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)
167 struct fib_result res = { 0 };
168 struct net_device *dev = NULL;
169 struct fib_table *local_table;
171 #ifdef CONFIG_IP_MULTIPLE_TABLES
176 local_table = fib_get_table(net, RT_TABLE_LOCAL);
178 fib_table_lookup(local_table, &fl, &res, FIB_LOOKUP_NOREF)) {
182 if (res.type != RTN_LOCAL)
184 dev = FIB_RES_DEV(res);
192 EXPORT_SYMBOL(__ip_dev_find);
195 * Find address type as if only "dev" was present in the system. If
196 * on_dev is NULL then all interfaces are taken into consideration.
198 static inline unsigned __inet_dev_addr_type(struct net *net,
199 const struct net_device *dev,
202 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
203 struct fib_result res;
204 unsigned ret = RTN_BROADCAST;
205 struct fib_table *local_table;
207 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
208 return RTN_BROADCAST;
209 if (ipv4_is_multicast(addr))
210 return RTN_MULTICAST;
212 #ifdef CONFIG_IP_MULTIPLE_TABLES
216 local_table = fib_get_table(net, RT_TABLE_LOCAL);
220 if (!fib_table_lookup(local_table, &fl, &res, FIB_LOOKUP_NOREF)) {
221 if (!dev || dev == res.fi->fib_dev)
229 unsigned int inet_addr_type(struct net *net, __be32 addr)
231 return __inet_dev_addr_type(net, NULL, addr);
233 EXPORT_SYMBOL(inet_addr_type);
235 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
238 return __inet_dev_addr_type(net, dev, addr);
240 EXPORT_SYMBOL(inet_dev_addr_type);
242 /* Given (packet source, input interface) and optional (dst, oif, tos):
243 * - (main) check, that source is valid i.e. not broadcast or our local
245 * - figure out what "logical" interface this packet arrived
246 * and calculate "specific destination" address.
247 * - check, that packet arrived from expected physical interface.
248 * called with rcu_read_lock()
250 int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
251 struct net_device *dev, __be32 *spec_dst,
254 struct in_device *in_dev;
266 struct fib_result res;
267 int no_addr, rpf, accept_local;
272 no_addr = rpf = accept_local = 0;
273 in_dev = __in_dev_get_rcu(dev);
275 no_addr = in_dev->ifa_list == NULL;
276 rpf = IN_DEV_RPFILTER(in_dev);
277 accept_local = IN_DEV_ACCEPT_LOCAL(in_dev);
278 if (mark && !IN_DEV_SRC_VMARK(in_dev))
286 if (fib_lookup(net, &fl, &res))
288 if (res.type != RTN_UNICAST) {
289 if (res.type != RTN_LOCAL || !accept_local)
292 *spec_dst = FIB_RES_PREFSRC(res);
293 fib_combine_itag(itag, &res);
296 #ifdef CONFIG_IP_ROUTE_MULTIPATH
297 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
298 struct fib_nh *nh = &res.fi->fib_nh[ret];
300 if (nh->nh_dev == dev) {
306 if (FIB_RES_DEV(res) == dev)
310 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
317 fl.oif = dev->ifindex;
320 if (fib_lookup(net, &fl, &res) == 0) {
321 if (res.type == RTN_UNICAST) {
322 *spec_dst = FIB_RES_PREFSRC(res);
323 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
331 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
341 static inline __be32 sk_extract_addr(struct sockaddr *addr)
343 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
346 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
350 nla = (struct nlattr *) ((char *) mx + len);
351 nla->nla_type = type;
352 nla->nla_len = nla_attr_size(4);
353 *(u32 *) nla_data(nla) = value;
355 return len + nla_total_size(4);
358 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
359 struct fib_config *cfg)
364 memset(cfg, 0, sizeof(*cfg));
365 cfg->fc_nlinfo.nl_net = net;
367 if (rt->rt_dst.sa_family != AF_INET)
368 return -EAFNOSUPPORT;
371 * Check mask for validity:
372 * a) it must be contiguous.
