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_sk_redirect(struct sk_buff *skb, struct sock *sk)
1182 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1184 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1186 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1188 struct net_device *dev = dst->dev;
1189 unsigned int mtu = dst_mtu(dst);
1190 struct net *net = dev_net(dev);
1192 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1194 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1195 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1198 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1199 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1200 * IPV6_MAXPLEN is also valid and means: "any MSS,
1201 * rely only on pmtu discovery"
1203 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1208 static unsigned int ip6_mtu(const struct dst_entry *dst)
1210 struct inet6_dev *idev;
1211 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1219 idev = __in6_dev_get(dst->dev);
1221 mtu = idev->cnf.mtu6;
1227 static struct dst_entry *icmp6_dst_gc_list;
1228 static DEFINE_SPINLOCK(icmp6_dst_lock);
1230 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1233 struct dst_entry *dst;
1234 struct rt6_info *rt;
1235 struct inet6_dev *idev = in6_dev_get(dev);
1236 struct net *net = dev_net(dev);
1238 if (unlikely(!idev))
1239 return ERR_PTR(-ENODEV);
1241 rt = ip6_dst_alloc(net, dev, 0, NULL);
1242 if (unlikely(!rt)) {
1244 dst = ERR_PTR(-ENOMEM);
1248 rt->dst.flags |= DST_HOST;
1249 rt->dst.output = ip6_output;
1250 atomic_set(&rt->dst.__refcnt, 1);
1251 rt->rt6i_dst.addr = fl6->daddr;
1252 rt->rt6i_dst.plen = 128;
1253 rt->rt6i_idev = idev;
1254 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1256 spin_lock_bh(&icmp6_dst_lock);
1257 rt->dst.next = icmp6_dst_gc_list;
1258 icmp6_dst_gc_list = &rt->dst;
1259 spin_unlock_bh(&icmp6_dst_lock);
1261 fib6_force_start_gc(net);
1263 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1269 int icmp6_dst_gc(void)
1271 struct dst_entry *dst, **pprev;
1274 spin_lock_bh(&icmp6_dst_lock);
1275 pprev = &icmp6_dst_gc_list;
1277 while ((dst = *pprev) != NULL) {
1278 if (!atomic_read(&dst->__refcnt)) {
1287 spin_unlock_bh(&icmp6_dst_lock);
1292 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1295 struct dst_entry *dst, **pprev;
1297 spin_lock_bh(&icmp6_dst_lock);
1298 pprev = &icmp6_dst_gc_list;
1299 while ((dst = *pprev) != NULL) {
1300 struct rt6_info *rt = (struct rt6_info *) dst;
1301 if (func(rt, arg)) {
1308 spin_unlock_bh(&icmp6_dst_lock);
1311 static int ip6_dst_gc(struct dst_ops *ops)
1313 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1314 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1315 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1316 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1317 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1318 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1321 entries = dst_entries_get_fast(ops);
1322 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1323 entries <= rt_max_size)
1326 net->ipv6.ip6_rt_gc_expire++;
1327 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1328 entries = dst_entries_get_slow(ops);
1329 if (entries < ops->gc_thresh)
1330 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1332 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1333 return entries > rt_max_size;
1336 int ip6_dst_hoplimit(struct dst_entry *dst)
1338 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1339 if (hoplimit == 0) {
1340 struct net_device *dev = dst->dev;
1341 struct inet6_dev *idev;
1344 idev = __in6_dev_get(dev);
1346 hoplimit = idev->cnf.hop_limit;
1348 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1353 EXPORT_SYMBOL(ip6_dst_hoplimit);
1359 int ip6_route_add(struct fib6_config *cfg)
1362 struct net *net = cfg->fc_nlinfo.nl_net;
1363 struct rt6_info *rt = NULL;
1364 struct net_device *dev = NULL;
1365 struct inet6_dev *idev = NULL;
1366 struct fib6_table *table;
1369 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1371 #ifndef CONFIG_IPV6_SUBTREES
1372 if (cfg->fc_src_len)
1375 if (cfg->fc_ifindex) {
1377 dev = dev_get_by_index(net, cfg->fc_ifindex);
1380 idev = in6_dev_get(dev);
1385 if (cfg->fc_metric == 0)
1386 cfg->fc_metric = IP6_RT_PRIO_USER;
1389 if (cfg->fc_nlinfo.nlh &&
1390 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1391 table = fib6_get_table(net, cfg->fc_table);
1393 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1394 table = fib6_new_table(net, cfg->fc_table);
1397 table = fib6_new_table(net, cfg->fc_table);
1403 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1410 if (cfg->fc_flags & RTF_EXPIRES)
1411 rt6_set_expires(rt, jiffies +
1412 clock_t_to_jiffies(cfg->fc_expires));
1414 rt6_clean_expires(rt);
1416 if (cfg->fc_protocol == RTPROT_UNSPEC)
1417 cfg->fc_protocol = RTPROT_BOOT;
1418 rt->rt6i_protocol = cfg->fc_protocol;
1420 addr_type = ipv6_addr_type(&cfg->fc_dst);
1422 if (addr_type & IPV6_ADDR_MULTICAST)
1423 rt->dst.input = ip6_mc_input;
1424 else if (cfg->fc_flags & RTF_LOCAL)
1425 rt->dst.input = ip6_input;
