1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * IPv4 Forwarding Information Base: FIB frontend.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 #include <linux/module.h>
13 #include <linux/uaccess.h>
14 #include <linux/bitops.h>
15 #include <linux/capability.h>
16 #include <linux/types.h>
17 #include <linux/kernel.h>
19 #include <linux/string.h>
20 #include <linux/socket.h>
21 #include <linux/sockios.h>
22 #include <linux/errno.h>
24 #include <linux/inet.h>
25 #include <linux/inetdevice.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_addr.h>
28 #include <linux/if_arp.h>
29 #include <linux/skbuff.h>
30 #include <linux/cache.h>
31 #include <linux/init.h>
32 #include <linux/list.h>
33 #include <linux/slab.h>
36 #include <net/protocol.h>
37 #include <net/route.h>
41 #include <net/ip_fib.h>
42 #include <net/nexthop.h>
43 #include <net/rtnetlink.h>
45 #include <net/l3mdev.h>
46 #include <net/lwtunnel.h>
47 #include <trace/events/fib.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 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
59 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
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]);
70 fib_free_table(main_table);
74 static bool fib4_has_custom_rules(struct net *net)
80 struct fib_table *fib_new_table(struct net *net, u32 id)
82 struct fib_table *tb, *alias = NULL;
87 tb = fib_get_table(net, id);
91 if (id == RT_TABLE_LOCAL && !net->ipv4.fib_has_custom_rules)
92 alias = fib_new_table(net, RT_TABLE_MAIN);
94 tb = fib_trie_table(id, alias);
100 rcu_assign_pointer(net->ipv4.fib_main, tb);
102 case RT_TABLE_DEFAULT:
103 rcu_assign_pointer(net->ipv4.fib_default, tb);
109 h = id & (FIB_TABLE_HASHSZ - 1);
110 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
113 EXPORT_SYMBOL_GPL(fib_new_table);
115 /* caller must hold either rtnl or rcu read lock */
116 struct fib_table *fib_get_table(struct net *net, u32 id)
118 struct fib_table *tb;
119 struct hlist_head *head;
124 h = id & (FIB_TABLE_HASHSZ - 1);
126 head = &net->ipv4.fib_table_hash[h];
127 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
134 static bool fib4_has_custom_rules(struct net *net)
136 return net->ipv4.fib_has_custom_rules;
138 #endif /* CONFIG_IP_MULTIPLE_TABLES */
140 static void fib_replace_table(struct net *net, struct fib_table *old,
141 struct fib_table *new)
143 #ifdef CONFIG_IP_MULTIPLE_TABLES
144 switch (new->tb_id) {
146 rcu_assign_pointer(net->ipv4.fib_main, new);
148 case RT_TABLE_DEFAULT:
149 rcu_assign_pointer(net->ipv4.fib_default, new);
156 /* replace the old table in the hlist */
157 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
160 int fib_unmerge(struct net *net)
162 struct fib_table *old, *new, *main_table;
164 /* attempt to fetch local table if it has been allocated */
165 old = fib_get_table(net, RT_TABLE_LOCAL);
169 new = fib_trie_unmerge(old);
173 /* table is already unmerged */
177 /* replace merged table with clean table */
178 fib_replace_table(net, old, new);
181 /* attempt to fetch main table if it has been allocated */
182 main_table = fib_get_table(net, RT_TABLE_MAIN);
186 /* flush local entries from main table */
187 fib_table_flush_external(main_table);
192 void fib_flush(struct net *net)
197 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
198 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
199 struct hlist_node *tmp;
200 struct fib_table *tb;
202 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
203 flushed += fib_table_flush(net, tb, false);
211 * Find address type as if only "dev" was present in the system. If
212 * on_dev is NULL then all interfaces are taken into consideration.
