dev
);
506 struct in_ifaddr
*ifa
;
507 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
511 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
)
512 if (strcmp(ifa
->ifa_label
, devname
) == 0)
516 cfg
->fc_prefsrc
= ifa
->ifa_local
;
520 addr
= sk_extract_addr(&rt
->rt_gateway
);
521 if (rt
->rt_gateway
.sa_family
== AF_INET
&& addr
) {
522 unsigned int addr_type
;
525 addr_type
= inet_addr_type_table(net
, addr
, cfg
->fc_table
);
526 if (rt
->rt_flags
& RTF_GATEWAY
&&
527 addr_type
== RTN_UNICAST
)
528 cfg
->fc_scope
= RT_SCOPE_UNIVERSE
;
531 if (cmd
== SIOCDELRT
)
534 if (rt
->rt_flags
& RTF_GATEWAY
&& !cfg
->fc_gw
)
537 if (cfg
->fc_scope
== RT_SCOPE_NOWHERE
)
538 cfg
->fc_scope
= RT_SCOPE_LINK
;
540 if (rt
->rt_flags
& (RTF_MTU
| RTF_WINDOW
| RTF_IRTT
)) {
544 mx
= kzalloc(3 * nla_total_size(4), GFP_KERNEL
);
548 if (rt
->rt_flags
& RTF_MTU
)
549 len
= put_rtax(mx
, len
, RTAX_ADVMSS
, rt
->rt_mtu
- 40);
551 if (rt
->rt_flags
& RTF_WINDOW
)
552 len
= put_rtax(mx
, len
, RTAX_WINDOW
, rt
->rt_window
);
554 if (rt
->rt_flags
& RTF_IRTT
)
555 len
= put_rtax(mx
, len
, RTAX_RTT
, rt
->rt_irtt
<< 3);
558 cfg
->fc_mx_len
= len
;
565 * Handle IP routing ioctl calls.
566 * These are used to manipulate the routing tables
568 int ip_rt_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
570 struct fib_config cfg
;
575 case SIOCADDRT
: /* Add a route */
576 case SIOCDELRT
: /* Delete a route */
577 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
580 if (copy_from_user(&rt
, arg
, sizeof(rt
)))
584 err
= rtentry_to_fib_config(net
, cmd
, &rt
, &cfg
);
586 struct fib_table
*tb
;
588 if (cmd
== SIOCDELRT
) {
589 tb
= fib_get_table(net
, cfg
.fc_table
);
591 err
= fib_table_delete(net
, tb
, &cfg
);
595 tb
= fib_new_table(net
, cfg
.fc_table
);
597 err
= fib_table_insert(net
, tb
, &cfg
);
602 /* allocated by rtentry_to_fib_config() */
611 const struct nla_policy rtm_ipv4_policy
[RTA_MAX
+ 1] = {
612 [RTA_DST
] = { .type
= NLA_U32
},
613 [RTA_SRC
] = { .type
= NLA_U32
},
614 [RTA_IIF
] = { .type
= NLA_U32
},
615 [RTA_OIF
] = { .type
= NLA_U32
},
616 [RTA_GATEWAY
] = { .type
= NLA_U32
},
617 [RTA_PRIORITY
] = { .type
= NLA_U32
},
618 [RTA_PREFSRC
] = { .type
= NLA_U32
},
619 [RTA_METRICS
] = { .type
= NLA_NESTED
},
620 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
621 [RTA_FLOW
] = { .type
= NLA_U32
},
622 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
623 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
624 [RTA_UID
] = { .type
= NLA_U32
},
625 [RTA_MARK
] = { .type
= NLA_U32
},
628 static int rtm_to_fib_config(struct net
*net
, struct sk_buff
*skb
,
629 struct nlmsghdr
*nlh
, struct fib_config
*cfg
)
635 err
= nlmsg_validate(nlh
, sizeof(*rtm
), RTA_MAX
, rtm_ipv4_policy
,
640 memset(cfg
, 0, sizeof(*cfg
));
642 rtm
= nlmsg_data(nlh
);
