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>
61 #include <asm/uaccess.h>
64 #include <linux/sysctl.h>
67 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
68 const struct in6_addr *dest);
69 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
70 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
71 static unsigned int ip6_mtu(const struct dst_entry *dst);
72 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
73 static void ip6_dst_destroy(struct dst_entry *);
74 static void ip6_dst_ifdown(struct dst_entry *,
75 struct net_device *dev, int how);
76 static int ip6_dst_gc(struct dst_ops *ops);
78 static int ip6_pkt_discard(struct sk_buff *skb);
79 static int ip6_pkt_discard_out(struct sk_buff *skb);
80 static void ip6_link_failure(struct sk_buff *skb);
81 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
83 #ifdef CONFIG_IPV6_ROUTE_INFO
84 static struct rt6_info *rt6_add_route_info(struct net *net,
85 const struct in6_addr *prefix, int prefixlen,
86 const struct in6_addr *gwaddr, int ifindex,
88 static struct rt6_info *rt6_get_route_info(struct net *net,
89 const struct in6_addr *prefix, int prefixlen,
90 const struct in6_addr *gwaddr, int ifindex);
93 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
95 struct rt6_info *rt = (struct rt6_info *) dst;
96 struct inet_peer *peer;
99 if (!(rt->dst.flags & DST_HOST))
102 peer = rt6_get_peer_create(rt);
104 u32 *old_p = __DST_METRICS_PTR(old);
105 unsigned long prev, new;
108 if (inet_metrics_new(peer))
109 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
111 new = (unsigned long) p;
112 prev = cmpxchg(&dst->_metrics, old, new);
115 p = __DST_METRICS_PTR(prev);
116 if (prev & DST_METRICS_READ_ONLY)
123 static inline const void *choose_neigh_daddr(struct rt6_info *rt, const void *daddr)
125 struct in6_addr *p = &rt->rt6i_gateway;
127 if (!ipv6_addr_any(p))
128 return (const void *) p;
132 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst, const void *daddr)
134 struct rt6_info *rt = (struct rt6_info *) dst;
137 daddr = choose_neigh_daddr(rt, daddr);
138 n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
141 return neigh_create(&nd_tbl, daddr, dst->dev);
144 static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
146 struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
148 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
152 dst_set_neighbour(&rt->dst, n);
157 static struct dst_ops ip6_dst_ops_template = {
159 .protocol = cpu_to_be16(ETH_P_IPV6),
162 .check = ip6_dst_check,
163 .default_advmss = ip6_default_advmss,
165 .cow_metrics = ipv6_cow_metrics,
166 .destroy = ip6_dst_destroy,
167 .ifdown = ip6_dst_ifdown,
168 .negative_advice = ip6_negative_advice,
169 .link_failure = ip6_link_failure,
170 .update_pmtu = ip6_rt_update_pmtu,
171 .local_out = __ip6_local_out,
172 .neigh_lookup = ip6_neigh_lookup,
175 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
177 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
179 return mtu ? : dst->dev->mtu;
182 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
186 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
192 static struct dst_ops ip6_dst_blackhole_ops = {
194 .protocol = cpu_to_be16(ETH_P_IPV6),
195 .destroy = ip6_dst_destroy,
196 .check = ip6_dst_check,
197 .mtu = ip6_blackhole_mtu,
198 .default_advmss = ip6_default_advmss,
199 .update_pmtu = ip6_rt_blackhole_update_pmtu,
200 .cow_metrics = ip6_rt_blackhole_cow_metrics,
201 .neigh_lookup = ip6_neigh_lookup,
204 static const u32 ip6_template_metrics[RTAX_MAX] = {
205 [RTAX_HOPLIMIT - 1] = 255,
208 static struct rt6_info ip6_null_entry_template = {
210 .__refcnt = ATOMIC_INIT(1),
213 .error = -ENETUNREACH,
214 .input = ip6_pkt_discard,
215 .output = ip6_pkt_discard_out,
217 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
218 .rt6i_protocol = RTPROT_KERNEL,
219 .rt6i_metric = ~(u32) 0,
220 .rt6i_ref = ATOMIC_INIT(1),
223 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
225 static int ip6_pkt_prohibit(struct sk_buff *skb);
226 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
228 static struct rt6_info ip6_prohibit_entry_template = {
230 .__refcnt = ATOMIC_INIT(1),
234 .input = ip6_pkt_prohibit,
235 .output = ip6_pkt_prohibit_out,
237 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
238 .rt6i_protocol = RTPROT_KERNEL,
239 .rt6i_metric = ~(u32) 0,
240 .rt6i_ref = ATOMIC_INIT(1),
243 static struct rt6_info ip6_blk_hole_entry_template = {
245 .__refcnt = ATOMIC_INIT(1),
249 .input = dst_discard,
250 .output = dst_discard,
252 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
253 .rt6i_protocol = RTPROT_KERNEL,
254 .rt6i_metric = ~(u32) 0,
255 .rt6i_ref = ATOMIC_INIT(1),
260 /* allocate dst with ip6_dst_ops */
261 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
262 struct net_device *dev,
264 struct fib6_table *table)
266 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
270 memset(&rt->rt6i_table, 0,
271 sizeof(*rt) - sizeof(struct dst_entry));
272 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
277 static void ip6_dst_destroy(struct dst_entry *dst)
279 struct rt6_info *rt = (struct rt6_info *)dst;
280 struct inet6_dev *idev = rt->rt6i_idev;
282 if (!(rt->dst.flags & DST_HOST))
283 dst_destroy_metrics_generic(dst);
286 rt->rt6i_idev = NULL;
290 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
291 dst_release(dst->from);
293 if (rt6_has_peer(rt)) {
294 struct inet_peer *peer = rt6_peer_ptr(rt);
299 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
301 static u32 rt6_peer_genid(void)
303 return atomic_read(&__rt6_peer_genid);
306 void rt6_bind_peer(struct rt6_info *rt, int create)
308 struct inet_peer_base *base;
309 struct inet_peer *peer;
311 base = inetpeer_base_ptr(rt->_rt6i_peer);
315 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
317 if (!rt6_set_peer(rt, peer))
320 rt->rt6i_peer_genid = rt6_peer_genid();
324 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
327 struct rt6_info *rt = (struct rt6_info *)dst;
328 struct inet6_dev *idev = rt->rt6i_idev;
329 struct net_device *loopback_dev =
330 dev_net(dev)->loopback_dev;
332 if (dev != loopback_dev && idev && idev->dev == dev) {
333 struct inet6_dev *loopback_idev =
334 in6_dev_get(loopback_dev);
336 rt->rt6i_idev = loopback_idev;
342 static bool rt6_check_expired(const struct rt6_info *rt)
344 struct rt6_info *ort = NULL;
346 if (rt->rt6i_flags & RTF_EXPIRES) {
347 if (time_after(jiffies, rt->dst.expires))
349 } else if (rt->dst.from) {
350 ort = (struct rt6_info *) rt->dst.from;
351 return (ort->rt6i_flags & RTF_EXPIRES) &&
352 time_after(jiffies, ort->dst.expires);
357 static bool rt6_need_strict(const struct in6_addr *daddr)
359 return ipv6_addr_type(daddr) &
360 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
364 * Route lookup. Any table->tb6_lock is implied.
367 static inline struct rt6_info *rt6_device_match(struct net *net,
369 const struct in6_addr *saddr,
373 struct rt6_info *local = NULL;
374 struct rt6_info *sprt;
376 if (!oif && ipv6_addr_any(saddr))
379 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
380 struct net_device *dev = sprt->dst.dev;
383 if (dev->ifindex == oif)
385 if (dev->flags & IFF_LOOPBACK) {
386 if (!sprt->rt6i_idev ||
387 sprt->rt6i_idev->dev->ifindex != oif) {
388 if (flags & RT6_LOOKUP_F_IFACE && oif)
390 if (local && (!oif ||
391 local->rt6i_idev->dev->ifindex == oif))
397 if (ipv6_chk_addr(net, saddr, dev,
398 flags & RT6_LOOKUP_F_IFACE))
407 if (flags & RT6_LOOKUP_F_IFACE)
408 return net->ipv6.ip6_null_entry;
414 #ifdef CONFIG_IPV6_ROUTER_PREF
415 static void rt6_probe(struct rt6_info *rt)
417 struct neighbour *neigh;
419 * Okay, this does not seem to be appropriate
420 * for now, however, we need to check if it
421 * is really so; aka Router Reachability Probing.
