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.
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <linux/module.h>
17 #include <asm/uaccess.h>
18 #include <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>
49 #ifndef CONFIG_IP_MULTIPLE_TABLES
51 static int __net_init fib4_rules_init(struct net *net)
53 struct fib_table *local_table, *main_table;
55 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
59 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
63 hlist_add_head_rcu(&local_table->tb_hlist,
64 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
65 hlist_add_head_rcu(&main_table->tb_hlist,
66 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
70 fib_free_table(main_table);
75 struct fib_table *fib_new_table(struct net *net, u32 id)
77 struct fib_table *tb, *alias = NULL;
82 tb = fib_get_table(net, id);
86 if (id == RT_TABLE_LOCAL)
87 alias = fib_new_table(net, RT_TABLE_MAIN);
89 tb = fib_trie_table(id, alias);
95 rcu_assign_pointer(net->ipv4.fib_local, tb);
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]);
112 /* caller must hold either rtnl or rcu read lock */
113 struct fib_table *fib_get_table(struct net *net, u32 id)
115 struct fib_table *tb;
116 struct hlist_head *head;
121 h = id & (FIB_TABLE_HASHSZ - 1);
123 head = &net->ipv4.fib_table_hash[h];
124 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
130 #endif /* CONFIG_IP_MULTIPLE_TABLES */
132 static void fib_replace_table(struct net *net, struct fib_table *old,
133 struct fib_table *new)
135 #ifdef CONFIG_IP_MULTIPLE_TABLES
136 switch (new->tb_id) {
138 rcu_assign_pointer(net->ipv4.fib_local, new);
141 rcu_assign_pointer(net->ipv4.fib_main, new);
143 case RT_TABLE_DEFAULT:
144 rcu_assign_pointer(net->ipv4.fib_default, new);
151 /* replace the old table in the hlist */
152 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
155 int fib_unmerge(struct net *net)
157 struct fib_table *old, *new;
159 /* attempt to fetch local table if it has been allocated */
160 old = fib_get_table(net, RT_TABLE_LOCAL);
164 new = fib_trie_unmerge(old);
168 /* replace merged table with clean table */
170 fib_replace_table(net, old, new);
177 static void fib_flush(struct net *net)
182 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
183 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
184 struct hlist_node *tmp;
185 struct fib_table *tb;
187 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
188 flushed += fib_table_flush(tb);
195 void fib_flush_external(struct net *net)
197 struct fib_table *tb;
198 struct hlist_head *head;
201 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
202 head = &net->ipv4.fib_table_hash[h];
203 hlist_for_each_entry(tb, head, tb_hlist)
204 fib_table_flush_external(tb);
209 * Find address type as if only "dev" was present in the system. If
210 * on_dev is NULL then all interfaces are taken into consideration.
212 static inline unsigned int __inet_dev_addr_type(struct net *net,
213 const struct net_device *dev,
216 struct flowi4 fl4 = { .daddr = addr };
217 struct fib_result res;
218 unsigned int ret = RTN_BROADCAST;
219 struct fib_table *local_table;
221 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
222 return RTN_BROADCAST;
223 if (ipv4_is_multicast(addr))
224 return RTN_MULTICAST;
228 local_table = fib_get_table(net, RT_TABLE_LOCAL);
231 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
232 if (!dev || dev == res.fi->fib_dev)
241 unsigned int inet_addr_type(struct net *net, __be32 addr)
243 return __inet_dev_addr_type(net, NULL, addr);
245 EXPORT_SYMBOL(inet_addr_type);
247 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
250 return __inet_dev_addr_type(net, dev, addr);
252 EXPORT_SYMBOL(inet_dev_addr_type);
254 __be32 fib_compute_spec_dst(struct sk_buff *skb)
256 struct net_device *dev = skb->dev;
257 struct in_device *in_dev;
258 struct fib_result res;
264 rt = skb_rtable(skb);
265 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
267 return ip_hdr(skb)->daddr;
269 in_dev = __in_dev_get_rcu(dev);
274 scope = RT_SCOPE_UNIVERSE;
275 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
277 fl4.flowi4_iif = LOOPBACK_IFINDEX;
278 fl4.daddr = ip_hdr(skb)->saddr;
280 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
281 fl4.flowi4_scope = scope;
282 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
283 if (!fib_lookup(net, &fl4, &res))
284 return FIB_RES_PREFSRC(net, res);
286 scope = RT_SCOPE_LINK;
289 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
292 /* Given (packet source, input interface) and optional (dst, oif, tos):
293 * - (main) check, that source is valid i.e. not broadcast or our local
295 * - figure out what "logical" interface this packet arrived
296 * and calculate "specific destination" address.
