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
, false);
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
);
288 scope
= RT_SCOPE_UNIVERSE
;
289 if (!ipv4_is_zeronet(ip_hdr(skb
)->saddr
)) {
290 bool vmark
= in_dev
&& IN_DEV_SRC_VMARK(in_dev
);
291 struct flowi4 fl4
= {
292 .flowi4_iif
= LOOPBACK_IFINDEX
,
293 .flowi4_oif
= l3mdev_master_ifindex_rcu(dev
),
294 .daddr
= ip_hdr(skb
)->saddr
,
295 .flowi4_tos
= RT_TOS(ip_hdr(skb
)->tos
),
296 .flowi4_scope
= scope
,
297 .flowi4_mark
= vmark
? skb
->mark
: 0,
299 if (!fib_lookup(net
, &fl4
, &res
, 0))
300 return FIB_RES_PREFSRC(net
, res
);
302 scope
= RT_SCOPE_LINK
;
305 return inet_select_addr(dev
, ip_hdr(skb
)->saddr
, scope
);
308 /* Given (packet source, input interface) and optional (dst, oif, tos):
309 * - (main) check, that source is valid i.e. not broadcast or our local
311 * - figure out what "logical" interface this packet arrived
312 * and calculate "specific destination" address.
313 * - check, that packet arrived from expected physical interface.
314 * called with rcu_read_lock()
316 static int __fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
317 u8 tos
, int oif
, struct net_device
*dev
,
318 int rpf
, struct in_device
*idev
, u32
*itag
)
321 struct fib_result res
;
323 struct net
*net
= dev_net(dev
);
327 fl4
.flowi4_iif
= l3mdev_master_ifindex_rcu(dev
);
329 fl4
.flowi4_iif
= oif
? : LOOPBACK_IFINDEX
;
332 fl4
.flowi4_tos
= tos
;
333 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
334 fl4
.flowi4_tun_key
.tun_id
= 0;
335 fl4
.flowi4_flags
= 0;
336 fl4
.flowi4_uid
= sock_net_uid(net
, NULL
);
338 no_addr
= idev
->ifa_list
== NULL
;
340 fl4
.flowi4_mark
= IN_DEV_SRC_VMARK(idev
) ? skb
->mark
: 0;
342 trace_fib_validate_source(dev
, &fl4
);
344 if (fib_lookup(net
, &fl4
, &res
, 0))
346 if (res
.type
!= RTN_UNICAST
&&
347 (res
.type
!= RTN_LOCAL
|| !IN_DEV_ACCEPT_LOCAL(idev
)))
349 if (!rpf
&& !fib_num_tclassid_users(net
) &&
350 (dev
->ifindex
!= oif
|| !IN_DEV_TX_REDIRECTS(idev
)))
352 fib_combine_itag(itag
, &res
);
355 #ifdef CONFIG_IP_ROUTE_MULTIPATH
356 for (ret
= 0; ret
< res
.fi
->fib_nhs
; ret
++) {
357 struct fib_nh
*nh
= &res
.fi
->fib_nh
[ret
];
359 if (nh
->nh_dev
== dev
) {
362 } else if (l3mdev_master_ifindex_rcu(nh
->nh_dev
) == dev
->ifindex
) {
368 if (FIB_RES_DEV(res
) == dev
)
372 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
379 fl4
.flowi4_oif
= dev
->ifindex
;
382 if (fib_lookup(net
, &fl4
, &res
, FIB_LOOKUP_IGNORE_LINKSTATE
) == 0) {
383 if (res
.type
== RTN_UNICAST
)
384 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
400 /* Ignore rp_filter for packets protected by IPsec. */
401 int fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
402 u8 tos
, int oif
, struct net_device
*dev
,
403 struct in_device
*idev
, u32
*itag
)
405 int r
= secpath_exists(skb
) ? 0 : IN_DEV_RPFILTER(idev
);
407 if (!r
&& !fib_num_tclassid_users(dev_net(dev
)) &&
408 IN_DEV_ACCEPT_LOCAL(idev
) &&
409 (dev
->ifindex
!= oif
|| !IN_DEV_TX_REDIRECTS(idev
))) {
413 return __fib_validate_source(skb
, src
, dst
, tos
, oif
, dev
, r
, idev
, itag
);
