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>
50 #ifndef CONFIG_IP_MULTIPLE_TABLES
52 static int __net_init fib4_rules_init(struct net *net)
54 struct fib_table *local_table, *main_table;
56 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
60 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
64 hlist_add_head_rcu(&local_table->tb_hlist,
65 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
66 hlist_add_head_rcu(&main_table->tb_hlist,
67 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
71 fib_free_table(main_table);
76 struct fib_table *fib_new_table(struct net *net, u32 id)
78 struct fib_table *tb, *alias = NULL;
83 tb = fib_get_table(net, id);
87 if (id == RT_TABLE_LOCAL)
88 alias = fib_new_table(net, RT_TABLE_MAIN);
90 tb = fib_trie_table(id, alias);
96 rcu_assign_pointer(net->ipv4.fib_local, tb);
99 rcu_assign_pointer(net->ipv4.fib_main, tb);
101 case RT_TABLE_DEFAULT:
102 rcu_assign_pointer(net->ipv4.fib_default, tb);
108 h = id & (FIB_TABLE_HASHSZ - 1);
109 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
113 /* caller must hold either rtnl or rcu read lock */
114 struct fib_table *fib_get_table(struct net *net, u32 id)
116 struct fib_table *tb;
117 struct hlist_head *head;
122 h = id & (FIB_TABLE_HASHSZ - 1);
124 head = &net->ipv4.fib_table_hash[h];
125 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
131 #endif /* CONFIG_IP_MULTIPLE_TABLES */
133 static void fib_replace_table(struct net *net, struct fib_table *old,
134 struct fib_table *new)
136 #ifdef CONFIG_IP_MULTIPLE_TABLES
137 switch (new->tb_id) {
139 rcu_assign_pointer(net->ipv4.fib_local, new);
142 rcu_assign_pointer(net->ipv4.fib_main, new);
144 case RT_TABLE_DEFAULT:
145 rcu_assign_pointer(net->ipv4.fib_default, new);
152 /* replace the old table in the hlist */
153 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
156 int fib_unmerge(struct net *net)
158 struct fib_table *old, *new;
160 /* attempt to fetch local table if it has been allocated */
161 old = fib_get_table(net, RT_TABLE_LOCAL);
165 new = fib_trie_unmerge(old);
169 /* replace merged table with clean table */
171 fib_replace_table(net, old, new);
178 static void fib_flush(struct net *net)
183 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
184 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
185 struct hlist_node *tmp;
186 struct fib_table *tb;
188 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
189 flushed += fib_table_flush(tb);
196 void fib_flush_external(struct net *net)
198 struct fib_table *tb;
199 struct hlist_head *head;
202 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
203 head = &net->ipv4.fib_table_hash[h];
204 hlist_for_each_entry(tb, head, tb_hlist)
205 fib_table_flush_external(tb);
210 * Find address type as if only "dev" was present in the system. If
211 * on_dev is NULL then all interfaces are taken into consideration.
213 static inline unsigned int __inet_dev_addr_type(struct net *net,
214 const struct net_device *dev,
215 __be32 addr, int tb_id)
217 struct flowi4 fl4 = { .daddr = addr };
218 struct fib_result res;
219 unsigned int ret = RTN_BROADCAST;
220 struct fib_table *table;
222 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
223 return RTN_BROADCAST;
224 if (ipv4_is_multicast(addr))
225 return RTN_MULTICAST;
229 table = fib_get_table(net, tb_id);
232 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
233 if (!dev || dev == res.fi->fib_dev)
242 unsigned int inet_addr_type_table(struct net *net, __be32 addr, int tb_id)
244 return __inet_dev_addr_type(net, NULL, addr, tb_id);
246 EXPORT_SYMBOL(inet_addr_type_table);
248 unsigned int inet_addr_type(struct net *net, __be32 addr)
250 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
252 EXPORT_SYMBOL(inet_addr_type);
254 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
257 int rt_table = vrf_dev_table(dev) ? : RT_TABLE_LOCAL;
259 return __inet_dev_addr_type(net, dev, addr, rt_table);
261 EXPORT_SYMBOL(inet_dev_addr_type);
263 __be32 fib_compute_spec_dst(struct sk_buff *skb)
265 struct net_device *dev = skb->dev;
266 struct in_device *in_dev;
267 struct fib_result res;
273 rt = skb_rtable(skb);
274 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
276 return ip_hdr(skb)->daddr;
278 in_dev = __in_dev_get_rcu(dev);
283 scope = RT_SCOPE_UNIVERSE;
284 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
286 fl4.flowi4_iif = LOOPBACK_IFINDEX;
287 fl4.daddr = ip_hdr(skb)->saddr;
289 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
290 fl4.flowi4_scope = scope;
291 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
292 fl4.flowi4_tun_key.tun_id = 0;
293 if (!fib_lookup(net, &fl4, &res, 0))
294 return FIB_RES_PREFSRC(net, res);
296 scope = RT_SCOPE_LINK;
299 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
302 /* Given (packet source, input interface) and optional (dst, oif, tos):
303 * - (main) check, that source is valid i.e. not broadcast or our local
305 * - figure out what "logical" interface this packet arrived
306 * and calculate "specific destination" address.
