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 static void rt6_probe(struct rt6_info *rt)
481 struct neighbour *neigh;
483 * Okay, this does not seem to be appropriate
484 * for now, however, we need to check if it
485 * is really so; aka Router Reachability Probing.
487 * Router Reachability Probe MUST be rate-limited
488 * to no more than one per minute.
490 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
493 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
495 write_lock(&neigh->lock);
496 if (neigh->nud_state & NUD_VALID)
501 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
502 struct in6_addr mcaddr;
503 struct in6_addr *target;
506 neigh->updated = jiffies;
507 write_unlock(&neigh->lock);
510 target = (struct in6_addr *)&rt->rt6i_gateway;
511 addrconf_addr_solict_mult(target, &mcaddr);
512 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
515 write_unlock(&neigh->lock);
517 rcu_read_unlock_bh();
520 static inline void rt6_probe(struct rt6_info *rt)
526 * Default Router Selection (RFC 2461 6.3.6)
528 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
530 struct net_device *dev = rt->dst.dev;
531 if (!oif || dev->ifindex == oif)
533 if ((dev->flags & IFF_LOOPBACK) &&
534 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
539 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
541 struct neighbour *neigh;
542 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
544 if (rt->rt6i_flags & RTF_NONEXTHOP ||
545 !(rt->rt6i_flags & RTF_GATEWAY))
546 return RT6_NUD_SUCCEED;
549 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
551 read_lock(&neigh->lock);
552 if (neigh->nud_state & NUD_VALID)
553 ret = RT6_NUD_SUCCEED;
554 #ifdef CONFIG_IPV6_ROUTER_PREF
555 else if (!(neigh->nud_state & NUD_FAILED))
556 ret = RT6_NUD_SUCCEED;
558 read_unlock(&neigh->lock);
560 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
561 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
563 rcu_read_unlock_bh();
568 static int rt6_score_route(struct rt6_info *rt, int oif,
573 m = rt6_check_dev(rt, oif);
574 if (!m && (strict & RT6_LOOKUP_F_IFACE))
575 return RT6_NUD_FAIL_HARD;
576 #ifdef CONFIG_IPV6_ROUTER_PREF
577 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
579 if (strict & RT6_LOOKUP_F_REACHABLE) {
580 int n = rt6_check_neigh(rt);
587 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
588 int *mpri, struct rt6_info *match,
592 bool match_do_rr = false;
594 if (rt6_check_expired(rt))
597 m = rt6_score_route(rt, oif, strict);
598 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
600 m = 0; /* lowest valid score */
605 if (strict & RT6_LOOKUP_F_REACHABLE)
609 *do_rr = match_do_rr;
617 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
618 struct rt6_info *rr_head,
619 u32 metric, int oif, int strict,
622 struct rt6_info *rt, *match;
626 for (rt = rr_head; rt && rt->rt6i_metric == metric;
627 rt = rt->dst.rt6_next)
628 match = find_match(rt, oif, strict, &mpri, match, do_rr);
629 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
630 rt = rt->dst.rt6_next)
631 match = find_match(rt, oif, strict, &mpri, match, do_rr);
636 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
638 struct rt6_info *match, *rt0;
644 fn->rr_ptr = rt0 = fn->leaf;
646 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
650 struct rt6_info *next = rt0->dst.rt6_next;
652 /* no entries matched; do round-robin */
653 if (!next || next->rt6i_metric != rt0->rt6i_metric)
660 net = dev_net(rt0->dst.dev);
661 return match ? match : net->ipv6.ip6_null_entry;
664 #ifdef CONFIG_IPV6_ROUTE_INFO
665 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
666 const struct in6_addr *gwaddr)
668 struct net *net = dev_net(dev);
669 struct route_info *rinfo = (struct route_info *) opt;
670 struct in6_addr prefix_buf, *prefix;
672 unsigned long lifetime;
675 if (len < sizeof(struct route_info)) {
679 /* Sanity check for prefix_len and length */
680 if (rinfo->length > 3) {
682 } else if (rinfo->prefix_len > 128) {
684 } else if (rinfo->prefix_len > 64) {
685 if (rinfo->length < 2) {
688 } else if (rinfo->prefix_len > 0) {
689 if (rinfo->length < 1) {
694 pref = rinfo->route_pref;
695 if (pref == ICMPV6_ROUTER_PREF_INVALID)
698 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
700 if (rinfo->length == 3)
701 prefix = (struct in6_addr *)rinfo->prefix;
703 /* this function is safe */
704 ipv6_addr_prefix(&prefix_buf,
705 (struct in6_addr *)rinfo->prefix,
707 prefix = &prefix_buf;
710 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
713 if (rt && !lifetime) {
719 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
722 rt->rt6i_flags = RTF_ROUTEINFO |
723 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
726 if (!addrconf_finite_timeout(lifetime))
727 rt6_clean_expires(rt);
729 rt6_set_expires(rt, jiffies + HZ * lifetime);
737 #define BACKTRACK(__net, saddr) \
739 if (rt == __net->ipv6.ip6_null_entry) { \
740 struct fib6_node *pn; \
742 if (fn->fn_flags & RTN_TL_ROOT) \
745 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
746 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
749 if (fn->fn_flags & RTN_RTINFO) \
755 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
756 struct fib6_table *table,
757 struct flowi6 *fl6, int flags)
759 struct fib6_node *fn;
762 read_lock_bh(&table->tb6_lock);
763 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
766 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
767 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
768 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
769 BACKTRACK(net, &fl6->saddr);
771 dst_use(&rt->dst, jiffies);
772 read_unlock_bh(&table->tb6_lock);
777 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
780 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
782 EXPORT_SYMBOL_GPL(ip6_route_lookup);
784 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
785 const struct in6_addr *saddr, int oif, int strict)
787 struct flowi6 fl6 = {
791 struct dst_entry *dst;
792 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
795 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
796 flags |= RT6_LOOKUP_F_HAS_SADDR;
799 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
801 return (struct rt6_info *) dst;
808 EXPORT_SYMBOL(rt6_lookup);
810 /* ip6_ins_rt is called with FREE table->tb6_lock.