416 static inline __be32
sk_extract_addr(struct sockaddr
*addr
)
418 return ((struct sockaddr_in
*) addr
)->sin_addr
.s_addr
;
421 static int put_rtax(struct nlattr
*mx
, int len
, int type
, u32 value
)
425 nla
= (struct nlattr
*) ((char *) mx
+ len
);
426 nla
->nla_type
= type
;
427 nla
->nla_len
= nla_attr_size(4);
428 *(u32
*) nla_data(nla
) = value
;
430 return len
+ nla_total_size(4);
433 static int rtentry_to_fib_config(struct net
*net
, int cmd
, struct rtentry
*rt
,
434 struct fib_config
*cfg
)
439 memset(cfg
, 0, sizeof(*cfg
));
440 cfg
->fc_nlinfo
.nl_net
= net
;
442 if (rt
->rt_dst
.sa_family
!= AF_INET
)
443 return -EAFNOSUPPORT
;
446 * Check mask for validity:
447 * a) it must be contiguous.
448 * b) destination must have all host bits clear.
449 * c) if application forgot to set correct family (AF_INET),
450 * reject request unless it is absolutely clear i.e.
451 * both family and mask are zero.
454 addr
= sk_extract_addr(&rt
->rt_dst
);
455 if (!(rt
->rt_flags
& RTF_HOST
)) {
456 __be32 mask
= sk_extract_addr(&rt
->rt_genmask
);
458 if (rt
->rt_genmask
.sa_family
!= AF_INET
) {
459 if (mask
|| rt
->rt_genmask
.sa_family
)
460 return -EAFNOSUPPORT
;
463 if (bad_mask(mask
, addr
))
466 plen
= inet_mask_len(mask
);
469 cfg
->fc_dst_len
= plen
;
472 if (cmd
!= SIOCDELRT
) {
473 cfg
->fc_nlflags
= NLM_F_CREATE
;
474 cfg
->fc_protocol
= RTPROT_BOOT
;
478 cfg
->fc_priority
= rt
->rt_metric
- 1;
480 if (rt
->rt_flags
& RTF_REJECT
) {
481 cfg
->fc_scope
= RT_SCOPE_HOST
;
482 cfg
->fc_type
= RTN_UNREACHABLE
;
486 cfg
->fc_scope
= RT_SCOPE_NOWHERE
;
487 cfg
->fc_type
= RTN_UNICAST
;
491 struct net_device
*dev
;
492 char devname
[IFNAMSIZ
];
494 if (copy_from_user(devname
, rt
->rt_dev
, IFNAMSIZ
-1))
497 devname
[IFNAMSIZ
-1] = 0;
498 colon
= strchr(devname
, ':');
501 dev
= __dev_get_by_name(net
, devname
);
504 cfg
->fc_oif
= dev
->ifindex
;
505 cfg
->fc_table
= l3mdev_fib_table(dev
);
507 struct in_ifaddr
*ifa
;
508 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
512 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
)
513 if (strcmp(ifa
->ifa_label
, devname
) == 0)
517 cfg
->fc_prefsrc
= ifa
->ifa_local
;
521 addr
= sk_extract_addr(&rt
->rt_gateway
);
522 if (rt
->rt_gateway
.sa_family
== AF_INET
&& addr
) {
523 unsigned int addr_type
;
526 addr_type
= inet_addr_type_table(net
, addr
, cfg
->fc_table
);
527 if (rt
->rt_flags
& RTF_GATEWAY
&&
528 addr_type
== RTN_UNICAST
)
529 cfg
->fc_scope
= RT_SCOPE_UNIVERSE
;
532 if (cmd
== SIOCDELRT
)
535 if (rt
->rt_flags
& RTF_GATEWAY
&& !cfg
->fc_gw
)
538 if (cfg
->fc_scope
== RT_SCOPE_NOWHERE
)
539 cfg
->fc_scope
= RT_SCOPE_LINK
;
541 if (rt
->rt_flags
& (RTF_MTU
| RTF_WINDOW
| RTF_IRTT
)) {
545 mx
= kzalloc(3 * nla_total_size(4), GFP_KERNEL
);
549 if (rt
->rt_flags
& RTF_MTU
)
550 len
= put_rtax(mx
, len
, RTAX_ADVMSS
, rt
->rt_mtu
- 40);
552 if (rt
->rt_flags
& RTF_WINDOW
)
553 len
= put_rtax(mx
, len
, RTAX_WINDOW
, rt
->rt_window
);
555 if (rt
->rt_flags
& RTF_IRTT
)
556 len
= put_rtax(mx
, len
, RTAX_RTT
, rt
->rt_irtt
<< 3);
559 cfg
->fc_mx_len
= len
;
566 * Handle IP routing ioctl calls.
