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/config.h>
19 #include <linux/module.h>
20 #include <asm/uaccess.h>
21 #include <asm/system.h>
22 #include <linux/bitops.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/netdevice.h>
34 #include <linux/if_arp.h>
35 #include <linux/skbuff.h>
36 #include <linux/netlink.h>
37 #include <linux/init.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
46 #include <net/ip_fib.h>
48 #define FFprint(a...) printk(KERN_DEBUG a)
50 #ifndef CONFIG_IP_MULTIPLE_TABLES
52 #define RT_TABLE_MIN RT_TABLE_MAIN
54 struct fib_table *ip_fib_local_table;
55 struct fib_table *ip_fib_main_table;
59 #define RT_TABLE_MIN 1
61 struct fib_table *fib_tables[RT_TABLE_MAX+1];
63 struct fib_table *__fib_new_table(int id)
67 tb = fib_hash_init(id);
75 #endif /* CONFIG_IP_MULTIPLE_TABLES */
78 static void fib_flush(void)
81 #ifdef CONFIG_IP_MULTIPLE_TABLES
85 for (id = RT_TABLE_MAX; id>0; id--) {
86 if ((tb = fib_get_table(id))==NULL)
88 flushed += tb->tb_flush(tb);
90 #else /* CONFIG_IP_MULTIPLE_TABLES */
91 flushed += ip_fib_main_table->tb_flush(ip_fib_main_table);
92 flushed += ip_fib_local_table->tb_flush(ip_fib_local_table);
93 #endif /* CONFIG_IP_MULTIPLE_TABLES */
100 * Find the first device with a given source address.
103 struct net_device * ip_dev_find(u32 addr)
105 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
106 struct fib_result res;
107 struct net_device *dev = NULL;
109 #ifdef CONFIG_IP_MULTIPLE_TABLES
113 if (!ip_fib_local_table ||
114 ip_fib_local_table->tb_lookup(ip_fib_local_table, &fl, &res))
116 if (res.type != RTN_LOCAL)
118 dev = FIB_RES_DEV(res);
127 unsigned inet_addr_type(u32 addr)
129 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
130 struct fib_result res;
131 unsigned ret = RTN_BROADCAST;
133 if (ZERONET(addr) || BADCLASS(addr))
134 return RTN_BROADCAST;
136 return RTN_MULTICAST;
138 #ifdef CONFIG_IP_MULTIPLE_TABLES
142 if (ip_fib_local_table) {
144 if (!ip_fib_local_table->tb_lookup(ip_fib_local_table,
153 /* Given (packet source, input interface) and optional (dst, oif, tos):
154 - (main) check, that source is valid i.e. not broadcast or our local
156 - figure out what "logical" interface this packet arrived
157 and calculate "specific destination" address.
158 - check, that packet arrived from expected physical interface.
161 int fib_validate_source(u32 src, u32 dst, u8 tos, int oif,
162 struct net_device *dev, u32 *spec_dst, u32 *itag)
164 struct in_device *in_dev;
165 struct flowi fl = { .nl_u = { .ip4_u =
170 struct fib_result res;
176 in_dev = __in_dev_get_rcu(dev);
178 no_addr = in_dev->ifa_list == NULL;
179 rpf = IN_DEV_RPFILTER(in_dev);
186 if (fib_lookup(&fl, &res))
188 if (res.type != RTN_UNICAST)
190 *spec_dst = FIB_RES_PREFSRC(res);
191 fib_combine_itag(itag, &res);
192 #ifdef CONFIG_IP_ROUTE_MULTIPATH
193 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1)
195 if (FIB_RES_DEV(res) == dev)
198 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
207 fl.oif = dev->ifindex;
210 if (fib_lookup(&fl, &res) == 0) {
211 if (res.type == RTN_UNICAST) {
212 *spec_dst = FIB_RES_PREFSRC(res);
213 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
222 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
232 #ifndef CONFIG_IP_NOSIOCRT
235 * Handle IP routing ioctl calls. These are used to manipulate the routing tables
238 int ip_rt_ioctl(unsigned int cmd, void __user *arg)
249 case SIOCADDRT: /* Add a route */
250 case SIOCDELRT: /* Delete a route */
251 if (!capable(CAP_NET_ADMIN))
253 if (copy_from_user(&r, arg, sizeof(struct rtentry)))
256 err = fib_convert_rtentry(cmd, &req.nlh, &req.rtm, &rta, &r);
258 if (cmd == SIOCDELRT) {
259 struct fib_table *tb = fib_get_table(req.rtm.rtm_table);
262 err = tb->tb_delete(tb, &req.rtm, &rta, &req.nlh, NULL);
264 struct fib_table *tb = fib_new_table(req.rtm.rtm_table);
267 err = tb->tb_insert(tb, &req.rtm, &rta, &req.nlh, NULL);
280 int ip_rt_ioctl(unsigned int cmd, void *arg)
287 static int inet_check_attr(struct rtmsg *r, struct rtattr **rta)
291 for (i=1; i<=RTA_MAX; i++) {
292 struct rtattr *attr = rta[i-1];
294 if (RTA_PAYLOAD(attr) < 4)
296 if (i != RTA_MULTIPATH && i != RTA_METRICS)