214 static inline unsigned int __inet_dev_addr_type(struct net *net,
215 const struct net_device *dev,
216 __be32 addr, u32 tb_id)
218 struct flowi4 fl4 = { .daddr = addr };
219 struct fib_result res;
220 unsigned int ret = RTN_BROADCAST;
221 struct fib_table *table;
223 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
224 return RTN_BROADCAST;
225 if (ipv4_is_multicast(addr))
226 return RTN_MULTICAST;
230 table = fib_get_table(net, tb_id);
233 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
234 struct fib_nh_common *nhc = fib_info_nhc(res.fi, 0);
236 if (!dev || dev == nhc->nhc_dev)
245 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id)
247 return __inet_dev_addr_type(net, NULL, addr, tb_id);
249 EXPORT_SYMBOL(inet_addr_type_table);
251 unsigned int inet_addr_type(struct net *net, __be32 addr)
253 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
255 EXPORT_SYMBOL(inet_addr_type);
257 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
260 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
262 return __inet_dev_addr_type(net, dev, addr, rt_table);
264 EXPORT_SYMBOL(inet_dev_addr_type);
266 /* inet_addr_type with dev == NULL but using the table from a dev
267 * if one is associated
269 unsigned int inet_addr_type_dev_table(struct net *net,
270 const struct net_device *dev,
273 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
275 return __inet_dev_addr_type(net, NULL, addr, rt_table);
277 EXPORT_SYMBOL(inet_addr_type_dev_table);
279 __be32 fib_compute_spec_dst(struct sk_buff *skb)
281 struct net_device *dev = skb->dev;
282 struct in_device *in_dev;
283 struct fib_result res;
288 rt = skb_rtable(skb);
289 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
291 return ip_hdr(skb)->daddr;
293 in_dev = __in_dev_get_rcu(dev);
297 scope = RT_SCOPE_UNIVERSE;
298 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
299 bool vmark = in_dev && IN_DEV_SRC_VMARK(in_dev);
300 struct flowi4 fl4 = {
301 .flowi4_iif = LOOPBACK_IFINDEX,
302 .flowi4_oif = l3mdev_master_ifindex_rcu(dev),
303 .daddr = ip_hdr(skb)->saddr,
304 .flowi4_tos = RT_TOS(ip_hdr(skb)->tos),
305 .flowi4_scope = scope,
306 .flowi4_mark = vmark ? skb->mark : 0,
308 if (!fib_lookup(net, &fl4, &res, 0))
309 return fib_result_prefsrc(net, &res);
311 scope = RT_SCOPE_LINK;
314 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
317 bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev)
319 bool dev_match = false;
320 #ifdef CONFIG_IP_ROUTE_MULTIPATH
323 for (ret = 0; ret < fib_info_num_path(fi); ret++) {
324 const struct fib_nh_common *nhc = fib_info_nhc(fi, ret);
326 if (nhc->nhc_dev == dev) {
329 } else if (l3mdev_master_ifindex_rcu(nhc->nhc_dev) == dev->ifindex) {
335 if (fib_info_nhc(fi, 0)->nhc_dev == dev)
341 EXPORT_SYMBOL_GPL(fib_info_nh_uses_dev);
343 /* Given (packet source, input interface) and optional (dst, oif, tos):
344 * - (main) check, that source is valid i.e. not broadcast or our local
346 * - figure out what "logical" interface this packet arrived
347 * and calculate "specific destination" address.
348 * - check, that packet arrived from expected physical interface.
349 * called with rcu_read_lock()
351 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
352 u8 tos, int oif, struct net_device *dev,
353 int rpf, struct in_device *idev, u32 *itag)
355 struct net *net = dev_net(dev);
356 struct flow_keys flkeys;
358 struct fib_result res;
363 fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev);
365 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
368 fl4.flowi4_tos = tos;
369 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
370 fl4.flowi4_tun_key.tun_id = 0;
371 fl4.flowi4_flags = 0;
372 fl4.flowi4_uid = sock_net_uid(net, NULL);
374 no_addr = idev->ifa_list == NULL;
376 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
377 if (!fib4_rules_early_flow_dissect(net, skb, &fl4, &flkeys)) {
378 fl4.flowi4_proto = 0;
383 if (fib_lookup(net, &fl4, &res, 0))
385 if (res.type != RTN_UNICAST &&
386 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
388 fib_combine_itag(itag, &res);
390 dev_match = fib_info_nh_uses_dev(res.fi, dev);
392 ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_HOST;
399 fl4.flowi4_oif = dev->ifindex;
402 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
403 if (res.type == RTN_UNICAST)
404 ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_HOST;
420 /* Ignore rp_filter for packets protected by IPsec. */
421 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
422 u8 tos, int oif, struct net_device *dev,
423 struct in_device *idev, u32 *itag)
425 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
426 struct net *net = dev_net(dev);
428 if (!r && !fib_num_tclassid_users(net) &&
429 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
430 if (IN_DEV_ACCEPT_LOCAL(idev))
432 /* with custom local routes in place, checking local addresses
433 * only will be too optimistic, with custom rules, checking
434 * local addresses only can be too strict, e.g. due to vrf
436 if (net->ipv4.fib_has_custom_local_routes ||
437 fib4_has_custom_rules(net))
439 if (inet_lookup_ifaddr_rcu(net, src))
448 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
451 static inline __be32 sk_extract_addr(struct sockaddr *addr)
453 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
456 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
460 nla = (struct nlattr *) ((char *) mx + len);
461 nla->nla_type = type;
462 nla->nla_len = nla_attr_size(4);
463 *(u32 *) nla_data(nla) = value;
465 return len + nla_total_size(4);
468 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
469 struct fib_config *cfg)
474 memset(cfg, 0, sizeof(*cfg));
475 cfg->fc_nlinfo.nl_net = net;
477 if (rt->rt_dst.sa_family != AF_INET)
478 return -EAFNOSUPPORT;
481 * Check mask for validity:
482 * a) it must be contiguous.
