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 <asm/system.h>
19 #include <linux/bitops.h>
20 #include <linux/capability.h>
21 #include <linux/types.h>
22 #include <linux/kernel.h>
24 #include <linux/string.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/errno.h>
29 #include <linux/inet.h>
30 #include <linux/inetdevice.h>
31 #include <linux/netdevice.h>
32 #include <linux/if_addr.h>
33 #include <linux/if_arp.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/list.h>
37 #include <linux/slab.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
45 #include <net/ip_fib.h>
46 #include <net/rtnetlink.h>
48 #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_hash_table(RT_TABLE_LOCAL);
55 if (local_table == NULL)
58 main_table = fib_hash_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_hash_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 void fib_select_default(struct net *net,
118 const struct flowi *flp, struct fib_result *res)
120 struct fib_table *tb;
121 int table = RT_TABLE_MAIN;
122 #ifdef CONFIG_IP_MULTIPLE_TABLES
123 if (res->r == NULL || res->r->action != FR_ACT_TO_TBL)
125 table = res->r->table;
127 tb = fib_get_table(net, table);
128 if (FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
129 fib_table_select_default(tb, flp, res);
132 static void fib_flush(struct net *net)
135 struct fib_table *tb;
136 struct hlist_node *node;
137 struct hlist_head *head;
140 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
141 head = &net->ipv4.fib_table_hash[h];
142 hlist_for_each_entry(tb, node, head, tb_hlist)
143 flushed += fib_table_flush(tb);
147 rt_cache_flush(net, -1);
151 * __ip_dev_find - find the first device with a given source address.
152 * @net: the net namespace
153 * @addr: the source address
154 * @devref: if true, take a reference on the found device
156 * If a caller uses devref=false, it should be protected by RCU, or RTNL
158 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
162 .flags = FLOWI_FLAG_MATCH_ANY_IIF
164 struct fib_result res = { 0 };
165 struct net_device *dev = NULL;
168 if (fib_lookup(net, &fl, &res)) {
172 if (res.type != RTN_LOCAL)
174 dev = FIB_RES_DEV(res);
182 EXPORT_SYMBOL(__ip_dev_find);
185 * Find address type as if only "dev" was present in the system. If
186 * on_dev is NULL then all interfaces are taken into consideration.
188 static inline unsigned __inet_dev_addr_type(struct net *net,
189 const struct net_device *dev,
192 struct flowi fl = { .fl4_dst = addr };
193 struct fib_result res;
194 unsigned ret = RTN_BROADCAST;
195 struct fib_table *local_table;
197 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
198 return RTN_BROADCAST;
199 if (ipv4_is_multicast(addr))
200 return RTN_MULTICAST;
202 #ifdef CONFIG_IP_MULTIPLE_TABLES
206 local_table = fib_get_table(net, RT_TABLE_LOCAL);
210 if (!fib_table_lookup(local_table, &fl, &res, FIB_LOOKUP_NOREF)) {
211 if (!dev || dev == res.fi->fib_dev)
219 unsigned int inet_addr_type(struct net *net, __be32 addr)
221 return __inet_dev_addr_type(net, NULL, addr);
223 EXPORT_SYMBOL(inet_addr_type);
225 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
228 return __inet_dev_addr_type(net, dev, addr);
230 EXPORT_SYMBOL(inet_dev_addr_type);
232 /* Given (packet source, input interface) and optional (dst, oif, tos):
233 * - (main) check, that source is valid i.e. not broadcast or our local
235 * - figure out what "logical" interface this packet arrived
236 * and calculate "specific destination" address.
237 * - check, that packet arrived from expected physical interface.
