2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/export.h>
30 #include <linux/types.h>
31 #include <linux/times.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/route.h>
36 #include <linux/netdevice.h>
37 #include <linux/in6.h>
38 #include <linux/mroute6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/nsproxy.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
53 #include <linux/rtnetlink.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
59 #include <asm/uaccess.h>
62 #include <linux/sysctl.h>
65 /* Set to 3 to get tracing. */
69 #define RDBG(x) printk x
70 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
73 #define RT6_TRACE(x...) do { ; } while (0)
76 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
77 const struct in6_addr *dest);
78 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
79 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
80 static unsigned int ip6_mtu(const struct dst_entry *dst);
81 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
82 static void ip6_dst_destroy(struct dst_entry *);
83 static void ip6_dst_ifdown(struct dst_entry *,
84 struct net_device *dev, int how);
85 static int ip6_dst_gc(struct dst_ops *ops);
87 static int ip6_pkt_discard(struct sk_buff *skb);
88 static int ip6_pkt_discard_out(struct sk_buff *skb);
89 static void ip6_link_failure(struct sk_buff *skb);
90 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
92 #ifdef CONFIG_IPV6_ROUTE_INFO
93 static struct rt6_info *rt6_add_route_info(struct net *net,
94 const struct in6_addr *prefix, int prefixlen,
95 const struct in6_addr *gwaddr, int ifindex,
97 static struct rt6_info *rt6_get_route_info(struct net *net,
98 const struct in6_addr *prefix, int prefixlen,
99 const struct in6_addr *gwaddr, int ifindex);
102 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
104 struct rt6_info *rt = (struct rt6_info *) dst;
105 struct inet_peer *peer;
108 if (!(rt->dst.flags & DST_HOST))
112 rt6_bind_peer(rt, 1);
114 peer = rt->rt6i_peer;
116 u32 *old_p = __DST_METRICS_PTR(old);
117 unsigned long prev, new;
120 if (inet_metrics_new(peer))
121 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
123 new = (unsigned long) p;
124 prev = cmpxchg(&dst->_metrics, old, new);
127 p = __DST_METRICS_PTR(prev);
128 if (prev & DST_METRICS_READ_ONLY)
135 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst, const void *daddr)
137 return __neigh_lookup_errno(&nd_tbl, daddr, dst->dev);
140 static struct dst_ops ip6_dst_ops_template = {
142 .protocol = cpu_to_be16(ETH_P_IPV6),
145 .check = ip6_dst_check,
146 .default_advmss = ip6_default_advmss,
148 .cow_metrics = ipv6_cow_metrics,
149 .destroy = ip6_dst_destroy,
150 .ifdown = ip6_dst_ifdown,
151 .negative_advice = ip6_negative_advice,
152 .link_failure = ip6_link_failure,
153 .update_pmtu = ip6_rt_update_pmtu,
154 .local_out = __ip6_local_out,
155 .neigh_lookup = ip6_neigh_lookup,
158 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
160 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
162 return mtu ? : dst->dev->mtu;
165 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
169 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
175 static struct dst_ops ip6_dst_blackhole_ops = {
177 .protocol = cpu_to_be16(ETH_P_IPV6),
178 .destroy = ip6_dst_destroy,
179 .check = ip6_dst_check,
180 .mtu = ip6_blackhole_mtu,
181 .default_advmss = ip6_default_advmss,
182 .update_pmtu = ip6_rt_blackhole_update_pmtu,
183 .cow_metrics = ip6_rt_blackhole_cow_metrics,
184 .neigh_lookup = ip6_neigh_lookup,
187 static const u32 ip6_template_metrics[RTAX_MAX] = {
188 [RTAX_HOPLIMIT - 1] = 255,
191 static struct rt6_info ip6_null_entry_template = {
193 .__refcnt = ATOMIC_INIT(1),
196 .error = -ENETUNREACH,
197 .input = ip6_pkt_discard,
198 .output = ip6_pkt_discard_out,
200 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
201 .rt6i_protocol = RTPROT_KERNEL,
202 .rt6i_metric = ~(u32) 0,
203 .rt6i_ref = ATOMIC_INIT(1),
206 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
208 static int ip6_pkt_prohibit(struct sk_buff *skb);
209 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
211 static struct rt6_info ip6_prohibit_entry_template = {
213 .__refcnt = ATOMIC_INIT(1),
217 .input = ip6_pkt_prohibit,
218 .output = ip6_pkt_prohibit_out,
220 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
221 .rt6i_protocol = RTPROT_KERNEL,
222 .rt6i_metric = ~(u32) 0,
223 .rt6i_ref = ATOMIC_INIT(1),
226 static struct rt6_info ip6_blk_hole_entry_template = {
228 .__refcnt = ATOMIC_INIT(1),
232 .input = dst_discard,
233 .output = dst_discard,
235 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
236 .rt6i_protocol = RTPROT_KERNEL,
237 .rt6i_metric = ~(u32) 0,
238 .rt6i_ref = ATOMIC_INIT(1),
243 /* allocate dst with ip6_dst_ops */
244 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops,
245 struct net_device *dev,
248 struct rt6_info *rt = dst_alloc(ops, dev, 0, 0, flags);
251 memset(&rt->rt6i_table, 0,
252 sizeof(*rt) - sizeof(struct dst_entry));
257 static void ip6_dst_destroy(struct dst_entry *dst)
259 struct rt6_info *rt = (struct rt6_info *)dst;
260 struct inet6_dev *idev = rt->rt6i_idev;
261 struct inet_peer *peer = rt->rt6i_peer;
263 if (!(rt->dst.flags & DST_HOST))
264 dst_destroy_metrics_generic(dst);
267 rt->rt6i_idev = NULL;
271 rt->rt6i_peer = NULL;
276 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
278 static u32 rt6_peer_genid(void)
280 return atomic_read(&__rt6_peer_genid);
283 void rt6_bind_peer(struct rt6_info *rt, int create)
285 struct inet_peer *peer;
287 peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
288 if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
291 rt->rt6i_peer_genid = rt6_peer_genid();
294 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
297 struct rt6_info *rt = (struct rt6_info *)dst;
298 struct inet6_dev *idev = rt->rt6i_idev;
299 struct net_device *loopback_dev =
300 dev_net(dev)->loopback_dev;
302 if (dev != loopback_dev && idev && idev->dev == dev) {
303 struct inet6_dev *loopback_idev =
304 in6_dev_get(loopback_dev);
306 rt->rt6i_idev = loopback_idev;
312 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
314 return (rt->rt6i_flags & RTF_EXPIRES) &&
315 time_after(jiffies, rt->rt6i_expires);
318 static inline int rt6_need_strict(const struct in6_addr *daddr)
320 return ipv6_addr_type(daddr) &
321 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
325 * Route lookup. Any table->tb6_lock is implied.
328 static inline struct rt6_info *rt6_device_match(struct net *net,
330 const struct in6_addr *saddr,
334 struct rt6_info *local = NULL;
335 struct rt6_info *sprt;
337 if (!oif && ipv6_addr_any(saddr))
340 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
341 struct net_device *dev = sprt->rt6i_dev;
344 if (dev->ifindex == oif)
346 if (dev->flags & IFF_LOOPBACK) {
347 if (!sprt->rt6i_idev ||
348 sprt->rt6i_idev->dev->ifindex != oif) {
349 if (flags & RT6_LOOKUP_F_IFACE && oif)
351 if (local && (!oif ||
352 local->rt6i_idev->dev->ifindex == oif))
358 if (ipv6_chk_addr(net, saddr, dev,
359 flags & RT6_LOOKUP_F_IFACE))
368 if (flags & RT6_LOOKUP_F_IFACE)
369 return net->ipv6.ip6_null_entry;
375 #ifdef CONFIG_IPV6_ROUTER_PREF
376 static void rt6_probe(struct rt6_info *rt)
378 struct neighbour *neigh;
380 * Okay, this does not seem to be appropriate
381 * for now, however, we need to check if it
382 * is really so; aka Router Reachability Probing.