811 It takes new route entry, the addition fails by any reason the
812 route is freed. In any case, if caller does not hold it, it may
816 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
819 struct fib6_table *table;
821 table = rt->rt6i_table;
822 write_lock_bh(&table->tb6_lock);
823 err = fib6_add(&table->tb6_root, rt, info);
824 write_unlock_bh(&table->tb6_lock);
829 int ip6_ins_rt(struct rt6_info *rt)
831 struct nl_info info = {
832 .nl_net = dev_net(rt->dst.dev),
834 return __ip6_ins_rt(rt, &info);
837 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
838 const struct in6_addr *daddr,
839 const struct in6_addr *saddr)
847 rt = ip6_rt_copy(ort, daddr);
850 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
851 if (ort->rt6i_dst.plen != 128 &&
852 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
853 rt->rt6i_flags |= RTF_ANYCAST;
856 rt->rt6i_flags |= RTF_CACHE;
858 #ifdef CONFIG_IPV6_SUBTREES
859 if (rt->rt6i_src.plen && saddr) {
860 rt->rt6i_src.addr = *saddr;
861 rt->rt6i_src.plen = 128;
869 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
870 const struct in6_addr *daddr)
872 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
875 rt->rt6i_flags |= RTF_CACHE;
879 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
880 struct flowi6 *fl6, int flags)
882 struct fib6_node *fn;
883 struct rt6_info *rt, *nrt;
887 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
889 strict |= flags & RT6_LOOKUP_F_IFACE;
892 read_lock_bh(&table->tb6_lock);
895 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
898 rt = rt6_select(fn, oif, strict | reachable);
899 if (rt->rt6i_nsiblings)
900 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
901 BACKTRACK(net, &fl6->saddr);
902 if (rt == net->ipv6.ip6_null_entry ||
903 rt->rt6i_flags & RTF_CACHE)
907 read_unlock_bh(&table->tb6_lock);
909 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
910 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
911 else if (!(rt->dst.flags & DST_HOST))
912 nrt = rt6_alloc_clone(rt, &fl6->daddr);
917 rt = nrt ? : net->ipv6.ip6_null_entry;
921 err = ip6_ins_rt(nrt);
930 * Race condition! In the gap, when table->tb6_lock was
931 * released someone could insert this route. Relookup.
942 read_unlock_bh(&table->tb6_lock);
944 rt->dst.lastuse = jiffies;
950 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
951 struct flowi6 *fl6, int flags)
953 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
956 static struct dst_entry *ip6_route_input_lookup(struct net *net,
957 struct net_device *dev,
958 struct flowi6 *fl6, int flags)
960 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
961 flags |= RT6_LOOKUP_F_IFACE;
963 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
966 void ip6_route_input(struct sk_buff *skb)
968 const struct ipv6hdr *iph = ipv6_hdr(skb);
969 struct net *net = dev_net(skb->dev);
970 int flags = RT6_LOOKUP_F_HAS_SADDR;
971 struct flowi6 fl6 = {
972 .flowi6_iif = skb->dev->ifindex,
975 .flowlabel = ip6_flowinfo(iph),
976 .flowi6_mark = skb->mark,
977 .flowi6_proto = iph->nexthdr,
980 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
983 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
984 struct flowi6 *fl6, int flags)
986 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
989 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
994 fl6->flowi6_iif = LOOPBACK_IFINDEX;
996 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
997 flags |= RT6_LOOKUP_F_IFACE;
999 if (!ipv6_addr_any(&fl6->saddr))
1000 flags |= RT6_LOOKUP_F_HAS_SADDR;
1002 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1004 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1007 EXPORT_SYMBOL(ip6_route_output);
1009 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1011 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1012 struct dst_entry *new = NULL;
1014 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1018 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1019 rt6_init_peer(rt, net->ipv6.peers);
1022 new->input = dst_discard;
1023 new->output = dst_discard;
1025 if (dst_metrics_read_only(&ort->dst))
1026 new->_metrics = ort->dst._metrics;
1028 dst_copy_metrics(new, &ort->dst);
1029 rt->rt6i_idev = ort->rt6i_idev;
1031 in6_dev_hold(rt->rt6i_idev);
1033 rt->rt6i_gateway = ort->rt6i_gateway;
1034 rt->rt6i_flags = ort->rt6i_flags;
1035 rt->rt6i_metric = 0;
1037 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1038 #ifdef CONFIG_IPV6_SUBTREES
1039 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1045 dst_release(dst_orig);
1046 return new ? new : ERR_PTR(-ENOMEM);
1050 * Destination cache support functions
1053 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1055 struct rt6_info *rt;
1057 rt = (struct rt6_info *) dst;
1059 /* All IPV6 dsts are created with ->obsolete set to the value
1060 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1061 * into this function always.
1063 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1066 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1072 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1074 struct rt6_info *rt = (struct rt6_info *) dst;
1077 if (rt->rt6i_flags & RTF_CACHE) {
1078 if (rt6_check_expired(rt)) {
1090 static void ip6_link_failure(struct sk_buff *skb)
1092 struct rt6_info *rt;
1094 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1096 rt = (struct rt6_info *) skb_dst(skb);
1098 if (rt->rt6i_flags & RTF_CACHE) {
1102 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1103 rt->rt6i_node->fn_sernum = -1;
1108 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1109 struct sk_buff *skb, u32 mtu)
1111 struct rt6_info *rt6 = (struct rt6_info*)dst;
1114 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1115 struct net *net = dev_net(dst->dev);
1117 rt6->rt6i_flags |= RTF_MODIFIED;
1118 if (mtu < IPV6_MIN_MTU) {
1119 u32 features = dst_metric(dst, RTAX_FEATURES);
1121 features |= RTAX_FEATURE_ALLFRAG;
1122 dst_metric_set(dst, RTAX_FEATURES, features);
1124 dst_metric_set(dst, RTAX_MTU, mtu);
1125 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1129 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1132 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1133 struct dst_entry *dst;
1136 memset(&fl6, 0, sizeof(fl6));
1137 fl6.flowi6_oif = oif;
1138 fl6.flowi6_mark = mark;
1139 fl6.flowi6_flags = 0;
1140 fl6.daddr = iph->daddr;
1141 fl6.saddr = iph->saddr;
1142 fl6.flowlabel = ip6_flowinfo(iph);
1144 dst = ip6_route_output(net, NULL, &fl6);
1146 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1149 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1151 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1153 ip6_update_pmtu(skb, sock_net(sk), mtu,
1154 sk->sk_bound_dev_if, sk->sk_mark);
1156 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1158 /* Handle redirects */
1159 struct ip6rd_flowi {
1161 struct in6_addr gateway;
1164 static struct rt6_info *__ip6_route_redirect(struct net *net,
1165 struct fib6_table *table,
1169 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1170 struct rt6_info *rt;
1171 struct fib6_node *fn;
1173 /* Get the "current" route for this destination and
1174 * check if the redirect has come from approriate router.
1176 * RFC 4861 specifies that redirects should only be
1177 * accepted if they come from the nexthop to the target.
