2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * IPv4 Forwarding Information Base: FIB frontend.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <linux/module.h>
17 #include <asm/uaccess.h>
18 #include <linux/bitops.h>
19 #include <linux/capability.h>
20 #include <linux/types.h>
21 #include <linux/kernel.h>
23 #include <linux/string.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/errno.h>
28 #include <linux/inet.h>
29 #include <linux/inetdevice.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_addr.h>
32 #include <linux/if_arp.h>
33 #include <linux/skbuff.h>
34 #include <linux/cache.h>
35 #include <linux/init.h>
36 #include <linux/list.h>
37 #include <linux/slab.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
45 #include <net/ip_fib.h>
46 #include <net/rtnetlink.h>
49 #ifndef CONFIG_IP_MULTIPLE_TABLES
51 static int __net_init fib4_rules_init(struct net *net)
53 struct fib_table *local_table, *main_table;
55 local_table = fib_trie_table(RT_TABLE_LOCAL);
56 if (local_table == NULL)
59 main_table = fib_trie_table(RT_TABLE_MAIN);
60 if (main_table == NULL)
63 hlist_add_head_rcu(&local_table->tb_hlist,
64 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
65 hlist_add_head_rcu(&main_table->tb_hlist,
66 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
75 struct fib_table *fib_new_table(struct net *net, u32 id)
82 tb = fib_get_table(net, id);
86 tb = fib_trie_table(id);
92 net->ipv4.fib_local = tb;
96 net->ipv4.fib_main = tb;
99 case RT_TABLE_DEFAULT:
100 net->ipv4.fib_default = tb;
107 h = id & (FIB_TABLE_HASHSZ - 1);
108 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
112 struct fib_table *fib_get_table(struct net *net, u32 id)
114 struct fib_table *tb;
115 struct hlist_head *head;
120 h = id & (FIB_TABLE_HASHSZ - 1);
123 head = &net->ipv4.fib_table_hash[h];
124 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
125 if (tb->tb_id == id) {
133 #endif /* CONFIG_IP_MULTIPLE_TABLES */
135 static void fib_flush(struct net *net)
138 struct fib_table *tb;
139 struct hlist_head *head;
142 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
143 head = &net->ipv4.fib_table_hash[h];
144 hlist_for_each_entry(tb, head, tb_hlist)
145 flushed += fib_table_flush(tb);
153 * Find address type as if only "dev" was present in the system. If
154 * on_dev is NULL then all interfaces are taken into consideration.
156 static inline unsigned int __inet_dev_addr_type(struct net *net,
157 const struct net_device *dev,
160 struct flowi4 fl4 = { .daddr = addr };
161 struct fib_result res;
162 unsigned int ret = RTN_BROADCAST;
163 struct fib_table *local_table;
165 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
166 return RTN_BROADCAST;
167 if (ipv4_is_multicast(addr))
168 return RTN_MULTICAST;
170 local_table = fib_get_table(net, RT_TABLE_LOCAL);
174 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
175 if (!dev || dev == res.fi->fib_dev)
183 unsigned int inet_addr_type(struct net *net, __be32 addr)
185 return __inet_dev_addr_type(net, NULL, addr);
187 EXPORT_SYMBOL(inet_addr_type);
189 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
192 return __inet_dev_addr_type(net, dev, addr);
194 EXPORT_SYMBOL(inet_dev_addr_type);
196 __be32 fib_compute_spec_dst(struct sk_buff *skb)
198 struct net_device *dev = skb->dev;
199 struct in_device *in_dev;
200 struct fib_result res;
206 rt = skb_rtable(skb);
207 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
209 return ip_hdr(skb)->daddr;
211 in_dev = __in_dev_get_rcu(dev);
216 scope = RT_SCOPE_UNIVERSE;
217 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
219 fl4.flowi4_iif = LOOPBACK_IFINDEX;
220 fl4.daddr = ip_hdr(skb)->saddr;
222 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
223 fl4.flowi4_scope = scope;
224 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
225 if (!fib_lookup(net, &fl4, &res))
226 return FIB_RES_PREFSRC(net, res);
228 scope = RT_SCOPE_LINK;
231 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
234 /* Given (packet source, input interface) and optional (dst, oif, tos):
235 * - (main) check, that source is valid i.e. not broadcast or our local
237 * - figure out what "logical" interface this packet arrived
238 * and calculate "specific destination" address.
