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>
25 #include <linux/sched.h>
27 #include <linux/string.h>
28 #include <linux/socket.h>
29 #include <linux/sockios.h>
30 #include <linux/errno.h>
32 #include <linux/inet.h>
33 #include <linux/inetdevice.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_addr.h>
36 #include <linux/if_arp.h>
37 #include <linux/skbuff.h>
38 #include <linux/netlink.h>
39 #include <linux/init.h>
40 #include <linux/list.h>
43 #include <net/protocol.h>
44 #include <net/route.h>
49 #include <net/ip_fib.h>
51 #define FFprint(a...) printk(KERN_DEBUG a)
53 #ifndef CONFIG_IP_MULTIPLE_TABLES
55 struct fib_table *ip_fib_local_table;
56 struct fib_table *ip_fib_main_table;
58 #define FIB_TABLE_HASHSZ 1
59 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
63 #define FIB_TABLE_HASHSZ 256
64 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
66 struct fib_table *fib_new_table(u32 id)
73 tb = fib_get_table(id);
76 tb = fib_hash_init(id);
79 h = id & (FIB_TABLE_HASHSZ - 1);
80 hlist_add_head_rcu(&tb->tb_hlist, &fib_table_hash[h]);
84 struct fib_table *fib_get_table(u32 id)
87 struct hlist_node *node;
92 h = id & (FIB_TABLE_HASHSZ - 1);
94 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb_hlist) {
95 if (tb->tb_id == id) {
103 #endif /* CONFIG_IP_MULTIPLE_TABLES */
105 static void fib_flush(void)
108 struct fib_table *tb;
109 struct hlist_node *node;
112 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
113 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist)
114 flushed += tb->tb_flush(tb);
122 * Find the first device with a given source address.
125 struct net_device * ip_dev_find(u32 addr)
127 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
128 struct fib_result res;
129 struct net_device *dev = NULL;
131 #ifdef CONFIG_IP_MULTIPLE_TABLES
135 if (!ip_fib_local_table ||
136 ip_fib_local_table->tb_lookup(ip_fib_local_table, &fl, &res))
138 if (res.type != RTN_LOCAL)
140 dev = FIB_RES_DEV(res);
149 unsigned inet_addr_type(u32 addr)
151 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
152 struct fib_result res;
153 unsigned ret = RTN_BROADCAST;
155 if (ZERONET(addr) || BADCLASS(addr))
156 return RTN_BROADCAST;
158 return RTN_MULTICAST;
160 #ifdef CONFIG_IP_MULTIPLE_TABLES
164 if (ip_fib_local_table) {
166 if (!ip_fib_local_table->tb_lookup(ip_fib_local_table,
175 /* Given (packet source, input interface) and optional (dst, oif, tos):
176 - (main) check, that source is valid i.e. not broadcast or our local
178 - figure out what "logical" interface this packet arrived
179 and calculate "specific destination" address.
180 - check, that packet arrived from expected physical interface.
183 int fib_validate_source(u32 src, u32 dst, u8 tos, int oif,
184 struct net_device *dev, u32 *spec_dst, u32 *itag)
186 struct in_device *in_dev;
187 struct flowi fl = { .nl_u = { .ip4_u =
192 struct fib_result res;
198 in_dev = __in_dev_get_rcu(dev);
200 no_addr = in_dev->ifa_list == NULL;
201 rpf = IN_DEV_RPFILTER(in_dev);
208 if (fib_lookup(&fl, &res))
210 if (res.type != RTN_UNICAST)
212 *spec_dst = FIB_RES_PREFSRC(res);
213 fib_combine_itag(itag, &res);
214 #ifdef CONFIG_IP_ROUTE_MULTIPATH
215 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1)
217 if (FIB_RES_DEV(res) == dev)
220 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
229 fl.oif = dev->ifindex;
232 if (fib_lookup(&fl, &res) == 0) {
233 if (res.type == RTN_UNICAST) {
234 *spec_dst = FIB_RES_PREFSRC(res);
235 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
244 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
254 #ifndef CONFIG_IP_NOSIOCRT
256 static inline u32 sk_extract_addr(struct sockaddr *addr)
258 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
261 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
265 nla = (struct nlattr *) ((char *) mx + len);
266 nla->nla_type = type;
267 nla->nla_len = nla_attr_size(4);
268 *(u32 *) nla_data(nla) = value;
270 return len + nla_total_size(4);
273 static int rtentry_to_fib_config(int cmd, struct rtentry *rt,
274 struct fib_config *cfg)
279 memset(cfg, 0, sizeof(*cfg));
281 if (rt->rt_dst.sa_family != AF_INET)
282 return -EAFNOSUPPORT;
285 * Check mask for validity:
286 * a) it must be contiguous.