238 * called with rcu_read_lock()
240 int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
241 struct net_device *dev, __be32 *spec_dst,
244 struct in_device *in_dev;
252 struct fib_result res;
253 int no_addr, rpf, accept_local;
258 no_addr = rpf = accept_local = 0;
259 in_dev = __in_dev_get_rcu(dev);
261 no_addr = in_dev->ifa_list == NULL;
262 rpf = IN_DEV_RPFILTER(in_dev);
263 accept_local = IN_DEV_ACCEPT_LOCAL(in_dev);
264 if (mark && !IN_DEV_SRC_VMARK(in_dev))
272 if (fib_lookup(net, &fl, &res))
274 if (res.type != RTN_UNICAST) {
275 if (res.type != RTN_LOCAL || !accept_local)
278 *spec_dst = FIB_RES_PREFSRC(res);
279 fib_combine_itag(itag, &res);
282 #ifdef CONFIG_IP_ROUTE_MULTIPATH
283 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
284 struct fib_nh *nh = &res.fi->fib_nh[ret];
286 if (nh->nh_dev == dev) {
292 if (FIB_RES_DEV(res) == dev)
296 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
303 fl.oif = dev->ifindex;
306 if (fib_lookup(net, &fl, &res) == 0) {
307 if (res.type == RTN_UNICAST) {
308 *spec_dst = FIB_RES_PREFSRC(res);
309 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
317 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
327 static inline __be32 sk_extract_addr(struct sockaddr *addr)
329 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
332 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
336 nla = (struct nlattr *) ((char *) mx + len);
337 nla->nla_type = type;
338 nla->nla_len = nla_attr_size(4);
339 *(u32 *) nla_data(nla) = value;
341 return len + nla_total_size(4);
344 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
345 struct fib_config *cfg)
350 memset(cfg, 0, sizeof(*cfg));
351 cfg->fc_nlinfo.nl_net = net;
353 if (rt->rt_dst.sa_family != AF_INET)
354 return -EAFNOSUPPORT;
357 * Check mask for validity:
358 * a) it must be contiguous.
359 * b) destination must have all host bits clear.
360 * c) if application forgot to set correct family (AF_INET),
361 * reject request unless it is absolutely clear i.e.
362 * both family and mask are zero.
365 addr = sk_extract_addr(&rt->rt_dst);
366 if (!(rt->rt_flags & RTF_HOST)) {
367 __be32 mask = sk_extract_addr(&rt->rt_genmask);
369 if (rt->rt_genmask.sa_family != AF_INET) {
370 if (mask || rt->rt_genmask.sa_family)
371 return -EAFNOSUPPORT;
374 if (bad_mask(mask, addr))
377 plen = inet_mask_len(mask);
380 cfg->fc_dst_len = plen;
383 if (cmd != SIOCDELRT) {
384 cfg->fc_nlflags = NLM_F_CREATE;
385 cfg->fc_protocol = RTPROT_BOOT;
389 cfg->fc_priority = rt->rt_metric - 1;
391 if (rt->rt_flags & RTF_REJECT) {
392 cfg->fc_scope = RT_SCOPE_HOST;
393 cfg->fc_type = RTN_UNREACHABLE;
397 cfg->fc_scope = RT_SCOPE_NOWHERE;
398 cfg->fc_type = RTN_UNICAST;
402 struct net_device *dev;
403 char devname[IFNAMSIZ];
405 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
408 devname[IFNAMSIZ-1] = 0;
409 colon = strchr(devname, ':');
412 dev = __dev_get_by_name(net, devname);
415 cfg->fc_oif = dev->ifindex;
417 struct in_ifaddr *ifa;
418 struct in_device *in_dev = __in_dev_get_rtnl(dev);
422 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
423 if (strcmp(ifa->ifa_label, devname) == 0)
427 cfg->fc_prefsrc = ifa->ifa_local;
431 addr = sk_extract_addr(&rt->rt_gateway);
432 if (rt->rt_gateway.sa_family == AF_INET && addr) {
434 if (rt->rt_flags & RTF_GATEWAY &&
435 inet_addr_type(net, addr) == RTN_UNICAST)
436 cfg->fc_scope = RT_SCOPE_UNIVERSE;
439 if (cmd == SIOCDELRT)
442 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
445 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
446 cfg->fc_scope = RT_SCOPE_LINK;
448 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
452 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
456 if (rt->rt_flags & RTF_MTU)
457 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
459 if (rt->rt_flags & RTF_WINDOW)
460 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
462 if (rt->rt_flags & RTF_IRTT)
463 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
466 cfg->fc_mx_len = len;
473 * Handle IP routing ioctl calls.