239 * - check, that packet arrived from expected physical interface.
240 * called with rcu_read_lock()
242 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
243 u8 tos, int oif, struct net_device *dev,
244 int rpf, struct in_device *idev, u32 *itag)
246 int ret, no_addr, accept_local;
247 struct fib_result res;
253 fl4.flowi4_iif = oif;
256 fl4.flowi4_tos = tos;
257 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
259 no_addr = idev->ifa_list == NULL;
261 accept_local = IN_DEV_ACCEPT_LOCAL(idev);
262 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
265 if (fib_lookup(net, &fl4, &res))
267 if (res.type != RTN_UNICAST) {
268 if (res.type != RTN_LOCAL || !accept_local)
271 fib_combine_itag(itag, &res);
274 #ifdef CONFIG_IP_ROUTE_MULTIPATH
275 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
276 struct fib_nh *nh = &res.fi->fib_nh[ret];
278 if (nh->nh_dev == dev) {
284 if (FIB_RES_DEV(res) == dev)
288 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
295 fl4.flowi4_oif = dev->ifindex;
298 if (fib_lookup(net, &fl4, &res) == 0) {
299 if (res.type == RTN_UNICAST)
300 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
316 /* Ignore rp_filter for packets protected by IPsec. */
317 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
318 u8 tos, int oif, struct net_device *dev,
319 struct in_device *idev, u32 *itag)
321 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
323 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
324 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
328 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
331 static inline __be32 sk_extract_addr(struct sockaddr *addr)
333 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
336 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
340 nla = (struct nlattr *) ((char *) mx + len);
341 nla->nla_type = type;
342 nla->nla_len = nla_attr_size(4);
343 *(u32 *) nla_data(nla) = value;
345 return len + nla_total_size(4);
348 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
349 struct fib_config *cfg)
354 memset(cfg, 0, sizeof(*cfg));
355 cfg->fc_nlinfo.nl_net = net;
357 if (rt->rt_dst.sa_family != AF_INET)
358 return -EAFNOSUPPORT;
361 * Check mask for validity:
362 * a) it must be contiguous.
363 * b) destination must have all host bits clear.
364 * c) if application forgot to set correct family (AF_INET),
365 * reject request unless it is absolutely clear i.e.
366 * both family and mask are zero.
369 addr = sk_extract_addr(&rt->rt_dst);
370 if (!(rt->rt_flags & RTF_HOST)) {
371 __be32 mask = sk_extract_addr(&rt->rt_genmask);
373 if (rt->rt_genmask.sa_family != AF_INET) {
374 if (mask || rt->rt_genmask.sa_family)
375 return -EAFNOSUPPORT;
378 if (bad_mask(mask, addr))
381 plen = inet_mask_len(mask);
384 cfg->fc_dst_len = plen;
387 if (cmd != SIOCDELRT) {
388 cfg->fc_nlflags = NLM_F_CREATE;
389 cfg->fc_protocol = RTPROT_BOOT;
393 cfg->fc_priority = rt->rt_metric - 1;
395 if (rt->rt_flags & RTF_REJECT) {
396 cfg->fc_scope = RT_SCOPE_HOST;
397 cfg->fc_type = RTN_UNREACHABLE;
401 cfg->fc_scope = RT_SCOPE_NOWHERE;
402 cfg->fc_type = RTN_UNICAST;
406 struct net_device *dev;
407 char devname[IFNAMSIZ];
409 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
412 devname[IFNAMSIZ-1] = 0;
413 colon = strchr(devname, ':');
416 dev = __dev_get_by_name(net, devname);
419 cfg->fc_oif = dev->ifindex;
421 struct in_ifaddr *ifa;
422 struct in_device *in_dev = __in_dev_get_rtnl(dev);
426 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
427 if (strcmp(ifa->ifa_label, devname) == 0)
431 cfg->fc_prefsrc = ifa->ifa_local;
435 addr = sk_extract_addr(&rt->rt_gateway);
436 if (rt->rt_gateway.sa_family == AF_INET && addr) {
438 if (rt->rt_flags & RTF_GATEWAY &&
439 inet_addr_type(net, addr) == RTN_UNICAST)
440 cfg->fc_scope = RT_SCOPE_UNIVERSE;
443 if (cmd == SIOCDELRT)
446 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
449 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
450 cfg->fc_scope = RT_SCOPE_LINK;
452 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
456 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
460 if (rt->rt_flags & RTF_MTU)
461 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
463 if (rt->rt_flags & RTF_WINDOW)
464 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
466 if (rt->rt_flags & RTF_IRTT)
467 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
470 cfg->fc_mx_len = len;
477 * Handle IP routing ioctl calls.
