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
);
77 static bool fib4_has_custom_rules(struct net
*net
)
83 struct fib_table
*fib_new_table(struct net
*net
, u32 id
)
85 struct fib_table
*tb
, *alias
= NULL
;
90 tb
= fib_get_table(net
, id
);
94 if (id
== RT_TABLE_LOCAL
&& !net
->ipv4
.fib_has_custom_rules
)
95 alias
= fib_new_table(net
, RT_TABLE_MAIN
);
97 tb
= fib_trie_table(id
, alias
);
103 rcu_assign_pointer(net
->ipv4
.fib_main
, tb
);
105 case RT_TABLE_DEFAULT
:
106 rcu_assign_pointer(net
->ipv4
.fib_default
, tb
);
112 h
= id
& (FIB_TABLE_HASHSZ
- 1);
113 hlist_add_head_rcu(&tb
->tb_hlist
, &net
->ipv4
.fib_table_hash
[h
]);
116 EXPORT_SYMBOL_GPL(fib_new_table
);
118 /* caller must hold either rtnl or rcu read lock */
119 struct fib_table
*fib_get_table(struct net
*net
, u32 id
)
121 struct fib_table
*tb
;
122 struct hlist_head
*head
;
127 h
= id
& (FIB_TABLE_HASHSZ
- 1);
129 head
= &net
->ipv4
.fib_table_hash
[h
];
130 hlist_for_each_entry_rcu(tb
, head
, tb_hlist
) {
137 static bool fib4_has_custom_rules(struct net
*net
)
139 return net
->ipv4
.fib_has_custom_rules
;
141 #endif /* CONFIG_IP_MULTIPLE_TABLES */
143 static void fib_replace_table(struct net
*net
, struct fib_table
*old
,
144 struct fib_table
*new)
146 #ifdef CONFIG_IP_MULTIPLE_TABLES
147 switch (new->tb_id
) {
149 rcu_assign_pointer(net
->ipv4
.fib_main
, new);
151 case RT_TABLE_DEFAULT
:
152 rcu_assign_pointer(net
->ipv4
.fib_default
, new);
159 /* replace the old table in the hlist */
160 hlist_replace_rcu(&old
->tb_hlist
, &new->tb_hlist
);
163 int fib_unmerge(struct net
*net
)
165 struct fib_table
*old
, *new, *main_table
;
167 /* attempt to fetch local table if it has been allocated */
168 old
= fib_get_table(net
, RT_TABLE_LOCAL
);
172 new = fib_trie_unmerge(old
);
176 /* table is already unmerged */
180 /* replace merged table with clean table */
181 fib_replace_table(net
, old
, new);
184 /* attempt to fetch main table if it has been allocated */
185 main_table
= fib_get_table(net
, RT_TABLE_MAIN
);
189 /* flush local entries from main table */
190 fib_table_flush_external(main_table
);
195 static void fib_flush(struct net
*net
)
200 for (h
= 0; h
< FIB_TABLE_HASHSZ
; h
++) {
201 struct hlist_head
*head
= &net
->ipv4
.fib_table_hash
[h
];
202 struct hlist_node
*tmp
;
203 struct fib_table
*tb
;
205 hlist_for_each_entry_safe(tb
, tmp
, head
, tb_hlist
)
206 flushed
+= fib_table_flush(net
, tb
);
214 * Find address type as if only "dev" was present in the system. If
215 * on_dev is NULL then all interfaces are taken into consideration.
