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_noref(&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_noref(&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 rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
662 const struct in6_addr *saddr, int oif, int strict)
664 struct flowi6 fl6 = {
668 struct dst_entry *dst;
669 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
672 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
673 flags |= RT6_LOOKUP_F_HAS_SADDR;
676 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
678 return (struct rt6_info *) dst;
685 EXPORT_SYMBOL(rt6_lookup);
687 /* ip6_ins_rt is called with FREE table->tb6_lock.
688 It takes new route entry, the addition fails by any reason the
689 route is freed. In any case, if caller does not hold it, it may
693 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
696 struct fib6_table *table;
698 table = rt->rt6i_table;
699 write_lock_bh(&table->tb6_lock);
700 err = fib6_add(&table->tb6_root, rt, info);
701 write_unlock_bh(&table->tb6_lock);
706 int ip6_ins_rt(struct rt6_info *rt)
708 struct nl_info info = {
709 .nl_net = dev_net(rt->rt6i_dev),
711 return __ip6_ins_rt(rt, &info);
714 static struct rt6_info *rt6_alloc_cow(const struct rt6_info *ort,
715 const struct in6_addr *daddr,
716 const struct in6_addr *saddr)
724 rt = ip6_rt_copy(ort, daddr);
727 struct neighbour *neigh;
728 int attempts = !in_softirq();
730 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
731 if (rt->rt6i_dst.plen != 128 &&
732 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
733 rt->rt6i_flags |= RTF_ANYCAST;
734 rt->rt6i_gateway = *daddr;
737 rt->rt6i_flags |= RTF_CACHE;
739 #ifdef CONFIG_IPV6_SUBTREES
740 if (rt->rt6i_src.plen && saddr) {
741 rt->rt6i_src.addr = *saddr;
742 rt->rt6i_src.plen = 128;
747 neigh = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway,
750 struct net *net = dev_net(rt->rt6i_dev);
751 int saved_rt_min_interval =
752 net->ipv6.sysctl.ip6_rt_gc_min_interval;
753 int saved_rt_elasticity =
754 net->ipv6.sysctl.ip6_rt_gc_elasticity;
756 if (attempts-- > 0) {
757 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
758 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
760 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
762 net->ipv6.sysctl.ip6_rt_gc_elasticity =
764 net->ipv6.sysctl.ip6_rt_gc_min_interval =
765 saved_rt_min_interval;
771 "ipv6: Neighbour table overflow.\n");
775 dst_set_neighbour(&rt->dst, neigh);
782 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
783 const struct in6_addr *daddr)
785 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
788 rt->rt6i_flags |= RTF_CACHE;
789 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_noref_raw(&ort->dst)));
794 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
795 struct flowi6 *fl6, int flags)
797 struct fib6_node *fn;
798 struct rt6_info *rt, *nrt;
802 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
804 strict |= flags & RT6_LOOKUP_F_IFACE;
807 read_lock_bh(&table->tb6_lock);
810 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
813 rt = rt6_select(fn, oif, strict | reachable);
815 BACKTRACK(net, &fl6->saddr);
816 if (rt == net->ipv6.ip6_null_entry ||
817 rt->rt6i_flags & RTF_CACHE)
821 read_unlock_bh(&table->tb6_lock);
823 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
824 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
825 else if (!(rt->dst.flags & DST_HOST))
826 nrt = rt6_alloc_clone(rt, &fl6->daddr);
830 dst_release(&rt->dst);
831 rt = nrt ? : net->ipv6.ip6_null_entry;
835 err = ip6_ins_rt(nrt);
844 * Race condition! In the gap, when table->tb6_lock was
845 * released someone could insert this route. Relookup.
847 dst_release(&rt->dst);
856 read_unlock_bh(&table->tb6_lock);
858 rt->dst.lastuse = jiffies;
864 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
865 struct flowi6 *fl6, int flags)
867 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
870 void ip6_route_input(struct sk_buff *skb)
872 const struct ipv6hdr *iph = ipv6_hdr(skb);
873 struct net *net = dev_net(skb->dev);
874 int flags = RT6_LOOKUP_F_HAS_SADDR;
875 struct flowi6 fl6 = {
876 .flowi6_iif = skb->dev->ifindex,
879 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
880 .flowi6_mark = skb->mark,
881 .flowi6_proto = iph->nexthdr,
884 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
885 flags |= RT6_LOOKUP_F_IFACE;
887 skb_dst_set(skb, fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_input));
890 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
891 struct flowi6 *fl6, int flags)
893 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
896 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
901 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
902 flags |= RT6_LOOKUP_F_IFACE;
904 if (!ipv6_addr_any(&fl6->saddr))
905 flags |= RT6_LOOKUP_F_HAS_SADDR;
907 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
909 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
912 EXPORT_SYMBOL(ip6_route_output);
914 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
916 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
917 struct dst_entry *new = NULL;
919 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
921 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
926 new->input = dst_discard;
927 new->output = dst_discard;
929 if (dst_metrics_read_only(&ort->dst))
930 new->_metrics = ort->dst._metrics;
932 dst_copy_metrics(new, &ort->dst);
933 rt->rt6i_idev = ort->rt6i_idev;
935 in6_dev_hold(rt->rt6i_idev);
936 rt->rt6i_expires = 0;
938 rt->rt6i_gateway = ort->rt6i_gateway;
939 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
942 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
943 #ifdef CONFIG_IPV6_SUBTREES
944 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
950 dst_release(dst_orig);
951 return new ? new : ERR_PTR(-ENOMEM);
955 * Destination cache support functions
958 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
962 rt = (struct rt6_info *) dst;
964 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
965 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
967 rt6_bind_peer(rt, 0);
968 rt->rt6i_peer_genid = rt6_peer_genid();
975 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
977 struct rt6_info *rt = (struct rt6_info *) dst;
980 if (rt->rt6i_flags & RTF_CACHE) {
981 if (rt6_check_expired(rt)) {
993 static void ip6_link_failure(struct sk_buff *skb)
997 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
999 rt = (struct rt6_info *) skb_dst(skb);
1001 if (rt->rt6i_flags & RTF_CACHE) {
1002 dst_set_expires(&rt->dst, 0);
1003 rt->rt6i_flags |= RTF_EXPIRES;
1004 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1005 rt->rt6i_node->fn_sernum = -1;
1009 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1011 struct rt6_info *rt6 = (struct rt6_info*)dst;
1013 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1014 rt6->rt6i_flags |= RTF_MODIFIED;
1015 if (mtu < IPV6_MIN_MTU) {
1016 u32 features = dst_metric(dst, RTAX_FEATURES);