478 * These are used to manipulate the routing tables
480 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
482 struct fib_config cfg;
487 case SIOCADDRT: /* Add a route */
488 case SIOCDELRT: /* Delete a route */
489 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
492 if (copy_from_user(&rt, arg, sizeof(rt)))
496 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
498 struct fib_table *tb;
500 if (cmd == SIOCDELRT) {
501 tb = fib_get_table(net, cfg.fc_table);
503 err = fib_table_delete(tb, &cfg);
507 tb = fib_new_table(net, cfg.fc_table);
509 err = fib_table_insert(tb, &cfg);
514 /* allocated by rtentry_to_fib_config() */
523 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
524 [RTA_DST] = { .type = NLA_U32 },
525 [RTA_SRC] = { .type = NLA_U32 },
526 [RTA_IIF] = { .type = NLA_U32 },
527 [RTA_OIF] = { .type = NLA_U32 },
528 [RTA_GATEWAY] = { .type = NLA_U32 },
529 [RTA_PRIORITY] = { .type = NLA_U32 },
530 [RTA_PREFSRC] = { .type = NLA_U32 },
531 [RTA_METRICS] = { .type = NLA_NESTED },
532 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
533 [RTA_FLOW] = { .type = NLA_U32 },
534 [RTA_UID] = { .type = NLA_U32 },
537 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
538 struct nlmsghdr *nlh, struct fib_config *cfg)
544 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
548 memset(cfg, 0, sizeof(*cfg));
550 rtm = nlmsg_data(nlh);
551 cfg->fc_dst_len = rtm->rtm_dst_len;
552 cfg->fc_tos = rtm->rtm_tos;
553 cfg->fc_table = rtm->rtm_table;
554 cfg->fc_protocol = rtm->rtm_protocol;
555 cfg->fc_scope = rtm->rtm_scope;
556 cfg->fc_type = rtm->rtm_type;
557 cfg->fc_flags = rtm->rtm_flags;
558 cfg->fc_nlflags = nlh->nlmsg_flags;
560 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
561 cfg->fc_nlinfo.nlh = nlh;
562 cfg->fc_nlinfo.nl_net = net;
564 if (cfg->fc_type > RTN_MAX) {
569 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
570 switch (nla_type(attr)) {
572 cfg->fc_dst = nla_get_be32(attr);
575 cfg->fc_oif = nla_get_u32(attr);
578 cfg->fc_gw = nla_get_be32(attr);
581 cfg->fc_priority = nla_get_u32(attr);
584 cfg->fc_prefsrc = nla_get_be32(attr);
587 cfg->fc_mx = nla_data(attr);
588 cfg->fc_mx_len = nla_len(attr);
591 cfg->fc_mp = nla_data(attr);
592 cfg->fc_mp_len = nla_len(attr);
595 cfg->fc_flow = nla_get_u32(attr);
598 cfg->fc_table = nla_get_u32(attr);
608 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
610 struct net *net = sock_net(skb->sk);
611 struct fib_config cfg;
612 struct fib_table *tb;
615 err = rtm_to_fib_config(net, skb, nlh, &cfg);
619 tb = fib_get_table(net, cfg.fc_table);
625 err = fib_table_delete(tb, &cfg);
630 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
632 struct net *net = sock_net(skb->sk);
633 struct fib_config cfg;
634 struct fib_table *tb;
637 err = rtm_to_fib_config(net, skb, nlh, &cfg);
641 tb = fib_new_table(net, cfg.fc_table);
647 err = fib_table_insert(tb, &cfg);
652 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
654 struct net *net = sock_net(skb->sk);
656 unsigned int e = 0, s_e;
657 struct fib_table *tb;
658 struct hlist_head *head;
661 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
662 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
663 return ip_rt_dump(skb, cb);
668 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
670 head = &net->ipv4.fib_table_hash[h];
671 hlist_for_each_entry(tb, head, tb_hlist) {
675 memset(&cb->args[2], 0, sizeof(cb->args) -
676 2 * sizeof(cb->args[0]));
677 if (fib_table_dump(tb, skb, cb) < 0)
691 /* Prepare and feed intra-kernel routing request.
