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(net);
287 INIT_LIST_HEAD(&rt->rt6i_siblings);
288 rt->rt6i_nsiblings = 0;
293 static void ip6_dst_destroy(struct dst_entry *dst)
295 struct rt6_info *rt = (struct rt6_info *)dst;
296 struct inet6_dev *idev = rt->rt6i_idev;
297 struct dst_entry *from = dst->from;
299 if (!(rt->dst.flags & DST_HOST))
300 dst_destroy_metrics_generic(dst);
303 rt->rt6i_idev = NULL;
310 if (rt6_has_peer(rt)) {
311 struct inet_peer *peer = rt6_peer_ptr(rt);
316 void rt6_bind_peer(struct rt6_info *rt, int create)
318 struct inet_peer_base *base;
319 struct inet_peer *peer;
321 base = inetpeer_base_ptr(rt->_rt6i_peer);
325 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
327 if (!rt6_set_peer(rt, peer))
332 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
335 struct rt6_info *rt = (struct rt6_info *)dst;
336 struct inet6_dev *idev = rt->rt6i_idev;
337 struct net_device *loopback_dev =
338 dev_net(dev)->loopback_dev;
340 if (dev != loopback_dev) {
341 if (idev && idev->dev == dev) {
342 struct inet6_dev *loopback_idev =
343 in6_dev_get(loopback_dev);
345 rt->rt6i_idev = loopback_idev;
352 static bool rt6_check_expired(const struct rt6_info *rt)
354 if (rt->rt6i_flags & RTF_EXPIRES) {
355 if (time_after(jiffies, rt->dst.expires))
357 } else if (rt->dst.from) {
358 return rt6_check_expired((struct rt6_info *) rt->dst.from);
363 static bool rt6_need_strict(const struct in6_addr *daddr)
365 return ipv6_addr_type(daddr) &
366 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
369 /* Multipath route selection:
370 * Hash based function using packet header and flowlabel.
371 * Adapted from fib_info_hashfn()
373 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
374 const struct flowi6 *fl6)
376 unsigned int val = fl6->flowi6_proto;
378 val ^= ipv6_addr_hash(&fl6->daddr);
379 val ^= ipv6_addr_hash(&fl6->saddr);
381 /* Work only if this not encapsulated */
382 switch (fl6->flowi6_proto) {
386 val ^= (__force u16)fl6->fl6_sport;
387 val ^= (__force u16)fl6->fl6_dport;
391 val ^= (__force u16)fl6->fl6_icmp_type;
392 val ^= (__force u16)fl6->fl6_icmp_code;
395 /* RFC6438 recommands to use flowlabel */
396 val ^= (__force u32)fl6->flowlabel;
398 /* Perhaps, we need to tune, this function? */
399 val = val ^ (val >> 7) ^ (val >> 12);
400 return val % candidate_count;
403 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
404 struct flowi6 *fl6, int oif,
407 struct rt6_info *sibling, *next_sibling;
410 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
411 /* Don't change the route, if route_choosen == 0
412 * (siblings does not include ourself)
415 list_for_each_entry_safe(sibling, next_sibling,
416 &match->rt6i_siblings, rt6i_siblings) {
418 if (route_choosen == 0) {
419 if (rt6_score_route(sibling, oif, strict) < 0)
429 * Route lookup. Any table->tb6_lock is implied.
432 static inline struct rt6_info *rt6_device_match(struct net *net,
434 const struct in6_addr *saddr,
438 struct rt6_info *local = NULL;
439 struct rt6_info *sprt;
441 if (!oif && ipv6_addr_any(saddr))
444 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
445 struct net_device *dev = sprt->dst.dev;
448 if (dev->ifindex == oif)
450 if (dev->flags & IFF_LOOPBACK) {
451 if (!sprt->rt6i_idev ||
452 sprt->rt6i_idev->dev->ifindex != oif) {
453 if (flags & RT6_LOOKUP_F_IFACE && oif)
455 if (local && (!oif ||
456 local->rt6i_idev->dev->ifindex == oif))
462 if (ipv6_chk_addr(net, saddr, dev,
463 flags & RT6_LOOKUP_F_IFACE))
472 if (flags & RT6_LOOKUP_F_IFACE)
473 return net->ipv6.ip6_null_entry;
479 #ifdef CONFIG_IPV6_ROUTER_PREF
480 static void rt6_probe(struct rt6_info *rt)
482 struct neighbour *neigh;
484 * Okay, this does not seem to be appropriate
485 * for now, however, we need to check if it
486 * is really so; aka Router Reachability Probing.
488 * Router Reachability Probe MUST be rate-limited
489 * to no more than one per minute.
491 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
494 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
496 write_lock(&neigh->lock);
497 if (neigh->nud_state & NUD_VALID)
502 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
503 struct in6_addr mcaddr;
504 struct in6_addr *target;
507 neigh->updated = jiffies;
508 write_unlock(&neigh->lock);
511 target = (struct in6_addr *)&rt->rt6i_gateway;
512 addrconf_addr_solict_mult(target, &mcaddr);
513 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
516 write_unlock(&neigh->lock);
518 rcu_read_unlock_bh();
521 static inline void rt6_probe(struct rt6_info *rt)
527 * Default Router Selection (RFC 2461 6.3.6)
529 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
531 struct net_device *dev = rt->dst.dev;
532 if (!oif || dev->ifindex == oif)
534 if ((dev->flags & IFF_LOOPBACK) &&
535 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
540 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
542 struct neighbour *neigh;
543 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
545 if (rt->rt6i_flags & RTF_NONEXTHOP ||
546 !(rt->rt6i_flags & RTF_GATEWAY))
547 return RT6_NUD_SUCCEED;
550 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
552 read_lock(&neigh->lock);
553 if (neigh->nud_state & NUD_VALID)
554 ret = RT6_NUD_SUCCEED;
555 #ifdef CONFIG_IPV6_ROUTER_PREF
556 else if (!(neigh->nud_state & NUD_FAILED))
557 ret = RT6_NUD_SUCCEED;
559 read_unlock(&neigh->lock);
561 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
562 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
564 rcu_read_unlock_bh();
569 static int rt6_score_route(struct rt6_info *rt, int oif,
574 m = rt6_check_dev(rt, oif);
575 if (!m && (strict & RT6_LOOKUP_F_IFACE))
576 return RT6_NUD_FAIL_HARD;
577 #ifdef CONFIG_IPV6_ROUTER_PREF
578 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
580 if (strict & RT6_LOOKUP_F_REACHABLE) {
581 int n = rt6_check_neigh(rt);
588 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
589 int *mpri, struct rt6_info *match,
593 bool match_do_rr = false;
595 if (rt6_check_expired(rt))
598 m = rt6_score_route(rt, oif, strict);
599 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
601 m = 0; /* lowest valid score */
606 if (strict & RT6_LOOKUP_F_REACHABLE)
610 *do_rr = match_do_rr;
618 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
619 struct rt6_info *rr_head,
620 u32 metric, int oif, int strict,
623 struct rt6_info *rt, *match;
627 for (rt = rr_head; rt && rt->rt6i_metric == metric;
628 rt = rt->dst.rt6_next)
629 match = find_match(rt, oif, strict, &mpri, match, do_rr);
630 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
631 rt = rt->dst.rt6_next)
632 match = find_match(rt, oif, strict, &mpri, match, do_rr);
637 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
639 struct rt6_info *match, *rt0;
645 fn->rr_ptr = rt0 = fn->leaf;
647 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
651 struct rt6_info *next = rt0->dst.rt6_next;
653 /* no entries matched; do round-robin */
654 if (!next || next->rt6i_metric != rt0->rt6i_metric)
661 net = dev_net(rt0->dst.dev);
662 return match ? match : net->ipv6.ip6_null_entry;
665 #ifdef CONFIG_IPV6_ROUTE_INFO
666 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
667 const struct in6_addr *gwaddr)
669 struct net *net = dev_net(dev);
670 struct route_info *rinfo = (struct route_info *) opt;
671 struct in6_addr prefix_buf, *prefix;
673 unsigned long lifetime;
676 if (len < sizeof(struct route_info)) {
