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>
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 #define RT_TABLE_MIN RT_TABLE_MAIN
56 struct fib_table *ip_fib_local_table;
57 struct fib_table *ip_fib_main_table;
61 #define RT_TABLE_MIN 1
63 struct fib_table *fib_tables[RT_TABLE_MAX+1];
65 struct fib_table *__fib_new_table(u32 id)
69 tb = fib_hash_init(id);
77 #endif /* CONFIG_IP_MULTIPLE_TABLES */
80 static void fib_flush(void)
83 #ifdef CONFIG_IP_MULTIPLE_TABLES
87 for (id = RT_TABLE_MAX; id>0; id--) {
88 if ((tb = fib_get_table(id))==NULL)
90 flushed += tb->tb_flush(tb);
92 #else /* CONFIG_IP_MULTIPLE_TABLES */
93 flushed += ip_fib_main_table->tb_flush(ip_fib_main_table);
94 flushed += ip_fib_local_table->tb_flush(ip_fib_local_table);
95 #endif /* CONFIG_IP_MULTIPLE_TABLES */
102 * Find the first device with a given source address.
105 struct net_device * ip_dev_find(u32 addr)
107 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
108 struct fib_result res;
109 struct net_device *dev = NULL;
111 #ifdef CONFIG_IP_MULTIPLE_TABLES
115 if (!ip_fib_local_table ||
116 ip_fib_local_table->tb_lookup(ip_fib_local_table, &fl, &res))
118 if (res.type != RTN_LOCAL)
120 dev = FIB_RES_DEV(res);
129 unsigned inet_addr_type(u32 addr)
131 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
132 struct fib_result res;
133 unsigned ret = RTN_BROADCAST;
135 if (ZERONET(addr) || BADCLASS(addr))
136 return RTN_BROADCAST;
138 return RTN_MULTICAST;
140 #ifdef CONFIG_IP_MULTIPLE_TABLES
144 if (ip_fib_local_table) {
146 if (!ip_fib_local_table->tb_lookup(ip_fib_local_table,
155 /* Given (packet source, input interface) and optional (dst, oif, tos):
156 - (main) check, that source is valid i.e. not broadcast or our local
158 - figure out what "logical" interface this packet arrived
159 and calculate "specific destination" address.
160 - check, that packet arrived from expected physical interface.
163 int fib_validate_source(u32 src, u32 dst, u8 tos, int oif,
164 struct net_device *dev, u32 *spec_dst, u32 *itag)
166 struct in_device *in_dev;
167 struct flowi fl = { .nl_u = { .ip4_u =
172 struct fib_result res;
178 in_dev = __in_dev_get_rcu(dev);
180 no_addr = in_dev->ifa_list == NULL;
181 rpf = IN_DEV_RPFILTER(in_dev);
188 if (fib_lookup(&fl, &res))
190 if (res.type != RTN_UNICAST)
192 *spec_dst = FIB_RES_PREFSRC(res);
193 fib_combine_itag(itag, &res);
194 #ifdef CONFIG_IP_ROUTE_MULTIPATH
195 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1)
197 if (FIB_RES_DEV(res) == dev)
200 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
209 fl.oif = dev->ifindex;
212 if (fib_lookup(&fl, &res) == 0) {
213 if (res.type == RTN_UNICAST) {
214 *spec_dst = FIB_RES_PREFSRC(res);
215 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
224 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
234 #ifndef CONFIG_IP_NOSIOCRT
237 * Handle IP routing ioctl calls. These are used to manipulate the routing tables
240 int ip_rt_ioctl(unsigned int cmd, void __user *arg)
251 case SIOCADDRT: /* Add a route */
252 case SIOCDELRT: /* Delete a route */
253 if (!capable(CAP_NET_ADMIN))
255 if (copy_from_user(&r, arg, sizeof(struct rtentry)))
258 err = fib_convert_rtentry(cmd, &req.nlh, &req.rtm, &rta, &r);
260 if (cmd == SIOCDELRT) {
261 struct fib_table *tb = fib_get_table(req.rtm.rtm_table);
264 err = tb->tb_delete(tb, &req.rtm, &rta, &req.nlh, NULL);
266 struct fib_table *tb = fib_new_table(req.rtm.rtm_table);
269 err = tb->tb_insert(tb, &req.rtm, &rta, &req.nlh, NULL);
281 int ip_rt_ioctl(unsigned int cmd, void *arg)
288 static int inet_check_attr(struct rtmsg *r, struct rtattr **rta)
292 for (i=1; i<=RTA_MAX; i++, rta++) {
293 struct rtattr *attr = *rta;
295 if (RTA_PAYLOAD(attr) < 4)
297 if (i != RTA_MULTIPATH && i != RTA_METRICS)
298 *rta = (struct rtattr*)RTA_DATA(attr);
304 int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
306 struct fib_table * tb;
307 struct rtattr **rta = arg;
308 struct rtmsg *r = NLMSG_DATA(nlh);
