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(&nd_tbl, 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(&nd_tbl, 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 ^= (__force u32)fl6->daddr.s6_addr32[0];
392 val ^= (__force u32)fl6->daddr.s6_addr32[1];
393 val ^= (__force u32)fl6->daddr.s6_addr32[2];
394 val ^= (__force u32)fl6->daddr.s6_addr32[3];
396 val ^= (__force u32)fl6->saddr.s6_addr32[0];
397 val ^= (__force u32)fl6->saddr.s6_addr32[1];
398 val ^= (__force u32)fl6->saddr.s6_addr32[2];
399 val ^= (__force u32)fl6->saddr.s6_addr32[3];
401 /* Work only if this not encapsulated */
402 switch (fl6->flowi6_proto) {
406 val ^= (__force u16)fl6->fl6_sport;
407 val ^= (__force u16)fl6->fl6_dport;
411 val ^= (__force u16)fl6->fl6_icmp_type;
412 val ^= (__force u16)fl6->fl6_icmp_code;
415 /* RFC6438 recommands to use flowlabel */
416 val ^= (__force u32)fl6->flowlabel;
418 /* Perhaps, we need to tune, this function? */
419 val = val ^ (val >> 7) ^ (val >> 12);
420 return val % candidate_count;
423 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
426 struct rt6_info *sibling, *next_sibling;
429 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
430 /* Don't change the route, if route_choosen == 0
431 * (siblings does not include ourself)
434 list_for_each_entry_safe(sibling, next_sibling,
435 &match->rt6i_siblings, rt6i_siblings) {
437 if (route_choosen == 0) {
446 * Route lookup. Any table->tb6_lock is implied.
449 static inline struct rt6_info *rt6_device_match(struct net *net,
451 const struct in6_addr *saddr,
455 struct rt6_info *local = NULL;
456 struct rt6_info *sprt;
458 if (!oif && ipv6_addr_any(saddr))
461 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
462 struct net_device *dev = sprt->dst.dev;
465 if (dev->ifindex == oif)
467 if (dev->flags & IFF_LOOPBACK) {
468 if (!sprt->rt6i_idev ||
469 sprt->rt6i_idev->dev->ifindex != oif) {
470 if (flags & RT6_LOOKUP_F_IFACE && oif)
472 if (local && (!oif ||
473 local->rt6i_idev->dev->ifindex == oif))
479 if (ipv6_chk_addr(net, saddr, dev,
480 flags & RT6_LOOKUP_F_IFACE))
489 if (flags & RT6_LOOKUP_F_IFACE)
490 return net->ipv6.ip6_null_entry;
496 #ifdef CONFIG_IPV6_ROUTER_PREF
497 static void rt6_probe(struct rt6_info *rt)
499 struct neighbour *neigh;
501 * Okay, this does not seem to be appropriate
502 * for now, however, we need to check if it
503 * is really so; aka Router Reachability Probing.
505 * Router Reachability Probe MUST be rate-limited
506 * to no more than one per minute.
508 neigh = rt ? rt->n : NULL;
509 if (!neigh || (neigh->nud_state & NUD_VALID))
511 read_lock_bh(&neigh->lock);
512 if (!(neigh->nud_state & NUD_VALID) &&
513 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
514 struct in6_addr mcaddr;
515 struct in6_addr *target;
517 neigh->updated = jiffies;
518 read_unlock_bh(&neigh->lock);
520 target = (struct in6_addr *)&neigh->primary_key;
521 addrconf_addr_solict_mult(target, &mcaddr);
522 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
524 read_unlock_bh(&neigh->lock);
528 static inline void rt6_probe(struct rt6_info *rt)
534 * Default Router Selection (RFC 2461 6.3.6)
536 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
538 struct net_device *dev = rt->dst.dev;
539 if (!oif || dev->ifindex == oif)
541 if ((dev->flags & IFF_LOOPBACK) &&
542 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
547 static inline bool rt6_check_neigh(struct rt6_info *rt)
549 struct neighbour *neigh;
553 if (rt->rt6i_flags & RTF_NONEXTHOP ||
554 !(rt->rt6i_flags & RTF_GATEWAY))
557 read_lock_bh(&neigh->lock);
558 if (neigh->nud_state & NUD_VALID)
560 #ifdef CONFIG_IPV6_ROUTER_PREF
561 else if (!(neigh->nud_state & NUD_FAILED))
564 read_unlock_bh(&neigh->lock);
569 static int rt6_score_route(struct rt6_info *rt, int oif,
574 m = rt6_check_dev(rt, oif);
575 if (!m && (strict & RT6_LOOKUP_F_IFACE))
577 #ifdef CONFIG_IPV6_ROUTER_PREF
578 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
580 if (!rt6_check_neigh(rt) && (strict & RT6_LOOKUP_F_REACHABLE))
585 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
586 int *mpri, struct rt6_info *match)
590 if (rt6_check_expired(rt))
593 m = rt6_score_route(rt, oif, strict);
598 if (strict & RT6_LOOKUP_F_REACHABLE)
602 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
610 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
611 struct rt6_info *rr_head,
612 u32 metric, int oif, int strict)
614 struct rt6_info *rt, *match;
618 for (rt = rr_head; rt && rt->rt6i_metric == metric;
619 rt = rt->dst.rt6_next)
620 match = find_match(rt, oif, strict, &mpri, match);
621 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
622 rt = rt->dst.rt6_next)
623 match = find_match(rt, oif, strict, &mpri, match);
628 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
630 struct rt6_info *match, *rt0;
635 fn->rr_ptr = rt0 = fn->leaf;
637 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
640 (strict & RT6_LOOKUP_F_REACHABLE)) {
641 struct rt6_info *next = rt0->dst.rt6_next;
643 /* no entries matched; do round-robin */
644 if (!next || next->rt6i_metric != rt0->rt6i_metric)
651 net = dev_net(rt0->dst.dev);
652 return match ? match : net->ipv6.ip6_null_entry;
655 #ifdef CONFIG_IPV6_ROUTE_INFO
656 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
657 const struct in6_addr *gwaddr)
659 struct net *net = dev_net(dev);
660 struct route_info *rinfo = (struct route_info *) opt;
661 struct in6_addr prefix_buf, *prefix;
663 unsigned long lifetime;
666 if (len < sizeof(struct route_info)) {
670 /* Sanity check for prefix_len and length */
671 if (rinfo->length > 3) {
673 } else if (rinfo->prefix_len > 128) {
675 } else if (rinfo->prefix_len > 64) {
676 if (rinfo->length < 2) {
679 } else if (rinfo->prefix_len > 0) {
680 if (rinfo->length < 1) {
685 pref = rinfo->route_pref;
686 if (pref == ICMPV6_ROUTER_PREF_INVALID)
689 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
691 if (rinfo->length == 3)
692 prefix = (struct in6_addr *)rinfo->prefix;
694 /* this function is safe */
695 ipv6_addr_prefix(&prefix_buf,
696 (struct in6_addr *)rinfo->prefix,
698 prefix = &prefix_buf;
701 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
704 if (rt && !lifetime) {
710 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
713 rt->rt6i_flags = RTF_ROUTEINFO |
714 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
717 if (!addrconf_finite_timeout(lifetime))
718 rt6_clean_expires(rt);
720 rt6_set_expires(rt, jiffies + HZ * lifetime);
728 #define BACKTRACK(__net, saddr) \
730 if (rt == __net->ipv6.ip6_null_entry) { \
731 struct fib6_node *pn; \
733 if (fn->fn_flags & RTN_TL_ROOT) \
736 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
737 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
740 if (fn->fn_flags & RTN_RTINFO) \
746 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
747 struct fib6_table *table,
748 struct flowi6 *fl6, int flags)
750 struct fib6_node *fn;
753 read_lock_bh(&table->tb6_lock);
754 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
757 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
758 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
759 rt = rt6_multipath_select(rt, fl6);
760 BACKTRACK(net, &fl6->saddr);
762 dst_use(&rt->dst, jiffies);
763 read_unlock_bh(&table->tb6_lock);
768 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
771 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
773 EXPORT_SYMBOL_GPL(ip6_route_lookup);
775 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
776 const struct in6_addr *saddr, int oif, int strict)
778 struct flowi6 fl6 = {
782 struct dst_entry *dst;
783 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
786 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
787 flags |= RT6_LOOKUP_F_HAS_SADDR;
790 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
792 return (struct rt6_info *) dst;
799 EXPORT_SYMBOL(rt6_lookup);
801 /* ip6_ins_rt is called with FREE table->tb6_lock.