680 /* Sanity check for prefix_len and length */
681 if (rinfo->length > 3) {
683 } else if (rinfo->prefix_len > 128) {
685 } else if (rinfo->prefix_len > 64) {
686 if (rinfo->length < 2) {
689 } else if (rinfo->prefix_len > 0) {
690 if (rinfo->length < 1) {
695 pref = rinfo->route_pref;
696 if (pref == ICMPV6_ROUTER_PREF_INVALID)
699 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
701 if (rinfo->length == 3)
702 prefix = (struct in6_addr *)rinfo->prefix;
704 /* this function is safe */
705 ipv6_addr_prefix(&prefix_buf,
706 (struct in6_addr *)rinfo->prefix,
708 prefix = &prefix_buf;
711 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
714 if (rt && !lifetime) {
720 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
723 rt->rt6i_flags = RTF_ROUTEINFO |
724 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
727 if (!addrconf_finite_timeout(lifetime))
728 rt6_clean_expires(rt);
730 rt6_set_expires(rt, jiffies + HZ * lifetime);
738 #define BACKTRACK(__net, saddr) \
740 if (rt == __net->ipv6.ip6_null_entry) { \
741 struct fib6_node *pn; \
743 if (fn->fn_flags & RTN_TL_ROOT) \
746 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
747 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
750 if (fn->fn_flags & RTN_RTINFO) \
756 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
757 struct fib6_table *table,
758 struct flowi6 *fl6, int flags)
760 struct fib6_node *fn;
763 read_lock_bh(&table->tb6_lock);
764 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
767 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
768 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
769 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
770 BACKTRACK(net, &fl6->saddr);
772 dst_use(&rt->dst, jiffies);
773 read_unlock_bh(&table->tb6_lock);
778 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
781 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
783 EXPORT_SYMBOL_GPL(ip6_route_lookup);
785 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
786 const struct in6_addr *saddr, int oif, int strict)
788 struct flowi6 fl6 = {
792 struct dst_entry *dst;
793 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
796 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
797 flags |= RT6_LOOKUP_F_HAS_SADDR;
800 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
802 return (struct rt6_info *) dst;
809 EXPORT_SYMBOL(rt6_lookup);
811 /* ip6_ins_rt is called with FREE table->tb6_lock.
812 It takes new route entry, the addition fails by any reason the
813 route is freed. In any case, if caller does not hold it, it may
817 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
820 struct fib6_table *table;
822 table = rt->rt6i_table;
823 write_lock_bh(&table->tb6_lock);
824 err = fib6_add(&table->tb6_root, rt, info);
825 write_unlock_bh(&table->tb6_lock);
830 int ip6_ins_rt(struct rt6_info *rt)
832 struct nl_info info = {
833 .nl_net = dev_net(rt->dst.dev),
835 return __ip6_ins_rt(rt, &info);
838 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
839 const struct in6_addr *daddr,
840 const struct in6_addr *saddr)
848 rt = ip6_rt_copy(ort, daddr);
851 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
852 if (ort->rt6i_dst.plen != 128 &&
853 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
854 rt->rt6i_flags |= RTF_ANYCAST;
855 rt->rt6i_gateway = *daddr;
858 rt->rt6i_flags |= RTF_CACHE;
860 #ifdef CONFIG_IPV6_SUBTREES
861 if (rt->rt6i_src.plen && saddr) {
862 rt->rt6i_src.addr = *saddr;
863 rt->rt6i_src.plen = 128;
871 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
872 const struct in6_addr *daddr)
874 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
877 rt->rt6i_flags |= RTF_CACHE;
881 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
882 struct flowi6 *fl6, int flags)
884 struct fib6_node *fn;
885 struct rt6_info *rt, *nrt;
889 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
891 strict |= flags & RT6_LOOKUP_F_IFACE;
894 read_lock_bh(&table->tb6_lock);
897 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
900 rt = rt6_select(fn, oif, strict | reachable);
901 if (rt->rt6i_nsiblings)
902 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
903 BACKTRACK(net, &fl6->saddr);
904 if (rt == net->ipv6.ip6_null_entry ||
905 rt->rt6i_flags & RTF_CACHE)
909 read_unlock_bh(&table->tb6_lock);
911 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
912 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
913 else if (!(rt->dst.flags & DST_HOST))
914 nrt = rt6_alloc_clone(rt, &fl6->daddr);
919 rt = nrt ? : net->ipv6.ip6_null_entry;
923 err = ip6_ins_rt(nrt);
932 * Race condition! In the gap, when table->tb6_lock was
933 * released someone could insert this route. Relookup.
944 read_unlock_bh(&table->tb6_lock);
946 rt->dst.lastuse = jiffies;
952 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
953 struct flowi6 *fl6, int flags)
955 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
958 static struct dst_entry *ip6_route_input_lookup(struct net *net,
959 struct net_device *dev,
960 struct flowi6 *fl6, int flags)
962 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
963 flags |= RT6_LOOKUP_F_IFACE;
965 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
968 void ip6_route_input(struct sk_buff *skb)
970 const struct ipv6hdr *iph = ipv6_hdr(skb);
971 struct net *net = dev_net(skb->dev);
972 int flags = RT6_LOOKUP_F_HAS_SADDR;
973 struct flowi6 fl6 = {
974 .flowi6_iif = skb->dev->ifindex,
977 .flowlabel = ip6_flowinfo(iph),
978 .flowi6_mark = skb->mark,
979 .flowi6_proto = iph->nexthdr,
982 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
985 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
986 struct flowi6 *fl6, int flags)
988 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
991 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
996 fl6->flowi6_iif = LOOPBACK_IFINDEX;
998 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
999 flags |= RT6_LOOKUP_F_IFACE;
1001 if (!ipv6_addr_any(&fl6->saddr))
1002 flags |= RT6_LOOKUP_F_HAS_SADDR;
1004 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1006 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1009 EXPORT_SYMBOL(ip6_route_output);
1011 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1013 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1014 struct dst_entry *new = NULL;
1016 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1020 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1021 rt6_init_peer(rt, net->ipv6.peers);
1024 new->input = dst_discard;
1025 new->output = dst_discard;
1027 if (dst_metrics_read_only(&ort->dst))
1028 new->_metrics = ort->dst._metrics;
1030 dst_copy_metrics(new, &ort->dst);
1031 rt->rt6i_idev = ort->rt6i_idev;
1033 in6_dev_hold(rt->rt6i_idev);
1035 rt->rt6i_gateway = ort->rt6i_gateway;
1036 rt->rt6i_flags = ort->rt6i_flags;
1037 rt->rt6i_metric = 0;
1039 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1040 #ifdef CONFIG_IPV6_SUBTREES
1041 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1047 dst_release(dst_orig);
1048 return new ? new : ERR_PTR(-ENOMEM);
1052 * Destination cache support functions
1055 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1057 struct rt6_info *rt;
1059 rt = (struct rt6_info *) dst;
1061 /* All IPV6 dsts are created with ->obsolete set to the value
1062 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1063 * into this function always.