217 static inline unsigned int __inet_dev_addr_type(struct net
*net
,
218 const struct net_device
*dev
,
219 __be32 addr
, u32 tb_id
)
221 struct flowi4 fl4
= { .daddr
= addr
};
222 struct fib_result res
;
223 unsigned int ret
= RTN_BROADCAST
;
224 struct fib_table
*table
;
226 if (ipv4_is_zeronet(addr
) || ipv4_is_lbcast(addr
))
227 return RTN_BROADCAST
;
228 if (ipv4_is_multicast(addr
))
229 return RTN_MULTICAST
;
233 table
= fib_get_table(net
, tb_id
);
236 if (!fib_table_lookup(table
, &fl4
, &res
, FIB_LOOKUP_NOREF
)) {
237 if (!dev
|| dev
== res
.fi
->fib_dev
)
246 unsigned int inet_addr_type_table(struct net
*net
, __be32 addr
, u32 tb_id
)
248 return __inet_dev_addr_type(net
, NULL
, addr
, tb_id
);
250 EXPORT_SYMBOL(inet_addr_type_table
);
252 unsigned int inet_addr_type(struct net
*net
, __be32 addr
)
254 return __inet_dev_addr_type(net
, NULL
, addr
, RT_TABLE_LOCAL
);
256 EXPORT_SYMBOL(inet_addr_type
);
258 unsigned int inet_dev_addr_type(struct net
*net
, const struct net_device
*dev
,
261 u32 rt_table
= l3mdev_fib_table(dev
) ? : RT_TABLE_LOCAL
;
263 return __inet_dev_addr_type(net
, dev
, addr
, rt_table
);
265 EXPORT_SYMBOL(inet_dev_addr_type
);
267 /* inet_addr_type with dev == NULL but using the table from a dev
268 * if one is associated
270 unsigned int inet_addr_type_dev_table(struct net
*net
,
271 const struct net_device
*dev
,
274 u32 rt_table
= l3mdev_fib_table(dev
) ? : RT_TABLE_LOCAL
;
276 return __inet_dev_addr_type(net
, NULL
, addr
, rt_table
);
278 EXPORT_SYMBOL(inet_addr_type_dev_table
);
280 __be32
fib_compute_spec_dst(struct sk_buff
*skb
)
282 struct net_device
*dev
= skb
->dev
;
283 struct in_device
*in_dev
;
284 struct fib_result res
;
289 rt
= skb_rtable(skb
);
290 if ((rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
| RTCF_LOCAL
)) ==
292 return ip_hdr(skb
)->daddr
;
294 in_dev
= __in_dev_get_rcu(dev
);
298 scope
= RT_SCOPE_UNIVERSE
;
299 if (!ipv4_is_zeronet(ip_hdr(skb
)->saddr
)) {
300 bool vmark
= in_dev
&& IN_DEV_SRC_VMARK(in_dev
);
301 struct flowi4 fl4
= {
302 .flowi4_iif
= LOOPBACK_IFINDEX
,
303 .flowi4_oif
= l3mdev_master_ifindex_rcu(dev
),
304 .daddr
= ip_hdr(skb
)->saddr
,
305 .flowi4_tos
= RT_TOS(ip_hdr(skb
)->tos
),
306 .flowi4_scope
= scope
,
307 .flowi4_mark
= vmark
? skb
->mark
: 0,
309 if (!fib_lookup(net
, &fl4
, &res
, 0))
310 return FIB_RES_PREFSRC(net
, res
);
312 scope
= RT_SCOPE_LINK
;
315 return inet_select_addr(dev
, ip_hdr(skb
)->saddr
, scope
);
318 /* Given (packet source, input interface) and optional (dst, oif, tos):
319 * - (main) check, that source is valid i.e. not broadcast or our local
321 * - figure out what "logical" interface this packet arrived
322 * and calculate "specific destination" address.
323 * - check, that packet arrived from expected physical interface.
324 * called with rcu_read_lock()
326 static int __fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
327 u8 tos
, int oif
, struct net_device
*dev
,
328 int rpf
, struct in_device
*idev
, u32
*itag
)
331 struct fib_result res
;
333 struct net
*net
= dev_net(dev
);
337 fl4
.flowi4_iif
= l3mdev_master_ifindex_rcu(dev
);
339 fl4
.flowi4_iif
= oif
? : LOOPBACK_IFINDEX
;
342 fl4
.flowi4_tos
= tos
;
343 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
344 fl4
.flowi4_tun_key
.tun_id
= 0;
345 fl4
.flowi4_flags
= 0;
346 fl4
.flowi4_uid
= sock_net_uid(net
, NULL
);
348 no_addr
= idev
->ifa_list
== NULL
;
350 fl4
.flowi4_mark
= IN_DEV_SRC_VMARK(idev
) ? skb
->mark
: 0;
352 trace_fib_validate_source(dev
, &fl4
);
354 if (fib_lookup(net
, &fl4
, &res
, 0))
356 if (res
.type
!= RTN_UNICAST
&&
357 (res
.type
!= RTN_LOCAL
|| !IN_DEV_ACCEPT_LOCAL(idev
)))
359 fib_combine_itag(itag
, &res
);
362 #ifdef CONFIG_IP_ROUTE_MULTIPATH
363 for (ret
= 0; ret
< res
.fi
->fib_nhs
; ret
++) {
364 struct fib_nh
*nh
= &res
.fi
->fib_nh
[ret
];
366 if (nh
->nh_dev
== dev
) {
369 } else if (l3mdev_master_ifindex_rcu(nh
->nh_dev
) == dev
->ifindex
) {
375 if (FIB_RES_DEV(res
) == dev
)
379 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
386 fl4
.flowi4_oif
= dev
->ifindex
;
389 if (fib_lookup(net
, &fl4
, &res
