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 = -3,
70 RT6_NUD_FAIL_PROBE = -2,
71 RT6_NUD_FAIL_DO_RR = -1,
75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
76 const struct in6_addr *dest);
77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
79 static unsigned int ip6_mtu(const struct dst_entry *dst);
80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
81 static void ip6_dst_destroy(struct dst_entry *);
82 static void ip6_dst_ifdown(struct dst_entry *,
83 struct net_device *dev, int how);
84 static int ip6_dst_gc(struct dst_ops *ops);
86 static int ip6_pkt_discard(struct sk_buff *skb);
87 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
88 static int ip6_pkt_prohibit(struct sk_buff *skb);
89 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
90 static void ip6_link_failure(struct sk_buff *skb);
91 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
92 struct sk_buff *skb, u32 mtu);
93 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
97 #ifdef CONFIG_IPV6_ROUTE_INFO
98 static struct rt6_info *rt6_add_route_info(struct net *net,
99 const struct in6_addr *prefix, int prefixlen,
100 const struct in6_addr *gwaddr, int ifindex,
102 static struct rt6_info *rt6_get_route_info(struct net *net,
103 const struct in6_addr *prefix, int prefixlen,
104 const struct in6_addr *gwaddr, int ifindex);
107 static void rt6_bind_peer(struct rt6_info *rt, int create)
109 struct inet_peer_base *base;
110 struct inet_peer *peer;
112 base = inetpeer_base_ptr(rt->_rt6i_peer);
116 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
118 if (!rt6_set_peer(rt, peer))
123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
125 if (rt6_has_peer(rt))
126 return rt6_peer_ptr(rt);
128 rt6_bind_peer(rt, create);
129 return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
134 return __rt6_get_peer(rt, 1);
137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
139 struct rt6_info *rt = (struct rt6_info *) dst;
140 struct inet_peer *peer;
143 if (!(rt->dst.flags & DST_HOST))
146 peer = rt6_get_peer_create(rt);
148 u32 *old_p = __DST_METRICS_PTR(old);
149 unsigned long prev, new;
152 if (inet_metrics_new(peer) ||
153 (old & DST_METRICS_FORCE_OVERWRITE))
154 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
156 new = (unsigned long) p;
157 prev = cmpxchg(&dst->_metrics, old, new);
160 p = __DST_METRICS_PTR(prev);
161 if (prev & DST_METRICS_READ_ONLY)
168 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
172 struct in6_addr *p = &rt->rt6i_gateway;
174 if (!ipv6_addr_any(p))
175 return (const void *) p;
177 return &ipv6_hdr(skb)->daddr;
181 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
185 struct rt6_info *rt = (struct rt6_info *) dst;
188 daddr = choose_neigh_daddr(rt, skb, daddr);
189 n = __ipv6_neigh_lookup(dst->dev, daddr);
192 return neigh_create(&nd_tbl, daddr, dst->dev);
195 static struct dst_ops ip6_dst_ops_template = {
197 .protocol = cpu_to_be16(ETH_P_IPV6),
200 .check = ip6_dst_check,
201 .default_advmss = ip6_default_advmss,
203 .cow_metrics = ipv6_cow_metrics,
204 .destroy = ip6_dst_destroy,
205 .ifdown = ip6_dst_ifdown,
206 .negative_advice = ip6_negative_advice,
207 .link_failure = ip6_link_failure,
208 .update_pmtu = ip6_rt_update_pmtu,
209 .redirect = rt6_do_redirect,
210 .local_out = __ip6_local_out,
211 .neigh_lookup = ip6_neigh_lookup,
214 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
216 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
218 return mtu ? : dst->dev->mtu;
221 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
222 struct sk_buff *skb, u32 mtu)
226 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
231 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
237 static struct dst_ops ip6_dst_blackhole_ops = {
239 .protocol = cpu_to_be16(ETH_P_IPV6),
240 .destroy = ip6_dst_destroy,
241 .check = ip6_dst_check,
242 .mtu = ip6_blackhole_mtu,
243 .default_advmss = ip6_default_advmss,
244 .update_pmtu = ip6_rt_blackhole_update_pmtu,
245 .redirect = ip6_rt_blackhole_redirect,
246 .cow_metrics = ip6_rt_blackhole_cow_metrics,
247 .neigh_lookup = ip6_neigh_lookup,
250 static const u32 ip6_template_metrics[RTAX_MAX] = {
251 [RTAX_HOPLIMIT - 1] = 0,
254 static const struct rt6_info ip6_null_entry_template = {
256 .__refcnt = ATOMIC_INIT(1),
258 .obsolete = DST_OBSOLETE_FORCE_CHK,
259 .error = -ENETUNREACH,
260 .input = ip6_pkt_discard,
261 .output = ip6_pkt_discard_out,
263 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
264 .rt6i_protocol = RTPROT_KERNEL,
265 .rt6i_metric = ~(u32) 0,
266 .rt6i_ref = ATOMIC_INIT(1),
269 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
271 static const struct rt6_info ip6_prohibit_entry_template = {
273 .__refcnt = ATOMIC_INIT(1),
275 .obsolete = DST_OBSOLETE_FORCE_CHK,
277 .input = ip6_pkt_prohibit,
278 .output = ip6_pkt_prohibit_out,
280 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
281 .rt6i_protocol = RTPROT_KERNEL,
282 .rt6i_metric = ~(u32) 0,
283 .rt6i_ref = ATOMIC_INIT(1),
286 static const struct rt6_info ip6_blk_hole_entry_template = {
288 .__refcnt = ATOMIC_INIT(1),
290 .obsolete = DST_OBSOLETE_FORCE_CHK,
292 .input = dst_discard,
293 .output = dst_discard_sk,
295 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .rt6i_protocol = RTPROT_KERNEL,
297 .rt6i_metric = ~(u32) 0,
298 .rt6i_ref = ATOMIC_INIT(1),
303 /* allocate dst with ip6_dst_ops */
304 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
305 struct net_device *dev,
307 struct fib6_table *table)
309 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
310 0, DST_OBSOLETE_FORCE_CHK, flags);
313 struct dst_entry *dst = &rt->dst;
315 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
316 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
317 INIT_LIST_HEAD(&rt->rt6i_siblings);
322 static void ip6_dst_destroy(struct dst_entry *dst)
324 struct rt6_info *rt = (struct rt6_info *)dst;
325 struct inet6_dev *idev = rt->rt6i_idev;
326 struct dst_entry *from = dst->from;
328 if (!(rt->dst.flags & DST_HOST))
329 dst_destroy_metrics_generic(dst);
332 rt->rt6i_idev = NULL;
339 if (rt6_has_peer(rt)) {
340 struct inet_peer *peer = rt6_peer_ptr(rt);
345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
348 struct rt6_info *rt = (struct rt6_info *)dst;
349 struct inet6_dev *idev = rt->rt6i_idev;
350 struct net_device *loopback_dev =
351 dev_net(dev)->loopback_dev;
353 if (dev != loopback_dev) {
354 if (idev && idev->dev == dev) {
355 struct inet6_dev *loopback_idev =
356 in6_dev_get(loopback_dev);
358 rt->rt6i_idev = loopback_idev;
365 static bool rt6_check_expired(const struct rt6_info *rt)
367 if (rt->rt6i_flags & RTF_EXPIRES) {
368 if (time_after(jiffies, rt->dst.expires))
370 } else if (rt->dst.from) {
371 return rt6_check_expired((struct rt6_info *) rt->dst.from);
376 /* Multipath route selection:
377 * Hash based function using packet header and flowlabel.