1018 features |= RTAX_FEATURE_ALLFRAG;
1019 dst_metric_set(dst, RTAX_FEATURES, features);
1021 dst_metric_set(dst, RTAX_MTU, mtu);
1025 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1027 struct net_device *dev = dst->dev;
1028 unsigned int mtu = dst_mtu(dst);
1029 struct net *net = dev_net(dev);
1031 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1033 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1034 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1037 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1038 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1039 * IPV6_MAXPLEN is also valid and means: "any MSS,
1040 * rely only on pmtu discovery"
1042 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1047 static unsigned int ip6_mtu(const struct dst_entry *dst)
1049 struct inet6_dev *idev;
1050 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1058 idev = __in6_dev_get(dst->dev);
1060 mtu = idev->cnf.mtu6;
1066 static struct dst_entry *icmp6_dst_gc_list;
1067 static DEFINE_SPINLOCK(icmp6_dst_lock);
1069 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1070 struct neighbour *neigh,
1073 struct dst_entry *dst;
1074 struct rt6_info *rt;
1075 struct inet6_dev *idev = in6_dev_get(dev);
1076 struct net *net = dev_net(dev);
1078 if (unlikely(!idev))
1081 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1082 if (unlikely(!rt)) {
1084 dst = ERR_PTR(-ENOMEM);
1091 neigh = __neigh_lookup_errno(&nd_tbl, &fl6->daddr, dev);
1096 rt->dst.flags |= DST_HOST;
1097 rt->dst.output = ip6_output;
1098 dst_set_neighbour(&rt->dst, neigh);
1099 atomic_set(&rt->dst.__refcnt, 1);
1100 rt->rt6i_dst.addr = fl6->daddr;
1101 rt->rt6i_dst.plen = 128;
1102 rt->rt6i_idev = idev;
1103 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1105 spin_lock_bh(&icmp6_dst_lock);
1106 rt->dst.next = icmp6_dst_gc_list;
1107 icmp6_dst_gc_list = &rt->dst;
1108 spin_unlock_bh(&icmp6_dst_lock);
1110 fib6_force_start_gc(net);
1112 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1118 int icmp6_dst_gc(void)
1120 struct dst_entry *dst, **pprev;
1123 spin_lock_bh(&icmp6_dst_lock);
1124 pprev = &icmp6_dst_gc_list;
1126 while ((dst = *pprev) != NULL) {
1127 if (!atomic_read(&dst->__refcnt)) {
1136 spin_unlock_bh(&icmp6_dst_lock);
1141 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1144 struct dst_entry *dst, **pprev;
1146 spin_lock_bh(&icmp6_dst_lock);
1147 pprev = &icmp6_dst_gc_list;
1148 while ((dst = *pprev) != NULL) {
1149 struct rt6_info *rt = (struct rt6_info *) dst;
1150 if (func(rt, arg)) {
1157 spin_unlock_bh(&icmp6_dst_lock);
1160 static int ip6_dst_gc(struct dst_ops *ops)
1162 unsigned long now = jiffies;
1163 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1164 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1165 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1166 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1167 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1168 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1171 entries = dst_entries_get_fast(ops);
1172 if (time_after(rt_last_gc + rt_min_interval, now) &&
1173 entries <= rt_max_size)
1176 net->ipv6.ip6_rt_gc_expire++;
1177 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1178 net->ipv6.ip6_rt_last_gc = now;
1179 entries = dst_entries_get_slow(ops);
1180 if (entries < ops->gc_thresh)
1181 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1183 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1184 return entries > rt_max_size;
1187 /* Clean host part of a prefix. Not necessary in radix tree,
1188 but results in cleaner routing tables.
1190 Remove it only when all the things will work!
1193 int ip6_dst_hoplimit(struct dst_entry *dst)
1195 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1196 if (hoplimit == 0) {
1197 struct net_device *dev = dst->dev;
1198 struct inet6_dev *idev;
1201 idev = __in6_dev_get(dev);
1203 hoplimit = idev->cnf.hop_limit;
1205 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1210 EXPORT_SYMBOL(ip6_dst_hoplimit);
1216 int ip6_route_add(struct fib6_config *cfg)
1219 struct net *net = cfg->fc_nlinfo.nl_net;
1220 struct rt6_info *rt = NULL;
1221 struct net_device *dev = NULL;
1222 struct inet6_dev *idev = NULL;
1223 struct fib6_table *table;
1226 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1228 #ifndef CONFIG_IPV6_SUBTREES
1229 if (cfg->fc_src_len)
1232 if (cfg->fc_ifindex) {
1234 dev = dev_get_by_index(net, cfg->fc_ifindex);
1237 idev = in6_dev_get(dev);
1242 if (cfg->fc_metric == 0)
1243 cfg->fc_metric = IP6_RT_PRIO_USER;
1246 if (cfg->fc_nlinfo.nlh &&
1247 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1248 table = fib6_get_table(net, cfg->fc_table);
1250 printk(KERN_WARNING "IPv6: NLM_F_CREATE should be specified when creating new route\n");
1251 table = fib6_new_table(net, cfg->fc_table);
1254 table = fib6_new_table(net, cfg->fc_table);
1260 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1267 rt->dst.obsolete = -1;
1268 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1269 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1272 if (cfg->fc_protocol == RTPROT_UNSPEC)
1273 cfg->fc_protocol = RTPROT_BOOT;
1274 rt->rt6i_protocol = cfg->fc_protocol;
1276 addr_type = ipv6_addr_type(&cfg->fc_dst);
1278 if (addr_type & IPV6_ADDR_MULTICAST)
1279 rt->dst.input = ip6_mc_input;
1280 else if (cfg->fc_flags & RTF_LOCAL)
1281 rt->dst.input = ip6_input;
1283 rt->dst.input = ip6_forward;
1285 rt->dst.output = ip6_output;
1287 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1288 rt->rt6i_dst.plen = cfg->fc_dst_len;
1289 if (rt->rt6i_dst.plen == 128)
1290 rt->dst.flags |= DST_HOST;
1292 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1293 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1298 dst_init_metrics(&rt->dst, metrics, 0);
1300 #ifdef CONFIG_IPV6_SUBTREES
1301 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1302 rt->rt6i_src.plen = cfg->fc_src_len;
1305 rt->rt6i_metric = cfg->fc_metric;
1307 /* We cannot add true routes via loopback here,
1308 they would result in kernel looping; promote them to reject routes
1310 if ((cfg->fc_flags & RTF_REJECT) ||
1311 (dev && (dev->flags & IFF_LOOPBACK) &&
1312 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1313 !(cfg->fc_flags & RTF_LOCAL))) {
1314 /* hold loopback dev/idev if we haven't done so. */
1315 if (dev != net->loopback_dev) {
1320 dev = net->loopback_dev;
1322 idev = in6_dev_get(dev);
1328 rt->dst.output = ip6_pkt_discard_out;
1329 rt->dst.input = ip6_pkt_discard;
1330 rt->dst.error = -ENETUNREACH;
1331 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1335 if (cfg->fc_flags & RTF_GATEWAY) {
1336 const struct in6_addr *gw_addr;
1339 gw_addr = &cfg->fc_gateway;
1340 rt->rt6i_gateway = *gw_addr;
1341 gwa_type = ipv6_addr_type(gw_addr);
1343 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1344 struct rt6_info *grt;
1346 /* IPv6 strictly inhibits using not link-local
1347 addresses as nexthop address.