474 * These are used to manipulate the routing tables
476 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
478 struct fib_config cfg;
483 case SIOCADDRT: /* Add a route */
484 case SIOCDELRT: /* Delete a route */
485 if (!capable(CAP_NET_ADMIN))
488 if (copy_from_user(&rt, arg, sizeof(rt)))
492 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
494 struct fib_table *tb;
496 if (cmd == SIOCDELRT) {
497 tb = fib_get_table(net, cfg.fc_table);
499 err = fib_table_delete(tb, &cfg);
503 tb = fib_new_table(net, cfg.fc_table);
505 err = fib_table_insert(tb, &cfg);
510 /* allocated by rtentry_to_fib_config() */
519 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
520 [RTA_DST] = { .type = NLA_U32 },
521 [RTA_SRC] = { .type = NLA_U32 },
522 [RTA_IIF] = { .type = NLA_U32 },
523 [RTA_OIF] = { .type = NLA_U32 },
524 [RTA_GATEWAY] = { .type = NLA_U32 },
525 [RTA_PRIORITY] = { .type = NLA_U32 },
526 [RTA_PREFSRC] = { .type = NLA_U32 },
527 [RTA_METRICS] = { .type = NLA_NESTED },
528 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
529 [RTA_FLOW] = { .type = NLA_U32 },
532 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
533 struct nlmsghdr *nlh, struct fib_config *cfg)
539 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
543 memset(cfg, 0, sizeof(*cfg));
545 rtm = nlmsg_data(nlh);
546 cfg->fc_dst_len = rtm->rtm_dst_len;
547 cfg->fc_tos = rtm->rtm_tos;
548 cfg->fc_table = rtm->rtm_table;
549 cfg->fc_protocol = rtm->rtm_protocol;
550 cfg->fc_scope = rtm->rtm_scope;
551 cfg->fc_type = rtm->rtm_type;
552 cfg->fc_flags = rtm->rtm_flags;
553 cfg->fc_nlflags = nlh->nlmsg_flags;
555 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
556 cfg->fc_nlinfo.nlh = nlh;
557 cfg->fc_nlinfo.nl_net = net;
559 if (cfg->fc_type > RTN_MAX) {
564 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
565 switch (nla_type(attr)) {
567 cfg->fc_dst = nla_get_be32(attr);
570 cfg->fc_oif = nla_get_u32(attr);
573 cfg->fc_gw = nla_get_be32(attr);
576 cfg->fc_priority = nla_get_u32(attr);
579 cfg->fc_prefsrc = nla_get_be32(attr);
582 cfg->fc_mx = nla_data(attr);
583 cfg->fc_mx_len = nla_len(attr);
586 cfg->fc_mp = nla_data(attr);
587 cfg->fc_mp_len = nla_len(attr);
590 cfg->fc_flow = nla_get_u32(attr);
593 cfg->fc_table = nla_get_u32(attr);
603 static int inet_rtm_delroute(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_get_table(net, cfg.fc_table);
620 err = fib_table_delete(tb, &cfg);
625 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
627 struct net *net = sock_net(skb->sk);
628 struct fib_config cfg;
629 struct fib_table *tb;
632 err = rtm_to_fib_config(net, skb, nlh, &cfg);
636 tb = fib_new_table(net, cfg.fc_table);
642 err = fib_table_insert(tb, &cfg);
647 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
649 struct net *net = sock_net(skb->sk);
651 unsigned int e = 0, s_e;
652 struct fib_table *tb;
653 struct hlist_node *node;
654 struct hlist_head *head;
657 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
658 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
659 return ip_rt_dump(skb, cb);
664 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
666 head = &net->ipv4.fib_table_hash[h];
667 hlist_for_each_entry(tb, node, head, tb_hlist) {
671 memset(&cb->args[2], 0, sizeof(cb->args) -
672 2 * sizeof(cb->args[0]));
673 if (fib_table_dump(tb, skb, cb) < 0)
687 /* Prepare and feed intra-kernel routing request.
688 * Really, it should be netlink message, but :-( netlink
689 * can be not configured, so that we feed it directly
690 * to fib engine. It is legal, because all events occur
691 * only when netlink is already locked.