567 * These are used to manipulate the routing tables
569 int ip_rt_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
571 struct fib_config cfg
;
576 case SIOCADDRT
: /* Add a route */
577 case SIOCDELRT
: /* Delete a route */
578 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
581 if (copy_from_user(&rt
, arg
, sizeof(rt
)))
585 err
= rtentry_to_fib_config(net
, cmd
, &rt
, &cfg
);
587 struct fib_table
*tb
;
589 if (cmd
== SIOCDELRT
) {
590 tb
= fib_get_table(net
, cfg
.fc_table
);
592 err
= fib_table_delete(net
, tb
, &cfg
,
597 tb
= fib_new_table(net
, cfg
.fc_table
);
599 err
= fib_table_insert(net
, tb
,
605 /* allocated by rtentry_to_fib_config() */
614 const struct nla_policy rtm_ipv4_policy
[RTA_MAX
+ 1] = {
615 [RTA_DST
] = { .type
= NLA_U32
},
616 [RTA_SRC
] = { .type
= NLA_U32
},
617 [RTA_IIF
] = { .type
= NLA_U32
},
618 [RTA_OIF
] = { .type
= NLA_U32
},
619 [RTA_GATEWAY
] = { .type
= NLA_U32
},
620 [RTA_PRIORITY
] = { .type
= NLA_U32
},
621 [RTA_PREFSRC
] = { .type
= NLA_U32
},
622 [RTA_METRICS
] = { .type
= NLA_NESTED
},
623 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
624 [RTA_FLOW
] = { .type
= NLA_U32
},
625 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
626 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
627 [RTA_UID
] = { .type
= NLA_U32
},
628 [RTA_MARK
] = { .type
= NLA_U32
},
629 [RTA_TABLE
] = { .type
= NLA_U32
},
632 static int rtm_to_fib_config(struct net
*net
, struct sk_buff
*skb
,
633 struct nlmsghdr
*nlh
, struct fib_config
*cfg
,
634 struct netlink_ext_ack
*extack
)
640 err
= nlmsg_validate(nlh
, sizeof(*rtm
), RTA_MAX
, rtm_ipv4_policy
,
645 memset(cfg
, 0, sizeof(*cfg
));
647 rtm
= nlmsg_data(nlh
);
648 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
649 cfg
->fc_tos
= rtm
->rtm_tos
;
650 cfg
->fc_table
= rtm
->rtm_table
;
651 cfg
->fc_protocol
= rtm
->rtm_protocol
;
652 cfg
->fc_scope
= rtm
->rtm_scope
;
653 cfg
->fc_type
= rtm
->rtm_type
;
654 cfg
->fc_flags
= rtm
->rtm_flags
;
655 cfg
->fc_nlflags
= nlh
->nlmsg_flags
;
657 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
658 cfg
->fc_nlinfo
.nlh
= nlh
;
659 cfg
->fc_nlinfo
.nl_net
= net
;
661 if (cfg
->fc_type
> RTN_MAX
) {
662 NL_SET_ERR_MSG(extack
, "Invalid route type");
667 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct rtmsg
), remaining
) {
668 switch (nla_type(attr
)) {
