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>
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
89 struct sk_buff *skb, u32 mtu);
90 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
94 #ifdef CONFIG_IPV6_ROUTE_INFO
95 static struct rt6_info *rt6_add_route_info(struct net *net,
96 const struct in6_addr *prefix, int prefixlen,
97 const struct in6_addr *gwaddr, int ifindex,
99 static struct rt6_info *rt6_get_route_info(struct net *net,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr, int ifindex);
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
110 if (!(rt->dst.flags & DST_HOST))
113 peer = rt6_get_peer_create(rt);
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
138 struct in6_addr *p = &rt->rt6i_gateway;
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
143 return &ipv6_hdr(skb)->daddr;
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
151 struct rt6_info *rt = (struct rt6_info *) dst;
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
158 return neigh_create(&nd_tbl, daddr, dst->dev);
161 static struct dst_ops ip6_dst_ops_template = {
163 .protocol = cpu_to_be16(ETH_P_IPV6),
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
184 return mtu ? : dst->dev->mtu;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
203 static struct dst_ops ip6_dst_blackhole_ops = {
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
220 static const struct rt6_info ip6_null_entry_template = {
222 .__refcnt = ATOMIC_INIT(1),
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static int ip6_pkt_prohibit(struct sk_buff *skb);
238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
240 static const struct rt6_info ip6_prohibit_entry_template = {
242 .__refcnt = ATOMIC_INIT(1),
244 .obsolete = DST_OBSOLETE_FORCE_CHK,
246 .input = ip6_pkt_prohibit,
247 .output = ip6_pkt_prohibit_out,
249 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
250 .rt6i_protocol = RTPROT_KERNEL,
251 .rt6i_metric = ~(u32) 0,
252 .rt6i_ref = ATOMIC_INIT(1),
255 static const struct rt6_info ip6_blk_hole_entry_template = {
257 .__refcnt = ATOMIC_INIT(1),
259 .obsolete = DST_OBSOLETE_FORCE_CHK,
261 .input = dst_discard,
262 .output = dst_discard,
264 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
265 .rt6i_protocol = RTPROT_KERNEL,
266 .rt6i_metric = ~(u32) 0,
267 .rt6i_ref = ATOMIC_INIT(1),
272 /* allocate dst with ip6_dst_ops */
273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
274 struct net_device *dev,
276 struct fib6_table *table)
278 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
279 0, DST_OBSOLETE_FORCE_CHK, flags);
282 struct dst_entry *dst = &rt->dst;
284 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
285 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
286 rt->rt6i_genid = rt_genid_ipv6(net);
287 INIT_LIST_HEAD(&rt->rt6i_siblings);
292 static void ip6_dst_destroy(struct dst_entry *dst)
294 struct rt6_info *rt = (struct rt6_info *)dst;
295 struct inet6_dev *idev = rt->rt6i_idev;
296 struct dst_entry *from = dst->from;
298 if (!(rt->dst.flags & DST_HOST))
299 dst_destroy_metrics_generic(dst);
302 rt->rt6i_idev = NULL;
309 if (rt6_has_peer(rt)) {
310 struct inet_peer *peer = rt6_peer_ptr(rt);
315 void rt6_bind_peer(struct rt6_info *rt, int create)
317 struct inet_peer_base *base;
318 struct inet_peer *peer;
320 base = inetpeer_base_ptr(rt->_rt6i_peer);
324 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
326 if (!rt6_set_peer(rt, peer))
331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
334 struct rt6_info *rt = (struct rt6_info *)dst;
335 struct inet6_dev *idev = rt->rt6i_idev;
336 struct net_device *loopback_dev =
337 dev_net(dev)->loopback_dev;
339 if (dev != loopback_dev) {
340 if (idev && idev->dev == dev) {
341 struct inet6_dev *loopback_idev =
342 in6_dev_get(loopback_dev);
344 rt->rt6i_idev = loopback_idev;
351 static bool rt6_check_expired(const struct rt6_info *rt)
353 if (rt->rt6i_flags & RTF_EXPIRES) {
354 if (time_after(jiffies, rt->dst.expires))
356 } else if (rt->dst.from) {
357 return rt6_check_expired((struct rt6_info *) rt->dst.from);
362 static bool rt6_need_strict(const struct in6_addr *daddr)
364 return ipv6_addr_type(daddr) &
365 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
368 /* Multipath route selection:
369 * Hash based function using packet header and flowlabel.
370 * Adapted from fib_info_hashfn()
372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
373 const struct flowi6 *fl6)
375 unsigned int val = fl6->flowi6_proto;
377 val ^= ipv6_addr_hash(&fl6->daddr);
378 val ^= ipv6_addr_hash(&fl6->saddr);
380 /* Work only if this not encapsulated */
381 switch (fl6->flowi6_proto) {
385 val ^= (__force u16)fl6->fl6_sport;
386 val ^= (__force u16)fl6->fl6_dport;
390 val ^= (__force u16)fl6->fl6_icmp_type;
391 val ^= (__force u16)fl6->fl6_icmp_code;
394 /* RFC6438 recommands to use flowlabel */
395 val ^= (__force u32)fl6->flowlabel;
397 /* Perhaps, we need to tune, this function? */
398 val = val ^ (val >> 7) ^ (val >> 12);
399 return val % candidate_count;
402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct flowi6 *fl6, int oif,
406 struct rt6_info *sibling, *next_sibling;
409 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
410 /* Don't change the route, if route_choosen == 0
411 * (siblings does not include ourself)
414 list_for_each_entry_safe(sibling, next_sibling,
415 &match->rt6i_siblings, rt6i_siblings) {
417 if (route_choosen == 0) {
418 if (rt6_score_route(sibling, oif, strict) < 0)
428 * Route lookup. Any table->tb6_lock is implied.
