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>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
68 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
69 const struct in6_addr *dest);
70 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
71 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
72 static unsigned int ip6_mtu(const struct dst_entry *dst);
73 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
74 static void ip6_dst_destroy(struct dst_entry *);
75 static void ip6_dst_ifdown(struct dst_entry *,
76 struct net_device *dev, int how);
77 static int ip6_dst_gc(struct dst_ops *ops);
79 static int ip6_pkt_discard(struct sk_buff *skb);
80 static int ip6_pkt_discard_out(struct sk_buff *skb);
81 static void ip6_link_failure(struct sk_buff *skb);
82 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
83 struct sk_buff *skb, u32 mtu);
84 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
87 #ifdef CONFIG_IPV6_ROUTE_INFO
88 static struct rt6_info *rt6_add_route_info(struct net *net,
89 const struct in6_addr *prefix, int prefixlen,
90 const struct in6_addr *gwaddr, int ifindex,
92 static struct rt6_info *rt6_get_route_info(struct net *net,
93 const struct in6_addr *prefix, int prefixlen,
94 const struct in6_addr *gwaddr, int ifindex);
97 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
99 struct rt6_info *rt = (struct rt6_info *) dst;
100 struct inet_peer *peer;
103 if (!(rt->dst.flags & DST_HOST))
106 peer = rt6_get_peer_create(rt);
108 u32 *old_p = __DST_METRICS_PTR(old);
109 unsigned long prev, new;
112 if (inet_metrics_new(peer))
113 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
115 new = (unsigned long) p;
116 prev = cmpxchg(&dst->_metrics, old, new);
119 p = __DST_METRICS_PTR(prev);
120 if (prev & DST_METRICS_READ_ONLY)
127 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
131 struct in6_addr *p = &rt->rt6i_gateway;
133 if (!ipv6_addr_any(p))
134 return (const void *) p;
136 return &ipv6_hdr(skb)->daddr;
140 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
144 struct rt6_info *rt = (struct rt6_info *) dst;
147 daddr = choose_neigh_daddr(rt, skb, daddr);
148 n = __ipv6_neigh_lookup(dst->dev, daddr);
151 return neigh_create(&nd_tbl, daddr, dst->dev);
154 static struct dst_ops ip6_dst_ops_template = {
156 .protocol = cpu_to_be16(ETH_P_IPV6),
159 .check = ip6_dst_check,
160 .default_advmss = ip6_default_advmss,
162 .cow_metrics = ipv6_cow_metrics,
163 .destroy = ip6_dst_destroy,
164 .ifdown = ip6_dst_ifdown,
165 .negative_advice = ip6_negative_advice,
166 .link_failure = ip6_link_failure,
167 .update_pmtu = ip6_rt_update_pmtu,
168 .redirect = rt6_do_redirect,
169 .local_out = __ip6_local_out,
170 .neigh_lookup = ip6_neigh_lookup,
173 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
175 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
177 return mtu ? : dst->dev->mtu;
180 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
181 struct sk_buff *skb, u32 mtu)
185 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
190 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
196 static struct dst_ops ip6_dst_blackhole_ops = {
198 .protocol = cpu_to_be16(ETH_P_IPV6),
199 .destroy = ip6_dst_destroy,
200 .check = ip6_dst_check,
201 .mtu = ip6_blackhole_mtu,
202 .default_advmss = ip6_default_advmss,
203 .update_pmtu = ip6_rt_blackhole_update_pmtu,
204 .redirect = ip6_rt_blackhole_redirect,
205 .cow_metrics = ip6_rt_blackhole_cow_metrics,
206 .neigh_lookup = ip6_neigh_lookup,
209 static const u32 ip6_template_metrics[RTAX_MAX] = {
210 [RTAX_HOPLIMIT - 1] = 0,
213 static const struct rt6_info ip6_null_entry_template = {
215 .__refcnt = ATOMIC_INIT(1),
217 .obsolete = DST_OBSOLETE_FORCE_CHK,
218 .error = -ENETUNREACH,
219 .input = ip6_pkt_discard,
220 .output = ip6_pkt_discard_out,
222 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
223 .rt6i_protocol = RTPROT_KERNEL,
224 .rt6i_metric = ~(u32) 0,
225 .rt6i_ref = ATOMIC_INIT(1),
228 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
230 static int ip6_pkt_prohibit(struct sk_buff *skb);
231 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
233 static const struct rt6_info ip6_prohibit_entry_template = {
235 .__refcnt = ATOMIC_INIT(1),
237 .obsolete = DST_OBSOLETE_FORCE_CHK,
239 .input = ip6_pkt_prohibit,
240 .output = ip6_pkt_prohibit_out,
242 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
243 .rt6i_protocol = RTPROT_KERNEL,
244 .rt6i_metric = ~(u32) 0,
245 .rt6i_ref = ATOMIC_INIT(1),
248 static const struct rt6_info ip6_blk_hole_entry_template = {
250 .__refcnt = ATOMIC_INIT(1),
252 .obsolete = DST_OBSOLETE_FORCE_CHK,
254 .input = dst_discard,
255 .output = dst_discard,
257 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
258 .rt6i_protocol = RTPROT_KERNEL,
259 .rt6i_metric = ~(u32) 0,
260 .rt6i_ref = ATOMIC_INIT(1),
265 /* allocate dst with ip6_dst_ops */
266 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
267 struct net_device *dev,
269 struct fib6_table *table)
271 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
272 0, DST_OBSOLETE_FORCE_CHK, flags);
275 struct dst_entry *dst = &rt->dst;
277 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
278 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
279 rt->rt6i_genid = rt_genid(net);
280 INIT_LIST_HEAD(&rt->rt6i_siblings);
281 rt->rt6i_nsiblings = 0;
286 static void ip6_dst_destroy(struct dst_entry *dst)
288 struct rt6_info *rt = (struct rt6_info *)dst;
289 struct inet6_dev *idev = rt->rt6i_idev;
291 if (!(rt->dst.flags & DST_HOST))
292 dst_destroy_metrics_generic(dst);
295 rt->rt6i_idev = NULL;
299 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
300 dst_release(dst->from);
302 if (rt6_has_peer(rt)) {
303 struct inet_peer *peer = rt6_peer_ptr(rt);
308 void rt6_bind_peer(struct rt6_info *rt, int create)
310 struct inet_peer_base *base;
311 struct inet_peer *peer;
313 base = inetpeer_base_ptr(rt->_rt6i_peer);
317 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
319 if (!rt6_set_peer(rt, peer))
324 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
327 struct rt6_info *rt = (struct rt6_info *)dst;
328 struct inet6_dev *idev = rt->rt6i_idev;
329 struct net_device *loopback_dev =
330 dev_net(dev)->loopback_dev;
332 if (dev != loopback_dev) {
333 if (idev && idev->dev == dev) {
334 struct inet6_dev *loopback_idev =
335 in6_dev_get(loopback_dev);
337 rt->rt6i_idev = loopback_idev;
344 static bool rt6_check_expired(const struct rt6_info *rt)
346 if (rt->rt6i_flags & RTF_EXPIRES) {
347 if (time_after(jiffies, rt->dst.expires))
349 } else if (rt->dst.from) {
350 return rt6_check_expired((struct rt6_info *) rt->dst.from);
355 static bool rt6_need_strict(const struct in6_addr *daddr)
357 return ipv6_addr_type(daddr) &
358 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
361 /* Multipath route selection:
362 * Hash based function using packet header and flowlabel.