287 * b) destination must have all host bits clear.
288 * c) if application forgot to set correct family (AF_INET),
289 * reject request unless it is absolutely clear i.e.
290 * both family and mask are zero.
293 addr = sk_extract_addr(&rt->rt_dst);
294 if (!(rt->rt_flags & RTF_HOST)) {
295 u32 mask = sk_extract_addr(&rt->rt_genmask);
297 if (rt->rt_genmask.sa_family != AF_INET) {
298 if (mask || rt->rt_genmask.sa_family)
299 return -EAFNOSUPPORT;
302 if (bad_mask(mask, addr))
305 plen = inet_mask_len(mask);
308 cfg->fc_dst_len = plen;
311 if (cmd != SIOCDELRT) {
312 cfg->fc_nlflags = NLM_F_CREATE;
313 cfg->fc_protocol = RTPROT_BOOT;
317 cfg->fc_priority = rt->rt_metric - 1;
319 if (rt->rt_flags & RTF_REJECT) {
320 cfg->fc_scope = RT_SCOPE_HOST;
321 cfg->fc_type = RTN_UNREACHABLE;
325 cfg->fc_scope = RT_SCOPE_NOWHERE;
326 cfg->fc_type = RTN_UNICAST;
330 struct net_device *dev;
331 char devname[IFNAMSIZ];
333 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
336 devname[IFNAMSIZ-1] = 0;
337 colon = strchr(devname, ':');
340 dev = __dev_get_by_name(devname);
343 cfg->fc_oif = dev->ifindex;
345 struct in_ifaddr *ifa;
346 struct in_device *in_dev = __in_dev_get_rtnl(dev);
350 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
351 if (strcmp(ifa->ifa_label, devname) == 0)
355 cfg->fc_prefsrc = ifa->ifa_local;
359 addr = sk_extract_addr(&rt->rt_gateway);
360 if (rt->rt_gateway.sa_family == AF_INET && addr) {
362 if (rt->rt_flags & RTF_GATEWAY &&
363 inet_addr_type(addr) == RTN_UNICAST)
364 cfg->fc_scope = RT_SCOPE_UNIVERSE;
367 if (cmd == SIOCDELRT)
370 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
373 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
374 cfg->fc_scope = RT_SCOPE_LINK;
376 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
380 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
384 if (rt->rt_flags & RTF_MTU)
385 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
387 if (rt->rt_flags & RTF_WINDOW)
388 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
390 if (rt->rt_flags & RTF_IRTT)
391 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
394 cfg->fc_mx_len = len;
401 * Handle IP routing ioctl calls. These are used to manipulate the routing tables
404 int ip_rt_ioctl(unsigned int cmd, void __user *arg)
406 struct fib_config cfg;
411 case SIOCADDRT: /* Add a route */
412 case SIOCDELRT: /* Delete a route */
413 if (!capable(CAP_NET_ADMIN))
416 if (copy_from_user(&rt, arg, sizeof(rt)))
420 err = rtentry_to_fib_config(cmd, &rt, &cfg);
422 struct fib_table *tb;
424 if (cmd == SIOCDELRT) {
425 tb = fib_get_table(cfg.fc_table);
427 err = tb->tb_delete(tb, &cfg);
431 tb = fib_new_table(cfg.fc_table);
433 err = tb->tb_insert(tb, &cfg);
438 /* allocated by rtentry_to_fib_config() */
449 int ip_rt_ioctl(unsigned int cmd, void *arg)
456 struct nla_policy rtm_ipv4_policy[RTA_MAX+1] __read_mostly = {
457 [RTA_DST] = { .type = NLA_U32 },
458 [RTA_SRC] = { .type = NLA_U32 },
459 [RTA_IIF] = { .type = NLA_U32 },
460 [RTA_OIF] = { .type = NLA_U32 },
461 [RTA_GATEWAY] = { .type = NLA_U32 },
462 [RTA_PRIORITY] = { .type = NLA_U32 },
463 [RTA_PREFSRC] = { .type = NLA_U32 },
464 [RTA_METRICS] = { .type = NLA_NESTED },
465 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
466 [RTA_PROTOINFO] = { .type = NLA_U32 },
467 [RTA_FLOW] = { .type = NLA_U32 },
468 [RTA_MP_ALGO] = { .type = NLA_U32 },
471 static int rtm_to_fib_config(struct sk_buff *skb, struct nlmsghdr *nlh,
472 struct fib_config *cfg)
478 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
482 memset(cfg, 0, sizeof(*cfg));
484 rtm = nlmsg_data(nlh);
485 cfg->fc_family = rtm->rtm_family;
486 cfg->fc_dst_len = rtm->rtm_dst_len;
487 cfg->fc_src_len = rtm->rtm_src_len;
488 cfg->fc_tos = rtm->rtm_tos;
489 cfg->fc_table = rtm->rtm_table;
490 cfg->fc_protocol = rtm->rtm_protocol;
491 cfg->fc_scope = rtm->rtm_scope;
492 cfg->fc_type = rtm->rtm_type;
493 cfg->fc_flags = rtm->rtm_flags;
494 cfg->fc_nlflags = nlh->nlmsg_flags;
496 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
497 cfg->fc_nlinfo.nlh = nlh;
499 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
500 switch (attr->nla_type) {
502 cfg->fc_dst = nla_get_u32(attr);
