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 local_table = fib_trie_table(RT_TABLE_LOCAL);
56 if (local_table == NULL)
59 main_table = fib_trie_table(RT_TABLE_MAIN);
60 if (main_table == NULL)
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]);
75 struct fib_table *fib_new_table(struct net *net, u32 id)
82 tb = fib_get_table(net, id);
86 tb = fib_trie_table(id);
89 h = id & (FIB_TABLE_HASHSZ - 1);
90 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
94 struct fib_table *fib_get_table(struct net *net, u32 id)
97 struct hlist_node *node;
98 struct hlist_head *head;
103 h = id & (FIB_TABLE_HASHSZ - 1);
106 head = &net->ipv4.fib_table_hash[h];
107 hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
108 if (tb->tb_id == id) {
116 #endif /* CONFIG_IP_MULTIPLE_TABLES */
118 static void fib_flush(struct net *net)
121 struct fib_table *tb;
122 struct hlist_node *node;
123 struct hlist_head *head;
126 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
127 head = &net->ipv4.fib_table_hash[h];
128 hlist_for_each_entry(tb, node, head, tb_hlist)
129 flushed += fib_table_flush(tb);
133 rt_cache_flush(net, -1);
137 * Find address type as if only "dev" was present in the system. If
138 * on_dev is NULL then all interfaces are taken into consideration.
140 static inline unsigned int __inet_dev_addr_type(struct net *net,
141 const struct net_device *dev,
144 struct flowi4 fl4 = { .daddr = addr };
145 struct fib_result res;
146 unsigned int ret = RTN_BROADCAST;
147 struct fib_table *local_table;
149 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
150 return RTN_BROADCAST;
151 if (ipv4_is_multicast(addr))
152 return RTN_MULTICAST;
154 #ifdef CONFIG_IP_MULTIPLE_TABLES
158 local_table = fib_get_table(net, RT_TABLE_LOCAL);
162 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
163 if (!dev || dev == res.fi->fib_dev)
171 unsigned int inet_addr_type(struct net *net, __be32 addr)
173 return __inet_dev_addr_type(net, NULL, addr);
175 EXPORT_SYMBOL(inet_addr_type);
177 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
180 return __inet_dev_addr_type(net, dev, addr);
182 EXPORT_SYMBOL(inet_dev_addr_type);
184 __be32 fib_compute_spec_dst(struct sk_buff *skb)
186 struct net_device *dev = skb->dev;
187 struct in_device *in_dev;
188 struct fib_result res;
194 rt = skb_rtable(skb);
195 if (!(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)))
196 return ip_hdr(skb)->daddr;
198 in_dev = __in_dev_get_rcu(dev);
203 scope = RT_SCOPE_UNIVERSE;
204 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
206 fl4.flowi4_iif = net->loopback_dev->ifindex;
207 fl4.daddr = ip_hdr(skb)->saddr;
209 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
210 fl4.flowi4_scope = scope;
211 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
212 if (!fib_lookup(net, &fl4, &res))
213 return FIB_RES_PREFSRC(net, res);
215 scope = RT_SCOPE_LINK;
218 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
221 #ifdef CONFIG_IP_ROUTE_CLASSID
222 int fib_num_tclassid_users __read_mostly;
225 /* Given (packet source, input interface) and optional (dst, oif, tos):
226 * - (main) check, that source is valid i.e. not broadcast or our local
228 * - figure out what "logical" interface this packet arrived
229 * and calculate "specific destination" address.
230 * - check, that packet arrived from expected physical interface.
231 * called with rcu_read_lock()
233 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
234 u8 tos, int oif, struct net_device *dev,
235 int rpf, struct in_device *idev, u32 *itag)
237 int ret, no_addr, accept_local;
238 struct fib_result res;
244 fl4.flowi4_iif = oif;
247 fl4.flowi4_tos = tos;
248 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
250 no_addr = accept_local = 0;
251 no_addr = idev->ifa_list == NULL;
253 accept_local = IN_DEV_ACCEPT_LOCAL(idev);
254 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
257 if (fib_lookup(net, &fl4, &res))
259 if (res.type != RTN_UNICAST) {
260 if (res.type != RTN_LOCAL || !accept_local)
263 fib_combine_itag(itag, &res);
266 #ifdef CONFIG_IP_ROUTE_MULTIPATH
267 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
268 struct fib_nh *nh = &res.fi->fib_nh[ret];
270 if (nh->nh_dev == dev) {
276 if (FIB_RES_DEV(res) == dev)
280 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
287 fl4.flowi4_oif = dev->ifindex;
290 if (fib_lookup(net, &fl4, &res) == 0) {
291 if (res.type == RTN_UNICAST)
292 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
308 /* Ignore rp_filter for packets protected by IPsec. */
309 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
310 u8 tos, int oif, struct net_device *dev,
311 struct in_device *idev, u32 *itag)
313 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
315 if (!r && !fib_num_tclassid_users) {
319 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
322 static inline __be32 sk_extract_addr(struct sockaddr *addr)
324 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
327 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
331 nla = (struct nlattr *) ((char *) mx + len);
332 nla->nla_type = type;
333 nla->nla_len = nla_attr_size(4);
334 *(u32 *) nla_data(nla) = value;
336 return len + nla_total_size(4);
339 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
340 struct fib_config *cfg)
345 memset(cfg, 0, sizeof(*cfg));
346 cfg->fc_nlinfo.nl_net = net;
348 if (rt->rt_dst.sa_family != AF_INET)
349 return -EAFNOSUPPORT;
352 * Check mask for validity:
353 * a) it must be contiguous.
