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 <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 fl4.flowi4_tun_key.tun_id = 0;
284 if (!fib_lookup(net, &fl4, &res, 0))
285 return FIB_RES_PREFSRC(net, res);
287 scope = RT_SCOPE_LINK;
290 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
293 /* Given (packet source, input interface) and optional (dst, oif, tos):
294 * - (main) check, that source is valid i.e. not broadcast or our local
296 * - figure out what "logical" interface this packet arrived
297 * and calculate "specific destination" address.
298 * - check, that packet arrived from expected physical interface.
299 * called with rcu_read_lock()
301 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
302 u8 tos, int oif, struct net_device *dev,
303 int rpf, struct in_device *idev, u32 *itag)
306 struct fib_result res;
312 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
315 fl4.flowi4_tos = tos;
316 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
317 fl4.flowi4_tun_key.tun_id = 0;
319 no_addr = idev->ifa_list == NULL;
321 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
324 if (fib_lookup(net, &fl4, &res, 0))
326 if (res.type != RTN_UNICAST &&
327 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
329 if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
330 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
332 fib_combine_itag(itag, &res);
335 #ifdef CONFIG_IP_ROUTE_MULTIPATH
336 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
337 struct fib_nh *nh = &res.fi->fib_nh[ret];
339 if (nh->nh_dev == dev) {
345 if (FIB_RES_DEV(res) == dev)
349 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
356 fl4.flowi4_oif = dev->ifindex;
359 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
360 if (res.type == RTN_UNICAST)
361 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
377 /* Ignore rp_filter for packets protected by IPsec. */
378 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
379 u8 tos, int oif, struct net_device *dev,
380 struct in_device *idev, u32 *itag)
382 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
384 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
385 IN_DEV_ACCEPT_LOCAL(idev) &&
386 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
390 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
393 static inline __be32 sk_extract_addr(struct sockaddr *addr)
395 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
398 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
402 nla = (struct nlattr *) ((char *) mx + len);
403 nla->nla_type = type;
404 nla->nla_len = nla_attr_size(4);
405 *(u32 *) nla_data(nla) = value;
407 return len + nla_total_size(4);
410 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
411 struct fib_config *cfg)
416 memset(cfg, 0, sizeof(*cfg));
417 cfg->fc_nlinfo.nl_net = net;
419 if (rt->rt_dst.sa_family != AF_INET)
420 return -EAFNOSUPPORT;
423 * Check mask for validity:
424 * a) it must be contiguous.
425 * b) destination must have all host bits clear.
426 * c) if application forgot to set correct family (AF_INET),
427 * reject request unless it is absolutely clear i.e.
428 * both family and mask are zero.
431 addr = sk_extract_addr(&rt->rt_dst);
432 if (!(rt->rt_flags & RTF_HOST)) {
433 __be32 mask = sk_extract_addr(&rt->rt_genmask);
435 if (rt->rt_genmask.sa_family != AF_INET) {
436 if (mask || rt->rt_genmask.sa_family)
437 return -EAFNOSUPPORT;
440 if (bad_mask(mask, addr))
443 plen = inet_mask_len(mask);
446 cfg->fc_dst_len = plen;
449 if (cmd != SIOCDELRT) {
450 cfg->fc_nlflags = NLM_F_CREATE;
451 cfg->fc_protocol = RTPROT_BOOT;
455 cfg->fc_priority = rt->rt_metric - 1;
457 if (rt->rt_flags & RTF_REJECT) {
458 cfg->fc_scope = RT_SCOPE_HOST;
459 cfg->fc_type = RTN_UNREACHABLE;
463 cfg->fc_scope = RT_SCOPE_NOWHERE;
464 cfg->fc_type = RTN_UNICAST;
468 struct net_device *dev;
469 char devname[IFNAMSIZ];
471 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
474 devname[IFNAMSIZ-1] = 0;
475 colon = strchr(devname, ':');
478 dev = __dev_get_by_name(net, devname);
481 cfg->fc_oif = dev->ifindex;
483 struct in_ifaddr *ifa;
484 struct in_device *in_dev = __in_dev_get_rtnl(dev);
488 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
489 if (strcmp(ifa->ifa_label, devname) == 0)
493 cfg->fc_prefsrc = ifa->ifa_local;
497 addr = sk_extract_addr(&rt->rt_gateway);
498 if (rt->rt_gateway.sa_family == AF_INET && addr) {
500 if (rt->rt_flags & RTF_GATEWAY &&
501 inet_addr_type(net, addr) == RTN_UNICAST)
502 cfg->fc_scope = RT_SCOPE_UNIVERSE;
505 if (cmd == SIOCDELRT)
508 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
511 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
512 cfg->fc_scope = RT_SCOPE_LINK;
514 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
518 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
522 if (rt->rt_flags & RTF_MTU)
523 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
525 if (rt->rt_flags & RTF_WINDOW)
526 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
528 if (rt->rt_flags & RTF_IRTT)
529 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
532 cfg->fc_mx_len = len;
539 * Handle IP routing ioctl calls.