378 * Adapted from fib_info_hashfn()
380 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
381 const struct flowi6 *fl6)
383 unsigned int val = fl6->flowi6_proto;
385 val ^= ipv6_addr_hash(&fl6->daddr);
386 val ^= ipv6_addr_hash(&fl6->saddr);
388 /* Work only if this not encapsulated */
389 switch (fl6->flowi6_proto) {
393 val ^= (__force u16)fl6->fl6_sport;
394 val ^= (__force u16)fl6->fl6_dport;
398 val ^= (__force u16)fl6->fl6_icmp_type;
399 val ^= (__force u16)fl6->fl6_icmp_code;
402 /* RFC6438 recommands to use flowlabel */
403 val ^= (__force u32)fl6->flowlabel;
405 /* Perhaps, we need to tune, this function? */
406 val = val ^ (val >> 7) ^ (val >> 12);
407 return val % candidate_count;
410 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
411 struct flowi6 *fl6, int oif,
414 struct rt6_info *sibling, *next_sibling;
417 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
418 /* Don't change the route, if route_choosen == 0
419 * (siblings does not include ourself)
422 list_for_each_entry_safe(sibling, next_sibling,
423 &match->rt6i_siblings, rt6i_siblings) {
425 if (route_choosen == 0) {
426 if (rt6_score_route(sibling, oif, strict) < 0)
436 * Route lookup. Any table->tb6_lock is implied.
439 static inline struct rt6_info *rt6_device_match(struct net *net,
441 const struct in6_addr *saddr,
445 struct rt6_info *local = NULL;
446 struct rt6_info *sprt;
448 if (!oif && ipv6_addr_any(saddr))
451 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
452 struct net_device *dev = sprt->dst.dev;
455 if (dev->ifindex == oif)
457 if (dev->flags & IFF_LOOPBACK) {
458 if (!sprt->rt6i_idev ||
459 sprt->rt6i_idev->dev->ifindex != oif) {
460 if (flags & RT6_LOOKUP_F_IFACE && oif)
462 if (local && (!oif ||
463 local->rt6i_idev->dev->ifindex == oif))
469 if (ipv6_chk_addr(net, saddr, dev,
470 flags & RT6_LOOKUP_F_IFACE))
479 if (flags & RT6_LOOKUP_F_IFACE)
480 return net->ipv6.ip6_null_entry;
486 #ifdef CONFIG_IPV6_ROUTER_PREF
487 struct __rt6_probe_work {
488 struct work_struct work;
489 struct in6_addr target;
490 struct net_device *dev;
493 static void rt6_probe_deferred(struct work_struct *w)
495 struct in6_addr mcaddr;
496 struct __rt6_probe_work *work =
497 container_of(w, struct __rt6_probe_work, work);
499 addrconf_addr_solict_mult(&work->target, &mcaddr);
500 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
505 static void rt6_probe(struct rt6_info *rt)
507 struct neighbour *neigh;
509 * Okay, this does not seem to be appropriate
510 * for now, however, we need to check if it
511 * is really so; aka Router Reachability Probing.
513 * Router Reachability Probe MUST be rate-limited
514 * to no more than one per minute.
516 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
519 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
521 write_lock(&neigh->lock);
522 if (neigh->nud_state & NUD_VALID)
527 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
528 struct __rt6_probe_work *work;
530 work = kmalloc(sizeof(*work), GFP_ATOMIC);
533 __neigh_set_probe_once(neigh);
536 write_unlock(&neigh->lock);
539 INIT_WORK(&work->work, rt6_probe_deferred);
540 work->target = rt->rt6i_gateway;
541 dev_hold(rt->dst.dev);
542 work->dev = rt->dst.dev;
543 schedule_work(&work->work);
547 write_unlock(&neigh->lock);
549 rcu_read_unlock_bh();
552 static inline void rt6_probe(struct rt6_info *rt)
558 * Default Router Selection (RFC 2461 6.3.6)
560 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
562 struct net_device *dev = rt->dst.dev;
563 if (!oif || dev->ifindex == oif)
565 if ((dev->flags & IFF_LOOPBACK) &&
566 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
571 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
573 struct neighbour *neigh;
574 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
576 if (rt->rt6i_flags & RTF_NONEXTHOP ||
577 !(rt->rt6i_flags & RTF_GATEWAY))
578 return RT6_NUD_SUCCEED;
581 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
583 read_lock(&neigh->lock);
584 if (neigh->nud_state & NUD_VALID)
585 ret = RT6_NUD_SUCCEED;
586 #ifdef CONFIG_IPV6_ROUTER_PREF
587 else if (!(neigh->nud_state & NUD_FAILED))
588 ret = RT6_NUD_SUCCEED;
590 ret = RT6_NUD_FAIL_PROBE;
592 read_unlock(&neigh->lock);
594 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
595 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
597 rcu_read_unlock_bh();
602 static int rt6_score_route(struct rt6_info *rt, int oif,
607 m = rt6_check_dev(rt, oif);
608 if (!m && (strict & RT6_LOOKUP_F_IFACE))
609 return RT6_NUD_FAIL_HARD;
610 #ifdef CONFIG_IPV6_ROUTER_PREF
611 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
613 if (strict & RT6_LOOKUP_F_REACHABLE) {
614 int n = rt6_check_neigh(rt);
621 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
622 int *mpri, struct rt6_info *match,
626 bool match_do_rr = false;
628 if (rt6_check_expired(rt))
631 m = rt6_score_route(rt, oif, strict);
632 if (m == RT6_NUD_FAIL_DO_RR) {
634 m = 0; /* lowest valid score */
635 } else if (m == RT6_NUD_FAIL_HARD) {
639 if (strict & RT6_LOOKUP_F_REACHABLE)
642 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
644 *do_rr = match_do_rr;
652 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
653 struct rt6_info *rr_head,
654 u32 metric, int oif, int strict,
657 struct rt6_info *rt, *match;
661 for (rt = rr_head; rt && rt->rt6i_metric == metric;
662 rt = rt->dst.rt6_next)
663 match = find_match(rt, oif, strict, &mpri, match, do_rr);
664 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
665 rt = rt->dst.rt6_next)
666 match = find_match(rt, oif, strict, &mpri, match, do_rr);
671 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
673 struct rt6_info *match, *rt0;
679 fn->rr_ptr = rt0 = fn->leaf;
681 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
685 struct rt6_info *next = rt0->dst.rt6_next;
687 /* no entries matched; do round-robin */
688 if (!next || next->rt6i_metric != rt0->rt6i_metric)
695 net = dev_net(rt0->dst.dev);
696 return match ? match : net->ipv6.ip6_null_entry;
699 #ifdef CONFIG_IPV6_ROUTE_INFO
700 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
701 const struct in6_addr *gwaddr)
703 struct net *net = dev_net(dev);
704 struct route_info *rinfo = (struct route_info *) opt;
705 struct in6_addr prefix_buf, *prefix;
707 unsigned long lifetime;
710 if (len < sizeof(struct route_info)) {
714 /* Sanity check for prefix_len and length */
715 if (rinfo->length > 3) {
717 } else if (rinfo->prefix_len > 128) {
719 } else if (rinfo->prefix_len > 64) {
720 if (rinfo->length < 2) {
723 } else if (rinfo->prefix_len > 0) {
724 if (rinfo->length < 1) {
729 pref = rinfo->route_pref;
730 if (pref == ICMPV6_ROUTER_PREF_INVALID)
733 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
735 if (rinfo->length == 3)
736 prefix = (struct in6_addr *)rinfo->prefix;
738 /* this function is safe */
739 ipv6_addr_prefix(&prefix_buf,
740 (struct in6_addr *)rinfo->prefix,
742 prefix = &prefix_buf;
745 if (rinfo->prefix_len == 0)
746 rt = rt6_get_dflt_router(gwaddr, dev);
748 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
749 gwaddr, dev->ifindex);
751 if (rt && !lifetime) {
757 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
760 rt->rt6i_flags = RTF_ROUTEINFO |
761 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
764 if (!addrconf_finite_timeout(lifetime))
765 rt6_clean_expires(rt);
767 rt6_set_expires(rt, jiffies + HZ * lifetime);
775 #define BACKTRACK(__net, saddr) \
777 if (rt == __net->ipv6.ip6_null_entry) { \
778 struct fib6_node *pn; \
780 if (fn->fn_flags & RTN_TL_ROOT) \
783 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
784 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
787 if (fn->fn_flags & RTN_RTINFO) \
793 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
794 struct fib6_table *table,
795 struct flowi6 *fl6, int flags)
797 struct fib6_node *fn;
800 read_lock_bh(&table->tb6_lock);
801 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
804 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
805 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
806 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
807 BACKTRACK(net, &fl6->saddr);
809 dst_use(&rt->dst, jiffies);
810 read_unlock_bh(&table->tb6_lock);
815 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
818 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
820 EXPORT_SYMBOL_GPL(ip6_route_lookup);
822 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
823 const struct in6_addr *saddr, int oif, int strict)
825 struct flowi6 fl6 = {
829 struct dst_entry *dst;
830 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
833 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
834 flags |= RT6_LOOKUP_F_HAS_SADDR;
837 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
839 return (struct rt6_info *) dst;
845 EXPORT_SYMBOL(rt6_lookup);
847 /* ip6_ins_rt is called with FREE table->tb6_lock.