1065 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1068 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1074 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1076 struct rt6_info *rt = (struct rt6_info *) dst;
1079 if (rt->rt6i_flags & RTF_CACHE) {
1080 if (rt6_check_expired(rt)) {
1092 static void ip6_link_failure(struct sk_buff *skb)
1094 struct rt6_info *rt;
1096 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1098 rt = (struct rt6_info *) skb_dst(skb);
1100 if (rt->rt6i_flags & RTF_CACHE) {
1104 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1105 rt->rt6i_node->fn_sernum = -1;
1110 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1111 struct sk_buff *skb, u32 mtu)
1113 struct rt6_info *rt6 = (struct rt6_info*)dst;
1116 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1117 struct net *net = dev_net(dst->dev);
1119 rt6->rt6i_flags |= RTF_MODIFIED;
1120 if (mtu < IPV6_MIN_MTU) {
1121 u32 features = dst_metric(dst, RTAX_FEATURES);
1123 features |= RTAX_FEATURE_ALLFRAG;
1124 dst_metric_set(dst, RTAX_FEATURES, features);
1126 dst_metric_set(dst, RTAX_MTU, mtu);
1127 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1131 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1134 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1135 struct dst_entry *dst;
1138 memset(&fl6, 0, sizeof(fl6));
1139 fl6.flowi6_oif = oif;
1140 fl6.flowi6_mark = mark;
1141 fl6.flowi6_flags = 0;
1142 fl6.daddr = iph->daddr;
1143 fl6.saddr = iph->saddr;
1144 fl6.flowlabel = ip6_flowinfo(iph);
1146 dst = ip6_route_output(net, NULL, &fl6);
1148 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1151 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1153 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1155 ip6_update_pmtu(skb, sock_net(sk), mtu,
1156 sk->sk_bound_dev_if, sk->sk_mark);
1158 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1160 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1162 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1163 struct dst_entry *dst;
1166 memset(&fl6, 0, sizeof(fl6));
1167 fl6.flowi6_oif = oif;
1168 fl6.flowi6_mark = mark;
1169 fl6.flowi6_flags = 0;
1170 fl6.daddr = iph->daddr;
1171 fl6.saddr = iph->saddr;
1172 fl6.flowlabel = ip6_flowinfo(iph);
1174 dst = ip6_route_output(net, NULL, &fl6);
1176 rt6_do_redirect(dst, NULL, skb);
1179 EXPORT_SYMBOL_GPL(ip6_redirect);
1181 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1184 const struct ipv6hdr *iph = ipv6_hdr(skb);
1185 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1186 struct dst_entry *dst;
1189 memset(&fl6, 0, sizeof(fl6));
1190 fl6.flowi6_oif = oif;
1191 fl6.flowi6_mark = mark;
1192 fl6.flowi6_flags = 0;
1193 fl6.daddr = msg->dest;
1194 fl6.saddr = iph->daddr;
1196 dst = ip6_route_output(net, NULL, &fl6);
1198 rt6_do_redirect(dst, NULL, skb);
1202 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1204 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1206 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1208 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1210 struct net_device *dev = dst->dev;
1211 unsigned int mtu = dst_mtu(dst);
1212 struct net *net = dev_net(dev);
1214 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1216 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1217 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1220 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1221 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1222 * IPV6_MAXPLEN is also valid and means: "any MSS,
1223 * rely only on pmtu discovery"
1225 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1230 static unsigned int ip6_mtu(const struct dst_entry *dst)
1232 struct inet6_dev *idev;
1233 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1241 idev = __in6_dev_get(dst->dev);
1243 mtu = idev->cnf.mtu6;
1249 static struct dst_entry *icmp6_dst_gc_list;
1250 static DEFINE_SPINLOCK(icmp6_dst_lock);
1252 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1255 struct dst_entry *dst;
1256 struct rt6_info *rt;
1257 struct inet6_dev *idev = in6_dev_get(dev);
1258 struct net *net = dev_net(dev);
1260 if (unlikely(!idev))
1261 return ERR_PTR(-ENODEV);
1263 rt = ip6_dst_alloc(net, dev, 0, NULL);
1264 if (unlikely(!rt)) {
1266 dst = ERR_PTR(-ENOMEM);
1270 rt->dst.flags |= DST_HOST;
1271 rt->dst.output = ip6_output;
1272 atomic_set(&rt->dst.__refcnt, 1);
1273 rt->rt6i_dst.addr = fl6->daddr;
1274 rt->rt6i_dst.plen = 128;
1275 rt->rt6i_idev = idev;
1276 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1278 spin_lock_bh(&icmp6_dst_lock);
1279 rt->dst.next = icmp6_dst_gc_list;
1280 icmp6_dst_gc_list = &rt->dst;
1281 spin_unlock_bh(&icmp6_dst_lock);
1283 fib6_force_start_gc(net);
1285 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1291 int icmp6_dst_gc(void)
1293 struct dst_entry *dst, **pprev;
1296 spin_lock_bh(&icmp6_dst_lock);
1297 pprev = &icmp6_dst_gc_list;
1299 while ((dst = *pprev) != NULL) {
1300 if (!atomic_read(&dst->__refcnt)) {
1309 spin_unlock_bh(&icmp6_dst_lock);
1314 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1317 struct dst_entry *dst, **pprev;
1319 spin_lock_bh(&icmp6_dst_lock);
1320 pprev = &icmp6_dst_gc_list;
1321 while ((dst = *pprev) != NULL) {
1322 struct rt6_info *rt = (struct rt6_info *) dst;
1323 if (func(rt, arg)) {
1330 spin_unlock_bh(&icmp6_dst_lock);
1333 static int ip6_dst_gc(struct dst_ops *ops)
1335 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1336 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1337 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1338 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1339 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1340 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1343 entries = dst_entries_get_fast(ops);
1344 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1345 entries <= rt_max_size)
1348 net->ipv6.ip6_rt_gc_expire++;
1349 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1350 entries = dst_entries_get_slow(ops);
1351 if (entries < ops->gc_thresh)
1352 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1354 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1355 return entries > rt_max_size;
1358 int ip6_dst_hoplimit(struct dst_entry *dst)
1360 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1361 if (hoplimit == 0) {
1362 struct net_device *dev = dst->dev;
1363 struct inet6_dev *idev;
1366 idev = __in6_dev_get(dev);
1368 hoplimit = idev->cnf.hop_limit;
1370 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1375 EXPORT_SYMBOL(ip6_dst_hoplimit);
1381 int ip6_route_add(struct fib6_config *cfg)
1384 struct net *net = cfg->fc_nlinfo.nl_net;
1385 struct rt6_info *rt = NULL;
1386 struct net_device *dev = NULL;
1387 struct inet6_dev *idev = NULL;
1388 struct fib6_table *table;
1391 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1393 #ifndef CONFIG_IPV6_SUBTREES
