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
, false);
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_result_prefsrc(net
, &res
);
312 scope
= RT_SCOPE_LINK
;
315 return inet_select_addr(dev
, ip_hdr(skb
)->saddr
, scope
);
318 bool fib_info_nh_uses_dev(struct fib_info
*fi
, const struct net_device
*dev
)
320 bool dev_match
= false;
321 #ifdef CONFIG_IP_ROUTE_MULTIPATH
324 for (ret
= 0; ret
< fi
->fib_nhs
; ret
++) {
325 struct fib_nh
*nh
= &fi
->fib_nh
[ret
];
327 if (nh
->fib_nh_dev
== dev
) {
330 } else if (l3mdev_master_ifindex_rcu(nh
->fib_nh_dev
) == dev
->ifindex
) {
336 if (fi
->fib_nh
[0].fib_nh_dev
== dev
)
342 EXPORT_SYMBOL_GPL(fib_info_nh_uses_dev
);
344 /* Given (packet source, input interface) and optional (dst, oif, tos):
345 * - (main) check, that source is valid i.e. not broadcast or our local
347 * - figure out what "logical" interface this packet arrived
348 * and calculate "specific destination" address.
349 * - check, that packet arrived from expected physical interface.
350 * called with rcu_read_lock()
352 static int __fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
353 u8 tos
, int oif
, struct net_device
*dev
,
354 int rpf
, struct in_device
*idev
, u32
*itag
)
356 struct net
*net
= dev_net(dev
);
357 struct flow_keys flkeys
;
359 struct fib_result res
;
364 fl4
.flowi4_iif
= l3mdev_master_ifindex_rcu(dev
);
366 fl4
.flowi4_iif
= oif
? : LOOPBACK_IFINDEX
;
369 fl4
.flowi4_tos
= tos
;
370 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
371 fl4
.flowi4_tun_key
.tun_id
= 0;
372 fl4
.flowi4_flags
= 0;
373 fl4
.flowi4_uid
= sock_net_uid(net
, NULL
);
375 no_addr
= idev
->ifa_list
== NULL
;
377 fl4
.flowi4_mark
= IN_DEV_SRC_VMARK(idev
) ? skb
->mark
: 0;
378 if (!fib4_rules_early_flow_dissect(net
, skb
, &fl4
, &flkeys
)) {
379 fl4
.flowi4_proto
= 0;
384 if (fib_lookup(net
, &fl4
, &res
, 0))
386 if (res
.type
!= RTN_UNICAST
&&
387 (res
.type
!= RTN_LOCAL
|| !IN_DEV_ACCEPT_LOCAL(idev
)))
389 fib_combine_itag(itag
, &res
);
391 dev_match
= fib_info_nh_uses_dev(res
.fi
, dev
);
393 ret
= FIB_RES_NHC(res
)->nhc_scope
>= RT_SCOPE_HOST
;
400 fl4
.flowi4_oif
= dev
->ifindex
;
403 if (fib_lookup(net
, &fl4
, &res
, FIB_LOOKUP_IGNORE_LINKSTATE
) == 0) {
404 if (res
.type
== RTN_UNICAST
)
405 ret
= FIB_RES_NHC(res
)->nhc_scope
>= RT_SCOPE_HOST
;
421 /* Ignore rp_filter for packets protected by IPsec. */
422 int fib_validate_source(struct sk_buff
*skb
, __be32 src
, __be32 dst
,
423 u8 tos
, int oif
, struct net_device
*dev
,
424 struct in_device
*idev
, u32
*itag
)
426 int r
= secpath_exists(skb
) ? 0 : IN_DEV_RPFILTER(idev
);
427 struct net
*net
= dev_net(dev
);
429 if (!r
&& !fib_num_tclassid_users(net
) &&
430 (dev
->ifindex
!= oif
|| !IN_DEV_TX_REDIRECTS(idev
))) {
431 if (IN_DEV_ACCEPT_LOCAL(idev
))
433 /* with custom local routes in place, checking local addresses
434 * only will be too optimistic, with custom rules, checking
435 * local addresses only can be too strict, e.g. due to vrf
437 if (net
->ipv4
.fib_has_custom_local_routes
||
438 fib4_has_custom_rules(net
))
440 if (inet_lookup_ifaddr_rcu(net
, src
))
449 return __fib_validate_source(skb
, src
, dst
, tos
, oif
, dev
, r
, idev
, itag
);
452 static inline __be32
sk_extract_addr(struct sockaddr
*addr
)
454 return ((struct sockaddr_in
*) addr
)->sin_addr
.s_addr
;
457 static int put_rtax(struct nlattr
*mx
, int len
, int type
, u32 value
)
461 nla
= (struct nlattr
*) ((char *) mx
+ len
);
462 nla
->nla_type
= type
;
463 nla
->nla_len
= nla_attr_size(4);
464 *(u32
*) nla_data(nla
) = value
;
466 return len
+ nla_total_size(4);
469 static int rtentry_to_fib_config(struct net
*net
, int cmd
, struct rtentry
*rt
,
470 struct fib_config
*cfg
)
475 memset(cfg
, 0, sizeof(*cfg
));
476 cfg
->fc_nlinfo
.nl_net
= net
;
478 if (rt
->rt_dst
.sa_family
!= AF_INET
)
479 return -EAFNOSUPPORT
;
482 * Check mask for validity:
483 * a) it must be contiguous.