540 * These are used to manipulate the routing tables
542 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
544 struct fib_config cfg;
549 case SIOCADDRT: /* Add a route */
550 case SIOCDELRT: /* Delete a route */
551 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
554 if (copy_from_user(&rt, arg, sizeof(rt)))
558 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
560 struct fib_table *tb;
562 if (cmd == SIOCDELRT) {
563 tb = fib_get_table(net, cfg.fc_table);
565 err = fib_table_delete(tb, &cfg);
569 tb = fib_new_table(net, cfg.fc_table);
571 err = fib_table_insert(tb, &cfg);
576 /* allocated by rtentry_to_fib_config() */
585 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
586 [RTA_DST] = { .type = NLA_U32 },
587 [RTA_SRC] = { .type = NLA_U32 },
588 [RTA_IIF] = { .type = NLA_U32 },
589 [RTA_OIF] = { .type = NLA_U32 },
590 [RTA_GATEWAY] = { .type = NLA_U32 },
591 [RTA_PRIORITY] = { .type = NLA_U32 },
592 [RTA_PREFSRC] = { .type = NLA_U32 },
593 [RTA_METRICS] = { .type = NLA_NESTED },
594 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
595 [RTA_FLOW] = { .type = NLA_U32 },
596 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
597 [RTA_ENCAP] = { .type = NLA_NESTED },
600 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
601 struct nlmsghdr *nlh, struct fib_config *cfg)
607 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
611 memset(cfg, 0, sizeof(*cfg));
613 rtm = nlmsg_data(nlh);
614 cfg->fc_dst_len = rtm->rtm_dst_len;
615 cfg->fc_tos = rtm->rtm_tos;
616 cfg->fc_table = rtm->rtm_table;
617 cfg->fc_protocol = rtm->rtm_protocol;
618 cfg->fc_scope = rtm->rtm_scope;
619 cfg->fc_type = rtm->rtm_type;
620 cfg->fc_flags = rtm->rtm_flags;
621 cfg->fc_nlflags = nlh->nlmsg_flags;
623 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
624 cfg->fc_nlinfo.nlh = nlh;
625 cfg->fc_nlinfo.nl_net = net;
627 if (cfg->fc_type > RTN_MAX) {
632 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
633 switch (nla_type(attr)) {
635 cfg->fc_dst = nla_get_be32(attr);
638 cfg->fc_oif = nla_get_u32(attr);
641 cfg->fc_gw = nla_get_be32(attr);
644 cfg->fc_priority = nla_get_u32(attr);
647 cfg->fc_prefsrc = nla_get_be32(attr);
650 cfg->fc_mx = nla_data(attr);
651 cfg->fc_mx_len = nla_len(attr);
654 cfg->fc_mp = nla_data(attr);
655 cfg->fc_mp_len = nla_len(attr);
658 cfg->fc_flow = nla_get_u32(attr);
661 cfg->fc_table = nla_get_u32(attr);
664 cfg->fc_encap = attr;
667 cfg->fc_encap_type = nla_get_u16(attr);
677 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
679 struct net *net = sock_net(skb->sk);
680 struct fib_config cfg;
681 struct fib_table *tb;
684 err = rtm_to_fib_config(net, skb, nlh, &cfg);
688 tb = fib_get_table(net, cfg.fc_table);
694 err = fib_table_delete(tb, &cfg);
699 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
701 struct net *net = sock_net(skb->sk);
702 struct fib_config cfg;
703 struct fib_table *tb;
706 err = rtm_to_fib_config(net, skb, nlh, &cfg);
710 tb = fib_new_table(net, cfg.fc_table);
716 err = fib_table_insert(tb, &cfg);
721 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
723 struct net *net = sock_net(skb->sk);
725 unsigned int e = 0, s_e;
726 struct fib_table *tb;
727 struct hlist_head *head;
730 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
731 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
739 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
741 head = &net->ipv4.fib_table_hash[h];
742 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
746 memset(&cb->args[2], 0, sizeof(cb->args) -
747 2 * sizeof(cb->args[0]));
748 if (fib_table_dump(tb, skb, cb) < 0)
764 /* Prepare and feed intra-kernel routing request.