1394 if (cfg->fc_src_len)
1397 if (cfg->fc_ifindex) {
1399 dev = dev_get_by_index(net, cfg->fc_ifindex);
1402 idev = in6_dev_get(dev);
1407 if (cfg->fc_metric == 0)
1408 cfg->fc_metric = IP6_RT_PRIO_USER;
1411 if (cfg->fc_nlinfo.nlh &&
1412 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1413 table = fib6_get_table(net, cfg->fc_table);
1415 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1416 table = fib6_new_table(net, cfg->fc_table);
1419 table = fib6_new_table(net, cfg->fc_table);
1425 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1432 if (cfg->fc_flags & RTF_EXPIRES)
1433 rt6_set_expires(rt, jiffies +
1434 clock_t_to_jiffies(cfg->fc_expires));
1436 rt6_clean_expires(rt);
1438 if (cfg->fc_protocol == RTPROT_UNSPEC)
1439 cfg->fc_protocol = RTPROT_BOOT;
1440 rt->rt6i_protocol = cfg->fc_protocol;
1442 addr_type = ipv6_addr_type(&cfg->fc_dst);
1444 if (addr_type & IPV6_ADDR_MULTICAST)
1445 rt->dst.input = ip6_mc_input;
1446 else if (cfg->fc_flags & RTF_LOCAL)
1447 rt->dst.input = ip6_input;
1449 rt->dst.input = ip6_forward;
1451 rt->dst.output = ip6_output;
1453 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1454 rt->rt6i_dst.plen = cfg->fc_dst_len;
1455 if (rt->rt6i_dst.plen == 128)
1456 rt->dst.flags |= DST_HOST;
1458 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1459 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1464 dst_init_metrics(&rt->dst, metrics, 0);
1466 #ifdef CONFIG_IPV6_SUBTREES
1467 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1468 rt->rt6i_src.plen = cfg->fc_src_len;
1471 rt->rt6i_metric = cfg->fc_metric;
1473 /* We cannot add true routes via loopback here,
1474 they would result in kernel looping; promote them to reject routes
1476 if ((cfg->fc_flags & RTF_REJECT) ||
1477 (dev && (dev->flags & IFF_LOOPBACK) &&
1478 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1479 !(cfg->fc_flags & RTF_LOCAL))) {
1480 /* hold loopback dev/idev if we haven't done so. */
1481 if (dev != net->loopback_dev) {
1486 dev = net->loopback_dev;
1488 idev = in6_dev_get(dev);
1494 rt->dst.output = ip6_pkt_discard_out;
1495 rt->dst.input = ip6_pkt_discard;
1496 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1497 switch (cfg->fc_type) {
1499 rt->dst.error = -EINVAL;
1502 rt->dst.error = -EACCES;
1505 rt->dst.error = -EAGAIN;
1508 rt->dst.error = -ENETUNREACH;
1514 if (cfg->fc_flags & RTF_GATEWAY) {
1515 const struct in6_addr *gw_addr;
1518 gw_addr = &cfg->fc_gateway;
1519 rt->rt6i_gateway = *gw_addr;
1520 gwa_type = ipv6_addr_type(gw_addr);
1522 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1523 struct rt6_info *grt;
1525 /* IPv6 strictly inhibits using not link-local
1526 addresses as nexthop address.
1527 Otherwise, router will not able to send redirects.
1528 It is very good, but in some (rare!) circumstances
1529 (SIT, PtP, NBMA NOARP links) it is handy to allow
1530 some exceptions. --ANK
1533 if (!(gwa_type & IPV6_ADDR_UNICAST))
1536 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1538 err = -EHOSTUNREACH;
1542 if (dev != grt->dst.dev) {
1548 idev = grt->rt6i_idev;
1550 in6_dev_hold(grt->rt6i_idev);
1552 if (!(grt->rt6i_flags & RTF_GATEWAY))
1560 if (!dev || (dev->flags & IFF_LOOPBACK))
1568 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1569 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1573 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1574 rt->rt6i_prefsrc.plen = 128;
1576 rt->rt6i_prefsrc.plen = 0;
1578 rt->rt6i_flags = cfg->fc_flags;
1585 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1586 int type = nla_type(nla);
1589 if (type > RTAX_MAX) {
1594 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1600 rt->rt6i_idev = idev;
1601 rt->rt6i_table = table;
1603 cfg->fc_nlinfo.nl_net = dev_net(dev);
1605 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1617 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1620 struct fib6_table *table;
1621 struct net *net = dev_net(rt->dst.dev);
1623 if (rt == net->ipv6.ip6_null_entry) {
1628 table = rt->rt6i_table;
1629 write_lock_bh(&table->tb6_lock);
1630 err = fib6_del(rt, info);
1631 write_unlock_bh(&table->tb6_lock);
1638 int ip6_del_rt(struct rt6_info *rt)
1640 struct nl_info info = {
1641 .nl_net = dev_net(rt->dst.dev),
1643 return __ip6_del_rt(rt, &info);
1646 static int ip6_route_del(struct fib6_config *cfg)
1648 struct fib6_table *table;
1649 struct fib6_node *fn;
1650 struct rt6_info *rt;
1653 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1657 read_lock_bh(&table->tb6_lock);
1659 fn = fib6_locate(&table->tb6_root,
1660 &cfg->fc_dst, cfg->fc_dst_len,
1661 &cfg->fc_src, cfg->fc_src_len);
1664 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1665 if (cfg->fc_ifindex &&
1667 rt->dst.dev->ifindex != cfg->fc_ifindex))
1669 if (cfg->fc_flags & RTF_GATEWAY &&
1670 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1672 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1675 read_unlock_bh(&table->tb6_lock);
1677 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1680 read_unlock_bh(&table->tb6_lock);
1685 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1687 struct net *net = dev_net(skb->dev);
1688 struct netevent_redirect netevent;
1689 struct rt6_info *rt, *nrt = NULL;
1690 struct ndisc_options ndopts;
1691 struct inet6_dev *in6_dev;
1692 struct neighbour *neigh;
1694 int optlen, on_link;
1697 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1698 optlen -= sizeof(*msg);
1701 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1705 msg = (struct rd_msg *)icmp6_hdr(skb);
1707 if (ipv6_addr_is_multicast(&msg->dest)) {
1708 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1713 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1715 } else if (ipv6_addr_type(&msg->target) !=
1716 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1717 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1721 in6_dev = __in6_dev_get(skb->dev);
1724 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1728 * The IP source address of the Redirect MUST be the same as the current
1729 * first-hop router for the specified ICMP Destination Address.
1732 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1733 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1738 if (ndopts.nd_opts_tgt_lladdr) {
1739 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1742 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1747 rt = (struct rt6_info *) dst;
1748 if (rt == net->ipv6.ip6_null_entry) {
1749 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1753 /* Redirect received -> path was valid.
1754 * Look, redirects are sent only in response to data packets,
1755 * so that this nexthop apparently is reachable. --ANK
1757 dst_confirm(&rt->dst);
1759 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1764 * We have finally decided to accept it.
