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 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
61 #include <asm/uaccess.h>
64 #include <linux/sysctl.h>
67 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
68 const struct in6_addr *dest);
69 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
70 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
71 static unsigned int ip6_mtu(const struct dst_entry *dst);
72 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
73 static void ip6_dst_destroy(struct dst_entry *);
74 static void ip6_dst_ifdown(struct dst_entry *,
75 struct net_device *dev, int how);
76 static int ip6_dst_gc(struct dst_ops *ops);
78 static int ip6_pkt_discard(struct sk_buff *skb);
79 static int ip6_pkt_discard_out(struct sk_buff *skb);
80 static void ip6_link_failure(struct sk_buff *skb);
81 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
82 static void rt6_do_redirect(struct dst_entry *dst, struct sk_buff *skb);
84 #ifdef CONFIG_IPV6_ROUTE_INFO
85 static struct rt6_info *rt6_add_route_info(struct net *net,
86 const struct in6_addr *prefix, int prefixlen,
87 const struct in6_addr *gwaddr, int ifindex,
89 static struct rt6_info *rt6_get_route_info(struct net *net,
90 const struct in6_addr *prefix, int prefixlen,
91 const struct in6_addr *gwaddr, int ifindex);
94 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
96 struct rt6_info *rt = (struct rt6_info *) dst;
97 struct inet_peer *peer;
100 if (!(rt->dst.flags & DST_HOST))
103 peer = rt6_get_peer_create(rt);
105 u32 *old_p = __DST_METRICS_PTR(old);
106 unsigned long prev, new;
109 if (inet_metrics_new(peer))
110 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
112 new = (unsigned long) p;
113 prev = cmpxchg(&dst->_metrics, old, new);
116 p = __DST_METRICS_PTR(prev);
117 if (prev & DST_METRICS_READ_ONLY)
124 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
128 struct in6_addr *p = &rt->rt6i_gateway;
130 if (!ipv6_addr_any(p))
131 return (const void *) p;
133 return &ipv6_hdr(skb)->daddr;
137 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
141 struct rt6_info *rt = (struct rt6_info *) dst;
144 daddr = choose_neigh_daddr(rt, skb, daddr);
145 n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
148 return neigh_create(&nd_tbl, daddr, dst->dev);
151 static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
153 struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
155 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
164 static struct dst_ops ip6_dst_ops_template = {
166 .protocol = cpu_to_be16(ETH_P_IPV6),
169 .check = ip6_dst_check,
170 .default_advmss = ip6_default_advmss,
172 .cow_metrics = ipv6_cow_metrics,
173 .destroy = ip6_dst_destroy,
174 .ifdown = ip6_dst_ifdown,
175 .negative_advice = ip6_negative_advice,
176 .link_failure = ip6_link_failure,
177 .update_pmtu = ip6_rt_update_pmtu,
178 .redirect = rt6_do_redirect,
179 .local_out = __ip6_local_out,
180 .neigh_lookup = ip6_neigh_lookup,
183 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
185 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
187 return mtu ? : dst->dev->mtu;
190 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
194 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sk_buff *skb)
198 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
204 static struct dst_ops ip6_dst_blackhole_ops = {
206 .protocol = cpu_to_be16(ETH_P_IPV6),
207 .destroy = ip6_dst_destroy,
208 .check = ip6_dst_check,
209 .mtu = ip6_blackhole_mtu,
210 .default_advmss = ip6_default_advmss,
211 .update_pmtu = ip6_rt_blackhole_update_pmtu,
212 .redirect = ip6_rt_blackhole_redirect,
213 .cow_metrics = ip6_rt_blackhole_cow_metrics,
214 .neigh_lookup = ip6_neigh_lookup,
217 static const u32 ip6_template_metrics[RTAX_MAX] = {
218 [RTAX_HOPLIMIT - 1] = 255,
221 static struct rt6_info ip6_null_entry_template = {
223 .__refcnt = ATOMIC_INIT(1),
226 .error = -ENETUNREACH,
227 .input = ip6_pkt_discard,
228 .output = ip6_pkt_discard_out,
230 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
231 .rt6i_protocol = RTPROT_KERNEL,
232 .rt6i_metric = ~(u32) 0,
233 .rt6i_ref = ATOMIC_INIT(1),
236 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
238 static int ip6_pkt_prohibit(struct sk_buff *skb);
239 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
241 static struct rt6_info ip6_prohibit_entry_template = {
243 .__refcnt = ATOMIC_INIT(1),
247 .input = ip6_pkt_prohibit,
248 .output = ip6_pkt_prohibit_out,
250 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
251 .rt6i_protocol = RTPROT_KERNEL,
252 .rt6i_metric = ~(u32) 0,
253 .rt6i_ref = ATOMIC_INIT(1),
256 static struct rt6_info ip6_blk_hole_entry_template = {
258 .__refcnt = ATOMIC_INIT(1),
262 .input = dst_discard,
263 .output = dst_discard,
265 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
266 .rt6i_protocol = RTPROT_KERNEL,
267 .rt6i_metric = ~(u32) 0,
268 .rt6i_ref = ATOMIC_INIT(1),
273 /* allocate dst with ip6_dst_ops */
274 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
275 struct net_device *dev,
277 struct fib6_table *table)
279 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
283 struct dst_entry *dst = &rt->dst;
285 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
286 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
291 static void ip6_dst_destroy(struct dst_entry *dst)
293 struct rt6_info *rt = (struct rt6_info *)dst;
294 struct inet6_dev *idev = rt->rt6i_idev;
297 neigh_release(rt->n);
299 if (!(rt->dst.flags & DST_HOST))
300 dst_destroy_metrics_generic(dst);
303 rt->rt6i_idev = NULL;
307 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
308 dst_release(dst->from);
310 if (rt6_has_peer(rt)) {
311 struct inet_peer *peer = rt6_peer_ptr(rt);
316 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
318 static u32 rt6_peer_genid(void)
320 return atomic_read(&__rt6_peer_genid);
323 void rt6_bind_peer(struct rt6_info *rt, int create)
325 struct inet_peer_base *base;
326 struct inet_peer *peer;
328 base = inetpeer_base_ptr(rt->_rt6i_peer);
332 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
334 if (!rt6_set_peer(rt, peer))
337 rt->rt6i_peer_genid = rt6_peer_genid();
341 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
344 struct rt6_info *rt = (struct rt6_info *)dst;
345 struct inet6_dev *idev = rt->rt6i_idev;
346 struct net_device *loopback_dev =
347 dev_net(dev)->loopback_dev;
349 if (dev != loopback_dev) {
350 if (idev && idev->dev == dev) {
351 struct inet6_dev *loopback_idev =
352 in6_dev_get(loopback_dev);
354 rt->rt6i_idev = loopback_idev;
358 if (rt->n && rt->n->dev == dev) {
359 rt->n->dev = loopback_dev;
360 dev_hold(loopback_dev);
366 static bool rt6_check_expired(const struct rt6_info *rt)
368 struct rt6_info *ort = NULL;
370 if (rt->rt6i_flags & RTF_EXPIRES) {
371 if (time_after(jiffies, rt->dst.expires))
373 } else if (rt->dst.from) {
374 ort = (struct rt6_info *) rt->dst.from;
375 return (ort->rt6i_flags & RTF_EXPIRES) &&
376 time_after(jiffies, ort->dst.expires);
381 static bool rt6_need_strict(const struct in6_addr *daddr)
383 return ipv6_addr_type(daddr) &
384 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
388 * Route lookup. Any table->tb6_lock is implied.
