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 /* inet_addr_type with dev == NULL but using the table from a dev
264 * if one is associated
266 unsigned int inet_addr_type_dev_table(struct net *net,
267 const struct net_device *dev,
270 int rt_table = vrf_dev_table(dev) ? : RT_TABLE_LOCAL;
272 return __inet_dev_addr_type(net, NULL, addr, rt_table);
274 EXPORT_SYMBOL(inet_addr_type_dev_table);
276 __be32 fib_compute_spec_dst(struct sk_buff *skb)
278 struct net_device *dev = skb->dev;
279 struct in_device *in_dev;
280 struct fib_result res;
286 rt = skb_rtable(skb);
287 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
289 return ip_hdr(skb)->daddr;
291 in_dev = __in_dev_get_rcu(dev);
296 scope = RT_SCOPE_UNIVERSE;
297 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
299 fl4.flowi4_iif = LOOPBACK_IFINDEX;
300 fl4.daddr = ip_hdr(skb)->saddr;
302 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
303 fl4.flowi4_scope = scope;
304 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
305 fl4.flowi4_tun_key.tun_id = 0;
306 if (!fib_lookup(net, &fl4, &res, 0))
307 return FIB_RES_PREFSRC(net, res);
309 scope = RT_SCOPE_LINK;
312 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
315 /* Given (packet source, input interface) and optional (dst, oif, tos):
316 * - (main) check, that source is valid i.e. not broadcast or our local
318 * - figure out what "logical" interface this packet arrived
319 * and calculate "specific destination" address.
320 * - check, that packet arrived from expected physical interface.
321 * called with rcu_read_lock()
323 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
324 u8 tos, int oif, struct net_device *dev,
325 int rpf, struct in_device *idev, u32 *itag)
328 struct fib_result res;
334 fl4.flowi4_iif = vrf_master_ifindex_rcu(dev);
336 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
339 fl4.flowi4_tos = tos;
340 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
341 fl4.flowi4_tun_key.tun_id = 0;
343 no_addr = idev->ifa_list == NULL;
345 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
348 if (fib_lookup(net, &fl4, &res, 0))
350 if (res.type != RTN_UNICAST &&
351 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
353 if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
354 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
356 fib_combine_itag(itag, &res);
359 #ifdef CONFIG_IP_ROUTE_MULTIPATH
360 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
361 struct fib_nh *nh = &res.fi->fib_nh[ret];
363 if (nh->nh_dev == dev) {
366 } else if (vrf_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
372 if (FIB_RES_DEV(res) == dev)
376 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
383 fl4.flowi4_oif = dev->ifindex;
386 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
387 if (res.type == RTN_UNICAST)
388 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
404 /* Ignore rp_filter for packets protected by IPsec. */
405 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
406 u8 tos, int oif, struct net_device *dev,
407 struct in_device *idev, u32 *itag)
409 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
411 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
412 IN_DEV_ACCEPT_LOCAL(idev) &&
413 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
417 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
420 static inline __be32 sk_extract_addr(struct sockaddr *addr)
422 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
425 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
429 nla = (struct nlattr *) ((char *) mx + len);
430 nla->nla_type = type;
431 nla->nla_len = nla_attr_size(4);
432 *(u32 *) nla_data(nla) = value;
434 return len + nla_total_size(4);
437 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
438 struct fib_config *cfg)
443 memset(cfg, 0, sizeof(*cfg));
444 cfg->fc_nlinfo.nl_net = net;
446 if (rt->rt_dst.sa_family != AF_INET)
447 return -EAFNOSUPPORT;
450 * Check mask for validity:
451 * a) it must be contiguous.
452 * b) destination must have all host bits clear.
453 * c) if application forgot to set correct family (AF_INET),
454 * reject request unless it is absolutely clear i.e.
455 * both family and mask are zero.