363 * Adapted from fib_info_hashfn()
365 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
366 const struct flowi6 *fl6)
368 unsigned int val = fl6->flowi6_proto;
370 val ^= ipv6_addr_hash(&fl6->daddr);
371 val ^= ipv6_addr_hash(&fl6->saddr);
373 /* Work only if this not encapsulated */
374 switch (fl6->flowi6_proto) {
378 val ^= (__force u16)fl6->fl6_sport;
379 val ^= (__force u16)fl6->fl6_dport;
383 val ^= (__force u16)fl6->fl6_icmp_type;
384 val ^= (__force u16)fl6->fl6_icmp_code;
387 /* RFC6438 recommands to use flowlabel */
388 val ^= (__force u32)fl6->flowlabel;
390 /* Perhaps, we need to tune, this function? */
391 val = val ^ (val >> 7) ^ (val >> 12);
392 return val % candidate_count;
395 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
398 struct rt6_info *sibling, *next_sibling;
401 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
402 /* Don't change the route, if route_choosen == 0
403 * (siblings does not include ourself)
406 list_for_each_entry_safe(sibling, next_sibling,
407 &match->rt6i_siblings, rt6i_siblings) {
409 if (route_choosen == 0) {
418 * Route lookup. Any table->tb6_lock is implied.
421 static inline struct rt6_info *rt6_device_match(struct net *net,
423 const struct in6_addr *saddr,
427 struct rt6_info *local = NULL;
428 struct rt6_info *sprt;
430 if (!oif && ipv6_addr_any(saddr))
433 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
434 struct net_device *dev = sprt->dst.dev;
437 if (dev->ifindex == oif)
439 if (dev->flags & IFF_LOOPBACK) {
440 if (!sprt->rt6i_idev ||
441 sprt->rt6i_idev->dev->ifindex != oif) {
442 if (flags & RT6_LOOKUP_F_IFACE && oif)
444 if (local && (!oif ||
445 local->rt6i_idev->dev->ifindex == oif))
451 if (ipv6_chk_addr(net, saddr, dev,
452 flags & RT6_LOOKUP_F_IFACE))
461 if (flags & RT6_LOOKUP_F_IFACE)
462 return net->ipv6.ip6_null_entry;
468 #ifdef CONFIG_IPV6_ROUTER_PREF
469 static void rt6_probe(struct rt6_info *rt)
471 struct neighbour *neigh;
473 * Okay, this does not seem to be appropriate
474 * for now, however, we need to check if it
475 * is really so; aka Router Reachability Probing.
477 * Router Reachability Probe MUST be rate-limited
478 * to no more than one per minute.
480 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
483 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
485 write_lock(&neigh->lock);
486 if (neigh->nud_state & NUD_VALID)
491 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
492 struct in6_addr mcaddr;
493 struct in6_addr *target;
496 neigh->updated = jiffies;
497 write_unlock(&neigh->lock);
500 target = (struct in6_addr *)&rt->rt6i_gateway;
501 addrconf_addr_solict_mult(target, &mcaddr);
502 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
505 write_unlock(&neigh->lock);
507 rcu_read_unlock_bh();
510 static inline void rt6_probe(struct rt6_info *rt)
516 * Default Router Selection (RFC 2461 6.3.6)
518 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
520 struct net_device *dev = rt->dst.dev;
521 if (!oif || dev->ifindex == oif)
523 if ((dev->flags & IFF_LOOPBACK) &&
524 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
529 static inline bool rt6_check_neigh(struct rt6_info *rt)
531 struct neighbour *neigh;
534 if (rt->rt6i_flags & RTF_NONEXTHOP ||
535 !(rt->rt6i_flags & RTF_GATEWAY))
539 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
541 read_lock(&neigh->lock);
542 if (neigh->nud_state & NUD_VALID)
544 #ifdef CONFIG_IPV6_ROUTER_PREF
545 else if (!(neigh->nud_state & NUD_FAILED))
548 read_unlock(&neigh->lock);
550 rcu_read_unlock_bh();
555 static int rt6_score_route(struct rt6_info *rt, int oif,
560 m = rt6_check_dev(rt, oif);
561 if (!m && (strict & RT6_LOOKUP_F_IFACE))
563 #ifdef CONFIG_IPV6_ROUTER_PREF
564 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
566 if (!rt6_check_neigh(rt) && (strict & RT6_LOOKUP_F_REACHABLE))
571 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
572 int *mpri, struct rt6_info *match)
576 if (rt6_check_expired(rt))
579 m = rt6_score_route(rt, oif, strict);
584 if (strict & RT6_LOOKUP_F_REACHABLE)
588 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
596 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
597 struct rt6_info *rr_head,
598 u32 metric, int oif, int strict)
600 struct rt6_info *rt, *match;
604 for (rt = rr_head; rt && rt->rt6i_metric == metric;
605 rt = rt->dst.rt6_next)
606 match = find_match(rt, oif, strict, &mpri, match);
607 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
608 rt = rt->dst.rt6_next)
609 match = find_match(rt, oif, strict, &mpri, match);
614 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
616 struct rt6_info *match, *rt0;
621 fn->rr_ptr = rt0 = fn->leaf;
623 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
626 (strict & RT6_LOOKUP_F_REACHABLE)) {
627 struct rt6_info *next = rt0->dst.rt6_next;
629 /* no entries matched; do round-robin */
630 if (!next || next->rt6i_metric != rt0->rt6i_metric)
637 net = dev_net(rt0->dst.dev);
638 return match ? match : net->ipv6.ip6_null_entry;
641 #ifdef CONFIG_IPV6_ROUTE_INFO
642 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
643 const struct in6_addr *gwaddr)
645 struct net *net = dev_net(dev);
646 struct route_info *rinfo = (struct route_info *) opt;
647 struct in6_addr prefix_buf, *prefix;
649 unsigned long lifetime;
652 if (len < sizeof(struct route_info)) {
656 /* Sanity check for prefix_len and length */
657 if (rinfo->length > 3) {
659 } else if (rinfo->prefix_len > 128) {
661 } else if (rinfo->prefix_len > 64) {
662 if (rinfo->length < 2) {
665 } else if (rinfo->prefix_len > 0) {
666 if (rinfo->length < 1) {
671 pref = rinfo->route_pref;
672 if (pref == ICMPV6_ROUTER_PREF_INVALID)
675 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
677 if (rinfo->length == 3)
678 prefix = (struct in6_addr *)rinfo->prefix;
680 /* this function is safe */
681 ipv6_addr_prefix(&prefix_buf,
682 (struct in6_addr *)rinfo->prefix,
684 prefix = &prefix_buf;
687 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
690 if (rt && !lifetime) {
696 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
699 rt->rt6i_flags = RTF_ROUTEINFO |
700 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
703 if (!addrconf_finite_timeout(lifetime))
704 rt6_clean_expires(rt);
706 rt6_set_expires(rt, jiffies + HZ * lifetime);
714 #define BACKTRACK(__net, saddr) \
716 if (rt == __net->ipv6.ip6_null_entry) { \
717 struct fib6_node *pn; \
719 if (fn->fn_flags & RTN_TL_ROOT) \
722 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
723 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
726 if (fn->fn_flags & RTN_RTINFO) \
732 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
733 struct fib6_table *table,
734 struct flowi6 *fl6, int flags)
736 struct fib6_node *fn;
739 read_lock_bh(&table->tb6_lock);
740 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
743 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
744 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
745 rt = rt6_multipath_select(rt, fl6);
746 BACKTRACK(net, &fl6->saddr);
748 dst_use(&rt->dst, jiffies);
749 read_unlock_bh(&table->tb6_lock);
754 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
757 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
759 EXPORT_SYMBOL_GPL(ip6_route_lookup);
761 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
762 const struct in6_addr *saddr, int oif, int strict)
764 struct flowi6 fl6 = {
768 struct dst_entry *dst;
769 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
772 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
773 flags |= RT6_LOOKUP_F_HAS_SADDR;
776 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
778 return (struct rt6_info *) dst;
785 EXPORT_SYMBOL(rt6_lookup);
787 /* ip6_ins_rt is called with FREE table->tb6_lock.