848 It takes new route entry, the addition fails by any reason the
849 route is freed. In any case, if caller does not hold it, it may
853 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
854 struct nlattr *mx, int mx_len)
857 struct fib6_table *table;
859 table = rt->rt6i_table;
860 write_lock_bh(&table->tb6_lock);
861 err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
862 write_unlock_bh(&table->tb6_lock);
867 int ip6_ins_rt(struct rt6_info *rt)
869 struct nl_info info = {
870 .nl_net = dev_net(rt->dst.dev),
872 return __ip6_ins_rt(rt, &info, NULL, 0);
875 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
876 const struct in6_addr *daddr,
877 const struct in6_addr *saddr)
885 rt = ip6_rt_copy(ort, daddr);
888 if (ort->rt6i_dst.plen != 128 &&
889 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
890 rt->rt6i_flags |= RTF_ANYCAST;
892 rt->rt6i_flags |= RTF_CACHE;
894 #ifdef CONFIG_IPV6_SUBTREES
895 if (rt->rt6i_src.plen && saddr) {
896 rt->rt6i_src.addr = *saddr;
897 rt->rt6i_src.plen = 128;
905 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
906 const struct in6_addr *daddr)
908 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
911 rt->rt6i_flags |= RTF_CACHE;
915 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
916 struct flowi6 *fl6, int flags)
918 struct fib6_node *fn;
919 struct rt6_info *rt, *nrt;
923 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
925 strict |= flags & RT6_LOOKUP_F_IFACE;
928 read_lock_bh(&table->tb6_lock);
931 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
934 rt = rt6_select(fn, oif, strict | reachable);
935 if (rt->rt6i_nsiblings)
936 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
937 BACKTRACK(net, &fl6->saddr);
938 if (rt == net->ipv6.ip6_null_entry ||
939 rt->rt6i_flags & RTF_CACHE)
943 read_unlock_bh(&table->tb6_lock);
945 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
946 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
947 else if (!(rt->dst.flags & DST_HOST))
948 nrt = rt6_alloc_clone(rt, &fl6->daddr);
953 rt = nrt ? : net->ipv6.ip6_null_entry;
957 err = ip6_ins_rt(nrt);
966 * Race condition! In the gap, when table->tb6_lock was
967 * released someone could insert this route. Relookup.
978 read_unlock_bh(&table->tb6_lock);
980 rt->dst.lastuse = jiffies;
986 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
987 struct flowi6 *fl6, int flags)
989 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
992 static struct dst_entry *ip6_route_input_lookup(struct net *net,
993 struct net_device *dev,
994 struct flowi6 *fl6, int flags)
996 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
997 flags |= RT6_LOOKUP_F_IFACE;
999 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1002 void ip6_route_input(struct sk_buff *skb)
1004 const struct ipv6hdr *iph = ipv6_hdr(skb);
1005 struct net *net = dev_net(skb->dev);
1006 int flags = RT6_LOOKUP_F_HAS_SADDR;
1007 struct flowi6 fl6 = {
1008 .flowi6_iif = skb->dev->ifindex,
1009 .daddr = iph->daddr,
1010 .saddr = iph->saddr,
1011 .flowlabel = ip6_flowinfo(iph),
1012 .flowi6_mark = skb->mark,
1013 .flowi6_proto = iph->nexthdr,
1016 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1019 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1020 struct flowi6 *fl6, int flags)
1022 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1025 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1030 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1032 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1033 flags |= RT6_LOOKUP_F_IFACE;
1035 if (!ipv6_addr_any(&fl6->saddr))
1036 flags |= RT6_LOOKUP_F_HAS_SADDR;
1038 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1040 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1042 EXPORT_SYMBOL(ip6_route_output);
1044 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1046 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1047 struct dst_entry *new = NULL;
1049 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1053 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1054 rt6_init_peer(rt, net->ipv6.peers);
1057 new->input = dst_discard;
1058 new->output = dst_discard_sk;
1060 if (dst_metrics_read_only(&ort->dst))
1061 new->_metrics = ort->dst._metrics;
1063 dst_copy_metrics(new, &ort->dst);
1064 rt->rt6i_idev = ort->rt6i_idev;
1066 in6_dev_hold(rt->rt6i_idev);
1068 rt->rt6i_gateway = ort->rt6i_gateway;
1069 rt->rt6i_flags = ort->rt6i_flags;
1070 rt->rt6i_metric = 0;
1072 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1073 #ifdef CONFIG_IPV6_SUBTREES
1074 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1080 dst_release(dst_orig);
1081 return new ? new : ERR_PTR(-ENOMEM);
1085 * Destination cache support functions
1088 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1090 struct rt6_info *rt;
1092 rt = (struct rt6_info *) dst;
1094 /* All IPV6 dsts are created with ->obsolete set to the value
1095 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1096 * into this function always.
1098 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1101 if (rt6_check_expired(rt))
1107 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1109 struct rt6_info *rt = (struct rt6_info *) dst;
1112 if (rt->rt6i_flags & RTF_CACHE) {
1113 if (rt6_check_expired(rt)) {
1125 static void ip6_link_failure(struct sk_buff *skb)
1127 struct rt6_info *rt;
1129 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1131 rt = (struct rt6_info *) skb_dst(skb);
1133 if (rt->rt6i_flags & RTF_CACHE) {
1137 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1138 rt->rt6i_node->fn_sernum = -1;
1143 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1144 struct sk_buff *skb, u32 mtu)
1146 struct rt6_info *rt6 = (struct rt6_info *)dst;
1149 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1150 struct net *net = dev_net(dst->dev);
1152 rt6->rt6i_flags |= RTF_MODIFIED;
1153 if (mtu < IPV6_MIN_MTU) {
1154 u32 features = dst_metric(dst, RTAX_FEATURES);
1156 features |= RTAX_FEATURE_ALLFRAG;
1157 dst_metric_set(dst, RTAX_FEATURES, features);
1159 dst_metric_set(dst, RTAX_MTU, mtu);
1160 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1164 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1167 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1168 struct dst_entry *dst;
1171 memset(&fl6, 0, sizeof(fl6));
1172 fl6.flowi6_oif = oif;
1173 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1174 fl6.daddr = iph->daddr;
1175 fl6.saddr = iph->saddr;
1176 fl6.flowlabel = ip6_flowinfo(iph);
1178 dst = ip6_route_output(net, NULL, &fl6);
1180 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1183 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1185 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1187 ip6_update_pmtu(skb, sock_net(sk), mtu,
1188 sk->sk_bound_dev_if, sk->sk_mark);
1190 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1192 /* Handle redirects */
1193 struct ip6rd_flowi {
1195 struct in6_addr gateway;
1198 static struct rt6_info *__ip6_route_redirect(struct net *net,
1199 struct fib6_table *table,
1203 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1204 struct rt6_info *rt;
1205 struct fib6_node *fn;
1207 /* Get the "current" route for this destination and
1208 * check if the redirect has come from approriate router.
1210 * RFC 4861 specifies that redirects should only be
1211 * accepted if they come from the nexthop to the target.