297 * - check, that packet arrived from expected physical interface.
298 * called with rcu_read_lock()
300 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
301 u8 tos, int oif, struct net_device *dev,
302 int rpf, struct in_device *idev, u32 *itag)
305 struct fib_result res;
311 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
314 fl4.flowi4_tos = tos;
315 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
317 no_addr = idev->ifa_list == NULL;
319 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
322 if (fib_lookup(net, &fl4, &res))
324 if (res.type != RTN_UNICAST &&
325 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
327 if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
328 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
330 fib_combine_itag(itag, &res);
333 #ifdef CONFIG_IP_ROUTE_MULTIPATH
334 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
335 struct fib_nh *nh = &res.fi->fib_nh[ret];
337 if (nh->nh_dev == dev) {
343 if (FIB_RES_DEV(res) == dev)
347 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
354 fl4.flowi4_oif = dev->ifindex;
357 if (fib_lookup(net, &fl4, &res) == 0) {
358 if (res.type == RTN_UNICAST)
359 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
375 /* Ignore rp_filter for packets protected by IPsec. */
376 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
377 u8 tos, int oif, struct net_device *dev,
378 struct in_device *idev, u32 *itag)
380 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
382 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
383 IN_DEV_ACCEPT_LOCAL(idev) &&
384 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
388 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
391 static inline __be32 sk_extract_addr(struct sockaddr *addr)
393 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
396 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
400 nla = (struct nlattr *) ((char *) mx + len);
401 nla->nla_type = type;
402 nla->nla_len = nla_attr_size(4);
403 *(u32 *) nla_data(nla) = value;
405 return len + nla_total_size(4);
408 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
409 struct fib_config *cfg)
414 memset(cfg, 0, sizeof(*cfg));
415 cfg->fc_nlinfo.nl_net = net;
417 if (rt->rt_dst.sa_family != AF_INET)
418 return -EAFNOSUPPORT;
421 * Check mask for validity:
422 * a) it must be contiguous.
423 * b) destination must have all host bits clear.
424 * c) if application forgot to set correct family (AF_INET),
425 * reject request unless it is absolutely clear i.e.
426 * both family and mask are zero.