, FIB_LOOKUP_IGNORE_LINKSTATE
) == 0) {
390 if (res
.type
== RTN_UNICAST
)
391 ret
= FIB_RES_NH(res
).nh_scope
>= RT_SCOPE_HOST
;
407 /* Ignore rp_filter for packets protected by IPsec. */
408 int fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
409 u8 tos
, int oif
, struct net_device
*dev
,
410 struct in_device
*idev
, u32
*itag
)
412 int r
= secpath_exists(skb
) ? 0 : IN_DEV_RPFILTER(idev
);
413 struct net
*net
= dev_net(dev
);
415 if (!r
&& !fib_num_tclassid_users(net
) &&
416 (dev
->ifindex
!= oif
|| !IN_DEV_TX_REDIRECTS(idev
))) {
417 if (IN_DEV_ACCEPT_LOCAL(idev
))
419 /* with custom local routes in place, checking local addresses
420 * only will be too optimistic, with custom rules, checking
421 * local addresses only can be too strict, e.g. due to vrf
423 if (net
->ipv4
.fib_has_custom_local_routes
||
424 fib4_has_custom_rules(net
))
426 if (inet_lookup_ifaddr_rcu(net
, src
))
435 return __fib_validate_source(skb
, src
, dst
, tos
, oif
, dev
, r
, idev
, itag
);
438 static inline __be32
sk_extract_addr(struct sockaddr
*addr
)
440 return ((struct sockaddr_in
*) addr
)->sin_addr
.s_addr
;
443 static int put_rtax(struct nlattr
*mx
, int len
, int type
, u32 value
)
447 nla
= (struct nlattr
*) ((char *) mx
+ len
);
448 nla
->nla_type
= type
;
449 nla
->nla_len
= nla_attr_size(4);
450 *(u32
*) nla_data(nla
) = value
;
452 return len
+ nla_total_size(4);
455 static int rtentry_to_fib_config(struct net
*net
, int cmd
, struct rtentry
*rt
,
456 struct fib_config
*cfg
)
461 memset(cfg
, 0, sizeof(*cfg
));
462 cfg
->fc_nlinfo
.nl_net
= net
;
464 if (rt
->rt_dst
.sa_family
!= AF_INET
)
465 return -EAFNOSUPPORT
;
468 * Check mask for validity:
469 * a) it must be contiguous.
470 * b) destination must have all host bits clear.
471 * c) if application forgot to set correct family (AF_INET),
472 * reject request unless it is absolutely clear i.e.
473 * both family and mask are zero.
476 addr
= sk_extract_addr(&rt
->rt_dst
);
477 if (!(rt
->rt_flags
& RTF_HOST
)) {
478 __be32 mask
= sk_extract_addr(&rt
->rt_genmask
);
480 if (rt
->rt_genmask
.sa_family
!= AF_INET
) {
481 if (mask
|| rt
->rt_genmask
.sa_family
)
482 return -EAFNOSUPPORT
;
485 if (bad_mask(mask
, addr
))
488 plen
= inet_mask_len(mask
);
491 cfg
->fc_dst_len
= plen
;
494 if (cmd
!= SIOCDELRT
) {
495 cfg
->fc_nlflags
= NLM_F_CREATE
;
496 cfg
->fc_protocol
= RTPROT_BOOT
;
500 cfg
->fc_priority
= rt
->rt_metric
- 1;
502 if (rt
->rt_flags
& RTF_REJECT
) {
503 cfg
->fc_scope
= RT_SCOPE_HOST
;
504 cfg
->fc_type
= RTN_UNREACHABLE
;
508 cfg
->fc_scope
= RT_SCOPE_NOWHERE
;
509 cfg
->fc_type
= RTN_UNICAST
;
513 struct net_device
*dev
;
514 char devname
[IFNAMSIZ
];
516 if (copy_from_user(devname
, rt
->rt_dev
, IFNAMSIZ
-1))
519 devname
[IFNAMSIZ
-1] = 0;
520 colon
= strchr(devname
, ':');
523 dev
= __dev_get_by_name(net
, devname
);
526 cfg
->fc_oif
= dev
->ifindex
;
527 cfg
->fc_table
= l3mdev_fib_table(dev
);
529 struct in_ifaddr
*ifa
;
530 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
534 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
)
535 if (strcmp(ifa
->ifa_label
, devname
) == 0)
539 cfg
->fc_prefsrc
= ifa
->ifa_local
;
543 addr
= sk_extract_addr(&rt
->rt_gateway
);
544 if (rt
->rt_gateway
.sa_family
== AF_INET
&& addr
) {
545 unsigned int addr_type
;
548 addr_type
= inet_addr_type_table(net
, addr
, cfg
->fc_table
);
549 if (rt
->rt_flags
& RTF_GATEWAY
&&
550 addr_type
== RTN_UNICAST
)
551 cfg
->fc_scope
= RT_SCOPE_UNIVERSE
;
554 if (cmd
== SIOCDELRT
)
557 if (rt
->rt_flags
& RTF_GATEWAY
&& !cfg
->fc_gw
)
560 if (cfg
->fc_scope
== RT_SCOPE_NOWHERE
)
561 cfg
->fc_scope
= RT_SCOPE_LINK
;
563 if (rt
->rt_flags
& (RTF_MTU
| RTF_WINDOW
| RTF_IRTT
)) {
567 mx
= kzalloc(3 * nla_total_size(4), GFP_KERNEL
);
571 if (rt
->rt_flags
& RTF_MTU
)
572 len
= put_rtax(mx
, len
, RTAX_ADVMSS
, rt
->rt_mtu
- 40);
574 if (rt
->rt_flags
& RTF_WINDOW
)
575 len
= put_rtax(mx
, len
, RTAX_WINDOW
, rt
->rt_window
);
577 if (rt
->rt_flags
& RTF_IRTT
)
578 len
= put_rtax(mx
, len
, RTAX_RTT
, rt
->rt_irtt
<< 3);
581 cfg
->fc_mx_len
= len
;
588 * Handle IP routing ioctl calls.