643 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
644 cfg
->fc_tos
= rtm
->rtm_tos
;
645 cfg
->fc_table
= rtm
->rtm_table
;
646 cfg
->fc_protocol
= rtm
->rtm_protocol
;
647 cfg
->fc_scope
= rtm
->rtm_scope
;
648 cfg
->fc_type
= rtm
->rtm_type
;
649 cfg
->fc_flags
= rtm
->rtm_flags
;
650 cfg
->fc_nlflags
= nlh
->nlmsg_flags
;
652 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
653 cfg
->fc_nlinfo
.nlh
= nlh
;
654 cfg
->fc_nlinfo
.nl_net
= net
;
656 if (cfg
->fc_type
> RTN_MAX
) {
661 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct rtmsg
), remaining
) {
662 switch (nla_type(attr
)) {
664 cfg
->fc_dst
= nla_get_be32(attr
);
667 cfg
->fc_oif
= nla_get_u32(attr
);
670 cfg
->fc_gw
= nla_get_be32(attr
);
673 cfg
->fc_priority
= nla_get_u32(attr
);
676 cfg
->fc_prefsrc
= nla_get_be32(attr
);
679 cfg
->fc_mx
= nla_data(attr
);
680 cfg
->fc_mx_len
= nla_len(attr
);
683 err
= lwtunnel_valid_encap_type_attr(nla_data(attr
),
687 cfg
->fc_mp
= nla_data(attr
);
688 cfg
->fc_mp_len
= nla_len(attr
);
691 cfg
->fc_flow
= nla_get_u32(attr
);
694 cfg
->fc_table
= nla_get_u32(attr
);
697 cfg
->fc_encap
= attr
;
700 cfg
->fc_encap_type
= nla_get_u16(attr
);
701 err
= lwtunnel_valid_encap_type(cfg
->fc_encap_type
);
713 static int inet_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
714 struct netlink_ext_ack
*extack
)
716 struct net
*net
= sock_net(skb
->sk
);
717 struct fib_config cfg
;
718 struct fib_table
*tb
;
721 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
);
725 tb
= fib_get_table(net
, cfg
.fc_table
);
731 err
= fib_table_delete(net
, tb
, &cfg
);
736 static int inet_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
737 struct netlink_ext_ack
*extack
)
739 struct net
*net
= sock_net(skb
->sk
);
740 struct fib_config cfg
;
741 struct fib_table
*tb
;
744 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
);
748 tb
= fib_new_table(net
, cfg
.fc_table
);
754 err
= fib_table_insert(net
, tb
, &cfg
);
759 static int inet_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
761 struct net
*net
= sock_net(skb
->sk
);
763 unsigned int e
= 0, s_e
;
764 struct fib_table
*tb
;
765 struct hlist_head
*head
;
768 if (nlmsg_len(cb
->nlh
) >= sizeof(struct rtmsg
) &&
769 ((struct rtmsg
*) nlmsg_data(cb
->nlh
))->rtm_flags
& RTM_F_CLONED
)
777 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
779 head
= &net
->ipv4
.fib_table_hash
[h
];
780 hlist_for_each_entry_rcu(tb
, head
, tb_hlist
) {
784 memset(&cb
->args
[2], 0, sizeof(cb
->args
) -
785 2 * sizeof(cb
->args
[0]));
786 if (fib_table_dump(tb
, skb
, cb
) < 0)
802 /* Prepare and feed intra-kernel routing request.
803 * Really, it should be netlink message, but :-( netlink
804 * can be not configured, so that we feed it directly
805 * to fib engine. It is legal, because all events occur
806 * only when netlink is already locked.