1348 Otherwise, router will not able to send redirects.
1349 It is very good, but in some (rare!) circumstances
1350 (SIT, PtP, NBMA NOARP links) it is handy to allow
1351 some exceptions. --ANK
1354 if (!(gwa_type & IPV6_ADDR_UNICAST))
1357 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1359 err = -EHOSTUNREACH;
1363 if (dev != grt->rt6i_dev) {
1364 dst_release(&grt->dst);
1368 dev = grt->rt6i_dev;
1369 idev = grt->rt6i_idev;
1371 in6_dev_hold(grt->rt6i_idev);
1373 if (!(grt->rt6i_flags & RTF_GATEWAY))
1375 dst_release(&grt->dst);
1381 if (!dev || (dev->flags & IFF_LOOPBACK))
1389 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1390 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1394 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1395 rt->rt6i_prefsrc.plen = 128;
1397 rt->rt6i_prefsrc.plen = 0;
1399 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1400 struct neighbour *n = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1405 dst_set_neighbour(&rt->dst, n);
1408 rt->rt6i_flags = cfg->fc_flags;
1415 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1416 int type = nla_type(nla);
1419 if (type > RTAX_MAX) {
1424 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1430 rt->rt6i_idev = idev;
1431 rt->rt6i_table = table;
1433 cfg->fc_nlinfo.nl_net = dev_net(dev);
1435 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1447 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1450 struct fib6_table *table;
1451 struct net *net = dev_net(rt->rt6i_dev);
1453 if (rt == net->ipv6.ip6_null_entry)
1456 table = rt->rt6i_table;
1457 write_lock_bh(&table->tb6_lock);
1459 err = fib6_del(rt, info);
1460 dst_release(&rt->dst);
1462 write_unlock_bh(&table->tb6_lock);
1467 int ip6_del_rt(struct rt6_info *rt)
1469 struct nl_info info = {
1470 .nl_net = dev_net(rt->rt6i_dev),
1472 return __ip6_del_rt(rt, &info);
1475 static int ip6_route_del(struct fib6_config *cfg)
1477 struct fib6_table *table;
1478 struct fib6_node *fn;
1479 struct rt6_info *rt;
1482 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1486 read_lock_bh(&table->tb6_lock);
1488 fn = fib6_locate(&table->tb6_root,
1489 &cfg->fc_dst, cfg->fc_dst_len,
1490 &cfg->fc_src, cfg->fc_src_len);
1493 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1494 if (cfg->fc_ifindex &&
1496 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1498 if (cfg->fc_flags & RTF_GATEWAY &&
1499 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1501 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1504 read_unlock_bh(&table->tb6_lock);
1506 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1509 read_unlock_bh(&table->tb6_lock);
1517 struct ip6rd_flowi {
1519 struct in6_addr gateway;
1522 static struct rt6_info *__ip6_route_redirect(struct net *net,
1523 struct fib6_table *table,
1527 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1528 struct rt6_info *rt;
1529 struct fib6_node *fn;
1532 * Get the "current" route for this destination and
1533 * check if the redirect has come from approriate router.
1535 * RFC 2461 specifies that redirects should only be
1536 * accepted if they come from the nexthop to the target.
1537 * Due to the way the routes are chosen, this notion
1538 * is a bit fuzzy and one might need to check all possible
1542 read_lock_bh(&table->tb6_lock);
1543 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1545 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1547 * Current route is on-link; redirect is always invalid.
1549 * Seems, previous statement is not true. It could
1550 * be node, which looks for us as on-link (f.e. proxy ndisc)
1551 * But then router serving it might decide, that we should
1552 * know truth 8)8) --ANK (980726).