483 * b) destination must have all host bits clear.
484 * c) if application forgot to set correct family (AF_INET),
485 * reject request unless it is absolutely clear i.e.
486 * both family and mask are zero.
489 addr = sk_extract_addr(&rt->rt_dst);
490 if (!(rt->rt_flags & RTF_HOST)) {
491 __be32 mask = sk_extract_addr(&rt->rt_genmask);
493 if (rt->rt_genmask.sa_family != AF_INET) {
494 if (mask || rt->rt_genmask.sa_family)
495 return -EAFNOSUPPORT;
498 if (bad_mask(mask, addr))
501 plen = inet_mask_len(mask);
504 cfg->fc_dst_len = plen;
507 if (cmd != SIOCDELRT) {
508 cfg->fc_nlflags = NLM_F_CREATE;
509 cfg->fc_protocol = RTPROT_BOOT;
513 cfg->fc_priority = rt->rt_metric - 1;
515 if (rt->rt_flags & RTF_REJECT) {
516 cfg->fc_scope = RT_SCOPE_HOST;
517 cfg->fc_type = RTN_UNREACHABLE;
521 cfg->fc_scope = RT_SCOPE_NOWHERE;
522 cfg->fc_type = RTN_UNICAST;
526 struct net_device *dev;
527 char devname[IFNAMSIZ];
529 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
532 devname[IFNAMSIZ-1] = 0;
533 colon = strchr(devname, ':');
536 dev = __dev_get_by_name(net, devname);
539 cfg->fc_oif = dev->ifindex;
540 cfg->fc_table = l3mdev_fib_table(dev);
542 const struct in_ifaddr *ifa;
543 struct in_device *in_dev;
545 in_dev = __in_dev_get_rtnl(dev);
552 in_dev_for_each_ifa_rcu(ifa, in_dev) {
553 if (strcmp(ifa->ifa_label, devname) == 0)
560 cfg->fc_prefsrc = ifa->ifa_local;
564 addr = sk_extract_addr(&rt->rt_gateway);
565 if (rt->rt_gateway.sa_family == AF_INET && addr) {
566 unsigned int addr_type;
569 cfg->fc_gw_family = AF_INET;
570 addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
571 if (rt->rt_flags & RTF_GATEWAY &&
572 addr_type == RTN_UNICAST)
573 cfg->fc_scope = RT_SCOPE_UNIVERSE;
576 if (cmd == SIOCDELRT)
579 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw_family)
582 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
583 cfg->fc_scope = RT_SCOPE_LINK;
585 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
589 mx = kcalloc(3, nla_total_size(4), GFP_KERNEL);
593 if (rt->rt_flags & RTF_MTU)
594 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
596 if (rt->rt_flags & RTF_WINDOW)
597 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
599 if (rt->rt_flags & RTF_IRTT)
600 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
603 cfg->fc_mx_len = len;
610 * Handle IP routing ioctl calls.