808 static void fib_magic(int cmd
, int type
, __be32 dst
, int dst_len
, struct in_ifaddr
*ifa
)
810 struct net
*net
= dev_net(ifa
->ifa_dev
->dev
);
811 u32 tb_id
= l3mdev_fib_table(ifa
->ifa_dev
->dev
);
812 struct fib_table
*tb
;
813 struct fib_config cfg
= {
814 .fc_protocol
= RTPROT_KERNEL
,
817 .fc_dst_len
= dst_len
,
818 .fc_prefsrc
= ifa
->ifa_local
,
819 .fc_oif
= ifa
->ifa_dev
->dev
->ifindex
,
820 .fc_nlflags
= NLM_F_CREATE
| NLM_F_APPEND
,
827 tb_id
= (type
== RTN_UNICAST
) ? RT_TABLE_MAIN
: RT_TABLE_LOCAL
;
829 tb
= fib_new_table(net
, tb_id
);
833 cfg
.fc_table
= tb
->tb_id
;
835 if (type
!= RTN_LOCAL
)
836 cfg
.fc_scope
= RT_SCOPE_LINK
;
838 cfg
.fc_scope
= RT_SCOPE_HOST
;
840 if (cmd
== RTM_NEWROUTE
)
841 fib_table_insert(net
, tb
, &cfg
);
843 fib_table_delete(net
, tb
, &cfg
);
846 void fib_add_ifaddr(struct in_ifaddr
*ifa
)
848 struct in_device
*in_dev
= ifa
->ifa_dev
;
849 struct net_device
*dev
= in_dev
->dev
;
850 struct in_ifaddr
*prim
= ifa
;
851 __be32 mask
= ifa
->ifa_mask
;
852 __be32 addr
= ifa
->ifa_local
;
853 __be32 prefix
= ifa
->ifa_address
& mask
;
855 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
856 prim
= inet_ifa_byprefix(in_dev
, prefix
, mask
);
858 pr_warn("%s: bug: prim == NULL\n", __func__
);
863 fib_magic(RTM_NEWROUTE
, RTN_LOCAL
, addr
, 32, prim
);
865 if (!(dev
->flags
& IFF_UP
))
868 /* Add broadcast address, if it is explicitly assigned. */
869 if (ifa
->ifa_broadcast
&& ifa
->ifa_broadcast
!= htonl(0xFFFFFFFF))
870 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
872 if (!ipv4_is_zeronet(prefix
) && !(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
873 (prefix
!= addr
|| ifa
->ifa_prefixlen
< 32)) {
874 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
875 fib_magic(RTM_NEWROUTE
,
876 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
877 prefix
, ifa
->ifa_prefixlen
, prim
);
879 /* Add network specific broadcasts, when it takes a sense */
880 if (ifa
->ifa_prefixlen
< 31) {
881 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
, 32, prim
);
882 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
| ~mask
,
888 /* Delete primary or secondary address.
889 * Optionally, on secondary address promotion consider the addresses
890 * from subnet iprim as deleted, even if they are in device list.
891 * In this case the secondary ifa can be in device list.
893 void fib_del_ifaddr(struct in_ifaddr
*ifa
, struct in_ifaddr
*iprim
)
895 struct in_device
*in_dev
= ifa
->ifa_dev
;
896 struct net_device
*dev
= in_dev
->dev
;
897 struct in_ifaddr
*ifa1
;
898 struct in_ifaddr
*prim
= ifa
, *prim1
= NULL
;
899 __be32 brd
= ifa
->ifa_address
| ~ifa
->ifa_mask
;
900 __be32 any
= ifa
->ifa_address
& ifa
->ifa_mask
;
906 int subnet
= 0; /* Primary network */
907 int gone
= 1; /* Address is missing */
908 int same_prefsrc
= 0; /* Another primary with same IP */
910 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
911 prim
= inet_ifa_byprefix(in_dev
, any
, ifa
->ifa_mask
);
913 /* if the device has been deleted, we don't perform
917 pr_warn("%s: bug: prim == NULL\n", __func__
);
920 if (iprim
&& iprim
!= prim
) {
921 pr_warn("%s: bug: iprim != prim\n", __func__
);
924 } else if (!ipv4_is_zeronet(any
) &&
925 (any
!= ifa
->ifa_local
|| ifa
->ifa_prefixlen
< 32)) {
926 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
927 fib_magic(RTM_DELROUTE
,
928 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
929 any
, ifa
->ifa_prefixlen
, prim
);
936 /* Deletion is more complicated than add.