1554 if (rt6_check_expired(rt))
1556 if (!(rt->rt6i_flags & RTF_GATEWAY))
1558 if (fl6->flowi6_oif != rt->rt6i_dev->ifindex)
1560 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1566 rt = net->ipv6.ip6_null_entry;
1567 BACKTRACK(net, &fl6->saddr);
1571 read_unlock_bh(&table->tb6_lock);
1576 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1577 const struct in6_addr *src,
1578 const struct in6_addr *gateway,
1579 struct net_device *dev)
1581 int flags = RT6_LOOKUP_F_HAS_SADDR;
1582 struct net *net = dev_net(dev);
1583 struct ip6rd_flowi rdfl = {
1585 .flowi6_oif = dev->ifindex,
1591 rdfl.gateway = *gateway;
1593 if (rt6_need_strict(dest))
1594 flags |= RT6_LOOKUP_F_IFACE;
1596 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1597 flags, __ip6_route_redirect);
1600 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1601 const struct in6_addr *saddr,
1602 struct neighbour *neigh, u8 *lladdr, int on_link)
1604 struct rt6_info *rt, *nrt = NULL;
1605 struct netevent_redirect netevent;
1606 struct net *net = dev_net(neigh->dev);
1608 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1610 if (rt == net->ipv6.ip6_null_entry) {
1611 if (net_ratelimit())
1612 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1613 "for redirect target\n");
1618 * We have finally decided to accept it.
1621 neigh_update(neigh, lladdr, NUD_STALE,
1622 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1623 NEIGH_UPDATE_F_OVERRIDE|
1624 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1625 NEIGH_UPDATE_F_ISROUTER))
1629 * Redirect received -> path was valid.
1630 * Look, redirects are sent only in response to data packets,
1631 * so that this nexthop apparently is reachable. --ANK
1633 dst_confirm(&rt->dst);
1635 /* Duplicate redirect: silently ignore. */
1636 if (neigh == dst_get_neighbour_noref_raw(&rt->dst))
1639 nrt = ip6_rt_copy(rt, dest);
1643 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1645 nrt->rt6i_flags &= ~RTF_GATEWAY;
1647 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1648 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1650 if (ip6_ins_rt(nrt))
1653 netevent.old = &rt->dst;
1654 netevent.new = &nrt->dst;
1655 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1657 if (rt->rt6i_flags & RTF_CACHE) {
1663 dst_release(&rt->dst);
1667 * Handle ICMP "packet too big" messages
1668 * i.e. Path MTU discovery
1671 static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1672 struct net *net, u32 pmtu, int ifindex)
1674 struct rt6_info *rt, *nrt;
1677 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1681 if (rt6_check_expired(rt)) {
1686 if (pmtu >= dst_mtu(&rt->dst))
1689 if (pmtu < IPV6_MIN_MTU) {
1691 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1692 * MTU (1280) and a fragment header should always be included
1693 * after a node receiving Too Big message reporting PMTU is
1694 * less than the IPv6 Minimum Link MTU.
1696 pmtu = IPV6_MIN_MTU;
1700 /* New mtu received -> path was valid.
1701 They are sent only in response to data packets,
1702 so that this nexthop apparently is reachable. --ANK
1704 dst_confirm(&rt->dst);
1706 /* Host route. If it is static, it would be better
1707 not to override it, but add new one, so that
1708 when cache entry will expire old pmtu
1709 would return automatically.
1711 if (rt->rt6i_flags & RTF_CACHE) {
1712 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1714 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1715 features |= RTAX_FEATURE_ALLFRAG;
1716 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1718 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1719 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1724 Two cases are possible:
1725 1. It is connected route. Action: COW
1726 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1728 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1729 nrt = rt6_alloc_cow(rt, daddr, saddr);
1731 nrt = rt6_alloc_clone(rt, daddr);
1734 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1736 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1737 features |= RTAX_FEATURE_ALLFRAG;
1738 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1741 /* According to RFC 1981, detecting PMTU increase shouldn't be
1742 * happened within 5 mins, the recommended timer is 10 mins.
1743 * Here this route expiration time is set to ip6_rt_mtu_expires
1744 * which is 10 mins. After 10 mins the decreased pmtu is expired
1745 * and detecting PMTU increase will be automatically happened.
1747 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1748 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1753 dst_release(&rt->dst);
1756 void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1757 struct net_device *dev, u32 pmtu)
1759 struct net *net = dev_net(dev);
1762 * RFC 1981 states that a node "MUST reduce the size of the packets it
1763 * is sending along the path" that caused the Packet Too Big message.
1764 * Since it's not possible in the general case to determine which
1765 * interface was used to send the original packet, we update the MTU
1766 * on the interface that will be used to send future packets. We also
1767 * update the MTU on the interface that received the Packet Too Big in
1768 * case the original packet was forced out that interface with
1769 * SO_BINDTODEVICE or similar. This is the next best thing to the
1770 * correct behaviour, which would be to update the MTU on all
1773 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1774 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1778 * Misc support functions
1781 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
1782 const struct in6_addr *dest)
1784 struct net *net = dev_net(ort->rt6i_dev);
1785 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1789 rt->dst.input = ort->dst.input;
1790 rt->dst.output = ort->dst.output;
1791 rt->dst.flags |= DST_HOST;
1793 rt->rt6i_dst.addr = *dest;
1794 rt->rt6i_dst.plen = 128;
1795 dst_copy_metrics(&rt->dst, &ort->dst);
1796 rt->dst.error = ort->dst.error;
1797 rt->rt6i_idev = ort->rt6i_idev;
1799 in6_dev_hold(rt->rt6i_idev);
1800 rt->dst.lastuse = jiffies;
1801 rt->rt6i_expires = 0;
1803 rt->rt6i_gateway = ort->rt6i_gateway;
1804 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1805 rt->rt6i_metric = 0;
1807 #ifdef CONFIG_IPV6_SUBTREES
1808 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1810 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1811 rt->rt6i_table = ort->rt6i_table;
1816 #ifdef CONFIG_IPV6_ROUTE_INFO
1817 static struct rt6_info *rt6_get_route_info(struct net *net,
1818 const struct in6_addr *prefix, int prefixlen,
1819 const struct in6_addr *gwaddr, int ifindex)
1821 struct fib6_node *fn;
1822 struct rt6_info *rt = NULL;
1823 struct fib6_table *table;
1825 table = fib6_get_table(net, RT6_TABLE_INFO);
1829 write_lock_bh(&table->tb6_lock);
1830 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1834 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1835 if (rt->rt6i_dev->ifindex != ifindex)
1837 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1839 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1845 write_unlock_bh(&table->tb6_lock);
1849 static struct rt6_info *rt6_add_route_info(struct net *net,
1850 const struct in6_addr *prefix, int prefixlen,
1851 const struct in6_addr *gwaddr, int ifindex,