1178 * Due to the way the routes are chosen, this notion
1179 * is a bit fuzzy and one might need to check all possible
1183 read_lock_bh(&table->tb6_lock);
1184 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1186 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1187 if (rt6_check_expired(rt))
1191 if (!(rt->rt6i_flags & RTF_GATEWAY))
1193 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1195 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1201 rt = net->ipv6.ip6_null_entry;
1202 else if (rt->dst.error) {
1203 rt = net->ipv6.ip6_null_entry;
1206 BACKTRACK(net, &fl6->saddr);
1210 read_unlock_bh(&table->tb6_lock);
1215 static struct dst_entry *ip6_route_redirect(struct net *net,
1216 const struct flowi6 *fl6,
1217 const struct in6_addr *gateway)
1219 int flags = RT6_LOOKUP_F_HAS_SADDR;
1220 struct ip6rd_flowi rdfl;
1223 rdfl.gateway = *gateway;
1225 return fib6_rule_lookup(net, &rdfl.fl6,
1226 flags, __ip6_route_redirect);
1229 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1231 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1232 struct dst_entry *dst;
1235 memset(&fl6, 0, sizeof(fl6));
1236 fl6.flowi6_oif = oif;
1237 fl6.flowi6_mark = mark;
1238 fl6.flowi6_flags = 0;
1239 fl6.daddr = iph->daddr;
1240 fl6.saddr = iph->saddr;
1241 fl6.flowlabel = ip6_flowinfo(iph);
1243 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1244 rt6_do_redirect(dst, NULL, skb);
1247 EXPORT_SYMBOL_GPL(ip6_redirect);
1249 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1252 const struct ipv6hdr *iph = ipv6_hdr(skb);
1253 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1254 struct dst_entry *dst;
1257 memset(&fl6, 0, sizeof(fl6));
1258 fl6.flowi6_oif = oif;
1259 fl6.flowi6_mark = mark;
1260 fl6.flowi6_flags = 0;
1261 fl6.daddr = msg->dest;
1262 fl6.saddr = iph->daddr;
1264 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1265 rt6_do_redirect(dst, NULL, skb);
1269 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1271 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1273 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1275 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1277 struct net_device *dev = dst->dev;
1278 unsigned int mtu = dst_mtu(dst);
1279 struct net *net = dev_net(dev);
1281 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1283 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1284 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1287 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1288 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1289 * IPV6_MAXPLEN is also valid and means: "any MSS,
1290 * rely only on pmtu discovery"
1292 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1297 static unsigned int ip6_mtu(const struct dst_entry *dst)
1299 struct inet6_dev *idev;
1300 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1308 idev = __in6_dev_get(dst->dev);
1310 mtu = idev->cnf.mtu6;
1316 static struct dst_entry *icmp6_dst_gc_list;
1317 static DEFINE_SPINLOCK(icmp6_dst_lock);
1319 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1322 struct dst_entry *dst;
1323 struct rt6_info *rt;
1324 struct inet6_dev *idev = in6_dev_get(dev);
1325 struct net *net = dev_net(dev);
1327 if (unlikely(!idev))
1328 return ERR_PTR(-ENODEV);
1330 rt = ip6_dst_alloc(net, dev, 0, NULL);
1331 if (unlikely(!rt)) {
1333 dst = ERR_PTR(-ENOMEM);
1337 rt->dst.flags |= DST_HOST;
1338 rt->dst.output = ip6_output;
1339 atomic_set(&rt->dst.__refcnt, 1);
1340 rt->rt6i_gateway = fl6->daddr;
1341 rt->rt6i_dst.addr = fl6->daddr;
1342 rt->rt6i_dst.plen = 128;
1343 rt->rt6i_idev = idev;
1344 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1346 spin_lock_bh(&icmp6_dst_lock);
1347 rt->dst.next = icmp6_dst_gc_list;
1348 icmp6_dst_gc_list = &rt->dst;
1349 spin_unlock_bh(&icmp6_dst_lock);
1351 fib6_force_start_gc(net);
1353 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1359 int icmp6_dst_gc(void)
1361 struct dst_entry *dst, **pprev;
1364 spin_lock_bh(&icmp6_dst_lock);
1365 pprev = &icmp6_dst_gc_list;
1367 while ((dst = *pprev) != NULL) {
1368 if (!atomic_read(&dst->__refcnt)) {
1377 spin_unlock_bh(&icmp6_dst_lock);
1382 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1385 struct dst_entry *dst, **pprev;
1387 spin_lock_bh(&icmp6_dst_lock);
1388 pprev = &icmp6_dst_gc_list;
1389 while ((dst = *pprev) != NULL) {
1390 struct rt6_info *rt = (struct rt6_info *) dst;
1391 if (func(rt, arg)) {
1398 spin_unlock_bh(&icmp6_dst_lock);
1401 static int ip6_dst_gc(struct dst_ops *ops)
1403 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1404 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1405 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1406 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1407 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1408 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1411 entries = dst_entries_get_fast(ops);
1412 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1413 entries <= rt_max_size)
1416 net->ipv6.ip6_rt_gc_expire++;
1417 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1418 entries = dst_entries_get_slow(ops);
1419 if (entries < ops->gc_thresh)
1420 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1422 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1423 return entries > rt_max_size;
1430 int ip6_route_add(struct fib6_config *cfg)
1433 struct net *net = cfg->fc_nlinfo.nl_net;
1434 struct rt6_info *rt = NULL;
1435 struct net_device *dev = NULL;
1436 struct inet6_dev *idev = NULL;
1437 struct fib6_table *table;
1440 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1442 #ifndef CONFIG_IPV6_SUBTREES
1443 if (cfg->fc_src_len)
1446 if (cfg->fc_ifindex) {
1448 dev = dev_get_by_index(net, cfg->fc_ifindex);
1451 idev = in6_dev_get(dev);
1456 if (cfg->fc_metric == 0)
1457 cfg->fc_metric = IP6_RT_PRIO_USER;
1460 if (cfg->fc_nlinfo.nlh &&
1461 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1462 table = fib6_get_table(net, cfg->fc_table);
1464 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1465 table = fib6_new_table(net, cfg->fc_table);
1468 table = fib6_new_table(net, cfg->fc_table);
1474 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1481 if (cfg->fc_flags & RTF_EXPIRES)
1482 rt6_set_expires(rt, jiffies +
1483 clock_t_to_jiffies(cfg->fc_expires));
1485 rt6_clean_expires(rt);
1487 if (cfg->fc_protocol == RTPROT_UNSPEC)
1488 cfg->fc_protocol = RTPROT_BOOT;
1489 rt->rt6i_protocol = cfg->fc_protocol;
1491 addr_type = ipv6_addr_type(&cfg->fc_dst);
1493 if (addr_type & IPV6_ADDR_MULTICAST)
1494 rt->dst.input = ip6_mc_input;
1495 else if (cfg->fc_flags & RTF_LOCAL)
1496 rt->dst.input = ip6_input;
1498 rt->dst.input = ip6_forward;
1500 rt->dst.output = ip6_output;
1502 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1503 rt->rt6i_dst.plen = cfg->fc_dst_len;
1504 if (rt->rt6i_dst.plen == 128)
1505 rt->dst.flags |= DST_HOST;
1507 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1508 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1513 dst_init_metrics(&rt->dst, metrics, 0);
1515 #ifdef CONFIG_IPV6_SUBTREES
1516 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1517 rt->rt6i_src.plen = cfg->fc_src_len;
1520 rt->rt6i_metric = cfg->fc_metric;
1522 /* We cannot add true routes via loopback here,
1523 they would result in kernel looping; promote them to reject routes
1525 if ((cfg->fc_flags & RTF_REJECT) ||
1526 (dev && (dev->flags & IFF_LOOPBACK) &&
1527 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1528 !(cfg->fc_flags & RTF_LOCAL))) {
1529 /* hold loopback dev/idev if we haven't done so. */
1530 if (dev != net->loopback_dev) {
1535 dev = net->loopback_dev;
1537 idev = in6_dev_get(dev);
1543 rt->dst.output = ip6_pkt_discard_out;
1544 rt->dst.input = ip6_pkt_discard;
1545 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1546 switch (cfg->fc_type) {
1548 rt->dst.error = -EINVAL;
1551 rt->dst.error = -EACCES;
1554 rt->dst.error = -EAGAIN;
1557 rt->dst.error = -ENETUNREACH;
1563 if (cfg->fc_flags & RTF_GATEWAY) {
1564 const struct in6_addr *gw_addr;
1567 gw_addr = &cfg->fc_gateway;
1568 rt->rt6i_gateway = *gw_addr;
1569 gwa_type = ipv6_addr_type(gw_addr);
1571 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1572 struct rt6_info *grt;
1574 /* IPv6 strictly inhibits using not link-local
1575 addresses as nexthop address.