310 if (inet_check_attr(r, rta))
313 tb = fib_get_table(r->rtm_table);
315 return tb->tb_delete(tb, r, (struct kern_rta*)rta, nlh, &NETLINK_CB(skb));
319 int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
321 struct fib_table * tb;
322 struct rtattr **rta = arg;
323 struct rtmsg *r = NLMSG_DATA(nlh);
325 if (inet_check_attr(r, rta))
328 tb = fib_new_table(r->rtm_table);
330 return tb->tb_insert(tb, r, (struct kern_rta*)rta, nlh, &NETLINK_CB(skb));
334 int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
338 struct fib_table *tb;
340 if (NLMSG_PAYLOAD(cb->nlh, 0) >= sizeof(struct rtmsg) &&
341 ((struct rtmsg*)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED)
342 return ip_rt_dump(skb, cb);
346 s_t = cb->args[0] = RT_TABLE_MIN;
348 for (t=s_t; t<=RT_TABLE_MAX; t++) {
349 if (t < s_t) continue;
351 memset(&cb->args[1], 0, sizeof(cb->args)-sizeof(cb->args[0]));
352 if ((tb = fib_get_table(t))==NULL)
354 if (tb->tb_dump(tb, skb, cb) < 0)
363 /* Prepare and feed intra-kernel routing request.
364 Really, it should be netlink message, but :-( netlink
365 can be not configured, so that we feed it directly
366 to fib engine. It is legal, because all events occur
367 only when netlink is already locked.
370 static void fib_magic(int cmd, int type, u32 dst, int dst_len, struct in_ifaddr *ifa)
372 struct fib_table * tb;
379 memset(&req.rtm, 0, sizeof(req.rtm));
380 memset(&rta, 0, sizeof(rta));
382 if (type == RTN_UNICAST)
383 tb = fib_new_table(RT_TABLE_MAIN);
385 tb = fib_new_table(RT_TABLE_LOCAL);
390 req.nlh.nlmsg_len = sizeof(req);
391 req.nlh.nlmsg_type = cmd;
392 req.nlh.nlmsg_flags = NLM_F_REQUEST|NLM_F_CREATE|NLM_F_APPEND;
393 req.nlh.nlmsg_pid = 0;
394 req.nlh.nlmsg_seq = 0;
396 req.rtm.rtm_dst_len = dst_len;
397 req.rtm.rtm_table = tb->tb_id;
398 req.rtm.rtm_protocol = RTPROT_KERNEL;
399 req.rtm.rtm_scope = (type != RTN_LOCAL ? RT_SCOPE_LINK : RT_SCOPE_HOST);
400 req.rtm.rtm_type = type;
403 rta.rta_prefsrc = &ifa->ifa_local;
404 rta.rta_oif = &ifa->ifa_dev->dev->ifindex;
406 if (cmd == RTM_NEWROUTE)
407 tb->tb_insert(tb, &req.rtm, &rta, &req.nlh, NULL);
409 tb->tb_delete(tb, &req.rtm, &rta, &req.nlh, NULL);
412 void fib_add_ifaddr(struct in_ifaddr *ifa)
414 struct in_device *in_dev = ifa->ifa_dev;
415 struct net_device *dev = in_dev->dev;
416 struct in_ifaddr *prim = ifa;
417 u32 mask = ifa->ifa_mask;
418 u32 addr = ifa->ifa_local;
419 u32 prefix = ifa->ifa_address&mask;
421 if (ifa->ifa_flags&IFA_F_SECONDARY) {
422 prim = inet_ifa_byprefix(in_dev, prefix, mask);
424 printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n");
429 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
431 if (!(dev->flags&IFF_UP))
434 /* Add broadcast address, if it is explicitly assigned. */
435 if (ifa->ifa_broadcast && ifa->ifa_broadcast != 0xFFFFFFFF)
436 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
438 if (!ZERONET(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) &&
439 (prefix != addr || ifa->ifa_prefixlen < 32)) {
440 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
441 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim);
443 /* Add network specific broadcasts, when it takes a sense */
444 if (ifa->ifa_prefixlen < 31) {
445 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
446 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim);
451 static void fib_del_ifaddr(struct in_ifaddr *ifa)
453 struct in_device *in_dev = ifa->ifa_dev;
454 struct net_device *dev = in_dev->dev;
455 struct in_ifaddr *ifa1;
456 struct in_ifaddr *prim = ifa;
457 u32 brd = ifa->ifa_address|~ifa->ifa_mask;
458 u32 any = ifa->ifa_address&ifa->ifa_mask;
465 if (!(ifa->ifa_flags&IFA_F_SECONDARY))
466 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
467 RTN_UNICAST, any, ifa->ifa_prefixlen, prim);
469 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
471 printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n");
476 /* Deletion is more complicated than add.