458 addr = sk_extract_addr(&rt->rt_dst);
459 if (!(rt->rt_flags & RTF_HOST)) {
460 __be32 mask = sk_extract_addr(&rt->rt_genmask);
462 if (rt->rt_genmask.sa_family != AF_INET) {
463 if (mask || rt->rt_genmask.sa_family)
464 return -EAFNOSUPPORT;
467 if (bad_mask(mask, addr))
470 plen = inet_mask_len(mask);
473 cfg->fc_dst_len = plen;
476 if (cmd != SIOCDELRT) {
477 cfg->fc_nlflags = NLM_F_CREATE;
478 cfg->fc_protocol = RTPROT_BOOT;
482 cfg->fc_priority = rt->rt_metric - 1;
484 if (rt->rt_flags & RTF_REJECT) {
485 cfg->fc_scope = RT_SCOPE_HOST;
486 cfg->fc_type = RTN_UNREACHABLE;
490 cfg->fc_scope = RT_SCOPE_NOWHERE;
491 cfg->fc_type = RTN_UNICAST;
495 struct net_device *dev;
496 char devname[IFNAMSIZ];
498 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
501 devname[IFNAMSIZ-1] = 0;
502 colon = strchr(devname, ':');
505 dev = __dev_get_by_name(net, devname);
508 cfg->fc_oif = dev->ifindex;
510 struct in_ifaddr *ifa;
511 struct in_device *in_dev = __in_dev_get_rtnl(dev);
515 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
516 if (strcmp(ifa->ifa_label, devname) == 0)
520 cfg->fc_prefsrc = ifa->ifa_local;
524 addr = sk_extract_addr(&rt->rt_gateway);
525 if (rt->rt_gateway.sa_family == AF_INET && addr) {
526 unsigned int addr_type;
529 addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
530 if (rt->rt_flags & RTF_GATEWAY &&
531 addr_type == RTN_UNICAST)
532 cfg->fc_scope = RT_SCOPE_UNIVERSE;
535 if (cmd == SIOCDELRT)
538 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
541 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
542 cfg->fc_scope = RT_SCOPE_LINK;
544 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
548 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
552 if (rt->rt_flags & RTF_MTU)
553 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
555 if (rt->rt_flags & RTF_WINDOW)
556 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
558 if (rt->rt_flags & RTF_IRTT)
559 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
562 cfg->fc_mx_len = len;
569 * Handle IP routing ioctl calls.
570 * These are used to manipulate the routing tables
572 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
574 struct fib_config cfg;
579 case SIOCADDRT: /* Add a route */
580 case SIOCDELRT: /* Delete a route */
581 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
584 if (copy_from_user(&rt, arg, sizeof(rt)))
588 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
590 struct fib_table *tb;
592 if (cmd == SIOCDELRT) {
593 tb = fib_get_table(net, cfg.fc_table);
595 err = fib_table_delete(tb, &cfg);
599 tb = fib_new_table(net, cfg.fc_table);
601 err = fib_table_insert(tb, &cfg);
606 /* allocated by rtentry_to_fib_config() */
615 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
616 [RTA_DST] = { .type = NLA_U32 },
617 [RTA_SRC] = { .type = NLA_U32 },
618 [RTA_IIF] = { .type = NLA_U32 },
619 [RTA_OIF] = { .type = NLA_U32 },
620 [RTA_GATEWAY] = { .type = NLA_U32 },
621 [RTA_PRIORITY] = { .type = NLA_U32 },
622 [RTA_PREFSRC] = { .type = NLA_U32 },
623 [RTA_METRICS] = { .type = NLA_NESTED },
624 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
625 [RTA_FLOW] = { .type = NLA_U32 },
626 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
627 [RTA_ENCAP] = { .type = NLA_NESTED },
630 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
631 struct nlmsghdr *nlh, struct fib_config *cfg)
637 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
641 memset(cfg, 0, sizeof(*cfg));
643 rtm = nlmsg_data(nlh);
644 cfg->fc_dst_len = rtm->rtm_dst_len;
645 cfg->fc_tos = rtm->rtm_tos;
646 cfg->fc_table = rtm->rtm_table;
647 cfg->fc_protocol = rtm->rtm_protocol;
648 cfg->fc_scope = rtm->rtm_scope;
649 cfg->fc_type = rtm->rtm_type;
650 cfg->fc_flags = rtm->rtm_flags;
651 cfg->fc_nlflags = nlh->nlmsg_flags;
653 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
654 cfg->fc_nlinfo.nlh = nlh;
655 cfg->fc_nlinfo.nl_net = net;
657 if (cfg->fc_type > RTN_MAX) {
662 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
663 switch (nla_type(attr)) {