692 * Really, it should be netlink message, but :-( netlink
693 * can be not configured, so that we feed it directly
694 * to fib engine. It is legal, because all events occur
695 * only when netlink is already locked.
697 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
699 struct net *net = dev_net(ifa->ifa_dev->dev);
700 struct fib_table *tb;
701 struct fib_config cfg = {
702 .fc_protocol = RTPROT_KERNEL,
705 .fc_dst_len = dst_len,
706 .fc_prefsrc = ifa->ifa_local,
707 .fc_oif = ifa->ifa_dev->dev->ifindex,
708 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
714 if (type == RTN_UNICAST)
715 tb = fib_new_table(net, RT_TABLE_MAIN);
717 tb = fib_new_table(net, RT_TABLE_LOCAL);
722 cfg.fc_table = tb->tb_id;
724 if (type != RTN_LOCAL)
725 cfg.fc_scope = RT_SCOPE_LINK;
727 cfg.fc_scope = RT_SCOPE_HOST;
729 if (cmd == RTM_NEWROUTE)
730 fib_table_insert(tb, &cfg);
732 fib_table_delete(tb, &cfg);
735 void fib_add_ifaddr(struct in_ifaddr *ifa)
737 struct in_device *in_dev = ifa->ifa_dev;
738 struct net_device *dev = in_dev->dev;
739 struct in_ifaddr *prim = ifa;
740 __be32 mask = ifa->ifa_mask;
741 __be32 addr = ifa->ifa_local;
742 __be32 prefix = ifa->ifa_address & mask;
744 if (ifa->ifa_flags & IFA_F_SECONDARY) {
745 prim = inet_ifa_byprefix(in_dev, prefix, mask);
747 pr_warn("%s: bug: prim == NULL\n", __func__);
752 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
754 if (!(dev->flags & IFF_UP))
757 /* Add broadcast address, if it is explicitly assigned. */
758 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
759 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
761 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
762 (prefix != addr || ifa->ifa_prefixlen < 32)) {
763 fib_magic(RTM_NEWROUTE,
764 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
765 prefix, ifa->ifa_prefixlen, prim);
767 /* Add network specific broadcasts, when it takes a sense */
768 if (ifa->ifa_prefixlen < 31) {
769 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
770 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
776 /* Delete primary or secondary address.
777 * Optionally, on secondary address promotion consider the addresses
778 * from subnet iprim as deleted, even if they are in device list.
779 * In this case the secondary ifa can be in device list.
