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/init.h>
35 #include <linux/list.h>
36 #include <linux/slab.h>
39 #include <net/protocol.h>
40 #include <net/route.h>
44 #include <net/ip_fib.h>
45 #include <net/rtnetlink.h>
48 #ifndef CONFIG_IP_MULTIPLE_TABLES
50 static int __net_init fib4_rules_init(struct net *net)
52 struct fib_table *local_table, *main_table;
54 local_table = fib_trie_table(RT_TABLE_LOCAL);
55 if (local_table == NULL)
58 main_table = fib_trie_table(RT_TABLE_MAIN);
59 if (main_table == NULL)
62 hlist_add_head_rcu(&local_table->tb_hlist,
63 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
64 hlist_add_head_rcu(&main_table->tb_hlist,
65 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
74 struct fib_table *fib_new_table(struct net *net, u32 id)
81 tb = fib_get_table(net, id);
85 tb = fib_trie_table(id);
88 h = id & (FIB_TABLE_HASHSZ - 1);
89 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
93 struct fib_table *fib_get_table(struct net *net, u32 id)
96 struct hlist_node *node;
97 struct hlist_head *head;
102 h = id & (FIB_TABLE_HASHSZ - 1);
105 head = &net->ipv4.fib_table_hash[h];
106 hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
107 if (tb->tb_id == id) {
115 #endif /* CONFIG_IP_MULTIPLE_TABLES */
117 static void fib_flush(struct net *net)
120 struct fib_table *tb;
121 struct hlist_node *node;
122 struct hlist_head *head;
125 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
126 head = &net->ipv4.fib_table_hash[h];
127 hlist_for_each_entry(tb, node, head, tb_hlist)
128 flushed += fib_table_flush(tb);
132 rt_cache_flush(net, -1);
136 * Find address type as if only "dev" was present in the system. If
137 * on_dev is NULL then all interfaces are taken into consideration.
139 static inline unsigned int __inet_dev_addr_type(struct net *net,
140 const struct net_device *dev,
143 struct flowi4 fl4 = { .daddr = addr };
144 struct fib_result res;
145 unsigned int ret = RTN_BROADCAST;
146 struct fib_table *local_table;
148 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
149 return RTN_BROADCAST;
150 if (ipv4_is_multicast(addr))
151 return RTN_MULTICAST;
153 #ifdef CONFIG_IP_MULTIPLE_TABLES
157 local_table = fib_get_table(net, RT_TABLE_LOCAL);
161 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
162 if (!dev || dev == res.fi->fib_dev)
170 unsigned int inet_addr_type(struct net *net, __be32 addr)
172 return __inet_dev_addr_type(net, NULL, addr);
174 EXPORT_SYMBOL(inet_addr_type);
176 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
179 return __inet_dev_addr_type(net, dev, addr);
181 EXPORT_SYMBOL(inet_dev_addr_type);
183 __be32 fib_compute_spec_dst(struct sk_buff *skb)
185 struct net_device *dev = skb->dev;
186 struct in_device *in_dev;
187 struct fib_result res;
191 if (skb->pkt_type != PACKET_BROADCAST &&
192 skb->pkt_type != PACKET_MULTICAST)
193 return ip_hdr(skb)->daddr;
195 in_dev = __in_dev_get_rcu(dev);
199 fl4.daddr = ip_hdr(skb)->saddr;
200 fl4.saddr = ip_hdr(skb)->daddr;
201 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
202 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
203 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
206 if (!fib_lookup(net, &fl4, &res))
207 return FIB_RES_PREFSRC(net, res);
209 return inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
212 /* Given (packet source, input interface) and optional (dst, oif, tos):
213 * - (main) check, that source is valid i.e. not broadcast or our local
215 * - figure out what "logical" interface this packet arrived
216 * and calculate "specific destination" address.
217 * - check, that packet arrived from expected physical interface.