307 * - check, that packet arrived from expected physical interface.
308 * called with rcu_read_lock()
310 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
311 u8 tos, int oif, struct net_device *dev,
312 int rpf, struct in_device *idev, u32 *itag)
315 struct fib_result res;
321 fl4.flowi4_iif = vrf_master_ifindex_rcu(dev);
323 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
326 fl4.flowi4_tos = tos;
327 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
328 fl4.flowi4_tun_key.tun_id = 0;
330 no_addr = idev->ifa_list == NULL;
332 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
335 if (fib_lookup(net, &fl4, &res, 0))
337 if (res.type != RTN_UNICAST &&
338 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
340 if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
341 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
343 fib_combine_itag(itag, &res);
346 #ifdef CONFIG_IP_ROUTE_MULTIPATH
347 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
348 struct fib_nh *nh = &res.fi->fib_nh[ret];
350 if (nh->nh_dev == dev) {
353 } else if (vrf_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
359 if (FIB_RES_DEV(res) == dev)
363 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
370 fl4.flowi4_oif = dev->ifindex;
373 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
374 if (res.type == RTN_UNICAST)
375 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
391 /* Ignore rp_filter for packets protected by IPsec. */
392 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
393 u8 tos, int oif, struct net_device *dev,
394 struct in_device *idev, u32 *itag)
396 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
398 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
399 IN_DEV_ACCEPT_LOCAL(idev) &&
400 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
404 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
407 static inline __be32 sk_extract_addr(struct sockaddr *addr)
409 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
412 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
416 nla = (struct nlattr *) ((char *) mx + len);
417 nla->nla_type = type;
418 nla->nla_len = nla_attr_size(4);
419 *(u32 *) nla_data(nla) = value;
421 return len + nla_total_size(4);
424 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
425 struct fib_config *cfg)
430 memset(cfg, 0, sizeof(*cfg));
431 cfg->fc_nlinfo.nl_net = net;
433 if (rt->rt_dst.sa_family != AF_INET)
434 return -EAFNOSUPPORT;
437 * Check mask for validity:
438 * a) it must be contiguous.
439 * b) destination must have all host bits clear.
440 * c) if application forgot to set correct family (AF_INET),
441 * reject request unless it is absolutely clear i.e.
442 * both family and mask are zero.