384 * Router Reachability Probe MUST be rate-limited
385 * to no more than one per minute.
388 neigh = rt ? dst_get_neighbour(&rt->dst) : NULL;
389 if (!neigh || (neigh->nud_state & NUD_VALID))
391 read_lock_bh(&neigh->lock);
392 if (!(neigh->nud_state & NUD_VALID) &&
393 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
394 struct in6_addr mcaddr;
395 struct in6_addr *target;
397 neigh->updated = jiffies;
398 read_unlock_bh(&neigh->lock);
400 target = (struct in6_addr *)&neigh->primary_key;
401 addrconf_addr_solict_mult(target, &mcaddr);
402 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
404 read_unlock_bh(&neigh->lock);
410 static inline void rt6_probe(struct rt6_info *rt)
416 * Default Router Selection (RFC 2461 6.3.6)
418 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
420 struct net_device *dev = rt->rt6i_dev;
421 if (!oif || dev->ifindex == oif)
423 if ((dev->flags & IFF_LOOPBACK) &&
424 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
429 static inline int rt6_check_neigh(struct rt6_info *rt)
431 struct neighbour *neigh;
435 neigh = dst_get_neighbour(&rt->dst);
436 if (rt->rt6i_flags & RTF_NONEXTHOP ||
437 !(rt->rt6i_flags & RTF_GATEWAY))
440 read_lock_bh(&neigh->lock);
441 if (neigh->nud_state & NUD_VALID)
443 #ifdef CONFIG_IPV6_ROUTER_PREF
444 else if (neigh->nud_state & NUD_FAILED)
449 read_unlock_bh(&neigh->lock);
456 static int rt6_score_route(struct rt6_info *rt, int oif,
461 m = rt6_check_dev(rt, oif);
462 if (!m && (strict & RT6_LOOKUP_F_IFACE))
464 #ifdef CONFIG_IPV6_ROUTER_PREF
465 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
467 n = rt6_check_neigh(rt);
468 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
473 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
474 int *mpri, struct rt6_info *match)
478 if (rt6_check_expired(rt))
481 m = rt6_score_route(rt, oif, strict);
486 if (strict & RT6_LOOKUP_F_REACHABLE)
490 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
498 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
499 struct rt6_info *rr_head,
500 u32 metric, int oif, int strict)
502 struct rt6_info *rt, *match;
506 for (rt = rr_head; rt && rt->rt6i_metric == metric;
507 rt = rt->dst.rt6_next)
508 match = find_match(rt, oif, strict, &mpri, match);
509 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
510 rt = rt->dst.rt6_next)
511 match = find_match(rt, oif, strict, &mpri, match);
516 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
518 struct rt6_info *match, *rt0;
521 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
522 __func__, fn->leaf, oif);
526 fn->rr_ptr = rt0 = fn->leaf;
528 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
531 (strict & RT6_LOOKUP_F_REACHABLE)) {
532 struct rt6_info *next = rt0->dst.rt6_next;
534 /* no entries matched; do round-robin */
535 if (!next || next->rt6i_metric != rt0->rt6i_metric)
542 RT6_TRACE("%s() => %p\n",
545 net = dev_net(rt0->rt6i_dev);
546 return match ? match : net->ipv6.ip6_null_entry;
549 #ifdef CONFIG_IPV6_ROUTE_INFO
550 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
551 const struct in6_addr *gwaddr)
553 struct net *net = dev_net(dev);
554 struct route_info *rinfo = (struct route_info *) opt;
555 struct in6_addr prefix_buf, *prefix;
557 unsigned long lifetime;
560 if (len < sizeof(struct route_info)) {
564 /* Sanity check for prefix_len and length */
565 if (rinfo->length > 3) {
567 } else if (rinfo->prefix_len > 128) {
569 } else if (rinfo->prefix_len > 64) {
570 if (rinfo->length < 2) {
573 } else if (rinfo->prefix_len > 0) {
574 if (rinfo->length < 1) {
579 pref = rinfo->route_pref;
580 if (pref == ICMPV6_ROUTER_PREF_INVALID)
583 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
585 if (rinfo->length == 3)
586 prefix = (struct in6_addr *)rinfo->prefix;
588 /* this function is safe */
589 ipv6_addr_prefix(&prefix_buf,
590 (struct in6_addr *)rinfo->prefix,
592 prefix = &prefix_buf;
595 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
598 if (rt && !lifetime) {
604 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
607 rt->rt6i_flags = RTF_ROUTEINFO |
608 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
611 if (!addrconf_finite_timeout(lifetime)) {
612 rt->rt6i_flags &= ~RTF_EXPIRES;
614 rt->rt6i_expires = jiffies + HZ * lifetime;
615 rt->rt6i_flags |= RTF_EXPIRES;
617 dst_release(&rt->dst);
623 #define BACKTRACK(__net, saddr) \
625 if (rt == __net->ipv6.ip6_null_entry) { \
626 struct fib6_node *pn; \
628 if (fn->fn_flags & RTN_TL_ROOT) \
631 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
632 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
635 if (fn->fn_flags & RTN_RTINFO) \
641 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
642 struct fib6_table *table,
643 struct flowi6 *fl6, int flags)
645 struct fib6_node *fn;
648 read_lock_bh(&table->tb6_lock);
649 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
652 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
653 BACKTRACK(net, &fl6->saddr);
655 dst_use(&rt->dst, jiffies);
656 read_unlock_bh(&table->tb6_lock);
661 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
664 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
666 EXPORT_SYMBOL_GPL(ip6_route_lookup);
668 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
669 const struct in6_addr *saddr, int oif, int strict)
671 struct flowi6 fl6 = {
675 struct dst_entry *dst;
676 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
679 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
680 flags |= RT6_LOOKUP_F_HAS_SADDR;
683 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
685 return (struct rt6_info *) dst;
692 EXPORT_SYMBOL(rt6_lookup);
694 /* ip6_ins_rt is called with FREE table->tb6_lock.
695 It takes new route entry, the addition fails by any reason the
696 route is freed. In any case, if caller does not hold it, it may
700 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
703 struct fib6_table *table;
705 table = rt->rt6i_table;
706 write_lock_bh(&table->tb6_lock);
707 err = fib6_add(&table->tb6_root, rt, info);
708 write_unlock_bh(&table->tb6_lock);
713 int ip6_ins_rt(struct rt6_info *rt)
715 struct nl_info info = {
716 .nl_net = dev_net(rt->rt6i_dev),
718 return __ip6_ins_rt(rt, &info);
721 static struct rt6_info *rt6_alloc_cow(const struct rt6_info *ort,
722 const struct in6_addr *daddr,
723 const struct in6_addr *saddr)
731 rt = ip6_rt_copy(ort, daddr);
734 struct neighbour *neigh;
735 int attempts = !in_softirq();
737 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
738 if (rt->rt6i_dst.plen != 128 &&
739 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
740 rt->rt6i_flags |= RTF_ANYCAST;
741 rt->rt6i_gateway = *daddr;
744 rt->rt6i_flags |= RTF_CACHE;
746 #ifdef CONFIG_IPV6_SUBTREES
747 if (rt->rt6i_src.plen && saddr) {
748 rt->rt6i_src.addr = *saddr;
749 rt->rt6i_src.plen = 128;
754 neigh = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway,
757 struct net *net = dev_net(rt->rt6i_dev);
758 int saved_rt_min_interval =
759 net->ipv6.sysctl.ip6_rt_gc_min_interval;
760 int saved_rt_elasticity =
761 net->ipv6.sysctl.ip6_rt_gc_elasticity;
763 if (attempts-- > 0) {
764 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
765 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
767 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
769 net->ipv6.sysctl.ip6_rt_gc_elasticity =
771 net->ipv6.sysctl.ip6_rt_gc_min_interval =
772 saved_rt_min_interval;
778 "ipv6: Neighbour table overflow.\n");
782 dst_set_neighbour(&rt->dst, neigh);
789 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
790 const struct in6_addr *daddr)
792 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
795 rt->rt6i_flags |= RTF_CACHE;
796 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_raw(&ort->dst)));
801 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
802 struct flowi6 *fl6, int flags)
804 struct fib6_node *fn;
805 struct rt6_info *rt, *nrt;
809 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
811 strict |= flags & RT6_LOOKUP_F_IFACE;
814 read_lock_bh(&table->tb6_lock);
817 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
820 rt = rt6_select(fn, oif, strict | reachable);
822 BACKTRACK(net, &fl6->saddr);
823 if (rt == net->ipv6.ip6_null_entry ||
824 rt->rt6i_flags & RTF_CACHE)
828 read_unlock_bh(&table->tb6_lock);
830 if (!dst_get_neighbour_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
831 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
832 else if (!(rt->dst.flags & DST_HOST))
833 nrt = rt6_alloc_clone(rt, &fl6->daddr);
837 dst_release(&rt->dst);
838 rt = nrt ? : net->ipv6.ip6_null_entry;
842 err = ip6_ins_rt(nrt);
851 * Race condition! In the gap, when table->tb6_lock was
852 * released someone could insert this route. Relookup.
