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);
82 static void rt6_do_redirect(struct dst_entry *dst, struct sk_buff *skb);
84 #ifdef CONFIG_IPV6_ROUTE_INFO
85 static struct rt6_info *rt6_add_route_info(struct net *net,
86 const struct in6_addr *prefix, int prefixlen,
87 const struct in6_addr *gwaddr, int ifindex,
89 static struct rt6_info *rt6_get_route_info(struct net *net,
90 const struct in6_addr *prefix, int prefixlen,
91 const struct in6_addr *gwaddr, int ifindex);
94 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
96 struct rt6_info *rt = (struct rt6_info *) dst;
97 struct inet_peer *peer;
100 if (!(rt->dst.flags & DST_HOST))
103 peer = rt6_get_peer_create(rt);
105 u32 *old_p = __DST_METRICS_PTR(old);
106 unsigned long prev, new;
109 if (inet_metrics_new(peer))
110 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
112 new = (unsigned long) p;
113 prev = cmpxchg(&dst->_metrics, old, new);
116 p = __DST_METRICS_PTR(prev);
117 if (prev & DST_METRICS_READ_ONLY)
124 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
128 struct in6_addr *p = &rt->rt6i_gateway;
130 if (!ipv6_addr_any(p))
131 return (const void *) p;
133 return &ipv6_hdr(skb)->daddr;
137 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
141 struct rt6_info *rt = (struct rt6_info *) dst;
144 daddr = choose_neigh_daddr(rt, skb, daddr);
145 n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
148 return neigh_create(&nd_tbl, daddr, dst->dev);
151 static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
153 struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
155 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
164 static struct dst_ops ip6_dst_ops_template = {
166 .protocol = cpu_to_be16(ETH_P_IPV6),
169 .check = ip6_dst_check,
170 .default_advmss = ip6_default_advmss,
172 .cow_metrics = ipv6_cow_metrics,
173 .destroy = ip6_dst_destroy,
174 .ifdown = ip6_dst_ifdown,
175 .negative_advice = ip6_negative_advice,
176 .link_failure = ip6_link_failure,
177 .update_pmtu = ip6_rt_update_pmtu,
178 .redirect = rt6_do_redirect,
179 .local_out = __ip6_local_out,
180 .neigh_lookup = ip6_neigh_lookup,
183 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
185 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
187 return mtu ? : dst->dev->mtu;
190 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
194 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
200 static struct dst_ops ip6_dst_blackhole_ops = {
202 .protocol = cpu_to_be16(ETH_P_IPV6),
203 .destroy = ip6_dst_destroy,
204 .check = ip6_dst_check,
205 .mtu = ip6_blackhole_mtu,
206 .default_advmss = ip6_default_advmss,
207 .update_pmtu = ip6_rt_blackhole_update_pmtu,
208 .cow_metrics = ip6_rt_blackhole_cow_metrics,
209 .neigh_lookup = ip6_neigh_lookup,
212 static const u32 ip6_template_metrics[RTAX_MAX] = {
213 [RTAX_HOPLIMIT - 1] = 255,
216 static struct rt6_info ip6_null_entry_template = {
218 .__refcnt = ATOMIC_INIT(1),
221 .error = -ENETUNREACH,
222 .input = ip6_pkt_discard,
223 .output = ip6_pkt_discard_out,
225 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
226 .rt6i_protocol = RTPROT_KERNEL,
227 .rt6i_metric = ~(u32) 0,
228 .rt6i_ref = ATOMIC_INIT(1),
231 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
233 static int ip6_pkt_prohibit(struct sk_buff *skb);
234 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
236 static struct rt6_info ip6_prohibit_entry_template = {
238 .__refcnt = ATOMIC_INIT(1),
242 .input = ip6_pkt_prohibit,
243 .output = ip6_pkt_prohibit_out,
245 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
246 .rt6i_protocol = RTPROT_KERNEL,
247 .rt6i_metric = ~(u32) 0,
248 .rt6i_ref = ATOMIC_INIT(1),
251 static struct rt6_info ip6_blk_hole_entry_template = {
253 .__refcnt = ATOMIC_INIT(1),
257 .input = dst_discard,
258 .output = dst_discard,
260 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
261 .rt6i_protocol = RTPROT_KERNEL,
262 .rt6i_metric = ~(u32) 0,
263 .rt6i_ref = ATOMIC_INIT(1),
268 /* allocate dst with ip6_dst_ops */
269 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
270 struct net_device *dev,
272 struct fib6_table *table)
274 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
279 sizeof(*rt) - sizeof(struct dst_entry));
280 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
285 static void ip6_dst_destroy(struct dst_entry *dst)
287 struct rt6_info *rt = (struct rt6_info *)dst;
288 struct inet6_dev *idev = rt->rt6i_idev;
291 neigh_release(rt->n);
293 if (!(rt->dst.flags & DST_HOST))
294 dst_destroy_metrics_generic(dst);
297 rt->rt6i_idev = NULL;
301 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
302 dst_release(dst->from);
304 if (rt6_has_peer(rt)) {
305 struct inet_peer *peer = rt6_peer_ptr(rt);
310 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
312 static u32 rt6_peer_genid(void)
314 return atomic_read(&__rt6_peer_genid);
317 void rt6_bind_peer(struct rt6_info *rt, int create)
319 struct inet_peer_base *base;
320 struct inet_peer *peer;
322 base = inetpeer_base_ptr(rt->_rt6i_peer);
326 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
328 if (!rt6_set_peer(rt, peer))
331 rt->rt6i_peer_genid = rt6_peer_genid();
335 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
338 struct rt6_info *rt = (struct rt6_info *)dst;
339 struct inet6_dev *idev = rt->rt6i_idev;
340 struct net_device *loopback_dev =
341 dev_net(dev)->loopback_dev;
343 if (dev != loopback_dev) {
344 if (idev && idev->dev == dev) {
345 struct inet6_dev *loopback_idev =
346 in6_dev_get(loopback_dev);
348 rt->rt6i_idev = loopback_idev;
352 if (rt->n && rt->n->dev == dev) {
353 rt->n->dev = loopback_dev;
354 dev_hold(loopback_dev);
360 static bool rt6_check_expired(const struct rt6_info *rt)
362 struct rt6_info *ort = NULL;
364 if (rt->rt6i_flags & RTF_EXPIRES) {
365 if (time_after(jiffies, rt->dst.expires))
367 } else if (rt->dst.from) {
368 ort = (struct rt6_info *) rt->dst.from;
369 return (ort->rt6i_flags & RTF_EXPIRES) &&
370 time_after(jiffies, ort->dst.expires);
375 static bool rt6_need_strict(const struct in6_addr *daddr)
377 return ipv6_addr_type(daddr) &
378 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
382 * Route lookup. Any table->tb6_lock is implied.
385 static inline struct rt6_info *rt6_device_match(struct net *net,
387 const struct in6_addr *saddr,
391 struct rt6_info *local = NULL;
392 struct rt6_info *sprt;
394 if (!oif && ipv6_addr_any(saddr))
397 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
398 struct net_device *dev = sprt->dst.dev;
401 if (dev->ifindex == oif)
403 if (dev->flags & IFF_LOOPBACK) {
404 if (!sprt->rt6i_idev ||
405 sprt->rt6i_idev->dev->ifindex != oif) {
406 if (flags & RT6_LOOKUP_F_IFACE && oif)
408 if (local && (!oif ||
409 local->rt6i_idev->dev->ifindex == oif))
415 if (ipv6_chk_addr(net, saddr, dev,
416 flags & RT6_LOOKUP_F_IFACE))
425 if (flags & RT6_LOOKUP_F_IFACE)
426 return net->ipv6.ip6_null_entry;
432 #ifdef CONFIG_IPV6_ROUTER_PREF
433 static void rt6_probe(struct rt6_info *rt)
435 struct neighbour *neigh;
437 * Okay, this does not seem to be appropriate
438 * for now, however, we need to check if it
439 * is really so; aka Router Reachability Probing.