937 * We should take care of not to delete too much :-)
939 * Scan address list to be sure that addresses are really gone.
942 for (ifa1
= in_dev
->ifa_list
; ifa1
; ifa1
= ifa1
->ifa_next
) {
944 /* promotion, keep the IP */
948 /* Ignore IFAs from our subnet */
949 if (iprim
&& ifa1
->ifa_mask
== iprim
->ifa_mask
&&
950 inet_ifa_match(ifa1
->ifa_address
, iprim
))
953 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
954 if (ifa1
->ifa_flags
& IFA_F_SECONDARY
) {
955 /* Another address from our subnet? */
956 if (ifa1
->ifa_mask
== prim
->ifa_mask
&&
957 inet_ifa_match(ifa1
->ifa_address
, prim
))
960 /* We reached the secondaries, so
961 * same_prefsrc should be determined.
965 /* Search new prim1 if ifa1 is not
966 * using the current prim1
969 ifa1
->ifa_mask
!= prim1
->ifa_mask
||
970 !inet_ifa_match(ifa1
->ifa_address
, prim1
))
971 prim1
= inet_ifa_byprefix(in_dev
,
976 if (prim1
->ifa_local
!= prim
->ifa_local
)
980 if (prim
->ifa_local
!= ifa1
->ifa_local
)
986 if (ifa
->ifa_local
== ifa1
->ifa_local
)
988 if (ifa
->ifa_broadcast
== ifa1
->ifa_broadcast
)
990 if (brd
== ifa1
->ifa_broadcast
)
992 if (any
== ifa1
->ifa_broadcast
)
994 /* primary has network specific broadcasts */
995 if (prim1
== ifa1
&& ifa1
->ifa_prefixlen
< 31) {
996 __be32 brd1
= ifa1
->ifa_address
| ~ifa1
->ifa_mask
;
997 __be32 any1
= ifa1
->ifa_address
& ifa1
->ifa_mask
;
999 if (!ipv4_is_zeronet(any1
)) {
1000 if (ifa
->ifa_broadcast
== brd1
||
1001 ifa
->ifa_broadcast
== any1
)
1003 if (brd
== brd1
|| brd
== any1
)
1005 if (any
== brd1
|| any
== any1
)
1013 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
1014 if (subnet
&& ifa
->ifa_prefixlen
< 31) {
1015 if (!(ok
& BRD1_OK
))
1016 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, brd
, 32, prim
);
1017 if (!(ok
& BRD0_OK
))
1018 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, any
, 32, prim
);
1020 if (!(ok
& LOCAL_OK
)) {
1021 unsigned int addr_type
;
1023 fib_magic(RTM_DELROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 32, prim
);
1025 /* Check, that this local address finally disappeared. */
1026 addr_type
= inet_addr_type_dev_table(dev_net(dev
), dev
,
1028 if (gone
&& addr_type
!= RTN_LOCAL
) {
1029 /* And the last, but not the least thing.
1030 * We must flush stray FIB entries.
1032 * First of all, we scan fib_info list searching
1033 * for stray nexthop entries, then ignite fib_flush.