802 It takes new route entry, the addition fails by any reason the
803 route is freed. In any case, if caller does not hold it, it may
807 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
810 struct fib6_table *table;
812 table = rt->rt6i_table;
813 write_lock_bh(&table->tb6_lock);
814 err = fib6_add(&table->tb6_root, rt, info);
815 write_unlock_bh(&table->tb6_lock);
820 int ip6_ins_rt(struct rt6_info *rt)
822 struct nl_info info = {
823 .nl_net = dev_net(rt->dst.dev),
825 return __ip6_ins_rt(rt, &info);
828 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
829 const struct in6_addr *daddr,
830 const struct in6_addr *saddr)
838 rt = ip6_rt_copy(ort, daddr);
841 int attempts = !in_softirq();
843 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
844 if (ort->rt6i_dst.plen != 128 &&
845 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
846 rt->rt6i_flags |= RTF_ANYCAST;
847 rt->rt6i_gateway = *daddr;
850 rt->rt6i_flags |= RTF_CACHE;
852 #ifdef CONFIG_IPV6_SUBTREES
853 if (rt->rt6i_src.plen && saddr) {
854 rt->rt6i_src.addr = *saddr;
855 rt->rt6i_src.plen = 128;
860 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
861 struct net *net = dev_net(rt->dst.dev);
862 int saved_rt_min_interval =
863 net->ipv6.sysctl.ip6_rt_gc_min_interval;
864 int saved_rt_elasticity =
865 net->ipv6.sysctl.ip6_rt_gc_elasticity;
867 if (attempts-- > 0) {
868 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
869 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
871 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
873 net->ipv6.sysctl.ip6_rt_gc_elasticity =
875 net->ipv6.sysctl.ip6_rt_gc_min_interval =
876 saved_rt_min_interval;
880 net_warn_ratelimited("Neighbour table overflow\n");
889 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
890 const struct in6_addr *daddr)
892 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
895 rt->rt6i_flags |= RTF_CACHE;
896 rt->n = neigh_clone(ort->n);
901 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
902 struct flowi6 *fl6, int flags)
904 struct fib6_node *fn;
905 struct rt6_info *rt, *nrt;
909 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
911 strict |= flags & RT6_LOOKUP_F_IFACE;
914 read_lock_bh(&table->tb6_lock);
917 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
920 rt = rt6_select(fn, oif, strict | reachable);
921 if (rt->rt6i_nsiblings && oif == 0)
922 rt = rt6_multipath_select(rt, fl6);
923 BACKTRACK(net, &fl6->saddr);
924 if (rt == net->ipv6.ip6_null_entry ||
925 rt->rt6i_flags & RTF_CACHE)
929 read_unlock_bh(&table->tb6_lock);
931 if (!rt->n && !(rt->rt6i_flags & RTF_NONEXTHOP))
932 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
933 else if (!(rt->dst.flags & DST_HOST))
934 nrt = rt6_alloc_clone(rt, &fl6->daddr);
939 rt = nrt ? : net->ipv6.ip6_null_entry;
943 err = ip6_ins_rt(nrt);
952 * Race condition! In the gap, when table->tb6_lock was
953 * released someone could insert this route. Relookup.
964 read_unlock_bh(&table->tb6_lock);
966 rt->dst.lastuse = jiffies;
972 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
973 struct flowi6 *fl6, int flags)
975 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
978 static struct dst_entry *ip6_route_input_lookup(struct net *net,
979 struct net_device *dev,
980 struct flowi6 *fl6, int flags)
982 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
983 flags |= RT6_LOOKUP_F_IFACE;
985 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
988 void ip6_route_input(struct sk_buff *skb)
990 const struct ipv6hdr *iph = ipv6_hdr(skb);
991 struct net *net = dev_net(skb->dev);
992 int flags = RT6_LOOKUP_F_HAS_SADDR;
993 struct flowi6 fl6 = {
994 .flowi6_iif = skb->dev->ifindex,
997 .flowlabel = ip6_flowinfo(iph),
998 .flowi6_mark = skb->mark,
999 .flowi6_proto = iph->nexthdr,
1002 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1005 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1006 struct flowi6 *fl6, int flags)
1008 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1011 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1016 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1018 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1019 flags |= RT6_LOOKUP_F_IFACE;
1021 if (!ipv6_addr_any(&fl6->saddr))
1022 flags |= RT6_LOOKUP_F_HAS_SADDR;
1024 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1026 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1029 EXPORT_SYMBOL(ip6_route_output);
1031 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1033 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1034 struct dst_entry *new = NULL;
1036 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1040 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1041 rt6_init_peer(rt, net->ipv6.peers);
1044 new->input = dst_discard;
1045 new->output = dst_discard;
1047 if (dst_metrics_read_only(&ort->dst))
1048 new->_metrics = ort->dst._metrics;
1050 dst_copy_metrics(new, &ort->dst);
1051 rt->rt6i_idev = ort->rt6i_idev;
1053 in6_dev_hold(rt->rt6i_idev);
1055 rt->rt6i_gateway = ort->rt6i_gateway;
1056 rt->rt6i_flags = ort->rt6i_flags;
1057 rt6_clean_expires(rt);
1058 rt->rt6i_metric = 0;
1060 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1061 #ifdef CONFIG_IPV6_SUBTREES
1062 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1068 dst_release(dst_orig);
1069 return new ? new : ERR_PTR(-ENOMEM);
1073 * Destination cache support functions
1076 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1078 struct rt6_info *rt;
1080 rt = (struct rt6_info *) dst;
1082 /* All IPV6 dsts are created with ->obsolete set to the value
1083 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1084 * into this function always.