670 cfg
->fc_dst
= nla_get_be32(attr
);
673 cfg
->fc_oif
= nla_get_u32(attr
);
676 cfg
->fc_gw
= nla_get_be32(attr
);
679 cfg
->fc_priority
= nla_get_u32(attr
);
682 cfg
->fc_prefsrc
= nla_get_be32(attr
);
685 cfg
->fc_mx
= nla_data(attr
);
686 cfg
->fc_mx_len
= nla_len(attr
);
689 err
= lwtunnel_valid_encap_type_attr(nla_data(attr
),
694 cfg
->fc_mp
= nla_data(attr
);
695 cfg
->fc_mp_len
= nla_len(attr
);
698 cfg
->fc_flow
= nla_get_u32(attr
);
701 cfg
->fc_table
= nla_get_u32(attr
);
704 cfg
->fc_encap
= attr
;
707 cfg
->fc_encap_type
= nla_get_u16(attr
);
708 err
= lwtunnel_valid_encap_type(cfg
->fc_encap_type
,
721 static int inet_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
722 struct netlink_ext_ack
*extack
)
724 struct net
*net
= sock_net(skb
->sk
);
725 struct fib_config cfg
;
726 struct fib_table
*tb
;
729 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
, extack
);
733 tb
= fib_get_table(net
, cfg
.fc_table
);
735 NL_SET_ERR_MSG(extack
, "FIB table does not exist");
740 err
= fib_table_delete(net
, tb
, &cfg
, extack
);
745 static int inet_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
746 struct netlink_ext_ack
*extack
)
748 struct net
*net
= sock_net(skb
->sk
);
749 struct fib_config cfg
;
750 struct fib_table
*tb
;
753 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
, extack
);
757 tb
= fib_new_table(net
, cfg
.fc_table
);
763 err
= fib_table_insert(net
, tb
, &cfg
, extack
);
768 static int inet_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
770 struct net
*net
= sock_net(skb
->sk
);
772 unsigned int e
= 0, s_e
;
773 struct fib_table
*tb
;
774 struct hlist_head
*head
;
777 if (nlmsg_len(cb
->nlh
) >= sizeof(struct rtmsg
) &&
778 ((struct rtmsg
*) nlmsg_data(cb
->nlh
))->rtm_flags
& RTM_F_CLONED
)
786 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
788 head
= &net
->ipv4
.fib_table_hash
[h
];
789 hlist_for_each_entry_rcu(tb
, head
, tb_hlist
) {
793 memset(&cb
->args
[2], 0, sizeof(cb
->args
) -
794 2 * sizeof(cb
->args
[0]));
795 err
= fib_table_dump(tb
, skb
, cb
);
797 if (likely(skb
->len
))
818 /* Prepare and feed intra-kernel routing request.
819 * Really, it should be netlink message, but :-( netlink
820 * can be not configured, so that we feed it directly
821 * to fib engine. It is legal, because all events occur
822 * only when netlink is already locked.