441 * Router Reachability Probe MUST be rate-limited
442 * to no more than one per minute.
445 neigh = rt ? rt->n : NULL;
446 if (!neigh || (neigh->nud_state & NUD_VALID))
448 read_lock_bh(&neigh->lock);
449 if (!(neigh->nud_state & NUD_VALID) &&
450 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
451 struct in6_addr mcaddr;
452 struct in6_addr *target;
454 neigh->updated = jiffies;
455 read_unlock_bh(&neigh->lock);
457 target = (struct in6_addr *)&neigh->primary_key;
458 addrconf_addr_solict_mult(target, &mcaddr);
459 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
461 read_unlock_bh(&neigh->lock);
467 static inline void rt6_probe(struct rt6_info *rt)
473 * Default Router Selection (RFC 2461 6.3.6)
475 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
477 struct net_device *dev = rt->dst.dev;
478 if (!oif || dev->ifindex == oif)
480 if ((dev->flags & IFF_LOOPBACK) &&
481 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
486 static inline int rt6_check_neigh(struct rt6_info *rt)
488 struct neighbour *neigh;
493 if (rt->rt6i_flags & RTF_NONEXTHOP ||
494 !(rt->rt6i_flags & RTF_GATEWAY))
497 read_lock_bh(&neigh->lock);
498 if (neigh->nud_state & NUD_VALID)
500 #ifdef CONFIG_IPV6_ROUTER_PREF
501 else if (neigh->nud_state & NUD_FAILED)
506 read_unlock_bh(&neigh->lock);
513 static int rt6_score_route(struct rt6_info *rt, int oif,
518 m = rt6_check_dev(rt, oif);
519 if (!m && (strict & RT6_LOOKUP_F_IFACE))
521 #ifdef CONFIG_IPV6_ROUTER_PREF
522 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
524 n = rt6_check_neigh(rt);
525 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
530 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
531 int *mpri, struct rt6_info *match)
535 if (rt6_check_expired(rt))
538 m = rt6_score_route(rt, oif, strict);
543 if (strict & RT6_LOOKUP_F_REACHABLE)
547 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
555 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
556 struct rt6_info *rr_head,
557 u32 metric, int oif, int strict)
559 struct rt6_info *rt, *match;
563 for (rt = rr_head; rt && rt->rt6i_metric == metric;
564 rt = rt->dst.rt6_next)
565 match = find_match(rt, oif, strict, &mpri, match);
566 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
567 rt = rt->dst.rt6_next)
568 match = find_match(rt, oif, strict, &mpri, match);
573 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
575 struct rt6_info *match, *rt0;
580 fn->rr_ptr = rt0 = fn->leaf;
582 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
585 (strict & RT6_LOOKUP_F_REACHABLE)) {
586 struct rt6_info *next = rt0->dst.rt6_next;
588 /* no entries matched; do round-robin */
589 if (!next || next->rt6i_metric != rt0->rt6i_metric)
596 net = dev_net(rt0->dst.dev);
597 return match ? match : net->ipv6.ip6_null_entry;
600 #ifdef CONFIG_IPV6_ROUTE_INFO
601 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
602 const struct in6_addr *gwaddr)
604 struct net *net = dev_net(dev);
605 struct route_info *rinfo = (struct route_info *) opt;
606 struct in6_addr prefix_buf, *prefix;
608 unsigned long lifetime;
611 if (len < sizeof(struct route_info)) {
615 /* Sanity check for prefix_len and length */
616 if (rinfo->length > 3) {
618 } else if (rinfo->prefix_len > 128) {
620 } else if (rinfo->prefix_len > 64) {
621 if (rinfo->length < 2) {
624 } else if (rinfo->prefix_len > 0) {
625 if (rinfo->length < 1) {
630 pref = rinfo->route_pref;
631 if (pref == ICMPV6_ROUTER_PREF_INVALID)
634 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
636 if (rinfo->length == 3)
637 prefix = (struct in6_addr *)rinfo->prefix;
639 /* this function is safe */
640 ipv6_addr_prefix(&prefix_buf,
641 (struct in6_addr *)rinfo->prefix,
643 prefix = &prefix_buf;
646 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
649 if (rt && !lifetime) {
655 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
658 rt->rt6i_flags = RTF_ROUTEINFO |
659 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
662 if (!addrconf_finite_timeout(lifetime))
663 rt6_clean_expires(rt);
665 rt6_set_expires(rt, jiffies + HZ * lifetime);
667 dst_release(&rt->dst);
673 #define BACKTRACK(__net, saddr) \
675 if (rt == __net->ipv6.ip6_null_entry) { \
676 struct fib6_node *pn; \
678 if (fn->fn_flags & RTN_TL_ROOT) \
681 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
682 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
685 if (fn->fn_flags & RTN_RTINFO) \
691 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
692 struct fib6_table *table,
693 struct flowi6 *fl6, int flags)
695 struct fib6_node *fn;
698 read_lock_bh(&table->tb6_lock);
699 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
702 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
703 BACKTRACK(net, &fl6->saddr);
705 dst_use(&rt->dst, jiffies);
706 read_unlock_bh(&table->tb6_lock);
711 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
714 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
716 EXPORT_SYMBOL_GPL(ip6_route_lookup);
718 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
719 const struct in6_addr *saddr, int oif, int strict)
721 struct flowi6 fl6 = {
725 struct dst_entry *dst;
726 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
729 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
730 flags |= RT6_LOOKUP_F_HAS_SADDR;
733 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
735 return (struct rt6_info *) dst;
742 EXPORT_SYMBOL(rt6_lookup);
744 /* ip6_ins_rt is called with FREE table->tb6_lock.
745 It takes new route entry, the addition fails by any reason the
746 route is freed. In any case, if caller does not hold it, it may
750 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
753 struct fib6_table *table;
755 table = rt->rt6i_table;
756 write_lock_bh(&table->tb6_lock);
757 err = fib6_add(&table->tb6_root, rt, info);
758 write_unlock_bh(&table->tb6_lock);
763 int ip6_ins_rt(struct rt6_info *rt)
765 struct nl_info info = {
766 .nl_net = dev_net(rt->dst.dev),
768 return __ip6_ins_rt(rt, &info);
771 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
772 const struct in6_addr *daddr,
773 const struct in6_addr *saddr)
781 rt = ip6_rt_copy(ort, daddr);
784 int attempts = !in_softirq();
786 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
787 if (ort->rt6i_dst.plen != 128 &&
788 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
789 rt->rt6i_flags |= RTF_ANYCAST;
790 rt->rt6i_gateway = *daddr;
793 rt->rt6i_flags |= RTF_CACHE;
795 #ifdef CONFIG_IPV6_SUBTREES
796 if (rt->rt6i_src.plen && saddr) {
797 rt->rt6i_src.addr = *saddr;
798 rt->rt6i_src.plen = 128;
803 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
804 struct net *net = dev_net(rt->dst.dev);
805 int saved_rt_min_interval =
806 net->ipv6.sysctl.ip6_rt_gc_min_interval;
807 int saved_rt_elasticity =
808 net->ipv6.sysctl.ip6_rt_gc_elasticity;
810 if (attempts-- > 0) {
811 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
812 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
814 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
816 net->ipv6.sysctl.ip6_rt_gc_elasticity =
818 net->ipv6.sysctl.ip6_rt_gc_min_interval =
819 saved_rt_min_interval;
823 net_warn_ratelimited("Neighbour table overflow\n");
832 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
833 const struct in6_addr *daddr)
835 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
838 rt->rt6i_flags |= RTF_CACHE;
839 rt->n = neigh_clone(ort->n);
844 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
845 struct flowi6 *fl6, int flags)
847 struct fib6_node *fn;
848 struct rt6_info *rt, *nrt;
852 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
854 strict |= flags & RT6_LOOKUP_F_IFACE;
857 read_lock_bh(&table->tb6_lock);
860 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
863 rt = rt6_select(fn, oif, strict | reachable);
865 BACKTRACK(net, &fl6->saddr);
866 if (rt == net->ipv6.ip6_null_entry ||
867 rt->rt6i_flags & RTF_CACHE)
871 read_unlock_bh(&table->tb6_lock);
873 if (!rt->n && !(rt->rt6i_flags & RTF_NONEXTHOP))
874 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
875 else if (!(rt->dst.flags & DST_HOST))
876 nrt = rt6_alloc_clone(rt, &fl6->daddr);
880 dst_release(&rt->dst);
881 rt = nrt ? : net->ipv6.ip6_null_entry;
885 err = ip6_ins_rt(nrt);
894 * Race condition! In the gap, when table->tb6_lock was
895 * released someone could insert this route. Relookup.