429 addr = sk_extract_addr(&rt->rt_dst);
430 if (!(rt->rt_flags & RTF_HOST)) {
431 __be32 mask = sk_extract_addr(&rt->rt_genmask);
433 if (rt->rt_genmask.sa_family != AF_INET) {
434 if (mask || rt->rt_genmask.sa_family)
435 return -EAFNOSUPPORT;
438 if (bad_mask(mask, addr))
441 plen = inet_mask_len(mask);
444 cfg->fc_dst_len = plen;
447 if (cmd != SIOCDELRT) {
448 cfg->fc_nlflags = NLM_F_CREATE;
449 cfg->fc_protocol = RTPROT_BOOT;
453 cfg->fc_priority = rt->rt_metric - 1;
455 if (rt->rt_flags & RTF_REJECT) {
456 cfg->fc_scope = RT_SCOPE_HOST;
457 cfg->fc_type = RTN_UNREACHABLE;
461 cfg->fc_scope = RT_SCOPE_NOWHERE;
462 cfg->fc_type = RTN_UNICAST;
466 struct net_device *dev;
467 char devname[IFNAMSIZ];
469 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
472 devname[IFNAMSIZ-1] = 0;
473 colon = strchr(devname, ':');
476 dev = __dev_get_by_name(net, devname);
479 cfg->fc_oif = dev->ifindex;
481 struct in_ifaddr *ifa;
482 struct in_device *in_dev = __in_dev_get_rtnl(dev);
486 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
487 if (strcmp(ifa->ifa_label, devname) == 0)
491 cfg->fc_prefsrc = ifa->ifa_local;
495 addr = sk_extract_addr(&rt->rt_gateway);
496 if (rt->rt_gateway.sa_family == AF_INET && addr) {
498 if (rt->rt_flags & RTF_GATEWAY &&
499 inet_addr_type(net, addr) == RTN_UNICAST)
500 cfg->fc_scope = RT_SCOPE_UNIVERSE;
503 if (cmd == SIOCDELRT)
506 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
509 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
510 cfg->fc_scope = RT_SCOPE_LINK;
512 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
516 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
520 if (rt->rt_flags & RTF_MTU)
521 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
523 if (rt->rt_flags & RTF_WINDOW)
524 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
526 if (rt->rt_flags & RTF_IRTT)
527 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
530 cfg->fc_mx_len = len;
537 * Handle IP routing ioctl calls.
538 * These are used to manipulate the routing tables
540 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
542 struct fib_config cfg;
547 case SIOCADDRT: /* Add a route */
548 case SIOCDELRT: /* Delete a route */
549 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
552 if (copy_from_user(&rt, arg, sizeof(rt)))
556 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
558 struct fib_table *tb;
560 if (cmd == SIOCDELRT) {
561 tb = fib_get_table(net, cfg.fc_table);
563 err = fib_table_delete(tb, &cfg);
567 tb = fib_new_table(net, cfg.fc_table);
569 err = fib_table_insert(tb, &cfg);
574 /* allocated by rtentry_to_fib_config() */
583 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
584 [RTA_DST] = { .type = NLA_U32 },
585 [RTA_SRC] = { .type = NLA_U32 },
586 [RTA_IIF] = { .type = NLA_U32 },
587 [RTA_OIF] = { .type = NLA_U32 },
588 [RTA_GATEWAY] = { .type = NLA_U32 },
589 [RTA_PRIORITY] = { .type = NLA_U32 },
590 [RTA_PREFSRC] = { .type = NLA_U32 },
591 [RTA_METRICS] = { .type = NLA_NESTED },
592 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
593 [RTA_FLOW] = { .type = NLA_U32 },
596 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
597 struct nlmsghdr *nlh, struct fib_config *cfg)
603 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
607 memset(cfg, 0, sizeof(*cfg));
609 rtm = nlmsg_data(nlh);
610 cfg->fc_dst_len = rtm->rtm_dst_len;
611 cfg->fc_tos = rtm->rtm_tos;
612 cfg->fc_table = rtm->rtm_table;
613 cfg->fc_protocol = rtm->rtm_protocol;
614 cfg->fc_scope = rtm->rtm_scope;
615 cfg->fc_type = rtm->rtm_type;
616 cfg->fc_flags = rtm->rtm_flags;
617 cfg->fc_nlflags = nlh->nlmsg_flags;
619 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
620 cfg->fc_nlinfo.nlh = nlh;
621 cfg->fc_nlinfo.nl_net = net;
623 if (cfg->fc_type > RTN_MAX) {
628 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
629 switch (nla_type(attr)) {
631 cfg->fc_dst = nla_get_be32(attr);
