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 * Version: $Id: fib_frontend.c,v 1.26 2001/10/31 21:55:54 davem Exp $
10 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
18 #include <linux/module.h>
19 #include <asm/uaccess.h>
20 #include <asm/system.h>
21 #include <linux/bitops.h>
22 #include <linux/capability.h>
23 #include <linux/types.h>
24 #include <linux/kernel.h>
26 #include <linux/string.h>
27 #include <linux/socket.h>
28 #include <linux/sockios.h>
29 #include <linux/errno.h>
31 #include <linux/inet.h>
32 #include <linux/inetdevice.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_arp.h>
36 #include <linux/skbuff.h>
37 #include <linux/netlink.h>
38 #include <linux/init.h>
39 #include <linux/list.h>
42 #include <net/protocol.h>
43 #include <net/route.h>
48 #include <net/ip_fib.h>
50 #define FFprint(a...) printk(KERN_DEBUG a)
52 #ifndef CONFIG_IP_MULTIPLE_TABLES
54 struct fib_table *ip_fib_local_table;
55 struct fib_table *ip_fib_main_table;
57 #define FIB_TABLE_HASHSZ 1
58 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
62 #define FIB_TABLE_HASHSZ 256
63 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
65 struct fib_table *fib_new_table(u32 id)
72 tb = fib_get_table(id);
75 tb = fib_hash_init(id);
78 h = id & (FIB_TABLE_HASHSZ - 1);
79 hlist_add_head_rcu(&tb->tb_hlist, &fib_table_hash[h]);
83 struct fib_table *fib_get_table(u32 id)
86 struct hlist_node *node;
91 h = id & (FIB_TABLE_HASHSZ - 1);
93 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb_hlist) {
94 if (tb->tb_id == id) {
102 #endif /* CONFIG_IP_MULTIPLE_TABLES */
104 static void fib_flush(void)
107 struct fib_table *tb;
108 struct hlist_node *node;
111 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
112 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist)
113 flushed += tb->tb_flush(tb);
121 * Find the first device with a given source address.
124 struct net_device * ip_dev_find(__be32 addr)
126 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
127 struct fib_result res;
128 struct net_device *dev = NULL;
130 #ifdef CONFIG_IP_MULTIPLE_TABLES
134 if (!ip_fib_local_table ||
135 ip_fib_local_table->tb_lookup(ip_fib_local_table, &fl, &res))
137 if (res.type != RTN_LOCAL)
139 dev = FIB_RES_DEV(res);
148 unsigned inet_addr_type(__be32 addr)
150 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
151 struct fib_result res;
152 unsigned ret = RTN_BROADCAST;
154 if (ZERONET(addr) || BADCLASS(addr))
155 return RTN_BROADCAST;
157 return RTN_MULTICAST;
159 #ifdef CONFIG_IP_MULTIPLE_TABLES
163 if (ip_fib_local_table) {
165 if (!ip_fib_local_table->tb_lookup(ip_fib_local_table,
174 /* Given (packet source, input interface) and optional (dst, oif, tos):
175 - (main) check, that source is valid i.e. not broadcast or our local
177 - figure out what "logical" interface this packet arrived
178 and calculate "specific destination" address.
179 - check, that packet arrived from expected physical interface.
182 int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
183 struct net_device *dev, __be32 *spec_dst, u32 *itag)
185 struct in_device *in_dev;
186 struct flowi fl = { .nl_u = { .ip4_u =
191 struct fib_result res;
197 in_dev = __in_dev_get_rcu(dev);
199 no_addr = in_dev->ifa_list == NULL;
200 rpf = IN_DEV_RPFILTER(in_dev);
207 if (fib_lookup(&fl, &res))
209 if (res.type != RTN_UNICAST)
211 *spec_dst = FIB_RES_PREFSRC(res);
212 fib_combine_itag(itag, &res);
213 #ifdef CONFIG_IP_ROUTE_MULTIPATH
214 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1)
216 if (FIB_RES_DEV(res) == dev)
219 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
228 fl.oif = dev->ifindex;
231 if (fib_lookup(&fl, &res) == 0) {
232 if (res.type == RTN_UNICAST) {
233 *spec_dst = FIB_RES_PREFSRC(res);
234 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
243 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
253 #ifndef CONFIG_IP_NOSIOCRT
255 static inline __be32 sk_extract_addr(struct sockaddr *addr)
257 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
260 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
264 nla = (struct nlattr *) ((char *) mx + len);
265 nla->nla_type = type;
266 nla->nla_len = nla_attr_size(4);
267 *(u32 *) nla_data(nla) = value;
269 return len + nla_total_size(4);
272 static int rtentry_to_fib_config(int cmd, struct rtentry *rt,
273 struct fib_config *cfg)
278 memset(cfg, 0, sizeof(*cfg));
280 if (rt->rt_dst.sa_family != AF_INET)
281 return -EAFNOSUPPORT;
284 * Check mask for validity:
285 * a) it must be contiguous.
