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 != NULL && idev->dev == dev) {
303 struct inet6_dev *loopback_idev =
304 in6_dev_get(loopback_dev);
305 if (loopback_idev != NULL) {
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 == NULL ||
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 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 = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
749 struct net *net = dev_net(rt->rt6i_dev);
750 int saved_rt_min_interval =
751 net->ipv6.sysctl.ip6_rt_gc_min_interval;
752 int saved_rt_elasticity =
753 net->ipv6.sysctl.ip6_rt_gc_elasticity;
755 if (attempts-- > 0) {
756 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
757 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
759 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
761 net->ipv6.sysctl.ip6_rt_gc_elasticity =
763 net->ipv6.sysctl.ip6_rt_gc_min_interval =
764 saved_rt_min_interval;
770 "ipv6: Neighbour table overflow.\n");
774 dst_set_neighbour(&rt->dst, neigh);
781 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
782 const struct in6_addr *daddr)
784 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
787 rt->rt6i_flags |= RTF_CACHE;
788 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_raw(&ort->dst)));
793 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
794 struct flowi6 *fl6, int flags)
796 struct fib6_node *fn;
797 struct rt6_info *rt, *nrt;
801 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
803 strict |= flags & RT6_LOOKUP_F_IFACE;
806 read_lock_bh(&table->tb6_lock);
809 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
812 rt = rt6_select(fn, oif, strict | reachable);
814 BACKTRACK(net, &fl6->saddr);
815 if (rt == net->ipv6.ip6_null_entry ||
816 rt->rt6i_flags & RTF_CACHE)
820 read_unlock_bh(&table->tb6_lock);
822 if (!dst_get_neighbour_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
823 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
824 else if (!(rt->dst.flags & DST_HOST))
825 nrt = rt6_alloc_clone(rt, &fl6->daddr);
829 dst_release(&rt->dst);
830 rt = nrt ? : net->ipv6.ip6_null_entry;
834 err = ip6_ins_rt(nrt);
843 * Race condition! In the gap, when table->tb6_lock was
844 * released someone could insert this route. Relookup.
846 dst_release(&rt->dst);
855 read_unlock_bh(&table->tb6_lock);
857 rt->dst.lastuse = jiffies;
863 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
864 struct flowi6 *fl6, int flags)
866 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
869 void ip6_route_input(struct sk_buff *skb)
871 const struct ipv6hdr *iph = ipv6_hdr(skb);
872 struct net *net = dev_net(skb->dev);
873 int flags = RT6_LOOKUP_F_HAS_SADDR;
874 struct flowi6 fl6 = {
875 .flowi6_iif = skb->dev->ifindex,
878 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
879 .flowi6_mark = skb->mark,
880 .flowi6_proto = iph->nexthdr,
883 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
884 flags |= RT6_LOOKUP_F_IFACE;
886 skb_dst_set(skb, fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_input));
889 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
890 struct flowi6 *fl6, int flags)
892 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
895 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
900 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
901 flags |= RT6_LOOKUP_F_IFACE;
903 if (!ipv6_addr_any(&fl6->saddr))
904 flags |= RT6_LOOKUP_F_HAS_SADDR;
906 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
908 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
911 EXPORT_SYMBOL(ip6_route_output);
913 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
915 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
916 struct dst_entry *new = NULL;
918 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
920 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
925 new->input = dst_discard;
926 new->output = dst_discard;
928 if (dst_metrics_read_only(&ort->dst))
929 new->_metrics = ort->dst._metrics;
931 dst_copy_metrics(new, &ort->dst);
932 rt->rt6i_idev = ort->rt6i_idev;
934 in6_dev_hold(rt->rt6i_idev);
935 rt->rt6i_expires = 0;
937 rt->rt6i_gateway = ort->rt6i_gateway;
938 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
941 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
942 #ifdef CONFIG_IPV6_SUBTREES
943 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
949 dst_release(dst_orig);
950 return new ? new : ERR_PTR(-ENOMEM);
954 * Destination cache support functions
957 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
961 rt = (struct rt6_info *) dst;
963 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
964 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
966 rt6_bind_peer(rt, 0);
967 rt->rt6i_peer_genid = rt6_peer_genid();
974 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
976 struct rt6_info *rt = (struct rt6_info *) dst;
979 if (rt->rt6i_flags & RTF_CACHE) {
980 if (rt6_check_expired(rt)) {
992 static void ip6_link_failure(struct sk_buff *skb)
996 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
998 rt = (struct rt6_info *) skb_dst(skb);
1000 if (rt->rt6i_flags&RTF_CACHE) {
1001 dst_set_expires(&rt->dst, 0);
1002 rt->rt6i_flags |= RTF_EXPIRES;
1003 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1004 rt->rt6i_node->fn_sernum = -1;
1008 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1010 struct rt6_info *rt6 = (struct rt6_info*)dst;
1012 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1013 rt6->rt6i_flags |= RTF_MODIFIED;
1014 if (mtu < IPV6_MIN_MTU) {
1015 u32 features = dst_metric(dst, RTAX_FEATURES);
1017 features |= RTAX_FEATURE_ALLFRAG;
1018 dst_metric_set(dst, RTAX_FEATURES, features);
1020 dst_metric_set(dst, RTAX_MTU, mtu);
1024 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1026 struct net_device *dev = dst->dev;
1027 unsigned int mtu = dst_mtu(dst);
1028 struct net *net = dev_net(dev);
1030 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1032 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1033 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1036 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1037 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1038 * IPV6_MAXPLEN is also valid and means: "any MSS,