897 dst_release(&rt->dst);
906 read_unlock_bh(&table->tb6_lock);
908 rt->dst.lastuse = jiffies;
914 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
915 struct flowi6 *fl6, int flags)
917 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
920 static struct dst_entry *ip6_route_input_lookup(struct net *net,
921 struct net_device *dev,
922 struct flowi6 *fl6, int flags)
924 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
925 flags |= RT6_LOOKUP_F_IFACE;
927 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
930 void ip6_route_input(struct sk_buff *skb)
932 const struct ipv6hdr *iph = ipv6_hdr(skb);
933 struct net *net = dev_net(skb->dev);
934 int flags = RT6_LOOKUP_F_HAS_SADDR;
935 struct flowi6 fl6 = {
936 .flowi6_iif = skb->dev->ifindex,
939 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
940 .flowi6_mark = skb->mark,
941 .flowi6_proto = iph->nexthdr,
944 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
947 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
948 struct flowi6 *fl6, int flags)
950 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
953 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
958 fl6->flowi6_iif = net->loopback_dev->ifindex;
960 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
961 flags |= RT6_LOOKUP_F_IFACE;
963 if (!ipv6_addr_any(&fl6->saddr))
964 flags |= RT6_LOOKUP_F_HAS_SADDR;
966 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
968 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
971 EXPORT_SYMBOL(ip6_route_output);
973 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
975 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
976 struct dst_entry *new = NULL;
978 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
980 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
981 rt6_init_peer(rt, net->ipv6.peers);
986 new->input = dst_discard;
987 new->output = dst_discard;
989 if (dst_metrics_read_only(&ort->dst))
990 new->_metrics = ort->dst._metrics;
992 dst_copy_metrics(new, &ort->dst);
993 rt->rt6i_idev = ort->rt6i_idev;
995 in6_dev_hold(rt->rt6i_idev);
997 rt->rt6i_gateway = ort->rt6i_gateway;
998 rt->rt6i_flags = ort->rt6i_flags;
999 rt6_clean_expires(rt);
1000 rt->rt6i_metric = 0;
1002 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1003 #ifdef CONFIG_IPV6_SUBTREES
1004 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1010 dst_release(dst_orig);
1011 return new ? new : ERR_PTR(-ENOMEM);
1015 * Destination cache support functions
1018 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1020 struct rt6_info *rt;
1022 rt = (struct rt6_info *) dst;
1024 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
1025 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
1026 if (!rt6_has_peer(rt))
1027 rt6_bind_peer(rt, 0);
1028 rt->rt6i_peer_genid = rt6_peer_genid();
1035 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1037 struct rt6_info *rt = (struct rt6_info *) dst;
1040 if (rt->rt6i_flags & RTF_CACHE) {
1041 if (rt6_check_expired(rt)) {
1053 static void ip6_link_failure(struct sk_buff *skb)
1055 struct rt6_info *rt;
1057 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1059 rt = (struct rt6_info *) skb_dst(skb);
1061 if (rt->rt6i_flags & RTF_CACHE)
1062 rt6_update_expires(rt, 0);
1063 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1064 rt->rt6i_node->fn_sernum = -1;
1068 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1070 struct rt6_info *rt6 = (struct rt6_info*)dst;
1073 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1074 struct net *net = dev_net(dst->dev);
1076 rt6->rt6i_flags |= RTF_MODIFIED;
1077 if (mtu < IPV6_MIN_MTU) {
1078 u32 features = dst_metric(dst, RTAX_FEATURES);
1080 features |= RTAX_FEATURE_ALLFRAG;
1081 dst_metric_set(dst, RTAX_FEATURES, features);
1083 dst_metric_set(dst, RTAX_MTU, mtu);
1084 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1088 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1091 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1092 struct dst_entry *dst;
1095 memset(&fl6, 0, sizeof(fl6));
1096 fl6.flowi6_oif = oif;
1097 fl6.flowi6_mark = mark;
1098 fl6.flowi6_flags = 0;
1099 fl6.daddr = iph->daddr;
1100 fl6.saddr = iph->saddr;
1101 fl6.flowlabel = (*(__be32 *) iph) & IPV6_FLOWINFO_MASK;
1103 dst = ip6_route_output(net, NULL, &fl6);
1105 ip6_rt_update_pmtu(dst, ntohl(mtu));
1108 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1110 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1112 ip6_update_pmtu(skb, sock_net(sk), mtu,
1113 sk->sk_bound_dev_if, sk->sk_mark);
1115 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1117 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1119 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1120 struct dst_entry *dst;
1123 memset(&fl6, 0, sizeof(fl6));
1124 fl6.flowi6_oif = oif;
1125 fl6.flowi6_mark = mark;
1126 fl6.flowi6_flags = 0;
1127 fl6.daddr = iph->daddr;
1128 fl6.saddr = iph->saddr;
1129 fl6.flowlabel = (*(__be32 *) iph) & IPV6_FLOWINFO_MASK;
1131 dst = ip6_route_output(net, NULL, &fl6);
1133 rt6_do_redirect(dst, skb);
1136 EXPORT_SYMBOL_GPL(ip6_redirect);
1138 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1140 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1142 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1144 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1146 struct net_device *dev = dst->dev;
1147 unsigned int mtu = dst_mtu(dst);
1148 struct net *net = dev_net(dev);
1150 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1152 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1153 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1156 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1157 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1158 * IPV6_MAXPLEN is also valid and means: "any MSS,
1159 * rely only on pmtu discovery"
1161 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1166 static unsigned int ip6_mtu(const struct dst_entry *dst)
1168 struct inet6_dev *idev;
1169 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1177 idev = __in6_dev_get(dst->dev);
1179 mtu = idev->cnf.mtu6;
1185 static struct dst_entry *icmp6_dst_gc_list;
1186 static DEFINE_SPINLOCK(icmp6_dst_lock);
1188 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1189 struct neighbour *neigh,
1192 struct dst_entry *dst;
1193 struct rt6_info *rt;
1194 struct inet6_dev *idev = in6_dev_get(dev);
1195 struct net *net = dev_net(dev);
1197 if (unlikely(!idev))
1198 return ERR_PTR(-ENODEV);
1200 rt = ip6_dst_alloc(net, dev, 0, NULL);
1201 if (unlikely(!rt)) {
1203 dst = ERR_PTR(-ENOMEM);
1210 neigh = ip6_neigh_lookup(&rt->dst, NULL, &fl6->daddr);
1211 if (IS_ERR(neigh)) {
1214 return ERR_CAST(neigh);
1218 rt->dst.flags |= DST_HOST;
1219 rt->dst.output = ip6_output;
1221 atomic_set(&rt->dst.__refcnt, 1);
1222 rt->rt6i_dst.addr = fl6->daddr;
1223 rt->rt6i_dst.plen = 128;
1224 rt->rt6i_idev = idev;
1225 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1227 spin_lock_bh(&icmp6_dst_lock);
1228 rt->dst.next = icmp6_dst_gc_list;
1229 icmp6_dst_gc_list = &rt->dst;
1230 spin_unlock_bh(&icmp6_dst_lock);
1232 fib6_force_start_gc(net);
1234 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1240 int icmp6_dst_gc(void)
1242 struct dst_entry *dst, **pprev;
1245 spin_lock_bh(&icmp6_dst_lock);
1246 pprev = &icmp6_dst_gc_list;
1248 while ((dst = *pprev) != NULL) {
1249 if (!atomic_read(&dst->__refcnt)) {
1258 spin_unlock_bh(&icmp6_dst_lock);
1263 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1266 struct dst_entry *dst, **pprev;
1268 spin_lock_bh(&icmp6_dst_lock);
1269 pprev = &icmp6_dst_gc_list;
1270 while ((dst = *pprev) != NULL) {
1271 struct rt6_info *rt = (struct rt6_info *) dst;
1272 if (func(rt, arg)) {
1279 spin_unlock_bh(&icmp6_dst_lock);
1282 static int ip6_dst_gc(struct dst_ops *ops)
1284 unsigned long now = jiffies;
1285 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1286 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1287 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1288 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1289 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1290 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1293 entries = dst_entries_get_fast(ops);
1294 if (time_after(rt_last_gc + rt_min_interval, now) &&
1295 entries <= rt_max_size)
1298 net->ipv6.ip6_rt_gc_expire++;
1299 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1300 net->ipv6.ip6_rt_last_gc = now;
1301 entries = dst_entries_get_slow(ops);
1302 if (entries < ops->gc_thresh)
1303 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1305 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1306 return entries > rt_max_size;
1309 /* Clean host part of a prefix. Not necessary in radix tree,
1310 but results in cleaner routing tables.