1427 rt->dst.input = ip6_forward;
1429 rt->dst.output = ip6_output;
1431 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1432 rt->rt6i_dst.plen = cfg->fc_dst_len;
1433 if (rt->rt6i_dst.plen == 128)
1434 rt->dst.flags |= DST_HOST;
1436 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1437 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1442 dst_init_metrics(&rt->dst, metrics, 0);
1444 #ifdef CONFIG_IPV6_SUBTREES
1445 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1446 rt->rt6i_src.plen = cfg->fc_src_len;
1449 rt->rt6i_metric = cfg->fc_metric;
1451 /* We cannot add true routes via loopback here,
1452 they would result in kernel looping; promote them to reject routes
1454 if ((cfg->fc_flags & RTF_REJECT) ||
1455 (dev && (dev->flags & IFF_LOOPBACK) &&
1456 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1457 !(cfg->fc_flags & RTF_LOCAL))) {
1458 /* hold loopback dev/idev if we haven't done so. */
1459 if (dev != net->loopback_dev) {
1464 dev = net->loopback_dev;
1466 idev = in6_dev_get(dev);
1472 rt->dst.output = ip6_pkt_discard_out;
1473 rt->dst.input = ip6_pkt_discard;
1474 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1475 switch (cfg->fc_type) {
1477 rt->dst.error = -EINVAL;
1480 rt->dst.error = -EACCES;
1483 rt->dst.error = -EAGAIN;
1486 rt->dst.error = -ENETUNREACH;
1492 if (cfg->fc_flags & RTF_GATEWAY) {
1493 const struct in6_addr *gw_addr;
1496 gw_addr = &cfg->fc_gateway;
1497 rt->rt6i_gateway = *gw_addr;
1498 gwa_type = ipv6_addr_type(gw_addr);
1500 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1501 struct rt6_info *grt;
1503 /* IPv6 strictly inhibits using not link-local
1504 addresses as nexthop address.
1505 Otherwise, router will not able to send redirects.
1506 It is very good, but in some (rare!) circumstances
1507 (SIT, PtP, NBMA NOARP links) it is handy to allow
1508 some exceptions. --ANK
1511 if (!(gwa_type & IPV6_ADDR_UNICAST))
1514 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1516 err = -EHOSTUNREACH;
1520 if (dev != grt->dst.dev) {
1526 idev = grt->rt6i_idev;
1528 in6_dev_hold(grt->rt6i_idev);
1530 if (!(grt->rt6i_flags & RTF_GATEWAY))
1538 if (!dev || (dev->flags & IFF_LOOPBACK))
1546 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1547 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1551 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1552 rt->rt6i_prefsrc.plen = 128;
1554 rt->rt6i_prefsrc.plen = 0;
1556 rt->rt6i_flags = cfg->fc_flags;
1563 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1564 int type = nla_type(nla);
1567 if (type > RTAX_MAX) {
1572 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1578 rt->rt6i_idev = idev;
1579 rt->rt6i_table = table;
1581 cfg->fc_nlinfo.nl_net = dev_net(dev);
1583 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1595 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1598 struct fib6_table *table;
1599 struct net *net = dev_net(rt->dst.dev);
1601 if (rt == net->ipv6.ip6_null_entry) {
1606 table = rt->rt6i_table;
1607 write_lock_bh(&table->tb6_lock);
1608 err = fib6_del(rt, info);
1609 write_unlock_bh(&table->tb6_lock);
1616 int ip6_del_rt(struct rt6_info *rt)
1618 struct nl_info info = {
1619 .nl_net = dev_net(rt->dst.dev),
1621 return __ip6_del_rt(rt, &info);
1624 static int ip6_route_del(struct fib6_config *cfg)
1626 struct fib6_table *table;
1627 struct fib6_node *fn;
1628 struct rt6_info *rt;
1631 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1635 read_lock_bh(&table->tb6_lock);
1637 fn = fib6_locate(&table->tb6_root,
1638 &cfg->fc_dst, cfg->fc_dst_len,
1639 &cfg->fc_src, cfg->fc_src_len);
1642 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1643 if (cfg->fc_ifindex &&
1645 rt->dst.dev->ifindex != cfg->fc_ifindex))
1647 if (cfg->fc_flags & RTF_GATEWAY &&
1648 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1650 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1653 read_unlock_bh(&table->tb6_lock);
1655 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1658 read_unlock_bh(&table->tb6_lock);
1663 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1665 struct net *net = dev_net(skb->dev);
1666 struct netevent_redirect netevent;
1667 struct rt6_info *rt, *nrt = NULL;
1668 struct ndisc_options ndopts;
1669 struct inet6_dev *in6_dev;
1670 struct neighbour *neigh;
1672 int optlen, on_link;
1675 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1676 optlen -= sizeof(*msg);
1679 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1683 msg = (struct rd_msg *)icmp6_hdr(skb);
1685 if (ipv6_addr_is_multicast(&msg->dest)) {
1686 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1691 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1693 } else if (ipv6_addr_type(&msg->target) !=
1694 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1695 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1699 in6_dev = __in6_dev_get(skb->dev);
1702 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1706 * The IP source address of the Redirect MUST be the same as the current
1707 * first-hop router for the specified ICMP Destination Address.