423 * Router Reachability Probe MUST be rate-limited
424 * to no more than one per minute.
427 neigh = rt ? dst_get_neighbour_noref(&rt->dst) : NULL;
428 if (!neigh || (neigh->nud_state & NUD_VALID))
430 read_lock_bh(&neigh->lock);
431 if (!(neigh->nud_state & NUD_VALID) &&
432 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
433 struct in6_addr mcaddr;
434 struct in6_addr *target;
436 neigh->updated = jiffies;
437 read_unlock_bh(&neigh->lock);
439 target = (struct in6_addr *)&neigh->primary_key;
440 addrconf_addr_solict_mult(target, &mcaddr);
441 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
443 read_unlock_bh(&neigh->lock);
449 static inline void rt6_probe(struct rt6_info *rt)
455 * Default Router Selection (RFC 2461 6.3.6)
457 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
459 struct net_device *dev = rt->dst.dev;
460 if (!oif || dev->ifindex == oif)
462 if ((dev->flags & IFF_LOOPBACK) &&
463 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
468 static inline int rt6_check_neigh(struct rt6_info *rt)
470 struct neighbour *neigh;
474 neigh = dst_get_neighbour_noref(&rt->dst);
475 if (rt->rt6i_flags & RTF_NONEXTHOP ||
476 !(rt->rt6i_flags & RTF_GATEWAY))
479 read_lock_bh(&neigh->lock);
480 if (neigh->nud_state & NUD_VALID)
482 #ifdef CONFIG_IPV6_ROUTER_PREF
483 else if (neigh->nud_state & NUD_FAILED)
488 read_unlock_bh(&neigh->lock);
495 static int rt6_score_route(struct rt6_info *rt, int oif,
500 m = rt6_check_dev(rt, oif);
501 if (!m && (strict & RT6_LOOKUP_F_IFACE))
503 #ifdef CONFIG_IPV6_ROUTER_PREF
504 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
506 n = rt6_check_neigh(rt);
507 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
512 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
513 int *mpri, struct rt6_info *match)
517 if (rt6_check_expired(rt))
520 m = rt6_score_route(rt, oif, strict);
525 if (strict & RT6_LOOKUP_F_REACHABLE)
529 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
537 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
538 struct rt6_info *rr_head,
539 u32 metric, int oif, int strict)
541 struct rt6_info *rt, *match;
545 for (rt = rr_head; rt && rt->rt6i_metric == metric;
546 rt = rt->dst.rt6_next)
547 match = find_match(rt, oif, strict, &mpri, match);
548 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
549 rt = rt->dst.rt6_next)
550 match = find_match(rt, oif, strict, &mpri, match);
555 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
557 struct rt6_info *match, *rt0;
562 fn->rr_ptr = rt0 = fn->leaf;
564 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
567 (strict & RT6_LOOKUP_F_REACHABLE)) {
568 struct rt6_info *next = rt0->dst.rt6_next;
570 /* no entries matched; do round-robin */
571 if (!next || next->rt6i_metric != rt0->rt6i_metric)
578 net = dev_net(rt0->dst.dev);
579 return match ? match : net->ipv6.ip6_null_entry;
582 #ifdef CONFIG_IPV6_ROUTE_INFO
583 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
584 const struct in6_addr *gwaddr)
586 struct net *net = dev_net(dev);
587 struct route_info *rinfo = (struct route_info *) opt;
588 struct in6_addr prefix_buf, *prefix;
590 unsigned long lifetime;
593 if (len < sizeof(struct route_info)) {
597 /* Sanity check for prefix_len and length */
598 if (rinfo->length > 3) {
600 } else if (rinfo->prefix_len > 128) {
602 } else if (rinfo->prefix_len > 64) {
603 if (rinfo->length < 2) {
606 } else if (rinfo->prefix_len > 0) {
607 if (rinfo->length < 1) {
612 pref = rinfo->route_pref;
613 if (pref == ICMPV6_ROUTER_PREF_INVALID)
616 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
618 if (rinfo->length == 3)
619 prefix = (struct in6_addr *)rinfo->prefix;
621 /* this function is safe */
622 ipv6_addr_prefix(&prefix_buf,
623 (struct in6_addr *)rinfo->prefix,
625 prefix = &prefix_buf;
628 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
631 if (rt && !lifetime) {
637 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
640 rt->rt6i_flags = RTF_ROUTEINFO |
641 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
644 if (!addrconf_finite_timeout(lifetime))
645 rt6_clean_expires(rt);
647 rt6_set_expires(rt, jiffies + HZ * lifetime);
649 dst_release(&rt->dst);
655 #define BACKTRACK(__net, saddr) \
657 if (rt == __net->ipv6.ip6_null_entry) { \
658 struct fib6_node *pn; \
660 if (fn->fn_flags & RTN_TL_ROOT) \
663 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
664 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
667 if (fn->fn_flags & RTN_RTINFO) \
673 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
674 struct fib6_table *table,
675 struct flowi6 *fl6, int flags)
677 struct fib6_node *fn;
680 read_lock_bh(&table->tb6_lock);
681 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
684 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
685 BACKTRACK(net, &fl6->saddr);
687 dst_use(&rt->dst, jiffies);
688 read_unlock_bh(&table->tb6_lock);
693 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
696 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
698 EXPORT_SYMBOL_GPL(ip6_route_lookup);
700 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
701 const struct in6_addr *saddr, int oif, int strict)
703 struct flowi6 fl6 = {
707 struct dst_entry *dst;
708 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
711 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
712 flags |= RT6_LOOKUP_F_HAS_SADDR;
715 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
717 return (struct rt6_info *) dst;
724 EXPORT_SYMBOL(rt6_lookup);
726 /* ip6_ins_rt is called with FREE table->tb6_lock.
727 It takes new route entry, the addition fails by any reason the
728 route is freed. In any case, if caller does not hold it, it may
732 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
735 struct fib6_table *table;
737 table = rt->rt6i_table;
738 write_lock_bh(&table->tb6_lock);
739 err = fib6_add(&table->tb6_root, rt, info);
740 write_unlock_bh(&table->tb6_lock);
745 int ip6_ins_rt(struct rt6_info *rt)
747 struct nl_info info = {
748 .nl_net = dev_net(rt->dst.dev),
750 return __ip6_ins_rt(rt, &info);
753 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
754 const struct in6_addr *daddr,
755 const struct in6_addr *saddr)
763 rt = ip6_rt_copy(ort, daddr);
766 int attempts = !in_softirq();
768 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
769 if (ort->rt6i_dst.plen != 128 &&
770 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
771 rt->rt6i_flags |= RTF_ANYCAST;
772 rt->rt6i_gateway = *daddr;
775 rt->rt6i_flags |= RTF_CACHE;
777 #ifdef CONFIG_IPV6_SUBTREES
778 if (rt->rt6i_src.plen && saddr) {
779 rt->rt6i_src.addr = *saddr;
780 rt->rt6i_src.plen = 128;
785 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
786 struct net *net = dev_net(rt->dst.dev);
787 int saved_rt_min_interval =
788 net->ipv6.sysctl.ip6_rt_gc_min_interval;
789 int saved_rt_elasticity =
790 net->ipv6.sysctl.ip6_rt_gc_elasticity;
792 if (attempts-- > 0) {
793 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
794 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
796 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
798 net->ipv6.sysctl.ip6_rt_gc_elasticity =
800 net->ipv6.sysctl.ip6_rt_gc_min_interval =
801 saved_rt_min_interval;
805 net_warn_ratelimited("Neighbour table overflow\n");
814 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
815 const struct in6_addr *daddr)
817 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
820 rt->rt6i_flags |= RTF_CACHE;
821 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_noref_raw(&ort->dst)));
826 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
827 struct flowi6 *fl6, int flags)
829 struct fib6_node *fn;
830 struct rt6_info *rt, *nrt;
834 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
836 strict |= flags & RT6_LOOKUP_F_IFACE;
839 read_lock_bh(&table->tb6_lock);
842 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
845 rt = rt6_select(fn, oif, strict | reachable);
847 BACKTRACK(net, &fl6->saddr);
848 if (rt == net->ipv6.ip6_null_entry ||
849 rt->rt6i_flags & RTF_CACHE)
853 read_unlock_bh(&table->tb6_lock);
855 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
856 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
857 else if (!(rt->dst.flags & DST_HOST))
858 nrt = rt6_alloc_clone(rt, &fl6->daddr);
862 dst_release(&rt->dst);
863 rt = nrt ? : net->ipv6.ip6_null_entry;
867 err = ip6_ins_rt(nrt);
876 * Race condition! In the gap, when table->tb6_lock was
877 * released someone could insert this route. Relookup.
