2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
18 * YOSHIFUJI Hideaki @USAGI
19 * reworked default router selection.
20 * - respect outgoing interface
21 * - select from (probably) reachable routers (i.e.
22 * routers in REACHABLE, STALE, DELAY or PROBE states).
23 * - always select the same router if it is (probably)
24 * reachable. otherwise, round-robin the list.
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/types.h>
30 #include <linux/times.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/route.h>
35 #include <linux/netdevice.h>
36 #include <linux/in6.h>
37 #include <linux/init.h>
38 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
58 #include <asm/uaccess.h>
61 #include <linux/sysctl.h>
64 /* Set to 3 to get tracing. */
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #define RT6_TRACE(x...) do { ; } while (0)
75 #define CLONE_OFFLINK_ROUTE 0
77 #define RT6_SELECT_F_IFACE 0x1
78 #define RT6_SELECT_F_REACHABLE 0x2
80 static int ip6_rt_max_size = 4096;
81 static int ip6_rt_gc_min_interval = HZ / 2;
82 static int ip6_rt_gc_timeout = 60*HZ;
83 int ip6_rt_gc_interval = 30*HZ;
84 static int ip6_rt_gc_elasticity = 9;
85 static int ip6_rt_mtu_expires = 10*60*HZ;
86 static int ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
88 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
89 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
90 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
91 static void ip6_dst_destroy(struct dst_entry *);
92 static void ip6_dst_ifdown(struct dst_entry *,
93 struct net_device *dev, int how);
94 static int ip6_dst_gc(void);
96 static int ip6_pkt_discard(struct sk_buff *skb);
97 static int ip6_pkt_discard_out(struct sk_buff *skb);
98 static void ip6_link_failure(struct sk_buff *skb);
99 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
101 #ifdef CONFIG_IPV6_ROUTE_INFO
102 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
103 struct in6_addr *gwaddr, int ifindex,
105 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
106 struct in6_addr *gwaddr, int ifindex);
109 static struct dst_ops ip6_dst_ops = {
111 .protocol = __constant_htons(ETH_P_IPV6),
114 .check = ip6_dst_check,
115 .destroy = ip6_dst_destroy,
116 .ifdown = ip6_dst_ifdown,
117 .negative_advice = ip6_negative_advice,
118 .link_failure = ip6_link_failure,
119 .update_pmtu = ip6_rt_update_pmtu,
120 .entry_size = sizeof(struct rt6_info),
123 struct rt6_info ip6_null_entry = {
126 .__refcnt = ATOMIC_INIT(1),
128 .dev = &loopback_dev,
130 .error = -ENETUNREACH,
131 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
132 .input = ip6_pkt_discard,
133 .output = ip6_pkt_discard_out,
135 .path = (struct dst_entry*)&ip6_null_entry,
138 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
139 .rt6i_metric = ~(u32) 0,
140 .rt6i_ref = ATOMIC_INIT(1),
143 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
145 struct rt6_info ip6_prohibit_entry = {
148 .__refcnt = ATOMIC_INIT(1),
150 .dev = &loopback_dev,
153 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
154 .input = ip6_pkt_discard,
155 .output = ip6_pkt_discard_out,
157 .path = (struct dst_entry*)&ip6_prohibit_entry,
160 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
161 .rt6i_metric = ~(u32) 0,
162 .rt6i_ref = ATOMIC_INIT(1),
165 struct rt6_info ip6_blk_hole_entry = {
168 .__refcnt = ATOMIC_INIT(1),
170 .dev = &loopback_dev,
173 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
174 .input = ip6_pkt_discard,
175 .output = ip6_pkt_discard_out,
177 .path = (struct dst_entry*)&ip6_blk_hole_entry,
180 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
181 .rt6i_metric = ~(u32) 0,
182 .rt6i_ref = ATOMIC_INIT(1),
187 /* allocate dst with ip6_dst_ops */
188 static __inline__ struct rt6_info *ip6_dst_alloc(void)
190 return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
193 static void ip6_dst_destroy(struct dst_entry *dst)
195 struct rt6_info *rt = (struct rt6_info *)dst;
196 struct inet6_dev *idev = rt->rt6i_idev;
199 rt->rt6i_idev = NULL;
204 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
207 struct rt6_info *rt = (struct rt6_info *)dst;
208 struct inet6_dev *idev = rt->rt6i_idev;
210 if (dev != &loopback_dev && idev != NULL && idev->dev == dev) {
211 struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
212 if (loopback_idev != NULL) {
213 rt->rt6i_idev = loopback_idev;
219 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
221 return (rt->rt6i_flags & RTF_EXPIRES &&
222 time_after(jiffies, rt->rt6i_expires));
225 static inline int rt6_need_strict(struct in6_addr *daddr)
227 return (ipv6_addr_type(daddr) &
228 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
232 * Route lookup. Any table->tb6_lock is implied.
235 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
239 struct rt6_info *local = NULL;
240 struct rt6_info *sprt;
243 for (sprt = rt; sprt; sprt = sprt->u.next) {
244 struct net_device *dev = sprt->rt6i_dev;
245 if (dev->ifindex == oif)
247 if (dev->flags & IFF_LOOPBACK) {
248 if (sprt->rt6i_idev == NULL ||
249 sprt->rt6i_idev->dev->ifindex != oif) {
252 if (local && (!oif ||
253 local->rt6i_idev->dev->ifindex == oif))
264 return &ip6_null_entry;
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 static void rt6_probe(struct rt6_info *rt)
272 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
274 * Okay, this does not seem to be appropriate
275 * for now, however, we need to check if it
276 * is really so; aka Router Reachability Probing.
278 * Router Reachability Probe MUST be rate-limited
279 * to no more than one per minute.