854 dst_release(&rt->dst);
863 read_unlock_bh(&table->tb6_lock);
865 rt->dst.lastuse = jiffies;
871 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
872 struct flowi6 *fl6, int flags)
874 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
877 void ip6_route_input(struct sk_buff *skb)
879 const struct ipv6hdr *iph = ipv6_hdr(skb);
880 struct net *net = dev_net(skb->dev);
881 int flags = RT6_LOOKUP_F_HAS_SADDR;
882 struct flowi6 fl6 = {
883 .flowi6_iif = skb->dev->ifindex,
886 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
887 .flowi6_mark = skb->mark,
888 .flowi6_proto = iph->nexthdr,
891 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
892 flags |= RT6_LOOKUP_F_IFACE;
894 skb_dst_set(skb, fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_input));
897 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
898 struct flowi6 *fl6, int flags)
900 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
903 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
908 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
909 flags |= RT6_LOOKUP_F_IFACE;
911 if (!ipv6_addr_any(&fl6->saddr))
912 flags |= RT6_LOOKUP_F_HAS_SADDR;
914 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
916 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
919 EXPORT_SYMBOL(ip6_route_output);
921 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
923 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
924 struct dst_entry *new = NULL;
926 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
928 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
933 new->input = dst_discard;
934 new->output = dst_discard;
936 if (dst_metrics_read_only(&ort->dst))
937 new->_metrics = ort->dst._metrics;
939 dst_copy_metrics(new, &ort->dst);
940 rt->rt6i_idev = ort->rt6i_idev;
942 in6_dev_hold(rt->rt6i_idev);
943 rt->rt6i_expires = 0;
945 rt->rt6i_gateway = ort->rt6i_gateway;
946 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
949 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
950 #ifdef CONFIG_IPV6_SUBTREES
951 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
957 dst_release(dst_orig);
958 return new ? new : ERR_PTR(-ENOMEM);
962 * Destination cache support functions
965 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
969 rt = (struct rt6_info *) dst;
971 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
972 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
974 rt6_bind_peer(rt, 0);
975 rt->rt6i_peer_genid = rt6_peer_genid();
982 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
984 struct rt6_info *rt = (struct rt6_info *) dst;
987 if (rt->rt6i_flags & RTF_CACHE) {
988 if (rt6_check_expired(rt)) {
1000 static void ip6_link_failure(struct sk_buff *skb)
1002 struct rt6_info *rt;
1004 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1006 rt = (struct rt6_info *) skb_dst(skb);
1008 if (rt->rt6i_flags & RTF_CACHE) {
1009 dst_set_expires(&rt->dst, 0);
1010 rt->rt6i_flags |= RTF_EXPIRES;
1011 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1012 rt->rt6i_node->fn_sernum = -1;
1016 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1018 struct rt6_info *rt6 = (struct rt6_info*)dst;
1020 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1021 rt6->rt6i_flags |= RTF_MODIFIED;
1022 if (mtu < IPV6_MIN_MTU) {
1023 u32 features = dst_metric(dst, RTAX_FEATURES);
1025 features |= RTAX_FEATURE_ALLFRAG;
1026 dst_metric_set(dst, RTAX_FEATURES, features);
1028 dst_metric_set(dst, RTAX_MTU, mtu);
1032 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1034 struct net_device *dev = dst->dev;
1035 unsigned int mtu = dst_mtu(dst);
1036 struct net *net = dev_net(dev);
1038 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1040 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1041 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1044 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1045 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1046 * IPV6_MAXPLEN is also valid and means: "any MSS,
1047 * rely only on pmtu discovery"
1049 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1054 static unsigned int ip6_mtu(const struct dst_entry *dst)
1056 struct inet6_dev *idev;
1057 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1065 idev = __in6_dev_get(dst->dev);
1067 mtu = idev->cnf.mtu6;
1073 static struct dst_entry *icmp6_dst_gc_list;
1074 static DEFINE_SPINLOCK(icmp6_dst_lock);
1076 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1077 struct neighbour *neigh,
1078 const struct in6_addr *addr)
1080 struct rt6_info *rt;
1081 struct inet6_dev *idev = in6_dev_get(dev);
1082 struct net *net = dev_net(dev);
1084 if (unlikely(!idev))
1087 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1088 if (unlikely(!rt)) {
1096 neigh = __neigh_lookup_errno(&nd_tbl, addr, dev);
1101 rt->dst.flags |= DST_HOST;
1102 rt->dst.output = ip6_output;
1103 dst_set_neighbour(&rt->dst, neigh);
1104 atomic_set(&rt->dst.__refcnt, 1);
1105 rt->rt6i_dst.addr = *addr;
1106 rt->rt6i_dst.plen = 128;
1107 rt->rt6i_idev = idev;
1108 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1110 spin_lock_bh(&icmp6_dst_lock);
1111 rt->dst.next = icmp6_dst_gc_list;
1112 icmp6_dst_gc_list = &rt->dst;
1113 spin_unlock_bh(&icmp6_dst_lock);
1115 fib6_force_start_gc(net);
1121 int icmp6_dst_gc(void)
1123 struct dst_entry *dst, **pprev;
1126 spin_lock_bh(&icmp6_dst_lock);
1127 pprev = &icmp6_dst_gc_list;
1129 while ((dst = *pprev) != NULL) {
1130 if (!atomic_read(&dst->__refcnt)) {
1139 spin_unlock_bh(&icmp6_dst_lock);
1144 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1147 struct dst_entry *dst, **pprev;
1149 spin_lock_bh(&icmp6_dst_lock);
1150 pprev = &icmp6_dst_gc_list;
1151 while ((dst = *pprev) != NULL) {
1152 struct rt6_info *rt = (struct rt6_info *) dst;
1153 if (func(rt, arg)) {
1160 spin_unlock_bh(&icmp6_dst_lock);
1163 static int ip6_dst_gc(struct dst_ops *ops)
1165 unsigned long now = jiffies;
1166 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1167 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1168 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1169 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1170 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1171 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1174 entries = dst_entries_get_fast(ops);
1175 if (time_after(rt_last_gc + rt_min_interval, now) &&
1176 entries <= rt_max_size)
1179 net->ipv6.ip6_rt_gc_expire++;
1180 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1181 net->ipv6.ip6_rt_last_gc = now;
1182 entries = dst_entries_get_slow(ops);
1183 if (entries < ops->gc_thresh)
1184 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1186 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1187 return entries > rt_max_size;
1190 /* Clean host part of a prefix. Not necessary in radix tree,
1191 but results in cleaner routing tables.