297 rta[i-1] = (struct rtattr*)RTA_DATA(attr);
303 int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
305 struct fib_table * tb;
306 struct rtattr **rta = arg;
307 struct rtmsg *r = NLMSG_DATA(nlh);
309 if (inet_check_attr(r, rta))
312 tb = fib_get_table(r->rtm_table);
314 return tb->tb_delete(tb, r, (struct kern_rta*)rta, nlh, &NETLINK_CB(skb));
318 int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
320 struct fib_table * tb;
321 struct rtattr **rta = arg;
322 struct rtmsg *r = NLMSG_DATA(nlh);
324 if (inet_check_attr(r, rta))
327 tb = fib_new_table(r->rtm_table);
329 return tb->tb_insert(tb, r, (struct kern_rta*)rta, nlh, &NETLINK_CB(skb));
333 int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
337 struct fib_table *tb;
339 if (NLMSG_PAYLOAD(cb->nlh, 0) >= sizeof(struct rtmsg) &&
340 ((struct rtmsg*)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED)
341 return ip_rt_dump(skb, cb);
345 s_t = cb->args[0] = RT_TABLE_MIN;
347 for (t=s_t; t<=RT_TABLE_MAX; t++) {
348 if (t < s_t) continue;
350 memset(&cb->args[1], 0, sizeof(cb->args)-sizeof(cb->args[0]));
351 if ((tb = fib_get_table(t))==NULL)
353 if (tb->tb_dump(tb, skb, cb) < 0)
362 /* Prepare and feed intra-kernel routing request.
363 Really, it should be netlink message, but :-( netlink
364 can be not configured, so that we feed it directly
365 to fib engine. It is legal, because all events occur
366 only when netlink is already locked.
369 static void fib_magic(int cmd, int type, u32 dst, int dst_len, struct in_ifaddr *ifa)
371 struct fib_table * tb;
378 memset(&req.rtm, 0, sizeof(req.rtm));
379 memset(&rta, 0, sizeof(rta));
381 if (type == RTN_UNICAST)
382 tb = fib_new_table(RT_TABLE_MAIN);
384 tb = fib_new_table(RT_TABLE_LOCAL);
389 req.nlh.nlmsg_len = sizeof(req);
390 req.nlh.nlmsg_type = cmd;
391 req.nlh.nlmsg_flags = NLM_F_REQUEST|NLM_F_CREATE|NLM_F_APPEND;
392 req.nlh.nlmsg_pid = 0;
393 req.nlh.nlmsg_seq = 0;
395 req.rtm.rtm_dst_len = dst_len;
396 req.rtm.rtm_table = tb->tb_id;
397 req.rtm.rtm_protocol = RTPROT_KERNEL;
398 req.rtm.rtm_scope = (type != RTN_LOCAL ? RT_SCOPE_LINK : RT_SCOPE_HOST);
399 req.rtm.rtm_type = type;
402 rta.rta_prefsrc = &ifa->ifa_local;
403 rta.rta_oif = &ifa->ifa_dev->dev->ifindex;
405 if (cmd == RTM_NEWROUTE)
406 tb->tb_insert(tb, &req.rtm, &rta, &req.nlh, NULL);
408 tb->tb_delete(tb, &req.rtm, &rta, &req.nlh, NULL);
411 static void fib_add_ifaddr(struct in_ifaddr *ifa)
413 struct in_device *in_dev = ifa->ifa_dev;
414 struct net_device *dev = in_dev->dev;
415 struct in_ifaddr *prim = ifa;
416 u32 mask = ifa->ifa_mask;
417 u32 addr = ifa->ifa_local;
418 u32 prefix = ifa->ifa_address&mask;
420 if (ifa->ifa_flags&IFA_F_SECONDARY) {
421 prim = inet_ifa_byprefix(in_dev, prefix, mask);
423 printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n");
428 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
430 if (!(dev->flags&IFF_UP))
433 /* Add broadcast address, if it is explicitly assigned. */
434 if (ifa->ifa_broadcast && ifa->ifa_broadcast != 0xFFFFFFFF)
435 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
437 if (!ZERONET(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) &&
438 (prefix != addr || ifa->ifa_prefixlen < 32)) {
439 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
440 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim);
442 /* Add network specific broadcasts, when it takes a sense */
443 if (ifa->ifa_prefixlen < 31) {
444 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
445 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim);
450 static void fib_del_ifaddr(struct in_ifaddr *ifa)
452 struct in_device *in_dev = ifa->ifa_dev;
453 struct net_device *dev = in_dev->dev;
454 struct in_ifaddr *ifa1;
455 struct in_ifaddr *prim = ifa;
456 u32 brd = ifa->ifa_address|~ifa->ifa_mask;
457 u32 any = ifa->ifa_address&ifa->ifa_mask;
464 if (!(ifa->ifa_flags&IFA_F_SECONDARY))
465 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
466 RTN_UNICAST, any, ifa->ifa_prefixlen, prim);
468 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
470 printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n");
475 /* Deletion is more complicated than add.