1212 * Due to the way the routes are chosen, this notion
1213 * is a bit fuzzy and one might need to check all possible
1217 read_lock_bh(&table->tb6_lock);
1218 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1220 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1221 if (rt6_check_expired(rt))
1225 if (!(rt->rt6i_flags & RTF_GATEWAY))
1227 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1229 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1235 rt = net->ipv6.ip6_null_entry;
1236 else if (rt->dst.error) {
1237 rt = net->ipv6.ip6_null_entry;
1240 BACKTRACK(net, &fl6->saddr);
1244 read_unlock_bh(&table->tb6_lock);
1249 static struct dst_entry *ip6_route_redirect(struct net *net,
1250 const struct flowi6 *fl6,
1251 const struct in6_addr *gateway)
1253 int flags = RT6_LOOKUP_F_HAS_SADDR;
1254 struct ip6rd_flowi rdfl;
1257 rdfl.gateway = *gateway;
1259 return fib6_rule_lookup(net, &rdfl.fl6,
1260 flags, __ip6_route_redirect);
1263 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1265 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1266 struct dst_entry *dst;
1269 memset(&fl6, 0, sizeof(fl6));
1270 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1271 fl6.flowi6_oif = oif;
1272 fl6.flowi6_mark = mark;
1273 fl6.daddr = iph->daddr;
1274 fl6.saddr = iph->saddr;
1275 fl6.flowlabel = ip6_flowinfo(iph);
1277 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1278 rt6_do_redirect(dst, NULL, skb);
1281 EXPORT_SYMBOL_GPL(ip6_redirect);
1283 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1286 const struct ipv6hdr *iph = ipv6_hdr(skb);
1287 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1288 struct dst_entry *dst;
1291 memset(&fl6, 0, sizeof(fl6));
1292 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1293 fl6.flowi6_oif = oif;
1294 fl6.flowi6_mark = mark;
1295 fl6.daddr = msg->dest;
1296 fl6.saddr = iph->daddr;
1298 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1299 rt6_do_redirect(dst, NULL, skb);
1303 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1305 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1307 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1309 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1311 struct net_device *dev = dst->dev;
1312 unsigned int mtu = dst_mtu(dst);
1313 struct net *net = dev_net(dev);
1315 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1317 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1318 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1321 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1322 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1323 * IPV6_MAXPLEN is also valid and means: "any MSS,
1324 * rely only on pmtu discovery"
1326 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1331 static unsigned int ip6_mtu(const struct dst_entry *dst)
1333 struct inet6_dev *idev;
1334 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1342 idev = __in6_dev_get(dst->dev);
1344 mtu = idev->cnf.mtu6;
1348 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1351 static struct dst_entry *icmp6_dst_gc_list;
1352 static DEFINE_SPINLOCK(icmp6_dst_lock);
1354 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1357 struct dst_entry *dst;
1358 struct rt6_info *rt;
1359 struct inet6_dev *idev = in6_dev_get(dev);
1360 struct net *net = dev_net(dev);
1362 if (unlikely(!idev))
1363 return ERR_PTR(-ENODEV);
1365 rt = ip6_dst_alloc(net, dev, 0, NULL);
1366 if (unlikely(!rt)) {
1368 dst = ERR_PTR(-ENOMEM);
1372 rt->dst.flags |= DST_HOST;
1373 rt->dst.output = ip6_output;
1374 atomic_set(&rt->dst.__refcnt, 1);
1375 rt->rt6i_gateway = fl6->daddr;
1376 rt->rt6i_dst.addr = fl6->daddr;
1377 rt->rt6i_dst.plen = 128;
1378 rt->rt6i_idev = idev;
1379 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1381 spin_lock_bh(&icmp6_dst_lock);
1382 rt->dst.next = icmp6_dst_gc_list;
1383 icmp6_dst_gc_list = &rt->dst;
1384 spin_unlock_bh(&icmp6_dst_lock);
1386 fib6_force_start_gc(net);
1388 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1394 int icmp6_dst_gc(void)
1396 struct dst_entry *dst, **pprev;
1399 spin_lock_bh(&icmp6_dst_lock);
1400 pprev = &icmp6_dst_gc_list;
1402 while ((dst = *pprev) != NULL) {
1403 if (!atomic_read(&dst->__refcnt)) {
1412 spin_unlock_bh(&icmp6_dst_lock);
1417 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1420 struct dst_entry *dst, **pprev;
1422 spin_lock_bh(&icmp6_dst_lock);
1423 pprev = &icmp6_dst_gc_list;
1424 while ((dst = *pprev) != NULL) {
1425 struct rt6_info *rt = (struct rt6_info *) dst;
1426 if (func(rt, arg)) {
1433 spin_unlock_bh(&icmp6_dst_lock);
1436 static int ip6_dst_gc(struct dst_ops *ops)
1438 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1439 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1440 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1441 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1442 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1443 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1446 entries = dst_entries_get_fast(ops);
1447 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1448 entries <= rt_max_size)
1451 net->ipv6.ip6_rt_gc_expire++;
1452 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1453 entries = dst_entries_get_slow(ops);
1454 if (entries < ops->gc_thresh)
1455 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1457 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1458 return entries > rt_max_size;
1465 int ip6_route_add(struct fib6_config *cfg)
1468 struct net *net = cfg->fc_nlinfo.nl_net;
1469 struct rt6_info *rt = NULL;
1470 struct net_device *dev = NULL;
1471 struct inet6_dev *idev = NULL;
1472 struct fib6_table *table;
1475 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1477 #ifndef CONFIG_IPV6_SUBTREES
1478 if (cfg->fc_src_len)
1481 if (cfg->fc_ifindex) {
1483 dev = dev_get_by_index(net, cfg->fc_ifindex);
1486 idev = in6_dev_get(dev);
1491 if (cfg->fc_metric == 0)
1492 cfg->fc_metric = IP6_RT_PRIO_USER;
1495 if (cfg->fc_nlinfo.nlh &&
1496 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1497 table = fib6_get_table(net, cfg->fc_table);
1499 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1500 table = fib6_new_table(net, cfg->fc_table);
1503 table = fib6_new_table(net, cfg->fc_table);
1509 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1516 if (cfg->fc_flags & RTF_EXPIRES)
1517 rt6_set_expires(rt, jiffies +
1518 clock_t_to_jiffies(cfg->fc_expires));
1520 rt6_clean_expires(rt);
1522 if (cfg->fc_protocol == RTPROT_UNSPEC)
1523 cfg->fc_protocol = RTPROT_BOOT;
1524 rt->rt6i_protocol = cfg->fc_protocol;
1526 addr_type = ipv6_addr_type(&cfg->fc_dst);
1528 if (addr_type & IPV6_ADDR_MULTICAST)
1529 rt->dst.input = ip6_mc_input;
1530 else if (cfg->fc_flags & RTF_LOCAL)
1531 rt->dst.input = ip6_input;
1533 rt->dst.input = ip6_forward;
1535 rt->dst.output = ip6_output;
1537 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1538 rt->rt6i_dst.plen = cfg->fc_dst_len;
1539 if (rt->rt6i_dst.plen == 128) {
1540 rt->dst.flags |= DST_HOST;
1541 dst_metrics_set_force_overwrite(&rt->dst);
1544 #ifdef CONFIG_IPV6_SUBTREES
1545 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1546 rt->rt6i_src.plen = cfg->fc_src_len;
1549 rt->rt6i_metric = cfg->fc_metric;
1551 /* We cannot add true routes via loopback here,
1552 they would result in kernel looping; promote them to reject routes
1554 if ((cfg->fc_flags & RTF_REJECT) ||
1555 (dev && (dev->flags & IFF_LOOPBACK) &&
1556 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1557 !(cfg->fc_flags & RTF_LOCAL))) {
1558 /* hold loopback dev/idev if we haven't done so. */
1559 if (dev != net->loopback_dev) {
1564 dev = net->loopback_dev;
1566 idev = in6_dev_get(dev);
1572 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1573 switch (cfg->fc_type) {
1575 rt->dst.error = -EINVAL;
1576 rt->dst.output = dst_discard_sk;
1577 rt->dst.input = dst_discard;
1580 rt->dst.error = -EACCES;
1581 rt->dst.output = ip6_pkt_prohibit_out;
1582 rt->dst.input = ip6_pkt_prohibit;
1586 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1588 rt->dst.output = ip6_pkt_discard_out;
1589 rt->dst.input = ip6_pkt_discard;
1595 if (cfg->fc_flags & RTF_GATEWAY) {
1596 const struct in6_addr *gw_addr;
1599 gw_addr = &cfg->fc_gateway;
1600 rt->rt6i_gateway = *gw_addr;
1601 gwa_type = ipv6_addr_type(gw_addr);
1603 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1604 struct rt6_info *grt;
1606 /* IPv6 strictly inhibits using not link-local
1607 addresses as nexthop address.