1854 struct fib6_config cfg = {
1855 .fc_table = RT6_TABLE_INFO,
1856 .fc_metric = IP6_RT_PRIO_USER,
1857 .fc_ifindex = ifindex,
1858 .fc_dst_len = prefixlen,
1859 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1860 RTF_UP | RTF_PREF(pref),
1862 .fc_nlinfo.nlh = NULL,
1863 .fc_nlinfo.nl_net = net,
1866 cfg.fc_dst = *prefix;
1867 cfg.fc_gateway = *gwaddr;
1869 /* We should treat it as a default route if prefix length is 0. */
1871 cfg.fc_flags |= RTF_DEFAULT;
1873 ip6_route_add(&cfg);
1875 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1879 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1881 struct rt6_info *rt;
1882 struct fib6_table *table;
1884 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1888 write_lock_bh(&table->tb6_lock);
1889 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1890 if (dev == rt->rt6i_dev &&
1891 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1892 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1897 write_unlock_bh(&table->tb6_lock);
1901 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1902 struct net_device *dev,
1905 struct fib6_config cfg = {
1906 .fc_table = RT6_TABLE_DFLT,
1907 .fc_metric = IP6_RT_PRIO_USER,
1908 .fc_ifindex = dev->ifindex,
1909 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1910 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1912 .fc_nlinfo.nlh = NULL,
1913 .fc_nlinfo.nl_net = dev_net(dev),
1916 cfg.fc_gateway = *gwaddr;
1918 ip6_route_add(&cfg);
1920 return rt6_get_dflt_router(gwaddr, dev);
1923 void rt6_purge_dflt_routers(struct net *net)
1925 struct rt6_info *rt;
1926 struct fib6_table *table;
1928 /* NOTE: Keep consistent with rt6_get_dflt_router */
1929 table = fib6_get_table(net, RT6_TABLE_DFLT);
1934 read_lock_bh(&table->tb6_lock);
1935 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1936 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1938 read_unlock_bh(&table->tb6_lock);
1943 read_unlock_bh(&table->tb6_lock);
1946 static void rtmsg_to_fib6_config(struct net *net,
1947 struct in6_rtmsg *rtmsg,
1948 struct fib6_config *cfg)
1950 memset(cfg, 0, sizeof(*cfg));
1952 cfg->fc_table = RT6_TABLE_MAIN;
1953 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1954 cfg->fc_metric = rtmsg->rtmsg_metric;
1955 cfg->fc_expires = rtmsg->rtmsg_info;
1956 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1957 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1958 cfg->fc_flags = rtmsg->rtmsg_flags;
1960 cfg->fc_nlinfo.nl_net = net;
1962 cfg->fc_dst = rtmsg->rtmsg_dst;
1963 cfg->fc_src = rtmsg->rtmsg_src;
1964 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1967 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1969 struct fib6_config cfg;
1970 struct in6_rtmsg rtmsg;
1974 case SIOCADDRT: /* Add a route */
1975 case SIOCDELRT: /* Delete a route */
1976 if (!capable(CAP_NET_ADMIN))
1978 err = copy_from_user(&rtmsg, arg,
1979 sizeof(struct in6_rtmsg));
1983 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1988 err = ip6_route_add(&cfg);
1991 err = ip6_route_del(&cfg);
2005 * Drop the packet on the floor
2008 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2011 struct dst_entry *dst = skb_dst(skb);
2012 switch (ipstats_mib_noroutes) {
2013 case IPSTATS_MIB_INNOROUTES:
2014 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2015 if (type == IPV6_ADDR_ANY) {
2016 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2017 IPSTATS_MIB_INADDRERRORS);
2021 case IPSTATS_MIB_OUTNOROUTES:
2022 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2023 ipstats_mib_noroutes);
2026 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2031 static int ip6_pkt_discard(struct sk_buff *skb)
2033 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2036 static int ip6_pkt_discard_out(struct sk_buff *skb)
2038 skb->dev = skb_dst(skb)->dev;
2039 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2042 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2044 static int ip6_pkt_prohibit(struct sk_buff *skb)
2046 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2049 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2051 skb->dev = skb_dst(skb)->dev;
2052 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2058 * Allocate a dst for local (unicast / anycast) address.
2061 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2062 const struct in6_addr *addr,
2065 struct net *net = dev_net(idev->dev);
2066 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2067 net->loopback_dev, 0);
2068 struct neighbour *neigh;
2071 if (net_ratelimit())
2072 pr_warning("IPv6: Maximum number of routes reached,"
2073 " consider increasing route/max_size.\n");
2074 return ERR_PTR(-ENOMEM);
2079 rt->dst.flags |= DST_HOST;
2080 rt->dst.input = ip6_input;
2081 rt->dst.output = ip6_output;
2082 rt->rt6i_idev = idev;
2083 rt->dst.obsolete = -1;
2085 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2087 rt->rt6i_flags |= RTF_ANYCAST;
2089 rt->rt6i_flags |= RTF_LOCAL;
2090 neigh = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, rt->rt6i_dev);
2091 if (IS_ERR(neigh)) {
2094 return ERR_CAST(neigh);
2096 dst_set_neighbour(&rt->dst, neigh);
2098 rt->rt6i_dst.addr = *addr;
2099 rt->rt6i_dst.plen = 128;
2100 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2102 atomic_set(&rt->dst.__refcnt, 1);
2107 int ip6_route_get_saddr(struct net *net,
2108 struct rt6_info *rt,
2109 const struct in6_addr *daddr,
2111 struct in6_addr *saddr)
2113 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2115 if (rt->rt6i_prefsrc.plen)
2116 *saddr = rt->rt6i_prefsrc.addr;
2118 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2119 daddr, prefs, saddr);
2123 /* remove deleted ip from prefsrc entries */
2124 struct arg_dev_net_ip {
2125 struct net_device *dev;
2127 struct in6_addr *addr;
2130 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2132 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2133 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2134 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2136 if (((void *)rt->rt6i_dev == dev || !dev) &&
2137 rt != net->ipv6.ip6_null_entry &&
2138 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2139 /* remove prefsrc entry */
2140 rt->rt6i_prefsrc.plen = 0;
2145 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2147 struct net *net = dev_net(ifp->idev->dev);
2148 struct arg_dev_net_ip adni = {
2149 .dev = ifp->idev->dev,
2153 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2156 struct arg_dev_net {
2157 struct net_device *dev;
2161 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2163 const struct arg_dev_net *adn = arg;
2164 const struct net_device *dev = adn->dev;
2166 if ((rt->rt6i_dev == dev || !dev) &&
2167 rt != adn->net->ipv6.ip6_null_entry) {
2168 RT6_TRACE("deleted by ifdown %p\n", rt);
2174 void rt6_ifdown(struct net *net, struct net_device *dev)
2176 struct arg_dev_net adn = {
2181 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2182 icmp6_clean_all(fib6_ifdown, &adn);
2185 struct rt6_mtu_change_arg
2187 struct net_device *dev;
2191 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2193 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2194 struct inet6_dev *idev;
2196 /* In IPv6 pmtu discovery is not optional,
2197 so that RTAX_MTU lock cannot disable it.