354 * b) destination must have all host bits clear.
355 * c) if application forgot to set correct family (AF_INET),
356 * reject request unless it is absolutely clear i.e.
357 * both family and mask are zero.
360 addr = sk_extract_addr(&rt->rt_dst);
361 if (!(rt->rt_flags & RTF_HOST)) {
362 __be32 mask = sk_extract_addr(&rt->rt_genmask);
364 if (rt->rt_genmask.sa_family != AF_INET) {
365 if (mask || rt->rt_genmask.sa_family)
366 return -EAFNOSUPPORT;
369 if (bad_mask(mask, addr))
372 plen = inet_mask_len(mask);
375 cfg->fc_dst_len = plen;
378 if (cmd != SIOCDELRT) {
379 cfg->fc_nlflags = NLM_F_CREATE;
380 cfg->fc_protocol = RTPROT_BOOT;
384 cfg->fc_priority = rt->rt_metric - 1;
386 if (rt->rt_flags & RTF_REJECT) {
387 cfg->fc_scope = RT_SCOPE_HOST;
388 cfg->fc_type = RTN_UNREACHABLE;
392 cfg->fc_scope = RT_SCOPE_NOWHERE;
393 cfg->fc_type = RTN_UNICAST;
397 struct net_device *dev;
398 char devname[IFNAMSIZ];
400 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
403 devname[IFNAMSIZ-1] = 0;
404 colon = strchr(devname, ':');
407 dev = __dev_get_by_name(net, devname);
410 cfg->fc_oif = dev->ifindex;
412 struct in_ifaddr *ifa;
413 struct in_device *in_dev = __in_dev_get_rtnl(dev);
417 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
418 if (strcmp(ifa->ifa_label, devname) == 0)
422 cfg->fc_prefsrc = ifa->ifa_local;
426 addr = sk_extract_addr(&rt->rt_gateway);
427 if (rt->rt_gateway.sa_family == AF_INET && addr) {
429 if (rt->rt_flags & RTF_GATEWAY &&
430 inet_addr_type(net, addr) == RTN_UNICAST)
431 cfg->fc_scope = RT_SCOPE_UNIVERSE;
434 if (cmd == SIOCDELRT)
437 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
440 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
441 cfg->fc_scope = RT_SCOPE_LINK;
443 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
447 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
451 if (rt->rt_flags & RTF_MTU)
452 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
454 if (rt->rt_flags & RTF_WINDOW)
455 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
457 if (rt->rt_flags & RTF_IRTT)
458 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
461 cfg->fc_mx_len = len;
468 * Handle IP routing ioctl calls.