1608 Otherwise, router will not able to send redirects.
1609 It is very good, but in some (rare!) circumstances
1610 (SIT, PtP, NBMA NOARP links) it is handy to allow
1611 some exceptions. --ANK
1614 if (!(gwa_type & IPV6_ADDR_UNICAST))
1617 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1619 err = -EHOSTUNREACH;
1623 if (dev != grt->dst.dev) {
1629 idev = grt->rt6i_idev;
1631 in6_dev_hold(grt->rt6i_idev);
1633 if (!(grt->rt6i_flags & RTF_GATEWAY))
1641 if (!dev || (dev->flags & IFF_LOOPBACK))
1649 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1650 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1654 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1655 rt->rt6i_prefsrc.plen = 128;
1657 rt->rt6i_prefsrc.plen = 0;
1659 rt->rt6i_flags = cfg->fc_flags;
1663 rt->rt6i_idev = idev;
1664 rt->rt6i_table = table;
1666 cfg->fc_nlinfo.nl_net = dev_net(dev);
1668 return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1680 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1683 struct fib6_table *table;
1684 struct net *net = dev_net(rt->dst.dev);
1686 if (rt == net->ipv6.ip6_null_entry) {
1691 table = rt->rt6i_table;
1692 write_lock_bh(&table->tb6_lock);
1693 err = fib6_del(rt, info);
1694 write_unlock_bh(&table->tb6_lock);
1701 int ip6_del_rt(struct rt6_info *rt)
1703 struct nl_info info = {
1704 .nl_net = dev_net(rt->dst.dev),
1706 return __ip6_del_rt(rt, &info);
1709 static int ip6_route_del(struct fib6_config *cfg)
1711 struct fib6_table *table;
1712 struct fib6_node *fn;
1713 struct rt6_info *rt;
1716 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1720 read_lock_bh(&table->tb6_lock);
1722 fn = fib6_locate(&table->tb6_root,
1723 &cfg->fc_dst, cfg->fc_dst_len,
1724 &cfg->fc_src, cfg->fc_src_len);
1727 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1728 if (cfg->fc_ifindex &&
1730 rt->dst.dev->ifindex != cfg->fc_ifindex))
1732 if (cfg->fc_flags & RTF_GATEWAY &&
1733 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1735 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1738 read_unlock_bh(&table->tb6_lock);
1740 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1743 read_unlock_bh(&table->tb6_lock);
1748 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1750 struct net *net = dev_net(skb->dev);
1751 struct netevent_redirect netevent;
1752 struct rt6_info *rt, *nrt = NULL;
1753 struct ndisc_options ndopts;
1754 struct inet6_dev *in6_dev;
1755 struct neighbour *neigh;
1757 int optlen, on_link;
1760 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1761 optlen -= sizeof(*msg);
1764 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1768 msg = (struct rd_msg *)icmp6_hdr(skb);
1770 if (ipv6_addr_is_multicast(&msg->dest)) {
1771 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1776 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1778 } else if (ipv6_addr_type(&msg->target) !=
1779 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1780 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1784 in6_dev = __in6_dev_get(skb->dev);
1787 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1791 * The IP source address of the Redirect MUST be the same as the current
1792 * first-hop router for the specified ICMP Destination Address.
1795 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1796 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1801 if (ndopts.nd_opts_tgt_lladdr) {
1802 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1805 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1810 rt = (struct rt6_info *) dst;
1811 if (rt == net->ipv6.ip6_null_entry) {
1812 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1816 /* Redirect received -> path was valid.
1817 * Look, redirects are sent only in response to data packets,
1818 * so that this nexthop apparently is reachable. --ANK
1820 dst_confirm(&rt->dst);
1822 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1827 * We have finally decided to accept it.
1830 neigh_update(neigh, lladdr, NUD_STALE,
1831 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1832 NEIGH_UPDATE_F_OVERRIDE|
1833 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1834 NEIGH_UPDATE_F_ISROUTER))
1837 nrt = ip6_rt_copy(rt, &msg->dest);
1841 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1843 nrt->rt6i_flags &= ~RTF_GATEWAY;
1845 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1847 if (ip6_ins_rt(nrt))
1850 netevent.old = &rt->dst;
1851 netevent.new = &nrt->dst;
1852 netevent.daddr = &msg->dest;
1853 netevent.neigh = neigh;
1854 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1856 if (rt->rt6i_flags & RTF_CACHE) {
1857 rt = (struct rt6_info *) dst_clone(&rt->dst);
1862 neigh_release(neigh);
1866 * Misc support functions
1869 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1870 const struct in6_addr *dest)
1872 struct net *net = dev_net(ort->dst.dev);
1873 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1877 rt->dst.input = ort->dst.input;
1878 rt->dst.output = ort->dst.output;
1879 rt->dst.flags |= DST_HOST;
1881 rt->rt6i_dst.addr = *dest;
1882 rt->rt6i_dst.plen = 128;
1883 dst_copy_metrics(&rt->dst, &ort->dst);
1884 rt->dst.error = ort->dst.error;
1885 rt->rt6i_idev = ort->rt6i_idev;
1887 in6_dev_hold(rt->rt6i_idev);
1888 rt->dst.lastuse = jiffies;
1890 if (ort->rt6i_flags & RTF_GATEWAY)
1891 rt->rt6i_gateway = ort->rt6i_gateway;
1893 rt->rt6i_gateway = *dest;
1894 rt->rt6i_flags = ort->rt6i_flags;
1895 rt6_set_from(rt, ort);
1896 rt->rt6i_metric = 0;
1898 #ifdef CONFIG_IPV6_SUBTREES
1899 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1901 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1902 rt->rt6i_table = ort->rt6i_table;
1907 #ifdef CONFIG_IPV6_ROUTE_INFO
1908 static struct rt6_info *rt6_get_route_info(struct net *net,
1909 const struct in6_addr *prefix, int prefixlen,
1910 const struct in6_addr *gwaddr, int ifindex)
1912 struct fib6_node *fn;
1913 struct rt6_info *rt = NULL;
1914 struct fib6_table *table;
1916 table = fib6_get_table(net, RT6_TABLE_INFO);
1920 read_lock_bh(&table->tb6_lock);
1921 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1925 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1926 if (rt->dst.dev->ifindex != ifindex)
1928 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1930 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1936 read_unlock_bh(&table->tb6_lock);
1940 static struct rt6_info *rt6_add_route_info(struct net *net,
1941 const struct in6_addr *prefix, int prefixlen,
1942 const struct in6_addr *gwaddr, int ifindex,
1945 struct fib6_config cfg = {
1946 .fc_table = RT6_TABLE_INFO,
1947 .fc_metric = IP6_RT_PRIO_USER,
1948 .fc_ifindex = ifindex,
1949 .fc_dst_len = prefixlen,
1950 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1951 RTF_UP | RTF_PREF(pref),
1952 .fc_nlinfo.portid = 0,
1953 .fc_nlinfo.nlh = NULL,
1954 .fc_nlinfo.nl_net = net,
1957 cfg.fc_dst = *prefix;
1958 cfg.fc_gateway = *gwaddr;
1960 /* We should treat it as a default route if prefix length is 0. */
1962 cfg.fc_flags |= RTF_DEFAULT;
1964 ip6_route_add(&cfg);
1966 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1970 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1972 struct rt6_info *rt;
1973 struct fib6_table *table;
1975 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1979 read_lock_bh(&table->tb6_lock);
1980 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1981 if (dev == rt->dst.dev &&
1982 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1983 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1988 read_unlock_bh(&table->tb6_lock);
1992 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1993 struct net_device *dev,
1996 struct fib6_config cfg = {
1997 .fc_table = RT6_TABLE_DFLT,
1998 .fc_metric = IP6_RT_PRIO_USER,
1999 .fc_ifindex = dev->ifindex,
2000 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2001 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2002 .fc_nlinfo.portid = 0,
2003 .fc_nlinfo.nlh = NULL,
2004 .fc_nlinfo.nl_net = dev_net(dev),
2007 cfg.fc_gateway = *gwaddr;
2009 ip6_route_add(&cfg);
2011 return rt6_get_dflt_router(gwaddr, dev);
2014 void rt6_purge_dflt_routers(struct net *net)
2016 struct rt6_info *rt;
2017 struct fib6_table *table;
2019 /* NOTE: Keep consistent with rt6_get_dflt_router */
2020 table = fib6_get_table(net, RT6_TABLE_DFLT);
2025 read_lock_bh(&table->tb6_lock);
2026 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2027 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2028 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2030 read_unlock_bh(&table->tb6_lock);
2035 read_unlock_bh(&table->tb6_lock);
2038 static void rtmsg_to_fib6_config(struct net *net,
2039 struct in6_rtmsg *rtmsg,
2040 struct fib6_config *cfg)
2042 memset(cfg, 0, sizeof(*cfg));
2044 cfg->fc_table = RT6_TABLE_MAIN;
2045 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2046 cfg->fc_metric = rtmsg->rtmsg_metric;
2047 cfg->fc_expires = rtmsg->rtmsg_info;
2048 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2049 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2050 cfg->fc_flags = rtmsg->rtmsg_flags;
2052 cfg->fc_nlinfo.nl_net = net;
2054 cfg->fc_dst = rtmsg->rtmsg_dst;
2055 cfg->fc_src = rtmsg->rtmsg_src;
2056 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2059 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2061 struct fib6_config cfg;
2062 struct in6_rtmsg rtmsg;
2066 case SIOCADDRT: /* Add a route */
2067 case SIOCDELRT: /* Delete a route */
2068 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2070 err = copy_from_user(&rtmsg, arg,
2071 sizeof(struct in6_rtmsg));
2075 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2080 err = ip6_route_add(&cfg);
2083 err = ip6_route_del(&cfg);
2097 * Drop the packet on the floor
2100 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2103 struct dst_entry *dst = skb_dst(skb);
2104 switch (ipstats_mib_noroutes) {
2105 case IPSTATS_MIB_INNOROUTES:
2106 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2107 if (type == IPV6_ADDR_ANY) {
2108 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2109 IPSTATS_MIB_INADDRERRORS);
2113 case IPSTATS_MIB_OUTNOROUTES:
2114 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2115 ipstats_mib_noroutes);
2118 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2123 static int ip6_pkt_discard(struct sk_buff *skb)
2125 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2128 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2130 skb->dev = skb_dst(skb)->dev;
2131 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2134 static int ip6_pkt_prohibit(struct sk_buff *skb)
2136 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2139 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2141 skb->dev = skb_dst(skb)->dev;
2142 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2146 * Allocate a dst for local (unicast / anycast) address.
