2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
89 struct sk_buff *skb, u32 mtu);
90 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
94 #ifdef CONFIG_IPV6_ROUTE_INFO
95 static struct rt6_info *rt6_add_route_info(struct net *net,
96 const struct in6_addr *prefix, int prefixlen,
97 const struct in6_addr *gwaddr, int ifindex,
99 static struct rt6_info *rt6_get_route_info(struct net *net,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr, int ifindex);
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
110 if (!(rt->dst.flags & DST_HOST))
113 peer = rt6_get_peer_create(rt);
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
138 struct in6_addr *p = &rt->rt6i_gateway;
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
143 return &ipv6_hdr(skb)->daddr;
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
151 struct rt6_info *rt = (struct rt6_info *) dst;
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
158 return neigh_create(&nd_tbl, daddr, dst->dev);
161 static struct dst_ops ip6_dst_ops_template = {
163 .protocol = cpu_to_be16(ETH_P_IPV6),
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
184 return mtu ? : dst->dev->mtu;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
203 static struct dst_ops ip6_dst_blackhole_ops = {
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
220 static const struct rt6_info ip6_null_entry_template = {
222 .__refcnt = ATOMIC_INIT(1),
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static int ip6_pkt_prohibit(struct sk_buff *skb);
238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
240 static const struct rt6_info ip6_prohibit_entry_template = {
242 .__refcnt = ATOMIC_INIT(1),
244 .obsolete = DST_OBSOLETE_FORCE_CHK,
246 .input = ip6_pkt_prohibit,
247 .output = ip6_pkt_prohibit_out,
249 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
250 .rt6i_protocol = RTPROT_KERNEL,
251 .rt6i_metric = ~(u32) 0,
252 .rt6i_ref = ATOMIC_INIT(1),
255 static const struct rt6_info ip6_blk_hole_entry_template = {
257 .__refcnt = ATOMIC_INIT(1),
259 .obsolete = DST_OBSOLETE_FORCE_CHK,
261 .input = dst_discard,
262 .output = dst_discard,
264 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
265 .rt6i_protocol = RTPROT_KERNEL,
266 .rt6i_metric = ~(u32) 0,
267 .rt6i_ref = ATOMIC_INIT(1),
272 /* allocate dst with ip6_dst_ops */
273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
274 struct net_device *dev,
276 struct fib6_table *table)
278 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
279 0, DST_OBSOLETE_FORCE_CHK, flags);
282 struct dst_entry *dst = &rt->dst;
284 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
285 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
286 rt->rt6i_genid = rt_genid_ipv6(net);
287 INIT_LIST_HEAD(&rt->rt6i_siblings);
292 static void ip6_dst_destroy(struct dst_entry *dst)
294 struct rt6_info *rt = (struct rt6_info *)dst;
295 struct inet6_dev *idev = rt->rt6i_idev;
296 struct dst_entry *from = dst->from;
298 if (!(rt->dst.flags & DST_HOST))
299 dst_destroy_metrics_generic(dst);
302 rt->rt6i_idev = NULL;
309 if (rt6_has_peer(rt)) {
310 struct inet_peer *peer = rt6_peer_ptr(rt);
315 void rt6_bind_peer(struct rt6_info *rt, int create)
317 struct inet_peer_base *base;
318 struct inet_peer *peer;
320 base = inetpeer_base_ptr(rt->_rt6i_peer);
324 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
326 if (!rt6_set_peer(rt, peer))
331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
334 struct rt6_info *rt = (struct rt6_info *)dst;
335 struct inet6_dev *idev = rt->rt6i_idev;
336 struct net_device *loopback_dev =
337 dev_net(dev)->loopback_dev;
339 if (dev != loopback_dev) {
340 if (idev && idev->dev == dev) {
341 struct inet6_dev *loopback_idev =
342 in6_dev_get(loopback_dev);
344 rt->rt6i_idev = loopback_idev;
351 static bool rt6_check_expired(const struct rt6_info *rt)
353 if (rt->rt6i_flags & RTF_EXPIRES) {
354 if (time_after(jiffies, rt->dst.expires))
356 } else if (rt->dst.from) {
357 return rt6_check_expired((struct rt6_info *) rt->dst.from);
362 static bool rt6_need_strict(const struct in6_addr *daddr)
364 return ipv6_addr_type(daddr) &
365 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
368 /* Multipath route selection:
369 * Hash based function using packet header and flowlabel.
370 * Adapted from fib_info_hashfn()
372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
373 const struct flowi6 *fl6)
375 unsigned int val = fl6->flowi6_proto;
377 val ^= ipv6_addr_hash(&fl6->daddr);
378 val ^= ipv6_addr_hash(&fl6->saddr);
380 /* Work only if this not encapsulated */
381 switch (fl6->flowi6_proto) {
385 val ^= (__force u16)fl6->fl6_sport;
386 val ^= (__force u16)fl6->fl6_dport;
390 val ^= (__force u16)fl6->fl6_icmp_type;
391 val ^= (__force u16)fl6->fl6_icmp_code;
394 /* RFC6438 recommands to use flowlabel */
395 val ^= (__force u32)fl6->flowlabel;
397 /* Perhaps, we need to tune, this function? */
398 val = val ^ (val >> 7) ^ (val >> 12);
399 return val % candidate_count;
402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct flowi6 *fl6, int oif,
406 struct rt6_info *sibling, *next_sibling;
409 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
410 /* Don't change the route, if route_choosen == 0
411 * (siblings does not include ourself)
414 list_for_each_entry_safe(sibling, next_sibling,
415 &match->rt6i_siblings, rt6i_siblings) {
417 if (route_choosen == 0) {
418 if (rt6_score_route(sibling, oif, strict) < 0)
428 * Route lookup. Any table->tb6_lock is implied.
431 static inline struct rt6_info *rt6_device_match(struct net *net,
433 const struct in6_addr *saddr,
437 struct rt6_info *local = NULL;
438 struct rt6_info *sprt;
440 if (!oif && ipv6_addr_any(saddr))
443 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
444 struct net_device *dev = sprt->dst.dev;
447 if (dev->ifindex == oif)
449 if (dev->flags & IFF_LOOPBACK) {
450 if (!sprt->rt6i_idev ||
451 sprt->rt6i_idev->dev->ifindex != oif) {
452 if (flags & RT6_LOOKUP_F_IFACE && oif)
454 if (local && (!oif ||
455 local->rt6i_idev->dev->ifindex == oif))
461 if (ipv6_chk_addr(net, saddr, dev,
462 flags & RT6_LOOKUP_F_IFACE))
471 if (flags & RT6_LOOKUP_F_IFACE)
472 return net->ipv6.ip6_null_entry;
478 #ifdef CONFIG_IPV6_ROUTER_PREF
479 static void rt6_probe(struct rt6_info *rt)
481 struct neighbour *neigh;
483 * Okay, this does not seem to be appropriate
484 * for now, however, we need to check if it
485 * is really so; aka Router Reachability Probing.
487 * Router Reachability Probe MUST be rate-limited
488 * to no more than one per minute.