1576 Otherwise, router will not able to send redirects.
1577 It is very good, but in some (rare!) circumstances
1578 (SIT, PtP, NBMA NOARP links) it is handy to allow
1579 some exceptions. --ANK
1582 if (!(gwa_type & IPV6_ADDR_UNICAST))
1585 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1587 err = -EHOSTUNREACH;
1591 if (dev != grt->dst.dev) {
1597 idev = grt->rt6i_idev;
1599 in6_dev_hold(grt->rt6i_idev);
1601 if (!(grt->rt6i_flags & RTF_GATEWAY))
1609 if (!dev || (dev->flags & IFF_LOOPBACK))
1617 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1618 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1622 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1623 rt->rt6i_prefsrc.plen = 128;
1625 rt->rt6i_prefsrc.plen = 0;
1627 rt->rt6i_flags = cfg->fc_flags;
1634 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1635 int type = nla_type(nla);
1638 if (type > RTAX_MAX) {
1643 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1649 rt->rt6i_idev = idev;
1650 rt->rt6i_table = table;
1652 cfg->fc_nlinfo.nl_net = dev_net(dev);
1654 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1666 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1669 struct fib6_table *table;
1670 struct net *net = dev_net(rt->dst.dev);
1672 if (rt == net->ipv6.ip6_null_entry) {
1677 table = rt->rt6i_table;
1678 write_lock_bh(&table->tb6_lock);
1679 err = fib6_del(rt, info);
1680 write_unlock_bh(&table->tb6_lock);
1687 int ip6_del_rt(struct rt6_info *rt)
1689 struct nl_info info = {
1690 .nl_net = dev_net(rt->dst.dev),
1692 return __ip6_del_rt(rt, &info);
1695 static int ip6_route_del(struct fib6_config *cfg)
1697 struct fib6_table *table;
1698 struct fib6_node *fn;
1699 struct rt6_info *rt;
1702 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1706 read_lock_bh(&table->tb6_lock);
1708 fn = fib6_locate(&table->tb6_root,
1709 &cfg->fc_dst, cfg->fc_dst_len,
1710 &cfg->fc_src, cfg->fc_src_len);
1713 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1714 if (cfg->fc_ifindex &&
1716 rt->dst.dev->ifindex != cfg->fc_ifindex))
1718 if (cfg->fc_flags & RTF_GATEWAY &&
1719 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1721 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1724 read_unlock_bh(&table->tb6_lock);
1726 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1729 read_unlock_bh(&table->tb6_lock);
1734 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1736 struct net *net = dev_net(skb->dev);
1737 struct netevent_redirect netevent;
1738 struct rt6_info *rt, *nrt = NULL;
1739 struct ndisc_options ndopts;
1740 struct inet6_dev *in6_dev;
1741 struct neighbour *neigh;
1743 int optlen, on_link;
1746 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1747 optlen -= sizeof(*msg);
1750 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1754 msg = (struct rd_msg *)icmp6_hdr(skb);
1756 if (ipv6_addr_is_multicast(&msg->dest)) {
1757 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1762 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1764 } else if (ipv6_addr_type(&msg->target) !=
1765 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1766 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1770 in6_dev = __in6_dev_get(skb->dev);
1773 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1777 * The IP source address of the Redirect MUST be the same as the current
1778 * first-hop router for the specified ICMP Destination Address.
1781 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1782 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1787 if (ndopts.nd_opts_tgt_lladdr) {
1788 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1791 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1796 rt = (struct rt6_info *) dst;
1797 if (rt == net->ipv6.ip6_null_entry) {
1798 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1802 /* Redirect received -> path was valid.
1803 * Look, redirects are sent only in response to data packets,
1804 * so that this nexthop apparently is reachable. --ANK
1806 dst_confirm(&rt->dst);
1808 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1813 * We have finally decided to accept it.
1816 neigh_update(neigh, lladdr, NUD_STALE,
1817 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1818 NEIGH_UPDATE_F_OVERRIDE|
1819 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1820 NEIGH_UPDATE_F_ISROUTER))
1823 nrt = ip6_rt_copy(rt, &msg->dest);
1827 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1829 nrt->rt6i_flags &= ~RTF_GATEWAY;
1831 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1833 if (ip6_ins_rt(nrt))
1836 netevent.old = &rt->dst;
1837 netevent.new = &nrt->dst;
1838 netevent.daddr = &msg->dest;
1839 netevent.neigh = neigh;
1840 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1842 if (rt->rt6i_flags & RTF_CACHE) {
1843 rt = (struct rt6_info *) dst_clone(&rt->dst);
1848 neigh_release(neigh);
1852 * Misc support functions
1855 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1856 const struct in6_addr *dest)
1858 struct net *net = dev_net(ort->dst.dev);
1859 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1863 rt->dst.input = ort->dst.input;
1864 rt->dst.output = ort->dst.output;
1865 rt->dst.flags |= DST_HOST;
1867 rt->rt6i_dst.addr = *dest;
1868 rt->rt6i_dst.plen = 128;
1869 dst_copy_metrics(&rt->dst, &ort->dst);
1870 rt->dst.error = ort->dst.error;
1871 rt->rt6i_idev = ort->rt6i_idev;
1873 in6_dev_hold(rt->rt6i_idev);
1874 rt->dst.lastuse = jiffies;
1876 if (ort->rt6i_flags & RTF_GATEWAY)
1877 rt->rt6i_gateway = ort->rt6i_gateway;
1879 rt->rt6i_gateway = *dest;
1880 rt->rt6i_flags = ort->rt6i_flags;
1881 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1882 (RTF_DEFAULT | RTF_ADDRCONF))
1883 rt6_set_from(rt, ort);
1884 rt->rt6i_metric = 0;
1886 #ifdef CONFIG_IPV6_SUBTREES
1887 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1889 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1890 rt->rt6i_table = ort->rt6i_table;
1895 #ifdef CONFIG_IPV6_ROUTE_INFO
1896 static struct rt6_info *rt6_get_route_info(struct net *net,
1897 const struct in6_addr *prefix, int prefixlen,
1898 const struct in6_addr *gwaddr, int ifindex)
1900 struct fib6_node *fn;
1901 struct rt6_info *rt = NULL;
1902 struct fib6_table *table;
1904 table = fib6_get_table(net, RT6_TABLE_INFO);
1908 read_lock_bh(&table->tb6_lock);
1909 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1913 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1914 if (rt->dst.dev->ifindex != ifindex)
1916 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1918 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1924 read_unlock_bh(&table->tb6_lock);
1928 static struct rt6_info *rt6_add_route_info(struct net *net,
1929 const struct in6_addr *prefix, int prefixlen,
1930 const struct in6_addr *gwaddr, int ifindex,
1933 struct fib6_config cfg = {
1934 .fc_table = RT6_TABLE_INFO,
1935 .fc_metric = IP6_RT_PRIO_USER,
1936 .fc_ifindex = ifindex,