693 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
695 struct net *net = dev_net(ifa->ifa_dev->dev);
696 struct fib_table *tb;
697 struct fib_config cfg = {
698 .fc_protocol = RTPROT_KERNEL,
701 .fc_dst_len = dst_len,
702 .fc_prefsrc = ifa->ifa_local,
703 .fc_oif = ifa->ifa_dev->dev->ifindex,
704 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
710 if (type == RTN_UNICAST)
711 tb = fib_new_table(net, RT_TABLE_MAIN);
713 tb = fib_new_table(net, RT_TABLE_LOCAL);
718 cfg.fc_table = tb->tb_id;
720 if (type != RTN_LOCAL)
721 cfg.fc_scope = RT_SCOPE_LINK;
723 cfg.fc_scope = RT_SCOPE_HOST;
725 if (cmd == RTM_NEWROUTE)
726 fib_table_insert(tb, &cfg);
728 fib_table_delete(tb, &cfg);
731 void fib_add_ifaddr(struct in_ifaddr *ifa)
733 struct in_device *in_dev = ifa->ifa_dev;
734 struct net_device *dev = in_dev->dev;
735 struct in_ifaddr *prim = ifa;
736 __be32 mask = ifa->ifa_mask;
737 __be32 addr = ifa->ifa_local;
738 __be32 prefix = ifa->ifa_address & mask;
740 if (ifa->ifa_flags & IFA_F_SECONDARY) {
741 prim = inet_ifa_byprefix(in_dev, prefix, mask);
743 printk(KERN_WARNING "fib_add_ifaddr: bug: prim == NULL\n");
748 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
750 if (!(dev->flags & IFF_UP))
753 /* Add broadcast address, if it is explicitly assigned. */
754 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
755 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
757 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
758 (prefix != addr || ifa->ifa_prefixlen < 32)) {
759 fib_magic(RTM_NEWROUTE,
760 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
761 prefix, ifa->ifa_prefixlen, prim);
763 /* Add network specific broadcasts, when it takes a sense */
764 if (ifa->ifa_prefixlen < 31) {
765 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
766 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
772 static void fib_del_ifaddr(struct in_ifaddr *ifa)
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;
778 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
779 __be32 any = ifa->ifa_address & ifa->ifa_mask;
786 if (!(ifa->ifa_flags & IFA_F_SECONDARY))
787 fib_magic(RTM_DELROUTE,
788 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
789 any, ifa->ifa_prefixlen, prim);
791 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
793 printk(KERN_WARNING "fib_del_ifaddr: bug: prim == NULL\n");
798 /* Deletion is more complicated than add.
799 * We should take care of not to delete too much :-)
801 * Scan address list to be sure that addresses are really gone.
804 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
805 if (ifa->ifa_local == ifa1->ifa_local)
807 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
809 if (brd == ifa1->ifa_broadcast)
811 if (any == ifa1->ifa_broadcast)
816 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
818 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
820 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
821 if (!(ok & LOCAL_OK)) {
822 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
824 /* Check, that this local address finally disappeared. */
825 if (inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
826 /* And the last, but not the least thing.
827 * We must flush stray FIB entries.
829 * First of all, we scan fib_info list searching
830 * for stray nexthop entries, then ignite fib_flush.