484 * b) destination must have all host bits clear.
485 * c) if application forgot to set correct family (AF_INET),
486 * reject request unless it is absolutely clear i.e.
487 * both family and mask are zero.
490 addr
= sk_extract_addr(&rt
->rt_dst
);
491 if (!(rt
->rt_flags
& RTF_HOST
)) {
492 __be32 mask
= sk_extract_addr(&rt
->rt_genmask
);
494 if (rt
->rt_genmask
.sa_family
!= AF_INET
) {
495 if (mask
|| rt
->rt_genmask
.sa_family
)
496 return -EAFNOSUPPORT
;
499 if (bad_mask(mask
, addr
))
502 plen
= inet_mask_len(mask
);
505 cfg
->fc_dst_len
= plen
;
508 if (cmd
!= SIOCDELRT
) {
509 cfg
->fc_nlflags
= NLM_F_CREATE
;
510 cfg
->fc_protocol
= RTPROT_BOOT
;
514 cfg
->fc_priority
= rt
->rt_metric
- 1;
516 if (rt
->rt_flags
& RTF_REJECT
) {
517 cfg
->fc_scope
= RT_SCOPE_HOST
;
518 cfg
->fc_type
= RTN_UNREACHABLE
;
522 cfg
->fc_scope
= RT_SCOPE_NOWHERE
;
523 cfg
->fc_type
= RTN_UNICAST
;
527 struct net_device
*dev
;
528 char devname
[IFNAMSIZ
];
530 if (copy_from_user(devname
, rt
->rt_dev
, IFNAMSIZ
-1))
533 devname
[IFNAMSIZ
-1] = 0;
534 colon
= strchr(devname
, ':');
537 dev
= __dev_get_by_name(net
, devname
);
540 cfg
->fc_oif
= dev
->ifindex
;
541 cfg
->fc_table
= l3mdev_fib_table(dev
);
543 struct in_ifaddr
*ifa
;
544 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
548 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
)
549 if (strcmp(ifa
->ifa_label
, devname
) == 0)
553 cfg
->fc_prefsrc
= ifa
->ifa_local
;
557 addr
= sk_extract_addr(&rt
->rt_gateway
);
558 if (rt
->rt_gateway
.sa_family
== AF_INET
&& addr
) {
559 unsigned int addr_type
;
562 cfg
->fc_gw_family
= AF_INET
;
563 addr_type
= inet_addr_type_table(net
, addr
, cfg
->fc_table
);
564 if (rt
->rt_flags
& RTF_GATEWAY
&&
565 addr_type
== RTN_UNICAST
)
566 cfg
->fc_scope
= RT_SCOPE_UNIVERSE
;
569 if (cmd
== SIOCDELRT
)
572 if (rt
->rt_flags
& RTF_GATEWAY
&& !cfg
->fc_gw_family
)
575 if (cfg
->fc_scope
== RT_SCOPE_NOWHERE
)
576 cfg
->fc_scope
= RT_SCOPE_LINK
;
578 if (rt
->rt_flags
& (RTF_MTU
| RTF_WINDOW
| RTF_IRTT
)) {
582 mx
= kcalloc(3, nla_total_size(4), GFP_KERNEL
);
586 if (rt
->rt_flags
& RTF_MTU
)
587 len
= put_rtax(mx
, len
, RTAX_ADVMSS
, rt
->rt_mtu
- 40);
589 if (rt
->rt_flags
& RTF_WINDOW
)
590 len
= put_rtax(mx
, len
, RTAX_WINDOW
, rt
->rt_window
);
592 if (rt
->rt_flags
& RTF_IRTT
)
593 len
= put_rtax(mx
, len
, RTAX_RTT
, rt
->rt_irtt
<< 3);
596 cfg
->fc_mx_len
= len
;
603 * Handle IP routing ioctl calls.