286 * b) destination must have all host bits clear.
287 * c) if application forgot to set correct family (AF_INET),
288 * reject request unless it is absolutely clear i.e.
289 * both family and mask are zero.
292 addr = sk_extract_addr(&rt->rt_dst);
293 if (!(rt->rt_flags & RTF_HOST)) {
294 __be32 mask = sk_extract_addr(&rt->rt_genmask);
296 if (rt->rt_genmask.sa_family != AF_INET) {
297 if (mask || rt->rt_genmask.sa_family)
298 return -EAFNOSUPPORT;
301 if (bad_mask(mask, addr))
304 plen = inet_mask_len(mask);
307 cfg->fc_dst_len = plen;
310 if (cmd != SIOCDELRT) {
311 cfg->fc_nlflags = NLM_F_CREATE;
312 cfg->fc_protocol = RTPROT_BOOT;
316 cfg->fc_priority = rt->rt_metric - 1;
318 if (rt->rt_flags & RTF_REJECT) {
319 cfg->fc_scope = RT_SCOPE_HOST;
320 cfg->fc_type = RTN_UNREACHABLE;
324 cfg->fc_scope = RT_SCOPE_NOWHERE;
325 cfg->fc_type = RTN_UNICAST;
329 struct net_device *dev;
330 char devname[IFNAMSIZ];
332 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
335 devname[IFNAMSIZ-1] = 0;
336 colon = strchr(devname, ':');
339 dev = __dev_get_by_name(devname);
342 cfg->fc_oif = dev->ifindex;
344 struct in_ifaddr *ifa;
345 struct in_device *in_dev = __in_dev_get_rtnl(dev);
349 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
350 if (strcmp(ifa->ifa_label, devname) == 0)
354 cfg->fc_prefsrc = ifa->ifa_local;
358 addr = sk_extract_addr(&rt->rt_gateway);
359 if (rt->rt_gateway.sa_family == AF_INET && addr) {
361 if (rt->rt_flags & RTF_GATEWAY &&
362 inet_addr_type(addr) == RTN_UNICAST)
363 cfg->fc_scope = RT_SCOPE_UNIVERSE;
366 if (cmd == SIOCDELRT)
369 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
372 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
373 cfg->fc_scope = RT_SCOPE_LINK;
375 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
379 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
383 if (rt->rt_flags & RTF_MTU)
384 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
386 if (rt->rt_flags & RTF_WINDOW)
387 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
389 if (rt->rt_flags & RTF_IRTT)
390 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
393 cfg->fc_mx_len = len;
400 * Handle IP routing ioctl calls. These are used to manipulate the routing tables
403 int ip_rt_ioctl(unsigned int cmd, void __user *arg)
405 struct fib_config cfg;
410 case SIOCADDRT: /* Add a route */
411 case SIOCDELRT: /* Delete a route */
412 if (!capable(CAP_NET_ADMIN))
415 if (copy_from_user(&rt, arg, sizeof(rt)))
419 err = rtentry_to_fib_config(cmd, &rt, &cfg);
421 struct fib_table *tb;
423 if (cmd == SIOCDELRT) {
424 tb = fib_get_table(cfg.fc_table);
426 err = tb->tb_delete(tb, &cfg);
430 tb = fib_new_table(cfg.fc_table);
432 err = tb->tb_insert(tb, &cfg);
437 /* allocated by rtentry_to_fib_config() */
448 int ip_rt_ioctl(unsigned int cmd, void *arg)
455 struct nla_policy rtm_ipv4_policy[RTA_MAX+1] __read_mostly = {
456 [RTA_DST] = { .type = NLA_U32 },
457 [RTA_SRC] = { .type = NLA_U32 },
458 [RTA_IIF] = { .type = NLA_U32 },
459 [RTA_OIF] = { .type = NLA_U32 },
460 [RTA_GATEWAY] = { .type = NLA_U32 },
461 [RTA_PRIORITY] = { .type = NLA_U32 },
462 [RTA_PREFSRC] = { .type = NLA_U32 },
463 [RTA_METRICS] = { .type = NLA_NESTED },
464 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
465 [RTA_PROTOINFO] = { .type = NLA_U32 },
466 [RTA_FLOW] = { .type = NLA_U32 },
467 [RTA_MP_ALGO] = { .type = NLA_U32 },
470 static int rtm_to_fib_config(struct sk_buff *skb, struct nlmsghdr *nlh,
471 struct fib_config *cfg)
477 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
481 memset(cfg, 0, sizeof(*cfg));
483 rtm = nlmsg_data(nlh);
484 cfg->fc_dst_len = rtm->rtm_dst_len;
485 cfg->fc_tos = rtm->rtm_tos;
486 cfg->fc_table = rtm->rtm_table;
487 cfg->fc_protocol = rtm->rtm_protocol;
488 cfg->fc_scope = rtm->rtm_scope;
489 cfg->fc_type = rtm->rtm_type;
490 cfg->fc_flags = rtm->rtm_flags;
491 cfg->fc_nlflags = nlh->nlmsg_flags;
493 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
494 cfg->fc_nlinfo.nlh = nlh;
496 if (cfg->fc_type > RTN_MAX) {
501 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
502 switch (attr->nla_type) {