1710 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1711 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1716 if (ndopts.nd_opts_tgt_lladdr) {
1717 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1720 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1725 rt = (struct rt6_info *) dst;
1726 if (rt == net->ipv6.ip6_null_entry) {
1727 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1731 /* Redirect received -> path was valid.
1732 * Look, redirects are sent only in response to data packets,
1733 * so that this nexthop apparently is reachable. --ANK
1735 dst_confirm(&rt->dst);
1737 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1742 * We have finally decided to accept it.
1745 neigh_update(neigh, lladdr, NUD_STALE,
1746 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1747 NEIGH_UPDATE_F_OVERRIDE|
1748 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1749 NEIGH_UPDATE_F_ISROUTER))
1752 nrt = ip6_rt_copy(rt, &msg->dest);
1756 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1758 nrt->rt6i_flags &= ~RTF_GATEWAY;
1760 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1762 if (ip6_ins_rt(nrt))
1765 netevent.old = &rt->dst;
1766 netevent.new = &nrt->dst;
1767 netevent.daddr = &msg->dest;
1768 netevent.neigh = neigh;
1769 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1771 if (rt->rt6i_flags & RTF_CACHE) {
1772 rt = (struct rt6_info *) dst_clone(&rt->dst);
1777 neigh_release(neigh);
1781 * Misc support functions
1784 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1785 const struct in6_addr *dest)
1787 struct net *net = dev_net(ort->dst.dev);
1788 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1792 rt->dst.input = ort->dst.input;
1793 rt->dst.output = ort->dst.output;
1794 rt->dst.flags |= DST_HOST;
1796 rt->rt6i_dst.addr = *dest;
1797 rt->rt6i_dst.plen = 128;
1798 dst_copy_metrics(&rt->dst, &ort->dst);
1799 rt->dst.error = ort->dst.error;
1800 rt->rt6i_idev = ort->rt6i_idev;
1802 in6_dev_hold(rt->rt6i_idev);
1803 rt->dst.lastuse = jiffies;
1805 rt->rt6i_gateway = ort->rt6i_gateway;
1806 rt->rt6i_flags = ort->rt6i_flags;
1807 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1808 (RTF_DEFAULT | RTF_ADDRCONF))
1809 rt6_set_from(rt, ort);
1810 rt->rt6i_metric = 0;
1812 #ifdef CONFIG_IPV6_SUBTREES
1813 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1815 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1816 rt->rt6i_table = ort->rt6i_table;
1821 #ifdef CONFIG_IPV6_ROUTE_INFO
1822 static struct rt6_info *rt6_get_route_info(struct net *net,
1823 const struct in6_addr *prefix, int prefixlen,
1824 const struct in6_addr *gwaddr, int ifindex)
1826 struct fib6_node *fn;
1827 struct rt6_info *rt = NULL;
1828 struct fib6_table *table;
1830 table = fib6_get_table(net, RT6_TABLE_INFO);
1834 read_lock_bh(&table->tb6_lock);
1835 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1839 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1840 if (rt->dst.dev->ifindex != ifindex)
1842 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1844 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1850 read_unlock_bh(&table->tb6_lock);
1854 static struct rt6_info *rt6_add_route_info(struct net *net,
1855 const struct in6_addr *prefix, int prefixlen,
1856 const struct in6_addr *gwaddr, int ifindex,
1859 struct fib6_config cfg = {
1860 .fc_table = RT6_TABLE_INFO,
1861 .fc_metric = IP6_RT_PRIO_USER,
1862 .fc_ifindex = ifindex,
1863 .fc_dst_len = prefixlen,
1864 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1865 RTF_UP | RTF_PREF(pref),
1866 .fc_nlinfo.portid = 0,
1867 .fc_nlinfo.nlh = NULL,
1868 .fc_nlinfo.nl_net = net,
1871 cfg.fc_dst = *prefix;
1872 cfg.fc_gateway = *gwaddr;
1874 /* We should treat it as a default route if prefix length is 0. */
1876 cfg.fc_flags |= RTF_DEFAULT;
1878 ip6_route_add(&cfg);
1880 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1884 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1886 struct rt6_info *rt;
1887 struct fib6_table *table;
1889 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1893 read_lock_bh(&table->tb6_lock);
1894 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1895 if (dev == rt->dst.dev &&
1896 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1897 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1902 read_unlock_bh(&table->tb6_lock);
1906 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1907 struct net_device *dev,
1910 struct fib6_config cfg = {
1911 .fc_table = RT6_TABLE_DFLT,
1912 .fc_metric = IP6_RT_PRIO_USER,
1913 .fc_ifindex = dev->ifindex,
1914 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1915 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1916 .fc_nlinfo.portid = 0,
1917 .fc_nlinfo.nlh = NULL,
1918 .fc_nlinfo.nl_net = dev_net(dev),
1921 cfg.fc_gateway = *gwaddr;
1923 ip6_route_add(&cfg);
1925 return rt6_get_dflt_router(gwaddr, dev);
1928 void rt6_purge_dflt_routers(struct net *net)
1930 struct rt6_info *rt;
1931 struct fib6_table *table;
1933 /* NOTE: Keep consistent with rt6_get_dflt_router */
1934 table = fib6_get_table(net, RT6_TABLE_DFLT);
1939 read_lock_bh(&table->tb6_lock);