589 * These are used to manipulate the routing tables
591 int ip_rt_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
593 struct fib_config cfg
;
598 case SIOCADDRT
: /* Add a route */
599 case SIOCDELRT
: /* Delete a route */
600 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
603 if (copy_from_user(&rt
, arg
, sizeof(rt
)))
607 err
= rtentry_to_fib_config(net
, cmd
, &rt
, &cfg
);
609 struct fib_table
*tb
;
611 if (cmd
== SIOCDELRT
) {
612 tb
= fib_get_table(net
, cfg
.fc_table
);
614 err
= fib_table_delete(net
, tb
, &cfg
,
619 tb
= fib_new_table(net
, cfg
.fc_table
);
621 err
= fib_table_insert(net
, tb
,
627 /* allocated by rtentry_to_fib_config() */
636 const struct nla_policy rtm_ipv4_policy
[RTA_MAX
+ 1] = {
637 [RTA_DST
] = { .type
= NLA_U32
},
638 [RTA_SRC
] = { .type
= NLA_U32
},
639 [RTA_IIF
] = { .type
= NLA_U32
},
640 [RTA_OIF
] = { .type
= NLA_U32
},
641 [RTA_GATEWAY
] = { .type
= NLA_U32
},
642 [RTA_PRIORITY
] = { .type
= NLA_U32
},
643 [RTA_PREFSRC
] = { .type
= NLA_U32
},
644 [RTA_METRICS
] = { .type
= NLA_NESTED
},
645 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
646 [RTA_FLOW
] = { .type
= NLA_U32
},
647 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
648 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
649 [RTA_UID
] = { .type
= NLA_U32
},
650 [RTA_MARK
] = { .type
= NLA_U32
},
651 [RTA_TABLE
] = { .type
= NLA_U32
},
654 static int rtm_to_fib_config(struct net
*net
, struct sk_buff
*skb
,
655 struct nlmsghdr
*nlh
, struct fib_config
*cfg
,
656 struct netlink_ext_ack
*extack
)
662 err
= nlmsg_validate(nlh
, sizeof(*rtm
), RTA_MAX
, rtm_ipv4_policy
,
667 memset(cfg
, 0, sizeof(*cfg
));
669 rtm
= nlmsg_data(nlh
);
670 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
671 cfg
->fc_tos
= rtm
->rtm_tos
;
672 cfg
->fc_table
= rtm
->rtm_table
;
673 cfg
->fc_protocol
= rtm
->rtm_protocol
;
674 cfg
->fc_scope
= rtm
->rtm_scope
;
675 cfg
->fc_type
= rtm
->rtm_type
;
676 cfg
->fc_flags
= rtm
->rtm_flags
;
677 cfg
->fc_nlflags
= nlh
->nlmsg_flags
;
679 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
680 cfg
->fc_nlinfo
.nlh
= nlh
;
681 cfg
->fc_nlinfo
.nl_net
= net
;
683 if (cfg
->fc_type
> RTN_MAX
) {
684 NL_SET_ERR_MSG(extack
, "Invalid route type");
689 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct rtmsg
), remaining
) {
690 switch (nla_type(attr
)) {
692 cfg
->fc_dst
= nla_get_be32(attr
);
695 cfg
->fc_oif
= nla_get_u32(attr
);
698 cfg
->fc_gw
= nla_get_be32(attr
);
701 cfg
->fc_priority
= nla_get_u32(attr
);
704 cfg
->fc_prefsrc
= nla_get_be32(attr
);
707 cfg
->fc_mx
= nla_data(attr
);
708 cfg
->fc_mx_len
= nla_len(attr
);
711 err
= lwtunnel_valid_encap_type_attr(nla_data(attr
),
716 cfg
->fc_mp
= nla_data(attr
);
717 cfg
->fc_mp_len
= nla_len(attr
);
720 cfg
->fc_flow
= nla_get_u32(attr
);
723 cfg
->fc_table
= nla_get_u32(attr
);
726 cfg
->fc_encap
= attr
;
729 cfg
->fc_encap_type
= nla_get_u16(attr
);
730 err
= lwtunnel_valid_encap_type(cfg
->fc_encap_type
,
743 static int inet_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
744 struct netlink_ext_ack
*extack
)
746 struct net
*net
= sock_net(skb
->sk
);
747 struct fib_config cfg
;
748 struct fib_table
*tb
;
751 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
, extack
);
755 tb
= fib_get_table(net
, cfg
.fc_table
);
757 NL_SET_ERR_MSG(extack
, "FIB table does not exist");
762 err
= fib_table_delete(net
, tb
, &cfg
, extack
);
767 static int inet_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