879 dst_release(&rt->dst);
888 read_unlock_bh(&table->tb6_lock);
890 rt->dst.lastuse = jiffies;
896 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
897 struct flowi6 *fl6, int flags)
899 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
902 static struct dst_entry *ip6_route_input_lookup(struct net *net,
903 struct net_device *dev,
904 struct flowi6 *fl6, int flags)
906 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
907 flags |= RT6_LOOKUP_F_IFACE;
909 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
912 void ip6_route_input(struct sk_buff *skb)
914 const struct ipv6hdr *iph = ipv6_hdr(skb);
915 struct net *net = dev_net(skb->dev);
916 int flags = RT6_LOOKUP_F_HAS_SADDR;
917 struct flowi6 fl6 = {
918 .flowi6_iif = skb->dev->ifindex,
921 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
922 .flowi6_mark = skb->mark,
923 .flowi6_proto = iph->nexthdr,
926 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
929 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
930 struct flowi6 *fl6, int flags)
932 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
935 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
940 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
941 flags |= RT6_LOOKUP_F_IFACE;
943 if (!ipv6_addr_any(&fl6->saddr))
944 flags |= RT6_LOOKUP_F_HAS_SADDR;
946 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
948 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
951 EXPORT_SYMBOL(ip6_route_output);
953 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
955 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
956 struct dst_entry *new = NULL;
958 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
960 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
961 rt6_init_peer(rt, net->ipv6.peers);
966 new->input = dst_discard;
967 new->output = dst_discard;
969 if (dst_metrics_read_only(&ort->dst))
970 new->_metrics = ort->dst._metrics;
972 dst_copy_metrics(new, &ort->dst);
973 rt->rt6i_idev = ort->rt6i_idev;
975 in6_dev_hold(rt->rt6i_idev);
977 rt->rt6i_gateway = ort->rt6i_gateway;
978 rt->rt6i_flags = ort->rt6i_flags;
979 rt6_clean_expires(rt);
982 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
983 #ifdef CONFIG_IPV6_SUBTREES
984 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
990 dst_release(dst_orig);
991 return new ? new : ERR_PTR(-ENOMEM);
995 * Destination cache support functions
998 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1000 struct rt6_info *rt;
1002 rt = (struct rt6_info *) dst;
1004 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
1005 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
1006 if (!rt6_has_peer(rt))
1007 rt6_bind_peer(rt, 0);
1008 rt->rt6i_peer_genid = rt6_peer_genid();
1015 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1017 struct rt6_info *rt = (struct rt6_info *) dst;
1020 if (rt->rt6i_flags & RTF_CACHE) {
1021 if (rt6_check_expired(rt)) {
1033 static void ip6_link_failure(struct sk_buff *skb)
1035 struct rt6_info *rt;
1037 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1039 rt = (struct rt6_info *) skb_dst(skb);
1041 if (rt->rt6i_flags & RTF_CACHE)
1042 rt6_update_expires(rt, 0);
1043 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1044 rt->rt6i_node->fn_sernum = -1;
1048 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1050 struct rt6_info *rt6 = (struct rt6_info*)dst;
1052 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1053 rt6->rt6i_flags |= RTF_MODIFIED;
1054 if (mtu < IPV6_MIN_MTU) {
1055 u32 features = dst_metric(dst, RTAX_FEATURES);
1057 features |= RTAX_FEATURE_ALLFRAG;
1058 dst_metric_set(dst, RTAX_FEATURES, features);
1060 dst_metric_set(dst, RTAX_MTU, mtu);
1064 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1066 struct net_device *dev = dst->dev;
1067 unsigned int mtu = dst_mtu(dst);
1068 struct net *net = dev_net(dev);
1070 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1072 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1073 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1076 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1077 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1078 * IPV6_MAXPLEN is also valid and means: "any MSS,
1079 * rely only on pmtu discovery"
1081 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1086 static unsigned int ip6_mtu(const struct dst_entry *dst)
1088 struct inet6_dev *idev;
1089 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1097 idev = __in6_dev_get(dst->dev);
1099 mtu = idev->cnf.mtu6;
1105 static struct dst_entry *icmp6_dst_gc_list;
1106 static DEFINE_SPINLOCK(icmp6_dst_lock);
1108 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1109 struct neighbour *neigh,
1112 struct dst_entry *dst;
1113 struct rt6_info *rt;
1114 struct inet6_dev *idev = in6_dev_get(dev);
1115 struct net *net = dev_net(dev);
1117 if (unlikely(!idev))
1118 return ERR_PTR(-ENODEV);
1120 rt = ip6_dst_alloc(net, dev, 0, NULL);
1121 if (unlikely(!rt)) {
1123 dst = ERR_PTR(-ENOMEM);
1130 neigh = ip6_neigh_lookup(&rt->dst, &fl6->daddr);
1131 if (IS_ERR(neigh)) {
1134 return ERR_CAST(neigh);
1138 rt->dst.flags |= DST_HOST;
1139 rt->dst.output = ip6_output;
1140 dst_set_neighbour(&rt->dst, neigh);
1141 atomic_set(&rt->dst.__refcnt, 1);
1142 rt->rt6i_dst.addr = fl6->daddr;
1143 rt->rt6i_dst.plen = 128;
1144 rt->rt6i_idev = idev;
1145 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1147 spin_lock_bh(&icmp6_dst_lock);
1148 rt->dst.next = icmp6_dst_gc_list;
1149 icmp6_dst_gc_list = &rt->dst;
1150 spin_unlock_bh(&icmp6_dst_lock);
1152 fib6_force_start_gc(net);
1154 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1160 int icmp6_dst_gc(void)
1162 struct dst_entry *dst, **pprev;
1165 spin_lock_bh(&icmp6_dst_lock);
1166 pprev = &icmp6_dst_gc_list;
1168 while ((dst = *pprev) != NULL) {
1169 if (!atomic_read(&dst->__refcnt)) {
1178 spin_unlock_bh(&icmp6_dst_lock);
1183 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1186 struct dst_entry *dst, **pprev;
1188 spin_lock_bh(&icmp6_dst_lock);
1189 pprev = &icmp6_dst_gc_list;
1190 while ((dst = *pprev) != NULL) {
1191 struct rt6_info *rt = (struct rt6_info *) dst;
1192 if (func(rt, arg)) {
1199 spin_unlock_bh(&icmp6_dst_lock);
1202 static int ip6_dst_gc(struct dst_ops *ops)
1204 unsigned long now = jiffies;
1205 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1206 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1207 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1208 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1209 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1210 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1213 entries = dst_entries_get_fast(ops);
1214 if (time_after(rt_last_gc + rt_min_interval, now) &&
1215 entries <= rt_max_size)
1218 net->ipv6.ip6_rt_gc_expire++;
1219 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1220 net->ipv6.ip6_rt_last_gc = now;
1221 entries = dst_entries_get_slow(ops);
1222 if (entries < ops->gc_thresh)
1223 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1225 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1226 return entries > rt_max_size;
1229 /* Clean host part of a prefix. Not necessary in radix tree,
1230 but results in cleaner routing tables.