391 static inline struct rt6_info *rt6_device_match(struct net *net,
393 const struct in6_addr *saddr,
397 struct rt6_info *local = NULL;
398 struct rt6_info *sprt;
400 if (!oif && ipv6_addr_any(saddr))
403 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
404 struct net_device *dev = sprt->dst.dev;
407 if (dev->ifindex == oif)
409 if (dev->flags & IFF_LOOPBACK) {
410 if (!sprt->rt6i_idev ||
411 sprt->rt6i_idev->dev->ifindex != oif) {
412 if (flags & RT6_LOOKUP_F_IFACE && oif)
414 if (local && (!oif ||
415 local->rt6i_idev->dev->ifindex == oif))
421 if (ipv6_chk_addr(net, saddr, dev,
422 flags & RT6_LOOKUP_F_IFACE))
431 if (flags & RT6_LOOKUP_F_IFACE)
432 return net->ipv6.ip6_null_entry;
438 #ifdef CONFIG_IPV6_ROUTER_PREF
439 static void rt6_probe(struct rt6_info *rt)
441 struct neighbour *neigh;
443 * Okay, this does not seem to be appropriate
444 * for now, however, we need to check if it
445 * is really so; aka Router Reachability Probing.
447 * Router Reachability Probe MUST be rate-limited
448 * to no more than one per minute.
451 neigh = rt ? rt->n : NULL;
452 if (!neigh || (neigh->nud_state & NUD_VALID))
454 read_lock_bh(&neigh->lock);
455 if (!(neigh->nud_state & NUD_VALID) &&
456 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
457 struct in6_addr mcaddr;
458 struct in6_addr *target;
460 neigh->updated = jiffies;
461 read_unlock_bh(&neigh->lock);
463 target = (struct in6_addr *)&neigh->primary_key;
464 addrconf_addr_solict_mult(target, &mcaddr);
465 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
467 read_unlock_bh(&neigh->lock);
473 static inline void rt6_probe(struct rt6_info *rt)
479 * Default Router Selection (RFC 2461 6.3.6)
481 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
483 struct net_device *dev = rt->dst.dev;
484 if (!oif || dev->ifindex == oif)
486 if ((dev->flags & IFF_LOOPBACK) &&
487 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
492 static inline int rt6_check_neigh(struct rt6_info *rt)
494 struct neighbour *neigh;
499 if (rt->rt6i_flags & RTF_NONEXTHOP ||
500 !(rt->rt6i_flags & RTF_GATEWAY))
503 read_lock_bh(&neigh->lock);
504 if (neigh->nud_state & NUD_VALID)
506 #ifdef CONFIG_IPV6_ROUTER_PREF
507 else if (neigh->nud_state & NUD_FAILED)
512 read_unlock_bh(&neigh->lock);
519 static int rt6_score_route(struct rt6_info *rt, int oif,
524 m = rt6_check_dev(rt, oif);
525 if (!m && (strict & RT6_LOOKUP_F_IFACE))
527 #ifdef CONFIG_IPV6_ROUTER_PREF
528 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
530 n = rt6_check_neigh(rt);
531 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
536 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
537 int *mpri, struct rt6_info *match)
541 if (rt6_check_expired(rt))
544 m = rt6_score_route(rt, oif, strict);
549 if (strict & RT6_LOOKUP_F_REACHABLE)
553 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
561 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
562 struct rt6_info *rr_head,
563 u32 metric, int oif, int strict)
565 struct rt6_info *rt, *match;
569 for (rt = rr_head; rt && rt->rt6i_metric == metric;
570 rt = rt->dst.rt6_next)
571 match = find_match(rt, oif, strict, &mpri, match);
572 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
573 rt = rt->dst.rt6_next)
574 match = find_match(rt, oif, strict, &mpri, match);
579 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
581 struct rt6_info *match, *rt0;
586 fn->rr_ptr = rt0 = fn->leaf;
588 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
591 (strict & RT6_LOOKUP_F_REACHABLE)) {
592 struct rt6_info *next = rt0->dst.rt6_next;
594 /* no entries matched; do round-robin */
595 if (!next || next->rt6i_metric != rt0->rt6i_metric)
602 net = dev_net(rt0->dst.dev);
603 return match ? match : net->ipv6.ip6_null_entry;
606 #ifdef CONFIG_IPV6_ROUTE_INFO
607 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
608 const struct in6_addr *gwaddr)
610 struct net *net = dev_net(dev);
611 struct route_info *rinfo = (struct route_info *) opt;
612 struct in6_addr prefix_buf, *prefix;
614 unsigned long lifetime;
617 if (len < sizeof(struct route_info)) {
621 /* Sanity check for prefix_len and length */
622 if (rinfo->length > 3) {
624 } else if (rinfo->prefix_len > 128) {
626 } else if (rinfo->prefix_len > 64) {
627 if (rinfo->length < 2) {
630 } else if (rinfo->prefix_len > 0) {
631 if (rinfo->length < 1) {
636 pref = rinfo->route_pref;
637 if (pref == ICMPV6_ROUTER_PREF_INVALID)
640 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
642 if (rinfo->length == 3)
643 prefix = (struct in6_addr *)rinfo->prefix;
645 /* this function is safe */
646 ipv6_addr_prefix(&prefix_buf,
647 (struct in6_addr *)rinfo->prefix,
649 prefix = &prefix_buf;
652 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
655 if (rt && !lifetime) {
661 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
664 rt->rt6i_flags = RTF_ROUTEINFO |
665 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
668 if (!addrconf_finite_timeout(lifetime))
669 rt6_clean_expires(rt);
671 rt6_set_expires(rt, jiffies + HZ * lifetime);
673 dst_release(&rt->dst);
679 #define BACKTRACK(__net, saddr) \
681 if (rt == __net->ipv6.ip6_null_entry) { \
682 struct fib6_node *pn; \
684 if (fn->fn_flags & RTN_TL_ROOT) \
687 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
688 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
691 if (fn->fn_flags & RTN_RTINFO) \
697 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
698 struct fib6_table *table,
699 struct flowi6 *fl6, int flags)
701 struct fib6_node *fn;
704 read_lock_bh(&table->tb6_lock);
705 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
708 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
709 BACKTRACK(net, &fl6->saddr);
711 dst_use(&rt->dst, jiffies);
712 read_unlock_bh(&table->tb6_lock);
717 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
720 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
722 EXPORT_SYMBOL_GPL(ip6_route_lookup);
724 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
725 const struct in6_addr *saddr, int oif, int strict)
727 struct flowi6 fl6 = {
731 struct dst_entry *dst;
732 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
735 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
736 flags |= RT6_LOOKUP_F_HAS_SADDR;
739 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
741 return (struct rt6_info *) dst;
748 EXPORT_SYMBOL(rt6_lookup);
750 /* ip6_ins_rt is called with FREE table->tb6_lock.