281 if (!neigh || (neigh->nud_state & NUD_VALID))
283 read_lock_bh(&neigh->lock);
284 if (!(neigh->nud_state & NUD_VALID) &&
285 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
286 struct in6_addr mcaddr;
287 struct in6_addr *target;
289 neigh->updated = jiffies;
290 read_unlock_bh(&neigh->lock);
292 target = (struct in6_addr *)&neigh->primary_key;
293 addrconf_addr_solict_mult(target, &mcaddr);
294 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
296 read_unlock_bh(&neigh->lock);
299 static inline void rt6_probe(struct rt6_info *rt)
306 * Default Router Selection (RFC 2461 6.3.6)
308 static int inline rt6_check_dev(struct rt6_info *rt, int oif)
310 struct net_device *dev = rt->rt6i_dev;
311 if (!oif || dev->ifindex == oif)
313 if ((dev->flags & IFF_LOOPBACK) &&
314 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
319 static int inline rt6_check_neigh(struct rt6_info *rt)
321 struct neighbour *neigh = rt->rt6i_nexthop;
323 if (rt->rt6i_flags & RTF_NONEXTHOP ||
324 !(rt->rt6i_flags & RTF_GATEWAY))
327 read_lock_bh(&neigh->lock);
328 if (neigh->nud_state & NUD_VALID)
330 read_unlock_bh(&neigh->lock);
335 static int rt6_score_route(struct rt6_info *rt, int oif,
340 m = rt6_check_dev(rt, oif);
341 if (!m && (strict & RT6_SELECT_F_IFACE))
343 #ifdef CONFIG_IPV6_ROUTER_PREF
344 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
346 n = rt6_check_neigh(rt);
349 else if (!n && strict & RT6_SELECT_F_REACHABLE)
354 static struct rt6_info *rt6_select(struct rt6_info **head, int oif,
357 struct rt6_info *match = NULL, *last = NULL;
358 struct rt6_info *rt, *rt0 = *head;
362 RT6_TRACE("%s(head=%p(*head=%p), oif=%d)\n",
363 __FUNCTION__, head, head ? *head : NULL, oif);
365 for (rt = rt0, metric = rt0->rt6i_metric;
366 rt && rt->rt6i_metric == metric && (!last || rt != rt0);
370 if (rt6_check_expired(rt))
375 m = rt6_score_route(rt, oif, strict);
389 (strict & RT6_SELECT_F_REACHABLE) &&
390 last && last != rt0) {
391 /* no entries matched; do round-robin */
392 static DEFINE_SPINLOCK(lock);
395 rt0->u.next = last->u.next;
400 RT6_TRACE("%s() => %p, score=%d\n",
401 __FUNCTION__, match, mpri);
403 return (match ? match : &ip6_null_entry);
406 #ifdef CONFIG_IPV6_ROUTE_INFO
407 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
408 struct in6_addr *gwaddr)
410 struct route_info *rinfo = (struct route_info *) opt;
411 struct in6_addr prefix_buf, *prefix;
416 if (len < sizeof(struct route_info)) {
420 /* Sanity check for prefix_len and length */
421 if (rinfo->length > 3) {
423 } else if (rinfo->prefix_len > 128) {
425 } else if (rinfo->prefix_len > 64) {
426 if (rinfo->length < 2) {
429 } else if (rinfo->prefix_len > 0) {
430 if (rinfo->length < 1) {
435 pref = rinfo->route_pref;
436 if (pref == ICMPV6_ROUTER_PREF_INVALID)
437 pref = ICMPV6_ROUTER_PREF_MEDIUM;
439 lifetime = htonl(rinfo->lifetime);
440 if (lifetime == 0xffffffff) {
442 } else if (lifetime > 0x7fffffff/HZ) {
443 /* Avoid arithmetic overflow */
444 lifetime = 0x7fffffff/HZ - 1;
447 if (rinfo->length == 3)
448 prefix = (struct in6_addr *)rinfo->prefix;
450 /* this function is safe */
451 ipv6_addr_prefix(&prefix_buf,
452 (struct in6_addr *)rinfo->prefix,
454 prefix = &prefix_buf;
457 rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
459 if (rt && !lifetime) {
465 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
468 rt->rt6i_flags = RTF_ROUTEINFO |
469 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
472 if (lifetime == 0xffffffff) {
473 rt->rt6i_flags &= ~RTF_EXPIRES;
475 rt->rt6i_expires = jiffies + HZ * lifetime;
476 rt->rt6i_flags |= RTF_EXPIRES;
478 dst_release(&rt->u.dst);
484 #define BACKTRACK() \
485 if (rt == &ip6_null_entry && flags & RT6_F_STRICT) { \
486 while ((fn = fn->parent) != NULL) { \
487 if (fn->fn_flags & RTN_TL_ROOT) { \
488 dst_hold(&rt->u.dst); \
491 if (fn->fn_flags & RTN_RTINFO) \
496 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
497 struct flowi *fl, int flags)
499 struct fib6_node *fn;
502 read_lock_bh(&table->tb6_lock);
503 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
506 rt = rt6_device_match(rt, fl->oif, flags & RT6_F_STRICT);
508 dst_hold(&rt->u.dst);
510 read_unlock_bh(&table->tb6_lock);
512 rt->u.dst.lastuse = jiffies;
519 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
531 struct dst_entry *dst;
532 int flags = strict ? RT6_F_STRICT : 0;
534 dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
536 return (struct rt6_info *) dst;
543 /* ip6_ins_rt is called with FREE table->tb6_lock.
544 It takes new route entry, the addition fails by any reason the
545 route is freed. In any case, if caller does not hold it, it may
549 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
552 struct fib6_table *table;
554 table = rt->rt6i_table;
555 write_lock_bh(&table->tb6_lock);
556 err = fib6_add(&table->tb6_root, rt, info);
557 write_unlock_bh(&table->tb6_lock);
562 int ip6_ins_rt(struct rt6_info *rt)
564 return __ip6_ins_rt(rt, NULL);
567 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
568 struct in6_addr *saddr)
576 rt = ip6_rt_copy(ort);
579 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
580 if (rt->rt6i_dst.plen != 128 &&
581 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
582 rt->rt6i_flags |= RTF_ANYCAST;
583 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
586 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
587 rt->rt6i_dst.plen = 128;
588 rt->rt6i_flags |= RTF_CACHE;
589 rt->u.dst.flags |= DST_HOST;
591 #ifdef CONFIG_IPV6_SUBTREES
592 if (rt->rt6i_src.plen && saddr) {
593 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
594 rt->rt6i_src.plen = 128;
598 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
605 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
607 struct rt6_info *rt = ip6_rt_copy(ort);
609 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
610 rt->rt6i_dst.plen = 128;
611 rt->rt6i_flags |= RTF_CACHE;
612 if (rt->rt6i_flags & RTF_REJECT)
613 rt->u.dst.error = ort->u.dst.error;
614 rt->u.dst.flags |= DST_HOST;
615 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
620 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
621 struct flowi *fl, int flags)
623 struct fib6_node *fn;
624 struct rt6_info *rt, *nrt;
628 int reachable = RT6_SELECT_F_REACHABLE;
630 if (flags & RT6_F_STRICT)
631 strict = RT6_SELECT_F_IFACE;
634 read_lock_bh(&table->tb6_lock);
637 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
640 rt = rt6_select(&fn->leaf, fl->iif, strict | reachable);
642 if (rt == &ip6_null_entry ||
643 rt->rt6i_flags & RTF_CACHE)
646 dst_hold(&rt->u.dst);
647 read_unlock_bh(&table->tb6_lock);
649 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
650 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
652 #if CLONE_OFFLINK_ROUTE
653 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
659 dst_release(&rt->u.dst);
660 rt = nrt ? : &ip6_null_entry;
662 dst_hold(&rt->u.dst);
664 err = ip6_ins_rt(nrt);
673 * Race condition! In the gap, when table->tb6_lock was
674 * released someone could insert this route. Relookup.