490 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
493 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
495 write_lock(&neigh->lock);
496 if (neigh->nud_state & NUD_VALID)
501 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
502 struct in6_addr mcaddr;
503 struct in6_addr *target;
506 neigh->updated = jiffies;
507 write_unlock(&neigh->lock);
510 target = (struct in6_addr *)&rt->rt6i_gateway;
511 addrconf_addr_solict_mult(target, &mcaddr);
512 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
515 write_unlock(&neigh->lock);
517 rcu_read_unlock_bh();
520 static inline void rt6_probe(struct rt6_info *rt)
526 * Default Router Selection (RFC 2461 6.3.6)
528 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
530 struct net_device *dev = rt->dst.dev;
531 if (!oif || dev->ifindex == oif)
533 if ((dev->flags & IFF_LOOPBACK) &&
534 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
539 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
541 struct neighbour *neigh;
542 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
544 if (rt->rt6i_flags & RTF_NONEXTHOP ||
545 !(rt->rt6i_flags & RTF_GATEWAY))
546 return RT6_NUD_SUCCEED;
549 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
551 read_lock(&neigh->lock);
552 if (neigh->nud_state & NUD_VALID)
553 ret = RT6_NUD_SUCCEED;
554 #ifdef CONFIG_IPV6_ROUTER_PREF
555 else if (!(neigh->nud_state & NUD_FAILED))
556 ret = RT6_NUD_SUCCEED;
558 read_unlock(&neigh->lock);
560 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
561 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
563 rcu_read_unlock_bh();
568 static int rt6_score_route(struct rt6_info *rt, int oif,
573 m = rt6_check_dev(rt, oif);
574 if (!m && (strict & RT6_LOOKUP_F_IFACE))
575 return RT6_NUD_FAIL_HARD;
576 #ifdef CONFIG_IPV6_ROUTER_PREF
577 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
579 if (strict & RT6_LOOKUP_F_REACHABLE) {
580 int n = rt6_check_neigh(rt);
587 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
588 int *mpri, struct rt6_info *match,
592 bool match_do_rr = false;
594 if (rt6_check_expired(rt))
597 m = rt6_score_route(rt, oif, strict);
598 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
600 m = 0; /* lowest valid score */
605 if (strict & RT6_LOOKUP_F_REACHABLE)
609 *do_rr = match_do_rr;
617 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
618 struct rt6_info *rr_head,
619 u32 metric, int oif, int strict,
622 struct rt6_info *rt, *match;
626 for (rt = rr_head; rt && rt->rt6i_metric == metric;
627 rt = rt->dst.rt6_next)
628 match = find_match(rt, oif, strict, &mpri, match, do_rr);
629 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
630 rt = rt->dst.rt6_next)
631 match = find_match(rt, oif, strict, &mpri, match, do_rr);
636 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
638 struct rt6_info *match, *rt0;
644 fn->rr_ptr = rt0 = fn->leaf;
646 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
650 struct rt6_info *next = rt0->dst.rt6_next;
652 /* no entries matched; do round-robin */
653 if (!next || next->rt6i_metric != rt0->rt6i_metric)
660 net = dev_net(rt0->dst.dev);
661 return match ? match : net->ipv6.ip6_null_entry;
664 #ifdef CONFIG_IPV6_ROUTE_INFO
665 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
666 const struct in6_addr *gwaddr)
668 struct net *net = dev_net(dev);
669 struct route_info *rinfo = (struct route_info *) opt;
670 struct in6_addr prefix_buf, *prefix;
672 unsigned long lifetime;
675 if (len < sizeof(struct route_info)) {
679 /* Sanity check for prefix_len and length */
680 if (rinfo->length > 3) {
682 } else if (rinfo->prefix_len > 128) {
684 } else if (rinfo->prefix_len > 64) {
685 if (rinfo->length < 2) {
688 } else if (rinfo->prefix_len > 0) {
689 if (rinfo->length < 1) {
694 pref = rinfo->route_pref;
695 if (pref == ICMPV6_ROUTER_PREF_INVALID)
698 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
700 if (rinfo->length == 3)
701 prefix = (struct in6_addr *)rinfo->prefix;
703 /* this function is safe */
704 ipv6_addr_prefix(&prefix_buf,
705 (struct in6_addr *)rinfo->prefix,
707 prefix = &prefix_buf;
710 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
713 if (rt && !lifetime) {
719 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
722 rt->rt6i_flags = RTF_ROUTEINFO |
723 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
726 if (!addrconf_finite_timeout(lifetime))
727 rt6_clean_expires(rt);
729 rt6_set_expires(rt, jiffies + HZ * lifetime);
737 #define BACKTRACK(__net, saddr) \
739 if (rt == __net->ipv6.ip6_null_entry) { \
740 struct fib6_node *pn; \
742 if (fn->fn_flags & RTN_TL_ROOT) \
745 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
746 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
749 if (fn->fn_flags & RTN_RTINFO) \
755 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
756 struct fib6_table *table,
757 struct flowi6 *fl6, int flags)
759 struct fib6_node *fn;
762 read_lock_bh(&table->tb6_lock);
763 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
766 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
767 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
768 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
769 BACKTRACK(net, &fl6->saddr);
771 dst_use(&rt->dst, jiffies);
772 read_unlock_bh(&table->tb6_lock);
777 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
780 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
782 EXPORT_SYMBOL_GPL(ip6_route_lookup);
784 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
785 const struct in6_addr *saddr, int oif, int strict)
787 struct flowi6 fl6 = {
791 struct dst_entry *dst;
792 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
795 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
796 flags |= RT6_LOOKUP_F_HAS_SADDR;
799 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
801 return (struct rt6_info *) dst;
808 EXPORT_SYMBOL(rt6_lookup);
810 /* ip6_ins_rt is called with FREE table->tb6_lock.
811 It takes new route entry, the addition fails by any reason the
812 route is freed. In any case, if caller does not hold it, it may
816 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
819 struct fib6_table *table;
821 table = rt->rt6i_table;
822 write_lock_bh(&table->tb6_lock);
823 err = fib6_add(&table->tb6_root, rt, info);
824 write_unlock_bh(&table->tb6_lock);
829 int ip6_ins_rt(struct rt6_info *rt)
831 struct nl_info info = {
832 .nl_net = dev_net(rt->dst.dev),
834 return __ip6_ins_rt(rt, &info);
837 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
838 const struct in6_addr *daddr,
839 const struct in6_addr *saddr)
847 rt = ip6_rt_copy(ort, daddr);
850 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
851 if (ort->rt6i_dst.plen != 128 &&
852 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
853 rt->rt6i_flags |= RTF_ANYCAST;
854 rt->rt6i_gateway = *daddr;
857 rt->rt6i_flags |= RTF_CACHE;
859 #ifdef CONFIG_IPV6_SUBTREES
860 if (rt->rt6i_src.plen && saddr) {
861 rt->rt6i_src.addr = *saddr;
862 rt->rt6i_src.plen = 128;
870 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
871 const struct in6_addr *daddr)
873 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
876 rt->rt6i_flags |= RTF_CACHE;
880 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
881 struct flowi6 *fl6, int flags)
883 struct fib6_node *fn;
884 struct rt6_info *rt, *nrt;
888 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
890 strict |= flags & RT6_LOOKUP_F_IFACE;
893 read_lock_bh(&table->tb6_lock);
896 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
899 rt = rt6_select(fn, oif, strict | reachable);
900 if (rt->rt6i_nsiblings)
901 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
902 BACKTRACK(net, &fl6->saddr);
903 if (rt == net->ipv6.ip6_null_entry ||
904 rt->rt6i_flags & RTF_CACHE)
908 read_unlock_bh(&table->tb6_lock);
910 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
911 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
912 else if (!(rt->dst.flags & DST_HOST))
913 nrt = rt6_alloc_clone(rt, &fl6->daddr);
918 rt = nrt ? : net->ipv6.ip6_null_entry;
922 err = ip6_ins_rt(nrt);
931 * Race condition! In the gap, when table->tb6_lock was
932 * released someone could insert this route. Relookup.