604 * These are used to manipulate the routing tables
606 int ip_rt_ioctl(struct net
*net
, unsigned int cmd
, struct rtentry
*rt
)
608 struct fib_config cfg
;
612 case SIOCADDRT
: /* Add a route */
613 case SIOCDELRT
: /* Delete a route */
614 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
618 err
= rtentry_to_fib_config(net
, cmd
, rt
, &cfg
);
620 struct fib_table
*tb
;
622 if (cmd
== SIOCDELRT
) {
623 tb
= fib_get_table(net
, cfg
.fc_table
);
625 err
= fib_table_delete(net
, tb
, &cfg
,
630 tb
= fib_new_table(net
, cfg
.fc_table
);
632 err
= fib_table_insert(net
, tb
,
638 /* allocated by rtentry_to_fib_config() */
647 const struct nla_policy rtm_ipv4_policy
[RTA_MAX
+ 1] = {
648 [RTA_DST
] = { .type
= NLA_U32
},
649 [RTA_SRC
] = { .type
= NLA_U32
},
650 [RTA_IIF
] = { .type
= NLA_U32
},
651 [RTA_OIF
] = { .type
= NLA_U32
},
652 [RTA_GATEWAY
] = { .type
= NLA_U32
},
653 [RTA_PRIORITY
] = { .type
= NLA_U32
},
654 [RTA_PREFSRC
] = { .type
= NLA_U32
},
655 [RTA_METRICS
] = { .type
= NLA_NESTED
},
656 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
657 [RTA_FLOW
] = { .type
= NLA_U32
},
658 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
659 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
660 [RTA_UID
] = { .type
= NLA_U32
},
661 [RTA_MARK
] = { .type
= NLA_U32
},
662 [RTA_TABLE
] = { .type
= NLA_U32
},
663 [RTA_IP_PROTO
] = { .type
= NLA_U8
},
664 [RTA_SPORT
] = { .type
= NLA_U16
},
665 [RTA_DPORT
] = { .type
= NLA_U16
},
668 int fib_gw_from_via(struct fib_config
*cfg
, struct nlattr
*nla
,
669 struct netlink_ext_ack
*extack
)
674 if (nla_len(nla
) < offsetof(struct rtvia
, rtvia_addr
)) {
675 NL_SET_ERR_MSG(extack
, "Invalid attribute length for RTA_VIA");
680 alen
= nla_len(nla
) - offsetof(struct rtvia
, rtvia_addr
);
682 switch (via
->rtvia_family
) {
684 if (alen
!= sizeof(__be32
)) {
685 NL_SET_ERR_MSG(extack
, "Invalid IPv4 address in RTA_VIA");
688 cfg
->fc_gw_family
= AF_INET
;
689 cfg
->fc_gw4
= *((__be32
*)via
->rtvia_addr
);
693 if (alen
!= sizeof(struct in6_addr
)) {
694 NL_SET_ERR_MSG(extack
, "Invalid IPv6 address in RTA_VIA");
697 cfg
->fc_gw_family
= AF_INET6
;
698 cfg
->fc_gw6
= *((struct in6_addr
*)via
->rtvia_addr
);
700 NL_SET_ERR_MSG(extack
, "IPv6 support not enabled in kernel");
705 NL_SET_ERR_MSG(extack
, "Unsupported address family in RTA_VIA");
712 static int rtm_to_fib_config(struct net
*net
, struct sk_buff
*skb
,
713 struct nlmsghdr
*nlh
, struct fib_config
*cfg
,
714 struct netlink_ext_ack
*extack
)
716 bool has_gw
= false, has_via
= false;
721 err
= nlmsg_validate_deprecated(nlh
, sizeof(*rtm
), RTA_MAX
,
722 rtm_ipv4_policy
, extack
);
726 memset(cfg
, 0, sizeof(*cfg
));
728 rtm
= nlmsg_data(nlh
);
729 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
730 cfg
->fc_tos
= rtm
->rtm_tos
;
731 cfg
->fc_table
= rtm
->rtm_table
;
732 cfg
->fc_protocol
= rtm
->rtm_protocol
;
733 cfg
->fc_scope
= rtm
->rtm_scope
;
734 cfg
->fc_type
= rtm
->rtm_type
;
735 cfg
->fc_flags
= rtm
->rtm_flags
;
736 cfg
->fc_nlflags
= nlh
->nlmsg_flags
;
738 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
739 cfg
->fc_nlinfo
.nlh
= nlh
;
740 cfg
->fc_nlinfo
.nl_net
= net
;
742 if (cfg
->fc_type
> RTN_MAX
) {
743 NL_SET_ERR_MSG(extack
, "Invalid route type");
748 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct rtmsg
), remaining
) {
749 switch (nla_type(attr
)) {
751 cfg
->fc_dst
= nla_get_be32(attr
);
754 cfg
->fc_oif
= nla_get_u32(attr
);
758 cfg
->fc_gw4
= nla_get_be32(attr
);
760 cfg
->fc_gw_family
= AF_INET
;
764 err
= fib_gw_from_via(cfg
, attr
, extack
);
769 cfg
->fc_priority
= nla_get_u32(attr
);
772 cfg
->fc_prefsrc
= nla_get_be32(attr
);
775 cfg
->fc_mx
= nla_data(attr
);
776 cfg
->fc_mx_len
= nla_len(attr
);
779 err
= lwtunnel_valid_encap_type_attr(nla_data(attr
),
784 cfg
->fc_mp
= nla_data(attr
);
785 cfg
->fc_mp_len
= nla_len(attr
);
788 cfg
->fc_flow
= nla_get_u32(attr
);
791 cfg
->fc_table
= nla_get_u32(attr
);
794 cfg