676 dst_release(&rt->u.dst);
684 dst_hold(&rt->u.dst);
685 read_unlock_bh(&table->tb6_lock);
687 rt->u.dst.lastuse = jiffies;
693 void ip6_route_input(struct sk_buff *skb)
695 struct ipv6hdr *iph = skb->nh.ipv6h;
697 .iif = skb->dev->ifindex,
702 .flowlabel = (* (u32 *) iph)&IPV6_FLOWINFO_MASK,
705 .proto = iph->nexthdr,
709 if (rt6_need_strict(&iph->daddr))
710 flags |= RT6_F_STRICT;
712 skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
715 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
716 struct flowi *fl, int flags)
718 struct fib6_node *fn;
719 struct rt6_info *rt, *nrt;
723 int reachable = RT6_SELECT_F_REACHABLE;
725 if (flags & RT6_F_STRICT)
726 strict = RT6_SELECT_F_IFACE;
729 read_lock_bh(&table->tb6_lock);
732 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
735 rt = rt6_select(&fn->leaf, fl->oif, strict | reachable);
737 if (rt == &ip6_null_entry ||
738 rt->rt6i_flags & RTF_CACHE)
741 dst_hold(&rt->u.dst);
742 read_unlock_bh(&table->tb6_lock);
744 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
745 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
747 #if CLONE_OFFLINK_ROUTE
748 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
754 dst_release(&rt->u.dst);
755 rt = nrt ? : &ip6_null_entry;
757 dst_hold(&rt->u.dst);
759 err = ip6_ins_rt(nrt);
768 * Race condition! In the gap, when table->tb6_lock was
769 * released someone could insert this route. Relookup.
771 dst_release(&rt->u.dst);
779 dst_hold(&rt->u.dst);
780 read_unlock_bh(&table->tb6_lock);
782 rt->u.dst.lastuse = jiffies;
787 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
791 if (rt6_need_strict(&fl->fl6_dst))
792 flags |= RT6_F_STRICT;
794 return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
799 * Destination cache support functions
802 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
806 rt = (struct rt6_info *) dst;
808 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
814 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
816 struct rt6_info *rt = (struct rt6_info *) dst;
819 if (rt->rt6i_flags & RTF_CACHE)
827 static void ip6_link_failure(struct sk_buff *skb)
831 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
833 rt = (struct rt6_info *) skb->dst;
835 if (rt->rt6i_flags&RTF_CACHE) {
836 dst_set_expires(&rt->u.dst, 0);
837 rt->rt6i_flags |= RTF_EXPIRES;
838 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
839 rt->rt6i_node->fn_sernum = -1;
843 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
845 struct rt6_info *rt6 = (struct rt6_info*)dst;
847 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
848 rt6->rt6i_flags |= RTF_MODIFIED;
849 if (mtu < IPV6_MIN_MTU) {
851 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
853 dst->metrics[RTAX_MTU-1] = mtu;
854 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
858 static int ipv6_get_mtu(struct net_device *dev);
860 static inline unsigned int ipv6_advmss(unsigned int mtu)
862 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
864 if (mtu < ip6_rt_min_advmss)
865 mtu = ip6_rt_min_advmss;
868 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
869 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
870 * IPV6_MAXPLEN is also valid and means: "any MSS,
871 * rely only on pmtu discovery"
873 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
878 static struct dst_entry *ndisc_dst_gc_list;
879 static DEFINE_SPINLOCK(ndisc_lock);
881 struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
882 struct neighbour *neigh,
883 struct in6_addr *addr,
884 int (*output)(struct sk_buff *))
887 struct inet6_dev *idev = in6_dev_get(dev);
889 if (unlikely(idev == NULL))
892 rt = ip6_dst_alloc();
893 if (unlikely(rt == NULL)) {
902 neigh = ndisc_get_neigh(dev, addr);
905 rt->rt6i_idev = idev;
906 rt->rt6i_nexthop = neigh;
907 atomic_set(&rt->u.dst.__refcnt, 1);
908 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
909 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
910 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
911 rt->u.dst.output = output;
913 #if 0 /* there's no chance to use these for ndisc */
914 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
917 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
918 rt->rt6i_dst.plen = 128;
921 spin_lock_bh(&ndisc_lock);
922 rt->u.dst.next = ndisc_dst_gc_list;
923 ndisc_dst_gc_list = &rt->u.dst;
924 spin_unlock_bh(&ndisc_lock);
926 fib6_force_start_gc();
929 return (struct dst_entry *)rt;
932 int ndisc_dst_gc(int *more)
934 struct dst_entry *dst, *next, **pprev;
940 spin_lock_bh(&ndisc_lock);
941 pprev = &ndisc_dst_gc_list;
943 while ((dst = *pprev) != NULL) {
944 if (!atomic_read(&dst->__refcnt)) {
954 spin_unlock_bh(&ndisc_lock);
959 static int ip6_dst_gc(void)
961 static unsigned expire = 30*HZ;
962 static unsigned long last_gc;
963 unsigned long now = jiffies;
965 if (time_after(last_gc + ip6_rt_gc_min_interval, now) &&
966 atomic_read(&ip6_dst_ops.entries) <= ip6_rt_max_size)
972 if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
973 expire = ip6_rt_gc_timeout>>1;
976 expire -= expire>>ip6_rt_gc_elasticity;
977 return (atomic_read(&ip6_dst_ops.entries) > ip6_rt_max_size);
980 /* Clean host part of a prefix. Not necessary in radix tree,
981 but results in cleaner routing tables.
