2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
89 struct sk_buff *skb, u32 mtu);
90 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
94 #ifdef CONFIG_IPV6_ROUTE_INFO
95 static struct rt6_info *rt6_add_route_info(struct net *net,
96 const struct in6_addr *prefix, int prefixlen,
97 const struct in6_addr *gwaddr, int ifindex,
99 static struct rt6_info *rt6_get_route_info(struct net *net,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr, int ifindex);
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
110 if (!(rt->dst.flags & DST_HOST))
113 peer = rt6_get_peer_create(rt);
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
138 struct in6_addr *p = &rt->rt6i_gateway;
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
143 return &ipv6_hdr(skb)->daddr;
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
151 struct rt6_info *rt = (struct rt6_info *) dst;
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
158 return neigh_create(&nd_tbl, daddr, dst->dev);
161 static struct dst_ops ip6_dst_ops_template = {
163 .protocol = cpu_to_be16(ETH_P_IPV6),
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
184 return mtu ? : dst->dev->mtu;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
203 static struct dst_ops ip6_dst_blackhole_ops = {
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
220 static const struct rt6_info ip6_null_entry_template = {
222 .__refcnt = ATOMIC_INIT(1),
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static int ip6_pkt_prohibit(struct sk_buff *skb);
238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
240 static const struct rt6_info ip6_prohibit_entry_template = {
242 .__refcnt = ATOMIC_INIT(1),
244 .obsolete = DST_OBSOLETE_FORCE_CHK,
246 .input = ip6_pkt_prohibit,
247 .output = ip6_pkt_prohibit_out,
249 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
250 .rt6i_protocol = RTPROT_KERNEL,
251 .rt6i_metric = ~(u32) 0,
252 .rt6i_ref = ATOMIC_INIT(1),
255 static const struct rt6_info ip6_blk_hole_entry_template = {
257 .__refcnt = ATOMIC_INIT(1),
259 .obsolete = DST_OBSOLETE_FORCE_CHK,
261 .input = dst_discard,
262 .output = dst_discard,
264 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
265 .rt6i_protocol = RTPROT_KERNEL,
266 .rt6i_metric = ~(u32) 0,
267 .rt6i_ref = ATOMIC_INIT(1),
272 /* allocate dst with ip6_dst_ops */
273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
274 struct net_device *dev,
276 struct fib6_table *table)
278 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
279 0, DST_OBSOLETE_FORCE_CHK, flags);
282 struct dst_entry *dst = &rt->dst;
284 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
285 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
286 rt->rt6i_genid = rt_genid_ipv6(net);
287 INIT_LIST_HEAD(&rt->rt6i_siblings);
292 static void ip6_dst_destroy(struct dst_entry *dst)
294 struct rt6_info *rt = (struct rt6_info *)dst;
295 struct inet6_dev *idev = rt->rt6i_idev;
296 struct dst_entry *from = dst->from;
298 if (!(rt->dst.flags & DST_HOST))
299 dst_destroy_metrics_generic(dst);
302 rt->rt6i_idev = NULL;
309 if (rt6_has_peer(rt)) {
310 struct inet_peer *peer = rt6_peer_ptr(rt);
315 void rt6_bind_peer(struct rt6_info *rt, int create)
317 struct inet_peer_base *base;
318 struct inet_peer *peer;
320 base = inetpeer_base_ptr(rt->_rt6i_peer);
324 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
326 if (!rt6_set_peer(rt, peer))
331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
334 struct rt6_info *rt = (struct rt6_info *)dst;
335 struct inet6_dev *idev = rt->rt6i_idev;
336 struct net_device *loopback_dev =
337 dev_net(dev)->loopback_dev;
339 if (dev != loopback_dev) {
340 if (idev && idev->dev == dev) {
341 struct inet6_dev *loopback_idev =
342 in6_dev_get(loopback_dev);
344 rt->rt6i_idev = loopback_idev;
351 static bool rt6_check_expired(const struct rt6_info *rt)
353 if (rt->rt6i_flags & RTF_EXPIRES) {
354 if (time_after(jiffies, rt->dst.expires))
356 } else if (rt->dst.from) {
357 return rt6_check_expired((struct rt6_info *) rt->dst.from);
362 static bool rt6_need_strict(const struct in6_addr *daddr)
364 return ipv6_addr_type(daddr) &
365 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
368 /* Multipath route selection:
369 * Hash based function using packet header and flowlabel.
370 * Adapted from fib_info_hashfn()
372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
373 const struct flowi6 *fl6)
375 unsigned int val = fl6->flowi6_proto;
377 val ^= ipv6_addr_hash(&fl6->daddr);
378 val ^= ipv6_addr_hash(&fl6->saddr);
380 /* Work only if this not encapsulated */
381 switch (fl6->flowi6_proto) {
385 val ^= (__force u16)fl6->fl6_sport;
386 val ^= (__force u16)fl6->fl6_dport;
390 val ^= (__force u16)fl6->fl6_icmp_type;
391 val ^= (__force u16)fl6->fl6_icmp_code;
394 /* RFC6438 recommands to use flowlabel */
395 val ^= (__force u32)fl6->flowlabel;
397 /* Perhaps, we need to tune, this function? */
398 val = val ^ (val >> 7) ^ (val >> 12);
399 return val % candidate_count;
402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct flowi6 *fl6, int oif,
406 struct rt6_info *sibling, *next_sibling;
409 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
410 /* Don't change the route, if route_choosen == 0
411 * (siblings does not include ourself)
414 list_for_each_entry_safe(sibling, next_sibling,
415 &match->rt6i_siblings, rt6i_siblings) {
417 if (route_choosen == 0) {
418 if (rt6_score_route(sibling, oif, strict) < 0)
428 * Route lookup. Any table->tb6_lock is implied.
431 static inline struct rt6_info *rt6_device_match(struct net *net,
433 const struct in6_addr *saddr,
437 struct rt6_info *local = NULL;
438 struct rt6_info *sprt;
440 if (!oif && ipv6_addr_any(saddr))
443 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
444 struct net_device *dev = sprt->dst.dev;
447 if (dev->ifindex == oif)
449 if (dev->flags & IFF_LOOPBACK) {
450 if (!sprt->rt6i_idev ||
451 sprt->rt6i_idev->dev->ifindex != oif) {
452 if (flags & RT6_LOOKUP_F_IFACE && oif)
454 if (local && (!oif ||
455 local->rt6i_idev->dev->ifindex == oif))
461 if (ipv6_chk_addr(net, saddr, dev,
462 flags & RT6_LOOKUP_F_IFACE))
471 if (flags & RT6_LOOKUP_F_IFACE)
472 return net->ipv6.ip6_null_entry;
478 #ifdef CONFIG_IPV6_ROUTER_PREF
479 struct __rt6_probe_work {
480 struct work_struct work;
481 struct in6_addr target;
482 struct net_device *dev;
485 static void rt6_probe_deferred(struct work_struct *w)
487 struct in6_addr mcaddr;
488 struct __rt6_probe_work *work =
489 container_of(w, struct __rt6_probe_work, work);
491 addrconf_addr_solict_mult(&work->target, &mcaddr);
492 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
497 static void rt6_probe(struct rt6_info *rt)
499 struct neighbour *neigh;
501 * Okay, this does not seem to be appropriate
502 * for now, however, we need to check if it
503 * is really so; aka Router Reachability Probing.
505 * Router Reachability Probe MUST be rate-limited
506 * to no more than one per minute.
