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 int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
156 struct neighbour *n = __ipv6_neigh_lookup(dev, &rt->rt6i_gateway);
158 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
167 static struct dst_ops ip6_dst_ops_template = {
169 .protocol = cpu_to_be16(ETH_P_IPV6),
172 .check = ip6_dst_check,
173 .default_advmss = ip6_default_advmss,
175 .cow_metrics = ipv6_cow_metrics,
176 .destroy = ip6_dst_destroy,
177 .ifdown = ip6_dst_ifdown,
178 .negative_advice = ip6_negative_advice,
179 .link_failure = ip6_link_failure,
180 .update_pmtu = ip6_rt_update_pmtu,
181 .redirect = rt6_do_redirect,
182 .local_out = __ip6_local_out,
183 .neigh_lookup = ip6_neigh_lookup,
186 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
188 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
190 return mtu ? : dst->dev->mtu;
193 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
194 struct sk_buff *skb, u32 mtu)
198 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
203 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
209 static struct dst_ops ip6_dst_blackhole_ops = {
211 .protocol = cpu_to_be16(ETH_P_IPV6),
212 .destroy = ip6_dst_destroy,
213 .check = ip6_dst_check,
214 .mtu = ip6_blackhole_mtu,
215 .default_advmss = ip6_default_advmss,
216 .update_pmtu = ip6_rt_blackhole_update_pmtu,
217 .redirect = ip6_rt_blackhole_redirect,
218 .cow_metrics = ip6_rt_blackhole_cow_metrics,
219 .neigh_lookup = ip6_neigh_lookup,
222 static const u32 ip6_template_metrics[RTAX_MAX] = {
223 [RTAX_HOPLIMIT - 1] = 0,
226 static const struct rt6_info ip6_null_entry_template = {
228 .__refcnt = ATOMIC_INIT(1),
230 .obsolete = DST_OBSOLETE_FORCE_CHK,
231 .error = -ENETUNREACH,
232 .input = ip6_pkt_discard,
233 .output = ip6_pkt_discard_out,
235 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
236 .rt6i_protocol = RTPROT_KERNEL,
237 .rt6i_metric = ~(u32) 0,
238 .rt6i_ref = ATOMIC_INIT(1),
241 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
243 static int ip6_pkt_prohibit(struct sk_buff *skb);
244 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
246 static const struct rt6_info ip6_prohibit_entry_template = {
248 .__refcnt = ATOMIC_INIT(1),
250 .obsolete = DST_OBSOLETE_FORCE_CHK,
252 .input = ip6_pkt_prohibit,
253 .output = ip6_pkt_prohibit_out,
255 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
256 .rt6i_protocol = RTPROT_KERNEL,
257 .rt6i_metric = ~(u32) 0,
258 .rt6i_ref = ATOMIC_INIT(1),
261 static const struct rt6_info ip6_blk_hole_entry_template = {
263 .__refcnt = ATOMIC_INIT(1),
265 .obsolete = DST_OBSOLETE_FORCE_CHK,
267 .input = dst_discard,
268 .output = dst_discard,
270 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
271 .rt6i_protocol = RTPROT_KERNEL,
272 .rt6i_metric = ~(u32) 0,
273 .rt6i_ref = ATOMIC_INIT(1),
278 /* allocate dst with ip6_dst_ops */
279 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
280 struct net_device *dev,
282 struct fib6_table *table)
284 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
285 0, DST_OBSOLETE_FORCE_CHK, flags);
288 struct dst_entry *dst = &rt->dst;
290 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
291 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
292 rt->rt6i_genid = rt_genid(net);
293 INIT_LIST_HEAD(&rt->rt6i_siblings);
294 rt->rt6i_nsiblings = 0;
299 static void ip6_dst_destroy(struct dst_entry *dst)
301 struct rt6_info *rt = (struct rt6_info *)dst;
302 struct inet6_dev *idev = rt->rt6i_idev;
305 neigh_release(rt->n);
307 if (!(rt->dst.flags & DST_HOST))
308 dst_destroy_metrics_generic(dst);
311 rt->rt6i_idev = NULL;
315 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
316 dst_release(dst->from);
318 if (rt6_has_peer(rt)) {
319 struct inet_peer *peer = rt6_peer_ptr(rt);
324 void rt6_bind_peer(struct rt6_info *rt, int create)
326 struct inet_peer_base *base;
327 struct inet_peer *peer;
329 base = inetpeer_base_ptr(rt->_rt6i_peer);
333 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
335 if (!rt6_set_peer(rt, peer))
340 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
343 struct rt6_info *rt = (struct rt6_info *)dst;
344 struct inet6_dev *idev = rt->rt6i_idev;
345 struct net_device *loopback_dev =
346 dev_net(dev)->loopback_dev;
348 if (dev != loopback_dev) {
349 if (idev && idev->dev == dev) {
350 struct inet6_dev *loopback_idev =
351 in6_dev_get(loopback_dev);
353 rt->rt6i_idev = loopback_idev;
357 if (rt->n && rt->n->dev == dev) {
358 rt->n->dev = loopback_dev;
359 dev_hold(loopback_dev);
365 static bool rt6_check_expired(const struct rt6_info *rt)
367 if (rt->rt6i_flags & RTF_EXPIRES) {
368 if (time_after(jiffies, rt->dst.expires))
370 } else if (rt->dst.from) {
371 return rt6_check_expired((struct rt6_info *) rt->dst.from);
376 static bool rt6_need_strict(const struct in6_addr *daddr)
378 return ipv6_addr_type(daddr) &
379 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
382 /* Multipath route selection:
383 * Hash based function using packet header and flowlabel.
384 * Adapted from fib_info_hashfn()
386 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
387 const struct flowi6 *fl6)
389 unsigned int val = fl6->flowi6_proto;
391 val ^= ipv6_addr_hash(&fl6->daddr);
392 val ^= ipv6_addr_hash(&fl6->saddr);
394 /* Work only if this not encapsulated */
395 switch (fl6->flowi6_proto) {
399 val ^= (__force u16)fl6->fl6_sport;
400 val ^= (__force u16)fl6->fl6_dport;
404 val ^= (__force u16)fl6->fl6_icmp_type;
405 val ^= (__force u16)fl6->fl6_icmp_code;
408 /* RFC6438 recommands to use flowlabel */
409 val ^= (__force u32)fl6->flowlabel;
411 /* Perhaps, we need to tune, this function? */
412 val = val ^ (val >> 7) ^ (val >> 12);
413 return val % candidate_count;
416 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
419 struct rt6_info *sibling, *next_sibling;
422 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
423 /* Don't change the route, if route_choosen == 0
424 * (siblings does not include ourself)
427 list_for_each_entry_safe(sibling, next_sibling,
428 &match->rt6i_siblings, rt6i_siblings) {
430 if (route_choosen == 0) {
439 * Route lookup. Any table->tb6_lock is implied.
442 static inline struct rt6_info *rt6_device_match(struct net *net,
444 const struct in6_addr *saddr,
448 struct rt6_info *local = NULL;
449 struct rt6_info *sprt;
451 if (!oif && ipv6_addr_any(saddr))
454 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
455 struct net_device *dev = sprt->dst.dev;
458 if (dev->ifindex == oif)
460 if (dev->flags & IFF_LOOPBACK) {
461 if (!sprt->rt6i_idev ||
462 sprt->rt6i_idev->dev->ifindex != oif) {
463 if (flags & RT6_LOOKUP_F_IFACE && oif)
465 if (local && (!oif ||
466 local->rt6i_idev->dev->ifindex == oif))
472 if (ipv6_chk_addr(net, saddr, dev,
473 flags & RT6_LOOKUP_F_IFACE))
482 if (flags & RT6_LOOKUP_F_IFACE)
483 return net->ipv6.ip6_null_entry;
489 #ifdef CONFIG_IPV6_ROUTER_PREF
490 static void rt6_probe(struct rt6_info *rt)
492 struct neighbour *neigh;
494 * Okay, this does not seem to be appropriate
495 * for now, however, we need to check if it
496 * is really so; aka Router Reachability Probing.