751 It takes new route entry, the addition fails by any reason the
752 route is freed. In any case, if caller does not hold it, it may
756 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
759 struct fib6_table *table;
761 table = rt->rt6i_table;
762 write_lock_bh(&table->tb6_lock);
763 err = fib6_add(&table->tb6_root, rt, info);
764 write_unlock_bh(&table->tb6_lock);
769 int ip6_ins_rt(struct rt6_info *rt)
771 struct nl_info info = {
772 .nl_net = dev_net(rt->dst.dev),
774 return __ip6_ins_rt(rt, &info);
777 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
778 const struct in6_addr *daddr,
779 const struct in6_addr *saddr)
787 rt = ip6_rt_copy(ort, daddr);
790 int attempts = !in_softirq();
792 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
793 if (ort->rt6i_dst.plen != 128 &&
794 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
795 rt->rt6i_flags |= RTF_ANYCAST;
796 rt->rt6i_gateway = *daddr;
799 rt->rt6i_flags |= RTF_CACHE;
801 #ifdef CONFIG_IPV6_SUBTREES
802 if (rt->rt6i_src.plen && saddr) {
803 rt->rt6i_src.addr = *saddr;
804 rt->rt6i_src.plen = 128;
809 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
810 struct net *net = dev_net(rt->dst.dev);
811 int saved_rt_min_interval =
812 net->ipv6.sysctl.ip6_rt_gc_min_interval;
813 int saved_rt_elasticity =
814 net->ipv6.sysctl.ip6_rt_gc_elasticity;
816 if (attempts-- > 0) {
817 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
818 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
820 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
822 net->ipv6.sysctl.ip6_rt_gc_elasticity =
824 net->ipv6.sysctl.ip6_rt_gc_min_interval =
825 saved_rt_min_interval;
829 net_warn_ratelimited("Neighbour table overflow\n");
838 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
839 const struct in6_addr *daddr)
841 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
844 rt->rt6i_flags |= RTF_CACHE;
845 rt->n = neigh_clone(ort->n);
850 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
851 struct flowi6 *fl6, int flags)
853 struct fib6_node *fn;
854 struct rt6_info *rt, *nrt;
858 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
860 strict |= flags & RT6_LOOKUP_F_IFACE;
863 read_lock_bh(&table->tb6_lock);
866 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
869 rt = rt6_select(fn, oif, strict | reachable);
871 BACKTRACK(net, &fl6->saddr);
872 if (rt == net->ipv6.ip6_null_entry ||
873 rt->rt6i_flags & RTF_CACHE)
877 read_unlock_bh(&table->tb6_lock);
879 if (!rt->n && !(rt->rt6i_flags & RTF_NONEXTHOP))
880 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
881 else if (!(rt->dst.flags & DST_HOST))
882 nrt = rt6_alloc_clone(rt, &fl6->daddr);
886 dst_release(&rt->dst);
887 rt = nrt ? : net->ipv6.ip6_null_entry;
891 err = ip6_ins_rt(nrt);
900 * Race condition! In the gap, when table->tb6_lock was
901 * released someone could insert this route. Relookup.
903 dst_release(&rt->dst);
912 read_unlock_bh(&table->tb6_lock);
914 rt->dst.lastuse = jiffies;
920 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
921 struct flowi6 *fl6, int flags)
923 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
926 static struct dst_entry *ip6_route_input_lookup(struct net *net,
927 struct net_device *dev,
928 struct flowi6 *fl6, int flags)
930 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
931 flags |= RT6_LOOKUP_F_IFACE;
933 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
936 void ip6_route_input(struct sk_buff *skb)
938 const struct ipv6hdr *iph = ipv6_hdr(skb);
939 struct net *net = dev_net(skb->dev);
940 int flags = RT6_LOOKUP_F_HAS_SADDR;
941 struct flowi6 fl6 = {
942 .flowi6_iif = skb->dev->ifindex,
945 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
946 .flowi6_mark = skb->mark,
947 .flowi6_proto = iph->nexthdr,
950 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
953 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
954 struct flowi6 *fl6, int flags)
956 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
959 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
964 fl6->flowi6_iif = net->loopback_dev->ifindex;
966 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
967 flags |= RT6_LOOKUP_F_IFACE;
969 if (!ipv6_addr_any(&fl6->saddr))
970 flags |= RT6_LOOKUP_F_HAS_SADDR;
972 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
974 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
977 EXPORT_SYMBOL(ip6_route_output);
979 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
981 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
982 struct dst_entry *new = NULL;
984 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
988 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
989 rt6_init_peer(rt, net->ipv6.peers);
992 new->input = dst_discard;
993 new->output = dst_discard;
995 if (dst_metrics_read_only(&ort->dst))
996 new->_metrics = ort->dst._metrics;
998 dst_copy_metrics(new, &ort->dst);
999 rt->rt6i_idev = ort->rt6i_idev;
1001 in6_dev_hold(rt->rt6i_idev);
1003 rt->rt6i_gateway = ort->rt6i_gateway;
1004 rt->rt6i_flags = ort->rt6i_flags;
1005 rt6_clean_expires(rt);
1006 rt->rt6i_metric = 0;
1008 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1009 #ifdef CONFIG_IPV6_SUBTREES
1010 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1016 dst_release(dst_orig);
1017 return new ? new : ERR_PTR(-ENOMEM);
1021 * Destination cache support functions
1024 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1026 struct rt6_info *rt;
1028 rt = (struct rt6_info *) dst;
1030 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
1031 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
1032 if (!rt6_has_peer(rt))
1033 rt6_bind_peer(rt, 0);
1034 rt->rt6i_peer_genid = rt6_peer_genid();
1041 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1043 struct rt6_info *rt = (struct rt6_info *) dst;
1046 if (rt->rt6i_flags & RTF_CACHE) {
1047 if (rt6_check_expired(rt)) {
1059 static void ip6_link_failure(struct sk_buff *skb)
1061 struct rt6_info *rt;
1063 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1065 rt = (struct rt6_info *) skb_dst(skb);
1067 if (rt->rt6i_flags & RTF_CACHE)
1068 rt6_update_expires(rt, 0);
1069 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1070 rt->rt6i_node->fn_sernum = -1;
1074 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1076 struct rt6_info *rt6 = (struct rt6_info*)dst;
1079 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1080 struct net *net = dev_net(dst->dev);
1082 rt6->rt6i_flags |= RTF_MODIFIED;
1083 if (mtu < IPV6_MIN_MTU) {
1084 u32 features = dst_metric(dst, RTAX_FEATURES);
1086 features |= RTAX_FEATURE_ALLFRAG;
1087 dst_metric_set(dst, RTAX_FEATURES, features);
1089 dst_metric_set(dst, RTAX_MTU, mtu);
1090 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1094 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1097 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1098 struct dst_entry *dst;
1101 memset(&fl6, 0, sizeof(fl6));
1102 fl6.flowi6_oif = oif;
1103 fl6.flowi6_mark = mark;
1104 fl6.flowi6_flags = 0;
1105 fl6.daddr = iph->daddr;
1106 fl6.saddr = iph->saddr;
1107 fl6.flowlabel = (*(__be32 *) iph) & IPV6_FLOWINFO_MASK;
1109 dst = ip6_route_output(net, NULL, &fl6);
1111 ip6_rt_update_pmtu(dst, ntohl(mtu));
1114 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1116 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1118 ip6_update_pmtu(skb, sock_net(sk), mtu,
1119 sk->sk_bound_dev_if, sk->sk_mark);