634 cfg->fc_oif = nla_get_u32(attr);
637 cfg->fc_gw = nla_get_be32(attr);
640 cfg->fc_priority = nla_get_u32(attr);
643 cfg->fc_prefsrc = nla_get_be32(attr);
646 cfg->fc_mx = nla_data(attr);
647 cfg->fc_mx_len = nla_len(attr);
650 cfg->fc_mp = nla_data(attr);
651 cfg->fc_mp_len = nla_len(attr);
654 cfg->fc_flow = nla_get_u32(attr);
657 cfg->fc_table = nla_get_u32(attr);
667 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
669 struct net *net = sock_net(skb->sk);
670 struct fib_config cfg;
671 struct fib_table *tb;
674 err = rtm_to_fib_config(net, skb, nlh, &cfg);
678 tb = fib_get_table(net, cfg.fc_table);
684 err = fib_table_delete(tb, &cfg);
689 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
691 struct net *net = sock_net(skb->sk);
692 struct fib_config cfg;
693 struct fib_table *tb;
696 err = rtm_to_fib_config(net, skb, nlh, &cfg);
700 tb = fib_new_table(net, cfg.fc_table);
706 err = fib_table_insert(tb, &cfg);
711 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
713 struct net *net = sock_net(skb->sk);
715 unsigned int e = 0, s_e;
716 struct fib_table *tb;
717 struct hlist_head *head;
720 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
721 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
729 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
731 head = &net->ipv4.fib_table_hash[h];
732 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
736 memset(&cb->args[2], 0, sizeof(cb->args) -
737 2 * sizeof(cb->args[0]));
738 if (fib_table_dump(tb, skb, cb) < 0)
754 /* Prepare and feed intra-kernel routing request.
755 * Really, it should be netlink message, but :-( netlink
756 * can be not configured, so that we feed it directly
757 * to fib engine. It is legal, because all events occur
758 * only when netlink is already locked.
760 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
762 struct net *net = dev_net(ifa->ifa_dev->dev);
763 struct fib_table *tb;
764 struct fib_config cfg = {
765 .fc_protocol = RTPROT_KERNEL,
768 .fc_dst_len = dst_len,
769 .fc_prefsrc = ifa->ifa_local,
770 .fc_oif = ifa->ifa_dev->dev->ifindex,
771 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
777 if (type == RTN_UNICAST)
778 tb = fib_new_table(net, RT_TABLE_MAIN);
780 tb = fib_new_table(net, RT_TABLE_LOCAL);
785 cfg.fc_table = tb->tb_id;
787 if (type != RTN_LOCAL)
788 cfg.fc_scope = RT_SCOPE_LINK;
790 cfg.fc_scope = RT_SCOPE_HOST;
792 if (cmd == RTM_NEWROUTE)
793 fib_table_insert(tb, &cfg);
795 fib_table_delete(tb, &cfg);
798 void fib_add_ifaddr(struct in_ifaddr *ifa)
800 struct in_device *in_dev = ifa->ifa_dev;
801 struct net_device *dev = in_dev->dev;
802 struct in_ifaddr *prim = ifa;
803 __be32 mask = ifa->ifa_mask;
804 __be32 addr = ifa->ifa_local;
805 __be32 prefix = ifa->ifa_address & mask;
807 if (ifa->ifa_flags & IFA_F_SECONDARY) {
808 prim = inet_ifa_byprefix(in_dev, prefix, mask);
810 pr_warn("%s: bug: prim == NULL\n", __func__);
815 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
817 if (!(dev->flags & IFF_UP))
820 /* Add broadcast address, if it is explicitly assigned. */
821 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
822 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
824 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
825 (prefix != addr || ifa->ifa_prefixlen < 32)) {
826 fib_magic(RTM_NEWROUTE,
827 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
828 prefix, ifa->ifa_prefixlen, prim);
830 /* Add network specific broadcasts, when it takes a sense */
831 if (ifa->ifa_prefixlen < 31) {
832 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
833 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
839 /* Delete primary or secondary address.
840 * Optionally, on secondary address promotion consider the addresses
841 * from subnet iprim as deleted, even if they are in device list.
842 * In this case the secondary ifa can be in device list.