498 * Router Reachability Probe MUST be rate-limited
499 * to no more than one per minute.
501 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
504 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
506 write_lock(&neigh->lock);
507 if (neigh->nud_state & NUD_VALID)
512 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
513 struct in6_addr mcaddr;
514 struct in6_addr *target;
516 neigh->updated = jiffies;
519 write_unlock(&neigh->lock);
521 target = (struct in6_addr *)&rt->rt6i_gateway;
522 addrconf_addr_solict_mult(target, &mcaddr);
523 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
526 write_unlock(&neigh->lock);
528 rcu_read_unlock_bh();
531 static inline void rt6_probe(struct rt6_info *rt)
537 * Default Router Selection (RFC 2461 6.3.6)
539 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
541 struct net_device *dev = rt->dst.dev;
542 if (!oif || dev->ifindex == oif)
544 if ((dev->flags & IFF_LOOPBACK) &&
545 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
550 static inline bool rt6_check_neigh(struct rt6_info *rt)
552 struct neighbour *neigh;
555 if (rt->rt6i_flags & RTF_NONEXTHOP ||
556 !(rt->rt6i_flags & RTF_GATEWAY))
560 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
562 read_lock(&neigh->lock);
563 if (neigh->nud_state & NUD_VALID)
565 #ifdef CONFIG_IPV6_ROUTER_PREF
566 else if (!(neigh->nud_state & NUD_FAILED))
569 read_unlock(&neigh->lock);
571 rcu_read_unlock_bh();
576 static int rt6_score_route(struct rt6_info *rt, int oif,
581 m = rt6_check_dev(rt, oif);
582 if (!m && (strict & RT6_LOOKUP_F_IFACE))
584 #ifdef CONFIG_IPV6_ROUTER_PREF
585 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
587 if (!rt6_check_neigh(rt) && (strict & RT6_LOOKUP_F_REACHABLE))
592 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
593 int *mpri, struct rt6_info *match)
597 if (rt6_check_expired(rt))
600 m = rt6_score_route(rt, oif, strict);
605 if (strict & RT6_LOOKUP_F_REACHABLE)
609 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
617 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
618 struct rt6_info *rr_head,
619 u32 metric, int oif, int strict)
621 struct rt6_info *rt, *match;
625 for (rt = rr_head; rt && rt->rt6i_metric == metric;
626 rt = rt->dst.rt6_next)
627 match = find_match(rt, oif, strict, &mpri, match);
628 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
629 rt = rt->dst.rt6_next)
630 match = find_match(rt, oif, strict, &mpri, match);
635 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
637 struct rt6_info *match, *rt0;
642 fn->rr_ptr = rt0 = fn->leaf;
644 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
647 (strict & RT6_LOOKUP_F_REACHABLE)) {
648 struct rt6_info *next = rt0->dst.rt6_next;
650 /* no entries matched; do round-robin */
651 if (!next || next->rt6i_metric != rt0->rt6i_metric)
658 net = dev_net(rt0->dst.dev);
659 return match ? match : net->ipv6.ip6_null_entry;
662 #ifdef CONFIG_IPV6_ROUTE_INFO
663 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
664 const struct in6_addr *gwaddr)
666 struct net *net = dev_net(dev);
667 struct route_info *rinfo = (struct route_info *) opt;
668 struct in6_addr prefix_buf, *prefix;
670 unsigned long lifetime;
673 if (len < sizeof(struct route_info)) {
677 /* Sanity check for prefix_len and length */
678 if (rinfo->length > 3) {
680 } else if (rinfo->prefix_len > 128) {
682 } else if (rinfo->prefix_len > 64) {
683 if (rinfo->length < 2) {
686 } else if (rinfo->prefix_len > 0) {
687 if (rinfo->length < 1) {
692 pref = rinfo->route_pref;
693 if (pref == ICMPV6_ROUTER_PREF_INVALID)
696 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
698 if (rinfo->length == 3)
699 prefix = (struct in6_addr *)rinfo->prefix;
701 /* this function is safe */
702 ipv6_addr_prefix(&prefix_buf,
703 (struct in6_addr *)rinfo->prefix,
705 prefix = &prefix_buf;
708 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
711 if (rt && !lifetime) {
717 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
720 rt->rt6i_flags = RTF_ROUTEINFO |
721 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
724 if (!addrconf_finite_timeout(lifetime))
725 rt6_clean_expires(rt);
727 rt6_set_expires(rt, jiffies + HZ * lifetime);
735 #define BACKTRACK(__net, saddr) \
737 if (rt == __net->ipv6.ip6_null_entry) { \
738 struct fib6_node *pn; \
740 if (fn->fn_flags & RTN_TL_ROOT) \
743 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
744 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
747 if (fn->fn_flags & RTN_RTINFO) \
753 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
754 struct fib6_table *table,
755 struct flowi6 *fl6, int flags)
757 struct fib6_node *fn;
760 read_lock_bh(&table->tb6_lock);
761 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
764 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
765 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
766 rt = rt6_multipath_select(rt, fl6);
767 BACKTRACK(net, &fl6->saddr);
769 dst_use(&rt->dst, jiffies);
770 read_unlock_bh(&table->tb6_lock);
775 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
778 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
780 EXPORT_SYMBOL_GPL(ip6_route_lookup);
782 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
783 const struct in6_addr *saddr, int oif, int strict)
785 struct flowi6 fl6 = {
789 struct dst_entry *dst;
790 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
793 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
794 flags |= RT6_LOOKUP_F_HAS_SADDR;
797 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
799 return (struct rt6_info *) dst;
806 EXPORT_SYMBOL(rt6_lookup);
808 /* ip6_ins_rt is called with FREE table->tb6_lock.
809 It takes new route entry, the addition fails by any reason the
810 route is freed. In any case, if caller does not hold it, it may
814 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
817 struct fib6_table *table;
819 table = rt->rt6i_table;
820 write_lock_bh(&table->tb6_lock);
821 err = fib6_add(&table->tb6_root, rt, info);
822 write_unlock_bh(&table->tb6_lock);
827 int ip6_ins_rt(struct rt6_info *rt)
829 struct nl_info info = {
830 .nl_net = dev_net(rt->dst.dev),
832 return __ip6_ins_rt(rt, &info);
835 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
836 const struct in6_addr *daddr,
837 const struct in6_addr *saddr)
845 rt = ip6_rt_copy(ort, daddr);
848 int attempts = !in_softirq();
850 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
851 if (ort->rt6i_dst.plen != 128 &&
852 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
853 rt->rt6i_flags |= RTF_ANYCAST;
854 rt->rt6i_gateway = *daddr;
857 rt->rt6i_flags |= RTF_CACHE;
859 #ifdef CONFIG_IPV6_SUBTREES
860 if (rt->rt6i_src.plen && saddr) {
861 rt->rt6i_src.addr = *saddr;
862 rt->rt6i_src.plen = 128;
867 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
868 struct net *net = dev_net(rt->dst.dev);
869 int saved_rt_min_interval =
870 net->ipv6.sysctl.ip6_rt_gc_min_interval;
871 int saved_rt_elasticity =
872 net->ipv6.sysctl.ip6_rt_gc_elasticity;
874 if (attempts-- > 0) {
875 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
876 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
878 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
880 net->ipv6.sysctl.ip6_rt_gc_elasticity =
882 net->ipv6.sysctl.ip6_rt_gc_min_interval =
883 saved_rt_min_interval;
887 net_warn_ratelimited("Neighbour table overflow\n");
896 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
897 const struct in6_addr *daddr)
899 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
902 rt->rt6i_flags |= RTF_CACHE;
903 rt->n = neigh_clone(ort->n);
908 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
909 struct flowi6 *fl6, int flags)
911 struct fib6_node *fn;
912 struct rt6_info *rt, *nrt;
916 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
918 strict |= flags & RT6_LOOKUP_F_IFACE;
921 read_lock_bh(&table->tb6_lock);
924 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
927 rt = rt6_select(fn, oif, strict | reachable);
928 if (rt->rt6i_nsiblings && oif == 0)
929 rt = rt6_multipath_select(rt, fl6);
930 BACKTRACK(net, &fl6->saddr);
931 if (rt == net->ipv6.ip6_null_entry ||
932 rt->rt6i_flags & RTF_CACHE)
936 read_unlock_bh(&table->tb6_lock);
938 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
939 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
940 else if (!(rt->dst.flags & DST_HOST))
941 nrt = rt6_alloc_clone(rt, &fl6->daddr);
946 rt = nrt ? : net->ipv6.ip6_null_entry;
950 err = ip6_ins_rt(nrt);
959 * Race condition! In the gap, when table->tb6_lock was
960 * released someone could insert this route. Relookup.