508 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
511 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
513 write_lock(&neigh->lock);
514 if (neigh->nud_state & NUD_VALID)
519 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
520 struct __rt6_probe_work *work;
522 work = kmalloc(sizeof(*work), GFP_ATOMIC);
525 neigh->updated = jiffies;
528 write_unlock(&neigh->lock);
531 INIT_WORK(&work->work, rt6_probe_deferred);
532 work->target = rt->rt6i_gateway;
533 dev_hold(rt->dst.dev);
534 work->dev = rt->dst.dev;
535 schedule_work(&work->work);
539 write_unlock(&neigh->lock);
541 rcu_read_unlock_bh();
544 static inline void rt6_probe(struct rt6_info *rt)
550 * Default Router Selection (RFC 2461 6.3.6)
552 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
554 struct net_device *dev = rt->dst.dev;
555 if (!oif || dev->ifindex == oif)
557 if ((dev->flags & IFF_LOOPBACK) &&
558 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
563 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
565 struct neighbour *neigh;
566 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
568 if (rt->rt6i_flags & RTF_NONEXTHOP ||
569 !(rt->rt6i_flags & RTF_GATEWAY))
570 return RT6_NUD_SUCCEED;
573 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
575 read_lock(&neigh->lock);
576 if (neigh->nud_state & NUD_VALID)
577 ret = RT6_NUD_SUCCEED;
578 #ifdef CONFIG_IPV6_ROUTER_PREF
579 else if (!(neigh->nud_state & NUD_FAILED))
580 ret = RT6_NUD_SUCCEED;
582 read_unlock(&neigh->lock);
584 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
585 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
587 rcu_read_unlock_bh();
592 static int rt6_score_route(struct rt6_info *rt, int oif,
597 m = rt6_check_dev(rt, oif);
598 if (!m && (strict & RT6_LOOKUP_F_IFACE))
599 return RT6_NUD_FAIL_HARD;
600 #ifdef CONFIG_IPV6_ROUTER_PREF
601 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
603 if (strict & RT6_LOOKUP_F_REACHABLE) {
604 int n = rt6_check_neigh(rt);
611 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
612 int *mpri, struct rt6_info *match,
616 bool match_do_rr = false;
618 if (rt6_check_expired(rt))
621 m = rt6_score_route(rt, oif, strict);
622 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
624 m = 0; /* lowest valid score */
629 if (strict & RT6_LOOKUP_F_REACHABLE)
633 *do_rr = match_do_rr;
641 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
642 struct rt6_info *rr_head,
643 u32 metric, int oif, int strict,
646 struct rt6_info *rt, *match;
650 for (rt = rr_head; rt && rt->rt6i_metric == metric;
651 rt = rt->dst.rt6_next)
652 match = find_match(rt, oif, strict, &mpri, match, do_rr);
653 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
654 rt = rt->dst.rt6_next)
655 match = find_match(rt, oif, strict, &mpri, match, do_rr);
660 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
662 struct rt6_info *match, *rt0;
668 fn->rr_ptr = rt0 = fn->leaf;
670 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
674 struct rt6_info *next = rt0->dst.rt6_next;
676 /* no entries matched; do round-robin */
677 if (!next || next->rt6i_metric != rt0->rt6i_metric)
684 net = dev_net(rt0->dst.dev);
685 return match ? match : net->ipv6.ip6_null_entry;
688 #ifdef CONFIG_IPV6_ROUTE_INFO
689 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
690 const struct in6_addr *gwaddr)
692 struct net *net = dev_net(dev);
693 struct route_info *rinfo = (struct route_info *) opt;
694 struct in6_addr prefix_buf, *prefix;
696 unsigned long lifetime;
699 if (len < sizeof(struct route_info)) {
703 /* Sanity check for prefix_len and length */
704 if (rinfo->length > 3) {
706 } else if (rinfo->prefix_len > 128) {
708 } else if (rinfo->prefix_len > 64) {
709 if (rinfo->length < 2) {
712 } else if (rinfo->prefix_len > 0) {
713 if (rinfo->length < 1) {
718 pref = rinfo->route_pref;
719 if (pref == ICMPV6_ROUTER_PREF_INVALID)
722 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
724 if (rinfo->length == 3)
725 prefix = (struct in6_addr *)rinfo->prefix;
727 /* this function is safe */
728 ipv6_addr_prefix(&prefix_buf,
729 (struct in6_addr *)rinfo->prefix,
731 prefix = &prefix_buf;
734 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
737 if (rt && !lifetime) {
743 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
746 rt->rt6i_flags = RTF_ROUTEINFO |
747 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
750 if (!addrconf_finite_timeout(lifetime))
751 rt6_clean_expires(rt);
753 rt6_set_expires(rt, jiffies + HZ * lifetime);
761 #define BACKTRACK(__net, saddr) \
763 if (rt == __net->ipv6.ip6_null_entry) { \
764 struct fib6_node *pn; \
766 if (fn->fn_flags & RTN_TL_ROOT) \
769 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
770 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
773 if (fn->fn_flags & RTN_RTINFO) \
779 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
780 struct fib6_table *table,
781 struct flowi6 *fl6, int flags)
783 struct fib6_node *fn;
786 read_lock_bh(&table->tb6_lock);
787 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
790 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
791 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
792 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
793 BACKTRACK(net, &fl6->saddr);
795 dst_use(&rt->dst, jiffies);
796 read_unlock_bh(&table->tb6_lock);
801 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
804 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
806 EXPORT_SYMBOL_GPL(ip6_route_lookup);
808 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
809 const struct in6_addr *saddr, int oif, int strict)
811 struct flowi6 fl6 = {
815 struct dst_entry *dst;
816 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
819 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
820 flags |= RT6_LOOKUP_F_HAS_SADDR;
823 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
825 return (struct rt6_info *) dst;
832 EXPORT_SYMBOL(rt6_lookup);
834 /* ip6_ins_rt is called with FREE table->tb6_lock.
835 It takes new route entry, the addition fails by any reason the
836 route is freed. In any case, if caller does not hold it, it may
840 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
843 struct fib6_table *table;
845 table = rt->rt6i_table;
846 write_lock_bh(&table->tb6_lock);
847 err = fib6_add(&table->tb6_root, rt, info);
848 write_unlock_bh(&table->tb6_lock);
853 int ip6_ins_rt(struct rt6_info *rt)
855 struct nl_info info = {
856 .nl_net = dev_net(rt->dst.dev),
858 return __ip6_ins_rt(rt, &info);
861 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
862 const struct in6_addr *daddr,
863 const struct in6_addr *saddr)
871 rt = ip6_rt_copy(ort, daddr);
874 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
875 if (ort->rt6i_dst.plen != 128 &&
876 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
877 rt->rt6i_flags |= RTF_ANYCAST;
880 rt->rt6i_flags |= RTF_CACHE;
882 #ifdef CONFIG_IPV6_SUBTREES
883 if (rt->rt6i_src.plen && saddr) {
884 rt->rt6i_src.addr = *saddr;
885 rt->rt6i_src.plen = 128;
893 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
894 const struct in6_addr *daddr)
896 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
899 rt->rt6i_flags |= RTF_CACHE;
903 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
904 struct flowi6 *fl6, int flags)
906 struct fib6_node *fn;
907 struct rt6_info *rt, *nrt;
911 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
913 strict |= flags & RT6_LOOKUP_F_IFACE;
916 read_lock_bh(&table->tb6_lock);
919 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
922 rt = rt6_select(fn, oif, strict | reachable);
923 if (rt->rt6i_nsiblings)
924 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
925 BACKTRACK(net, &fl6->saddr);
926 if (rt == net->ipv6.ip6_null_entry ||
927 rt->rt6i_flags & RTF_CACHE)
931 read_unlock_bh(&table->tb6_lock);
933 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
934 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
935 else if (!(rt->dst.flags & DST_HOST))
936 nrt = rt6_alloc_clone(rt, &fl6->daddr);
941 rt = nrt ? : net->ipv6.ip6_null_entry;
945 err = ip6_ins_rt(nrt);
954 * Race condition! In the gap, when table->tb6_lock was
955 * released someone could insert this route. Relookup.