824 static void fib_magic(int cmd
, int type
, __be32 dst
, int dst_len
, struct in_ifaddr
*ifa
)
826 struct net
*net
= dev_net(ifa
->ifa_dev
->dev
);
827 u32 tb_id
= l3mdev_fib_table(ifa
->ifa_dev
->dev
);
828 struct fib_table
*tb
;
829 struct fib_config cfg
= {
830 .fc_protocol
= RTPROT_KERNEL
,
833 .fc_dst_len
= dst_len
,
834 .fc_prefsrc
= ifa
->ifa_local
,
835 .fc_oif
= ifa
->ifa_dev
->dev
->ifindex
,
836 .fc_nlflags
= NLM_F_CREATE
| NLM_F_APPEND
,
843 tb_id
= (type
== RTN_UNICAST
) ? RT_TABLE_MAIN
: RT_TABLE_LOCAL
;
845 tb
= fib_new_table(net
, tb_id
);
849 cfg
.fc_table
= tb
->tb_id
;
851 if (type
!= RTN_LOCAL
)
852 cfg
.fc_scope
= RT_SCOPE_LINK
;
854 cfg
.fc_scope
= RT_SCOPE_HOST
;
856 if (cmd
== RTM_NEWROUTE
)
857 fib_table_insert(net
, tb
, &cfg
, NULL
);
859 fib_table_delete(net
, tb
, &cfg
, NULL
);
862 void fib_add_ifaddr(struct in_ifaddr
*ifa
)
864 struct in_device
*in_dev
= ifa
->ifa_dev
;
865 struct net_device
*dev
= in_dev
->dev
;
866 struct in_ifaddr
*prim
= ifa
;
867 __be32 mask
= ifa
->ifa_mask
;
868 __be32 addr
= ifa
->ifa_local
;
869 __be32 prefix
= ifa
->ifa_address
& mask
;
871 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
872 prim
= inet_ifa_byprefix(in_dev
, prefix
, mask
);
874 pr_warn("%s: bug: prim == NULL\n", __func__
);
879 fib_magic(RTM_NEWROUTE
, RTN_LOCAL
, addr
, 32, prim
);
881 if (!(dev
->flags
& IFF_UP
))
884 /* Add broadcast address, if it is explicitly assigned. */
885 if (ifa
->ifa_broadcast
&& ifa
->ifa_broadcast
!= htonl(0xFFFFFFFF))
886 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
888 if (!ipv4_is_zeronet(prefix
) && !(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
889 (prefix
!= addr
|| ifa
->ifa_prefixlen
< 32)) {
890 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
891 fib_magic(RTM_NEWROUTE
,
892 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
893 prefix
, ifa
->ifa_prefixlen
, prim
);
895 /* Add network specific broadcasts, when it takes a sense */
896 if (ifa
->ifa_prefixlen
< 31) {
897 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
, 32, prim
);
898 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
| ~mask
,
904 /* Delete primary or secondary address.
905 * Optionally, on secondary address promotion consider the addresses
906 * from subnet iprim as deleted, even if they are in device list.
907 * In this case the secondary ifa can be in device list.
909 void fib_del_ifaddr(struct in_ifaddr
*ifa
, struct in_ifaddr
*iprim
)
911 struct in_device
*in_dev
= ifa
->ifa_dev
;
912 struct net_device
*dev
= in_dev
->dev
;
913 struct in_ifaddr
*ifa1
;
914 struct in_ifaddr
*prim
= ifa
, *prim1
= NULL
;
915 __be32 brd
= ifa
->ifa_address
| ~ifa
->ifa_mask
;
916 __be32 any
= ifa
->ifa_address
& ifa
->ifa_mask
;
922 int subnet
= 0; /* Primary network */
923 int gone
= 1; /* Address is missing */
924 int same_prefsrc
= 0; /* Another primary with same IP */
926 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
927 prim
= inet_ifa_byprefix(in_dev
, any
, ifa
->ifa_mask
);
929 /* if the device has been deleted, we don't perform
933 pr_warn("%s: bug: prim == NULL\n", __func__
);
936 if (iprim
&& iprim
!= prim
) {
937 pr_warn("%s: bug: iprim != prim\n", __func__
);
940 } else if (!ipv4_is_zeronet(any
) &&
941 (any
!= ifa
->ifa_local
|| ifa
->ifa_prefixlen
< 32)) {
942 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
943 fib_magic(RTM_DELROUTE
,
944 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
945 any
, ifa
->ifa_prefixlen
, prim
);
952 /* Deletion is more complicated than add.
953 * We should take care of not to delete too much :-)
955 * Scan address list to be sure that addresses are really gone.