971 read_unlock_bh(&table->tb6_lock);
973 rt->dst.lastuse = jiffies;
979 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
980 struct flowi6 *fl6, int flags)
982 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
985 static struct dst_entry *ip6_route_input_lookup(struct net *net,
986 struct net_device *dev,
987 struct flowi6 *fl6, int flags)
989 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
990 flags |= RT6_LOOKUP_F_IFACE;
992 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
995 void ip6_route_input(struct sk_buff *skb)
997 const struct ipv6hdr *iph = ipv6_hdr(skb);
998 struct net *net = dev_net(skb->dev);
999 int flags = RT6_LOOKUP_F_HAS_SADDR;
1000 struct flowi6 fl6 = {
1001 .flowi6_iif = skb->dev->ifindex,
1002 .daddr = iph->daddr,
1003 .saddr = iph->saddr,
1004 .flowlabel = ip6_flowinfo(iph),
1005 .flowi6_mark = skb->mark,
1006 .flowi6_proto = iph->nexthdr,
1009 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1012 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1013 struct flowi6 *fl6, int flags)
1015 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1018 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1023 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1025 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1026 flags |= RT6_LOOKUP_F_IFACE;
1028 if (!ipv6_addr_any(&fl6->saddr))
1029 flags |= RT6_LOOKUP_F_HAS_SADDR;
1031 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1033 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1036 EXPORT_SYMBOL(ip6_route_output);
1038 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1040 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1041 struct dst_entry *new = NULL;
1043 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1047 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1048 rt6_init_peer(rt, net->ipv6.peers);
1051 new->input = dst_discard;
1052 new->output = dst_discard;
1054 if (dst_metrics_read_only(&ort->dst))
1055 new->_metrics = ort->dst._metrics;
1057 dst_copy_metrics(new, &ort->dst);
1058 rt->rt6i_idev = ort->rt6i_idev;
1060 in6_dev_hold(rt->rt6i_idev);
1062 rt->rt6i_gateway = ort->rt6i_gateway;
1063 rt->rt6i_flags = ort->rt6i_flags;
1064 rt6_clean_expires(rt);
1065 rt->rt6i_metric = 0;
1067 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1068 #ifdef CONFIG_IPV6_SUBTREES
1069 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1075 dst_release(dst_orig);
1076 return new ? new : ERR_PTR(-ENOMEM);
1080 * Destination cache support functions
1083 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1085 struct rt6_info *rt;
1087 rt = (struct rt6_info *) dst;
1089 /* All IPV6 dsts are created with ->obsolete set to the value
1090 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1091 * into this function always.
1093 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1096 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1102 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1104 struct rt6_info *rt = (struct rt6_info *) dst;
1107 if (rt->rt6i_flags & RTF_CACHE) {
1108 if (rt6_check_expired(rt)) {
1120 static void ip6_link_failure(struct sk_buff *skb)
1122 struct rt6_info *rt;
1124 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1126 rt = (struct rt6_info *) skb_dst(skb);
1128 if (rt->rt6i_flags & RTF_CACHE)
1129 rt6_update_expires(rt, 0);
1130 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1131 rt->rt6i_node->fn_sernum = -1;
1135 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1136 struct sk_buff *skb, u32 mtu)
1138 struct rt6_info *rt6 = (struct rt6_info*)dst;
1141 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1142 struct net *net = dev_net(dst->dev);
1144 rt6->rt6i_flags |= RTF_MODIFIED;
1145 if (mtu < IPV6_MIN_MTU) {
1146 u32 features = dst_metric(dst, RTAX_FEATURES);
1148 features |= RTAX_FEATURE_ALLFRAG;
1149 dst_metric_set(dst, RTAX_FEATURES, features);
1151 dst_metric_set(dst, RTAX_MTU, mtu);
1152 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1156 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1159 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1160 struct dst_entry *dst;
1163 memset(&fl6, 0, sizeof(fl6));
1164 fl6.flowi6_oif = oif;
1165 fl6.flowi6_mark = mark;
1166 fl6.flowi6_flags = 0;
1167 fl6.daddr = iph->daddr;
1168 fl6.saddr = iph->saddr;
1169 fl6.flowlabel = ip6_flowinfo(iph);
1171 dst = ip6_route_output(net, NULL, &fl6);
1173 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1176 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1178 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1180 ip6_update_pmtu(skb, sock_net(sk), mtu,
1181 sk->sk_bound_dev_if, sk->sk_mark);
1183 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1185 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1187 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1188 struct dst_entry *dst;
1191 memset(&fl6, 0, sizeof(fl6));
1192 fl6.flowi6_oif = oif;
1193 fl6.flowi6_mark = mark;
1194 fl6.flowi6_flags = 0;
1195 fl6.daddr = iph->daddr;
1196 fl6.saddr = iph->saddr;
1197 fl6.flowlabel = ip6_flowinfo(iph);
1199 dst = ip6_route_output(net, NULL, &fl6);
1201 rt6_do_redirect(dst, NULL, skb);
1204 EXPORT_SYMBOL_GPL(ip6_redirect);
1206 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1208 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1210 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1212 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1214 struct net_device *dev = dst->dev;
1215 unsigned int mtu = dst_mtu(dst);
1216 struct net *net = dev_net(dev);
1218 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1220 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1221 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1224 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1225 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1226 * IPV6_MAXPLEN is also valid and means: "any MSS,
1227 * rely only on pmtu discovery"
1229 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1234 static unsigned int ip6_mtu(const struct dst_entry *dst)
1236 struct inet6_dev *idev;
1237 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1245 idev = __in6_dev_get(dst->dev);
1247 mtu = idev->cnf.mtu6;
1253 static struct dst_entry *icmp6_dst_gc_list;
1254 static DEFINE_SPINLOCK(icmp6_dst_lock);
1256 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1257 struct neighbour *neigh,
1260 struct dst_entry *dst;
1261 struct rt6_info *rt;
1262 struct inet6_dev *idev = in6_dev_get(dev);
1263 struct net *net = dev_net(dev);
1265 if (unlikely(!idev))
1266 return ERR_PTR(-ENODEV);
1268 rt = ip6_dst_alloc(net, dev, 0, NULL);
1269 if (unlikely(!rt)) {
1271 dst = ERR_PTR(-ENOMEM);
1278 neigh = ip6_neigh_lookup(&rt->dst, NULL, &fl6->daddr);
1279 if (IS_ERR(neigh)) {
1282 return ERR_CAST(neigh);
1286 rt->dst.flags |= DST_HOST;
1287 rt->dst.output = ip6_output;
1289 atomic_set(&rt->dst.__refcnt, 1);
1290 rt->rt6i_dst.addr = fl6->daddr;
1291 rt->rt6i_dst.plen = 128;
1292 rt->rt6i_idev = idev;
1293 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1295 spin_lock_bh(&icmp6_dst_lock);
1296 rt->dst.next = icmp6_dst_gc_list;
1297 icmp6_dst_gc_list = &rt->dst;
1298 spin_unlock_bh(&icmp6_dst_lock);
1300 fib6_force_start_gc(net);
1302 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1308 int icmp6_dst_gc(void)
1310 struct dst_entry *dst, **pprev;
1313 spin_lock_bh(&icmp6_dst_lock);
1314 pprev = &icmp6_dst_gc_list;
1316 while ((dst = *pprev) != NULL) {
1317 if (!atomic_read(&dst->__refcnt)) {
1326 spin_unlock_bh(&icmp6_dst_lock);
1331 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1334 struct dst_entry *dst, **pprev;
1336 spin_lock_bh(&icmp6_dst_lock);
1337 pprev = &icmp6_dst_gc_list;
1338 while ((dst = *pprev) != NULL) {
1339 struct rt6_info *rt = (struct rt6_info *) dst;
1340 if (func(rt, arg)) {
1347 spin_unlock_bh(&icmp6_dst_lock);