469 * These are used to manipulate the routing tables
471 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
473 struct fib_config cfg;
478 case SIOCADDRT: /* Add a route */
479 case SIOCDELRT: /* Delete a route */
480 if (!capable(CAP_NET_ADMIN))
483 if (copy_from_user(&rt, arg, sizeof(rt)))
487 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
489 struct fib_table *tb;
491 if (cmd == SIOCDELRT) {
492 tb = fib_get_table(net, cfg.fc_table);
494 err = fib_table_delete(tb, &cfg);
498 tb = fib_new_table(net, cfg.fc_table);
500 err = fib_table_insert(tb, &cfg);
505 /* allocated by rtentry_to_fib_config() */
514 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
515 [RTA_DST] = { .type = NLA_U32 },
516 [RTA_SRC] = { .type = NLA_U32 },
517 [RTA_IIF] = { .type = NLA_U32 },
518 [RTA_OIF] = { .type = NLA_U32 },
519 [RTA_GATEWAY] = { .type = NLA_U32 },
520 [RTA_PRIORITY] = { .type = NLA_U32 },
521 [RTA_PREFSRC] = { .type = NLA_U32 },
522 [RTA_METRICS] = { .type = NLA_NESTED },
523 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
524 [RTA_FLOW] = { .type = NLA_U32 },
527 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
528 struct nlmsghdr *nlh, struct fib_config *cfg)
534 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
538 memset(cfg, 0, sizeof(*cfg));
540 rtm = nlmsg_data(nlh);
541 cfg->fc_dst_len = rtm->rtm_dst_len;
542 cfg->fc_tos = rtm->rtm_tos;
543 cfg->fc_table = rtm->rtm_table;
544 cfg->fc_protocol = rtm->rtm_protocol;
545 cfg->fc_scope = rtm->rtm_scope;
546 cfg->fc_type = rtm->rtm_type;
547 cfg->fc_flags = rtm->rtm_flags;
548 cfg->fc_nlflags = nlh->nlmsg_flags;
550 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
551 cfg->fc_nlinfo.nlh = nlh;
552 cfg->fc_nlinfo.nl_net = net;
554 if (cfg->fc_type > RTN_MAX) {
559 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
560 switch (nla_type(attr)) {
562 cfg->fc_dst = nla_get_be32(attr);
565 cfg->fc_oif = nla_get_u32(attr);
568 cfg->fc_gw = nla_get_be32(attr);
571 cfg->fc_priority = nla_get_u32(attr);
574 cfg->fc_prefsrc = nla_get_be32(attr);
577 cfg->fc_mx = nla_data(attr);
578 cfg->fc_mx_len = nla_len(attr);
581 cfg->fc_mp = nla_data(attr);
582 cfg->fc_mp_len = nla_len(attr);
585 cfg->fc_flow = nla_get_u32(attr);
588 cfg->fc_table = nla_get_u32(attr);
598 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
600 struct net *net = sock_net(skb->sk);
601 struct fib_config cfg;
602 struct fib_table *tb;
605 err = rtm_to_fib_config(net, skb, nlh, &cfg);
609 tb = fib_get_table(net, cfg.fc_table);
615 err = fib_table_delete(tb, &cfg);
620 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
622 struct net *net = sock_net(skb->sk);
623 struct fib_config cfg;
624 struct fib_table *tb;
627 err = rtm_to_fib_config(net, skb, nlh, &cfg);
631 tb = fib_new_table(net, cfg.fc_table);
637 err = fib_table_insert(tb, &cfg);
642 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
644 struct net *net = sock_net(skb->sk);
646 unsigned int e = 0, s_e;
647 struct fib_table *tb;
648 struct hlist_node *node;
649 struct hlist_head *head;
652 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
653 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
654 return ip_rt_dump(skb, cb);
659 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
661 head = &net->ipv4.fib_table_hash[h];
662 hlist_for_each_entry(tb, node, head, tb_hlist) {
666 memset(&cb->args[2], 0, sizeof(cb->args) -
667 2 * sizeof(cb->args[0]));
668 if (fib_table_dump(tb, skb, cb) < 0)
682 /* Prepare and feed intra-kernel routing request.
683 * Really, it should be netlink message, but :-( netlink
684 * can be not configured, so that we feed it directly
685 * to fib engine. It is legal, because all events occur
686 * only when netlink is already locked.
688 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
690 struct net *net = dev_net(ifa->ifa_dev->dev);
691 struct fib_table *tb;
692 struct fib_config cfg = {
693 .fc_protocol = RTPROT_KERNEL,
696 .fc_dst_len = dst_len,
697 .fc_prefsrc = ifa->ifa_local,
698 .fc_oif = ifa->ifa_dev->dev->ifindex,
699 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
705 if (type == RTN_UNICAST)
706 tb = fib_new_table(net, RT_TABLE_MAIN);
708 tb = fib_new_table(net, RT_TABLE_LOCAL);
713 cfg.fc_table = tb->tb_id;
715 if (type != RTN_LOCAL)
716 cfg.fc_scope = RT_SCOPE_LINK;
718 cfg.fc_scope = RT_SCOPE_HOST;
720 if (cmd == RTM_NEWROUTE)
721 fib_table_insert(tb, &cfg);
723 fib_table_delete(tb, &cfg);
726 void fib_add_ifaddr(struct in_ifaddr *ifa)
728 struct in_device *in_dev = ifa->ifa_dev;
729 struct net_device *dev = in_dev->dev;
730 struct in_ifaddr *prim = ifa;
731 __be32 mask = ifa->ifa_mask;
732 __be32 addr = ifa->ifa_local;
733 __be32 prefix = ifa->ifa_address & mask;
735 if (ifa->ifa_flags & IFA_F_SECONDARY) {
736 prim = inet_ifa_byprefix(in_dev, prefix, mask);
738 pr_warn("%s: bug: prim == NULL\n", __func__);
743 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
745 if (!(dev->flags & IFF_UP))
748 /* Add broadcast address, if it is explicitly assigned. */
749 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
750 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
752 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
753 (prefix != addr || ifa->ifa_prefixlen < 32)) {
754 fib_magic(RTM_NEWROUTE,
755 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
756 prefix, ifa->ifa_prefixlen, prim);
758 /* Add network specific broadcasts, when it takes a sense */
759 if (ifa->ifa_prefixlen < 31) {
760 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
761 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
767 /* Delete primary or secondary address.