504 cfg->fc_dst = nla_get_be32(attr);
507 cfg->fc_oif = nla_get_u32(attr);
510 cfg->fc_gw = nla_get_be32(attr);
513 cfg->fc_priority = nla_get_u32(attr);
516 cfg->fc_prefsrc = nla_get_be32(attr);
519 cfg->fc_mx = nla_data(attr);
520 cfg->fc_mx_len = nla_len(attr);
523 cfg->fc_mp = nla_data(attr);
524 cfg->fc_mp_len = nla_len(attr);
527 cfg->fc_flow = nla_get_u32(attr);
530 cfg->fc_mp_alg = nla_get_u32(attr);
533 cfg->fc_table = nla_get_u32(attr);
543 int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
545 struct fib_config cfg;
546 struct fib_table *tb;
549 err = rtm_to_fib_config(skb, nlh, &cfg);
553 tb = fib_get_table(cfg.fc_table);
559 err = tb->tb_delete(tb, &cfg);
564 int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
566 struct fib_config cfg;
567 struct fib_table *tb;
570 err = rtm_to_fib_config(skb, nlh, &cfg);
574 tb = fib_new_table(cfg.fc_table);
580 err = tb->tb_insert(tb, &cfg);
585 int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
588 unsigned int e = 0, s_e;
589 struct fib_table *tb;
590 struct hlist_node *node;
593 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
594 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
595 return ip_rt_dump(skb, cb);
600 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
602 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) {
606 memset(&cb->args[2], 0, sizeof(cb->args) -
607 2 * sizeof(cb->args[0]));
608 if (tb->tb_dump(tb, skb, cb) < 0)
622 /* Prepare and feed intra-kernel routing request.
623 Really, it should be netlink message, but :-( netlink
624 can be not configured, so that we feed it directly
625 to fib engine. It is legal, because all events occur
626 only when netlink is already locked.
629 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
631 struct fib_table *tb;
632 struct fib_config cfg = {
633 .fc_protocol = RTPROT_KERNEL,
636 .fc_dst_len = dst_len,
637 .fc_prefsrc = ifa->ifa_local,
638 .fc_oif = ifa->ifa_dev->dev->ifindex,
639 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
642 if (type == RTN_UNICAST)
643 tb = fib_new_table(RT_TABLE_MAIN);
645 tb = fib_new_table(RT_TABLE_LOCAL);
650 cfg.fc_table = tb->tb_id;
652 if (type != RTN_LOCAL)
653 cfg.fc_scope = RT_SCOPE_LINK;
655 cfg.fc_scope = RT_SCOPE_HOST;
657 if (cmd == RTM_NEWROUTE)
658 tb->tb_insert(tb, &cfg);
660 tb->tb_delete(tb, &cfg);
663 void fib_add_ifaddr(struct in_ifaddr *ifa)
665 struct in_device *in_dev = ifa->ifa_dev;
666 struct net_device *dev = in_dev->dev;
667 struct in_ifaddr *prim = ifa;
668 __be32 mask = ifa->ifa_mask;
669 __be32 addr = ifa->ifa_local;
670 __be32 prefix = ifa->ifa_address&mask;
672 if (ifa->ifa_flags&IFA_F_SECONDARY) {
673 prim = inet_ifa_byprefix(in_dev, prefix, mask);
675 printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n");
680 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
682 if (!(dev->flags&IFF_UP))
685 /* Add broadcast address, if it is explicitly assigned. */
686 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
687 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
689 if (!ZERONET(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) &&
690 (prefix != addr || ifa->ifa_prefixlen < 32)) {
691 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
692 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim);
694 /* Add network specific broadcasts, when it takes a sense */
695 if (ifa->ifa_prefixlen < 31) {
696 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
697 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim);
702 static void fib_del_ifaddr(struct in_ifaddr *ifa)
704 struct in_device *in_dev = ifa->ifa_dev;
705 struct net_device *dev = in_dev->dev;
706 struct in_ifaddr *ifa1;
707 struct in_ifaddr *prim = ifa;
708 __be32 brd = ifa->ifa_address|~ifa->ifa_mask;
709 __be32 any = ifa->ifa_address&ifa->ifa_mask;
716 if (!(ifa->ifa_flags&IFA_F_SECONDARY))
717 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
718 RTN_UNICAST, any, ifa->ifa_prefixlen, prim);
720 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
722 printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n");
727 /* Deletion is more complicated than add.