1039 * rely only on pmtu discovery"
1041 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1046 static unsigned int ip6_mtu(const struct dst_entry *dst)
1048 struct inet6_dev *idev;
1049 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1057 idev = __in6_dev_get(dst->dev);
1059 mtu = idev->cnf.mtu6;
1065 static struct dst_entry *icmp6_dst_gc_list;
1066 static DEFINE_SPINLOCK(icmp6_dst_lock);
1068 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1069 struct neighbour *neigh,
1070 const struct in6_addr *addr)
1072 struct rt6_info *rt;
1073 struct inet6_dev *idev = in6_dev_get(dev);
1074 struct net *net = dev_net(dev);
1076 if (unlikely(idev == NULL))
1079 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1080 if (unlikely(rt == NULL)) {
1088 neigh = ndisc_get_neigh(dev, addr);
1093 rt->dst.flags |= DST_HOST;
1094 rt->dst.output = ip6_output;
1095 dst_set_neighbour(&rt->dst, neigh);
1096 atomic_set(&rt->dst.__refcnt, 1);
1097 rt->rt6i_dst.addr = *addr;
1098 rt->rt6i_dst.plen = 128;
1099 rt->rt6i_idev = idev;
1100 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1102 spin_lock_bh(&icmp6_dst_lock);
1103 rt->dst.next = icmp6_dst_gc_list;
1104 icmp6_dst_gc_list = &rt->dst;
1105 spin_unlock_bh(&icmp6_dst_lock);
1107 fib6_force_start_gc(net);
1113 int icmp6_dst_gc(void)
1115 struct dst_entry *dst, **pprev;
1118 spin_lock_bh(&icmp6_dst_lock);
1119 pprev = &icmp6_dst_gc_list;
1121 while ((dst = *pprev) != NULL) {
1122 if (!atomic_read(&dst->__refcnt)) {
1131 spin_unlock_bh(&icmp6_dst_lock);
1136 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1139 struct dst_entry *dst, **pprev;
1141 spin_lock_bh(&icmp6_dst_lock);
1142 pprev = &icmp6_dst_gc_list;
1143 while ((dst = *pprev) != NULL) {
1144 struct rt6_info *rt = (struct rt6_info *) dst;
1145 if (func(rt, arg)) {
1152 spin_unlock_bh(&icmp6_dst_lock);
1155 static int ip6_dst_gc(struct dst_ops *ops)
1157 unsigned long now = jiffies;
1158 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1159 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1160 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1161 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1162 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1163 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1166 entries = dst_entries_get_fast(ops);
1167 if (time_after(rt_last_gc + rt_min_interval, now) &&
1168 entries <= rt_max_size)
1171 net->ipv6.ip6_rt_gc_expire++;
1172 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1173 net->ipv6.ip6_rt_last_gc = now;
1174 entries = dst_entries_get_slow(ops);
1175 if (entries < ops->gc_thresh)
1176 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1178 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1179 return entries > rt_max_size;
1182 /* Clean host part of a prefix. Not necessary in radix tree,
1183 but results in cleaner routing tables.
1185 Remove it only when all the things will work!
1188 int ip6_dst_hoplimit(struct dst_entry *dst)
1190 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1191 if (hoplimit == 0) {
1192 struct net_device *dev = dst->dev;
1193 struct inet6_dev *idev;
1196 idev = __in6_dev_get(dev);
1198 hoplimit = idev->cnf.hop_limit;
1200 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1205 EXPORT_SYMBOL(ip6_dst_hoplimit);
1211 int ip6_route_add(struct fib6_config *cfg)
1214 struct net *net = cfg->fc_nlinfo.nl_net;
1215 struct rt6_info *rt = NULL;
1216 struct net_device *dev = NULL;
1217 struct inet6_dev *idev = NULL;
1218 struct fib6_table *table;
1221 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1223 #ifndef CONFIG_IPV6_SUBTREES
1224 if (cfg->fc_src_len)
1227 if (cfg->fc_ifindex) {
1229 dev = dev_get_by_index(net, cfg->fc_ifindex);
1232 idev = in6_dev_get(dev);
1237 if (cfg->fc_metric == 0)
1238 cfg->fc_metric = IP6_RT_PRIO_USER;
1241 if (NULL != cfg->fc_nlinfo.nlh &&
1242 !(cfg->fc_nlinfo.nlh->nlmsg_flags&NLM_F_CREATE)) {
1243 table = fib6_get_table(net, cfg->fc_table);
1244 if (table == NULL) {
1245 printk(KERN_WARNING "IPv6: NLM_F_CREATE should be specified when creating new route\n");
1246 table = fib6_new_table(net, cfg->fc_table);
1249 table = fib6_new_table(net, cfg->fc_table);
1251 if (table == NULL) {
1255 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1262 rt->dst.obsolete = -1;
1263 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1264 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1267 if (cfg->fc_protocol == RTPROT_UNSPEC)
1268 cfg->fc_protocol = RTPROT_BOOT;
1269 rt->rt6i_protocol = cfg->fc_protocol;
1271 addr_type = ipv6_addr_type(&cfg->fc_dst);
1273 if (addr_type & IPV6_ADDR_MULTICAST)
1274 rt->dst.input = ip6_mc_input;
1275 else if (cfg->fc_flags & RTF_LOCAL)
1276 rt->dst.input = ip6_input;
1278 rt->dst.input = ip6_forward;
1280 rt->dst.output = ip6_output;
1282 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1283 rt->rt6i_dst.plen = cfg->fc_dst_len;
1284 if (rt->rt6i_dst.plen == 128)
1285 rt->dst.flags |= DST_HOST;
1287 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1288 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1293 dst_init_metrics(&rt->dst, metrics, 0);
1295 #ifdef CONFIG_IPV6_SUBTREES
1296 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1297 rt->rt6i_src.plen = cfg->fc_src_len;
1300 rt->rt6i_metric = cfg->fc_metric;
1302 /* We cannot add true routes via loopback here,
1303 they would result in kernel looping; promote them to reject routes
1305 if ((cfg->fc_flags & RTF_REJECT) ||
1306 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK)
1307 && !(cfg->fc_flags&RTF_LOCAL))) {
1308 /* hold loopback dev/idev if we haven't done so. */
1309 if (dev != net->loopback_dev) {
1314 dev = net->loopback_dev;
1316 idev = in6_dev_get(dev);
1322 rt->dst.output = ip6_pkt_discard_out;
1323 rt->dst.input = ip6_pkt_discard;
1324 rt->dst.error = -ENETUNREACH;
1325 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1329 if (cfg->fc_flags & RTF_GATEWAY) {
1330 const struct in6_addr *gw_addr;
1333 gw_addr = &cfg->fc_gateway;
1334 rt->rt6i_gateway = *gw_addr;
1335 gwa_type = ipv6_addr_type(gw_addr);
1337 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1338 struct rt6_info *grt;
1340 /* IPv6 strictly inhibits using not link-local
1341 addresses as nexthop address.