966 read_unlock_bh(&table->tb6_lock);
968 rt->dst.lastuse = jiffies;
974 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
975 struct flowi6 *fl6, int flags)
977 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
980 static struct dst_entry *ip6_route_input_lookup(struct net *net,
981 struct net_device *dev,
982 struct flowi6 *fl6, int flags)
984 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
985 flags |= RT6_LOOKUP_F_IFACE;
987 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
990 void ip6_route_input(struct sk_buff *skb)
992 const struct ipv6hdr *iph = ipv6_hdr(skb);
993 struct net *net = dev_net(skb->dev);
994 int flags = RT6_LOOKUP_F_HAS_SADDR;
995 struct flowi6 fl6 = {
996 .flowi6_iif = skb->dev->ifindex,
999 .flowlabel = ip6_flowinfo(iph),
1000 .flowi6_mark = skb->mark,
1001 .flowi6_proto = iph->nexthdr,
1004 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1007 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1008 struct flowi6 *fl6, int flags)
1010 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1013 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1018 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1020 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1021 flags |= RT6_LOOKUP_F_IFACE;
1023 if (!ipv6_addr_any(&fl6->saddr))
1024 flags |= RT6_LOOKUP_F_HAS_SADDR;
1026 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1028 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1031 EXPORT_SYMBOL(ip6_route_output);
1033 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1035 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1036 struct dst_entry *new = NULL;
1038 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1042 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1043 rt6_init_peer(rt, net->ipv6.peers);
1046 new->input = dst_discard;
1047 new->output = dst_discard;
1049 if (dst_metrics_read_only(&ort->dst))
1050 new->_metrics = ort->dst._metrics;
1052 dst_copy_metrics(new, &ort->dst);
1053 rt->rt6i_idev = ort->rt6i_idev;
1055 in6_dev_hold(rt->rt6i_idev);
1057 rt->rt6i_gateway = ort->rt6i_gateway;
1058 rt->rt6i_flags = ort->rt6i_flags;
1059 rt->rt6i_metric = 0;
1061 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1062 #ifdef CONFIG_IPV6_SUBTREES
1063 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1069 dst_release(dst_orig);
1070 return new ? new : ERR_PTR(-ENOMEM);
1074 * Destination cache support functions
1077 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1079 struct rt6_info *rt;
1081 rt = (struct rt6_info *) dst;
1083 /* All IPV6 dsts are created with ->obsolete set to the value
1084 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1085 * into this function always.
1087 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1090 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1096 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1098 struct rt6_info *rt = (struct rt6_info *) dst;
1101 if (rt->rt6i_flags & RTF_CACHE) {
1102 if (rt6_check_expired(rt)) {
1114 static void ip6_link_failure(struct sk_buff *skb)
1116 struct rt6_info *rt;
1118 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1120 rt = (struct rt6_info *) skb_dst(skb);
1122 if (rt->rt6i_flags & RTF_CACHE) {
1126 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1127 rt->rt6i_node->fn_sernum = -1;
1132 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1133 struct sk_buff *skb, u32 mtu)
1135 struct rt6_info *rt6 = (struct rt6_info*)dst;
1138 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1139 struct net *net = dev_net(dst->dev);
1141 rt6->rt6i_flags |= RTF_MODIFIED;
1142 if (mtu < IPV6_MIN_MTU) {
1143 u32 features = dst_metric(dst, RTAX_FEATURES);
1145 features |= RTAX_FEATURE_ALLFRAG;
1146 dst_metric_set(dst, RTAX_FEATURES, features);
1148 dst_metric_set(dst, RTAX_MTU, mtu);
1149 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1153 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1156 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1157 struct dst_entry *dst;
1160 memset(&fl6, 0, sizeof(fl6));
1161 fl6.flowi6_oif = oif;
1162 fl6.flowi6_mark = mark;
1163 fl6.flowi6_flags = 0;
1164 fl6.daddr = iph->daddr;
1165 fl6.saddr = iph->saddr;
1166 fl6.flowlabel = ip6_flowinfo(iph);
1168 dst = ip6_route_output(net, NULL, &fl6);
1170 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1173 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1175 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1177 ip6_update_pmtu(skb, sock_net(sk), mtu,
1178 sk->sk_bound_dev_if, sk->sk_mark);
1180 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1182 /* Handle redirects */
1183 struct ip6rd_flowi {
1185 struct in6_addr gateway;
1188 static struct rt6_info *__ip6_route_redirect(struct net *net,
1189 struct fib6_table *table,
1193 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1194 struct rt6_info *rt;
1195 struct fib6_node *fn;
1197 /* Get the "current" route for this destination and
1198 * check if the redirect has come from approriate router.
1200 * RFC 4861 specifies that redirects should only be
1201 * accepted if they come from the nexthop to the target.
1202 * Due to the way the routes are chosen, this notion
1203 * is a bit fuzzy and one might need to check all possible
1207 read_lock_bh(&table->tb6_lock);
1208 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1210 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1211 if (rt6_check_expired(rt))
1215 if (!(rt->rt6i_flags & RTF_GATEWAY))
1217 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1219 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1225 rt = net->ipv6.ip6_null_entry;
1226 else if (rt->dst.error) {
1227 rt = net->ipv6.ip6_null_entry;
1230 BACKTRACK(net, &fl6->saddr);
1234 read_unlock_bh(&table->tb6_lock);
1239 static struct dst_entry *ip6_route_redirect(struct net *net,
1240 const struct flowi6 *fl6,
1241 const struct in6_addr *gateway)
1243 int flags = RT6_LOOKUP_F_HAS_SADDR;
1244 struct ip6rd_flowi rdfl;
1247 rdfl.gateway = *gateway;
1249 return fib6_rule_lookup(net, &rdfl.fl6,
1250 flags, __ip6_route_redirect);
1253 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1255 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1256 struct dst_entry *dst;
1259 memset(&fl6, 0, sizeof(fl6));
1260 fl6.flowi6_oif = oif;
1261 fl6.flowi6_mark = mark;
1262 fl6.flowi6_flags = 0;
1263 fl6.daddr = iph->daddr;
1264 fl6.saddr = iph->saddr;
1265 fl6.flowlabel = ip6_flowinfo(iph);
1267 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1268 rt6_do_redirect(dst, NULL, skb);
1271 EXPORT_SYMBOL_GPL(ip6_redirect);
1273 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1276 const struct ipv6hdr *iph = ipv6_hdr(skb);
1277 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1278 struct dst_entry *dst;
1281 memset(&fl6, 0, sizeof(fl6));
1282 fl6.flowi6_oif = oif;
1283 fl6.flowi6_mark = mark;
1284 fl6.flowi6_flags = 0;
1285 fl6.daddr = msg->dest;
1286 fl6.saddr = iph->daddr;
1288 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1289 rt6_do_redirect(dst, NULL, skb);
1293 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1295 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1297 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1299 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1301 struct net_device *dev = dst->dev;
1302 unsigned int mtu = dst_mtu(dst);
1303 struct net *net = dev_net(dev);
1305 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1307 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1308 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1311 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1312 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1313 * IPV6_MAXPLEN is also valid and means: "any MSS,
1314 * rely only on pmtu discovery"
1316 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1321 static unsigned int ip6_mtu(const struct dst_entry *dst)
1323 struct inet6_dev *idev;
1324 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1332 idev = __in6_dev_get(dst->dev);
1334 mtu = idev->cnf.mtu6;
1340 static struct dst_entry *icmp6_dst_gc_list;
1341 static DEFINE_SPINLOCK(icmp6_dst_lock);
1343 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1346 struct dst_entry *dst;
1347 struct rt6_info *rt;
1348 struct inet6_dev *idev = in6_dev_get(dev);
1349 struct net *net = dev_net(dev);
1351 if (unlikely(!idev))
1352 return ERR_PTR(-ENODEV);
1354 rt = ip6_dst_alloc(net, dev, 0, NULL);