943 read_unlock_bh(&table->tb6_lock);
945 rt->dst.lastuse = jiffies;
951 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
952 struct flowi6 *fl6, int flags)
954 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
957 static struct dst_entry *ip6_route_input_lookup(struct net *net,
958 struct net_device *dev,
959 struct flowi6 *fl6, int flags)
961 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
962 flags |= RT6_LOOKUP_F_IFACE;
964 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
967 void ip6_route_input(struct sk_buff *skb)
969 const struct ipv6hdr *iph = ipv6_hdr(skb);
970 struct net *net = dev_net(skb->dev);
971 int flags = RT6_LOOKUP_F_HAS_SADDR;
972 struct flowi6 fl6 = {
973 .flowi6_iif = skb->dev->ifindex,
976 .flowlabel = ip6_flowinfo(iph),
977 .flowi6_mark = skb->mark,
978 .flowi6_proto = iph->nexthdr,
981 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
984 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
985 struct flowi6 *fl6, int flags)
987 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
990 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
995 fl6->flowi6_iif = LOOPBACK_IFINDEX;
997 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
998 flags |= RT6_LOOKUP_F_IFACE;
1000 if (!ipv6_addr_any(&fl6->saddr))
1001 flags |= RT6_LOOKUP_F_HAS_SADDR;
1003 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1005 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1008 EXPORT_SYMBOL(ip6_route_output);
1010 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1012 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1013 struct dst_entry *new = NULL;
1015 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1019 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1020 rt6_init_peer(rt, net->ipv6.peers);
1023 new->input = dst_discard;
1024 new->output = dst_discard;
1026 if (dst_metrics_read_only(&ort->dst))
1027 new->_metrics = ort->dst._metrics;
1029 dst_copy_metrics(new, &ort->dst);
1030 rt->rt6i_idev = ort->rt6i_idev;
1032 in6_dev_hold(rt->rt6i_idev);
1034 rt->rt6i_gateway = ort->rt6i_gateway;
1035 rt->rt6i_flags = ort->rt6i_flags;
1036 rt->rt6i_metric = 0;
1038 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1039 #ifdef CONFIG_IPV6_SUBTREES
1040 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1046 dst_release(dst_orig);
1047 return new ? new : ERR_PTR(-ENOMEM);
1051 * Destination cache support functions
1054 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1056 struct rt6_info *rt;
1058 rt = (struct rt6_info *) dst;
1060 /* All IPV6 dsts are created with ->obsolete set to the value
1061 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1062 * into this function always.
1064 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1067 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1073 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1075 struct rt6_info *rt = (struct rt6_info *) dst;
1078 if (rt->rt6i_flags & RTF_CACHE) {
1079 if (rt6_check_expired(rt)) {
1091 static void ip6_link_failure(struct sk_buff *skb)
1093 struct rt6_info *rt;
1095 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1097 rt = (struct rt6_info *) skb_dst(skb);
1099 if (rt->rt6i_flags & RTF_CACHE) {
1103 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1104 rt->rt6i_node->fn_sernum = -1;
1109 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1110 struct sk_buff *skb, u32 mtu)
1112 struct rt6_info *rt6 = (struct rt6_info*)dst;
1115 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1116 struct net *net = dev_net(dst->dev);
1118 rt6->rt6i_flags |= RTF_MODIFIED;
1119 if (mtu < IPV6_MIN_MTU) {
1120 u32 features = dst_metric(dst, RTAX_FEATURES);
1122 features |= RTAX_FEATURE_ALLFRAG;
1123 dst_metric_set(dst, RTAX_FEATURES, features);
1125 dst_metric_set(dst, RTAX_MTU, mtu);
1126 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1130 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1133 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1134 struct dst_entry *dst;
1137 memset(&fl6, 0, sizeof(fl6));
1138 fl6.flowi6_oif = oif;
1139 fl6.flowi6_mark = mark;
1140 fl6.flowi6_flags = 0;
1141 fl6.daddr = iph->daddr;
1142 fl6.saddr = iph->saddr;
1143 fl6.flowlabel = ip6_flowinfo(iph);
1145 dst = ip6_route_output(net, NULL, &fl6);
1147 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1150 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1152 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1154 ip6_update_pmtu(skb, sock_net(sk), mtu,
1155 sk->sk_bound_dev_if, sk->sk_mark);
1157 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1159 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1161 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1162 struct dst_entry *dst;
1165 memset(&fl6, 0, sizeof(fl6));
1166 fl6.flowi6_oif = oif;
1167 fl6.flowi6_mark = mark;
1168 fl6.flowi6_flags = 0;
1169 fl6.daddr = iph->daddr;
1170 fl6.saddr = iph->saddr;
1171 fl6.flowlabel = ip6_flowinfo(iph);
1173 dst = ip6_route_output(net, NULL, &fl6);
1175 rt6_do_redirect(dst, NULL, skb);
1178 EXPORT_SYMBOL_GPL(ip6_redirect);
1180 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1183 const struct ipv6hdr *iph = ipv6_hdr(skb);
1184 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1185 struct dst_entry *dst;
1188 memset(&fl6, 0, sizeof(fl6));
1189 fl6.flowi6_oif = oif;
1190 fl6.flowi6_mark = mark;
1191 fl6.flowi6_flags = 0;
1192 fl6.daddr = msg->dest;
1193 fl6.saddr = iph->daddr;
1195 dst = ip6_route_output(net, NULL, &fl6);
1197 rt6_do_redirect(dst, NULL, skb);
1201 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1203 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1205 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1207 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1209 struct net_device *dev = dst->dev;
1210 unsigned int mtu = dst_mtu(dst);
1211 struct net *net = dev_net(dev);
1213 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1215 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1216 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1219 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1220 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1221 * IPV6_MAXPLEN is also valid and means: "any MSS,
1222 * rely only on pmtu discovery"
1224 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1229 static unsigned int ip6_mtu(const struct dst_entry *dst)
1231 struct inet6_dev *idev;
1232 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1240 idev = __in6_dev_get(dst->dev);
1242 mtu = idev->cnf.mtu6;
1248 static struct dst_entry *icmp6_dst_gc_list;
1249 static DEFINE_SPINLOCK(icmp6_dst_lock);
1251 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1254 struct dst_entry *dst;
1255 struct rt6_info *rt;
1256 struct inet6_dev *idev = in6_dev_get(dev);
1257 struct net *net = dev_net(dev);
1259 if (unlikely(!idev))
1260 return ERR_PTR(-ENODEV);
1262 rt = ip6_dst_alloc(net, dev, 0, NULL);
1263 if (unlikely(!rt)) {
1265 dst = ERR_PTR(-ENOMEM);
1269 rt->dst.flags |= DST_HOST;
1270 rt->dst.output = ip6_output;
1271 atomic_set(&rt->dst.__refcnt, 1);
1272 rt->rt6i_dst.addr = fl6->daddr;
1273 rt->rt6i_dst.plen = 128;
1274 rt->rt6i_idev = idev;
1275 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1277 spin_lock_bh(&icmp6_dst_lock);
1278 rt->dst.next = icmp6_dst_gc_list;
1279 icmp6_dst_gc_list = &rt->dst;
1280 spin_unlock_bh(&icmp6_dst_lock);
1282 fib6_force_start_gc(net);
1284 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1290 int icmp6_dst_gc(void)
1292 struct dst_entry *dst, **pprev;
1295 spin_lock_bh(&icmp6_dst_lock);
1296 pprev = &icmp6_dst_gc_list;
1298 while ((dst = *pprev) != NULL) {
1299 if (!atomic_read(&dst->__refcnt)) {
1308 spin_unlock_bh(&icmp6_dst_lock);
1313 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1316 struct dst_entry *dst, **pprev;
1318 spin_lock_bh(&icmp6_dst_lock);