788 It takes new route entry, the addition fails by any reason the
789 route is freed. In any case, if caller does not hold it, it may
793 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
796 struct fib6_table *table;
798 table = rt->rt6i_table;
799 write_lock_bh(&table->tb6_lock);
800 err = fib6_add(&table->tb6_root, rt, info);
801 write_unlock_bh(&table->tb6_lock);
806 int ip6_ins_rt(struct rt6_info *rt)
808 struct nl_info info = {
809 .nl_net = dev_net(rt->dst.dev),
811 return __ip6_ins_rt(rt, &info);
814 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
815 const struct in6_addr *daddr,
816 const struct in6_addr *saddr)
824 rt = ip6_rt_copy(ort, daddr);
827 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
828 if (ort->rt6i_dst.plen != 128 &&
829 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
830 rt->rt6i_flags |= RTF_ANYCAST;
831 rt->rt6i_gateway = *daddr;
834 rt->rt6i_flags |= RTF_CACHE;
836 #ifdef CONFIG_IPV6_SUBTREES
837 if (rt->rt6i_src.plen && saddr) {
838 rt->rt6i_src.addr = *saddr;
839 rt->rt6i_src.plen = 128;
847 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
848 const struct in6_addr *daddr)
850 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
853 rt->rt6i_flags |= RTF_CACHE;
857 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
858 struct flowi6 *fl6, int flags)
860 struct fib6_node *fn;
861 struct rt6_info *rt, *nrt;
865 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
867 strict |= flags & RT6_LOOKUP_F_IFACE;
870 read_lock_bh(&table->tb6_lock);
873 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
876 rt = rt6_select(fn, oif, strict | reachable);
877 if (rt->rt6i_nsiblings && oif == 0)
878 rt = rt6_multipath_select(rt, fl6);
879 BACKTRACK(net, &fl6->saddr);
880 if (rt == net->ipv6.ip6_null_entry ||
881 rt->rt6i_flags & RTF_CACHE)
885 read_unlock_bh(&table->tb6_lock);
887 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
888 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
889 else if (!(rt->dst.flags & DST_HOST))
890 nrt = rt6_alloc_clone(rt, &fl6->daddr);
895 rt = nrt ? : net->ipv6.ip6_null_entry;
899 err = ip6_ins_rt(nrt);
908 * Race condition! In the gap, when table->tb6_lock was
909 * released someone could insert this route. Relookup.
920 read_unlock_bh(&table->tb6_lock);
922 rt->dst.lastuse = jiffies;
928 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
929 struct flowi6 *fl6, int flags)
931 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
934 static struct dst_entry *ip6_route_input_lookup(struct net *net,
935 struct net_device *dev,
936 struct flowi6 *fl6, int flags)
938 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
939 flags |= RT6_LOOKUP_F_IFACE;
941 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
944 void ip6_route_input(struct sk_buff *skb)
946 const struct ipv6hdr *iph = ipv6_hdr(skb);
947 struct net *net = dev_net(skb->dev);
948 int flags = RT6_LOOKUP_F_HAS_SADDR;
949 struct flowi6 fl6 = {
950 .flowi6_iif = skb->dev->ifindex,
953 .flowlabel = ip6_flowinfo(iph),
954 .flowi6_mark = skb->mark,
955 .flowi6_proto = iph->nexthdr,
958 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
961 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
962 struct flowi6 *fl6, int flags)
964 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
967 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
972 fl6->flowi6_iif = LOOPBACK_IFINDEX;
974 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
975 flags |= RT6_LOOKUP_F_IFACE;
977 if (!ipv6_addr_any(&fl6->saddr))
978 flags |= RT6_LOOKUP_F_HAS_SADDR;
980 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
982 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
985 EXPORT_SYMBOL(ip6_route_output);
987 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
989 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
990 struct dst_entry *new = NULL;
992 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
996 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
997 rt6_init_peer(rt, net->ipv6.peers);
1000 new->input = dst_discard;
1001 new->output = dst_discard;
1003 if (dst_metrics_read_only(&ort->dst))
1004 new->_metrics = ort->dst._metrics;
1006 dst_copy_metrics(new, &ort->dst);
1007 rt->rt6i_idev = ort->rt6i_idev;
1009 in6_dev_hold(rt->rt6i_idev);
1011 rt->rt6i_gateway = ort->rt6i_gateway;
1012 rt->rt6i_flags = ort->rt6i_flags;
1013 rt6_clean_expires(rt);
1014 rt->rt6i_metric = 0;
1016 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1017 #ifdef CONFIG_IPV6_SUBTREES
1018 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1024 dst_release(dst_orig);
1025 return new ? new : ERR_PTR(-ENOMEM);
1029 * Destination cache support functions
1032 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1034 struct rt6_info *rt;
1036 rt = (struct rt6_info *) dst;
1038 /* All IPV6 dsts are created with ->obsolete set to the value
1039 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1040 * into this function always.
1042 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1045 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1051 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1053 struct rt6_info *rt = (struct rt6_info *) dst;
1056 if (rt->rt6i_flags & RTF_CACHE) {
1057 if (rt6_check_expired(rt)) {
1069 static void ip6_link_failure(struct sk_buff *skb)
1071 struct rt6_info *rt;
1073 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1075 rt = (struct rt6_info *) skb_dst(skb);
1077 if (rt->rt6i_flags & RTF_CACHE)
1078 rt6_update_expires(rt, 0);
1079 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1080 rt->rt6i_node->fn_sernum = -1;
1084 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1085 struct sk_buff *skb, u32 mtu)
1087 struct rt6_info *rt6 = (struct rt6_info*)dst;
1090 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1091 struct net *net = dev_net(dst->dev);
1093 rt6->rt6i_flags |= RTF_MODIFIED;
1094 if (mtu < IPV6_MIN_MTU) {
1095 u32 features = dst_metric(dst, RTAX_FEATURES);
1097 features |= RTAX_FEATURE_ALLFRAG;
1098 dst_metric_set(dst, RTAX_FEATURES, features);
1100 dst_metric_set(dst, RTAX_MTU, mtu);
1101 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1105 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1108 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1109 struct dst_entry *dst;
1112 memset(&fl6, 0, sizeof(fl6));
1113 fl6.flowi6_oif = oif;
1114 fl6.flowi6_mark = mark;
1115 fl6.flowi6_flags = 0;
1116 fl6.daddr = iph->daddr;
1117 fl6.saddr = iph->saddr;
1118 fl6.flowlabel = ip6_flowinfo(iph);
1120 dst = ip6_route_output(net, NULL, &fl6);
1122 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1125 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1127 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1129 ip6_update_pmtu(skb, sock_net(sk), mtu,
1130 sk->sk_bound_dev_if, sk->sk_mark);
1132 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1134 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1136 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1137 struct dst_entry *dst;
1140 memset(&fl6, 0, sizeof(fl6));
1141 fl6.flowi6_oif = oif;
1142 fl6.flowi6_mark = mark;
1143 fl6.flowi6_flags = 0;
1144 fl6.daddr = iph->daddr;
1145 fl6.saddr = iph->saddr;
1146 fl6.flowlabel = ip6_flowinfo(iph);
1148 dst = ip6_route_output(net, NULL, &fl6);
1150 rt6_do_redirect(dst, NULL, skb);
1153 EXPORT_SYMBOL_GPL(ip6_redirect);
1155 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1157 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1159 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1161 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1163 struct net_device *dev = dst->dev;
1164 unsigned int mtu = dst_mtu(dst);
1165 struct net *net = dev_net(dev);
1167 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1169 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1170 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1173 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1174 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1175 * IPV6_MAXPLEN is also valid and means: "any MSS,
1176 * rely only on pmtu discovery"
1178 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1183 static unsigned int ip6_mtu(const struct dst_entry *dst)
1185 struct inet6_dev *idev;
1186 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1194 idev = __in6_dev_get(dst->dev);
1196 mtu = idev->cnf.mtu6;
1202 static struct dst_entry *icmp6_dst_gc_list;
1203 static DEFINE_SPINLOCK(icmp6_dst_lock);
1205 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1208 struct dst_entry *dst;
1209 struct rt6_info *rt;
1210 struct inet6_dev *idev = in6_dev_get(dev);
1211 struct net *net = dev_net(dev);
1213 if (unlikely(!idev))
1214 return ERR_PTR(-ENODEV);
1216 rt = ip6_dst_alloc(net, dev, 0, NULL);
1217 if (unlikely(!rt)) {