1940 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1941 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1942 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
1944 read_unlock_bh(&table->tb6_lock);
1949 read_unlock_bh(&table->tb6_lock);
1952 static void rtmsg_to_fib6_config(struct net *net,
1953 struct in6_rtmsg *rtmsg,
1954 struct fib6_config *cfg)
1956 memset(cfg, 0, sizeof(*cfg));
1958 cfg->fc_table = RT6_TABLE_MAIN;
1959 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1960 cfg->fc_metric = rtmsg->rtmsg_metric;
1961 cfg->fc_expires = rtmsg->rtmsg_info;
1962 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1963 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1964 cfg->fc_flags = rtmsg->rtmsg_flags;
1966 cfg->fc_nlinfo.nl_net = net;
1968 cfg->fc_dst = rtmsg->rtmsg_dst;
1969 cfg->fc_src = rtmsg->rtmsg_src;
1970 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1973 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1975 struct fib6_config cfg;
1976 struct in6_rtmsg rtmsg;
1980 case SIOCADDRT: /* Add a route */
1981 case SIOCDELRT: /* Delete a route */
1982 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1984 err = copy_from_user(&rtmsg, arg,
1985 sizeof(struct in6_rtmsg));
1989 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1994 err = ip6_route_add(&cfg);
1997 err = ip6_route_del(&cfg);
2011 * Drop the packet on the floor
2014 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2017 struct dst_entry *dst = skb_dst(skb);
2018 switch (ipstats_mib_noroutes) {
2019 case IPSTATS_MIB_INNOROUTES:
2020 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2021 if (type == IPV6_ADDR_ANY) {
2022 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2023 IPSTATS_MIB_INADDRERRORS);
2027 case IPSTATS_MIB_OUTNOROUTES:
2028 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2029 ipstats_mib_noroutes);
2032 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2037 static int ip6_pkt_discard(struct sk_buff *skb)
2039 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2042 static int ip6_pkt_discard_out(struct sk_buff *skb)
2044 skb->dev = skb_dst(skb)->dev;
2045 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2048 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2050 static int ip6_pkt_prohibit(struct sk_buff *skb)
2052 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2055 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2057 skb->dev = skb_dst(skb)->dev;
2058 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2064 * Allocate a dst for local (unicast / anycast) address.
2067 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2068 const struct in6_addr *addr,
2071 struct net *net = dev_net(idev->dev);
2072 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2075 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2076 return ERR_PTR(-ENOMEM);
2081 rt->dst.flags |= DST_HOST;
2082 rt->dst.input = ip6_input;
2083 rt->dst.output = ip6_output;
2084 rt->rt6i_idev = idev;
2086 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2088 rt->rt6i_flags |= RTF_ANYCAST;
2090 rt->rt6i_flags |= RTF_LOCAL;
2092 rt->rt6i_dst.addr = *addr;
2093 rt->rt6i_dst.plen = 128;
2094 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2096 atomic_set(&rt->dst.__refcnt, 1);
2101 int ip6_route_get_saddr(struct net *net,
2102 struct rt6_info *rt,
2103 const struct in6_addr *daddr,
2105 struct in6_addr *saddr)
2107 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2109 if (rt->rt6i_prefsrc.plen)
2110 *saddr = rt->rt6i_prefsrc.addr;
2112 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2113 daddr, prefs, saddr);
2117 /* remove deleted ip from prefsrc entries */
2118 struct arg_dev_net_ip {
2119 struct net_device *dev;
2121 struct in6_addr *addr;
2124 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2126 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2127 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2128 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2130 if (((void *)rt->dst.dev == dev || !dev) &&
2131 rt != net->ipv6.ip6_null_entry &&
2132 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2133 /* remove prefsrc entry */
2134 rt->rt6i_prefsrc.plen = 0;
2139 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2141 struct net *net = dev_net(ifp->idev->dev);
2142 struct arg_dev_net_ip adni = {
2143 .dev = ifp->idev->dev,
2147 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2150 struct arg_dev_net {
2151 struct net_device *dev;
2155 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2157 const struct arg_dev_net *adn = arg;
2158 const struct net_device *dev = adn->dev;
2160 if ((rt->dst.dev == dev || !dev) &&
2161 rt != adn->net->ipv6.ip6_null_entry)
2167 void rt6_ifdown(struct net *net, struct net_device *dev)
2169 struct arg_dev_net adn = {
2174 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2175 icmp6_clean_all(fib6_ifdown, &adn);
2178 struct rt6_mtu_change_arg {
2179 struct net_device *dev;
2183 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2185 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2186 struct inet6_dev *idev;
2188 /* In IPv6 pmtu discovery is not optional,
2189 so that RTAX_MTU lock cannot disable it.