768 * Optionally, on secondary address promotion consider the addresses
769 * from subnet iprim as deleted, even if they are in device list.
770 * In this case the secondary ifa can be in device list.
772 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
774 struct in_device *in_dev = ifa->ifa_dev;
775 struct net_device *dev = in_dev->dev;
776 struct in_ifaddr *ifa1;
777 struct in_ifaddr *prim = ifa, *prim1 = NULL;
778 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
779 __be32 any = ifa->ifa_address & ifa->ifa_mask;
785 int subnet = 0; /* Primary network */
786 int gone = 1; /* Address is missing */
787 int same_prefsrc = 0; /* Another primary with same IP */
789 if (ifa->ifa_flags & IFA_F_SECONDARY) {
790 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
792 pr_warn("%s: bug: prim == NULL\n", __func__);
795 if (iprim && iprim != prim) {
796 pr_warn("%s: bug: iprim != prim\n", __func__);
799 } else if (!ipv4_is_zeronet(any) &&
800 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
801 fib_magic(RTM_DELROUTE,
802 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
803 any, ifa->ifa_prefixlen, prim);
807 /* Deletion is more complicated than add.
808 * We should take care of not to delete too much :-)
810 * Scan address list to be sure that addresses are really gone.
813 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
815 /* promotion, keep the IP */
819 /* Ignore IFAs from our subnet */
820 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
821 inet_ifa_match(ifa1->ifa_address, iprim))
824 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
825 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
826 /* Another address from our subnet? */
827 if (ifa1->ifa_mask == prim->ifa_mask &&
828 inet_ifa_match(ifa1->ifa_address, prim))
831 /* We reached the secondaries, so
832 * same_prefsrc should be determined.
836 /* Search new prim1 if ifa1 is not
837 * using the current prim1
840 ifa1->ifa_mask != prim1->ifa_mask ||
841 !inet_ifa_match(ifa1->ifa_address, prim1))
842 prim1 = inet_ifa_byprefix(in_dev,
847 if (prim1->ifa_local != prim->ifa_local)
851 if (prim->ifa_local != ifa1->ifa_local)
857 if (ifa->ifa_local == ifa1->ifa_local)
859 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
861 if (brd == ifa1->ifa_broadcast)
863 if (any == ifa1->ifa_broadcast)
865 /* primary has network specific broadcasts */
866 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
867 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
868 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
870 if (!ipv4_is_zeronet(any1)) {
871 if (ifa->ifa_broadcast == brd1 ||
872 ifa->ifa_broadcast == any1)
874 if (brd == brd1 || brd == any1)
876 if (any == brd1 || any == any1)
883 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
884 if (subnet && ifa->ifa_prefixlen < 31) {
886 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
888 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
890 if (!(ok & LOCAL_OK)) {
891 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
893 /* Check, that this local address finally disappeared. */
895 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
896 /* And the last, but not the least thing.
897 * We must flush stray FIB entries.
899 * First of all, we scan fib_info list searching
900 * for stray nexthop entries, then ignite fib_flush.