->fc_encap
= attr
;
797 cfg
->fc_encap_type
= nla_get_u16(attr
);
798 err
= lwtunnel_valid_encap_type(cfg
->fc_encap_type
,
806 if (has_gw
&& has_via
) {
807 NL_SET_ERR_MSG(extack
,
808 "Nexthop configuration can not contain both GATEWAY and VIA");
817 static int inet_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
818 struct netlink_ext_ack
*extack
)
820 struct net
*net
= sock_net(skb
->sk
);
821 struct fib_config cfg
;
822 struct fib_table
*tb
;
825 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
, extack
);
829 tb
= fib_get_table(net
, cfg
.fc_table
);
831 NL_SET_ERR_MSG(extack
, "FIB table does not exist");
836 err
= fib_table_delete(net
, tb
, &cfg
, extack
);
841 static int inet_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
842 struct netlink_ext_ack
*extack
)
844 struct net
*net
= sock_net(skb
->sk
);
845 struct fib_config cfg
;
846 struct fib_table
*tb
;
849 err
= rtm_to_fib_config(net
, skb
, nlh
, &cfg
, extack
);
853 tb
= fib_new_table(net
, cfg
.fc_table
);
859 err
= fib_table_insert(net
, tb
, &cfg
, extack
);
860 if (!err
&& cfg
.fc_type
== RTN_LOCAL
)
861 net
->ipv4
.fib_has_custom_local_routes
= true;
866 int ip_valid_fib_dump_req(struct net
*net
, const struct nlmsghdr
*nlh
,
867 struct fib_dump_filter
*filter
,
868 struct netlink_callback
*cb
)
870 struct netlink_ext_ack
*extack
= cb
->extack
;
871 struct nlattr
*tb
[RTA_MAX
+ 1];
877 if (nlh
->nlmsg_len
< nlmsg_msg_size(sizeof(*rtm
))) {
878 NL_SET_ERR_MSG(extack
, "Invalid header for FIB dump request");
882 rtm
= nlmsg_data(nlh
);
883 if (rtm
->rtm_dst_len
|| rtm
->rtm_src_len
|| rtm
->rtm_tos
||
885 NL_SET_ERR_MSG(extack
, "Invalid values in header for FIB dump request");
888 if (rtm
->rtm_flags
& ~(RTM_F_CLONED
| RTM_F_PREFIX
)) {
889 NL_SET_ERR_MSG(extack
, "Invalid flags for FIB dump request");
893 filter
->dump_all_families
= (rtm
->rtm_family
== AF_UNSPEC
);
894 filter
->flags
= rtm
->rtm_flags
;
895 filter
->protocol
= rtm
->rtm_protocol
;
896 filter
->rt_type
= rtm
->rtm_type
;
897 filter
->table_id
= rtm
->rtm_table
;
899 err
= nlmsg_parse_deprecated_strict(nlh
, sizeof(*rtm
), tb
, RTA_MAX
,
900 rtm_ipv4_policy
, extack
);
904 for (i
= 0; i
<= RTA_MAX
; ++i
) {
912 filter
->table_id
= nla_get_u32(tb
[i
]);
915 ifindex
= nla_get_u32(tb
[i
]);
916 filter
->dev
= __dev_get_by_index(net
, ifindex
);
921 NL_SET_ERR_MSG(extack
, "Unsupported attribute in dump request");
926 if (filter
->flags
|| filter
->protocol
|| filter
->rt_type
||
927 filter
->table_id
|| filter
->dev
) {
928 filter
->filter_set
= 1;
929 cb
->answer_flags
= NLM_F_DUMP_FILTERED
;
934 EXPORT_SYMBOL_GPL(ip_valid_fib_dump_req
);
936 static int inet_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
938 const struct nlmsghdr
*nlh
= cb
->nlh
;
939 struct net
*net
= sock_net(skb
->sk
);
940 struct fib_dump_filter filter
= {};
942 unsigned int e
= 0, s_e
;
943 struct fib_table
*tb
;
944 struct hlist_head
*head
;
947 if (cb
->strict_check
) {
948 err
= ip_valid_fib_dump_req(net
, nlh
, &filter
, cb
);
951 } else if (nlmsg_len(nlh
) >= sizeof(struct rtmsg
)) {
952 struct rtmsg
*rtm
= nlmsg_data(nlh
);
954 filter
.flags
= rtm
->rtm_flags
& (RTM_F_PREFIX
| RTM_F_CLONED
);
957 /* fib entries are never clones and ipv4 does not use prefix flag */
958 if (filter
.flags
& (RTM_F_PREFIX
| RTM_F_CLONED
))
961 if (filter
.table_id
) {
962 tb
= fib_get_table(net
, filter
.table_id
);
964 if (filter
.dump_all_families
)
967 NL_SET_ERR_MSG(cb
->extack
, "ipv4: FIB table does not exist");
971 err
= fib_table_dump(tb
, skb
, cb
, &filter
);
972 return skb
->len
? : err
;
980 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
982 head
= &net
->ipv4
.fib_table_hash
[h
];
983 hlist_for_each_entry_rcu(tb
, head
, tb_hlist
) {
987 memset(&cb
->args
[2], 0, sizeof(cb
->args
) -
988 2 * sizeof(cb
->args
[0]));
989 err
= fib_table_dump(tb
, skb
, cb
, &filter
);
991 if (likely(skb
->len
))
1012 /* Prepare and feed intra-kernel routing request.