431 static inline struct rt6_info *rt6_device_match(struct net *net,
433 const struct in6_addr *saddr,
437 struct rt6_info *local = NULL;
438 struct rt6_info *sprt;
440 if (!oif && ipv6_addr_any(saddr))
443 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
444 struct net_device *dev = sprt->dst.dev;
447 if (dev->ifindex == oif)
449 if (dev->flags & IFF_LOOPBACK) {
450 if (!sprt->rt6i_idev ||
451 sprt->rt6i_idev->dev->ifindex != oif) {
452 if (flags & RT6_LOOKUP_F_IFACE && oif)
454 if (local && (!oif ||
455 local->rt6i_idev->dev->ifindex == oif))
461 if (ipv6_chk_addr(net, saddr, dev,
462 flags & RT6_LOOKUP_F_IFACE))
471 if (flags & RT6_LOOKUP_F_IFACE)
472 return net->ipv6.ip6_null_entry;
478 #ifdef CONFIG_IPV6_ROUTER_PREF
479 struct __rt6_probe_work {
480 struct work_struct work;
481 struct in6_addr target;
482 struct net_device *dev;
485 static void rt6_probe_deferred(struct work_struct *w)
487 struct in6_addr mcaddr;
488 struct __rt6_probe_work *work =
489 container_of(w, struct __rt6_probe_work, work);
491 addrconf_addr_solict_mult(&work->target, &mcaddr);
492 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
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 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
511 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
513 write_lock(&neigh->lock);
514 if (neigh->nud_state & NUD_VALID)
519 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
520 struct __rt6_probe_work *work;
522 work = kmalloc(sizeof(*work), GFP_ATOMIC);
525 neigh->updated = jiffies;
528 write_unlock(&neigh->lock);
531 INIT_WORK(&work->work, rt6_probe_deferred);
532 work->target = rt->rt6i_gateway;
533 dev_hold(rt->dst.dev);
534 work->dev = rt->dst.dev;
535 schedule_work(&work->work);
539 write_unlock(&neigh->lock);
541 rcu_read_unlock_bh();
544 static inline void rt6_probe(struct rt6_info *rt)
550 * Default Router Selection (RFC 2461 6.3.6)
552 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
554 struct net_device *dev = rt->dst.dev;
555 if (!oif || dev->ifindex == oif)
557 if ((dev->flags & IFF_LOOPBACK) &&
558 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
563 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
565 struct neighbour *neigh;
566 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
568 if (rt->rt6i_flags & RTF_NONEXTHOP ||
569 !(rt->rt6i_flags & RTF_GATEWAY))
570 return RT6_NUD_SUCCEED;
573 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
575 read_lock(&neigh->lock);
576 if (neigh->nud_state & NUD_VALID)
577 ret = RT6_NUD_SUCCEED;
578 #ifdef CONFIG_IPV6_ROUTER_PREF
579 else if (!(neigh->nud_state & NUD_FAILED))
580 ret = RT6_NUD_SUCCEED;
582 read_unlock(&neigh->lock);
584 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
585 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
587 rcu_read_unlock_bh();
592 static int rt6_score_route(struct rt6_info *rt, int oif,
597 m = rt6_check_dev(rt, oif);
598 if (!m && (strict & RT6_LOOKUP_F_IFACE))
599 return RT6_NUD_FAIL_HARD;
600 #ifdef CONFIG_IPV6_ROUTER_PREF
601 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
603 if (strict & RT6_LOOKUP_F_REACHABLE) {
604 int n = rt6_check_neigh(rt);
611 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
612 int *mpri, struct rt6_info *match,
616 bool match_do_rr = false;
618 if (rt6_check_expired(rt))
621 m = rt6_score_route(rt, oif, strict);
622 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
624 m = 0; /* lowest valid score */
629 if (strict & RT6_LOOKUP_F_REACHABLE)
633 *do_rr = match_do_rr;
641 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
642 struct rt6_info *rr_head,
643 u32 metric, int oif, int strict,
646 struct rt6_info *rt, *match;
650 for (rt = rr_head; rt && rt->rt6i_metric == metric;
651 rt = rt->dst.rt6_next)
652 match = find_match(rt, oif, strict, &mpri, match, do_rr);
653 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
654 rt = rt->dst.rt6_next)
655 match = find_match(rt, oif, strict, &mpri, match, do_rr);
660 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
662 struct rt6_info *match, *rt0;
668 fn->rr_ptr = rt0 = fn->leaf;
670 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
674 struct rt6_info *next = rt0->dst.rt6_next;
676 /* no entries matched; do round-robin */
677 if (!next || next->rt6i_metric != rt0->rt6i_metric)
684 net = dev_net(rt0->dst.dev);
685 return match ? match : net->ipv6.ip6_null_entry;
688 #ifdef CONFIG_IPV6_ROUTE_INFO
689 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
690 const struct in6_addr *gwaddr)
692 struct net *net = dev_net(dev);
693 struct route_info *rinfo = (struct route_info *) opt;
694 struct in6_addr prefix_buf, *prefix;
696 unsigned long lifetime;
699 if (len < sizeof(struct route_info)) {
703 /* Sanity check for prefix_len and length */
704 if (rinfo->length > 3) {
706 } else if (rinfo->prefix_len > 128) {
708 } else if (rinfo->prefix_len > 64) {
709 if (rinfo->length < 2) {
712 } else if (rinfo->prefix_len > 0) {
713 if (rinfo->length < 1) {
718 pref = rinfo->route_pref;
719 if (pref == ICMPV6_ROUTER_PREF_INVALID)
722 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
724 if (rinfo->length == 3)
725 prefix = (struct in6_addr *)rinfo->prefix;
727 /* this function is safe */
728 ipv6_addr_prefix(&prefix_buf,
729 (struct in6_addr *)rinfo->prefix,
731 prefix = &prefix_buf;
734 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
737 if (rt && !lifetime) {
743 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
746 rt->rt6i_flags = RTF_ROUTEINFO |
747 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
750 if (!addrconf_finite_timeout(lifetime))
751 rt6_clean_expires(rt);
753 rt6_set_expires(rt, jiffies + HZ * lifetime);
761 #define BACKTRACK(__net, saddr) \
763 if (rt == __net->ipv6.ip6_null_entry) { \
764 struct fib6_node *pn; \
766 if (fn->fn_flags & RTN_TL_ROOT) \
769 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
770 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
773 if (fn->fn_flags & RTN_RTINFO) \
779 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
780 struct fib6_table *table,
781 struct flowi6 *fl6, int flags)
783 struct fib6_node *fn;
786 read_lock_bh(&table->tb6_lock);
787 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
790 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
791 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
792 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
793 BACKTRACK(net, &fl6->saddr);
795 dst_use(&rt->dst, jiffies);
796 read_unlock_bh(&table->tb6_lock);
801 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
804 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
806 EXPORT_SYMBOL_GPL(ip6_route_lookup);
808 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
809 const struct in6_addr *saddr, int oif, int strict)
811 struct flowi6 fl6 = {
815 struct dst_entry *dst;
816 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
819 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
820 flags |= RT6_LOOKUP_F_HAS_SADDR;
823 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
825 return (struct rt6_info *) dst;
832 EXPORT_SYMBOL(rt6_lookup);
834 /* ip6_ins_rt is called with FREE table->tb6_lock.
835 It takes new route entry, the addition fails by any reason the
836 route is freed. In any case, if caller does not hold it, it may
840 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
843 struct fib6_table *table;
845 table = rt->rt6i_table;
846 write_lock_bh(&table->tb6_lock);
847 err = fib6_add(&table->tb6_root, rt, info);
848 write_unlock_bh(&table->tb6_lock);
853 int ip6_ins_rt(struct rt6_info *rt)
855 struct nl_info info = {
856 .nl_net = dev_net(rt->dst.dev),
858 return __ip6_ins_rt(rt, &info);
861 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
862 const struct in6_addr *daddr,
863 const struct in6_addr *saddr)
871 rt = ip6_rt_copy(ort, daddr);
874 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
875 if (ort->rt6i_dst.plen != 128 &&
876 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
877 rt->rt6i_flags |= RTF_ANYCAST;
880 rt->rt6i_flags |= RTF_CACHE;
882 #ifdef CONFIG_IPV6_SUBTREES
883 if (rt->rt6i_src.plen && saddr) {
884 rt->rt6i_src.addr = *saddr;
885 rt->rt6i_src.plen = 128;
893 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
894 const struct in6_addr *daddr)
896 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
899 rt->rt6i_flags |= RTF_CACHE;
903 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
904 struct flowi6 *fl6, int flags)
906 struct fib6_node *fn;
907 struct rt6_info *rt, *nrt;
911 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
913 strict |= flags & RT6_LOOKUP_F_IFACE;
916 read_lock_bh(&table->tb6_lock);
919 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
922 rt = rt6_select(fn, oif, strict | reachable);
923 if (rt->rt6i_nsiblings)
924 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
925 BACKTRACK(net, &fl6->saddr);
926 if (rt == net->ipv6.ip6_null_entry ||
927 rt->rt6i_flags & RTF_CACHE)
931 read_unlock_bh(&table->tb6_lock);
933 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
934 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
935 else if (!(rt->dst.flags & DST_HOST))
936 nrt = rt6_alloc_clone(rt, &fl6->daddr);
941 rt = nrt ? : net->ipv6.ip6_null_entry;
945 err = ip6_ins_rt(nrt);
954 * Race condition! In the gap, when table->tb6_lock was
955 * released someone could insert this route. Relookup.