505 cfg->fc_src = nla_get_u32(attr);
508 cfg->fc_oif = nla_get_u32(attr);
511 cfg->fc_gw = nla_get_u32(attr);
514 cfg->fc_priority = nla_get_u32(attr);
517 cfg->fc_prefsrc = nla_get_u32(attr);
520 cfg->fc_mx = nla_data(attr);
521 cfg->fc_mx_len = nla_len(attr);
524 cfg->fc_mp = nla_data(attr);
525 cfg->fc_mp_len = nla_len(attr);
528 cfg->fc_flow = nla_get_u32(attr);
531 cfg->fc_mp_alg = nla_get_u32(attr);
534 cfg->fc_table = nla_get_u32(attr);
544 int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
546 struct fib_config cfg;
547 struct fib_table *tb;
550 err = rtm_to_fib_config(skb, nlh, &cfg);
554 tb = fib_get_table(cfg.fc_table);
560 err = tb->tb_delete(tb, &cfg);
565 int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
567 struct fib_config cfg;
568 struct fib_table *tb;
571 err = rtm_to_fib_config(skb, nlh, &cfg);
575 tb = fib_new_table(cfg.fc_table);
581 err = tb->tb_insert(tb, &cfg);
586 int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
589 unsigned int e = 0, s_e;
590 struct fib_table *tb;
591 struct hlist_node *node;
594 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
595 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
596 return ip_rt_dump(skb, cb);
601 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
603 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) {
607 memset(&cb->args[2], 0, sizeof(cb->args) -
608 2 * sizeof(cb->args[0]));
609 if (tb->tb_dump(tb, skb, cb) < 0)
623 /* Prepare and feed intra-kernel routing request.
624 Really, it should be netlink message, but :-( netlink
625 can be not configured, so that we feed it directly
626 to fib engine. It is legal, because all events occur
627 only when netlink is already locked.
630 static void fib_magic(int cmd, int type, u32 dst, int dst_len,
631 struct in_ifaddr *ifa)
633 struct fib_table *tb;
634 struct fib_config cfg = {
635 .fc_protocol = RTPROT_KERNEL,
638 .fc_dst_len = dst_len,
639 .fc_prefsrc = ifa->ifa_local,
640 .fc_oif = ifa->ifa_dev->dev->ifindex,
641 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
644 if (type == RTN_UNICAST)
645 tb = fib_new_table(RT_TABLE_MAIN);
647 tb = fib_new_table(RT_TABLE_LOCAL);
652 cfg.fc_table = tb->tb_id;
654 if (type != RTN_LOCAL)
655 cfg.fc_scope = RT_SCOPE_LINK;
657 cfg.fc_scope = RT_SCOPE_HOST;
659 if (cmd == RTM_NEWROUTE)
660 tb->tb_insert(tb, &cfg);
662 tb->tb_delete(tb, &cfg);
665 void fib_add_ifaddr(struct in_ifaddr *ifa)
667 struct in_device *in_dev = ifa->ifa_dev;
668 struct net_device *dev = in_dev->dev;
669 struct in_ifaddr *prim = ifa;
670 u32 mask = ifa->ifa_mask;
671 u32 addr = ifa->ifa_local;
672 u32 prefix = ifa->ifa_address&mask;
674 if (ifa->ifa_flags&IFA_F_SECONDARY) {
675 prim = inet_ifa_byprefix(in_dev, prefix, mask);
677 printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n");
682 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
684 if (!(dev->flags&IFF_UP))
687 /* Add broadcast address, if it is explicitly assigned. */
688 if (ifa->ifa_broadcast && ifa->ifa_broadcast != 0xFFFFFFFF)
689 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
691 if (!ZERONET(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) &&
692 (prefix != addr || ifa->ifa_prefixlen < 32)) {
693 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
694 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim);
696 /* Add network specific broadcasts, when it takes a sense */
697 if (ifa->ifa_prefixlen < 31) {
698 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
699 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim);
704 static void fib_del_ifaddr(struct in_ifaddr *ifa)
706 struct in_device *in_dev = ifa->ifa_dev;
707 struct net_device *dev = in_dev->dev;
708 struct in_ifaddr *ifa1;
709 struct in_ifaddr *prim = ifa;
710 u32 brd = ifa->ifa_address|~ifa->ifa_mask;
711 u32 any = ifa->ifa_address&ifa->ifa_mask;
718 if (!(ifa->ifa_flags&IFA_F_SECONDARY))
719 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
720 RTN_UNICAST, any, ifa->ifa_prefixlen, prim);
722 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
724 printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n");
729 /* Deletion is more complicated than add.