983 Remove it only when all the things will work!
986 static int ipv6_get_mtu(struct net_device *dev)
988 int mtu = IPV6_MIN_MTU;
989 struct inet6_dev *idev;
991 idev = in6_dev_get(dev);
993 mtu = idev->cnf.mtu6;
999 int ipv6_get_hoplimit(struct net_device *dev)
1001 int hoplimit = ipv6_devconf.hop_limit;
1002 struct inet6_dev *idev;
1004 idev = in6_dev_get(dev);
1006 hoplimit = idev->cnf.hop_limit;
1016 int ip6_route_add(struct fib6_config *cfg)
1019 struct rt6_info *rt = NULL;
1020 struct net_device *dev = NULL;
1021 struct inet6_dev *idev = NULL;
1022 struct fib6_table *table;
1025 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1027 #ifndef CONFIG_IPV6_SUBTREES
1028 if (cfg->fc_src_len)
1031 if (cfg->fc_ifindex) {
1033 dev = dev_get_by_index(cfg->fc_ifindex);
1036 idev = in6_dev_get(dev);
1041 if (cfg->fc_metric == 0)
1042 cfg->fc_metric = IP6_RT_PRIO_USER;
1044 table = fib6_new_table(cfg->fc_table);
1045 if (table == NULL) {
1050 rt = ip6_dst_alloc();
1057 rt->u.dst.obsolete = -1;
1058 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1060 if (cfg->fc_protocol == RTPROT_UNSPEC)
1061 cfg->fc_protocol = RTPROT_BOOT;
1062 rt->rt6i_protocol = cfg->fc_protocol;
1064 addr_type = ipv6_addr_type(&cfg->fc_dst);
1066 if (addr_type & IPV6_ADDR_MULTICAST)
1067 rt->u.dst.input = ip6_mc_input;
1069 rt->u.dst.input = ip6_forward;
1071 rt->u.dst.output = ip6_output;
1073 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1074 rt->rt6i_dst.plen = cfg->fc_dst_len;
1075 if (rt->rt6i_dst.plen == 128)
1076 rt->u.dst.flags = DST_HOST;
1078 #ifdef CONFIG_IPV6_SUBTREES
1079 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1080 rt->rt6i_src.plen = cfg->fc_src_len;
1083 rt->rt6i_metric = cfg->fc_metric;
1085 /* We cannot add true routes via loopback here,
1086 they would result in kernel looping; promote them to reject routes
1088 if ((cfg->fc_flags & RTF_REJECT) ||
1089 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1090 /* hold loopback dev/idev if we haven't done so. */
1091 if (dev != &loopback_dev) {
1096 dev = &loopback_dev;
1098 idev = in6_dev_get(dev);
1104 rt->u.dst.output = ip6_pkt_discard_out;
1105 rt->u.dst.input = ip6_pkt_discard;
1106 rt->u.dst.error = -ENETUNREACH;
1107 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1111 if (cfg->fc_flags & RTF_GATEWAY) {
1112 struct in6_addr *gw_addr;
1115 gw_addr = &cfg->fc_gateway;
1116 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1117 gwa_type = ipv6_addr_type(gw_addr);
1119 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1120 struct rt6_info *grt;
1122 /* IPv6 strictly inhibits using not link-local
1123 addresses as nexthop address.
1124 Otherwise, router will not able to send redirects.
1125 It is very good, but in some (rare!) circumstances
1126 (SIT, PtP, NBMA NOARP links) it is handy to allow
1127 some exceptions. --ANK
1130 if (!(gwa_type&IPV6_ADDR_UNICAST))
1133 grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1135 err = -EHOSTUNREACH;
1139 if (dev != grt->rt6i_dev) {
1140 dst_release(&grt->u.dst);
1144 dev = grt->rt6i_dev;
1145 idev = grt->rt6i_idev;
1147 in6_dev_hold(grt->rt6i_idev);
1149 if (!(grt->rt6i_flags&RTF_GATEWAY))
1151 dst_release(&grt->u.dst);
1157 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1165 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1166 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1167 if (IS_ERR(rt->rt6i_nexthop)) {
1168 err = PTR_ERR(rt->rt6i_nexthop);
1169 rt->rt6i_nexthop = NULL;
1174 rt->rt6i_flags = cfg->fc_flags;
1181 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1182 int type = nla->nla_type;
1185 if (type > RTAX_MAX) {
1190 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1195 if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1196 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1197 if (!rt->u.dst.metrics[RTAX_MTU-1])
1198 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1199 if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1200 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1201 rt->u.dst.dev = dev;
1202 rt->rt6i_idev = idev;
1203 rt->rt6i_table = table;
1204 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1212 dst_free((struct dst_entry *) rt);
1216 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1219 struct fib6_table *table;
1221 if (rt == &ip6_null_entry)
1224 table = rt->rt6i_table;
1225 write_lock_bh(&table->tb6_lock);
1227 err = fib6_del(rt, info);
1228 dst_release(&rt->u.dst);
1230 write_unlock_bh(&table->tb6_lock);
1235 int ip6_del_rt(struct rt6_info *rt)
1237 return __ip6_del_rt(rt, NULL);
1240 static int ip6_route_del(struct fib6_config *cfg)
1242 struct fib6_table *table;
1243 struct fib6_node *fn;
1244 struct rt6_info *rt;
1247 table = fib6_get_table(cfg->fc_table);
1251 read_lock_bh(&table->tb6_lock);
1253 fn = fib6_locate(&table->tb6_root,
1254 &cfg->fc_dst, cfg->fc_dst_len,
1255 &cfg->fc_src, cfg->fc_src_len);
1258 for (rt = fn->leaf; rt; rt = rt->u.next) {
1259 if (cfg->fc_ifindex &&
1260 (rt->rt6i_dev == NULL ||
1261 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1263 if (cfg->fc_flags & RTF_GATEWAY &&
1264 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1266 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1268 dst_hold(&rt->u.dst);
1269 read_unlock_bh(&table->tb6_lock);
1271 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1274 read_unlock_bh(&table->tb6_lock);
1282 struct ip6rd_flowi {
1284 struct in6_addr gateway;
1287 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1291 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1292 struct rt6_info *rt;
1293 struct fib6_node *fn;
1296 * Get the "current" route for this destination and
1297 * check if the redirect has come from approriate router.
1299 * RFC 2461 specifies that redirects should only be
1300 * accepted if they come from the nexthop to the target.