1232 Remove it only when all the things will work!
1235 int ip6_dst_hoplimit(struct dst_entry *dst)
1237 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1238 if (hoplimit == 0) {
1239 struct net_device *dev = dst->dev;
1240 struct inet6_dev *idev;
1243 idev = __in6_dev_get(dev);
1245 hoplimit = idev->cnf.hop_limit;
1247 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1252 EXPORT_SYMBOL(ip6_dst_hoplimit);
1258 int ip6_route_add(struct fib6_config *cfg)
1261 struct net *net = cfg->fc_nlinfo.nl_net;
1262 struct rt6_info *rt = NULL;
1263 struct net_device *dev = NULL;
1264 struct inet6_dev *idev = NULL;
1265 struct fib6_table *table;
1268 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1270 #ifndef CONFIG_IPV6_SUBTREES
1271 if (cfg->fc_src_len)
1274 if (cfg->fc_ifindex) {
1276 dev = dev_get_by_index(net, cfg->fc_ifindex);
1279 idev = in6_dev_get(dev);
1284 if (cfg->fc_metric == 0)
1285 cfg->fc_metric = IP6_RT_PRIO_USER;
1288 if (cfg->fc_nlinfo.nlh &&
1289 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1290 table = fib6_get_table(net, cfg->fc_table);
1292 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1293 table = fib6_new_table(net, cfg->fc_table);
1296 table = fib6_new_table(net, cfg->fc_table);
1302 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1309 rt->dst.obsolete = -1;
1311 if (cfg->fc_flags & RTF_EXPIRES)
1312 rt6_set_expires(rt, jiffies +
1313 clock_t_to_jiffies(cfg->fc_expires));
1315 rt6_clean_expires(rt);
1317 if (cfg->fc_protocol == RTPROT_UNSPEC)
1318 cfg->fc_protocol = RTPROT_BOOT;
1319 rt->rt6i_protocol = cfg->fc_protocol;
1321 addr_type = ipv6_addr_type(&cfg->fc_dst);
1323 if (addr_type & IPV6_ADDR_MULTICAST)
1324 rt->dst.input = ip6_mc_input;
1325 else if (cfg->fc_flags & RTF_LOCAL)
1326 rt->dst.input = ip6_input;
1328 rt->dst.input = ip6_forward;
1330 rt->dst.output = ip6_output;
1332 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1333 rt->rt6i_dst.plen = cfg->fc_dst_len;
1334 if (rt->rt6i_dst.plen == 128)
1335 rt->dst.flags |= DST_HOST;
1337 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1338 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1343 dst_init_metrics(&rt->dst, metrics, 0);
1345 #ifdef CONFIG_IPV6_SUBTREES
1346 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1347 rt->rt6i_src.plen = cfg->fc_src_len;
1350 rt->rt6i_metric = cfg->fc_metric;
1352 /* We cannot add true routes via loopback here,
1353 they would result in kernel looping; promote them to reject routes
1355 if ((cfg->fc_flags & RTF_REJECT) ||
1356 (dev && (dev->flags & IFF_LOOPBACK) &&
1357 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1358 !(cfg->fc_flags & RTF_LOCAL))) {
1359 /* hold loopback dev/idev if we haven't done so. */
1360 if (dev != net->loopback_dev) {
1365 dev = net->loopback_dev;
1367 idev = in6_dev_get(dev);
1373 rt->dst.output = ip6_pkt_discard_out;
1374 rt->dst.input = ip6_pkt_discard;
1375 rt->dst.error = -ENETUNREACH;
1376 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1380 if (cfg->fc_flags & RTF_GATEWAY) {
1381 const struct in6_addr *gw_addr;
1384 gw_addr = &cfg->fc_gateway;
1385 rt->rt6i_gateway = *gw_addr;
1386 gwa_type = ipv6_addr_type(gw_addr);
1388 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1389 struct rt6_info *grt;
1391 /* IPv6 strictly inhibits using not link-local
1392 addresses as nexthop address.
1393 Otherwise, router will not able to send redirects.
1394 It is very good, but in some (rare!) circumstances
1395 (SIT, PtP, NBMA NOARP links) it is handy to allow
1396 some exceptions. --ANK
1399 if (!(gwa_type & IPV6_ADDR_UNICAST))
1402 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1404 err = -EHOSTUNREACH;
1408 if (dev != grt->dst.dev) {
1409 dst_release(&grt->dst);
1414 idev = grt->rt6i_idev;
1416 in6_dev_hold(grt->rt6i_idev);
1418 if (!(grt->rt6i_flags & RTF_GATEWAY))
1420 dst_release(&grt->dst);
1426 if (!dev || (dev->flags & IFF_LOOPBACK))
1434 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1435 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1439 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1440 rt->rt6i_prefsrc.plen = 128;
1442 rt->rt6i_prefsrc.plen = 0;
1444 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1445 err = rt6_bind_neighbour(rt, dev);
1450 rt->rt6i_flags = cfg->fc_flags;
1457 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1458 int type = nla_type(nla);
1461 if (type > RTAX_MAX) {
1466 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1472 rt->rt6i_idev = idev;
1473 rt->rt6i_table = table;
1475 cfg->fc_nlinfo.nl_net = dev_net(dev);
1477 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1489 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1492 struct fib6_table *table;
1493 struct net *net = dev_net(rt->dst.dev);
1495 if (rt == net->ipv6.ip6_null_entry)
1498 table = rt->rt6i_table;
1499 write_lock_bh(&table->tb6_lock);
1501 err = fib6_del(rt, info);
1502 dst_release(&rt->dst);
1504 write_unlock_bh(&table->tb6_lock);
1509 int ip6_del_rt(struct rt6_info *rt)
1511 struct nl_info info = {
1512 .nl_net = dev_net(rt->dst.dev),
1514 return __ip6_del_rt(rt, &info);
1517 static int ip6_route_del(struct fib6_config *cfg)
1519 struct fib6_table *table;
1520 struct fib6_node *fn;
1521 struct rt6_info *rt;
1524 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1528 read_lock_bh(&table->tb6_lock);
1530 fn = fib6_locate(&table->tb6_root,
1531 &cfg->fc_dst, cfg->fc_dst_len,
1532 &cfg->fc_src, cfg->fc_src_len);
1535 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1536 if (cfg->fc_ifindex &&
1538 rt->dst.dev->ifindex != cfg->fc_ifindex))
1540 if (cfg->fc_flags & RTF_GATEWAY &&
1541 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1543 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1546 read_unlock_bh(&table->tb6_lock);
1548 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1551 read_unlock_bh(&table->tb6_lock);
1559 struct ip6rd_flowi {
1561 struct in6_addr gateway;
1564 static struct rt6_info *__ip6_route_redirect(struct net *net,
1565 struct fib6_table *table,
1569 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1570 struct rt6_info *rt;
1571 struct fib6_node *fn;
1574 * Get the "current" route for this destination and
1575 * check if the redirect has come from approriate router.
1577 * RFC 2461 specifies that redirects should only be
1578 * accepted if they come from the nexthop to the target.
1579 * Due to the way the routes are chosen, this notion
1580 * is a bit fuzzy and one might need to check all possible
1584 read_lock_bh(&table->tb6_lock);
1585 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1587 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1589 * Current route is on-link; redirect is always invalid.
1591 * Seems, previous statement is not true. It could
1592 * be node, which looks for us as on-link (f.e. proxy ndisc)
1593 * But then router serving it might decide, that we should
1594 * know truth 8)8) --ANK (980726).