958 for (ifa1
= in_dev
->ifa_list
; ifa1
; ifa1
= ifa1
->ifa_next
) {
960 /* promotion, keep the IP */
964 /* Ignore IFAs from our subnet */
965 if (iprim
&& ifa1
->ifa_mask
== iprim
->ifa_mask
&&
966 inet_ifa_match(ifa1
->ifa_address
, iprim
))
969 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
970 if (ifa1
->ifa_flags
& IFA_F_SECONDARY
) {
971 /* Another address from our subnet? */
972 if (ifa1
->ifa_mask
== prim
->ifa_mask
&&
973 inet_ifa_match(ifa1
->ifa_address
, prim
))
976 /* We reached the secondaries, so
977 * same_prefsrc should be determined.
981 /* Search new prim1 if ifa1 is not
982 * using the current prim1
985 ifa1
->ifa_mask
!= prim1
->ifa_mask
||
986 !inet_ifa_match(ifa1
->ifa_address
, prim1
))
987 prim1
= inet_ifa_byprefix(in_dev
,
992 if (prim1
->ifa_local
!= prim
->ifa_local
)
996 if (prim
->ifa_local
!= ifa1
->ifa_local
)
1002 if (ifa
->ifa_local
== ifa1
->ifa_local
)
1004 if (ifa
->ifa_broadcast
== ifa1
->ifa_broadcast
)
1006 if (brd
== ifa1
->ifa_broadcast
)
1008 if (any
== ifa1
->ifa_broadcast
)
1010 /* primary has network specific broadcasts */
1011 if (prim1
== ifa1
&& ifa1
->ifa_prefixlen
< 31) {
1012 __be32 brd1
= ifa1
->ifa_address
| ~ifa1
->ifa_mask
;
1013 __be32 any1
= ifa1
->ifa_address
& ifa1
->ifa_mask
;
1015 if (!ipv4_is_zeronet(any1
)) {
1016 if (ifa
->ifa_broadcast
== brd1
||
1017 ifa
->ifa_broadcast
== any1
)
1019 if (brd
== brd1
|| brd
== any1
)
1021 if (any
== brd1
|| any
== any1
)
1029 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
1030 if (subnet
&& ifa
->ifa_prefixlen
< 31) {
1031 if (!(ok
& BRD1_OK
))
1032 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, brd
, 32, prim
);
1033 if (!(ok
& BRD0_OK
))
1034 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, any
, 32, prim
);
1036 if (!(ok
& LOCAL_OK
)) {
1037 unsigned int addr_type
;
1039 fib_magic(RTM_DELROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 32, prim
);
1041 /* Check, that this local address finally disappeared. */
1042 addr_type
= inet_addr_type_dev_table(dev_net(dev
), dev
,
1044 if (gone
&& addr_type
!= RTN_LOCAL
) {
1045 /* And the last, but not the least thing.
1046 * We must flush stray FIB entries.
1048 * First of all, we scan fib_info list searching
1049 * for stray nexthop entries, then ignite fib_flush.