2190 We still use this lock to block changes
2191 caused by addrconf/ndisc.
2194 idev = __in6_dev_get(arg->dev);
2198 /* For administrative MTU increase, there is no way to discover
2199 IPv6 PMTU increase, so PMTU increase should be updated here.
2200 Since RFC 1981 doesn't include administrative MTU increase
2201 update PMTU increase is a MUST. (i.e. jumbo frame)
2204 If new MTU is less than route PMTU, this new MTU will be the
2205 lowest MTU in the path, update the route PMTU to reflect PMTU
2206 decreases; if new MTU is greater than route PMTU, and the
2207 old MTU is the lowest MTU in the path, update the route PMTU
2208 to reflect the increase. In this case if the other nodes' MTU
2209 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2212 if (rt->dst.dev == arg->dev &&
2213 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2214 (dst_mtu(&rt->dst) >= arg->mtu ||
2215 (dst_mtu(&rt->dst) < arg->mtu &&
2216 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2217 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2222 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2224 struct rt6_mtu_change_arg arg = {
2229 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2232 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2233 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2234 [RTA_OIF] = { .type = NLA_U32 },
2235 [RTA_IIF] = { .type = NLA_U32 },
2236 [RTA_PRIORITY] = { .type = NLA_U32 },
2237 [RTA_METRICS] = { .type = NLA_NESTED },
2238 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2241 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2242 struct fib6_config *cfg)
2245 struct nlattr *tb[RTA_MAX+1];
2248 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2253 rtm = nlmsg_data(nlh);
2254 memset(cfg, 0, sizeof(*cfg));
2256 cfg->fc_table = rtm->rtm_table;
2257 cfg->fc_dst_len = rtm->rtm_dst_len;
2258 cfg->fc_src_len = rtm->rtm_src_len;
2259 cfg->fc_flags = RTF_UP;
2260 cfg->fc_protocol = rtm->rtm_protocol;
2261 cfg->fc_type = rtm->rtm_type;
2263 if (rtm->rtm_type == RTN_UNREACHABLE ||
2264 rtm->rtm_type == RTN_BLACKHOLE ||
2265 rtm->rtm_type == RTN_PROHIBIT ||
2266 rtm->rtm_type == RTN_THROW)
2267 cfg->fc_flags |= RTF_REJECT;
2269 if (rtm->rtm_type == RTN_LOCAL)
2270 cfg->fc_flags |= RTF_LOCAL;
2272 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2273 cfg->fc_nlinfo.nlh = nlh;
2274 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2276 if (tb[RTA_GATEWAY]) {
2277 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2278 cfg->fc_flags |= RTF_GATEWAY;
2282 int plen = (rtm->rtm_dst_len + 7) >> 3;
2284 if (nla_len(tb[RTA_DST]) < plen)
2287 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2291 int plen = (rtm->rtm_src_len + 7) >> 3;
2293 if (nla_len(tb[RTA_SRC]) < plen)
2296 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2299 if (tb[RTA_PREFSRC])
2300 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2303 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2305 if (tb[RTA_PRIORITY])
2306 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2308 if (tb[RTA_METRICS]) {
2309 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2310 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2314 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2316 if (tb[RTA_MULTIPATH]) {
2317 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2318 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2326 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2328 struct fib6_config r_cfg;
2329 struct rtnexthop *rtnh;
2332 int err = 0, last_err = 0;
2335 rtnh = (struct rtnexthop *)cfg->fc_mp;
2336 remaining = cfg->fc_mp_len;
2338 /* Parse a Multipath Entry */
2339 while (rtnh_ok(rtnh, remaining)) {
2340 memcpy(&r_cfg, cfg, sizeof(*cfg));
2341 if (rtnh->rtnh_ifindex)
2342 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2344 attrlen = rtnh_attrlen(rtnh);
2346 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2348 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2350 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2351 r_cfg.fc_flags |= RTF_GATEWAY;
2354 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2357 /* If we are trying to remove a route, do not stop the
2358 * loop when ip6_route_del() fails (because next hop is
2359 * already gone), we should try to remove all next hops.
2362 /* If add fails, we should try to delete all
2363 * next hops that have been already added.
2369 /* Because each route is added like a single route we remove
2370 * this flag after the first nexthop (if there is a collision,
2371 * we have already fail to add the first nexthop:
2372 * fib6_add_rt2node() has reject it).