665 cfg->fc_dst = nla_get_be32(attr);
668 cfg->fc_oif = nla_get_u32(attr);
671 cfg->fc_gw = nla_get_be32(attr);
674 cfg->fc_priority = nla_get_u32(attr);
677 cfg->fc_prefsrc = nla_get_be32(attr);
680 cfg->fc_mx = nla_data(attr);
681 cfg->fc_mx_len = nla_len(attr);
684 cfg->fc_mp = nla_data(attr);
685 cfg->fc_mp_len = nla_len(attr);
688 cfg->fc_flow = nla_get_u32(attr);
691 cfg->fc_table = nla_get_u32(attr);
694 cfg->fc_encap = attr;
697 cfg->fc_encap_type = nla_get_u16(attr);
707 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
709 struct net *net = sock_net(skb->sk);
710 struct fib_config cfg;
711 struct fib_table *tb;
714 err = rtm_to_fib_config(net, skb, nlh, &cfg);
718 tb = fib_get_table(net, cfg.fc_table);
724 err = fib_table_delete(tb, &cfg);
729 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
731 struct net *net = sock_net(skb->sk);
732 struct fib_config cfg;
733 struct fib_table *tb;
736 err = rtm_to_fib_config(net, skb, nlh, &cfg);
740 tb = fib_new_table(net, cfg.fc_table);
746 err = fib_table_insert(tb, &cfg);
751 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
753 struct net *net = sock_net(skb->sk);
755 unsigned int e = 0, s_e;
756 struct fib_table *tb;
757 struct hlist_head *head;
760 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
761 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
769 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
771 head = &net->ipv4.fib_table_hash[h];
772 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
776 memset(&cb->args[2], 0, sizeof(cb->args) -
777 2 * sizeof(cb->args[0]));
778 if (fib_table_dump(tb, skb, cb) < 0)
794 /* Prepare and feed intra-kernel routing request.
795 * Really, it should be netlink message, but :-( netlink
796 * can be not configured, so that we feed it directly
797 * to fib engine. It is legal, because all events occur
798 * only when netlink is already locked.
800 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
802 struct net *net = dev_net(ifa->ifa_dev->dev);
803 int tb_id = vrf_dev_table_rtnl(ifa->ifa_dev->dev);
804 struct fib_table *tb;
805 struct fib_config cfg = {
806 .fc_protocol = RTPROT_KERNEL,
809 .fc_dst_len = dst_len,
810 .fc_prefsrc = ifa->ifa_local,
811 .fc_oif = ifa->ifa_dev->dev->ifindex,
812 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
819 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
821 tb = fib_new_table(net, tb_id);
825 cfg.fc_table = tb->tb_id;
827 if (type != RTN_LOCAL)
828 cfg.fc_scope = RT_SCOPE_LINK;
830 cfg.fc_scope = RT_SCOPE_HOST;
832 if (cmd == RTM_NEWROUTE)
833 fib_table_insert(tb, &cfg);
835 fib_table_delete(tb, &cfg);
838 void fib_add_ifaddr(struct in_ifaddr *ifa)
840 struct in_device *in_dev = ifa->ifa_dev;
841 struct net_device *dev = in_dev->dev;
842 struct in_ifaddr *prim = ifa;
843 __be32 mask = ifa->ifa_mask;
844 __be32 addr = ifa->ifa_local;
845 __be32 prefix = ifa->ifa_address & mask;
847 if (ifa->ifa_flags & IFA_F_SECONDARY) {
848 prim = inet_ifa_byprefix(in_dev, prefix, mask);
850 pr_warn("%s: bug: prim == NULL\n", __func__);
855 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
857 if (!(dev->flags & IFF_UP))
860 /* Add broadcast address, if it is explicitly assigned. */
861 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
862 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
864 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
865 (prefix != addr || ifa->ifa_prefixlen < 32)) {
866 fib_magic(RTM_NEWROUTE,
867 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
868 prefix, ifa->ifa_prefixlen, prim);
870 /* Add network specific broadcasts, when it takes a sense */
871 if (ifa->ifa_prefixlen < 31) {
872 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
873 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
879 /* Delete primary or secondary address.
880 * Optionally, on secondary address promotion consider the addresses
881 * from subnet iprim as deleted, even if they are in device list.
882 * In this case the secondary ifa can be in device list.