768 struct netlink_ext_ack
*extack
)
770 struct net
*net
= sock_net(skb
->sk
);
771 struct fib_config cfg
;
772 struct fib_table
*tb
;
775 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
, extack
);
779 tb
= fib_new_table(net
, cfg
.fc_table
);
785 err
= fib_table_insert(net
, tb
, &cfg
, extack
);
786 if (!err
&& cfg
.fc_type
== RTN_LOCAL
)
787 net
->ipv4
.fib_has_custom_local_routes
= true;
792 static int inet_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
794 struct net
*net
= sock_net(skb
->sk
);
796 unsigned int e
= 0, s_e
;
797 struct fib_table
*tb
;
798 struct hlist_head
*head
;
801 if (nlmsg_len(cb
->nlh
) >= sizeof(struct rtmsg
) &&
802 ((struct rtmsg
*) nlmsg_data(cb
->nlh
))->rtm_flags
& RTM_F_CLONED
)
810 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
812 head
= &net
->ipv4
.fib_table_hash
[h
];
813 hlist_for_each_entry_rcu(tb
, head
, tb_hlist
) {
817 memset(&cb
->args
[2], 0, sizeof(cb
->args
) -
818 2 * sizeof(cb
->args
[0]));
819 err
= fib_table_dump(tb
, skb
, cb
);
821 if (likely(skb
->len
))
842 /* Prepare and feed intra-kernel routing request.
843 * Really, it should be netlink message, but :-( netlink
844 * can be not configured, so that we feed it directly
845 * to fib engine. It is legal, because all events occur
846 * only when netlink is already locked.
848 static void fib_magic(int cmd
, int type
, __be32 dst
, int dst_len
, struct in_ifaddr
*ifa
)
850 struct net
*net
= dev_net(ifa
->ifa_dev
->dev
);
851 u32 tb_id
= l3mdev_fib_table(ifa
->ifa_dev
->dev
);
852 struct fib_table
*tb
;
853 struct fib_config cfg
= {
854 .fc_protocol
= RTPROT_KERNEL
,
857 .fc_dst_len
= dst_len
,
858 .fc_prefsrc
= ifa
->ifa_local
,
859 .fc_oif
= ifa
->ifa_dev
->dev
->ifindex
,
860 .fc_nlflags
= NLM_F_CREATE
| NLM_F_APPEND
,
867 tb_id
= (type
== RTN_UNICAST
) ? RT_TABLE_MAIN
: RT_TABLE_LOCAL
;
869 tb
= fib_new_table(net
, tb_id
);
873 cfg
.fc_table
= tb
->tb_id
;
875 if (type
!= RTN_LOCAL
)
876 cfg
.fc_scope
= RT_SCOPE_LINK
;
878 cfg
.fc_scope
= RT_SCOPE_HOST
;
880 if (cmd
== RTM_NEWROUTE
)
881 fib_table_insert(net
, tb
, &cfg
, NULL
);
883 fib_table_delete(net
, tb
, &cfg
, NULL
);
886 void fib_add_ifaddr(struct in_ifaddr
*ifa
)
888 struct in_device
*in_dev
= ifa
->ifa_dev
;
889 struct net_device
*dev
= in_dev
->dev
;
890 struct in_ifaddr
*prim
= ifa
;
891 __be32 mask
= ifa
->ifa_mask
;
892 __be32 addr
= ifa
->ifa_local
;
893 __be32 prefix
= ifa
->ifa_address
& mask
;
895 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
896 prim
= inet_ifa_byprefix(in_dev
, prefix
, mask
);
898 pr_warn("%s: bug: prim == NULL\n", __func__
);
903 fib_magic(RTM_NEWROUTE
, RTN_LOCAL
, addr
, 32, prim
);
905 if (!(dev
->flags
& IFF_UP
))
908 /* Add broadcast address, if it is explicitly assigned. */
909 if (ifa
->ifa_broadcast
&& ifa
->ifa_broadcast
!= htonl(0xFFFFFFFF))
910 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
912 if (!ipv4_is_zeronet(prefix
) && !(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
913 (prefix
!= addr
|| ifa
->ifa_prefixlen
< 32)) {
914 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
915 fib_magic(RTM_NEWROUTE
,
916 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
917 prefix
, ifa
->ifa_prefixlen
, prim
);
919 /* Add network specific broadcasts, when it takes a sense */
920 if (ifa
->ifa_prefixlen
< 31) {
921 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
, 32, prim
);
922 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
| ~mask
,
928 /* Delete primary or secondary address.