1301 * Due to the way the routes are chosen, this notion
1302 * is a bit fuzzy and one might need to check all possible
1306 read_lock_bh(&table->tb6_lock);
1307 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1309 for (rt = fn->leaf; rt; rt = rt->u.next) {
1311 * Current route is on-link; redirect is always invalid.
1313 * Seems, previous statement is not true. It could
1314 * be node, which looks for us as on-link (f.e. proxy ndisc)
1315 * But then router serving it might decide, that we should
1316 * know truth 8)8) --ANK (980726).
1318 if (rt6_check_expired(rt))
1320 if (!(rt->rt6i_flags & RTF_GATEWAY))
1322 if (fl->oif != rt->rt6i_dev->ifindex)
1324 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1330 if (rt6_need_strict(&fl->fl6_dst)) {
1331 while ((fn = fn->parent) != NULL) {
1332 if (fn->fn_flags & RTN_ROOT)
1334 if (fn->fn_flags & RTN_RTINFO)
1338 rt = &ip6_null_entry;
1340 dst_hold(&rt->u.dst);
1342 read_unlock_bh(&table->tb6_lock);
1347 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1348 struct in6_addr *src,
1349 struct in6_addr *gateway,
1350 struct net_device *dev)
1352 struct ip6rd_flowi rdfl = {
1354 .oif = dev->ifindex,
1362 .gateway = *gateway,
1364 int flags = rt6_need_strict(dest) ? RT6_F_STRICT : 0;
1366 return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
1369 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1370 struct in6_addr *saddr,
1371 struct neighbour *neigh, u8 *lladdr, int on_link)
1373 struct rt6_info *rt, *nrt = NULL;
1374 struct netevent_redirect netevent;
1376 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1378 if (rt == &ip6_null_entry) {
1379 if (net_ratelimit())
1380 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1381 "for redirect target\n");
1386 * We have finally decided to accept it.
1389 neigh_update(neigh, lladdr, NUD_STALE,
1390 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1391 NEIGH_UPDATE_F_OVERRIDE|
1392 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1393 NEIGH_UPDATE_F_ISROUTER))
1397 * Redirect received -> path was valid.
1398 * Look, redirects are sent only in response to data packets,
1399 * so that this nexthop apparently is reachable. --ANK
1401 dst_confirm(&rt->u.dst);
1403 /* Duplicate redirect: silently ignore. */
1404 if (neigh == rt->u.dst.neighbour)
1407 nrt = ip6_rt_copy(rt);
1411 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1413 nrt->rt6i_flags &= ~RTF_GATEWAY;
1415 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1416 nrt->rt6i_dst.plen = 128;
1417 nrt->u.dst.flags |= DST_HOST;
1419 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1420 nrt->rt6i_nexthop = neigh_clone(neigh);
1421 /* Reset pmtu, it may be better */
1422 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1423 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1425 if (ip6_ins_rt(nrt))
1428 netevent.old = &rt->u.dst;
1429 netevent.new = &nrt->u.dst;
1430 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1432 if (rt->rt6i_flags&RTF_CACHE) {
1438 dst_release(&rt->u.dst);
1443 * Handle ICMP "packet too big" messages
1444 * i.e. Path MTU discovery
1447 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1448 struct net_device *dev, u32 pmtu)
1450 struct rt6_info *rt, *nrt;
1453 rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1457 if (pmtu >= dst_mtu(&rt->u.dst))
1460 if (pmtu < IPV6_MIN_MTU) {
1462 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1463 * MTU (1280) and a fragment header should always be included
1464 * after a node receiving Too Big message reporting PMTU is
1465 * less than the IPv6 Minimum Link MTU.
1467 pmtu = IPV6_MIN_MTU;
1471 /* New mtu received -> path was valid.
1472 They are sent only in response to data packets,
1473 so that this nexthop apparently is reachable. --ANK
1475 dst_confirm(&rt->u.dst);
1477 /* Host route. If it is static, it would be better
1478 not to override it, but add new one, so that
1479 when cache entry will expire old pmtu
1480 would return automatically.
1482 if (rt->rt6i_flags & RTF_CACHE) {
1483 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1485 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1486 dst_set_expires(&rt->u.dst, ip6_rt_mtu_expires);
1487 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1492 Two cases are possible:
1493 1. It is connected route. Action: COW
1494 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1496 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1497 nrt = rt6_alloc_cow(rt, daddr, saddr);
1499 nrt = rt6_alloc_clone(rt, daddr);
1502 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1504 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1506 /* According to RFC 1981, detecting PMTU increase shouldn't be
1507 * happened within 5 mins, the recommended timer is 10 mins.
1508 * Here this route expiration time is set to ip6_rt_mtu_expires
1509 * which is 10 mins. After 10 mins the decreased pmtu is expired
1510 * and detecting PMTU increase will be automatically happened.