1767 neigh_update(neigh, lladdr, NUD_STALE,
1768 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1769 NEIGH_UPDATE_F_OVERRIDE|
1770 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1771 NEIGH_UPDATE_F_ISROUTER))
1774 nrt = ip6_rt_copy(rt, &msg->dest);
1778 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1780 nrt->rt6i_flags &= ~RTF_GATEWAY;
1782 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1784 if (ip6_ins_rt(nrt))
1787 netevent.old = &rt->dst;
1788 netevent.new = &nrt->dst;
1789 netevent.daddr = &msg->dest;
1790 netevent.neigh = neigh;
1791 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1793 if (rt->rt6i_flags & RTF_CACHE) {
1794 rt = (struct rt6_info *) dst_clone(&rt->dst);
1799 neigh_release(neigh);
1803 * Misc support functions
1806 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1807 const struct in6_addr *dest)
1809 struct net *net = dev_net(ort->dst.dev);
1810 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1814 rt->dst.input = ort->dst.input;
1815 rt->dst.output = ort->dst.output;
1816 rt->dst.flags |= DST_HOST;
1818 rt->rt6i_dst.addr = *dest;
1819 rt->rt6i_dst.plen = 128;
1820 dst_copy_metrics(&rt->dst, &ort->dst);
1821 rt->dst.error = ort->dst.error;
1822 rt->rt6i_idev = ort->rt6i_idev;
1824 in6_dev_hold(rt->rt6i_idev);
1825 rt->dst.lastuse = jiffies;
1827 rt->rt6i_gateway = ort->rt6i_gateway;
1828 rt->rt6i_flags = ort->rt6i_flags;
1829 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1830 (RTF_DEFAULT | RTF_ADDRCONF))
1831 rt6_set_from(rt, ort);
1832 rt->rt6i_metric = 0;
1834 #ifdef CONFIG_IPV6_SUBTREES
1835 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1837 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1838 rt->rt6i_table = ort->rt6i_table;
1843 #ifdef CONFIG_IPV6_ROUTE_INFO
1844 static struct rt6_info *rt6_get_route_info(struct net *net,
1845 const struct in6_addr *prefix, int prefixlen,
1846 const struct in6_addr *gwaddr, int ifindex)
1848 struct fib6_node *fn;
1849 struct rt6_info *rt = NULL;
1850 struct fib6_table *table;
1852 table = fib6_get_table(net, RT6_TABLE_INFO);
1856 read_lock_bh(&table->tb6_lock);
1857 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1861 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1862 if (rt->dst.dev->ifindex != ifindex)
1864 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1866 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1872 read_unlock_bh(&table->tb6_lock);
1876 static struct rt6_info *rt6_add_route_info(struct net *net,
1877 const struct in6_addr *prefix, int prefixlen,
1878 const struct in6_addr *gwaddr, int ifindex,
1881 struct fib6_config cfg = {
1882 .fc_table = RT6_TABLE_INFO,
1883 .fc_metric = IP6_RT_PRIO_USER,
1884 .fc_ifindex = ifindex,
1885 .fc_dst_len = prefixlen,
1886 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1887 RTF_UP | RTF_PREF(pref),
1888 .fc_nlinfo.portid = 0,
1889 .fc_nlinfo.nlh = NULL,
1890 .fc_nlinfo.nl_net = net,
1893 cfg.fc_dst = *prefix;
1894 cfg.fc_gateway = *gwaddr;
1896 /* We should treat it as a default route if prefix length is 0. */
1898 cfg.fc_flags |= RTF_DEFAULT;
1900 ip6_route_add(&cfg);
1902 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1906 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1908 struct rt6_info *rt;
1909 struct fib6_table *table;
1911 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1915 read_lock_bh(&table->tb6_lock);
1916 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1917 if (dev == rt->dst.dev &&
1918 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1919 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1924 read_unlock_bh(&table->tb6_lock);
1928 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1929 struct net_device *dev,
1932 struct fib6_config cfg = {
1933 .fc_table = RT6_TABLE_DFLT,
1934 .fc_metric = IP6_RT_PRIO_USER,
1935 .fc_ifindex = dev->ifindex,
1936 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1937 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1938 .fc_nlinfo.portid = 0,
1939 .fc_nlinfo.nlh = NULL,
1940 .fc_nlinfo.nl_net = dev_net(dev),
1943 cfg.fc_gateway = *gwaddr;
1945 ip6_route_add(&cfg);
1947 return rt6_get_dflt_router(gwaddr, dev);
1950 void rt6_purge_dflt_routers(struct net *net)
1952 struct rt6_info *rt;
1953 struct fib6_table *table;
1955 /* NOTE: Keep consistent with rt6_get_dflt_router */
1956 table = fib6_get_table(net, RT6_TABLE_DFLT);
1961 read_lock_bh(&table->tb6_lock);
1962 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1963 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1964 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
1966 read_unlock_bh(&table->tb6_lock);
1971 read_unlock_bh(&table->tb6_lock);
1974 static void rtmsg_to_fib6_config(struct net *net,
1975 struct in6_rtmsg *rtmsg,
1976 struct fib6_config *cfg)
1978 memset(cfg, 0, sizeof(*cfg));
1980 cfg->fc_table = RT6_TABLE_MAIN;
1981 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1982 cfg->fc_metric = rtmsg->rtmsg_metric;
1983 cfg->fc_expires = rtmsg->rtmsg_info;
1984 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1985 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1986 cfg->fc_flags = rtmsg->rtmsg_flags;
1988 cfg->fc_nlinfo.nl_net = net;
1990 cfg->fc_dst = rtmsg->rtmsg_dst;
1991 cfg->fc_src = rtmsg->rtmsg_src;
1992 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1995 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1997 struct fib6_config cfg;
1998 struct in6_rtmsg rtmsg;
2002 case SIOCADDRT: /* Add a route */
2003 case SIOCDELRT: /* Delete a route */
2004 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2006 err = copy_from_user(&rtmsg, arg,
2007 sizeof(struct in6_rtmsg));
2011 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2016 err = ip6_route_add(&cfg);
2019 err = ip6_route_del(&cfg);
2033 * Drop the packet on the floor
2036 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2039 struct dst_entry *dst = skb_dst(skb);
2040 switch (ipstats_mib_noroutes) {
2041 case IPSTATS_MIB_INNOROUTES:
2042 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2043 if (type == IPV6_ADDR_ANY) {
2044 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2045 IPSTATS_MIB_INADDRERRORS);
2049 case IPSTATS_MIB_OUTNOROUTES:
2050 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2051 ipstats_mib_noroutes);
2054 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2059 static int ip6_pkt_discard(struct sk_buff *skb)
2061 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2064 static int ip6_pkt_discard_out(struct sk_buff *skb)
2066 skb->dev = skb_dst(skb)->dev;
2067 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2070 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2072 static int ip6_pkt_prohibit(struct sk_buff *skb)
2074 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2077 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2079 skb->dev = skb_dst(skb)->dev;
2080 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2086 * Allocate a dst for local (unicast / anycast) address.