929 * Optionally, on secondary address promotion consider the addresses
930 * from subnet iprim as deleted, even if they are in device list.
931 * In this case the secondary ifa can be in device list.
933 void fib_del_ifaddr(struct in_ifaddr
*ifa
, struct in_ifaddr
*iprim
)
935 struct in_device
*in_dev
= ifa
->ifa_dev
;
936 struct net_device
*dev
= in_dev
->dev
;
937 struct in_ifaddr
*ifa1
;
938 struct in_ifaddr
*prim
= ifa
, *prim1
= NULL
;
939 __be32 brd
= ifa
->ifa_address
| ~ifa
->ifa_mask
;
940 __be32 any
= ifa
->ifa_address
& ifa
->ifa_mask
;
946 int subnet
= 0; /* Primary network */
947 int gone
= 1; /* Address is missing */
948 int same_prefsrc
= 0; /* Another primary with same IP */
950 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
951 prim
= inet_ifa_byprefix(in_dev
, any
, ifa
->ifa_mask
);
953 /* if the device has been deleted, we don't perform
957 pr_warn("%s: bug: prim == NULL\n", __func__
);
960 if (iprim
&& iprim
!= prim
) {
961 pr_warn("%s: bug: iprim != prim\n", __func__
);
964 } else if (!ipv4_is_zeronet(any
) &&
965 (any
!= ifa
->ifa_local
|| ifa
->ifa_prefixlen
< 32)) {
966 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
967 fib_magic(RTM_DELROUTE
,
968 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
969 any
, ifa
->ifa_prefixlen
, prim
);
976 /* Deletion is more complicated than add.
977 * We should take care of not to delete too much :-)
979 * Scan address list to be sure that addresses are really gone.
982 for (ifa1
= in_dev
->ifa_list
; ifa1
; ifa1
= ifa1
->ifa_next
) {
984 /* promotion, keep the IP */
988 /* Ignore IFAs from our subnet */
989 if (iprim
&& ifa1
->ifa_mask
== iprim
->ifa_mask
&&
990 inet_ifa_match(ifa1
->ifa_address
, iprim
))
993 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
994 if (ifa1
->ifa_flags
& IFA_F_SECONDARY
) {
995 /* Another address from our subnet? */
996 if (ifa1
->ifa_mask
== prim
->ifa_mask
&&
997 inet_ifa_match(ifa1
->ifa_address
, prim
))
1000 /* We reached the secondaries, so
1001 * same_prefsrc should be determined.
1005 /* Search new prim1 if ifa1 is not
1006 * using the current prim1
1009 ifa1
->ifa_mask
!= prim1
->ifa_mask
||
1010 !inet_ifa_match(ifa1
->ifa_address
, prim1
))
1011 prim1
= inet_ifa_byprefix(in_dev
,
1016 if (prim1
->ifa_local
!= prim
->ifa_local
)
1020 if (prim
->ifa_local
!= ifa1
->ifa_local
)
1026 if (ifa
->ifa_local
== ifa1
->ifa_local
)
1028 if (ifa
->ifa_broadcast
== ifa1
->ifa_broadcast
)
1030 if (brd
== ifa1
->ifa_broadcast
)
1032 if (any
== ifa1
->ifa_broadcast
)
1034 /* primary has network specific broadcasts */
1035 if (prim1
== ifa1
&& ifa1
->ifa_prefixlen
< 31) {
1036 __be32 brd1
= ifa1
->ifa_address
| ~ifa1
->ifa_mask
;
1037 __be32 any1
= ifa1
->ifa_address
& ifa1
->ifa_mask
;
1039 if (!ipv4_is_zeronet(any1
)) {
1040 if (ifa
->ifa_broadcast
== brd1
||
1041 ifa
->ifa_broadcast
== any1
)
1043 if (brd
== brd1
|| brd
== any1
)
1045 if (any
== brd1
|| any
== any1
)
1053 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32, prim
);
1054 if (subnet
&& ifa
->ifa_prefixlen
< 31) {
1055 if (!(ok
& BRD1_OK
))
1056 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, brd
, 32, prim
);
1057 if (!(ok
& BRD0_OK
))
1058 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, any
, 32, prim
);
1060 if (!(ok
& LOCAL_OK
)) {
1061 unsigned int addr_type
;
1063 fib_magic(RTM_DELROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 32, prim
);
1065 /* Check, that this local address finally disappeared. */
1066 addr_type
= inet_addr_type_dev_table(dev_net(dev
), dev
,
1068 if (gone
&& addr_type
!= RTN_LOCAL
) {
1069 /* And the last, but not the least thing.