884 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
886 struct in_device *in_dev = ifa->ifa_dev;
887 struct net_device *dev = in_dev->dev;
888 struct in_ifaddr *ifa1;
889 struct in_ifaddr *prim = ifa, *prim1 = NULL;
890 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
891 __be32 any = ifa->ifa_address & ifa->ifa_mask;
897 int subnet = 0; /* Primary network */
898 int gone = 1; /* Address is missing */
899 int same_prefsrc = 0; /* Another primary with same IP */
901 if (ifa->ifa_flags & IFA_F_SECONDARY) {
902 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
904 pr_warn("%s: bug: prim == NULL\n", __func__);
907 if (iprim && iprim != prim) {
908 pr_warn("%s: bug: iprim != prim\n", __func__);
911 } else if (!ipv4_is_zeronet(any) &&
912 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
913 fib_magic(RTM_DELROUTE,
914 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
915 any, ifa->ifa_prefixlen, prim);
919 /* Deletion is more complicated than add.
920 * We should take care of not to delete too much :-)
922 * Scan address list to be sure that addresses are really gone.
925 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
927 /* promotion, keep the IP */
931 /* Ignore IFAs from our subnet */
932 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
933 inet_ifa_match(ifa1->ifa_address, iprim))
936 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
937 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
938 /* Another address from our subnet? */
939 if (ifa1->ifa_mask == prim->ifa_mask &&
940 inet_ifa_match(ifa1->ifa_address, prim))
943 /* We reached the secondaries, so
944 * same_prefsrc should be determined.
948 /* Search new prim1 if ifa1 is not
949 * using the current prim1
952 ifa1->ifa_mask != prim1->ifa_mask ||
953 !inet_ifa_match(ifa1->ifa_address, prim1))
954 prim1 = inet_ifa_byprefix(in_dev,
959 if (prim1->ifa_local != prim->ifa_local)
963 if (prim->ifa_local != ifa1->ifa_local)
969 if (ifa->ifa_local == ifa1->ifa_local)
971 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
973 if (brd == ifa1->ifa_broadcast)
975 if (any == ifa1->ifa_broadcast)
977 /* primary has network specific broadcasts */
978 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
979 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
980 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
982 if (!ipv4_is_zeronet(any1)) {
983 if (ifa->ifa_broadcast == brd1 ||
984 ifa->ifa_broadcast == any1)
986 if (brd == brd1 || brd == any1)
988 if (any == brd1 || any == any1)
995 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
996 if (subnet && ifa->ifa_prefixlen < 31) {
998 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
1000 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
1002 if (!(ok & LOCAL_OK)) {
1003 unsigned int addr_type;
1005 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
1007 /* Check, that this local address finally disappeared. */
1008 addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1010 if (gone && addr_type != RTN_LOCAL) {
1011 /* And the last, but not the least thing.
1012 * We must flush stray FIB entries.
1014 * First of all, we scan fib_info list searching
1015 * for stray nexthop entries, then ignite fib_flush.
1017 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
1018 fib_flush(dev_net(dev));
1027 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1030 struct fib_result res;
1031 struct flowi4 fl4 = {
1032 .flowi4_mark = frn->fl_mark,
1033 .daddr = frn->fl_addr,
1034 .flowi4_tos = frn->fl_tos,
1035 .flowi4_scope = frn->fl_scope,
1037 struct fib_table *tb;
1041 tb = fib_get_table(net, frn->tb_id_in);
1047 frn->tb_id = tb->tb_id;
1048 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1051 frn->prefixlen = res.prefixlen;
1052 frn->nh_sel = res.nh_sel;
1053 frn->type = res.type;
1054 frn->scope = res.scope;
1062 static void nl_fib_input(struct sk_buff *skb)
1065 struct fib_result_nl *frn;
1066 struct nlmsghdr *nlh;
1069 net = sock_net(skb->sk);
1070 nlh = nlmsg_hdr(skb);
1071 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1072 nlmsg_len(nlh) < sizeof(*frn))
1075 skb = netlink_skb_clone(skb, GFP_KERNEL);
1078 nlh = nlmsg_hdr(skb);
1080 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1081 nl_fib_lookup(net, frn);
1083 portid = NETLINK_CB(skb).portid; /* netlink portid */
1084 NETLINK_CB(skb).portid = 0; /* from kernel */
1085 NETLINK_CB(skb).dst_group = 0; /* unicast */
1086 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1089 static int __net_init nl_fib_lookup_init(struct net *net)
1092 struct netlink_kernel_cfg cfg = {
1093 .input = nl_fib_input,
1096 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1098 return -EAFNOSUPPORT;
1099 net->ipv4.fibnl = sk;
1103 static void nl_fib_lookup_exit(struct net *net)
1105 netlink_kernel_release(net->ipv4.fibnl);
1106 net->ipv4.fibnl = NULL;
1109 static void fib_disable_ip(struct net_device *dev, unsigned long event)
1111 if (fib_sync_down_dev(dev, event))
1112 fib_flush(dev_net(dev));
1113 rt_cache_flush(dev_net(dev));
1117 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1119 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1120 struct net_device *dev = ifa->ifa_dev->dev;
1121 struct net *net = dev_net(dev);
1125 fib_add_ifaddr(ifa);
1126 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1127 fib_sync_up(dev, RTNH_F_DEAD);
1129 atomic_inc(&net->ipv4.dev_addr_genid);
1130 rt_cache_flush(dev_net(dev));
1133 fib_del_ifaddr(ifa, NULL);
1134 atomic_inc(&net->ipv4.dev_addr_genid);
1135 if (!ifa->ifa_dev->ifa_list) {
1136 /* Last address was deleted from this interface.