218 * called with rcu_read_lock()
220 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, u8 tos,
221 int oif, struct net_device *dev, u32 *itag)
223 struct in_device *in_dev;
225 struct fib_result res;
226 int no_addr, rpf, accept_local;
232 fl4.flowi4_iif = oif;
235 fl4.flowi4_tos = tos;
236 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
238 no_addr = rpf = accept_local = 0;
239 in_dev = __in_dev_get_rcu(dev);
241 no_addr = in_dev->ifa_list == NULL;
243 /* Ignore rp_filter for packets protected by IPsec. */
244 rpf = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(in_dev);
246 accept_local = IN_DEV_ACCEPT_LOCAL(in_dev);
247 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
254 if (fib_lookup(net, &fl4, &res))
256 if (res.type != RTN_UNICAST) {
257 if (res.type != RTN_LOCAL || !accept_local)
260 fib_combine_itag(itag, &res);
263 #ifdef CONFIG_IP_ROUTE_MULTIPATH
264 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
265 struct fib_nh *nh = &res.fi->fib_nh[ret];
267 if (nh->nh_dev == dev) {
273 if (FIB_RES_DEV(res) == dev)
277 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
284 fl4.flowi4_oif = dev->ifindex;
287 if (fib_lookup(net, &fl4, &res) == 0) {
288 if (res.type == RTN_UNICAST)
289 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
305 static inline __be32 sk_extract_addr(struct sockaddr *addr)
307 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
310 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
314 nla = (struct nlattr *) ((char *) mx + len);
315 nla->nla_type = type;
316 nla->nla_len = nla_attr_size(4);
317 *(u32 *) nla_data(nla) = value;
319 return len + nla_total_size(4);
322 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
323 struct fib_config *cfg)
328 memset(cfg, 0, sizeof(*cfg));
329 cfg->fc_nlinfo.nl_net = net;
331 if (rt->rt_dst.sa_family != AF_INET)
332 return -EAFNOSUPPORT;
335 * Check mask for validity:
336 * a) it must be contiguous.
337 * b) destination must have all host bits clear.
338 * c) if application forgot to set correct family (AF_INET),
339 * reject request unless it is absolutely clear i.e.
340 * both family and mask are zero.
343 addr = sk_extract_addr(&rt->rt_dst);
344 if (!(rt->rt_flags & RTF_HOST)) {
345 __be32 mask = sk_extract_addr(&rt->rt_genmask);
347 if (rt->rt_genmask.sa_family != AF_INET) {
348 if (mask || rt->rt_genmask.sa_family)
349 return -EAFNOSUPPORT;
352 if (bad_mask(mask, addr))
355 plen = inet_mask_len(mask);
358 cfg->fc_dst_len = plen;
361 if (cmd != SIOCDELRT) {
362 cfg->fc_nlflags = NLM_F_CREATE;
363 cfg->fc_protocol = RTPROT_BOOT;
367 cfg->fc_priority = rt->rt_metric - 1;
369 if (rt->rt_flags & RTF_REJECT) {
370 cfg->fc_scope = RT_SCOPE_HOST;
371 cfg->fc_type = RTN_UNREACHABLE;
375 cfg->fc_scope = RT_SCOPE_NOWHERE;
376 cfg->fc_type = RTN_UNICAST;
380 struct net_device *dev;
381 char devname[IFNAMSIZ];
383 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
386 devname[IFNAMSIZ-1] = 0;
387 colon = strchr(devname, ':');
390 dev = __dev_get_by_name(net, devname);
393 cfg->fc_oif = dev->ifindex;
395 struct in_ifaddr *ifa;
396 struct in_device *in_dev = __in_dev_get_rtnl(dev);
400 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
401 if (strcmp(ifa->ifa_label, devname) == 0)
405 cfg->fc_prefsrc = ifa->ifa_local;
409 addr = sk_extract_addr(&rt->rt_gateway);
410 if (rt->rt_gateway.sa_family == AF_INET && addr) {
412 if (rt->rt_flags & RTF_GATEWAY &&
413 inet_addr_type(net, addr) == RTN_UNICAST)
414 cfg->fc_scope = RT_SCOPE_UNIVERSE;
417 if (cmd == SIOCDELRT)
420 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
423 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
424 cfg->fc_scope = RT_SCOPE_LINK;
426 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
430 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
434 if (rt->rt_flags & RTF_MTU)
435 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
437 if (rt->rt_flags & RTF_WINDOW)
438 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
440 if (rt->rt_flags & RTF_IRTT)
441 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
444 cfg->fc_mx_len = len;
451 * Handle IP routing ioctl calls.