2198 We still use this lock to block changes
2199 caused by addrconf/ndisc.
2202 idev = __in6_dev_get(arg->dev);
2206 /* For administrative MTU increase, there is no way to discover
2207 IPv6 PMTU increase, so PMTU increase should be updated here.
2208 Since RFC 1981 doesn't include administrative MTU increase
2209 update PMTU increase is a MUST. (i.e. jumbo frame)
2212 If new MTU is less than route PMTU, this new MTU will be the
2213 lowest MTU in the path, update the route PMTU to reflect PMTU
2214 decreases; if new MTU is greater than route PMTU, and the
2215 old MTU is the lowest MTU in the path, update the route PMTU
2216 to reflect the increase. In this case if the other nodes' MTU
2217 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2220 if (rt->rt6i_dev == arg->dev &&
2221 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2222 (dst_mtu(&rt->dst) >= arg->mtu ||
2223 (dst_mtu(&rt->dst) < arg->mtu &&
2224 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2225 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2230 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2232 struct rt6_mtu_change_arg arg = {
2237 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2240 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2241 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2242 [RTA_OIF] = { .type = NLA_U32 },
2243 [RTA_IIF] = { .type = NLA_U32 },
2244 [RTA_PRIORITY] = { .type = NLA_U32 },
2245 [RTA_METRICS] = { .type = NLA_NESTED },
2248 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2249 struct fib6_config *cfg)
2252 struct nlattr *tb[RTA_MAX+1];
2255 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2260 rtm = nlmsg_data(nlh);
2261 memset(cfg, 0, sizeof(*cfg));
2263 cfg->fc_table = rtm->rtm_table;
2264 cfg->fc_dst_len = rtm->rtm_dst_len;
2265 cfg->fc_src_len = rtm->rtm_src_len;
2266 cfg->fc_flags = RTF_UP;
2267 cfg->fc_protocol = rtm->rtm_protocol;
2269 if (rtm->rtm_type == RTN_UNREACHABLE)
2270 cfg->fc_flags |= RTF_REJECT;
2272 if (rtm->rtm_type == RTN_LOCAL)
2273 cfg->fc_flags |= RTF_LOCAL;
2275 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2276 cfg->fc_nlinfo.nlh = nlh;
2277 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2279 if (tb[RTA_GATEWAY]) {
2280 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2281 cfg->fc_flags |= RTF_GATEWAY;
2285 int plen = (rtm->rtm_dst_len + 7) >> 3;
2287 if (nla_len(tb[RTA_DST]) < plen)
2290 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2294 int plen = (rtm->rtm_src_len + 7) >> 3;
2296 if (nla_len(tb[RTA_SRC]) < plen)
2299 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2302 if (tb[RTA_PREFSRC])
2303 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2306 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2308 if (tb[RTA_PRIORITY])
2309 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2311 if (tb[RTA_METRICS]) {
2312 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2313 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2317 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2324 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2326 struct fib6_config cfg;
2329 err = rtm_to_fib6_config(skb, nlh, &cfg);
2333 return ip6_route_del(&cfg);
2336 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2338 struct fib6_config cfg;
2341 err = rtm_to_fib6_config(skb, nlh, &cfg);
2345 return ip6_route_add(&cfg);
2348 static inline size_t rt6_nlmsg_size(void)
2350 return NLMSG_ALIGN(sizeof(struct rtmsg))
2351 + nla_total_size(16) /* RTA_SRC */
2352 + nla_total_size(16) /* RTA_DST */
2353 + nla_total_size(16) /* RTA_GATEWAY */
2354 + nla_total_size(16) /* RTA_PREFSRC */
2355 + nla_total_size(4) /* RTA_TABLE */
2356 + nla_total_size(4) /* RTA_IIF */
2357 + nla_total_size(4) /* RTA_OIF */
2358 + nla_total_size(4) /* RTA_PRIORITY */
2359 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2360 + nla_total_size(sizeof(struct rta_cacheinfo));
2363 static int rt6_fill_node(struct net *net,
2364 struct sk_buff *skb, struct rt6_info *rt,
2365 struct in6_addr *dst, struct in6_addr *src,
2366 int iif, int type, u32 pid, u32 seq,
2367 int prefix, int nowait, unsigned int flags)
2370 struct nlmsghdr *nlh;
2373 struct neighbour *n;
2375 if (prefix) { /* user wants prefix routes only */
2376 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2377 /* success since this is not a prefix route */
2382 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2386 rtm = nlmsg_data(nlh);
2387 rtm->rtm_family = AF_INET6;
2388 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2389 rtm->rtm_src_len = rt->rt6i_src.plen;
2392 table = rt->rt6i_table->tb6_id;
2394 table = RT6_TABLE_UNSPEC;
2395 rtm->rtm_table = table;
2396 NLA_PUT_U32(skb, RTA_TABLE, table);
2397 if (rt->rt6i_flags & RTF_REJECT)
2398 rtm->rtm_type = RTN_UNREACHABLE;
2399 else if (rt->rt6i_flags & RTF_LOCAL)
2400 rtm->rtm_type = RTN_LOCAL;
2401 else if (rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
2402 rtm->rtm_type = RTN_LOCAL;
2404 rtm->rtm_type = RTN_UNICAST;
2406 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2407 rtm->rtm_protocol = rt->rt6i_protocol;
2408 if (rt->rt6i_flags & RTF_DYNAMIC)
2409 rtm->rtm_protocol = RTPROT_REDIRECT;
2410 else if (rt->rt6i_flags & RTF_ADDRCONF)
2411 rtm->rtm_protocol = RTPROT_KERNEL;
2412 else if (rt->rt6i_flags & RTF_DEFAULT)
2413 rtm->rtm_protocol = RTPROT_RA;
2415 if (rt->rt6i_flags & RTF_CACHE)
2416 rtm->rtm_flags |= RTM_F_CLONED;
2419 NLA_PUT(skb, RTA_DST, 16, dst);
2420 rtm->rtm_dst_len = 128;
2421 } else if (rtm->rtm_dst_len)
2422 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2423 #ifdef CONFIG_IPV6_SUBTREES
2425 NLA_PUT(skb, RTA_SRC, 16, src);
2426 rtm->rtm_src_len = 128;
2427 } else if (rtm->rtm_src_len)
2428 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2431 #ifdef CONFIG_IPV6_MROUTE
2432 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2433 int err = ip6mr_get_route(net, skb, rtm, nowait);
2438 goto nla_put_failure;
2440 if (err == -EMSGSIZE)
2441 goto nla_put_failure;
2446 NLA_PUT_U32(skb, RTA_IIF, iif);
2448 struct in6_addr saddr_buf;
2449 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0)
2450 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2453 if (rt->rt6i_prefsrc.plen) {
2454 struct in6_addr saddr_buf;
2455 saddr_buf = rt->rt6i_prefsrc.addr;
2456 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2459 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2460 goto nla_put_failure;
2463 n = dst_get_neighbour_noref(&rt->dst);
2465 NLA_PUT(skb, RTA_GATEWAY, 16, &n->primary_key);