1219 dst = ERR_PTR(-ENOMEM);
1223 rt->dst.flags |= DST_HOST;
1224 rt->dst.output = ip6_output;
1225 atomic_set(&rt->dst.__refcnt, 1);
1226 rt->rt6i_dst.addr = fl6->daddr;
1227 rt->rt6i_dst.plen = 128;
1228 rt->rt6i_idev = idev;
1229 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1231 spin_lock_bh(&icmp6_dst_lock);
1232 rt->dst.next = icmp6_dst_gc_list;
1233 icmp6_dst_gc_list = &rt->dst;
1234 spin_unlock_bh(&icmp6_dst_lock);
1236 fib6_force_start_gc(net);
1238 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1244 int icmp6_dst_gc(void)
1246 struct dst_entry *dst, **pprev;
1249 spin_lock_bh(&icmp6_dst_lock);
1250 pprev = &icmp6_dst_gc_list;
1252 while ((dst = *pprev) != NULL) {
1253 if (!atomic_read(&dst->__refcnt)) {
1262 spin_unlock_bh(&icmp6_dst_lock);
1267 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1270 struct dst_entry *dst, **pprev;
1272 spin_lock_bh(&icmp6_dst_lock);
1273 pprev = &icmp6_dst_gc_list;
1274 while ((dst = *pprev) != NULL) {
1275 struct rt6_info *rt = (struct rt6_info *) dst;
1276 if (func(rt, arg)) {
1283 spin_unlock_bh(&icmp6_dst_lock);
1286 static int ip6_dst_gc(struct dst_ops *ops)
1288 unsigned long now = jiffies;
1289 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1290 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1291 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1292 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1293 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1294 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1297 entries = dst_entries_get_fast(ops);
1298 if (time_after(rt_last_gc + rt_min_interval, now) &&
1299 entries <= rt_max_size)
1302 net->ipv6.ip6_rt_gc_expire++;
1303 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1304 net->ipv6.ip6_rt_last_gc = now;
1305 entries = dst_entries_get_slow(ops);
1306 if (entries < ops->gc_thresh)
1307 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1309 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1310 return entries > rt_max_size;
1313 int ip6_dst_hoplimit(struct dst_entry *dst)
1315 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1316 if (hoplimit == 0) {
1317 struct net_device *dev = dst->dev;
1318 struct inet6_dev *idev;
1321 idev = __in6_dev_get(dev);
1323 hoplimit = idev->cnf.hop_limit;
1325 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1330 EXPORT_SYMBOL(ip6_dst_hoplimit);
1336 int ip6_route_add(struct fib6_config *cfg)
1339 struct net *net = cfg->fc_nlinfo.nl_net;
1340 struct rt6_info *rt = NULL;
1341 struct net_device *dev = NULL;
1342 struct inet6_dev *idev = NULL;
1343 struct fib6_table *table;
1346 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1348 #ifndef CONFIG_IPV6_SUBTREES
1349 if (cfg->fc_src_len)
1352 if (cfg->fc_ifindex) {
1354 dev = dev_get_by_index(net, cfg->fc_ifindex);
1357 idev = in6_dev_get(dev);
1362 if (cfg->fc_metric == 0)
1363 cfg->fc_metric = IP6_RT_PRIO_USER;
1366 if (cfg->fc_nlinfo.nlh &&
1367 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1368 table = fib6_get_table(net, cfg->fc_table);
1370 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1371 table = fib6_new_table(net, cfg->fc_table);
1374 table = fib6_new_table(net, cfg->fc_table);
1380 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1387 if (cfg->fc_flags & RTF_EXPIRES)
1388 rt6_set_expires(rt, jiffies +
1389 clock_t_to_jiffies(cfg->fc_expires));
1391 rt6_clean_expires(rt);
1393 if (cfg->fc_protocol == RTPROT_UNSPEC)
1394 cfg->fc_protocol = RTPROT_BOOT;
1395 rt->rt6i_protocol = cfg->fc_protocol;
1397 addr_type = ipv6_addr_type(&cfg->fc_dst);
1399 if (addr_type & IPV6_ADDR_MULTICAST)
1400 rt->dst.input = ip6_mc_input;
1401 else if (cfg->fc_flags & RTF_LOCAL)
1402 rt->dst.input = ip6_input;
1404 rt->dst.input = ip6_forward;
1406 rt->dst.output = ip6_output;
1408 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1409 rt->rt6i_dst.plen = cfg->fc_dst_len;
1410 if (rt->rt6i_dst.plen == 128)
1411 rt->dst.flags |= DST_HOST;
1413 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1414 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1419 dst_init_metrics(&rt->dst, metrics, 0);
1421 #ifdef CONFIG_IPV6_SUBTREES
1422 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1423 rt->rt6i_src.plen = cfg->fc_src_len;
1426 rt->rt6i_metric = cfg->fc_metric;
1428 /* We cannot add true routes via loopback here,
1429 they would result in kernel looping; promote them to reject routes
1431 if ((cfg->fc_flags & RTF_REJECT) ||
1432 (dev && (dev->flags & IFF_LOOPBACK) &&
1433 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1434 !(cfg->fc_flags & RTF_LOCAL))) {
1435 /* hold loopback dev/idev if we haven't done so. */
1436 if (dev != net->loopback_dev) {
1441 dev = net->loopback_dev;
1443 idev = in6_dev_get(dev);
1449 rt->dst.output = ip6_pkt_discard_out;
1450 rt->dst.input = ip6_pkt_discard;
1451 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1452 switch (cfg->fc_type) {
1454 rt->dst.error = -EINVAL;
1457 rt->dst.error = -EACCES;
1460 rt->dst.error = -EAGAIN;
1463 rt->dst.error = -ENETUNREACH;
1469 if (cfg->fc_flags & RTF_GATEWAY) {
1470 const struct in6_addr *gw_addr;
1473 gw_addr = &cfg->fc_gateway;
1474 rt->rt6i_gateway = *gw_addr;
1475 gwa_type = ipv6_addr_type(gw_addr);
1477 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1478 struct rt6_info *grt;
1480 /* IPv6 strictly inhibits using not link-local
1481 addresses as nexthop address.
1482 Otherwise, router will not able to send redirects.
1483 It is very good, but in some (rare!) circumstances
1484 (SIT, PtP, NBMA NOARP links) it is handy to allow
1485 some exceptions. --ANK
1488 if (!(gwa_type & IPV6_ADDR_UNICAST))
1491 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1493 err = -EHOSTUNREACH;
1497 if (dev != grt->dst.dev) {
1503 idev = grt->rt6i_idev;
1505 in6_dev_hold(grt->rt6i_idev);
1507 if (!(grt->rt6i_flags & RTF_GATEWAY))
1515 if (!dev || (dev->flags & IFF_LOOPBACK))
1523 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1524 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1528 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1529 rt->rt6i_prefsrc.plen = 128;
1531 rt->rt6i_prefsrc.plen = 0;
1533 rt->rt6i_flags = cfg->fc_flags;
1540 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1541 int type = nla_type(nla);
1544 if (type > RTAX_MAX) {
1549 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1555 rt->rt6i_idev = idev;
1556 rt->rt6i_table = table;
1558 cfg->fc_nlinfo.nl_net = dev_net(dev);
1560 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1572 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1575 struct fib6_table *table;
1576 struct net *net = dev_net(rt->dst.dev);
1578 if (rt == net->ipv6.ip6_null_entry) {
1583 table = rt->rt6i_table;
1584 write_lock_bh(&table->tb6_lock);
1585 err = fib6_del(rt, info);
1586 write_unlock_bh(&table->tb6_lock);
1593 int ip6_del_rt(struct rt6_info *rt)
1595 struct nl_info info = {
1596 .nl_net = dev_net(rt->dst.dev),
1598 return __ip6_del_rt(rt, &info);
1601 static int ip6_route_del(struct fib6_config *cfg)
1603 struct fib6_table *table;
1604 struct fib6_node *fn;
1605 struct rt6_info *rt;
1608 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1612 read_lock_bh(&table->tb6_lock);
1614 fn = fib6_locate(&table->tb6_root,
1615 &cfg->fc_dst, cfg->fc_dst_len,
1616 &cfg->fc_src, cfg->fc_src_len);
1619 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1620 if (cfg->fc_ifindex &&
1622 rt->dst.dev->ifindex != cfg->fc_ifindex))
1624 if (cfg->fc_flags & RTF_GATEWAY &&
1625 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1627 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1630 read_unlock_bh(&table->tb6_lock);
1632 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1635 read_unlock_bh(&table->tb6_lock);
1640 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1642 struct net *net = dev_net(skb->dev);
1643 struct netevent_redirect netevent;
1644 struct rt6_info *rt, *nrt = NULL;
1645 struct ndisc_options ndopts;
1646 struct inet6_dev *in6_dev;
1647 struct neighbour *neigh;
1649 int optlen, on_link;
1652 optlen = skb->tail - skb->transport_header;
1653 optlen -= sizeof(*msg);
1656 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1660 msg = (struct rd_msg *)icmp6_hdr(skb);
1662 if (ipv6_addr_is_multicast(&msg->dest)) {
1663 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1668 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1670 } else if (ipv6_addr_type(&msg->target) !=
1671 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1672 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1676 in6_dev = __in6_dev_get(skb->dev);
1679 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1683 * The IP source address of the Redirect MUST be the same as the current
1684 * first-hop router for the specified ICMP Destination Address.