1312 Remove it only when all the things will work!
1315 int ip6_dst_hoplimit(struct dst_entry *dst)
1317 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1318 if (hoplimit == 0) {
1319 struct net_device *dev = dst->dev;
1320 struct inet6_dev *idev;
1323 idev = __in6_dev_get(dev);
1325 hoplimit = idev->cnf.hop_limit;
1327 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1332 EXPORT_SYMBOL(ip6_dst_hoplimit);
1338 int ip6_route_add(struct fib6_config *cfg)
1341 struct net *net = cfg->fc_nlinfo.nl_net;
1342 struct rt6_info *rt = NULL;
1343 struct net_device *dev = NULL;
1344 struct inet6_dev *idev = NULL;
1345 struct fib6_table *table;
1348 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1350 #ifndef CONFIG_IPV6_SUBTREES
1351 if (cfg->fc_src_len)
1354 if (cfg->fc_ifindex) {
1356 dev = dev_get_by_index(net, cfg->fc_ifindex);
1359 idev = in6_dev_get(dev);
1364 if (cfg->fc_metric == 0)
1365 cfg->fc_metric = IP6_RT_PRIO_USER;
1368 if (cfg->fc_nlinfo.nlh &&
1369 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1370 table = fib6_get_table(net, cfg->fc_table);
1372 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1373 table = fib6_new_table(net, cfg->fc_table);
1376 table = fib6_new_table(net, cfg->fc_table);
1382 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1389 rt->dst.obsolete = -1;
1391 if (cfg->fc_flags & RTF_EXPIRES)
1392 rt6_set_expires(rt, jiffies +
1393 clock_t_to_jiffies(cfg->fc_expires));
1395 rt6_clean_expires(rt);
1397 if (cfg->fc_protocol == RTPROT_UNSPEC)
1398 cfg->fc_protocol = RTPROT_BOOT;
1399 rt->rt6i_protocol = cfg->fc_protocol;
1401 addr_type = ipv6_addr_type(&cfg->fc_dst);
1403 if (addr_type & IPV6_ADDR_MULTICAST)
1404 rt->dst.input = ip6_mc_input;
1405 else if (cfg->fc_flags & RTF_LOCAL)
1406 rt->dst.input = ip6_input;
1408 rt->dst.input = ip6_forward;
1410 rt->dst.output = ip6_output;
1412 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1413 rt->rt6i_dst.plen = cfg->fc_dst_len;
1414 if (rt->rt6i_dst.plen == 128)
1415 rt->dst.flags |= DST_HOST;
1417 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1418 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1423 dst_init_metrics(&rt->dst, metrics, 0);
1425 #ifdef CONFIG_IPV6_SUBTREES
1426 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1427 rt->rt6i_src.plen = cfg->fc_src_len;
1430 rt->rt6i_metric = cfg->fc_metric;
1432 /* We cannot add true routes via loopback here,
1433 they would result in kernel looping; promote them to reject routes
1435 if ((cfg->fc_flags & RTF_REJECT) ||
1436 (dev && (dev->flags & IFF_LOOPBACK) &&
1437 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1438 !(cfg->fc_flags & RTF_LOCAL))) {
1439 /* hold loopback dev/idev if we haven't done so. */
1440 if (dev != net->loopback_dev) {
1445 dev = net->loopback_dev;
1447 idev = in6_dev_get(dev);
1453 rt->dst.output = ip6_pkt_discard_out;
1454 rt->dst.input = ip6_pkt_discard;
1455 rt->dst.error = -ENETUNREACH;
1456 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1460 if (cfg->fc_flags & RTF_GATEWAY) {
1461 const struct in6_addr *gw_addr;
1464 gw_addr = &cfg->fc_gateway;
1465 rt->rt6i_gateway = *gw_addr;
1466 gwa_type = ipv6_addr_type(gw_addr);
1468 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1469 struct rt6_info *grt;
1471 /* IPv6 strictly inhibits using not link-local
1472 addresses as nexthop address.