966 read_unlock_bh(&table->tb6_lock);
968 rt->dst.lastuse = jiffies;
974 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
975 struct flowi6 *fl6, int flags)
977 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
980 static struct dst_entry *ip6_route_input_lookup(struct net *net,
981 struct net_device *dev,
982 struct flowi6 *fl6, int flags)
984 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
985 flags |= RT6_LOOKUP_F_IFACE;
987 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
990 void ip6_route_input(struct sk_buff *skb)
992 const struct ipv6hdr *iph = ipv6_hdr(skb);
993 struct net *net = dev_net(skb->dev);
994 int flags = RT6_LOOKUP_F_HAS_SADDR;
995 struct flowi6 fl6 = {
996 .flowi6_iif = skb->dev->ifindex,
999 .flowlabel = ip6_flowinfo(iph),
1000 .flowi6_mark = skb->mark,
1001 .flowi6_proto = iph->nexthdr,
1004 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1007 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1008 struct flowi6 *fl6, int flags)
1010 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1013 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1018 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1020 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1021 flags |= RT6_LOOKUP_F_IFACE;
1023 if (!ipv6_addr_any(&fl6->saddr))
1024 flags |= RT6_LOOKUP_F_HAS_SADDR;
1026 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1028 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1031 EXPORT_SYMBOL(ip6_route_output);
1033 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1035 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1036 struct dst_entry *new = NULL;
1038 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1042 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1043 rt6_init_peer(rt, net->ipv6.peers);
1046 new->input = dst_discard;
1047 new->output = dst_discard;
1049 if (dst_metrics_read_only(&ort->dst))
1050 new->_metrics = ort->dst._metrics;
1052 dst_copy_metrics(new, &ort->dst);
1053 rt->rt6i_idev = ort->rt6i_idev;
1055 in6_dev_hold(rt->rt6i_idev);
1057 rt->rt6i_gateway = ort->rt6i_gateway;
1058 rt->rt6i_flags = ort->rt6i_flags;
1059 rt->rt6i_metric = 0;
1061 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1062 #ifdef CONFIG_IPV6_SUBTREES
1063 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1069 dst_release(dst_orig);
1070 return new ? new : ERR_PTR(-ENOMEM);
1074 * Destination cache support functions
1077 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1079 struct rt6_info *rt;
1081 rt = (struct rt6_info *) dst;
1083 /* All IPV6 dsts are created with ->obsolete set to the value
1084 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1085 * into this function always.
1087 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1090 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1096 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1098 struct rt6_info *rt = (struct rt6_info *) dst;
1101 if (rt->rt6i_flags & RTF_CACHE) {
1102 if (rt6_check_expired(rt)) {
1114 static void ip6_link_failure(struct sk_buff *skb)
1116 struct rt6_info *rt;
1118 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1120 rt = (struct rt6_info *) skb_dst(skb);
1122 if (rt->rt6i_flags & RTF_CACHE) {
1126 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1127 rt->rt6i_node->fn_sernum = -1;
1132 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1133 struct sk_buff *skb, u32 mtu)
1135 struct rt6_info *rt6 = (struct rt6_info*)dst;
1138 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1139 struct net *net = dev_net(dst->dev);
1141 rt6->rt6i_flags |= RTF_MODIFIED;
1142 if (mtu < IPV6_MIN_MTU) {
1143 u32 features = dst_metric(dst, RTAX_FEATURES);
1145 features |= RTAX_FEATURE_ALLFRAG;
1146 dst_metric_set(dst, RTAX_FEATURES, features);
1148 dst_metric_set(dst, RTAX_MTU, mtu);
1149 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1153 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1156 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1157 struct dst_entry *dst;
1160 memset(&fl6, 0, sizeof(fl6));
1161 fl6.flowi6_oif = oif;
1162 fl6.flowi6_mark = mark;
1163 fl6.daddr = iph->daddr;
1164 fl6.saddr = iph->saddr;
1165 fl6.flowlabel = ip6_flowinfo(iph);
1167 dst = ip6_route_output(net, NULL, &fl6);
1169 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1172 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1174 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1176 ip6_update_pmtu(skb, sock_net(sk), mtu,
1177 sk->sk_bound_dev_if, sk->sk_mark);
1179 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1181 /* Handle redirects */
1182 struct ip6rd_flowi {
1184 struct in6_addr gateway;
1187 static struct rt6_info *__ip6_route_redirect(struct net *net,
1188 struct fib6_table *table,
1192 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1193 struct rt6_info *rt;
1194 struct fib6_node *fn;
1196 /* Get the "current" route for this destination and
1197 * check if the redirect has come from approriate router.
1199 * RFC 4861 specifies that redirects should only be
1200 * accepted if they come from the nexthop to the target.
1201 * Due to the way the routes are chosen, this notion
1202 * is a bit fuzzy and one might need to check all possible
1206 read_lock_bh(&table->tb6_lock);
1207 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1209 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1210 if (rt6_check_expired(rt))
1214 if (!(rt->rt6i_flags & RTF_GATEWAY))
1216 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1218 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1224 rt = net->ipv6.ip6_null_entry;
1225 else if (rt->dst.error) {
1226 rt = net->ipv6.ip6_null_entry;
1229 BACKTRACK(net, &fl6->saddr);
1233 read_unlock_bh(&table->tb6_lock);
1238 static struct dst_entry *ip6_route_redirect(struct net *net,
1239 const struct flowi6 *fl6,
1240 const struct in6_addr *gateway)
1242 int flags = RT6_LOOKUP_F_HAS_SADDR;
1243 struct ip6rd_flowi rdfl;
1246 rdfl.gateway = *gateway;
1248 return fib6_rule_lookup(net, &rdfl.fl6,
1249 flags, __ip6_route_redirect);
1252 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1254 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1255 struct dst_entry *dst;
1258 memset(&fl6, 0, sizeof(fl6));
1259 fl6.flowi6_oif = oif;
1260 fl6.flowi6_mark = mark;
1261 fl6.daddr = iph->daddr;
1262 fl6.saddr = iph->saddr;
1263 fl6.flowlabel = ip6_flowinfo(iph);
1265 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1266 rt6_do_redirect(dst, NULL, skb);
1269 EXPORT_SYMBOL_GPL(ip6_redirect);
1271 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1274 const struct ipv6hdr *iph = ipv6_hdr(skb);
1275 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1276 struct dst_entry *dst;
1279 memset(&fl6, 0, sizeof(fl6));
1280 fl6.flowi6_oif = oif;
1281 fl6.flowi6_mark = mark;
1282 fl6.daddr = msg->dest;
1283 fl6.saddr = iph->daddr;
1285 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1286 rt6_do_redirect(dst, NULL, skb);
1290 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1292 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1294 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1296 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1298 struct net_device *dev = dst->dev;
1299 unsigned int mtu = dst_mtu(dst);
1300 struct net *net = dev_net(dev);
1302 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1304 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1305 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1308 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1309 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1310 * IPV6_MAXPLEN is also valid and means: "any MSS,
1311 * rely only on pmtu discovery"
1313 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1318 static unsigned int ip6_mtu(const struct dst_entry *dst)
1320 struct inet6_dev *idev;
1321 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1329 idev = __in6_dev_get(dst->dev);
1331 mtu = idev->cnf.mtu6;
1337 static struct dst_entry *icmp6_dst_gc_list;
1338 static DEFINE_SPINLOCK(icmp6_dst_lock);
1340 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1343 struct dst_entry *dst;
1344 struct rt6_info *rt;
1345 struct inet6_dev *idev = in6_dev_get(dev);
1346 struct net *net = dev_net(dev);
1348 if (unlikely(!idev))
1349 return ERR_PTR(-ENODEV);
1351 rt = ip6_dst_alloc(net, dev, 0, NULL);
1352 if (unlikely(!rt)) {
1354 dst = ERR_PTR(-ENOMEM);