611 * These are used to manipulate the routing tables
613 int ip_rt_ioctl(struct net *net, unsigned int cmd, struct rtentry *rt)
615 struct fib_config cfg;
619 case SIOCADDRT: /* Add a route */
620 case SIOCDELRT: /* Delete a route */
621 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
625 err = rtentry_to_fib_config(net, cmd, rt, &cfg);
627 struct fib_table *tb;
629 if (cmd == SIOCDELRT) {
630 tb = fib_get_table(net, cfg.fc_table);
632 err = fib_table_delete(net, tb, &cfg,
637 tb = fib_new_table(net, cfg.fc_table);
639 err = fib_table_insert(net, tb,
645 /* allocated by rtentry_to_fib_config() */
654 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
655 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
656 [RTA_DST] = { .type = NLA_U32 },
657 [RTA_SRC] = { .type = NLA_U32 },
658 [RTA_IIF] = { .type = NLA_U32 },
659 [RTA_OIF] = { .type = NLA_U32 },
660 [RTA_GATEWAY] = { .type = NLA_U32 },
661 [RTA_PRIORITY] = { .type = NLA_U32 },
662 [RTA_PREFSRC] = { .type = NLA_U32 },
663 [RTA_METRICS] = { .type = NLA_NESTED },
664 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
665 [RTA_FLOW] = { .type = NLA_U32 },
666 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
667 [RTA_ENCAP] = { .type = NLA_NESTED },
668 [RTA_UID] = { .type = NLA_U32 },
669 [RTA_MARK] = { .type = NLA_U32 },
670 [RTA_TABLE] = { .type = NLA_U32 },
671 [RTA_IP_PROTO] = { .type = NLA_U8 },
672 [RTA_SPORT] = { .type = NLA_U16 },
673 [RTA_DPORT] = { .type = NLA_U16 },
674 [RTA_NH_ID] = { .type = NLA_U32 },
677 int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla,
678 struct netlink_ext_ack *extack)
683 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
684 NL_SET_ERR_MSG(extack, "Invalid attribute length for RTA_VIA");
689 alen = nla_len(nla) - offsetof(struct rtvia, rtvia_addr);
691 switch (via->rtvia_family) {
693 if (alen != sizeof(__be32)) {
694 NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_VIA");
697 cfg->fc_gw_family = AF_INET;
698 cfg->fc_gw4 = *((__be32 *)via->rtvia_addr);
702 if (alen != sizeof(struct in6_addr)) {
703 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_VIA");
706 cfg->fc_gw_family = AF_INET6;
707 cfg->fc_gw6 = *((struct in6_addr *)via->rtvia_addr);
709 NL_SET_ERR_MSG(extack, "IPv6 support not enabled in kernel");
714 NL_SET_ERR_MSG(extack, "Unsupported address family in RTA_VIA");
721 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
722 struct nlmsghdr *nlh, struct fib_config *cfg,
723 struct netlink_ext_ack *extack)
725 bool has_gw = false, has_via = false;
730 err = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
731 rtm_ipv4_policy, extack);
735 memset(cfg, 0, sizeof(*cfg));
737 rtm = nlmsg_data(nlh);
738 cfg->fc_dst_len = rtm->rtm_dst_len;
739 cfg->fc_tos = rtm->rtm_tos;
740 cfg->fc_table = rtm->rtm_table;
741 cfg->fc_protocol = rtm->rtm_protocol;
742 cfg->fc_scope = rtm->rtm_scope;
743 cfg->fc_type = rtm->rtm_type;
744 cfg->fc_flags = rtm->rtm_flags;
745 cfg->fc_nlflags = nlh->nlmsg_flags;
747 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
748 cfg->fc_nlinfo.nlh = nlh;
749 cfg->fc_nlinfo.nl_net = net;
751 if (cfg->fc_type > RTN_MAX) {
752 NL_SET_ERR_MSG(extack, "Invalid route type");
757 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
758 switch (nla_type(attr)) {
760 cfg->fc_dst = nla_get_be32(attr);
763 cfg->fc_oif = nla_get_u32(attr);
767 cfg->fc_gw4 = nla_get_be32(attr);
769 cfg->fc_gw_family = AF_INET;
773 err = fib_gw_from_via(cfg, attr, extack);
778 cfg->fc_priority = nla_get_u32(attr);
781 cfg->fc_prefsrc = nla_get_be32(attr);
784 cfg->fc_mx = nla_data(attr);
785 cfg->fc_mx_len = nla_len(attr);
788 err = lwtunnel_valid_encap_type_attr(nla_data(attr),
793 cfg->fc_mp = nla_data(attr);
794 cfg->fc_mp_len = nla_len(attr);
797 cfg->fc_flow = nla_get_u32(attr);
800 cfg->fc_table = nla_get_u32(attr);
803 cfg->fc_encap = attr;
806 cfg->fc_encap_type = nla_get_u16(attr);
807 err = lwtunnel_valid_encap_type(cfg->fc_encap_type,
813 cfg->fc_nh_id = nla_get_u32(attr);
819 if (cfg->fc_oif || cfg->fc_gw_family ||
820 cfg->fc_encap || cfg->fc_mp) {
821 NL_SET_ERR_MSG(extack,