1473 Otherwise, router will not able to send redirects.
1474 It is very good, but in some (rare!) circumstances
1475 (SIT, PtP, NBMA NOARP links) it is handy to allow
1476 some exceptions. --ANK
1479 if (!(gwa_type & IPV6_ADDR_UNICAST))
1482 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1484 err = -EHOSTUNREACH;
1488 if (dev != grt->dst.dev) {
1489 dst_release(&grt->dst);
1494 idev = grt->rt6i_idev;
1496 in6_dev_hold(grt->rt6i_idev);
1498 if (!(grt->rt6i_flags & RTF_GATEWAY))
1500 dst_release(&grt->dst);
1506 if (!dev || (dev->flags & IFF_LOOPBACK))
1514 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1515 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1519 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1520 rt->rt6i_prefsrc.plen = 128;
1522 rt->rt6i_prefsrc.plen = 0;
1524 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1525 err = rt6_bind_neighbour(rt, dev);
1530 rt->rt6i_flags = cfg->fc_flags;
1537 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1538 int type = nla_type(nla);
1541 if (type > RTAX_MAX) {
1546 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1552 rt->rt6i_idev = idev;
1553 rt->rt6i_table = table;
1555 cfg->fc_nlinfo.nl_net = dev_net(dev);
1557 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1569 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1572 struct fib6_table *table;
1573 struct net *net = dev_net(rt->dst.dev);
1575 if (rt == net->ipv6.ip6_null_entry)
1578 table = rt->rt6i_table;
1579 write_lock_bh(&table->tb6_lock);
1581 err = fib6_del(rt, info);
1582 dst_release(&rt->dst);
1584 write_unlock_bh(&table->tb6_lock);
1589 int ip6_del_rt(struct rt6_info *rt)
1591 struct nl_info info = {
1592 .nl_net = dev_net(rt->dst.dev),
1594 return __ip6_del_rt(rt, &info);
1597 static int ip6_route_del(struct fib6_config *cfg)
1599 struct fib6_table *table;
1600 struct fib6_node *fn;
1601 struct rt6_info *rt;
1604 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1608 read_lock_bh(&table->tb6_lock);
1610 fn = fib6_locate(&table->tb6_root,
1611 &cfg->fc_dst, cfg->fc_dst_len,
1612 &cfg->fc_src, cfg->fc_src_len);
1615 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1616 if (cfg->fc_ifindex &&
1618 rt->dst.dev->ifindex != cfg->fc_ifindex))
1620 if (cfg->fc_flags & RTF_GATEWAY &&
1621 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1623 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1626 read_unlock_bh(&table->tb6_lock);
1628 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1631 read_unlock_bh(&table->tb6_lock);
1636 static void rt6_do_redirect(struct dst_entry *dst, struct sk_buff *skb)
1638 struct net *net = dev_net(skb->dev);
1639 struct netevent_redirect netevent;
1640 struct rt6_info *rt, *nrt = NULL;
1641 const struct in6_addr *target;
1642 struct ndisc_options ndopts;
1643 const struct in6_addr *dest;
1644 struct neighbour *old_neigh;
1645 struct inet6_dev *in6_dev;
1646 struct neighbour *neigh;
1647 struct icmp6hdr *icmph;
1648 int optlen, on_link;
1651 optlen = skb->tail - skb->transport_header;
1652 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1655 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1659 icmph = icmp6_hdr(skb);
1660 target = (const struct in6_addr *) (icmph + 1);
1663 if (ipv6_addr_is_multicast(dest)) {
1664 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1669 if (ipv6_addr_equal(dest, target)) {
1671 } else if (ipv6_addr_type(target) !=
1672 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1673 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1677 in6_dev = __in6_dev_get(skb->dev);
1680 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1684 * The IP source address of the Redirect MUST be the same as the current
1685 * first-hop router for the specified ICMP Destination Address.
1688 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1689 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1694 if (ndopts.nd_opts_tgt_lladdr) {
1695 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1698 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1703 rt = (struct rt6_info *) dst;
1704 if (rt == net->ipv6.ip6_null_entry) {
1705 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1709 /* Redirect received -> path was valid.
1710 * Look, redirects are sent only in response to data packets,
1711 * so that this nexthop apparently is reachable. --ANK
1713 dst_confirm(&rt->dst);
1715 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1719 /* Duplicate redirect: silently ignore. */
1721 if (neigh == old_neigh)
1725 * We have finally decided to accept it.
1728 neigh_update(neigh, lladdr, NUD_STALE,
1729 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1730 NEIGH_UPDATE_F_OVERRIDE|
1731 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1732 NEIGH_UPDATE_F_ISROUTER))
1735 nrt = ip6_rt_copy(rt, dest);
1739 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1741 nrt->rt6i_flags &= ~RTF_GATEWAY;
1743 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1744 nrt->n = neigh_clone(neigh);
1746 if (ip6_ins_rt(nrt))
1749 netevent.old = &rt->dst;
1750 netevent.old_neigh = old_neigh;
1751 netevent.new = &nrt->dst;
1752 netevent.new_neigh = neigh;
1753 netevent.daddr = dest;
1754 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1756 if (rt->rt6i_flags & RTF_CACHE) {
1757 rt = (struct rt6_info *) dst_clone(&rt->dst);
1762 neigh_release(neigh);
1766 * Misc support functions
1769 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1770 const struct in6_addr *dest)
1772 struct net *net = dev_net(ort->dst.dev);
1773 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1777 rt->dst.input = ort->dst.input;
1778 rt->dst.output = ort->dst.output;
1779 rt->dst.flags |= DST_HOST;
1781 rt->rt6i_dst.addr = *dest;
1782 rt->rt6i_dst.plen = 128;