1319 pprev = &icmp6_dst_gc_list;
1320 while ((dst = *pprev) != NULL) {
1321 struct rt6_info *rt = (struct rt6_info *) dst;
1322 if (func(rt, arg)) {
1329 spin_unlock_bh(&icmp6_dst_lock);
1332 static int ip6_dst_gc(struct dst_ops *ops)
1334 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1335 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1336 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1337 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1338 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1339 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1342 entries = dst_entries_get_fast(ops);
1343 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1344 entries <= rt_max_size)
1347 net->ipv6.ip6_rt_gc_expire++;
1348 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1349 entries = dst_entries_get_slow(ops);
1350 if (entries < ops->gc_thresh)
1351 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1353 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1354 return entries > rt_max_size;
1357 int ip6_dst_hoplimit(struct dst_entry *dst)
1359 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1360 if (hoplimit == 0) {
1361 struct net_device *dev = dst->dev;
1362 struct inet6_dev *idev;
1365 idev = __in6_dev_get(dev);
1367 hoplimit = idev->cnf.hop_limit;
1369 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1374 EXPORT_SYMBOL(ip6_dst_hoplimit);
1380 int ip6_route_add(struct fib6_config *cfg)
1383 struct net *net = cfg->fc_nlinfo.nl_net;
1384 struct rt6_info *rt = NULL;
1385 struct net_device *dev = NULL;
1386 struct inet6_dev *idev = NULL;
1387 struct fib6_table *table;
1390 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1392 #ifndef CONFIG_IPV6_SUBTREES
1393 if (cfg->fc_src_len)
1396 if (cfg->fc_ifindex) {
1398 dev = dev_get_by_index(net, cfg->fc_ifindex);
1401 idev = in6_dev_get(dev);
1406 if (cfg->fc_metric == 0)
1407 cfg->fc_metric = IP6_RT_PRIO_USER;
1410 if (cfg->fc_nlinfo.nlh &&
1411 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1412 table = fib6_get_table(net, cfg->fc_table);
1414 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1415 table = fib6_new_table(net, cfg->fc_table);
1418 table = fib6_new_table(net, cfg->fc_table);
1424 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1431 if (cfg->fc_flags & RTF_EXPIRES)
1432 rt6_set_expires(rt, jiffies +
1433 clock_t_to_jiffies(cfg->fc_expires));
1435 rt6_clean_expires(rt);
1437 if (cfg->fc_protocol == RTPROT_UNSPEC)
1438 cfg->fc_protocol = RTPROT_BOOT;
1439 rt->rt6i_protocol = cfg->fc_protocol;
1441 addr_type = ipv6_addr_type(&cfg->fc_dst);
1443 if (addr_type & IPV6_ADDR_MULTICAST)
1444 rt->dst.input = ip6_mc_input;
1445 else if (cfg->fc_flags & RTF_LOCAL)
1446 rt->dst.input = ip6_input;
1448 rt->dst.input = ip6_forward;
1450 rt->dst.output = ip6_output;
1452 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1453 rt->rt6i_dst.plen = cfg->fc_dst_len;
1454 if (rt->rt6i_dst.plen == 128)
1455 rt->dst.flags |= DST_HOST;
1457 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1458 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1463 dst_init_metrics(&rt->dst, metrics, 0);
1465 #ifdef CONFIG_IPV6_SUBTREES
1466 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1467 rt->rt6i_src.plen = cfg->fc_src_len;
1470 rt->rt6i_metric = cfg->fc_metric;
1472 /* We cannot add true routes via loopback here,
1473 they would result in kernel looping; promote them to reject routes
1475 if ((cfg->fc_flags & RTF_REJECT) ||
1476 (dev && (dev->flags & IFF_LOOPBACK) &&
1477 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1478 !(cfg->fc_flags & RTF_LOCAL))) {
1479 /* hold loopback dev/idev if we haven't done so. */
1480 if (dev != net->loopback_dev) {
1485 dev = net->loopback_dev;
1487 idev = in6_dev_get(dev);
1493 rt->dst.output = ip6_pkt_discard_out;
1494 rt->dst.input = ip6_pkt_discard;
1495 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1496 switch (cfg->fc_type) {
1498 rt->dst.error = -EINVAL;
1501 rt->dst.error = -EACCES;
1504 rt->dst.error = -EAGAIN;
1507 rt->dst.error = -ENETUNREACH;
1513 if (cfg->fc_flags & RTF_GATEWAY) {
1514 const struct in6_addr *gw_addr;
1517 gw_addr = &cfg->fc_gateway;
1518 rt->rt6i_gateway = *gw_addr;
1519 gwa_type = ipv6_addr_type(gw_addr);
1521 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1522 struct rt6_info *grt;
1524 /* IPv6 strictly inhibits using not link-local
1525 addresses as nexthop address.
1526 Otherwise, router will not able to send redirects.
1527 It is very good, but in some (rare!) circumstances
1528 (SIT, PtP, NBMA NOARP links) it is handy to allow
1529 some exceptions. --ANK
1532 if (!(gwa_type & IPV6_ADDR_UNICAST))
1535 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1537 err = -EHOSTUNREACH;
1541 if (dev != grt->dst.dev) {
1547 idev = grt->rt6i_idev;
1549 in6_dev_hold(grt->rt6i_idev);
1551 if (!(grt->rt6i_flags & RTF_GATEWAY))
1559 if (!dev || (dev->flags & IFF_LOOPBACK))
1567 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1568 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1572 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1573 rt->rt6i_prefsrc.plen = 128;
1575 rt->rt6i_prefsrc.plen = 0;
1577 rt->rt6i_flags = cfg->fc_flags;
1584 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1585 int type = nla_type(nla);
1588 if (type > RTAX_MAX) {
1593 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1599 rt->rt6i_idev = idev;
1600 rt->rt6i_table = table;
1602 cfg->fc_nlinfo.nl_net = dev_net(dev);
1604 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1616 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1619 struct fib6_table *table;
1620 struct net *net = dev_net(rt->dst.dev);
1622 if (rt == net->ipv6.ip6_null_entry) {
1627 table = rt->rt6i_table;
1628 write_lock_bh(&table->tb6_lock);
1629 err = fib6_del(rt, info);
1630 write_unlock_bh(&table->tb6_lock);
1637 int ip6_del_rt(struct rt6_info *rt)
1639 struct nl_info info = {
1640 .nl_net = dev_net(rt->dst.dev),
1642 return __ip6_del_rt(rt, &info);
1645 static int ip6_route_del(struct fib6_config *cfg)
1647 struct fib6_table *table;
1648 struct fib6_node *fn;
1649 struct rt6_info *rt;
1652 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1656 read_lock_bh(&table->tb6_lock);
1658 fn = fib6_locate(&table->tb6_root,
1659 &cfg->fc_dst, cfg->fc_dst_len,
1660 &cfg->fc_src, cfg->fc_src_len);
1663 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1664 if (cfg->fc_ifindex &&
1666 rt->dst.dev->ifindex != cfg->fc_ifindex))
1668 if (cfg->fc_flags & RTF_GATEWAY &&
1669 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1671 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1674 read_unlock_bh(&table->tb6_lock);
1676 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1679 read_unlock_bh(&table->tb6_lock);
1684 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1686 struct net *net = dev_net(skb->dev);
1687 struct netevent_redirect netevent;
1688 struct rt6_info *rt, *nrt = NULL;
1689 struct ndisc_options ndopts;
1690 struct inet6_dev *in6_dev;
1691 struct neighbour *neigh;
1693 int optlen, on_link;
1696 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1697 optlen -= sizeof(*msg);
1700 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1704 msg = (struct rd_msg *)icmp6_hdr(skb);
1706 if (ipv6_addr_is_multicast(&msg->dest)) {
1707 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1712 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1714 } else if (ipv6_addr_type(&msg->target) !=
1715 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1716 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1720 in6_dev = __in6_dev_get(skb->dev);
1723 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1727 * The IP source address of the Redirect MUST be the same as the current
1728 * first-hop router for the specified ICMP Destination Address.