1350 static int ip6_dst_gc(struct dst_ops *ops)
1352 unsigned long now = jiffies;
1353 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1354 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1355 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1356 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1357 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1358 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1361 entries = dst_entries_get_fast(ops);
1362 if (time_after(rt_last_gc + rt_min_interval, now) &&
1363 entries <= rt_max_size)
1366 net->ipv6.ip6_rt_gc_expire++;
1367 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1368 net->ipv6.ip6_rt_last_gc = now;
1369 entries = dst_entries_get_slow(ops);
1370 if (entries < ops->gc_thresh)
1371 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1373 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1374 return entries > rt_max_size;
1377 int ip6_dst_hoplimit(struct dst_entry *dst)
1379 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1380 if (hoplimit == 0) {
1381 struct net_device *dev = dst->dev;
1382 struct inet6_dev *idev;
1385 idev = __in6_dev_get(dev);
1387 hoplimit = idev->cnf.hop_limit;
1389 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1394 EXPORT_SYMBOL(ip6_dst_hoplimit);
1400 int ip6_route_add(struct fib6_config *cfg)
1403 struct net *net = cfg->fc_nlinfo.nl_net;
1404 struct rt6_info *rt = NULL;
1405 struct net_device *dev = NULL;
1406 struct inet6_dev *idev = NULL;
1407 struct fib6_table *table;
1410 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1412 #ifndef CONFIG_IPV6_SUBTREES
1413 if (cfg->fc_src_len)
1416 if (cfg->fc_ifindex) {
1418 dev = dev_get_by_index(net, cfg->fc_ifindex);
1421 idev = in6_dev_get(dev);
1426 if (cfg->fc_metric == 0)
1427 cfg->fc_metric = IP6_RT_PRIO_USER;
1430 if (cfg->fc_nlinfo.nlh &&
1431 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1432 table = fib6_get_table(net, cfg->fc_table);
1434 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1435 table = fib6_new_table(net, cfg->fc_table);
1438 table = fib6_new_table(net, cfg->fc_table);
1444 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1451 if (cfg->fc_flags & RTF_EXPIRES)
1452 rt6_set_expires(rt, jiffies +
1453 clock_t_to_jiffies(cfg->fc_expires));
1455 rt6_clean_expires(rt);
1457 if (cfg->fc_protocol == RTPROT_UNSPEC)
1458 cfg->fc_protocol = RTPROT_BOOT;
1459 rt->rt6i_protocol = cfg->fc_protocol;
1461 addr_type = ipv6_addr_type(&cfg->fc_dst);
1463 if (addr_type & IPV6_ADDR_MULTICAST)
1464 rt->dst.input = ip6_mc_input;
1465 else if (cfg->fc_flags & RTF_LOCAL)
1466 rt->dst.input = ip6_input;
1468 rt->dst.input = ip6_forward;
1470 rt->dst.output = ip6_output;
1472 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1473 rt->rt6i_dst.plen = cfg->fc_dst_len;
1474 if (rt->rt6i_dst.plen == 128)
1475 rt->dst.flags |= DST_HOST;
1477 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1478 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1483 dst_init_metrics(&rt->dst, metrics, 0);
1485 #ifdef CONFIG_IPV6_SUBTREES
1486 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1487 rt->rt6i_src.plen = cfg->fc_src_len;
1490 rt->rt6i_metric = cfg->fc_metric;
1492 /* We cannot add true routes via loopback here,
1493 they would result in kernel looping; promote them to reject routes
1495 if ((cfg->fc_flags & RTF_REJECT) ||
1496 (dev && (dev->flags & IFF_LOOPBACK) &&
1497 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1498 !(cfg->fc_flags & RTF_LOCAL))) {
1499 /* hold loopback dev/idev if we haven't done so. */
1500 if (dev != net->loopback_dev) {
1505 dev = net->loopback_dev;
1507 idev = in6_dev_get(dev);
1513 rt->dst.output = ip6_pkt_discard_out;
1514 rt->dst.input = ip6_pkt_discard;
1515 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1516 switch (cfg->fc_type) {
1518 rt->dst.error = -EINVAL;
1521 rt->dst.error = -EACCES;
1524 rt->dst.error = -EAGAIN;
1527 rt->dst.error = -ENETUNREACH;
1533 if (cfg->fc_flags & RTF_GATEWAY) {
1534 const struct in6_addr *gw_addr;
1537 gw_addr = &cfg->fc_gateway;
1538 rt->rt6i_gateway = *gw_addr;
1539 gwa_type = ipv6_addr_type(gw_addr);
1541 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1542 struct rt6_info *grt;
1544 /* IPv6 strictly inhibits using not link-local
1545 addresses as nexthop address.
1546 Otherwise, router will not able to send redirects.
1547 It is very good, but in some (rare!) circumstances
1548 (SIT, PtP, NBMA NOARP links) it is handy to allow
1549 some exceptions. --ANK
1552 if (!(gwa_type & IPV6_ADDR_UNICAST))
1555 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1557 err = -EHOSTUNREACH;
1561 if (dev != grt->dst.dev) {
1567 idev = grt->rt6i_idev;
1569 in6_dev_hold(grt->rt6i_idev);
1571 if (!(grt->rt6i_flags & RTF_GATEWAY))
1579 if (!dev || (dev->flags & IFF_LOOPBACK))
1587 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1588 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1592 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1593 rt->rt6i_prefsrc.plen = 128;
1595 rt->rt6i_prefsrc.plen = 0;
1597 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1598 err = rt6_bind_neighbour(rt, dev);
1603 rt->rt6i_flags = cfg->fc_flags;
1610 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1611 int type = nla_type(nla);
1614 if (type > RTAX_MAX) {
1619 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1625 rt->rt6i_idev = idev;
1626 rt->rt6i_table = table;
1628 cfg->fc_nlinfo.nl_net = dev_net(dev);
1630 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1642 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1645 struct fib6_table *table;
1646 struct net *net = dev_net(rt->dst.dev);
1648 if (rt == net->ipv6.ip6_null_entry) {
1653 table = rt->rt6i_table;
1654 write_lock_bh(&table->tb6_lock);
1655 err = fib6_del(rt, info);
1656 write_unlock_bh(&table->tb6_lock);
1663 int ip6_del_rt(struct rt6_info *rt)
1665 struct nl_info info = {
1666 .nl_net = dev_net(rt->dst.dev),
1668 return __ip6_del_rt(rt, &info);
1671 static int ip6_route_del(struct fib6_config *cfg)
1673 struct fib6_table *table;
1674 struct fib6_node *fn;
1675 struct rt6_info *rt;
1678 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1682 read_lock_bh(&table->tb6_lock);
1684 fn = fib6_locate(&table->tb6_root,
1685 &cfg->fc_dst, cfg->fc_dst_len,
1686 &cfg->fc_src, cfg->fc_src_len);
1689 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1690 if (cfg->fc_ifindex &&
1692 rt->dst.dev->ifindex != cfg->fc_ifindex))
1694 if (cfg->fc_flags & RTF_GATEWAY &&
1695 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1697 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1700 read_unlock_bh(&table->tb6_lock);
1702 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1705 read_unlock_bh(&table->tb6_lock);
1710 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1712 struct net *net = dev_net(skb->dev);
1713 struct netevent_redirect netevent;
1714 struct rt6_info *rt, *nrt = NULL;
1715 struct ndisc_options ndopts;
1716 struct neighbour *old_neigh;
1717 struct inet6_dev *in6_dev;
1718 struct neighbour *neigh;
1720 int optlen, on_link;
1723 optlen = skb->tail - skb->transport_header;
1724 optlen -= sizeof(*msg);
1727 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1731 msg = (struct rd_msg *)icmp6_hdr(skb);
1733 if (ipv6_addr_is_multicast(&msg->dest)) {
1734 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1739 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1741 } else if (ipv6_addr_type(&msg->target) !=
1742 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1743 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1747 in6_dev = __in6_dev_get(skb->dev);
1750 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1754 * The IP source address of the Redirect MUST be the same as the current
1755 * first-hop router for the specified ICMP Destination Address.