1596 if (rt6_check_expired(rt))
1598 if (!(rt->rt6i_flags & RTF_GATEWAY))
1600 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1602 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1608 rt = net->ipv6.ip6_null_entry;
1609 BACKTRACK(net, &fl6->saddr);
1613 read_unlock_bh(&table->tb6_lock);
1618 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1619 const struct in6_addr *src,
1620 const struct in6_addr *gateway,
1621 struct net_device *dev)
1623 int flags = RT6_LOOKUP_F_HAS_SADDR;
1624 struct net *net = dev_net(dev);
1625 struct ip6rd_flowi rdfl = {
1627 .flowi6_oif = dev->ifindex,
1633 rdfl.gateway = *gateway;
1635 if (rt6_need_strict(dest))
1636 flags |= RT6_LOOKUP_F_IFACE;
1638 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1639 flags, __ip6_route_redirect);
1642 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1643 const struct in6_addr *saddr,
1644 struct neighbour *neigh, u8 *lladdr, int on_link)
1646 struct rt6_info *rt, *nrt = NULL;
1647 struct netevent_redirect netevent;
1648 struct net *net = dev_net(neigh->dev);
1650 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1652 if (rt == net->ipv6.ip6_null_entry) {
1653 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1658 * We have finally decided to accept it.
1661 neigh_update(neigh, lladdr, NUD_STALE,
1662 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1663 NEIGH_UPDATE_F_OVERRIDE|
1664 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1665 NEIGH_UPDATE_F_ISROUTER))
1669 * Redirect received -> path was valid.
1670 * Look, redirects are sent only in response to data packets,
1671 * so that this nexthop apparently is reachable. --ANK
1673 dst_confirm(&rt->dst);
1675 /* Duplicate redirect: silently ignore. */
1676 if (neigh == dst_get_neighbour_noref_raw(&rt->dst))
1679 nrt = ip6_rt_copy(rt, dest);
1683 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1685 nrt->rt6i_flags &= ~RTF_GATEWAY;
1687 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1688 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1690 if (ip6_ins_rt(nrt))
1693 netevent.old = &rt->dst;
1694 netevent.new = &nrt->dst;
1695 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1697 if (rt->rt6i_flags & RTF_CACHE) {
1703 dst_release(&rt->dst);
1707 * Handle ICMP "packet too big" messages
1708 * i.e. Path MTU discovery
1711 static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1712 struct net *net, u32 pmtu, int ifindex)
1714 struct rt6_info *rt, *nrt;
1717 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1721 if (rt6_check_expired(rt)) {
1726 if (pmtu >= dst_mtu(&rt->dst))
1729 if (pmtu < IPV6_MIN_MTU) {
1731 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1732 * MTU (1280) and a fragment header should always be included
1733 * after a node receiving Too Big message reporting PMTU is
1734 * less than the IPv6 Minimum Link MTU.
1736 pmtu = IPV6_MIN_MTU;
1740 /* New mtu received -> path was valid.
1741 They are sent only in response to data packets,
1742 so that this nexthop apparently is reachable. --ANK
1744 dst_confirm(&rt->dst);
1746 /* Host route. If it is static, it would be better
1747 not to override it, but add new one, so that
1748 when cache entry will expire old pmtu
1749 would return automatically.
1751 if (rt->rt6i_flags & RTF_CACHE) {
1752 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1754 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1755 features |= RTAX_FEATURE_ALLFRAG;
1756 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1758 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1759 rt->rt6i_flags |= RTF_MODIFIED;
1764 Two cases are possible:
1765 1. It is connected route. Action: COW
1766 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1768 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1769 nrt = rt6_alloc_cow(rt, daddr, saddr);
1771 nrt = rt6_alloc_clone(rt, daddr);
1774 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1776 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1777 features |= RTAX_FEATURE_ALLFRAG;
1778 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1781 /* According to RFC 1981, detecting PMTU increase shouldn't be
1782 * happened within 5 mins, the recommended timer is 10 mins.
1783 * Here this route expiration time is set to ip6_rt_mtu_expires
1784 * which is 10 mins. After 10 mins the decreased pmtu is expired
1785 * and detecting PMTU increase will be automatically happened.
1787 rt6_update_expires(nrt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1788 nrt->rt6i_flags |= RTF_DYNAMIC;
1792 dst_release(&rt->dst);
1795 void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1796 struct net_device *dev, u32 pmtu)
1798 struct net *net = dev_net(dev);
1801 * RFC 1981 states that a node "MUST reduce the size of the packets it
1802 * is sending along the path" that caused the Packet Too Big message.
1803 * Since it's not possible in the general case to determine which
1804 * interface was used to send the original packet, we update the MTU
1805 * on the interface that will be used to send future packets. We also
1806 * update the MTU on the interface that received the Packet Too Big in
1807 * case the original packet was forced out that interface with
1808 * SO_BINDTODEVICE or similar. This is the next best thing to the
1809 * correct behaviour, which would be to update the MTU on all
1812 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1813 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1817 * Misc support functions
1820 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1821 const struct in6_addr *dest)
1823 struct net *net = dev_net(ort->dst.dev);
1824 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1828 rt->dst.input = ort->dst.input;
1829 rt->dst.output = ort->dst.output;
1830 rt->dst.flags |= DST_HOST;
1832 rt->rt6i_dst.addr = *dest;
1833 rt->rt6i_dst.plen = 128;
1834 dst_copy_metrics(&rt->dst, &ort->dst);
1835 rt->dst.error = ort->dst.error;
1836 rt->rt6i_idev = ort->rt6i_idev;
1838 in6_dev_hold(rt->rt6i_idev);
1839 rt->dst.lastuse = jiffies;
1841 rt->rt6i_gateway = ort->rt6i_gateway;
1842 rt->rt6i_flags = ort->rt6i_flags;
1843 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1844 (RTF_DEFAULT | RTF_ADDRCONF))
1845 rt6_set_from(rt, ort);
1847 rt6_clean_expires(rt);
1848 rt->rt6i_metric = 0;
1850 #ifdef CONFIG_IPV6_SUBTREES
1851 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1853 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1854 rt->rt6i_table = ort->rt6i_table;
1859 #ifdef CONFIG_IPV6_ROUTE_INFO
1860 static struct rt6_info *rt6_get_route_info(struct net *net,
1861 const struct in6_addr *prefix, int prefixlen,
1862 const struct in6_addr *gwaddr, int ifindex)
1864 struct fib6_node *fn;
1865 struct rt6_info *rt = NULL;
1866 struct fib6_table *table;
1868 table = fib6_get_table(net, RT6_TABLE_INFO);
1872 write_lock_bh(&table->tb6_lock);
1873 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1877 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1878 if (rt->dst.dev->ifindex != ifindex)
1880 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1882 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1888 write_unlock_bh(&table->tb6_lock);
1892 static struct rt6_info *rt6_add_route_info(struct net *net,