445 addr = sk_extract_addr(&rt->rt_dst);
446 if (!(rt->rt_flags & RTF_HOST)) {
447 __be32 mask = sk_extract_addr(&rt->rt_genmask);
449 if (rt->rt_genmask.sa_family != AF_INET) {
450 if (mask || rt->rt_genmask.sa_family)
451 return -EAFNOSUPPORT;
454 if (bad_mask(mask, addr))
457 plen = inet_mask_len(mask);
460 cfg->fc_dst_len = plen;
463 if (cmd != SIOCDELRT) {
464 cfg->fc_nlflags = NLM_F_CREATE;
465 cfg->fc_protocol = RTPROT_BOOT;
469 cfg->fc_priority = rt->rt_metric - 1;
471 if (rt->rt_flags & RTF_REJECT) {
472 cfg->fc_scope = RT_SCOPE_HOST;
473 cfg->fc_type = RTN_UNREACHABLE;
477 cfg->fc_scope = RT_SCOPE_NOWHERE;
478 cfg->fc_type = RTN_UNICAST;
482 struct net_device *dev;
483 char devname[IFNAMSIZ];
485 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
488 devname[IFNAMSIZ-1] = 0;
489 colon = strchr(devname, ':');
492 dev = __dev_get_by_name(net, devname);
495 cfg->fc_oif = dev->ifindex;
497 struct in_ifaddr *ifa;
498 struct in_device *in_dev = __in_dev_get_rtnl(dev);
502 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
503 if (strcmp(ifa->ifa_label, devname) == 0)
507 cfg->fc_prefsrc = ifa->ifa_local;
511 addr = sk_extract_addr(&rt->rt_gateway);
512 if (rt->rt_gateway.sa_family == AF_INET && addr) {
514 if (rt->rt_flags & RTF_GATEWAY &&
515 inet_addr_type(net, addr) == RTN_UNICAST)
516 cfg->fc_scope = RT_SCOPE_UNIVERSE;
519 if (cmd == SIOCDELRT)
522 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
525 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
526 cfg->fc_scope = RT_SCOPE_LINK;
528 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
532 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
536 if (rt->rt_flags & RTF_MTU)
537 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
539 if (rt->rt_flags & RTF_WINDOW)
540 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
542 if (rt->rt_flags & RTF_IRTT)
543 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
546 cfg->fc_mx_len = len;
553 * Handle IP routing ioctl calls.
554 * These are used to manipulate the routing tables
556 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
558 struct fib_config cfg;
563 case SIOCADDRT: /* Add a route */
564 case SIOCDELRT: /* Delete a route */
565 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
568 if (copy_from_user(&rt, arg, sizeof(rt)))
572 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
574 struct fib_table *tb;
576 if (cmd == SIOCDELRT) {
577 tb = fib_get_table(net, cfg.fc_table);
579 err = fib_table_delete(tb, &cfg);
583 tb = fib_new_table(net, cfg.fc_table);
585 err = fib_table_insert(tb, &cfg);
590 /* allocated by rtentry_to_fib_config() */
599 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
600 [RTA_DST] = { .type = NLA_U32 },
601 [RTA_SRC] = { .type = NLA_U32 },
602 [RTA_IIF] = { .type = NLA_U32 },
603 [RTA_OIF] = { .type = NLA_U32 },
604 [RTA_GATEWAY] = { .type = NLA_U32 },
605 [RTA_PRIORITY] = { .type = NLA_U32 },
606 [RTA_PREFSRC] = { .type = NLA_U32 },
607 [RTA_METRICS] = { .type = NLA_NESTED },
608 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
609 [RTA_FLOW] = { .type = NLA_U32 },
610 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
611 [RTA_ENCAP] = { .type = NLA_NESTED },
614 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
615 struct nlmsghdr *nlh, struct fib_config *cfg)
621 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
625 memset(cfg, 0, sizeof(*cfg));
627 rtm = nlmsg_data(nlh);
628 cfg->fc_dst_len = rtm->rtm_dst_len;
629 cfg->fc_tos = rtm->rtm_tos;
630 cfg->fc_table = rtm->rtm_table;
631 cfg->fc_protocol = rtm->rtm_protocol;
632 cfg->fc_scope = rtm->rtm_scope;
633 cfg->fc_type = rtm->rtm_type;
634 cfg->fc_flags = rtm->rtm_flags;
635 cfg->fc_nlflags = nlh->nlmsg_flags;
637 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
638 cfg->fc_nlinfo.nlh = nlh;
639 cfg->fc_nlinfo.nl_net = net;
641 if (cfg->fc_type > RTN_MAX) {
646 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
647 switch (nla_type(attr)) {