1512 dst_set_expires(&nrt->u.dst, ip6_rt_mtu_expires);
1513 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1518 dst_release(&rt->u.dst);
1522 * Misc support functions
1525 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1527 struct rt6_info *rt = ip6_dst_alloc();
1530 rt->u.dst.input = ort->u.dst.input;
1531 rt->u.dst.output = ort->u.dst.output;
1533 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1534 rt->u.dst.dev = ort->u.dst.dev;
1536 dev_hold(rt->u.dst.dev);
1537 rt->rt6i_idev = ort->rt6i_idev;
1539 in6_dev_hold(rt->rt6i_idev);
1540 rt->u.dst.lastuse = jiffies;
1541 rt->rt6i_expires = 0;
1543 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1544 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1545 rt->rt6i_metric = 0;
1547 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1548 #ifdef CONFIG_IPV6_SUBTREES
1549 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1551 rt->rt6i_table = ort->rt6i_table;
1556 #ifdef CONFIG_IPV6_ROUTE_INFO
1557 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1558 struct in6_addr *gwaddr, int ifindex)
1560 struct fib6_node *fn;
1561 struct rt6_info *rt = NULL;
1562 struct fib6_table *table;
1564 table = fib6_get_table(RT6_TABLE_INFO);
1568 write_lock_bh(&table->tb6_lock);
1569 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1573 for (rt = fn->leaf; rt; rt = rt->u.next) {
1574 if (rt->rt6i_dev->ifindex != ifindex)
1576 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1578 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1580 dst_hold(&rt->u.dst);
1584 write_unlock_bh(&table->tb6_lock);
1588 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1589 struct in6_addr *gwaddr, int ifindex,
1592 struct fib6_config cfg = {
1593 .fc_table = RT6_TABLE_INFO,
1595 .fc_ifindex = ifindex,
1596 .fc_dst_len = prefixlen,
1597 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1598 RTF_UP | RTF_PREF(pref),
1601 ipv6_addr_copy(&cfg.fc_dst, prefix);
1602 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1604 /* We should treat it as a default route if prefix length is 0. */
1606 cfg.fc_flags |= RTF_DEFAULT;
1608 ip6_route_add(&cfg);
1610 return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1614 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1616 struct rt6_info *rt;
1617 struct fib6_table *table;
1619 table = fib6_get_table(RT6_TABLE_DFLT);
1623 write_lock_bh(&table->tb6_lock);
1624 for (rt = table->tb6_root.leaf; rt; rt=rt->u.next) {
1625 if (dev == rt->rt6i_dev &&
1626 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1627 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1631 dst_hold(&rt->u.dst);
1632 write_unlock_bh(&table->tb6_lock);
1636 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1637 struct net_device *dev,
1640 struct fib6_config cfg = {
1641 .fc_table = RT6_TABLE_DFLT,
1643 .fc_ifindex = dev->ifindex,
1644 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1645 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1648 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1650 ip6_route_add(&cfg);
1652 return rt6_get_dflt_router(gwaddr, dev);
1655 void rt6_purge_dflt_routers(void)
1657 struct rt6_info *rt;
1658 struct fib6_table *table;
1660 /* NOTE: Keep consistent with rt6_get_dflt_router */
1661 table = fib6_get_table(RT6_TABLE_DFLT);
1666 read_lock_bh(&table->tb6_lock);
1667 for (rt = table->tb6_root.leaf; rt; rt = rt->u.next) {
1668 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1669 dst_hold(&rt->u.dst);
1670 read_unlock_bh(&table->tb6_lock);
1675 read_unlock_bh(&table->tb6_lock);
1678 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1679 struct fib6_config *cfg)
1681 memset(cfg, 0, sizeof(*cfg));
1683 cfg->fc_table = RT6_TABLE_MAIN;
1684 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1685 cfg->fc_metric = rtmsg->rtmsg_metric;
1686 cfg->fc_expires = rtmsg->rtmsg_info;
1687 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1688 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1689 cfg->fc_flags = rtmsg->rtmsg_flags;
1691 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1692 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1693 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1696 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1698 struct fib6_config cfg;
1699 struct in6_rtmsg rtmsg;
1703 case SIOCADDRT: /* Add a route */
1704 case SIOCDELRT: /* Delete a route */
1705 if (!capable(CAP_NET_ADMIN))
1707 err = copy_from_user(&rtmsg, arg,
1708 sizeof(struct in6_rtmsg));
1712 rtmsg_to_fib6_config(&rtmsg, &cfg);
1717 err = ip6_route_add(&cfg);
1720 err = ip6_route_del(&cfg);
1734 * Drop the packet on the floor
1737 static int ip6_pkt_discard(struct sk_buff *skb)
1739 int type = ipv6_addr_type(&skb->nh.ipv6h->daddr);
1740 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED)
1741 IP6_INC_STATS(IPSTATS_MIB_INADDRERRORS);
1743 IP6_INC_STATS(IPSTATS_MIB_OUTNOROUTES);
1744 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_NOROUTE, 0, skb->dev);
1749 static int ip6_pkt_discard_out(struct sk_buff *skb)
1751 skb->dev = skb->dst->dev;
1752 return ip6_pkt_discard(skb);
1756 * Allocate a dst for local (unicast / anycast) address.
1759 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1760 const struct in6_addr *addr,
1763 struct rt6_info *rt = ip6_dst_alloc();
1766 return ERR_PTR(-ENOMEM);
1768 dev_hold(&loopback_dev);
1771 rt->u.dst.flags = DST_HOST;
1772 rt->u.dst.input = ip6_input;
1773 rt->u.dst.output = ip6_output;
1774 rt->rt6i_dev = &loopback_dev;
1775 rt->rt6i_idev = idev;
1776 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1777 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1778 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1779 rt->u.dst.obsolete = -1;
1781 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1783 rt->rt6i_flags |= RTF_ANYCAST;
1785 rt->rt6i_flags |= RTF_LOCAL;
1786 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1787 if (rt->rt6i_nexthop == NULL) {
1788 dst_free((struct dst_entry *) rt);
1789 return ERR_PTR(-ENOMEM);
1792 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1793 rt->rt6i_dst.plen = 128;
1794 rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1796 atomic_set(&rt->u.dst.__refcnt, 1);
1801 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1803 if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1804 rt != &ip6_null_entry) {
1805 RT6_TRACE("deleted by ifdown %p\n", rt);
1811 void rt6_ifdown(struct net_device *dev)
1813 fib6_clean_all(fib6_ifdown, 0, dev);
1816 struct rt6_mtu_change_arg
1818 struct net_device *dev;
1822 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1824 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1825 struct inet6_dev *idev;
1827 /* In IPv6 pmtu discovery is not optional,
1828 so that RTAX_MTU lock cannot disable it.
1829 We still use this lock to block changes
1830 caused by addrconf/ndisc.
1833 idev = __in6_dev_get(arg->dev);