1355 if (unlikely(!rt)) {
1357 dst = ERR_PTR(-ENOMEM);
1361 rt->dst.flags |= DST_HOST;
1362 rt->dst.output = ip6_output;
1363 atomic_set(&rt->dst.__refcnt, 1);
1364 rt->rt6i_gateway = fl6->daddr;
1365 rt->rt6i_dst.addr = fl6->daddr;
1366 rt->rt6i_dst.plen = 128;
1367 rt->rt6i_idev = idev;
1368 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1370 spin_lock_bh(&icmp6_dst_lock);
1371 rt->dst.next = icmp6_dst_gc_list;
1372 icmp6_dst_gc_list = &rt->dst;
1373 spin_unlock_bh(&icmp6_dst_lock);
1375 fib6_force_start_gc(net);
1377 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1383 int icmp6_dst_gc(void)
1385 struct dst_entry *dst, **pprev;
1388 spin_lock_bh(&icmp6_dst_lock);
1389 pprev = &icmp6_dst_gc_list;
1391 while ((dst = *pprev) != NULL) {
1392 if (!atomic_read(&dst->__refcnt)) {
1401 spin_unlock_bh(&icmp6_dst_lock);
1406 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1409 struct dst_entry *dst, **pprev;
1411 spin_lock_bh(&icmp6_dst_lock);
1412 pprev = &icmp6_dst_gc_list;
1413 while ((dst = *pprev) != NULL) {
1414 struct rt6_info *rt = (struct rt6_info *) dst;
1415 if (func(rt, arg)) {
1422 spin_unlock_bh(&icmp6_dst_lock);
1425 static int ip6_dst_gc(struct dst_ops *ops)
1427 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1428 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1429 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1430 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1431 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1432 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1435 entries = dst_entries_get_fast(ops);
1436 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1437 entries <= rt_max_size)
1440 net->ipv6.ip6_rt_gc_expire++;
1441 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1442 entries = dst_entries_get_slow(ops);
1443 if (entries < ops->gc_thresh)
1444 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1446 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1447 return entries > rt_max_size;
1454 int ip6_route_add(struct fib6_config *cfg)
1457 struct net *net = cfg->fc_nlinfo.nl_net;
1458 struct rt6_info *rt = NULL;
1459 struct net_device *dev = NULL;
1460 struct inet6_dev *idev = NULL;
1461 struct fib6_table *table;
1464 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1466 #ifndef CONFIG_IPV6_SUBTREES
1467 if (cfg->fc_src_len)
1470 if (cfg->fc_ifindex) {
1472 dev = dev_get_by_index(net, cfg->fc_ifindex);
1475 idev = in6_dev_get(dev);
1480 if (cfg->fc_metric == 0)
1481 cfg->fc_metric = IP6_RT_PRIO_USER;
1484 if (cfg->fc_nlinfo.nlh &&
1485 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1486 table = fib6_get_table(net, cfg->fc_table);
1488 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1489 table = fib6_new_table(net, cfg->fc_table);
1492 table = fib6_new_table(net, cfg->fc_table);
1498 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1505 if (cfg->fc_flags & RTF_EXPIRES)
1506 rt6_set_expires(rt, jiffies +
1507 clock_t_to_jiffies(cfg->fc_expires));
1509 rt6_clean_expires(rt);
1511 if (cfg->fc_protocol == RTPROT_UNSPEC)
1512 cfg->fc_protocol = RTPROT_BOOT;
1513 rt->rt6i_protocol = cfg->fc_protocol;
1515 addr_type = ipv6_addr_type(&cfg->fc_dst);
1517 if (addr_type & IPV6_ADDR_MULTICAST)
1518 rt->dst.input = ip6_mc_input;
1519 else if (cfg->fc_flags & RTF_LOCAL)
1520 rt->dst.input = ip6_input;
1522 rt->dst.input = ip6_forward;
1524 rt->dst.output = ip6_output;
1526 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1527 rt->rt6i_dst.plen = cfg->fc_dst_len;
1528 if (rt->rt6i_dst.plen == 128)
1529 rt->dst.flags |= DST_HOST;
1531 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1532 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1537 dst_init_metrics(&rt->dst, metrics, 0);
1539 #ifdef CONFIG_IPV6_SUBTREES
1540 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1541 rt->rt6i_src.plen = cfg->fc_src_len;
1544 rt->rt6i_metric = cfg->fc_metric;
1546 /* We cannot add true routes via loopback here,
1547 they would result in kernel looping; promote them to reject routes
1549 if ((cfg->fc_flags & RTF_REJECT) ||
1550 (dev && (dev->flags & IFF_LOOPBACK) &&
1551 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1552 !(cfg->fc_flags & RTF_LOCAL))) {
1553 /* hold loopback dev/idev if we haven't done so. */
1554 if (dev != net->loopback_dev) {
1559 dev = net->loopback_dev;
1561 idev = in6_dev_get(dev);
1567 rt->dst.output = ip6_pkt_discard_out;
1568 rt->dst.input = ip6_pkt_discard;
1569 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1570 switch (cfg->fc_type) {
1572 rt->dst.error = -EINVAL;
1575 rt->dst.error = -EACCES;
1578 rt->dst.error = -EAGAIN;
1581 rt->dst.error = -ENETUNREACH;
1587 if (cfg->fc_flags & RTF_GATEWAY) {
1588 const struct in6_addr *gw_addr;
1591 gw_addr = &cfg->fc_gateway;
1592 rt->rt6i_gateway = *gw_addr;
1593 gwa_type = ipv6_addr_type(gw_addr);
1595 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1596 struct rt6_info *grt;
1598 /* IPv6 strictly inhibits using not link-local
1599 addresses as nexthop address.
1600 Otherwise, router will not able to send redirects.
1601 It is very good, but in some (rare!) circumstances
1602 (SIT, PtP, NBMA NOARP links) it is handy to allow
1603 some exceptions. --ANK
1606 if (!(gwa_type & IPV6_ADDR_UNICAST))
1609 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1611 err = -EHOSTUNREACH;
1615 if (dev != grt->dst.dev) {
1621 idev = grt->rt6i_idev;
1623 in6_dev_hold(grt->rt6i_idev);
1625 if (!(grt->rt6i_flags & RTF_GATEWAY))
1633 if (!dev || (dev->flags & IFF_LOOPBACK))
1641 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1642 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1646 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1647 rt->rt6i_prefsrc.plen = 128;
1649 rt->rt6i_prefsrc.plen = 0;
1651 rt->rt6i_flags = cfg->fc_flags;
1658 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1659 int type = nla_type(nla);
1662 if (type > RTAX_MAX) {
1667 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1673 rt->rt6i_idev = idev;
1674 rt->rt6i_table = table;
1676 cfg->fc_nlinfo.nl_net = dev_net(dev);
1678 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1690 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1693 struct fib6_table *table;
1694 struct net *net = dev_net(rt->dst.dev);
1696 if (rt == net->ipv6.ip6_null_entry) {
1701 table = rt->rt6i_table;
1702 write_lock_bh(&table->tb6_lock);
1703 err = fib6_del(rt, info);
1704 write_unlock_bh(&table->tb6_lock);
1711 int ip6_del_rt(struct rt6_info *rt)
1713 struct nl_info info = {
1714 .nl_net = dev_net(rt->dst.dev),
1716 return __ip6_del_rt(rt, &info);
1719 static int ip6_route_del(struct fib6_config *cfg)
1721 struct fib6_table *table;
1722 struct fib6_node *fn;
1723 struct rt6_info *rt;
1726 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1730 read_lock_bh(&table->tb6_lock);
1732 fn = fib6_locate(&table->tb6_root,
1733 &cfg->fc_dst, cfg->fc_dst_len,
1734 &cfg->fc_src, cfg->fc_src_len);
1737 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1738 if (cfg->fc_ifindex &&
1740 rt->dst.dev->ifindex != cfg->fc_ifindex))
1742 if (cfg->fc_flags & RTF_GATEWAY &&
1743 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1745 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1748 read_unlock_bh(&table->tb6_lock);
1750 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1753 read_unlock_bh(&table->tb6_lock);
1758 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1760 struct net *net = dev_net(skb->dev);
1761 struct netevent_redirect netevent;
1762 struct rt6_info *rt, *nrt = NULL;
1763 struct ndisc_options ndopts;
1764 struct inet6_dev *in6_dev;
1765 struct neighbour *neigh;
1767 int optlen, on_link;
1770 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1771 optlen -= sizeof(*msg);
1774 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1778 msg = (struct rd_msg *)icmp6_hdr(skb);
1780 if (ipv6_addr_is_multicast(&msg->dest)) {
1781 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1786 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1788 } else if (ipv6_addr_type(&msg->target) !=
1789 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1790 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1794 in6_dev = __in6_dev_get(skb->dev);
1797 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1801 * The IP source address of the Redirect MUST be the same as the current
1802 * first-hop router for the specified ICMP Destination Address.