2374 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2375 rtnh = rtnh_next(rtnh, &remaining);
2381 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2383 struct fib6_config cfg;
2386 err = rtm_to_fib6_config(skb, nlh, &cfg);
2391 return ip6_route_multipath(&cfg, 0);
2393 return ip6_route_del(&cfg);
2396 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2398 struct fib6_config cfg;
2401 err = rtm_to_fib6_config(skb, nlh, &cfg);
2406 return ip6_route_multipath(&cfg, 1);
2408 return ip6_route_add(&cfg);
2411 static inline size_t rt6_nlmsg_size(void)
2413 return NLMSG_ALIGN(sizeof(struct rtmsg))
2414 + nla_total_size(16) /* RTA_SRC */
2415 + nla_total_size(16) /* RTA_DST */
2416 + nla_total_size(16) /* RTA_GATEWAY */
2417 + nla_total_size(16) /* RTA_PREFSRC */
2418 + nla_total_size(4) /* RTA_TABLE */
2419 + nla_total_size(4) /* RTA_IIF */
2420 + nla_total_size(4) /* RTA_OIF */
2421 + nla_total_size(4) /* RTA_PRIORITY */
2422 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2423 + nla_total_size(sizeof(struct rta_cacheinfo));
2426 static int rt6_fill_node(struct net *net,
2427 struct sk_buff *skb, struct rt6_info *rt,
2428 struct in6_addr *dst, struct in6_addr *src,
2429 int iif, int type, u32 portid, u32 seq,
2430 int prefix, int nowait, unsigned int flags)
2433 struct nlmsghdr *nlh;
2437 if (prefix) { /* user wants prefix routes only */
2438 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2439 /* success since this is not a prefix route */
2444 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2448 rtm = nlmsg_data(nlh);
2449 rtm->rtm_family = AF_INET6;
2450 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2451 rtm->rtm_src_len = rt->rt6i_src.plen;
2454 table = rt->rt6i_table->tb6_id;
2456 table = RT6_TABLE_UNSPEC;
2457 rtm->rtm_table = table;
2458 if (nla_put_u32(skb, RTA_TABLE, table))
2459 goto nla_put_failure;
2460 if (rt->rt6i_flags & RTF_REJECT) {
2461 switch (rt->dst.error) {
2463 rtm->rtm_type = RTN_BLACKHOLE;
2466 rtm->rtm_type = RTN_PROHIBIT;
2469 rtm->rtm_type = RTN_THROW;
2472 rtm->rtm_type = RTN_UNREACHABLE;
2476 else if (rt->rt6i_flags & RTF_LOCAL)
2477 rtm->rtm_type = RTN_LOCAL;
2478 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2479 rtm->rtm_type = RTN_LOCAL;
2481 rtm->rtm_type = RTN_UNICAST;
2483 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2484 rtm->rtm_protocol = rt->rt6i_protocol;
2485 if (rt->rt6i_flags & RTF_DYNAMIC)
2486 rtm->rtm_protocol = RTPROT_REDIRECT;
2487 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2488 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2489 rtm->rtm_protocol = RTPROT_RA;
2491 rtm->rtm_protocol = RTPROT_KERNEL;
2494 if (rt->rt6i_flags & RTF_CACHE)
2495 rtm->rtm_flags |= RTM_F_CLONED;
2498 if (nla_put(skb, RTA_DST, 16, dst))
2499 goto nla_put_failure;
2500 rtm->rtm_dst_len = 128;
2501 } else if (rtm->rtm_dst_len)
2502 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2503 goto nla_put_failure;
2504 #ifdef CONFIG_IPV6_SUBTREES
2506 if (nla_put(skb, RTA_SRC, 16, src))
2507 goto nla_put_failure;
2508 rtm->rtm_src_len = 128;
2509 } else if (rtm->rtm_src_len &&
2510 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2511 goto nla_put_failure;
2514 #ifdef CONFIG_IPV6_MROUTE
2515 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2516 int err = ip6mr_get_route(net, skb, rtm, nowait);
2521 goto nla_put_failure;
2523 if (err == -EMSGSIZE)
2524 goto nla_put_failure;
2529 if (nla_put_u32(skb, RTA_IIF, iif))
2530 goto nla_put_failure;
2532 struct in6_addr saddr_buf;
2533 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2534 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2535 goto nla_put_failure;
2538 if (rt->rt6i_prefsrc.plen) {
2539 struct in6_addr saddr_buf;
2540 saddr_buf = rt->rt6i_prefsrc.addr;
2541 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2542 goto nla_put_failure;
2545 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2546 goto nla_put_failure;
2548 if (rt->rt6i_flags & RTF_GATEWAY) {
2549 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2550 goto nla_put_failure;
2554 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2555 goto nla_put_failure;
2556 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2557 goto nla_put_failure;
2559 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2561 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2562 goto nla_put_failure;
2564 return nlmsg_end(skb, nlh);
2567 nlmsg_cancel(skb, nlh);
2571 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2573 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2576 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2577 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2578 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2582 return rt6_fill_node(arg->net,
2583 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2584 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2585 prefix, 0, NLM_F_MULTI);
2588 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2590 struct net *net = sock_net(in_skb->sk);
2591 struct nlattr *tb[RTA_MAX+1];
2592 struct rt6_info *rt;
2593 struct sk_buff *skb;
2596 int err, iif = 0, oif = 0;
2598 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2603 memset(&fl6, 0, sizeof(fl6));
2606 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2609 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2613 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2616 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2620 iif = nla_get_u32(tb[RTA_IIF]);
2623 oif = nla_get_u32(tb[RTA_OIF]);
2626 struct net_device *dev;
2629 dev = __dev_get_by_index(net, iif);
2635 fl6.flowi6_iif = iif;
2637 if (!ipv6_addr_any(&fl6.saddr))
2638 flags |= RT6_LOOKUP_F_HAS_SADDR;
2640 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2643 fl6.flowi6_oif = oif;
2645 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2648 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2655 /* Reserve room for dummy headers, this skb can pass
2656 through good chunk of routing engine.