1731 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1732 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1737 if (ndopts.nd_opts_tgt_lladdr) {
1738 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1741 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1746 rt = (struct rt6_info *) dst;
1747 if (rt == net->ipv6.ip6_null_entry) {
1748 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1752 /* Redirect received -> path was valid.
1753 * Look, redirects are sent only in response to data packets,
1754 * so that this nexthop apparently is reachable. --ANK
1756 dst_confirm(&rt->dst);
1758 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1763 * We have finally decided to accept it.
1766 neigh_update(neigh, lladdr, NUD_STALE,
1767 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1768 NEIGH_UPDATE_F_OVERRIDE|
1769 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1770 NEIGH_UPDATE_F_ISROUTER))
1773 nrt = ip6_rt_copy(rt, &msg->dest);
1777 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1779 nrt->rt6i_flags &= ~RTF_GATEWAY;
1781 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1783 if (ip6_ins_rt(nrt))
1786 netevent.old = &rt->dst;
1787 netevent.new = &nrt->dst;
1788 netevent.daddr = &msg->dest;
1789 netevent.neigh = neigh;
1790 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1792 if (rt->rt6i_flags & RTF_CACHE) {
1793 rt = (struct rt6_info *) dst_clone(&rt->dst);
1798 neigh_release(neigh);
1802 * Misc support functions
1805 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1806 const struct in6_addr *dest)
1808 struct net *net = dev_net(ort->dst.dev);
1809 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1813 rt->dst.input = ort->dst.input;
1814 rt->dst.output = ort->dst.output;
1815 rt->dst.flags |= DST_HOST;
1817 rt->rt6i_dst.addr = *dest;
1818 rt->rt6i_dst.plen = 128;
1819 dst_copy_metrics(&rt->dst, &ort->dst);
1820 rt->dst.error = ort->dst.error;
1821 rt->rt6i_idev = ort->rt6i_idev;
1823 in6_dev_hold(rt->rt6i_idev);
1824 rt->dst.lastuse = jiffies;
1826 rt->rt6i_gateway = ort->rt6i_gateway;
1827 rt->rt6i_flags = ort->rt6i_flags;
1828 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1829 (RTF_DEFAULT | RTF_ADDRCONF))
1830 rt6_set_from(rt, ort);
1831 rt->rt6i_metric = 0;
1833 #ifdef CONFIG_IPV6_SUBTREES
1834 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1836 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1837 rt->rt6i_table = ort->rt6i_table;
1842 #ifdef CONFIG_IPV6_ROUTE_INFO
1843 static struct rt6_info *rt6_get_route_info(struct net *net,
1844 const struct in6_addr *prefix, int prefixlen,
1845 const struct in6_addr *gwaddr, int ifindex)
1847 struct fib6_node *fn;
1848 struct rt6_info *rt = NULL;
1849 struct fib6_table *table;
1851 table = fib6_get_table(net, RT6_TABLE_INFO);
1855 read_lock_bh(&table->tb6_lock);
1856 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1860 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1861 if (rt->dst.dev->ifindex != ifindex)
1863 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1865 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1871 read_unlock_bh(&table->tb6_lock);
1875 static struct rt6_info *rt6_add_route_info(struct net *net,
1876 const struct in6_addr *prefix, int prefixlen,
1877 const struct in6_addr *gwaddr, int ifindex,
1880 struct fib6_config cfg = {
1881 .fc_table = RT6_TABLE_INFO,
1882 .fc_metric = IP6_RT_PRIO_USER,
1883 .fc_ifindex = ifindex,
1884 .fc_dst_len = prefixlen,
1885 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1886 RTF_UP | RTF_PREF(pref),
1887 .fc_nlinfo.portid = 0,
1888 .fc_nlinfo.nlh = NULL,
1889 .fc_nlinfo.nl_net = net,
1892 cfg.fc_dst = *prefix;
1893 cfg.fc_gateway = *gwaddr;
1895 /* We should treat it as a default route if prefix length is 0. */
1897 cfg.fc_flags |= RTF_DEFAULT;
1899 ip6_route_add(&cfg);
1901 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1905 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1907 struct rt6_info *rt;
1908 struct fib6_table *table;
1910 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1914 read_lock_bh(&table->tb6_lock);
1915 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1916 if (dev == rt->dst.dev &&
1917 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1918 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1923 read_unlock_bh(&table->tb6_lock);
1927 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1928 struct net_device *dev,
1931 struct fib6_config cfg = {
1932 .fc_table = RT6_TABLE_DFLT,
1933 .fc_metric = IP6_RT_PRIO_USER,
1934 .fc_ifindex = dev->ifindex,
1935 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1936 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1937 .fc_nlinfo.portid = 0,
1938 .fc_nlinfo.nlh = NULL,
1939 .fc_nlinfo.nl_net = dev_net(dev),
1942 cfg.fc_gateway = *gwaddr;
1944 ip6_route_add(&cfg);
1946 return rt6_get_dflt_router(gwaddr, dev);
1949 void rt6_purge_dflt_routers(struct net *net)
1951 struct rt6_info *rt;
1952 struct fib6_table *table;
1954 /* NOTE: Keep consistent with rt6_get_dflt_router */
1955 table = fib6_get_table(net, RT6_TABLE_DFLT);
1960 read_lock_bh(&table->tb6_lock);
1961 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1962 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1963 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
1965 read_unlock_bh(&table->tb6_lock);
1970 read_unlock_bh(&table->tb6_lock);
1973 static void rtmsg_to_fib6_config(struct net *net,
1974 struct in6_rtmsg *rtmsg,
1975 struct fib6_config *cfg)
1977 memset(cfg, 0, sizeof(*cfg));
1979 cfg->fc_table = RT6_TABLE_MAIN;
1980 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1981 cfg->fc_metric = rtmsg->rtmsg_metric;
1982 cfg->fc_expires = rtmsg->rtmsg_info;
1983 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1984 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1985 cfg->fc_flags = rtmsg->rtmsg_flags;
1987 cfg->fc_nlinfo.nl_net = net;
1989 cfg->fc_dst = rtmsg->rtmsg_dst;
1990 cfg->fc_src = rtmsg->rtmsg_src;
1991 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1994 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1996 struct fib6_config cfg;
1997 struct in6_rtmsg rtmsg;
2001 case SIOCADDRT: /* Add a route */
2002 case SIOCDELRT: /* Delete a route */
2003 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2005 err = copy_from_user(&rtmsg, arg,
2006 sizeof(struct in6_rtmsg));
2010 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2015 err = ip6_route_add(&cfg);
2018 err = ip6_route_del(&cfg);
2032 * Drop the packet on the floor
2035 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2038 struct dst_entry *dst = skb_dst(skb);
2039 switch (ipstats_mib_noroutes) {
2040 case IPSTATS_MIB_INNOROUTES:
2041 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2042 if (type == IPV6_ADDR_ANY) {
2043 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2044 IPSTATS_MIB_INADDRERRORS);
2048 case IPSTATS_MIB_OUTNOROUTES:
2049 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2050 ipstats_mib_noroutes);
2053 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2058 static int ip6_pkt_discard(struct sk_buff *skb)
2060 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2063 static int ip6_pkt_discard_out(struct sk_buff *skb)
2065 skb->dev = skb_dst(skb)->dev;
2066 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2069 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2071 static int ip6_pkt_prohibit(struct sk_buff *skb)
2073 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2076 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2078 skb->dev = skb_dst(skb)->dev;
2079 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2085 * Allocate a dst for local (unicast / anycast) address.