1358 rt->dst.flags |= DST_HOST;
1359 rt->dst.output = ip6_output;
1360 atomic_set(&rt->dst.__refcnt, 1);
1361 rt->rt6i_gateway = fl6->daddr;
1362 rt->rt6i_dst.addr = fl6->daddr;
1363 rt->rt6i_dst.plen = 128;
1364 rt->rt6i_idev = idev;
1365 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1367 spin_lock_bh(&icmp6_dst_lock);
1368 rt->dst.next = icmp6_dst_gc_list;
1369 icmp6_dst_gc_list = &rt->dst;
1370 spin_unlock_bh(&icmp6_dst_lock);
1372 fib6_force_start_gc(net);
1374 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1380 int icmp6_dst_gc(void)
1382 struct dst_entry *dst, **pprev;
1385 spin_lock_bh(&icmp6_dst_lock);
1386 pprev = &icmp6_dst_gc_list;
1388 while ((dst = *pprev) != NULL) {
1389 if (!atomic_read(&dst->__refcnt)) {
1398 spin_unlock_bh(&icmp6_dst_lock);
1403 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1406 struct dst_entry *dst, **pprev;
1408 spin_lock_bh(&icmp6_dst_lock);
1409 pprev = &icmp6_dst_gc_list;
1410 while ((dst = *pprev) != NULL) {
1411 struct rt6_info *rt = (struct rt6_info *) dst;
1412 if (func(rt, arg)) {
1419 spin_unlock_bh(&icmp6_dst_lock);
1422 static int ip6_dst_gc(struct dst_ops *ops)
1424 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1425 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1426 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1427 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1428 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1429 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1432 entries = dst_entries_get_fast(ops);
1433 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1434 entries <= rt_max_size)
1437 net->ipv6.ip6_rt_gc_expire++;
1438 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1439 entries = dst_entries_get_slow(ops);
1440 if (entries < ops->gc_thresh)
1441 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1443 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1444 return entries > rt_max_size;
1451 int ip6_route_add(struct fib6_config *cfg)
1454 struct net *net = cfg->fc_nlinfo.nl_net;
1455 struct rt6_info *rt = NULL;
1456 struct net_device *dev = NULL;
1457 struct inet6_dev *idev = NULL;
1458 struct fib6_table *table;
1461 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1463 #ifndef CONFIG_IPV6_SUBTREES
1464 if (cfg->fc_src_len)
1467 if (cfg->fc_ifindex) {
1469 dev = dev_get_by_index(net, cfg->fc_ifindex);
1472 idev = in6_dev_get(dev);
1477 if (cfg->fc_metric == 0)
1478 cfg->fc_metric = IP6_RT_PRIO_USER;
1481 if (cfg->fc_nlinfo.nlh &&
1482 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1483 table = fib6_get_table(net, cfg->fc_table);
1485 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1486 table = fib6_new_table(net, cfg->fc_table);
1489 table = fib6_new_table(net, cfg->fc_table);
1495 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1502 if (cfg->fc_flags & RTF_EXPIRES)
1503 rt6_set_expires(rt, jiffies +
1504 clock_t_to_jiffies(cfg->fc_expires));
1506 rt6_clean_expires(rt);
1508 if (cfg->fc_protocol == RTPROT_UNSPEC)
1509 cfg->fc_protocol = RTPROT_BOOT;
1510 rt->rt6i_protocol = cfg->fc_protocol;
1512 addr_type = ipv6_addr_type(&cfg->fc_dst);
1514 if (addr_type & IPV6_ADDR_MULTICAST)
1515 rt->dst.input = ip6_mc_input;
1516 else if (cfg->fc_flags & RTF_LOCAL)
1517 rt->dst.input = ip6_input;
1519 rt->dst.input = ip6_forward;
1521 rt->dst.output = ip6_output;
1523 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1524 rt->rt6i_dst.plen = cfg->fc_dst_len;
1525 if (rt->rt6i_dst.plen == 128)
1526 rt->dst.flags |= DST_HOST;
1528 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1529 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1534 dst_init_metrics(&rt->dst, metrics, 0);
1536 #ifdef CONFIG_IPV6_SUBTREES
1537 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1538 rt->rt6i_src.plen = cfg->fc_src_len;
1541 rt->rt6i_metric = cfg->fc_metric;
1543 /* We cannot add true routes via loopback here,
1544 they would result in kernel looping; promote them to reject routes
1546 if ((cfg->fc_flags & RTF_REJECT) ||
1547 (dev && (dev->flags & IFF_LOOPBACK) &&
1548 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1549 !(cfg->fc_flags & RTF_LOCAL))) {
1550 /* hold loopback dev/idev if we haven't done so. */
1551 if (dev != net->loopback_dev) {
1556 dev = net->loopback_dev;
1558 idev = in6_dev_get(dev);
1564 rt->dst.output = ip6_pkt_discard_out;
1565 rt->dst.input = ip6_pkt_discard;
1566 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1567 switch (cfg->fc_type) {
1569 rt->dst.error = -EINVAL;
1572 rt->dst.error = -EACCES;
1575 rt->dst.error = -EAGAIN;
1578 rt->dst.error = -ENETUNREACH;
1584 if (cfg->fc_flags & RTF_GATEWAY) {
1585 const struct in6_addr *gw_addr;
1588 gw_addr = &cfg->fc_gateway;
1589 rt->rt6i_gateway = *gw_addr;
1590 gwa_type = ipv6_addr_type(gw_addr);
1592 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1593 struct rt6_info *grt;
1595 /* IPv6 strictly inhibits using not link-local
1596 addresses as nexthop address.
1597 Otherwise, router will not able to send redirects.
1598 It is very good, but in some (rare!) circumstances
1599 (SIT, PtP, NBMA NOARP links) it is handy to allow
1600 some exceptions. --ANK
1603 if (!(gwa_type & IPV6_ADDR_UNICAST))
1606 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1608 err = -EHOSTUNREACH;
1612 if (dev != grt->dst.dev) {
1618 idev = grt->rt6i_idev;
1620 in6_dev_hold(grt->rt6i_idev);
1622 if (!(grt->rt6i_flags & RTF_GATEWAY))
1630 if (!dev || (dev->flags & IFF_LOOPBACK))
1638 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1639 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1643 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1644 rt->rt6i_prefsrc.plen = 128;
1646 rt->rt6i_prefsrc.plen = 0;
1648 rt->rt6i_flags = cfg->fc_flags;
1655 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1656 int type = nla_type(nla);
1659 if (type > RTAX_MAX) {
1664 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1670 rt->rt6i_idev = idev;
1671 rt->rt6i_table = table;
1673 cfg->fc_nlinfo.nl_net = dev_net(dev);
1675 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1687 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1690 struct fib6_table *table;
1691 struct net *net = dev_net(rt->dst.dev);
1693 if (rt == net->ipv6.ip6_null_entry) {
1698 table = rt->rt6i_table;
1699 write_lock_bh(&table->tb6_lock);
1700 err = fib6_del(rt, info);
1701 write_unlock_bh(&table->tb6_lock);
1708 int ip6_del_rt(struct rt6_info *rt)
1710 struct nl_info info = {
1711 .nl_net = dev_net(rt->dst.dev),
1713 return __ip6_del_rt(rt, &info);
1716 static int ip6_route_del(struct fib6_config *cfg)
1718 struct fib6_table *table;
1719 struct fib6_node *fn;
1720 struct rt6_info *rt;
1723 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1727 read_lock_bh(&table->tb6_lock);
1729 fn = fib6_locate(&table->tb6_root,
1730 &cfg->fc_dst, cfg->fc_dst_len,
1731 &cfg->fc_src, cfg->fc_src_len);
1734 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1735 if (cfg->fc_ifindex &&
1737 rt->dst.dev->ifindex != cfg->fc_ifindex))
1739 if (cfg->fc_flags & RTF_GATEWAY &&
1740 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1742 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1745 read_unlock_bh(&table->tb6_lock);
1747 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1750 read_unlock_bh(&table->tb6_lock);
1755 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1757 struct net *net = dev_net(skb->dev);
1758 struct netevent_redirect netevent;
1759 struct rt6_info *rt, *nrt = NULL;
1760 struct ndisc_options ndopts;
1761 struct inet6_dev *in6_dev;
1762 struct neighbour *neigh;
1764 int optlen, on_link;
1767 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1768 optlen -= sizeof(*msg);
1771 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1775 msg = (struct rd_msg *)icmp6_hdr(skb);
1777 if (ipv6_addr_is_multicast(&msg->dest)) {
1778 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1783 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1785 } else if (ipv6_addr_type(&msg->target) !=
1786 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1787 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1791 in6_dev = __in6_dev_get(skb->dev);
1794 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1798 * The IP source address of the Redirect MUST be the same as the current
1799 * first-hop router for the specified ICMP Destination Address.