373 * b) destination must have all host bits clear.
374 * c) if application forgot to set correct family (AF_INET),
375 * reject request unless it is absolutely clear i.e.
376 * both family and mask are zero.
379 addr = sk_extract_addr(&rt->rt_dst);
380 if (!(rt->rt_flags & RTF_HOST)) {
381 __be32 mask = sk_extract_addr(&rt->rt_genmask);
383 if (rt->rt_genmask.sa_family != AF_INET) {
384 if (mask || rt->rt_genmask.sa_family)
385 return -EAFNOSUPPORT;
388 if (bad_mask(mask, addr))
391 plen = inet_mask_len(mask);
394 cfg->fc_dst_len = plen;
397 if (cmd != SIOCDELRT) {
398 cfg->fc_nlflags = NLM_F_CREATE;
399 cfg->fc_protocol = RTPROT_BOOT;
403 cfg->fc_priority = rt->rt_metric - 1;
405 if (rt->rt_flags & RTF_REJECT) {
406 cfg->fc_scope = RT_SCOPE_HOST;
407 cfg->fc_type = RTN_UNREACHABLE;
411 cfg->fc_scope = RT_SCOPE_NOWHERE;
412 cfg->fc_type = RTN_UNICAST;
416 struct net_device *dev;
417 char devname[IFNAMSIZ];
419 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
422 devname[IFNAMSIZ-1] = 0;
423 colon = strchr(devname, ':');
426 dev = __dev_get_by_name(net, devname);
429 cfg->fc_oif = dev->ifindex;
431 struct in_ifaddr *ifa;
432 struct in_device *in_dev = __in_dev_get_rtnl(dev);
436 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
437 if (strcmp(ifa->ifa_label, devname) == 0)
441 cfg->fc_prefsrc = ifa->ifa_local;
445 addr = sk_extract_addr(&rt->rt_gateway);
446 if (rt->rt_gateway.sa_family == AF_INET && addr) {
448 if (rt->rt_flags & RTF_GATEWAY &&
449 inet_addr_type(net, addr) == RTN_UNICAST)
450 cfg->fc_scope = RT_SCOPE_UNIVERSE;
453 if (cmd == SIOCDELRT)
456 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
459 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
460 cfg->fc_scope = RT_SCOPE_LINK;
462 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
466 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
470 if (rt->rt_flags & RTF_MTU)
471 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
473 if (rt->rt_flags & RTF_WINDOW)
474 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
476 if (rt->rt_flags & RTF_IRTT)
477 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
480 cfg->fc_mx_len = len;
487 * Handle IP routing ioctl calls.