1342 Otherwise, router will not able to send redirects.
1343 It is very good, but in some (rare!) circumstances
1344 (SIT, PtP, NBMA NOARP links) it is handy to allow
1345 some exceptions. --ANK
1348 if (!(gwa_type&IPV6_ADDR_UNICAST))
1351 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1353 err = -EHOSTUNREACH;
1357 if (dev != grt->rt6i_dev) {
1358 dst_release(&grt->dst);
1362 dev = grt->rt6i_dev;
1363 idev = grt->rt6i_idev;
1365 in6_dev_hold(grt->rt6i_idev);
1367 if (!(grt->rt6i_flags&RTF_GATEWAY))
1369 dst_release(&grt->dst);
1375 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1383 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1384 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1388 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1389 rt->rt6i_prefsrc.plen = 128;
1391 rt->rt6i_prefsrc.plen = 0;
1393 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1394 struct neighbour *n = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1399 dst_set_neighbour(&rt->dst, n);
1402 rt->rt6i_flags = cfg->fc_flags;
1409 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1410 int type = nla_type(nla);
1413 if (type > RTAX_MAX) {
1418 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1424 rt->rt6i_idev = idev;
1425 rt->rt6i_table = table;
1427 cfg->fc_nlinfo.nl_net = dev_net(dev);
1429 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1441 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1444 struct fib6_table *table;
1445 struct net *net = dev_net(rt->rt6i_dev);
1447 if (rt == net->ipv6.ip6_null_entry)
1450 table = rt->rt6i_table;
1451 write_lock_bh(&table->tb6_lock);
1453 err = fib6_del(rt, info);
1454 dst_release(&rt->dst);
1456 write_unlock_bh(&table->tb6_lock);
1461 int ip6_del_rt(struct rt6_info *rt)
1463 struct nl_info info = {
1464 .nl_net = dev_net(rt->rt6i_dev),
1466 return __ip6_del_rt(rt, &info);
1469 static int ip6_route_del(struct fib6_config *cfg)
1471 struct fib6_table *table;
1472 struct fib6_node *fn;
1473 struct rt6_info *rt;
1476 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1480 read_lock_bh(&table->tb6_lock);
1482 fn = fib6_locate(&table->tb6_root,
1483 &cfg->fc_dst, cfg->fc_dst_len,
1484 &cfg->fc_src, cfg->fc_src_len);
1487 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1488 if (cfg->fc_ifindex &&
1489 (rt->rt6i_dev == NULL ||
1490 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1492 if (cfg->fc_flags & RTF_GATEWAY &&
1493 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1495 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1498 read_unlock_bh(&table->tb6_lock);
1500 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1503 read_unlock_bh(&table->tb6_lock);
1511 struct ip6rd_flowi {
1513 struct in6_addr gateway;
1516 static struct rt6_info *__ip6_route_redirect(struct net *net,
1517 struct fib6_table *table,
1521 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1522 struct rt6_info *rt;
1523 struct fib6_node *fn;
1526 * Get the "current" route for this destination and
1527 * check if the redirect has come from approriate router.
1529 * RFC 2461 specifies that redirects should only be
1530 * accepted if they come from the nexthop to the target.
1531 * Due to the way the routes are chosen, this notion
1532 * is a bit fuzzy and one might need to check all possible
1536 read_lock_bh(&table->tb6_lock);
1537 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1539 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1541 * Current route is on-link; redirect is always invalid.
1543 * Seems, previous statement is not true. It could
1544 * be node, which looks for us as on-link (f.e. proxy ndisc)
1545 * But then router serving it might decide, that we should
1546 * know truth 8)8) --ANK (980726).
1548 if (rt6_check_expired(rt))
1550 if (!(rt->rt6i_flags & RTF_GATEWAY))
1552 if (fl6->flowi6_oif != rt->rt6i_dev->ifindex)
1554 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1560 rt = net->ipv6.ip6_null_entry;
1561 BACKTRACK(net, &fl6->saddr);
1565 read_unlock_bh(&table->tb6_lock);
1570 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1571 const struct in6_addr *src,
1572 const struct in6_addr *gateway,
1573 struct net_device *dev)
1575 int flags = RT6_LOOKUP_F_HAS_SADDR;
1576 struct net *net = dev_net(dev);
1577 struct ip6rd_flowi rdfl = {
1579 .flowi6_oif = dev->ifindex,
1585 rdfl.gateway = *gateway;
1587 if (rt6_need_strict(dest))
1588 flags |= RT6_LOOKUP_F_IFACE;
1590 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1591 flags, __ip6_route_redirect);
1594 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1595 const struct in6_addr *saddr,
1596 struct neighbour *neigh, u8 *lladdr, int on_link)
1598 struct rt6_info *rt, *nrt = NULL;
1599 struct netevent_redirect netevent;
1600 struct net *net = dev_net(neigh->dev);
1602 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1604 if (rt == net->ipv6.ip6_null_entry) {
1605 if (net_ratelimit())
1606 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1607 "for redirect target\n");
1612 * We have finally decided to accept it.
1615 neigh_update(neigh, lladdr, NUD_STALE,
1616 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1617 NEIGH_UPDATE_F_OVERRIDE|
1618 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1619 NEIGH_UPDATE_F_ISROUTER))
1623 * Redirect received -> path was valid.