1802 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1803 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1808 if (ndopts.nd_opts_tgt_lladdr) {
1809 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1812 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1817 rt = (struct rt6_info *) dst;
1818 if (rt == net->ipv6.ip6_null_entry) {
1819 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1823 /* Redirect received -> path was valid.
1824 * Look, redirects are sent only in response to data packets,
1825 * so that this nexthop apparently is reachable. --ANK
1827 dst_confirm(&rt->dst);
1829 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1834 * We have finally decided to accept it.
1837 neigh_update(neigh, lladdr, NUD_STALE,
1838 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1839 NEIGH_UPDATE_F_OVERRIDE|
1840 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1841 NEIGH_UPDATE_F_ISROUTER))
1844 nrt = ip6_rt_copy(rt, &msg->dest);
1848 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1850 nrt->rt6i_flags &= ~RTF_GATEWAY;
1852 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1854 if (ip6_ins_rt(nrt))
1857 netevent.old = &rt->dst;
1858 netevent.new = &nrt->dst;
1859 netevent.daddr = &msg->dest;
1860 netevent.neigh = neigh;
1861 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1863 if (rt->rt6i_flags & RTF_CACHE) {
1864 rt = (struct rt6_info *) dst_clone(&rt->dst);
1869 neigh_release(neigh);
1873 * Misc support functions
1876 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1877 const struct in6_addr *dest)
1879 struct net *net = dev_net(ort->dst.dev);
1880 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1884 rt->dst.input = ort->dst.input;
1885 rt->dst.output = ort->dst.output;
1886 rt->dst.flags |= DST_HOST;
1888 rt->rt6i_dst.addr = *dest;
1889 rt->rt6i_dst.plen = 128;
1890 dst_copy_metrics(&rt->dst, &ort->dst);
1891 rt->dst.error = ort->dst.error;
1892 rt->rt6i_idev = ort->rt6i_idev;
1894 in6_dev_hold(rt->rt6i_idev);
1895 rt->dst.lastuse = jiffies;
1897 if (ort->rt6i_flags & RTF_GATEWAY)
1898 rt->rt6i_gateway = ort->rt6i_gateway;
1900 rt->rt6i_gateway = *dest;
1901 rt->rt6i_flags = ort->rt6i_flags;
1902 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1903 (RTF_DEFAULT | RTF_ADDRCONF))
1904 rt6_set_from(rt, ort);
1905 rt->rt6i_metric = 0;
1907 #ifdef CONFIG_IPV6_SUBTREES
1908 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1910 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1911 rt->rt6i_table = ort->rt6i_table;
1916 #ifdef CONFIG_IPV6_ROUTE_INFO
1917 static struct rt6_info *rt6_get_route_info(struct net *net,
1918 const struct in6_addr *prefix, int prefixlen,
1919 const struct in6_addr *gwaddr, int ifindex)
1921 struct fib6_node *fn;
1922 struct rt6_info *rt = NULL;
1923 struct fib6_table *table;
1925 table = fib6_get_table(net, RT6_TABLE_INFO);
1929 read_lock_bh(&table->tb6_lock);
1930 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1934 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1935 if (rt->dst.dev->ifindex != ifindex)
1937 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1939 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1945 read_unlock_bh(&table->tb6_lock);
1949 static struct rt6_info *rt6_add_route_info(struct net *net,
1950 const struct in6_addr *prefix, int prefixlen,
1951 const struct in6_addr *gwaddr, int ifindex,
1954 struct fib6_config cfg = {
1955 .fc_table = RT6_TABLE_INFO,
1956 .fc_metric = IP6_RT_PRIO_USER,
1957 .fc_ifindex = ifindex,
1958 .fc_dst_len = prefixlen,
1959 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1960 RTF_UP | RTF_PREF(pref),
1961 .fc_nlinfo.portid = 0,
1962 .fc_nlinfo.nlh = NULL,
1963 .fc_nlinfo.nl_net = net,
1966 cfg.fc_dst = *prefix;
1967 cfg.fc_gateway = *gwaddr;
1969 /* We should treat it as a default route if prefix length is 0. */
1971 cfg.fc_flags |= RTF_DEFAULT;
1973 ip6_route_add(&cfg);
1975 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1979 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1981 struct rt6_info *rt;
1982 struct fib6_table *table;
1984 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1988 read_lock_bh(&table->tb6_lock);
1989 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1990 if (dev == rt->dst.dev &&
1991 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1992 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1997 read_unlock_bh(&table->tb6_lock);
2001 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2002 struct net_device *dev,
2005 struct fib6_config cfg = {
2006 .fc_table = RT6_TABLE_DFLT,
2007 .fc_metric = IP6_RT_PRIO_USER,
2008 .fc_ifindex = dev->ifindex,
2009 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2010 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2011 .fc_nlinfo.portid = 0,
2012 .fc_nlinfo.nlh = NULL,
2013 .fc_nlinfo.nl_net = dev_net(dev),
2016 cfg.fc_gateway = *gwaddr;
2018 ip6_route_add(&cfg);
2020 return rt6_get_dflt_router(gwaddr, dev);
2023 void rt6_purge_dflt_routers(struct net *net)
2025 struct rt6_info *rt;
2026 struct fib6_table *table;
2028 /* NOTE: Keep consistent with rt6_get_dflt_router */
2029 table = fib6_get_table(net, RT6_TABLE_DFLT);
2034 read_lock_bh(&table->tb6_lock);
2035 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2036 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2037 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2039 read_unlock_bh(&table->tb6_lock);
2044 read_unlock_bh(&table->tb6_lock);
2047 static void rtmsg_to_fib6_config(struct net *net,
2048 struct in6_rtmsg *rtmsg,
2049 struct fib6_config *cfg)
2051 memset(cfg, 0, sizeof(*cfg));
2053 cfg->fc_table = RT6_TABLE_MAIN;
2054 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2055 cfg->fc_metric = rtmsg->rtmsg_metric;
2056 cfg->fc_expires = rtmsg->rtmsg_info;
2057 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2058 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2059 cfg->fc_flags = rtmsg->rtmsg_flags;
2061 cfg->fc_nlinfo.nl_net = net;
2063 cfg->fc_dst = rtmsg->rtmsg_dst;
2064 cfg->fc_src = rtmsg->rtmsg_src;
2065 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2068 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2070 struct fib6_config cfg;
2071 struct in6_rtmsg rtmsg;
2075 case SIOCADDRT: /* Add a route */
2076 case SIOCDELRT: /* Delete a route */
2077 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2079 err = copy_from_user(&rtmsg, arg,
2080 sizeof(struct in6_rtmsg));
2084 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2089 err = ip6_route_add(&cfg);
2092 err = ip6_route_del(&cfg);
2106 * Drop the packet on the floor
2109 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2112 struct dst_entry *dst = skb_dst(skb);
2113 switch (ipstats_mib_noroutes) {
2114 case IPSTATS_MIB_INNOROUTES:
2115 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2116 if (type == IPV6_ADDR_ANY) {
2117 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2118 IPSTATS_MIB_INADDRERRORS);
2122 case IPSTATS_MIB_OUTNOROUTES:
2123 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2124 ipstats_mib_noroutes);
2127 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2132 static int ip6_pkt_discard(struct sk_buff *skb)
2134 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2137 static int ip6_pkt_discard_out(struct sk_buff *skb)
2139 skb->dev = skb_dst(skb)->dev;
2140 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2143 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2145 static int ip6_pkt_prohibit(struct sk_buff *skb)
2147 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2150 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2152 skb->dev = skb_dst(skb)->dev;
2153 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2159 * Allocate a dst for local (unicast / anycast) address.