1783 dst_copy_metrics(&rt->dst, &ort->dst);
1784 rt->dst.error = ort->dst.error;
1785 rt->rt6i_idev = ort->rt6i_idev;
1787 in6_dev_hold(rt->rt6i_idev);
1788 rt->dst.lastuse = jiffies;
1790 rt->rt6i_gateway = ort->rt6i_gateway;
1791 rt->rt6i_flags = ort->rt6i_flags;
1792 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1793 (RTF_DEFAULT | RTF_ADDRCONF))
1794 rt6_set_from(rt, ort);
1796 rt6_clean_expires(rt);
1797 rt->rt6i_metric = 0;
1799 #ifdef CONFIG_IPV6_SUBTREES
1800 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1802 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1803 rt->rt6i_table = ort->rt6i_table;
1808 #ifdef CONFIG_IPV6_ROUTE_INFO
1809 static struct rt6_info *rt6_get_route_info(struct net *net,
1810 const struct in6_addr *prefix, int prefixlen,
1811 const struct in6_addr *gwaddr, int ifindex)
1813 struct fib6_node *fn;
1814 struct rt6_info *rt = NULL;
1815 struct fib6_table *table;
1817 table = fib6_get_table(net, RT6_TABLE_INFO);
1821 write_lock_bh(&table->tb6_lock);
1822 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1826 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1827 if (rt->dst.dev->ifindex != ifindex)
1829 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1831 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1837 write_unlock_bh(&table->tb6_lock);
1841 static struct rt6_info *rt6_add_route_info(struct net *net,
1842 const struct in6_addr *prefix, int prefixlen,
1843 const struct in6_addr *gwaddr, int ifindex,
1846 struct fib6_config cfg = {
1847 .fc_table = RT6_TABLE_INFO,
1848 .fc_metric = IP6_RT_PRIO_USER,
1849 .fc_ifindex = ifindex,
1850 .fc_dst_len = prefixlen,
1851 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1852 RTF_UP | RTF_PREF(pref),
1854 .fc_nlinfo.nlh = NULL,
1855 .fc_nlinfo.nl_net = net,
1858 cfg.fc_dst = *prefix;
1859 cfg.fc_gateway = *gwaddr;
1861 /* We should treat it as a default route if prefix length is 0. */
1863 cfg.fc_flags |= RTF_DEFAULT;
1865 ip6_route_add(&cfg);
1867 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1871 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1873 struct rt6_info *rt;
1874 struct fib6_table *table;
1876 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1880 write_lock_bh(&table->tb6_lock);
1881 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1882 if (dev == rt->dst.dev &&
1883 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1884 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1889 write_unlock_bh(&table->tb6_lock);
1893 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1894 struct net_device *dev,
1897 struct fib6_config cfg = {
1898 .fc_table = RT6_TABLE_DFLT,
1899 .fc_metric = IP6_RT_PRIO_USER,
1900 .fc_ifindex = dev->ifindex,
1901 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1902 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1904 .fc_nlinfo.nlh = NULL,
1905 .fc_nlinfo.nl_net = dev_net(dev),
1908 cfg.fc_gateway = *gwaddr;
1910 ip6_route_add(&cfg);
1912 return rt6_get_dflt_router(gwaddr, dev);
1915 void rt6_purge_dflt_routers(struct net *net)
1917 struct rt6_info *rt;
1918 struct fib6_table *table;
1920 /* NOTE: Keep consistent with rt6_get_dflt_router */
1921 table = fib6_get_table(net, RT6_TABLE_DFLT);
1926 read_lock_bh(&table->tb6_lock);
1927 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1928 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1930 read_unlock_bh(&table->tb6_lock);
1935 read_unlock_bh(&table->tb6_lock);
1938 static void rtmsg_to_fib6_config(struct net *net,
1939 struct in6_rtmsg *rtmsg,
1940 struct fib6_config *cfg)
1942 memset(cfg, 0, sizeof(*cfg));
1944 cfg->fc_table = RT6_TABLE_MAIN;
1945 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1946 cfg->fc_metric = rtmsg->rtmsg_metric;
1947 cfg->fc_expires = rtmsg->rtmsg_info;
1948 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1949 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1950 cfg->fc_flags = rtmsg->rtmsg_flags;
1952 cfg->fc_nlinfo.nl_net = net;
1954 cfg->fc_dst = rtmsg->rtmsg_dst;
1955 cfg->fc_src = rtmsg->rtmsg_src;
1956 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1959 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1961 struct fib6_config cfg;
1962 struct in6_rtmsg rtmsg;
1966 case SIOCADDRT: /* Add a route */
1967 case SIOCDELRT: /* Delete a route */
1968 if (!capable(CAP_NET_ADMIN))
1970 err = copy_from_user(&rtmsg, arg,
1971 sizeof(struct in6_rtmsg));
1975 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1980 err = ip6_route_add(&cfg);
1983 err = ip6_route_del(&cfg);
1997 * Drop the packet on the floor
2000 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2003 struct dst_entry *dst = skb_dst(skb);
2004 switch (ipstats_mib_noroutes) {
2005 case IPSTATS_MIB_INNOROUTES:
2006 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2007 if (type == IPV6_ADDR_ANY) {
2008 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2009 IPSTATS_MIB_INADDRERRORS);
2013 case IPSTATS_MIB_OUTNOROUTES:
2014 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2015 ipstats_mib_noroutes);
2018 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2023 static int ip6_pkt_discard(struct sk_buff *skb)
2025 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2028 static int ip6_pkt_discard_out(struct sk_buff *skb)
2030 skb->dev = skb_dst(skb)->dev;
2031 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2034 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2036 static int ip6_pkt_prohibit(struct sk_buff *skb)
2038 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2041 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2043 skb->dev = skb_dst(skb)->dev;
2044 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2050 * Allocate a dst for local (unicast / anycast) address.