1687 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1688 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1693 if (ndopts.nd_opts_tgt_lladdr) {
1694 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1697 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1702 rt = (struct rt6_info *) dst;
1703 if (rt == net->ipv6.ip6_null_entry) {
1704 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1708 /* Redirect received -> path was valid.
1709 * Look, redirects are sent only in response to data packets,
1710 * so that this nexthop apparently is reachable. --ANK
1712 dst_confirm(&rt->dst);
1714 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1719 * We have finally decided to accept it.
1722 neigh_update(neigh, lladdr, NUD_STALE,
1723 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1724 NEIGH_UPDATE_F_OVERRIDE|
1725 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1726 NEIGH_UPDATE_F_ISROUTER))
1729 nrt = ip6_rt_copy(rt, &msg->dest);
1733 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1735 nrt->rt6i_flags &= ~RTF_GATEWAY;
1737 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1739 if (ip6_ins_rt(nrt))
1742 netevent.old = &rt->dst;
1743 netevent.new = &nrt->dst;
1744 netevent.daddr = &msg->dest;
1745 netevent.neigh = neigh;
1746 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1748 if (rt->rt6i_flags & RTF_CACHE) {
1749 rt = (struct rt6_info *) dst_clone(&rt->dst);
1754 neigh_release(neigh);
1758 * Misc support functions
1761 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1762 const struct in6_addr *dest)
1764 struct net *net = dev_net(ort->dst.dev);
1765 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1769 rt->dst.input = ort->dst.input;
1770 rt->dst.output = ort->dst.output;
1771 rt->dst.flags |= DST_HOST;
1773 rt->rt6i_dst.addr = *dest;
1774 rt->rt6i_dst.plen = 128;
1775 dst_copy_metrics(&rt->dst, &ort->dst);
1776 rt->dst.error = ort->dst.error;
1777 rt->rt6i_idev = ort->rt6i_idev;
1779 in6_dev_hold(rt->rt6i_idev);
1780 rt->dst.lastuse = jiffies;
1782 rt->rt6i_gateway = ort->rt6i_gateway;
1783 rt->rt6i_flags = ort->rt6i_flags;
1784 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1785 (RTF_DEFAULT | RTF_ADDRCONF))
1786 rt6_set_from(rt, ort);
1788 rt6_clean_expires(rt);
1789 rt->rt6i_metric = 0;
1791 #ifdef CONFIG_IPV6_SUBTREES
1792 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1794 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1795 rt->rt6i_table = ort->rt6i_table;
1800 #ifdef CONFIG_IPV6_ROUTE_INFO
1801 static struct rt6_info *rt6_get_route_info(struct net *net,
1802 const struct in6_addr *prefix, int prefixlen,
1803 const struct in6_addr *gwaddr, int ifindex)
1805 struct fib6_node *fn;
1806 struct rt6_info *rt = NULL;
1807 struct fib6_table *table;
1809 table = fib6_get_table(net, RT6_TABLE_INFO);
1813 read_lock_bh(&table->tb6_lock);
1814 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1818 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1819 if (rt->dst.dev->ifindex != ifindex)
1821 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1823 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1829 read_unlock_bh(&table->tb6_lock);
1833 static struct rt6_info *rt6_add_route_info(struct net *net,
1834 const struct in6_addr *prefix, int prefixlen,
1835 const struct in6_addr *gwaddr, int ifindex,
1838 struct fib6_config cfg = {
1839 .fc_table = RT6_TABLE_INFO,
1840 .fc_metric = IP6_RT_PRIO_USER,
1841 .fc_ifindex = ifindex,
1842 .fc_dst_len = prefixlen,
1843 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1844 RTF_UP | RTF_PREF(pref),
1845 .fc_nlinfo.portid = 0,
1846 .fc_nlinfo.nlh = NULL,
1847 .fc_nlinfo.nl_net = net,
1850 cfg.fc_dst = *prefix;
1851 cfg.fc_gateway = *gwaddr;
1853 /* We should treat it as a default route if prefix length is 0. */
1855 cfg.fc_flags |= RTF_DEFAULT;
1857 ip6_route_add(&cfg);
1859 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1863 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1865 struct rt6_info *rt;
1866 struct fib6_table *table;
1868 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1872 read_lock_bh(&table->tb6_lock);
1873 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1874 if (dev == rt->dst.dev &&
1875 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1876 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1881 read_unlock_bh(&table->tb6_lock);
1885 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1886 struct net_device *dev,
1889 struct fib6_config cfg = {
1890 .fc_table = RT6_TABLE_DFLT,
1891 .fc_metric = IP6_RT_PRIO_USER,
1892 .fc_ifindex = dev->ifindex,
1893 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1894 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1895 .fc_nlinfo.portid = 0,
1896 .fc_nlinfo.nlh = NULL,
1897 .fc_nlinfo.nl_net = dev_net(dev),
1900 cfg.fc_gateway = *gwaddr;
1902 ip6_route_add(&cfg);
1904 return rt6_get_dflt_router(gwaddr, dev);
1907 void rt6_purge_dflt_routers(struct net *net)
1909 struct rt6_info *rt;
1910 struct fib6_table *table;
1912 /* NOTE: Keep consistent with rt6_get_dflt_router */
1913 table = fib6_get_table(net, RT6_TABLE_DFLT);
1918 read_lock_bh(&table->tb6_lock);
1919 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1920 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1922 read_unlock_bh(&table->tb6_lock);
1927 read_unlock_bh(&table->tb6_lock);
1930 static void rtmsg_to_fib6_config(struct net *net,
1931 struct in6_rtmsg *rtmsg,
1932 struct fib6_config *cfg)
1934 memset(cfg, 0, sizeof(*cfg));
1936 cfg->fc_table = RT6_TABLE_MAIN;
1937 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1938 cfg->fc_metric = rtmsg->rtmsg_metric;
1939 cfg->fc_expires = rtmsg->rtmsg_info;
1940 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1941 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1942 cfg->fc_flags = rtmsg->rtmsg_flags;
1944 cfg->fc_nlinfo.nl_net = net;
1946 cfg->fc_dst = rtmsg->rtmsg_dst;
1947 cfg->fc_src = rtmsg->rtmsg_src;
1948 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1951 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1953 struct fib6_config cfg;
1954 struct in6_rtmsg rtmsg;
1958 case SIOCADDRT: /* Add a route */
1959 case SIOCDELRT: /* Delete a route */
1960 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1962 err = copy_from_user(&rtmsg, arg,
1963 sizeof(struct in6_rtmsg));
1967 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1972 err = ip6_route_add(&cfg);
1975 err = ip6_route_del(&cfg);
1989 * Drop the packet on the floor
1992 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1995 struct dst_entry *dst = skb_dst(skb);
1996 switch (ipstats_mib_noroutes) {
1997 case IPSTATS_MIB_INNOROUTES:
1998 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1999 if (type == IPV6_ADDR_ANY) {
2000 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2001 IPSTATS_MIB_INADDRERRORS);
2005 case IPSTATS_MIB_OUTNOROUTES:
2006 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2007 ipstats_mib_noroutes);
2010 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2015 static int ip6_pkt_discard(struct sk_buff *skb)
2017 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2020 static int ip6_pkt_discard_out(struct sk_buff *skb)
2022 skb->dev = skb_dst(skb)->dev;
2023 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2026 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2028 static int ip6_pkt_prohibit(struct sk_buff *skb)
2030 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2033 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2035 skb->dev = skb_dst(skb)->dev;
2036 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2042 * Allocate a dst for local (unicast / anycast) address.