2469 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2471 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2473 if (!(rt->rt6i_flags & RTF_EXPIRES))
2475 else if (rt->rt6i_expires - jiffies < INT_MAX)
2476 expires = rt->rt6i_expires - jiffies;
2480 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2481 expires, rt->dst.error) < 0)
2482 goto nla_put_failure;
2484 return nlmsg_end(skb, nlh);
2487 nlmsg_cancel(skb, nlh);
2491 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2493 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2496 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2497 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2498 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2502 return rt6_fill_node(arg->net,
2503 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2504 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2505 prefix, 0, NLM_F_MULTI);
2508 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2510 struct net *net = sock_net(in_skb->sk);
2511 struct nlattr *tb[RTA_MAX+1];
2512 struct rt6_info *rt;
2513 struct sk_buff *skb;
2518 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2523 memset(&fl6, 0, sizeof(fl6));
2526 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2529 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2533 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2536 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2540 iif = nla_get_u32(tb[RTA_IIF]);
2543 fl6.flowi6_oif = nla_get_u32(tb[RTA_OIF]);
2546 struct net_device *dev;
2547 dev = __dev_get_by_index(net, iif);
2554 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2560 /* Reserve room for dummy headers, this skb can pass
2561 through good chunk of routing engine.
2563 skb_reset_mac_header(skb);
2564 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2566 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl6);
2567 skb_dst_set(skb, &rt->dst);
2569 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2570 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2571 nlh->nlmsg_seq, 0, 0, 0);
2577 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2582 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2584 struct sk_buff *skb;
2585 struct net *net = info->nl_net;
2590 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2592 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2596 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2597 event, info->pid, seq, 0, 0, 0);
2599 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2600 WARN_ON(err == -EMSGSIZE);
2604 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2605 info->nlh, gfp_any());
2609 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2612 static int ip6_route_dev_notify(struct notifier_block *this,
2613 unsigned long event, void *data)
2615 struct net_device *dev = (struct net_device *)data;
2616 struct net *net = dev_net(dev);
2618 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2619 net->ipv6.ip6_null_entry->dst.dev = dev;
2620 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2621 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2622 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2623 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2624 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2625 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2636 #ifdef CONFIG_PROC_FS
2647 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2649 struct seq_file *m = p_arg;
2650 struct neighbour *n;
2652 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2654 #ifdef CONFIG_IPV6_SUBTREES
2655 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2657 seq_puts(m, "00000000000000000000000000000000 00 ");
2660 n = dst_get_neighbour_noref(&rt->dst);
2662 seq_printf(m, "%pi6", n->primary_key);
2664 seq_puts(m, "00000000000000000000000000000000");
2667 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2668 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2669 rt->dst.__use, rt->rt6i_flags,
2670 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2674 static int ipv6_route_show(struct seq_file *m, void *v)
2676 struct net *net = (struct net *)m->private;
2677 fib6_clean_all(net, rt6_info_route, 0, m);
2681 static int ipv6_route_open(struct inode *inode, struct file *file)
2683 return single_open_net(inode, file, ipv6_route_show);
2686 static const struct file_operations ipv6_route_proc_fops = {
2687 .owner = THIS_MODULE,
2688 .open = ipv6_route_open,
2690 .llseek = seq_lseek,
2691 .release = single_release_net,
2694 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2696 struct net *net = (struct net *)seq->private;
2697 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2698 net->ipv6.rt6_stats->fib_nodes,
2699 net->ipv6.rt6_stats->fib_route_nodes,
2700 net->ipv6.rt6_stats->fib_rt_alloc,
2701 net->ipv6.rt6_stats->fib_rt_entries,
2702 net->ipv6.rt6_stats->fib_rt_cache,
2703 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2704 net->ipv6.rt6_stats->fib_discarded_routes);
2709 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2711 return single_open_net(inode, file, rt6_stats_seq_show);
2714 static const struct file_operations rt6_stats_seq_fops = {
2715 .owner = THIS_MODULE,
2716 .open = rt6_stats_seq_open,
2718 .llseek = seq_lseek,
2719 .release = single_release_net,
2721 #endif /* CONFIG_PROC_FS */
2723 #ifdef CONFIG_SYSCTL
2726 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2727 void __user *buffer, size_t *lenp, loff_t *ppos)
2734 net = (struct net *)ctl->extra1;
2735 delay = net->ipv6.sysctl.flush_delay;
2736 proc_dointvec(ctl, write, buffer, lenp, ppos);
2737 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2741 ctl_table ipv6_route_table_template[] = {
2743 .procname = "flush",
2744 .data = &init_net.ipv6.sysctl.flush_delay,
2745 .maxlen = sizeof(int),
2747 .proc_handler = ipv6_sysctl_rtcache_flush
2750 .procname = "gc_thresh",
2751 .data = &ip6_dst_ops_template.gc_thresh,
2752 .maxlen = sizeof(int),
2754 .proc_handler = proc_dointvec,
2757 .procname = "max_size",
2758 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2759 .maxlen = sizeof(int),
2761 .proc_handler = proc_dointvec,
2764 .procname = "gc_min_interval",
2765 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2766 .maxlen = sizeof(int),
2768 .proc_handler = proc_dointvec_jiffies,
2771 .procname = "gc_timeout",
2772 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2773 .maxlen = sizeof(int),
2775 .proc_handler = proc_dointvec_jiffies,
2778 .procname = "gc_interval",