1013 * Really, it should be netlink message, but :-( netlink
1014 * can be not configured, so that we feed it directly
1015 * to fib engine. It is legal, because all events occur
1016 * only when netlink is already locked.
1018 static void fib_magic(int cmd
, int type
, __be32 dst
, int dst_len
,
1019 struct in_ifaddr
*ifa
, u32 rt_priority
)
1021 struct net
*net
= dev_net(ifa
->ifa_dev
->dev
);
1022 u32 tb_id
= l3mdev_fib_table(ifa
->ifa_dev
->dev
);
1023 struct fib_table
*tb
;
1024 struct fib_config cfg
= {
1025 .fc_protocol
= RTPROT_KERNEL
,
1028 .fc_dst_len
= dst_len
,
1029 .fc_priority
= rt_priority
,
1030 .fc_prefsrc
= ifa
->ifa_local
,
1031 .fc_oif
= ifa
->ifa_dev
->dev
->ifindex
,
1032 .fc_nlflags
= NLM_F_CREATE
| NLM_F_APPEND
,
1039 tb_id
= (type
== RTN_UNICAST
) ? RT_TABLE_MAIN
: RT_TABLE_LOCAL
;
1041 tb
= fib_new_table(net
, tb_id
);
1045 cfg
.fc_table
= tb
->tb_id
;
1047 if (type
!= RTN_LOCAL
)
1048 cfg
.fc_scope
= RT_SCOPE_LINK
;
1050 cfg
.fc_scope
= RT_SCOPE_HOST
;
1052 if (cmd
== RTM_NEWROUTE
)
1053 fib_table_insert(net
, tb
, &cfg
, NULL
);
1055 fib_table_delete(net
, tb
, &cfg
, NULL
);
1058 void fib_add_ifaddr(struct in_ifaddr
*ifa
)
1060 struct in_device
*in_dev
= ifa
->ifa_dev
;
1061 struct net_device
*dev
= in_dev
->dev
;
1062 struct in_ifaddr
*prim
= ifa
;
1063 __be32 mask
= ifa
->ifa_mask
;
1064 __be32 addr
= ifa
->ifa_local
;
1065 __be32 prefix
= ifa
->ifa_address
& mask
;
1067 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
1068 prim
= inet_ifa_byprefix(in_dev
, prefix
, mask
);
1070 pr_warn("%s: bug: prim == NULL\n", __func__
);
1075 fib_magic(RTM_NEWROUTE
, RTN_LOCAL
, addr
, 32, prim
, 0);
1077 if (!(dev
->flags
& IFF_UP
))
1080 /* Add broadcast address, if it is explicitly assigned. */
1081 if (ifa
->ifa_broadcast
&& ifa
->ifa_broadcast
!= htonl(0xFFFFFFFF))
1082 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32,
1085 if (!ipv4_is_zeronet(prefix
) && !(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
1086 (prefix
!= addr
|| ifa
->ifa_prefixlen
< 32)) {
1087 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
1088 fib_magic(RTM_NEWROUTE
,
1089 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
1090 prefix
, ifa
->ifa_prefixlen
, prim
,
1091 ifa
->ifa_rt_priority
);
1093 /* Add network specific broadcasts, when it takes a sense */
1094 if (ifa
->ifa_prefixlen
< 31) {
1095 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
, 32,
1097 fib_magic(RTM_NEWROUTE
, RTN_BROADCAST
, prefix
| ~mask
,
1103 void fib_modify_prefix_metric(struct in_ifaddr
*ifa
, u32 new_metric
)
1105 __be32 prefix
= ifa
->ifa_address
& ifa
->ifa_mask
;
1106 struct in_device
*in_dev
= ifa
->ifa_dev
;
1107 struct net_device
*dev
= in_dev
->dev
;
1109 if (!(dev
->flags
& IFF_UP
) ||
1110 ifa
->ifa_flags
& (IFA_F_SECONDARY
| IFA_F_NOPREFIXROUTE
) ||
1111 ipv4_is_zeronet(prefix
) ||
1112 prefix
== ifa
->ifa_local
|| ifa
->ifa_prefixlen
== 32)
1116 fib_magic(RTM_NEWROUTE
,
1117 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
1118 prefix
, ifa
->ifa_prefixlen
, ifa
, new_metric
);
1120 /* delete the old */
1121 fib_magic(RTM_DELROUTE
,
1122 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
1123 prefix
, ifa
->ifa_prefixlen
, ifa
, ifa
->ifa_rt_priority
);
1126 /* Delete primary or secondary address.