1193 Remove it only when all the things will work!
1196 int ip6_dst_hoplimit(struct dst_entry *dst)
1198 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1199 if (hoplimit == 0) {
1200 struct net_device *dev = dst->dev;
1201 struct inet6_dev *idev;
1204 idev = __in6_dev_get(dev);
1206 hoplimit = idev->cnf.hop_limit;
1208 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1213 EXPORT_SYMBOL(ip6_dst_hoplimit);
1219 int ip6_route_add(struct fib6_config *cfg)
1222 struct net *net = cfg->fc_nlinfo.nl_net;
1223 struct rt6_info *rt = NULL;
1224 struct net_device *dev = NULL;
1225 struct inet6_dev *idev = NULL;
1226 struct fib6_table *table;
1229 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1231 #ifndef CONFIG_IPV6_SUBTREES
1232 if (cfg->fc_src_len)
1235 if (cfg->fc_ifindex) {
1237 dev = dev_get_by_index(net, cfg->fc_ifindex);
1240 idev = in6_dev_get(dev);
1245 if (cfg->fc_metric == 0)
1246 cfg->fc_metric = IP6_RT_PRIO_USER;
1249 if (cfg->fc_nlinfo.nlh &&
1250 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1251 table = fib6_get_table(net, cfg->fc_table);
1253 printk(KERN_WARNING "IPv6: NLM_F_CREATE should be specified when creating new route\n");
1254 table = fib6_new_table(net, cfg->fc_table);
1257 table = fib6_new_table(net, cfg->fc_table);
1263 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1270 rt->dst.obsolete = -1;
1271 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1272 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1275 if (cfg->fc_protocol == RTPROT_UNSPEC)
1276 cfg->fc_protocol = RTPROT_BOOT;
1277 rt->rt6i_protocol = cfg->fc_protocol;
1279 addr_type = ipv6_addr_type(&cfg->fc_dst);
1281 if (addr_type & IPV6_ADDR_MULTICAST)
1282 rt->dst.input = ip6_mc_input;
1283 else if (cfg->fc_flags & RTF_LOCAL)
1284 rt->dst.input = ip6_input;
1286 rt->dst.input = ip6_forward;
1288 rt->dst.output = ip6_output;
1290 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1291 rt->rt6i_dst.plen = cfg->fc_dst_len;
1292 if (rt->rt6i_dst.plen == 128)
1293 rt->dst.flags |= DST_HOST;
1295 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1296 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1301 dst_init_metrics(&rt->dst, metrics, 0);
1303 #ifdef CONFIG_IPV6_SUBTREES
1304 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1305 rt->rt6i_src.plen = cfg->fc_src_len;
1308 rt->rt6i_metric = cfg->fc_metric;
1310 /* We cannot add true routes via loopback here,
1311 they would result in kernel looping; promote them to reject routes
1313 if ((cfg->fc_flags & RTF_REJECT) ||
1314 (dev && (dev->flags & IFF_LOOPBACK) &&
1315 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1316 !(cfg->fc_flags & RTF_LOCAL))) {
1317 /* hold loopback dev/idev if we haven't done so. */
1318 if (dev != net->loopback_dev) {
1323 dev = net->loopback_dev;
1325 idev = in6_dev_get(dev);
1331 rt->dst.output = ip6_pkt_discard_out;
1332 rt->dst.input = ip6_pkt_discard;
1333 rt->dst.error = -ENETUNREACH;
1334 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1338 if (cfg->fc_flags & RTF_GATEWAY) {
1339 const struct in6_addr *gw_addr;
1342 gw_addr = &cfg->fc_gateway;
1343 rt->rt6i_gateway = *gw_addr;
1344 gwa_type = ipv6_addr_type(gw_addr);
1346 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1347 struct rt6_info *grt;
1349 /* IPv6 strictly inhibits using not link-local
1350 addresses as nexthop address.
1351 Otherwise, router will not able to send redirects.
1352 It is very good, but in some (rare!) circumstances
1353 (SIT, PtP, NBMA NOARP links) it is handy to allow
1354 some exceptions. --ANK
1357 if (!(gwa_type & IPV6_ADDR_UNICAST))
1360 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1362 err = -EHOSTUNREACH;
1366 if (dev != grt->rt6i_dev) {
1367 dst_release(&grt->dst);
1371 dev = grt->rt6i_dev;
1372 idev = grt->rt6i_idev;
1374 in6_dev_hold(grt->rt6i_idev);
1376 if (!(grt->rt6i_flags & RTF_GATEWAY))
1378 dst_release(&grt->dst);
1384 if (!dev || (dev->flags & IFF_LOOPBACK))
1392 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1393 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1397 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1398 rt->rt6i_prefsrc.plen = 128;
1400 rt->rt6i_prefsrc.plen = 0;
1402 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1403 struct neighbour *n = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1408 dst_set_neighbour(&rt->dst, n);
1411 rt->rt6i_flags = cfg->fc_flags;
1418 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1419 int type = nla_type(nla);
1422 if (type > RTAX_MAX) {
1427 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1433 rt->rt6i_idev = idev;
1434 rt->rt6i_table = table;
1436 cfg->fc_nlinfo.nl_net = dev_net(dev);
1438 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1450 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1453 struct fib6_table *table;
1454 struct net *net = dev_net(rt->rt6i_dev);
1456 if (rt == net->ipv6.ip6_null_entry)
1459 table = rt->rt6i_table;
1460 write_lock_bh(&table->tb6_lock);
1462 err = fib6_del(rt, info);
1463 dst_release(&rt->dst);
1465 write_unlock_bh(&table->tb6_lock);
1470 int ip6_del_rt(struct rt6_info *rt)
1472 struct nl_info info = {
1473 .nl_net = dev_net(rt->rt6i_dev),
1475 return __ip6_del_rt(rt, &info);
1478 static int ip6_route_del(struct fib6_config *cfg)
1480 struct fib6_table *table;
1481 struct fib6_node *fn;
1482 struct rt6_info *rt;
1485 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1489 read_lock_bh(&table->tb6_lock);
1491 fn = fib6_locate(&table->tb6_root,
1492 &cfg->fc_dst, cfg->fc_dst_len,
1493 &cfg->fc_src, cfg->fc_src_len);
1496 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1497 if (cfg->fc_ifindex &&
1499 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1501 if (cfg->fc_flags & RTF_GATEWAY &&
1502 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1504 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1507 read_unlock_bh(&table->tb6_lock);
1509 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1512 read_unlock_bh(&table->tb6_lock);
1520 struct ip6rd_flowi {
1522 struct in6_addr gateway;
1525 static struct rt6_info *__ip6_route_redirect(struct net *net,
1526 struct fib6_table *table,
1530 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1531 struct rt6_info *rt;
1532 struct fib6_node *fn;
1535 * Get the "current" route for this destination and
1536 * check if the redirect has come from approriate router.
1538 * RFC 2461 specifies that redirects should only be
1539 * accepted if they come from the nexthop to the target.
1540 * Due to the way the routes are chosen, this notion
1541 * is a bit fuzzy and one might need to check all possible
1545 read_lock_bh(&table->tb6_lock);
1546 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1548 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1550 * Current route is on-link; redirect is always invalid.
1552 * Seems, previous statement is not true. It could
1553 * be node, which looks for us as on-link (f.e. proxy ndisc)
1554 * But then router serving it might decide, that we should
1555 * know truth 8)8) --ANK (980726).