2149 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2150 const struct in6_addr *addr,
2153 struct net *net = dev_net(idev->dev);
2154 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2157 return ERR_PTR(-ENOMEM);
2161 rt->dst.flags |= DST_HOST;
2162 rt->dst.input = ip6_input;
2163 rt->dst.output = ip6_output;
2164 rt->rt6i_idev = idev;
2166 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2168 rt->rt6i_flags |= RTF_ANYCAST;
2170 rt->rt6i_flags |= RTF_LOCAL;
2172 rt->rt6i_gateway = *addr;
2173 rt->rt6i_dst.addr = *addr;
2174 rt->rt6i_dst.plen = 128;
2175 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2177 atomic_set(&rt->dst.__refcnt, 1);
2182 int ip6_route_get_saddr(struct net *net,
2183 struct rt6_info *rt,
2184 const struct in6_addr *daddr,
2186 struct in6_addr *saddr)
2188 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
2190 if (rt->rt6i_prefsrc.plen)
2191 *saddr = rt->rt6i_prefsrc.addr;
2193 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2194 daddr, prefs, saddr);
2198 /* remove deleted ip from prefsrc entries */
2199 struct arg_dev_net_ip {
2200 struct net_device *dev;
2202 struct in6_addr *addr;
2205 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2207 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2208 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2209 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2211 if (((void *)rt->dst.dev == dev || !dev) &&
2212 rt != net->ipv6.ip6_null_entry &&
2213 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2214 /* remove prefsrc entry */
2215 rt->rt6i_prefsrc.plen = 0;
2220 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2222 struct net *net = dev_net(ifp->idev->dev);
2223 struct arg_dev_net_ip adni = {
2224 .dev = ifp->idev->dev,
2228 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2231 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2232 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2234 /* Remove routers and update dst entries when gateway turn into host. */
2235 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2237 struct in6_addr *gateway = (struct in6_addr *)arg;
2239 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2240 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2241 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2247 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2249 fib6_clean_all(net, fib6_clean_tohost, gateway);
2252 struct arg_dev_net {
2253 struct net_device *dev;
2257 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2259 const struct arg_dev_net *adn = arg;
2260 const struct net_device *dev = adn->dev;
2262 if ((rt->dst.dev == dev || !dev) &&
2263 rt != adn->net->ipv6.ip6_null_entry)
2269 void rt6_ifdown(struct net *net, struct net_device *dev)
2271 struct arg_dev_net adn = {
2276 fib6_clean_all(net, fib6_ifdown, &adn);
2277 icmp6_clean_all(fib6_ifdown, &adn);
2280 struct rt6_mtu_change_arg {
2281 struct net_device *dev;
2285 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2287 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2288 struct inet6_dev *idev;
2290 /* In IPv6 pmtu discovery is not optional,
2291 so that RTAX_MTU lock cannot disable it.
2292 We still use this lock to block changes
2293 caused by addrconf/ndisc.
2296 idev = __in6_dev_get(arg->dev);
2300 /* For administrative MTU increase, there is no way to discover
2301 IPv6 PMTU increase, so PMTU increase should be updated here.
2302 Since RFC 1981 doesn't include administrative MTU increase
2303 update PMTU increase is a MUST. (i.e. jumbo frame)
2306 If new MTU is less than route PMTU, this new MTU will be the
2307 lowest MTU in the path, update the route PMTU to reflect PMTU
2308 decreases; if new MTU is greater than route PMTU, and the
2309 old MTU is the lowest MTU in the path, update the route PMTU
2310 to reflect the increase. In this case if the other nodes' MTU
2311 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2314 if (rt->dst.dev == arg->dev &&
2315 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2316 (dst_mtu(&rt->dst) >= arg->mtu ||
2317 (dst_mtu(&rt->dst) < arg->mtu &&
2318 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2319 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2324 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2326 struct rt6_mtu_change_arg arg = {
2331 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2334 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2335 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2336 [RTA_OIF] = { .type = NLA_U32 },
2337 [RTA_IIF] = { .type = NLA_U32 },
2338 [RTA_PRIORITY] = { .type = NLA_U32 },
2339 [RTA_METRICS] = { .type = NLA_NESTED },
2340 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2343 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2344 struct fib6_config *cfg)
2347 struct nlattr *tb[RTA_MAX+1];
2350 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2355 rtm = nlmsg_data(nlh);
2356 memset(cfg, 0, sizeof(*cfg));
2358 cfg->fc_table = rtm->rtm_table;
2359 cfg->fc_dst_len = rtm->rtm_dst_len;
2360 cfg->fc_src_len = rtm->rtm_src_len;
2361 cfg->fc_flags = RTF_UP;
2362 cfg->fc_protocol = rtm->rtm_protocol;
2363 cfg->fc_type = rtm->rtm_type;
2365 if (rtm->rtm_type == RTN_UNREACHABLE ||
2366 rtm->rtm_type == RTN_BLACKHOLE ||
2367 rtm->rtm_type == RTN_PROHIBIT ||
2368 rtm->rtm_type == RTN_THROW)
2369 cfg->fc_flags |= RTF_REJECT;
2371 if (rtm->rtm_type == RTN_LOCAL)
2372 cfg->fc_flags |= RTF_LOCAL;
2374 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2375 cfg->fc_nlinfo.nlh = nlh;
2376 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2378 if (tb[RTA_GATEWAY]) {
2379 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2380 cfg->fc_flags |= RTF_GATEWAY;
2384 int plen = (rtm->rtm_dst_len + 7) >> 3;
2386 if (nla_len(tb[RTA_DST]) < plen)
2389 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2393 int plen = (rtm->rtm_src_len + 7) >> 3;
2395 if (nla_len(tb[RTA_SRC]) < plen)
2398 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2401 if (tb[RTA_PREFSRC])
2402 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2405 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2407 if (tb[RTA_PRIORITY])
2408 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2410 if (tb[RTA_METRICS]) {
2411 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2412 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2416 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2418 if (tb[RTA_MULTIPATH]) {
2419 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2420 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2428 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2430 struct fib6_config r_cfg;
2431 struct rtnexthop *rtnh;
2434 int err = 0, last_err = 0;
2437 rtnh = (struct rtnexthop *)cfg->fc_mp;
2438 remaining = cfg->fc_mp_len;
2440 /* Parse a Multipath Entry */
2441 while (rtnh_ok(rtnh, remaining)) {
2442 memcpy(&r_cfg, cfg, sizeof(*cfg));
2443 if (rtnh->rtnh_ifindex)
2444 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2446 attrlen = rtnh_attrlen(rtnh);
2448 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2450 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2452 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2453 r_cfg.fc_flags |= RTF_GATEWAY;
2456 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2459 /* If we are trying to remove a route, do not stop the
2460 * loop when ip6_route_del() fails (because next hop is
2461 * already gone), we should try to remove all next hops.