2658 skb_reset_mac_header(skb);
2659 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2661 skb_dst_set(skb, &rt->dst);
2663 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2664 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2665 nlh->nlmsg_seq, 0, 0, 0);
2671 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2676 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2678 struct sk_buff *skb;
2679 struct net *net = info->nl_net;
2684 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2686 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2690 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2691 event, info->portid, seq, 0, 0, 0);
2693 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2694 WARN_ON(err == -EMSGSIZE);
2698 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2699 info->nlh, gfp_any());
2703 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2706 static int ip6_route_dev_notify(struct notifier_block *this,
2707 unsigned long event, void *ptr)
2709 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2710 struct net *net = dev_net(dev);
2712 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2713 net->ipv6.ip6_null_entry->dst.dev = dev;
2714 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2715 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2716 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2717 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2718 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2719 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2730 #ifdef CONFIG_PROC_FS
2741 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2743 struct seq_file *m = p_arg;
2745 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2747 #ifdef CONFIG_IPV6_SUBTREES
2748 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2750 seq_puts(m, "00000000000000000000000000000000 00 ");
2752 if (rt->rt6i_flags & RTF_GATEWAY) {
2753 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2755 seq_puts(m, "00000000000000000000000000000000");
2757 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2758 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2759 rt->dst.__use, rt->rt6i_flags,
2760 rt->dst.dev ? rt->dst.dev->name : "");
2764 static int ipv6_route_show(struct seq_file *m, void *v)
2766 struct net *net = (struct net *)m->private;
2767 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2771 static int ipv6_route_open(struct inode *inode, struct file *file)
2773 return single_open_net(inode, file, ipv6_route_show);
2776 static const struct file_operations ipv6_route_proc_fops = {
2777 .owner = THIS_MODULE,
2778 .open = ipv6_route_open,
2780 .llseek = seq_lseek,
2781 .release = single_release_net,
2784 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2786 struct net *net = (struct net *)seq->private;
2787 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2788 net->ipv6.rt6_stats->fib_nodes,
2789 net->ipv6.rt6_stats->fib_route_nodes,
2790 net->ipv6.rt6_stats->fib_rt_alloc,
2791 net->ipv6.rt6_stats->fib_rt_entries,
2792 net->ipv6.rt6_stats->fib_rt_cache,
2793 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2794 net->ipv6.rt6_stats->fib_discarded_routes);
2799 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2801 return single_open_net(inode, file, rt6_stats_seq_show);
2804 static const struct file_operations rt6_stats_seq_fops = {
2805 .owner = THIS_MODULE,
2806 .open = rt6_stats_seq_open,
2808 .llseek = seq_lseek,
2809 .release = single_release_net,
2811 #endif /* CONFIG_PROC_FS */
2813 #ifdef CONFIG_SYSCTL
2816 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2817 void __user *buffer, size_t *lenp, loff_t *ppos)
2824 net = (struct net *)ctl->extra1;
2825 delay = net->ipv6.sysctl.flush_delay;
2826 proc_dointvec(ctl, write, buffer, lenp, ppos);
2827 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2831 struct ctl_table ipv6_route_table_template[] = {
2833 .procname = "flush",
2834 .data = &init_net.ipv6.sysctl.flush_delay,
2835 .maxlen = sizeof(int),
2837 .proc_handler = ipv6_sysctl_rtcache_flush
2840 .procname = "gc_thresh",
2841 .data = &ip6_dst_ops_template.gc_thresh,
2842 .maxlen = sizeof(int),
2844 .proc_handler = proc_dointvec,
2847 .procname = "max_size",
2848 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2849 .maxlen = sizeof(int),
2851 .proc_handler = proc_dointvec,
2854 .procname = "gc_min_interval",
2855 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2856 .maxlen = sizeof(int),
2858 .proc_handler = proc_dointvec_jiffies,
2861 .procname = "gc_timeout",
2862 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2863 .maxlen = sizeof(int),
2865 .proc_handler = proc_dointvec_jiffies,
2868 .procname = "gc_interval",
2869 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2870 .maxlen = sizeof(int),
2872 .proc_handler = proc_dointvec_jiffies,
2875 .procname = "gc_elasticity",
2876 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2877 .maxlen = sizeof(int),
2879 .proc_handler = proc_dointvec,
2882 .procname = "mtu_expires",
2883 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2884 .maxlen = sizeof(int),
2886 .proc_handler = proc_dointvec_jiffies,
2889 .procname = "min_adv_mss",
2890 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2891 .maxlen = sizeof(int),
2893 .proc_handler = proc_dointvec,
2896 .procname = "gc_min_interval_ms",
2897 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2898 .maxlen = sizeof(int),
2900 .proc_handler = proc_dointvec_ms_jiffies,
2905 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2907 struct ctl_table *table;
2909 table = kmemdup(ipv6_route_table_template,
2910 sizeof(ipv6_route_table_template),
2914 table[0].data = &net->ipv6.sysctl.flush_delay;
2915 table[0].extra1 = net;
2916 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2917 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2918 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2919 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2920 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2921 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2922 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2923 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2924 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2926 /* Don't export sysctls to unprivileged users */