2053 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2054 const struct in6_addr *addr,
2057 struct net *net = dev_net(idev->dev);
2058 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2062 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2063 return ERR_PTR(-ENOMEM);
2068 rt->dst.flags |= DST_HOST;
2069 rt->dst.input = ip6_input;
2070 rt->dst.output = ip6_output;
2071 rt->rt6i_idev = idev;
2072 rt->dst.obsolete = -1;
2074 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2076 rt->rt6i_flags |= RTF_ANYCAST;
2078 rt->rt6i_flags |= RTF_LOCAL;
2079 err = rt6_bind_neighbour(rt, rt->dst.dev);
2082 return ERR_PTR(err);
2085 rt->rt6i_dst.addr = *addr;
2086 rt->rt6i_dst.plen = 128;
2087 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2089 atomic_set(&rt->dst.__refcnt, 1);
2094 int ip6_route_get_saddr(struct net *net,
2095 struct rt6_info *rt,
2096 const struct in6_addr *daddr,
2098 struct in6_addr *saddr)
2100 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2102 if (rt->rt6i_prefsrc.plen)
2103 *saddr = rt->rt6i_prefsrc.addr;
2105 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2106 daddr, prefs, saddr);
2110 /* remove deleted ip from prefsrc entries */
2111 struct arg_dev_net_ip {
2112 struct net_device *dev;
2114 struct in6_addr *addr;
2117 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2119 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2120 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2121 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2123 if (((void *)rt->dst.dev == dev || !dev) &&
2124 rt != net->ipv6.ip6_null_entry &&
2125 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2126 /* remove prefsrc entry */
2127 rt->rt6i_prefsrc.plen = 0;
2132 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2134 struct net *net = dev_net(ifp->idev->dev);
2135 struct arg_dev_net_ip adni = {
2136 .dev = ifp->idev->dev,
2140 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2143 struct arg_dev_net {
2144 struct net_device *dev;
2148 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2150 const struct arg_dev_net *adn = arg;
2151 const struct net_device *dev = adn->dev;
2153 if ((rt->dst.dev == dev || !dev) &&
2154 rt != adn->net->ipv6.ip6_null_entry)
2160 void rt6_ifdown(struct net *net, struct net_device *dev)
2162 struct arg_dev_net adn = {
2167 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2168 icmp6_clean_all(fib6_ifdown, &adn);
2171 struct rt6_mtu_change_arg {
2172 struct net_device *dev;
2176 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2178 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2179 struct inet6_dev *idev;
2181 /* In IPv6 pmtu discovery is not optional,
2182 so that RTAX_MTU lock cannot disable it.
2183 We still use this lock to block changes
2184 caused by addrconf/ndisc.
2187 idev = __in6_dev_get(arg->dev);
2191 /* For administrative MTU increase, there is no way to discover
2192 IPv6 PMTU increase, so PMTU increase should be updated here.
2193 Since RFC 1981 doesn't include administrative MTU increase
2194 update PMTU increase is a MUST. (i.e. jumbo frame)
2197 If new MTU is less than route PMTU, this new MTU will be the
2198 lowest MTU in the path, update the route PMTU to reflect PMTU
2199 decreases; if new MTU is greater than route PMTU, and the
2200 old MTU is the lowest MTU in the path, update the route PMTU
2201 to reflect the increase. In this case if the other nodes' MTU
2202 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2205 if (rt->dst.dev == arg->dev &&
2206 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2207 (dst_mtu(&rt->dst) >= arg->mtu ||
2208 (dst_mtu(&rt->dst) < arg->mtu &&
2209 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2210 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2215 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2217 struct rt6_mtu_change_arg arg = {
2222 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2225 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2226 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2227 [RTA_OIF] = { .type = NLA_U32 },
2228 [RTA_IIF] = { .type = NLA_U32 },
2229 [RTA_PRIORITY] = { .type = NLA_U32 },
2230 [RTA_METRICS] = { .type = NLA_NESTED },
2233 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2234 struct fib6_config *cfg)
2237 struct nlattr *tb[RTA_MAX+1];
2240 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2245 rtm = nlmsg_data(nlh);
2246 memset(cfg, 0, sizeof(*cfg));
2248 cfg->fc_table = rtm->rtm_table;
2249 cfg->fc_dst_len = rtm->rtm_dst_len;
2250 cfg->fc_src_len = rtm->rtm_src_len;
2251 cfg->fc_flags = RTF_UP;
2252 cfg->fc_protocol = rtm->rtm_protocol;
2254 if (rtm->rtm_type == RTN_UNREACHABLE)
2255 cfg->fc_flags |= RTF_REJECT;
2257 if (rtm->rtm_type == RTN_LOCAL)
2258 cfg->fc_flags |= RTF_LOCAL;
2260 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2261 cfg->fc_nlinfo.nlh = nlh;
2262 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2264 if (tb[RTA_GATEWAY]) {
2265 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2266 cfg->fc_flags |= RTF_GATEWAY;
2270 int plen = (rtm->rtm_dst_len + 7) >> 3;
2272 if (nla_len(tb[RTA_DST]) < plen)
2275 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2279 int plen = (rtm->rtm_src_len + 7) >> 3;
2281 if (nla_len(tb[RTA_SRC]) < plen)
2284 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2287 if (tb[RTA_PREFSRC])
2288 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2291 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2293 if (tb[RTA_PRIORITY])
2294 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2296 if (tb[RTA_METRICS]) {
2297 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2298 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2302 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2309 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2311 struct fib6_config cfg;
2314 err = rtm_to_fib6_config(skb, nlh, &cfg);
2318 return ip6_route_del(&cfg);
2321 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2323 struct fib6_config cfg;
2326 err = rtm_to_fib6_config(skb, nlh, &cfg);
2330 return ip6_route_add(&cfg);
2333 static inline size_t rt6_nlmsg_size(void)
2335 return NLMSG_ALIGN(sizeof(struct rtmsg))
2336 + nla_total_size(16) /* RTA_SRC */
2337 + nla_total_size(16) /* RTA_DST */
2338 + nla_total_size(16) /* RTA_GATEWAY */
2339 + nla_total_size(16) /* RTA_PREFSRC */
2340 + nla_total_size(4) /* RTA_TABLE */
2341 + nla_total_size(4) /* RTA_IIF */
2342 + nla_total_size(4) /* RTA_OIF */
2343 + nla_total_size(4) /* RTA_PRIORITY */
2344 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2345 + nla_total_size(sizeof(struct rta_cacheinfo));
2348 static int rt6_fill_node(struct net *net,
2349 struct sk_buff *skb, struct rt6_info *rt,
2350 struct in6_addr *dst, struct in6_addr *src,
2351 int iif, int type, u32 pid, u32 seq,
2352 int prefix, int nowait, unsigned int flags)
2355 struct nlmsghdr *nlh;
2358 struct neighbour *n;
2360 if (prefix) { /* user wants prefix routes only */
2361 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2362 /* success since this is not a prefix route */
2367 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2371 rtm = nlmsg_data(nlh);
2372 rtm->rtm_family = AF_INET6;
2373 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2374 rtm->rtm_src_len = rt->rt6i_src.plen;
2377 table = rt->rt6i_table->tb6_id;
2379 table = RT6_TABLE_UNSPEC;
2380 rtm->rtm_table = table;
2381 if (nla_put_u32(skb, RTA_TABLE, table))
2382 goto nla_put_failure;
2383 if (rt->rt6i_flags & RTF_REJECT)
2384 rtm->rtm_type = RTN_UNREACHABLE;
2385 else if (rt->rt6i_flags & RTF_LOCAL)
2386 rtm->rtm_type = RTN_LOCAL;
2387 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2388 rtm->rtm_type = RTN_LOCAL;
2390 rtm->rtm_type = RTN_UNICAST;
2392 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2393 rtm->rtm_protocol = rt->rt6i_protocol;
2394 if (rt->rt6i_flags & RTF_DYNAMIC)
2395 rtm->rtm_protocol = RTPROT_REDIRECT;
2396 else if (rt->rt6i_flags & RTF_ADDRCONF)
2397 rtm->rtm_protocol = RTPROT_KERNEL;
2398 else if (rt->rt6i_flags & RTF_DEFAULT)
2399 rtm->rtm_protocol = RTPROT_RA;
2401 if (rt->rt6i_flags & RTF_CACHE)
2402 rtm->rtm_flags |= RTM_F_CLONED;
2405 if (nla_put(skb, RTA_DST, 16, dst))
2406 goto nla_put_failure;
2407 rtm->rtm_dst_len = 128;
2408 } else if (rtm->rtm_dst_len)
2409 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2410 goto nla_put_failure;
2411 #ifdef CONFIG_IPV6_SUBTREES
2413 if (nla_put(skb, RTA_SRC, 16, src))
2414 goto nla_put_failure;
2415 rtm->rtm_src_len = 128;
2416 } else if (rtm->rtm_src_len &&
2417 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2418 goto nla_put_failure;
2421 #ifdef CONFIG_IPV6_MROUTE
2422 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2423 int err = ip6mr_get_route(net, skb, rtm, nowait);
2428 goto nla_put_failure;
2430 if (err == -EMSGSIZE)
2431 goto nla_put_failure;
2436 if (nla_put_u32(skb, RTA_IIF, iif))
2437 goto nla_put_failure;
2439 struct in6_addr saddr_buf;
2440 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2441 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2442 goto nla_put_failure;
2445 if (rt->rt6i_prefsrc.plen) {
2446 struct in6_addr saddr_buf;
2447 saddr_buf = rt->rt6i_prefsrc.addr;
2448 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2449 goto nla_put_failure;
2452 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2453 goto nla_put_failure;
2458 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0) {
2460 goto nla_put_failure;
2466 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2467 goto nla_put_failure;
2468 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2469 goto nla_put_failure;
2470 if (!(rt->rt6i_flags & RTF_EXPIRES))
2472 else if (rt->dst.expires - jiffies < INT_MAX)
2473 expires = rt->dst.expires - jiffies;
2477 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2478 goto nla_put_failure;
2480 return nlmsg_end(skb, nlh);
2483 nlmsg_cancel(skb, nlh);
2487 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2489 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2492 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2493 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2494 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2498 return rt6_fill_node(arg->net,
2499 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2500 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2501 prefix, 0, NLM_F_MULTI);
2504 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2506 struct net *net = sock_net(in_skb->sk);
2507 struct nlattr *tb[RTA_MAX+1];
2508 struct rt6_info *rt;
2509 struct sk_buff *skb;
2512 int err, iif = 0, oif = 0;
2514 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2519 memset(&fl6, 0, sizeof(fl6));
2522 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2525 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2529 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2532 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2536 iif = nla_get_u32(tb[RTA_IIF]);
2539 oif = nla_get_u32(tb[RTA_OIF]);
2542 struct net_device *dev;
2545 dev = __dev_get_by_index(net, iif);
2551 fl6.flowi6_iif = iif;
2553 if (!ipv6_addr_any(&fl6.saddr))
2554 flags |= RT6_LOOKUP_F_HAS_SADDR;
2556 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2559 fl6.flowi6_oif = oif;
2561 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2564 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2566 dst_release(&rt->dst);
2571 /* Reserve room for dummy headers, this skb can pass
2572 through good chunk of routing engine.