1070 * We must flush stray FIB entries.
1072 * First of all, we scan fib_info list searching
1073 * for stray nexthop entries, then ignite fib_flush.
1075 if (fib_sync_down_addr(dev
, ifa
->ifa_local
))
1076 fib_flush(dev_net(dev
));
1085 static void nl_fib_lookup(struct net
*net
, struct fib_result_nl
*frn
)
1088 struct fib_result res
;
1089 struct flowi4 fl4
= {
1090 .flowi4_mark
= frn
->fl_mark
,
1091 .daddr
= frn
->fl_addr
,
1092 .flowi4_tos
= frn
->fl_tos
,
1093 .flowi4_scope
= frn
->fl_scope
,
1095 struct fib_table
*tb
;
1099 tb
= fib_get_table(net
, frn
->tb_id_in
);
1105 frn
->tb_id
= tb
->tb_id
;
1106 frn
->err
= fib_table_lookup(tb
, &fl4
, &res
, FIB_LOOKUP_NOREF
);
1109 frn
->prefixlen
= res
.prefixlen
;
1110 frn
->nh_sel
= res
.nh_sel
;
1111 frn
->type
= res
.type
;
1112 frn
->scope
= res
.scope
;
1120 static void nl_fib_input(struct sk_buff
*skb
)
1123 struct fib_result_nl
*frn
;
1124 struct nlmsghdr
*nlh
;
1127 net
= sock_net(skb
->sk
);
1128 nlh
= nlmsg_hdr(skb
);
1129 if (skb
->len
< nlmsg_total_size(sizeof(*frn
)) ||
1130 skb
->len
< nlh
->nlmsg_len
||
1131 nlmsg_len(nlh
) < sizeof(*frn
))
1134 skb
= netlink_skb_clone(skb
, GFP_KERNEL
);
1137 nlh
= nlmsg_hdr(skb
);
1139 frn
= (struct fib_result_nl
*) nlmsg_data(nlh
);
1140 nl_fib_lookup(net
, frn
);
1142 portid
= NETLINK_CB(skb
).portid
; /* netlink portid */
1143 NETLINK_CB(skb
).portid
= 0; /* from kernel */
1144 NETLINK_CB(skb
).dst_group
= 0; /* unicast */
1145 netlink_unicast(net
->ipv4
.fibnl
, skb
, portid
, MSG_DONTWAIT
);
1148 static int __net_init
nl_fib_lookup_init(struct net
*net
)
1151 struct netlink_kernel_cfg cfg
= {
1152 .input
= nl_fib_input
,
1155 sk
= netlink_kernel_create(net
, NETLINK_FIB_LOOKUP
, &cfg
);
1157 return -EAFNOSUPPORT
;
1158 net
->ipv4
.fibnl
= sk
;
1162 static void nl_fib_lookup_exit(struct net
*net
)
1164 netlink_kernel_release(net
->ipv4
.fibnl
);
1165 net
->ipv4
.fibnl
= NULL
;
1168 static void fib_disable_ip(struct net_device
*dev
, unsigned long event
,
1171 if (fib_sync_down_dev(dev
, event
, force
))
1172 fib_flush(dev_net(dev
));
1174 rt_cache_flush(dev_net(dev
));
1178 static int fib_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1180 struct in_ifaddr
*ifa
= (struct in_ifaddr
*)ptr
;
1181 struct net_device
*dev
= ifa
->ifa_dev
->dev
;
1182 struct net
*net
= dev_net(dev
);
1186 fib_add_ifaddr(ifa
);
1187 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1188 fib_sync_up(dev
, RTNH_F_DEAD
);
1190 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1191 rt_cache_flush(dev_net(dev
));
1194 fib_del_ifaddr(ifa
, NULL
);
1195 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1196 if (!ifa
->ifa_dev
->ifa_list
) {
1197 /* Last address was deleted from this interface.