1937 .fc_dst_len = prefixlen,
1938 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1939 RTF_UP | RTF_PREF(pref),
1940 .fc_nlinfo.portid = 0,
1941 .fc_nlinfo.nlh = NULL,
1942 .fc_nlinfo.nl_net = net,
1945 cfg.fc_dst = *prefix;
1946 cfg.fc_gateway = *gwaddr;
1948 /* We should treat it as a default route if prefix length is 0. */
1950 cfg.fc_flags |= RTF_DEFAULT;
1952 ip6_route_add(&cfg);
1954 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1958 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1960 struct rt6_info *rt;
1961 struct fib6_table *table;
1963 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1967 read_lock_bh(&table->tb6_lock);
1968 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1969 if (dev == rt->dst.dev &&
1970 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1971 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1976 read_unlock_bh(&table->tb6_lock);
1980 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1981 struct net_device *dev,
1984 struct fib6_config cfg = {
1985 .fc_table = RT6_TABLE_DFLT,
1986 .fc_metric = IP6_RT_PRIO_USER,
1987 .fc_ifindex = dev->ifindex,
1988 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1989 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1990 .fc_nlinfo.portid = 0,
1991 .fc_nlinfo.nlh = NULL,
1992 .fc_nlinfo.nl_net = dev_net(dev),
1995 cfg.fc_gateway = *gwaddr;
1997 ip6_route_add(&cfg);
1999 return rt6_get_dflt_router(gwaddr, dev);
2002 void rt6_purge_dflt_routers(struct net *net)
2004 struct rt6_info *rt;
2005 struct fib6_table *table;
2007 /* NOTE: Keep consistent with rt6_get_dflt_router */
2008 table = fib6_get_table(net, RT6_TABLE_DFLT);
2013 read_lock_bh(&table->tb6_lock);
2014 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2015 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2016 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2018 read_unlock_bh(&table->tb6_lock);
2023 read_unlock_bh(&table->tb6_lock);
2026 static void rtmsg_to_fib6_config(struct net *net,
2027 struct in6_rtmsg *rtmsg,
2028 struct fib6_config *cfg)
2030 memset(cfg, 0, sizeof(*cfg));
2032 cfg->fc_table = RT6_TABLE_MAIN;
2033 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2034 cfg->fc_metric = rtmsg->rtmsg_metric;
2035 cfg->fc_expires = rtmsg->rtmsg_info;
2036 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2037 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2038 cfg->fc_flags = rtmsg->rtmsg_flags;
2040 cfg->fc_nlinfo.nl_net = net;
2042 cfg->fc_dst = rtmsg->rtmsg_dst;
2043 cfg->fc_src = rtmsg->rtmsg_src;
2044 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2047 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2049 struct fib6_config cfg;
2050 struct in6_rtmsg rtmsg;
2054 case SIOCADDRT: /* Add a route */
2055 case SIOCDELRT: /* Delete a route */
2056 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2058 err = copy_from_user(&rtmsg, arg,
2059 sizeof(struct in6_rtmsg));
2063 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2068 err = ip6_route_add(&cfg);
2071 err = ip6_route_del(&cfg);
2085 * Drop the packet on the floor
2088 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2091 struct dst_entry *dst = skb_dst(skb);
2092 switch (ipstats_mib_noroutes) {
2093 case IPSTATS_MIB_INNOROUTES:
2094 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2095 if (type == IPV6_ADDR_ANY) {
2096 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2097 IPSTATS_MIB_INADDRERRORS);
2101 case IPSTATS_MIB_OUTNOROUTES:
2102 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2103 ipstats_mib_noroutes);
2106 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2111 static int ip6_pkt_discard(struct sk_buff *skb)
2113 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2116 static int ip6_pkt_discard_out(struct sk_buff *skb)
2118 skb->dev = skb_dst(skb)->dev;
2119 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2122 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2124 static int ip6_pkt_prohibit(struct sk_buff *skb)
2126 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2129 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2131 skb->dev = skb_dst(skb)->dev;
2132 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2138 * Allocate a dst for local (unicast / anycast) address.
2141 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2142 const struct in6_addr *addr,
2145 struct net *net = dev_net(idev->dev);
2146 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2149 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2150 return ERR_PTR(-ENOMEM);
2155 rt->dst.flags |= DST_HOST;
2156 rt->dst.input = ip6_input;
2157 rt->dst.output = ip6_output;
2158 rt->rt6i_idev = idev;
2160 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2162 rt->rt6i_flags |= RTF_ANYCAST;
2164 rt->rt6i_flags |= RTF_LOCAL;
2166 rt->rt6i_gateway = *addr;
2167 rt->rt6i_dst.addr = *addr;
2168 rt->rt6i_dst.plen = 128;
2169 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2171 atomic_set(&rt->dst.__refcnt, 1);
2176 int ip6_route_get_saddr(struct net *net,
2177 struct rt6_info *rt,
2178 const struct in6_addr *daddr,
2180 struct in6_addr *saddr)
2182 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2184 if (rt->rt6i_prefsrc.plen)
2185 *saddr = rt->rt6i_prefsrc.addr;
2187 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2188 daddr, prefs, saddr);
2192 /* remove deleted ip from prefsrc entries */
2193 struct arg_dev_net_ip {
2194 struct net_device *dev;
2196 struct in6_addr *addr;
2199 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2201 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2202 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2203 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2205 if (((void *)rt->dst.dev == dev || !dev) &&
2206 rt != net->ipv6.ip6_null_entry &&
2207 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2208 /* remove prefsrc entry */
2209 rt->rt6i_prefsrc.plen = 0;
2214 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2216 struct net *net = dev_net(ifp->idev->dev);
2217 struct arg_dev_net_ip adni = {
2218 .dev = ifp->idev->dev,
2222 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2225 struct arg_dev_net {
2226 struct net_device *dev;
2230 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2232 const struct arg_dev_net *adn = arg;
2233 const struct net_device *dev = adn->dev;
2235 if ((rt->dst.dev == dev || !dev) &&
2236 rt != adn->net->ipv6.ip6_null_entry)
2242 void rt6_ifdown(struct net *net, struct net_device *dev)
2244 struct arg_dev_net adn = {
2249 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2250 icmp6_clean_all(fib6_ifdown, &adn);
2253 struct rt6_mtu_change_arg {
2254 struct net_device *dev;
2258 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2260 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2261 struct inet6_dev *idev;
2263 /* In IPv6 pmtu discovery is not optional,
2264 so that RTAX_MTU lock cannot disable it.