2162 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2163 const struct in6_addr *addr,
2166 struct net *net = dev_net(idev->dev);
2167 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2170 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2171 return ERR_PTR(-ENOMEM);
2176 rt->dst.flags |= DST_HOST;
2177 rt->dst.input = ip6_input;
2178 rt->dst.output = ip6_output;
2179 rt->rt6i_idev = idev;
2181 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2183 rt->rt6i_flags |= RTF_ANYCAST;
2185 rt->rt6i_flags |= RTF_LOCAL;
2187 rt->rt6i_gateway = *addr;
2188 rt->rt6i_dst.addr = *addr;
2189 rt->rt6i_dst.plen = 128;
2190 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2192 atomic_set(&rt->dst.__refcnt, 1);
2197 int ip6_route_get_saddr(struct net *net,
2198 struct rt6_info *rt,
2199 const struct in6_addr *daddr,
2201 struct in6_addr *saddr)
2203 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2205 if (rt->rt6i_prefsrc.plen)
2206 *saddr = rt->rt6i_prefsrc.addr;
2208 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2209 daddr, prefs, saddr);
2213 /* remove deleted ip from prefsrc entries */
2214 struct arg_dev_net_ip {
2215 struct net_device *dev;
2217 struct in6_addr *addr;
2220 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2222 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2223 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2224 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2226 if (((void *)rt->dst.dev == dev || !dev) &&
2227 rt != net->ipv6.ip6_null_entry &&
2228 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2229 /* remove prefsrc entry */
2230 rt->rt6i_prefsrc.plen = 0;
2235 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2237 struct net *net = dev_net(ifp->idev->dev);
2238 struct arg_dev_net_ip adni = {
2239 .dev = ifp->idev->dev,
2243 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2246 struct arg_dev_net {
2247 struct net_device *dev;
2251 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2253 const struct arg_dev_net *adn = arg;
2254 const struct net_device *dev = adn->dev;
2256 if ((rt->dst.dev == dev || !dev) &&
2257 rt != adn->net->ipv6.ip6_null_entry)
2263 void rt6_ifdown(struct net *net, struct net_device *dev)
2265 struct arg_dev_net adn = {
2270 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2271 icmp6_clean_all(fib6_ifdown, &adn);
2274 struct rt6_mtu_change_arg {
2275 struct net_device *dev;
2279 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2281 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2282 struct inet6_dev *idev;
2284 /* In IPv6 pmtu discovery is not optional,
2285 so that RTAX_MTU lock cannot disable it.
2286 We still use this lock to block changes
2287 caused by addrconf/ndisc.
2290 idev = __in6_dev_get(arg->dev);
2294 /* For administrative MTU increase, there is no way to discover
2295 IPv6 PMTU increase, so PMTU increase should be updated here.
2296 Since RFC 1981 doesn't include administrative MTU increase
2297 update PMTU increase is a MUST. (i.e. jumbo frame)
2300 If new MTU is less than route PMTU, this new MTU will be the
2301 lowest MTU in the path, update the route PMTU to reflect PMTU
2302 decreases; if new MTU is greater than route PMTU, and the
2303 old MTU is the lowest MTU in the path, update the route PMTU
2304 to reflect the increase. In this case if the other nodes' MTU
2305 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2308 if (rt->dst.dev == arg->dev &&
2309 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2310 (dst_mtu(&rt->dst) >= arg->mtu ||
2311 (dst_mtu(&rt->dst) < arg->mtu &&
2312 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2313 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2318 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2320 struct rt6_mtu_change_arg arg = {
2325 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2328 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2329 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2330 [RTA_OIF] = { .type = NLA_U32 },
2331 [RTA_IIF] = { .type = NLA_U32 },
2332 [RTA_PRIORITY] = { .type = NLA_U32 },
2333 [RTA_METRICS] = { .type = NLA_NESTED },
2334 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2337 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2338 struct fib6_config *cfg)
2341 struct nlattr *tb[RTA_MAX+1];
2344 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2349 rtm = nlmsg_data(nlh);
2350 memset(cfg, 0, sizeof(*cfg));
2352 cfg->fc_table = rtm->rtm_table;
2353 cfg->fc_dst_len = rtm->rtm_dst_len;
2354 cfg->fc_src_len = rtm->rtm_src_len;
2355 cfg->fc_flags = RTF_UP;
2356 cfg->fc_protocol = rtm->rtm_protocol;
2357 cfg->fc_type = rtm->rtm_type;
2359 if (rtm->rtm_type == RTN_UNREACHABLE ||
2360 rtm->rtm_type == RTN_BLACKHOLE ||
2361 rtm->rtm_type == RTN_PROHIBIT ||
2362 rtm->rtm_type == RTN_THROW)
2363 cfg->fc_flags |= RTF_REJECT;
2365 if (rtm->rtm_type == RTN_LOCAL)
2366 cfg->fc_flags |= RTF_LOCAL;
2368 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2369 cfg->fc_nlinfo.nlh = nlh;
2370 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2372 if (tb[RTA_GATEWAY]) {
2373 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2374 cfg->fc_flags |= RTF_GATEWAY;
2378 int plen = (rtm->rtm_dst_len + 7) >> 3;
2380 if (nla_len(tb[RTA_DST]) < plen)
2383 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2387 int plen = (rtm->rtm_src_len + 7) >> 3;
2389 if (nla_len(tb[RTA_SRC]) < plen)
2392 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2395 if (tb[RTA_PREFSRC])
2396 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2399 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2401 if (tb[RTA_PRIORITY])
2402 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2404 if (tb[RTA_METRICS]) {
2405 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2406 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2410 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2412 if (tb[RTA_MULTIPATH]) {
2413 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2414 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2422 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2424 struct fib6_config r_cfg;
2425 struct rtnexthop *rtnh;
2428 int err = 0, last_err = 0;
2431 rtnh = (struct rtnexthop *)cfg->fc_mp;
2432 remaining = cfg->fc_mp_len;
2434 /* Parse a Multipath Entry */
2435 while (rtnh_ok(rtnh, remaining)) {
2436 memcpy(&r_cfg, cfg, sizeof(*cfg));
2437 if (rtnh->rtnh_ifindex)
2438 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2440 attrlen = rtnh_attrlen(rtnh);
2442 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2444 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2446 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2447 r_cfg.fc_flags |= RTF_GATEWAY;
2450 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2453 /* If we are trying to remove a route, do not stop the
2454 * loop when ip6_route_del() fails (because next hop is
2455 * already gone), we should try to remove all next hops.
2458 /* If add fails, we should try to delete all
2459 * next hops that have been already added.