832 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
833 fib_flush(dev_net(dev));
842 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
845 struct fib_result res;
847 .mark = frn->fl_mark,
848 .fl4_dst = frn->fl_addr,
849 .fl4_tos = frn->fl_tos,
850 .fl4_scope = frn->fl_scope,
853 #ifdef CONFIG_IP_MULTIPLE_TABLES
861 frn->tb_id = tb->tb_id;
863 frn->err = fib_table_lookup(tb, &fl, &res, FIB_LOOKUP_NOREF);
866 frn->prefixlen = res.prefixlen;
867 frn->nh_sel = res.nh_sel;
868 frn->type = res.type;
869 frn->scope = res.scope;
876 static void nl_fib_input(struct sk_buff *skb)
879 struct fib_result_nl *frn;
880 struct nlmsghdr *nlh;
881 struct fib_table *tb;
884 net = sock_net(skb->sk);
885 nlh = nlmsg_hdr(skb);
886 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
887 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
890 skb = skb_clone(skb, GFP_KERNEL);
893 nlh = nlmsg_hdr(skb);
895 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
896 tb = fib_get_table(net, frn->tb_id_in);
898 nl_fib_lookup(frn, tb);
900 pid = NETLINK_CB(skb).pid; /* pid of sending process */
901 NETLINK_CB(skb).pid = 0; /* from kernel */
902 NETLINK_CB(skb).dst_group = 0; /* unicast */
903 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
906 static int __net_init nl_fib_lookup_init(struct net *net)
909 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
910 nl_fib_input, NULL, THIS_MODULE);
912 return -EAFNOSUPPORT;
913 net->ipv4.fibnl = sk;
917 static void nl_fib_lookup_exit(struct net *net)
919 netlink_kernel_release(net->ipv4.fibnl);
920 net->ipv4.fibnl = NULL;
923 static void fib_disable_ip(struct net_device *dev, int force, int delay)
925 if (fib_sync_down_dev(dev, force))
926 fib_flush(dev_net(dev));
927 rt_cache_flush(dev_net(dev), delay);
931 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
933 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
934 struct net_device *dev = ifa->ifa_dev->dev;
939 #ifdef CONFIG_IP_ROUTE_MULTIPATH
942 rt_cache_flush(dev_net(dev), -1);
946 if (ifa->ifa_dev->ifa_list == NULL) {
947 /* Last address was deleted from this interface.
950 fib_disable_ip(dev, 1, 0);
952 rt_cache_flush(dev_net(dev), -1);
959 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
961 struct net_device *dev = ptr;
962 struct in_device *in_dev = __in_dev_get_rtnl(dev);
964 if (event == NETDEV_UNREGISTER) {
965 fib_disable_ip(dev, 2, -1);
976 } endfor_ifa(in_dev);
977 #ifdef CONFIG_IP_ROUTE_MULTIPATH
980 rt_cache_flush(dev_net(dev), -1);
983 fib_disable_ip(dev, 0, 0);
985 case NETDEV_CHANGEMTU:
987 rt_cache_flush(dev_net(dev), 0);
989 case NETDEV_UNREGISTER_BATCH:
990 rt_cache_flush_batch();
996 static struct notifier_block fib_inetaddr_notifier = {
997 .notifier_call = fib_inetaddr_event,
1000 static struct notifier_block fib_netdev_notifier = {
1001 .notifier_call = fib_netdev_event,
1004 static int __net_init ip_fib_net_init(struct net *net)
1007 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1009 /* Avoid false sharing : Use at least a full cache line */
1010 size = max_t(size_t, size, L1_CACHE_BYTES);
1012 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1013 if (net->ipv4.fib_table_hash == NULL)
1016 err = fib4_rules_init(net);
1022 kfree(net->ipv4.fib_table_hash);
1026 static void ip_fib_net_exit(struct net *net)
1030 #ifdef CONFIG_IP_MULTIPLE_TABLES
1031 fib4_rules_exit(net);
1034 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1035 struct fib_table *tb;
1036 struct hlist_head *head;
1037 struct hlist_node *node, *tmp;
1039 head = &net->ipv4.fib_table_hash[i];
1040 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1042 fib_table_flush(tb);
1046 kfree(net->ipv4.fib_table_hash);
1049 static int __net_init fib_net_init(struct net *net)
1053 error = ip_fib_net_init(net);
1056 error = nl_fib_lookup_init(net);
1059 error = fib_proc_init(net);
1066 nl_fib_lookup_exit(net);
1068 ip_fib_net_exit(net);
1072 static void __net_exit fib_net_exit(struct net *net)
1075 nl_fib_lookup_exit(net);
1076 ip_fib_net_exit(net);
1079 static struct pernet_operations fib_net_ops = {
1080 .init = fib_net_init,
1081 .exit = fib_net_exit,
1084 void __init ip_fib_init(void)
1086 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1087 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1088 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1090 register_pernet_subsys(&fib_net_ops);
1091 register_netdevice_notifier(&fib_netdev_notifier);
1092 register_inetaddr_notifier(&fib_inetaddr_notifier);