2265 We still use this lock to block changes
2266 caused by addrconf/ndisc.
2269 idev = __in6_dev_get(arg->dev);
2273 /* For administrative MTU increase, there is no way to discover
2274 IPv6 PMTU increase, so PMTU increase should be updated here.
2275 Since RFC 1981 doesn't include administrative MTU increase
2276 update PMTU increase is a MUST. (i.e. jumbo frame)
2279 If new MTU is less than route PMTU, this new MTU will be the
2280 lowest MTU in the path, update the route PMTU to reflect PMTU
2281 decreases; if new MTU is greater than route PMTU, and the
2282 old MTU is the lowest MTU in the path, update the route PMTU
2283 to reflect the increase. In this case if the other nodes' MTU
2284 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2287 if (rt->dst.dev == arg->dev &&
2288 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2289 (dst_mtu(&rt->dst) >= arg->mtu ||
2290 (dst_mtu(&rt->dst) < arg->mtu &&
2291 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2292 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2297 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2299 struct rt6_mtu_change_arg arg = {
2304 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2307 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2308 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2309 [RTA_OIF] = { .type = NLA_U32 },
2310 [RTA_IIF] = { .type = NLA_U32 },
2311 [RTA_PRIORITY] = { .type = NLA_U32 },
2312 [RTA_METRICS] = { .type = NLA_NESTED },
2313 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2316 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2317 struct fib6_config *cfg)
2320 struct nlattr *tb[RTA_MAX+1];
2323 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2328 rtm = nlmsg_data(nlh);
2329 memset(cfg, 0, sizeof(*cfg));
2331 cfg->fc_table = rtm->rtm_table;
2332 cfg->fc_dst_len = rtm->rtm_dst_len;
2333 cfg->fc_src_len = rtm->rtm_src_len;
2334 cfg->fc_flags = RTF_UP;
2335 cfg->fc_protocol = rtm->rtm_protocol;
2336 cfg->fc_type = rtm->rtm_type;
2338 if (rtm->rtm_type == RTN_UNREACHABLE ||
2339 rtm->rtm_type == RTN_BLACKHOLE ||
2340 rtm->rtm_type == RTN_PROHIBIT ||
2341 rtm->rtm_type == RTN_THROW)
2342 cfg->fc_flags |= RTF_REJECT;
2344 if (rtm->rtm_type == RTN_LOCAL)
2345 cfg->fc_flags |= RTF_LOCAL;
2347 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2348 cfg->fc_nlinfo.nlh = nlh;
2349 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2351 if (tb[RTA_GATEWAY]) {
2352 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2353 cfg->fc_flags |= RTF_GATEWAY;
2357 int plen = (rtm->rtm_dst_len + 7) >> 3;
2359 if (nla_len(tb[RTA_DST]) < plen)
2362 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2366 int plen = (rtm->rtm_src_len + 7) >> 3;
2368 if (nla_len(tb[RTA_SRC]) < plen)
2371 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2374 if (tb[RTA_PREFSRC])
2375 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2378 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2380 if (tb[RTA_PRIORITY])
2381 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2383 if (tb[RTA_METRICS]) {
2384 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2385 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2389 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2391 if (tb[RTA_MULTIPATH]) {
2392 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2393 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2401 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2403 struct fib6_config r_cfg;
2404 struct rtnexthop *rtnh;
2407 int err = 0, last_err = 0;
2410 rtnh = (struct rtnexthop *)cfg->fc_mp;
2411 remaining = cfg->fc_mp_len;
2413 /* Parse a Multipath Entry */
2414 while (rtnh_ok(rtnh, remaining)) {
2415 memcpy(&r_cfg, cfg, sizeof(*cfg));
2416 if (rtnh->rtnh_ifindex)
2417 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2419 attrlen = rtnh_attrlen(rtnh);
2421 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2423 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2425 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2426 r_cfg.fc_flags |= RTF_GATEWAY;
2429 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2432 /* If we are trying to remove a route, do not stop the
2433 * loop when ip6_route_del() fails (because next hop is
2434 * already gone), we should try to remove all next hops.
2437 /* If add fails, we should try to delete all
2438 * next hops that have been already added.
2444 /* Because each route is added like a single route we remove
2445 * this flag after the first nexthop (if there is a collision,
2446 * we have already fail to add the first nexthop:
2447 * fib6_add_rt2node() has reject it).
2449 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2450 rtnh = rtnh_next(rtnh, &remaining);
2456 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2458 struct fib6_config cfg;
2461 err = rtm_to_fib6_config(skb, nlh, &cfg);
2466 return ip6_route_multipath(&cfg, 0);
2468 return ip6_route_del(&cfg);
2471 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2473 struct fib6_config cfg;
2476 err = rtm_to_fib6_config(skb, nlh, &cfg);
2481 return ip6_route_multipath(&cfg, 1);
2483 return ip6_route_add(&cfg);
2486 static inline size_t rt6_nlmsg_size(void)
2488 return NLMSG_ALIGN(sizeof(struct rtmsg))
2489 + nla_total_size(16) /* RTA_SRC */
2490 + nla_total_size(16) /* RTA_DST */
2491 + nla_total_size(16) /* RTA_GATEWAY */
2492 + nla_total_size(16) /* RTA_PREFSRC */
2493 + nla_total_size(4) /* RTA_TABLE */
2494 + nla_total_size(4) /* RTA_IIF */
2495 + nla_total_size(4) /* RTA_OIF */
2496 + nla_total_size(4) /* RTA_PRIORITY */
2497 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2498 + nla_total_size(sizeof(struct rta_cacheinfo));
2501 static int rt6_fill_node(struct net *net,
2502 struct sk_buff *skb, struct rt6_info *rt,
2503 struct in6_addr *dst, struct in6_addr *src,
2504 int iif, int type, u32 portid, u32 seq,
2505 int prefix, int nowait, unsigned int flags)
2508 struct nlmsghdr *nlh;
2512 if (prefix) { /* user wants prefix routes only */
2513 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2514 /* success since this is not a prefix route */
2519 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2523 rtm = nlmsg_data(nlh);
2524 rtm->rtm_family = AF_INET6;
2525 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2526 rtm->rtm_src_len = rt->rt6i_src.plen;
2529 table = rt->rt6i_table->tb6_id;
2531 table = RT6_TABLE_UNSPEC;
2532 rtm->rtm_table = table;
2533 if (nla_put_u32(skb, RTA_TABLE, table))
2534 goto nla_put_failure;
2535 if (rt->rt6i_flags & RTF_REJECT) {
2536 switch (rt->dst.error) {
2538 rtm->rtm_type = RTN_BLACKHOLE;
2541 rtm->rtm_type = RTN_PROHIBIT;
2544 rtm->rtm_type = RTN_THROW;
2547 rtm->rtm_type = RTN_UNREACHABLE;
2551 else if (rt->rt6i_flags & RTF_LOCAL)
2552 rtm->rtm_type = RTN_LOCAL;
2553 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2554 rtm->rtm_type = RTN_LOCAL;
2556 rtm->rtm_type = RTN_UNICAST;
2558 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2559 rtm->rtm_protocol = rt->rt6i_protocol;
2560 if (rt->rt6i_flags & RTF_DYNAMIC)
2561 rtm->rtm_protocol = RTPROT_REDIRECT;
2562 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2563 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2564 rtm->rtm_protocol = RTPROT_RA;
2566 rtm->rtm_protocol = RTPROT_KERNEL;
2569 if (rt->rt6i_flags & RTF_CACHE)
2570 rtm->rtm_flags |= RTM_F_CLONED;
2573 if (nla_put(skb, RTA_DST, 16, dst))
2574 goto nla_put_failure;
2575 rtm->rtm_dst_len = 128;
2576 } else if (rtm->rtm_dst_len)
2577 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2578 goto nla_put_failure;
2579 #ifdef CONFIG_IPV6_SUBTREES
2581 if (nla_put(skb, RTA_SRC, 16, src))
2582 goto nla_put_failure;
2583 rtm->rtm_src_len = 128;
2584 } else if (rtm->rtm_src_len &&