2089 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2090 const struct in6_addr *addr,
2093 struct net *net = dev_net(idev->dev);
2094 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2097 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2098 return ERR_PTR(-ENOMEM);
2103 rt->dst.flags |= DST_HOST;
2104 rt->dst.input = ip6_input;
2105 rt->dst.output = ip6_output;
2106 rt->rt6i_idev = idev;
2108 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2110 rt->rt6i_flags |= RTF_ANYCAST;
2112 rt->rt6i_flags |= RTF_LOCAL;
2114 rt->rt6i_dst.addr = *addr;
2115 rt->rt6i_dst.plen = 128;
2116 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2118 atomic_set(&rt->dst.__refcnt, 1);
2123 int ip6_route_get_saddr(struct net *net,
2124 struct rt6_info *rt,
2125 const struct in6_addr *daddr,
2127 struct in6_addr *saddr)
2129 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2131 if (rt->rt6i_prefsrc.plen)
2132 *saddr = rt->rt6i_prefsrc.addr;
2134 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2135 daddr, prefs, saddr);
2139 /* remove deleted ip from prefsrc entries */
2140 struct arg_dev_net_ip {
2141 struct net_device *dev;
2143 struct in6_addr *addr;
2146 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2148 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2149 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2150 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2152 if (((void *)rt->dst.dev == dev || !dev) &&
2153 rt != net->ipv6.ip6_null_entry &&
2154 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2155 /* remove prefsrc entry */
2156 rt->rt6i_prefsrc.plen = 0;
2161 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2163 struct net *net = dev_net(ifp->idev->dev);
2164 struct arg_dev_net_ip adni = {
2165 .dev = ifp->idev->dev,
2169 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2172 struct arg_dev_net {
2173 struct net_device *dev;
2177 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2179 const struct arg_dev_net *adn = arg;
2180 const struct net_device *dev = adn->dev;
2182 if ((rt->dst.dev == dev || !dev) &&
2183 rt != adn->net->ipv6.ip6_null_entry)
2189 void rt6_ifdown(struct net *net, struct net_device *dev)
2191 struct arg_dev_net adn = {
2196 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2197 icmp6_clean_all(fib6_ifdown, &adn);
2200 struct rt6_mtu_change_arg {
2201 struct net_device *dev;
2205 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2207 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2208 struct inet6_dev *idev;
2210 /* In IPv6 pmtu discovery is not optional,
2211 so that RTAX_MTU lock cannot disable it.
2212 We still use this lock to block changes
2213 caused by addrconf/ndisc.
2216 idev = __in6_dev_get(arg->dev);
2220 /* For administrative MTU increase, there is no way to discover
2221 IPv6 PMTU increase, so PMTU increase should be updated here.
2222 Since RFC 1981 doesn't include administrative MTU increase
2223 update PMTU increase is a MUST. (i.e. jumbo frame)
2226 If new MTU is less than route PMTU, this new MTU will be the
2227 lowest MTU in the path, update the route PMTU to reflect PMTU
2228 decreases; if new MTU is greater than route PMTU, and the
2229 old MTU is the lowest MTU in the path, update the route PMTU
2230 to reflect the increase. In this case if the other nodes' MTU
2231 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2234 if (rt->dst.dev == arg->dev &&
2235 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2236 (dst_mtu(&rt->dst) >= arg->mtu ||
2237 (dst_mtu(&rt->dst) < arg->mtu &&
2238 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2239 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2244 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2246 struct rt6_mtu_change_arg arg = {
2251 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2254 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2255 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2256 [RTA_OIF] = { .type = NLA_U32 },
2257 [RTA_IIF] = { .type = NLA_U32 },
2258 [RTA_PRIORITY] = { .type = NLA_U32 },
2259 [RTA_METRICS] = { .type = NLA_NESTED },
2260 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2263 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2264 struct fib6_config *cfg)
2267 struct nlattr *tb[RTA_MAX+1];
2270 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2275 rtm = nlmsg_data(nlh);
2276 memset(cfg, 0, sizeof(*cfg));
2278 cfg->fc_table = rtm->rtm_table;
2279 cfg->fc_dst_len = rtm->rtm_dst_len;
2280 cfg->fc_src_len = rtm->rtm_src_len;
2281 cfg->fc_flags = RTF_UP;
2282 cfg->fc_protocol = rtm->rtm_protocol;
2283 cfg->fc_type = rtm->rtm_type;
2285 if (rtm->rtm_type == RTN_UNREACHABLE ||
2286 rtm->rtm_type == RTN_BLACKHOLE ||
2287 rtm->rtm_type == RTN_PROHIBIT ||
2288 rtm->rtm_type == RTN_THROW)
2289 cfg->fc_flags |= RTF_REJECT;
2291 if (rtm->rtm_type == RTN_LOCAL)
2292 cfg->fc_flags |= RTF_LOCAL;
2294 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2295 cfg->fc_nlinfo.nlh = nlh;
2296 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2298 if (tb[RTA_GATEWAY]) {
2299 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2300 cfg->fc_flags |= RTF_GATEWAY;
2304 int plen = (rtm->rtm_dst_len + 7) >> 3;
2306 if (nla_len(tb[RTA_DST]) < plen)
2309 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2313 int plen = (rtm->rtm_src_len + 7) >> 3;
2315 if (nla_len(tb[RTA_SRC]) < plen)
2318 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2321 if (tb[RTA_PREFSRC])
2322 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2325 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2327 if (tb[RTA_PRIORITY])
2328 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2330 if (tb[RTA_METRICS]) {
2331 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2332 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2336 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2338 if (tb[RTA_MULTIPATH]) {
2339 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2340 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2348 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2350 struct fib6_config r_cfg;
2351 struct rtnexthop *rtnh;
2354 int err = 0, last_err = 0;
2357 rtnh = (struct rtnexthop *)cfg->fc_mp;
2358 remaining = cfg->fc_mp_len;
2360 /* Parse a Multipath Entry */
2361 while (rtnh_ok(rtnh, remaining)) {
2362 memcpy(&r_cfg, cfg, sizeof(*cfg));
2363 if (rtnh->rtnh_ifindex)
2364 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2366 attrlen = rtnh_attrlen(rtnh);
2368 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2370 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2372 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2373 r_cfg.fc_flags |= RTF_GATEWAY;
2376 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2379 /* If we are trying to remove a route, do not stop the
2380 * loop when ip6_route_del() fails (because next hop is
2381 * already gone), we should try to remove all next hops.
2384 /* If add fails, we should try to delete all
2385 * next hops that have been already added.
2391 /* Because each route is added like a single route we remove
2392 * this flag after the first nexthop (if there is a collision,
2393 * we have already fail to add the first nexthop:
2394 * fib6_add_rt2node() has reject it).