765 * Really, it should be netlink message, but :-( netlink
766 * can be not configured, so that we feed it directly
767 * to fib engine. It is legal, because all events occur
768 * only when netlink is already locked.
770 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
772 struct net *net = dev_net(ifa->ifa_dev->dev);
773 struct fib_table *tb;
774 struct fib_config cfg = {
775 .fc_protocol = RTPROT_KERNEL,
778 .fc_dst_len = dst_len,
779 .fc_prefsrc = ifa->ifa_local,
780 .fc_oif = ifa->ifa_dev->dev->ifindex,
781 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
787 if (type == RTN_UNICAST)
788 tb = fib_new_table(net, RT_TABLE_MAIN);
790 tb = fib_new_table(net, RT_TABLE_LOCAL);
795 cfg.fc_table = tb->tb_id;
797 if (type != RTN_LOCAL)
798 cfg.fc_scope = RT_SCOPE_LINK;
800 cfg.fc_scope = RT_SCOPE_HOST;
802 if (cmd == RTM_NEWROUTE)
803 fib_table_insert(tb, &cfg);
805 fib_table_delete(tb, &cfg);
808 void fib_add_ifaddr(struct in_ifaddr *ifa)
810 struct in_device *in_dev = ifa->ifa_dev;
811 struct net_device *dev = in_dev->dev;
812 struct in_ifaddr *prim = ifa;
813 __be32 mask = ifa->ifa_mask;
814 __be32 addr = ifa->ifa_local;
815 __be32 prefix = ifa->ifa_address & mask;
817 if (ifa->ifa_flags & IFA_F_SECONDARY) {
818 prim = inet_ifa_byprefix(in_dev, prefix, mask);
820 pr_warn("%s: bug: prim == NULL\n", __func__);
825 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
827 if (!(dev->flags & IFF_UP))
830 /* Add broadcast address, if it is explicitly assigned. */
831 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
832 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
834 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
835 (prefix != addr || ifa->ifa_prefixlen < 32)) {
836 fib_magic(RTM_NEWROUTE,
837 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
838 prefix, ifa->ifa_prefixlen, prim);
840 /* Add network specific broadcasts, when it takes a sense */
841 if (ifa->ifa_prefixlen < 31) {
842 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
843 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
849 /* Delete primary or secondary address.
850 * Optionally, on secondary address promotion consider the addresses
851 * from subnet iprim as deleted, even if they are in device list.
852 * In this case the secondary ifa can be in device list.
854 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
856 struct in_device *in_dev = ifa->ifa_dev;
857 struct net_device *dev = in_dev->dev;
858 struct in_ifaddr *ifa1;
859 struct in_ifaddr *prim = ifa, *prim1 = NULL;
860 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
861 __be32 any = ifa->ifa_address & ifa->ifa_mask;
867 int subnet = 0; /* Primary network */
868 int gone = 1; /* Address is missing */
869 int same_prefsrc = 0; /* Another primary with same IP */
871 if (ifa->ifa_flags & IFA_F_SECONDARY) {
872 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
874 pr_warn("%s: bug: prim == NULL\n", __func__);
877 if (iprim && iprim != prim) {
878 pr_warn("%s: bug: iprim != prim\n", __func__);
881 } else if (!ipv4_is_zeronet(any) &&
882 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
883 fib_magic(RTM_DELROUTE,
884 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
885 any, ifa->ifa_prefixlen, prim);
889 /* Deletion is more complicated than add.