649 cfg->fc_dst = nla_get_be32(attr);
652 cfg->fc_oif = nla_get_u32(attr);
655 cfg->fc_gw = nla_get_be32(attr);
658 cfg->fc_priority = nla_get_u32(attr);
661 cfg->fc_prefsrc = nla_get_be32(attr);
664 cfg->fc_mx = nla_data(attr);
665 cfg->fc_mx_len = nla_len(attr);
668 cfg->fc_mp = nla_data(attr);
669 cfg->fc_mp_len = nla_len(attr);
672 cfg->fc_flow = nla_get_u32(attr);
675 cfg->fc_table = nla_get_u32(attr);
678 cfg->fc_encap = attr;
681 cfg->fc_encap_type = nla_get_u16(attr);
691 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
693 struct net *net = sock_net(skb->sk);
694 struct fib_config cfg;
695 struct fib_table *tb;
698 err = rtm_to_fib_config(net, skb, nlh, &cfg);
702 tb = fib_get_table(net, cfg.fc_table);
708 err = fib_table_delete(tb, &cfg);
713 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
715 struct net *net = sock_net(skb->sk);
716 struct fib_config cfg;
717 struct fib_table *tb;
720 err = rtm_to_fib_config(net, skb, nlh, &cfg);
724 tb = fib_new_table(net, cfg.fc_table);
730 err = fib_table_insert(tb, &cfg);
735 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
737 struct net *net = sock_net(skb->sk);
739 unsigned int e = 0, s_e;
740 struct fib_table *tb;
741 struct hlist_head *head;
744 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
745 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
753 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
755 head = &net->ipv4.fib_table_hash[h];
756 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
760 memset(&cb->args[2], 0, sizeof(cb->args) -
761 2 * sizeof(cb->args[0]));
762 if (fib_table_dump(tb, skb, cb) < 0)
778 /* Prepare and feed intra-kernel routing request.
779 * Really, it should be netlink message, but :-( netlink
780 * can be not configured, so that we feed it directly
781 * to fib engine. It is legal, because all events occur
782 * only when netlink is already locked.
784 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
786 struct net *net = dev_net(ifa->ifa_dev->dev);
787 struct fib_table *tb;
788 struct fib_config cfg = {
789 .fc_protocol = RTPROT_KERNEL,
792 .fc_dst_len = dst_len,
793 .fc_prefsrc = ifa->ifa_local,
794 .fc_oif = ifa->ifa_dev->dev->ifindex,
795 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
801 if (type == RTN_UNICAST)
802 tb = fib_new_table(net, RT_TABLE_MAIN);
804 tb = fib_new_table(net, RT_TABLE_LOCAL);
809 cfg.fc_table = tb->tb_id;
811 if (type != RTN_LOCAL)
812 cfg.fc_scope = RT_SCOPE_LINK;
814 cfg.fc_scope = RT_SCOPE_HOST;
816 if (cmd == RTM_NEWROUTE)
817 fib_table_insert(tb, &cfg);
819 fib_table_delete(tb, &cfg);
822 void fib_add_ifaddr(struct in_ifaddr *ifa)
824 struct in_device *in_dev = ifa->ifa_dev;
825 struct net_device *dev = in_dev->dev;
826 struct in_ifaddr *prim = ifa;
827 __be32 mask = ifa->ifa_mask;
828 __be32 addr = ifa->ifa_local;
829 __be32 prefix = ifa->ifa_address & mask;
831 if (ifa->ifa_flags & IFA_F_SECONDARY) {
832 prim = inet_ifa_byprefix(in_dev, prefix, mask);
834 pr_warn("%s: bug: prim == NULL\n", __func__);
839 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
841 if (!(dev->flags & IFF_UP))
844 /* Add broadcast address, if it is explicitly assigned. */
845 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
846 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
848 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
849 (prefix != addr || ifa->ifa_prefixlen < 32)) {
850 fib_magic(RTM_NEWROUTE,
851 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
852 prefix, ifa->ifa_prefixlen, prim);
854 /* Add network specific broadcasts, when it takes a sense */
855 if (ifa->ifa_prefixlen < 31) {
856 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
857 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
863 /* Delete primary or secondary address.
864 * Optionally, on secondary address promotion consider the addresses
865 * from subnet iprim as deleted, even if they are in device list.
866 * In this case the secondary ifa can be in device list.