2045 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2046 const struct in6_addr *addr,
2049 struct net *net = dev_net(idev->dev);
2050 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2053 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2054 return ERR_PTR(-ENOMEM);
2059 rt->dst.flags |= DST_HOST;
2060 rt->dst.input = ip6_input;
2061 rt->dst.output = ip6_output;
2062 rt->rt6i_idev = idev;
2064 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2066 rt->rt6i_flags |= RTF_ANYCAST;
2068 rt->rt6i_flags |= RTF_LOCAL;
2070 rt->rt6i_dst.addr = *addr;
2071 rt->rt6i_dst.plen = 128;
2072 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2074 atomic_set(&rt->dst.__refcnt, 1);
2079 int ip6_route_get_saddr(struct net *net,
2080 struct rt6_info *rt,
2081 const struct in6_addr *daddr,
2083 struct in6_addr *saddr)
2085 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2087 if (rt->rt6i_prefsrc.plen)
2088 *saddr = rt->rt6i_prefsrc.addr;
2090 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2091 daddr, prefs, saddr);
2095 /* remove deleted ip from prefsrc entries */
2096 struct arg_dev_net_ip {
2097 struct net_device *dev;
2099 struct in6_addr *addr;
2102 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2104 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2105 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2106 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2108 if (((void *)rt->dst.dev == dev || !dev) &&
2109 rt != net->ipv6.ip6_null_entry &&
2110 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2111 /* remove prefsrc entry */
2112 rt->rt6i_prefsrc.plen = 0;
2117 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2119 struct net *net = dev_net(ifp->idev->dev);
2120 struct arg_dev_net_ip adni = {
2121 .dev = ifp->idev->dev,
2125 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2128 struct arg_dev_net {
2129 struct net_device *dev;
2133 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2135 const struct arg_dev_net *adn = arg;
2136 const struct net_device *dev = adn->dev;
2138 if ((rt->dst.dev == dev || !dev) &&
2139 rt != adn->net->ipv6.ip6_null_entry)
2145 void rt6_ifdown(struct net *net, struct net_device *dev)
2147 struct arg_dev_net adn = {
2152 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2153 icmp6_clean_all(fib6_ifdown, &adn);
2156 struct rt6_mtu_change_arg {
2157 struct net_device *dev;
2161 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2163 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2164 struct inet6_dev *idev;
2166 /* In IPv6 pmtu discovery is not optional,
2167 so that RTAX_MTU lock cannot disable it.
2168 We still use this lock to block changes
2169 caused by addrconf/ndisc.
2172 idev = __in6_dev_get(arg->dev);
2176 /* For administrative MTU increase, there is no way to discover
2177 IPv6 PMTU increase, so PMTU increase should be updated here.
2178 Since RFC 1981 doesn't include administrative MTU increase
2179 update PMTU increase is a MUST. (i.e. jumbo frame)
2182 If new MTU is less than route PMTU, this new MTU will be the
2183 lowest MTU in the path, update the route PMTU to reflect PMTU
2184 decreases; if new MTU is greater than route PMTU, and the
2185 old MTU is the lowest MTU in the path, update the route PMTU
2186 to reflect the increase. In this case if the other nodes' MTU
2187 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2190 if (rt->dst.dev == arg->dev &&
2191 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2192 (dst_mtu(&rt->dst) >= arg->mtu ||
2193 (dst_mtu(&rt->dst) < arg->mtu &&
2194 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2195 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2200 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2202 struct rt6_mtu_change_arg arg = {
2207 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2210 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2211 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2212 [RTA_OIF] = { .type = NLA_U32 },
2213 [RTA_IIF] = { .type = NLA_U32 },
2214 [RTA_PRIORITY] = { .type = NLA_U32 },
2215 [RTA_METRICS] = { .type = NLA_NESTED },
2216 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2219 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2220 struct fib6_config *cfg)
2223 struct nlattr *tb[RTA_MAX+1];
2226 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2231 rtm = nlmsg_data(nlh);
2232 memset(cfg, 0, sizeof(*cfg));
2234 cfg->fc_table = rtm->rtm_table;
2235 cfg->fc_dst_len = rtm->rtm_dst_len;
2236 cfg->fc_src_len = rtm->rtm_src_len;
2237 cfg->fc_flags = RTF_UP;
2238 cfg->fc_protocol = rtm->rtm_protocol;
2239 cfg->fc_type = rtm->rtm_type;
2241 if (rtm->rtm_type == RTN_UNREACHABLE ||
2242 rtm->rtm_type == RTN_BLACKHOLE ||
2243 rtm->rtm_type == RTN_PROHIBIT ||
2244 rtm->rtm_type == RTN_THROW)
2245 cfg->fc_flags |= RTF_REJECT;
2247 if (rtm->rtm_type == RTN_LOCAL)
2248 cfg->fc_flags |= RTF_LOCAL;
2250 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2251 cfg->fc_nlinfo.nlh = nlh;
2252 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2254 if (tb[RTA_GATEWAY]) {
2255 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2256 cfg->fc_flags |= RTF_GATEWAY;
2260 int plen = (rtm->rtm_dst_len + 7) >> 3;
2262 if (nla_len(tb[RTA_DST]) < plen)
2265 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2269 int plen = (rtm->rtm_src_len + 7) >> 3;
2271 if (nla_len(tb[RTA_SRC]) < plen)
2274 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2277 if (tb[RTA_PREFSRC])
2278 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2281 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2283 if (tb[RTA_PRIORITY])
2284 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2286 if (tb[RTA_METRICS]) {
2287 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2288 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2292 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2294 if (tb[RTA_MULTIPATH]) {
2295 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2296 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2304 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2306 struct fib6_config r_cfg;
2307 struct rtnexthop *rtnh;
2310 int err = 0, last_err = 0;
2313 rtnh = (struct rtnexthop *)cfg->fc_mp;
2314 remaining = cfg->fc_mp_len;
2316 /* Parse a Multipath Entry */
2317 while (rtnh_ok(rtnh, remaining)) {
2318 memcpy(&r_cfg, cfg, sizeof(*cfg));
2319 if (rtnh->rtnh_ifindex)
2320 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2322 attrlen = rtnh_attrlen(rtnh);
2324 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2326 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2328 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2329 r_cfg.fc_flags |= RTF_GATEWAY;
2332 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2335 /* If we are trying to remove a route, do not stop the
2336 * loop when ip6_route_del() fails (because next hop is
2337 * already gone), we should try to remove all next hops.
2340 /* If add fails, we should try to delete all
2341 * next hops that have been already added.
2347 /* Because each route is added like a single route we remove
2348 * this flag after the first nexthop (if there is a collision,
2349 * we have already fail to add the first nexthop:
2350 * fib6_add_rt2node() has reject it).