1758 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1759 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1764 if (ndopts.nd_opts_tgt_lladdr) {
1765 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1768 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1773 rt = (struct rt6_info *) dst;
1774 if (rt == net->ipv6.ip6_null_entry) {
1775 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1779 /* Redirect received -> path was valid.
1780 * Look, redirects are sent only in response to data packets,
1781 * so that this nexthop apparently is reachable. --ANK
1783 dst_confirm(&rt->dst);
1785 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1789 /* Duplicate redirect: silently ignore. */
1791 if (neigh == old_neigh)
1795 * We have finally decided to accept it.
1798 neigh_update(neigh, lladdr, NUD_STALE,
1799 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1800 NEIGH_UPDATE_F_OVERRIDE|
1801 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1802 NEIGH_UPDATE_F_ISROUTER))
1805 nrt = ip6_rt_copy(rt, &msg->dest);
1809 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1811 nrt->rt6i_flags &= ~RTF_GATEWAY;
1813 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1814 nrt->n = neigh_clone(neigh);
1816 if (ip6_ins_rt(nrt))
1819 netevent.old = &rt->dst;
1820 netevent.new = &nrt->dst;
1821 netevent.daddr = &msg->dest;
1822 netevent.neigh = neigh;
1823 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1825 if (rt->rt6i_flags & RTF_CACHE) {
1826 rt = (struct rt6_info *) dst_clone(&rt->dst);
1831 neigh_release(neigh);
1835 * Misc support functions
1838 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1839 const struct in6_addr *dest)
1841 struct net *net = dev_net(ort->dst.dev);
1842 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1846 rt->dst.input = ort->dst.input;
1847 rt->dst.output = ort->dst.output;
1848 rt->dst.flags |= DST_HOST;
1850 rt->rt6i_dst.addr = *dest;
1851 rt->rt6i_dst.plen = 128;
1852 dst_copy_metrics(&rt->dst, &ort->dst);
1853 rt->dst.error = ort->dst.error;
1854 rt->rt6i_idev = ort->rt6i_idev;
1856 in6_dev_hold(rt->rt6i_idev);
1857 rt->dst.lastuse = jiffies;
1859 rt->rt6i_gateway = ort->rt6i_gateway;
1860 rt->rt6i_flags = ort->rt6i_flags;
1861 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1862 (RTF_DEFAULT | RTF_ADDRCONF))
1863 rt6_set_from(rt, ort);
1865 rt6_clean_expires(rt);
1866 rt->rt6i_metric = 0;
1868 #ifdef CONFIG_IPV6_SUBTREES
1869 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1871 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1872 rt->rt6i_table = ort->rt6i_table;
1877 #ifdef CONFIG_IPV6_ROUTE_INFO
1878 static struct rt6_info *rt6_get_route_info(struct net *net,
1879 const struct in6_addr *prefix, int prefixlen,
1880 const struct in6_addr *gwaddr, int ifindex)
1882 struct fib6_node *fn;
1883 struct rt6_info *rt = NULL;
1884 struct fib6_table *table;
1886 table = fib6_get_table(net, RT6_TABLE_INFO);
1890 read_lock_bh(&table->tb6_lock);
1891 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1895 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1896 if (rt->dst.dev->ifindex != ifindex)
1898 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1900 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1906 read_unlock_bh(&table->tb6_lock);
1910 static struct rt6_info *rt6_add_route_info(struct net *net,
1911 const struct in6_addr *prefix, int prefixlen,
1912 const struct in6_addr *gwaddr, int ifindex,
1915 struct fib6_config cfg = {
1916 .fc_table = RT6_TABLE_INFO,
1917 .fc_metric = IP6_RT_PRIO_USER,
1918 .fc_ifindex = ifindex,
1919 .fc_dst_len = prefixlen,
1920 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1921 RTF_UP | RTF_PREF(pref),
1922 .fc_nlinfo.portid = 0,
1923 .fc_nlinfo.nlh = NULL,
1924 .fc_nlinfo.nl_net = net,
1927 cfg.fc_dst = *prefix;
1928 cfg.fc_gateway = *gwaddr;
1930 /* We should treat it as a default route if prefix length is 0. */
1932 cfg.fc_flags |= RTF_DEFAULT;
1934 ip6_route_add(&cfg);
1936 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1940 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1942 struct rt6_info *rt;
1943 struct fib6_table *table;
1945 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1949 read_lock_bh(&table->tb6_lock);
1950 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1951 if (dev == rt->dst.dev &&
1952 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1953 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1958 read_unlock_bh(&table->tb6_lock);
1962 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1963 struct net_device *dev,
1966 struct fib6_config cfg = {
1967 .fc_table = RT6_TABLE_DFLT,
1968 .fc_metric = IP6_RT_PRIO_USER,
1969 .fc_ifindex = dev->ifindex,
1970 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1971 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1972 .fc_nlinfo.portid = 0,
1973 .fc_nlinfo.nlh = NULL,
1974 .fc_nlinfo.nl_net = dev_net(dev),
1977 cfg.fc_gateway = *gwaddr;
1979 ip6_route_add(&cfg);
1981 return rt6_get_dflt_router(gwaddr, dev);
1984 void rt6_purge_dflt_routers(struct net *net)
1986 struct rt6_info *rt;
1987 struct fib6_table *table;
1989 /* NOTE: Keep consistent with rt6_get_dflt_router */
1990 table = fib6_get_table(net, RT6_TABLE_DFLT);
1995 read_lock_bh(&table->tb6_lock);
1996 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1997 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1999 read_unlock_bh(&table->tb6_lock);
2004 read_unlock_bh(&table->tb6_lock);
2007 static void rtmsg_to_fib6_config(struct net *net,
2008 struct in6_rtmsg *rtmsg,
2009 struct fib6_config *cfg)
2011 memset(cfg, 0, sizeof(*cfg));
2013 cfg->fc_table = RT6_TABLE_MAIN;
2014 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2015 cfg->fc_metric = rtmsg->rtmsg_metric;
2016 cfg->fc_expires = rtmsg->rtmsg_info;
2017 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2018 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2019 cfg->fc_flags = rtmsg->rtmsg_flags;
2021 cfg->fc_nlinfo.nl_net = net;
2023 cfg->fc_dst = rtmsg->rtmsg_dst;
2024 cfg->fc_src = rtmsg->rtmsg_src;
2025 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2028 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2030 struct fib6_config cfg;
2031 struct in6_rtmsg rtmsg;
2035 case SIOCADDRT: /* Add a route */
2036 case SIOCDELRT: /* Delete a route */
2037 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2039 err = copy_from_user(&rtmsg, arg,
2040 sizeof(struct in6_rtmsg));
2044 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2049 err = ip6_route_add(&cfg);
2052 err = ip6_route_del(&cfg);
2066 * Drop the packet on the floor
2069 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2072 struct dst_entry *dst = skb_dst(skb);
2073 switch (ipstats_mib_noroutes) {
2074 case IPSTATS_MIB_INNOROUTES:
2075 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2076 if (type == IPV6_ADDR_ANY) {
2077 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2078 IPSTATS_MIB_INADDRERRORS);
2082 case IPSTATS_MIB_OUTNOROUTES:
2083 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2084 ipstats_mib_noroutes);
2087 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2092 static int ip6_pkt_discard(struct sk_buff *skb)
2094 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2097 static int ip6_pkt_discard_out(struct sk_buff *skb)
2099 skb->dev = skb_dst(skb)->dev;
2100 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2103 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2105 static int ip6_pkt_prohibit(struct sk_buff *skb)
2107 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2110 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2112 skb->dev = skb_dst(skb)->dev;
2113 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2119 * Allocate a dst for local (unicast / anycast) address.