1805 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1806 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1811 if (ndopts.nd_opts_tgt_lladdr) {
1812 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1815 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1820 rt = (struct rt6_info *) dst;
1821 if (rt == net->ipv6.ip6_null_entry) {
1822 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1826 /* Redirect received -> path was valid.
1827 * Look, redirects are sent only in response to data packets,
1828 * so that this nexthop apparently is reachable. --ANK
1830 dst_confirm(&rt->dst);
1832 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1837 * We have finally decided to accept it.
1840 neigh_update(neigh, lladdr, NUD_STALE,
1841 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1842 NEIGH_UPDATE_F_OVERRIDE|
1843 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1844 NEIGH_UPDATE_F_ISROUTER))
1847 nrt = ip6_rt_copy(rt, &msg->dest);
1851 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1853 nrt->rt6i_flags &= ~RTF_GATEWAY;
1855 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1857 if (ip6_ins_rt(nrt))
1860 netevent.old = &rt->dst;
1861 netevent.new = &nrt->dst;
1862 netevent.daddr = &msg->dest;
1863 netevent.neigh = neigh;
1864 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1866 if (rt->rt6i_flags & RTF_CACHE) {
1867 rt = (struct rt6_info *) dst_clone(&rt->dst);
1872 neigh_release(neigh);
1876 * Misc support functions
1879 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1880 const struct in6_addr *dest)
1882 struct net *net = dev_net(ort->dst.dev);
1883 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1887 rt->dst.input = ort->dst.input;
1888 rt->dst.output = ort->dst.output;
1889 rt->dst.flags |= DST_HOST;
1891 rt->rt6i_dst.addr = *dest;
1892 rt->rt6i_dst.plen = 128;
1893 dst_copy_metrics(&rt->dst, &ort->dst);
1894 rt->dst.error = ort->dst.error;
1895 rt->rt6i_idev = ort->rt6i_idev;
1897 in6_dev_hold(rt->rt6i_idev);
1898 rt->dst.lastuse = jiffies;
1900 if (ort->rt6i_flags & RTF_GATEWAY)
1901 rt->rt6i_gateway = ort->rt6i_gateway;
1903 rt->rt6i_gateway = *dest;
1904 rt->rt6i_flags = ort->rt6i_flags;
1905 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1906 (RTF_DEFAULT | RTF_ADDRCONF))
1907 rt6_set_from(rt, ort);
1908 rt->rt6i_metric = 0;
1910 #ifdef CONFIG_IPV6_SUBTREES
1911 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1913 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1914 rt->rt6i_table = ort->rt6i_table;
1919 #ifdef CONFIG_IPV6_ROUTE_INFO
1920 static struct rt6_info *rt6_get_route_info(struct net *net,
1921 const struct in6_addr *prefix, int prefixlen,
1922 const struct in6_addr *gwaddr, int ifindex)
1924 struct fib6_node *fn;
1925 struct rt6_info *rt = NULL;
1926 struct fib6_table *table;
1928 table = fib6_get_table(net, RT6_TABLE_INFO);
1932 read_lock_bh(&table->tb6_lock);
1933 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1937 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1938 if (rt->dst.dev->ifindex != ifindex)
1940 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1942 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1948 read_unlock_bh(&table->tb6_lock);
1952 static struct rt6_info *rt6_add_route_info(struct net *net,
1953 const struct in6_addr *prefix, int prefixlen,
1954 const struct in6_addr *gwaddr, int ifindex,
1957 struct fib6_config cfg = {
1958 .fc_table = RT6_TABLE_INFO,
1959 .fc_metric = IP6_RT_PRIO_USER,
1960 .fc_ifindex = ifindex,
1961 .fc_dst_len = prefixlen,
1962 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1963 RTF_UP | RTF_PREF(pref),
1964 .fc_nlinfo.portid = 0,
1965 .fc_nlinfo.nlh = NULL,
1966 .fc_nlinfo.nl_net = net,
1969 cfg.fc_dst = *prefix;
1970 cfg.fc_gateway = *gwaddr;
1972 /* We should treat it as a default route if prefix length is 0. */
1974 cfg.fc_flags |= RTF_DEFAULT;
1976 ip6_route_add(&cfg);
1978 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1982 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1984 struct rt6_info *rt;
1985 struct fib6_table *table;
1987 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1991 read_lock_bh(&table->tb6_lock);
1992 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1993 if (dev == rt->dst.dev &&
1994 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1995 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2000 read_unlock_bh(&table->tb6_lock);
2004 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2005 struct net_device *dev,
2008 struct fib6_config cfg = {
2009 .fc_table = RT6_TABLE_DFLT,
2010 .fc_metric = IP6_RT_PRIO_USER,
2011 .fc_ifindex = dev->ifindex,
2012 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2013 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2014 .fc_nlinfo.portid = 0,
2015 .fc_nlinfo.nlh = NULL,
2016 .fc_nlinfo.nl_net = dev_net(dev),
2019 cfg.fc_gateway = *gwaddr;
2021 ip6_route_add(&cfg);
2023 return rt6_get_dflt_router(gwaddr, dev);
2026 void rt6_purge_dflt_routers(struct net *net)
2028 struct rt6_info *rt;
2029 struct fib6_table *table;
2031 /* NOTE: Keep consistent with rt6_get_dflt_router */
2032 table = fib6_get_table(net, RT6_TABLE_DFLT);
2037 read_lock_bh(&table->tb6_lock);
2038 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2039 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2040 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2042 read_unlock_bh(&table->tb6_lock);
2047 read_unlock_bh(&table->tb6_lock);
2050 static void rtmsg_to_fib6_config(struct net *net,
2051 struct in6_rtmsg *rtmsg,
2052 struct fib6_config *cfg)
2054 memset(cfg, 0, sizeof(*cfg));
2056 cfg->fc_table = RT6_TABLE_MAIN;
2057 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2058 cfg->fc_metric = rtmsg->rtmsg_metric;
2059 cfg->fc_expires = rtmsg->rtmsg_info;
2060 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2061 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2062 cfg->fc_flags = rtmsg->rtmsg_flags;
2064 cfg->fc_nlinfo.nl_net = net;
2066 cfg->fc_dst = rtmsg->rtmsg_dst;
2067 cfg->fc_src = rtmsg->rtmsg_src;
2068 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2071 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2073 struct fib6_config cfg;
2074 struct in6_rtmsg rtmsg;
2078 case SIOCADDRT: /* Add a route */
2079 case SIOCDELRT: /* Delete a route */
2080 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2082 err = copy_from_user(&rtmsg, arg,
2083 sizeof(struct in6_rtmsg));
2087 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2092 err = ip6_route_add(&cfg);
2095 err = ip6_route_del(&cfg);
2109 * Drop the packet on the floor
2112 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2115 struct dst_entry *dst = skb_dst(skb);
2116 switch (ipstats_mib_noroutes) {
2117 case IPSTATS_MIB_INNOROUTES:
2118 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2119 if (type == IPV6_ADDR_ANY) {
2120 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121 IPSTATS_MIB_INADDRERRORS);
2125 case IPSTATS_MIB_OUTNOROUTES:
2126 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2127 ipstats_mib_noroutes);
2130 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2135 static int ip6_pkt_discard(struct sk_buff *skb)
2137 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2140 static int ip6_pkt_discard_out(struct sk_buff *skb)
2142 skb->dev = skb_dst(skb)->dev;
2143 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2146 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2148 static int ip6_pkt_prohibit(struct sk_buff *skb)
2150 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2153 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2155 skb->dev = skb_dst(skb)->dev;
2156 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2162 * Allocate a dst for local (unicast / anycast) address.