1557 if (rt6_check_expired(rt))
1559 if (!(rt->rt6i_flags & RTF_GATEWAY))
1561 if (fl6->flowi6_oif != rt->rt6i_dev->ifindex)
1563 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1569 rt = net->ipv6.ip6_null_entry;
1570 BACKTRACK(net, &fl6->saddr);
1574 read_unlock_bh(&table->tb6_lock);
1579 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1580 const struct in6_addr *src,
1581 const struct in6_addr *gateway,
1582 struct net_device *dev)
1584 int flags = RT6_LOOKUP_F_HAS_SADDR;
1585 struct net *net = dev_net(dev);
1586 struct ip6rd_flowi rdfl = {
1588 .flowi6_oif = dev->ifindex,
1594 rdfl.gateway = *gateway;
1596 if (rt6_need_strict(dest))
1597 flags |= RT6_LOOKUP_F_IFACE;
1599 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1600 flags, __ip6_route_redirect);
1603 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1604 const struct in6_addr *saddr,
1605 struct neighbour *neigh, u8 *lladdr, int on_link)
1607 struct rt6_info *rt, *nrt = NULL;
1608 struct netevent_redirect netevent;
1609 struct net *net = dev_net(neigh->dev);
1611 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1613 if (rt == net->ipv6.ip6_null_entry) {
1614 if (net_ratelimit())
1615 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1616 "for redirect target\n");
1621 * We have finally decided to accept it.
1624 neigh_update(neigh, lladdr, NUD_STALE,
1625 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1626 NEIGH_UPDATE_F_OVERRIDE|
1627 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1628 NEIGH_UPDATE_F_ISROUTER))
1632 * Redirect received -> path was valid.
1633 * Look, redirects are sent only in response to data packets,
1634 * so that this nexthop apparently is reachable. --ANK
1636 dst_confirm(&rt->dst);
1638 /* Duplicate redirect: silently ignore. */
1639 if (neigh == dst_get_neighbour_raw(&rt->dst))
1642 nrt = ip6_rt_copy(rt, dest);
1646 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1648 nrt->rt6i_flags &= ~RTF_GATEWAY;
1650 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1651 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1653 if (ip6_ins_rt(nrt))
1656 netevent.old = &rt->dst;
1657 netevent.new = &nrt->dst;
1658 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1660 if (rt->rt6i_flags & RTF_CACHE) {
1666 dst_release(&rt->dst);
1670 * Handle ICMP "packet too big" messages
1671 * i.e. Path MTU discovery
1674 static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1675 struct net *net, u32 pmtu, int ifindex)
1677 struct rt6_info *rt, *nrt;
1680 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1684 if (rt6_check_expired(rt)) {
1689 if (pmtu >= dst_mtu(&rt->dst))
1692 if (pmtu < IPV6_MIN_MTU) {
1694 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1695 * MTU (1280) and a fragment header should always be included
1696 * after a node receiving Too Big message reporting PMTU is
1697 * less than the IPv6 Minimum Link MTU.
1699 pmtu = IPV6_MIN_MTU;
1703 /* New mtu received -> path was valid.
1704 They are sent only in response to data packets,
1705 so that this nexthop apparently is reachable. --ANK
1707 dst_confirm(&rt->dst);
1709 /* Host route. If it is static, it would be better
1710 not to override it, but add new one, so that
1711 when cache entry will expire old pmtu
1712 would return automatically.
1714 if (rt->rt6i_flags & RTF_CACHE) {
1715 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1717 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1718 features |= RTAX_FEATURE_ALLFRAG;
1719 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1721 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1722 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1727 Two cases are possible:
1728 1. It is connected route. Action: COW
1729 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1731 if (!dst_get_neighbour_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1732 nrt = rt6_alloc_cow(rt, daddr, saddr);
1734 nrt = rt6_alloc_clone(rt, daddr);
1737 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1739 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1740 features |= RTAX_FEATURE_ALLFRAG;
1741 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1744 /* According to RFC 1981, detecting PMTU increase shouldn't be
1745 * happened within 5 mins, the recommended timer is 10 mins.
1746 * Here this route expiration time is set to ip6_rt_mtu_expires
1747 * which is 10 mins. After 10 mins the decreased pmtu is expired
1748 * and detecting PMTU increase will be automatically happened.
1750 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1751 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1756 dst_release(&rt->dst);
1759 void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1760 struct net_device *dev, u32 pmtu)
1762 struct net *net = dev_net(dev);
1765 * RFC 1981 states that a node "MUST reduce the size of the packets it
1766 * is sending along the path" that caused the Packet Too Big message.
1767 * Since it's not possible in the general case to determine which
1768 * interface was used to send the original packet, we update the MTU
1769 * on the interface that will be used to send future packets. We also
1770 * update the MTU on the interface that received the Packet Too Big in
1771 * case the original packet was forced out that interface with
1772 * SO_BINDTODEVICE or similar. This is the next best thing to the
1773 * correct behaviour, which would be to update the MTU on all
1776 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1777 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1781 * Misc support functions
1784 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
1785 const struct in6_addr *dest)
1787 struct net *net = dev_net(ort->rt6i_dev);
1788 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1792 rt->dst.input = ort->dst.input;
1793 rt->dst.output = ort->dst.output;
1794 rt->dst.flags |= DST_HOST;
1796 rt->rt6i_dst.addr = *dest;
1797 rt->rt6i_dst.plen = 128;
1798 dst_copy_metrics(&rt->dst, &ort->dst);
1799 rt->dst.error = ort->dst.error;
1800 rt->rt6i_idev = ort->rt6i_idev;
1802 in6_dev_hold(rt->rt6i_idev);
1803 rt->dst.lastuse = jiffies;
1804 rt->rt6i_expires = 0;
1806 rt->rt6i_gateway = ort->rt6i_gateway;
1807 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1808 rt->rt6i_metric = 0;
1810 #ifdef CONFIG_IPV6_SUBTREES
1811 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1813 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1814 rt->rt6i_table = ort->rt6i_table;
1819 #ifdef CONFIG_IPV6_ROUTE_INFO
1820 static struct rt6_info *rt6_get_route_info(struct net *net,
1821 const struct in6_addr *prefix, int prefixlen,
1822 const struct in6_addr *gwaddr, int ifindex)
1824 struct fib6_node *fn;
1825 struct rt6_info *rt = NULL;
1826 struct fib6_table *table;
1828 table = fib6_get_table(net, RT6_TABLE_INFO);
1832 write_lock_bh(&table->tb6_lock);
1833 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1837 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1838 if (rt->rt6i_dev->ifindex != ifindex)
1840 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1842 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1848 write_unlock_bh(&table->tb6_lock);
1852 static struct rt6_info *rt6_add_route_info(struct net *net,
1853 const struct in6_addr *prefix, int prefixlen,
1854 const struct in6_addr *gwaddr, int ifindex,
1857 struct fib6_config cfg = {