868 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
870 struct in_device *in_dev = ifa->ifa_dev;
871 struct net_device *dev = in_dev->dev;
872 struct in_ifaddr *ifa1;
873 struct in_ifaddr *prim = ifa, *prim1 = NULL;
874 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
875 __be32 any = ifa->ifa_address & ifa->ifa_mask;
881 int subnet = 0; /* Primary network */
882 int gone = 1; /* Address is missing */
883 int same_prefsrc = 0; /* Another primary with same IP */
885 if (ifa->ifa_flags & IFA_F_SECONDARY) {
886 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
888 pr_warn("%s: bug: prim == NULL\n", __func__);
891 if (iprim && iprim != prim) {
892 pr_warn("%s: bug: iprim != prim\n", __func__);
895 } else if (!ipv4_is_zeronet(any) &&
896 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
897 fib_magic(RTM_DELROUTE,
898 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
899 any, ifa->ifa_prefixlen, prim);
903 /* Deletion is more complicated than add.
904 * We should take care of not to delete too much :-)
906 * Scan address list to be sure that addresses are really gone.
909 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
911 /* promotion, keep the IP */
915 /* Ignore IFAs from our subnet */
916 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
917 inet_ifa_match(ifa1->ifa_address, iprim))
920 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
921 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
922 /* Another address from our subnet? */
923 if (ifa1->ifa_mask == prim->ifa_mask &&
924 inet_ifa_match(ifa1->ifa_address, prim))
927 /* We reached the secondaries, so
928 * same_prefsrc should be determined.
932 /* Search new prim1 if ifa1 is not
933 * using the current prim1
936 ifa1->ifa_mask != prim1->ifa_mask ||
937 !inet_ifa_match(ifa1->ifa_address, prim1))
938 prim1 = inet_ifa_byprefix(in_dev,
943 if (prim1->ifa_local != prim->ifa_local)
947 if (prim->ifa_local != ifa1->ifa_local)
953 if (ifa->ifa_local == ifa1->ifa_local)
955 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
957 if (brd == ifa1->ifa_broadcast)
959 if (any == ifa1->ifa_broadcast)
961 /* primary has network specific broadcasts */
962 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
963 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
964 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
966 if (!ipv4_is_zeronet(any1)) {
967 if (ifa->ifa_broadcast == brd1 ||
968 ifa->ifa_broadcast == any1)
970 if (brd == brd1 || brd == any1)
972 if (any == brd1 || any == any1)
979 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
980 if (subnet && ifa->ifa_prefixlen < 31) {
982 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
984 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
986 if (!(ok & LOCAL_OK)) {
987 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
989 /* Check, that this local address finally disappeared. */
991 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
992 /* And the last, but not the least thing.
993 * We must flush stray FIB entries.
995 * First of all, we scan fib_info list searching
996 * for stray nexthop entries, then ignite fib_flush.
998 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
999 fib_flush(dev_net(dev));
1008 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1011 struct fib_result res;
1012 struct flowi4 fl4 = {
1013 .flowi4_mark = frn->fl_mark,
1014 .daddr = frn->fl_addr,
1015 .flowi4_tos = frn->fl_tos,
1016 .flowi4_scope = frn->fl_scope,
1018 struct fib_table *tb;
1022 tb = fib_get_table(net, frn->tb_id_in);
1028 frn->tb_id = tb->tb_id;
1029 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1032 frn->prefixlen = res.prefixlen;
1033 frn->nh_sel = res.nh_sel;
1034 frn->type = res.type;
1035 frn->scope = res.scope;
1043 static void nl_fib_input(struct sk_buff *skb)
1046 struct fib_result_nl *frn;
1047 struct nlmsghdr *nlh;
1050 net = sock_net(skb->sk);
1051 nlh = nlmsg_hdr(skb);
1052 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1053 nlmsg_len(nlh) < sizeof(*frn))
1056 skb = netlink_skb_clone(skb, GFP_KERNEL);
1059 nlh = nlmsg_hdr(skb);
1061 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1062 nl_fib_lookup(net, frn);
1064 portid = NETLINK_CB(skb).portid; /* netlink portid */
1065 NETLINK_CB(skb).portid = 0; /* from kernel */
1066 NETLINK_CB(skb).dst_group = 0; /* unicast */
1067 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1070 static int __net_init nl_fib_lookup_init(struct net *net)
1073 struct netlink_kernel_cfg cfg = {
1074 .input = nl_fib_input,
1077 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1079 return -EAFNOSUPPORT;
1080 net->ipv4.fibnl = sk;
1084 static void nl_fib_lookup_exit(struct net *net)
1086 netlink_kernel_release(net->ipv4.fibnl);
1087 net->ipv4.fibnl = NULL;
1090 static void fib_disable_ip(struct net_device *dev, unsigned long event)
1092 if (fib_sync_down_dev(dev, event))
1093 fib_flush(dev_net(dev));
1094 rt_cache_flush(dev_net(dev));
1098 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1100 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1101 struct net_device *dev = ifa->ifa_dev->dev;
1102 struct net *net = dev_net(dev);
1106 fib_add_ifaddr(ifa);
1107 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1108 fib_sync_up(dev, RTNH_F_DEAD);
1110 atomic_inc(&net->ipv4.dev_addr_genid);
1111 rt_cache_flush(dev_net(dev));
1114 fib_del_ifaddr(ifa, NULL);
1115 atomic_inc(&net->ipv4.dev_addr_genid);
1116 if (!ifa->ifa_dev->ifa_list) {
1117 /* Last address was deleted from this interface.