452 * These are used to manipulate the routing tables
454 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
456 struct fib_config cfg;
461 case SIOCADDRT: /* Add a route */
462 case SIOCDELRT: /* Delete a route */
463 if (!capable(CAP_NET_ADMIN))
466 if (copy_from_user(&rt, arg, sizeof(rt)))
470 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
472 struct fib_table *tb;
474 if (cmd == SIOCDELRT) {
475 tb = fib_get_table(net, cfg.fc_table);
477 err = fib_table_delete(tb, &cfg);
481 tb = fib_new_table(net, cfg.fc_table);
483 err = fib_table_insert(tb, &cfg);
488 /* allocated by rtentry_to_fib_config() */
497 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
498 [RTA_DST] = { .type = NLA_U32 },
499 [RTA_SRC] = { .type = NLA_U32 },
500 [RTA_IIF] = { .type = NLA_U32 },
501 [RTA_OIF] = { .type = NLA_U32 },
502 [RTA_GATEWAY] = { .type = NLA_U32 },
503 [RTA_PRIORITY] = { .type = NLA_U32 },
504 [RTA_PREFSRC] = { .type = NLA_U32 },
505 [RTA_METRICS] = { .type = NLA_NESTED },
506 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
507 [RTA_FLOW] = { .type = NLA_U32 },
510 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
511 struct nlmsghdr *nlh, struct fib_config *cfg)
517 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
521 memset(cfg, 0, sizeof(*cfg));
523 rtm = nlmsg_data(nlh);
524 cfg->fc_dst_len = rtm->rtm_dst_len;
525 cfg->fc_tos = rtm->rtm_tos;
526 cfg->fc_table = rtm->rtm_table;
527 cfg->fc_protocol = rtm->rtm_protocol;
528 cfg->fc_scope = rtm->rtm_scope;
529 cfg->fc_type = rtm->rtm_type;
530 cfg->fc_flags = rtm->rtm_flags;
531 cfg->fc_nlflags = nlh->nlmsg_flags;
533 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
534 cfg->fc_nlinfo.nlh = nlh;
535 cfg->fc_nlinfo.nl_net = net;
537 if (cfg->fc_type > RTN_MAX) {
542 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
543 switch (nla_type(attr)) {
545 cfg->fc_dst = nla_get_be32(attr);
548 cfg->fc_oif = nla_get_u32(attr);
551 cfg->fc_gw = nla_get_be32(attr);
554 cfg->fc_priority = nla_get_u32(attr);
557 cfg->fc_prefsrc = nla_get_be32(attr);
560 cfg->fc_mx = nla_data(attr);
561 cfg->fc_mx_len = nla_len(attr);
564 cfg->fc_mp = nla_data(attr);
565 cfg->fc_mp_len = nla_len(attr);
568 cfg->fc_flow = nla_get_u32(attr);
571 cfg->fc_table = nla_get_u32(attr);
581 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
583 struct net *net = sock_net(skb->sk);
584 struct fib_config cfg;
585 struct fib_table *tb;
588 err = rtm_to_fib_config(net, skb, nlh, &cfg);
592 tb = fib_get_table(net, cfg.fc_table);
598 err = fib_table_delete(tb, &cfg);
603 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
605 struct net *net = sock_net(skb->sk);
606 struct fib_config cfg;
607 struct fib_table *tb;
610 err = rtm_to_fib_config(net, skb, nlh, &cfg);
614 tb = fib_new_table(net, cfg.fc_table);
620 err = fib_table_insert(tb, &cfg);
625 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
627 struct net *net = sock_net(skb->sk);
629 unsigned int e = 0, s_e;
630 struct fib_table *tb;
631 struct hlist_node *node;
632 struct hlist_head *head;
635 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
636 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
637 return ip_rt_dump(skb, cb);
642 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
644 head = &net->ipv4.fib_table_hash[h];
645 hlist_for_each_entry(tb, node, head, tb_hlist) {
649 memset(&cb->args[2], 0, sizeof(cb->args) -
650 2 * sizeof(cb->args[0]));
651 if (fib_table_dump(tb, skb, cb) < 0)
665 /* Prepare and feed intra-kernel routing request.