1893 const struct in6_addr *prefix, int prefixlen,
1894 const struct in6_addr *gwaddr, int ifindex,
1897 struct fib6_config cfg = {
1898 .fc_table = RT6_TABLE_INFO,
1899 .fc_metric = IP6_RT_PRIO_USER,
1900 .fc_ifindex = ifindex,
1901 .fc_dst_len = prefixlen,
1902 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1903 RTF_UP | RTF_PREF(pref),
1905 .fc_nlinfo.nlh = NULL,
1906 .fc_nlinfo.nl_net = net,
1909 cfg.fc_dst = *prefix;
1910 cfg.fc_gateway = *gwaddr;
1912 /* We should treat it as a default route if prefix length is 0. */
1914 cfg.fc_flags |= RTF_DEFAULT;
1916 ip6_route_add(&cfg);
1918 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1922 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1924 struct rt6_info *rt;
1925 struct fib6_table *table;
1927 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1931 write_lock_bh(&table->tb6_lock);
1932 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1933 if (dev == rt->dst.dev &&
1934 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1935 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1940 write_unlock_bh(&table->tb6_lock);
1944 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1945 struct net_device *dev,
1948 struct fib6_config cfg = {
1949 .fc_table = RT6_TABLE_DFLT,
1950 .fc_metric = IP6_RT_PRIO_USER,
1951 .fc_ifindex = dev->ifindex,
1952 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1953 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1955 .fc_nlinfo.nlh = NULL,
1956 .fc_nlinfo.nl_net = dev_net(dev),
1959 cfg.fc_gateway = *gwaddr;
1961 ip6_route_add(&cfg);
1963 return rt6_get_dflt_router(gwaddr, dev);
1966 void rt6_purge_dflt_routers(struct net *net)
1968 struct rt6_info *rt;
1969 struct fib6_table *table;
1971 /* NOTE: Keep consistent with rt6_get_dflt_router */
1972 table = fib6_get_table(net, RT6_TABLE_DFLT);
1977 read_lock_bh(&table->tb6_lock);
1978 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1979 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1981 read_unlock_bh(&table->tb6_lock);
1986 read_unlock_bh(&table->tb6_lock);
1989 static void rtmsg_to_fib6_config(struct net *net,
1990 struct in6_rtmsg *rtmsg,
1991 struct fib6_config *cfg)
1993 memset(cfg, 0, sizeof(*cfg));
1995 cfg->fc_table = RT6_TABLE_MAIN;
1996 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1997 cfg->fc_metric = rtmsg->rtmsg_metric;
1998 cfg->fc_expires = rtmsg->rtmsg_info;
1999 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2000 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2001 cfg->fc_flags = rtmsg->rtmsg_flags;
2003 cfg->fc_nlinfo.nl_net = net;
2005 cfg->fc_dst = rtmsg->rtmsg_dst;
2006 cfg->fc_src = rtmsg->rtmsg_src;
2007 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2010 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2012 struct fib6_config cfg;
2013 struct in6_rtmsg rtmsg;
2017 case SIOCADDRT: /* Add a route */
2018 case SIOCDELRT: /* Delete a route */
2019 if (!capable(CAP_NET_ADMIN))
2021 err = copy_from_user(&rtmsg, arg,
2022 sizeof(struct in6_rtmsg));
2026 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2031 err = ip6_route_add(&cfg);
2034 err = ip6_route_del(&cfg);
2048 * Drop the packet on the floor
2051 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2054 struct dst_entry *dst = skb_dst(skb);
2055 switch (ipstats_mib_noroutes) {
2056 case IPSTATS_MIB_INNOROUTES:
2057 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2058 if (type == IPV6_ADDR_ANY) {
2059 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2060 IPSTATS_MIB_INADDRERRORS);
2064 case IPSTATS_MIB_OUTNOROUTES:
2065 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2066 ipstats_mib_noroutes);
2069 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2074 static int ip6_pkt_discard(struct sk_buff *skb)
2076 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2079 static int ip6_pkt_discard_out(struct sk_buff *skb)
2081 skb->dev = skb_dst(skb)->dev;
2082 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2085 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2087 static int ip6_pkt_prohibit(struct sk_buff *skb)
2089 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2092 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2094 skb->dev = skb_dst(skb)->dev;
2095 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2101 * Allocate a dst for local (unicast / anycast) address.
2104 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2105 const struct in6_addr *addr,
2108 struct net *net = dev_net(idev->dev);
2109 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2113 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2114 return ERR_PTR(-ENOMEM);
2119 rt->dst.flags |= DST_HOST;
2120 rt->dst.input = ip6_input;
2121 rt->dst.output = ip6_output;
2122 rt->rt6i_idev = idev;
2123 rt->dst.obsolete = -1;
2125 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2127 rt->rt6i_flags |= RTF_ANYCAST;
2129 rt->rt6i_flags |= RTF_LOCAL;
2130 err = rt6_bind_neighbour(rt, rt->dst.dev);
2133 return ERR_PTR(err);
2136 rt->rt6i_dst.addr = *addr;
2137 rt->rt6i_dst.plen = 128;
2138 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2140 atomic_set(&rt->dst.__refcnt, 1);
2145 int ip6_route_get_saddr(struct net *net,
2146 struct rt6_info *rt,
2147 const struct in6_addr *daddr,
2149 struct in6_addr *saddr)
2151 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2153 if (rt->rt6i_prefsrc.plen)
2154 *saddr = rt->rt6i_prefsrc.addr;
2156 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2157 daddr, prefs, saddr);
2161 /* remove deleted ip from prefsrc entries */
2162 struct arg_dev_net_ip {
2163 struct net_device *dev;
2165 struct in6_addr *addr;
2168 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2170 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2171 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2172 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2174 if (((void *)rt->dst.dev == dev || !dev) &&
2175 rt != net->ipv6.ip6_null_entry &&
2176 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2177 /* remove prefsrc entry */
2178 rt->rt6i_prefsrc.plen = 0;
2183 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2185 struct net *net = dev_net(ifp->idev->dev);
2186 struct arg_dev_net_ip adni = {
2187 .dev = ifp->idev->dev,
2191 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2194 struct arg_dev_net {
2195 struct net_device *dev;
2199 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2201 const struct arg_dev_net *adn = arg;
2202 const struct net_device *dev = adn->dev;
2204 if ((rt->dst.dev == dev || !dev) &&
2205 rt != adn->net->ipv6.ip6_null_entry)
2211 void rt6_ifdown(struct net *net, struct net_device *dev)
2213 struct arg_dev_net adn = {
2218 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2219 icmp6_clean_all(fib6_ifdown, &adn);
2222 struct rt6_mtu_change_arg {
2223 struct net_device *dev;
2227 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2229 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2230 struct inet6_dev *idev;
2232 /* In IPv6 pmtu discovery is not optional,
2233 so that RTAX_MTU lock cannot disable it.
2234 We still use this lock to block changes
2235 caused by addrconf/ndisc.
2238 idev = __in6_dev_get(arg->dev);