1086 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1089 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1095 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1097 struct rt6_info *rt = (struct rt6_info *) dst;
1100 if (rt->rt6i_flags & RTF_CACHE) {
1101 if (rt6_check_expired(rt)) {
1113 static void ip6_link_failure(struct sk_buff *skb)
1115 struct rt6_info *rt;
1117 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1119 rt = (struct rt6_info *) skb_dst(skb);
1121 if (rt->rt6i_flags & RTF_CACHE)
1122 rt6_update_expires(rt, 0);
1123 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1124 rt->rt6i_node->fn_sernum = -1;
1128 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1129 struct sk_buff *skb, u32 mtu)
1131 struct rt6_info *rt6 = (struct rt6_info*)dst;
1134 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1135 struct net *net = dev_net(dst->dev);
1137 rt6->rt6i_flags |= RTF_MODIFIED;
1138 if (mtu < IPV6_MIN_MTU) {
1139 u32 features = dst_metric(dst, RTAX_FEATURES);
1141 features |= RTAX_FEATURE_ALLFRAG;
1142 dst_metric_set(dst, RTAX_FEATURES, features);
1144 dst_metric_set(dst, RTAX_MTU, mtu);
1145 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1149 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1152 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1153 struct dst_entry *dst;
1156 memset(&fl6, 0, sizeof(fl6));
1157 fl6.flowi6_oif = oif;
1158 fl6.flowi6_mark = mark;
1159 fl6.flowi6_flags = 0;
1160 fl6.daddr = iph->daddr;
1161 fl6.saddr = iph->saddr;
1162 fl6.flowlabel = ip6_flowinfo(iph);
1164 dst = ip6_route_output(net, NULL, &fl6);
1166 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1169 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1171 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1173 ip6_update_pmtu(skb, sock_net(sk), mtu,
1174 sk->sk_bound_dev_if, sk->sk_mark);
1176 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1178 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1180 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1181 struct dst_entry *dst;
1184 memset(&fl6, 0, sizeof(fl6));
1185 fl6.flowi6_oif = oif;
1186 fl6.flowi6_mark = mark;
1187 fl6.flowi6_flags = 0;
1188 fl6.daddr = iph->daddr;
1189 fl6.saddr = iph->saddr;
1190 fl6.flowlabel = ip6_flowinfo(iph);
1192 dst = ip6_route_output(net, NULL, &fl6);
1194 rt6_do_redirect(dst, NULL, skb);
1197 EXPORT_SYMBOL_GPL(ip6_redirect);
1199 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1201 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1203 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1205 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1207 struct net_device *dev = dst->dev;
1208 unsigned int mtu = dst_mtu(dst);
1209 struct net *net = dev_net(dev);
1211 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1213 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1214 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1217 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1218 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1219 * IPV6_MAXPLEN is also valid and means: "any MSS,
1220 * rely only on pmtu discovery"
1222 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1227 static unsigned int ip6_mtu(const struct dst_entry *dst)
1229 struct inet6_dev *idev;
1230 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1238 idev = __in6_dev_get(dst->dev);
1240 mtu = idev->cnf.mtu6;
1246 static struct dst_entry *icmp6_dst_gc_list;
1247 static DEFINE_SPINLOCK(icmp6_dst_lock);
1249 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1250 struct neighbour *neigh,
1253 struct dst_entry *dst;
1254 struct rt6_info *rt;
1255 struct inet6_dev *idev = in6_dev_get(dev);
1256 struct net *net = dev_net(dev);
1258 if (unlikely(!idev))
1259 return ERR_PTR(-ENODEV);
1261 rt = ip6_dst_alloc(net, dev, 0, NULL);
1262 if (unlikely(!rt)) {
1264 dst = ERR_PTR(-ENOMEM);
1271 neigh = ip6_neigh_lookup(&rt->dst, NULL, &fl6->daddr);
1272 if (IS_ERR(neigh)) {
1275 return ERR_CAST(neigh);
1279 rt->dst.flags |= DST_HOST;
1280 rt->dst.output = ip6_output;
1282 atomic_set(&rt->dst.__refcnt, 1);
1283 rt->rt6i_dst.addr = fl6->daddr;
1284 rt->rt6i_dst.plen = 128;
1285 rt->rt6i_idev = idev;
1286 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1288 spin_lock_bh(&icmp6_dst_lock);
1289 rt->dst.next = icmp6_dst_gc_list;
1290 icmp6_dst_gc_list = &rt->dst;
1291 spin_unlock_bh(&icmp6_dst_lock);
1293 fib6_force_start_gc(net);
1295 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1301 int icmp6_dst_gc(void)
1303 struct dst_entry *dst, **pprev;
1306 spin_lock_bh(&icmp6_dst_lock);
1307 pprev = &icmp6_dst_gc_list;
1309 while ((dst = *pprev) != NULL) {
1310 if (!atomic_read(&dst->__refcnt)) {
1319 spin_unlock_bh(&icmp6_dst_lock);
1324 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1327 struct dst_entry *dst, **pprev;
1329 spin_lock_bh(&icmp6_dst_lock);
1330 pprev = &icmp6_dst_gc_list;
1331 while ((dst = *pprev) != NULL) {
1332 struct rt6_info *rt = (struct rt6_info *) dst;
1333 if (func(rt, arg)) {
1340 spin_unlock_bh(&icmp6_dst_lock);
1343 static int ip6_dst_gc(struct dst_ops *ops)
1345 unsigned long now = jiffies;
1346 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1347 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1348 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1349 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1350 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1351 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1354 entries = dst_entries_get_fast(ops);
1355 if (time_after(rt_last_gc + rt_min_interval, now) &&
1356 entries <= rt_max_size)
1359 net->ipv6.ip6_rt_gc_expire++;
1360 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1361 net->ipv6.ip6_rt_last_gc = now;
1362 entries = dst_entries_get_slow(ops);
1363 if (entries < ops->gc_thresh)
1364 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1366 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1367 return entries > rt_max_size;
1370 int ip6_dst_hoplimit(struct dst_entry *dst)
1372 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1373 if (hoplimit == 0) {
1374 struct net_device *dev = dst->dev;
1375 struct inet6_dev *idev;
1378 idev = __in6_dev_get(dev);
1380 hoplimit = idev->cnf.hop_limit;
1382 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1387 EXPORT_SYMBOL(ip6_dst_hoplimit);
1393 int ip6_route_add(struct fib6_config *cfg)
1396 struct net *net = cfg->fc_nlinfo.nl_net;
1397 struct rt6_info *rt = NULL;
1398 struct net_device *dev = NULL;
1399 struct inet6_dev *idev = NULL;
1400 struct fib6_table *table;
1403 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1405 #ifndef CONFIG_IPV6_SUBTREES
1406 if (cfg->fc_src_len)
1409 if (cfg->fc_ifindex) {
1411 dev = dev_get_by_index(net, cfg->fc_ifindex);
1414 idev = in6_dev_get(dev);
1419 if (cfg->fc_metric == 0)