1121 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1123 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1125 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1126 struct dst_entry *dst;
1129 memset(&fl6, 0, sizeof(fl6));
1130 fl6.flowi6_oif = oif;
1131 fl6.flowi6_mark = mark;
1132 fl6.flowi6_flags = 0;
1133 fl6.daddr = iph->daddr;
1134 fl6.saddr = iph->saddr;
1135 fl6.flowlabel = (*(__be32 *) iph) & IPV6_FLOWINFO_MASK;
1137 dst = ip6_route_output(net, NULL, &fl6);
1139 rt6_do_redirect(dst, skb);
1142 EXPORT_SYMBOL_GPL(ip6_redirect);
1144 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1146 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1148 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1150 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1152 struct net_device *dev = dst->dev;
1153 unsigned int mtu = dst_mtu(dst);
1154 struct net *net = dev_net(dev);
1156 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1158 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1159 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1162 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1163 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1164 * IPV6_MAXPLEN is also valid and means: "any MSS,
1165 * rely only on pmtu discovery"
1167 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1172 static unsigned int ip6_mtu(const struct dst_entry *dst)
1174 struct inet6_dev *idev;
1175 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1183 idev = __in6_dev_get(dst->dev);
1185 mtu = idev->cnf.mtu6;
1191 static struct dst_entry *icmp6_dst_gc_list;
1192 static DEFINE_SPINLOCK(icmp6_dst_lock);
1194 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1195 struct neighbour *neigh,
1198 struct dst_entry *dst;
1199 struct rt6_info *rt;
1200 struct inet6_dev *idev = in6_dev_get(dev);
1201 struct net *net = dev_net(dev);
1203 if (unlikely(!idev))
1204 return ERR_PTR(-ENODEV);
1206 rt = ip6_dst_alloc(net, dev, 0, NULL);
1207 if (unlikely(!rt)) {
1209 dst = ERR_PTR(-ENOMEM);
1216 neigh = ip6_neigh_lookup(&rt->dst, NULL, &fl6->daddr);
1217 if (IS_ERR(neigh)) {
1220 return ERR_CAST(neigh);
1224 rt->dst.flags |= DST_HOST;
1225 rt->dst.output = ip6_output;
1227 atomic_set(&rt->dst.__refcnt, 1);
1228 rt->rt6i_dst.addr = fl6->daddr;
1229 rt->rt6i_dst.plen = 128;
1230 rt->rt6i_idev = idev;
1231 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1233 spin_lock_bh(&icmp6_dst_lock);
1234 rt->dst.next = icmp6_dst_gc_list;
1235 icmp6_dst_gc_list = &rt->dst;
1236 spin_unlock_bh(&icmp6_dst_lock);
1238 fib6_force_start_gc(net);
1240 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1246 int icmp6_dst_gc(void)
1248 struct dst_entry *dst, **pprev;
1251 spin_lock_bh(&icmp6_dst_lock);
1252 pprev = &icmp6_dst_gc_list;
1254 while ((dst = *pprev) != NULL) {
1255 if (!atomic_read(&dst->__refcnt)) {
1264 spin_unlock_bh(&icmp6_dst_lock);
1269 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1272 struct dst_entry *dst, **pprev;
1274 spin_lock_bh(&icmp6_dst_lock);
1275 pprev = &icmp6_dst_gc_list;
1276 while ((dst = *pprev) != NULL) {
1277 struct rt6_info *rt = (struct rt6_info *) dst;
1278 if (func(rt, arg)) {
1285 spin_unlock_bh(&icmp6_dst_lock);
1288 static int ip6_dst_gc(struct dst_ops *ops)
1290 unsigned long now = jiffies;
1291 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1292 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1293 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1294 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1295 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1296 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1299 entries = dst_entries_get_fast(ops);
1300 if (time_after(rt_last_gc + rt_min_interval, now) &&
1301 entries <= rt_max_size)
1304 net->ipv6.ip6_rt_gc_expire++;
1305 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1306 net->ipv6.ip6_rt_last_gc = now;
1307 entries = dst_entries_get_slow(ops);
1308 if (entries < ops->gc_thresh)
1309 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1311 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1312 return entries > rt_max_size;
1315 /* Clean host part of a prefix. Not necessary in radix tree,
1316 but results in cleaner routing tables.
1318 Remove it only when all the things will work!
1321 int ip6_dst_hoplimit(struct dst_entry *dst)
1323 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1324 if (hoplimit == 0) {
1325 struct net_device *dev = dst->dev;
1326 struct inet6_dev *idev;
1329 idev = __in6_dev_get(dev);
1331 hoplimit = idev->cnf.hop_limit;
1333 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1338 EXPORT_SYMBOL(ip6_dst_hoplimit);
1344 int ip6_route_add(struct fib6_config *cfg)
1347 struct net *net = cfg->fc_nlinfo.nl_net;
1348 struct rt6_info *rt = NULL;
1349 struct net_device *dev = NULL;
1350 struct inet6_dev *idev = NULL;
1351 struct fib6_table *table;
1354 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1356 #ifndef CONFIG_IPV6_SUBTREES
1357 if (cfg->fc_src_len)
1360 if (cfg->fc_ifindex) {
1362 dev = dev_get_by_index(net, cfg->fc_ifindex);
1365 idev = in6_dev_get(dev);
1370 if (cfg->fc_metric == 0)
1371 cfg->fc_metric = IP6_RT_PRIO_USER;
1374 if (cfg->fc_nlinfo.nlh &&
1375 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1376 table = fib6_get_table(net, cfg->fc_table);
1378 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1379 table = fib6_new_table(net, cfg->fc_table);
1382 table = fib6_new_table(net, cfg->fc_table);
1388 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1395 rt->dst.obsolete = -1;
1397 if (cfg->fc_flags & RTF_EXPIRES)
1398 rt6_set_expires(rt, jiffies +
1399 clock_t_to_jiffies(cfg->fc_expires));
1401 rt6_clean_expires(rt);
1403 if (cfg->fc_protocol == RTPROT_UNSPEC)
1404 cfg->fc_protocol = RTPROT_BOOT;
1405 rt->rt6i_protocol = cfg->fc_protocol;
1407 addr_type = ipv6_addr_type(&cfg->fc_dst);
1409 if (addr_type & IPV6_ADDR_MULTICAST)
1410 rt->dst.input = ip6_mc_input;
1411 else if (cfg->fc_flags & RTF_LOCAL)
1412 rt->dst.input = ip6_input;
1414 rt->dst.input = ip6_forward;
1416 rt->dst.output = ip6_output;
1418 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1419 rt->rt6i_dst.plen = cfg->fc_dst_len;
1420 if (rt->rt6i_dst.plen == 128)
1421 rt->dst.flags |= DST_HOST;
1423 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1424 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1429 dst_init_metrics(&rt->dst, metrics, 0);
1431 #ifdef CONFIG_IPV6_SUBTREES
1432 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1433 rt->rt6i_src.plen = cfg->fc_src_len;
1436 rt->rt6i_metric = cfg->fc_metric;
1438 /* We cannot add true routes via loopback here,
1439 they would result in kernel looping; promote them to reject routes
1441 if ((cfg->fc_flags & RTF_REJECT) ||
1442 (dev && (dev->flags & IFF_LOOPBACK) &&
1443 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1444 !(cfg->fc_flags & RTF_LOCAL))) {
1445 /* hold loopback dev/idev if we haven't done so. */
1446 if (dev != net->loopback_dev) {
1451 dev = net->loopback_dev;
1453 idev = in6_dev_get(dev);
1459 rt->dst.output = ip6_pkt_discard_out;
1460 rt->dst.input = ip6_pkt_discard;
1461 rt->dst.error = -ENETUNREACH;
1462 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1466 if (cfg->fc_flags & RTF_GATEWAY) {
1467 const struct in6_addr *gw_addr;
1470 gw_addr = &cfg->fc_gateway;
1471 rt->rt6i_gateway = *gw_addr;
1472 gwa_type = ipv6_addr_type(gw_addr);
1474 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1475 struct rt6_info *grt;
1477 /* IPv6 strictly inhibits using not link-local
1478 addresses as nexthop address.