2242 /* For administrative MTU increase, there is no way to discover
2243 IPv6 PMTU increase, so PMTU increase should be updated here.
2244 Since RFC 1981 doesn't include administrative MTU increase
2245 update PMTU increase is a MUST. (i.e. jumbo frame)
2248 If new MTU is less than route PMTU, this new MTU will be the
2249 lowest MTU in the path, update the route PMTU to reflect PMTU
2250 decreases; if new MTU is greater than route PMTU, and the
2251 old MTU is the lowest MTU in the path, update the route PMTU
2252 to reflect the increase. In this case if the other nodes' MTU
2253 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2256 if (rt->dst.dev == arg->dev &&
2257 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2258 (dst_mtu(&rt->dst) >= arg->mtu ||
2259 (dst_mtu(&rt->dst) < arg->mtu &&
2260 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2261 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2266 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2268 struct rt6_mtu_change_arg arg = {
2273 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2276 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2277 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2278 [RTA_OIF] = { .type = NLA_U32 },
2279 [RTA_IIF] = { .type = NLA_U32 },
2280 [RTA_PRIORITY] = { .type = NLA_U32 },
2281 [RTA_METRICS] = { .type = NLA_NESTED },
2284 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2285 struct fib6_config *cfg)
2288 struct nlattr *tb[RTA_MAX+1];
2291 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2296 rtm = nlmsg_data(nlh);
2297 memset(cfg, 0, sizeof(*cfg));
2299 cfg->fc_table = rtm->rtm_table;
2300 cfg->fc_dst_len = rtm->rtm_dst_len;
2301 cfg->fc_src_len = rtm->rtm_src_len;
2302 cfg->fc_flags = RTF_UP;
2303 cfg->fc_protocol = rtm->rtm_protocol;
2305 if (rtm->rtm_type == RTN_UNREACHABLE)
2306 cfg->fc_flags |= RTF_REJECT;
2308 if (rtm->rtm_type == RTN_LOCAL)
2309 cfg->fc_flags |= RTF_LOCAL;
2311 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2312 cfg->fc_nlinfo.nlh = nlh;
2313 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2315 if (tb[RTA_GATEWAY]) {
2316 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2317 cfg->fc_flags |= RTF_GATEWAY;
2321 int plen = (rtm->rtm_dst_len + 7) >> 3;
2323 if (nla_len(tb[RTA_DST]) < plen)
2326 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2330 int plen = (rtm->rtm_src_len + 7) >> 3;
2332 if (nla_len(tb[RTA_SRC]) < plen)
2335 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2338 if (tb[RTA_PREFSRC])
2339 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2342 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2344 if (tb[RTA_PRIORITY])
2345 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2347 if (tb[RTA_METRICS]) {
2348 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2349 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2353 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2360 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2362 struct fib6_config cfg;
2365 err = rtm_to_fib6_config(skb, nlh, &cfg);
2369 return ip6_route_del(&cfg);
2372 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2374 struct fib6_config cfg;
2377 err = rtm_to_fib6_config(skb, nlh, &cfg);
2381 return ip6_route_add(&cfg);
2384 static inline size_t rt6_nlmsg_size(void)
2386 return NLMSG_ALIGN(sizeof(struct rtmsg))
2387 + nla_total_size(16) /* RTA_SRC */
2388 + nla_total_size(16) /* RTA_DST */
2389 + nla_total_size(16) /* RTA_GATEWAY */
2390 + nla_total_size(16) /* RTA_PREFSRC */
2391 + nla_total_size(4) /* RTA_TABLE */
2392 + nla_total_size(4) /* RTA_IIF */
2393 + nla_total_size(4) /* RTA_OIF */
2394 + nla_total_size(4) /* RTA_PRIORITY */
2395 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2396 + nla_total_size(sizeof(struct rta_cacheinfo));
2399 static int rt6_fill_node(struct net *net,
2400 struct sk_buff *skb, struct rt6_info *rt,
2401 struct in6_addr *dst, struct in6_addr *src,
2402 int iif, int type, u32 pid, u32 seq,
2403 int prefix, int nowait, unsigned int flags)
2405 const struct inet_peer *peer;
2407 struct nlmsghdr *nlh;
2410 struct neighbour *n;
2413 if (prefix) { /* user wants prefix routes only */
2414 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2415 /* success since this is not a prefix route */
2420 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2424 rtm = nlmsg_data(nlh);
2425 rtm->rtm_family = AF_INET6;
2426 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2427 rtm->rtm_src_len = rt->rt6i_src.plen;
2430 table = rt->rt6i_table->tb6_id;
2432 table = RT6_TABLE_UNSPEC;
2433 rtm->rtm_table = table;
2434 if (nla_put_u32(skb, RTA_TABLE, table))
2435 goto nla_put_failure;
2436 if (rt->rt6i_flags & RTF_REJECT)
2437 rtm->rtm_type = RTN_UNREACHABLE;
2438 else if (rt->rt6i_flags & RTF_LOCAL)
2439 rtm->rtm_type = RTN_LOCAL;
2440 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2441 rtm->rtm_type = RTN_LOCAL;
2443 rtm->rtm_type = RTN_UNICAST;
2445 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2446 rtm->rtm_protocol = rt->rt6i_protocol;
2447 if (rt->rt6i_flags & RTF_DYNAMIC)
2448 rtm->rtm_protocol = RTPROT_REDIRECT;
2449 else if (rt->rt6i_flags & RTF_ADDRCONF)
2450 rtm->rtm_protocol = RTPROT_KERNEL;
2451 else if (rt->rt6i_flags & RTF_DEFAULT)
2452 rtm->rtm_protocol = RTPROT_RA;
2454 if (rt->rt6i_flags & RTF_CACHE)
2455 rtm->rtm_flags |= RTM_F_CLONED;
2458 if (nla_put(skb, RTA_DST, 16, dst))
2459 goto nla_put_failure;
2460 rtm->rtm_dst_len = 128;
2461 } else if (rtm->rtm_dst_len)
2462 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2463 goto nla_put_failure;
2464 #ifdef CONFIG_IPV6_SUBTREES
2466 if (nla_put(skb, RTA_SRC, 16, src))
2467 goto nla_put_failure;
2468 rtm->rtm_src_len = 128;
2469 } else if (rtm->rtm_src_len &&
2470 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2471 goto nla_put_failure;
2474 #ifdef CONFIG_IPV6_MROUTE
2475 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2476 int err = ip6mr_get_route(net, skb, rtm, nowait);
2481 goto nla_put_failure;
2483 if (err == -EMSGSIZE)
2484 goto nla_put_failure;
2489 if (nla_put_u32(skb, RTA_IIF, iif))
2490 goto nla_put_failure;
2492 struct in6_addr saddr_buf;
2493 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2494 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2495 goto nla_put_failure;
2498 if (rt->rt6i_prefsrc.plen) {
2499 struct in6_addr saddr_buf;
2500 saddr_buf = rt->rt6i_prefsrc.addr;
2501 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2502 goto nla_put_failure;
2505 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2506 goto nla_put_failure;
2509 n = dst_get_neighbour_noref(&rt->dst);
2511 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0) {
2513 goto nla_put_failure;
2519 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2520 goto nla_put_failure;
2521 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2522 goto nla_put_failure;
2523 if (!(rt->rt6i_flags & RTF_EXPIRES))
2525 else if (rt->dst.expires - jiffies < INT_MAX)
2526 expires = rt->dst.expires - jiffies;
2531 if (rt6_has_peer(rt))
2532 peer = rt6_peer_ptr(rt);
2534 if (peer && peer->tcp_ts_stamp) {
2536 tsage = get_seconds() - peer->tcp_ts_stamp;
2539 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, ts, tsage,
2540 expires, rt->dst.error) < 0)
2541 goto nla_put_failure;
2543 return nlmsg_end(skb, nlh);
2546 nlmsg_cancel(skb, nlh);
2550 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2552 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2555 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2556 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2557 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2561 return rt6_fill_node(arg->net,
2562 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2563 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2564 prefix, 0, NLM_F_MULTI);
2567 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2569 struct net *net = sock_net(in_skb->sk);
2570 struct nlattr *tb[RTA_MAX+1];
2571 struct rt6_info *rt;
2572 struct sk_buff *skb;
2575 int err, iif = 0, oif = 0;
2577 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2582 memset(&fl6, 0, sizeof(fl6));
2585 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2588 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2592 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2595 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2599 iif = nla_get_u32(tb[RTA_IIF]);
2602 oif = nla_get_u32(tb[RTA_OIF]);
2605 struct net_device *dev;
2608 dev = __dev_get_by_index(net, iif);
2614 fl6.flowi6_iif = iif;
2616 if (!ipv6_addr_any(&fl6.saddr))
2617 flags |= RT6_LOOKUP_F_HAS_SADDR;
2619 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2622 fl6.flowi6_oif = oif;
2624 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2627 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2629 dst_release(&rt->dst);
2634 /* Reserve room for dummy headers, this skb can pass
2635 through good chunk of routing engine.