2464 /* If add fails, we should try to delete all
2465 * next hops that have been already added.
2471 /* Because each route is added like a single route we remove
2472 * this flag after the first nexthop (if there is a collision,
2473 * we have already fail to add the first nexthop:
2474 * fib6_add_rt2node() has reject it).
2476 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2477 rtnh = rtnh_next(rtnh, &remaining);
2483 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2485 struct fib6_config cfg;
2488 err = rtm_to_fib6_config(skb, nlh, &cfg);
2493 return ip6_route_multipath(&cfg, 0);
2495 return ip6_route_del(&cfg);
2498 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2500 struct fib6_config cfg;
2503 err = rtm_to_fib6_config(skb, nlh, &cfg);
2508 return ip6_route_multipath(&cfg, 1);
2510 return ip6_route_add(&cfg);
2513 static inline size_t rt6_nlmsg_size(void)
2515 return NLMSG_ALIGN(sizeof(struct rtmsg))
2516 + nla_total_size(16) /* RTA_SRC */
2517 + nla_total_size(16) /* RTA_DST */
2518 + nla_total_size(16) /* RTA_GATEWAY */
2519 + nla_total_size(16) /* RTA_PREFSRC */
2520 + nla_total_size(4) /* RTA_TABLE */
2521 + nla_total_size(4) /* RTA_IIF */
2522 + nla_total_size(4) /* RTA_OIF */
2523 + nla_total_size(4) /* RTA_PRIORITY */
2524 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2525 + nla_total_size(sizeof(struct rta_cacheinfo));
2528 static int rt6_fill_node(struct net *net,
2529 struct sk_buff *skb, struct rt6_info *rt,
2530 struct in6_addr *dst, struct in6_addr *src,
2531 int iif, int type, u32 portid, u32 seq,
2532 int prefix, int nowait, unsigned int flags)
2535 struct nlmsghdr *nlh;
2539 if (prefix) { /* user wants prefix routes only */
2540 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2541 /* success since this is not a prefix route */
2546 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2550 rtm = nlmsg_data(nlh);
2551 rtm->rtm_family = AF_INET6;
2552 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2553 rtm->rtm_src_len = rt->rt6i_src.plen;
2556 table = rt->rt6i_table->tb6_id;
2558 table = RT6_TABLE_UNSPEC;
2559 rtm->rtm_table = table;
2560 if (nla_put_u32(skb, RTA_TABLE, table))
2561 goto nla_put_failure;
2562 if (rt->rt6i_flags & RTF_REJECT) {
2563 switch (rt->dst.error) {
2565 rtm->rtm_type = RTN_BLACKHOLE;
2568 rtm->rtm_type = RTN_PROHIBIT;
2571 rtm->rtm_type = RTN_THROW;
2574 rtm->rtm_type = RTN_UNREACHABLE;
2578 else if (rt->rt6i_flags & RTF_LOCAL)
2579 rtm->rtm_type = RTN_LOCAL;
2580 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2581 rtm->rtm_type = RTN_LOCAL;
2583 rtm->rtm_type = RTN_UNICAST;
2585 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2586 rtm->rtm_protocol = rt->rt6i_protocol;
2587 if (rt->rt6i_flags & RTF_DYNAMIC)
2588 rtm->rtm_protocol = RTPROT_REDIRECT;
2589 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2590 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2591 rtm->rtm_protocol = RTPROT_RA;
2593 rtm->rtm_protocol = RTPROT_KERNEL;
2596 if (rt->rt6i_flags & RTF_CACHE)
2597 rtm->rtm_flags |= RTM_F_CLONED;
2600 if (nla_put(skb, RTA_DST, 16, dst))
2601 goto nla_put_failure;
2602 rtm->rtm_dst_len = 128;
2603 } else if (rtm->rtm_dst_len)
2604 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2605 goto nla_put_failure;
2606 #ifdef CONFIG_IPV6_SUBTREES
2608 if (nla_put(skb, RTA_SRC, 16, src))
2609 goto nla_put_failure;
2610 rtm->rtm_src_len = 128;
2611 } else if (rtm->rtm_src_len &&
2612 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2613 goto nla_put_failure;
2616 #ifdef CONFIG_IPV6_MROUTE
2617 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2618 int err = ip6mr_get_route(net, skb, rtm, nowait);
2623 goto nla_put_failure;
2625 if (err == -EMSGSIZE)
2626 goto nla_put_failure;
2631 if (nla_put_u32(skb, RTA_IIF, iif))
2632 goto nla_put_failure;
2634 struct in6_addr saddr_buf;
2635 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2636 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2637 goto nla_put_failure;
2640 if (rt->rt6i_prefsrc.plen) {
2641 struct in6_addr saddr_buf;
2642 saddr_buf = rt->rt6i_prefsrc.addr;
2643 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2644 goto nla_put_failure;
2647 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2648 goto nla_put_failure;
2650 if (rt->rt6i_flags & RTF_GATEWAY) {
2651 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2652 goto nla_put_failure;
2656 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2657 goto nla_put_failure;
2658 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2659 goto nla_put_failure;
2661 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2663 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2664 goto nla_put_failure;
2666 return nlmsg_end(skb, nlh);
2669 nlmsg_cancel(skb, nlh);
2673 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2675 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2678 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2679 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2680 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2684 return rt6_fill_node(arg->net,
2685 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2686 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2687 prefix, 0, NLM_F_MULTI);
2690 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2692 struct net *net = sock_net(in_skb->sk);
2693 struct nlattr *tb[RTA_MAX+1];
2694 struct rt6_info *rt;
2695 struct sk_buff *skb;
2698 int err, iif = 0, oif = 0;
2700 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2705 memset(&fl6, 0, sizeof(fl6));
2708 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2711 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2715 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2718 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2722 iif = nla_get_u32(tb[RTA_IIF]);
2725 oif = nla_get_u32(tb[RTA_OIF]);
2728 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2731 struct net_device *dev;
2734 dev = __dev_get_by_index(net, iif);
2740 fl6.flowi6_iif = iif;
2742 if (!ipv6_addr_any(&fl6.saddr))
2743 flags |= RT6_LOOKUP_F_HAS_SADDR;
2745 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2748 fl6.flowi6_oif = oif;
2750 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2753 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2760 /* Reserve room for dummy headers, this skb can pass
2761 through good chunk of routing engine.