1624 * Look, redirects are sent only in response to data packets,
1625 * so that this nexthop apparently is reachable. --ANK
1627 dst_confirm(&rt->dst);
1629 /* Duplicate redirect: silently ignore. */
1630 if (neigh == dst_get_neighbour_raw(&rt->dst))
1633 nrt = ip6_rt_copy(rt, dest);
1637 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1639 nrt->rt6i_flags &= ~RTF_GATEWAY;
1641 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1642 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1644 if (ip6_ins_rt(nrt))
1647 netevent.old = &rt->dst;
1648 netevent.new = &nrt->dst;
1649 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1651 if (rt->rt6i_flags&RTF_CACHE) {
1657 dst_release(&rt->dst);
1661 * Handle ICMP "packet too big" messages
1662 * i.e. Path MTU discovery
1665 static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1666 struct net *net, u32 pmtu, int ifindex)
1668 struct rt6_info *rt, *nrt;
1671 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1675 if (rt6_check_expired(rt)) {
1680 if (pmtu >= dst_mtu(&rt->dst))
1683 if (pmtu < IPV6_MIN_MTU) {
1685 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1686 * MTU (1280) and a fragment header should always be included
1687 * after a node receiving Too Big message reporting PMTU is
1688 * less than the IPv6 Minimum Link MTU.
1690 pmtu = IPV6_MIN_MTU;
1694 /* New mtu received -> path was valid.
1695 They are sent only in response to data packets,
1696 so that this nexthop apparently is reachable. --ANK
1698 dst_confirm(&rt->dst);
1700 /* Host route. If it is static, it would be better
1701 not to override it, but add new one, so that
1702 when cache entry will expire old pmtu
1703 would return automatically.
1705 if (rt->rt6i_flags & RTF_CACHE) {
1706 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1708 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1709 features |= RTAX_FEATURE_ALLFRAG;
1710 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1712 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1713 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1718 Two cases are possible:
1719 1. It is connected route. Action: COW
1720 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1722 if (!dst_get_neighbour_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1723 nrt = rt6_alloc_cow(rt, daddr, saddr);
1725 nrt = rt6_alloc_clone(rt, daddr);
1728 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1730 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1731 features |= RTAX_FEATURE_ALLFRAG;
1732 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1735 /* According to RFC 1981, detecting PMTU increase shouldn't be
1736 * happened within 5 mins, the recommended timer is 10 mins.
1737 * Here this route expiration time is set to ip6_rt_mtu_expires
1738 * which is 10 mins. After 10 mins the decreased pmtu is expired
1739 * and detecting PMTU increase will be automatically happened.
1741 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1742 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1747 dst_release(&rt->dst);
1750 void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1751 struct net_device *dev, u32 pmtu)
1753 struct net *net = dev_net(dev);
1756 * RFC 1981 states that a node "MUST reduce the size of the packets it
1757 * is sending along the path" that caused the Packet Too Big message.
1758 * Since it's not possible in the general case to determine which
1759 * interface was used to send the original packet, we update the MTU
1760 * on the interface that will be used to send future packets. We also
1761 * update the MTU on the interface that received the Packet Too Big in
1762 * case the original packet was forced out that interface with
1763 * SO_BINDTODEVICE or similar. This is the next best thing to the
1764 * correct behaviour, which would be to update the MTU on all
1767 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1768 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1772 * Misc support functions
1775 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
1776 const struct in6_addr *dest)
1778 struct net *net = dev_net(ort->rt6i_dev);
1779 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1783 rt->dst.input = ort->dst.input;
1784 rt->dst.output = ort->dst.output;
1785 rt->dst.flags |= DST_HOST;
1787 rt->rt6i_dst.addr = *dest;
1788 rt->rt6i_dst.plen = 128;
1789 dst_copy_metrics(&rt->dst, &ort->dst);
1790 rt->dst.error = ort->dst.error;
1791 rt->rt6i_idev = ort->rt6i_idev;
1793 in6_dev_hold(rt->rt6i_idev);
1794 rt->dst.lastuse = jiffies;
1795 rt->rt6i_expires = 0;
1797 rt->rt6i_gateway = ort->rt6i_gateway;
1798 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1799 rt->rt6i_metric = 0;
1801 #ifdef CONFIG_IPV6_SUBTREES
1802 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1804 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1805 rt->rt6i_table = ort->rt6i_table;
1810 #ifdef CONFIG_IPV6_ROUTE_INFO
1811 static struct rt6_info *rt6_get_route_info(struct net *net,
1812 const struct in6_addr *prefix, int prefixlen,
1813 const struct in6_addr *gwaddr, int ifindex)
1815 struct fib6_node *fn;
1816 struct rt6_info *rt = NULL;
1817 struct fib6_table *table;
1819 table = fib6_get_table(net, RT6_TABLE_INFO);
1823 write_lock_bh(&table->tb6_lock);
1824 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1828 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1829 if (rt->rt6i_dev->ifindex != ifindex)
1831 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1833 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1839 write_unlock_bh(&table->tb6_lock);
1843 static struct rt6_info *rt6_add_route_info(struct net *net,
1844 const struct in6_addr *prefix, int prefixlen,
1845 const struct in6_addr *gwaddr, int ifindex,
1848 struct fib6_config cfg = {
1849 .fc_table = RT6_TABLE_INFO,
1850 .fc_metric = IP6_RT_PRIO_USER,
1851 .fc_ifindex = ifindex,
1852 .fc_dst_len = prefixlen,
1853 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1854 RTF_UP | RTF_PREF(pref),
1856 .fc_nlinfo.nlh = NULL,
1857 .fc_nlinfo.nl_net = net,
1860 cfg.fc_dst = *prefix;
1861 cfg.fc_gateway = *gwaddr;
1863 /* We should treat it as a default route if prefix length is 0. */
1865 cfg.fc_flags |= RTF_DEFAULT;
1867 ip6_route_add(&cfg);
1869 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1873 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1875 struct rt6_info *rt;
1876 struct fib6_table *table;
1878 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1882 write_lock_bh(&table->tb6_lock);