844 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
846 struct in_device *in_dev = ifa->ifa_dev;
847 struct net_device *dev = in_dev->dev;
848 struct in_ifaddr *ifa1;
849 struct in_ifaddr *prim = ifa, *prim1 = NULL;
850 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
851 __be32 any = ifa->ifa_address & ifa->ifa_mask;
857 int subnet = 0; /* Primary network */
858 int gone = 1; /* Address is missing */
859 int same_prefsrc = 0; /* Another primary with same IP */
861 if (ifa->ifa_flags & IFA_F_SECONDARY) {
862 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
864 pr_warn("%s: bug: prim == NULL\n", __func__);
867 if (iprim && iprim != prim) {
868 pr_warn("%s: bug: iprim != prim\n", __func__);
871 } else if (!ipv4_is_zeronet(any) &&
872 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
873 fib_magic(RTM_DELROUTE,
874 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
875 any, ifa->ifa_prefixlen, prim);
879 /* Deletion is more complicated than add.
880 * We should take care of not to delete too much :-)
882 * Scan address list to be sure that addresses are really gone.
885 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
887 /* promotion, keep the IP */
891 /* Ignore IFAs from our subnet */
892 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
893 inet_ifa_match(ifa1->ifa_address, iprim))
896 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
897 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
898 /* Another address from our subnet? */
899 if (ifa1->ifa_mask == prim->ifa_mask &&
900 inet_ifa_match(ifa1->ifa_address, prim))
903 /* We reached the secondaries, so
904 * same_prefsrc should be determined.
908 /* Search new prim1 if ifa1 is not
909 * using the current prim1
912 ifa1->ifa_mask != prim1->ifa_mask ||
913 !inet_ifa_match(ifa1->ifa_address, prim1))
914 prim1 = inet_ifa_byprefix(in_dev,
919 if (prim1->ifa_local != prim->ifa_local)
923 if (prim->ifa_local != ifa1->ifa_local)
929 if (ifa->ifa_local == ifa1->ifa_local)
931 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
933 if (brd == ifa1->ifa_broadcast)
935 if (any == ifa1->ifa_broadcast)
937 /* primary has network specific broadcasts */
938 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
939 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
940 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
942 if (!ipv4_is_zeronet(any1)) {
943 if (ifa->ifa_broadcast == brd1 ||
944 ifa->ifa_broadcast == any1)
946 if (brd == brd1 || brd == any1)
948 if (any == brd1 || any == any1)
955 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
956 if (subnet && ifa->ifa_prefixlen < 31) {
958 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
960 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
962 if (!(ok & LOCAL_OK)) {
963 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
965 /* Check, that this local address finally disappeared. */
967 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
968 /* And the last, but not the least thing.
969 * We must flush stray FIB entries.
971 * First of all, we scan fib_info list searching
972 * for stray nexthop entries, then ignite fib_flush.
974 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
975 fib_flush(dev_net(dev));
984 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
987 struct fib_result res;
988 struct flowi4 fl4 = {
989 .flowi4_mark = frn->fl_mark,
990 .daddr = frn->fl_addr,
991 .flowi4_tos = frn->fl_tos,
992 .flowi4_scope = frn->fl_scope,
994 struct fib_table *tb;
998 tb = fib_get_table(net, frn->tb_id_in);
1004 frn->tb_id = tb->tb_id;
1005 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1008 frn->prefixlen = res.prefixlen;
1009 frn->nh_sel = res.nh_sel;
1010 frn->type = res.type;
1011 frn->scope = res.scope;
1019 static void nl_fib_input(struct sk_buff *skb)
1022 struct fib_result_nl *frn;
1023 struct nlmsghdr *nlh;
1026 net = sock_net(skb->sk);
1027 nlh = nlmsg_hdr(skb);
1028 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1029 nlmsg_len(nlh) < sizeof(*frn))
1032 skb = netlink_skb_clone(skb, GFP_KERNEL);
1035 nlh = nlmsg_hdr(skb);
1037 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1038 nl_fib_lookup(net, frn);
1040 portid = NETLINK_CB(skb).portid; /* netlink portid */
1041 NETLINK_CB(skb).portid = 0; /* from kernel */
1042 NETLINK_CB(skb).dst_group = 0; /* unicast */
1043 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1046 static int __net_init nl_fib_lookup_init(struct net *net)
1049 struct netlink_kernel_cfg cfg = {
1050 .input = nl_fib_input,
1053 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1055 return -EAFNOSUPPORT;
1056 net->ipv4.fibnl = sk;
1060 static void nl_fib_lookup_exit(struct net *net)
1062 netlink_kernel_release(net->ipv4.fibnl);
1063 net->ipv4.fibnl = NULL;
1066 static void fib_disable_ip(struct net_device *dev, int force)
1068 if (fib_sync_down_dev(dev, force))
1069 fib_flush(dev_net(dev));
1070 rt_cache_flush(dev_net(dev));
1074 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1076 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1077 struct net_device *dev = ifa->ifa_dev->dev;
1078 struct net *net = dev_net(dev);
1082 fib_add_ifaddr(ifa);
1083 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1086 atomic_inc(&net->ipv4.dev_addr_genid);
1087 rt_cache_flush(dev_net(dev));
1090 fib_del_ifaddr(ifa, NULL);
1091 atomic_inc(&net->ipv4.dev_addr_genid);
1092 if (!ifa->ifa_dev->ifa_list) {
1093 /* Last address was deleted from this interface.