2088 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2089 const struct in6_addr *addr,
2092 struct net *net = dev_net(idev->dev);
2093 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2096 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2097 return ERR_PTR(-ENOMEM);
2102 rt->dst.flags |= DST_HOST;
2103 rt->dst.input = ip6_input;
2104 rt->dst.output = ip6_output;
2105 rt->rt6i_idev = idev;
2107 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2109 rt->rt6i_flags |= RTF_ANYCAST;
2111 rt->rt6i_flags |= RTF_LOCAL;
2113 rt->rt6i_dst.addr = *addr;
2114 rt->rt6i_dst.plen = 128;
2115 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2117 atomic_set(&rt->dst.__refcnt, 1);
2122 int ip6_route_get_saddr(struct net *net,
2123 struct rt6_info *rt,
2124 const struct in6_addr *daddr,
2126 struct in6_addr *saddr)
2128 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2130 if (rt->rt6i_prefsrc.plen)
2131 *saddr = rt->rt6i_prefsrc.addr;
2133 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2134 daddr, prefs, saddr);
2138 /* remove deleted ip from prefsrc entries */
2139 struct arg_dev_net_ip {
2140 struct net_device *dev;
2142 struct in6_addr *addr;
2145 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2147 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2148 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2149 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2151 if (((void *)rt->dst.dev == dev || !dev) &&
2152 rt != net->ipv6.ip6_null_entry &&
2153 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2154 /* remove prefsrc entry */
2155 rt->rt6i_prefsrc.plen = 0;
2160 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2162 struct net *net = dev_net(ifp->idev->dev);
2163 struct arg_dev_net_ip adni = {
2164 .dev = ifp->idev->dev,
2168 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2171 struct arg_dev_net {
2172 struct net_device *dev;
2176 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2178 const struct arg_dev_net *adn = arg;
2179 const struct net_device *dev = adn->dev;
2181 if ((rt->dst.dev == dev || !dev) &&
2182 rt != adn->net->ipv6.ip6_null_entry)
2188 void rt6_ifdown(struct net *net, struct net_device *dev)
2190 struct arg_dev_net adn = {
2195 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2196 icmp6_clean_all(fib6_ifdown, &adn);
2199 struct rt6_mtu_change_arg {
2200 struct net_device *dev;
2204 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2206 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2207 struct inet6_dev *idev;
2209 /* In IPv6 pmtu discovery is not optional,
2210 so that RTAX_MTU lock cannot disable it.
2211 We still use this lock to block changes
2212 caused by addrconf/ndisc.
2215 idev = __in6_dev_get(arg->dev);
2219 /* For administrative MTU increase, there is no way to discover
2220 IPv6 PMTU increase, so PMTU increase should be updated here.
2221 Since RFC 1981 doesn't include administrative MTU increase
2222 update PMTU increase is a MUST. (i.e. jumbo frame)
2225 If new MTU is less than route PMTU, this new MTU will be the
2226 lowest MTU in the path, update the route PMTU to reflect PMTU
2227 decreases; if new MTU is greater than route PMTU, and the
2228 old MTU is the lowest MTU in the path, update the route PMTU
2229 to reflect the increase. In this case if the other nodes' MTU
2230 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2233 if (rt->dst.dev == arg->dev &&
2234 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2235 (dst_mtu(&rt->dst) >= arg->mtu ||
2236 (dst_mtu(&rt->dst) < arg->mtu &&
2237 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2238 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2243 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2245 struct rt6_mtu_change_arg arg = {
2250 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2253 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2254 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2255 [RTA_OIF] = { .type = NLA_U32 },
2256 [RTA_IIF] = { .type = NLA_U32 },
2257 [RTA_PRIORITY] = { .type = NLA_U32 },
2258 [RTA_METRICS] = { .type = NLA_NESTED },
2259 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2262 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2263 struct fib6_config *cfg)
2266 struct nlattr *tb[RTA_MAX+1];
2269 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2274 rtm = nlmsg_data(nlh);
2275 memset(cfg, 0, sizeof(*cfg));
2277 cfg->fc_table = rtm->rtm_table;
2278 cfg->fc_dst_len = rtm->rtm_dst_len;
2279 cfg->fc_src_len = rtm->rtm_src_len;
2280 cfg->fc_flags = RTF_UP;
2281 cfg->fc_protocol = rtm->rtm_protocol;
2282 cfg->fc_type = rtm->rtm_type;
2284 if (rtm->rtm_type == RTN_UNREACHABLE ||
2285 rtm->rtm_type == RTN_BLACKHOLE ||
2286 rtm->rtm_type == RTN_PROHIBIT ||
2287 rtm->rtm_type == RTN_THROW)
2288 cfg->fc_flags |= RTF_REJECT;
2290 if (rtm->rtm_type == RTN_LOCAL)
2291 cfg->fc_flags |= RTF_LOCAL;
2293 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2294 cfg->fc_nlinfo.nlh = nlh;
2295 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2297 if (tb[RTA_GATEWAY]) {
2298 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2299 cfg->fc_flags |= RTF_GATEWAY;
2303 int plen = (rtm->rtm_dst_len + 7) >> 3;
2305 if (nla_len(tb[RTA_DST]) < plen)
2308 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2312 int plen = (rtm->rtm_src_len + 7) >> 3;
2314 if (nla_len(tb[RTA_SRC]) < plen)
2317 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2320 if (tb[RTA_PREFSRC])
2321 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2324 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2326 if (tb[RTA_PRIORITY])
2327 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2329 if (tb[RTA_METRICS]) {
2330 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2331 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2335 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2337 if (tb[RTA_MULTIPATH]) {
2338 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2339 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2347 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2349 struct fib6_config r_cfg;
2350 struct rtnexthop *rtnh;
2353 int err = 0, last_err = 0;
2356 rtnh = (struct rtnexthop *)cfg->fc_mp;
2357 remaining = cfg->fc_mp_len;
2359 /* Parse a Multipath Entry */
2360 while (rtnh_ok(rtnh, remaining)) {
2361 memcpy(&r_cfg, cfg, sizeof(*cfg));
2362 if (rtnh->rtnh_ifindex)
2363 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2365 attrlen = rtnh_attrlen(rtnh);
2367 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2369 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2371 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2372 r_cfg.fc_flags |= RTF_GATEWAY;
2375 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2378 /* If we are trying to remove a route, do not stop the
2379 * loop when ip6_route_del() fails (because next hop is
2380 * already gone), we should try to remove all next hops.
2383 /* If add fails, we should try to delete all
2384 * next hops that have been already added.
2390 /* Because each route is added like a single route we remove
2391 * this flag after the first nexthop (if there is a collision,
2392 * we have already fail to add the first nexthop:
2393 * fib6_add_rt2node() has reject it).