1883 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1884 if (dev == rt->rt6i_dev &&
1885 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1886 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1891 write_unlock_bh(&table->tb6_lock);
1895 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1896 struct net_device *dev,
1899 struct fib6_config cfg = {
1900 .fc_table = RT6_TABLE_DFLT,
1901 .fc_metric = IP6_RT_PRIO_USER,
1902 .fc_ifindex = dev->ifindex,
1903 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1904 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1906 .fc_nlinfo.nlh = NULL,
1907 .fc_nlinfo.nl_net = dev_net(dev),
1910 cfg.fc_gateway = *gwaddr;
1912 ip6_route_add(&cfg);
1914 return rt6_get_dflt_router(gwaddr, dev);
1917 void rt6_purge_dflt_routers(struct net *net)
1919 struct rt6_info *rt;
1920 struct fib6_table *table;
1922 /* NOTE: Keep consistent with rt6_get_dflt_router */
1923 table = fib6_get_table(net, RT6_TABLE_DFLT);
1928 read_lock_bh(&table->tb6_lock);
1929 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1930 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1932 read_unlock_bh(&table->tb6_lock);
1937 read_unlock_bh(&table->tb6_lock);
1940 static void rtmsg_to_fib6_config(struct net *net,
1941 struct in6_rtmsg *rtmsg,
1942 struct fib6_config *cfg)
1944 memset(cfg, 0, sizeof(*cfg));
1946 cfg->fc_table = RT6_TABLE_MAIN;
1947 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1948 cfg->fc_metric = rtmsg->rtmsg_metric;
1949 cfg->fc_expires = rtmsg->rtmsg_info;
1950 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1951 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1952 cfg->fc_flags = rtmsg->rtmsg_flags;
1954 cfg->fc_nlinfo.nl_net = net;
1956 cfg->fc_dst = rtmsg->rtmsg_dst;
1957 cfg->fc_src = rtmsg->rtmsg_src;
1958 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1961 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1963 struct fib6_config cfg;
1964 struct in6_rtmsg rtmsg;
1968 case SIOCADDRT: /* Add a route */
1969 case SIOCDELRT: /* Delete a route */
1970 if (!capable(CAP_NET_ADMIN))
1972 err = copy_from_user(&rtmsg, arg,
1973 sizeof(struct in6_rtmsg));
1977 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1982 err = ip6_route_add(&cfg);
1985 err = ip6_route_del(&cfg);
1999 * Drop the packet on the floor
2002 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2005 struct dst_entry *dst = skb_dst(skb);
2006 switch (ipstats_mib_noroutes) {
2007 case IPSTATS_MIB_INNOROUTES:
2008 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2009 if (type == IPV6_ADDR_ANY) {
2010 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2011 IPSTATS_MIB_INADDRERRORS);
2015 case IPSTATS_MIB_OUTNOROUTES:
2016 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2017 ipstats_mib_noroutes);
2020 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2025 static int ip6_pkt_discard(struct sk_buff *skb)
2027 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2030 static int ip6_pkt_discard_out(struct sk_buff *skb)
2032 skb->dev = skb_dst(skb)->dev;
2033 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2036 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2038 static int ip6_pkt_prohibit(struct sk_buff *skb)
2040 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2043 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2045 skb->dev = skb_dst(skb)->dev;
2046 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2052 * Allocate a dst for local (unicast / anycast) address.
2055 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2056 const struct in6_addr *addr,
2059 struct net *net = dev_net(idev->dev);
2060 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2061 net->loopback_dev, 0);
2062 struct neighbour *neigh;
2065 if (net_ratelimit())
2066 pr_warning("IPv6: Maximum number of routes reached,"
2067 " consider increasing route/max_size.\n");
2068 return ERR_PTR(-ENOMEM);
2073 rt->dst.flags |= DST_HOST;
2074 rt->dst.input = ip6_input;
2075 rt->dst.output = ip6_output;
2076 rt->rt6i_idev = idev;
2077 rt->dst.obsolete = -1;
2079 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2081 rt->rt6i_flags |= RTF_ANYCAST;
2083 rt->rt6i_flags |= RTF_LOCAL;
2084 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
2085 if (IS_ERR(neigh)) {
2088 return ERR_CAST(neigh);
2090 dst_set_neighbour(&rt->dst, neigh);
2092 rt->rt6i_dst.addr = *addr;
2093 rt->rt6i_dst.plen = 128;
2094 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2096 atomic_set(&rt->dst.__refcnt, 1);
2101 int ip6_route_get_saddr(struct net *net,
2102 struct rt6_info *rt,
2103 const struct in6_addr *daddr,
2105 struct in6_addr *saddr)
2107 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2109 if (rt->rt6i_prefsrc.plen)
2110 *saddr = rt->rt6i_prefsrc.addr;
2112 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2113 daddr, prefs, saddr);
2117 /* remove deleted ip from prefsrc entries */
2118 struct arg_dev_net_ip {
2119 struct net_device *dev;
2121 struct in6_addr *addr;
2124 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2126 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2127 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2128 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2130 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
2131 rt != net->ipv6.ip6_null_entry &&
2132 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2133 /* remove prefsrc entry */
2134 rt->rt6i_prefsrc.plen = 0;
2139 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2141 struct net *net = dev_net(ifp->idev->dev);
2142 struct arg_dev_net_ip adni = {
2143 .dev = ifp->idev->dev,
2147 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2150 struct arg_dev_net {
2151 struct net_device *dev;
2155 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2157 const struct arg_dev_net *adn = arg;
2158 const struct net_device *dev = adn->dev;
2160 if ((rt->rt6i_dev == dev || dev == NULL) &&
2161 rt != adn->net->ipv6.ip6_null_entry) {
2162 RT6_TRACE("deleted by ifdown %p\n", rt);
2168 void rt6_ifdown(struct net *net, struct net_device *dev)
2170 struct arg_dev_net adn = {
2175 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2176 icmp6_clean_all(fib6_ifdown, &adn);
2179 struct rt6_mtu_change_arg
2181 struct net_device *dev;
2185 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2187 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2188 struct inet6_dev *idev;
2190 /* In IPv6 pmtu discovery is not optional,
2191 so that RTAX_MTU lock cannot disable it.