2637 skb_reset_mac_header(skb);
2638 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2640 skb_dst_set(skb, &rt->dst);
2642 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2643 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2644 nlh->nlmsg_seq, 0, 0, 0);
2650 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2655 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2657 struct sk_buff *skb;
2658 struct net *net = info->nl_net;
2663 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2665 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2669 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2670 event, info->pid, seq, 0, 0, 0);
2672 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2673 WARN_ON(err == -EMSGSIZE);
2677 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2678 info->nlh, gfp_any());
2682 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2685 static int ip6_route_dev_notify(struct notifier_block *this,
2686 unsigned long event, void *data)
2688 struct net_device *dev = (struct net_device *)data;
2689 struct net *net = dev_net(dev);
2691 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2692 net->ipv6.ip6_null_entry->dst.dev = dev;
2693 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2694 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2695 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2696 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2697 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2698 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2709 #ifdef CONFIG_PROC_FS
2720 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2722 struct seq_file *m = p_arg;
2723 struct neighbour *n;
2725 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2727 #ifdef CONFIG_IPV6_SUBTREES
2728 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2730 seq_puts(m, "00000000000000000000000000000000 00 ");
2733 n = dst_get_neighbour_noref(&rt->dst);
2735 seq_printf(m, "%pi6", n->primary_key);
2737 seq_puts(m, "00000000000000000000000000000000");
2740 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2741 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2742 rt->dst.__use, rt->rt6i_flags,
2743 rt->dst.dev ? rt->dst.dev->name : "");
2747 static int ipv6_route_show(struct seq_file *m, void *v)
2749 struct net *net = (struct net *)m->private;
2750 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2754 static int ipv6_route_open(struct inode *inode, struct file *file)
2756 return single_open_net(inode, file, ipv6_route_show);
2759 static const struct file_operations ipv6_route_proc_fops = {
2760 .owner = THIS_MODULE,
2761 .open = ipv6_route_open,
2763 .llseek = seq_lseek,
2764 .release = single_release_net,
2767 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2769 struct net *net = (struct net *)seq->private;
2770 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2771 net->ipv6.rt6_stats->fib_nodes,
2772 net->ipv6.rt6_stats->fib_route_nodes,
2773 net->ipv6.rt6_stats->fib_rt_alloc,
2774 net->ipv6.rt6_stats->fib_rt_entries,
2775 net->ipv6.rt6_stats->fib_rt_cache,
2776 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2777 net->ipv6.rt6_stats->fib_discarded_routes);
2782 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2784 return single_open_net(inode, file, rt6_stats_seq_show);
2787 static const struct file_operations rt6_stats_seq_fops = {
2788 .owner = THIS_MODULE,
2789 .open = rt6_stats_seq_open,
2791 .llseek = seq_lseek,
2792 .release = single_release_net,
2794 #endif /* CONFIG_PROC_FS */
2796 #ifdef CONFIG_SYSCTL
2799 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2800 void __user *buffer, size_t *lenp, loff_t *ppos)
2807 net = (struct net *)ctl->extra1;
2808 delay = net->ipv6.sysctl.flush_delay;
2809 proc_dointvec(ctl, write, buffer, lenp, ppos);
2810 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2814 ctl_table ipv6_route_table_template[] = {
2816 .procname = "flush",
2817 .data = &init_net.ipv6.sysctl.flush_delay,
2818 .maxlen = sizeof(int),
2820 .proc_handler = ipv6_sysctl_rtcache_flush
2823 .procname = "gc_thresh",
2824 .data = &ip6_dst_ops_template.gc_thresh,
2825 .maxlen = sizeof(int),
2827 .proc_handler = proc_dointvec,
2830 .procname = "max_size",
2831 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2832 .maxlen = sizeof(int),
2834 .proc_handler = proc_dointvec,
2837 .procname = "gc_min_interval",
2838 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2839 .maxlen = sizeof(int),
2841 .proc_handler = proc_dointvec_jiffies,
2844 .procname = "gc_timeout",
2845 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2846 .maxlen = sizeof(int),
2848 .proc_handler = proc_dointvec_jiffies,
2851 .procname = "gc_interval",
2852 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2853 .maxlen = sizeof(int),
2855 .proc_handler = proc_dointvec_jiffies,
2858 .procname = "gc_elasticity",
2859 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2860 .maxlen = sizeof(int),
2862 .proc_handler = proc_dointvec,
2865 .procname = "mtu_expires",
2866 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2867 .maxlen = sizeof(int),
2869 .proc_handler = proc_dointvec_jiffies,
2872 .procname = "min_adv_mss",
2873 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2874 .maxlen = sizeof(int),
2876 .proc_handler = proc_dointvec,
2879 .procname = "gc_min_interval_ms",
2880 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2881 .maxlen = sizeof(int),
2883 .proc_handler = proc_dointvec_ms_jiffies,
2888 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2890 struct ctl_table *table;
2892 table = kmemdup(ipv6_route_table_template,
2893 sizeof(ipv6_route_table_template),
2897 table[0].data = &net->ipv6.sysctl.flush_delay;
2898 table[0].extra1 = net;
2899 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2900 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2901 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2902 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2903 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2904 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2905 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2906 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2907 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2914 static int __net_init ip6_route_net_init(struct net *net)
2918 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2919 sizeof(net->ipv6.ip6_dst_ops));
2921 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2922 goto out_ip6_dst_ops;
2924 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2925 sizeof(*net->ipv6.ip6_null_entry),
2927 if (!net->ipv6.ip6_null_entry)
2928 goto out_ip6_dst_entries;
2929 net->ipv6.ip6_null_entry->dst.path =
2930 (struct dst_entry *)net->ipv6.ip6_null_entry;
2931 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2932 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2933 ip6_template_metrics, true);
2935 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2936 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2937 sizeof(*net->ipv6.ip6_prohibit_entry),
2939 if (!net->ipv6.ip6_prohibit_entry)
2940 goto out_ip6_null_entry;
2941 net->ipv6.ip6_prohibit_entry->dst.path =
2942 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2943 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2944 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2945 ip6_template_metrics, true);
2947 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2948 sizeof(*net->ipv6.ip6_blk_hole_entry),
2950 if (!net->ipv6.ip6_blk_hole_entry)
2951 goto out_ip6_prohibit_entry;
2952 net->ipv6.ip6_blk_hole_entry->dst.path =
2953 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2954 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2955 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2956 ip6_template_metrics, true);
2959 net->ipv6.sysctl.flush_delay = 0;
2960 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2961 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2962 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2963 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2964 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2965 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2966 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2968 #ifdef CONFIG_PROC_FS
2969 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2970 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2972 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2978 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2979 out_ip6_prohibit_entry:
2980 kfree(net->ipv6.ip6_prohibit_entry);
2982 kfree(net->ipv6.ip6_null_entry);
2984 out_ip6_dst_entries:
2985 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2990 static void __net_exit ip6_route_net_exit(struct net *net)
2992 #ifdef CONFIG_PROC_FS
2993 proc_net_remove(net, "ipv6_route");
2994 proc_net_remove(net, "rt6_stats");
2996 kfree(net->ipv6.ip6_null_entry);
2997 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2998 kfree(net->ipv6.ip6_prohibit_entry);
2999 kfree(net->ipv6.ip6_blk_hole_entry);
3001 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3004 static struct pernet_operations ip6_route_net_ops = {
3005 .init = ip6_route_net_init,
3006 .exit = ip6_route_net_exit,
3009 static int __net_init ipv6_inetpeer_init(struct net *net)
3011 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3015 inet_peer_base_init(bp);
3016 net->ipv6.peers = bp;
3020 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3022 struct inet_peer_base *bp = net->ipv6.peers;
3024 net->ipv6.peers = NULL;
3025 inetpeer_invalidate_tree(bp);
3029 static struct pernet_operations ipv6_inetpeer_ops = {
3030 .init = ipv6_inetpeer_init,
3031 .exit = ipv6_inetpeer_exit,
3034 static struct notifier_block ip6_route_dev_notifier = {
3035 .notifier_call = ip6_route_dev_notify,
3039 int __init ip6_route_init(void)
3044 ip6_dst_ops_template.kmem_cachep =
3045 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3046 SLAB_HWCACHE_ALIGN, NULL);
3047 if (!ip6_dst_ops_template.kmem_cachep)
3050 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3052 goto out_kmem_cache;
3054 ret = register_pernet_subsys(&ip6_route_net_ops);
3056 goto out_dst_entries;
3058 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3060 goto out_register_subsys;
3062 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3064 /* Registering of the loopback is done before this portion of code,
3065 * the loopback reference in rt6_info will not be taken, do it
3066 * manually for init_net */
3067 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3068 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3069 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3070 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3071 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3072 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3073 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3077 goto out_register_inetpeer;
3083 ret = fib6_rules_init();
3088 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3089 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3090 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3091 goto fib6_rules_init;
3093 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3095 goto fib6_rules_init;
3101 fib6_rules_cleanup();
3106 out_register_inetpeer:
3107 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3108 out_register_subsys:
3109 unregister_pernet_subsys(&ip6_route_net_ops);
3111 dst_entries_destroy(&ip6_dst_blackhole_ops);
3113 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3117 void ip6_route_cleanup(void)
3119 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3120 fib6_rules_cleanup();
3123 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3124 unregister_pernet_subsys(&ip6_route_net_ops);
3125 dst_entries_destroy(&ip6_dst_blackhole_ops);
3126 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);