1051 if (fib_sync_down_addr(dev
, ifa
->ifa_local
))
1052 fib_flush(dev_net(dev
));
1061 static void nl_fib_lookup(struct net
*net
, struct fib_result_nl
*frn
)
1064 struct fib_result res
;
1065 struct flowi4 fl4
= {
1066 .flowi4_mark
= frn
->fl_mark
,
1067 .daddr
= frn
->fl_addr
,
1068 .flowi4_tos
= frn
->fl_tos
,
1069 .flowi4_scope
= frn
->fl_scope
,
1071 struct fib_table
*tb
;
1075 tb
= fib_get_table(net
, frn
->tb_id_in
);
1081 frn
->tb_id
= tb
->tb_id
;
1082 frn
->err
= fib_table_lookup(tb
, &fl4
, &res
, FIB_LOOKUP_NOREF
);
1085 frn
->prefixlen
= res
.prefixlen
;
1086 frn
->nh_sel
= res
.nh_sel
;
1087 frn
->type
= res
.type
;
1088 frn
->scope
= res
.scope
;
1096 static void nl_fib_input(struct sk_buff
*skb
)
1099 struct fib_result_nl
*frn
;
1100 struct nlmsghdr
*nlh
;
1103 net
= sock_net(skb
->sk
);
1104 nlh
= nlmsg_hdr(skb
);
1105 if (skb
->len
< nlmsg_total_size(sizeof(*frn
)) ||
1106 skb
->len
< nlh
->nlmsg_len
||
1107 nlmsg_len(nlh
) < sizeof(*frn
))
1110 skb
= netlink_skb_clone(skb
, GFP_KERNEL
);
1113 nlh
= nlmsg_hdr(skb
);
1115 frn
= (struct fib_result_nl
*) nlmsg_data(nlh
);
1116 nl_fib_lookup(net
, frn
);
1118 portid
= NETLINK_CB(skb
).portid
; /* netlink portid */
1119 NETLINK_CB(skb
).portid
= 0; /* from kernel */
1120 NETLINK_CB(skb
).dst_group
= 0; /* unicast */
1121 netlink_unicast(net
->ipv4
.fibnl
, skb
, portid
, MSG_DONTWAIT
);
1124 static int __net_init
nl_fib_lookup_init(struct net
*net
)
1127 struct netlink_kernel_cfg cfg
= {
1128 .input
= nl_fib_input
,
1131 sk
= netlink_kernel_create(net
, NETLINK_FIB_LOOKUP
, &cfg
);
1133 return -EAFNOSUPPORT
;
1134 net
->ipv4
.fibnl
= sk
;
1138 static void nl_fib_lookup_exit(struct net
*net
)
1140 netlink_kernel_release(net
->ipv4
.fibnl
);
1141 net
->ipv4
.fibnl
= NULL
;
1144 static void fib_disable_ip(struct net_device
*dev
, unsigned long event
,
1147 if (fib_sync_down_dev(dev
, event
, force
))
1148 fib_flush(dev_net(dev
));
1150 rt_cache_flush(dev_net(dev
));
1154 static int fib_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1156 struct in_ifaddr
*ifa
= (struct in_ifaddr
*)ptr
;
1157 struct net_device
*dev
= ifa
->ifa_dev
->dev
;
1158 struct net
*net
= dev_net(dev
);
1162 fib_add_ifaddr(ifa
);
1163 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1164 fib_sync_up(dev
, RTNH_F_DEAD
);
1166 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1167 rt_cache_flush(dev_net(dev
));
1170 fib_del_ifaddr(ifa
, NULL
);
1171 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1172 if (!ifa
->ifa_dev
->ifa_list
) {
1173 /* Last address was deleted from this interface.
1176 fib_disable_ip(dev
, event
, true);
1178 rt_cache_flush(dev_net(dev
));
1185 static int fib_netdev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1187 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1188 struct netdev_notifier_changeupper_info
*upper_info
= ptr
;
1189 struct netdev_notifier_info_ext
*info_ext
= ptr
;
1190 struct in_device
*in_dev
;
1191 struct net
*net
= dev_net(dev
);
1194 if (event
== NETDEV_UNREGISTER
) {
1195 fib_disable_ip(dev
, event
, true);
1200 in_dev
= __in_dev_get_rtnl(dev
);
1207 fib_add_ifaddr(ifa
);
1208 } endfor_ifa(in_dev
);
1209 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1210 fib_sync_up(dev
, RTNH_F_DEAD
);