2122 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2123 const struct in6_addr *addr,
2126 struct net *net = dev_net(idev->dev);
2127 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2131 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2132 return ERR_PTR(-ENOMEM);
2137 rt->dst.flags |= DST_HOST;
2138 rt->dst.input = ip6_input;
2139 rt->dst.output = ip6_output;
2140 rt->rt6i_idev = idev;
2142 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2144 rt->rt6i_flags |= RTF_ANYCAST;
2146 rt->rt6i_flags |= RTF_LOCAL;
2147 err = rt6_bind_neighbour(rt, rt->dst.dev);
2150 return ERR_PTR(err);
2153 rt->rt6i_dst.addr = *addr;
2154 rt->rt6i_dst.plen = 128;
2155 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2157 atomic_set(&rt->dst.__refcnt, 1);
2162 int ip6_route_get_saddr(struct net *net,
2163 struct rt6_info *rt,
2164 const struct in6_addr *daddr,
2166 struct in6_addr *saddr)
2168 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2170 if (rt->rt6i_prefsrc.plen)
2171 *saddr = rt->rt6i_prefsrc.addr;
2173 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2174 daddr, prefs, saddr);
2178 /* remove deleted ip from prefsrc entries */
2179 struct arg_dev_net_ip {
2180 struct net_device *dev;
2182 struct in6_addr *addr;
2185 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2187 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2188 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2189 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2191 if (((void *)rt->dst.dev == dev || !dev) &&
2192 rt != net->ipv6.ip6_null_entry &&
2193 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2194 /* remove prefsrc entry */
2195 rt->rt6i_prefsrc.plen = 0;
2200 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2202 struct net *net = dev_net(ifp->idev->dev);
2203 struct arg_dev_net_ip adni = {
2204 .dev = ifp->idev->dev,
2208 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2211 struct arg_dev_net {
2212 struct net_device *dev;
2216 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2218 const struct arg_dev_net *adn = arg;
2219 const struct net_device *dev = adn->dev;
2221 if ((rt->dst.dev == dev || !dev) &&
2222 rt != adn->net->ipv6.ip6_null_entry)
2228 void rt6_ifdown(struct net *net, struct net_device *dev)
2230 struct arg_dev_net adn = {
2235 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2236 icmp6_clean_all(fib6_ifdown, &adn);
2239 struct rt6_mtu_change_arg {
2240 struct net_device *dev;
2244 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2246 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2247 struct inet6_dev *idev;
2249 /* In IPv6 pmtu discovery is not optional,
2250 so that RTAX_MTU lock cannot disable it.
2251 We still use this lock to block changes
2252 caused by addrconf/ndisc.
2255 idev = __in6_dev_get(arg->dev);
2259 /* For administrative MTU increase, there is no way to discover
2260 IPv6 PMTU increase, so PMTU increase should be updated here.
2261 Since RFC 1981 doesn't include administrative MTU increase
2262 update PMTU increase is a MUST. (i.e. jumbo frame)
2265 If new MTU is less than route PMTU, this new MTU will be the
2266 lowest MTU in the path, update the route PMTU to reflect PMTU
2267 decreases; if new MTU is greater than route PMTU, and the
2268 old MTU is the lowest MTU in the path, update the route PMTU
2269 to reflect the increase. In this case if the other nodes' MTU
2270 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2273 if (rt->dst.dev == arg->dev &&
2274 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2275 (dst_mtu(&rt->dst) >= arg->mtu ||
2276 (dst_mtu(&rt->dst) < arg->mtu &&
2277 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2278 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2283 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2285 struct rt6_mtu_change_arg arg = {
2290 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2293 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2294 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2295 [RTA_OIF] = { .type = NLA_U32 },
2296 [RTA_IIF] = { .type = NLA_U32 },
2297 [RTA_PRIORITY] = { .type = NLA_U32 },
2298 [RTA_METRICS] = { .type = NLA_NESTED },
2299 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2302 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2303 struct fib6_config *cfg)
2306 struct nlattr *tb[RTA_MAX+1];
2309 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2314 rtm = nlmsg_data(nlh);
2315 memset(cfg, 0, sizeof(*cfg));
2317 cfg->fc_table = rtm->rtm_table;
2318 cfg->fc_dst_len = rtm->rtm_dst_len;
2319 cfg->fc_src_len = rtm->rtm_src_len;
2320 cfg->fc_flags = RTF_UP;
2321 cfg->fc_protocol = rtm->rtm_protocol;
2322 cfg->fc_type = rtm->rtm_type;
2324 if (rtm->rtm_type == RTN_UNREACHABLE ||
2325 rtm->rtm_type == RTN_BLACKHOLE ||
2326 rtm->rtm_type == RTN_PROHIBIT ||
2327 rtm->rtm_type == RTN_THROW)
2328 cfg->fc_flags |= RTF_REJECT;
2330 if (rtm->rtm_type == RTN_LOCAL)
2331 cfg->fc_flags |= RTF_LOCAL;
2333 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2334 cfg->fc_nlinfo.nlh = nlh;
2335 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2337 if (tb[RTA_GATEWAY]) {
2338 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2339 cfg->fc_flags |= RTF_GATEWAY;
2343 int plen = (rtm->rtm_dst_len + 7) >> 3;
2345 if (nla_len(tb[RTA_DST]) < plen)
2348 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2352 int plen = (rtm->rtm_src_len + 7) >> 3;
2354 if (nla_len(tb[RTA_SRC]) < plen)
2357 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2360 if (tb[RTA_PREFSRC])
2361 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2364 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2366 if (tb[RTA_PRIORITY])
2367 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2369 if (tb[RTA_METRICS]) {
2370 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2371 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2375 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2377 if (tb[RTA_MULTIPATH]) {
2378 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2379 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2387 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2389 struct fib6_config r_cfg;
2390 struct rtnexthop *rtnh;
2393 int err = 0, last_err = 0;
2396 rtnh = (struct rtnexthop *)cfg->fc_mp;
2397 remaining = cfg->fc_mp_len;
2399 /* Parse a Multipath Entry */
2400 while (rtnh_ok(rtnh, remaining)) {
2401 memcpy(&r_cfg, cfg, sizeof(*cfg));
2402 if (rtnh->rtnh_ifindex)
2403 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2405 attrlen = rtnh_attrlen(rtnh);
2407 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2409 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2411 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2412 r_cfg.fc_flags |= RTF_GATEWAY;
2415 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2418 /* If we are trying to remove a route, do not stop the
2419 * loop when ip6_route_del() fails (because next hop is
2420 * already gone), we should try to remove all next hops.
2423 /* If add fails, we should try to delete all
2424 * next hops that have been already added.
2430 /* Because each route is added like a single route we remove
2431 * this flag after the first nexthop (if there is a collision,
2432 * we have already fail to add the first nexthop:
2433 * fib6_add_rt2node() has reject it).