1035 if (fib_sync_down_addr(dev
, ifa
->ifa_local
))
1036 fib_flush(dev_net(dev
));
1045 static void nl_fib_lookup(struct net
*net
, struct fib_result_nl
*frn
)
1048 struct fib_result res
;
1049 struct flowi4 fl4
= {
1050 .flowi4_mark
= frn
->fl_mark
,
1051 .daddr
= frn
->fl_addr
,
1052 .flowi4_tos
= frn
->fl_tos
,
1053 .flowi4_scope
= frn
->fl_scope
,
1055 struct fib_table
*tb
;
1059 tb
= fib_get_table(net
, frn
->tb_id_in
);
1065 frn
->tb_id
= tb
->tb_id
;
1066 frn
->err
= fib_table_lookup(tb
, &fl4
, &res
, FIB_LOOKUP_NOREF
);
1069 frn
->prefixlen
= res
.prefixlen
;
1070 frn
->nh_sel
= res
.nh_sel
;
1071 frn
->type
= res
.type
;
1072 frn
->scope
= res
.scope
;
1080 static void nl_fib_input(struct sk_buff
*skb
)
1083 struct fib_result_nl
*frn
;
1084 struct nlmsghdr
*nlh
;
1087 net
= sock_net(skb
->sk
);
1088 nlh
= nlmsg_hdr(skb
);
1089 if (skb
->len
< nlmsg_total_size(sizeof(*frn
)) ||
1090 skb
->len
< nlh
->nlmsg_len
||
1091 nlmsg_len(nlh
) < sizeof(*frn
))
1094 skb
= netlink_skb_clone(skb
, GFP_KERNEL
);
1097 nlh
= nlmsg_hdr(skb
);
1099 frn
= (struct fib_result_nl
*) nlmsg_data(nlh
);
1100 nl_fib_lookup(net
, frn
);
1102 portid
= NETLINK_CB(skb
).portid
; /* netlink portid */
1103 NETLINK_CB(skb
).portid
= 0; /* from kernel */
1104 NETLINK_CB(skb
).dst_group
= 0; /* unicast */
1105 netlink_unicast(net
->ipv4
.fibnl
, skb
, portid
, MSG_DONTWAIT
);
1108 static int __net_init
nl_fib_lookup_init(struct net
*net
)
1111 struct netlink_kernel_cfg cfg
= {
1112 .input
= nl_fib_input
,
1115 sk
= netlink_kernel_create(net
, NETLINK_FIB_LOOKUP
, &cfg
);
1117 return -EAFNOSUPPORT
;
1118 net
->ipv4
.fibnl
= sk
;
1122 static void nl_fib_lookup_exit(struct net
*net
)
1124 netlink_kernel_release(net
->ipv4
.fibnl
);
1125 net
->ipv4
.fibnl
= NULL
;
1128 static void fib_disable_ip(struct net_device
*dev
, unsigned long event
,
1131 if (fib_sync_down_dev(dev
, event
, force
))
1132 fib_flush(dev_net(dev
));
1134 rt_cache_flush(dev_net(dev
));
1138 static int fib_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1140 struct in_ifaddr
*ifa
= (struct in_ifaddr
*)ptr
;
1141 struct net_device
*dev
= ifa
->ifa_dev
->dev
;
1142 struct net
*net
= dev_net(dev
);
1146 fib_add_ifaddr(ifa
);
1147 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1148 fib_sync_up(dev
, RTNH_F_DEAD
);
1150 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1151 rt_cache_flush(dev_net(dev
));
1154 fib_del_ifaddr(ifa
, NULL
);
1155 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1156 if (!ifa
->ifa_dev
->ifa_list
) {
1157 /* Last address was deleted from this interface.