1127 * Optionally, on secondary address promotion consider the addresses
1128 * from subnet iprim as deleted, even if they are in device list.
1129 * In this case the secondary ifa can be in device list.
1131 void fib_del_ifaddr(struct in_ifaddr
*ifa
, struct in_ifaddr
*iprim
)
1133 struct in_device
*in_dev
= ifa
->ifa_dev
;
1134 struct net_device
*dev
= in_dev
->dev
;
1135 struct in_ifaddr
*ifa1
;
1136 struct in_ifaddr
*prim
= ifa
, *prim1
= NULL
;
1137 __be32 brd
= ifa
->ifa_address
| ~ifa
->ifa_mask
;
1138 __be32 any
= ifa
->ifa_address
& ifa
->ifa_mask
;
1143 unsigned int ok
= 0;
1144 int subnet
= 0; /* Primary network */
1145 int gone
= 1; /* Address is missing */
1146 int same_prefsrc
= 0; /* Another primary with same IP */
1148 if (ifa
->ifa_flags
& IFA_F_SECONDARY
) {
1149 prim
= inet_ifa_byprefix(in_dev
, any
, ifa
->ifa_mask
);
1151 /* if the device has been deleted, we don't perform
1155 pr_warn("%s: bug: prim == NULL\n", __func__
);
1158 if (iprim
&& iprim
!= prim
) {
1159 pr_warn("%s: bug: iprim != prim\n", __func__
);
1162 } else if (!ipv4_is_zeronet(any
) &&
1163 (any
!= ifa
->ifa_local
|| ifa
->ifa_prefixlen
< 32)) {
1164 if (!(ifa
->ifa_flags
& IFA_F_NOPREFIXROUTE
))
1165 fib_magic(RTM_DELROUTE
,
1166 dev
->flags
& IFF_LOOPBACK
? RTN_LOCAL
: RTN_UNICAST
,
1167 any
, ifa
->ifa_prefixlen
, prim
, 0);
1174 /* Deletion is more complicated than add.
1175 * We should take care of not to delete too much :-)
1177 * Scan address list to be sure that addresses are really gone.
1180 for (ifa1
= in_dev
->ifa_list
; ifa1
; ifa1
= ifa1
->ifa_next
) {
1182 /* promotion, keep the IP */
1186 /* Ignore IFAs from our subnet */
1187 if (iprim
&& ifa1
->ifa_mask
== iprim
->ifa_mask
&&
1188 inet_ifa_match(ifa1
->ifa_address
, iprim
))
1191 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
1192 if (ifa1
->ifa_flags
& IFA_F_SECONDARY
) {
1193 /* Another address from our subnet? */
1194 if (ifa1
->ifa_mask
== prim
->ifa_mask
&&
1195 inet_ifa_match(ifa1
->ifa_address
, prim
))
1198 /* We reached the secondaries, so
1199 * same_prefsrc should be determined.
1203 /* Search new prim1 if ifa1 is not
1204 * using the current prim1
1207 ifa1
->ifa_mask
!= prim1
->ifa_mask
||
1208 !inet_ifa_match(ifa1
->ifa_address
, prim1
))
1209 prim1
= inet_ifa_byprefix(in_dev
,
1214 if (prim1
->ifa_local
!= prim
->ifa_local
)
1218 if (prim
->ifa_local
!= ifa1
->ifa_local
)
1224 if (ifa
->ifa_local
== ifa1
->ifa_local
)
1226 if (ifa
->ifa_broadcast
== ifa1
->ifa_broadcast
)
1228 if (brd
== ifa1
->ifa_broadcast
)
1230 if (any
== ifa1
->ifa_broadcast
)
1232 /* primary has network specific broadcasts */
1233 if (prim1
== ifa1
&& ifa1
->ifa_prefixlen
< 31) {
1234 __be32 brd1
= ifa1
->ifa_address
| ~ifa1
->ifa_mask
;
1235 __be32 any1
= ifa1
->ifa_address
& ifa1
->ifa_mask
;
1237 if (!ipv4_is_zeronet(any1
)) {
1238 if (ifa
->ifa_broadcast
== brd1
||
1239 ifa
->ifa_broadcast
== any1
)
1241 if (brd
== brd1
|| brd
== any1
)
1243 if (any
== brd1
|| any
== any1
)
1251 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, ifa
->ifa_broadcast
, 32,
1253 if (subnet
&& ifa
->ifa_prefixlen
< 31) {
1254 if (!(ok
& BRD1_OK
))
1255 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, brd
, 32,
1257 if (!(ok
& BRD0_OK
))
1258 fib_magic(RTM_DELROUTE
, RTN_BROADCAST
, any
, 32,
1261 if (!(ok
& LOCAL_OK
)) {
1262 unsigned int addr_type
;
1264 fib_magic(RTM_DELROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 32, prim
, 0);
1266 /* Check, that this local address finally disappeared. */
1267 addr_type
= inet_addr_type_dev_table(dev_net(dev
), dev
,
1269 if (gone
&& addr_type
!= RTN_LOCAL
) {
1270 /* And the last, but not the least thing.