2165 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2166 const struct in6_addr *addr,
2169 struct net *net = dev_net(idev->dev);
2170 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2173 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2174 return ERR_PTR(-ENOMEM);
2179 rt->dst.flags |= DST_HOST;
2180 rt->dst.input = ip6_input;
2181 rt->dst.output = ip6_output;
2182 rt->rt6i_idev = idev;
2184 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2186 rt->rt6i_flags |= RTF_ANYCAST;
2188 rt->rt6i_flags |= RTF_LOCAL;
2190 rt->rt6i_gateway = *addr;
2191 rt->rt6i_dst.addr = *addr;
2192 rt->rt6i_dst.plen = 128;
2193 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2195 atomic_set(&rt->dst.__refcnt, 1);
2200 int ip6_route_get_saddr(struct net *net,
2201 struct rt6_info *rt,
2202 const struct in6_addr *daddr,
2204 struct in6_addr *saddr)
2206 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2208 if (rt->rt6i_prefsrc.plen)
2209 *saddr = rt->rt6i_prefsrc.addr;
2211 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2212 daddr, prefs, saddr);
2216 /* remove deleted ip from prefsrc entries */
2217 struct arg_dev_net_ip {
2218 struct net_device *dev;
2220 struct in6_addr *addr;
2223 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2225 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2226 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2227 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2229 if (((void *)rt->dst.dev == dev || !dev) &&
2230 rt != net->ipv6.ip6_null_entry &&
2231 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2232 /* remove prefsrc entry */
2233 rt->rt6i_prefsrc.plen = 0;
2238 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2240 struct net *net = dev_net(ifp->idev->dev);
2241 struct arg_dev_net_ip adni = {
2242 .dev = ifp->idev->dev,
2246 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2249 struct arg_dev_net {
2250 struct net_device *dev;
2254 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2256 const struct arg_dev_net *adn = arg;
2257 const struct net_device *dev = adn->dev;
2259 if ((rt->dst.dev == dev || !dev) &&
2260 rt != adn->net->ipv6.ip6_null_entry)
2266 void rt6_ifdown(struct net *net, struct net_device *dev)
2268 struct arg_dev_net adn = {
2273 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2274 icmp6_clean_all(fib6_ifdown, &adn);
2277 struct rt6_mtu_change_arg {
2278 struct net_device *dev;
2282 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2284 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2285 struct inet6_dev *idev;
2287 /* In IPv6 pmtu discovery is not optional,
2288 so that RTAX_MTU lock cannot disable it.
2289 We still use this lock to block changes
2290 caused by addrconf/ndisc.
2293 idev = __in6_dev_get(arg->dev);
2297 /* For administrative MTU increase, there is no way to discover
2298 IPv6 PMTU increase, so PMTU increase should be updated here.
2299 Since RFC 1981 doesn't include administrative MTU increase
2300 update PMTU increase is a MUST. (i.e. jumbo frame)
2303 If new MTU is less than route PMTU, this new MTU will be the
2304 lowest MTU in the path, update the route PMTU to reflect PMTU
2305 decreases; if new MTU is greater than route PMTU, and the
2306 old MTU is the lowest MTU in the path, update the route PMTU
2307 to reflect the increase. In this case if the other nodes' MTU
2308 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2311 if (rt->dst.dev == arg->dev &&
2312 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2313 (dst_mtu(&rt->dst) >= arg->mtu ||
2314 (dst_mtu(&rt->dst) < arg->mtu &&
2315 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2316 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2321 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2323 struct rt6_mtu_change_arg arg = {
2328 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2331 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2332 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2333 [RTA_OIF] = { .type = NLA_U32 },
2334 [RTA_IIF] = { .type = NLA_U32 },
2335 [RTA_PRIORITY] = { .type = NLA_U32 },
2336 [RTA_METRICS] = { .type = NLA_NESTED },
2337 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2340 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2341 struct fib6_config *cfg)
2344 struct nlattr *tb[RTA_MAX+1];
2347 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2352 rtm = nlmsg_data(nlh);
2353 memset(cfg, 0, sizeof(*cfg));
2355 cfg->fc_table = rtm->rtm_table;
2356 cfg->fc_dst_len = rtm->rtm_dst_len;
2357 cfg->fc_src_len = rtm->rtm_src_len;
2358 cfg->fc_flags = RTF_UP;
2359 cfg->fc_protocol = rtm->rtm_protocol;
2360 cfg->fc_type = rtm->rtm_type;
2362 if (rtm->rtm_type == RTN_UNREACHABLE ||
2363 rtm->rtm_type == RTN_BLACKHOLE ||
2364 rtm->rtm_type == RTN_PROHIBIT ||
2365 rtm->rtm_type == RTN_THROW)
2366 cfg->fc_flags |= RTF_REJECT;
2368 if (rtm->rtm_type == RTN_LOCAL)
2369 cfg->fc_flags |= RTF_LOCAL;
2371 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2372 cfg->fc_nlinfo.nlh = nlh;
2373 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2375 if (tb[RTA_GATEWAY]) {
2376 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2377 cfg->fc_flags |= RTF_GATEWAY;
2381 int plen = (rtm->rtm_dst_len + 7) >> 3;
2383 if (nla_len(tb[RTA_DST]) < plen)
2386 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2390 int plen = (rtm->rtm_src_len + 7) >> 3;
2392 if (nla_len(tb[RTA_SRC]) < plen)
2395 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2398 if (tb[RTA_PREFSRC])
2399 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2402 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2404 if (tb[RTA_PRIORITY])
2405 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2407 if (tb[RTA_METRICS]) {
2408 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2409 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2413 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2415 if (tb[RTA_MULTIPATH]) {
2416 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2417 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2425 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2427 struct fib6_config r_cfg;
2428 struct rtnexthop *rtnh;
2431 int err = 0, last_err = 0;
2434 rtnh = (struct rtnexthop *)cfg->fc_mp;
2435 remaining = cfg->fc_mp_len;
2437 /* Parse a Multipath Entry */
2438 while (rtnh_ok(rtnh, remaining)) {
2439 memcpy(&r_cfg, cfg, sizeof(*cfg));
2440 if (rtnh->rtnh_ifindex)
2441 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2443 attrlen = rtnh_attrlen(rtnh);
2445 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2447 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2449 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2450 r_cfg.fc_flags |= RTF_GATEWAY;
2453 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2456 /* If we are trying to remove a route, do not stop the
2457 * loop when ip6_route_del() fails (because next hop is
2458 * already gone), we should try to remove all next hops.
2461 /* If add fails, we should try to delete all
2462 * next hops that have been already added.
2468 /* Because each route is added like a single route we remove
2469 * this flag after the first nexthop (if there is a collision,
2470 * we have already fail to add the first nexthop:
2471 * fib6_add_rt2node() has reject it).