822 "Nexthop specification and nexthop id are mutually exclusive");
827 if (has_gw && has_via) {
828 NL_SET_ERR_MSG(extack,
829 "Nexthop configuration can not contain both GATEWAY and VIA");
838 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
839 struct netlink_ext_ack *extack)
841 struct net *net = sock_net(skb->sk);
842 struct fib_config cfg;
843 struct fib_table *tb;
846 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
850 if (cfg.fc_nh_id && !nexthop_find_by_id(net, cfg.fc_nh_id)) {
851 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
856 tb = fib_get_table(net, cfg.fc_table);
858 NL_SET_ERR_MSG(extack, "FIB table does not exist");
863 err = fib_table_delete(net, tb, &cfg, extack);
868 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
869 struct netlink_ext_ack *extack)
871 struct net *net = sock_net(skb->sk);
872 struct fib_config cfg;
873 struct fib_table *tb;
876 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
880 tb = fib_new_table(net, cfg.fc_table);
886 err = fib_table_insert(net, tb, &cfg, extack);
887 if (!err && cfg.fc_type == RTN_LOCAL)
888 net->ipv4.fib_has_custom_local_routes = true;
893 int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
894 struct fib_dump_filter *filter,
895 struct netlink_callback *cb)
897 struct netlink_ext_ack *extack = cb->extack;
898 struct nlattr *tb[RTA_MAX + 1];
904 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
905 NL_SET_ERR_MSG(extack, "Invalid header for FIB dump request");
909 rtm = nlmsg_data(nlh);
910 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos ||
912 NL_SET_ERR_MSG(extack, "Invalid values in header for FIB dump request");
916 if (rtm->rtm_flags & ~(RTM_F_CLONED | RTM_F_PREFIX)) {
917 NL_SET_ERR_MSG(extack, "Invalid flags for FIB dump request");
920 if (rtm->rtm_flags & RTM_F_CLONED)
921 filter->dump_routes = false;
923 filter->dump_exceptions = false;
925 filter->dump_all_families = (rtm->rtm_family == AF_UNSPEC);
926 filter->flags = rtm->rtm_flags;
927 filter->protocol = rtm->rtm_protocol;
928 filter->rt_type = rtm->rtm_type;
929 filter->table_id = rtm->rtm_table;
931 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
932 rtm_ipv4_policy, extack);
936 for (i = 0; i <= RTA_MAX; ++i) {
944 filter->table_id = nla_get_u32(tb[i]);
947 ifindex = nla_get_u32(tb[i]);
948 filter->dev = __dev_get_by_index(net, ifindex);
953 NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
958 if (filter->flags || filter->protocol || filter->rt_type ||
959 filter->table_id || filter->dev) {
960 filter->filter_set = 1;
961 cb->answer_flags = NLM_F_DUMP_FILTERED;
966 EXPORT_SYMBOL_GPL(ip_valid_fib_dump_req);
968 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
970 struct fib_dump_filter filter = { .dump_routes = true,
971 .dump_exceptions = true };
972 const struct nlmsghdr *nlh = cb->nlh;
973 struct net *net = sock_net(skb->sk);
975 unsigned int e = 0, s_e;
976 struct fib_table *tb;
977 struct hlist_head *head;
980 if (cb->strict_check) {
981 err = ip_valid_fib_dump_req(net, nlh, &filter, cb);
984 } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
985 struct rtmsg *rtm = nlmsg_data(nlh);
987 filter.flags = rtm->rtm_flags & (RTM_F_PREFIX | RTM_F_CLONED);
990 /* ipv4 does not use prefix flag */
991 if (filter.flags & RTM_F_PREFIX)
994 if (filter.table_id) {
995 tb = fib_get_table(net, filter.table_id);
997 if (filter.dump_all_families)
1000 NL_SET_ERR_MSG(cb->extack, "ipv4: FIB table does not exist");
1004 err = fib_table_dump(tb, skb, cb, &filter);
1005 return skb->len ? : err;
1013 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
1015 head = &net->ipv4.fib_table_hash[h];
1016 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
1020 memset(&cb->args[2], 0, sizeof(cb->args) -
1021 2 * sizeof(cb->args[0]));
1022 err = fib_table_dump(tb, skb, cb, &filter);
1024 if (likely(skb->len))
1045 /* Prepare and feed intra-kernel routing request.
1046 * Really, it should be netlink message, but :-( netlink
1047 * can be not configured, so that we feed it directly
1048 * to fib engine. It is legal, because all events occur
1049 * only when netlink is already locked.