1271 * We must flush stray FIB entries.
1273 * First of all, we scan fib_info list searching
1274 * for stray nexthop entries, then ignite fib_flush.
1276 if (fib_sync_down_addr(dev
, ifa
->ifa_local
))
1277 fib_flush(dev_net(dev
));
1286 static void nl_fib_lookup(struct net
*net
, struct fib_result_nl
*frn
)
1289 struct fib_result res
;
1290 struct flowi4 fl4
= {
1291 .flowi4_mark
= frn
->fl_mark
,
1292 .daddr
= frn
->fl_addr
,
1293 .flowi4_tos
= frn
->fl_tos
,
1294 .flowi4_scope
= frn
->fl_scope
,
1296 struct fib_table
*tb
;
1300 tb
= fib_get_table(net
, frn
->tb_id_in
);
1306 frn
->tb_id
= tb
->tb_id
;
1307 frn
->err
= fib_table_lookup(tb
, &fl4
, &res
, FIB_LOOKUP_NOREF
);
1310 frn
->prefixlen
= res
.prefixlen
;
1311 frn
->nh_sel
= res
.nh_sel
;
1312 frn
->type
= res
.type
;
1313 frn
->scope
= res
.scope
;
1321 static void nl_fib_input(struct sk_buff
*skb
)
1324 struct fib_result_nl
*frn
;
1325 struct nlmsghdr
*nlh
;
1328 net
= sock_net(skb
->sk
);
1329 nlh
= nlmsg_hdr(skb
);
1330 if (skb
->len
< nlmsg_total_size(sizeof(*frn
)) ||
1331 skb
->len
< nlh
->nlmsg_len
||
1332 nlmsg_len(nlh
) < sizeof(*frn
))
1335 skb
= netlink_skb_clone(skb
, GFP_KERNEL
);
1338 nlh
= nlmsg_hdr(skb
);
1340 frn
= (struct fib_result_nl
*) nlmsg_data(nlh
);
1341 nl_fib_lookup(net
, frn
);
1343 portid
= NETLINK_CB(skb
).portid
; /* netlink portid */
1344 NETLINK_CB(skb
).portid
= 0; /* from kernel */
1345 NETLINK_CB(skb
).dst_group
= 0; /* unicast */
1346 netlink_unicast(net
->ipv4
.fibnl
, skb
, portid
, MSG_DONTWAIT
);
1349 static int __net_init
nl_fib_lookup_init(struct net
*net
)
1352 struct netlink_kernel_cfg cfg
= {
1353 .input
= nl_fib_input
,
1356 sk
= netlink_kernel_create(net
, NETLINK_FIB_LOOKUP
, &cfg
);
1358 return -EAFNOSUPPORT
;
1359 net
->ipv4
.fibnl
= sk
;
1363 static void nl_fib_lookup_exit(struct net
*net
)
1365 netlink_kernel_release(net
->ipv4
.fibnl
);
1366 net
->ipv4
.fibnl
= NULL
;
1369 static void fib_disable_ip(struct net_device
*dev
, unsigned long event
,
1372 if (fib_sync_down_dev(dev
, event
, force
))
1373 fib_flush(dev_net(dev
));
1375 rt_cache_flush(dev_net(dev
));
1379 static int fib_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1381 struct in_ifaddr
*ifa
= (struct in_ifaddr
*)ptr
;
1382 struct net_device
*dev
= ifa
->ifa_dev
->dev
;
1383 struct net
*net
= dev_net(dev
);
1387 fib_add_ifaddr(ifa
);
1388 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1389 fib_sync_up(dev
, RTNH_F_DEAD
);
1391 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1392 rt_cache_flush(dev_net(dev
));
1395 fib_del_ifaddr(ifa
, NULL
);
1396 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1397 if (!ifa
->ifa_dev
->ifa_list
) {
1398 /* Last address was deleted from this interface.