781 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
783 struct in_device *in_dev = ifa->ifa_dev;
784 struct net_device *dev = in_dev->dev;
785 struct in_ifaddr *ifa1;
786 struct in_ifaddr *prim = ifa, *prim1 = NULL;
787 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
788 __be32 any = ifa->ifa_address & ifa->ifa_mask;
794 int subnet = 0; /* Primary network */
795 int gone = 1; /* Address is missing */
796 int same_prefsrc = 0; /* Another primary with same IP */
798 if (ifa->ifa_flags & IFA_F_SECONDARY) {
799 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
801 pr_warn("%s: bug: prim == NULL\n", __func__);
804 if (iprim && iprim != prim) {
805 pr_warn("%s: bug: iprim != prim\n", __func__);
808 } else if (!ipv4_is_zeronet(any) &&
809 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
810 fib_magic(RTM_DELROUTE,
811 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
812 any, ifa->ifa_prefixlen, prim);
816 /* Deletion is more complicated than add.
817 * We should take care of not to delete too much :-)
819 * Scan address list to be sure that addresses are really gone.
822 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
824 /* promotion, keep the IP */
828 /* Ignore IFAs from our subnet */
829 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
830 inet_ifa_match(ifa1->ifa_address, iprim))
833 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
834 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
835 /* Another address from our subnet? */
836 if (ifa1->ifa_mask == prim->ifa_mask &&
837 inet_ifa_match(ifa1->ifa_address, prim))
840 /* We reached the secondaries, so
841 * same_prefsrc should be determined.
845 /* Search new prim1 if ifa1 is not
846 * using the current prim1
849 ifa1->ifa_mask != prim1->ifa_mask ||
850 !inet_ifa_match(ifa1->ifa_address, prim1))
851 prim1 = inet_ifa_byprefix(in_dev,
856 if (prim1->ifa_local != prim->ifa_local)
860 if (prim->ifa_local != ifa1->ifa_local)
866 if (ifa->ifa_local == ifa1->ifa_local)
868 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
870 if (brd == ifa1->ifa_broadcast)
872 if (any == ifa1->ifa_broadcast)
874 /* primary has network specific broadcasts */
875 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
876 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
877 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
879 if (!ipv4_is_zeronet(any1)) {
880 if (ifa->ifa_broadcast == brd1 ||
881 ifa->ifa_broadcast == any1)
883 if (brd == brd1 || brd == any1)
885 if (any == brd1 || any == any1)
892 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
893 if (subnet && ifa->ifa_prefixlen < 31) {
895 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
897 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
899 if (!(ok & LOCAL_OK)) {
900 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
902 /* Check, that this local address finally disappeared. */
904 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
905 /* And the last, but not the least thing.
906 * We must flush stray FIB entries.
908 * First of all, we scan fib_info list searching
909 * for stray nexthop entries, then ignite fib_flush.
911 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
912 fib_flush(dev_net(dev));
921 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
924 struct fib_result res;
925 struct flowi4 fl4 = {
926 .flowi4_mark = frn->fl_mark,
927 .daddr = frn->fl_addr,
928 .flowi4_tos = frn->fl_tos,
929 .flowi4_scope = frn->fl_scope,
936 frn->tb_id = tb->tb_id;
938 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
941 frn->prefixlen = res.prefixlen;
942 frn->nh_sel = res.nh_sel;
943 frn->type = res.type;
944 frn->scope = res.scope;
951 static void nl_fib_input(struct sk_buff *skb)
954 struct fib_result_nl *frn;
955 struct nlmsghdr *nlh;
956 struct fib_table *tb;
959 net = sock_net(skb->sk);
960 nlh = nlmsg_hdr(skb);
961 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
962 nlmsg_len(nlh) < sizeof(*frn))
965 skb = skb_clone(skb, GFP_KERNEL);
968 nlh = nlmsg_hdr(skb);
970 frn = (struct fib_result_nl *) nlmsg_data(nlh);
971 tb = fib_get_table(net, frn->tb_id_in);
973 nl_fib_lookup(frn, tb);
975 portid = NETLINK_CB(skb).portid; /* netlink portid */
976 NETLINK_CB(skb).portid = 0; /* from kernel */
977 NETLINK_CB(skb).dst_group = 0; /* unicast */
978 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
981 static int __net_init nl_fib_lookup_init(struct net *net)
984 struct netlink_kernel_cfg cfg = {
985 .input = nl_fib_input,
988 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
990 return -EAFNOSUPPORT;
991 net->ipv4.fibnl = sk;
995 static void nl_fib_lookup_exit(struct net *net)
997 netlink_kernel_release(net->ipv4.fibnl);
998 net->ipv4.fibnl = NULL;
1001 static void fib_disable_ip(struct net_device *dev, int force)
1003 if (fib_sync_down_dev(dev, force))
1004 fib_flush(dev_net(dev));
1005 rt_cache_flush(dev_net(dev));
1009 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1011 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1012 struct net_device *dev = ifa->ifa_dev->dev;
1013 struct net *net = dev_net(dev);
1017 fib_add_ifaddr(ifa);
1018 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1021 atomic_inc(&net->ipv4.dev_addr_genid);
1022 rt_cache_flush(dev_net(dev));
1025 fib_del_ifaddr(ifa, NULL);
1026 atomic_inc(&net->ipv4.dev_addr_genid);
1027 if (ifa->ifa_dev->ifa_list == NULL) {
1028 /* Last address was deleted from this interface.