1051 static void fib_magic(int cmd, int type, __be32 dst, int dst_len,
1052 struct in_ifaddr *ifa, u32 rt_priority)
1054 struct net *net = dev_net(ifa->ifa_dev->dev);
1055 u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev);
1056 struct fib_table *tb;
1057 struct fib_config cfg = {
1058 .fc_protocol = RTPROT_KERNEL,
1061 .fc_dst_len = dst_len,
1062 .fc_priority = rt_priority,
1063 .fc_prefsrc = ifa->ifa_local,
1064 .fc_oif = ifa->ifa_dev->dev->ifindex,
1065 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
1072 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
1074 tb = fib_new_table(net, tb_id);
1078 cfg.fc_table = tb->tb_id;
1080 if (type != RTN_LOCAL)
1081 cfg.fc_scope = RT_SCOPE_LINK;
1083 cfg.fc_scope = RT_SCOPE_HOST;
1085 if (cmd == RTM_NEWROUTE)
1086 fib_table_insert(net, tb, &cfg, NULL);
1088 fib_table_delete(net, tb, &cfg, NULL);
1091 void fib_add_ifaddr(struct in_ifaddr *ifa)
1093 struct in_device *in_dev = ifa->ifa_dev;
1094 struct net_device *dev = in_dev->dev;
1095 struct in_ifaddr *prim = ifa;
1096 __be32 mask = ifa->ifa_mask;
1097 __be32 addr = ifa->ifa_local;
1098 __be32 prefix = ifa->ifa_address & mask;
1100 if (ifa->ifa_flags & IFA_F_SECONDARY) {
1101 prim = inet_ifa_byprefix(in_dev, prefix, mask);
1103 pr_warn("%s: bug: prim == NULL\n", __func__);
1108 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim, 0);
1110 if (!(dev->flags & IFF_UP))
1113 /* Add broadcast address, if it is explicitly assigned. */
1114 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
1115 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
1118 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
1119 (prefix != addr || ifa->ifa_prefixlen < 32)) {
1120 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
1121 fib_magic(RTM_NEWROUTE,
1122 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1123 prefix, ifa->ifa_prefixlen, prim,
1124 ifa->ifa_rt_priority);
1126 /* Add network specific broadcasts, when it takes a sense */
1127 if (ifa->ifa_prefixlen < 31) {
1128 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32,
1130 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
1136 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric)
1138 __be32 prefix = ifa->ifa_address & ifa->ifa_mask;
1139 struct in_device *in_dev = ifa->ifa_dev;
1140 struct net_device *dev = in_dev->dev;
1142 if (!(dev->flags & IFF_UP) ||
1143 ifa->ifa_flags & (IFA_F_SECONDARY | IFA_F_NOPREFIXROUTE) ||
1144 ipv4_is_zeronet(prefix) ||
1145 prefix == ifa->ifa_local || ifa->ifa_prefixlen == 32)
1149 fib_magic(RTM_NEWROUTE,
1150 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1151 prefix, ifa->ifa_prefixlen, ifa, new_metric);
1153 /* delete the old */
1154 fib_magic(RTM_DELROUTE,
1155 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1156 prefix, ifa->ifa_prefixlen, ifa, ifa->ifa_rt_priority);
1159 /* Delete primary or secondary address.
1160 * Optionally, on secondary address promotion consider the addresses
1161 * from subnet iprim as deleted, even if they are in device list.
1162 * In this case the secondary ifa can be in device list.
1164 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
1166 struct in_device *in_dev = ifa->ifa_dev;
1167 struct net_device *dev = in_dev->dev;
1168 struct in_ifaddr *ifa1;
1169 struct in_ifaddr *prim = ifa, *prim1 = NULL;
1170 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
1171 __be32 any = ifa->ifa_address & ifa->ifa_mask;
1176 unsigned int ok = 0;
1177 int subnet = 0; /* Primary network */
1178 int gone = 1; /* Address is missing */
1179 int same_prefsrc = 0; /* Another primary with same IP */
1181 if (ifa->ifa_flags & IFA_F_SECONDARY) {
1182 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
1184 /* if the device has been deleted, we don't perform
1188 pr_warn("%s: bug: prim == NULL\n", __func__);
1191 if (iprim && iprim != prim) {
1192 pr_warn("%s: bug: iprim != prim\n", __func__);
1195 } else if (!ipv4_is_zeronet(any) &&
1196 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
1197 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
1198 fib_magic(RTM_DELROUTE,
1199 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1200 any, ifa->ifa_prefixlen, prim, 0);
1207 /* Deletion is more complicated than add.
1208 * We should take care of not to delete too much :-)
1210 * Scan address list to be sure that addresses are really gone.