666 * Really, it should be netlink message, but :-( netlink
667 * can be not configured, so that we feed it directly
668 * to fib engine. It is legal, because all events occur
669 * only when netlink is already locked.
671 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
673 struct net *net = dev_net(ifa->ifa_dev->dev);
674 struct fib_table *tb;
675 struct fib_config cfg = {
676 .fc_protocol = RTPROT_KERNEL,
679 .fc_dst_len = dst_len,
680 .fc_prefsrc = ifa->ifa_local,
681 .fc_oif = ifa->ifa_dev->dev->ifindex,
682 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
688 if (type == RTN_UNICAST)
689 tb = fib_new_table(net, RT_TABLE_MAIN);
691 tb = fib_new_table(net, RT_TABLE_LOCAL);
696 cfg.fc_table = tb->tb_id;
698 if (type != RTN_LOCAL)
699 cfg.fc_scope = RT_SCOPE_LINK;
701 cfg.fc_scope = RT_SCOPE_HOST;
703 if (cmd == RTM_NEWROUTE)
704 fib_table_insert(tb, &cfg);
706 fib_table_delete(tb, &cfg);
709 void fib_add_ifaddr(struct in_ifaddr *ifa)
711 struct in_device *in_dev = ifa->ifa_dev;
712 struct net_device *dev = in_dev->dev;
713 struct in_ifaddr *prim = ifa;
714 __be32 mask = ifa->ifa_mask;
715 __be32 addr = ifa->ifa_local;
716 __be32 prefix = ifa->ifa_address & mask;
718 if (ifa->ifa_flags & IFA_F_SECONDARY) {
719 prim = inet_ifa_byprefix(in_dev, prefix, mask);
721 pr_warn("%s: bug: prim == NULL\n", __func__);
726 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
728 if (!(dev->flags & IFF_UP))
731 /* Add broadcast address, if it is explicitly assigned. */
732 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
733 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
735 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
736 (prefix != addr || ifa->ifa_prefixlen < 32)) {
737 fib_magic(RTM_NEWROUTE,
738 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
739 prefix, ifa->ifa_prefixlen, prim);
741 /* Add network specific broadcasts, when it takes a sense */
742 if (ifa->ifa_prefixlen < 31) {
743 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
744 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
750 /* Delete primary or secondary address.
751 * Optionally, on secondary address promotion consider the addresses
752 * from subnet iprim as deleted, even if they are in device list.
753 * In this case the secondary ifa can be in device list.
755 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
757 struct in_device *in_dev = ifa->ifa_dev;
758 struct net_device *dev = in_dev->dev;
759 struct in_ifaddr *ifa1;
760 struct in_ifaddr *prim = ifa, *prim1 = NULL;
761 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
762 __be32 any = ifa->ifa_address & ifa->ifa_mask;
768 int subnet = 0; /* Primary network */
769 int gone = 1; /* Address is missing */
770 int same_prefsrc = 0; /* Another primary with same IP */
772 if (ifa->ifa_flags & IFA_F_SECONDARY) {
773 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
775 pr_warn("%s: bug: prim == NULL\n", __func__);
778 if (iprim && iprim != prim) {
779 pr_warn("%s: bug: iprim != prim\n", __func__);
782 } else if (!ipv4_is_zeronet(any) &&
783 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
784 fib_magic(RTM_DELROUTE,
785 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
786 any, ifa->ifa_prefixlen, prim);
790 /* Deletion is more complicated than add.