1479 Otherwise, router will not able to send redirects.
1480 It is very good, but in some (rare!) circumstances
1481 (SIT, PtP, NBMA NOARP links) it is handy to allow
1482 some exceptions. --ANK
1485 if (!(gwa_type & IPV6_ADDR_UNICAST))
1488 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1490 err = -EHOSTUNREACH;
1494 if (dev != grt->dst.dev) {
1495 dst_release(&grt->dst);
1500 idev = grt->rt6i_idev;
1502 in6_dev_hold(grt->rt6i_idev);
1504 if (!(grt->rt6i_flags & RTF_GATEWAY))
1506 dst_release(&grt->dst);
1512 if (!dev || (dev->flags & IFF_LOOPBACK))
1520 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1521 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1525 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1526 rt->rt6i_prefsrc.plen = 128;
1528 rt->rt6i_prefsrc.plen = 0;
1530 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1531 err = rt6_bind_neighbour(rt, dev);
1536 rt->rt6i_flags = cfg->fc_flags;
1543 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1544 int type = nla_type(nla);
1547 if (type > RTAX_MAX) {
1552 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1558 rt->rt6i_idev = idev;
1559 rt->rt6i_table = table;
1561 cfg->fc_nlinfo.nl_net = dev_net(dev);
1563 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1575 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1578 struct fib6_table *table;
1579 struct net *net = dev_net(rt->dst.dev);
1581 if (rt == net->ipv6.ip6_null_entry)
1584 table = rt->rt6i_table;
1585 write_lock_bh(&table->tb6_lock);
1587 err = fib6_del(rt, info);
1588 dst_release(&rt->dst);
1590 write_unlock_bh(&table->tb6_lock);
1595 int ip6_del_rt(struct rt6_info *rt)
1597 struct nl_info info = {
1598 .nl_net = dev_net(rt->dst.dev),
1600 return __ip6_del_rt(rt, &info);
1603 static int ip6_route_del(struct fib6_config *cfg)
1605 struct fib6_table *table;
1606 struct fib6_node *fn;
1607 struct rt6_info *rt;
1610 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1614 read_lock_bh(&table->tb6_lock);
1616 fn = fib6_locate(&table->tb6_root,
1617 &cfg->fc_dst, cfg->fc_dst_len,
1618 &cfg->fc_src, cfg->fc_src_len);
1621 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1622 if (cfg->fc_ifindex &&
1624 rt->dst.dev->ifindex != cfg->fc_ifindex))
1626 if (cfg->fc_flags & RTF_GATEWAY &&
1627 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1629 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1632 read_unlock_bh(&table->tb6_lock);
1634 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1637 read_unlock_bh(&table->tb6_lock);
1642 static void rt6_do_redirect(struct dst_entry *dst, struct sk_buff *skb)
1644 struct net *net = dev_net(skb->dev);
1645 struct netevent_redirect netevent;
1646 struct rt6_info *rt, *nrt = NULL;
1647 const struct in6_addr *target;
1648 struct ndisc_options ndopts;
1649 const struct in6_addr *dest;
1650 struct neighbour *old_neigh;
1651 struct inet6_dev *in6_dev;
1652 struct neighbour *neigh;
1653 struct icmp6hdr *icmph;
1654 int optlen, on_link;
1657 optlen = skb->tail - skb->transport_header;
1658 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1661 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1665 icmph = icmp6_hdr(skb);
1666 target = (const struct in6_addr *) (icmph + 1);
1669 if (ipv6_addr_is_multicast(dest)) {
1670 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1675 if (ipv6_addr_equal(dest, target)) {
1677 } else if (ipv6_addr_type(target) !=
1678 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1679 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1683 in6_dev = __in6_dev_get(skb->dev);
1686 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1690 * The IP source address of the Redirect MUST be the same as the current
1691 * first-hop router for the specified ICMP Destination Address.
1694 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1695 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1700 if (ndopts.nd_opts_tgt_lladdr) {
1701 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1704 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1709 rt = (struct rt6_info *) dst;
1710 if (rt == net->ipv6.ip6_null_entry) {
1711 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1715 /* Redirect received -> path was valid.
1716 * Look, redirects are sent only in response to data packets,
1717 * so that this nexthop apparently is reachable. --ANK
1719 dst_confirm(&rt->dst);
1721 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1725 /* Duplicate redirect: silently ignore. */
1727 if (neigh == old_neigh)
1731 * We have finally decided to accept it.
1734 neigh_update(neigh, lladdr, NUD_STALE,
1735 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1736 NEIGH_UPDATE_F_OVERRIDE|
1737 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1738 NEIGH_UPDATE_F_ISROUTER))
1741 nrt = ip6_rt_copy(rt, dest);
1745 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1747 nrt->rt6i_flags &= ~RTF_GATEWAY;
1749 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1750 nrt->n = neigh_clone(neigh);
1752 if (ip6_ins_rt(nrt))
1755 netevent.old = &rt->dst;
1756 netevent.old_neigh = old_neigh;
1757 netevent.new = &nrt->dst;
1758 netevent.new_neigh = neigh;
1759 netevent.daddr = dest;
1760 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1762 if (rt->rt6i_flags & RTF_CACHE) {
1763 rt = (struct rt6_info *) dst_clone(&rt->dst);
1768 neigh_release(neigh);
1772 * Misc support functions
1775 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1776 const struct in6_addr *dest)
1778 struct net *net = dev_net(ort->dst.dev);
1779 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
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;
1796 rt->rt6i_gateway = ort->rt6i_gateway;
1797 rt->rt6i_flags = ort->rt6i_flags;
1798 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1799 (RTF_DEFAULT | RTF_ADDRCONF))
1800 rt6_set_from(rt, ort);
1802 rt6_clean_expires(rt);
1803 rt->rt6i_metric = 0;
1805 #ifdef CONFIG_IPV6_SUBTREES
1806 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1808 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1809 rt->rt6i_table = ort->rt6i_table;
1814 #ifdef CONFIG_IPV6_ROUTE_INFO
1815 static struct rt6_info *rt6_get_route_info(struct net *net,
1816 const struct in6_addr *prefix, int prefixlen,
1817 const struct in6_addr *gwaddr, int ifindex)
1819 struct fib6_node *fn;
1820 struct rt6_info *rt = NULL;
1821 struct fib6_table *table;
1823 table = fib6_get_table(net, RT6_TABLE_INFO);
1827 write_lock_bh(&table->tb6_lock);
1828 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1832 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1833 if (rt->dst.dev->ifindex != ifindex)
1835 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1837 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1843 write_unlock_bh(&table->tb6_lock);
1847 static struct rt6_info *rt6_add_route_info(struct net *net,
1848 const struct in6_addr *prefix, int prefixlen,
1849 const struct in6_addr *gwaddr, int ifindex,
1852 struct fib6_config cfg = {
1853 .fc_table = RT6_TABLE_INFO,
1854 .fc_metric = IP6_RT_PRIO_USER,
1855 .fc_ifindex = ifindex,
1856 .fc_dst_len = prefixlen,
1857 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1858 RTF_UP | RTF_PREF(pref),
1860 .fc_nlinfo.nlh = NULL,
1861 .fc_nlinfo.nl_net = net,
1864 cfg.fc_dst = *prefix;
1865 cfg.fc_gateway = *gwaddr;
1867 /* We should treat it as a default route if prefix length is 0. */
1869 cfg.fc_flags |= RTF_DEFAULT;
1871 ip6_route_add(&cfg);
1873 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1877 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1879 struct rt6_info *rt;
1880 struct fib6_table *table;
1882 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1886 write_lock_bh(&table->tb6_lock);
1887 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1888 if (dev == rt->dst.dev &&
1889 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1890 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1895 write_unlock_bh(&table->tb6_lock);
1899 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1900 struct net_device *dev,
1903 struct fib6_config cfg = {
1904 .fc_table = RT6_TABLE_DFLT,
1905 .fc_metric = IP6_RT_PRIO_USER,
1906 .fc_ifindex = dev->ifindex,
1907 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1908 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1910 .fc_nlinfo.nlh = NULL,
1911 .fc_nlinfo.nl_net = dev_net(dev),
1914 cfg.fc_gateway = *gwaddr;
1916 ip6_route_add(&cfg);
1918 return rt6_get_dflt_router(gwaddr, dev);
1921 void rt6_purge_dflt_routers(struct net *net)
1923 struct rt6_info *rt;
1924 struct fib6_table *table;
1926 /* NOTE: Keep consistent with rt6_get_dflt_router */
1927 table = fib6_get_table(net, RT6_TABLE_DFLT);
1932 read_lock_bh(&table->tb6_lock);
1933 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1934 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1936 read_unlock_bh(&table->tb6_lock);
1941 read_unlock_bh(&table->tb6_lock);
1944 static void rtmsg_to_fib6_config(struct net *net,
1945 struct in6_rtmsg *rtmsg,
1946 struct fib6_config *cfg)
1948 memset(cfg, 0, sizeof(*cfg));
1950 cfg->fc_table = RT6_TABLE_MAIN;
1951 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1952 cfg->fc_metric = rtmsg->rtmsg_metric;
1953 cfg->fc_expires = rtmsg->rtmsg_info;
1954 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1955 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1956 cfg->fc_flags = rtmsg->rtmsg_flags;
1958 cfg->fc_nlinfo.nl_net = net;
1960 cfg->fc_dst = rtmsg->rtmsg_dst;
1961 cfg->fc_src = rtmsg->rtmsg_src;
1962 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1965 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1967 struct fib6_config cfg;
1968 struct in6_rtmsg rtmsg;
1972 case SIOCADDRT: /* Add a route */
1973 case SIOCDELRT: /* Delete a route */
1974 if (!capable(CAP_NET_ADMIN))
1976 err = copy_from_user(&rtmsg, arg,
1977 sizeof(struct in6_rtmsg));
1981 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1986 err = ip6_route_add(&cfg);
1989 err = ip6_route_del(&cfg);
2003 * Drop the packet on the floor
2006 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2009 struct dst_entry *dst = skb_dst(skb);
2010 switch (ipstats_mib_noroutes) {
2011 case IPSTATS_MIB_INNOROUTES:
2012 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2013 if (type == IPV6_ADDR_ANY) {
2014 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2015 IPSTATS_MIB_INADDRERRORS);
2019 case IPSTATS_MIB_OUTNOROUTES:
2020 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2021 ipstats_mib_noroutes);
2024 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2029 static int ip6_pkt_discard(struct sk_buff *skb)
2031 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2034 static int ip6_pkt_discard_out(struct sk_buff *skb)
2036 skb->dev = skb_dst(skb)->dev;
2037 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2040 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2042 static int ip6_pkt_prohibit(struct sk_buff *skb)
2044 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2047 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2049 skb->dev = skb_dst(skb)->dev;
2050 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2056 * Allocate a dst for local (unicast / anycast) address.