2192 We still use this lock to block changes
2193 caused by addrconf/ndisc.
2196 idev = __in6_dev_get(arg->dev);
2200 /* For administrative MTU increase, there is no way to discover
2201 IPv6 PMTU increase, so PMTU increase should be updated here.
2202 Since RFC 1981 doesn't include administrative MTU increase
2203 update PMTU increase is a MUST. (i.e. jumbo frame)
2206 If new MTU is less than route PMTU, this new MTU will be the
2207 lowest MTU in the path, update the route PMTU to reflect PMTU
2208 decreases; if new MTU is greater than route PMTU, and the
2209 old MTU is the lowest MTU in the path, update the route PMTU
2210 to reflect the increase. In this case if the other nodes' MTU
2211 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2214 if (rt->rt6i_dev == arg->dev &&
2215 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2216 (dst_mtu(&rt->dst) >= arg->mtu ||
2217 (dst_mtu(&rt->dst) < arg->mtu &&
2218 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2219 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2224 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2226 struct rt6_mtu_change_arg arg = {
2231 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2234 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2235 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2236 [RTA_OIF] = { .type = NLA_U32 },
2237 [RTA_IIF] = { .type = NLA_U32 },
2238 [RTA_PRIORITY] = { .type = NLA_U32 },
2239 [RTA_METRICS] = { .type = NLA_NESTED },
2242 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2243 struct fib6_config *cfg)
2246 struct nlattr *tb[RTA_MAX+1];
2249 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2254 rtm = nlmsg_data(nlh);
2255 memset(cfg, 0, sizeof(*cfg));
2257 cfg->fc_table = rtm->rtm_table;
2258 cfg->fc_dst_len = rtm->rtm_dst_len;
2259 cfg->fc_src_len = rtm->rtm_src_len;
2260 cfg->fc_flags = RTF_UP;
2261 cfg->fc_protocol = rtm->rtm_protocol;
2263 if (rtm->rtm_type == RTN_UNREACHABLE)
2264 cfg->fc_flags |= RTF_REJECT;
2266 if (rtm->rtm_type == RTN_LOCAL)
2267 cfg->fc_flags |= RTF_LOCAL;
2269 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2270 cfg->fc_nlinfo.nlh = nlh;
2271 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2273 if (tb[RTA_GATEWAY]) {
2274 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2275 cfg->fc_flags |= RTF_GATEWAY;
2279 int plen = (rtm->rtm_dst_len + 7) >> 3;
2281 if (nla_len(tb[RTA_DST]) < plen)
2284 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2288 int plen = (rtm->rtm_src_len + 7) >> 3;
2290 if (nla_len(tb[RTA_SRC]) < plen)
2293 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2296 if (tb[RTA_PREFSRC])
2297 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2300 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2302 if (tb[RTA_PRIORITY])
2303 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2305 if (tb[RTA_METRICS]) {
2306 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2307 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2311 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2318 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2320 struct fib6_config cfg;
2323 err = rtm_to_fib6_config(skb, nlh, &cfg);
2327 return ip6_route_del(&cfg);
2330 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2332 struct fib6_config cfg;
2335 err = rtm_to_fib6_config(skb, nlh, &cfg);
2339 return ip6_route_add(&cfg);
2342 static inline size_t rt6_nlmsg_size(void)
2344 return NLMSG_ALIGN(sizeof(struct rtmsg))
2345 + nla_total_size(16) /* RTA_SRC */
2346 + nla_total_size(16) /* RTA_DST */
2347 + nla_total_size(16) /* RTA_GATEWAY */
2348 + nla_total_size(16) /* RTA_PREFSRC */
2349 + nla_total_size(4) /* RTA_TABLE */
2350 + nla_total_size(4) /* RTA_IIF */
2351 + nla_total_size(4) /* RTA_OIF */
2352 + nla_total_size(4) /* RTA_PRIORITY */
2353 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2354 + nla_total_size(sizeof(struct rta_cacheinfo));
2357 static int rt6_fill_node(struct net *net,
2358 struct sk_buff *skb, struct rt6_info *rt,
2359 struct in6_addr *dst, struct in6_addr *src,
2360 int iif, int type, u32 pid, u32 seq,
2361 int prefix, int nowait, unsigned int flags)
2364 struct nlmsghdr *nlh;
2367 struct neighbour *n;
2369 if (prefix) { /* user wants prefix routes only */
2370 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2371 /* success since this is not a prefix route */
2376 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2380 rtm = nlmsg_data(nlh);
2381 rtm->rtm_family = AF_INET6;
2382 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2383 rtm->rtm_src_len = rt->rt6i_src.plen;
2386 table = rt->rt6i_table->tb6_id;
2388 table = RT6_TABLE_UNSPEC;
2389 rtm->rtm_table = table;
2390 NLA_PUT_U32(skb, RTA_TABLE, table);
2391 if (rt->rt6i_flags&RTF_REJECT)
2392 rtm->rtm_type = RTN_UNREACHABLE;
2393 else if (rt->rt6i_flags&RTF_LOCAL)
2394 rtm->rtm_type = RTN_LOCAL;
2395 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2396 rtm->rtm_type = RTN_LOCAL;
2398 rtm->rtm_type = RTN_UNICAST;
2400 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2401 rtm->rtm_protocol = rt->rt6i_protocol;
2402 if (rt->rt6i_flags&RTF_DYNAMIC)
2403 rtm->rtm_protocol = RTPROT_REDIRECT;
2404 else if (rt->rt6i_flags & RTF_ADDRCONF)
2405 rtm->rtm_protocol = RTPROT_KERNEL;
2406 else if (rt->rt6i_flags&RTF_DEFAULT)
2407 rtm->rtm_protocol = RTPROT_RA;
2409 if (rt->rt6i_flags&RTF_CACHE)
2410 rtm->rtm_flags |= RTM_F_CLONED;
2413 NLA_PUT(skb, RTA_DST, 16, dst);
2414 rtm->rtm_dst_len = 128;
2415 } else if (rtm->rtm_dst_len)
2416 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2417 #ifdef CONFIG_IPV6_SUBTREES
2419 NLA_PUT(skb, RTA_SRC, 16, src);
2420 rtm->rtm_src_len = 128;
2421 } else if (rtm->rtm_src_len)
2422 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2425 #ifdef CONFIG_IPV6_MROUTE
2426 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2427 int err = ip6mr_get_route(net, skb, rtm, nowait);
2432 goto nla_put_failure;
2434 if (err == -EMSGSIZE)
2435 goto nla_put_failure;
2440 NLA_PUT_U32(skb, RTA_IIF, iif);
2442 struct in6_addr saddr_buf;
2443 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0)
2444 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2447 if (rt->rt6i_prefsrc.plen) {
2448 struct in6_addr saddr_buf;
2449 saddr_buf = rt->rt6i_prefsrc.addr;
2450 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2453 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2454 goto nla_put_failure;
2457 n = dst_get_neighbour(&rt->dst);
2459 NLA_PUT(skb, RTA_GATEWAY, 16, &n->primary_key);