1858 .fc_table = RT6_TABLE_INFO,
1859 .fc_metric = IP6_RT_PRIO_USER,
1860 .fc_ifindex = ifindex,
1861 .fc_dst_len = prefixlen,
1862 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1863 RTF_UP | RTF_PREF(pref),
1865 .fc_nlinfo.nlh = NULL,
1866 .fc_nlinfo.nl_net = net,
1869 cfg.fc_dst = *prefix;
1870 cfg.fc_gateway = *gwaddr;
1872 /* We should treat it as a default route if prefix length is 0. */
1874 cfg.fc_flags |= RTF_DEFAULT;
1876 ip6_route_add(&cfg);
1878 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1882 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1884 struct rt6_info *rt;
1885 struct fib6_table *table;
1887 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1891 write_lock_bh(&table->tb6_lock);
1892 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1893 if (dev == rt->rt6i_dev &&
1894 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1895 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1900 write_unlock_bh(&table->tb6_lock);
1904 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1905 struct net_device *dev,
1908 struct fib6_config cfg = {
1909 .fc_table = RT6_TABLE_DFLT,
1910 .fc_metric = IP6_RT_PRIO_USER,
1911 .fc_ifindex = dev->ifindex,
1912 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1913 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1915 .fc_nlinfo.nlh = NULL,
1916 .fc_nlinfo.nl_net = dev_net(dev),
1919 cfg.fc_gateway = *gwaddr;
1921 ip6_route_add(&cfg);
1923 return rt6_get_dflt_router(gwaddr, dev);
1926 void rt6_purge_dflt_routers(struct net *net)
1928 struct rt6_info *rt;
1929 struct fib6_table *table;
1931 /* NOTE: Keep consistent with rt6_get_dflt_router */
1932 table = fib6_get_table(net, RT6_TABLE_DFLT);
1937 read_lock_bh(&table->tb6_lock);
1938 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1939 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1941 read_unlock_bh(&table->tb6_lock);
1946 read_unlock_bh(&table->tb6_lock);
1949 static void rtmsg_to_fib6_config(struct net *net,
1950 struct in6_rtmsg *rtmsg,
1951 struct fib6_config *cfg)
1953 memset(cfg, 0, sizeof(*cfg));
1955 cfg->fc_table = RT6_TABLE_MAIN;
1956 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1957 cfg->fc_metric = rtmsg->rtmsg_metric;
1958 cfg->fc_expires = rtmsg->rtmsg_info;
1959 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1960 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1961 cfg->fc_flags = rtmsg->rtmsg_flags;
1963 cfg->fc_nlinfo.nl_net = net;
1965 cfg->fc_dst = rtmsg->rtmsg_dst;
1966 cfg->fc_src = rtmsg->rtmsg_src;
1967 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1970 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1972 struct fib6_config cfg;
1973 struct in6_rtmsg rtmsg;
1977 case SIOCADDRT: /* Add a route */
1978 case SIOCDELRT: /* Delete a route */
1979 if (!capable(CAP_NET_ADMIN))
1981 err = copy_from_user(&rtmsg, arg,
1982 sizeof(struct in6_rtmsg));
1986 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1991 err = ip6_route_add(&cfg);
1994 err = ip6_route_del(&cfg);
2008 * Drop the packet on the floor
2011 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2014 struct dst_entry *dst = skb_dst(skb);
2015 switch (ipstats_mib_noroutes) {
2016 case IPSTATS_MIB_INNOROUTES:
2017 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2018 if (type == IPV6_ADDR_ANY) {
2019 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2020 IPSTATS_MIB_INADDRERRORS);
2024 case IPSTATS_MIB_OUTNOROUTES:
2025 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2026 ipstats_mib_noroutes);
2029 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2034 static int ip6_pkt_discard(struct sk_buff *skb)
2036 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2039 static int ip6_pkt_discard_out(struct sk_buff *skb)
2041 skb->dev = skb_dst(skb)->dev;
2042 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2045 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2047 static int ip6_pkt_prohibit(struct sk_buff *skb)
2049 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2052 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2054 skb->dev = skb_dst(skb)->dev;
2055 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2061 * Allocate a dst for local (unicast / anycast) address.
2064 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2065 const struct in6_addr *addr,
2068 struct net *net = dev_net(idev->dev);
2069 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2070 net->loopback_dev, 0);
2071 struct neighbour *neigh;
2074 if (net_ratelimit())
2075 pr_warning("IPv6: Maximum number of routes reached,"
2076 " consider increasing route/max_size.\n");
2077 return ERR_PTR(-ENOMEM);
2082 rt->dst.flags |= DST_HOST;
2083 rt->dst.input = ip6_input;
2084 rt->dst.output = ip6_output;
2085 rt->rt6i_idev = idev;
2086 rt->dst.obsolete = -1;
2088 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2090 rt->rt6i_flags |= RTF_ANYCAST;
2092 rt->rt6i_flags |= RTF_LOCAL;
2093 neigh = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, rt->rt6i_dev);
2094 if (IS_ERR(neigh)) {
2097 return ERR_CAST(neigh);
2099 dst_set_neighbour(&rt->dst, neigh);
2101 rt->rt6i_dst.addr = *addr;
2102 rt->rt6i_dst.plen = 128;
2103 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2105 atomic_set(&rt->dst.__refcnt, 1);
2110 int ip6_route_get_saddr(struct net *net,
2111 struct rt6_info *rt,
2112 const struct in6_addr *daddr,
2114 struct in6_addr *saddr)
2116 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2118 if (rt->rt6i_prefsrc.plen)
2119 *saddr = rt->rt6i_prefsrc.addr;
2121 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2122 daddr, prefs, saddr);
2126 /* remove deleted ip from prefsrc entries */
2127 struct arg_dev_net_ip {
2128 struct net_device *dev;
2130 struct in6_addr *addr;
2133 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2135 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2136 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2137 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2139 if (((void *)rt->rt6i_dev == dev || !dev) &&
2140 rt != net->ipv6.ip6_null_entry &&
2141 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2142 /* remove prefsrc entry */
2143 rt->rt6i_prefsrc.plen = 0;
2148 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2150 struct net *net = dev_net(ifp->idev->dev);
2151 struct arg_dev_net_ip adni = {
2152 .dev = ifp->idev->dev,
2156 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2159 struct arg_dev_net {
2160 struct net_device *dev;
2164 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2166 const struct arg_dev_net *adn = arg;
2167 const struct net_device *dev = adn->dev;
2169 if ((rt->rt6i_dev == dev || !dev) &&
2170 rt != adn->net->ipv6.ip6_null_entry) {
2171 RT6_TRACE("deleted by ifdown %p\n", rt);
2177 void rt6_ifdown(struct net *net, struct net_device *dev)
2179 struct arg_dev_net adn = {
2184 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2185 icmp6_clean_all(fib6_ifdown, &adn);
2188 struct rt6_mtu_change_arg
2190 struct net_device *dev;
2194 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2196 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2197 struct inet6_dev *idev;
2199 /* In IPv6 pmtu discovery is not optional,
2200 so that RTAX_MTU lock cannot disable it.
2201 We still use this lock to block changes
2202 caused by addrconf/ndisc.
2205 idev = __in6_dev_get(arg->dev);