2396 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2397 rtnh = rtnh_next(rtnh, &remaining);
2403 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2405 struct fib6_config cfg;
2408 err = rtm_to_fib6_config(skb, nlh, &cfg);
2413 return ip6_route_multipath(&cfg, 0);
2415 return ip6_route_del(&cfg);
2418 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2420 struct fib6_config cfg;
2423 err = rtm_to_fib6_config(skb, nlh, &cfg);
2428 return ip6_route_multipath(&cfg, 1);
2430 return ip6_route_add(&cfg);
2433 static inline size_t rt6_nlmsg_size(void)
2435 return NLMSG_ALIGN(sizeof(struct rtmsg))
2436 + nla_total_size(16) /* RTA_SRC */
2437 + nla_total_size(16) /* RTA_DST */
2438 + nla_total_size(16) /* RTA_GATEWAY */
2439 + nla_total_size(16) /* RTA_PREFSRC */
2440 + nla_total_size(4) /* RTA_TABLE */
2441 + nla_total_size(4) /* RTA_IIF */
2442 + nla_total_size(4) /* RTA_OIF */
2443 + nla_total_size(4) /* RTA_PRIORITY */
2444 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2445 + nla_total_size(sizeof(struct rta_cacheinfo));
2448 static int rt6_fill_node(struct net *net,
2449 struct sk_buff *skb, struct rt6_info *rt,
2450 struct in6_addr *dst, struct in6_addr *src,
2451 int iif, int type, u32 portid, u32 seq,
2452 int prefix, int nowait, unsigned int flags)
2455 struct nlmsghdr *nlh;
2459 if (prefix) { /* user wants prefix routes only */
2460 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2461 /* success since this is not a prefix route */
2466 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2470 rtm = nlmsg_data(nlh);
2471 rtm->rtm_family = AF_INET6;
2472 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2473 rtm->rtm_src_len = rt->rt6i_src.plen;
2476 table = rt->rt6i_table->tb6_id;
2478 table = RT6_TABLE_UNSPEC;
2479 rtm->rtm_table = table;
2480 if (nla_put_u32(skb, RTA_TABLE, table))
2481 goto nla_put_failure;
2482 if (rt->rt6i_flags & RTF_REJECT) {
2483 switch (rt->dst.error) {
2485 rtm->rtm_type = RTN_BLACKHOLE;
2488 rtm->rtm_type = RTN_PROHIBIT;
2491 rtm->rtm_type = RTN_THROW;
2494 rtm->rtm_type = RTN_UNREACHABLE;
2498 else if (rt->rt6i_flags & RTF_LOCAL)
2499 rtm->rtm_type = RTN_LOCAL;
2500 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2501 rtm->rtm_type = RTN_LOCAL;
2503 rtm->rtm_type = RTN_UNICAST;
2505 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2506 rtm->rtm_protocol = rt->rt6i_protocol;
2507 if (rt->rt6i_flags & RTF_DYNAMIC)
2508 rtm->rtm_protocol = RTPROT_REDIRECT;
2509 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2510 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2511 rtm->rtm_protocol = RTPROT_RA;
2513 rtm->rtm_protocol = RTPROT_KERNEL;
2516 if (rt->rt6i_flags & RTF_CACHE)
2517 rtm->rtm_flags |= RTM_F_CLONED;
2520 if (nla_put(skb, RTA_DST, 16, dst))
2521 goto nla_put_failure;
2522 rtm->rtm_dst_len = 128;
2523 } else if (rtm->rtm_dst_len)
2524 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2525 goto nla_put_failure;
2526 #ifdef CONFIG_IPV6_SUBTREES
2528 if (nla_put(skb, RTA_SRC, 16, src))
2529 goto nla_put_failure;
2530 rtm->rtm_src_len = 128;
2531 } else if (rtm->rtm_src_len &&
2532 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2533 goto nla_put_failure;
2536 #ifdef CONFIG_IPV6_MROUTE
2537 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2538 int err = ip6mr_get_route(net, skb, rtm, nowait);
2543 goto nla_put_failure;
2545 if (err == -EMSGSIZE)
2546 goto nla_put_failure;
2551 if (nla_put_u32(skb, RTA_IIF, iif))
2552 goto nla_put_failure;
2554 struct in6_addr saddr_buf;
2555 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2556 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2557 goto nla_put_failure;
2560 if (rt->rt6i_prefsrc.plen) {
2561 struct in6_addr saddr_buf;
2562 saddr_buf = rt->rt6i_prefsrc.addr;
2563 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2564 goto nla_put_failure;
2567 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2568 goto nla_put_failure;
2570 if (rt->rt6i_flags & RTF_GATEWAY) {
2571 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2572 goto nla_put_failure;
2576 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2577 goto nla_put_failure;
2578 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2579 goto nla_put_failure;
2581 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2583 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2584 goto nla_put_failure;
2586 return nlmsg_end(skb, nlh);
2589 nlmsg_cancel(skb, nlh);
2593 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2595 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2598 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2599 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2600 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2604 return rt6_fill_node(arg->net,
2605 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2606 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2607 prefix, 0, NLM_F_MULTI);
2610 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2612 struct net *net = sock_net(in_skb->sk);
2613 struct nlattr *tb[RTA_MAX+1];
2614 struct rt6_info *rt;
2615 struct sk_buff *skb;
2618 int err, iif = 0, oif = 0;
2620 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2625 memset(&fl6, 0, sizeof(fl6));
2628 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2631 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2635 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2638 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2642 iif = nla_get_u32(tb[RTA_IIF]);
2645 oif = nla_get_u32(tb[RTA_OIF]);
2648 struct net_device *dev;
2651 dev = __dev_get_by_index(net, iif);
2657 fl6.flowi6_iif = iif;
2659 if (!ipv6_addr_any(&fl6.saddr))
2660 flags |= RT6_LOOKUP_F_HAS_SADDR;
2662 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2665 fl6.flowi6_oif = oif;
2667 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2670 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2677 /* Reserve room for dummy headers, this skb can pass
2678 through good chunk of routing engine.
2680 skb_reset_mac_header(skb);
2681 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2683 skb_dst_set(skb, &rt->dst);
2685 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2686 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2687 nlh->nlmsg_seq, 0, 0, 0);
2693 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2698 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2700 struct sk_buff *skb;
2701 struct net *net = info->nl_net;
2706 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2708 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2712 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2713 event, info->portid, seq, 0, 0, 0);
2715 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2716 WARN_ON(err == -EMSGSIZE);
2720 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2721 info->nlh, gfp_any());
2725 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2728 static int ip6_route_dev_notify(struct notifier_block *this,
2729 unsigned long event, void *ptr)
2731 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2732 struct net *net = dev_net(dev);
2734 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2735 net->ipv6.ip6_null_entry->dst.dev = dev;
2736 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2737 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2738 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2739 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2740 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2741 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2752 #ifdef CONFIG_PROC_FS
2763 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2765 struct seq_file *m = p_arg;
2767 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2769 #ifdef CONFIG_IPV6_SUBTREES
2770 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2772 seq_puts(m, "00000000000000000000000000000000 00 ");
2774 if (rt->rt6i_flags & RTF_GATEWAY) {
2775 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2777 seq_puts(m, "00000000000000000000000000000000");
2779 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2780 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2781 rt->dst.__use, rt->rt6i_flags,
2782 rt->dst.dev ? rt->dst.dev->name : "");
2786 static int ipv6_route_show(struct seq_file *m, void *v)
2788 struct net *net = (struct net *)m->private;
2789 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2793 static int ipv6_route_open(struct inode *inode, struct file *file)
2795 return single_open_net(inode, file, ipv6_route_show);
2798 static const struct file_operations ipv6_route_proc_fops = {
2799 .owner = THIS_MODULE,
2800 .open = ipv6_route_open,
2802 .llseek = seq_lseek,
2803 .release = single_release_net,
2806 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2808 struct net *net = (struct net *)seq->private;
2809 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2810 net->ipv6.rt6_stats->fib_nodes,
2811 net->ipv6.rt6_stats->fib_route_nodes,
2812 net->ipv6.rt6_stats->fib_rt_alloc,
2813 net->ipv6.rt6_stats->fib_rt_entries,
2814 net->ipv6.rt6_stats->fib_rt_cache,
2815 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2816 net->ipv6.rt6_stats->fib_discarded_routes);
2821 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2823 return single_open_net(inode, file, rt6_stats_seq_show);