1139 fib_disable_ip(dev, event);
1141 rt_cache_flush(dev_net(dev));
1148 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1150 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1151 struct in_device *in_dev;
1152 struct net *net = dev_net(dev);
1155 if (event == NETDEV_UNREGISTER) {
1156 fib_disable_ip(dev, event);
1161 in_dev = __in_dev_get_rtnl(dev);
1168 fib_add_ifaddr(ifa);
1169 } endfor_ifa(in_dev);
1170 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1171 fib_sync_up(dev, RTNH_F_DEAD);
1173 atomic_inc(&net->ipv4.dev_addr_genid);
1174 rt_cache_flush(net);
1177 fib_disable_ip(dev, event);
1180 flags = dev_get_flags(dev);
1181 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1182 fib_sync_up(dev, RTNH_F_LINKDOWN);
1184 fib_sync_down_dev(dev, event);
1186 case NETDEV_CHANGEMTU:
1187 rt_cache_flush(net);
1193 static struct notifier_block fib_inetaddr_notifier = {
1194 .notifier_call = fib_inetaddr_event,
1197 static struct notifier_block fib_netdev_notifier = {
1198 .notifier_call = fib_netdev_event,
1201 static int __net_init ip_fib_net_init(struct net *net)
1204 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1206 /* Avoid false sharing : Use at least a full cache line */
1207 size = max_t(size_t, size, L1_CACHE_BYTES);
1209 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1210 if (!net->ipv4.fib_table_hash)
1213 err = fib4_rules_init(net);
1219 kfree(net->ipv4.fib_table_hash);
1223 static void ip_fib_net_exit(struct net *net)
1228 #ifdef CONFIG_IP_MULTIPLE_TABLES
1229 RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1230 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1231 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1233 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1234 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1235 struct hlist_node *tmp;
1236 struct fib_table *tb;
1238 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1239 hlist_del(&tb->tb_hlist);
1240 fib_table_flush(tb);
1245 #ifdef CONFIG_IP_MULTIPLE_TABLES
1246 fib4_rules_exit(net);
1249 kfree(net->ipv4.fib_table_hash);
1252 static int __net_init fib_net_init(struct net *net)
1256 #ifdef CONFIG_IP_ROUTE_CLASSID
1257 net->ipv4.fib_num_tclassid_users = 0;
1259 error = ip_fib_net_init(net);
1262 error = nl_fib_lookup_init(net);
1265 error = fib_proc_init(net);
1272 nl_fib_lookup_exit(net);
1274 ip_fib_net_exit(net);
1278 static void __net_exit fib_net_exit(struct net *net)
1281 nl_fib_lookup_exit(net);
1282 ip_fib_net_exit(net);
1285 static struct pernet_operations fib_net_ops = {
1286 .init = fib_net_init,
1287 .exit = fib_net_exit,
1290 void __init ip_fib_init(void)
1292 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1293 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1294 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1296 register_pernet_subsys(&fib_net_ops);
1297 register_netdevice_notifier(&fib_netdev_notifier);
1298 register_inetaddr_notifier(&fib_inetaddr_notifier);