1837 /* For administrative MTU increase, there is no way to discover
1838 IPv6 PMTU increase, so PMTU increase should be updated here.
1839 Since RFC 1981 doesn't include administrative MTU increase
1840 update PMTU increase is a MUST. (i.e. jumbo frame)
1843 If new MTU is less than route PMTU, this new MTU will be the
1844 lowest MTU in the path, update the route PMTU to reflect PMTU
1845 decreases; if new MTU is greater than route PMTU, and the
1846 old MTU is the lowest MTU in the path, update the route PMTU
1847 to reflect the increase. In this case if the other nodes' MTU
1848 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1851 if (rt->rt6i_dev == arg->dev &&
1852 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1853 (dst_mtu(&rt->u.dst) > arg->mtu ||
1854 (dst_mtu(&rt->u.dst) < arg->mtu &&
1855 dst_mtu(&rt->u.dst) == idev->cnf.mtu6)))
1856 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1857 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1861 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1863 struct rt6_mtu_change_arg arg = {
1868 fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1871 static struct nla_policy rtm_ipv6_policy[RTA_MAX+1] __read_mostly = {
1872 [RTA_GATEWAY] = { .minlen = sizeof(struct in6_addr) },
1873 [RTA_OIF] = { .type = NLA_U32 },
1874 [RTA_IIF] = { .type = NLA_U32 },
1875 [RTA_PRIORITY] = { .type = NLA_U32 },
1876 [RTA_METRICS] = { .type = NLA_NESTED },
1879 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1880 struct fib6_config *cfg)
1883 struct nlattr *tb[RTA_MAX+1];
1886 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1891 rtm = nlmsg_data(nlh);
1892 memset(cfg, 0, sizeof(*cfg));
1894 cfg->fc_table = rtm->rtm_table;
1895 cfg->fc_dst_len = rtm->rtm_dst_len;
1896 cfg->fc_src_len = rtm->rtm_src_len;
1897 cfg->fc_flags = RTF_UP;
1898 cfg->fc_protocol = rtm->rtm_protocol;
1900 if (rtm->rtm_type == RTN_UNREACHABLE)
1901 cfg->fc_flags |= RTF_REJECT;
1903 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1904 cfg->fc_nlinfo.nlh = nlh;
1906 if (tb[RTA_GATEWAY]) {
1907 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1908 cfg->fc_flags |= RTF_GATEWAY;
1912 int plen = (rtm->rtm_dst_len + 7) >> 3;
1914 if (nla_len(tb[RTA_DST]) < plen)
1917 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1921 int plen = (rtm->rtm_src_len + 7) >> 3;
1923 if (nla_len(tb[RTA_SRC]) < plen)
1926 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1930 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1932 if (tb[RTA_PRIORITY])
1933 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
1935 if (tb[RTA_METRICS]) {
1936 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
1937 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1941 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
1948 int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1950 struct fib6_config cfg;
1953 err = rtm_to_fib6_config(skb, nlh, &cfg);
1957 return ip6_route_del(&cfg);
1960 int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1962 struct fib6_config cfg;
1965 err = rtm_to_fib6_config(skb, nlh, &cfg);
1969 return ip6_route_add(&cfg);
1972 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
1973 struct in6_addr *dst, struct in6_addr *src,
1974 int iif, int type, u32 pid, u32 seq,
1975 int prefix, unsigned int flags)
1978 struct nlmsghdr *nlh;
1979 struct rta_cacheinfo ci;
1982 if (prefix) { /* user wants prefix routes only */
1983 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
1984 /* success since this is not a prefix route */
1989 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
1993 rtm = nlmsg_data(nlh);
1994 rtm->rtm_family = AF_INET6;
1995 rtm->rtm_dst_len = rt->rt6i_dst.plen;
1996 rtm->rtm_src_len = rt->rt6i_src.plen;
1999 table = rt->rt6i_table->tb6_id;
2001 table = RT6_TABLE_UNSPEC;
2002 rtm->rtm_table = table;
2003 NLA_PUT_U32(skb, RTA_TABLE, table);
2004 if (rt->rt6i_flags&RTF_REJECT)
2005 rtm->rtm_type = RTN_UNREACHABLE;
2006 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2007 rtm->rtm_type = RTN_LOCAL;
2009 rtm->rtm_type = RTN_UNICAST;
2011 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2012 rtm->rtm_protocol = rt->rt6i_protocol;
2013 if (rt->rt6i_flags&RTF_DYNAMIC)
2014 rtm->rtm_protocol = RTPROT_REDIRECT;
2015 else if (rt->rt6i_flags & RTF_ADDRCONF)
2016 rtm->rtm_protocol = RTPROT_KERNEL;
2017 else if (rt->rt6i_flags&RTF_DEFAULT)
2018 rtm->rtm_protocol = RTPROT_RA;
2020 if (rt->rt6i_flags&RTF_CACHE)
2021 rtm->rtm_flags |= RTM_F_CLONED;
2024 NLA_PUT(skb, RTA_DST, 16, dst);
2025 rtm->rtm_dst_len = 128;
2026 } else if (rtm->rtm_dst_len)
2027 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2028 #ifdef CONFIG_IPV6_SUBTREES
2030 NLA_PUT(skb, RTA_SRC, 16, src);
2031 rtm->rtm_src_len = 128;
2032 } else if (rtm->rtm_src_len)
2033 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2036 NLA_PUT_U32(skb, RTA_IIF, iif);
2038 struct in6_addr saddr_buf;
2039 if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
2040 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2043 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2044 goto nla_put_failure;
2046 if (rt->u.dst.neighbour)
2047 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2050 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2052 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2053 ci.rta_lastuse = jiffies_to_clock_t(jiffies - rt->u.dst.lastuse);
2054 if (rt->rt6i_expires)
2055 ci.rta_expires = jiffies_to_clock_t(rt->rt6i_expires - jiffies);
2058 ci.rta_used = rt->u.dst.__use;
2059 ci.rta_clntref = atomic_read(&rt->u.dst.__refcnt);
2060 ci.rta_error = rt->u.dst.error;
2064 NLA_PUT(skb, RTA_CACHEINFO, sizeof(ci), &ci);
2066 return nlmsg_end(skb, nlh);
2069 return nlmsg_cancel(skb, nlh);
2072 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2074 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2077 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2078 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2079 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2083 return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2084 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2085 prefix, NLM_F_MULTI);
2088 int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2090 struct nlattr *tb[RTA_MAX+1];
2091 struct rt6_info *rt;
2092 struct sk_buff *skb;
2097 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2102 memset(&fl, 0, sizeof(fl));
2105 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2108 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2112 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2115 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2119 iif = nla_get_u32(tb[RTA_IIF]);
2122 fl.oif = nla_get_u32(tb[RTA_OIF]);
2125 struct net_device *dev;
2126 dev = __dev_get_by_index(iif);
2133 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2139 /* Reserve room for dummy headers, this skb can pass
2140 through good chunk of routing engine.