488 * These are used to manipulate the routing tables
490 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
492 struct fib_config cfg;
497 case SIOCADDRT: /* Add a route */
498 case SIOCDELRT: /* Delete a route */
499 if (!capable(CAP_NET_ADMIN))
502 if (copy_from_user(&rt, arg, sizeof(rt)))
506 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
508 struct fib_table *tb;
510 if (cmd == SIOCDELRT) {
511 tb = fib_get_table(net, cfg.fc_table);
513 err = fib_table_delete(tb, &cfg);
517 tb = fib_new_table(net, cfg.fc_table);
519 err = fib_table_insert(tb, &cfg);
524 /* allocated by rtentry_to_fib_config() */
533 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
534 [RTA_DST] = { .type = NLA_U32 },
535 [RTA_SRC] = { .type = NLA_U32 },
536 [RTA_IIF] = { .type = NLA_U32 },
537 [RTA_OIF] = { .type = NLA_U32 },
538 [RTA_GATEWAY] = { .type = NLA_U32 },
539 [RTA_PRIORITY] = { .type = NLA_U32 },
540 [RTA_PREFSRC] = { .type = NLA_U32 },
541 [RTA_METRICS] = { .type = NLA_NESTED },
542 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
543 [RTA_FLOW] = { .type = NLA_U32 },
546 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
547 struct nlmsghdr *nlh, struct fib_config *cfg)
553 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
557 memset(cfg, 0, sizeof(*cfg));
559 rtm = nlmsg_data(nlh);
560 cfg->fc_dst_len = rtm->rtm_dst_len;
561 cfg->fc_tos = rtm->rtm_tos;
562 cfg->fc_table = rtm->rtm_table;
563 cfg->fc_protocol = rtm->rtm_protocol;
564 cfg->fc_scope = rtm->rtm_scope;
565 cfg->fc_type = rtm->rtm_type;
566 cfg->fc_flags = rtm->rtm_flags;
567 cfg->fc_nlflags = nlh->nlmsg_flags;
569 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
570 cfg->fc_nlinfo.nlh = nlh;
571 cfg->fc_nlinfo.nl_net = net;
573 if (cfg->fc_type > RTN_MAX) {
578 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
579 switch (nla_type(attr)) {
581 cfg->fc_dst = nla_get_be32(attr);
584 cfg->fc_oif = nla_get_u32(attr);
587 cfg->fc_gw = nla_get_be32(attr);
590 cfg->fc_priority = nla_get_u32(attr);
593 cfg->fc_prefsrc = nla_get_be32(attr);
596 cfg->fc_mx = nla_data(attr);
597 cfg->fc_mx_len = nla_len(attr);
600 cfg->fc_mp = nla_data(attr);
601 cfg->fc_mp_len = nla_len(attr);
604 cfg->fc_flow = nla_get_u32(attr);
607 cfg->fc_table = nla_get_u32(attr);
617 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
619 struct net *net = sock_net(skb->sk);
620 struct fib_config cfg;
621 struct fib_table *tb;
624 err = rtm_to_fib_config(net, skb, nlh, &cfg);
628 tb = fib_get_table(net, cfg.fc_table);
634 err = fib_table_delete(tb, &cfg);
639 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
641 struct net *net = sock_net(skb->sk);
642 struct fib_config cfg;
643 struct fib_table *tb;
646 err = rtm_to_fib_config(net, skb, nlh, &cfg);
650 tb = fib_new_table(net, cfg.fc_table);
656 err = fib_table_insert(tb, &cfg);
661 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
663 struct net *net = sock_net(skb->sk);
665 unsigned int e = 0, s_e;
666 struct fib_table *tb;
667 struct hlist_node *node;
668 struct hlist_head *head;
671 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
672 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
673 return ip_rt_dump(skb, cb);
678 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
680 head = &net->ipv4.fib_table_hash[h];
681 hlist_for_each_entry(tb, node, head, tb_hlist) {
685 memset(&cb->args[2], 0, sizeof(cb->args) -
686 2 * sizeof(cb->args[0]));
687 if (fib_table_dump(tb, skb, cb) < 0)
701 /* Prepare and feed intra-kernel routing request.
702 * Really, it should be netlink message, but :-( netlink
703 * can be not configured, so that we feed it directly
704 * to fib engine. It is legal, because all events occur
705 * only when netlink is already locked.