1401 fib_disable_ip(dev
, event
, true);
1403 rt_cache_flush(dev_net(dev
));
1410 static int fib_netdev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1412 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1413 struct netdev_notifier_changeupper_info
*upper_info
= ptr
;
1414 struct netdev_notifier_info_ext
*info_ext
= ptr
;
1415 struct in_device
*in_dev
;
1416 struct net
*net
= dev_net(dev
);
1419 if (event
== NETDEV_UNREGISTER
) {
1420 fib_disable_ip(dev
, event
, true);
1425 in_dev
= __in_dev_get_rtnl(dev
);
1432 fib_add_ifaddr(ifa
);
1433 } endfor_ifa(in_dev
);
1434 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1435 fib_sync_up(dev
, RTNH_F_DEAD
);
1437 atomic_inc(&net
->ipv4
.dev_addr_genid
);
1438 rt_cache_flush(net
);
1441 fib_disable_ip(dev
, event
, false);
1444 flags
= dev_get_flags(dev
);
1445 if (flags
& (IFF_RUNNING
| IFF_LOWER_UP
))
1446 fib_sync_up(dev
, RTNH_F_LINKDOWN
);
1448 fib_sync_down_dev(dev
, event
, false);
1449 rt_cache_flush(net
);
1451 case NETDEV_CHANGEMTU
:
1452 fib_sync_mtu(dev
, info_ext
->ext
.mtu
);
1453 rt_cache_flush(net
);
1455 case NETDEV_CHANGEUPPER
:
1457 /* flush all routes if dev is linked to or unlinked from
1458 * an L3 master device (e.g., VRF)
1460 if (upper_info
->upper_dev
&&
1461 netif_is_l3_master(upper_info
->upper_dev
))
1462 fib_disable_ip(dev
, NETDEV_DOWN
, true);
1468 static struct notifier_block fib_inetaddr_notifier
= {
1469 .notifier_call
= fib_inetaddr_event
,
1472 static struct notifier_block fib_netdev_notifier
= {
1473 .notifier_call
= fib_netdev_event
,
1476 static int __net_init
ip_fib_net_init(struct net
*net
)
1479 size_t size
= sizeof(struct hlist_head
) * FIB_TABLE_HASHSZ
;
1481 err
= fib4_notifier_init(net
);
1485 /* Avoid false sharing : Use at least a full cache line */
1486 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1488 net
->ipv4
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1489 if (!net
->ipv4
.fib_table_hash
) {
1491 goto err_table_hash_alloc
;
1494 err
= fib4_rules_init(net
);
1496 goto err_rules_init
;
1500 kfree(net
->ipv4
.fib_table_hash
);
1501 err_table_hash_alloc
:
1502 fib4_notifier_exit(net
);
1506 static void ip_fib_net_exit(struct net
*net
)
1511 #ifdef CONFIG_IP_MULTIPLE_TABLES
1512 RCU_INIT_POINTER(net
->ipv4
.fib_main
, NULL
);
1513 RCU_INIT_POINTER(net
->ipv4
.fib_default
, NULL
);
1515 /* Destroy the tables in reverse order to guarantee that the
1516 * local table, ID 255, is destroyed before the main table, ID
1517 * 254. This is necessary as the local table may contain
1518 * references to data contained in the main table.
1520 for (i
= FIB_TABLE_HASHSZ
- 1; i
>= 0; i
--) {
1521 struct hlist_head
*head
= &net
->ipv4
.fib_table_hash
[i
];
1522 struct hlist_node
*tmp
;
1523 struct fib_table
*tb
;
1525 hlist_for_each_entry_safe(tb
, tmp
, head
, tb_hlist
) {
1526 hlist_del(&tb
->tb_hlist
);
1527 fib_table_flush(net
, tb
, true);
1532 #ifdef CONFIG_IP_MULTIPLE_TABLES
1533 fib4_rules_exit(net
);
1536 kfree(net
->ipv4
.fib_table_hash
);
1537 fib4_notifier_exit(net
);
1540 static int __net_init
fib_net_init(struct net
*net
)
1544 #ifdef CONFIG_IP_ROUTE_CLASSID
1545 net
->ipv4
.fib_num_tclassid_users
= 0;
1547 error
= ip_fib_net_init(net
);
1550 error
= nl_fib_lookup_init(net
);
1553 error
= fib_proc_init(net
);
1560 nl_fib_lookup_exit(net
);
1562 ip_fib_net_exit(net
);
1566 static void __net_exit
fib_net_exit(struct net
*net
)
1569 nl_fib_lookup_exit(net
);
1570 ip_fib_net_exit(net
);
1573 static struct pernet_operations fib_net_ops
= {
1574 .init
= fib_net_init
,
1575 .exit
= fib_net_exit
,
1578 void __init
ip_fib_init(void)
1582 register_pernet_subsys(&fib_net_ops
);
1584 register_netdevice_notifier(&fib_netdev_notifier
);
1585 register_inetaddr_notifier(&fib_inetaddr_notifier
);
1587 rtnl_register(PF_INET
, RTM_NEWROUTE
, inet_rtm_newroute
, NULL
, 0);
1588 rtnl_register(PF_INET
, RTM_DELROUTE
, inet_rtm_delroute
, NULL
, 0);
1589 rtnl_register(PF_INET
, RTM_GETROUTE
, NULL
, inet_dump_fib
, 0);