730 We should take care of not to delete too much :-)
732 Scan address list to be sure that addresses are really gone.
735 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
736 if (ifa->ifa_local == ifa1->ifa_local)
738 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
740 if (brd == ifa1->ifa_broadcast)
742 if (any == ifa1->ifa_broadcast)
747 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
749 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
751 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
752 if (!(ok&LOCAL_OK)) {
753 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
755 /* Check, that this local address finally disappeared. */
756 if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) {
757 /* And the last, but not the least thing.
758 We must flush stray FIB entries.
760 First of all, we scan fib_info list searching
761 for stray nexthop entries, then ignite fib_flush.
763 if (fib_sync_down(ifa->ifa_local, NULL, 0))
773 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb )
776 struct fib_result res;
777 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = frn->fl_addr,
778 .fwmark = frn->fl_fwmark,
780 .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;
797 static void nl_fib_input(struct sock *sk, int len)
799 struct sk_buff *skb = NULL;
800 struct nlmsghdr *nlh = NULL;
801 struct fib_result_nl *frn;
803 struct fib_table *tb;
805 skb = skb_dequeue(&sk->sk_receive_queue);
806 nlh = (struct nlmsghdr *)skb->data;
807 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
808 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn))) {
813 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
814 tb = fib_get_table(frn->tb_id_in);
816 nl_fib_lookup(frn, tb);
818 pid = nlh->nlmsg_pid; /*pid of sending process */
819 NETLINK_CB(skb).pid = 0; /* from kernel */
820 NETLINK_CB(skb).dst_pid = pid;
821 NETLINK_CB(skb).dst_group = 0; /* unicast */
822 netlink_unicast(sk, skb, pid, MSG_DONTWAIT);
825 static void nl_fib_lookup_init(void)
827 netlink_kernel_create(NETLINK_FIB_LOOKUP, 0, nl_fib_input, THIS_MODULE);
830 static void fib_disable_ip(struct net_device *dev, int force)
832 if (fib_sync_down(0, dev, force))
838 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
840 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr;
845 #ifdef CONFIG_IP_ROUTE_MULTIPATH
846 fib_sync_up(ifa->ifa_dev->dev);
852 if (ifa->ifa_dev->ifa_list == NULL) {
853 /* Last address was deleted from this interface.
856 fib_disable_ip(ifa->ifa_dev->dev, 1);
865 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
867 struct net_device *dev = ptr;
868 struct in_device *in_dev = __in_dev_get_rtnl(dev);
870 if (event == NETDEV_UNREGISTER) {
871 fib_disable_ip(dev, 2);
882 } endfor_ifa(in_dev);
883 #ifdef CONFIG_IP_ROUTE_MULTIPATH
889 fib_disable_ip(dev, 0);
891 case NETDEV_CHANGEMTU:
899 static struct notifier_block fib_inetaddr_notifier = {
900 .notifier_call =fib_inetaddr_event,
903 static struct notifier_block fib_netdev_notifier = {
904 .notifier_call =fib_netdev_event,
907 void __init ip_fib_init(void)
911 for (i = 0; i < FIB_TABLE_HASHSZ; i++)
912 INIT_HLIST_HEAD(&fib_table_hash[i]);
913 #ifndef CONFIG_IP_MULTIPLE_TABLES
914 ip_fib_local_table = fib_hash_init(RT_TABLE_LOCAL);
915 hlist_add_head_rcu(&ip_fib_local_table->tb_hlist, &fib_table_hash[0]);
916 ip_fib_main_table = fib_hash_init(RT_TABLE_MAIN);
917 hlist_add_head_rcu(&ip_fib_main_table->tb_hlist, &fib_table_hash[0]);
922 register_netdevice_notifier(&fib_netdev_notifier);
923 register_inetaddr_notifier(&fib_inetaddr_notifier);
924 nl_fib_lookup_init();
927 EXPORT_SYMBOL(inet_addr_type);
928 EXPORT_SYMBOL(ip_dev_find);