707 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
709 struct net *net = dev_net(ifa->ifa_dev->dev);
710 struct fib_table *tb;
711 struct fib_config cfg = {
712 .fc_protocol = RTPROT_KERNEL,
715 .fc_dst_len = dst_len,
716 .fc_prefsrc = ifa->ifa_local,
717 .fc_oif = ifa->ifa_dev->dev->ifindex,
718 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
724 if (type == RTN_UNICAST)
725 tb = fib_new_table(net, RT_TABLE_MAIN);
727 tb = fib_new_table(net, RT_TABLE_LOCAL);
732 cfg.fc_table = tb->tb_id;
734 if (type != RTN_LOCAL)
735 cfg.fc_scope = RT_SCOPE_LINK;
737 cfg.fc_scope = RT_SCOPE_HOST;
739 if (cmd == RTM_NEWROUTE)
740 fib_table_insert(tb, &cfg);
742 fib_table_delete(tb, &cfg);
745 void fib_add_ifaddr(struct in_ifaddr *ifa)
747 struct in_device *in_dev = ifa->ifa_dev;
748 struct net_device *dev = in_dev->dev;
749 struct in_ifaddr *prim = ifa;
750 __be32 mask = ifa->ifa_mask;
751 __be32 addr = ifa->ifa_local;
752 __be32 prefix = ifa->ifa_address & mask;
754 if (ifa->ifa_flags & IFA_F_SECONDARY) {
755 prim = inet_ifa_byprefix(in_dev, prefix, mask);
757 printk(KERN_WARNING "fib_add_ifaddr: bug: prim == NULL\n");
762 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
764 if (!(dev->flags & IFF_UP))
767 /* Add broadcast address, if it is explicitly assigned. */
768 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
769 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
771 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
772 (prefix != addr || ifa->ifa_prefixlen < 32)) {
773 fib_magic(RTM_NEWROUTE,
774 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
775 prefix, ifa->ifa_prefixlen, prim);
777 /* Add network specific broadcasts, when it takes a sense */
778 if (ifa->ifa_prefixlen < 31) {
779 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
780 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
786 static void fib_del_ifaddr(struct in_ifaddr *ifa)
788 struct in_device *in_dev = ifa->ifa_dev;
789 struct net_device *dev = in_dev->dev;
790 struct in_ifaddr *ifa1;
791 struct in_ifaddr *prim = ifa;
792 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
793 __be32 any = ifa->ifa_address & ifa->ifa_mask;
800 if (!(ifa->ifa_flags & IFA_F_SECONDARY))
801 fib_magic(RTM_DELROUTE,
802 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
803 any, ifa->ifa_prefixlen, prim);
805 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
807 printk(KERN_WARNING "fib_del_ifaddr: bug: prim == NULL\n");
812 /* Deletion is more complicated than add.
813 * We should take care of not to delete too much :-)
815 * Scan address list to be sure that addresses are really gone.
818 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
819 if (ifa->ifa_local == ifa1->ifa_local)
821 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
823 if (brd == ifa1->ifa_broadcast)
825 if (any == ifa1->ifa_broadcast)
830 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
832 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
834 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
835 if (!(ok & LOCAL_OK)) {
836 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
838 /* Check, that this local address finally disappeared. */
839 if (inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
840 /* And the last, but not the least thing.
841 * We must flush stray FIB entries.
843 * First of all, we scan fib_info list searching
844 * for stray nexthop entries, then ignite fib_flush.