477 We should take care of not to delete too much :-)
479 Scan address list to be sure that addresses are really gone.
482 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
483 if (ifa->ifa_local == ifa1->ifa_local)
485 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
487 if (brd == ifa1->ifa_broadcast)
489 if (any == ifa1->ifa_broadcast)
494 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
496 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
498 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
499 if (!(ok&LOCAL_OK)) {
500 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
502 /* Check, that this local address finally disappeared. */
503 if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) {
504 /* And the last, but not the least thing.
505 We must flush stray FIB entries.
507 First of all, we scan fib_info list searching
508 for stray nexthop entries, then ignite fib_flush.
510 if (fib_sync_down(ifa->ifa_local, NULL, 0))
520 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb )
523 struct fib_result res;
524 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = frn->fl_addr,
525 .fwmark = frn->fl_fwmark,
527 .scope = frn->fl_scope } } };
531 frn->tb_id = tb->tb_id;
532 frn->err = tb->tb_lookup(tb, &fl, &res);
535 frn->prefixlen = res.prefixlen;
536 frn->nh_sel = res.nh_sel;
537 frn->type = res.type;
538 frn->scope = res.scope;
544 static void nl_fib_input(struct sock *sk, int len)
546 struct sk_buff *skb = NULL;
547 struct nlmsghdr *nlh = NULL;
548 struct fib_result_nl *frn;
550 struct fib_table *tb;
552 skb = skb_dequeue(&sk->sk_receive_queue);
553 nlh = (struct nlmsghdr *)skb->data;
554 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
555 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn))) {
560 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
561 tb = fib_get_table(frn->tb_id_in);
563 nl_fib_lookup(frn, tb);
565 pid = nlh->nlmsg_pid; /*pid of sending process */
566 NETLINK_CB(skb).pid = 0; /* from kernel */
567 NETLINK_CB(skb).dst_pid = pid;
568 NETLINK_CB(skb).dst_group = 0; /* unicast */
569 netlink_unicast(sk, skb, pid, MSG_DONTWAIT);
572 static void nl_fib_lookup_init(void)
574 netlink_kernel_create(NETLINK_FIB_LOOKUP, 0, nl_fib_input, THIS_MODULE);
577 static void fib_disable_ip(struct net_device *dev, int force)
579 if (fib_sync_down(0, dev, force))
585 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
587 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr;
592 #ifdef CONFIG_IP_ROUTE_MULTIPATH
593 fib_sync_up(ifa->ifa_dev->dev);
599 if (ifa->ifa_dev->ifa_list == NULL) {
600 /* Last address was deleted from this interface.
603 fib_disable_ip(ifa->ifa_dev->dev, 1);
612 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
614 struct net_device *dev = ptr;
615 struct in_device *in_dev = __in_dev_get_rtnl(dev);
617 if (event == NETDEV_UNREGISTER) {
618 fib_disable_ip(dev, 2);
629 } endfor_ifa(in_dev);
630 #ifdef CONFIG_IP_ROUTE_MULTIPATH
636 fib_disable_ip(dev, 0);
638 case NETDEV_CHANGEMTU:
646 static struct notifier_block fib_inetaddr_notifier = {
647 .notifier_call =fib_inetaddr_event,
650 static struct notifier_block fib_netdev_notifier = {
651 .notifier_call =fib_netdev_event,
654 void __init ip_fib_init(void)
656 #ifndef CONFIG_IP_MULTIPLE_TABLES
657 ip_fib_local_table = fib_hash_init(RT_TABLE_LOCAL);
658 ip_fib_main_table = fib_hash_init(RT_TABLE_MAIN);
663 register_netdevice_notifier(&fib_netdev_notifier);
664 register_inetaddr_notifier(&fib_inetaddr_notifier);
665 nl_fib_lookup_init();
668 EXPORT_SYMBOL(inet_addr_type);
669 EXPORT_SYMBOL(ip_dev_find);