1160 fib_disable_ip(dev
, event
, true);
1162 rt_cache_flush(dev_net(dev
));
1169 static int fib_netdev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1171 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1172 struct netdev_notifier_changeupper_info
*info
;
1173 struct in_device
*in_dev
;
1174 struct net
*net
= dev_net(dev
);
1177 if (event
== NETDEV_UNREGISTER
) {
1178 fib_disable_ip(dev
, event
, true);
1183 in_dev
= __in_dev_get_rtnl(dev
);
1190 fib_add_ifaddr(ifa
);
1191 } endfor_ifa(in_dev
);
1192 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1193 fib_sync_up(dev
, RTNH_F_DEAD
);
1195 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1196 rt_cache_flush(net
);
1199 fib_disable_ip(dev
, event
, false);
1202 flags
= dev_get_flags(dev
);
1203 if (flags
& (IFF_RUNNING
| IFF_LOWER_UP
))
1204 fib_sync_up(dev
, RTNH_F_LINKDOWN
);
1206 fib_sync_down_dev(dev
, event
, false);
1208 case NETDEV_CHANGEMTU
:
1209 rt_cache_flush(net
);
1211 case NETDEV_CHANGEUPPER
:
1213 /* flush all routes if dev is linked to or unlinked from
1214 * an L3 master device (e.g., VRF)
1216 if (info
->upper_dev
&& netif_is_l3_master(info
->upper_dev
))
1217 fib_disable_ip(dev
, NETDEV_DOWN
, true);
1223 static struct notifier_block fib_inetaddr_notifier
= {
1224 .notifier_call
= fib_inetaddr_event
,
1227 static struct notifier_block fib_netdev_notifier
= {
1228 .notifier_call
= fib_netdev_event
,
1231 static int __net_init
ip_fib_net_init(struct net
*net
)
1234 size_t size
= sizeof(struct hlist_head
) * FIB_TABLE_HASHSZ
;
1236 net
->ipv4
.fib_seq
= 0;
1238 /* Avoid false sharing : Use at least a full cache line */
1239 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1241 net
->ipv4
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1242 if (!net
->ipv4
.fib_table_hash
)
1245 err
= fib4_rules_init(net
);
1251 kfree(net
->ipv4
.fib_table_hash
);
1255 static void ip_fib_net_exit(struct net
*net
)
1260 #ifdef CONFIG_IP_MULTIPLE_TABLES
1261 RCU_INIT_POINTER(net
->ipv4
.fib_main
, NULL
);
1262 RCU_INIT_POINTER(net
->ipv4
.fib_default
, NULL
);
1264 for (i
= 0; i
< FIB_TABLE_HASHSZ
; i
++) {
1265 struct hlist_head
*head
= &net
->ipv4
.fib_table_hash
[i
];
1266 struct hlist_node
*tmp
;
1267 struct fib_table
*tb
;
1269 hlist_for_each_entry_safe(tb
, tmp
, head
, tb_hlist
) {
1270 hlist_del(&tb
->tb_hlist
);
1271 fib_table_flush(net
, tb
);
1276 #ifdef CONFIG_IP_MULTIPLE_TABLES
1277 fib4_rules_exit(net
);
1280 kfree(net
->ipv4
.fib_table_hash
);
1283 static int __net_init
fib_net_init(struct net
*net
)
1287 #ifdef CONFIG_IP_ROUTE_CLASSID
1288 net
->ipv4
.fib_num_tclassid_users
= 0;
1290 error
= ip_fib_net_init(net
);
1293 error
= nl_fib_lookup_init(net
);
1296 error
= fib_proc_init(net
);
1303 nl_fib_lookup_exit(net
);
1305 ip_fib_net_exit(net
);
1309 static void __net_exit
fib_net_exit(struct net
*net
)
1312 nl_fib_lookup_exit(net
);
1313 ip_fib_net_exit(net
);
1316 static struct pernet_operations fib_net_ops
= {
1317 .init
= fib_net_init
,
1318 .exit
= fib_net_exit
,
1321 void __init
ip_fib_init(void)
1323 rtnl_register(PF_INET
, RTM_NEWROUTE
, inet_rtm_newroute
, NULL
, NULL
);
1324 rtnl_register(PF_INET
, RTM_DELROUTE
, inet_rtm_delroute
, NULL
, NULL
);
1325 rtnl_register(PF_INET
, RTM_GETROUTE
, NULL
, inet_dump_fib
, NULL
);
1327 register_pernet_subsys(&fib_net_ops
);
1328 register_netdevice_notifier(&fib_netdev_notifier
);
1329 register_inetaddr_notifier(&fib_inetaddr_notifier
);