2826 static const struct file_operations rt6_stats_seq_fops = {
2827 .owner = THIS_MODULE,
2828 .open = rt6_stats_seq_open,
2830 .llseek = seq_lseek,
2831 .release = single_release_net,
2833 #endif /* CONFIG_PROC_FS */
2835 #ifdef CONFIG_SYSCTL
2838 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2839 void __user *buffer, size_t *lenp, loff_t *ppos)
2846 net = (struct net *)ctl->extra1;
2847 delay = net->ipv6.sysctl.flush_delay;
2848 proc_dointvec(ctl, write, buffer, lenp, ppos);
2849 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2853 struct ctl_table ipv6_route_table_template[] = {
2855 .procname = "flush",
2856 .data = &init_net.ipv6.sysctl.flush_delay,
2857 .maxlen = sizeof(int),
2859 .proc_handler = ipv6_sysctl_rtcache_flush
2862 .procname = "gc_thresh",
2863 .data = &ip6_dst_ops_template.gc_thresh,
2864 .maxlen = sizeof(int),
2866 .proc_handler = proc_dointvec,
2869 .procname = "max_size",
2870 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2871 .maxlen = sizeof(int),
2873 .proc_handler = proc_dointvec,
2876 .procname = "gc_min_interval",
2877 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2878 .maxlen = sizeof(int),
2880 .proc_handler = proc_dointvec_jiffies,
2883 .procname = "gc_timeout",
2884 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2885 .maxlen = sizeof(int),
2887 .proc_handler = proc_dointvec_jiffies,
2890 .procname = "gc_interval",
2891 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2892 .maxlen = sizeof(int),
2894 .proc_handler = proc_dointvec_jiffies,
2897 .procname = "gc_elasticity",
2898 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2899 .maxlen = sizeof(int),
2901 .proc_handler = proc_dointvec,
2904 .procname = "mtu_expires",
2905 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2906 .maxlen = sizeof(int),
2908 .proc_handler = proc_dointvec_jiffies,
2911 .procname = "min_adv_mss",
2912 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2913 .maxlen = sizeof(int),
2915 .proc_handler = proc_dointvec,
2918 .procname = "gc_min_interval_ms",
2919 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2920 .maxlen = sizeof(int),
2922 .proc_handler = proc_dointvec_ms_jiffies,
2927 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2929 struct ctl_table *table;
2931 table = kmemdup(ipv6_route_table_template,
2932 sizeof(ipv6_route_table_template),
2936 table[0].data = &net->ipv6.sysctl.flush_delay;
2937 table[0].extra1 = net;
2938 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2939 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2940 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2941 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2942 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2943 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2944 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2945 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2946 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2948 /* Don't export sysctls to unprivileged users */
2949 if (net->user_ns != &init_user_ns)
2950 table[0].procname = NULL;
2957 static int __net_init ip6_route_net_init(struct net *net)
2961 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2962 sizeof(net->ipv6.ip6_dst_ops));
2964 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2965 goto out_ip6_dst_ops;
2967 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2968 sizeof(*net->ipv6.ip6_null_entry),
2970 if (!net->ipv6.ip6_null_entry)
2971 goto out_ip6_dst_entries;
2972 net->ipv6.ip6_null_entry->dst.path =
2973 (struct dst_entry *)net->ipv6.ip6_null_entry;
2974 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2975 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2976 ip6_template_metrics, true);
2978 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2979 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2980 sizeof(*net->ipv6.ip6_prohibit_entry),
2982 if (!net->ipv6.ip6_prohibit_entry)
2983 goto out_ip6_null_entry;
2984 net->ipv6.ip6_prohibit_entry->dst.path =
2985 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2986 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2987 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2988 ip6_template_metrics, true);
2990 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2991 sizeof(*net->ipv6.ip6_blk_hole_entry),
2993 if (!net->ipv6.ip6_blk_hole_entry)
2994 goto out_ip6_prohibit_entry;
2995 net->ipv6.ip6_blk_hole_entry->dst.path =
2996 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2997 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2998 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2999 ip6_template_metrics, true);
3002 net->ipv6.sysctl.flush_delay = 0;
3003 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3004 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3005 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3006 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3007 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3008 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3009 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3011 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3017 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3018 out_ip6_prohibit_entry:
3019 kfree(net->ipv6.ip6_prohibit_entry);
3021 kfree(net->ipv6.ip6_null_entry);
3023 out_ip6_dst_entries:
3024 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3029 static void __net_exit ip6_route_net_exit(struct net *net)
3031 kfree(net->ipv6.ip6_null_entry);
3032 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3033 kfree(net->ipv6.ip6_prohibit_entry);
3034 kfree(net->ipv6.ip6_blk_hole_entry);
3036 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3039 static int __net_init ip6_route_net_init_late(struct net *net)
3041 #ifdef CONFIG_PROC_FS
3042 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3043 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3048 static void __net_exit ip6_route_net_exit_late(struct net *net)
3050 #ifdef CONFIG_PROC_FS
3051 remove_proc_entry("ipv6_route", net->proc_net);
3052 remove_proc_entry("rt6_stats", net->proc_net);
3056 static struct pernet_operations ip6_route_net_ops = {
3057 .init = ip6_route_net_init,
3058 .exit = ip6_route_net_exit,
3061 static int __net_init ipv6_inetpeer_init(struct net *net)
3063 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3067 inet_peer_base_init(bp);
3068 net->ipv6.peers = bp;
3072 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3074 struct inet_peer_base *bp = net->ipv6.peers;
3076 net->ipv6.peers = NULL;
3077 inetpeer_invalidate_tree(bp);
3081 static struct pernet_operations ipv6_inetpeer_ops = {
3082 .init = ipv6_inetpeer_init,
3083 .exit = ipv6_inetpeer_exit,
3086 static struct pernet_operations ip6_route_net_late_ops = {
3087 .init = ip6_route_net_init_late,
3088 .exit = ip6_route_net_exit_late,
3091 static struct notifier_block ip6_route_dev_notifier = {
3092 .notifier_call = ip6_route_dev_notify,
3096 int __init ip6_route_init(void)
3101 ip6_dst_ops_template.kmem_cachep =
3102 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3103 SLAB_HWCACHE_ALIGN, NULL);
3104 if (!ip6_dst_ops_template.kmem_cachep)
3107 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3109 goto out_kmem_cache;
3111 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3113 goto out_dst_entries;
3115 ret = register_pernet_subsys(&ip6_route_net_ops);
3117 goto out_register_inetpeer;
3119 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3121 /* Registering of the loopback is done before this portion of code,
3122 * the loopback reference in rt6_info will not be taken, do it
3123 * manually for init_net */
3124 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3125 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3126 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3127 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3128 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3129 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3130 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3134 goto out_register_subsys;
3140 ret = fib6_rules_init();
3144 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3146 goto fib6_rules_init;
3149 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3150 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3151 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3152 goto out_register_late_subsys;
3154 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3156 goto out_register_late_subsys;
3161 out_register_late_subsys:
3162 unregister_pernet_subsys(&ip6_route_net_late_ops);
3164 fib6_rules_cleanup();
3169 out_register_subsys:
3170 unregister_pernet_subsys(&ip6_route_net_ops);
3171 out_register_inetpeer:
3172 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3174 dst_entries_destroy(&ip6_dst_blackhole_ops);
3176 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3180 void ip6_route_cleanup(void)
3182 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3183 unregister_pernet_subsys(&ip6_route_net_late_ops);
3184 fib6_rules_cleanup();
3187 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3188 unregister_pernet_subsys(&ip6_route_net_ops);
3189 dst_entries_destroy(&ip6_dst_blackhole_ops);
3190 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);