1420 cfg->fc_metric = IP6_RT_PRIO_USER;
1423 if (cfg->fc_nlinfo.nlh &&
1424 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1425 table = fib6_get_table(net, cfg->fc_table);
1427 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1428 table = fib6_new_table(net, cfg->fc_table);
1431 table = fib6_new_table(net, cfg->fc_table);
1437 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1444 if (cfg->fc_flags & RTF_EXPIRES)
1445 rt6_set_expires(rt, jiffies +
1446 clock_t_to_jiffies(cfg->fc_expires));
1448 rt6_clean_expires(rt);
1450 if (cfg->fc_protocol == RTPROT_UNSPEC)
1451 cfg->fc_protocol = RTPROT_BOOT;
1452 rt->rt6i_protocol = cfg->fc_protocol;
1454 addr_type = ipv6_addr_type(&cfg->fc_dst);
1456 if (addr_type & IPV6_ADDR_MULTICAST)
1457 rt->dst.input = ip6_mc_input;
1458 else if (cfg->fc_flags & RTF_LOCAL)
1459 rt->dst.input = ip6_input;
1461 rt->dst.input = ip6_forward;
1463 rt->dst.output = ip6_output;
1465 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1466 rt->rt6i_dst.plen = cfg->fc_dst_len;
1467 if (rt->rt6i_dst.plen == 128)
1468 rt->dst.flags |= DST_HOST;
1470 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1471 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1476 dst_init_metrics(&rt->dst, metrics, 0);
1478 #ifdef CONFIG_IPV6_SUBTREES
1479 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1480 rt->rt6i_src.plen = cfg->fc_src_len;
1483 rt->rt6i_metric = cfg->fc_metric;
1485 /* We cannot add true routes via loopback here,
1486 they would result in kernel looping; promote them to reject routes
1488 if ((cfg->fc_flags & RTF_REJECT) ||
1489 (dev && (dev->flags & IFF_LOOPBACK) &&
1490 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1491 !(cfg->fc_flags & RTF_LOCAL))) {
1492 /* hold loopback dev/idev if we haven't done so. */
1493 if (dev != net->loopback_dev) {
1498 dev = net->loopback_dev;
1500 idev = in6_dev_get(dev);
1506 rt->dst.output = ip6_pkt_discard_out;
1507 rt->dst.input = ip6_pkt_discard;
1508 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1509 switch (cfg->fc_type) {
1511 rt->dst.error = -EINVAL;
1514 rt->dst.error = -EACCES;
1517 rt->dst.error = -EAGAIN;
1520 rt->dst.error = -ENETUNREACH;
1526 if (cfg->fc_flags & RTF_GATEWAY) {
1527 const struct in6_addr *gw_addr;
1530 gw_addr = &cfg->fc_gateway;
1531 rt->rt6i_gateway = *gw_addr;
1532 gwa_type = ipv6_addr_type(gw_addr);
1534 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1535 struct rt6_info *grt;
1537 /* IPv6 strictly inhibits using not link-local
1538 addresses as nexthop address.
1539 Otherwise, router will not able to send redirects.
1540 It is very good, but in some (rare!) circumstances
1541 (SIT, PtP, NBMA NOARP links) it is handy to allow
1542 some exceptions. --ANK
1545 if (!(gwa_type & IPV6_ADDR_UNICAST))
1548 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1550 err = -EHOSTUNREACH;
1554 if (dev != grt->dst.dev) {
1560 idev = grt->rt6i_idev;
1562 in6_dev_hold(grt->rt6i_idev);
1564 if (!(grt->rt6i_flags & RTF_GATEWAY))
1572 if (!dev || (dev->flags & IFF_LOOPBACK))
1580 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1581 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1585 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1586 rt->rt6i_prefsrc.plen = 128;
1588 rt->rt6i_prefsrc.plen = 0;
1590 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1591 err = rt6_bind_neighbour(rt, dev);
1596 rt->rt6i_flags = cfg->fc_flags;
1603 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1604 int type = nla_type(nla);
1607 if (type > RTAX_MAX) {
1612 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1618 rt->rt6i_idev = idev;
1619 rt->rt6i_table = table;
1621 cfg->fc_nlinfo.nl_net = dev_net(dev);
1623 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1635 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1638 struct fib6_table *table;
1639 struct net *net = dev_net(rt->dst.dev);
1641 if (rt == net->ipv6.ip6_null_entry) {
1646 table = rt->rt6i_table;
1647 write_lock_bh(&table->tb6_lock);
1648 err = fib6_del(rt, info);
1649 write_unlock_bh(&table->tb6_lock);
1656 int ip6_del_rt(struct rt6_info *rt)
1658 struct nl_info info = {
1659 .nl_net = dev_net(rt->dst.dev),
1661 return __ip6_del_rt(rt, &info);
1664 static int ip6_route_del(struct fib6_config *cfg)
1666 struct fib6_table *table;
1667 struct fib6_node *fn;
1668 struct rt6_info *rt;
1671 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1675 read_lock_bh(&table->tb6_lock);
1677 fn = fib6_locate(&table->tb6_root,
1678 &cfg->fc_dst, cfg->fc_dst_len,
1679 &cfg->fc_src, cfg->fc_src_len);
1682 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1683 if (cfg->fc_ifindex &&
1685 rt->dst.dev->ifindex != cfg->fc_ifindex))
1687 if (cfg->fc_flags & RTF_GATEWAY &&
1688 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1690 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1693 read_unlock_bh(&table->tb6_lock);
1695 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1698 read_unlock_bh(&table->tb6_lock);
1703 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1705 struct net *net = dev_net(skb->dev);
1706 struct netevent_redirect netevent;
1707 struct rt6_info *rt, *nrt = NULL;
1708 struct ndisc_options ndopts;
1709 struct neighbour *old_neigh;
1710 struct inet6_dev *in6_dev;
1711 struct neighbour *neigh;
1713 int optlen, on_link;
1716 optlen = skb->tail - skb->transport_header;
1717 optlen -= sizeof(*msg);
1720 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1724 msg = (struct rd_msg *)icmp6_hdr(skb);
1726 if (ipv6_addr_is_multicast(&msg->dest)) {
1727 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1732 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1734 } else if (ipv6_addr_type(&msg->target) !=
1735 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1736 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1740 in6_dev = __in6_dev_get(skb->dev);
1743 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1747 * The IP source address of the Redirect MUST be the same as the current
1748 * first-hop router for the specified ICMP Destination Address.
1751 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1752 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1757 if (ndopts.nd_opts_tgt_lladdr) {
1758 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1761 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1766 rt = (struct rt6_info *) dst;
1767 if (rt == net->ipv6.ip6_null_entry) {
1768 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1772 /* Redirect received -> path was valid.
1773 * Look, redirects are sent only in response to data packets,
1774 * so that this nexthop apparently is reachable. --ANK
1776 dst_confirm(&rt->dst);
1778 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1782 /* Duplicate redirect: silently ignore. */
1784 if (neigh == old_neigh)
1788 * We have finally decided to accept it.