2209 /* For administrative MTU increase, there is no way to discover
2210 IPv6 PMTU increase, so PMTU increase should be updated here.
2211 Since RFC 1981 doesn't include administrative MTU increase
2212 update PMTU increase is a MUST. (i.e. jumbo frame)
2215 If new MTU is less than route PMTU, this new MTU will be the
2216 lowest MTU in the path, update the route PMTU to reflect PMTU
2217 decreases; if new MTU is greater than route PMTU, and the
2218 old MTU is the lowest MTU in the path, update the route PMTU
2219 to reflect the increase. In this case if the other nodes' MTU
2220 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2223 if (rt->rt6i_dev == arg->dev &&
2224 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2225 (dst_mtu(&rt->dst) >= arg->mtu ||
2226 (dst_mtu(&rt->dst) < arg->mtu &&
2227 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2228 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2233 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2235 struct rt6_mtu_change_arg arg = {
2240 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2243 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2244 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2245 [RTA_OIF] = { .type = NLA_U32 },
2246 [RTA_IIF] = { .type = NLA_U32 },
2247 [RTA_PRIORITY] = { .type = NLA_U32 },
2248 [RTA_METRICS] = { .type = NLA_NESTED },
2251 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2252 struct fib6_config *cfg)
2255 struct nlattr *tb[RTA_MAX+1];
2258 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2263 rtm = nlmsg_data(nlh);
2264 memset(cfg, 0, sizeof(*cfg));
2266 cfg->fc_table = rtm->rtm_table;
2267 cfg->fc_dst_len = rtm->rtm_dst_len;
2268 cfg->fc_src_len = rtm->rtm_src_len;
2269 cfg->fc_flags = RTF_UP;
2270 cfg->fc_protocol = rtm->rtm_protocol;
2272 if (rtm->rtm_type == RTN_UNREACHABLE)
2273 cfg->fc_flags |= RTF_REJECT;
2275 if (rtm->rtm_type == RTN_LOCAL)
2276 cfg->fc_flags |= RTF_LOCAL;
2278 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2279 cfg->fc_nlinfo.nlh = nlh;
2280 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2282 if (tb[RTA_GATEWAY]) {
2283 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2284 cfg->fc_flags |= RTF_GATEWAY;
2288 int plen = (rtm->rtm_dst_len + 7) >> 3;
2290 if (nla_len(tb[RTA_DST]) < plen)
2293 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2297 int plen = (rtm->rtm_src_len + 7) >> 3;
2299 if (nla_len(tb[RTA_SRC]) < plen)
2302 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2305 if (tb[RTA_PREFSRC])
2306 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2309 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2311 if (tb[RTA_PRIORITY])
2312 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2314 if (tb[RTA_METRICS]) {
2315 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2316 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2320 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2327 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2329 struct fib6_config cfg;
2332 err = rtm_to_fib6_config(skb, nlh, &cfg);
2336 return ip6_route_del(&cfg);
2339 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2341 struct fib6_config cfg;
2344 err = rtm_to_fib6_config(skb, nlh, &cfg);
2348 return ip6_route_add(&cfg);
2351 static inline size_t rt6_nlmsg_size(void)
2353 return NLMSG_ALIGN(sizeof(struct rtmsg))
2354 + nla_total_size(16) /* RTA_SRC */
2355 + nla_total_size(16) /* RTA_DST */
2356 + nla_total_size(16) /* RTA_GATEWAY */
2357 + nla_total_size(16) /* RTA_PREFSRC */
2358 + nla_total_size(4) /* RTA_TABLE */
2359 + nla_total_size(4) /* RTA_IIF */
2360 + nla_total_size(4) /* RTA_OIF */
2361 + nla_total_size(4) /* RTA_PRIORITY */
2362 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2363 + nla_total_size(sizeof(struct rta_cacheinfo));
2366 static int rt6_fill_node(struct net *net,
2367 struct sk_buff *skb, struct rt6_info *rt,
2368 struct in6_addr *dst, struct in6_addr *src,
2369 int iif, int type, u32 pid, u32 seq,
2370 int prefix, int nowait, unsigned int flags)
2373 struct nlmsghdr *nlh;
2376 struct neighbour *n;
2378 if (prefix) { /* user wants prefix routes only */
2379 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2380 /* success since this is not a prefix route */
2385 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2389 rtm = nlmsg_data(nlh);
2390 rtm->rtm_family = AF_INET6;
2391 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2392 rtm->rtm_src_len = rt->rt6i_src.plen;
2395 table = rt->rt6i_table->tb6_id;
2397 table = RT6_TABLE_UNSPEC;
2398 rtm->rtm_table = table;
2399 NLA_PUT_U32(skb, RTA_TABLE, table);
2400 if (rt->rt6i_flags & RTF_REJECT)
2401 rtm->rtm_type = RTN_UNREACHABLE;
2402 else if (rt->rt6i_flags & RTF_LOCAL)
2403 rtm->rtm_type = RTN_LOCAL;
2404 else if (rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
2405 rtm->rtm_type = RTN_LOCAL;
2407 rtm->rtm_type = RTN_UNICAST;
2409 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2410 rtm->rtm_protocol = rt->rt6i_protocol;
2411 if (rt->rt6i_flags & RTF_DYNAMIC)
2412 rtm->rtm_protocol = RTPROT_REDIRECT;
2413 else if (rt->rt6i_flags & RTF_ADDRCONF)
2414 rtm->rtm_protocol = RTPROT_KERNEL;
2415 else if (rt->rt6i_flags & RTF_DEFAULT)
2416 rtm->rtm_protocol = RTPROT_RA;
2418 if (rt->rt6i_flags & RTF_CACHE)
2419 rtm->rtm_flags |= RTM_F_CLONED;
2422 NLA_PUT(skb, RTA_DST, 16, dst);
2423 rtm->rtm_dst_len = 128;
2424 } else if (rtm->rtm_dst_len)
2425 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2426 #ifdef CONFIG_IPV6_SUBTREES
2428 NLA_PUT(skb, RTA_SRC, 16, src);
2429 rtm->rtm_src_len = 128;
2430 } else if (rtm->rtm_src_len)
2431 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2434 #ifdef CONFIG_IPV6_MROUTE
2435 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2436 int err = ip6mr_get_route(net, skb, rtm, nowait);
2441 goto nla_put_failure;
2443 if (err == -EMSGSIZE)
2444 goto nla_put_failure;
2449 NLA_PUT_U32(skb, RTA_IIF, iif);
2451 struct in6_addr saddr_buf;
2452 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0)
2453 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2456 if (rt->rt6i_prefsrc.plen) {
2457 struct in6_addr saddr_buf;
2458 saddr_buf = rt->rt6i_prefsrc.addr;
2459 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2462 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2463 goto nla_put_failure;
2466 n = dst_get_neighbour(&rt->dst);
2468 NLA_PUT(skb, RTA_GATEWAY, 16, &n->primary_key);
2472 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2474 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2476 if (!(rt->rt6i_flags & RTF_EXPIRES))
2478 else if (rt->rt6i_expires - jiffies < INT_MAX)
2479 expires = rt->rt6i_expires - jiffies;
2483 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2484 expires, rt->dst.error) < 0)
2485 goto nla_put_failure;
2487 return nlmsg_end(skb, nlh);
2490 nlmsg_cancel(skb, nlh);
2494 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2496 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2499 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2500 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2501 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2505 return rt6_fill_node(arg->net,
2506 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2507 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2508 prefix, 0, NLM_F_MULTI);
2511 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2513 struct net *net = sock_net(in_skb->sk);
2514 struct nlattr *tb[RTA_MAX+1];
2515 struct rt6_info *rt;
2516 struct sk_buff *skb;
2521 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2526 memset(&fl6, 0, sizeof(fl6));
2529 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2532 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2536 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2539 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2543 iif = nla_get_u32(tb[RTA_IIF]);
2546 fl6.flowi6_oif = nla_get_u32(tb[RTA_OIF]);
2549 struct net_device *dev;
2550 dev = __dev_get_by_index(net, iif);
2557 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2563 /* Reserve room for dummy headers, this skb can pass
2564 through good chunk of routing engine.