2585 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2586 goto nla_put_failure;
2589 #ifdef CONFIG_IPV6_MROUTE
2590 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2591 int err = ip6mr_get_route(net, skb, rtm, nowait);
2596 goto nla_put_failure;
2598 if (err == -EMSGSIZE)
2599 goto nla_put_failure;
2604 if (nla_put_u32(skb, RTA_IIF, iif))
2605 goto nla_put_failure;
2607 struct in6_addr saddr_buf;
2608 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2609 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2610 goto nla_put_failure;
2613 if (rt->rt6i_prefsrc.plen) {
2614 struct in6_addr saddr_buf;
2615 saddr_buf = rt->rt6i_prefsrc.addr;
2616 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2617 goto nla_put_failure;
2620 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2621 goto nla_put_failure;
2623 if (rt->rt6i_flags & RTF_GATEWAY) {
2624 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2625 goto nla_put_failure;
2629 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2630 goto nla_put_failure;
2631 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2632 goto nla_put_failure;
2634 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2636 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2637 goto nla_put_failure;
2639 return nlmsg_end(skb, nlh);
2642 nlmsg_cancel(skb, nlh);
2646 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2648 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2651 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2652 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2653 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2657 return rt6_fill_node(arg->net,
2658 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2659 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2660 prefix, 0, NLM_F_MULTI);
2663 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2665 struct net *net = sock_net(in_skb->sk);
2666 struct nlattr *tb[RTA_MAX+1];
2667 struct rt6_info *rt;
2668 struct sk_buff *skb;
2671 int err, iif = 0, oif = 0;
2673 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2678 memset(&fl6, 0, sizeof(fl6));
2681 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2684 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2688 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2691 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2695 iif = nla_get_u32(tb[RTA_IIF]);
2698 oif = nla_get_u32(tb[RTA_OIF]);
2701 struct net_device *dev;
2704 dev = __dev_get_by_index(net, iif);
2710 fl6.flowi6_iif = iif;
2712 if (!ipv6_addr_any(&fl6.saddr))
2713 flags |= RT6_LOOKUP_F_HAS_SADDR;
2715 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2718 fl6.flowi6_oif = oif;
2720 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2723 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2730 /* Reserve room for dummy headers, this skb can pass
2731 through good chunk of routing engine.
2733 skb_reset_mac_header(skb);
2734 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2736 skb_dst_set(skb, &rt->dst);
2738 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2739 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2740 nlh->nlmsg_seq, 0, 0, 0);
2746 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2751 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2753 struct sk_buff *skb;
2754 struct net *net = info->nl_net;
2759 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2761 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2765 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2766 event, info->portid, seq, 0, 0, 0);
2768 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2769 WARN_ON(err == -EMSGSIZE);
2773 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2774 info->nlh, gfp_any());
2778 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2781 static int ip6_route_dev_notify(struct notifier_block *this,
2782 unsigned long event, void *ptr)
2784 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2785 struct net *net = dev_net(dev);
2787 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2788 net->ipv6.ip6_null_entry->dst.dev = dev;
2789 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2790 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2791 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2792 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2793 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2794 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2805 #ifdef CONFIG_PROC_FS
2816 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2818 struct seq_file *m = p_arg;
2820 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2822 #ifdef CONFIG_IPV6_SUBTREES
2823 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2825 seq_puts(m, "00000000000000000000000000000000 00 ");
2827 if (rt->rt6i_flags & RTF_GATEWAY) {
2828 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2830 seq_puts(m, "00000000000000000000000000000000");
2832 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2833 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2834 rt->dst.__use, rt->rt6i_flags,
2835 rt->dst.dev ? rt->dst.dev->name : "");
2839 static int ipv6_route_show(struct seq_file *m, void *v)
2841 struct net *net = (struct net *)m->private;
2842 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2846 static int ipv6_route_open(struct inode *inode, struct file *file)
2848 return single_open_net(inode, file, ipv6_route_show);
2851 static const struct file_operations ipv6_route_proc_fops = {
2852 .owner = THIS_MODULE,
2853 .open = ipv6_route_open,
2855 .llseek = seq_lseek,
2856 .release = single_release_net,
2859 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2861 struct net *net = (struct net *)seq->private;
2862 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2863 net->ipv6.rt6_stats->fib_nodes,
2864 net->ipv6.rt6_stats->fib_route_nodes,
2865 net->ipv6.rt6_stats->fib_rt_alloc,
2866 net->ipv6.rt6_stats->fib_rt_entries,
2867 net->ipv6.rt6_stats->fib_rt_cache,
2868 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2869 net->ipv6.rt6_stats->fib_discarded_routes);
2874 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2876 return single_open_net(inode, file, rt6_stats_seq_show);
2879 static const struct file_operations rt6_stats_seq_fops = {
2880 .owner = THIS_MODULE,
2881 .open = rt6_stats_seq_open,
2883 .llseek = seq_lseek,
2884 .release = single_release_net,
2886 #endif /* CONFIG_PROC_FS */
2888 #ifdef CONFIG_SYSCTL
2891 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2892 void __user *buffer, size_t *lenp, loff_t *ppos)
2899 net = (struct net *)ctl->extra1;
2900 delay = net->ipv6.sysctl.flush_delay;
2901 proc_dointvec(ctl, write, buffer, lenp, ppos);
2902 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2906 struct ctl_table ipv6_route_table_template[] = {
2908 .procname = "flush",
2909 .data = &init_net.ipv6.sysctl.flush_delay,
2910 .maxlen = sizeof(int),
2912 .proc_handler = ipv6_sysctl_rtcache_flush
2915 .procname = "gc_thresh",
2916 .data = &ip6_dst_ops_template.gc_thresh,
2917 .maxlen = sizeof(int),
2919 .proc_handler = proc_dointvec,
2922 .procname = "max_size",
2923 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2924 .maxlen = sizeof(int),
2926 .proc_handler = proc_dointvec,
2929 .procname = "gc_min_interval",
2930 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2931 .maxlen = sizeof(int),
2933 .proc_handler = proc_dointvec_jiffies,
2936 .procname = "gc_timeout",
2937 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2938 .maxlen = sizeof(int),
2940 .proc_handler = proc_dointvec_jiffies,
2943 .procname = "gc_interval",
2944 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2945 .maxlen = sizeof(int),
2947 .proc_handler = proc_dointvec_jiffies,