846 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
847 fib_flush(dev_net(dev));
856 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
859 struct fib_result res;
861 .mark = frn->fl_mark,
864 .daddr = frn->fl_addr,
866 .scope = frn->fl_scope
871 #ifdef CONFIG_IP_MULTIPLE_TABLES
879 frn->tb_id = tb->tb_id;
881 frn->err = fib_table_lookup(tb, &fl, &res, FIB_LOOKUP_NOREF);
884 frn->prefixlen = res.prefixlen;
885 frn->nh_sel = res.nh_sel;
886 frn->type = res.type;
887 frn->scope = res.scope;
894 static void nl_fib_input(struct sk_buff *skb)
897 struct fib_result_nl *frn;
898 struct nlmsghdr *nlh;
899 struct fib_table *tb;
902 net = sock_net(skb->sk);
903 nlh = nlmsg_hdr(skb);
904 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
905 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
908 skb = skb_clone(skb, GFP_KERNEL);
911 nlh = nlmsg_hdr(skb);
913 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
914 tb = fib_get_table(net, frn->tb_id_in);
916 nl_fib_lookup(frn, tb);
918 pid = NETLINK_CB(skb).pid; /* pid of sending process */
919 NETLINK_CB(skb).pid = 0; /* from kernel */
920 NETLINK_CB(skb).dst_group = 0; /* unicast */
921 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
924 static int __net_init nl_fib_lookup_init(struct net *net)
927 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
928 nl_fib_input, NULL, THIS_MODULE);
930 return -EAFNOSUPPORT;
931 net->ipv4.fibnl = sk;
935 static void nl_fib_lookup_exit(struct net *net)
937 netlink_kernel_release(net->ipv4.fibnl);
938 net->ipv4.fibnl = NULL;
941 static void fib_disable_ip(struct net_device *dev, int force, int delay)
943 if (fib_sync_down_dev(dev, force))
944 fib_flush(dev_net(dev));
945 rt_cache_flush(dev_net(dev), delay);
949 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
951 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
952 struct net_device *dev = ifa->ifa_dev->dev;
957 #ifdef CONFIG_IP_ROUTE_MULTIPATH
960 rt_cache_flush(dev_net(dev), -1);
964 if (ifa->ifa_dev->ifa_list == NULL) {
965 /* Last address was deleted from this interface.
968 fib_disable_ip(dev, 1, 0);
970 rt_cache_flush(dev_net(dev), -1);
977 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
979 struct net_device *dev = ptr;
980 struct in_device *in_dev = __in_dev_get_rtnl(dev);
982 if (event == NETDEV_UNREGISTER) {
983 fib_disable_ip(dev, 2, -1);
994 } endfor_ifa(in_dev);
995 #ifdef CONFIG_IP_ROUTE_MULTIPATH
998 rt_cache_flush(dev_net(dev), -1);
1001 fib_disable_ip(dev, 0, 0);
1003 case NETDEV_CHANGEMTU:
1005 rt_cache_flush(dev_net(dev), 0);
1007 case NETDEV_UNREGISTER_BATCH:
1008 rt_cache_flush_batch();
1014 static struct notifier_block fib_inetaddr_notifier = {
1015 .notifier_call = fib_inetaddr_event,
1018 static struct notifier_block fib_netdev_notifier = {
1019 .notifier_call = fib_netdev_event,
1022 static int __net_init ip_fib_net_init(struct net *net)
1025 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1027 /* Avoid false sharing : Use at least a full cache line */
1028 size = max_t(size_t, size, L1_CACHE_BYTES);
1030 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1031 if (net->ipv4.fib_table_hash == NULL)
1034 err = fib4_rules_init(net);
1040 kfree(net->ipv4.fib_table_hash);
1044 static void ip_fib_net_exit(struct net *net)
1048 #ifdef CONFIG_IP_MULTIPLE_TABLES
1049 fib4_rules_exit(net);
1052 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1053 struct fib_table *tb;
1054 struct hlist_head *head;
1055 struct hlist_node *node, *tmp;
1057 head = &net->ipv4.fib_table_hash[i];
1058 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1060 fib_table_flush(tb);
1064 kfree(net->ipv4.fib_table_hash);
1067 static int __net_init fib_net_init(struct net *net)
1071 error = ip_fib_net_init(net);
1074 error = nl_fib_lookup_init(net);
1077 error = fib_proc_init(net);
1084 nl_fib_lookup_exit(net);
1086 ip_fib_net_exit(net);
1090 static void __net_exit fib_net_exit(struct net *net)
1093 nl_fib_lookup_exit(net);
1094 ip_fib_net_exit(net);
1097 static struct pernet_operations fib_net_ops = {
1098 .init = fib_net_init,
1099 .exit = fib_net_exit,
1102 void __init ip_fib_init(void)
1104 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1105 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1106 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1108 register_pernet_subsys(&fib_net_ops);
1109 register_netdevice_notifier(&fib_netdev_notifier);
1110 register_inetaddr_notifier(&fib_inetaddr_notifier);