2059 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2060 const struct in6_addr *addr,
2063 struct net *net = dev_net(idev->dev);
2064 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2068 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2069 return ERR_PTR(-ENOMEM);
2074 rt->dst.flags |= DST_HOST;
2075 rt->dst.input = ip6_input;
2076 rt->dst.output = ip6_output;
2077 rt->rt6i_idev = idev;
2078 rt->dst.obsolete = -1;
2080 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2082 rt->rt6i_flags |= RTF_ANYCAST;
2084 rt->rt6i_flags |= RTF_LOCAL;
2085 err = rt6_bind_neighbour(rt, rt->dst.dev);
2088 return ERR_PTR(err);
2091 rt->rt6i_dst.addr = *addr;
2092 rt->rt6i_dst.plen = 128;
2093 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2095 atomic_set(&rt->dst.__refcnt, 1);
2100 int ip6_route_get_saddr(struct net *net,
2101 struct rt6_info *rt,
2102 const struct in6_addr *daddr,
2104 struct in6_addr *saddr)
2106 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2108 if (rt->rt6i_prefsrc.plen)
2109 *saddr = rt->rt6i_prefsrc.addr;
2111 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2112 daddr, prefs, saddr);
2116 /* remove deleted ip from prefsrc entries */
2117 struct arg_dev_net_ip {
2118 struct net_device *dev;
2120 struct in6_addr *addr;
2123 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2125 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2126 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2127 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2129 if (((void *)rt->dst.dev == dev || !dev) &&
2130 rt != net->ipv6.ip6_null_entry &&
2131 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2132 /* remove prefsrc entry */
2133 rt->rt6i_prefsrc.plen = 0;
2138 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2140 struct net *net = dev_net(ifp->idev->dev);
2141 struct arg_dev_net_ip adni = {
2142 .dev = ifp->idev->dev,
2146 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2149 struct arg_dev_net {
2150 struct net_device *dev;
2154 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2156 const struct arg_dev_net *adn = arg;
2157 const struct net_device *dev = adn->dev;
2159 if ((rt->dst.dev == dev || !dev) &&
2160 rt != adn->net->ipv6.ip6_null_entry)
2166 void rt6_ifdown(struct net *net, struct net_device *dev)
2168 struct arg_dev_net adn = {
2173 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2174 icmp6_clean_all(fib6_ifdown, &adn);
2177 struct rt6_mtu_change_arg {
2178 struct net_device *dev;
2182 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2184 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2185 struct inet6_dev *idev;
2187 /* In IPv6 pmtu discovery is not optional,
2188 so that RTAX_MTU lock cannot disable it.
2189 We still use this lock to block changes
2190 caused by addrconf/ndisc.
2193 idev = __in6_dev_get(arg->dev);
2197 /* For administrative MTU increase, there is no way to discover
2198 IPv6 PMTU increase, so PMTU increase should be updated here.
2199 Since RFC 1981 doesn't include administrative MTU increase
2200 update PMTU increase is a MUST. (i.e. jumbo frame)
2203 If new MTU is less than route PMTU, this new MTU will be the
2204 lowest MTU in the path, update the route PMTU to reflect PMTU
2205 decreases; if new MTU is greater than route PMTU, and the
2206 old MTU is the lowest MTU in the path, update the route PMTU
2207 to reflect the increase. In this case if the other nodes' MTU
2208 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2211 if (rt->dst.dev == arg->dev &&
2212 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2213 (dst_mtu(&rt->dst) >= arg->mtu ||
2214 (dst_mtu(&rt->dst) < arg->mtu &&
2215 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2216 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2221 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2223 struct rt6_mtu_change_arg arg = {
2228 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2231 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2232 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2233 [RTA_OIF] = { .type = NLA_U32 },
2234 [RTA_IIF] = { .type = NLA_U32 },
2235 [RTA_PRIORITY] = { .type = NLA_U32 },
2236 [RTA_METRICS] = { .type = NLA_NESTED },
2239 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2240 struct fib6_config *cfg)
2243 struct nlattr *tb[RTA_MAX+1];
2246 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2251 rtm = nlmsg_data(nlh);
2252 memset(cfg, 0, sizeof(*cfg));
2254 cfg->fc_table = rtm->rtm_table;
2255 cfg->fc_dst_len = rtm->rtm_dst_len;
2256 cfg->fc_src_len = rtm->rtm_src_len;
2257 cfg->fc_flags = RTF_UP;
2258 cfg->fc_protocol = rtm->rtm_protocol;
2260 if (rtm->rtm_type == RTN_UNREACHABLE)
2261 cfg->fc_flags |= RTF_REJECT;
2263 if (rtm->rtm_type == RTN_LOCAL)
2264 cfg->fc_flags |= RTF_LOCAL;
2266 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2267 cfg->fc_nlinfo.nlh = nlh;
2268 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2270 if (tb[RTA_GATEWAY]) {
2271 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2272 cfg->fc_flags |= RTF_GATEWAY;
2276 int plen = (rtm->rtm_dst_len + 7) >> 3;
2278 if (nla_len(tb[RTA_DST]) < plen)
2281 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2285 int plen = (rtm->rtm_src_len + 7) >> 3;
2287 if (nla_len(tb[RTA_SRC]) < plen)
2290 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2293 if (tb[RTA_PREFSRC])
2294 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2297 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2299 if (tb[RTA_PRIORITY])
2300 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2302 if (tb[RTA_METRICS]) {
2303 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2304 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2308 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2315 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2317 struct fib6_config cfg;