2352 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2353 rtnh = rtnh_next(rtnh, &remaining);
2359 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2361 struct fib6_config cfg;
2364 err = rtm_to_fib6_config(skb, nlh, &cfg);
2369 return ip6_route_multipath(&cfg, 0);
2371 return ip6_route_del(&cfg);
2374 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2376 struct fib6_config cfg;
2379 err = rtm_to_fib6_config(skb, nlh, &cfg);
2384 return ip6_route_multipath(&cfg, 1);
2386 return ip6_route_add(&cfg);
2389 static inline size_t rt6_nlmsg_size(void)
2391 return NLMSG_ALIGN(sizeof(struct rtmsg))
2392 + nla_total_size(16) /* RTA_SRC */
2393 + nla_total_size(16) /* RTA_DST */
2394 + nla_total_size(16) /* RTA_GATEWAY */
2395 + nla_total_size(16) /* RTA_PREFSRC */
2396 + nla_total_size(4) /* RTA_TABLE */
2397 + nla_total_size(4) /* RTA_IIF */
2398 + nla_total_size(4) /* RTA_OIF */
2399 + nla_total_size(4) /* RTA_PRIORITY */
2400 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2401 + nla_total_size(sizeof(struct rta_cacheinfo));
2404 static int rt6_fill_node(struct net *net,
2405 struct sk_buff *skb, struct rt6_info *rt,
2406 struct in6_addr *dst, struct in6_addr *src,
2407 int iif, int type, u32 portid, u32 seq,
2408 int prefix, int nowait, unsigned int flags)
2411 struct nlmsghdr *nlh;
2415 if (prefix) { /* user wants prefix routes only */
2416 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2417 /* success since this is not a prefix route */
2422 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2426 rtm = nlmsg_data(nlh);
2427 rtm->rtm_family = AF_INET6;
2428 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2429 rtm->rtm_src_len = rt->rt6i_src.plen;
2432 table = rt->rt6i_table->tb6_id;
2434 table = RT6_TABLE_UNSPEC;
2435 rtm->rtm_table = table;
2436 if (nla_put_u32(skb, RTA_TABLE, table))
2437 goto nla_put_failure;
2438 if (rt->rt6i_flags & RTF_REJECT) {
2439 switch (rt->dst.error) {
2441 rtm->rtm_type = RTN_BLACKHOLE;
2444 rtm->rtm_type = RTN_PROHIBIT;
2447 rtm->rtm_type = RTN_THROW;
2450 rtm->rtm_type = RTN_UNREACHABLE;
2454 else if (rt->rt6i_flags & RTF_LOCAL)
2455 rtm->rtm_type = RTN_LOCAL;
2456 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2457 rtm->rtm_type = RTN_LOCAL;
2459 rtm->rtm_type = RTN_UNICAST;
2461 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2462 rtm->rtm_protocol = rt->rt6i_protocol;
2463 if (rt->rt6i_flags & RTF_DYNAMIC)
2464 rtm->rtm_protocol = RTPROT_REDIRECT;
2465 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2466 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2467 rtm->rtm_protocol = RTPROT_RA;
2469 rtm->rtm_protocol = RTPROT_KERNEL;
2472 if (rt->rt6i_flags & RTF_CACHE)
2473 rtm->rtm_flags |= RTM_F_CLONED;
2476 if (nla_put(skb, RTA_DST, 16, dst))
2477 goto nla_put_failure;
2478 rtm->rtm_dst_len = 128;
2479 } else if (rtm->rtm_dst_len)
2480 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2481 goto nla_put_failure;
2482 #ifdef CONFIG_IPV6_SUBTREES
2484 if (nla_put(skb, RTA_SRC, 16, src))
2485 goto nla_put_failure;
2486 rtm->rtm_src_len = 128;
2487 } else if (rtm->rtm_src_len &&
2488 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2489 goto nla_put_failure;
2492 #ifdef CONFIG_IPV6_MROUTE
2493 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2494 int err = ip6mr_get_route(net, skb, rtm, nowait);
2499 goto nla_put_failure;
2501 if (err == -EMSGSIZE)
2502 goto nla_put_failure;
2507 if (nla_put_u32(skb, RTA_IIF, iif))
2508 goto nla_put_failure;
2510 struct in6_addr saddr_buf;
2511 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2512 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2513 goto nla_put_failure;
2516 if (rt->rt6i_prefsrc.plen) {
2517 struct in6_addr saddr_buf;
2518 saddr_buf = rt->rt6i_prefsrc.addr;
2519 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2520 goto nla_put_failure;
2523 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2524 goto nla_put_failure;
2526 if (rt->rt6i_flags & RTF_GATEWAY) {
2527 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2528 goto nla_put_failure;
2532 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2533 goto nla_put_failure;
2534 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2535 goto nla_put_failure;
2537 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2539 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2540 goto nla_put_failure;
2542 return nlmsg_end(skb, nlh);
2545 nlmsg_cancel(skb, nlh);
2549 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2551 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2554 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2555 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2556 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2560 return rt6_fill_node(arg->net,
2561 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2562 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2563 prefix, 0, NLM_F_MULTI);
2566 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2568 struct net *net = sock_net(in_skb->sk);
2569 struct nlattr *tb[RTA_MAX+1];
2570 struct rt6_info *rt;
2571 struct sk_buff *skb;
2574 int err, iif = 0, oif = 0;
2576 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2581 memset(&fl6, 0, sizeof(fl6));
2584 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2587 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2591 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2594 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2598 iif = nla_get_u32(tb[RTA_IIF]);
2601 oif = nla_get_u32(tb[RTA_OIF]);
2604 struct net_device *dev;
2607 dev = __dev_get_by_index(net, iif);
2613 fl6.flowi6_iif = iif;
2615 if (!ipv6_addr_any(&fl6.saddr))
2616 flags |= RT6_LOOKUP_F_HAS_SADDR;
2618 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2621 fl6.flowi6_oif = oif;
2623 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2626 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2633 /* Reserve room for dummy headers, this skb can pass
2634 through good chunk of routing engine.