2435 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2436 rtnh = rtnh_next(rtnh, &remaining);
2442 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2444 struct fib6_config cfg;
2447 err = rtm_to_fib6_config(skb, nlh, &cfg);
2452 return ip6_route_multipath(&cfg, 0);
2454 return ip6_route_del(&cfg);
2457 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2459 struct fib6_config cfg;
2462 err = rtm_to_fib6_config(skb, nlh, &cfg);
2467 return ip6_route_multipath(&cfg, 1);
2469 return ip6_route_add(&cfg);
2472 static inline size_t rt6_nlmsg_size(void)
2474 return NLMSG_ALIGN(sizeof(struct rtmsg))
2475 + nla_total_size(16) /* RTA_SRC */
2476 + nla_total_size(16) /* RTA_DST */
2477 + nla_total_size(16) /* RTA_GATEWAY */
2478 + nla_total_size(16) /* RTA_PREFSRC */
2479 + nla_total_size(4) /* RTA_TABLE */
2480 + nla_total_size(4) /* RTA_IIF */
2481 + nla_total_size(4) /* RTA_OIF */
2482 + nla_total_size(4) /* RTA_PRIORITY */
2483 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2484 + nla_total_size(sizeof(struct rta_cacheinfo));
2487 static int rt6_fill_node(struct net *net,
2488 struct sk_buff *skb, struct rt6_info *rt,
2489 struct in6_addr *dst, struct in6_addr *src,
2490 int iif, int type, u32 portid, u32 seq,
2491 int prefix, int nowait, unsigned int flags)
2494 struct nlmsghdr *nlh;
2498 if (prefix) { /* user wants prefix routes only */
2499 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2500 /* success since this is not a prefix route */
2505 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2509 rtm = nlmsg_data(nlh);
2510 rtm->rtm_family = AF_INET6;
2511 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2512 rtm->rtm_src_len = rt->rt6i_src.plen;
2515 table = rt->rt6i_table->tb6_id;
2517 table = RT6_TABLE_UNSPEC;
2518 rtm->rtm_table = table;
2519 if (nla_put_u32(skb, RTA_TABLE, table))
2520 goto nla_put_failure;
2521 if (rt->rt6i_flags & RTF_REJECT) {
2522 switch (rt->dst.error) {
2524 rtm->rtm_type = RTN_BLACKHOLE;
2527 rtm->rtm_type = RTN_PROHIBIT;
2530 rtm->rtm_type = RTN_THROW;
2533 rtm->rtm_type = RTN_UNREACHABLE;
2537 else if (rt->rt6i_flags & RTF_LOCAL)
2538 rtm->rtm_type = RTN_LOCAL;
2539 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2540 rtm->rtm_type = RTN_LOCAL;
2542 rtm->rtm_type = RTN_UNICAST;
2544 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2545 rtm->rtm_protocol = rt->rt6i_protocol;
2546 if (rt->rt6i_flags & RTF_DYNAMIC)
2547 rtm->rtm_protocol = RTPROT_REDIRECT;
2548 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2549 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2550 rtm->rtm_protocol = RTPROT_RA;
2552 rtm->rtm_protocol = RTPROT_KERNEL;
2555 if (rt->rt6i_flags & RTF_CACHE)
2556 rtm->rtm_flags |= RTM_F_CLONED;
2559 if (nla_put(skb, RTA_DST, 16, dst))
2560 goto nla_put_failure;
2561 rtm->rtm_dst_len = 128;
2562 } else if (rtm->rtm_dst_len)
2563 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2564 goto nla_put_failure;
2565 #ifdef CONFIG_IPV6_SUBTREES
2567 if (nla_put(skb, RTA_SRC, 16, src))
2568 goto nla_put_failure;
2569 rtm->rtm_src_len = 128;
2570 } else if (rtm->rtm_src_len &&
2571 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2572 goto nla_put_failure;
2575 #ifdef CONFIG_IPV6_MROUTE
2576 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2577 int err = ip6mr_get_route(net, skb, rtm, nowait);
2582 goto nla_put_failure;
2584 if (err == -EMSGSIZE)
2585 goto nla_put_failure;
2590 if (nla_put_u32(skb, RTA_IIF, iif))
2591 goto nla_put_failure;
2593 struct in6_addr saddr_buf;
2594 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2595 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2596 goto nla_put_failure;
2599 if (rt->rt6i_prefsrc.plen) {
2600 struct in6_addr saddr_buf;
2601 saddr_buf = rt->rt6i_prefsrc.addr;
2602 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2603 goto nla_put_failure;
2606 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2607 goto nla_put_failure;
2609 if (rt->rt6i_flags & RTF_GATEWAY) {
2610 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2611 goto nla_put_failure;
2615 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2616 goto nla_put_failure;
2617 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2618 goto nla_put_failure;
2620 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2622 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2623 goto nla_put_failure;
2625 return nlmsg_end(skb, nlh);
2628 nlmsg_cancel(skb, nlh);
2632 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2634 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2637 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2638 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2639 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2643 return rt6_fill_node(arg->net,
2644 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2645 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2646 prefix, 0, NLM_F_MULTI);
2649 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2651 struct net *net = sock_net(in_skb->sk);
2652 struct nlattr *tb[RTA_MAX+1];
2653 struct rt6_info *rt;
2654 struct sk_buff *skb;
2657 int err, iif = 0, oif = 0;
2659 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2664 memset(&fl6, 0, sizeof(fl6));
2667 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2670 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2674 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2677 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2681 iif = nla_get_u32(tb[RTA_IIF]);
2684 oif = nla_get_u32(tb[RTA_OIF]);
2687 struct net_device *dev;
2690 dev = __dev_get_by_index(net, iif);
2696 fl6.flowi6_iif = iif;
2698 if (!ipv6_addr_any(&fl6.saddr))
2699 flags |= RT6_LOOKUP_F_HAS_SADDR;
2701 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2704 fl6.flowi6_oif = oif;
2706 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2709 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2716 /* Reserve room for dummy headers, this skb can pass
2717 through good chunk of routing engine.