2465 /* Because each route is added like a single route we remove
2466 * this flag after the first nexthop (if there is a collision,
2467 * we have already fail to add the first nexthop:
2468 * fib6_add_rt2node() has reject it).
2470 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2471 rtnh = rtnh_next(rtnh, &remaining);
2477 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2479 struct fib6_config cfg;
2482 err = rtm_to_fib6_config(skb, nlh, &cfg);
2487 return ip6_route_multipath(&cfg, 0);
2489 return ip6_route_del(&cfg);
2492 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2494 struct fib6_config cfg;
2497 err = rtm_to_fib6_config(skb, nlh, &cfg);
2502 return ip6_route_multipath(&cfg, 1);
2504 return ip6_route_add(&cfg);
2507 static inline size_t rt6_nlmsg_size(void)
2509 return NLMSG_ALIGN(sizeof(struct rtmsg))
2510 + nla_total_size(16) /* RTA_SRC */
2511 + nla_total_size(16) /* RTA_DST */
2512 + nla_total_size(16) /* RTA_GATEWAY */
2513 + nla_total_size(16) /* RTA_PREFSRC */
2514 + nla_total_size(4) /* RTA_TABLE */
2515 + nla_total_size(4) /* RTA_IIF */
2516 + nla_total_size(4) /* RTA_OIF */
2517 + nla_total_size(4) /* RTA_PRIORITY */
2518 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2519 + nla_total_size(sizeof(struct rta_cacheinfo));
2522 static int rt6_fill_node(struct net *net,
2523 struct sk_buff *skb, struct rt6_info *rt,
2524 struct in6_addr *dst, struct in6_addr *src,
2525 int iif, int type, u32 portid, u32 seq,
2526 int prefix, int nowait, unsigned int flags)
2529 struct nlmsghdr *nlh;
2533 if (prefix) { /* user wants prefix routes only */
2534 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2535 /* success since this is not a prefix route */
2540 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2544 rtm = nlmsg_data(nlh);
2545 rtm->rtm_family = AF_INET6;
2546 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2547 rtm->rtm_src_len = rt->rt6i_src.plen;
2550 table = rt->rt6i_table->tb6_id;
2552 table = RT6_TABLE_UNSPEC;
2553 rtm->rtm_table = table;
2554 if (nla_put_u32(skb, RTA_TABLE, table))
2555 goto nla_put_failure;
2556 if (rt->rt6i_flags & RTF_REJECT) {
2557 switch (rt->dst.error) {
2559 rtm->rtm_type = RTN_BLACKHOLE;
2562 rtm->rtm_type = RTN_PROHIBIT;
2565 rtm->rtm_type = RTN_THROW;
2568 rtm->rtm_type = RTN_UNREACHABLE;
2572 else if (rt->rt6i_flags & RTF_LOCAL)
2573 rtm->rtm_type = RTN_LOCAL;
2574 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2575 rtm->rtm_type = RTN_LOCAL;
2577 rtm->rtm_type = RTN_UNICAST;
2579 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2580 rtm->rtm_protocol = rt->rt6i_protocol;
2581 if (rt->rt6i_flags & RTF_DYNAMIC)
2582 rtm->rtm_protocol = RTPROT_REDIRECT;
2583 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2584 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2585 rtm->rtm_protocol = RTPROT_RA;
2587 rtm->rtm_protocol = RTPROT_KERNEL;
2590 if (rt->rt6i_flags & RTF_CACHE)
2591 rtm->rtm_flags |= RTM_F_CLONED;
2594 if (nla_put(skb, RTA_DST, 16, dst))
2595 goto nla_put_failure;
2596 rtm->rtm_dst_len = 128;
2597 } else if (rtm->rtm_dst_len)
2598 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2599 goto nla_put_failure;
2600 #ifdef CONFIG_IPV6_SUBTREES
2602 if (nla_put(skb, RTA_SRC, 16, src))
2603 goto nla_put_failure;
2604 rtm->rtm_src_len = 128;
2605 } else if (rtm->rtm_src_len &&
2606 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2607 goto nla_put_failure;
2610 #ifdef CONFIG_IPV6_MROUTE
2611 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2612 int err = ip6mr_get_route(net, skb, rtm, nowait);
2617 goto nla_put_failure;
2619 if (err == -EMSGSIZE)
2620 goto nla_put_failure;
2625 if (nla_put_u32(skb, RTA_IIF, iif))
2626 goto nla_put_failure;
2628 struct in6_addr saddr_buf;
2629 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2630 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2631 goto nla_put_failure;
2634 if (rt->rt6i_prefsrc.plen) {
2635 struct in6_addr saddr_buf;
2636 saddr_buf = rt->rt6i_prefsrc.addr;
2637 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2638 goto nla_put_failure;
2641 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2642 goto nla_put_failure;
2644 if (rt->rt6i_flags & RTF_GATEWAY) {
2645 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2646 goto nla_put_failure;
2650 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2651 goto nla_put_failure;
2652 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2653 goto nla_put_failure;
2655 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2657 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2658 goto nla_put_failure;
2660 return nlmsg_end(skb, nlh);
2663 nlmsg_cancel(skb, nlh);
2667 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2669 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2672 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2673 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2674 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2678 return rt6_fill_node(arg->net,
2679 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2680 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2681 prefix, 0, NLM_F_MULTI);
2684 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2686 struct net *net = sock_net(in_skb->sk);
2687 struct nlattr *tb[RTA_MAX+1];
2688 struct rt6_info *rt;
2689 struct sk_buff *skb;
2692 int err, iif = 0, oif = 0;
2694 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2699 memset(&fl6, 0, sizeof(fl6));
2702 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2705 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2709 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2712 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2716 iif = nla_get_u32(tb[RTA_IIF]);
2719 oif = nla_get_u32(tb[RTA_OIF]);
2722 struct net_device *dev;
2725 dev = __dev_get_by_index(net, iif);
2731 fl6.flowi6_iif = iif;
2733 if (!ipv6_addr_any(&fl6.saddr))
2734 flags |= RT6_LOOKUP_F_HAS_SADDR;
2736 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2739 fl6.flowi6_oif = oif;
2741 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2744 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2751 /* Reserve room for dummy headers, this skb can pass
2752 through good chunk of routing engine.