728 We should take care of not to delete too much :-)
730 Scan address list to be sure that addresses are really gone.
733 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
734 if (ifa->ifa_local == ifa1->ifa_local)
736 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
738 if (brd == ifa1->ifa_broadcast)
740 if (any == ifa1->ifa_broadcast)
745 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
747 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
749 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
750 if (!(ok&LOCAL_OK)) {
751 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
753 /* Check, that this local address finally disappeared. */
754 if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) {
755 /* And the last, but not the least thing.
756 We must flush stray FIB entries.
758 First of all, we scan fib_info list searching
759 for stray nexthop entries, then ignite fib_flush.
761 if (fib_sync_down(ifa->ifa_local, NULL, 0))
771 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb )
774 struct fib_result res;
775 struct flowi fl = { .mark = frn->fl_mark,
776 .nl_u = { .ip4_u = { .daddr = frn->fl_addr,
778 .scope = frn->fl_scope } } };
784 frn->tb_id = tb->tb_id;
785 frn->err = tb->tb_lookup(tb, &fl, &res);
788 frn->prefixlen = res.prefixlen;
789 frn->nh_sel = res.nh_sel;
790 frn->type = res.type;
791 frn->scope = res.scope;
798 static void nl_fib_input(struct sock *sk, int len)
800 struct sk_buff *skb = NULL;
801 struct nlmsghdr *nlh = NULL;
802 struct fib_result_nl *frn;
804 struct fib_table *tb;
806 skb = skb_dequeue(&sk->sk_receive_queue);
810 nlh = (struct nlmsghdr *)skb->data;
811 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
812 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn))) {
817 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
818 tb = fib_get_table(frn->tb_id_in);
820 nl_fib_lookup(frn, tb);
822 pid = NETLINK_CB(skb).pid; /* pid of sending process */
823 NETLINK_CB(skb).pid = 0; /* from kernel */
824 NETLINK_CB(skb).dst_group = 0; /* unicast */
825 netlink_unicast(sk, skb, pid, MSG_DONTWAIT);
828 static void nl_fib_lookup_init(void)
830 netlink_kernel_create(NETLINK_FIB_LOOKUP, 0, nl_fib_input, THIS_MODULE);
833 static void fib_disable_ip(struct net_device *dev, int force)
835 if (fib_sync_down(0, dev, force))
841 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
843 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr;
848 #ifdef CONFIG_IP_ROUTE_MULTIPATH
849 fib_sync_up(ifa->ifa_dev->dev);
855 if (ifa->ifa_dev->ifa_list == NULL) {
856 /* Last address was deleted from this interface.
859 fib_disable_ip(ifa->ifa_dev->dev, 1);
868 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
870 struct net_device *dev = ptr;
871 struct in_device *in_dev = __in_dev_get_rtnl(dev);
873 if (event == NETDEV_UNREGISTER) {
874 fib_disable_ip(dev, 2);
885 } endfor_ifa(in_dev);
886 #ifdef CONFIG_IP_ROUTE_MULTIPATH
892 fib_disable_ip(dev, 0);
894 case NETDEV_CHANGEMTU:
902 static struct notifier_block fib_inetaddr_notifier = {
903 .notifier_call =fib_inetaddr_event,
906 static struct notifier_block fib_netdev_notifier = {
907 .notifier_call =fib_netdev_event,
910 void __init ip_fib_init(void)
914 for (i = 0; i < FIB_TABLE_HASHSZ; i++)
915 INIT_HLIST_HEAD(&fib_table_hash[i]);
916 #ifndef CONFIG_IP_MULTIPLE_TABLES
917 ip_fib_local_table = fib_hash_init(RT_TABLE_LOCAL);
918 hlist_add_head_rcu(&ip_fib_local_table->tb_hlist, &fib_table_hash[0]);
919 ip_fib_main_table = fib_hash_init(RT_TABLE_MAIN);
920 hlist_add_head_rcu(&ip_fib_main_table->tb_hlist, &fib_table_hash[0]);
925 register_netdevice_notifier(&fib_netdev_notifier);
926 register_inetaddr_notifier(&fib_inetaddr_notifier);
927 nl_fib_lookup_init();
930 EXPORT_SYMBOL(inet_addr_type);
931 EXPORT_SYMBOL(ip_dev_find);