902 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
903 fib_flush(dev_net(dev));
912 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
915 struct fib_result res;
916 struct flowi4 fl4 = {
917 .flowi4_mark = frn->fl_mark,
918 .daddr = frn->fl_addr,
919 .flowi4_tos = frn->fl_tos,
920 .flowi4_scope = frn->fl_scope,
923 #ifdef CONFIG_IP_MULTIPLE_TABLES
931 frn->tb_id = tb->tb_id;
933 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
936 frn->prefixlen = res.prefixlen;
937 frn->nh_sel = res.nh_sel;
938 frn->type = res.type;
939 frn->scope = res.scope;
946 static void nl_fib_input(struct sk_buff *skb)
949 struct fib_result_nl *frn;
950 struct nlmsghdr *nlh;
951 struct fib_table *tb;
954 net = sock_net(skb->sk);
955 nlh = nlmsg_hdr(skb);
956 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
957 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
960 skb = skb_clone(skb, GFP_KERNEL);
963 nlh = nlmsg_hdr(skb);
965 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
966 tb = fib_get_table(net, frn->tb_id_in);
968 nl_fib_lookup(frn, tb);
970 pid = NETLINK_CB(skb).pid; /* pid of sending process */
971 NETLINK_CB(skb).pid = 0; /* from kernel */
972 NETLINK_CB(skb).dst_group = 0; /* unicast */
973 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
976 static int __net_init nl_fib_lookup_init(struct net *net)
979 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
980 nl_fib_input, NULL, THIS_MODULE);
982 return -EAFNOSUPPORT;
983 net->ipv4.fibnl = sk;
987 static void nl_fib_lookup_exit(struct net *net)
989 netlink_kernel_release(net->ipv4.fibnl);
990 net->ipv4.fibnl = NULL;
993 static void fib_disable_ip(struct net_device *dev, int force, int delay)
995 if (fib_sync_down_dev(dev, force))
996 fib_flush(dev_net(dev));
997 rt_cache_flush(dev_net(dev), delay);
1001 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1003 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1004 struct net_device *dev = ifa->ifa_dev->dev;
1005 struct net *net = dev_net(dev);
1009 fib_add_ifaddr(ifa);
1010 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1013 atomic_inc(&net->ipv4.dev_addr_genid);
1014 rt_cache_flush(dev_net(dev), -1);
1017 fib_del_ifaddr(ifa, NULL);
1018 atomic_inc(&net->ipv4.dev_addr_genid);
1019 if (ifa->ifa_dev->ifa_list == NULL) {
1020 /* Last address was deleted from this interface.
1023 fib_disable_ip(dev, 1, 0);
1025 rt_cache_flush(dev_net(dev), -1);
1032 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1034 struct net_device *dev = ptr;
1035 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1036 struct net *net = dev_net(dev);
1038 if (event == NETDEV_UNREGISTER) {
1039 fib_disable_ip(dev, 2, -1);
1049 fib_add_ifaddr(ifa);
1050 } endfor_ifa(in_dev);
1051 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1054 atomic_inc(&net->ipv4.dev_addr_genid);
1055 rt_cache_flush(dev_net(dev), -1);
1058 fib_disable_ip(dev, 0, 0);
1060 case NETDEV_CHANGEMTU:
1062 rt_cache_flush(dev_net(dev), 0);
1064 case NETDEV_UNREGISTER_BATCH:
1065 /* The batch unregister is only called on the first
1066 * device in the list of devices being unregistered.
1067 * Therefore we should not pass dev_net(dev) in here.
1069 rt_cache_flush_batch(NULL);
1075 static struct notifier_block fib_inetaddr_notifier = {
1076 .notifier_call = fib_inetaddr_event,
1079 static struct notifier_block fib_netdev_notifier = {
1080 .notifier_call = fib_netdev_event,
1083 static int __net_init ip_fib_net_init(struct net *net)
1086 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1088 /* Avoid false sharing : Use at least a full cache line */
1089 size = max_t(size_t, size, L1_CACHE_BYTES);
1091 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1092 if (net->ipv4.fib_table_hash == NULL)
1095 err = fib4_rules_init(net);
1101 kfree(net->ipv4.fib_table_hash);
1105 static void ip_fib_net_exit(struct net *net)
1109 #ifdef CONFIG_IP_MULTIPLE_TABLES
1110 fib4_rules_exit(net);
1114 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1115 struct fib_table *tb;
1116 struct hlist_head *head;
1117 struct hlist_node *node, *tmp;
1119 head = &net->ipv4.fib_table_hash[i];
1120 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1122 fib_table_flush(tb);
1127 kfree(net->ipv4.fib_table_hash);
1130 static int __net_init fib_net_init(struct net *net)
1134 error = ip_fib_net_init(net);
1137 error = nl_fib_lookup_init(net);
1140 error = fib_proc_init(net);
1147 nl_fib_lookup_exit(net);
1149 ip_fib_net_exit(net);
1153 static void __net_exit fib_net_exit(struct net *net)
1156 nl_fib_lookup_exit(net);
1157 ip_fib_net_exit(net);
1160 static struct pernet_operations fib_net_ops = {
1161 .init = fib_net_init,
1162 .exit = fib_net_exit,
1165 void __init ip_fib_init(void)
1167 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1168 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1169 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1171 register_pernet_subsys(&fib_net_ops);
1172 register_netdevice_notifier(&fib_netdev_notifier);
1173 register_inetaddr_notifier(&fib_inetaddr_notifier);