2395 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2396 rtnh = rtnh_next(rtnh, &remaining);
2402 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2404 struct fib6_config cfg;
2407 err = rtm_to_fib6_config(skb, nlh, &cfg);
2412 return ip6_route_multipath(&cfg, 0);
2414 return ip6_route_del(&cfg);
2417 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2419 struct fib6_config cfg;
2422 err = rtm_to_fib6_config(skb, nlh, &cfg);
2427 return ip6_route_multipath(&cfg, 1);
2429 return ip6_route_add(&cfg);
2432 static inline size_t rt6_nlmsg_size(void)
2434 return NLMSG_ALIGN(sizeof(struct rtmsg))
2435 + nla_total_size(16) /* RTA_SRC */
2436 + nla_total_size(16) /* RTA_DST */
2437 + nla_total_size(16) /* RTA_GATEWAY */
2438 + nla_total_size(16) /* RTA_PREFSRC */
2439 + nla_total_size(4) /* RTA_TABLE */
2440 + nla_total_size(4) /* RTA_IIF */
2441 + nla_total_size(4) /* RTA_OIF */
2442 + nla_total_size(4) /* RTA_PRIORITY */
2443 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2444 + nla_total_size(sizeof(struct rta_cacheinfo));
2447 static int rt6_fill_node(struct net *net,
2448 struct sk_buff *skb, struct rt6_info *rt,
2449 struct in6_addr *dst, struct in6_addr *src,
2450 int iif, int type, u32 portid, u32 seq,
2451 int prefix, int nowait, unsigned int flags)
2454 struct nlmsghdr *nlh;
2458 if (prefix) { /* user wants prefix routes only */
2459 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2460 /* success since this is not a prefix route */
2465 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2469 rtm = nlmsg_data(nlh);
2470 rtm->rtm_family = AF_INET6;
2471 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2472 rtm->rtm_src_len = rt->rt6i_src.plen;
2475 table = rt->rt6i_table->tb6_id;
2477 table = RT6_TABLE_UNSPEC;
2478 rtm->rtm_table = table;
2479 if (nla_put_u32(skb, RTA_TABLE, table))
2480 goto nla_put_failure;
2481 if (rt->rt6i_flags & RTF_REJECT) {
2482 switch (rt->dst.error) {
2484 rtm->rtm_type = RTN_BLACKHOLE;
2487 rtm->rtm_type = RTN_PROHIBIT;
2490 rtm->rtm_type = RTN_THROW;
2493 rtm->rtm_type = RTN_UNREACHABLE;
2497 else if (rt->rt6i_flags & RTF_LOCAL)
2498 rtm->rtm_type = RTN_LOCAL;
2499 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2500 rtm->rtm_type = RTN_LOCAL;
2502 rtm->rtm_type = RTN_UNICAST;
2504 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2505 rtm->rtm_protocol = rt->rt6i_protocol;
2506 if (rt->rt6i_flags & RTF_DYNAMIC)
2507 rtm->rtm_protocol = RTPROT_REDIRECT;
2508 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2509 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2510 rtm->rtm_protocol = RTPROT_RA;
2512 rtm->rtm_protocol = RTPROT_KERNEL;
2515 if (rt->rt6i_flags & RTF_CACHE)
2516 rtm->rtm_flags |= RTM_F_CLONED;
2519 if (nla_put(skb, RTA_DST, 16, dst))
2520 goto nla_put_failure;
2521 rtm->rtm_dst_len = 128;
2522 } else if (rtm->rtm_dst_len)
2523 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2524 goto nla_put_failure;
2525 #ifdef CONFIG_IPV6_SUBTREES
2527 if (nla_put(skb, RTA_SRC, 16, src))
2528 goto nla_put_failure;
2529 rtm->rtm_src_len = 128;
2530 } else if (rtm->rtm_src_len &&
2531 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2532 goto nla_put_failure;
2535 #ifdef CONFIG_IPV6_MROUTE
2536 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2537 int err = ip6mr_get_route(net, skb, rtm, nowait);
2542 goto nla_put_failure;
2544 if (err == -EMSGSIZE)
2545 goto nla_put_failure;
2550 if (nla_put_u32(skb, RTA_IIF, iif))
2551 goto nla_put_failure;
2553 struct in6_addr saddr_buf;
2554 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2555 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2556 goto nla_put_failure;
2559 if (rt->rt6i_prefsrc.plen) {
2560 struct in6_addr saddr_buf;
2561 saddr_buf = rt->rt6i_prefsrc.addr;
2562 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2563 goto nla_put_failure;
2566 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2567 goto nla_put_failure;
2569 if (rt->rt6i_flags & RTF_GATEWAY) {
2570 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2571 goto nla_put_failure;
2575 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2576 goto nla_put_failure;
2577 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2578 goto nla_put_failure;
2580 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2582 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2583 goto nla_put_failure;
2585 return nlmsg_end(skb, nlh);
2588 nlmsg_cancel(skb, nlh);
2592 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2594 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2597 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2598 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2599 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2603 return rt6_fill_node(arg->net,
2604 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2605 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2606 prefix, 0, NLM_F_MULTI);
2609 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2611 struct net *net = sock_net(in_skb->sk);
2612 struct nlattr *tb[RTA_MAX+1];
2613 struct rt6_info *rt;
2614 struct sk_buff *skb;
2617 int err, iif = 0, oif = 0;
2619 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2624 memset(&fl6, 0, sizeof(fl6));
2627 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2630 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2634 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2637 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2641 iif = nla_get_u32(tb[RTA_IIF]);
2644 oif = nla_get_u32(tb[RTA_OIF]);
2647 struct net_device *dev;
2650 dev = __dev_get_by_index(net, iif);
2656 fl6.flowi6_iif = iif;
2658 if (!ipv6_addr_any(&fl6.saddr))
2659 flags |= RT6_LOOKUP_F_HAS_SADDR;
2661 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2664 fl6.flowi6_oif = oif;
2666 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2669 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2676 /* Reserve room for dummy headers, this skb can pass
2677 through good chunk of routing engine.
2679 skb_reset_mac_header(skb);
2680 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2682 skb_dst_set(skb, &rt->dst);
2684 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2685 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2686 nlh->nlmsg_seq, 0, 0, 0);
2692 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2697 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2699 struct sk_buff *skb;
2700 struct net *net = info->nl_net;
2705 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2707 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2711 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2712 event, info->portid, seq, 0, 0, 0);
2714 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2715 WARN_ON(err == -EMSGSIZE);
2719 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2720 info->nlh, gfp_any());
2724 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2727 static int ip6_route_dev_notify(struct notifier_block *this,
2728 unsigned long event, void *ptr)
2730 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2731 struct net *net = dev_net(dev);
2733 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2734 net->ipv6.ip6_null_entry->dst.dev = dev;
2735 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2736 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2737 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2738 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2739 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2740 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2751 #ifdef CONFIG_PROC_FS
2762 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2764 struct seq_file *m = p_arg;
2766 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2768 #ifdef CONFIG_IPV6_SUBTREES
2769 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2771 seq_puts(m, "00000000000000000000000000000000 00 ");
2773 if (rt->rt6i_flags & RTF_GATEWAY) {
2774 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2776 seq_puts(m, "00000000000000000000000000000000");
2778 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2779 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2780 rt->dst.__use, rt->rt6i_flags,
2781 rt->dst.dev ? rt->dst.dev->name : "");
2785 static int ipv6_route_show(struct seq_file *m, void *v)
2787 struct net *net = (struct net *)m->private;
2788 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2792 static int ipv6_route_open(struct inode *inode, struct file *file)
2794 return single_open_net(inode, file, ipv6_route_show);
2797 static const struct file_operations ipv6_route_proc_fops = {
2798 .owner = THIS_MODULE,
2799 .open = ipv6_route_open,
2801 .llseek = seq_lseek,
2802 .release = single_release_net,
2805 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2807 struct net *net = (struct net *)seq->private;
2808 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2809 net->ipv6.rt6_stats->fib_nodes,
2810 net->ipv6.rt6_stats->fib_route_nodes,
2811 net->ipv6.rt6_stats->fib_rt_alloc,
2812 net->ipv6.rt6_stats->fib_rt_entries,
2813 net->ipv6.rt6_stats->fib_rt_cache,
2814 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2815 net->ipv6.rt6_stats->fib_discarded_routes);
2820 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2822 return single_open_net(inode, file, rt6_stats_seq_show);