1120 fib_disable_ip(dev, event);
1122 rt_cache_flush(dev_net(dev));
1129 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1131 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1132 struct in_device *in_dev;
1133 struct net *net = dev_net(dev);
1136 if (event == NETDEV_UNREGISTER) {
1137 fib_disable_ip(dev, event);
1142 in_dev = __in_dev_get_rtnl(dev);
1149 fib_add_ifaddr(ifa);
1150 } endfor_ifa(in_dev);
1151 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1152 fib_sync_up(dev, RTNH_F_DEAD);
1154 atomic_inc(&net->ipv4.dev_addr_genid);
1155 rt_cache_flush(net);
1158 fib_disable_ip(dev, event);
1161 flags = dev_get_flags(dev);
1162 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1163 fib_sync_up(dev, RTNH_F_LINKDOWN);
1165 fib_sync_down_dev(dev, event);
1167 case NETDEV_CHANGEMTU:
1168 rt_cache_flush(net);
1174 static struct notifier_block fib_inetaddr_notifier = {
1175 .notifier_call = fib_inetaddr_event,
1178 static struct notifier_block fib_netdev_notifier = {
1179 .notifier_call = fib_netdev_event,
1182 static int __net_init ip_fib_net_init(struct net *net)
1185 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1187 /* Avoid false sharing : Use at least a full cache line */
1188 size = max_t(size_t, size, L1_CACHE_BYTES);
1190 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1191 if (!net->ipv4.fib_table_hash)
1194 err = fib4_rules_init(net);
1200 kfree(net->ipv4.fib_table_hash);
1204 static void ip_fib_net_exit(struct net *net)
1209 #ifdef CONFIG_IP_MULTIPLE_TABLES
1210 RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1211 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1212 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1214 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1215 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1216 struct hlist_node *tmp;
1217 struct fib_table *tb;
1219 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1220 hlist_del(&tb->tb_hlist);
1221 fib_table_flush(tb);
1226 #ifdef CONFIG_IP_MULTIPLE_TABLES
1227 fib4_rules_exit(net);
1230 kfree(net->ipv4.fib_table_hash);
1233 static int __net_init fib_net_init(struct net *net)
1237 #ifdef CONFIG_IP_ROUTE_CLASSID
1238 net->ipv4.fib_num_tclassid_users = 0;
1240 error = ip_fib_net_init(net);
1243 error = nl_fib_lookup_init(net);
1246 error = fib_proc_init(net);
1253 nl_fib_lookup_exit(net);
1255 ip_fib_net_exit(net);
1259 static void __net_exit fib_net_exit(struct net *net)
1262 nl_fib_lookup_exit(net);
1263 ip_fib_net_exit(net);
1266 static struct pernet_operations fib_net_ops = {
1267 .init = fib_net_init,
1268 .exit = fib_net_exit,
1271 void __init ip_fib_init(void)
1273 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1274 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1275 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1277 register_pernet_subsys(&fib_net_ops);
1278 register_netdevice_notifier(&fib_netdev_notifier);
1279 register_inetaddr_notifier(&fib_inetaddr_notifier);