1200 fib_disable_ip(dev
, event
, true);
1202 rt_cache_flush(dev_net(dev
));
1209 static int fib_netdev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1211 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1212 struct netdev_notifier_changeupper_info
*info
;
1213 struct in_device
*in_dev
;
1214 struct net
*net
= dev_net(dev
);
1217 if (event
== NETDEV_UNREGISTER
) {
1218 fib_disable_ip(dev
, event
, true);
1223 in_dev
= __in_dev_get_rtnl(dev
);
1230 fib_add_ifaddr(ifa
);
1231 } endfor_ifa(in_dev
);
1232 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1233 fib_sync_up(dev
, RTNH_F_DEAD
);
1235 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1236 rt_cache_flush(net
);
1239 fib_disable_ip(dev
, event
, false);
1242 flags
= dev_get_flags(dev
);
1243 if (flags
& (IFF_RUNNING
| IFF_LOWER_UP
))
1244 fib_sync_up(dev
, RTNH_F_LINKDOWN
);
1246 fib_sync_down_dev(dev
, event
, false);
1248 case NETDEV_CHANGEMTU
:
1249 rt_cache_flush(net
);
1251 case NETDEV_CHANGEUPPER
:
1253 /* flush all routes if dev is linked to or unlinked from
1254 * an L3 master device (e.g., VRF)
1256 if (info
->upper_dev
&& netif_is_l3_master(info
->upper_dev
))
1257 fib_disable_ip(dev
, NETDEV_DOWN
, true);
1263 static struct notifier_block fib_inetaddr_notifier
= {
1264 .notifier_call
= fib_inetaddr_event
,
1267 static struct notifier_block fib_netdev_notifier
= {
1268 .notifier_call
= fib_netdev_event
,
1271 static int __net_init
ip_fib_net_init(struct net
*net
)
1274 size_t size
= sizeof(struct hlist_head
) * FIB_TABLE_HASHSZ
;
1276 err
= fib4_notifier_init(net
);
1280 /* Avoid false sharing : Use at least a full cache line */
1281 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1283 net
->ipv4
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1284 if (!net
->ipv4
.fib_table_hash
) {
1286 goto err_table_hash_alloc
;
1289 err
= fib4_rules_init(net
);
1291 goto err_rules_init
;
1295 kfree(net
->ipv4
.fib_table_hash
);
1296 err_table_hash_alloc
:
1297 fib4_notifier_exit(net
);
1301 static void ip_fib_net_exit(struct net
*net
)
1306 #ifdef CONFIG_IP_MULTIPLE_TABLES
1307 RCU_INIT_POINTER(net
->ipv4
.fib_main
, NULL
);
1308 RCU_INIT_POINTER(net
->ipv4
.fib_default
, NULL
);
1310 /* Destroy the tables in reverse order to guarantee that the
1311 * local table, ID 255, is destroyed before the main table, ID
1312 * 254. This is necessary as the local table may contain
1313 * references to data contained in the main table.
1315 for (i
= FIB_TABLE_HASHSZ
- 1; i
>= 0; i
--) {
1316 struct hlist_head
*head
= &net
->ipv4
.fib_table_hash
[i
];
1317 struct hlist_node
*tmp
;
1318 struct fib_table
*tb
;
1320 hlist_for_each_entry_safe(tb
, tmp
, head
, tb_hlist
) {
1321 hlist_del(&tb
->tb_hlist
);
1322 fib_table_flush(net
, tb
);
1327 #ifdef CONFIG_IP_MULTIPLE_TABLES
1328 fib4_rules_exit(net
);
1331 kfree(net
->ipv4
.fib_table_hash
);
1332 fib4_notifier_exit(net
);
1335 static int __net_init
fib_net_init(struct net
*net
)
1339 #ifdef CONFIG_IP_ROUTE_CLASSID
1340 net
->ipv4
.fib_num_tclassid_users
= 0;
1342 error
= ip_fib_net_init(net
);
1345 error
= nl_fib_lookup_init(net
);
1348 error
= fib_proc_init(net
);
1355 nl_fib_lookup_exit(net
);
1357 ip_fib_net_exit(net
);
1361 static void __net_exit
fib_net_exit(struct net
*net
)
1364 nl_fib_lookup_exit(net
);
1365 ip_fib_net_exit(net
);
1368 static struct pernet_operations fib_net_ops
= {
1369 .init
= fib_net_init
,
1370 .exit
= fib_net_exit
,
1373 void __init
ip_fib_init(void)
1377 register_pernet_subsys(&fib_net_ops
);
1379 register_netdevice_notifier(&fib_netdev_notifier
);
1380 register_inetaddr_notifier(&fib_inetaddr_notifier
);
1382 rtnl_register(PF_INET
, RTM_NEWROUTE
, inet_rtm_newroute
, NULL
, 0);
1383 rtnl_register(PF_INET
, RTM_DELROUTE
, inet_rtm_delroute
, NULL
, 0);
1384 rtnl_register(PF_INET
, RTM_GETROUTE
, NULL
, inet_dump_fib
, 0);