2320 err = rtm_to_fib6_config(skb, nlh, &cfg);
2324 return ip6_route_del(&cfg);
2327 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2329 struct fib6_config cfg;
2332 err = rtm_to_fib6_config(skb, nlh, &cfg);
2336 return ip6_route_add(&cfg);
2339 static inline size_t rt6_nlmsg_size(void)
2341 return NLMSG_ALIGN(sizeof(struct rtmsg))
2342 + nla_total_size(16) /* RTA_SRC */
2343 + nla_total_size(16) /* RTA_DST */
2344 + nla_total_size(16) /* RTA_GATEWAY */
2345 + nla_total_size(16) /* RTA_PREFSRC */
2346 + nla_total_size(4) /* RTA_TABLE */
2347 + nla_total_size(4) /* RTA_IIF */
2348 + nla_total_size(4) /* RTA_OIF */
2349 + nla_total_size(4) /* RTA_PRIORITY */
2350 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2351 + nla_total_size(sizeof(struct rta_cacheinfo));
2354 static int rt6_fill_node(struct net *net,
2355 struct sk_buff *skb, struct rt6_info *rt,
2356 struct in6_addr *dst, struct in6_addr *src,
2357 int iif, int type, u32 pid, u32 seq,
2358 int prefix, int nowait, unsigned int flags)
2361 struct nlmsghdr *nlh;
2364 struct neighbour *n;
2366 if (prefix) { /* user wants prefix routes only */
2367 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2368 /* success since this is not a prefix route */
2373 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2377 rtm = nlmsg_data(nlh);
2378 rtm->rtm_family = AF_INET6;
2379 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2380 rtm->rtm_src_len = rt->rt6i_src.plen;
2383 table = rt->rt6i_table->tb6_id;
2385 table = RT6_TABLE_UNSPEC;
2386 rtm->rtm_table = table;
2387 if (nla_put_u32(skb, RTA_TABLE, table))
2388 goto nla_put_failure;
2389 if (rt->rt6i_flags & RTF_REJECT)
2390 rtm->rtm_type = RTN_UNREACHABLE;
2391 else if (rt->rt6i_flags & RTF_LOCAL)
2392 rtm->rtm_type = RTN_LOCAL;
2393 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2394 rtm->rtm_type = RTN_LOCAL;
2396 rtm->rtm_type = RTN_UNICAST;
2398 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2399 rtm->rtm_protocol = rt->rt6i_protocol;
2400 if (rt->rt6i_flags & RTF_DYNAMIC)
2401 rtm->rtm_protocol = RTPROT_REDIRECT;
2402 else if (rt->rt6i_flags & RTF_ADDRCONF)
2403 rtm->rtm_protocol = RTPROT_KERNEL;
2404 else if (rt->rt6i_flags & RTF_DEFAULT)
2405 rtm->rtm_protocol = RTPROT_RA;
2407 if (rt->rt6i_flags & RTF_CACHE)
2408 rtm->rtm_flags |= RTM_F_CLONED;
2411 if (nla_put(skb, RTA_DST, 16, dst))
2412 goto nla_put_failure;
2413 rtm->rtm_dst_len = 128;
2414 } else if (rtm->rtm_dst_len)
2415 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2416 goto nla_put_failure;
2417 #ifdef CONFIG_IPV6_SUBTREES
2419 if (nla_put(skb, RTA_SRC, 16, src))
2420 goto nla_put_failure;
2421 rtm->rtm_src_len = 128;
2422 } else if (rtm->rtm_src_len &&
2423 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2424 goto nla_put_failure;
2427 #ifdef CONFIG_IPV6_MROUTE
2428 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2429 int err = ip6mr_get_route(net, skb, rtm, nowait);
2434 goto nla_put_failure;
2436 if (err == -EMSGSIZE)
2437 goto nla_put_failure;
2442 if (nla_put_u32(skb, RTA_IIF, iif))
2443 goto nla_put_failure;
2445 struct in6_addr saddr_buf;
2446 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2447 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2448 goto nla_put_failure;
2451 if (rt->rt6i_prefsrc.plen) {
2452 struct in6_addr saddr_buf;
2453 saddr_buf = rt->rt6i_prefsrc.addr;
2454 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2455 goto nla_put_failure;
2458 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2459 goto nla_put_failure;
2464 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0) {
2466 goto nla_put_failure;
2472 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2473 goto nla_put_failure;
2474 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2475 goto nla_put_failure;
2476 if (!(rt->rt6i_flags & RTF_EXPIRES))
2478 else if (rt->dst.expires - jiffies < INT_MAX)
2479 expires = rt->dst.expires - jiffies;
2483 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2484 goto nla_put_failure;
2486 return nlmsg_end(skb, nlh);
2489 nlmsg_cancel(skb, nlh);
2493 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2495 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2498 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2499 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2500 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2504 return rt6_fill_node(arg->net,
2505 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2506 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2507 prefix, 0, NLM_F_MULTI);
2510 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2512 struct net *net = sock_net(in_skb->sk);
2513 struct nlattr *tb[RTA_MAX+1];
2514 struct rt6_info *rt;
2515 struct sk_buff *skb;
2518 int err, iif = 0, oif = 0;
2520 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2525 memset(&fl6, 0, sizeof(fl6));
2528 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2531 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2535 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2538 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2542 iif = nla_get_u32(tb[RTA_IIF]);
2545 oif = nla_get_u32(tb[RTA_OIF]);
2548 struct net_device *dev;
2551 dev = __dev_get_by_index(net, iif);
2557 fl6.flowi6_iif = iif;
2559 if (!ipv6_addr_any(&fl6.saddr))
2560 flags |= RT6_LOOKUP_F_HAS_SADDR;
2562 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2565 fl6.flowi6_oif = oif;
2567 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2570 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2572 dst_release(&rt->dst);
2577 /* Reserve room for dummy headers, this skb can pass
2578 through good chunk of routing engine.