1213 in_dev_for_each_ifa_rcu(ifa1, in_dev) {
1215 /* promotion, keep the IP */
1219 /* Ignore IFAs from our subnet */
1220 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
1221 inet_ifa_match(ifa1->ifa_address, iprim))
1224 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
1225 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
1226 /* Another address from our subnet? */
1227 if (ifa1->ifa_mask == prim->ifa_mask &&
1228 inet_ifa_match(ifa1->ifa_address, prim))
1231 /* We reached the secondaries, so
1232 * same_prefsrc should be determined.
1236 /* Search new prim1 if ifa1 is not
1237 * using the current prim1
1240 ifa1->ifa_mask != prim1->ifa_mask ||
1241 !inet_ifa_match(ifa1->ifa_address, prim1))
1242 prim1 = inet_ifa_byprefix(in_dev,
1247 if (prim1->ifa_local != prim->ifa_local)
1251 if (prim->ifa_local != ifa1->ifa_local)
1257 if (ifa->ifa_local == ifa1->ifa_local)
1259 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
1261 if (brd == ifa1->ifa_broadcast)
1263 if (any == ifa1->ifa_broadcast)
1265 /* primary has network specific broadcasts */
1266 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
1267 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
1268 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
1270 if (!ipv4_is_zeronet(any1)) {
1271 if (ifa->ifa_broadcast == brd1 ||
1272 ifa->ifa_broadcast == any1)
1274 if (brd == brd1 || brd == any1)
1276 if (any == brd1 || any == any1)
1285 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
1287 if (subnet && ifa->ifa_prefixlen < 31) {
1288 if (!(ok & BRD1_OK))
1289 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32,
1291 if (!(ok & BRD0_OK))
1292 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32,
1295 if (!(ok & LOCAL_OK)) {
1296 unsigned int addr_type;
1298 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim, 0);
1300 /* Check, that this local address finally disappeared. */
1301 addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1303 if (gone && addr_type != RTN_LOCAL) {
1304 /* And the last, but not the least thing.
1305 * We must flush stray FIB entries.
1307 * First of all, we scan fib_info list searching
1308 * for stray nexthop entries, then ignite fib_flush.
1310 if (fib_sync_down_addr(dev, ifa->ifa_local))
1311 fib_flush(dev_net(dev));
1320 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1323 struct fib_result res;
1324 struct flowi4 fl4 = {
1325 .flowi4_mark = frn->fl_mark,
1326 .daddr = frn->fl_addr,
1327 .flowi4_tos = frn->fl_tos,
1328 .flowi4_scope = frn->fl_scope,
1330 struct fib_table *tb;
1334 tb = fib_get_table(net, frn->tb_id_in);
1340 frn->tb_id = tb->tb_id;
1341 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1344 frn->prefixlen = res.prefixlen;
1345 frn->nh_sel = res.nh_sel;
1346 frn->type = res.type;
1347 frn->scope = res.scope;
1355 static void nl_fib_input(struct sk_buff *skb)
1358 struct fib_result_nl *frn;
1359 struct nlmsghdr *nlh;
1362 net = sock_net(skb->sk);
1363 nlh = nlmsg_hdr(skb);
1364 if (skb->len < nlmsg_total_size(sizeof(*frn)) ||
1365 skb->len < nlh->nlmsg_len ||
1366 nlmsg_len(nlh) < sizeof(*frn))
1369 skb = netlink_skb_clone(skb, GFP_KERNEL);
1372 nlh = nlmsg_hdr(skb);
1374 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1375 nl_fib_lookup(net, frn);
1377 portid = NETLINK_CB(skb).portid; /* netlink portid */
1378 NETLINK_CB(skb).portid = 0; /* from kernel */
1379 NETLINK_CB(skb).dst_group = 0; /* unicast */
1380 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1383 static int __net_init nl_fib_lookup_init(struct net *net)
1386 struct netlink_kernel_cfg cfg = {
1387 .input = nl_fib_input,
1390 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1392 return -EAFNOSUPPORT;
1393 net->ipv4.fibnl = sk;
1397 static void nl_fib_lookup_exit(struct net *net)
1399 netlink_kernel_release(net->ipv4.fibnl);
1400 net->ipv4.fibnl = NULL;
1403 static void fib_disable_ip(struct net_device *dev, unsigned long event,
1406 if (fib_sync_down_dev(dev, event, force))
1407 fib_flush(dev_net(dev));
1409 rt_cache_flush(dev_net(dev));
1413 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1415 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1416 struct net_device *dev = ifa->ifa_dev->dev;
1417 struct net *net = dev_net(dev);
1421 fib_add_ifaddr(ifa);
1422 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1423 fib_sync_up(dev, RTNH_F_DEAD);
1425 atomic_inc(&net->ipv4.dev_addr_genid);
1426 rt_cache_flush(dev_net(dev));
1429 fib_del_ifaddr(ifa, NULL);
1430 atomic_inc(&net->ipv4.dev_addr_genid);
1431 if (!ifa->ifa_dev->ifa_list) {
1432 /* Last address was deleted from this interface.