2574 skb_reset_mac_header(skb);
2575 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2577 skb_dst_set(skb, &rt->dst);
2579 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2580 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2581 nlh->nlmsg_seq, 0, 0, 0);
2587 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2592 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2594 struct sk_buff *skb;
2595 struct net *net = info->nl_net;
2600 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2602 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2606 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2607 event, info->pid, seq, 0, 0, 0);
2609 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2610 WARN_ON(err == -EMSGSIZE);
2614 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2615 info->nlh, gfp_any());
2619 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2622 static int ip6_route_dev_notify(struct notifier_block *this,
2623 unsigned long event, void *data)
2625 struct net_device *dev = (struct net_device *)data;
2626 struct net *net = dev_net(dev);
2628 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2629 net->ipv6.ip6_null_entry->dst.dev = dev;
2630 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2631 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2632 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2633 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2634 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2635 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2646 #ifdef CONFIG_PROC_FS
2657 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2659 struct seq_file *m = p_arg;
2660 struct neighbour *n;
2662 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2664 #ifdef CONFIG_IPV6_SUBTREES
2665 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2667 seq_puts(m, "00000000000000000000000000000000 00 ");
2672 seq_printf(m, "%pi6", n->primary_key);
2674 seq_puts(m, "00000000000000000000000000000000");
2677 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2678 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2679 rt->dst.__use, rt->rt6i_flags,
2680 rt->dst.dev ? rt->dst.dev->name : "");
2684 static int ipv6_route_show(struct seq_file *m, void *v)
2686 struct net *net = (struct net *)m->private;
2687 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2691 static int ipv6_route_open(struct inode *inode, struct file *file)
2693 return single_open_net(inode, file, ipv6_route_show);
2696 static const struct file_operations ipv6_route_proc_fops = {
2697 .owner = THIS_MODULE,
2698 .open = ipv6_route_open,
2700 .llseek = seq_lseek,
2701 .release = single_release_net,
2704 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2706 struct net *net = (struct net *)seq->private;
2707 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2708 net->ipv6.rt6_stats->fib_nodes,
2709 net->ipv6.rt6_stats->fib_route_nodes,
2710 net->ipv6.rt6_stats->fib_rt_alloc,
2711 net->ipv6.rt6_stats->fib_rt_entries,
2712 net->ipv6.rt6_stats->fib_rt_cache,
2713 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2714 net->ipv6.rt6_stats->fib_discarded_routes);
2719 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2721 return single_open_net(inode, file, rt6_stats_seq_show);
2724 static const struct file_operations rt6_stats_seq_fops = {
2725 .owner = THIS_MODULE,
2726 .open = rt6_stats_seq_open,
2728 .llseek = seq_lseek,
2729 .release = single_release_net,
2731 #endif /* CONFIG_PROC_FS */
2733 #ifdef CONFIG_SYSCTL
2736 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2737 void __user *buffer, size_t *lenp, loff_t *ppos)
2744 net = (struct net *)ctl->extra1;
2745 delay = net->ipv6.sysctl.flush_delay;
2746 proc_dointvec(ctl, write, buffer, lenp, ppos);
2747 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2751 ctl_table ipv6_route_table_template[] = {
2753 .procname = "flush",
2754 .data = &init_net.ipv6.sysctl.flush_delay,
2755 .maxlen = sizeof(int),
2757 .proc_handler = ipv6_sysctl_rtcache_flush
2760 .procname = "gc_thresh",
2761 .data = &ip6_dst_ops_template.gc_thresh,
2762 .maxlen = sizeof(int),
2764 .proc_handler = proc_dointvec,
2767 .procname = "max_size",
2768 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2769 .maxlen = sizeof(int),
2771 .proc_handler = proc_dointvec,
2774 .procname = "gc_min_interval",
2775 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2776 .maxlen = sizeof(int),
2778 .proc_handler = proc_dointvec_jiffies,
2781 .procname = "gc_timeout",
2782 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2783 .maxlen = sizeof(int),
2785 .proc_handler = proc_dointvec_jiffies,
2788 .procname = "gc_interval",
2789 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2790 .maxlen = sizeof(int),
2792 .proc_handler = proc_dointvec_jiffies,
2795 .procname = "gc_elasticity",
2796 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2797 .maxlen = sizeof(int),
2799 .proc_handler = proc_dointvec,
2802 .procname = "mtu_expires",
2803 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2804 .maxlen = sizeof(int),
2806 .proc_handler = proc_dointvec_jiffies,
2809 .procname = "min_adv_mss",
2810 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2811 .maxlen = sizeof(int),
2813 .proc_handler = proc_dointvec,
2816 .procname = "gc_min_interval_ms",
2817 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2818 .maxlen = sizeof(int),
2820 .proc_handler = proc_dointvec_ms_jiffies,
2825 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2827 struct ctl_table *table;
2829 table = kmemdup(ipv6_route_table_template,
2830 sizeof(ipv6_route_table_template),
2834 table[0].data = &net->ipv6.sysctl.flush_delay;
2835 table[0].extra1 = net;
2836 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2837 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2838 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2839 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2840 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2841 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2842 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2843 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2844 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2851 static int __net_init ip6_route_net_init(struct net *net)
2855 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2856 sizeof(net->ipv6.ip6_dst_ops));
2858 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2859 goto out_ip6_dst_ops;
2861 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2862 sizeof(*net->ipv6.ip6_null_entry),
2864 if (!net->ipv6.ip6_null_entry)
2865 goto out_ip6_dst_entries;
2866 net->ipv6.ip6_null_entry->dst.path =
2867 (struct dst_entry *)net->ipv6.ip6_null_entry;
2868 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2869 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2870 ip6_template_metrics, true);
2872 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2873 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2874 sizeof(*net->ipv6.ip6_prohibit_entry),
2876 if (!net->ipv6.ip6_prohibit_entry)
2877 goto out_ip6_null_entry;
2878 net->ipv6.ip6_prohibit_entry->dst.path =
2879 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2880 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2881 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2882 ip6_template_metrics, true);
2884 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2885 sizeof(*net->ipv6.ip6_blk_hole_entry),
2887 if (!net->ipv6.ip6_blk_hole_entry)
2888 goto out_ip6_prohibit_entry;
2889 net->ipv6.ip6_blk_hole_entry->dst.path =
2890 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2891 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2892 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2893 ip6_template_metrics, true);
2896 net->ipv6.sysctl.flush_delay = 0;
2897 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2898 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2899 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2900 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2901 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2902 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2903 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2905 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2911 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2912 out_ip6_prohibit_entry:
2913 kfree(net->ipv6.ip6_prohibit_entry);
2915 kfree(net->ipv6.ip6_null_entry);
2917 out_ip6_dst_entries:
2918 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2923 static void __net_exit ip6_route_net_exit(struct net *net)
2925 kfree(net->ipv6.ip6_null_entry);
2926 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2927 kfree(net->ipv6.ip6_prohibit_entry);
2928 kfree(net->ipv6.ip6_blk_hole_entry);
2930 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2933 static int __net_init ip6_route_net_init_late(struct net *net)
2935 #ifdef CONFIG_PROC_FS
2936 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2937 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2942 static void __net_exit ip6_route_net_exit_late(struct net *net)
2944 #ifdef CONFIG_PROC_FS
2945 proc_net_remove(net, "ipv6_route");
2946 proc_net_remove(net, "rt6_stats");
2950 static struct pernet_operations ip6_route_net_ops = {
2951 .init = ip6_route_net_init,
2952 .exit = ip6_route_net_exit,
2955 static int __net_init ipv6_inetpeer_init(struct net *net)
2957 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2961 inet_peer_base_init(bp);
2962 net->ipv6.peers = bp;
2966 static void __net_exit ipv6_inetpeer_exit(struct net *net)
2968 struct inet_peer_base *bp = net->ipv6.peers;
2970 net->ipv6.peers = NULL;
2971 inetpeer_invalidate_tree(bp);
2975 static struct pernet_operations ipv6_inetpeer_ops = {
2976 .init = ipv6_inetpeer_init,
2977 .exit = ipv6_inetpeer_exit,
2980 static struct pernet_operations ip6_route_net_late_ops = {
2981 .init = ip6_route_net_init_late,
2982 .exit = ip6_route_net_exit_late,
2985 static struct notifier_block ip6_route_dev_notifier = {
2986 .notifier_call = ip6_route_dev_notify,
2990 int __init ip6_route_init(void)
2995 ip6_dst_ops_template.kmem_cachep =
2996 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2997 SLAB_HWCACHE_ALIGN, NULL);
2998 if (!ip6_dst_ops_template.kmem_cachep)
3001 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3003 goto out_kmem_cache;
3005 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3007 goto out_dst_entries;
3009 ret = register_pernet_subsys(&ip6_route_net_ops);
3011 goto out_register_inetpeer;
3013 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3015 /* Registering of the loopback is done before this portion of code,
3016 * the loopback reference in rt6_info will not be taken, do it
3017 * manually for init_net */
3018 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3019 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3020 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3021 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3022 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3023 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3024 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3028 goto out_register_subsys;
3034 ret = fib6_rules_init();
3038 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3040 goto fib6_rules_init;
3043 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3044 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3045 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3046 goto out_register_late_subsys;
3048 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3050 goto out_register_late_subsys;
3055 out_register_late_subsys:
3056 unregister_pernet_subsys(&ip6_route_net_late_ops);
3058 fib6_rules_cleanup();
3063 out_register_subsys:
3064 unregister_pernet_subsys(&ip6_route_net_ops);
3065 out_register_inetpeer:
3066 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3068 dst_entries_destroy(&ip6_dst_blackhole_ops);
3070 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3074 void ip6_route_cleanup(void)
3076 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3077 unregister_pernet_subsys(&ip6_route_net_late_ops);
3078 fib6_rules_cleanup();
3081 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3082 unregister_pernet_subsys(&ip6_route_net_ops);
3083 dst_entries_destroy(&ip6_dst_blackhole_ops);
3084 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);