2566 skb_reset_mac_header(skb);
2567 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2569 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl6);
2570 skb_dst_set(skb, &rt->dst);
2572 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2573 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2574 nlh->nlmsg_seq, 0, 0, 0);
2580 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2585 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2587 struct sk_buff *skb;
2588 struct net *net = info->nl_net;
2593 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2595 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2599 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2600 event, info->pid, seq, 0, 0, 0);
2602 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2603 WARN_ON(err == -EMSGSIZE);
2607 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2608 info->nlh, gfp_any());
2612 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2615 static int ip6_route_dev_notify(struct notifier_block *this,
2616 unsigned long event, void *data)
2618 struct net_device *dev = (struct net_device *)data;
2619 struct net *net = dev_net(dev);
2621 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2622 net->ipv6.ip6_null_entry->dst.dev = dev;
2623 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2624 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2625 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2626 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2627 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2628 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2639 #ifdef CONFIG_PROC_FS
2650 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2652 struct seq_file *m = p_arg;
2653 struct neighbour *n;
2655 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2657 #ifdef CONFIG_IPV6_SUBTREES
2658 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2660 seq_puts(m, "00000000000000000000000000000000 00 ");
2663 n = dst_get_neighbour(&rt->dst);
2665 seq_printf(m, "%pi6", n->primary_key);
2667 seq_puts(m, "00000000000000000000000000000000");
2670 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2671 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2672 rt->dst.__use, rt->rt6i_flags,
2673 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2677 static int ipv6_route_show(struct seq_file *m, void *v)
2679 struct net *net = (struct net *)m->private;
2680 fib6_clean_all(net, rt6_info_route, 0, m);
2684 static int ipv6_route_open(struct inode *inode, struct file *file)
2686 return single_open_net(inode, file, ipv6_route_show);
2689 static const struct file_operations ipv6_route_proc_fops = {
2690 .owner = THIS_MODULE,
2691 .open = ipv6_route_open,
2693 .llseek = seq_lseek,
2694 .release = single_release_net,
2697 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2699 struct net *net = (struct net *)seq->private;
2700 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2701 net->ipv6.rt6_stats->fib_nodes,
2702 net->ipv6.rt6_stats->fib_route_nodes,
2703 net->ipv6.rt6_stats->fib_rt_alloc,
2704 net->ipv6.rt6_stats->fib_rt_entries,
2705 net->ipv6.rt6_stats->fib_rt_cache,
2706 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2707 net->ipv6.rt6_stats->fib_discarded_routes);
2712 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2714 return single_open_net(inode, file, rt6_stats_seq_show);
2717 static const struct file_operations rt6_stats_seq_fops = {
2718 .owner = THIS_MODULE,
2719 .open = rt6_stats_seq_open,
2721 .llseek = seq_lseek,
2722 .release = single_release_net,
2724 #endif /* CONFIG_PROC_FS */
2726 #ifdef CONFIG_SYSCTL
2729 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2730 void __user *buffer, size_t *lenp, loff_t *ppos)
2737 net = (struct net *)ctl->extra1;
2738 delay = net->ipv6.sysctl.flush_delay;
2739 proc_dointvec(ctl, write, buffer, lenp, ppos);
2740 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2744 ctl_table ipv6_route_table_template[] = {
2746 .procname = "flush",
2747 .data = &init_net.ipv6.sysctl.flush_delay,
2748 .maxlen = sizeof(int),
2750 .proc_handler = ipv6_sysctl_rtcache_flush
2753 .procname = "gc_thresh",
2754 .data = &ip6_dst_ops_template.gc_thresh,
2755 .maxlen = sizeof(int),
2757 .proc_handler = proc_dointvec,
2760 .procname = "max_size",
2761 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2762 .maxlen = sizeof(int),
2764 .proc_handler = proc_dointvec,
2767 .procname = "gc_min_interval",
2768 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2769 .maxlen = sizeof(int),
2771 .proc_handler = proc_dointvec_jiffies,
2774 .procname = "gc_timeout",
2775 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2776 .maxlen = sizeof(int),
2778 .proc_handler = proc_dointvec_jiffies,
2781 .procname = "gc_interval",
2782 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2783 .maxlen = sizeof(int),
2785 .proc_handler = proc_dointvec_jiffies,
2788 .procname = "gc_elasticity",
2789 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2790 .maxlen = sizeof(int),
2792 .proc_handler = proc_dointvec,
2795 .procname = "mtu_expires",
2796 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2797 .maxlen = sizeof(int),
2799 .proc_handler = proc_dointvec_jiffies,
2802 .procname = "min_adv_mss",
2803 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2804 .maxlen = sizeof(int),
2806 .proc_handler = proc_dointvec,
2809 .procname = "gc_min_interval_ms",
2810 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2811 .maxlen = sizeof(int),
2813 .proc_handler = proc_dointvec_ms_jiffies,
2818 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2820 struct ctl_table *table;
2822 table = kmemdup(ipv6_route_table_template,
2823 sizeof(ipv6_route_table_template),
2827 table[0].data = &net->ipv6.sysctl.flush_delay;
2828 table[0].extra1 = net;
2829 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2830 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2831 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2832 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2833 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2834 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2835 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2836 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2837 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2844 static int __net_init ip6_route_net_init(struct net *net)
2848 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2849 sizeof(net->ipv6.ip6_dst_ops));
2851 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2852 goto out_ip6_dst_ops;
2854 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2855 sizeof(*net->ipv6.ip6_null_entry),
2857 if (!net->ipv6.ip6_null_entry)
2858 goto out_ip6_dst_entries;
2859 net->ipv6.ip6_null_entry->dst.path =
2860 (struct dst_entry *)net->ipv6.ip6_null_entry;
2861 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2862 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2863 ip6_template_metrics, true);
2865 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2866 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2867 sizeof(*net->ipv6.ip6_prohibit_entry),
2869 if (!net->ipv6.ip6_prohibit_entry)
2870 goto out_ip6_null_entry;
2871 net->ipv6.ip6_prohibit_entry->dst.path =
2872 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2873 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2874 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2875 ip6_template_metrics, true);
2877 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2878 sizeof(*net->ipv6.ip6_blk_hole_entry),
2880 if (!net->ipv6.ip6_blk_hole_entry)
2881 goto out_ip6_prohibit_entry;
2882 net->ipv6.ip6_blk_hole_entry->dst.path =
2883 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2884 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2885 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2886 ip6_template_metrics, true);
2889 net->ipv6.sysctl.flush_delay = 0;
2890 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2891 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2892 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2893 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2894 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2895 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2896 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2898 #ifdef CONFIG_PROC_FS
2899 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2900 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2902 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2908 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2909 out_ip6_prohibit_entry:
2910 kfree(net->ipv6.ip6_prohibit_entry);
2912 kfree(net->ipv6.ip6_null_entry);
2914 out_ip6_dst_entries:
2915 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2920 static void __net_exit ip6_route_net_exit(struct net *net)
2922 #ifdef CONFIG_PROC_FS
2923 proc_net_remove(net, "ipv6_route");
2924 proc_net_remove(net, "rt6_stats");
2926 kfree(net->ipv6.ip6_null_entry);
2927 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2928 kfree(net->ipv6.ip6_prohibit_entry);
2929 kfree(net->ipv6.ip6_blk_hole_entry);
2931 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2934 static struct pernet_operations ip6_route_net_ops = {
2935 .init = ip6_route_net_init,
2936 .exit = ip6_route_net_exit,
2939 static struct notifier_block ip6_route_dev_notifier = {
2940 .notifier_call = ip6_route_dev_notify,
2944 int __init ip6_route_init(void)
2949 ip6_dst_ops_template.kmem_cachep =
2950 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2951 SLAB_HWCACHE_ALIGN, NULL);
2952 if (!ip6_dst_ops_template.kmem_cachep)
2955 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2957 goto out_kmem_cache;
2959 ret = register_pernet_subsys(&ip6_route_net_ops);
2961 goto out_dst_entries;
2963 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2965 /* Registering of the loopback is done before this portion of code,
2966 * the loopback reference in rt6_info will not be taken, do it
2967 * manually for init_net */
2968 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2969 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2970 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2971 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2972 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2973 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2974 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2978 goto out_register_subsys;
2984 ret = fib6_rules_init();
2989 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
2990 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
2991 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
2992 goto fib6_rules_init;
2994 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2996 goto fib6_rules_init;
3002 fib6_rules_cleanup();
3007 out_register_subsys:
3008 unregister_pernet_subsys(&ip6_route_net_ops);
3010 dst_entries_destroy(&ip6_dst_blackhole_ops);
3012 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3016 void ip6_route_cleanup(void)
3018 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3019 fib6_rules_cleanup();
3022 unregister_pernet_subsys(&ip6_route_net_ops);
3023 dst_entries_destroy(&ip6_dst_blackhole_ops);
3024 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);