2763 skb_reset_mac_header(skb);
2764 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2766 skb_dst_set(skb, &rt->dst);
2768 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2769 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2770 nlh->nlmsg_seq, 0, 0, 0);
2776 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2781 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2783 struct sk_buff *skb;
2784 struct net *net = info->nl_net;
2789 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2791 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2795 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2796 event, info->portid, seq, 0, 0, 0);
2798 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2799 WARN_ON(err == -EMSGSIZE);
2803 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2804 info->nlh, gfp_any());
2808 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2811 static int ip6_route_dev_notify(struct notifier_block *this,
2812 unsigned long event, void *ptr)
2814 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2815 struct net *net = dev_net(dev);
2817 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2818 net->ipv6.ip6_null_entry->dst.dev = dev;
2819 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2820 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2821 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2822 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2823 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2824 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2835 #ifdef CONFIG_PROC_FS
2837 static const struct file_operations ipv6_route_proc_fops = {
2838 .owner = THIS_MODULE,
2839 .open = ipv6_route_open,
2841 .llseek = seq_lseek,
2842 .release = seq_release_net,
2845 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2847 struct net *net = (struct net *)seq->private;
2848 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2849 net->ipv6.rt6_stats->fib_nodes,
2850 net->ipv6.rt6_stats->fib_route_nodes,
2851 net->ipv6.rt6_stats->fib_rt_alloc,
2852 net->ipv6.rt6_stats->fib_rt_entries,
2853 net->ipv6.rt6_stats->fib_rt_cache,
2854 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2855 net->ipv6.rt6_stats->fib_discarded_routes);
2860 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2862 return single_open_net(inode, file, rt6_stats_seq_show);
2865 static const struct file_operations rt6_stats_seq_fops = {
2866 .owner = THIS_MODULE,
2867 .open = rt6_stats_seq_open,
2869 .llseek = seq_lseek,
2870 .release = single_release_net,
2872 #endif /* CONFIG_PROC_FS */
2874 #ifdef CONFIG_SYSCTL
2877 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2878 void __user *buffer, size_t *lenp, loff_t *ppos)
2885 net = (struct net *)ctl->extra1;
2886 delay = net->ipv6.sysctl.flush_delay;
2887 proc_dointvec(ctl, write, buffer, lenp, ppos);
2888 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2892 struct ctl_table ipv6_route_table_template[] = {
2894 .procname = "flush",
2895 .data = &init_net.ipv6.sysctl.flush_delay,
2896 .maxlen = sizeof(int),
2898 .proc_handler = ipv6_sysctl_rtcache_flush
2901 .procname = "gc_thresh",
2902 .data = &ip6_dst_ops_template.gc_thresh,
2903 .maxlen = sizeof(int),
2905 .proc_handler = proc_dointvec,
2908 .procname = "max_size",
2909 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2910 .maxlen = sizeof(int),
2912 .proc_handler = proc_dointvec,
2915 .procname = "gc_min_interval",
2916 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2917 .maxlen = sizeof(int),
2919 .proc_handler = proc_dointvec_jiffies,
2922 .procname = "gc_timeout",
2923 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2924 .maxlen = sizeof(int),
2926 .proc_handler = proc_dointvec_jiffies,
2929 .procname = "gc_interval",
2930 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2931 .maxlen = sizeof(int),
2933 .proc_handler = proc_dointvec_jiffies,
2936 .procname = "gc_elasticity",
2937 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2938 .maxlen = sizeof(int),
2940 .proc_handler = proc_dointvec,
2943 .procname = "mtu_expires",
2944 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2945 .maxlen = sizeof(int),
2947 .proc_handler = proc_dointvec_jiffies,
2950 .procname = "min_adv_mss",
2951 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2952 .maxlen = sizeof(int),
2954 .proc_handler = proc_dointvec,
2957 .procname = "gc_min_interval_ms",
2958 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2959 .maxlen = sizeof(int),
2961 .proc_handler = proc_dointvec_ms_jiffies,
2966 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2968 struct ctl_table *table;
2970 table = kmemdup(ipv6_route_table_template,
2971 sizeof(ipv6_route_table_template),
2975 table[0].data = &net->ipv6.sysctl.flush_delay;
2976 table[0].extra1 = net;
2977 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2978 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2979 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2980 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2981 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2982 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2983 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2984 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2985 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2987 /* Don't export sysctls to unprivileged users */
2988 if (net->user_ns != &init_user_ns)
2989 table[0].procname = NULL;
2996 static int __net_init ip6_route_net_init(struct net *net)
3000 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3001 sizeof(net->ipv6.ip6_dst_ops));
3003 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3004 goto out_ip6_dst_ops;
3006 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3007 sizeof(*net->ipv6.ip6_null_entry),
3009 if (!net->ipv6.ip6_null_entry)
3010 goto out_ip6_dst_entries;
3011 net->ipv6.ip6_null_entry->dst.path =
3012 (struct dst_entry *)net->ipv6.ip6_null_entry;
3013 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3014 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3015 ip6_template_metrics, true);
3017 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3018 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3019 sizeof(*net->ipv6.ip6_prohibit_entry),
3021 if (!net->ipv6.ip6_prohibit_entry)
3022 goto out_ip6_null_entry;
3023 net->ipv6.ip6_prohibit_entry->dst.path =
3024 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3025 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3026 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3027 ip6_template_metrics, true);
3029 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3030 sizeof(*net->ipv6.ip6_blk_hole_entry),
3032 if (!net->ipv6.ip6_blk_hole_entry)
3033 goto out_ip6_prohibit_entry;
3034 net->ipv6.ip6_blk_hole_entry->dst.path =
3035 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3036 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3037 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3038 ip6_template_metrics, true);
3041 net->ipv6.sysctl.flush_delay = 0;
3042 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3043 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3044 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3045 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3046 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3047 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3048 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3050 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3056 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3057 out_ip6_prohibit_entry:
3058 kfree(net->ipv6.ip6_prohibit_entry);
3060 kfree(net->ipv6.ip6_null_entry);
3062 out_ip6_dst_entries:
3063 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3068 static void __net_exit ip6_route_net_exit(struct net *net)
3070 kfree(net->ipv6.ip6_null_entry);
3071 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3072 kfree(net->ipv6.ip6_prohibit_entry);
3073 kfree(net->ipv6.ip6_blk_hole_entry);
3075 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3078 static int __net_init ip6_route_net_init_late(struct net *net)
3080 #ifdef CONFIG_PROC_FS
3081 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3082 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3087 static void __net_exit ip6_route_net_exit_late(struct net *net)
3089 #ifdef CONFIG_PROC_FS
3090 remove_proc_entry("ipv6_route", net->proc_net);
3091 remove_proc_entry("rt6_stats", net->proc_net);
3095 static struct pernet_operations ip6_route_net_ops = {
3096 .init = ip6_route_net_init,
3097 .exit = ip6_route_net_exit,
3100 static int __net_init ipv6_inetpeer_init(struct net *net)
3102 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3106 inet_peer_base_init(bp);
3107 net->ipv6.peers = bp;
3111 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3113 struct inet_peer_base *bp = net->ipv6.peers;
3115 net->ipv6.peers = NULL;
3116 inetpeer_invalidate_tree(bp);
3120 static struct pernet_operations ipv6_inetpeer_ops = {
3121 .init = ipv6_inetpeer_init,
3122 .exit = ipv6_inetpeer_exit,
3125 static struct pernet_operations ip6_route_net_late_ops = {
3126 .init = ip6_route_net_init_late,
3127 .exit = ip6_route_net_exit_late,
3130 static struct notifier_block ip6_route_dev_notifier = {
3131 .notifier_call = ip6_route_dev_notify,
3135 int __init ip6_route_init(void)
3140 ip6_dst_ops_template.kmem_cachep =
3141 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3142 SLAB_HWCACHE_ALIGN, NULL);
3143 if (!ip6_dst_ops_template.kmem_cachep)
3146 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3148 goto out_kmem_cache;
3150 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3152 goto out_dst_entries;
3154 ret = register_pernet_subsys(&ip6_route_net_ops);
3156 goto out_register_inetpeer;
3158 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3160 /* Registering of the loopback is done before this portion of code,
3161 * the loopback reference in rt6_info will not be taken, do it
3162 * manually for init_net */
3163 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3164 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3165 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3166 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3167 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3168 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3169 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3173 goto out_register_subsys;
3179 ret = fib6_rules_init();
3183 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3185 goto fib6_rules_init;
3188 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3189 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3190 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3191 goto out_register_late_subsys;
3193 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3195 goto out_register_late_subsys;
3200 out_register_late_subsys:
3201 unregister_pernet_subsys(&ip6_route_net_late_ops);
3203 fib6_rules_cleanup();
3208 out_register_subsys:
3209 unregister_pernet_subsys(&ip6_route_net_ops);
3210 out_register_inetpeer:
3211 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3213 dst_entries_destroy(&ip6_dst_blackhole_ops);
3215 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3219 void ip6_route_cleanup(void)
3221 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3222 unregister_pernet_subsys(&ip6_route_net_late_ops);
3223 fib6_rules_cleanup();
3226 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3227 unregister_pernet_subsys(&ip6_route_net_ops);
3228 dst_entries_destroy(&ip6_dst_blackhole_ops);
3229 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);