2950 .procname = "gc_elasticity",
2951 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2952 .maxlen = sizeof(int),
2954 .proc_handler = proc_dointvec,
2957 .procname = "mtu_expires",
2958 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2959 .maxlen = sizeof(int),
2961 .proc_handler = proc_dointvec_jiffies,
2964 .procname = "min_adv_mss",
2965 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2966 .maxlen = sizeof(int),
2968 .proc_handler = proc_dointvec,
2971 .procname = "gc_min_interval_ms",
2972 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2973 .maxlen = sizeof(int),
2975 .proc_handler = proc_dointvec_ms_jiffies,
2980 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2982 struct ctl_table *table;
2984 table = kmemdup(ipv6_route_table_template,
2985 sizeof(ipv6_route_table_template),
2989 table[0].data = &net->ipv6.sysctl.flush_delay;
2990 table[0].extra1 = net;
2991 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2992 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2993 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2994 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2995 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2996 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2997 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2998 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2999 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3001 /* Don't export sysctls to unprivileged users */
3002 if (net->user_ns != &init_user_ns)
3003 table[0].procname = NULL;
3010 static int __net_init ip6_route_net_init(struct net *net)
3014 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3015 sizeof(net->ipv6.ip6_dst_ops));
3017 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3018 goto out_ip6_dst_ops;
3020 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3021 sizeof(*net->ipv6.ip6_null_entry),
3023 if (!net->ipv6.ip6_null_entry)
3024 goto out_ip6_dst_entries;
3025 net->ipv6.ip6_null_entry->dst.path =
3026 (struct dst_entry *)net->ipv6.ip6_null_entry;
3027 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3028 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3029 ip6_template_metrics, true);
3031 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3032 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3033 sizeof(*net->ipv6.ip6_prohibit_entry),
3035 if (!net->ipv6.ip6_prohibit_entry)
3036 goto out_ip6_null_entry;
3037 net->ipv6.ip6_prohibit_entry->dst.path =
3038 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3039 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3040 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3041 ip6_template_metrics, true);
3043 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3044 sizeof(*net->ipv6.ip6_blk_hole_entry),
3046 if (!net->ipv6.ip6_blk_hole_entry)
3047 goto out_ip6_prohibit_entry;
3048 net->ipv6.ip6_blk_hole_entry->dst.path =
3049 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3050 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3051 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3052 ip6_template_metrics, true);
3055 net->ipv6.sysctl.flush_delay = 0;
3056 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3057 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3058 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3059 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3060 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3061 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3062 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3064 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3070 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3071 out_ip6_prohibit_entry:
3072 kfree(net->ipv6.ip6_prohibit_entry);
3074 kfree(net->ipv6.ip6_null_entry);
3076 out_ip6_dst_entries:
3077 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3082 static void __net_exit ip6_route_net_exit(struct net *net)
3084 kfree(net->ipv6.ip6_null_entry);
3085 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3086 kfree(net->ipv6.ip6_prohibit_entry);
3087 kfree(net->ipv6.ip6_blk_hole_entry);
3089 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3092 static int __net_init ip6_route_net_init_late(struct net *net)
3094 #ifdef CONFIG_PROC_FS
3095 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3096 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3101 static void __net_exit ip6_route_net_exit_late(struct net *net)
3103 #ifdef CONFIG_PROC_FS
3104 remove_proc_entry("ipv6_route", net->proc_net);
3105 remove_proc_entry("rt6_stats", net->proc_net);
3109 static struct pernet_operations ip6_route_net_ops = {
3110 .init = ip6_route_net_init,
3111 .exit = ip6_route_net_exit,
3114 static int __net_init ipv6_inetpeer_init(struct net *net)
3116 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3120 inet_peer_base_init(bp);
3121 net->ipv6.peers = bp;
3125 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3127 struct inet_peer_base *bp = net->ipv6.peers;
3129 net->ipv6.peers = NULL;
3130 inetpeer_invalidate_tree(bp);
3134 static struct pernet_operations ipv6_inetpeer_ops = {
3135 .init = ipv6_inetpeer_init,
3136 .exit = ipv6_inetpeer_exit,
3139 static struct pernet_operations ip6_route_net_late_ops = {
3140 .init = ip6_route_net_init_late,
3141 .exit = ip6_route_net_exit_late,
3144 static struct notifier_block ip6_route_dev_notifier = {
3145 .notifier_call = ip6_route_dev_notify,
3149 int __init ip6_route_init(void)
3154 ip6_dst_ops_template.kmem_cachep =
3155 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3156 SLAB_HWCACHE_ALIGN, NULL);
3157 if (!ip6_dst_ops_template.kmem_cachep)
3160 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3162 goto out_kmem_cache;
3164 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3166 goto out_dst_entries;
3168 ret = register_pernet_subsys(&ip6_route_net_ops);
3170 goto out_register_inetpeer;
3172 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3174 /* Registering of the loopback is done before this portion of code,
3175 * the loopback reference in rt6_info will not be taken, do it
3176 * manually for init_net */
3177 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3178 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3179 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3180 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3181 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3182 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3183 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3187 goto out_register_subsys;
3193 ret = fib6_rules_init();
3197 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3199 goto fib6_rules_init;
3202 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3203 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3204 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3205 goto out_register_late_subsys;
3207 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3209 goto out_register_late_subsys;
3214 out_register_late_subsys:
3215 unregister_pernet_subsys(&ip6_route_net_late_ops);
3217 fib6_rules_cleanup();
3222 out_register_subsys:
3223 unregister_pernet_subsys(&ip6_route_net_ops);
3224 out_register_inetpeer:
3225 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3227 dst_entries_destroy(&ip6_dst_blackhole_ops);
3229 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3233 void ip6_route_cleanup(void)
3235 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3236 unregister_pernet_subsys(&ip6_route_net_late_ops);
3237 fib6_rules_cleanup();
3240 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3241 unregister_pernet_subsys(&ip6_route_net_ops);
3242 dst_entries_destroy(&ip6_dst_blackhole_ops);
3243 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);