2636 skb_reset_mac_header(skb);
2637 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2639 skb_dst_set(skb, &rt->dst);
2641 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2642 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2643 nlh->nlmsg_seq, 0, 0, 0);
2649 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2654 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2656 struct sk_buff *skb;
2657 struct net *net = info->nl_net;
2662 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2664 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2668 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2669 event, info->portid, seq, 0, 0, 0);
2671 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2672 WARN_ON(err == -EMSGSIZE);
2676 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2677 info->nlh, gfp_any());
2681 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2684 static int ip6_route_dev_notify(struct notifier_block *this,
2685 unsigned long event, void *data)
2687 struct net_device *dev = (struct net_device *)data;
2688 struct net *net = dev_net(dev);
2690 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2691 net->ipv6.ip6_null_entry->dst.dev = dev;
2692 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2693 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2694 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2695 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2696 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2697 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2708 #ifdef CONFIG_PROC_FS
2719 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2721 struct seq_file *m = p_arg;
2723 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2725 #ifdef CONFIG_IPV6_SUBTREES
2726 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2728 seq_puts(m, "00000000000000000000000000000000 00 ");
2730 if (rt->rt6i_flags & RTF_GATEWAY) {
2731 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2733 seq_puts(m, "00000000000000000000000000000000");
2735 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2736 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2737 rt->dst.__use, rt->rt6i_flags,
2738 rt->dst.dev ? rt->dst.dev->name : "");
2742 static int ipv6_route_show(struct seq_file *m, void *v)
2744 struct net *net = (struct net *)m->private;
2745 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2749 static int ipv6_route_open(struct inode *inode, struct file *file)
2751 return single_open_net(inode, file, ipv6_route_show);
2754 static const struct file_operations ipv6_route_proc_fops = {
2755 .owner = THIS_MODULE,
2756 .open = ipv6_route_open,
2758 .llseek = seq_lseek,
2759 .release = single_release_net,
2762 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2764 struct net *net = (struct net *)seq->private;
2765 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2766 net->ipv6.rt6_stats->fib_nodes,
2767 net->ipv6.rt6_stats->fib_route_nodes,
2768 net->ipv6.rt6_stats->fib_rt_alloc,
2769 net->ipv6.rt6_stats->fib_rt_entries,
2770 net->ipv6.rt6_stats->fib_rt_cache,
2771 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2772 net->ipv6.rt6_stats->fib_discarded_routes);
2777 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2779 return single_open_net(inode, file, rt6_stats_seq_show);
2782 static const struct file_operations rt6_stats_seq_fops = {
2783 .owner = THIS_MODULE,
2784 .open = rt6_stats_seq_open,
2786 .llseek = seq_lseek,
2787 .release = single_release_net,
2789 #endif /* CONFIG_PROC_FS */
2791 #ifdef CONFIG_SYSCTL
2794 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2795 void __user *buffer, size_t *lenp, loff_t *ppos)
2802 net = (struct net *)ctl->extra1;
2803 delay = net->ipv6.sysctl.flush_delay;
2804 proc_dointvec(ctl, write, buffer, lenp, ppos);
2805 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2809 ctl_table ipv6_route_table_template[] = {
2811 .procname = "flush",
2812 .data = &init_net.ipv6.sysctl.flush_delay,
2813 .maxlen = sizeof(int),
2815 .proc_handler = ipv6_sysctl_rtcache_flush
2818 .procname = "gc_thresh",
2819 .data = &ip6_dst_ops_template.gc_thresh,
2820 .maxlen = sizeof(int),
2822 .proc_handler = proc_dointvec,
2825 .procname = "max_size",
2826 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2827 .maxlen = sizeof(int),
2829 .proc_handler = proc_dointvec,
2832 .procname = "gc_min_interval",
2833 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2834 .maxlen = sizeof(int),
2836 .proc_handler = proc_dointvec_jiffies,
2839 .procname = "gc_timeout",
2840 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2841 .maxlen = sizeof(int),
2843 .proc_handler = proc_dointvec_jiffies,
2846 .procname = "gc_interval",
2847 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2848 .maxlen = sizeof(int),
2850 .proc_handler = proc_dointvec_jiffies,
2853 .procname = "gc_elasticity",
2854 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2855 .maxlen = sizeof(int),
2857 .proc_handler = proc_dointvec,
2860 .procname = "mtu_expires",
2861 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2862 .maxlen = sizeof(int),
2864 .proc_handler = proc_dointvec_jiffies,
2867 .procname = "min_adv_mss",
2868 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2869 .maxlen = sizeof(int),
2871 .proc_handler = proc_dointvec,
2874 .procname = "gc_min_interval_ms",
2875 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2876 .maxlen = sizeof(int),
2878 .proc_handler = proc_dointvec_ms_jiffies,
2883 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2885 struct ctl_table *table;
2887 table = kmemdup(ipv6_route_table_template,
2888 sizeof(ipv6_route_table_template),
2892 table[0].data = &net->ipv6.sysctl.flush_delay;
2893 table[0].extra1 = net;
2894 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2895 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2896 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2897 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2898 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2899 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2900 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2901 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2902 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2904 /* Don't export sysctls to unprivileged users */
2905 if (net->user_ns != &init_user_ns)
2906 table[0].procname = NULL;
2913 static int __net_init ip6_route_net_init(struct net *net)
2917 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2918 sizeof(net->ipv6.ip6_dst_ops));
2920 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2921 goto out_ip6_dst_ops;
2923 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2924 sizeof(*net->ipv6.ip6_null_entry),
2926 if (!net->ipv6.ip6_null_entry)
2927 goto out_ip6_dst_entries;
2928 net->ipv6.ip6_null_entry->dst.path =
2929 (struct dst_entry *)net->ipv6.ip6_null_entry;
2930 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2931 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2932 ip6_template_metrics, true);
2934 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2935 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2936 sizeof(*net->ipv6.ip6_prohibit_entry),
2938 if (!net->ipv6.ip6_prohibit_entry)
2939 goto out_ip6_null_entry;
2940 net->ipv6.ip6_prohibit_entry->dst.path =
2941 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2942 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2943 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2944 ip6_template_metrics, true);
2946 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2947 sizeof(*net->ipv6.ip6_blk_hole_entry),
2949 if (!net->ipv6.ip6_blk_hole_entry)
2950 goto out_ip6_prohibit_entry;
2951 net->ipv6.ip6_blk_hole_entry->dst.path =
2952 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2953 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2954 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2955 ip6_template_metrics, true);
2958 net->ipv6.sysctl.flush_delay = 0;
2959 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2960 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2961 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2962 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2963 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2964 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2965 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2967 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2973 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2974 out_ip6_prohibit_entry:
2975 kfree(net->ipv6.ip6_prohibit_entry);
2977 kfree(net->ipv6.ip6_null_entry);
2979 out_ip6_dst_entries:
2980 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2985 static void __net_exit ip6_route_net_exit(struct net *net)
2987 kfree(net->ipv6.ip6_null_entry);
2988 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2989 kfree(net->ipv6.ip6_prohibit_entry);
2990 kfree(net->ipv6.ip6_blk_hole_entry);
2992 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2995 static int __net_init ip6_route_net_init_late(struct net *net)
2997 #ifdef CONFIG_PROC_FS
2998 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
2999 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3004 static void __net_exit ip6_route_net_exit_late(struct net *net)
3006 #ifdef CONFIG_PROC_FS
3007 remove_proc_entry("ipv6_route", net->proc_net);
3008 remove_proc_entry("rt6_stats", net->proc_net);
3012 static struct pernet_operations ip6_route_net_ops = {
3013 .init = ip6_route_net_init,
3014 .exit = ip6_route_net_exit,
3017 static int __net_init ipv6_inetpeer_init(struct net *net)
3019 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3023 inet_peer_base_init(bp);
3024 net->ipv6.peers = bp;
3028 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3030 struct inet_peer_base *bp = net->ipv6.peers;
3032 net->ipv6.peers = NULL;
3033 inetpeer_invalidate_tree(bp);
3037 static struct pernet_operations ipv6_inetpeer_ops = {
3038 .init = ipv6_inetpeer_init,
3039 .exit = ipv6_inetpeer_exit,
3042 static struct pernet_operations ip6_route_net_late_ops = {
3043 .init = ip6_route_net_init_late,
3044 .exit = ip6_route_net_exit_late,
3047 static struct notifier_block ip6_route_dev_notifier = {
3048 .notifier_call = ip6_route_dev_notify,
3052 int __init ip6_route_init(void)
3057 ip6_dst_ops_template.kmem_cachep =
3058 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3059 SLAB_HWCACHE_ALIGN, NULL);
3060 if (!ip6_dst_ops_template.kmem_cachep)
3063 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3065 goto out_kmem_cache;
3067 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3069 goto out_dst_entries;
3071 ret = register_pernet_subsys(&ip6_route_net_ops);
3073 goto out_register_inetpeer;
3075 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3077 /* Registering of the loopback is done before this portion of code,
3078 * the loopback reference in rt6_info will not be taken, do it
3079 * manually for init_net */
3080 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3081 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3082 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3083 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3084 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3085 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3086 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3090 goto out_register_subsys;
3096 ret = fib6_rules_init();
3100 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3102 goto fib6_rules_init;
3105 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3106 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3107 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3108 goto out_register_late_subsys;
3110 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3112 goto out_register_late_subsys;
3117 out_register_late_subsys:
3118 unregister_pernet_subsys(&ip6_route_net_late_ops);
3120 fib6_rules_cleanup();
3125 out_register_subsys:
3126 unregister_pernet_subsys(&ip6_route_net_ops);
3127 out_register_inetpeer:
3128 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3130 dst_entries_destroy(&ip6_dst_blackhole_ops);
3132 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3136 void ip6_route_cleanup(void)
3138 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3139 unregister_pernet_subsys(&ip6_route_net_late_ops);
3140 fib6_rules_cleanup();
3143 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3144 unregister_pernet_subsys(&ip6_route_net_ops);
3145 dst_entries_destroy(&ip6_dst_blackhole_ops);
3146 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);