2719 skb_reset_mac_header(skb);
2720 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2722 skb_dst_set(skb, &rt->dst);
2724 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2725 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2726 nlh->nlmsg_seq, 0, 0, 0);
2732 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2737 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2739 struct sk_buff *skb;
2740 struct net *net = info->nl_net;
2745 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2747 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2751 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2752 event, info->portid, seq, 0, 0, 0);
2754 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2755 WARN_ON(err == -EMSGSIZE);
2759 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2760 info->nlh, gfp_any());
2764 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2767 static int ip6_route_dev_notify(struct notifier_block *this,
2768 unsigned long event, void *data)
2770 struct net_device *dev = (struct net_device *)data;
2771 struct net *net = dev_net(dev);
2773 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2774 net->ipv6.ip6_null_entry->dst.dev = dev;
2775 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2776 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2777 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2778 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2779 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2780 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2791 #ifdef CONFIG_PROC_FS
2802 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2804 struct seq_file *m = p_arg;
2806 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2808 #ifdef CONFIG_IPV6_SUBTREES
2809 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2811 seq_puts(m, "00000000000000000000000000000000 00 ");
2813 if (rt->rt6i_flags & RTF_GATEWAY) {
2814 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2816 seq_puts(m, "00000000000000000000000000000000");
2818 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2819 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2820 rt->dst.__use, rt->rt6i_flags,
2821 rt->dst.dev ? rt->dst.dev->name : "");
2825 static int ipv6_route_show(struct seq_file *m, void *v)
2827 struct net *net = (struct net *)m->private;
2828 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2832 static int ipv6_route_open(struct inode *inode, struct file *file)
2834 return single_open_net(inode, file, ipv6_route_show);
2837 static const struct file_operations ipv6_route_proc_fops = {
2838 .owner = THIS_MODULE,
2839 .open = ipv6_route_open,
2841 .llseek = seq_lseek,
2842 .release = single_release_net,
2845 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2847 struct net *net = (struct net *)seq->private;
2848 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2849 net->ipv6.rt6_stats->fib_nodes,
2850 net->ipv6.rt6_stats->fib_route_nodes,
2851 net->ipv6.rt6_stats->fib_rt_alloc,
2852 net->ipv6.rt6_stats->fib_rt_entries,
2853 net->ipv6.rt6_stats->fib_rt_cache,
2854 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2855 net->ipv6.rt6_stats->fib_discarded_routes);
2860 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2862 return single_open_net(inode, file, rt6_stats_seq_show);
2865 static const struct file_operations rt6_stats_seq_fops = {
2866 .owner = THIS_MODULE,
2867 .open = rt6_stats_seq_open,
2869 .llseek = seq_lseek,
2870 .release = single_release_net,
2872 #endif /* CONFIG_PROC_FS */
2874 #ifdef CONFIG_SYSCTL
2877 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2878 void __user *buffer, size_t *lenp, loff_t *ppos)
2885 net = (struct net *)ctl->extra1;
2886 delay = net->ipv6.sysctl.flush_delay;
2887 proc_dointvec(ctl, write, buffer, lenp, ppos);
2888 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2892 ctl_table ipv6_route_table_template[] = {
2894 .procname = "flush",
2895 .data = &init_net.ipv6.sysctl.flush_delay,
2896 .maxlen = sizeof(int),
2898 .proc_handler = ipv6_sysctl_rtcache_flush
2901 .procname = "gc_thresh",
2902 .data = &ip6_dst_ops_template.gc_thresh,
2903 .maxlen = sizeof(int),
2905 .proc_handler = proc_dointvec,
2908 .procname = "max_size",
2909 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2910 .maxlen = sizeof(int),
2912 .proc_handler = proc_dointvec,
2915 .procname = "gc_min_interval",
2916 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2917 .maxlen = sizeof(int),
2919 .proc_handler = proc_dointvec_jiffies,
2922 .procname = "gc_timeout",
2923 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2924 .maxlen = sizeof(int),
2926 .proc_handler = proc_dointvec_jiffies,
2929 .procname = "gc_interval",
2930 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2931 .maxlen = sizeof(int),
2933 .proc_handler = proc_dointvec_jiffies,
2936 .procname = "gc_elasticity",
2937 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2938 .maxlen = sizeof(int),
2940 .proc_handler = proc_dointvec,
2943 .procname = "mtu_expires",
2944 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2945 .maxlen = sizeof(int),
2947 .proc_handler = proc_dointvec_jiffies,
2950 .procname = "min_adv_mss",
2951 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2952 .maxlen = sizeof(int),
2954 .proc_handler = proc_dointvec,
2957 .procname = "gc_min_interval_ms",
2958 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2959 .maxlen = sizeof(int),
2961 .proc_handler = proc_dointvec_ms_jiffies,
2966 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2968 struct ctl_table *table;
2970 table = kmemdup(ipv6_route_table_template,
2971 sizeof(ipv6_route_table_template),
2975 table[0].data = &net->ipv6.sysctl.flush_delay;
2976 table[0].extra1 = net;
2977 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2978 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2979 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2980 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2981 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2982 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2983 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2984 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2985 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2987 /* Don't export sysctls to unprivileged users */
2988 if (net->user_ns != &init_user_ns)
2989 table[0].procname = NULL;
2996 static int __net_init ip6_route_net_init(struct net *net)
3000 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3001 sizeof(net->ipv6.ip6_dst_ops));
3003 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3004 goto out_ip6_dst_ops;
3006 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3007 sizeof(*net->ipv6.ip6_null_entry),
3009 if (!net->ipv6.ip6_null_entry)
3010 goto out_ip6_dst_entries;
3011 net->ipv6.ip6_null_entry->dst.path =
3012 (struct dst_entry *)net->ipv6.ip6_null_entry;
3013 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3014 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3015 ip6_template_metrics, true);
3017 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3018 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3019 sizeof(*net->ipv6.ip6_prohibit_entry),
3021 if (!net->ipv6.ip6_prohibit_entry)
3022 goto out_ip6_null_entry;
3023 net->ipv6.ip6_prohibit_entry->dst.path =
3024 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3025 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3026 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3027 ip6_template_metrics, true);
3029 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3030 sizeof(*net->ipv6.ip6_blk_hole_entry),
3032 if (!net->ipv6.ip6_blk_hole_entry)
3033 goto out_ip6_prohibit_entry;
3034 net->ipv6.ip6_blk_hole_entry->dst.path =
3035 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3036 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3037 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3038 ip6_template_metrics, true);
3041 net->ipv6.sysctl.flush_delay = 0;
3042 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3043 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3044 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3045 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3046 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3047 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3048 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3050 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3056 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3057 out_ip6_prohibit_entry:
3058 kfree(net->ipv6.ip6_prohibit_entry);
3060 kfree(net->ipv6.ip6_null_entry);
3062 out_ip6_dst_entries:
3063 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3068 static void __net_exit ip6_route_net_exit(struct net *net)
3070 kfree(net->ipv6.ip6_null_entry);
3071 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3072 kfree(net->ipv6.ip6_prohibit_entry);
3073 kfree(net->ipv6.ip6_blk_hole_entry);
3075 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3078 static int __net_init ip6_route_net_init_late(struct net *net)
3080 #ifdef CONFIG_PROC_FS
3081 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
3082 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
3087 static void __net_exit ip6_route_net_exit_late(struct net *net)
3089 #ifdef CONFIG_PROC_FS
3090 proc_net_remove(net, "ipv6_route");
3091 proc_net_remove(net, "rt6_stats");
3095 static struct pernet_operations ip6_route_net_ops = {
3096 .init = ip6_route_net_init,
3097 .exit = ip6_route_net_exit,
3100 static int __net_init ipv6_inetpeer_init(struct net *net)
3102 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3106 inet_peer_base_init(bp);
3107 net->ipv6.peers = bp;
3111 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3113 struct inet_peer_base *bp = net->ipv6.peers;
3115 net->ipv6.peers = NULL;
3116 inetpeer_invalidate_tree(bp);
3120 static struct pernet_operations ipv6_inetpeer_ops = {
3121 .init = ipv6_inetpeer_init,
3122 .exit = ipv6_inetpeer_exit,
3125 static struct pernet_operations ip6_route_net_late_ops = {
3126 .init = ip6_route_net_init_late,
3127 .exit = ip6_route_net_exit_late,
3130 static struct notifier_block ip6_route_dev_notifier = {
3131 .notifier_call = ip6_route_dev_notify,
3135 int __init ip6_route_init(void)
3140 ip6_dst_ops_template.kmem_cachep =
3141 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3142 SLAB_HWCACHE_ALIGN, NULL);
3143 if (!ip6_dst_ops_template.kmem_cachep)
3146 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3148 goto out_kmem_cache;
3150 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3152 goto out_dst_entries;
3154 ret = register_pernet_subsys(&ip6_route_net_ops);
3156 goto out_register_inetpeer;
3158 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3160 /* Registering of the loopback is done before this portion of code,
3161 * the loopback reference in rt6_info will not be taken, do it
3162 * manually for init_net */
3163 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3164 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3165 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3166 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3167 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3168 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3169 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3173 goto out_register_subsys;
3179 ret = fib6_rules_init();
3183 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3185 goto fib6_rules_init;
3188 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3189 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3190 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3191 goto out_register_late_subsys;
3193 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3195 goto out_register_late_subsys;
3200 out_register_late_subsys:
3201 unregister_pernet_subsys(&ip6_route_net_late_ops);
3203 fib6_rules_cleanup();
3208 out_register_subsys:
3209 unregister_pernet_subsys(&ip6_route_net_ops);
3210 out_register_inetpeer:
3211 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3213 dst_entries_destroy(&ip6_dst_blackhole_ops);
3215 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3219 void ip6_route_cleanup(void)
3221 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3222 unregister_pernet_subsys(&ip6_route_net_late_ops);
3223 fib6_rules_cleanup();
3226 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3227 unregister_pernet_subsys(&ip6_route_net_ops);
3228 dst_entries_destroy(&ip6_dst_blackhole_ops);
3229 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);