890 * We should take care of not to delete too much :-)
892 * Scan address list to be sure that addresses are really gone.
895 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
897 /* promotion, keep the IP */
901 /* Ignore IFAs from our subnet */
902 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
903 inet_ifa_match(ifa1->ifa_address, iprim))
906 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
907 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
908 /* Another address from our subnet? */
909 if (ifa1->ifa_mask == prim->ifa_mask &&
910 inet_ifa_match(ifa1->ifa_address, prim))
913 /* We reached the secondaries, so
914 * same_prefsrc should be determined.
918 /* Search new prim1 if ifa1 is not
919 * using the current prim1
922 ifa1->ifa_mask != prim1->ifa_mask ||
923 !inet_ifa_match(ifa1->ifa_address, prim1))
924 prim1 = inet_ifa_byprefix(in_dev,
929 if (prim1->ifa_local != prim->ifa_local)
933 if (prim->ifa_local != ifa1->ifa_local)
939 if (ifa->ifa_local == ifa1->ifa_local)
941 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
943 if (brd == ifa1->ifa_broadcast)
945 if (any == ifa1->ifa_broadcast)
947 /* primary has network specific broadcasts */
948 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
949 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
950 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
952 if (!ipv4_is_zeronet(any1)) {
953 if (ifa->ifa_broadcast == brd1 ||
954 ifa->ifa_broadcast == any1)
956 if (brd == brd1 || brd == any1)
958 if (any == brd1 || any == any1)
965 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
966 if (subnet && ifa->ifa_prefixlen < 31) {
968 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
970 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
972 if (!(ok & LOCAL_OK)) {
973 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
975 /* Check, that this local address finally disappeared. */
977 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
978 /* And the last, but not the least thing.
979 * We must flush stray FIB entries.
981 * First of all, we scan fib_info list searching
982 * for stray nexthop entries, then ignite fib_flush.
984 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
985 fib_flush(dev_net(dev));
994 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
997 struct fib_result res;
998 struct flowi4 fl4 = {
999 .flowi4_mark = frn->fl_mark,
1000 .daddr = frn->fl_addr,
1001 .flowi4_tos = frn->fl_tos,
1002 .flowi4_scope = frn->fl_scope,
1004 struct fib_table *tb;
1008 tb = fib_get_table(net, frn->tb_id_in);
1014 frn->tb_id = tb->tb_id;
1015 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1018 frn->prefixlen = res.prefixlen;
1019 frn->nh_sel = res.nh_sel;
1020 frn->type = res.type;
1021 frn->scope = res.scope;
1029 static void nl_fib_input(struct sk_buff *skb)
1032 struct fib_result_nl *frn;
1033 struct nlmsghdr *nlh;
1036 net = sock_net(skb->sk);
1037 nlh = nlmsg_hdr(skb);
1038 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1039 nlmsg_len(nlh) < sizeof(*frn))
1042 skb = netlink_skb_clone(skb, GFP_KERNEL);
1045 nlh = nlmsg_hdr(skb);
1047 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1048 nl_fib_lookup(net, frn);
1050 portid = NETLINK_CB(skb).portid; /* netlink portid */
1051 NETLINK_CB(skb).portid = 0; /* from kernel */
1052 NETLINK_CB(skb).dst_group = 0; /* unicast */
1053 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1056 static int __net_init nl_fib_lookup_init(struct net *net)
1059 struct netlink_kernel_cfg cfg = {
1060 .input = nl_fib_input,
1063 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1065 return -EAFNOSUPPORT;
1066 net->ipv4.fibnl = sk;
1070 static void nl_fib_lookup_exit(struct net *net)
1072 netlink_kernel_release(net->ipv4.fibnl);
1073 net->ipv4.fibnl = NULL;
1076 static void fib_disable_ip(struct net_device *dev, unsigned long event)
1078 if (fib_sync_down_dev(dev, event))
1079 fib_flush(dev_net(dev));
1080 rt_cache_flush(dev_net(dev));
1084 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1086 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1087 struct net_device *dev = ifa->ifa_dev->dev;
1088 struct net *net = dev_net(dev);
1092 fib_add_ifaddr(ifa);
1093 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1094 fib_sync_up(dev, RTNH_F_DEAD);
1096 atomic_inc(&net->ipv4.dev_addr_genid);
1097 rt_cache_flush(dev_net(dev));
1100 fib_del_ifaddr(ifa, NULL);
1101 atomic_inc(&net->ipv4.dev_addr_genid);
1102 if (!ifa->ifa_dev->ifa_list) {
1103 /* Last address was deleted from this interface.