791 * We should take care of not to delete too much :-)
793 * Scan address list to be sure that addresses are really gone.
796 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
798 /* promotion, keep the IP */
802 /* Ignore IFAs from our subnet */
803 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
804 inet_ifa_match(ifa1->ifa_address, iprim))
807 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
808 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
809 /* Another address from our subnet? */
810 if (ifa1->ifa_mask == prim->ifa_mask &&
811 inet_ifa_match(ifa1->ifa_address, prim))
814 /* We reached the secondaries, so
815 * same_prefsrc should be determined.
819 /* Search new prim1 if ifa1 is not
820 * using the current prim1
823 ifa1->ifa_mask != prim1->ifa_mask ||
824 !inet_ifa_match(ifa1->ifa_address, prim1))
825 prim1 = inet_ifa_byprefix(in_dev,
830 if (prim1->ifa_local != prim->ifa_local)
834 if (prim->ifa_local != ifa1->ifa_local)
840 if (ifa->ifa_local == ifa1->ifa_local)
842 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
844 if (brd == ifa1->ifa_broadcast)
846 if (any == ifa1->ifa_broadcast)
848 /* primary has network specific broadcasts */
849 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
850 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
851 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
853 if (!ipv4_is_zeronet(any1)) {
854 if (ifa->ifa_broadcast == brd1 ||
855 ifa->ifa_broadcast == any1)
857 if (brd == brd1 || brd == any1)
859 if (any == brd1 || any == any1)
866 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
867 if (subnet && ifa->ifa_prefixlen < 31) {
869 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
871 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
873 if (!(ok & LOCAL_OK)) {
874 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
876 /* Check, that this local address finally disappeared. */
878 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
879 /* And the last, but not the least thing.
880 * We must flush stray FIB entries.
882 * First of all, we scan fib_info list searching
883 * for stray nexthop entries, then ignite fib_flush.
885 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
886 fib_flush(dev_net(dev));
895 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
898 struct fib_result res;
899 struct flowi4 fl4 = {
900 .flowi4_mark = frn->fl_mark,
901 .daddr = frn->fl_addr,
902 .flowi4_tos = frn->fl_tos,
903 .flowi4_scope = frn->fl_scope,
906 #ifdef CONFIG_IP_MULTIPLE_TABLES
914 frn->tb_id = tb->tb_id;
916 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
919 frn->prefixlen = res.prefixlen;
920 frn->nh_sel = res.nh_sel;
921 frn->type = res.type;
922 frn->scope = res.scope;
929 static void nl_fib_input(struct sk_buff *skb)
932 struct fib_result_nl *frn;
933 struct nlmsghdr *nlh;
934 struct fib_table *tb;
937 net = sock_net(skb->sk);
938 nlh = nlmsg_hdr(skb);
939 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
940 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
943 skb = skb_clone(skb, GFP_KERNEL);
946 nlh = nlmsg_hdr(skb);
948 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
949 tb = fib_get_table(net, frn->tb_id_in);
951 nl_fib_lookup(frn, tb);
953 pid = NETLINK_CB(skb).pid; /* pid of sending process */
954 NETLINK_CB(skb).pid = 0; /* from kernel */
955 NETLINK_CB(skb).dst_group = 0; /* unicast */
956 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
959 static int __net_init nl_fib_lookup_init(struct net *net)
962 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
963 nl_fib_input, NULL, THIS_MODULE);
965 return -EAFNOSUPPORT;
966 net->ipv4.fibnl = sk;
970 static void nl_fib_lookup_exit(struct net *net)
972 netlink_kernel_release(net->ipv4.fibnl);
973 net->ipv4.fibnl = NULL;
976 static void fib_disable_ip(struct net_device *dev, int force, int delay)
978 if (fib_sync_down_dev(dev, force))
979 fib_flush(dev_net(dev));
980 rt_cache_flush(dev_net(dev), delay);
984 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
986 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
987 struct net_device *dev = ifa->ifa_dev->dev;
988 struct net *net = dev_net(dev);
993 #ifdef CONFIG_IP_ROUTE_MULTIPATH
996 atomic_inc(&net->ipv4.dev_addr_genid);
997 rt_cache_flush(dev_net(dev), -1);
1000 fib_del_ifaddr(ifa, NULL);
1001 atomic_inc(&net->ipv4.dev_addr_genid);
1002 if (ifa->ifa_dev->ifa_list == NULL) {
1003 /* Last address was deleted from this interface.