476 We should take care of not to delete too much :-)
478 Scan address list to be sure that addresses are really gone.
481 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
482 if (ifa->ifa_local == ifa1->ifa_local)
484 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
486 if (brd == ifa1->ifa_broadcast)
488 if (any == ifa1->ifa_broadcast)
493 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
495 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
497 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
498 if (!(ok&LOCAL_OK)) {
499 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
501 /* Check, that this local address finally disappeared. */
502 if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) {
503 /* And the last, but not the least thing.
504 We must flush stray FIB entries.
506 First of all, we scan fib_info list searching
507 for stray nexthop entries, then ignite fib_flush.
509 if (fib_sync_down(ifa->ifa_local, NULL, 0))
519 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb )
522 struct fib_result res;
523 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = frn->fl_addr,
524 .fwmark = frn->fl_fwmark,
526 .scope = frn->fl_scope } } };
530 frn->tb_id = tb->tb_id;
531 frn->err = tb->tb_lookup(tb, &fl, &res);
534 frn->prefixlen = res.prefixlen;
535 frn->nh_sel = res.nh_sel;
536 frn->type = res.type;
537 frn->scope = res.scope;
543 static void nl_fib_input(struct sock *sk, int len)
545 struct sk_buff *skb = NULL;
546 struct nlmsghdr *nlh = NULL;
547 struct fib_result_nl *frn;
550 struct fib_table *tb;
552 skb = skb_recv_datagram(sk, 0, 0, &err);
553 nlh = (struct nlmsghdr *)skb->data;
555 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
556 tb = fib_get_table(frn->tb_id_in);
558 nl_fib_lookup(frn, tb);
560 pid = nlh->nlmsg_pid; /*pid of sending process */
561 NETLINK_CB(skb).pid = 0; /* from kernel */
562 NETLINK_CB(skb).dst_pid = pid;
563 NETLINK_CB(skb).dst_group = 0; /* unicast */
564 netlink_unicast(sk, skb, pid, MSG_DONTWAIT);
567 static void nl_fib_lookup_init(void)
569 netlink_kernel_create(NETLINK_FIB_LOOKUP, 0, nl_fib_input, THIS_MODULE);
572 static void fib_disable_ip(struct net_device *dev, int force)
574 if (fib_sync_down(0, dev, force))
580 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
582 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr;
587 #ifdef CONFIG_IP_ROUTE_MULTIPATH
588 fib_sync_up(ifa->ifa_dev->dev);
594 if (ifa->ifa_dev && ifa->ifa_dev->ifa_list == NULL) {
595 /* Last address was deleted from this interface.
598 fib_disable_ip(ifa->ifa_dev->dev, 1);
607 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
609 struct net_device *dev = ptr;
610 struct in_device *in_dev = __in_dev_get_rtnl(dev);
612 if (event == NETDEV_UNREGISTER) {
613 fib_disable_ip(dev, 2);
624 } endfor_ifa(in_dev);
625 #ifdef CONFIG_IP_ROUTE_MULTIPATH
631 fib_disable_ip(dev, 0);
633 case NETDEV_CHANGEMTU:
641 static struct notifier_block fib_inetaddr_notifier = {
642 .notifier_call =fib_inetaddr_event,
645 static struct notifier_block fib_netdev_notifier = {
646 .notifier_call =fib_netdev_event,
649 void __init ip_fib_init(void)
651 #ifndef CONFIG_IP_MULTIPLE_TABLES
652 ip_fib_local_table = fib_hash_init(RT_TABLE_LOCAL);
653 ip_fib_main_table = fib_hash_init(RT_TABLE_MAIN);
658 register_netdevice_notifier(&fib_netdev_notifier);
659 register_inetaddr_notifier(&fib_inetaddr_notifier);
660 nl_fib_lookup_init();
663 EXPORT_SYMBOL(inet_addr_type);
664 EXPORT_SYMBOL(ip_rt_ioctl);