1791 neigh_update(neigh, lladdr, NUD_STALE,
1792 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1793 NEIGH_UPDATE_F_OVERRIDE|
1794 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1795 NEIGH_UPDATE_F_ISROUTER))
1798 nrt = ip6_rt_copy(rt, &msg->dest);
1802 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1804 nrt->rt6i_flags &= ~RTF_GATEWAY;
1806 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1807 nrt->n = neigh_clone(neigh);
1809 if (ip6_ins_rt(nrt))
1812 netevent.old = &rt->dst;
1813 netevent.old_neigh = old_neigh;
1814 netevent.new = &nrt->dst;
1815 netevent.new_neigh = neigh;
1816 netevent.daddr = &msg->dest;
1817 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1819 if (rt->rt6i_flags & RTF_CACHE) {
1820 rt = (struct rt6_info *) dst_clone(&rt->dst);
1825 neigh_release(neigh);
1829 * Misc support functions
1832 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1833 const struct in6_addr *dest)
1835 struct net *net = dev_net(ort->dst.dev);
1836 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1840 rt->dst.input = ort->dst.input;
1841 rt->dst.output = ort->dst.output;
1842 rt->dst.flags |= DST_HOST;
1844 rt->rt6i_dst.addr = *dest;
1845 rt->rt6i_dst.plen = 128;
1846 dst_copy_metrics(&rt->dst, &ort->dst);
1847 rt->dst.error = ort->dst.error;
1848 rt->rt6i_idev = ort->rt6i_idev;
1850 in6_dev_hold(rt->rt6i_idev);
1851 rt->dst.lastuse = jiffies;
1853 rt->rt6i_gateway = ort->rt6i_gateway;
1854 rt->rt6i_flags = ort->rt6i_flags;
1855 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1856 (RTF_DEFAULT | RTF_ADDRCONF))
1857 rt6_set_from(rt, ort);
1859 rt6_clean_expires(rt);
1860 rt->rt6i_metric = 0;
1862 #ifdef CONFIG_IPV6_SUBTREES
1863 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1865 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1866 rt->rt6i_table = ort->rt6i_table;
1871 #ifdef CONFIG_IPV6_ROUTE_INFO
1872 static struct rt6_info *rt6_get_route_info(struct net *net,
1873 const struct in6_addr *prefix, int prefixlen,
1874 const struct in6_addr *gwaddr, int ifindex)
1876 struct fib6_node *fn;
1877 struct rt6_info *rt = NULL;
1878 struct fib6_table *table;
1880 table = fib6_get_table(net, RT6_TABLE_INFO);
1884 read_lock_bh(&table->tb6_lock);
1885 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1889 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1890 if (rt->dst.dev->ifindex != ifindex)
1892 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1894 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1900 read_unlock_bh(&table->tb6_lock);
1904 static struct rt6_info *rt6_add_route_info(struct net *net,
1905 const struct in6_addr *prefix, int prefixlen,
1906 const struct in6_addr *gwaddr, int ifindex,
1909 struct fib6_config cfg = {
1910 .fc_table = RT6_TABLE_INFO,
1911 .fc_metric = IP6_RT_PRIO_USER,
1912 .fc_ifindex = ifindex,
1913 .fc_dst_len = prefixlen,
1914 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1915 RTF_UP | RTF_PREF(pref),
1916 .fc_nlinfo.portid = 0,
1917 .fc_nlinfo.nlh = NULL,
1918 .fc_nlinfo.nl_net = net,
1921 cfg.fc_dst = *prefix;
1922 cfg.fc_gateway = *gwaddr;
1924 /* We should treat it as a default route if prefix length is 0. */
1926 cfg.fc_flags |= RTF_DEFAULT;
1928 ip6_route_add(&cfg);
1930 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1934 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1936 struct rt6_info *rt;
1937 struct fib6_table *table;
1939 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1943 read_lock_bh(&table->tb6_lock);
1944 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1945 if (dev == rt->dst.dev &&
1946 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1947 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1952 read_unlock_bh(&table->tb6_lock);
1956 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1957 struct net_device *dev,
1960 struct fib6_config cfg = {
1961 .fc_table = RT6_TABLE_DFLT,
1962 .fc_metric = IP6_RT_PRIO_USER,
1963 .fc_ifindex = dev->ifindex,
1964 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1965 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1966 .fc_nlinfo.portid = 0,
1967 .fc_nlinfo.nlh = NULL,
1968 .fc_nlinfo.nl_net = dev_net(dev),
1971 cfg.fc_gateway = *gwaddr;
1973 ip6_route_add(&cfg);
1975 return rt6_get_dflt_router(gwaddr, dev);
1978 void rt6_purge_dflt_routers(struct net *net)
1980 struct rt6_info *rt;
1981 struct fib6_table *table;
1983 /* NOTE: Keep consistent with rt6_get_dflt_router */
1984 table = fib6_get_table(net, RT6_TABLE_DFLT);
1989 read_lock_bh(&table->tb6_lock);
1990 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1991 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1993 read_unlock_bh(&table->tb6_lock);
1998 read_unlock_bh(&table->tb6_lock);
2001 static void rtmsg_to_fib6_config(struct net *net,
2002 struct in6_rtmsg *rtmsg,
2003 struct fib6_config *cfg)
2005 memset(cfg, 0, sizeof(*cfg));
2007 cfg->fc_table = RT6_TABLE_MAIN;
2008 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2009 cfg->fc_metric = rtmsg->rtmsg_metric;
2010 cfg->fc_expires = rtmsg->rtmsg_info;
2011 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2012 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2013 cfg->fc_flags = rtmsg->rtmsg_flags;
2015 cfg->fc_nlinfo.nl_net = net;
2017 cfg->fc_dst = rtmsg->rtmsg_dst;
2018 cfg->fc_src = rtmsg->rtmsg_src;
2019 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2022 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2024 struct fib6_config cfg;
2025 struct in6_rtmsg rtmsg;
2029 case SIOCADDRT: /* Add a route */
2030 case SIOCDELRT: /* Delete a route */
2031 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2033 err = copy_from_user(&rtmsg, arg,
2034 sizeof(struct in6_rtmsg));
2038 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2043 err = ip6_route_add(&cfg);
2046 err = ip6_route_del(&cfg);
2060 * Drop the packet on the floor
2063 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2066 struct dst_entry *dst = skb_dst(skb);
2067 switch (ipstats_mib_noroutes) {
2068 case IPSTATS_MIB_INNOROUTES:
2069 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2070 if (type == IPV6_ADDR_ANY) {
2071 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2072 IPSTATS_MIB_INADDRERRORS);
2076 case IPSTATS_MIB_OUTNOROUTES:
2077 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2078 ipstats_mib_noroutes);
2081 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2086 static int ip6_pkt_discard(struct sk_buff *skb)
2088 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2091 static int ip6_pkt_discard_out(struct sk_buff *skb)
2093 skb->dev = skb_dst(skb)->dev;
2094 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2097 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2099 static int ip6_pkt_prohibit(struct sk_buff *skb)
2101 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2104 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2106 skb->dev = skb_dst(skb)->dev;
2107 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2113 * Allocate a dst for local (unicast / anycast) address.