2580 skb_reset_mac_header(skb);
2581 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2583 skb_dst_set(skb, &rt->dst);
2585 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2586 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2587 nlh->nlmsg_seq, 0, 0, 0);
2593 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2598 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2600 struct sk_buff *skb;
2601 struct net *net = info->nl_net;
2606 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2608 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2612 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2613 event, info->pid, seq, 0, 0, 0);
2615 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2616 WARN_ON(err == -EMSGSIZE);
2620 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2621 info->nlh, gfp_any());
2625 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2628 static int ip6_route_dev_notify(struct notifier_block *this,
2629 unsigned long event, void *data)
2631 struct net_device *dev = (struct net_device *)data;
2632 struct net *net = dev_net(dev);
2634 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2635 net->ipv6.ip6_null_entry->dst.dev = dev;
2636 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2637 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2638 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2639 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2640 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2641 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2652 #ifdef CONFIG_PROC_FS
2663 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2665 struct seq_file *m = p_arg;
2666 struct neighbour *n;
2668 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2670 #ifdef CONFIG_IPV6_SUBTREES
2671 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2673 seq_puts(m, "00000000000000000000000000000000 00 ");
2678 seq_printf(m, "%pi6", n->primary_key);
2680 seq_puts(m, "00000000000000000000000000000000");
2683 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2684 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2685 rt->dst.__use, rt->rt6i_flags,
2686 rt->dst.dev ? rt->dst.dev->name : "");
2690 static int ipv6_route_show(struct seq_file *m, void *v)
2692 struct net *net = (struct net *)m->private;
2693 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2697 static int ipv6_route_open(struct inode *inode, struct file *file)
2699 return single_open_net(inode, file, ipv6_route_show);
2702 static const struct file_operations ipv6_route_proc_fops = {
2703 .owner = THIS_MODULE,
2704 .open = ipv6_route_open,
2706 .llseek = seq_lseek,
2707 .release = single_release_net,
2710 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2712 struct net *net = (struct net *)seq->private;
2713 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2714 net->ipv6.rt6_stats->fib_nodes,
2715 net->ipv6.rt6_stats->fib_route_nodes,
2716 net->ipv6.rt6_stats->fib_rt_alloc,
2717 net->ipv6.rt6_stats->fib_rt_entries,
2718 net->ipv6.rt6_stats->fib_rt_cache,
2719 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2720 net->ipv6.rt6_stats->fib_discarded_routes);
2725 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2727 return single_open_net(inode, file, rt6_stats_seq_show);
2730 static const struct file_operations rt6_stats_seq_fops = {
2731 .owner = THIS_MODULE,
2732 .open = rt6_stats_seq_open,
2734 .llseek = seq_lseek,
2735 .release = single_release_net,
2737 #endif /* CONFIG_PROC_FS */
2739 #ifdef CONFIG_SYSCTL
2742 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2743 void __user *buffer, size_t *lenp, loff_t *ppos)
2750 net = (struct net *)ctl->extra1;
2751 delay = net->ipv6.sysctl.flush_delay;
2752 proc_dointvec(ctl, write, buffer, lenp, ppos);
2753 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2757 ctl_table ipv6_route_table_template[] = {
2759 .procname = "flush",
2760 .data = &init_net.ipv6.sysctl.flush_delay,
2761 .maxlen = sizeof(int),
2763 .proc_handler = ipv6_sysctl_rtcache_flush
2766 .procname = "gc_thresh",
2767 .data = &ip6_dst_ops_template.gc_thresh,
2768 .maxlen = sizeof(int),
2770 .proc_handler = proc_dointvec,
2773 .procname = "max_size",
2774 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2775 .maxlen = sizeof(int),
2777 .proc_handler = proc_dointvec,
2780 .procname = "gc_min_interval",
2781 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2782 .maxlen = sizeof(int),
2784 .proc_handler = proc_dointvec_jiffies,
2787 .procname = "gc_timeout",
2788 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2789 .maxlen = sizeof(int),
2791 .proc_handler = proc_dointvec_jiffies,
2794 .procname = "gc_interval",
2795 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2796 .maxlen = sizeof(int),
2798 .proc_handler = proc_dointvec_jiffies,
2801 .procname = "gc_elasticity",
2802 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2803 .maxlen = sizeof(int),
2805 .proc_handler = proc_dointvec,
2808 .procname = "mtu_expires",
2809 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2810 .maxlen = sizeof(int),
2812 .proc_handler = proc_dointvec_jiffies,
2815 .procname = "min_adv_mss",
2816 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2817 .maxlen = sizeof(int),
2819 .proc_handler = proc_dointvec,
2822 .procname = "gc_min_interval_ms",
2823 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2824 .maxlen = sizeof(int),
2826 .proc_handler = proc_dointvec_ms_jiffies,
2831 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2833 struct ctl_table *table;
2835 table = kmemdup(ipv6_route_table_template,
2836 sizeof(ipv6_route_table_template),
2840 table[0].data = &net->ipv6.sysctl.flush_delay;
2841 table[0].extra1 = net;
2842 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2843 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2844 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2845 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2846 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2847 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2848 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2849 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2850 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2857 static int __net_init ip6_route_net_init(struct net *net)
2861 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2862 sizeof(net->ipv6.ip6_dst_ops));
2864 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2865 goto out_ip6_dst_ops;
2867 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2868 sizeof(*net->ipv6.ip6_null_entry),
2870 if (!net->ipv6.ip6_null_entry)
2871 goto out_ip6_dst_entries;
2872 net->ipv6.ip6_null_entry->dst.path =
2873 (struct dst_entry *)net->ipv6.ip6_null_entry;
2874 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2875 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2876 ip6_template_metrics, true);
2878 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2879 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2880 sizeof(*net->ipv6.ip6_prohibit_entry),
2882 if (!net->ipv6.ip6_prohibit_entry)
2883 goto out_ip6_null_entry;
2884 net->ipv6.ip6_prohibit_entry->dst.path =
2885 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2886 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2887 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2888 ip6_template_metrics, true);
2890 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2891 sizeof(*net->ipv6.ip6_blk_hole_entry),
2893 if (!net->ipv6.ip6_blk_hole_entry)
2894 goto out_ip6_prohibit_entry;
2895 net->ipv6.ip6_blk_hole_entry->dst.path =
2896 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2897 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2898 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2899 ip6_template_metrics, true);
2902 net->ipv6.sysctl.flush_delay = 0;
2903 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2904 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2905 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2906 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2907 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2908 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2909 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2911 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2917 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2918 out_ip6_prohibit_entry:
2919 kfree(net->ipv6.ip6_prohibit_entry);
2921 kfree(net->ipv6.ip6_null_entry);
2923 out_ip6_dst_entries:
2924 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2929 static void __net_exit ip6_route_net_exit(struct net *net)
2931 kfree(net->ipv6.ip6_null_entry);
2932 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2933 kfree(net->ipv6.ip6_prohibit_entry);
2934 kfree(net->ipv6.ip6_blk_hole_entry);
2936 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2939 static int __net_init ip6_route_net_init_late(struct net *net)
2941 #ifdef CONFIG_PROC_FS
2942 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2943 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2948 static void __net_exit ip6_route_net_exit_late(struct net *net)
2950 #ifdef CONFIG_PROC_FS
2951 proc_net_remove(net, "ipv6_route");
2952 proc_net_remove(net, "rt6_stats");
2956 static struct pernet_operations ip6_route_net_ops = {
2957 .init = ip6_route_net_init,
2958 .exit = ip6_route_net_exit,
2961 static int __net_init ipv6_inetpeer_init(struct net *net)
2963 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2967 inet_peer_base_init(bp);
2968 net->ipv6.peers = bp;
2972 static void __net_exit ipv6_inetpeer_exit(struct net *net)
2974 struct inet_peer_base *bp = net->ipv6.peers;
2976 net->ipv6.peers = NULL;
2977 inetpeer_invalidate_tree(bp);
2981 static struct pernet_operations ipv6_inetpeer_ops = {
2982 .init = ipv6_inetpeer_init,
2983 .exit = ipv6_inetpeer_exit,
2986 static struct pernet_operations ip6_route_net_late_ops = {
2987 .init = ip6_route_net_init_late,
2988 .exit = ip6_route_net_exit_late,
2991 static struct notifier_block ip6_route_dev_notifier = {
2992 .notifier_call = ip6_route_dev_notify,
2996 int __init ip6_route_init(void)
3001 ip6_dst_ops_template.kmem_cachep =
3002 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3003 SLAB_HWCACHE_ALIGN, NULL);
3004 if (!ip6_dst_ops_template.kmem_cachep)
3007 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3009 goto out_kmem_cache;
3011 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3013 goto out_dst_entries;
3015 ret = register_pernet_subsys(&ip6_route_net_ops);
3017 goto out_register_inetpeer;
3019 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3021 /* Registering of the loopback is done before this portion of code,
3022 * the loopback reference in rt6_info will not be taken, do it
3023 * manually for init_net */
3024 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3025 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3026 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3027 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3028 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3029 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3030 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3034 goto out_register_subsys;
3040 ret = fib6_rules_init();
3044 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3046 goto fib6_rules_init;
3049 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3050 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3051 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3052 goto out_register_late_subsys;
3054 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3056 goto out_register_late_subsys;
3061 out_register_late_subsys:
3062 unregister_pernet_subsys(&ip6_route_net_late_ops);
3064 fib6_rules_cleanup();
3069 out_register_subsys:
3070 unregister_pernet_subsys(&ip6_route_net_ops);
3071 out_register_inetpeer:
3072 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3074 dst_entries_destroy(&ip6_dst_blackhole_ops);
3076 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3080 void ip6_route_cleanup(void)
3082 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3083 unregister_pernet_subsys(&ip6_route_net_late_ops);
3084 fib6_rules_cleanup();
3087 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3088 unregister_pernet_subsys(&ip6_route_net_ops);
3089 dst_entries_destroy(&ip6_dst_blackhole_ops);
3090 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);