2754 skb_reset_mac_header(skb);
2755 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2757 skb_dst_set(skb, &rt->dst);
2759 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2760 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2761 nlh->nlmsg_seq, 0, 0, 0);
2767 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2772 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2774 struct sk_buff *skb;
2775 struct net *net = info->nl_net;
2780 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2782 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2786 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2787 event, info->portid, seq, 0, 0, 0);
2789 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2790 WARN_ON(err == -EMSGSIZE);
2794 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2795 info->nlh, gfp_any());
2799 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2802 static int ip6_route_dev_notify(struct notifier_block *this,
2803 unsigned long event, void *ptr)
2805 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2806 struct net *net = dev_net(dev);
2808 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2809 net->ipv6.ip6_null_entry->dst.dev = dev;
2810 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2811 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2812 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2813 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2814 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2815 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2826 #ifdef CONFIG_PROC_FS
2828 static const struct file_operations ipv6_route_proc_fops = {
2829 .owner = THIS_MODULE,
2830 .open = ipv6_route_open,
2832 .llseek = seq_lseek,
2833 .release = seq_release_net,
2836 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2838 struct net *net = (struct net *)seq->private;
2839 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2840 net->ipv6.rt6_stats->fib_nodes,
2841 net->ipv6.rt6_stats->fib_route_nodes,
2842 net->ipv6.rt6_stats->fib_rt_alloc,
2843 net->ipv6.rt6_stats->fib_rt_entries,
2844 net->ipv6.rt6_stats->fib_rt_cache,
2845 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2846 net->ipv6.rt6_stats->fib_discarded_routes);
2851 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2853 return single_open_net(inode, file, rt6_stats_seq_show);
2856 static const struct file_operations rt6_stats_seq_fops = {
2857 .owner = THIS_MODULE,
2858 .open = rt6_stats_seq_open,
2860 .llseek = seq_lseek,
2861 .release = single_release_net,
2863 #endif /* CONFIG_PROC_FS */
2865 #ifdef CONFIG_SYSCTL
2868 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2869 void __user *buffer, size_t *lenp, loff_t *ppos)
2876 net = (struct net *)ctl->extra1;
2877 delay = net->ipv6.sysctl.flush_delay;
2878 proc_dointvec(ctl, write, buffer, lenp, ppos);
2879 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2883 struct ctl_table ipv6_route_table_template[] = {
2885 .procname = "flush",
2886 .data = &init_net.ipv6.sysctl.flush_delay,
2887 .maxlen = sizeof(int),
2889 .proc_handler = ipv6_sysctl_rtcache_flush
2892 .procname = "gc_thresh",
2893 .data = &ip6_dst_ops_template.gc_thresh,
2894 .maxlen = sizeof(int),
2896 .proc_handler = proc_dointvec,
2899 .procname = "max_size",
2900 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2901 .maxlen = sizeof(int),
2903 .proc_handler = proc_dointvec,
2906 .procname = "gc_min_interval",
2907 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2908 .maxlen = sizeof(int),
2910 .proc_handler = proc_dointvec_jiffies,
2913 .procname = "gc_timeout",
2914 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2915 .maxlen = sizeof(int),
2917 .proc_handler = proc_dointvec_jiffies,
2920 .procname = "gc_interval",
2921 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2922 .maxlen = sizeof(int),
2924 .proc_handler = proc_dointvec_jiffies,
2927 .procname = "gc_elasticity",
2928 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2929 .maxlen = sizeof(int),
2931 .proc_handler = proc_dointvec,
2934 .procname = "mtu_expires",
2935 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2936 .maxlen = sizeof(int),
2938 .proc_handler = proc_dointvec_jiffies,
2941 .procname = "min_adv_mss",
2942 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2943 .maxlen = sizeof(int),
2945 .proc_handler = proc_dointvec,
2948 .procname = "gc_min_interval_ms",
2949 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2950 .maxlen = sizeof(int),
2952 .proc_handler = proc_dointvec_ms_jiffies,
2957 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2959 struct ctl_table *table;
2961 table = kmemdup(ipv6_route_table_template,
2962 sizeof(ipv6_route_table_template),
2966 table[0].data = &net->ipv6.sysctl.flush_delay;
2967 table[0].extra1 = net;
2968 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2969 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2970 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2971 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2972 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2973 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2974 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2975 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2976 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2978 /* Don't export sysctls to unprivileged users */
2979 if (net->user_ns != &init_user_ns)
2980 table[0].procname = NULL;
2987 static int __net_init ip6_route_net_init(struct net *net)
2991 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2992 sizeof(net->ipv6.ip6_dst_ops));
2994 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2995 goto out_ip6_dst_ops;
2997 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2998 sizeof(*net->ipv6.ip6_null_entry),
3000 if (!net->ipv6.ip6_null_entry)
3001 goto out_ip6_dst_entries;
3002 net->ipv6.ip6_null_entry->dst.path =
3003 (struct dst_entry *)net->ipv6.ip6_null_entry;
3004 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3005 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3006 ip6_template_metrics, true);
3008 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3009 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3010 sizeof(*net->ipv6.ip6_prohibit_entry),
3012 if (!net->ipv6.ip6_prohibit_entry)
3013 goto out_ip6_null_entry;
3014 net->ipv6.ip6_prohibit_entry->dst.path =
3015 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3016 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3017 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3018 ip6_template_metrics, true);
3020 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3021 sizeof(*net->ipv6.ip6_blk_hole_entry),
3023 if (!net->ipv6.ip6_blk_hole_entry)
3024 goto out_ip6_prohibit_entry;
3025 net->ipv6.ip6_blk_hole_entry->dst.path =
3026 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3027 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3028 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3029 ip6_template_metrics, true);
3032 net->ipv6.sysctl.flush_delay = 0;
3033 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3034 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3035 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3036 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3037 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3038 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3039 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3041 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3047 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3048 out_ip6_prohibit_entry:
3049 kfree(net->ipv6.ip6_prohibit_entry);
3051 kfree(net->ipv6.ip6_null_entry);
3053 out_ip6_dst_entries:
3054 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3059 static void __net_exit ip6_route_net_exit(struct net *net)
3061 kfree(net->ipv6.ip6_null_entry);
3062 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3063 kfree(net->ipv6.ip6_prohibit_entry);
3064 kfree(net->ipv6.ip6_blk_hole_entry);
3066 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3069 static int __net_init ip6_route_net_init_late(struct net *net)
3071 #ifdef CONFIG_PROC_FS
3072 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3073 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3078 static void __net_exit ip6_route_net_exit_late(struct net *net)
3080 #ifdef CONFIG_PROC_FS
3081 remove_proc_entry("ipv6_route", net->proc_net);
3082 remove_proc_entry("rt6_stats", net->proc_net);
3086 static struct pernet_operations ip6_route_net_ops = {
3087 .init = ip6_route_net_init,
3088 .exit = ip6_route_net_exit,
3091 static int __net_init ipv6_inetpeer_init(struct net *net)
3093 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3097 inet_peer_base_init(bp);
3098 net->ipv6.peers = bp;
3102 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3104 struct inet_peer_base *bp = net->ipv6.peers;
3106 net->ipv6.peers = NULL;
3107 inetpeer_invalidate_tree(bp);
3111 static struct pernet_operations ipv6_inetpeer_ops = {
3112 .init = ipv6_inetpeer_init,
3113 .exit = ipv6_inetpeer_exit,
3116 static struct pernet_operations ip6_route_net_late_ops = {
3117 .init = ip6_route_net_init_late,
3118 .exit = ip6_route_net_exit_late,
3121 static struct notifier_block ip6_route_dev_notifier = {
3122 .notifier_call = ip6_route_dev_notify,
3126 int __init ip6_route_init(void)
3131 ip6_dst_ops_template.kmem_cachep =
3132 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3133 SLAB_HWCACHE_ALIGN, NULL);
3134 if (!ip6_dst_ops_template.kmem_cachep)
3137 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3139 goto out_kmem_cache;
3141 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3143 goto out_dst_entries;
3145 ret = register_pernet_subsys(&ip6_route_net_ops);
3147 goto out_register_inetpeer;
3149 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3151 /* Registering of the loopback is done before this portion of code,
3152 * the loopback reference in rt6_info will not be taken, do it
3153 * manually for init_net */
3154 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3155 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3156 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3157 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3158 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3159 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3160 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3164 goto out_register_subsys;
3170 ret = fib6_rules_init();
3174 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3176 goto fib6_rules_init;
3179 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3180 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3181 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3182 goto out_register_late_subsys;
3184 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3186 goto out_register_late_subsys;
3191 out_register_late_subsys:
3192 unregister_pernet_subsys(&ip6_route_net_late_ops);
3194 fib6_rules_cleanup();
3199 out_register_subsys:
3200 unregister_pernet_subsys(&ip6_route_net_ops);
3201 out_register_inetpeer:
3202 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3204 dst_entries_destroy(&ip6_dst_blackhole_ops);
3206 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3210 void ip6_route_cleanup(void)
3212 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3213 unregister_pernet_subsys(&ip6_route_net_late_ops);
3214 fib6_rules_cleanup();
3217 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3218 unregister_pernet_subsys(&ip6_route_net_ops);
3219 dst_entries_destroy(&ip6_dst_blackhole_ops);
3220 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);