1212 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1213 rt_cache_flush(net
);
1216 fib_disable_ip(dev
, event
, false);
1219 flags
= dev_get_flags(dev
);
1220 if (flags
& (IFF_RUNNING
| IFF_LOWER_UP
))
1221 fib_sync_up(dev
, RTNH_F_LINKDOWN
);
1223 fib_sync_down_dev(dev
, event
, false);
1224 rt_cache_flush(net
);
1226 case NETDEV_CHANGEMTU
:
1227 fib_sync_mtu(dev
, info_ext
->ext
.mtu
);
1228 rt_cache_flush(net
);
1230 case NETDEV_CHANGEUPPER
:
1232 /* flush all routes if dev is linked to or unlinked from
1233 * an L3 master device (e.g., VRF)
1235 if (upper_info
->upper_dev
&&
1236 netif_is_l3_master(upper_info
->upper_dev
))
1237 fib_disable_ip(dev
, NETDEV_DOWN
, true);
1243 static struct notifier_block fib_inetaddr_notifier
= {
1244 .notifier_call
= fib_inetaddr_event
,
1247 static struct notifier_block fib_netdev_notifier
= {
1248 .notifier_call
= fib_netdev_event
,
1251 static int __net_init
ip_fib_net_init(struct net
*net
)
1254 size_t size
= sizeof(struct hlist_head
) * FIB_TABLE_HASHSZ
;
1256 err
= fib4_notifier_init(net
);
1260 /* Avoid false sharing : Use at least a full cache line */
1261 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1263 net
->ipv4
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1264 if (!net
->ipv4
.fib_table_hash
) {
1266 goto err_table_hash_alloc
;
1269 err
= fib4_rules_init(net
);
1271 goto err_rules_init
;
1275 kfree(net
->ipv4
.fib_table_hash
);
1276 err_table_hash_alloc
:
1277 fib4_notifier_exit(net
);
1281 static void ip_fib_net_exit(struct net
*net
)
1286 #ifdef CONFIG_IP_MULTIPLE_TABLES
1287 RCU_INIT_POINTER(net
->ipv4
.fib_main
, NULL
);
1288 RCU_INIT_POINTER(net
->ipv4
.fib_default
, NULL
);
1290 /* Destroy the tables in reverse order to guarantee that the
1291 * local table, ID 255, is destroyed before the main table, ID
1292 * 254. This is necessary as the local table may contain
1293 * references to data contained in the main table.
1295 for (i
= FIB_TABLE_HASHSZ
- 1; i
>= 0; i
--) {
1296 struct hlist_head
*head
= &net
->ipv4
.fib_table_hash
[i
];
1297 struct hlist_node
*tmp
;
1298 struct fib_table
*tb
;
1300 hlist_for_each_entry_safe(tb
, tmp
, head
, tb_hlist
) {
1301 hlist_del(&tb
->tb_hlist
);
1302 fib_table_flush(net
, tb
, true);
1307 #ifdef CONFIG_IP_MULTIPLE_TABLES
1308 fib4_rules_exit(net
);
1311 kfree(net
->ipv4
.fib_table_hash
);
1312 fib4_notifier_exit(net
);
1315 static int __net_init
fib_net_init(struct net
*net
)
1319 #ifdef CONFIG_IP_ROUTE_CLASSID
1320 net
->ipv4
.fib_num_tclassid_users
= 0;
1322 error
= ip_fib_net_init(net
);
1325 error
= nl_fib_lookup_init(net
);
1328 error
= fib_proc_init(net
);
1335 nl_fib_lookup_exit(net
);
1337 ip_fib_net_exit(net
);
1341 static void __net_exit
fib_net_exit(struct net
*net
)
1344 nl_fib_lookup_exit(net
);
1345 ip_fib_net_exit(net
);
1348 static struct pernet_operations fib_net_ops
= {
1349 .init
= fib_net_init
,
1350 .exit
= fib_net_exit
,
1353 void __init
ip_fib_init(void)
1357 register_pernet_subsys(&fib_net_ops
);
1359 register_netdevice_notifier(&fib_netdev_notifier
);
1360 register_inetaddr_notifier(&fib_inetaddr_notifier
);
1362 rtnl_register(PF_INET
, RTM_NEWROUTE
, inet_rtm_newroute
, NULL
, 0);
1363 rtnl_register(PF_INET
, RTM_DELROUTE
, inet_rtm_delroute
, NULL
, 0);
1364 rtnl_register(PF_INET
, RTM_GETROUTE
, NULL
, inet_dump_fib
, 0);