2463 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2465 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2467 if (!(rt->rt6i_flags & RTF_EXPIRES))
2469 else if (rt->rt6i_expires - jiffies < INT_MAX)
2470 expires = rt->rt6i_expires - jiffies;
2474 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2475 expires, rt->dst.error) < 0)
2476 goto nla_put_failure;
2478 return nlmsg_end(skb, nlh);
2481 nlmsg_cancel(skb, nlh);
2485 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2487 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2490 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2491 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2492 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2496 return rt6_fill_node(arg->net,
2497 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2498 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2499 prefix, 0, NLM_F_MULTI);
2502 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2504 struct net *net = sock_net(in_skb->sk);
2505 struct nlattr *tb[RTA_MAX+1];
2506 struct rt6_info *rt;
2507 struct sk_buff *skb;
2512 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2517 memset(&fl6, 0, sizeof(fl6));
2520 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2523 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2527 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2530 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2534 iif = nla_get_u32(tb[RTA_IIF]);
2537 fl6.flowi6_oif = nla_get_u32(tb[RTA_OIF]);
2540 struct net_device *dev;
2541 dev = __dev_get_by_index(net, iif);
2548 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2554 /* Reserve room for dummy headers, this skb can pass
2555 through good chunk of routing engine.
2557 skb_reset_mac_header(skb);
2558 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2560 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl6);
2561 skb_dst_set(skb, &rt->dst);
2563 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2564 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2565 nlh->nlmsg_seq, 0, 0, 0);
2571 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2576 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2578 struct sk_buff *skb;
2579 struct net *net = info->nl_net;
2584 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2586 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2590 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2591 event, info->pid, seq, 0, 0, 0);
2593 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2594 WARN_ON(err == -EMSGSIZE);
2598 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2599 info->nlh, gfp_any());
2603 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2606 static int ip6_route_dev_notify(struct notifier_block *this,
2607 unsigned long event, void *data)
2609 struct net_device *dev = (struct net_device *)data;
2610 struct net *net = dev_net(dev);
2612 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2613 net->ipv6.ip6_null_entry->dst.dev = dev;
2614 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2615 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2616 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2617 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2618 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2619 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2630 #ifdef CONFIG_PROC_FS
2641 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2643 struct seq_file *m = p_arg;
2644 struct neighbour *n;
2646 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2648 #ifdef CONFIG_IPV6_SUBTREES
2649 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2651 seq_puts(m, "00000000000000000000000000000000 00 ");
2654 n = dst_get_neighbour(&rt->dst);
2656 seq_printf(m, "%pi6", n->primary_key);
2658 seq_puts(m, "00000000000000000000000000000000");
2661 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2662 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2663 rt->dst.__use, rt->rt6i_flags,
2664 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2668 static int ipv6_route_show(struct seq_file *m, void *v)
2670 struct net *net = (struct net *)m->private;
2671 fib6_clean_all(net, rt6_info_route, 0, m);
2675 static int ipv6_route_open(struct inode *inode, struct file *file)
2677 return single_open_net(inode, file, ipv6_route_show);
2680 static const struct file_operations ipv6_route_proc_fops = {
2681 .owner = THIS_MODULE,
2682 .open = ipv6_route_open,
2684 .llseek = seq_lseek,
2685 .release = single_release_net,
2688 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2690 struct net *net = (struct net *)seq->private;
2691 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2692 net->ipv6.rt6_stats->fib_nodes,
2693 net->ipv6.rt6_stats->fib_route_nodes,
2694 net->ipv6.rt6_stats->fib_rt_alloc,
2695 net->ipv6.rt6_stats->fib_rt_entries,
2696 net->ipv6.rt6_stats->fib_rt_cache,
2697 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2698 net->ipv6.rt6_stats->fib_discarded_routes);
2703 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2705 return single_open_net(inode, file, rt6_stats_seq_show);
2708 static const struct file_operations rt6_stats_seq_fops = {
2709 .owner = THIS_MODULE,
2710 .open = rt6_stats_seq_open,
2712 .llseek = seq_lseek,
2713 .release = single_release_net,
2715 #endif /* CONFIG_PROC_FS */
2717 #ifdef CONFIG_SYSCTL
2720 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2721 void __user *buffer, size_t *lenp, loff_t *ppos)
2728 net = (struct net *)ctl->extra1;
2729 delay = net->ipv6.sysctl.flush_delay;
2730 proc_dointvec(ctl, write, buffer, lenp, ppos);
2731 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2735 ctl_table ipv6_route_table_template[] = {
2737 .procname = "flush",
2738 .data = &init_net.ipv6.sysctl.flush_delay,
2739 .maxlen = sizeof(int),
2741 .proc_handler = ipv6_sysctl_rtcache_flush
2744 .procname = "gc_thresh",
2745 .data = &ip6_dst_ops_template.gc_thresh,
2746 .maxlen = sizeof(int),
2748 .proc_handler = proc_dointvec,
2751 .procname = "max_size",
2752 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2753 .maxlen = sizeof(int),
2755 .proc_handler = proc_dointvec,
2758 .procname = "gc_min_interval",
2759 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2760 .maxlen = sizeof(int),
2762 .proc_handler = proc_dointvec_jiffies,
2765 .procname = "gc_timeout",
2766 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2767 .maxlen = sizeof(int),
2769 .proc_handler = proc_dointvec_jiffies,
2772 .procname = "gc_interval",