2142 skb->mac.raw = skb->data;
2143 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2145 rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2146 skb->dst = &rt->u.dst;
2148 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2149 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2150 nlh->nlmsg_seq, 0, 0);
2156 err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
2161 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2163 struct sk_buff *skb;
2164 u32 pid = 0, seq = 0;
2165 struct nlmsghdr *nlh = NULL;
2166 int payload = sizeof(struct rtmsg) + 256;
2173 seq = nlh->nlmsg_seq;
2176 skb = nlmsg_new(nlmsg_total_size(payload), gfp_any());
2180 err = rt6_fill_node(skb, rt, NULL, NULL, 0, event, pid, seq, 0, 0);
2186 err = rtnl_notify(skb, pid, RTNLGRP_IPV6_ROUTE, nlh, gfp_any());
2189 rtnl_set_sk_err(RTNLGRP_IPV6_ROUTE, err);
2196 #ifdef CONFIG_PROC_FS
2198 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2209 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2211 struct rt6_proc_arg *arg = (struct rt6_proc_arg *) p_arg;
2214 if (arg->skip < arg->offset / RT6_INFO_LEN) {
2219 if (arg->len >= arg->length)
2222 for (i=0; i<16; i++) {
2223 sprintf(arg->buffer + arg->len, "%02x",
2224 rt->rt6i_dst.addr.s6_addr[i]);
2227 arg->len += sprintf(arg->buffer + arg->len, " %02x ",
2230 #ifdef CONFIG_IPV6_SUBTREES
2231 for (i=0; i<16; i++) {
2232 sprintf(arg->buffer + arg->len, "%02x",
2233 rt->rt6i_src.addr.s6_addr[i]);
2236 arg->len += sprintf(arg->buffer + arg->len, " %02x ",
2239 sprintf(arg->buffer + arg->len,
2240 "00000000000000000000000000000000 00 ");
2244 if (rt->rt6i_nexthop) {
2245 for (i=0; i<16; i++) {
2246 sprintf(arg->buffer + arg->len, "%02x",
2247 rt->rt6i_nexthop->primary_key[i]);
2251 sprintf(arg->buffer + arg->len,
2252 "00000000000000000000000000000000");
2255 arg->len += sprintf(arg->buffer + arg->len,
2256 " %08x %08x %08x %08x %8s\n",
2257 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2258 rt->u.dst.__use, rt->rt6i_flags,
2259 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2263 static int rt6_proc_info(char *buffer, char **start, off_t offset, int length)
2265 struct rt6_proc_arg arg = {
2271 fib6_clean_all(rt6_info_route, 0, &arg);
2275 *start += offset % RT6_INFO_LEN;
2277 arg.len -= offset % RT6_INFO_LEN;
2279 if (arg.len > length)
2287 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2289 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2290 rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2291 rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2292 rt6_stats.fib_rt_cache,
2293 atomic_read(&ip6_dst_ops.entries),
2294 rt6_stats.fib_discarded_routes);
2299 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2301 return single_open(file, rt6_stats_seq_show, NULL);
2304 static struct file_operations rt6_stats_seq_fops = {
2305 .owner = THIS_MODULE,
2306 .open = rt6_stats_seq_open,
2308 .llseek = seq_lseek,
2309 .release = single_release,
2311 #endif /* CONFIG_PROC_FS */
2313 #ifdef CONFIG_SYSCTL
2315 static int flush_delay;
2318 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2319 void __user *buffer, size_t *lenp, loff_t *ppos)
2322 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2323 fib6_run_gc(flush_delay <= 0 ? ~0UL : (unsigned long)flush_delay);
2329 ctl_table ipv6_route_table[] = {
2331 .ctl_name = NET_IPV6_ROUTE_FLUSH,
2332 .procname = "flush",
2333 .data = &flush_delay,
2334 .maxlen = sizeof(int),
2336 .proc_handler = &ipv6_sysctl_rtcache_flush
2339 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2340 .procname = "gc_thresh",
2341 .data = &ip6_dst_ops.gc_thresh,
2342 .maxlen = sizeof(int),
2344 .proc_handler = &proc_dointvec,
2347 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2348 .procname = "max_size",
2349 .data = &ip6_rt_max_size,
2350 .maxlen = sizeof(int),
2352 .proc_handler = &proc_dointvec,
2355 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2356 .procname = "gc_min_interval",
2357 .data = &ip6_rt_gc_min_interval,
2358 .maxlen = sizeof(int),
2360 .proc_handler = &proc_dointvec_jiffies,
2361 .strategy = &sysctl_jiffies,
2364 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2365 .procname = "gc_timeout",
2366 .data = &ip6_rt_gc_timeout,
2367 .maxlen = sizeof(int),
2369 .proc_handler = &proc_dointvec_jiffies,
2370 .strategy = &sysctl_jiffies,
2373 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2374 .procname = "gc_interval",
2375 .data = &ip6_rt_gc_interval,
2376 .maxlen = sizeof(int),
2378 .proc_handler = &proc_dointvec_jiffies,
2379 .strategy = &sysctl_jiffies,
2382 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2383 .procname = "gc_elasticity",
2384 .data = &ip6_rt_gc_elasticity,
2385 .maxlen = sizeof(int),
2387 .proc_handler = &proc_dointvec_jiffies,
2388 .strategy = &sysctl_jiffies,
2391 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2392 .procname = "mtu_expires",
2393 .data = &ip6_rt_mtu_expires,
2394 .maxlen = sizeof(int),
2396 .proc_handler = &proc_dointvec_jiffies,
2397 .strategy = &sysctl_jiffies,
2400 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2401 .procname = "min_adv_mss",
2402 .data = &ip6_rt_min_advmss,
2403 .maxlen = sizeof(int),
2405 .proc_handler = &proc_dointvec_jiffies,
2406 .strategy = &sysctl_jiffies,
2409 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2410 .procname = "gc_min_interval_ms",
2411 .data = &ip6_rt_gc_min_interval,
2412 .maxlen = sizeof(int),
2414 .proc_handler = &proc_dointvec_ms_jiffies,
2415 .strategy = &sysctl_ms_jiffies,
2422 void __init ip6_route_init(void)
2424 struct proc_dir_entry *p;
2426 ip6_dst_ops.kmem_cachep = kmem_cache_create("ip6_dst_cache",
2427 sizeof(struct rt6_info),
2428 0, SLAB_HWCACHE_ALIGN,
2430 if (!ip6_dst_ops.kmem_cachep)
2431 panic("cannot create ip6_dst_cache");
2434 #ifdef CONFIG_PROC_FS
2435 p = proc_net_create("ipv6_route", 0, rt6_proc_info);
2437 p->owner = THIS_MODULE;
2439 proc_net_fops_create("rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2444 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2449 void ip6_route_cleanup(void)
2451 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2452 fib6_rules_cleanup();
2454 #ifdef CONFIG_PROC_FS
2455 proc_net_remove("ipv6_route");
2456 proc_net_remove("rt6_stats");
2463 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);