2473 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2474 rtnh = rtnh_next(rtnh, &remaining);
2480 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2482 struct fib6_config cfg;
2485 err = rtm_to_fib6_config(skb, nlh, &cfg);
2490 return ip6_route_multipath(&cfg, 0);
2492 return ip6_route_del(&cfg);
2495 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2497 struct fib6_config cfg;
2500 err = rtm_to_fib6_config(skb, nlh, &cfg);
2505 return ip6_route_multipath(&cfg, 1);
2507 return ip6_route_add(&cfg);
2510 static inline size_t rt6_nlmsg_size(void)
2512 return NLMSG_ALIGN(sizeof(struct rtmsg))
2513 + nla_total_size(16) /* RTA_SRC */
2514 + nla_total_size(16) /* RTA_DST */
2515 + nla_total_size(16) /* RTA_GATEWAY */
2516 + nla_total_size(16) /* RTA_PREFSRC */
2517 + nla_total_size(4) /* RTA_TABLE */
2518 + nla_total_size(4) /* RTA_IIF */
2519 + nla_total_size(4) /* RTA_OIF */
2520 + nla_total_size(4) /* RTA_PRIORITY */
2521 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2522 + nla_total_size(sizeof(struct rta_cacheinfo));
2525 static int rt6_fill_node(struct net *net,
2526 struct sk_buff *skb, struct rt6_info *rt,
2527 struct in6_addr *dst, struct in6_addr *src,
2528 int iif, int type, u32 portid, u32 seq,
2529 int prefix, int nowait, unsigned int flags)
2532 struct nlmsghdr *nlh;
2536 if (prefix) { /* user wants prefix routes only */
2537 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2538 /* success since this is not a prefix route */
2543 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2547 rtm = nlmsg_data(nlh);
2548 rtm->rtm_family = AF_INET6;
2549 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2550 rtm->rtm_src_len = rt->rt6i_src.plen;
2553 table = rt->rt6i_table->tb6_id;
2555 table = RT6_TABLE_UNSPEC;
2556 rtm->rtm_table = table;
2557 if (nla_put_u32(skb, RTA_TABLE, table))
2558 goto nla_put_failure;
2559 if (rt->rt6i_flags & RTF_REJECT) {
2560 switch (rt->dst.error) {
2562 rtm->rtm_type = RTN_BLACKHOLE;
2565 rtm->rtm_type = RTN_PROHIBIT;
2568 rtm->rtm_type = RTN_THROW;
2571 rtm->rtm_type = RTN_UNREACHABLE;
2575 else if (rt->rt6i_flags & RTF_LOCAL)
2576 rtm->rtm_type = RTN_LOCAL;
2577 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2578 rtm->rtm_type = RTN_LOCAL;
2580 rtm->rtm_type = RTN_UNICAST;
2582 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2583 rtm->rtm_protocol = rt->rt6i_protocol;
2584 if (rt->rt6i_flags & RTF_DYNAMIC)
2585 rtm->rtm_protocol = RTPROT_REDIRECT;
2586 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2587 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2588 rtm->rtm_protocol = RTPROT_RA;
2590 rtm->rtm_protocol = RTPROT_KERNEL;
2593 if (rt->rt6i_flags & RTF_CACHE)
2594 rtm->rtm_flags |= RTM_F_CLONED;
2597 if (nla_put(skb, RTA_DST, 16, dst))
2598 goto nla_put_failure;
2599 rtm->rtm_dst_len = 128;
2600 } else if (rtm->rtm_dst_len)
2601 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2602 goto nla_put_failure;
2603 #ifdef CONFIG_IPV6_SUBTREES
2605 if (nla_put(skb, RTA_SRC, 16, src))
2606 goto nla_put_failure;
2607 rtm->rtm_src_len = 128;
2608 } else if (rtm->rtm_src_len &&
2609 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2610 goto nla_put_failure;
2613 #ifdef CONFIG_IPV6_MROUTE
2614 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2615 int err = ip6mr_get_route(net, skb, rtm, nowait);
2620 goto nla_put_failure;
2622 if (err == -EMSGSIZE)
2623 goto nla_put_failure;
2628 if (nla_put_u32(skb, RTA_IIF, iif))
2629 goto nla_put_failure;
2631 struct in6_addr saddr_buf;
2632 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2633 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2634 goto nla_put_failure;
2637 if (rt->rt6i_prefsrc.plen) {
2638 struct in6_addr saddr_buf;
2639 saddr_buf = rt->rt6i_prefsrc.addr;
2640 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2641 goto nla_put_failure;
2644 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2645 goto nla_put_failure;
2647 if (rt->rt6i_flags & RTF_GATEWAY) {
2648 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2649 goto nla_put_failure;
2653 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2654 goto nla_put_failure;
2655 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2656 goto nla_put_failure;
2658 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2660 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2661 goto nla_put_failure;
2663 return nlmsg_end(skb, nlh);
2666 nlmsg_cancel(skb, nlh);
2670 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2672 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2675 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2676 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2677 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2681 return rt6_fill_node(arg->net,
2682 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2683 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2684 prefix, 0, NLM_F_MULTI);
2687 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2689 struct net *net = sock_net(in_skb->sk);
2690 struct nlattr *tb[RTA_MAX+1];
2691 struct rt6_info *rt;
2692 struct sk_buff *skb;
2695 int err, iif = 0, oif = 0;
2697 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2702 memset(&fl6, 0, sizeof(fl6));
2705 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2708 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2712 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2715 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2719 iif = nla_get_u32(tb[RTA_IIF]);
2722 oif = nla_get_u32(tb[RTA_OIF]);
2725 struct net_device *dev;
2728 dev = __dev_get_by_index(net, iif);
2734 fl6.flowi6_iif = iif;
2736 if (!ipv6_addr_any(&fl6.saddr))
2737 flags |= RT6_LOOKUP_F_HAS_SADDR;
2739 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2742 fl6.flowi6_oif = oif;
2744 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2747 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2754 /* Reserve room for dummy headers, this skb can pass
2755 through good chunk of routing engine.