1096 fib_disable_ip(dev, 1);
1098 rt_cache_flush(dev_net(dev));
1105 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1107 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1108 struct in_device *in_dev;
1109 struct net *net = dev_net(dev);
1111 if (event == NETDEV_UNREGISTER) {
1112 fib_disable_ip(dev, 2);
1117 in_dev = __in_dev_get_rtnl(dev);
1124 fib_add_ifaddr(ifa);
1125 } endfor_ifa(in_dev);
1126 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1129 atomic_inc(&net->ipv4.dev_addr_genid);
1130 rt_cache_flush(net);
1133 fib_disable_ip(dev, 0);
1135 case NETDEV_CHANGEMTU:
1137 rt_cache_flush(net);
1143 static struct notifier_block fib_inetaddr_notifier = {
1144 .notifier_call = fib_inetaddr_event,
1147 static struct notifier_block fib_netdev_notifier = {
1148 .notifier_call = fib_netdev_event,
1151 static int __net_init ip_fib_net_init(struct net *net)
1154 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1156 /* Avoid false sharing : Use at least a full cache line */
1157 size = max_t(size_t, size, L1_CACHE_BYTES);
1159 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1160 if (!net->ipv4.fib_table_hash)
1163 err = fib4_rules_init(net);
1169 kfree(net->ipv4.fib_table_hash);
1173 static void ip_fib_net_exit(struct net *net)
1178 #ifdef CONFIG_IP_MULTIPLE_TABLES
1179 RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1180 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1181 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1183 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1184 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1185 struct hlist_node *tmp;
1186 struct fib_table *tb;
1188 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1189 hlist_del(&tb->tb_hlist);
1190 fib_table_flush(tb);
1195 #ifdef CONFIG_IP_MULTIPLE_TABLES
1196 fib4_rules_exit(net);
1199 kfree(net->ipv4.fib_table_hash);
1202 static int __net_init fib_net_init(struct net *net)
1206 #ifdef CONFIG_IP_ROUTE_CLASSID
1207 net->ipv4.fib_num_tclassid_users = 0;
1209 error = ip_fib_net_init(net);
1212 error = nl_fib_lookup_init(net);
1215 error = fib_proc_init(net);
1222 nl_fib_lookup_exit(net);
1224 ip_fib_net_exit(net);
1228 static void __net_exit fib_net_exit(struct net *net)
1231 nl_fib_lookup_exit(net);
1232 ip_fib_net_exit(net);
1235 static struct pernet_operations fib_net_ops = {
1236 .init = fib_net_init,
1237 .exit = fib_net_exit,
1240 void __init ip_fib_init(void)
1242 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1243 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1244 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1246 register_pernet_subsys(&fib_net_ops);
1247 register_netdevice_notifier(&fib_netdev_notifier);
1248 register_inetaddr_notifier(&fib_inetaddr_notifier);