2116 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2117 const struct in6_addr *addr,
2120 struct net *net = dev_net(idev->dev);
2121 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2125 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2126 return ERR_PTR(-ENOMEM);
2131 rt->dst.flags |= DST_HOST;
2132 rt->dst.input = ip6_input;
2133 rt->dst.output = ip6_output;
2134 rt->rt6i_idev = idev;
2136 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2138 rt->rt6i_flags |= RTF_ANYCAST;
2140 rt->rt6i_flags |= RTF_LOCAL;
2141 err = rt6_bind_neighbour(rt, rt->dst.dev);
2144 return ERR_PTR(err);
2147 rt->rt6i_dst.addr = *addr;
2148 rt->rt6i_dst.plen = 128;
2149 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2151 atomic_set(&rt->dst.__refcnt, 1);
2156 int ip6_route_get_saddr(struct net *net,
2157 struct rt6_info *rt,
2158 const struct in6_addr *daddr,
2160 struct in6_addr *saddr)
2162 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2164 if (rt->rt6i_prefsrc.plen)
2165 *saddr = rt->rt6i_prefsrc.addr;
2167 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2168 daddr, prefs, saddr);
2172 /* remove deleted ip from prefsrc entries */
2173 struct arg_dev_net_ip {
2174 struct net_device *dev;
2176 struct in6_addr *addr;
2179 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2181 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2182 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2183 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2185 if (((void *)rt->dst.dev == dev || !dev) &&
2186 rt != net->ipv6.ip6_null_entry &&
2187 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2188 /* remove prefsrc entry */
2189 rt->rt6i_prefsrc.plen = 0;
2194 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2196 struct net *net = dev_net(ifp->idev->dev);
2197 struct arg_dev_net_ip adni = {
2198 .dev = ifp->idev->dev,
2202 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2205 struct arg_dev_net {
2206 struct net_device *dev;
2210 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2212 const struct arg_dev_net *adn = arg;
2213 const struct net_device *dev = adn->dev;
2215 if ((rt->dst.dev == dev || !dev) &&
2216 rt != adn->net->ipv6.ip6_null_entry)
2222 void rt6_ifdown(struct net *net, struct net_device *dev)
2224 struct arg_dev_net adn = {
2229 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2230 icmp6_clean_all(fib6_ifdown, &adn);
2233 struct rt6_mtu_change_arg {
2234 struct net_device *dev;
2238 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2240 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2241 struct inet6_dev *idev;
2243 /* In IPv6 pmtu discovery is not optional,
2244 so that RTAX_MTU lock cannot disable it.
2245 We still use this lock to block changes
2246 caused by addrconf/ndisc.
2249 idev = __in6_dev_get(arg->dev);
2253 /* For administrative MTU increase, there is no way to discover
2254 IPv6 PMTU increase, so PMTU increase should be updated here.
2255 Since RFC 1981 doesn't include administrative MTU increase
2256 update PMTU increase is a MUST. (i.e. jumbo frame)
2259 If new MTU is less than route PMTU, this new MTU will be the
2260 lowest MTU in the path, update the route PMTU to reflect PMTU
2261 decreases; if new MTU is greater than route PMTU, and the
2262 old MTU is the lowest MTU in the path, update the route PMTU
2263 to reflect the increase. In this case if the other nodes' MTU
2264 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2267 if (rt->dst.dev == arg->dev &&
2268 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2269 (dst_mtu(&rt->dst) >= arg->mtu ||
2270 (dst_mtu(&rt->dst) < arg->mtu &&
2271 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2272 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2277 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2279 struct rt6_mtu_change_arg arg = {
2284 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2287 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2288 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2289 [RTA_OIF] = { .type = NLA_U32 },
2290 [RTA_IIF] = { .type = NLA_U32 },
2291 [RTA_PRIORITY] = { .type = NLA_U32 },
2292 [RTA_METRICS] = { .type = NLA_NESTED },
2293 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2296 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2297 struct fib6_config *cfg)
2300 struct nlattr *tb[RTA_MAX+1];
2303 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2308 rtm = nlmsg_data(nlh);
2309 memset(cfg, 0, sizeof(*cfg));
2311 cfg->fc_table = rtm->rtm_table;
2312 cfg->fc_dst_len = rtm->rtm_dst_len;
2313 cfg->fc_src_len = rtm->rtm_src_len;
2314 cfg->fc_flags = RTF_UP;
2315 cfg->fc_protocol = rtm->rtm_protocol;
2316 cfg->fc_type = rtm->rtm_type;
2318 if (rtm->rtm_type == RTN_UNREACHABLE ||
2319 rtm->rtm_type == RTN_BLACKHOLE ||
2320 rtm->rtm_type == RTN_PROHIBIT ||
2321 rtm->rtm_type == RTN_THROW)
2322 cfg->fc_flags |= RTF_REJECT;
2324 if (rtm->rtm_type == RTN_LOCAL)
2325 cfg->fc_flags |= RTF_LOCAL;
2327 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2328 cfg->fc_nlinfo.nlh = nlh;
2329 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2331 if (tb[RTA_GATEWAY]) {
2332 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2333 cfg->fc_flags |= RTF_GATEWAY;
2337 int plen = (rtm->rtm_dst_len + 7) >> 3;
2339 if (nla_len(tb[RTA_DST]) < plen)
2342 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2346 int plen = (rtm->rtm_src_len + 7) >> 3;
2348 if (nla_len(tb[RTA_SRC]) < plen)
2351 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2354 if (tb[RTA_PREFSRC])
2355 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2358 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2360 if (tb[RTA_PRIORITY])
2361 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2363 if (tb[RTA_METRICS]) {
2364 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2365 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2369 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2371 if (tb[RTA_MULTIPATH]) {
2372 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2373 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2381 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2383 struct fib6_config r_cfg;
2384 struct rtnexthop *rtnh;
2387 int err = 0, last_err = 0;
2390 rtnh = (struct rtnexthop *)cfg->fc_mp;
2391 remaining = cfg->fc_mp_len;
2393 /* Parse a Multipath Entry */
2394 while (rtnh_ok(rtnh, remaining)) {
2395 memcpy(&r_cfg, cfg, sizeof(*cfg));
2396 if (rtnh->rtnh_ifindex)
2397 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2399 attrlen = rtnh_attrlen(rtnh);
2401 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2403 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2405 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2406 r_cfg.fc_flags |= RTF_GATEWAY;
2409 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2412 /* If we are trying to remove a route, do not stop the
2413 * loop when ip6_route_del() fails (because next hop is
2414 * already gone), we should try to remove all next hops.
2417 /* If add fails, we should try to delete all
2418 * next hops that have been already added.
2424 /* Because each route is added like a single route we remove
2425 * this flag after the first nexthop (if there is a collision,
2426 * we have already fail to add the first nexthop:
2427 * fib6_add_rt2node() has reject it).