1006 fib_disable_ip(dev, 1, 0);
1008 rt_cache_flush(dev_net(dev), -1);
1015 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1017 struct net_device *dev = ptr;
1018 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1019 struct net *net = dev_net(dev);
1021 if (event == NETDEV_UNREGISTER) {
1022 fib_disable_ip(dev, 2, -1);
1032 fib_add_ifaddr(ifa);
1033 } endfor_ifa(in_dev);
1034 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1037 atomic_inc(&net->ipv4.dev_addr_genid);
1038 rt_cache_flush(dev_net(dev), -1);
1041 fib_disable_ip(dev, 0, 0);
1043 case NETDEV_CHANGEMTU:
1045 rt_cache_flush(dev_net(dev), 0);
1047 case NETDEV_UNREGISTER_BATCH:
1048 /* The batch unregister is only called on the first
1049 * device in the list of devices being unregistered.
1050 * Therefore we should not pass dev_net(dev) in here.
1052 rt_cache_flush_batch(NULL);
1058 static struct notifier_block fib_inetaddr_notifier = {
1059 .notifier_call = fib_inetaddr_event,
1062 static struct notifier_block fib_netdev_notifier = {
1063 .notifier_call = fib_netdev_event,
1066 static int __net_init ip_fib_net_init(struct net *net)
1069 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1071 /* Avoid false sharing : Use at least a full cache line */
1072 size = max_t(size_t, size, L1_CACHE_BYTES);
1074 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1075 if (net->ipv4.fib_table_hash == NULL)
1078 err = fib4_rules_init(net);
1084 kfree(net->ipv4.fib_table_hash);
1088 static void ip_fib_net_exit(struct net *net)
1092 #ifdef CONFIG_IP_MULTIPLE_TABLES
1093 fib4_rules_exit(net);
1097 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1098 struct fib_table *tb;
1099 struct hlist_head *head;
1100 struct hlist_node *node, *tmp;
1102 head = &net->ipv4.fib_table_hash[i];
1103 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1105 fib_table_flush(tb);
1110 kfree(net->ipv4.fib_table_hash);
1113 static int __net_init fib_net_init(struct net *net)
1117 error = ip_fib_net_init(net);
1120 error = nl_fib_lookup_init(net);
1123 error = fib_proc_init(net);
1130 nl_fib_lookup_exit(net);
1132 ip_fib_net_exit(net);
1136 static void __net_exit fib_net_exit(struct net *net)
1139 nl_fib_lookup_exit(net);
1140 ip_fib_net_exit(net);
1143 static struct pernet_operations fib_net_ops = {
1144 .init = fib_net_init,
1145 .exit = fib_net_exit,
1148 void __init ip_fib_init(void)
1150 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1151 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1152 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1154 register_pernet_subsys(&fib_net_ops);
1155 register_netdevice_notifier(&fib_netdev_notifier);
1156 register_inetaddr_notifier(&fib_inetaddr_notifier);