2927 if (net->user_ns != &init_user_ns)
2928 table[0].procname = NULL;
2935 static int __net_init ip6_route_net_init(struct net *net)
2939 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2940 sizeof(net->ipv6.ip6_dst_ops));
2942 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2943 goto out_ip6_dst_ops;
2945 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2946 sizeof(*net->ipv6.ip6_null_entry),
2948 if (!net->ipv6.ip6_null_entry)
2949 goto out_ip6_dst_entries;
2950 net->ipv6.ip6_null_entry->dst.path =
2951 (struct dst_entry *)net->ipv6.ip6_null_entry;
2952 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2953 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2954 ip6_template_metrics, true);
2956 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2957 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2958 sizeof(*net->ipv6.ip6_prohibit_entry),
2960 if (!net->ipv6.ip6_prohibit_entry)
2961 goto out_ip6_null_entry;
2962 net->ipv6.ip6_prohibit_entry->dst.path =
2963 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2964 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2965 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2966 ip6_template_metrics, true);
2968 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2969 sizeof(*net->ipv6.ip6_blk_hole_entry),
2971 if (!net->ipv6.ip6_blk_hole_entry)
2972 goto out_ip6_prohibit_entry;
2973 net->ipv6.ip6_blk_hole_entry->dst.path =
2974 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2975 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2976 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2977 ip6_template_metrics, true);
2980 net->ipv6.sysctl.flush_delay = 0;
2981 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2982 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2983 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2984 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2985 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2986 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2987 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2989 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2995 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2996 out_ip6_prohibit_entry:
2997 kfree(net->ipv6.ip6_prohibit_entry);
2999 kfree(net->ipv6.ip6_null_entry);
3001 out_ip6_dst_entries:
3002 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3007 static void __net_exit ip6_route_net_exit(struct net *net)
3009 kfree(net->ipv6.ip6_null_entry);
3010 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3011 kfree(net->ipv6.ip6_prohibit_entry);
3012 kfree(net->ipv6.ip6_blk_hole_entry);
3014 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3017 static int __net_init ip6_route_net_init_late(struct net *net)
3019 #ifdef CONFIG_PROC_FS
3020 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3021 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3026 static void __net_exit ip6_route_net_exit_late(struct net *net)
3028 #ifdef CONFIG_PROC_FS
3029 remove_proc_entry("ipv6_route", net->proc_net);
3030 remove_proc_entry("rt6_stats", net->proc_net);
3034 static struct pernet_operations ip6_route_net_ops = {
3035 .init = ip6_route_net_init,
3036 .exit = ip6_route_net_exit,
3039 static int __net_init ipv6_inetpeer_init(struct net *net)
3041 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3045 inet_peer_base_init(bp);
3046 net->ipv6.peers = bp;
3050 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3052 struct inet_peer_base *bp = net->ipv6.peers;
3054 net->ipv6.peers = NULL;
3055 inetpeer_invalidate_tree(bp);
3059 static struct pernet_operations ipv6_inetpeer_ops = {
3060 .init = ipv6_inetpeer_init,
3061 .exit = ipv6_inetpeer_exit,
3064 static struct pernet_operations ip6_route_net_late_ops = {
3065 .init = ip6_route_net_init_late,
3066 .exit = ip6_route_net_exit_late,
3069 static struct notifier_block ip6_route_dev_notifier = {
3070 .notifier_call = ip6_route_dev_notify,
3074 int __init ip6_route_init(void)
3079 ip6_dst_ops_template.kmem_cachep =
3080 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3081 SLAB_HWCACHE_ALIGN, NULL);
3082 if (!ip6_dst_ops_template.kmem_cachep)
3085 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3087 goto out_kmem_cache;
3089 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3091 goto out_dst_entries;
3093 ret = register_pernet_subsys(&ip6_route_net_ops);
3095 goto out_register_inetpeer;
3097 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3099 /* Registering of the loopback is done before this portion of code,
3100 * the loopback reference in rt6_info will not be taken, do it
3101 * manually for init_net */
3102 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3103 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3104 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3105 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3106 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3107 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3108 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3112 goto out_register_subsys;
3118 ret = fib6_rules_init();
3122 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3124 goto fib6_rules_init;
3127 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3128 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3129 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3130 goto out_register_late_subsys;
3132 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3134 goto out_register_late_subsys;
3139 out_register_late_subsys:
3140 unregister_pernet_subsys(&ip6_route_net_late_ops);
3142 fib6_rules_cleanup();
3147 out_register_subsys:
3148 unregister_pernet_subsys(&ip6_route_net_ops);
3149 out_register_inetpeer:
3150 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3152 dst_entries_destroy(&ip6_dst_blackhole_ops);
3154 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3158 void ip6_route_cleanup(void)
3160 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3161 unregister_pernet_subsys(&ip6_route_net_late_ops);
3162 fib6_rules_cleanup();
3165 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3166 unregister_pernet_subsys(&ip6_route_net_ops);
3167 dst_entries_destroy(&ip6_dst_blackhole_ops);
3168 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);