2779 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2780 .maxlen = sizeof(int),
2782 .proc_handler = proc_dointvec_jiffies,
2785 .procname = "gc_elasticity",
2786 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2787 .maxlen = sizeof(int),
2789 .proc_handler = proc_dointvec,
2792 .procname = "mtu_expires",
2793 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2794 .maxlen = sizeof(int),
2796 .proc_handler = proc_dointvec_jiffies,
2799 .procname = "min_adv_mss",
2800 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2801 .maxlen = sizeof(int),
2803 .proc_handler = proc_dointvec,
2806 .procname = "gc_min_interval_ms",
2807 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2808 .maxlen = sizeof(int),
2810 .proc_handler = proc_dointvec_ms_jiffies,
2815 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2817 struct ctl_table *table;
2819 table = kmemdup(ipv6_route_table_template,
2820 sizeof(ipv6_route_table_template),
2824 table[0].data = &net->ipv6.sysctl.flush_delay;
2825 table[0].extra1 = net;
2826 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2827 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2828 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2829 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2830 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2831 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2832 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2833 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2834 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2841 static int __net_init ip6_route_net_init(struct net *net)
2845 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2846 sizeof(net->ipv6.ip6_dst_ops));
2848 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2849 goto out_ip6_dst_ops;
2851 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2852 sizeof(*net->ipv6.ip6_null_entry),
2854 if (!net->ipv6.ip6_null_entry)
2855 goto out_ip6_dst_entries;
2856 net->ipv6.ip6_null_entry->dst.path =
2857 (struct dst_entry *)net->ipv6.ip6_null_entry;
2858 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2859 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2860 ip6_template_metrics, true);
2862 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2863 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2864 sizeof(*net->ipv6.ip6_prohibit_entry),
2866 if (!net->ipv6.ip6_prohibit_entry)
2867 goto out_ip6_null_entry;
2868 net->ipv6.ip6_prohibit_entry->dst.path =
2869 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2870 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2871 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2872 ip6_template_metrics, true);
2874 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2875 sizeof(*net->ipv6.ip6_blk_hole_entry),
2877 if (!net->ipv6.ip6_blk_hole_entry)
2878 goto out_ip6_prohibit_entry;
2879 net->ipv6.ip6_blk_hole_entry->dst.path =
2880 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2881 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2882 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2883 ip6_template_metrics, true);
2886 net->ipv6.sysctl.flush_delay = 0;
2887 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2888 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2889 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2890 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2891 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2892 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2893 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2895 #ifdef CONFIG_PROC_FS
2896 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2897 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2899 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2905 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2906 out_ip6_prohibit_entry:
2907 kfree(net->ipv6.ip6_prohibit_entry);
2909 kfree(net->ipv6.ip6_null_entry);
2911 out_ip6_dst_entries:
2912 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2917 static void __net_exit ip6_route_net_exit(struct net *net)
2919 #ifdef CONFIG_PROC_FS
2920 proc_net_remove(net, "ipv6_route");
2921 proc_net_remove(net, "rt6_stats");
2923 kfree(net->ipv6.ip6_null_entry);
2924 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2925 kfree(net->ipv6.ip6_prohibit_entry);
2926 kfree(net->ipv6.ip6_blk_hole_entry);
2928 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2931 static struct pernet_operations ip6_route_net_ops = {
2932 .init = ip6_route_net_init,
2933 .exit = ip6_route_net_exit,
2936 static struct notifier_block ip6_route_dev_notifier = {
2937 .notifier_call = ip6_route_dev_notify,
2941 int __init ip6_route_init(void)
2946 ip6_dst_ops_template.kmem_cachep =
2947 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2948 SLAB_HWCACHE_ALIGN, NULL);
2949 if (!ip6_dst_ops_template.kmem_cachep)
2952 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2954 goto out_kmem_cache;
2956 ret = register_pernet_subsys(&ip6_route_net_ops);
2958 goto out_dst_entries;
2960 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2962 /* Registering of the loopback is done before this portion of code,
2963 * the loopback reference in rt6_info will not be taken, do it
2964 * manually for init_net */
2965 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2966 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2967 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2968 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2969 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2970 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2971 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2975 goto out_register_subsys;
2981 ret = fib6_rules_init();
2986 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
2987 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
2988 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
2989 goto fib6_rules_init;
2991 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2993 goto fib6_rules_init;
2999 fib6_rules_cleanup();
3004 out_register_subsys:
3005 unregister_pernet_subsys(&ip6_route_net_ops);
3007 dst_entries_destroy(&ip6_dst_blackhole_ops);
3009 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3013 void ip6_route_cleanup(void)
3015 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3016 fib6_rules_cleanup();
3019 unregister_pernet_subsys(&ip6_route_net_ops);
3020 dst_entries_destroy(&ip6_dst_blackhole_ops);
3021 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);