1106 fib_disable_ip(dev, event);
1108 rt_cache_flush(dev_net(dev));
1115 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1117 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1118 struct in_device *in_dev;
1119 struct net *net = dev_net(dev);
1122 if (event == NETDEV_UNREGISTER) {
1123 fib_disable_ip(dev, event);
1128 in_dev = __in_dev_get_rtnl(dev);
1135 fib_add_ifaddr(ifa);
1136 } endfor_ifa(in_dev);
1137 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1138 fib_sync_up(dev, RTNH_F_DEAD);
1140 atomic_inc(&net->ipv4.dev_addr_genid);
1141 rt_cache_flush(net);
1144 fib_disable_ip(dev, event);
1147 flags = dev_get_flags(dev);
1148 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1149 fib_sync_up(dev, RTNH_F_LINKDOWN);
1151 fib_sync_down_dev(dev, event);
1153 case NETDEV_CHANGEMTU:
1154 rt_cache_flush(net);
1160 static struct notifier_block fib_inetaddr_notifier = {
1161 .notifier_call = fib_inetaddr_event,
1164 static struct notifier_block fib_netdev_notifier = {
1165 .notifier_call = fib_netdev_event,
1168 static int __net_init ip_fib_net_init(struct net *net)
1171 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1173 /* Avoid false sharing : Use at least a full cache line */
1174 size = max_t(size_t, size, L1_CACHE_BYTES);
1176 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1177 if (!net->ipv4.fib_table_hash)
1180 err = fib4_rules_init(net);
1186 kfree(net->ipv4.fib_table_hash);
1190 static void ip_fib_net_exit(struct net *net)
1195 #ifdef CONFIG_IP_MULTIPLE_TABLES
1196 RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1197 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1198 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1200 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1201 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1202 struct hlist_node *tmp;
1203 struct fib_table *tb;
1205 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1206 hlist_del(&tb->tb_hlist);
1207 fib_table_flush(tb);
1212 #ifdef CONFIG_IP_MULTIPLE_TABLES
1213 fib4_rules_exit(net);
1216 kfree(net->ipv4.fib_table_hash);
1219 static int __net_init fib_net_init(struct net *net)
1223 #ifdef CONFIG_IP_ROUTE_CLASSID
1224 net->ipv4.fib_num_tclassid_users = 0;
1226 error = ip_fib_net_init(net);
1229 error = nl_fib_lookup_init(net);
1232 error = fib_proc_init(net);
1239 nl_fib_lookup_exit(net);
1241 ip_fib_net_exit(net);
1245 static void __net_exit fib_net_exit(struct net *net)
1248 nl_fib_lookup_exit(net);
1249 ip_fib_net_exit(net);
1252 static struct pernet_operations fib_net_ops = {
1253 .init = fib_net_init,
1254 .exit = fib_net_exit,
1257 void __init ip_fib_init(void)
1259 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1260 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1261 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1263 register_pernet_subsys(&fib_net_ops);
1264 register_netdevice_notifier(&fib_netdev_notifier);
1265 register_inetaddr_notifier(&fib_inetaddr_notifier);