2825 static const struct file_operations rt6_stats_seq_fops = {
2826 .owner = THIS_MODULE,
2827 .open = rt6_stats_seq_open,
2829 .llseek = seq_lseek,
2830 .release = single_release_net,
2832 #endif /* CONFIG_PROC_FS */
2834 #ifdef CONFIG_SYSCTL
2837 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2838 void __user *buffer, size_t *lenp, loff_t *ppos)
2845 net = (struct net *)ctl->extra1;
2846 delay = net->ipv6.sysctl.flush_delay;
2847 proc_dointvec(ctl, write, buffer, lenp, ppos);
2848 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2852 struct ctl_table ipv6_route_table_template[] = {
2854 .procname = "flush",
2855 .data = &init_net.ipv6.sysctl.flush_delay,
2856 .maxlen = sizeof(int),
2858 .proc_handler = ipv6_sysctl_rtcache_flush
2861 .procname = "gc_thresh",
2862 .data = &ip6_dst_ops_template.gc_thresh,
2863 .maxlen = sizeof(int),
2865 .proc_handler = proc_dointvec,
2868 .procname = "max_size",
2869 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2870 .maxlen = sizeof(int),
2872 .proc_handler = proc_dointvec,
2875 .procname = "gc_min_interval",
2876 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2877 .maxlen = sizeof(int),
2879 .proc_handler = proc_dointvec_jiffies,
2882 .procname = "gc_timeout",
2883 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2884 .maxlen = sizeof(int),
2886 .proc_handler = proc_dointvec_jiffies,
2889 .procname = "gc_interval",
2890 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2891 .maxlen = sizeof(int),
2893 .proc_handler = proc_dointvec_jiffies,
2896 .procname = "gc_elasticity",
2897 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2898 .maxlen = sizeof(int),
2900 .proc_handler = proc_dointvec,
2903 .procname = "mtu_expires",
2904 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2905 .maxlen = sizeof(int),
2907 .proc_handler = proc_dointvec_jiffies,
2910 .procname = "min_adv_mss",
2911 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2912 .maxlen = sizeof(int),
2914 .proc_handler = proc_dointvec,
2917 .procname = "gc_min_interval_ms",
2918 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2919 .maxlen = sizeof(int),
2921 .proc_handler = proc_dointvec_ms_jiffies,
2926 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2928 struct ctl_table *table;
2930 table = kmemdup(ipv6_route_table_template,
2931 sizeof(ipv6_route_table_template),
2935 table[0].data = &net->ipv6.sysctl.flush_delay;
2936 table[0].extra1 = net;
2937 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2938 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2939 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2940 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2941 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2942 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2943 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2944 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2945 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2947 /* Don't export sysctls to unprivileged users */
2948 if (net->user_ns != &init_user_ns)
2949 table[0].procname = NULL;
2956 static int __net_init ip6_route_net_init(struct net *net)
2960 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2961 sizeof(net->ipv6.ip6_dst_ops));
2963 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2964 goto out_ip6_dst_ops;
2966 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2967 sizeof(*net->ipv6.ip6_null_entry),
2969 if (!net->ipv6.ip6_null_entry)
2970 goto out_ip6_dst_entries;
2971 net->ipv6.ip6_null_entry->dst.path =
2972 (struct dst_entry *)net->ipv6.ip6_null_entry;
2973 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2974 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2975 ip6_template_metrics, true);
2977 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2978 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2979 sizeof(*net->ipv6.ip6_prohibit_entry),
2981 if (!net->ipv6.ip6_prohibit_entry)
2982 goto out_ip6_null_entry;
2983 net->ipv6.ip6_prohibit_entry->dst.path =
2984 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2985 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2986 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2987 ip6_template_metrics, true);
2989 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2990 sizeof(*net->ipv6.ip6_blk_hole_entry),
2992 if (!net->ipv6.ip6_blk_hole_entry)
2993 goto out_ip6_prohibit_entry;
2994 net->ipv6.ip6_blk_hole_entry->dst.path =
2995 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2996 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2997 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2998 ip6_template_metrics, true);
3001 net->ipv6.sysctl.flush_delay = 0;
3002 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3003 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3004 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3005 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3006 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3007 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3008 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3010 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3016 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3017 out_ip6_prohibit_entry:
3018 kfree(net->ipv6.ip6_prohibit_entry);
3020 kfree(net->ipv6.ip6_null_entry);
3022 out_ip6_dst_entries:
3023 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3028 static void __net_exit ip6_route_net_exit(struct net *net)
3030 kfree(net->ipv6.ip6_null_entry);
3031 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3032 kfree(net->ipv6.ip6_prohibit_entry);
3033 kfree(net->ipv6.ip6_blk_hole_entry);
3035 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3038 static int __net_init ip6_route_net_init_late(struct net *net)
3040 #ifdef CONFIG_PROC_FS
3041 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3042 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3047 static void __net_exit ip6_route_net_exit_late(struct net *net)
3049 #ifdef CONFIG_PROC_FS
3050 remove_proc_entry("ipv6_route", net->proc_net);
3051 remove_proc_entry("rt6_stats", net->proc_net);
3055 static struct pernet_operations ip6_route_net_ops = {
3056 .init = ip6_route_net_init,
3057 .exit = ip6_route_net_exit,
3060 static int __net_init ipv6_inetpeer_init(struct net *net)
3062 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3066 inet_peer_base_init(bp);
3067 net->ipv6.peers = bp;
3071 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3073 struct inet_peer_base *bp = net->ipv6.peers;
3075 net->ipv6.peers = NULL;
3076 inetpeer_invalidate_tree(bp);
3080 static struct pernet_operations ipv6_inetpeer_ops = {
3081 .init = ipv6_inetpeer_init,
3082 .exit = ipv6_inetpeer_exit,
3085 static struct pernet_operations ip6_route_net_late_ops = {
3086 .init = ip6_route_net_init_late,
3087 .exit = ip6_route_net_exit_late,
3090 static struct notifier_block ip6_route_dev_notifier = {
3091 .notifier_call = ip6_route_dev_notify,
3095 int __init ip6_route_init(void)
3100 ip6_dst_ops_template.kmem_cachep =
3101 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3102 SLAB_HWCACHE_ALIGN, NULL);
3103 if (!ip6_dst_ops_template.kmem_cachep)
3106 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3108 goto out_kmem_cache;
3110 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3112 goto out_dst_entries;
3114 ret = register_pernet_subsys(&ip6_route_net_ops);
3116 goto out_register_inetpeer;
3118 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3120 /* Registering of the loopback is done before this portion of code,
3121 * the loopback reference in rt6_info will not be taken, do it
3122 * manually for init_net */
3123 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3124 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3125 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3126 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3127 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3128 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3129 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3133 goto out_register_subsys;
3139 ret = fib6_rules_init();
3143 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3145 goto fib6_rules_init;
3148 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3149 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3150 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3151 goto out_register_late_subsys;
3153 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3155 goto out_register_late_subsys;
3160 out_register_late_subsys:
3161 unregister_pernet_subsys(&ip6_route_net_late_ops);
3163 fib6_rules_cleanup();
3168 out_register_subsys:
3169 unregister_pernet_subsys(&ip6_route_net_ops);
3170 out_register_inetpeer:
3171 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3173 dst_entries_destroy(&ip6_dst_blackhole_ops);
3175 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3179 void ip6_route_cleanup(void)
3181 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3182 unregister_pernet_subsys(&ip6_route_net_late_ops);
3183 fib6_rules_cleanup();
3186 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3187 unregister_pernet_subsys(&ip6_route_net_ops);
3188 dst_entries_destroy(&ip6_dst_blackhole_ops);
3189 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);