2773 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2774 .maxlen = sizeof(int),
2776 .proc_handler = proc_dointvec_jiffies,
2779 .procname = "gc_elasticity",
2780 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2781 .maxlen = sizeof(int),
2783 .proc_handler = proc_dointvec,
2786 .procname = "mtu_expires",
2787 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2788 .maxlen = sizeof(int),
2790 .proc_handler = proc_dointvec_jiffies,
2793 .procname = "min_adv_mss",
2794 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2795 .maxlen = sizeof(int),
2797 .proc_handler = proc_dointvec,
2800 .procname = "gc_min_interval_ms",
2801 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2802 .maxlen = sizeof(int),
2804 .proc_handler = proc_dointvec_ms_jiffies,
2809 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2811 struct ctl_table *table;
2813 table = kmemdup(ipv6_route_table_template,
2814 sizeof(ipv6_route_table_template),
2818 table[0].data = &net->ipv6.sysctl.flush_delay;
2819 table[0].extra1 = net;
2820 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2821 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2822 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2823 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2824 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2825 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2826 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2827 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2828 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2835 static int __net_init ip6_route_net_init(struct net *net)
2839 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2840 sizeof(net->ipv6.ip6_dst_ops));
2842 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2843 goto out_ip6_dst_ops;
2845 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2846 sizeof(*net->ipv6.ip6_null_entry),
2848 if (!net->ipv6.ip6_null_entry)
2849 goto out_ip6_dst_entries;
2850 net->ipv6.ip6_null_entry->dst.path =
2851 (struct dst_entry *)net->ipv6.ip6_null_entry;
2852 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2853 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2854 ip6_template_metrics, true);
2856 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2857 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2858 sizeof(*net->ipv6.ip6_prohibit_entry),
2860 if (!net->ipv6.ip6_prohibit_entry)
2861 goto out_ip6_null_entry;
2862 net->ipv6.ip6_prohibit_entry->dst.path =
2863 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2864 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2865 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2866 ip6_template_metrics, true);
2868 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2869 sizeof(*net->ipv6.ip6_blk_hole_entry),
2871 if (!net->ipv6.ip6_blk_hole_entry)
2872 goto out_ip6_prohibit_entry;
2873 net->ipv6.ip6_blk_hole_entry->dst.path =
2874 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2875 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2876 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2877 ip6_template_metrics, true);
2880 net->ipv6.sysctl.flush_delay = 0;
2881 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2882 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2883 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2884 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2885 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2886 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2887 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2889 #ifdef CONFIG_PROC_FS
2890 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2891 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2893 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2899 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2900 out_ip6_prohibit_entry:
2901 kfree(net->ipv6.ip6_prohibit_entry);
2903 kfree(net->ipv6.ip6_null_entry);
2905 out_ip6_dst_entries:
2906 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2911 static void __net_exit ip6_route_net_exit(struct net *net)
2913 #ifdef CONFIG_PROC_FS
2914 proc_net_remove(net, "ipv6_route");
2915 proc_net_remove(net, "rt6_stats");
2917 kfree(net->ipv6.ip6_null_entry);
2918 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2919 kfree(net->ipv6.ip6_prohibit_entry);
2920 kfree(net->ipv6.ip6_blk_hole_entry);
2922 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2925 static struct pernet_operations ip6_route_net_ops = {
2926 .init = ip6_route_net_init,
2927 .exit = ip6_route_net_exit,
2930 static struct notifier_block ip6_route_dev_notifier = {
2931 .notifier_call = ip6_route_dev_notify,
2935 int __init ip6_route_init(void)
2940 ip6_dst_ops_template.kmem_cachep =
2941 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2942 SLAB_HWCACHE_ALIGN, NULL);
2943 if (!ip6_dst_ops_template.kmem_cachep)
2946 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2948 goto out_kmem_cache;
2950 ret = register_pernet_subsys(&ip6_route_net_ops);
2952 goto out_dst_entries;
2954 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2956 /* Registering of the loopback is done before this portion of code,
2957 * the loopback reference in rt6_info will not be taken, do it
2958 * manually for init_net */
2959 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2960 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2961 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2962 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2963 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2964 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2965 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2969 goto out_register_subsys;
2975 ret = fib6_rules_init();
2980 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
2981 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
2982 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
2983 goto fib6_rules_init;
2985 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2987 goto fib6_rules_init;
2993 fib6_rules_cleanup();
2998 out_register_subsys:
2999 unregister_pernet_subsys(&ip6_route_net_ops);
3001 dst_entries_destroy(&ip6_dst_blackhole_ops);
3003 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3007 void ip6_route_cleanup(void)
3009 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3010 fib6_rules_cleanup();
3013 unregister_pernet_subsys(&ip6_route_net_ops);
3014 dst_entries_destroy(&ip6_dst_blackhole_ops);
3015 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);