1031 fib_disable_ip(dev, 1);
1033 rt_cache_flush(dev_net(dev));
1040 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1042 struct net_device *dev = ptr;
1043 struct in_device *in_dev;
1044 struct net *net = dev_net(dev);
1046 if (event == NETDEV_UNREGISTER) {
1047 fib_disable_ip(dev, 2);
1052 in_dev = __in_dev_get_rtnl(dev);
1059 fib_add_ifaddr(ifa);
1060 } endfor_ifa(in_dev);
1061 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1064 atomic_inc(&net->ipv4.dev_addr_genid);
1065 rt_cache_flush(net);
1068 fib_disable_ip(dev, 0);
1070 case NETDEV_CHANGEMTU:
1072 rt_cache_flush(net);
1078 static struct notifier_block fib_inetaddr_notifier = {
1079 .notifier_call = fib_inetaddr_event,
1082 static struct notifier_block fib_netdev_notifier = {
1083 .notifier_call = fib_netdev_event,
1086 static int __net_init ip_fib_net_init(struct net *net)
1089 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1091 /* Avoid false sharing : Use at least a full cache line */
1092 size = max_t(size_t, size, L1_CACHE_BYTES);
1094 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1095 if (net->ipv4.fib_table_hash == NULL)
1098 err = fib4_rules_init(net);
1104 kfree(net->ipv4.fib_table_hash);
1108 static void ip_fib_net_exit(struct net *net)
1112 #ifdef CONFIG_IP_MULTIPLE_TABLES
1113 fib4_rules_exit(net);
1117 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1118 struct fib_table *tb;
1119 struct hlist_head *head;
1120 struct hlist_node *tmp;
1122 head = &net->ipv4.fib_table_hash[i];
1123 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1124 hlist_del(&tb->tb_hlist);
1125 fib_table_flush(tb);
1130 kfree(net->ipv4.fib_table_hash);
1133 static int __net_init fib_net_init(struct net *net)
1137 #ifdef CONFIG_IP_ROUTE_CLASSID
1138 net->ipv4.fib_num_tclassid_users = 0;
1140 error = ip_fib_net_init(net);
1143 error = nl_fib_lookup_init(net);
1146 error = fib_proc_init(net);
1153 nl_fib_lookup_exit(net);
1155 ip_fib_net_exit(net);
1159 static void __net_exit fib_net_exit(struct net *net)
1162 nl_fib_lookup_exit(net);
1163 ip_fib_net_exit(net);
1166 static struct pernet_operations fib_net_ops = {
1167 .init = fib_net_init,
1168 .exit = fib_net_exit,
1171 void __init ip_fib_init(void)
1173 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1174 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1175 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1177 register_pernet_subsys(&fib_net_ops);
1178 register_netdevice_notifier(&fib_netdev_notifier);
1179 register_inetaddr_notifier(&fib_inetaddr_notifier);