2757 skb_reset_mac_header(skb);
2758 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2760 skb_dst_set(skb, &rt->dst);
2762 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2763 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2764 nlh->nlmsg_seq, 0, 0, 0);
2770 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2775 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2777 struct sk_buff *skb;
2778 struct net *net = info->nl_net;
2783 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2785 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2789 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2790 event, info->portid, seq, 0, 0, 0);
2792 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2793 WARN_ON(err == -EMSGSIZE);
2797 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2798 info->nlh, gfp_any());
2802 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2805 static int ip6_route_dev_notify(struct notifier_block *this,
2806 unsigned long event, void *ptr)
2808 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2809 struct net *net = dev_net(dev);
2811 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2812 net->ipv6.ip6_null_entry->dst.dev = dev;
2813 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2814 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2815 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2816 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2817 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2818 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2829 #ifdef CONFIG_PROC_FS
2840 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2842 struct seq_file *m = p_arg;
2844 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2846 #ifdef CONFIG_IPV6_SUBTREES
2847 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2849 seq_puts(m, "00000000000000000000000000000000 00 ");
2851 if (rt->rt6i_flags & RTF_GATEWAY) {
2852 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2854 seq_puts(m, "00000000000000000000000000000000");
2856 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2857 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2858 rt->dst.__use, rt->rt6i_flags,
2859 rt->dst.dev ? rt->dst.dev->name : "");
2863 static int ipv6_route_show(struct seq_file *m, void *v)
2865 struct net *net = (struct net *)m->private;
2866 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2870 static int ipv6_route_open(struct inode *inode, struct file *file)
2872 return single_open_net(inode, file, ipv6_route_show);
2875 static const struct file_operations ipv6_route_proc_fops = {
2876 .owner = THIS_MODULE,
2877 .open = ipv6_route_open,
2879 .llseek = seq_lseek,
2880 .release = single_release_net,
2883 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2885 struct net *net = (struct net *)seq->private;
2886 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2887 net->ipv6.rt6_stats->fib_nodes,
2888 net->ipv6.rt6_stats->fib_route_nodes,
2889 net->ipv6.rt6_stats->fib_rt_alloc,
2890 net->ipv6.rt6_stats->fib_rt_entries,
2891 net->ipv6.rt6_stats->fib_rt_cache,
2892 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2893 net->ipv6.rt6_stats->fib_discarded_routes);
2898 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2900 return single_open_net(inode, file, rt6_stats_seq_show);
2903 static const struct file_operations rt6_stats_seq_fops = {
2904 .owner = THIS_MODULE,
2905 .open = rt6_stats_seq_open,
2907 .llseek = seq_lseek,
2908 .release = single_release_net,
2910 #endif /* CONFIG_PROC_FS */
2912 #ifdef CONFIG_SYSCTL
2915 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2916 void __user *buffer, size_t *lenp, loff_t *ppos)
2923 net = (struct net *)ctl->extra1;
2924 delay = net->ipv6.sysctl.flush_delay;
2925 proc_dointvec(ctl, write, buffer, lenp, ppos);
2926 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2930 struct ctl_table ipv6_route_table_template[] = {
2932 .procname = "flush",
2933 .data = &init_net.ipv6.sysctl.flush_delay,
2934 .maxlen = sizeof(int),
2936 .proc_handler = ipv6_sysctl_rtcache_flush
2939 .procname = "gc_thresh",
2940 .data = &ip6_dst_ops_template.gc_thresh,
2941 .maxlen = sizeof(int),
2943 .proc_handler = proc_dointvec,
2946 .procname = "max_size",
2947 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2948 .maxlen = sizeof(int),
2950 .proc_handler = proc_dointvec,
2953 .procname = "gc_min_interval",
2954 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2955 .maxlen = sizeof(int),
2957 .proc_handler = proc_dointvec_jiffies,
2960 .procname = "gc_timeout",
2961 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2962 .maxlen = sizeof(int),
2964 .proc_handler = proc_dointvec_jiffies,
2967 .procname = "gc_interval",
2968 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2969 .maxlen = sizeof(int),
2971 .proc_handler = proc_dointvec_jiffies,
2974 .procname = "gc_elasticity",
2975 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2976 .maxlen = sizeof(int),
2978 .proc_handler = proc_dointvec,
2981 .procname = "mtu_expires",
2982 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2983 .maxlen = sizeof(int),
2985 .proc_handler = proc_dointvec_jiffies,
2988 .procname = "min_adv_mss",
2989 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2990 .maxlen = sizeof(int),
2992 .proc_handler = proc_dointvec,
2995 .procname = "gc_min_interval_ms",
2996 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2997 .maxlen = sizeof(int),
2999 .proc_handler = proc_dointvec_ms_jiffies,
3004 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3006 struct ctl_table *table;
3008 table = kmemdup(ipv6_route_table_template,
3009 sizeof(ipv6_route_table_template),
3013 table[0].data = &net->ipv6.sysctl.flush_delay;
3014 table[0].extra1 = net;
3015 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3016 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3017 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3018 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3019 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3020 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3021 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3022 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3023 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3025 /* Don't export sysctls to unprivileged users */
3026 if (net->user_ns != &init_user_ns)
3027 table[0].procname = NULL;
3034 static int __net_init ip6_route_net_init(struct net *net)
3038 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3039 sizeof(net->ipv6.ip6_dst_ops));
3041 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3042 goto out_ip6_dst_ops;
3044 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3045 sizeof(*net->ipv6.ip6_null_entry),
3047 if (!net->ipv6.ip6_null_entry)
3048 goto out_ip6_dst_entries;
3049 net->ipv6.ip6_null_entry->dst.path =
3050 (struct dst_entry *)net->ipv6.ip6_null_entry;
3051 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3052 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3053 ip6_template_metrics, true);
3055 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3056 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3057 sizeof(*net->ipv6.ip6_prohibit_entry),
3059 if (!net->ipv6.ip6_prohibit_entry)
3060 goto out_ip6_null_entry;
3061 net->ipv6.ip6_prohibit_entry->dst.path =
3062 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3063 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3064 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3065 ip6_template_metrics, true);
3067 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3068 sizeof(*net->ipv6.ip6_blk_hole_entry),
3070 if (!net->ipv6.ip6_blk_hole_entry)
3071 goto out_ip6_prohibit_entry;
3072 net->ipv6.ip6_blk_hole_entry->dst.path =
3073 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3074 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3075 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3076 ip6_template_metrics, true);
3079 net->ipv6.sysctl.flush_delay = 0;
3080 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3081 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3082 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3083 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3084 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3085 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3086 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3088 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3094 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3095 out_ip6_prohibit_entry:
3096 kfree(net->ipv6.ip6_prohibit_entry);
3098 kfree(net->ipv6.ip6_null_entry);
3100 out_ip6_dst_entries:
3101 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3106 static void __net_exit ip6_route_net_exit(struct net *net)
3108 kfree(net->ipv6.ip6_null_entry);
3109 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3110 kfree(net->ipv6.ip6_prohibit_entry);
3111 kfree(net->ipv6.ip6_blk_hole_entry);
3113 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3116 static int __net_init ip6_route_net_init_late(struct net *net)
3118 #ifdef CONFIG_PROC_FS
3119 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3120 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3125 static void __net_exit ip6_route_net_exit_late(struct net *net)
3127 #ifdef CONFIG_PROC_FS
3128 remove_proc_entry("ipv6_route", net->proc_net);
3129 remove_proc_entry("rt6_stats", net->proc_net);
3133 static struct pernet_operations ip6_route_net_ops = {
3134 .init = ip6_route_net_init,
3135 .exit = ip6_route_net_exit,
3138 static int __net_init ipv6_inetpeer_init(struct net *net)
3140 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3144 inet_peer_base_init(bp);
3145 net->ipv6.peers = bp;
3149 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3151 struct inet_peer_base *bp = net->ipv6.peers;
3153 net->ipv6.peers = NULL;
3154 inetpeer_invalidate_tree(bp);
3158 static struct pernet_operations ipv6_inetpeer_ops = {
3159 .init = ipv6_inetpeer_init,
3160 .exit = ipv6_inetpeer_exit,
3163 static struct pernet_operations ip6_route_net_late_ops = {
3164 .init = ip6_route_net_init_late,
3165 .exit = ip6_route_net_exit_late,
3168 static struct notifier_block ip6_route_dev_notifier = {
3169 .notifier_call = ip6_route_dev_notify,
3173 int __init ip6_route_init(void)
3178 ip6_dst_ops_template.kmem_cachep =
3179 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3180 SLAB_HWCACHE_ALIGN, NULL);
3181 if (!ip6_dst_ops_template.kmem_cachep)
3184 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3186 goto out_kmem_cache;
3188 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3190 goto out_dst_entries;
3192 ret = register_pernet_subsys(&ip6_route_net_ops);
3194 goto out_register_inetpeer;
3196 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3198 /* Registering of the loopback is done before this portion of code,
3199 * the loopback reference in rt6_info will not be taken, do it
3200 * manually for init_net */
3201 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3202 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3203 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3204 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3205 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3206 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3207 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3211 goto out_register_subsys;
3217 ret = fib6_rules_init();
3221 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3223 goto fib6_rules_init;
3226 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3227 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3228 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3229 goto out_register_late_subsys;
3231 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3233 goto out_register_late_subsys;
3238 out_register_late_subsys:
3239 unregister_pernet_subsys(&ip6_route_net_late_ops);
3241 fib6_rules_cleanup();
3246 out_register_subsys:
3247 unregister_pernet_subsys(&ip6_route_net_ops);
3248 out_register_inetpeer:
3249 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3251 dst_entries_destroy(&ip6_dst_blackhole_ops);
3253 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3257 void ip6_route_cleanup(void)
3259 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3260 unregister_pernet_subsys(&ip6_route_net_late_ops);
3261 fib6_rules_cleanup();
3264 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3265 unregister_pernet_subsys(&ip6_route_net_ops);
3266 dst_entries_destroy(&ip6_dst_blackhole_ops);
3267 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);