1435 fib_disable_ip(dev, event, true);
1437 rt_cache_flush(dev_net(dev));
1444 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1446 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1447 struct netdev_notifier_changeupper_info *upper_info = ptr;
1448 struct netdev_notifier_info_ext *info_ext = ptr;
1449 struct in_device *in_dev;
1450 struct net *net = dev_net(dev);
1451 struct in_ifaddr *ifa;
1454 if (event == NETDEV_UNREGISTER) {
1455 fib_disable_ip(dev, event, true);
1460 in_dev = __in_dev_get_rtnl(dev);
1466 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1467 fib_add_ifaddr(ifa);
1469 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1470 fib_sync_up(dev, RTNH_F_DEAD);
1472 atomic_inc(&net->ipv4.dev_addr_genid);
1473 rt_cache_flush(net);
1476 fib_disable_ip(dev, event, false);
1479 flags = dev_get_flags(dev);
1480 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1481 fib_sync_up(dev, RTNH_F_LINKDOWN);
1483 fib_sync_down_dev(dev, event, false);
1484 rt_cache_flush(net);
1486 case NETDEV_CHANGEMTU:
1487 fib_sync_mtu(dev, info_ext->ext.mtu);
1488 rt_cache_flush(net);
1490 case NETDEV_CHANGEUPPER:
1492 /* flush all routes if dev is linked to or unlinked from
1493 * an L3 master device (e.g., VRF)
1495 if (upper_info->upper_dev &&
1496 netif_is_l3_master(upper_info->upper_dev))
1497 fib_disable_ip(dev, NETDEV_DOWN, true);
1503 static struct notifier_block fib_inetaddr_notifier = {
1504 .notifier_call = fib_inetaddr_event,
1507 static struct notifier_block fib_netdev_notifier = {
1508 .notifier_call = fib_netdev_event,
1511 static int __net_init ip_fib_net_init(struct net *net)
1514 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1516 err = fib4_notifier_init(net);
1520 /* Avoid false sharing : Use at least a full cache line */
1521 size = max_t(size_t, size, L1_CACHE_BYTES);
1523 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1524 if (!net->ipv4.fib_table_hash) {
1526 goto err_table_hash_alloc;
1529 err = fib4_rules_init(net);
1531 goto err_rules_init;
1535 kfree(net->ipv4.fib_table_hash);
1536 err_table_hash_alloc:
1537 fib4_notifier_exit(net);
1541 static void ip_fib_net_exit(struct net *net)
1546 #ifdef CONFIG_IP_MULTIPLE_TABLES
1547 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1548 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1550 /* Destroy the tables in reverse order to guarantee that the
1551 * local table, ID 255, is destroyed before the main table, ID
1552 * 254. This is necessary as the local table may contain
1553 * references to data contained in the main table.
1555 for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) {
1556 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1557 struct hlist_node *tmp;
1558 struct fib_table *tb;
1560 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1561 hlist_del(&tb->tb_hlist);
1562 fib_table_flush(net, tb, true);
1567 #ifdef CONFIG_IP_MULTIPLE_TABLES
1568 fib4_rules_exit(net);
1571 kfree(net->ipv4.fib_table_hash);
1572 fib4_notifier_exit(net);
1575 static int __net_init fib_net_init(struct net *net)
1579 #ifdef CONFIG_IP_ROUTE_CLASSID
1580 net->ipv4.fib_num_tclassid_users = 0;
1582 error = ip_fib_net_init(net);
1585 error = nl_fib_lookup_init(net);
1588 error = fib_proc_init(net);
1595 nl_fib_lookup_exit(net);
1597 ip_fib_net_exit(net);
1601 static void __net_exit fib_net_exit(struct net *net)
1604 nl_fib_lookup_exit(net);
1605 ip_fib_net_exit(net);
1608 static struct pernet_operations fib_net_ops = {
1609 .init = fib_net_init,
1610 .exit = fib_net_exit,
1613 void __init ip_fib_init(void)
1617 register_pernet_subsys(&fib_net_ops);
1619 register_netdevice_notifier(&fib_netdev_notifier);
1620 register_inetaddr_notifier(&fib_inetaddr_notifier);
1622 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, 0);
1623 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, 0);
1624 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, 0);