2429 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2430 rtnh = rtnh_next(rtnh, &remaining);
2436 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2438 struct fib6_config cfg;
2441 err = rtm_to_fib6_config(skb, nlh, &cfg);
2446 return ip6_route_multipath(&cfg, 0);
2448 return ip6_route_del(&cfg);
2451 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2453 struct fib6_config cfg;
2456 err = rtm_to_fib6_config(skb, nlh, &cfg);
2461 return ip6_route_multipath(&cfg, 1);
2463 return ip6_route_add(&cfg);
2466 static inline size_t rt6_nlmsg_size(void)
2468 return NLMSG_ALIGN(sizeof(struct rtmsg))
2469 + nla_total_size(16) /* RTA_SRC */
2470 + nla_total_size(16) /* RTA_DST */
2471 + nla_total_size(16) /* RTA_GATEWAY */
2472 + nla_total_size(16) /* RTA_PREFSRC */
2473 + nla_total_size(4) /* RTA_TABLE */
2474 + nla_total_size(4) /* RTA_IIF */
2475 + nla_total_size(4) /* RTA_OIF */
2476 + nla_total_size(4) /* RTA_PRIORITY */
2477 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2478 + nla_total_size(sizeof(struct rta_cacheinfo));
2481 static int rt6_fill_node(struct net *net,
2482 struct sk_buff *skb, struct rt6_info *rt,
2483 struct in6_addr *dst, struct in6_addr *src,
2484 int iif, int type, u32 portid, u32 seq,
2485 int prefix, int nowait, unsigned int flags)
2488 struct nlmsghdr *nlh;
2491 struct neighbour *n;
2493 if (prefix) { /* user wants prefix routes only */
2494 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2495 /* success since this is not a prefix route */
2500 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2504 rtm = nlmsg_data(nlh);
2505 rtm->rtm_family = AF_INET6;
2506 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2507 rtm->rtm_src_len = rt->rt6i_src.plen;
2510 table = rt->rt6i_table->tb6_id;
2512 table = RT6_TABLE_UNSPEC;
2513 rtm->rtm_table = table;
2514 if (nla_put_u32(skb, RTA_TABLE, table))
2515 goto nla_put_failure;
2516 if (rt->rt6i_flags & RTF_REJECT) {
2517 switch (rt->dst.error) {
2519 rtm->rtm_type = RTN_BLACKHOLE;
2522 rtm->rtm_type = RTN_PROHIBIT;
2525 rtm->rtm_type = RTN_THROW;
2528 rtm->rtm_type = RTN_UNREACHABLE;
2532 else if (rt->rt6i_flags & RTF_LOCAL)
2533 rtm->rtm_type = RTN_LOCAL;
2534 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2535 rtm->rtm_type = RTN_LOCAL;
2537 rtm->rtm_type = RTN_UNICAST;
2539 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2540 rtm->rtm_protocol = rt->rt6i_protocol;
2541 if (rt->rt6i_flags & RTF_DYNAMIC)
2542 rtm->rtm_protocol = RTPROT_REDIRECT;
2543 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2544 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2545 rtm->rtm_protocol = RTPROT_RA;
2547 rtm->rtm_protocol = RTPROT_KERNEL;
2550 if (rt->rt6i_flags & RTF_CACHE)
2551 rtm->rtm_flags |= RTM_F_CLONED;
2554 if (nla_put(skb, RTA_DST, 16, dst))
2555 goto nla_put_failure;
2556 rtm->rtm_dst_len = 128;
2557 } else if (rtm->rtm_dst_len)
2558 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2559 goto nla_put_failure;
2560 #ifdef CONFIG_IPV6_SUBTREES
2562 if (nla_put(skb, RTA_SRC, 16, src))
2563 goto nla_put_failure;
2564 rtm->rtm_src_len = 128;
2565 } else if (rtm->rtm_src_len &&
2566 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2567 goto nla_put_failure;
2570 #ifdef CONFIG_IPV6_MROUTE
2571 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2572 int err = ip6mr_get_route(net, skb, rtm, nowait);
2577 goto nla_put_failure;
2579 if (err == -EMSGSIZE)
2580 goto nla_put_failure;
2585 if (nla_put_u32(skb, RTA_IIF, iif))
2586 goto nla_put_failure;
2588 struct in6_addr saddr_buf;
2589 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2590 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2591 goto nla_put_failure;
2594 if (rt->rt6i_prefsrc.plen) {
2595 struct in6_addr saddr_buf;
2596 saddr_buf = rt->rt6i_prefsrc.addr;
2597 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2598 goto nla_put_failure;
2601 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2602 goto nla_put_failure;
2606 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0)
2607 goto nla_put_failure;
2611 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2612 goto nla_put_failure;
2613 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2614 goto nla_put_failure;
2616 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2618 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2619 goto nla_put_failure;
2621 return nlmsg_end(skb, nlh);
2624 nlmsg_cancel(skb, nlh);
2628 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2630 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2633 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2634 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2635 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2639 return rt6_fill_node(arg->net,
2640 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2641 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2642 prefix, 0, NLM_F_MULTI);
2645 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2647 struct net *net = sock_net(in_skb->sk);
2648 struct nlattr *tb[RTA_MAX+1];
2649 struct rt6_info *rt;
2650 struct sk_buff *skb;
2653 int err, iif = 0, oif = 0;
2655 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2660 memset(&fl6, 0, sizeof(fl6));
2663 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2666 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2670 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2673 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2677 iif = nla_get_u32(tb[RTA_IIF]);
2680 oif = nla_get_u32(tb[RTA_OIF]);
2683 struct net_device *dev;
2686 dev = __dev_get_by_index(net, iif);
2692 fl6.flowi6_iif = iif;
2694 if (!ipv6_addr_any(&fl6.saddr))
2695 flags |= RT6_LOOKUP_F_HAS_SADDR;
2697 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2700 fl6.flowi6_oif = oif;
2702 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2705 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2712 /* Reserve room for dummy headers, this skb can pass
2713 through good chunk of routing engine.
2715 skb_reset_mac_header(skb);
2716 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2718 skb_dst_set(skb, &rt->dst);
2720 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2721 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2722 nlh->nlmsg_seq, 0, 0, 0);
2728 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2733 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2735 struct sk_buff *skb;
2736 struct net *net = info->nl_net;
2741 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2743 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2747 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2748 event, info->portid, seq, 0, 0, 0);
2750 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2751 WARN_ON(err == -EMSGSIZE);
2755 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2756 info->nlh, gfp_any());
2760 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2763 static int ip6_route_dev_notify(struct notifier_block *this,
2764 unsigned long event, void *data)
2766 struct net_device *dev = (struct net_device *)data;
2767 struct net *net = dev_net(dev);
2769 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2770 net->ipv6.ip6_null_entry->dst.dev = dev;
2771 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2772 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2773 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2774 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2775 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2776 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2787 #ifdef CONFIG_PROC_FS
2798 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2800 struct seq_file *m = p_arg;
2801 struct neighbour *n;
2803 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2805 #ifdef CONFIG_IPV6_SUBTREES
2806 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2808 seq_puts(m, "00000000000000000000000000000000 00 ");
2812 seq_printf(m, "%pi6", n->primary_key);
2814 seq_puts(m, "00000000000000000000000000000000");
2816 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2817 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2818 rt->dst.__use, rt->rt6i_flags,
2819 rt->dst.dev ? rt->dst.dev->name : "");
2823 static int ipv6_route_show(struct seq_file *m, void *v)
2825 struct net *net = (struct net *)m->private;
2826 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2830 static int ipv6_route_open(struct inode *inode, struct file *file)
2832 return single_open_net(inode, file, ipv6_route_show);
2835 static const struct file_operations ipv6_route_proc_fops = {
2836 .owner = THIS_MODULE,
2837 .open = ipv6_route_open,
2839 .llseek = seq_lseek,
2840 .release = single_release_net,
2843 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2845 struct net *net = (struct net *)seq->private;
2846 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2847 net->ipv6.rt6_stats->fib_nodes,
2848 net->ipv6.rt6_stats->fib_route_nodes,
2849 net->ipv6.rt6_stats->fib_rt_alloc,
2850 net->ipv6.rt6_stats->fib_rt_entries,
2851 net->ipv6.rt6_stats->fib_rt_cache,
2852 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2853 net->ipv6.rt6_stats->fib_discarded_routes);
2858 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2860 return single_open_net(inode, file, rt6_stats_seq_show);
2863 static const struct file_operations rt6_stats_seq_fops = {
2864 .owner = THIS_MODULE,
2865 .open = rt6_stats_seq_open,
2867 .llseek = seq_lseek,
2868 .release = single_release_net,
2870 #endif /* CONFIG_PROC_FS */
2872 #ifdef CONFIG_SYSCTL
2875 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2876 void __user *buffer, size_t *lenp, loff_t *ppos)
2883 net = (struct net *)ctl->extra1;
2884 delay = net->ipv6.sysctl.flush_delay;
2885 proc_dointvec(ctl, write, buffer, lenp, ppos);
2886 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2890 ctl_table ipv6_route_table_template[] = {
2892 .procname = "flush",
2893 .data = &init_net.ipv6.sysctl.flush_delay,
2894 .maxlen = sizeof(int),
2896 .proc_handler = ipv6_sysctl_rtcache_flush
2899 .procname = "gc_thresh",
2900 .data = &ip6_dst_ops_template.gc_thresh,
2901 .maxlen = sizeof(int),
2903 .proc_handler = proc_dointvec,
2906 .procname = "max_size",
2907 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2908 .maxlen = sizeof(int),
2910 .proc_handler = proc_dointvec,
2913 .procname = "gc_min_interval",
2914 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2915 .maxlen = sizeof(int),
2917 .proc_handler = proc_dointvec_jiffies,
2920 .procname = "gc_timeout",
2921 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2922 .maxlen = sizeof(int),
2924 .proc_handler = proc_dointvec_jiffies,
2927 .procname = "gc_interval",
2928 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2929 .maxlen = sizeof(int),
2931 .proc_handler = proc_dointvec_jiffies,
2934 .procname = "gc_elasticity",
2935 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2936 .maxlen = sizeof(int),
2938 .proc_handler = proc_dointvec,
2941 .procname = "mtu_expires",
2942 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2943 .maxlen = sizeof(int),
2945 .proc_handler = proc_dointvec_jiffies,
2948 .procname = "min_adv_mss",
2949 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2950 .maxlen = sizeof(int),
2952 .proc_handler = proc_dointvec,
2955 .procname = "gc_min_interval_ms",
2956 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2957 .maxlen = sizeof(int),
2959 .proc_handler = proc_dointvec_ms_jiffies,
2964 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2966 struct ctl_table *table;
2968 table = kmemdup(ipv6_route_table_template,
2969 sizeof(ipv6_route_table_template),
2973 table[0].data = &net->ipv6.sysctl.flush_delay;
2974 table[0].extra1 = net;
2975 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2976 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2977 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2978 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2979 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2980 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2981 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2982 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2983 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2985 /* Don't export sysctls to unprivileged users */
2986 if (net->user_ns != &init_user_ns)
2987 table[0].procname = NULL;
2994 static int __net_init ip6_route_net_init(struct net *net)
2998 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2999 sizeof(net->ipv6.ip6_dst_ops));
3001 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3002 goto out_ip6_dst_ops;
3004 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3005 sizeof(*net->ipv6.ip6_null_entry),
3007 if (!net->ipv6.ip6_null_entry)
3008 goto out_ip6_dst_entries;
3009 net->ipv6.ip6_null_entry->dst.path =
3010 (struct dst_entry *)net->ipv6.ip6_null_entry;
3011 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3012 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3013 ip6_template_metrics, true);
3015 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3016 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3017 sizeof(*net->ipv6.ip6_prohibit_entry),
3019 if (!net->ipv6.ip6_prohibit_entry)
3020 goto out_ip6_null_entry;
3021 net->ipv6.ip6_prohibit_entry->dst.path =
3022 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3023 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3024 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3025 ip6_template_metrics, true);
3027 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3028 sizeof(*net->ipv6.ip6_blk_hole_entry),
3030 if (!net->ipv6.ip6_blk_hole_entry)
3031 goto out_ip6_prohibit_entry;
3032 net->ipv6.ip6_blk_hole_entry->dst.path =
3033 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3034 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3035 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3036 ip6_template_metrics, true);
3039 net->ipv6.sysctl.flush_delay = 0;
3040 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3041 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3042 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3043 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3044 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3045 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3046 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3048 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3054 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3055 out_ip6_prohibit_entry:
3056 kfree(net->ipv6.ip6_prohibit_entry);
3058 kfree(net->ipv6.ip6_null_entry);
3060 out_ip6_dst_entries:
3061 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3066 static void __net_exit ip6_route_net_exit(struct net *net)
3068 kfree(net->ipv6.ip6_null_entry);
3069 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3070 kfree(net->ipv6.ip6_prohibit_entry);
3071 kfree(net->ipv6.ip6_blk_hole_entry);
3073 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3076 static int __net_init ip6_route_net_init_late(struct net *net)
3078 #ifdef CONFIG_PROC_FS
3079 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
3080 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
3085 static void __net_exit ip6_route_net_exit_late(struct net *net)
3087 #ifdef CONFIG_PROC_FS
3088 proc_net_remove(net, "ipv6_route");
3089 proc_net_remove(net, "rt6_stats");
3093 static struct pernet_operations ip6_route_net_ops = {
3094 .init = ip6_route_net_init,
3095 .exit = ip6_route_net_exit,
3098 static int __net_init ipv6_inetpeer_init(struct net *net)
3100 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3104 inet_peer_base_init(bp);
3105 net->ipv6.peers = bp;
3109 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3111 struct inet_peer_base *bp = net->ipv6.peers;
3113 net->ipv6.peers = NULL;
3114 inetpeer_invalidate_tree(bp);
3118 static struct pernet_operations ipv6_inetpeer_ops = {
3119 .init = ipv6_inetpeer_init,
3120 .exit = ipv6_inetpeer_exit,
3123 static struct pernet_operations ip6_route_net_late_ops = {
3124 .init = ip6_route_net_init_late,
3125 .exit = ip6_route_net_exit_late,
3128 static struct notifier_block ip6_route_dev_notifier = {
3129 .notifier_call = ip6_route_dev_notify,
3133 int __init ip6_route_init(void)
3138 ip6_dst_ops_template.kmem_cachep =
3139 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3140 SLAB_HWCACHE_ALIGN, NULL);
3141 if (!ip6_dst_ops_template.kmem_cachep)
3144 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3146 goto out_kmem_cache;
3148 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3150 goto out_dst_entries;
3152 ret = register_pernet_subsys(&ip6_route_net_ops);
3154 goto out_register_inetpeer;
3156 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3158 /* Registering of the loopback is done before this portion of code,
3159 * the loopback reference in rt6_info will not be taken, do it
3160 * manually for init_net */
3161 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3162 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3163 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3164 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3165 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3166 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3167 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3171 goto out_register_subsys;
3177 ret = fib6_rules_init();
3181 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3183 goto fib6_rules_init;
3186 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3187 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3188 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3189 goto out_register_late_subsys;
3191 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3193 goto out_register_late_subsys;
3198 out_register_late_subsys:
3199 unregister_pernet_subsys(&ip6_route_net_late_ops);
3201 fib6_rules_cleanup();
3206 out_register_subsys:
3207 unregister_pernet_subsys(&ip6_route_net_ops);
3208 out_register_inetpeer:
3209 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3211 dst_entries_destroy(&ip6_dst_blackhole_ops);
3213 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3217 void ip6_route_cleanup(void)
3219 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3220 unregister_pernet_subsys(&ip6_route_net_late_ops);
3221 fib6_rules_cleanup();
3224 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3225 unregister_pernet_subsys(&ip6_route_net_ops);
3226 dst_entries_destroy(&ip6_dst_blackhole_ops);
3227 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);